At91sam9g45 Ds

Features
• 400 MHz ARM926EJ-S™ ARM® Thumb® Processor

– 32 KBytes Data Cache, 32 KBytes Instruction Cache, MMU
• Memories
– Dual External Bus Interface supporting 4-bank DDR2/LPDDR, SDRAM/LPSDR,
Static Memories, CompactFlash, SLC NAND Flash with ECC
– One 64-kbyte internal SRAM, single-cycle access at system speed or processor
speed through TCM interface
– One 64-kbyte internal ROM, embedding bootstrap routine
• Peripherals
– LCD Controller supporting STN and TFT displays up to 1280*860
AT91 ARM
– ITU-R BT. 601/656 Image Sensor Interface Thumb-based
– USB Device High Speed, USB Host High Speed and USB Host Full Speed with On-
Chip Transceiver Microcontrollers
– 10/100 Mbps Ethernet MAC Controller
– Two High Speed Memory Card Hosts (SDIO, SDCard, MMC)
– AC'97 controller
– Two Master/Slave Serial Peripheral Interfaces AT91SAM9G45
– Two Three-channel 32-bit Timer/Counters
– Two Synchronous Serial Controllers (I2S mode)
– Four-channel 16-bit PWM Controller
– Two Two-wire Interfaces
Preliminary
– Four USARTs with ISO7816, IrDA, Manchester and SPI modes
– 8-channel 10-bit ADC with 4-wire Touch Screen support
• System
– 133 MHz twelve 32-bit layer AHB Bus Matrix
– 37 DMA Channels
– Boot from NAND Flash, SDCard, DataFlash® or serial DataFlash
– Reset Controller with on-chip Power-on Reset
– Selectable 32768 Hz Low-power and 12 MHz Crystal Oscillators
– Internal Low-power 32 kHz RC Oscillator
– One PLL for the system and one 480 MHz PLL optimized for USB High Speed
– Two Programmable External Clock Signals
– Advanced Interrupt Controller and Debug Unit
– Periodic Interval Timer, Watchdog Timer, Real Time Timer and Real Time Clock
• I/O
– Five 32-bit Parallel Input/Output Controllers
– 160 Programmable I/O Lines Multiplexed with up to Two Peripheral I/Os with
Schmitt trigger input
• Package
– 324-ball LFBGA, pitch 0.8 mm
6438B–ATARM–29-Jul-09
1. Description
The ARM926EJ-S based AT91SAM9G45 features the frequently demanded combination of user
interface functionality and high data rate connectivity, including LCD Controller, resistive touch-
screen, camera interface, audio, Ethernet 10/100 and high speed USB and SDIO. With the pro-
cessor running at 400MHz and multiple 100+ Mbps data rate peripherals, the AT91SAM9G45
has the performance and bandwidth to the network or local storage media to provide an ade-
quate user experience.
The AT91SAM9G45 supports the latest generation of DDR2 and NAND Flash memory inter-
faces for program and data storage. An internal 133 MHz multi-layer bus architecture associated
with 37 DMA channels, a dual external bus interface and distributed memory including a 64-
Kbyte SRAM which can be configured as a tightly coupled memory (TCM) sustains the high
bandwidth required by the processor and the high speed peripherals.
The I/Os support 1.8V or 3.3V operation, which are independently configurable for the memory
interface and peripheral I/Os. This feature completely eliminates the need for any external level
shifters. In addition it supports 0.8 ball pitch package for low cost PCB manufacturing.
The AT91SAM9G45 power management controller features efficient clock gating and a battery
backup section minimizing power consumption in active and standby modes.
2 AT91SAM9G45
B
DM
S
HS B
Figure 2-1.
M /H DM
D F
NC
HS /H
SY
A
B,D DM
M
DM
11
DP ,DFS
E
HS A
H B
C
, C
,H SDM
E
IS
A
-
EL
-E C
P F
P/H DP
T
C
O
GS
S
DI
SD ,H
SD FS
CK
RS
I
O
K
R
X -
S
S
X
D
D -L
D YN D2
D T ,LC
D ,LC
X
X R D
_ Y
I_
I_
I_
I
I_M YN
C
D
VB
IS DO- LCD
IS PCK I_D OD
IS HS
IS VS NC
ET CK
EC EN RX
ER S-E TX K
ERXER COL R
ET 0-E ERX
EMDC X3
LC
LC D0
L VS CD
L DO C 3
LC EN CK DH
IS WR DC
ET
HF
DF
EM 0-ET X3 V
HH DPA
DH DP/H
NT
TD
TD
TM
TC S
BM
RT
JTA
2. Block Diagram
JTAG / Boundary Scan HS HS PIO

TST System Controller Transceiver Transceiver
PCK0-PCK1
DDR_A0-DDR_A13
FIQ In-Circuit Emulator DDR_D0-DDR_D15
AIC PA PB
IRQ DDR_VREF
DRXD DBGU ARM926EJ-S DDR_DQM[0..1]
LCD ISI EMAC 8-CH
PIO
HS EHCI HS DDR_DQS[0..1]
DTXD PDC DMA
DCache USB HOST USB
ICache
32 Kbytes
MMU 32 Kbytes DDR_CS
PLLRCA DDR2
AT91SAM9G45 Block Diagram
PLLA DMA DMA DMA DMA DMA DDR_CLK,#DDR_CLK

ITCM DTCM Bus Interface
LPDDR DDR_CKE
PLLUTMI PMC I D DDR_RAS, DDR_CAS
XIN DDR_WE
OSC12M
XOUT DDR_BA0, DDR_BA1
SRAM
WDT PIT 64KB
4 D0-D15
RC A0/NBS0
XIN32 GPBR EBI
OSC 32K A1/NBS2/NWR2
XOUT32 RTT A2-A15, A18
SHDN A16/BA0
SHDC RTC Multi-Layer AHB Matrix DDR2/
WKUP A17/BA1
LPDDR/
NCS1/SDCS
VDDBU POR
SDRAM
RSTC SDCK, #SDCK, SDCKE
NRST Controller
RAS, CAS
VDDCORE POR SDWE, SDA10
PIOA PIOD DQM[0..1]
Peripheral Peripheral NAND Flash DQS[0..1]
ROM
PIOB PIOE RNG DMA Controller
64KB Bridge NRD
Controller ECC NWR0/NWE
PIOC
NWR1/NBS1
NWR3/NBS3
APB NCS0
NANDOE, NANDWE
CF
PDC D16-D31
FIFO PDC PDC PDC
TC0 TC3 PDC NWAIT
TWI0 USART0 4-CH DQM[2..3]
MCI0/MCI1 TC1 TC4
TWI1 USART1 PWM SPI0 SSC0 A19-A24
SD/SDIO TC2 TC5 AC97 8-CH
USART2 SPI1 SSC1 NCS4/CFCS0
CE ATA 10Bit ADC Static
USART3 NCS5/CFCS1
TouchScreen Memory
A25/CFRNW
Controller
CFCE1-CFCE2
NCS2
PIO NCS3/NANDCS
7
3
1
S
X
N
D
D3
M
C
Y
DA
0X
D
C 3
C
2Y
DA
IO
IO
9 K
TR
TX
M K
M SI
D
SP S0
0
0
0- F1
0 D
DA F
PW
1_ K
1_ 7
-T K2
-T
-T A5
I
NP 5
NPCS
NP S2
NPCS
A IG
0- X3
AD XM
DV AD7
D
AC97C
AD P
AC97R
AD X
GN NA
AC RK1
AC 7F
0-
TF -TK
T ISO
C
S CT
S C
TD 0-T 1
RF -RD1
K T
D
X C
RK -R 1
A
R 0-T 1
0-
A0 L
VD RE
B3 TIOK5
4
M
TS 97T
-M 1_C
O -TC
O - B
LK
O
L
O -
,M _
RT 0- K1
CT TW 1
TX 0-R K3
I
SC 0-R S3
RD 0-S S3
TS -GP M
CK -TW
C
TW
AD AD3 P
T 0
TI K3 IO
TI
TI A3 TCL 2
PW
A0 CI A
SPI0_, SPI1_ SSC0_, SSC1_
TC
TC B0 A2
TW D0
C M
GP
_D K,M CD
C
1
0_ A0-
M DA CI0
CI
CI
CI
M
M I0_ CI DA
M 0_D
3
AT91SAM9G45
3. Signal Description
Table 3-1 gives details on the signal names classified by peripheral.
Table 3-1. Signal Description List

Active Reference
Signal Name Function Type Level Voltage Comments
Power Supplies
VDDIOM0 DDR2 I/O Lines Power Supply Power 1.65V to 1.95V
VDDIOM1 EBI I/O Lines Power Supply Power 1.65V to 1.95V or 3.0V to3.6V
VDDIOP0 Peripherals I/O Lines Power Supply Power 1.65V to 3.6V
VDDIOP1 Peripherals I/O Lines Power Supply Power 1.65V to 3.6V
VDDIOP2 ISI I/O Lines Power Supply Power 1.65V to 3.6V
VDDBU Backup I/O Lines Power Supply Power 1.8V to 3.6V
VDDANA Analog Power Supply Power 3.0V to 3.6V
VDDPLLA PLLA Power Supply Power 0.9V to 1.1V
VDDPLLUTMI PLLUTMI Power Supply Power 0.9V to 1.1V
VDDOSC Oscillator Power Supply Power 1.65V to 3.6V
VDDCORE Core Chip Power Supply Power 0.9V to 1.1V
UDPHS and UHPHS UTMI+ Core
VDDUTMIC Power 0.9V to 1.1V
Power Supply
UDPHS and UHPHS UTMI+ interface
VDDUTMII Power 3.0V to 3.6V
Power Supply
GNDIOM DDR2 and EBI I/O Lines Ground Ground
GNDIOP Peripherals and ISI I/O lines Ground Ground
GNDCORE Core Chip Ground Ground
PLLA, PLLUTMI and Oscillator
GNDOSC Ground
Ground
GNDBU Backup Ground Ground
UDPHS and UHPHS UTMI+ Core and
GNDUTMI Ground
interface Ground
GNDANA Analog Ground Ground
Clocks, Oscillators and PLLs
XIN Main Oscillator Input Input
XOUT Main Oscillator Output Output
XIN32 Slow Clock Oscillator Input Input
XOUT32 Slow Clock Oscillator Output Output
VBG Bias Voltage Reference for USB Analog
(1)
PCK0 - PCK1 Programmable Clock Output Output
4 AT91SAM9G45
AT91SAM9G45
Table 3-1. Signal Description List (Continued)

Active Reference
Shutdown, Wakeup Logic
Driven at 0V only.
0: The device is in backup
SHDN Shut-Down Control Output VDDBU mode
1: The device is running (not in
backup mode).
Accept between 0V and
WKUP Wake-Up Input Input VDDBU
VDDBU.
ICE and JTAG
No pull-up resistor, Schmitt
TCK Test Clock Input VDDIOP0
trigger
TDI Test Data In Input VDDIOP0
trigger
TDO Test Data Out Output VDDIOP0
TMS Test Mode Select Input VDDIOP0
trigger
JTAGSEL JTAG Selection Input VDDBU Pull-down resistor (15 kΩ).
RTCK Return Test Clock Output VDDIOP0
Reset/Test
Pull-Up resistor (100 kΩ),
NRST Microcontroller Reset I/O Low VDDIOP0
Schmitt trigger
Pull-down resistor (15 kΩ),
TST Test Mode Select Input VDDBU
Schmitt trigger
Pull-Up resistor (100 kΩ),
NTRST Test Reset Signal Input VDDIOP0
Schmitt trigger
must be connected to GND or
BMS Boot Mode Select Input VDDIOP0
VDDIOP.
Debug Unit - DBGU
(1)
DRXD Debug Receive Data Input
(1)
DTXD Debug Transmit Data Output
Advanced Interrupt Controller - AIC
(1)
IRQ External Interrupt Input Input
(1)
FIQ Fast Interrupt Input Input
PIO Controller - PIOA- PIOB - PIOC - PIOD - PIOE
(1) Pulled-up input at reset

PA0 - PA31 Parallel IO Controller A I/O
(100kΩ)(2), Schmitt trigger

PB0 - PB31 Parallel IO Controller B I/O

PC0 - PC31 Parallel IO Controller C I/O
5
Active Reference

PD0 - PD31 Parallel IO Controller D I/O

PE0 - PE31 Parallel IO Controller E I/O
DDR Memory Interface- DDR2/SDRAM/LPDDR Controller
DDR_D0 -
Data Bus I/O VDDIOM0 Pulled-up input at reset
DDR_D15
DDR_A0 -
Address Bus Output VDDIOM0 0 at reset
DDR_A13
DDR_CLK-
DDR differential clock input Output VDDIOM0
#DDR_CLK
DDR_CKE DDR Clock Enable Output High VDDIOM0
DDR_CS DDR Chip Select Output Low VDDIOM0
DDR_WE DDR Write Enable Output Low VDDIOM0
DDR_RAS-
Row and Column Signal Output Low VDDIOM0
DDR_CAS
DDR_DQM[0..1] Write Data Mask Output VDDIOM0
DDR_DQS[0..1] Data Strobe Output VDDIOM0
DDR_BA0 -
Bank Select Output VDDIOM0
DDR_BA1
DDR_VREF Reference Voltage Input VDDIOM0
External Bus Interface - EBI
D0 -D31 Data Bus I/O VDDIOM1 Pulled-up input at reset
A0 - A25 Address Bus Output VDDIOM1 0 at reset
NWAIT External Wait Signal Input Low VDDIOM1
Static Memory Controller - SMC
NCS0 - NCS5 Chip Select Lines Output Low VDDIOM1
NWR0 - NWR3 Write Signal Output Low VDDIOM1
NRD Read Signal Output Low VDDIOM1
NWE Write Enable Output Low VDDIOM1
NBS0 - NBS3 Byte Mask Signal Output Low VDDIOM1
CompactFlash Support
CFCE1 - CFCE2 CompactFlash Chip Enable Output Low VDDIOM1
CFOE CompactFlash Output Enable Output Low VDDIOM1
CFWE CompactFlash Write Enable Output Low VDDIOM1
CFIOR CompactFlash IO Read Output Low VDDIOM1
CFIOW CompactFlash IO Write Output Low VDDIOM1
CFRNW CompactFlash Read Not Write Output VDDIOM1
6 AT91SAM9G45
AT91SAM9G45

Active Reference
CFCS0 -CFCS1 CompactFlash Chip Select Lines Output Low VDDIOM1
NAND Flash Support
NANDCS NAND Flash Chip Select Output Low VDDIOM1
NANDOE NAND Flash Output Enable Output Low VDDIOM1
NANDWE NAND Flash Write Enable Output Low VDDIOM1
DDR2/SDRAM/LPDDR Controller
SDCK,#SDCK DDR2/SDRAM differential clock Output VDDIOM1
SDCKE DDR2/SDRAM Clock Enable Output High VDDIOM1
SDCS DDR2/SDRAM Controller Chip Select Output Low VDDIOM1
BA0 - BA1 Bank Select Output VDDIOM1
SDWE DDR2/SDRAM Write Enable Output Low VDDIOM1
RAS - CAS Row and Column Signal Output Low VDDIOM1
SDA10 SDRAM Address 10 Line Output VDDIOM1
DQS[0..1] Data Strobe Output VDDIOM1
DQM[0..3] Write Data Mask Output VDDIOM1
High Speed Multimedia Card Interface - HSMCIx
(1)
MCIx_CK Multimedia Card Clock I/O
(1)
MCIx_CDA Multimedia Card Slot A Command I/O
MCIx_DA0 - (1)
Multimedia Card Slot A Data I/O
MCIx_DA7
Universal Synchronous Asynchronous Receiver Transmitter - USARTx
(1)
SCKx USARTx Serial Clock I/O
(1)
TXDx USARTx Transmit Data Output
(1)
RXDx USARTx Receive Data Input
(1)
RTSx USARTx Request To Send Output
(1)
CTSx USARTx Clear To Send Input
Synchronous Serial Controller - SSCx
(1)
TDx SSC Transmit Data Output
(1)
RDx SSC Receive Data Input
(1)
TKx SSC Transmit Clock I/O
(1)
RKx SSC Receive Clock I/O
(1)
TFx SSC Transmit Frame Sync I/O
(1)
RFx SSC Receive Frame Sync I/O
7
Active Reference
AC97 Controller - AC97C
(1)
AC97RX AC97 Receive Signal Input
(1)
AC97TX AC97 Transmit Signal Output
(1)
AC97FS AC97 Frame Synchronization Signal Output
(1)
AC97CK AC97 Clock signal Input
Time Counter - TCx
(1)
TCLKx TC Channel x External Clock Input Input
(1)
TIOAx TC Channel x I/O Line A I/O
(1)
TIOBx TC Channel x I/O Line B I/O
Pulse Width Modulation Controller - PWM
(1)
PWMx Pulse Width Modulation Output Output
Serial Peripheral Interface - SPIx_
(1)
SPIx_MISO Master In Slave Out I/O
(1)
SPIx_MOSI Master Out Slave In I/O
(1)
SPIx_SPCK SPI Serial Clock I/O
(1)
SPIx_NPCS0 SPI Peripheral Chip Select 0 I/O Low
SPIx_NPCS1- (1)
SPI Peripheral Chip Select Output Low
SPIx_NPCS3
Two-Wire Interface
(1)
TWDx Two-wire Serial Data I/O
(1)
TWCKx Two-wire Serial Clock I/O
USB Host High Speed Port - UHPHS
HFSDPA USB Host Port A Full Speed Data + Analog VDDUTMII
HFSDMA USB Host Port A Full Speed Data - Analog VDDUTMII
HHSDPA USB Host Port A High Speed Data + Analog VDDUTMII
HHSDMA USB Host Port A High Speed Data - Analog VDDUTMII
HFSDPB USB Host Port B Full Speed Data + Analog VDDUTMII Multiplexed with DFSDP
HFSDMB USB Host Port B Full Speed Data - Analog VDDUTMII Multiplexed with DFSDM
HHSDPB USB Host Port B High Speed Data + Analog VDDUTMII Multiplexed with DHSDP
HHSDMB USB Host Port B High Speed Data - Analog VDDUTMII Multiplexed with DHSDM
USB Device High Speed Port - UDPHS
DFSDM USB Device Full Speed Data - Analog VDDUTMII
DFSDP USB Device Full Speed Data + Analog VDDUTMII
DHSDM USB Device High Speed Data - Analog VDDUTMII
DHSDP USB Device High Speed Data + Analog VDDUTMII
8 AT91SAM9G45
AT91SAM9G45

Active Reference
Ethernet 10/100
(1)
ETXCK Transmit Clock or Reference Clock Input MII only, REFCK in RMII
(1)
ERXCK Receive Clock Input MII only
(1)
ETXEN Transmit Enable Output
(1)
ETX0-ETX3 Transmit Data Output ETX0-ETX1 only in RMII
(1)
ETXER Transmit Coding Error Output MII only
(1)
ERXDV Receive Data Valid Input RXDV in MII, CRSDV in RMII
(1)
ERX0-ERX3 Receive Data Input ERX0-ERX1 only in RMII
(1)
ERXER Receive Error Input
(1)
ECRS Carrier Sense and Data Valid Input MII only
(1)
ECOL Collision Detect Input MII only
(1)
EMDC Management Data Clock Output
(1)
EMDIO Management Data Input/Output I/O
Image Sensor Interface
ISI_D0-ISI_D11 Image Sensor Data Input VDDIOP2
ISI_MCK Image sensor Reference clock output VDDIOP2
ISI_HSYNC Image Sensor Horizontal Synchro input VDDIOP2
ISI_VSYNC Image Sensor Vertical Synchro input VDDIOP2
ISI_PCK Image Sensor Data clock input VDDIOP2
LCD Controller - LCDC
LCDD0 -
LCD Data Bus Output VDDIOP1
LCDD23
LCDVSYNC LCD Vertical Synchronization Output VDDIOP1
LCDHSYNC LCD Horizontal Synchronization Output VDDIOP1
LCDDOTCK LCD Dot Clock Output VDDIOP1
LCDDEN LCD Data Enable Output VDDIOP1
LCDCC LCD Contrast Control Output VDDIOP1
LCDPWR LCD panel Power enable control Output VDDIOP1
LCDMOD LCD Modulation signal Output VDDIOP1
Touch Screen Analog-to-Digital Converter
AD0XP Touch Panel Right side Analog VDDANA Multiplexed with AD0
AD1XM Touch Panel Left side Analog VDDANA Multiplexed with AD1
AD2YP Touch Panel Top side Analog VDDANA Multiplexed with AD2
AD3YM Touch Panel Bottom side Analog VDDANA Multiplexed with AD3
9
Active Reference
GPAD4-GPAD7 Analog Inputs Analog VDDANA
TSADTRG ADC Trigger Input VDDANA
TSADVREF ADC Reference Analog VDDANA
Notes: 1. Refer to peripheral multiplexing tables in Section 8.4 “Peripheral Signals Multiplexing on I/O Lines” for these signals.
2. Programming of this pull-up resistor is performed independently for each I/O line through the PIO Controllers. After reset, all
the I/O lines default as inputs with pull-up resistors enabled, except those which are multiplexed with the External Bus Inter-
face signals that require to be enabled as Peripheral at reset. This is explicitly indicated in the column “Reset State” of the
peripheral multiplexing tables.
10 AT91SAM9G45
AT91SAM9G45
4. Package and Pinout

The AT91SAM9G45 is delivered in a 324-ball LFBGA package.
4.1 Mechanical Overview of the 324-ball LFBGA Package

Figure 4-1 shows the orientation of the 324-ball LFBGA Package
Figure 4-1. Orientation of the 324-ball LFBGA Package

Bottom VIEW
V
U
T
R
P
N
M
L
K
J
H
G
F
E
D
C
B
A
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
11
4.2 324-ball TFBGA Package Pinout
Table 4-1. AT91SAM9G45 Pinout for 324-ball BGA Package

Pin Signal Name Pin Signal Name Pin Signal Name Pin Signal Name
A1 PC27 E10 NANDWE K1 PE21 P10 TMS
A2 PC28 E11 DQS1 K2 PE23 P11 VDDPLLA
A3 PC25 E12 D13 K3 PE26 P12 PB20
A4 PC20 E13 D11 K4 PE22 P13 PB31
A5 PC12 E14 A4 K5 PE24 P14 DDR_D7
A6 PC7 E15 A8 K6 PE25 P15 DDR_D3
A9 NWR3/NBS3 E18 VDDCORE K9 VDDIOP0 P18 DDR_D10
A10 NCS0 F1 PD22 K10 VDDIOP0 R1 PA18
A11 DQS0 F2 PD24 K11 GNDIOM R2 PA20
A12 RAS F3 SHDN K12 GNDIOM R3 PA24
A13 SDCK F4 PE1 K13 VDDIOM0 R4 PA30
A14 NSDCK F5 PE3 K14 DDR_A7 R5 PB4
A15 D7 F6 VDDIOM1 K15 DDR_A8 R6 PB13
A16 DDR_VREF F7 PC19 K16 DDR_A9 R7 PD0
A17 D0 F8 PC14 K17 DDR_A11 R8 PD9
A18 A14 F9 PC4 K18 DDR_A10 R9 PD18
B1 PC31 F10 NCS1/SDCS L1 PA0 R10 TDI
B2 PC29 F11 NRD L2 PE30 R11 RTCK
B3 PC30 F12 SDWE L3 PE29 R12 PB22
B4 PC22 F13 A0/NBS0 L4 PE31 R13 PB29
B5 PC17 F14 A1/NBS2/NWR2 L5 PA2 R14 DDR_D6
B6 PC10 F15 A3 L6 PA4 R15 DDR_D1
B7 PC11 F16 A6 L7 PA8 R16 DDR_D0
B8 PC2 F17 A5 L8 PD2 R17 HHSDMA/DHSDM
B9 SDA10 F18 A2 L9 PD13 R18 HFSDMA/DFSDM
B10 A17/BA1 G1 PD25 L10 PD29 T1 PA22
B11 DQM0 G2 PD23 L11 PD31 T2 PA25
B12 SDCKE G3 PE6 L12 VDDIOM0 T3 PA26
B13 D12 G4 PE0 L13 VDDIOM0 T4 PB0
B14 D8 G5 PE2 L14 DDR_A1 T5 PB6
B15 D4 G6 PE8 L15 DDR_A3 T6 PB16
B16 D3 G7 PE4 L16 DDR_A4 T7 PD1
B17 A15 G8 PE11 L17 DDR_A6 T8 PD11
B18 A13 G9 GNDCORE L18 DDR_A5 T9 PD19
C1 XIN32 G10 VDDIOM1 M1 PA1 T10 PD30
C2 GNDANA G11 VDDIOM1 M2 PA5 T11 BMS
C3 WKUP G12 VDDCORE M3 PA6 T12 PB8
C4 PC26 G13 VDDCORE M4 PA7 T13 PB30
C5 PC21 G14 DDR_DQM0 M5 PA10 T14 DDR_D2
C6 PC15 G15 DDR_DQS1 M6 PA14 T15 PB21
C7 PC9 G16 DDR_BA1 M7 PB14 T16 PB23
C8 PC3 G17 DDR_BA0 M8 PD4 T17 HHSDPA/DHSDP
C9 NWR0/NWE G18 DDR_DQS0 M9 PD15 T18 HFSDPA/DFSDP
C10 A16/BA0 H1 PD26 M10 NRST U1 PA27
C11 CAS H2 PD27 M11 PB11 U2 PA29
C12 D15 H3 VDDIOP1 M12 PB25 U3 PA28
12 AT91SAM9G45
AT91SAM9G45
Table 4-1. AT91SAM9G45 Pinout for 324-ball BGA Package

Pin Signal Name Pin Signal Name Pin Signal Name Pin Signal Name
C13 D10 H4 PE13 M13 PB27 U4 PB3
C14 D6 H5 PE5 M14 VDDIOM0 U5 PB7
C15 D2 H6 PE7 M15 DDR_D14 U6 PB17
C16 GNDIOM H7 PE9 M16 DDR_D15 U7 PD7
C17 A18 H8 PE10 M17 DDR_A0 U8 PD10
C18 A12 H9 GNDCORE M18 DDR_A2 U9 PD14
D1 XOUT32 H10 GNDIOP N1 PA3 U10 TCK
D2 PD20 H11 VDDCORE N2 PA9 U11 VDDOSC
D3 GNDBU H12 GNDIOM N3 PA12 U12 GNDOSC
D4 VDDBU H13 GNDIOM N4 PA15 U13 PB10
D5 PC24 H14 DDR_CS N5 PA16 U14 PB26
D6 PC18 H15 DDR_WE N6 PA17 U15 HHSDPB
D7 PC13 H16 DDR_DQM1 N7 PB18 U16 HHSDMB
D8 PC6 H17 DDR_CAS N8 PD6 U17 GNDUTMI
D9 NWR1/NBS1 H18 DDR_NCLK N9 PD16 U18 VDDUTMIC
D10 NANDOE J1 PE19 N10 NTRST V1 PA31
D11 DQM1 J2 PE16 N11 PB9 V2 PB1
D12 D14 J3 PE14 N12 PB24 V3 PB2
D13 D9 J4 PE15 N13 PB28 V4 PB5
D14 D5 J5 PE12 N14 DDR_D13 V5 PB15
D15 D1 J6 PE17 N15 DDR_D8 V6 PD3
D16 VDDIOM1 J7 PE18 N16 DDR_D9 V7 PD5
D17 A11 J8 PE20 N17 DDR_D11 V8 PD12
D18 A10 J9 GNDCORE N18 DDR_D12 V9 PD17
E1 PD21 J10 GNDCORE P1 PA11 V10 TDO
E2 TSADVREF J11 GNDIOP P2 PA13 V11 XOUT
E3 VDDANA J12 GNDIOM P3 PA19 V12 XIN
E4 JTAGSEL J13 GNDIOM P4 PA21 V13 VDDPLLUTMI
E5 TST J14 DDR_A12 P5 PA23 V14 VDDIOP2
E6 PC23 J15 DDR_A13 P6 PB12 V15 HFSDPB
E7 PC16 J16 DDR_CKE P7 PB19 V16 HFSDMB
E8 PC8 J17 DDR_RAS P8 PD8 V17 VDDUTMII
E9 PC1 J18 DDR_CLK P9 PD28 V18 VBG
13
5. Power Considerations
5.1 Power Supplies

The AT91SAM9G45 has several types of power supply pins:
• VDDCORE pins: Power the core, including the processor, the embedded memories and the
peripherals; voltage ranges from 0.9V to 1.1V, 1.0V nominal.
• VDDIOM0 pins: Power the DDR2/LPDDR I/O lines; voltage ranges between 1.65V and 1.95V
(1.8V typical).
• VDDIOM1 pins: Power the External Bus Interface 1 I/O lines; voltage ranges between 1.65V
and 1.95V (1.8V typical) or between 3.0V and 3.6V (3.3V nominal).
• VDDIOP0, VDDIOP1, VDDIOP2 pins: Power the Peripherals I/O lines; voltage ranges from
1.65V to 3.6V.
• VDDBU pin: Powers the Slow Clock oscillator, the internal RC oscillator and a part of the
System Controller; voltage ranges from 1.8V to 3.6V.
• VDDPLLUTMI Powers the PLLUTMI cell; voltage range from 0.9V to 1.1V.
• VDDUTMIC pin: Powers the USB device and host UTMI+ core; voltage range from 0.9V to
1.1V, 1.0V nominal.
• VDDUTMII pin: Powers the USB device and host UTMI+ interface; voltage range from 3.0V to
3.6V, 3.3V nominal.
• VDDPLLA pin: Powers the PLLA cell; voltage ranges from 0.9V to 1.1V.
• VDDOSC pin: Powers the Main Oscillator cells; voltage ranges from 1.65V to 3.6V
• VDDANA pin: Powers the Analog to Digital Converter; voltage ranges from 3.0V to 3.6V, 3.3V
nominal.
Ground pins GND are common to VDDCORE, VDDIOM0, VDDIOM1, VDDIOP0, VDDIOP1 and
VDDIOP2 power supplies. Separated ground pins are provided for VDDUTMIC, VDDUTMII,
VDDBU, VDDOSC, VDDPLLA, VDDPLLUTMI and VDDANA. These ground pins are respec-
tively GNDUTMIC, GNDUTMII, GNDBU, GNDOSC, GNDPLLA, GNDPLLUTMI and GNDANA.
14 AT91SAM9G45
AT91SAM9G45
6. Memories
Figure 6-1. AT91SAM9G45 Memory Mapping
Address Memory Space Internal Memories System Controller
0x00000000 0x00000000 0xFFFF0000
Boot Memory Reserved
Internal Memories 0x00100000 0xFFFFE400
ITCM DDRSDRC1
0x10000000 0x00200000 0xFFFFE600
DTCM DDRSDRC0
EBI Chip Select 0 0x00300000 0xFFFFE800
SRAM SMC
0x20000000 0x00400000 0xFFFFEA00
EBI Chip Select 1 ROM MATRIX
0x00500000 0xFFFFEC00
DDR2-LPDDR-SDRAM LCDC DMAC
0x30000000 0x00600000 23 0xFFFFEE00 21
UDPHS (DMA) DBGU
EBI Chip Select 2 0x00700000 0xFFFFF000
UHP OHCI AIC
0x40000000 0x00800000 0xFFFFF200 0;31
UHP EHCI PIOA
EBI Chip Select 3 0x00900000 0xFFFFF400 2
NANDFlash Reserved PIOB
0x50000000 0x00A00000 0xFFFFF600 3
EBI Chip Select 4 Undefined (Abort) PIOC
0x0FFFFFFF 0xFFFFF800 4
Compact Flash Slot 0 PIOD
Internal Peripherals +5
0x60000000 0xF0000000 0xFFFFFA00
EBI Chip Select 5 Reserved PIOE
0xFFF78000 0xFFFFFC00 +5
Compact Flash Slot 1 UDPHS PMC
0x70000000 0xFFF7C000 27 0xFFFFFD00
TC0 TC0 SYSC
RSTC
DDR2-LPDDR +0x40 +18 +0x10 1
TC0 TC1 SYSC
Chip Select SHDWC
0x80000000 +0x80 +18 +0x20 1
TC0 TC2 SYSC
RTT
Undefined (Abort) 0xFFF80000 +0x30 1
HSMCI0 SYSC
PIT
0xF0000000 0xFFF84000 11 +0x40 1
TWI0 SYSC
WDT
Internal Peripherals 0xFFF88000 12 +0x50 1
TWI1 SYSC
SCKCR
0xFFFFFFFF 0xFFF8C000 13 +0x60 1
USART0 SYSC
GPBR
0xFFF90000 7 +0x70 1
SYSC
USART1 Reserved
offset 0xFFF94000 8 0xFFFFFDB0
block peripheral USART2
ID 9 RTC
0xFFF98000 0xFFFFFDC0
(+ : wired-or)
USART3 Reserved
0xFFF9C000 10 0xFFFFFFFF
SSC0
0xFFFA0000 16
SSC1
0xFFFA4000 17
SPI0
0xFFFA8000 14
SPI1
0xFFFAC000 15
AC97C
0xFFFB0000 24
TSADCC
0xFFFB4000 20
ISI
0xFFFB8000 26
PWM
0xFFFBC000 19
EMAC
0xFFFC0000 25
Reserved
0xFFFC4000
Reserved
0xFFFC8000
Reserved
0xFFFCC000
TRNG
0xFFFD0000 6
HSMCI1
0xFFFD4000 29
TC1 TC3
+0x40
TC1 TC4
+0x80
TC1 TC5
0xFFFD8000
Reserved
0xFFFFC000
System controller
0xFFFFFFFF
15
6.1 Memory Mapping
A first level of address decoding is performed by the AHB Bus Matrix, i.e., the implementation of
the Advanced High performance Bus (AHB) for its Master and Slave interfaces with additional
features.
Decoding breaks up the 4 Gbytes of address space into 16 banks of 256 Mbytes. The banks 1 to
6 are directed to the EBI that associates these banks to the external chip selects NCS0 to
NCS5.
The bank 7 is directed to the DDRLPDDR that associates this bank to DDR_NCS chip select
and so dedicated to the 4-port DDR2/ LPDDR controller.
The bank 0 is reserved for the addressing of the internal memories, and a second level of
decoding provides 1 Mbyte of internal memory area. The bank 15 is reserved for the peripherals
and provides access to the Advanced Peripheral Bus (APB).
Other areas are unused and performing an access within them provides an abort to the master
requesting such an access.
6.2 Embedded Memories
6.2.1 Internal SRAM

The AT91SAM9G45 product embeds a total of 64 Kbytes high-speed SRAM split in 4 blocks of
16 KBytes connected to one slave of the matrix. After reset and until the Remap Command is
performed, the four SRAM blocks are contiguous and only accessible at address 0x00300000.
After Remap, the SRAM also becomes available at address 0x0.
Figure 6-2. Internal SRAM Reset

RAM RAM
Remap
64K 64K
0x00300000 0x00000000
The AT91SAM9G45 device embeds two memory features. The processor Tightly Coupled Mem-
ory Interface (TCM) that allows the processor to access the memory up to processor speed
(PCK) and the interface on the AHB side allowing masters to access the memory at AHB speed
(MCK).
A wait state is necessary to access the TCM at 400 MHz. Setting the bit NWS_TCM in the bus
Matrix TCM Configuration Register of the matrix inserts a wait state on the ITCM and DTCM
accesses.
16 AT91SAM9G45
AT91SAM9G45
6.2.2 TCM Interface

On the processor side, this Internal SRAM can be allocated to two areas.
• Internal SRAM A is the ARM926EJ-S Instruction TCM. The user can map this SRAM block
anywhere in the ARM926 instruction memory space using CP15 instructions and the TCR
configuration register located in the Chip Configuration User Interface. This SRAM block is
also accessible by the ARM926 Masters and by the AHB Masters through the AHB bus
• Internal SRAM B is the ARM926EJ-S Data TCM. The user can map this SRAM block
anywhere in the ARM926 data memory space using CP15 instructions. This SRAM block is
also accessible by the ARM926 Data Master and by the AHB Masters through the AHB bus.
• Internal SRAM C is only accessible by all the AHB Masters. After reset and until the Remap
Command is performed, this SRAM block is accessible through the AHB bus at address
0x0030 0000 by all the AHB Masters. After Remap, this SRAM block also becomes
accessible through the AHB bus at address 0x0 by the ARM926 Instruction and the ARM926
Data Masters.
Figure 6-3. RAM Mapping

Processor Side AHB Side
(through TCM interface) (Multi-slave mode)
RAM
16K
16K
DTCM @ 0x00200000
0x00300000
(32K)
16K
16K
ITCM @ 0x00100000
(32K)
0
Within the 64 Kbyte SRAM size available, the amount of memory assigned to each block is soft-
ware programmable according to Table 6-1.
Table 6-1. ITCM and DTCM Memory Configuration

SRAMC AHB
SRAM A ITCM size (KBytes) SRAM B DTCM size (KBytes) (KBytes)
0 0 64
0 64 0
32 32 0
17
6.2.3 Internal ROM
The AT91SAM9G45 embeds an Internal ROM, which contains the boot ROM and SAM-BA®
program.
At any time, the ROM is mapped at address 0x0050 0000. It is also accessible at address 0x0
(BMS =1) after the reset and before the Remap Command.
7. System Controller
The System Controller is a set of peripherals that allows handling of key elements of the system,
such as power, resets, clocks, time, interrupts, watchdog, etc.
The System Controller User Interface also embeds the registers that configure the Matrix and a
set of registers for the chip configuration. The chip configuration registers configure the EBI chip
select assignment and voltage range for external memories.
7.1 System Controller Mapping

The System Controller’s peripherals are all mapped within the highest 16 KBytes of address
space, between addresses 0xFFFF E800 and 0xFFFF FFFF.
However, all the registers of the System Controller are mapped on the top of the address space.
All the registers of the System Controller can be addressed from a single pointer by using the
standard ARM instruction set, as the Load/Store instruction have an indexing mode of ±4 KB.
Figure 7-1 on page 19 shows the System Controller block diagram.
Figure 6-1 on page 15 shows the mapping of the User Interfaces of the System Controller
peripherals.
18 AT91SAM9G45
AT91SAM9G45
7.2 System Controller Block Diagram
Figure 7-1. AT91SAM9G45 System Controller Block Diagram

System Controller
VDDCORE Powered
irq0-irq2 nirq
fiq Advanced nfiq
Interrupt
periph_irq[2..24] Controller
pit_irq int
rtt_irq
wdt_irq ntrst
dbgu_irq por_ntrst ARM926EJ-S
pmc_irq
rstc_irq
MCK proc_nreset
periph_nreset Debug dbgu_irq
Unit dbgu_txd PCK
dbgu_rxd
debug
MCK Periodic
debug Interval pit_irq
periph_nreset Timer
jtag_nreset Boundary Scan
SLCK
debug Watchdog TAP Controller
idle wdt_irq
Timer
proc_nreset
MCK
wdt_fault
WDRPROC Bus Matrix
periph_nreset
NRST
rstc_irq
por_ntrst periph_nreset
VDDCORE jtag_nreset
POR Reset proc_nreset
Controller
backup_nreset
VDDBU
VDDBU Powered
VDDBU UPLLCK
POR SLCK
UHP48M
SLCK Real-Time rtc_irq
UHP12M
backup_nreset Clock rtc_alarm USB High Speed
Host Port
rtt_irq periph_nreset
SLCK Real-Time
backup_nreset Timer rtt_alarm periph_irq[25]
SLCK
SHDN
WKUP UPLLCK
Shut-Down
backup_nreset Controller
RC rtt0_alarm USB High Speed
periph_nreset Device Port
OSC 4 General-purpose
Backup Registers periph_irq[24]
XIN32 SLOW
CLOCK
XOUT32 OSC SCKCR
SLCK periph_clk[2..30]
int pck[0-1]
XIN UHP48M
12MHz MAINCK
UHP12M
XOUT MAIN OSC PCK
Power
MCK
UPLL UPLLCK Management
DDR sysclk
Controller
pmc_irq
PLLA PLLACK idle

periph_clk[6..30]
periph_nreset
periph_nreset
periph_nreset periph_irq[2..6] Embedded
periph_clk[2..6] irq Peripherals
dbgu_rxd PIO fiq periph_irq[6..30]
PA0-PA31 Controllers dbgu_txd
PB0-PB31 in
PC0-PC31 out
enable
PD0-PD31
PE0-PE31
19
7.3 Chip Identification
The AT91SAM9G45 Chip ID is defined in the Debug Unit Chip ID Register and Debug Unit Chip
ID Extension Register.
• Chip ID: 0x819B05A2
• Ext ID: 0x00000004
• JTAG ID: 05B2_703F
• ARM926 TAP ID: 0x0792603F
7.4 Backup Section

The AT91SAM9G45 features a Backup Section that embeds:
• RC Oscillator
• Slow Clock Oscillator
• SCKR register
• RTT
• RTC
• Shutdown Controller
• 4 backup registers
• A part of RSTC
This section is powered by the VDDBU rail.
20 AT91SAM9G45
AT91SAM9G45
8. Peripherals
8.1 Peripheral Mapping

As shown in Figure 6-1, the Peripherals are mapped in the upper 256 Mbytes of the address
space between the addresses 0xFFF7 8000 and 0xFFFC FFFF.
Each User Peripheral is allocated 16K bytes of address space.
8.2 Peripheral Identifiers

Table 8-1 defines the Peripheral Identifiers of the AT91SAM9G45. A peripheral identifier is
required for the control of the peripheral interrupt with the Advanced Interrupt Controller and for
the control of the peripheral clock with the Power Management Controller.
Table 8-1. AT91SAM9G45 Peripheral Identifiers

Peripheral ID Peripheral Mnemonic Peripheral Name External Interrupt
0 AIC Advanced Interrupt Controller FIQ
1 SYSC System Controller Interrupt
2 PIOA Parallel I/O Controller A,
3 PIOB Parallel I/O Controller B
4 PIOC Parallel I/O Controller C
5 PIOD/PIOE Parallel I/O Controller D/E
6 RNG True Random Number Generator
7 US0 USART 0
8 US1 USART 1
9 US2 USART 2
10 US3 USART 3
11 MCI0 High Speed Multimedia Card Interface 0
12 TWI0 Two-Wire Interface 0
13 TWI1 Two-Wire Interface 1
14 SPI0 Serial Peripheral Interface
15 SPI1 Serial Peripheral Interface
16 SSC0 Synchronous Serial Controller 0
17 SSC1 Synchronous Serial Controller 1
18 TC0..TC5 Timer Counter 0,1,2,3,4,5
19 PWMC Pulse Width Modulation Controller
20 TSADCC Touch Screen ADC Controller
21 DMA DMA Controller
22 UHPHS USB Host High Speed
23 LCDC LCD Controller
24 AC97 AC97 Controller
25 EMAC Ethernet MAC
26 ISI Image Sensor Interface
27 UDPHS USB Device High Speed
29 MCI1 High Speed Multimedia Card Interface 1
30 Reserved
31 AIC Advanced Interrupt Controller IRQ
21
8.3 Peripheral Interrupts and Clock Control
8.3.1 System Interrupt

The System Interrupt in Source 1 is the wired-OR of the interrupt signals coming from:
• the DDR2/LPDDR Controller
• the Debug Unit
• the Periodic Interval Timer
• the Real-Time Timer
• the Real-Time Clock
• the Watchdog Timer
• the Reset Controller
• the Power Management Controller
The clock of these peripherals cannot be deactivated and Peripheral ID 1 can only be used
within the Advanced Interrupt Controller.
8.3.2 External Interrupts

All external interrupt signals, i.e., the Fast Interrupt signal FIQ or the Interrupt signal IRQ, use a
dedicated Peripheral ID. However, there is no clock control associated with these peripheral IDs.
8.4 Peripheral Signals Multiplexing on I/O Lines

The AT91SAM9G45 features 5 PIO controllers, PIOA, PIOB, PIOC, PIOD and PIOE, which mul-
tiplexes the I/O lines of the peripheral set.
Each PIO Controller controls up to 32 lines. Each line can be assigned to one of two peripheral
functions, A or B. The multiplexing tables in the following paragraphs define how the I/O lines of
the peripherals A and B are multiplexed on the PIO Controllers. The two columns “Function” and
“Comments” have been inserted in this table for the user’s own comments; they may be used to
track how pins are defined in an application.
Note that some peripheral function which are output only, might be duplicated within the both
tables.
The column “Reset State” indicates whether the PIO Line resets in I/O mode or in peripheral
mode. If I/O is mentioned, the PIO Line resets in input with the pull-up enabled, so that the
device is maintained in a static state as soon as the reset is released. As a result, the bit corre-
sponding to the PIO Line in the register PIO_PSR (Peripheral Status Register) resets low.
If a signal name is mentioned in the “Reset State” column, the PIO Line is assigned to this func-
tion and the corresponding bit in PIO_PSR resets high. This is the case of pins controlling
memories, in particular the address lines, which require the pin to be driven as soon as the reset
is released. Note that the pull-up resistor is also enabled in this case.
To amend EMC, programmable delay has been inserted on PIO lines able to run at high speed.
22 AT91SAM9G45
AT91SAM9G45
8.4.1 PIO Controller A Multiplexing
Table 8-2. Multiplexing on PIO Controller A (PIOA)

Reset Power
I/O Line Peripheral A Peripheral B State Supply Function Comments
PA0 MCI0_CK TCLK3 I/O VDDIOP0
PA1 MCI0_CDA TIOA3 I/O VDDIOP0
PA2 MCI0_DA0 TIOB3 I/O VDDIOP0
PA3 MCI0_DA1 TCKL4 I/O VDDIOP0
PA4 MCI0_DA2 TIOA4 I/O VDDIOP0
PA6 MCI0_DA4 ETX2 I/O VDDIOP0
PA7 MCI0_DA5 ETX3 I/O VDDIOP0
PA8 MCI0_DA6 ERX2 I/O VDDIOP0
PA9 MCI0_DA7 ERX3 I/O VDDIOP0
PA10 ETX0 I/O VDDIOP0
PA11 ETX1 I/O VDDIOP0
PA12 ERX0 I/O VDDIOP0
PA13 ERX1 I/O VDDIOP0
PA14 ETXEN I/O VDDIOP0
PA15 ERXDV I/O VDDIOP0
PA16 ERXER I/O VDDIOP0
PA17 ETXCK I/O VDDIOP0
PA18 EMDC I/O VDDIOP0
PA19 EMDIO I/O VDDIOP0
PA20 TWD0 I/O VDDIOP0
PA21 TWCK0 I/O VDDIOP0
PA22 MCI1_CDA SCK3 I/O VDDIOP0
PA23 MCI1_DA0 RTS3 I/O VDDIOP0
PA24 MCI1_DA1 CTS3 I/O VDDIOP0
PA25 MCI1_DA2 PWM3 I/O VDDIOP0
PA27 MCI1_DA4 ETXER I/O VDDIOP0
PA28 MCI1_DA5 ERXCK I/O VDDIOP0
PA29 MCI1_DA6 ECRS I/O VDDIOP0
PA30 MCI1_DA7 ECOL I/O VDDIOP0
PA31 MCI1_CK PCK0 I/O VDDIOP0
23
8.4.2 PIO Controller B Multiplexing
Table 8-3. Multiplexing on PIO Controller B (PIOB)

Reset Power
PB0 SPI0_MISO I/O VDDIOP0
PB1 SPI0_MOSI I/O VDDIOP0
PB2 SPI0_SPCK I/O VDDIOP0
PB3 SPI0_NPCS0 I/O VDDIOP0
PB4 TXD1 I/O VDDIOP0
PB5 RXD1 I/O VDDIOP0
PB6 TXD2 I/O VDDIOP0
PB7 RXD2 I/O VDDIOP0
PB8 TXD3 ISI_D8 I/O VDDIOP2
PB9 RXD3 ISI_D9 I/O VDDIOP2
PB10 TWD1 ISI_D10 I/O VDDIOP2
PB11 TWCK1 ISI_D11 I/O VDDIOP2
PB12 DRXD I/O VDDIOP0
PB13 DTXD I/O VDDIOP0
PB14 SPI1_MISO I/O VDDIOP0
PB15 SPI1_MOSI CTS0 I/O VDDIOP0
PB16 SPI1_SPCK SCK0 I/O VDDIOP0
PB17 SPI1_NPCS0 RTS0 I/O VDDIOP0
PB18 RXD0 SPI0_NPCS1 I/O VDDIOP0
PB19 TXD0 SPI0_NPCS2 I/O VDDIOP0
PB20 ISI_D0 I/O VDDIOP2
PB28 ISI_PCK I/O VDDIOP2
PB29 ISI_VSYNC I/O VDDIOP2
PB30 ISI_HSYNC I/O VDDIOP2
PB31 ISI_MCK PCK1 I/O VDDIOP2
24 AT91SAM9G45
AT91SAM9G45
8.4.3 PIO Controller C Multiplexing
Table 8-4. Multiplexing on PIO Controller C (PIOC)

Reset Power
PC0 DQM2 DQM2 VDDIOM1
PC1 DQM3 DQM3 VDDIOM1
PC2 A19 A19 VDDIOM1
PC3 A20 A20 VDDIOM1
PC4 A21/NANDALE A21 VDDIOM1
PC5 A22/NANDCLE A22 VDDIOM1
PC6 A23 A23 VDDIOM1
PC7 A24 A24 VDDIOM1
PC8 CFCE1 I/O VDDIOM1
PC9 CFCE2 RTS2 I/O VDDIOM1
PC10 NCS4/CFCS0 TCLK2 I/O VDDIOM1
PC11 NCS5/CFCS1 CTS2 I/O VDDIOM1
PC12 A25/CFRNW A25 VDDIOM1
PC13 NCS2 I/O VDDIOM1
PC14 NCS3/NANDCS I/O VDDIOM1
PC15 NWAIT I/O VDDIOM1
PC16 D16 I/O VDDIOM1
25
8.4.4 PIO Controller D Multiplexing
Table 8-5. Multiplexing on PIO Controller D (PIOD)

Reset Power
PD0 TK0 PWM3 I/O VDDIOP0
PD1 TF0 I/O VDDIOP0
PD2 TD0 I/O VDDIOP0
PD3 RD0 I/O VDDIOP0
PD4 RK0 I/O VDDIOP0
PD5 RF0 I/O VDDIOP0
PD6 AC97RX I/O VDDIOP0
PD7 AC97TX TIOA5 I/O VDDIOP0
PD8 AC97FS TIOB5 I/O VDDIOP0
PD9 AC97CK TCLK5 I/O VDDIOP0
PD10 TD1 I/O VDDIOP0
PD11 RD1 I/O VDDIOP0
PD12 TK1 PCK0 I/O VDDIOP0
PD13 RK1 I/O VDDIOP0
PD14 TF1 I/O VDDIOP0
PD15 RF1 I/O VDDIOP0
PD16 RTS1 I/O VDDIOP0
PD17 CTS1 I/O VDDIOP0
PD18 SPI1_NPCS2 IRQ I/O VDDIOP0
PD19 SPI1_NPCS3 FIQ I/O VDDIOP0
PD20 TIOA0 I/O VDDANA TSAD0
PD23 TCLK0 I/O VDDANA TSAD3
PD24 SPI0_NPCS1 PWM0 I/O VDDANA GPAD4
PD25 SPI0_NPCS2 PWM1 I/O VDDANA GPAD5
PD26 PCK0 PWM2 I/O VDDANA GPAD6
PD27 PCK1 SPI0_NPCS3 I/O VDDANA GPAD7
PD28 TSADTRG SPI1_NPCS1 I/O VDDIOP0
PD29 TCLK1 SCK1 I/O VDDIOP0
PD30 TIOB0 SCK2 I/O VDDIOP0
PD31 TIOB1 PWM1 I/O VDDIOP0
26 AT91SAM9G45
AT91SAM9G45
8.4.5 PIO Controller E Multiplexing
Table 8-6. Multiplexing on PIO Controller E (PIOE)

Reset Power
PE0 LCDPWR PCK0 I/O VDDIOP1
PE1 LCDMOD I/O VDDIOP1
PE2 LCDCC I/O VDDIOP1
PE3 LCDVSYNC I/O VDDIOP1
PE4 LCDHSYNC I/O VDDIOP1
PE5 LCDDOTCK I/O VDDIOP1
PE6 LCDDEN I/O VDDIOP1
PE7 LCDD0 LCDD2 I/O VDDIOP1
PE25 LCDD18 I/O VDDIOP1
PE31 PWM2 PCK1 I/O VDDIOP1
27
28 AT91SAM9G45
AT91SAM9G45
9. ARM926EJ-S Processor Overview
9.1 Description
The ARM926EJ-S™ processor is a member of the ARM9™ family of general-purpose micropro-
cessors. The ARM926EJ-S implements ARM architecture version 5TEJ and is targeted at multi-
tasking applications where full memory management, high performance, low die size and low
power are all important features.
The ARM926EJ-S processor supports the 32-bit ARM and 16-bit THUMB instruction sets,
enabling the user to trade off between high performance and high code density. It also supports
8-bit Java instruction set and includes features for efficient execution of Java bytecode, provid-
ing a Java performance similar to a JIT (Just-In-Time compilers), for the next generation of Java-
powered wireless and embedded devices. It includes an enhanced multiplier design for
improved DSP performance.
The ARM926EJ-S processor supports the ARM debug architecture and includes logic to assist
in both hardware and software debug.
The ARM926EJ-S provides a complete high performance processor subsystem, including:
• an ARM9EJ-S integer core
• a Memory Management Unit (MMU)
• separate instruction and data AMBA AHB bus interfaces
• separate instruction and data TCM interfaces
29
9.2 Embedded Characteristics
• RISC Processor Based on ARM v5TEJ Architecture with Jazelle technology for Java
acceleration
• Two Instruction Sets
– ARM High-performance 32-bit Instruction Set
– Thumb High Code Density 16-bit Instruction Set
• DSP Instruction Extensions
• 5-Stage Pipeline Architecture:
– Instruction Fetch (F)
– Instruction Decode (D)
– Execute (E)
– Data Memory (M)
– Register Write (W)
• 32-KByte Data Cache, 32-KByte Instruction Cache
– Virtually-addressed 4-way Associative Cache
– Eight words per line
– Write-through and Write-back Operation
– Pseudo-random or Round-robin Replacement
• Write Buffer
– Main Write Buffer with 16-word Data Buffer and 4-address Buffer
– DCache Write-back Buffer with 8-word Entries and a Single Address Entry
– Software Control Drain
• Standard ARM v4 and v5 Memory Management Unit (MMU)
– Access Permission for Sections
– Access Permission for large pages and small pages can be specified separately for
each quarter of the page
– 16 embedded domains
• Bus Interface Unit (BIU)
– Arbitrates and Schedules AHB Requests
– Separate Masters for both instruction and data access providing complete Matrix
system flexibility
– Separate Address and Data Buses for both the 32-bit instruction interface and the
32-bit data interface
– On Address and Data Buses, data can be 8-bit (Bytes), 16-bit (Half-words) or 32-bit
(Words)
• TCM Interface
30 AT91SAM9G45
AT91SAM9G45
9.3 Block Diagram
Figure 9-1. ARM926EJ-S Internal Functional Block Diagram

External Coprocessors ETM9
CP15 System External Trace Port

Configuration Coprocessor Interface
Coprocessor Interface
Write Data
ARM9EJ-S
Processor Core
Instruction
Read Fetches
Data
Data Instruction
Address Address
MMU
Instruction
DTCM Data TLB TLB ITCM
Interface Interface
Data TCM Instruction TCM
Data Instruction
Address Address
AHB Interface Instruction
Data Cache and Cache
Write Buffer
AMBA AHB
31
9.4 ARM9EJ-S Processor
9.4.1 ARM9EJ-S Operating States

The ARM9EJ-S processor can operate in three different states, each with a specific instruction
set:
• ARM state: 32-bit, word-aligned ARM instructions.
• THUMB state: 16-bit, halfword-aligned Thumb instructions.
• Jazelle state: variable length, byte-aligned Jazelle instructions.
In Jazelle state, all instruction Fetches are in words.
9.4.2 Switching State

The operating state of the ARM9EJ-S core can be switched between:
• ARM state and THUMB state using the BX and BLX instructions, and loads to the PC
• ARM state and Jazelle state using the BXJ instruction
All exceptions are entered, handled and exited in ARM state. If an exception occurs in Thumb or
Jazelle states, the processor reverts to ARM state. The transition back to Thumb or Jazelle
states occurs automatically on return from the exception handler.
9.4.3 Instruction Pipelines

The ARM9EJ-S core uses two kinds of pipelines to increase the speed of the flow of instructions
to the processor.
A five-stage (five clock cycles) pipeline is used for ARM and Thumb states. It consists of Fetch,
Decode, Execute, Memory and Writeback stages.
A six-stage (six clock cycles) pipeline is used for Jazelle state It consists of Fetch,
Jazelle/Decode (two clock cycles), Execute, Memory and Writeback stages.
9.4.4 Memory Access

The ARM9EJ-S core supports byte (8-bit), half-word (16-bit) and word (32-bit) access. Words
must be aligned to four-byte boundaries, half-words must be aligned to two-byte boundaries and
bytes can be placed on any byte boundary.
Because of the nature of the pipelines, it is possible for a value to be required for use before it
has been placed in the register bank by the actions of an earlier instruction. The ARM9EJ-S con-
trol logic automatically detects these cases and stalls the core or forward data.
9.4.5 Jazelle Technology

The Jazelle technology enables direct and efficient execution of Java byte codes on ARM pro-
cessors, providing high performance for the next generation of Java-powered wireless and
embedded devices.
The new Java feature of ARM9EJ-S can be described as a hardware emulation of a JVM (Java
Virtual Machine). Java mode will appear as another state: instead of executing ARM or Thumb
instructions, it executes Java byte codes. The Java byte code decoder logic implemented in
ARM9EJ-S decodes 95% of executed byte codes and turns them into ARM instructions without
any overhead, while less frequently used byte codes are broken down into optimized sequences
of ARM instructions. The hardware/software split is invisible to the programmer, invisible to the
application and invisible to the operating system. All existing ARM registers are re-used in
Jazelle state and all registers then have particular functions in this mode.
32 AT91SAM9G45
AT91SAM9G45
Minimum interrupt latency is maintained across both ARM state and Java state. Since byte
codes execution can be restarted, an interrupt automatically triggers the core to switch from
Java state to ARM state for the execution of the interrupt handler. This means that no special
provision has to be made for handling interrupts while executing byte codes, whether in hard-
ware or in software.
9.4.6 ARM9EJ-S Operating Modes

In all states, there are seven operation modes:
• User mode is the usual ARM program execution state. It is used for executing most
application programs
• Fast Interrupt (FIQ) mode is used for handling fast interrupts. It is suitable for high-speed data
transfer or channel process
• Interrupt (IRQ) mode is used for general-purpose interrupt handling
• Supervisor mode is a protected mode for the operating system
• Abort mode is entered after a data or instruction prefetch abort
• System mode is a privileged user mode for the operating system
• Undefined mode is entered when an undefined instruction exception occurs
Mode changes may be made under software control, or may be brought about by external inter-
rupts or exception processing. Most application programs execute in User Mode. The non-user
modes, known as privileged modes, are entered in order to service interrupts or exceptions or to
access protected resources.
9.4.7 ARM9EJ-S Registers

The ARM9EJ-S core has a total of 37 registers.
• 31 general-purpose 32-bit registers
• 6 32-bit status registers
Table 9-1 shows all the registers in all modes.
Table 9-1. ARM9TDMI Modes and Registers Layout

User and Supervisor Undefined Interrupt Fast Interrupt
System Mode Mode Abort Mode Mode Mode Mode
R0 R0 R0 R0 R0 R0
R1 R1 R1 R1 R1 R1
R2 R2 R2 R2 R2 R2
R3 R3 R3 R3 R3 R3
R4 R4 R4 R4 R4 R4
R5 R5 R5 R5 R5 R5
R6 R6 R6 R6 R6 R6
R7 R7 R7 R7 R7 R7
R8 R8 R8 R8 R8 R8_FIQ
R9 R9 R9 R9 R9 R9_FIQ
R10 R10 R10 R10 R10 R10_FIQ
R11 R11 R11 R11 R11 R11_FIQ
33
Table 9-1. ARM9TDMI Modes and Registers Layout
User and Supervisor Undefined Interrupt Fast Interrupt
System Mode Mode Abort Mode Mode Mode Mode
R12 R12 R12 R12 R12 R12_FIQ
R13 R13_SVC R13_ABORT R13_UNDEF R13_IRQ R13_FIQ
R14 R14_SVC R14_ABORT R14_UNDEF R14_IRQ R14_FIQ
PC PC PC PC PC PC
CPSR CPSR CPSR CPSR CPSR CPSR

SPSR_ABOR SPSR_UNDE
SPSR_SVC SPSR_IRQ SPSR_FIQ
T F
Mode-specific banked registers
The ARM state register set contains 16 directly-accessible registers, r0 to r15, and an additional
register, the Current Program Status Register (CPSR). Registers r0 to r13 are general-purpose
registers used to hold either data or address values. Register r14 is used as a Link register that
holds a value (return address) of r15 when BL or BLX is executed. Register r15 is used as a pro-
gram counter (PC), whereas the Current Program Status Register (CPSR) contains condition
code flags and the current mode bits.
In privileged modes (FIQ, Supervisor, Abort, IRQ, Undefined), mode-specific banked registers
(r8 to r14 in FIQ mode or r13 to r14 in the other modes) become available. The corresponding
banked registers r14_fiq, r14_svc, r14_abt, r14_irq, r14_und are similarly used to hold the val-
ues (return address for each mode) of r15 (PC) when interrupts and exceptions arise, or when
BL or BLX instructions are executed within interrupt or exception routines. There is another reg-
ister called Saved Program Status Register (SPSR) that becomes available in privileged modes
instead of CPSR. This register contains condition code flags and the current mode bits saved as
a result of the exception that caused entry to the current (privileged) mode.
In all modes and due to a software agreement, register r13 is used as stack pointer.
The use and the function of all the registers described above should obey ARM Procedure Call
Standard (APCS) which defines:
• constraints on the use of registers
• stack conventions
• argument passing and result return
For more details, refer to ARM Software Development Kit.
The Thumb state register set is a subset of the ARM state set. The programmer has direct
access to:
• Eight general-purpose registers r0-r7
• Stack pointer, SP
• Link register, LR (ARM r14)
• PC
34 AT91SAM9G45
AT91SAM9G45
• CPSR
There are banked registers SPs, LRs and SPSRs for each privileged mode (for more details see
the ARM9EJ-S Technical Reference Manual, revision r1p2 page 2-12).
9.4.7.1 Status Registers

The ARM9EJ-S core contains one CPSR, and five SPSRs for exception handlers to use. The
program status registers:
• hold information about the most recently performed ALU operation
• control the enabling and disabling of interrupts
• set the processor operation mode
Figure 9-2. Status Register Format

31 30 29 28 27 24 7 6 5 0
N Z C V Q J Reserved I F T Mode
Jazelle state bit

Reserved Mode bits
Sticky Overflow
Overflow Thumb state bit
Carry/Borrow/Extend
Zero FIQ disable
Negative/Less than
IRQ disable
Figure 9-2 shows the status register format, where:

• N: Negative, Z: Zero, C: Carry, and V: Overflow are the four ALU flags
• The Sticky Overflow (Q) flag can be set by certain multiply and fractional arithmetic
instructions like QADD, QDADD, QSUB, QDSUB, SMLAxy, and SMLAWy needed to achieve
DSP operations.
The Q flag is sticky in that, when set by an instruction, it remains set until explicitly cleared by
an MSR instruction writing to the CPSR. Instructions cannot execute conditionally on the
status of the Q flag.
• The J bit in the CPSR indicates when the ARM9EJ-S core is in Jazelle state, where:
– J = 0: The processor is in ARM or Thumb state, depending on the T bit
– J = 1: The processor is in Jazelle state.
• Mode: five bits to encode the current processor mode
9.4.7.2 Exceptions
9.4.7.3 Exception Types and Priorities

The ARM9EJ-S supports five types of exceptions. Each type drives the ARM9EJ-S in a privi-
leged mode. The types of exceptions are:
• Fast interrupt (FIQ)
• Normal interrupt (IRQ)
• Data and Prefetched aborts (Abort)
• Undefined instruction (Undefined)
• Software interrupt and Reset (Supervisor)
35
When an exception occurs, the banked version of R14 and the SPSR for the exception mode
are used to save the state.
More than one exception can happen at a time, therefore the ARM9EJ-S takes the arisen excep-
tions according to the following priority order:
• Reset (highest priority)
• Data Abort
• FIQ
• IRQ
• Prefetch Abort
• BKPT, Undefined instruction, and Software Interrupt (SWI) (Lowest priority)
The BKPT, or Undefined instruction, and SWI exceptions are mutually exclusive.
Note that there is one exception in the priority scheme: when FIQs are enabled and a Data Abort
occurs at the same time as an FIQ, the ARM9EJ-S core enters the Data Abort handler, and pro-
ceeds immediately to FIQ vector. A normal return from the FIQ causes the Data Abort handler to
resume execution. Data Aborts must have higher priority than FIQs to ensure that the transfer
error does not escape detection.
9.4.7.4 Exception Modes and Handling

Exceptions arise whenever the normal flow of a program must be halted temporarily, for exam-
ple, to service an interrupt from a peripheral.
When handling an ARM exception, the ARM9EJ-S core performs the following operations:
1. Preserves the address of the next instruction in the appropriate Link Register that cor-
responds to the new mode that has been entered. When the exception entry is from:
– ARM and Jazelle states, the ARM9EJ-S copies the address of the next instruction
into LR (current PC(r15) + 4 or PC + 8 depending on the exception).
– THUMB state, the ARM9EJ-S writes the value of the PC into LR, offset by a value
(current PC + 2, PC + 4 or PC + 8 depending on the exception) that causes the
program to resume from the correct place on return.
2. Copies the CPSR into the appropriate SPSR.
3. Forces the CPSR mode bits to a value that depends on the exception.
4. Forces the PC to fetch the next instruction from the relevant exception vector.
The register r13 is also banked across exception modes to provide each exception handler with
private stack pointer.
The ARM9EJ-S can also set the interrupt disable flags to prevent otherwise unmanageable
nesting of exceptions.
When an exception has completed, the exception handler must move both the return value in
the banked LR minus an offset to the PC and the SPSR to the CPSR. The offset value varies
according to the type of exception. This action restores both PC and the CPSR.
The fast interrupt mode has seven private registers r8 to r14 (banked registers) to reduce or
remove the requirement for register saving which minimizes the overhead of context switching.
The Prefetch Abort is one of the aborts that indicates that the current memory access cannot be
completed. When a Prefetch Abort occurs, the ARM9EJ-S marks the prefetched instruction as
invalid, but does not take the exception until the instruction reaches the Execute stage in the
36 AT91SAM9G45
AT91SAM9G45
pipeline. If the instruction is not executed, for example because a branch occurs while it is in the
pipeline, the abort does not take place.
The breakpoint (BKPT) instruction is a new feature of ARM9EJ-S that is destined to solve the
problem of the Prefetch Abort. A breakpoint instruction operates as though the instruction
caused a Prefetch Abort.
A breakpoint instruction does not cause the ARM9EJ-S to take the Prefetch Abort exception until
the instruction reaches the Execute stage of the pipeline. If the instruction is not executed, for
example because a branch occurs while it is in the pipeline, the breakpoint does not take place.
9.4.8 ARM Instruction Set Overview

The ARM instruction set is divided into:
• Branch instructions
• Data processing instructions
• Status register transfer instructions
• Load and Store instructions
• Coprocessor instructions
• Exception-generating instructions
ARM instructions can be executed conditionally. Every instruction contains a 4-bit condition
code field (bits[31:28]).
For further details, see the ARM Technical Reference Manual.
Table 9-2 gives the ARM instruction mnemonic list.
Table 9-2. ARM Instruction Mnemonic List

Mnemonic Operation Mnemonic Operation
MOV Move MVN Move Not
ADD Add ADC Add with Carry
SUB Subtract SBC Subtract with Carry

RSB Reverse Subtract RSC Reverse Subtract with Carry
CMP Compare CMN Compare Negated
TST Test TEQ Test Equivalence
AND Logical AND BIC Bit Clear
EOR Logical Exclusive OR ORR Logical (inclusive) OR
MUL Multiply MLA Multiply Accumulate
SMULL Sign Long Multiply UMULL Unsigned Long Multiply
Unsigned Long Multiply

SMLAL Signed Long Multiply Accumulate UMLAL
Accumulate
MSR Move to Status Register MRS Move From Status Register
B Branch BL Branch and Link
BX Branch and Exchange SWI Software Interrupt
LDR Load Word STR Store Word
LDRSH Load Signed Halfword
LDRSB Load Signed Byte
37
Table 9-2. ARM Instruction Mnemonic List (Continued)
LDRH Load Half Word STRH Store Half Word
LDRB Load Byte STRB Store Byte
Load Register Byte with Store Register Byte with

LDRBT STRBT
Translation Translation
LDRT Load Register with Translation STRT Store Register with Translation
LDM Load Multiple STM Store Multiple
SWP Swap Word SWPB Swap Byte
MCR Move To Coprocessor MRC Move From Coprocessor
LDC Load To Coprocessor STC Store From Coprocessor
CDP Coprocessor Data Processing
9.4.9 New ARM Instruction Set

.
Table 9-3. New ARM Instruction Mnemonic List

BXJ Branch and exchange to Java MRRC Move double from coprocessor
Alternative move of ARM reg to
BLX (1) Branch, Link and exchange MCR2
coprocessor
Signed Multiply Accumulate 16 *

SMLAxy MCRR Move double to coprocessor
16 bit
Alternative Coprocessor Data

SMLAL Signed Multiply Accumulate Long CDP2
Processing
Signed Multiply Accumulate 32 *
SMLAWy BKPT Breakpoint
16 bit
Soft Preload, Memory prepare to

SMULxy Signed Multiply 16 * 16 bit PLD
load from address
SMULWy Signed Multiply 32 * 16 bit STRD Store Double
Alternative Store from

QADD Saturated Add STC2
Coprocessor
QDADD Saturated Add with Double LDRD Load Double
QSUB Saturated subtract LDC2 Alternative Load to Coprocessor

QDSUB Saturated Subtract with double CLZ Count Leading Zeroes
Notes: 1. A Thumb BLX contains two consecutive Thumb instructions, and takes four cycles.
9.4.10 Thumb Instruction Set Overview

The Thumb instruction set is a re-encoded subset of the ARM instruction set.
The Thumb instruction set is divided into:
• Branch instructions
• Data processing instructions
• Load and Store instructions
• Load and Store multiple instructions
38 AT91SAM9G45
AT91SAM9G45
• Exception-generating instruction
Table 5 shows the Thumb instruction set, for further details, see the ARM Technical Reference
Manual.
Table 9-4 gives the Thumb instruction mnemonic list.
Table 9-4. Thumb Instruction Mnemonic List

MOV Move MVN Move Not
ADD Add ADC Add with Carry
SUB Subtract SBC Subtract with Carry
CMP Compare CMN Compare Negated

TST Test NEG Negate
AND Logical AND BIC Bit Clear
EOR Logical Exclusive OR ORR Logical (inclusive) OR
LSL Logical Shift Left LSR Logical Shift Right
ASR Arithmetic Shift Right ROR Rotate Right
MUL Multiply BLX Branch, Link, and Exchange
B Branch BL Branch and Link
BX Branch and Exchange SWI Software Interrupt
LDR Load Word STR Store Word
LDRH Load Half Word STRH Store Half Word

LDRB Load Byte STRB Store Byte
LDRSH Load Signed Halfword LDRSB Load Signed Byte
LDMIA Load Multiple STMIA Store Multiple

PUSH Push Register to stack POP Pop Register from stack
BCC Conditional Branch BKPT Breakpoint
39
9.5 CP15 Coprocessor
Coprocessor 15, or System Control Coprocessor CP15, is used to configure and control all the
items in the list below:
• ARM9EJ-S
• Caches (ICache, DCache and write buffer)
• TCM
• MMU
• Other system options
To control these features, CP15 provides 16 additional registers. See Table 9-5.
Table 9-5. CP15 Registers

Register Name Read/Write
(1)
0 ID Code Read/Unpredictable
(1)
0 Cache type Read/Unpredictable
0 TCM status(1) Read/Unpredictable
1 Control Read/write
2 Translation Table Base Read/write
3 Domain Access Control Read/write
4 Reserved None
(1)
5 Data fault Status Read/write
(1)
5 Instruction fault status Read/write
6 Fault Address Read/write
7 Cache Operations Read/Write
8 TLB operations Unpredictable/Write
(2)
9 cache lockdown Read/write
9 TCM region Read/write
10 TLB lockdown Read/write
11 Reserved None
12 Reserved None
(1)
13 FCSE PID Read/write
13 Context ID(1) Read/Write
14 Reserved None
15 Test configuration Read/Write
Notes: 1. Register locations 0,5, and 13 each provide access to more than one register. The register
accessed depends on the value of the opcode_2 field.
2. Register location 9 provides access to more than one register. The register accessed depends
on the value of the CRm field.
40 AT91SAM9G45
AT91SAM9G45
9.5.1 CP15 Registers Access

CP15 registers can only be accessed in privileged mode by:
• MCR (Move to Coprocessor from ARM Register) instruction is used to write an ARM register
to CP15.
• MRC (Move to ARM Register from Coprocessor) instruction is used to read the value of
CP15 to an ARM register.
Other instructions like CDP, LDC, STC can cause an undefined instruction exception.
The assembler code for these instructions is:
MCR/MRC{cond} p15, opcode_1, Rd, CRn, CRm, opcode_2.
The MCR, MRC instructions bit pattern is shown below:
31 30 29 28 27 26 25 24
cond 1 1 1 0
23 22 21 20 19 18 17 16
opcode_1 L CRn
15 14 13 12 11 10 9 8
Rd 1 1 1 1
7 6 5 4 3 2 1 0
opcode_2 1 CRm
• CRm[3:0]: Specified Coprocessor Action

Determines specific coprocessor action. Its value is dependent on the CP15 register used. For details, refer to CP15 spe-
cific register behavior.
• opcode_2[7:5]
Determines specific coprocessor operation code. By default, set to 0.
• Rd[15:12]: ARM Register
Defines the ARM register whose value is transferred to the coprocessor. If R15 is chosen, the result is unpredictable.
• CRn[19:16]: Coprocessor Register
Determines the destination coprocessor register.
• L: Instruction Bit
0 = MCR instruction
1 = MRC instruction
• opcode_1[23:20]: Coprocessor Code
Defines the coprocessor specific code. Value is c15 for CP15.
• cond [31:28]: Condition
For more details, see Chapter 2 in ARM926EJ-S TRM.
41
9.6 Memory Management Unit (MMU)
The ARM926EJ-S processor implements an enhanced ARM architecture v5 MMU to provide vir-
tual memory features required by operating systems like Symbian OS®, Windows CE®, and
Linux®. These virtual memory features are memory access permission controls and virtual to
physical address translations.
The Virtual Address generated by the CPU core is converted to a Modified Virtual Address
(MVA) by the FCSE (Fast Context Switch Extension) using the value in CP15 register13. The
MMU translates modified virtual addresses to physical addresses by using a single, two-level
page table set stored in physical memory. Each entry in the set contains the access permissions
and the physical address that correspond to the virtual address.
The first level translation tables contain 4096 entries indexed by bits [31:20] of the MVA. These
entries contain a pointer to either a 1 MB section of physical memory along with attribute infor-
mation (access permissions, domain, etc.) or an entry in the second level translation tables;
coarse table and fine table.
The second level translation tables contain two subtables, coarse table and fine table. An entry
in the coarse table contains a pointer to both large pages and small pages along with access
permissions. An entry in the fine table contains a pointer to large, small and tiny pages.
Table 7 shows the different attributes of each page in the physical memory.
Table 9-6. Mapping Details

Mapping Name Mapping Size Access Permission By Subpage Size
Section 1M byte Section -
Large Page 64K bytes 4 separated subpages 16K bytes
Small Page 4K bytes 4 separated subpages 1K byte
Tiny Page 1K byte Tiny Page -
The MMU consists of:

• Access control logic
• Translation Look-aside Buffer (TLB)
• Translation table walk hardware
9.6.1 Access Control Logic

The access control logic controls access information for every entry in the translation table. The
access control logic checks two pieces of access information: domain and access permissions.
The domain is the primary access control mechanism for a memory region; there are 16 of them.
It defines the conditions necessary for an access to proceed. The domain determines whether
the access permissions are used to qualify the access or whether they should be ignored.
The second access control mechanism is access permissions that are defined for sections and
for large, small and tiny pages. Sections and tiny pages have a single set of access permissions
whereas large and small pages can be associated with 4 sets of access permissions, one for
each subpage (quarter of a page).
9.6.2 Translation Look-aside Buffer (TLB)

The Translation Look-aside Buffer (TLB) caches translated entries and thus avoids going
through the translation process every time. When the TLB contains an entry for the MVA (Modi-
42 AT91SAM9G45
AT91SAM9G45
fied Virtual Address), the access control logic determines if the access is permitted and outputs
the appropriate physical address corresponding to the MVA. If access is not permitted, the MMU
signals the CPU core to abort.
If the TLB does not contain an entry for the MVA, the translation table walk hardware is invoked
to retrieve the translation information from the translation table in physical memory.
9.6.3 Translation Table Walk Hardware

The translation table walk hardware is a logic that traverses the translation tables located in
physical memory, gets the physical address and access permissions and updates the TLB.
The number of stages in the hardware table walking is one or two depending whether the
address is marked as a section-mapped access or a page-mapped access.
There are three sizes of page-mapped accesses and one size of section-mapped access. Page-
mapped accesses are for large pages, small pages and tiny pages. The translation process
always begins with a level one fetch. A section-mapped access requires only a level one fetch,
but a page-mapped access requires an additional level two fetch. For further details on the
MMU, please refer to chapter 3 in ARM926EJ-S Technical Reference Manual.
9.6.4 MMU Faults

The MMU generates an abort on the following types of faults:
• Alignment faults (for data accesses only)
• Translation faults
• Domain faults
• Permission faults
The access control mechanism of the MMU detects the conditions that produce these faults. If
the fault is a result of memory access, the MMU aborts the access and signals the fault to the
CPU core.The MMU retains status and address information about faults generated by the data
accesses in the data fault status register and fault address register. It also retains the status of
faults generated by instruction fetches in the instruction fault status register.
The fault status register (register 5 in CP15) indicates the cause of a data or prefetch abort, and
the domain number of the aborted access when it happens. The fault address register (register 6
in CP15) holds the MVA associated with the access that caused the Data Abort. For further
details on MMU faults, please refer to chapter 3 in ARM926EJ-S Technical Reference Manual.
43
9.7 Caches and Write Buffer
The ARM926EJ-S contains a 32K Byte Instruction Cache (ICache), a 32K Byte Data Cache
(DCache), and a write buffer. Although the ICache and DCache share common features, each
still has some specific mechanisms.
The caches (ICache and DCache) are four-way set associative, addressed, indexed and tagged
using the Modified Virtual Address (MVA), with a cache line length of eight words with two dirty
bits for the DCache. The ICache and DCache provide mechanisms for cache lockdown, cache
pollution control, and line replacement.
A new feature is now supported by ARM926EJ-S caches called allocate on read-miss commonly
known as wrapping. This feature enables the caches to perform critical word first cache refilling.
This means that when a request for a word causes a read-miss, the cache performs an AHB
access. Instead of loading the whole line (eight words), the cache loads the critical word first, so
the processor can reach it quickly, and then the remaining words, no matter where the word is
located in the line.
The caches and the write buffer are controlled by the CP15 register 1 (Control), CP15 register 7
(cache operations) and CP15 register 9 (cache lockdown).
9.7.1 Instruction Cache (ICache)

The ICache caches fetched instructions to be executed by the processor. The ICache can be
enabled by writing 1 to I bit of the CP15 Register 1 and disabled by writing 0 to this same bit.
When the MMU is enabled, all instruction fetches are subject to translation and permission
checks. If the MMU is disabled, all instructions fetches are cachable, no protection checks are
made and the physical address is flat-mapped to the modified virtual address. With the MVA use
disabled, context switching incurs ICache cleaning and/or invalidating.
When the ICache is disabled, all instruction fetches appear on external memory (AHB) (see
Tables 4-1 and 4-2 in page 4-4 in ARM926EJ-S TRM).
On reset, the ICache entries are invalidated and the ICache is disabled. For best performance,
ICache should be enabled as soon as possible after reset.
9.7.2 Data Cache (DCache) and Write Buffer

ARM926EJ-S includes a DCache and a write buffer to reduce the effect of main memory band-
width and latency on data access performance. The operations of DCache and write buffer are
closely connected.
9.7.2.1 DCache
The DCache needs the MMU to be enabled. All data accesses are subject to MMU permission
and translation checks. Data accesses that are aborted by the MMU do not cause linefills or data
accesses to appear on the AMBA ASB interface. If the MMU is disabled, all data accesses are
noncachable, nonbufferable, with no protection checks, and appear on the AHB bus. All
addresses are flat-mapped, VA = MVA = PA, which incurs DCache cleaning and/or invalidating
every time a context switch occurs.
The DCache stores the Physical Address Tag (PA Tag) from which every line was loaded and
uses it when writing modified lines back to external memory. This means that the MMU is not
involved in write-back operations.
Each line (8 words) in the DCache has two dirty bits, one for the first four words and the other
one for the second four words. These bits, if set, mark the associated half-lines as dirty. If the
44 AT91SAM9G45
AT91SAM9G45
cache line is replaced due to a linefill or a cache clean operation, the dirty bits are used to decide
whether all, half or none is written back to memory.
DCache can be enabled or disabled by writing either 1 or 0 to bit C in register 1 of CP15 (see
Tables 4-3 and 4-4 on page 4-5 in ARM926EJ-S TRM).
The DCache supports write-through and write-back cache operations, selected by memory
region using the C and B bits in the MMU translation tables.
The DCache contains an eight data word entry, single address entry write-back buffer used to
hold write-back data for cache line eviction or cleaning of dirty cache lines.
The Write Buffer can hold up to 16 words of data and four separate addresses. DCache and
Write Buffer operations are closely connected as their configuration is set in each section by the
page descriptor in the MMU translation table.
9.7.2.2 Write Buffer

The ARM926EJ-S contains a write buffer that has a 16-word data buffer and a four- address buf-
fer. The write buffer is used for all writes to a bufferable region, write-through region and write-
back region. It also allows to avoid stalling the processor when writes to external memory are
performed. When a store occurs, data is written to the write buffer at core speed (high speed).
The write buffer then completes the store to external memory at bus speed (typically slower than
the core speed). During this time, the ARM9EJ-S processor can preform other tasks.
DCache and Write Buffer support write-back and write-through memory regions, controlled by C
and B bits in each section and page descriptor within the MMU translation tables.
9.7.2.3 Write-though Operation

When a cache write hit occurs, the DCache line is updated. The updated data is then written to
the write buffer which transfers it to external memory.
When a cache write miss occurs, a line, chosen by round robin or another algorithm, is stored in
9.7.2.4 Write-back Operation

When a cache write hit occurs, the cache line or half line is marked as dirty, meaning that its
contents are not up-to-date with those in the external memory.
When a cache write miss occurs, a line, chosen by round robin or another algorithm, is stored in
45
9.8 Tightly-Coupled Memory Interface
9.8.1 TCM Description

The ARM926EJ-S processor features a Tightly-coupled Memory (TCM) interface, which enables
separate instruction and data TCMs (ITCM and DTCM) to be directly reached by the processor.
TCMs are used to store real-time and performance critical code, they also provide a DMA sup-
port mechanism. Unlike AHB accesses to external memories, accesses to TCMs are fast and
deterministic and do not incur bus penalties.
The user has the possibility to independently configure each TCM size with values within the fol-
lowing ranges, [0K Byte, 64K Bytes] for ITCM size and [0K Byte, 64K Bytes] for DTCM size.
TCMs can be configured by two means: HMATRIX TCM register and TCM region register (regis-
ter 9) in CP15 and both steps should be performed. HMATRIX TCM register sets TCM size
whereas TCM region register (register 9) in CP15 maps TCMs and enables them.
The data side of the ARM9EJ-S core is able to access the ITCM. This is necessary to enable
code to be loaded into the ITCM, for SWI and emulated instruction handlers, and for accesses to
PC-relative literal pools.
9.8.2 Enabling and Disabling TCMs

Prior to any enabling step, the user should configure the TCM sizes in HMATRIX TCM register.
Then enabling TCMs is performed by using TCM region register (register 9) in CP15. The user
should use the same sizes as those put in HMATRIX TCM register. For further details and pro-
gramming tips, please refer to chapter 2.3 in ARM926EJ-S TRM.
9.8.3 TCM Mapping

The TCMs can be located anywhere in the memory map, with a single region available for ITCM
and a separate region available for DTCM. The TCMs are physically addressed and can be
placed anywhere in physical address space. However, the base address of a TCM must be
aligned to its size, and the DTCM and ITCM regions must not overlap. TCM mapping is per-
formed by using TCM region register (register 9) in CP15. The user should input the right
mapping address for TCMs.
46 AT91SAM9G45
AT91SAM9G45
9.9 Bus Interface Unit

The ARM926EJ-S features a Bus Interface Unit (BIU) that arbitrates and schedules AHB
requests. The BIU implements a multi-layer AHB, based on the AHB-Lite protocol, that enables
parallel access paths between multiple AHB masters and slaves in a system. This is achieved by
using a more complex interconnection matrix and gives the benefit of increased overall bus
bandwidth, and a more flexible system architecture.
The multi-master bus architecture has a number of benefits:
• It allows the development of multi-master systems with an increased bus bandwidth and a
flexible architecture.
• Each AHB layer becomes simple because it only has one master, so no arbitration or master-
to-slave muxing is required. AHB layers, implementing AHB-Lite protocol, do not have to
support request and grant, nor do they have to support retry and split transactions.
• The arbitration becomes effective when more than one master wants to access the same
slave simultaneously.
9.9.1 Supported Transfers

The ARM926EJ-S processor performs all AHB accesses as single word, bursts of four words, or
bursts of eight words. Any ARM9EJ-S core request that is not 1, 4, 8 words in size is split into
packets of these sizes. Note that the Atmel bus is AHB-Lite protocol compliant, hence it does not
support split and retry requests.
Table 8 gives an overview of the supported transfers and different kinds of transactions they are
used for.
Table 9-7. Supported Transfers

HBurst[2:0] Description
Single transfer of word, half word, or byte:
• data write (NCNB, NCB, WT, or WB that has missed in DCache)
SINGLE Single transfer • data read (NCNB or NCB)
• NC instruction fetch (prefetched and non-prefetched)
• page table walk read
Half-line cache write-back, Instruction prefetch, if enabled. Four-word burst NCNB,
INCR4 Four-word incrementing burst
NCB, WT, or WB write.
INCR8 Eight-word incrementing burst Full-line cache write-back, eight-word burst NCNB, NCB, WT, or WB write.
WRAP8 Eight-word wrapping burst Cache linefill
9.9.2 Thumb Instruction Fetches

All instructions fetches, regardless of the state of ARM9EJ-S core, are made as 32-bit accesses
on the AHB. If the ARM9EJ-S is in Thumb state, then two instructions can be fetched at a time.
9.9.3 Address Alignment

The ARM926EJ-S BIU performs address alignment checking and aligns AHB addresses to the
necessary boundary. 16-bit accesses are aligned to halfword boundaries, and 32-bit accesses
are aligned to word boundaries.
47
48 AT91SAM9G45
AT91SAM9G45
10. AT91SAM9G45 Debug and Test
10.1 Description
The AT91SAM9G45 features a number of complementary debug and test capabilities. A com-
mon JTAG/ICE (In-Circuit Emulator) port is used for standard debugging functions, such as
downloading code and single-stepping through programs. The Debug Unit provides a two-pin
UART that can be used to upload an application into internal SRAM. It manages the interrupt
handling of the internal COMMTX and COMMRX signals that trace the activity of the Debug
Communication Channel.
A set of dedicated debug and test input/output pins gives direct access to these capabilities from
a PC-based test environment.

• ARM926 Real-time In-circuit Emulator
– Two real-time Watchpoint Units
– Two Independent Registers: Debug Control Register and Debug Status Register
– Test Access Port Accessible through JTAG Protocol
– Debug Communications Channel
• Debug Unit
– Two-pin UART
– Debug Communication Channel Interrupt Handling
– Chip ID Register
• IEEE1149.1 JTAG Boundary-scan on All Digital Pins.
49
10.3 Block Diagram
Figure 10-1. Debug and Test Block Diagram
TMS
TCK
TDI
NTRST
ICE/JTAG JTAGSEL
Boundary TAP
Port
TDO
RTCK
POR
Reset
and
Test TST
ARM9EJ-S ICE-RT
ARM926EJ-S
DTXD
PIO
PDC DBGU
DRXD
TAP: Test Access Port
50 AT91SAM9G45
AT91SAM9G45
10.4 Application Examples
10.4.1 Debug Environment

Figure 10-2 on page 51 shows a complete debug environment example. The ICE/JTAG inter-
face is used for standard debugging functions, such as downloading code and single-stepping
through the program. A software debugger running on a personal computer provides the user
interface for configuring a Trace Port interface utilizing the ICE/JTAG interface.
Figure 10-2. Application Debug and Trace Environment Example
Host Debugger PC
ICE/JTAG
Interface
ICE/JTAG
Connector
RS232
AT91SAM9G45 Terminal
Connector
AT91SAM9G45-based Application Board
51
10.4.2 Test Environment
Figure 10-3 on page 52 shows a test environment example. Test vectors are sent and inter-
preted by the tester. In this example, the “board in test” is designed using a number of JTAG-
compliant devices. These devices can be connected to form a single scan chain.
Figure 10-3. Application Test Environment Example
Test Adaptor
Tester
JTAG
Interface
ICE/JTAG Chip n Chip 2
AT91SAM9G45 Chip 1
AT91SAM9G45-based Application Board In Test
10.5 Debug and Test Pin Description
Table 10-1. Debug and Test Pin List

Pin Name Function Type Active Level
Reset/Test
NRST Microcontroller Reset Input/Output Low
TST Test Mode Select Input High
ICE and JTAG
NTRST Test Reset Signal Input Low
TCK Test Clock Input
TDI Test Data In Input
TDO Test Data Out Output
TMS Test Mode Select Input
RTCK Returned Test Clock Output
JTAGSEL JTAG Selection Input
Debug Unit
52 AT91SAM9G45
AT91SAM9G45
10.6 Functional Description
10.6.1 Test Pin

One dedicated pin, TST, is used to define the device operating mode. The user must make sure
that this pin is tied at low level to ensure normal operating conditions. Other values associated
with this pin are reserved for manufacturing test.
10.6.2 EmbeddedICE
The ARM9EJ-S EmbeddedICE-RT™ is supported via the ICE/JTAG port. It is connected to a
host computer via an ICE interface. Debug support is implemented using an ARM9EJ-S core
embedded within the ARM926EJ-S. The internal state of the ARM926EJ-S is examined through
an ICE/JTAG port which allows instructions to be serially inserted into the pipeline of the core
without using the external data bus. Therefore, when in debug state, a store-multiple (STM) can
be inserted into the instruction pipeline. This exports the contents of the ARM9EJ-S registers.
This data can be serially shifted out without affecting the rest of the system.
There are two scan chains inside the ARM9EJ-S processor which support testing, debugging,
and programming of the EmbeddedICE-RT. The scan chains are controlled by the ICE/JTAG
port.
EmbeddedICE mode is selected when JTAGSEL is low. It is not possible to switch directly
between ICE and JTAG operations. A chip reset must be performed after JTAGSEL is changed.
For further details on the EmbeddedICE-RT, see the ARM document:
ARM9EJ-S Technical Reference Manual (DDI 0222A).
10.6.3 JTAG Signal Description

TMS is the Test Mode Select input which controls the transitions of the test interface state
machine.
TDI is the Test Data Input line which supplies the data to the JTAG registers (Boundary Scan
Register, Instruction Register, or other data registers).
TDO is the Test Data Output line which is used to serially output the data from the JTAG regis-
ters to the equipment controlling the test. It carries the sampled values from the boundary scan
chain (or other JTAG registers) and propagates them to the next chip in the serial test circuit.
NTRST (optional in IEEE Standard 1149.1) is a Test-ReSeT input which is mandatory in ARM
cores and used to reset the debug logic. On Atmel ARM926EJ-S-based cores, NTRST is a
Power On Reset output. It is asserted on power on. If necessary, the user can also reset the
debug logic with the NTRST pin assertion during 2.5 MCK periods.
TCK is the Test ClocK input which enables the test interface. TCK is pulsed by the equipment
controlling the test and not by the tested device. It can be pulsed at any frequency. Note the
maximum JTAG clock rate on ARM926EJ-S cores is 1/6th the clock of the CPU. This gives 5.45
kHz maximum initial JTAG clock rate for an ARM9E running from the 32.768 kHz slow clock.
RTCK is the Return Test Clock. Not an IEEE Standard 1149.1 signal added for a better clock
handling by emulators. From some ICE Interface probes, this return signal can be used to syn-
chronize the TCK clock and take not care about the given ratio between the ICE Interface clock
and system clock equal to 1/6th. This signal is only available in JTAG ICE Mode and not in
boundary scan mode.
53
10.6.4 Debug Unit
The Debug Unit provides a two-pin (DXRD and TXRD) USART that can be used for several
debug and trace purposes and offers an ideal means for in-situ programming solutions and
debug monitor communication. Moreover, the association with two peripheral data controller
channels permits packet handling of these tasks with processor time reduced to a minimum.
The Debug Unit also manages the interrupt handling of the COMMTX and COMMRX signals
that come from the ICE and that trace the activity of the Debug Communication Channel.The
Debug Unit allows blockage of access to the system through the ICE interface.
A specific register, the Debug Unit Chip ID Register, gives information about the product version
and its internal configuration.
The AT91SAM9G45 Debug Unit Chip ID value is 0x819B 05A1 and the extended ID is
0x00000004 on 32-bit width.
For further details on the Debug Unit, see the Debug Unit section.
10.6.5 IEEE 1149.1 JTAG Boundary Scan

IEEE 1149.1 JTAG Boundary Scan allows pin-level access independent of the device packaging
technology.
IEEE 1149.1 JTAG Boundary Scan is enabled when JTAGSEL is high. The SAMPLE, EXTEST
and BYPASS functions are implemented. In ICE debug mode, the ARM processor responds
with a non-JTAG chip ID that identifies the processor to the ICE system. This is not IEEE 1149.1
JTAG-compliant.
It is not possible to switch directly between JTAG and ICE operations. A chip reset must be per-
formed after JTAGSEL is changed.
A Boundary-scan Descriptor Language (BSDL) file is provided to set up test.
10.6.6 JID Code Register
Access: Read-only
31 30 29 28 27 26 25 24
VERSION PART NUMBER
23 22 21 20 19 18 17 16
PART NUMBER
15 14 13 12 11 10 9 8
PART NUMBER MANUFACTURER IDENTITY
7 6 5 4 3 2 1 0
MANUFACTURER IDENTITY 1
• VERSION[31:28]: Product Version Number

Set to 0x0.
• PART NUMBER[27:12]: Product Part Number
Product part Number is 5B27
• MANUFACTURER IDENTITY[11:1]
Set to 0x01F.
Bit[0] required by IEEE Std. 1149.1.
Set to 0x1.
JTAG ID Code value is 05B2_703F.
54 AT91SAM9G45
AT91SAM9G45
11. Boot Strategies

The system always boots at address 0x0. To ensure maximum boot possibilities the memory
layout can be changed with two parameters.
• REMAP allows the user to layout the internal SRAM bank to 0x0 to ease the development.
This is done by software once the system has boot.
• BMS allows the user to layout to 0x0, when convenient, the ROM or an external memory. This
is done by hardware at reset.
Note: All the memory blocks can always be seen at their specified base addresses that are not
concerned by these parameters.
The AT91SAM9G45 manages a boot memory that depends on the level on the BMS pin at
reset. The internal memory area mapped between address 0x0 and 0x000F FFFF is reserved to
this effect.
If BMS is detected at 0, the boot memory is the memory connected on the Chip Select 0 of the
External Bus Interface.
• Boot on on-chip RC
• Boot with the default configuration for the Static Memory Controller, byte select mode, 16-bit
data bus, Read/Write controlled by Chip Select, allows boot on 16-bit non-volatile memory.
For optimization purpose, nothing else is done. To speed up the boot sequence user pro-
grammed software should perform a complete configuration:
• Enable the 32768 Hz oscillator if best accuracy is needed
• Program the PMC (main oscillator enable or bypass mode)
• Program and Start the PLL
• Reprogram the SMC setup, cycle, hold, mode timings registers for EBI CS0 to adapt them to
the new clock
• Switch the system clock to the new value
If BMS is detected at 1, the boot memory is the embedded ROM and the boot programdescribed
below is executed.
11.1 Boot Program

The Boot Program is contained in the embedded ROM. It is also called: “Rom Code” or “First
level bootloader”. At power on, if the BMS pin is detected at 1, the boot memory is the embed-
ded ROM and the Boot Program is executed.
The Boot Program consists of several steps. First, it performs device initialization. Then it
attempts to boot from external non volatile memories (NVM). And finally, if no valid program is
found in NVM, it executes a monitor called SAM-BA® Monitor.
55
11.2 Flow Diagram
The Boot Program implements the algorithm shown below in Figure 11-1.
Figure 11-1. Boot Program Algorithm Flow Diagram
Device Setup
Valid boot code Yes Copy and run it

found in one in internal SRAM
NVM
No
SAM-BA Monitor
56 AT91SAM9G45
AT91SAM9G45
11.3 Device Initialization
11.3.1 Clock at Start Up

At boot start up, the processor clock (PCK) and the master clock (MCK) are found on the slow
clock. The slow clock can be an external 32 kHz crystal oscillator or the internal RC oscillator. By
default the slow clock is the internal RC oscillator. Its frequency is not precise and is between 20
kHz and 40 kHz. Its start up is much faster than an external 32 kHz quartz. If a battery supplies
the backup power and if the external 32 kHz clock was previously started up and selected, the
slow clock at boot is the external 32 kHz quartz oscillator. Refer to the Slow Clock Crystal Oscil-
lator description in the Clock Generator section of the datasheet.
11.3.2 Initialization Sequence

Initialization follows the steps described below:
1. Stack setup for ARM supervisor mode.
2. Main Oscillator Detection: (External crystal or external clock on XIN). The Main Oscil-
lator is disabled at startup (MOSCEN = 0). First it is bypassed (OSCBYPASS set at 1).
Then the MAINRDY bit is polled. Since this bit is raised, the Main Clock Frequency field
is analyzed (MAINF). If the value is bigger than 16, an external clock connected on XIN
is detected. If not, an external quartz connected between XIN and XOUT (whose fre-
quency is unknown at this moment) is detected.
3. Main Oscillator Enabling: if an external clock is connected on XIN, the Main Oscillator
does not need to be started. Otherwise, the OSCBYPASS bit is not set. The Main Oscil-
lator is enabled (MOSCEN = 1) with the maximum start-up time and the MOSC bit is
polled to wait for stabilization.
4. Main Oscillator Selection: the Master Clock source is switched from Slow Clock to the
Main Oscillator without prescaler. The PMC Status Register is polled to wait for MCK
Ready. PCK and MCK are now the Main Oscillator clock.
5. C variable initialization: non zero-initialized data are initialized in RAM (copy from
ROM to RAM). Zero-initialized data are set to 0 in RAM.
6. PLLA initialization: PLLA is configured to allow communication on the USB link for the
SAM-BA Monitor. Its configuration depends on the Main Oscillator source (external
clock or crystal) and on its frequency.
57
AT91SAM9G45
11.4 NVM Boot
11.4.1 NVM Bootloader Program Description
Figure 11-2. NVM bootloader program diagram
Start
Initialize NVM
No Restore the reset values

Initialization OK ? for the peripherals and
Jump to next boot solution
Yes
Valid code detection in NVM
No
NVM contains valid code
Yes
Copy the valid code

from external NVM to internal SRAM.
Restore the reset values for the peripherals.

Perform the REMAP and set the PC to 0
to jump to the downloaded application
End
58
AT91SAM9G45
Figure 11-3. Remap Action after Download Completion
0x0000_0000 0x0000_0000
REMAP
Internal Internal
ROM SRAM
0x0030_0000 0x0030_0000
Internal Internal
SRAM SRAM
0x0040_0000 0x0040_0000
Internal Internal
ROM ROM
The NVM bootloader program initializes the NVM. It initializes the required PIO. It sets the right
peripheral depending on the NVM and tries to access the memory. If the initialization fails, it
restores the reset values for the PIO and peripherals and then the next NVM bootloader program
is executed.
If the initialization is successful, the NVM bootloader program reads the beginning of the NVM
and determines if the NVM contains valid code.
If the NVM does not contain valid code, the NVM bootloader program restores the reset value for
the peripherals and then the next NVM bootloader program is executed.
If valid code is found, this code is loaded from NVM into internal SRAM and executed by branch-
ing at address 0x0000_0000 after remap. This code may be the application code or a second-
level bootloader. All the calls to functions are PC relative and do not use absolute addresses.
11.4.2 Valid Code Detection

There are two kinds of valid code detection. Depending on the NVM bootloader, either one or
both of them is used.
11.4.2.1 ARM Exception Vectors Check

The NVM bootloader program reads and analyzes the first 28 bytes corresponding to the first
seven ARM exception vectors. Except for the sixth vector, these bytes must implement the ARM
instructions for either branch or load PC with PC relative addressing.
Figure 11-4. LDR Opcode
31 28 27 24 23 20 19 16 15 12 11 0
1 1 1 0 0 1 I P U 1 W 0 Rn Rd O set
59
AT91SAM9G45
Figure 11-5. B Opcode
31 28 27 24 23 0
1 1 1 0 1 0 1 0 O set (24 bits)
Unconditional instruction: 0xE for bits 31 to 28

Load PC with PC relative addressing instruction:
– Rn = Rd = PC = 0xF
– I==0 (12-bit immediate value)
– P==1 (pre-indexed)
– U offset added (U==1) or subtracted (U==0)
– W==1
The sixth vector, at offset 0x14, contains the size of the image to download. The user must
replace this vector with his/her own vector. This information is described below.
Figure 11-6. Structure of the ARM Vector 6
31 0
Size of the code to download in bytes
The value has to be smaller than 60 KBytes. 60 KBytes is the maximum size for a valid code.
This size is the internal SRAM size minus the stack size used by the ROM Code at the end of
the internal SRAM.
Example
An example of valid vectors follows:
00 ea000006 B 0x20
04 eafffffe B 0x04
08 ea00002f B _main
0c eafffffe B 0x0c
10 eafffffe B 0x10
14 00001234 B 0x14 <- Code size = 4660 bytes < 60kB
18 eafffffe B 0x18
11.4.2.2 boot.bin file check

The NVM bootloader program looks for a boot.bin file in the root directory of a FAT12/16/32 for-
matted NVM Flash.
60
AT91SAM9G45
11.4.3 NVM Bootloader Sequence
Figure 11-7. NVM Bootloader Sequence Diagram
Device
Setup
Yes Copy from

NAND Flash Boot Run NAND Flash Bootloader
NAND Flash to SRAM
No
Yes Copy from

SD Card Boot Run SD Card Bootloader
SD Card to SRAM
No
Yes Copy from

SPI Flash Boot Run SPI Flash Bootloader
SPI Flash to SRAM
No
Yes Copy from

TWI EEPROM Boot Run TWI EEPROM Bootloader
TWI EEPROM to SRAM
No
SAM-BA
Monitor
11.4.3.1 NAND Flash Boot

The NAND Flash bootloader program uses the EBI CS0. It uses both valid code detections. First
it searches a boot.bin file. Then it analyzes the ARM exception vectors.
The first block must be guaranteed by the manufacturer. There is no ECC check.
After NAND Flash interface configuration, the Manufacturer ID is read. If it is different from 0xFF,
the Device ID is read, else, the NAND Flash boot is aborted. The Boot program contains a list of
SLC small block Device ID with their characteristics (size, bus width, voltage) (see Table 11-1). If
the device ID is not found in this list, the NAND Flash device is considered as an SLC large block
and its characteristics are obtained by reading the Extended Device ID byte 3.
61
AT91SAM9G45
Supported NAND Flash Devices

The supportedSLC small block NAND Flash devices that are are described below inTable 11-1.
Table 11-1. Supported SLC Small Block NAND Flash

Size PageSize BlockSsize
Device ID (MBytes) (Bytes) (Bytes) Bus Width Voltage (V)
0x6E 1 256 4096 8 5
0x64 2 256 4096 8 5
0x6B 4 512 8196 8 5
0xE8 1 256 4096 8 3.3
0xEC 1 256 4096 8 3.3
0xEA 2 256 4096 8 3.3
0xE3 4 512 8196 8 3.3

0xE5 4 512 8196 8 3.3
0xD6 8 512 8196 8 3.3
0xE6 8 512 8196 8 3.3
0x33 16 512 16384 8 1.8
0x73 16 512 16384 8 3.3
0x43 16 512 16384 16 1.8
0x53 16 512 16384 16 3.3
0x45 32 512 16384 16 1.8
0x55 32 512 16384 16 3.3
0x36 64 512 16384 8 1.8
0x76 64 512 16384 8 3.3
0x46 64 512 16384 16 1.8
0x56 64 512 16384 16 3.3
0x78 128 512 16384 8 1.8
0x79 128 512 16384 8 3.3
0x72 128 512 16384 16 1.8
0x74 128 512 16384 16 3.3
The NAND Flash boot also supports all the SLC large block NAND Flash devices.
11.4.3.2 SD Card Boot

The SD Card bootloader uses MCI0. It uses only one valid code detection. It searches a boot.bin
file.
Supported SD Card devices
SD Card Boot supports all SD Card memories compliant with SD Memory Card Specification
V2.0. This includes SDHC cards.
62
AT91SAM9G45
11.4.3.3 SPI Flash Boot

Two kinds of SPI Flash are supported, SPI Serial Flash and SPI DataFlash.
The SPI Flash bootloader tries to boot on SPI0 Chip Select 0, first looking for SPI Serial flash,
and then for SPI DataFlash.
It uses only one valid code detection: analysis of ARM exception vectors.
The SPI Flash read is done thanks to a Continuous Read command from address 0x0. This
command is 0xE8 for DataFlash and 0x0B for Serial Flash devices.
Supported DataFlash Devices
The SPI Flash Boot program supports all Atmel DataFlash devices.
Table 11-2. DataFlash Device

Device Density Page Size (bytes) Number of Pages
AT45DB011 1 Mbit 264 512
AT45DB021 2 Mbits 264 1024
AT45DB041 4 Mbits 264 2048
AT45DB081 8 Mbits 264 4096
AT45DB161 16 Mbits 528 4096
AT45DB321 32 Mbits 528 8192
AT45DB642 64 Mbits 1056 8192
Supported Serial Flash Devices

The SPI Flash Boot program supports all Serial Flash devices.
11.4.3.4 TWI EEPROM Boot

The TWI EEPROM Bootloader uses the TWI. It uses only one valid code detection. It analyzes
the ARM exception vectors.
Supported TWI EEPROM Devices
TWI EEPROM Boot supports all I2C-compatibleTWI EEPROM memories using 7 bits device
address 0x50.
11.4.4 Hardware and Software Constraints

The NVM drivers use several PIOs in peripheral mode to communicate with devices. Care must
be taken when these PIOs are used by the application. The devices connected could be unin-
tentionally driven at boot time, and electrical conflicts between output pins used by the NVM
drivers and the connected devices may occur.
To assure correct functionality, it is recommended to plug in critical devices to other pins not
used by NVM.
Table 11-3 contains a list of pins that are driven during the boot program execution. These pins
are driven during the boot sequence for a period of less than 1 second if no correct boot program
is found.
63
AT91SAM9G45
Before performing the jump to the application in internal SRAM, all the PIOs and peripherals
used in the boot program are set to their reset state.
Table 11-3. PIO Driven during Boot Program Execution

NVM Bootloader Peripheral Pin PIO Line
EBI CS0 SMC NANDCS PIOC14
EBI CS0 SMC NAND ALE A21
NAND
EBI CS0 SMC NAND CLE A22
EBI CS0 SMC Cmd/Addr/Data D[16:0]
MCI0 MCI0_CK PIOA0
MCI0 MCI0_CD PIOA1
MCI0 MCI0_D0 PIOA2
SD Card
MCI0 MCI0_D1 PIOA3
MCI0 MCI0_D2 PIOA4
MCI0 MCI0_D3 PIOA5
SPI0 MOSI PIOB1
SPI0 MISO PIOB0
SPI Flash
SPI0 SPCK PIOB2
SPI0 NPCS0 PIOB3
TWI TWD PIOA20
TWI EEPROM
TWI TWCK PIOA21
DBGU DRXD PIOB12
SAM-BA Monitor
DBGU DTXD PIOB13
11.5 SAM-BA Monitor

If no valid code has been found in NVM during the NVM bootloader sequence, the SAM-BA
Monitor program is launched.
The SAM-BA Monitor principle is to:
– Initialize DBGU and USB
– Check if USB Device enumeration has occurred.
– Check if characters have been received on the DBGU.
– Once the communication interface is identified, the application runs in an infinite
loop waiting for different commands as listed in Table .
64
AT91SAM9G45
Figure 11-8. SAM-BA Monitor Diagram
No valid code in NVM
Init DBGU and USB

No
No
USB Enumeration Character(s) received
Successful ? on DBGU ?
Yes Yes
Run monitor Run monitor

Wait for command Wait for command
on the USB link on the DBGU link
11.5.1 Command List
Table 11-4. Commands Available through the SAM-BA Monitor

Command Action Argument(s) Example
N set Normal mode No argument N#

T set Terminal mode No argument T#
O write a byte Address, Value# O200001,CA#
o read a byte Address,# o200001,#
H write a half word Address, Value# H200002,CAFE#
h read a half word Address,# h200002,#
W write a word Address, Value# W200000,CAFEDECA#
w read a word Address,# w200000,#
S send a file Address,# S200000,#
R receive a file Address, NbOfBytes# R200000,1234#
G go Address# G200200#
V display version No argument V#
• Mode commands:
– Normal mode configures SAM-BA Monitor to send / receive data in binary format,
– Terminal mode configures SAM-BA Monitor to send / receive data in ascii format.
• Write commands: Write a byte (O), a halfword (H) or a word (W) to the target.
– Address: Address in hexadecimal.
– Value: Byte, halfword or word to write in hexadecimal.
65
– Output: ‘>’.
• Read commands: Read a byte (o), a halfword (h) or a word (w) from the target.
– Address: Address in hexadecimal
– Output: The byte, halfword or word read in hexadecimal following by ‘>’
• Send a file (S): Send a file to a specified address
– Output: ‘>’.
Note: There is a time-out on this command which is reached when the prompt ‘>’ appears before the
end of the command execution.
• Receive a file (R): Receive data into a file from a specified address
– NbOfBytes: Number of bytes in hexadecimal to receive
– Output: ‘>’
• Go (G): Jump to a specified address and execute the code
– Address: Address to jump in hexadecimal
– Output: ‘>’once returned from the program execution. If the executed program does
not handle the link register at its entry and does not return, the prompt will not be
displayed.
• Get Version (V): Return the Boot Program version
– Output: version, date and time of ROM code followed by the prompt: ‘>’.
11.5.2 DBGU Serial Port

Communication is performed through the DBGU serial port initialized to 115200 Baud, 8 bits of
data, no parity, 1 stop bit.
11.5.2.1 Supported External Crystal/External Clocks

The supported frequencies by the SAM-BA Monitor to allow the DBGU communication for both
external crystal and external clock are:
• 4 MHz,
• 12 MHz,
• 28 MHz.
11.5.2.2 Xmodem Protocol

The Send and Receive File commands use the Xmodem protocol to communicate. Any terminal
performing this protocol can be used to send the application file to the target. The size of the
binary file to send depends on the SRAM size embedded in the product. In all cases, the size of
the binary file must be lower than the SRAM size because the Xmodem protocol requires some
SRAM memory in order to work.
The Xmodem protocol supported is the 128-byte length block. This protocol uses a two-charac-
ter CRC-16 to guarantee detection of a maximum bit error.
Xmodem protocol with CRC is accurate provided both sender and receiver report successful
transmission. Each block of the transfer looks like:
<SOH><blk #><255-blk #><--128 data bytes--><checksum> in which:
66 AT91SAM9G45
AT91SAM9G45
– <SOH> = 01 hex
– <blk #> = binary number, starts at 01, increments by 1, and wraps 0FFH to 00H (not
to 01)
– <255-blk #> = 1’s complement of the blk#.
– <checksum> = 2 bytes CRC16
Figure 11-9 shows a transmission using this protocol.
Figure 11-9. Xmodem Transfer Example
Host Device
SOH 01 FE Data[128] CRC CRC
ACK
SOH 02 FD Data[128] CRC CRC
ACK
SOH 03 FC Data[100] CRC CRC
ACK
EOT
ACK
11.5.3 USB Device Port
11.5.3.1 Supported external crystal / external clocks

The only frequency supported by SAM-BA Monitor to allow USB communication is a 12 MHz
crystal or external clock.
11.5.3.2 USB class

The device uses the USB communication device class (CDC) drivers to take advantage of the
installed PC RS-232 software to talk over the USB. The CDC class is implemented in all
releases of Windows®, from Windows 98SE® to Windows XP®. The CDC document, available at
www.usb.org, describes how to implement devices such as ISDN modems and virtual COM
ports.
The Vendor ID is Atmel’s vendor ID 0x03EB. The product ID is 0x6124. These references are
used by the host operating system to mount the correct driver. On Windows systems, the INF
files contain the correspondence between vendor ID and product ID.
67
11.5.3.3 Enumeration Process
The USB protocol is a master/slave protocol. The host starts the enumeration, sending requests
to the device through the control endpoint. The device handles standard requests as defined in
the USB Specification.
Table 11-5. Handled Standard Requests

Request Definition
GET_DESCRIPTOR Returns the current device configuration value.
SET_ADDRESS Sets the device address for all future device access.
SET_CONFIGURATION Sets the device configuration.
GET_CONFIGURATION Returns the current device configuration value.
GET_STATUS Returns status for the specified recipient.
SET_FEATURE Used to set or enable a specific feature.
CLEAR_FEATURE Used to clear or disable a specific feature.
The device also handles some class requests defined in the CDC class.
Table 11-6. Handled Class Requests

Request Definition
Configures DTE rate, stop bits, parity and number of
SET_LINE_CODING
character bits.
Requests current DTE rate, stop bits, parity and number
GET_LINE_CODING
of character bits.
RS-232 signal used to tell the DCE device the DTE
SET_CONTROL_LINE_STATE
device is now present.
Unhandled requests are STALLed.
11.5.3.4 Communication Endpoints

There are two communication endpoints and endpoint 0 is used for the enumeration process.
Endpoint 1 is a 64-byte Bulk OUT endpoint and endpoint 2 is a 64-byte Bulk IN endpoint. SAM-
BA Boot commands are sent by the host through endpoint 1. If required, the message is split by
the host into several data payloads by the host driver.
If the command requires a response, the host can send IN transactions to pick up the response.
68 AT91SAM9G45
AT91SAM9G45
12. Reset Controller (RSTC)
12.1 Description
The Reset Controller (RSTC), based on power-on reset cells, handles all the resets of the sys-
tem without any external components. It reports which reset occurred last.
The Reset Controller also drives independently or simultaneously the external reset and the
peripheral and processor resets.

The Reset Controller is based on two Power-on-Reset cells, one on VDDBU and one on
VDDCORE.
The Reset Controller is capable to return to the software the source of the last reset, either a
general reset (VDDBU rising), a wake-up reset (VDDCORE rising), a software reset, a user
reset or a watchdog reset.
The Reset Controller controls the internal resets of the system and the NRST pin. The NRST pin
is bidirectional. It is handled by the on-chip reset controller and can be driven low to provide a
reset signal to the external components or asserted low externally to reset the microcontroller. It
will reset the Core and the peripherals except the Backup region. There is no constraint on the
length of the reset pulse and the reset controller can guarantee a minimum pulse length.
The NRST pin integrates a permanent pull-up resistor to VDDIOP0 of about 100 kOhms.
The configuration of the Reset Controller is saved as supplied on VDDBU.
12.3 Block Diagram
Figure 12-1. Reset Controller Block Diagram
Reset Controller
Main Supply
POR
rstc_irq
Backup Supply Startup
POR Counter
Reset
State
Manager
proc_nreset
user_reset
NRST
NRST periph_nreset
Manager
nrst_out
exter_nreset
backup_neset
WDRPROC
wd_fault
SLCK
69
12.4.1 Reset Controller Overview

The Reset Controller is made up of an NRST Manager, a Startup Counter and a Reset State
Manager. It runs at Slow Clock and generates the following reset signals:
• proc_nreset: Processor reset line. It also resets the Watchdog Timer.
• backup_nreset: Affects all the peripherals powered by VDDBU.
• periph_nreset: Affects the whole set of embedded peripherals.
• nrst_out: Drives the NRST pin.
These reset signals are asserted by the Reset Controller, either on external events or on soft-
ware action. The Reset State Manager controls the generation of reset signals and provides a
signal to the NRST Manager when an assertion of the NRST pin is required.
The NRST Manager shapes the NRST assertion during a programmable time, thus controlling
external device resets.
The startup counter waits for the complete crystal oscillator startup. The wait delay is given by
the crystal oscillator startup time maximum value that can be found in the section Crystal Oscil-
lator Characteristics in the Electrical Characteristics section of the product documentation.
The Reset Controller Mode Register (RSTC_MR), allowing the configuration of the Reset Con-
troller, is powered with VDDBU, so that its configuration is saved as long as VDDBU is on.
12.4.2 NRST Manager

The NRST Manager samples the NRST input pin and drives this pin low when required by the
Reset State Manager. Figure 12-2 shows the block diagram of the NRST Manager.
Figure 12-2. NRST Manager

RSTC_MR
URSTIEN
RSTC_SR
URSTS
rstc_irq
NRSTL RSTC_MR Other
interrupt
URSTEN
sources
user_reset
NRST RSTC_MR
ERSTL
nrst_out
External Reset Timer exter_nreset
12.4.2.1 NRST Signal or Interrupt

The NRST Manager samples the NRST pin at Slow Clock speed. When the line is detected low,
a User Reset is reported to the Reset State Manager.
However, the NRST Manager can be programmed to not trigger a reset when an assertion of
NRST occurs. Writing the bit URSTEN at 0 in RSTC_MR disables the User Reset trigger.
70 AT91SAM9G45
AT91SAM9G45
The level of the pin NRST can be read at any time in the bit NRSTL (NRST level) in RSTC_SR.
As soon as the pin NRST is asserted, the bit URSTS in RSTC_SR is set. This bit clears only
when RSTC_SR is read.
The Reset Controller can also be programmed to generate an interrupt instead of generating a
reset. To do so, the bit URSTIEN in RSTC_MR must be written at 1.
12.4.2.2 NRST External Reset Control

The Reset State Manager asserts the signal ext_nreset to assert the NRST pin. When this
occurs, the “nrst_out” signal is driven low by the NRST Manager for a time programmed by the
field ERSTL in RSTC_MR. This assertion duration, named EXTERNAL_RESET_LENGTH, lasts
2(ERSTL+1) Slow Clock cycles. This gives the approximate duration of an assertion between 60 μs
and 2 seconds. Note that ERSTL at 0 defines a two-cycle duration for the NRST pulse.
This feature allows the Reset Controller to shape the NRST pin level, and thus to guarantee that
the NRST line is driven low for a time compliant with potential external devices connected on the
system reset.
As the field is within RSTC_MR, which is backed-up, this field can be used to shape the system
power-up reset for devices requiring a longer startup time than the Slow Clock Oscillator.
12.4.3 BMS Sampling

The product matrix manages a boot memory that depends on the level on the BMS pin at reset.
The BMS signal is sampled three slow clock cycles after the Core Power-On-Reset output rising
edge.
Figure 12-3. BMS Sampling
SLCK
Core Supply
POR output
XXX H or L
BMS Signal
BMS sampling delay
= 3 cycles
proc_nreset
12.4.4 Reset States

The Reset State Manager handles the different reset sources and generates the internal reset
signals. It reports the reset status in the field RSTTYP of the Status Register (RSTC_SR). The
update of the field RSTTYP is performed when the processor reset is released.
12.4.4.1 General Reset

A general reset occurs when VDDBU and VDDCORE are powered on. The backup supply POR
cell output rises and is filtered with a Startup Counter, which operates at Slow Clock. The pur-
pose of this counter is to make sure the Slow Clock oscillator is stable before starting up the
71
device. The length of startup time is hardcoded to comply with the Slow Clock Oscillator startup
time.
After this time, the processor clock is released at Slow Clock and all the other signals remain
valid for 3 cycles for proper processor and logic reset. Then, all the reset signals are released
and the field RSTTYP in RSTC_SR reports a General Reset. As the RSTC_MR is reset, the
NRST line rises 2 cycles after the backup_nreset, as ERSTL defaults at value 0x0.
When VDDBU is detected low by the Backup Supply POR Cell, all resets signals are immedi-
ately asserted, even if the Main Supply POR Cell does not report a Main Supply shutdown.
VDDBU only activates the backup_nreset signal.
The backup_nreset must be released so that any other reset can be generated by VDDCORE
(Main Supply POR output).
Figure 12-4 shows how the General Reset affects the reset signals.
Figure 12-4. General Reset State
SLCK
Any
MCK Freq.
Backup Supply
POR output
Startup Time
Main Supply
POR output
backup_nreset
Processor Startup
= 3 cycles
proc_nreset
RSTTYP XXX 0x0 = General Reset XXX
periph_nreset
NRST
(nrst_out)
BMS Sampling
EXTERNAL RESET LENGTH

= 2 cycles
72 AT91SAM9G45
AT91SAM9G45
12.4.4.2 Wake-up Reset

The Wake-up Reset occurs when the Main Supply is down. When the Main Supply POR output
is active, all the reset signals are asserted except backup_nreset. When the Main Supply pow-
ers up, the POR output is resynchronized on Slow Clock. The processor clock is then re-enabled
during 3 Slow Clock cycles, depending on the requirements of the ARM processor.
At the end of this delay, the processor and other reset signals rise. The field RSTTYP in
RSTC_SR is updated to report a Wake-up Reset.
The “nrst_out” remains asserted for EXTERNAL_RESET_LENGTH cycles. As RSTC_MR is
backed-up, the programmed number of cycles is applicable.
When the Main Supply is detected falling, the reset signals are immediately asserted. This tran-
sition is synchronous with the output of the Main Supply POR.
Figure 12-5. Wake-up State
SLCK
Any
MCK Freq.
Main Supply
POR output
backup_nreset
Resynch. Processor Startup
2 cycles = 3 cycles
proc_nreset
RSTTYP XXX 0x1 = WakeUp Reset XXX
periph_nreset
NRST
(nrst_out)

= 4 cycles (ERSTL = 1)
12.4.4.3 User Reset

The User Reset is entered when a low level is detected on the NRST pin and the bit URSTEN in
RSTC_MR is at 1. The NRST input signal is resynchronized with SLCK to insure proper behav-
ior of the system.
The User Reset is entered as soon as a low level is detected on NRST. The Processor Reset
and the Peripheral Reset are asserted.
The User Reset is left when NRST rises, after a two-cycle resynchronization time and a 3-cycle
processor startup. The processor clock is re-enabled as soon as NRST is confirmed high.
73
When the processor reset signal is released, the RSTTYP field of the Status Register
(RSTC_SR) is loaded with the value 0x4, indicating a User Reset.
The NRST Manager guarantees that the NRST line is asserted for
EXTERNAL_RESET_LENGTH Slow Clock cycles, as programmed in the field ERSTL. How-
ever, if NRST does not rise after EXTERNAL_RESET_LENGTH because it is driven low
externally, the internal reset lines remain asserted until NRST actually rises.
Figure 12-6. User Reset State
SLCK
Any
MCK Freq.
NRST
Resynch. Resynch. Processor Startup

2 cycles 2 cycles = 3 cycles
proc_nreset
RSTTYP Any XXX 0x4 = User Reset
periph_nreset
NRST
(nrst_out)
>= EXTERNAL RESET LENGTH
12.4.4.4 Software Reset

The Reset Controller offers several commands used to assert the different reset signals. These
commands are performed by writing the Control Register (RSTC_CR) with the following bits at
1:
• PROCRST: Writing PROCRST at 1 resets the processor and the watchdog timer.
• PERRST: Writing PERRST at 1 resets all the embedded peripherals, including the memory
system, and, in particular, the Remap Command. The Peripheral Reset is generally used for
debug purposes.
Except for Debug purposes, PERRST must always be used in conjunction with PROCRST
(PERRST and PROCRST set both at 1 simultaneously.)
• EXTRST: Writing EXTRST at 1 asserts low the NRST pin during a time defined by the field
ERSTL in the Mode Register (RSTC_MR).
The software reset is entered if at least one of these bits is set by the software. All these com-
mands can be performed independently or simultaneously. The software reset lasts 3 Slow
Clock cycles.
74 AT91SAM9G45
AT91SAM9G45
The internal reset signals are asserted as soon as the register write is performed. This is
detected on the Master Clock (MCK). They are released when the software reset is left, i.e.; syn-
chronously to SLCK.
If EXTRST is set, the nrst_out signal is asserted depending on the programming of the field
ERSTL. However, the resulting falling edge on NRST does not lead to a User Reset.
If and only if the PROCRST bit is set, the Reset Controller reports the software status in the field
RSTTYP of the Status Register (RSTC_SR). Other Software Resets are not reported in
RSTTYP.
As soon as a software operation is detected, the bit SRCMP (Software Reset Command in Prog-
ress) is set in the Status Register (RSTC_SR). It is cleared as soon as the software reset is left.
No other software reset can be performed while the SRCMP bit is set, and writing any value in
RSTC_CR has no effect.
Figure 12-7. Software Reset
SLCK
Any
MCK Freq.
Write RSTC_CR
Resynch. Processor Startup
1 cycle = 3 cycles
proc_nreset
if PROCRST=1
RSTTYP Any XXX 0x3 = Software Reset
periph_nreset
if PERRST=1
NRST
(nrst_out)
if EXTRST=1
8 cycles (ERSTL=2)
SRCMP in RSTC_SR
12.4.4.5 Watchdog Reset

The Watchdog Reset is entered when a watchdog fault occurs. This state lasts 3 Slow Clock
cycles.
When in Watchdog Reset, assertion of the reset signals depends on the WDRPROC bit in
WDT_MR:
• If WDRPROC is 0, the Processor Reset and the Peripheral Reset are asserted. The NRST
line is also asserted, depending on the programming of the field ERSTL. However, the
resulting low level on NRST does not result in a User Reset state.
75
• If WDRPROC = 1, only the processor reset is asserted.
The Watchdog Timer is reset by the proc_nreset signal. As the watchdog fault always causes a
processor reset if WDRSTEN is set, the Watchdog Timer is always reset after a Watchdog
Reset, and the Watchdog is enabled by default and with a period set to a maximum.
When the WDRSTEN in WDT_MR bit is reset, the watchdog fault has no impact on the reset
controller.
Figure 12-8. Watchdog Reset
SLCK
Any
MCK Freq.
wd_fault
Processor Startup
= 3 cycles
proc_nreset
RSTTYP Any XXX 0x2 = Watchdog Reset
periph_nreset
Only if
WDRPROC = 0
NRST
(nrst_out)

8 cycles (ERSTL=2)
12.4.5 Reset State Priorities

The Reset State Manager manages the following priorities between the different reset sources,
given in descending order:
• Backup Reset
• Wake-up Reset
• Watchdog Reset
• Software Reset
• User Reset
Particular cases are listed below:
• When in User Reset:
– A watchdog event is impossible because the Watchdog Timer is being reset by the
proc_nreset signal.
– A software reset is impossible, since the processor reset is being activated.
• When in Software Reset:
– A watchdog event has priority over the current state.
– The NRST has no effect.
76 AT91SAM9G45
AT91SAM9G45
• When in Watchdog Reset:

– The processor reset is active and so a Software Reset cannot be programmed.
– A User Reset cannot be entered.
12.4.6 Reset Controller Status Register

The Reset Controller status register (RSTC_SR) provides several status fields:
• RSTTYP field: This field gives the type of the last reset, as explained in previous sections.
• SRCMP bit: This field indicates that a Software Reset Command is in progress and that no
further software reset should be performed until the end of the current one. This bit is
automatically cleared at the end of the current software reset.
• NRSTL bit: The NRSTL bit of the Status Register gives the level of the NRST pin sampled on
each MCK rising edge.
• URSTS bit: A high-to-low transition of the NRST pin sets the URSTS bit of the RSTC_SR
register. This transition is also detected on the Master Clock (MCK) rising edge (see Figure
12-9). If the User Reset is disabled (URSTEN = 0) and if the interruption is enabled by the
URSTIEN bit in the RSTC_MR register, the URSTS bit triggers an interrupt. Reading the
RSTC_SR status register resets the URSTS bit and clears the interrupt.
Figure 12-9. Reset Controller Status and Interrupt
MCK
read
Peripheral Access RSTC_SR
2 cycle 2 cycle
resynchronization resynchronization
NRST
NRSTL
URSTS
rstc_irq
if (URSTEN = 0) and
(URSTIEN = 1)
77
12.5 Reset Controller (RSTC) User Interface
Table 12-1. Register Mapping

Offset Register Name Access Reset Back-up Reset
0x00 Control Register RSTC_CR Write-only -
0x04 Status Register RSTC_SR Read-only 0x0000_0001 0x0000_0000
0x08 Mode Register RSTC_MR Read-write - 0x0000_0000
Note: 1. The reset value of RSTC_SR either reports a General Reset or a Wake-up Reset depending on last rising power supply.
78 AT91SAM9G45
AT91SAM9G45
12.5.1 Reset Controller Control Register

Name: RSTC_CR
Address: 0xFFFFFD00
Access Type: Write-only
31 30 29 28 27 26 25 24
KEY
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – –
7 6 5 4 3 2 1 0
– – – – EXTRST PERRST – PROCRST
• PROCRST: Processor Reset

0 = No effect.
1 = If KEY is correct, resets the processor.
• PERRST: Peripheral Reset

0 = No effect.
1 = If KEY is correct, resets the peripherals.
• EXTRST: External Reset

0 = No effect.
1 = If KEY is correct, asserts the NRST pin.
• KEY: Password
Should be written at value 0xA5. Writing any other value in this field aborts the write operation.
79
12.5.2 Reset Controller Status Register
Name: RSTC_SR
Address: 0xFFFFFD04
Access Type: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – SRCMP NRSTL
15 14 13 12 11 10 9 8
– – – – – RSTTYP
7 6 5 4 3 2 1 0
– – – – – – – URSTS
• URSTS: User Reset Status

0 = No high-to-low edge on NRST happened since the last read of RSTC_SR.
1 = At least one high-to-low transition of NRST has been detected since the last read of RSTC_SR.
• RSTTYP: Reset Type

Reports the cause of the last processor reset. Reading this RSTC_SR does not reset this field.
RSTTYP Reset Type Comments

0 0 0 General Reset Both VDDCORE and VDDBU rising
0 0 1 Wake Up Reset VDDCORE rising
0 1 0 Watchdog Reset Watchdog fault occurred
0 1 1 Software Reset Processor reset required by the software
1 0 0 User Reset NRST pin detected low
• NRSTL: NRST Pin Level

Registers the NRST Pin Level at Master Clock (MCK).
• SRCMP: Software Reset Command in Progress

0 = No software command is being performed by the reset controller. The reset controller is ready for a software command.
1 = A software reset command is being performed by the reset controller. The reset controller is busy.
80 AT91SAM9G45
AT91SAM9G45
12.5.3 Reset Controller Mode Register

Name: RSTC_MR
Address: 0xFFFFFD08
Access Type: Read-write
31 30 29 28 27 26 25 24
KEY
23 22 21 20 19 18 17 16
– – – – – – –
15 14 13 12 11 10 9 8
– – – – ERSTL
7 6 5 4 3 2 1 0
– – URSTIEN – – – URSTEN
• URSTEN: User Reset Enable

0 = The detection of a low level on the pin NRST does not generate a User Reset.
1 = The detection of a low level on the pin NRST triggers a User Reset.
• URSTIEN: User Reset Interrupt Enable

0 = USRTS bit in RSTC_SR at 1 has no effect on rstc_irq.
1 = USRTS bit in RSTC_SR at 1 asserts rstc_irq if URSTEN = 0.
• ERSTL: External Reset Length

This field defines the external reset length. The external reset is asserted during a time of 2(ERSTL+1) Slow Clock cycles. This
allows assertion duration to be programmed between 60 μs and 2 seconds.
• KEY: Password
81
82 AT91SAM9G45
AT91SAM9G45
13. Real-time Timer (RTT)
13.1 Description
The Real-time Timer is built around a 32-bit counter and used to count elapsed seconds. It gen-
erates a periodic interrupt and/or triggers an alarm on a programmed value.

• Real-Time Timer, allowing backup of time with different accuracies
– 32-bit Free-running back-up Counter
– Integrates a 16-bit programmable prescaler running on slow clock
– Alarm Register capable to generate a wake-up of the system through the Shut Down
Controller
13.3 Block Diagram
Figure 13-1. Real-time Timer

RTT_MR RTT_MR
RTTRST RTPRES
RTT_MR
reload RTTINCIEN
SLCK 16-bit
Divider
0 set
RTT_MR RTT_SR RTTINC
RTTRST 1 0 reset
rtt_int
32-bit
Counter read
RTT_MR
RTT_SR
ALMIEN
reset
RTT_VR CRTV
RTT_SR ALMS
set
rtt_alarm
=
RTT_AR ALMV

The Real-time Timer is used to count elapsed seconds. It is built around a 32-bit counter fed by
Slow Clock divided by a programmable 16-bit value. The value can be programmed in the field
RTPRES of the Real-time Mode Register (RTT_MR).
Programming RTPRES at 0x00008000 corresponds to feeding the real-time counter with a 1 Hz
signal (if the Slow Clock is 32.768 kHz). The 32-bit counter can count up to 232 seconds, corre-
sponding to more than 136 years, then roll over to 0.
83
The Real-time Timer can also be used as a free-running timer with a lower time-base. The best
accuracy is achieved by writing RTPRES to 3. Programming RTPRES to 1 or 2 is possible, but
may result in losing status events because the status register is cleared two Slow Clock cycles
after read. Thus if the RTT is configured to trigger an interrupt, the interrupt occurs during 2 Slow
Clock cycles after reading RTT_SR. To prevent several executions of the interrupt handler, the
interrupt must be disabled in the interrupt handler and re-enabled when the status register is
clear.
The Real-time Timer value (CRTV) can be read at any time in the register RTT_VR (Real-time
Value Register). As this value can be updated asynchronously from the Master Clock, it is advis-
able to read this register twice at the same value to improve accuracy of the returned value.
The current value of the counter is compared with the value written in the alarm register
RTT_AR (Real-time Alarm Register). If the counter value matches the alarm, the bit ALMS in
RTT_SR is set. The alarm register is set to its maximum value, corresponding to 0xFFFF_FFFF,
after a reset.
The bit RTTINC in RTT_SR is set each time the Real-time Timer counter is incremented. This bit
can be used to start a periodic interrupt, the period being one second when the RTPRES is pro-
grammed with 0x8000 and Slow Clock equal to 32.768 Hz.
Reading the RTT_SR status register resets the RTTINC and ALMS fields.
Writing the bit RTTRST in RTT_MR immediately reloads and restarts the clock divider with the
new programmed value. This also resets the 32-bit counter.
Note: Because of the asynchronism between the Slow Clock (SCLK) and the System Clock (MCK):
1) The restart of the counter and the reset of the RTT_VR current value register is effective only 2
slow clock cycles after the write of the RTTRST bit in the RTT_MR register.
2) The status register flags reset is taken into account only 2 slow clock cycles after the read of the
RTT_SR (Status Register).
84 AT91SAM9G45
AT91SAM9G45
Figure 13-2. RTT Counting

APB cycle APB cycle
SCLK
RTPRES - 1
Prescaler
RTT 0 ... ALMV-1 ALMV ALMV+1 ALMV+2 ALMV+3
RTTINC (RTT_SR)
ALMS (RTT_SR)
APB Interface
read RTT_SR
85
13.5 Real-time Timer (RTT) User Interface

Offset Register Name Access Reset
0x00 Mode Register RTT_MR Read-write 0x0000_8000
0x04 Alarm Register RTT_AR Read-write 0xFFFF_FFFF
0x08 Value Register RTT_VR Read-only 0x0000_0000
0x0C Status Register RTT_SR Read-only 0x0000_0000
86 AT91SAM9G45
AT91SAM9G45
13.5.1 Real-time Timer Mode Register

Register Name: RTT_MR
Address: 0xFFFFFD20
Access Type: Read/Write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – RTTRST RTTINCIEN ALMIEN
15 14 13 12 11 10 9 8
RTPRES
7 6 5 4 3 2 1 0
RTPRES
• RTPRES: Real-time Timer Prescaler Value

Defines the number of SLCK periods required to increment the Real-time timer. RTPRES is defined as follows:
RTPRES = 0: The prescaler period is equal to 216.
RTPRES ≠ 0: The prescaler period is equal to RTPRES.
• ALMIEN: Alarm Interrupt Enable

0 = The bit ALMS in RTT_SR has no effect on interrupt.
1 = The bit ALMS in RTT_SR asserts interrupt.
• RTTINCIEN: Real-time Timer Increment Interrupt Enable

0 = The bit RTTINC in RTT_SR has no effect on interrupt.
1 = The bit RTTINC in RTT_SR asserts interrupt.
• RTTRST: Real-time Timer Restart

1 = Reloads and restarts the clock divider with the new programmed value. This also resets the 32-bit counter.
87
13.5.2 Real-time Timer Alarm Register
Register Name: RTT_AR
Address: 0xFFFFFD24
31 30 29 28 27 26 25 24
ALMV
23 22 21 20 19 18 17 16
ALMV
15 14 13 12 11 10 9 8
ALMV
7 6 5 4 3 2 1 0
ALMV
• ALMV: Alarm Value

Defines the alarm value (ALMV+1) compared with the Real-time Timer.
13.5.3 Real-time Timer Value Register

Register Name: RTT_VR
Address: 0xFFFFFD28
31 30 29 28 27 26 25 24
CRTV
23 22 21 20 19 18 17 16
CRTV
15 14 13 12 11 10 9 8
CRTV
7 6 5 4 3 2 1 0
CRTV
• CRTV: Current Real-time Value

Returns the current value of the Real-time Timer.
88 AT91SAM9G45
AT91SAM9G45
13.5.4 Real-time Timer Status Register

Register Name: RTT_SR
Address: 0xFFFFFD2C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – RTTINC ALMS
• ALMS: Real-time Alarm Status

0 = The Real-time Alarm has not occured since the last read of RTT_SR.
1 = The Real-time Alarm occured since the last read of RTT_SR.
• RTTINC: Real-time Timer Increment

0 = The Real-time Timer has not been incremented since the last read of the RTT_SR.
1 = The Real-time Timer has been incremented since the last read of the RTT_SR.
89
90 AT91SAM9G45
AT91SAM9G45
13.6 Real-time Clock (RTC)
13.7 Description
The Real-time Clock (RTC) peripheral is designed for very low power consumption.
It combines a complete time-of-day clock with alarm and a two-hundred-year Gregorian calen-
dar, complemented by a programmable periodic interrupt. The alarm and calendar registers are
accessed by a 32-bit data bus.
The time and calendar values are coded in binary-coded decimal (BCD) format. The time format
can be 24-hour mode or 12-hour mode with an AM/PM indicator.
Updating time and calendar fields and configuring the alarm fields are performed by a parallel
capture on the 32-bit data bus. An entry control is performed to avoid loading registers with
incompatible BCD format data or with an incompatible date according to the current
month/year/century.

• Low power consumption
• Full asynchronous design
• Two hundred year calendar
• Programmable Periodic Interrupt
• Alarm and update parallel load
• Control of alarm and update Time/Calendar Data In
13.9 Block Diagram
Figure 13-3. RTC Block Diagram
Crystal Oscillator: SLCK 32768 Divider Time Date
Bus Interface Bus Interface
Entry Interrupt RTC Interrupt

Control Control
91
13.10 Product Dependencies
13.10.1 Power Management

The Real-time Clock is continuously clocked at 32768 Hz. The Power Management Controller
has no effect on RTC behavior.
13.10.2 Interrupt
The RTC Interrupt is connected to interrupt source 1 (IRQ1) of the advanced interrupt controller.
This interrupt line is due to the OR-wiring of the system peripheral interrupt lines (System Timer,
Real Time Clock, Power Management Controller, Memory Controller, etc.). When a system
interrupt occurs, the service routine must first determine the cause of the interrupt. This is done
by reading the status registers of the above system peripherals successively.

The RTC provides a full binary-coded decimal (BCD) clock that includes century (19/20), year
(with leap years), month, date, day, hours, minutes and seconds.
The valid year range is 1900 to 2099, a two-hundred-year Gregorian calendar achieving full Y2K
compliance.
The RTC can operate in 24-hour mode or in 12-hour mode with an AM/PM indicator.
Corrections for leap years are included (all years divisible by 4 being leap years, including year
2000). This is correct up to the year 2099.
After hardware reset, the calendar is initialized to Thursday, January 1, 1998.
13.11.1 Reference Clock

The reference clock is Slow Clock (SLCK). It can be driven internally or by an external 32.768
kHz crystal.
During low power modes of the processor (idle mode), the oscillator runs and power consump-
tion is critical. The crystal selection has to take into account the current consumption for power
saving and the frequency drift due to temperature effect on the circuit for time accuracy.
13.11.2 Timing
The RTC is updated in real time at one-second intervals in normal mode for the counters of sec-
onds, at one-minute intervals for the counter of minutes and so on.
Due to the asynchronous operation of the RTC with respect to the rest of the chip, to be certain
that the value read in the RTC registers (century, year, month, date, day, hours, minutes, sec-
onds) are valid and stable, it is necessary to read these registers twice. If the data is the same
both times, then it is valid. Therefore, a minimum of two and a maximum of three accesses are
required.
13.11.3 Alarm
The RTC has five programmable fields: month, date, hours, minutes and seconds.
Each of these fields can be enabled or disabled to match the alarm condition:
• If all the fields are enabled, an alarm flag is generated (the corresponding flag is asserted
and an interrupt generated if enabled) at a given month, date, hour/minute/second.
• If only the “seconds” field is enabled, then an alarm is generated every minute.
92 AT91SAM9G45
Depending on the combination of fields enabled, a large number of possibilities are available to
the user ranging from minutes to 365/366 days.
13.11.4 Error Checking

Verification on user interface data is performed when accessing the century, year, month, date,
day, hours, minutes, seconds and alarms. A check is performed on illegal BCD entries such as
illegal date of the month with regard to the year and century configured.
If one of the time fields is not correct, the data is not loaded into the register/counter and a flag is
set in the validity register. The user can not reset this flag. It is reset as soon as an acceptable
value is programmed. This avoids any further side effects in the hardware. The same procedure
is done for the alarm.
The following checks are performed:
1. Century (check if it is in range 19 - 20)
2. Year (BCD entry check)
3. Date (check range 01 - 31)
4. Month (check if it is in BCD range 01 - 12, check validity regarding “date”)
5. Day (check range 1 - 7)
6. Hour (BCD checks: in 24-hour mode, check range 00 - 23 and check that AM/PM flag is
not set if RTC is set in 24-hour mode; in 12-hour mode check range 01 - 12)
7. Minute (check BCD and range 00 - 59)
8. Second (check BCD and range 00 - 59)
Note: If the 12-hour mode is selected by means of the RTC_MODE register, a 12-hour value can be pro-
grammed and the returned value on RTC_TIME will be the corresponding 24-hour value. The
entry control checks the value of the AM/PM indicator (bit 22 of RTC_TIME register) to determine
the range to be checked.
13.11.5 Updating Time/Calendar

To update any of the time/calendar fields, the user must first stop the RTC by setting the corre-
sponding field in the Control Register. Bit UPDTIM must be set to update time fields (hour,
minute, second) and bit UPDCAL must be set to update calendar fields (century, year, month,
date, day).
Then the user must poll or wait for the interrupt (if enabled) of bit ACKUPD in the Status Regis-
ter. Once the bit reads 1, it is mandatory to clear this flag by writing the corresponding bit in
RTC_SCCR. The user can now write to the appropriate Time and Calendar register.
Once the update is finished, the user must reset (0) UPDTIM and/or UPDCAL in the Control
When entering programming mode of the calendar fields, the time fields remain enabled. When
entering the programming mode of the time fields, both time and calendar fields are stopped.
This is due to the location of the calendar logic circuity (downstream for low-power consider-
ations). It is highly recommended to prepare all the fields to be updated before entering
programming mode. In successive update operations, the user must wait at least one second
after resetting the UPDTIM/UPDCAL bit in the RTC_CR (Control Register) before setting these
bits again. This is done by waiting for the SEC flag in the Status Register before setting
UPDTIM/UPDCAL bit. After resetting UPDTIM/UPDCAL, the SEC flag must also be cleared.
93 AT91SAM9G45
Figure 13-4. Update Sequence
Begin
Prepare TIme or Calendar Fields
Set UPDTIM and/or UPDCAL

bit(s) in RTC_CR
Read RTC_SR
Polling or
IRQ (if enabled)
No
ACKUPD
=1?
Yes
Clear ACKUPD bit in RTC_SCCR
Update Time andor Calendar values in

RTC_TIMR/RTC_CALR
Clear UPDTIM and/or UPDCAL bit in

RTC_CR
End
94 AT91SAM9G45
13.12 Real Time Clock (RTC) User Interface

0x00 Control Register RTC_CR Read-write 0x0
0x04 Mode Register RTC_MR Read-write 0x0
0x08 Time Register RTC_TIMR Read-write 0x0
0x0C Calendar Register RTC_CALR Read-write 0x01819819
0x10 Time Alarm Register RTC_TIMALR Read-write 0x0
0x14 Calendar Alarm Register RTC_CALALR Read-write 0x01010000
0x18 Status Register RTC_SR Read-only 0x0
0x1C Status Clear Command Register RTC_SCCR Write-only ---
0x20 Interrupt Enable Register RTC_IER Write-only ---
0x24 Interrupt Disable Register RTC_IDR Write-only ---
0x28 Interrupt Mask Register RTC_IMR Read-only 0x0
0x2C Valid Entry Register RTC_VER Read-only 0x0
0xFC Reserved Register – – –
95 AT91SAM9G45
13.12.1 RTC Control Register
Name: RTC_CR
Address: 0xFFFFFDB0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – CALEVSEL
15 14 13 12 11 10 9 8
– – – – – – TIMEVSEL
7 6 5 4 3 2 1 0
– – – – – – UPDCAL UPDTIM
• UPDTIM: Update Request Time Register

0 = No effect.
1 = Stops the RTC time counting.
Time counting consists of second, minute and hour counters. Time counters can be programmed once this bit is set and
acknowledged by the bit ACKUPD of the Status Register.
• UPDCAL: Update Request Calendar Register

0 = No effect.
1 = Stops the RTC calendar counting.
Calendar counting consists of day, date, month, year and century counters. Calendar counters can be programmed once
this bit is set.
• TIMEVSEL: Time Event Selection

The event that generates the flag TIMEV in RTC_SR (Status Register) depends on the value of TIMEVSEL.
0 = Minute change.
1 = Hour change.
2 = Every day at midnight.
3 = Every day at noon.
• CALEVSEL: Calendar Event Selection

The event that generates the flag CALEV in RTC_SR depends on the value of CALEVSEL.
0 = Week change (every Monday at time 00:00:00).
1 = Month change (every 01 of each month at time 00:00:00).
2, 3 = Year change (every January 1 at time 00:00:00).
96 AT91SAM9G45
13.12.2 RTC Mode Register
Name: RTC_MR
Address: 0xFFFFFDB4
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – HRMOD
• HRMOD: 12-/24-hour Mode

0 = 24-hour mode is selected.
1 = 12-hour mode is selected.
All non-significant bits read zero.
97 AT91SAM9G45
13.12.3 RTC Time Register
Name: RTC_TIMR
Address: 0xFFFFFDB8
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– AMPM HOUR
15 14 13 12 11 10 9 8
– MIN
7 6 5 4 3 2 1 0
– SEC
• SEC: Current Second

The range that can be set is 0 - 59 (BCD).
The lowest four bits encode the units. The higher bits encode the tens.
• MIN: Current Minute

• HOUR: Current Hour

The range that can be set is 1 - 12 (BCD) in 12-hour mode or 0 - 23 (BCD) in 24-hour mode.
• AMPM: Ante Meridiem Post Meridiem Indicator

This bit is the AM/PM indicator in 12-hour mode.
0 = AM.
1 = PM.
98 AT91SAM9G45
13.12.4 RTC Calendar Register
Name: RTC_CALR
Address: 0xFFFFFDBC
31 30 29 28 27 26 25 24
– – DATE
23 22 21 20 19 18 17 16
DAY MONTH
15 14 13 12 11 10 9 8
YEAR
7 6 5 4 3 2 1 0
– CENT
• CENT: Current Century

• YEAR: Current Year

• MONTH: Current Month

• DAY: Current Day in Current Week

The coding of the number (which number represents which day) is user-defined as it has no effect on the date counter.
• DATE: Current Day in Current Month

99 AT91SAM9G45
13.12.5 RTC Time Alarm Register
Name: RTC_TIMALR
Address: 0xFFFFFDC0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
HOUREN AMPM HOUR
15 14 13 12 11 10 9 8
MINEN MIN
7 6 5 4 3 2 1 0
SECEN SEC
• SEC: Second Alarm

This field is the alarm field corresponding to the BCD-coded second counter.
• SECEN: Second Alarm Enable

0 = The second-matching alarm is disabled.
1 = The second-matching alarm is enabled.
• MIN: Minute Alarm

This field is the alarm field corresponding to the BCD-coded minute counter.
• MINEN: Minute Alarm Enable

0 = The minute-matching alarm is disabled.
1 = The minute-matching alarm is enabled.
• HOUR: Hour Alarm

This field is the alarm field corresponding to the BCD-coded hour counter.
• AMPM: AM/PM Indicator

This field is the alarm field corresponding to the BCD-coded hour counter.
• HOUREN: Hour Alarm Enable

0 = The hour-matching alarm is disabled.
1 = The hour-matching alarm is enabled.
100 AT91SAM9G45
13.12.6 RTC Calendar Alarm Register
Name: RTC_CALALR
Address: 0xFFFFFDC4
31 30 29 28 27 26 25 24
DATEEN – DATE
23 22 21 20 19 18 17 16
MTHEN – – MONTH
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – –
• MONTH: Month Alarm

This field is the alarm field corresponding to the BCD-coded month counter.
• MTHEN: Month Alarm Enable

0 = The month-matching alarm is disabled.
1 = The month-matching alarm is enabled.
• DATE: Date Alarm

This field is the alarm field corresponding to the BCD-coded date counter.
• DATEEN: Date Alarm Enable

0 = The date-matching alarm is disabled.
1 = The date-matching alarm is enabled.
101 AT91SAM9G45
13.12.7 RTC Status Register
Name: RTC_SR
Address: 0xFFFFFDC8
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – CALEV TIMEV SEC ALARM ACKUPD
• ACKUPD: Acknowledge for Update

0 = Time and calendar registers cannot be updated.
1 = Time and calendar registers can be updated.
• ALARM: Alarm Flag

0 = No alarm matching condition occurred.
1 = An alarm matching condition has occurred.
• SEC: Second Event

0 = No second event has occurred since the last clear.
1 = At least one second event has occurred since the last clear.
• TIMEV: Time Event

0 = No time event has occurred since the last clear.
1 = At least one time event has occurred since the last clear.
The time event is selected in the TIMEVSEL field in RTC_CTRL (Control Register) and can be any one of the following
events: minute change, hour change, noon, midnight (day change).
• CALEV: Calendar Event

0 = No calendar event has occurred since the last clear.
1 = At least one calendar event has occurred since the last clear.
The calendar event is selected in the CALEVSEL field in RTC_CR and can be any one of the following events: week
change, month change and year change.
102 AT91SAM9G45
13.12.8 RTC Status Clear Command Register
Name: RTC_SCCR
Address: 0xFFFFFDCC
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – CALCLR TIMCLR SECCLR ALRCLR ACKCLR
• ACKCLR: Acknowledge Clear

0 = No effect.
1 = Clears corresponding status flag in the Status Register (RTC_SR).
• ALRCLR: Alarm Clear

0 = No effect.
• SECCLR: Second Clear

0 = No effect.
• TIMCLR: Time Clear

0 = No effect.
• CALCLR: Calendar Clear

0 = No effect.
103 AT91SAM9G45
13.12.9 RTC Interrupt Enable Register
Name: RTC_IER
Address: 0xFFFFFDD0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – CALEN TIMEN SECEN ALREN ACKEN
• ACKEN: Acknowledge Update Interrupt Enable

0 = No effect.
1 = The acknowledge for update interrupt is enabled.
• ALREN: Alarm Interrupt Enable

0 = No effect.
1 = The alarm interrupt is enabled.
• SECEN: Second Event Interrupt Enable

0 = No effect.
1 = The second periodic interrupt is enabled.
• TIMEN: Time Event Interrupt Enable

0 = No effect.
1 = The selected time event interrupt is enabled.
• CALEN: Calendar Event Interrupt Enable

0 = No effect.
• 1 = The selected calendar event interrupt is enabled.
104 AT91SAM9G45
13.12.10 RTC Interrupt Disable Register
Name: RTC_IDR
Address: 0xFFFFFDD4
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – CALDIS TIMDIS SECDIS ALRDIS ACKDIS
• ACKDIS: Acknowledge Update Interrupt Disable

0 = No effect.
1 = The acknowledge for update interrupt is disabled.
• ALRDIS: Alarm Interrupt Disable

0 = No effect.
1 = The alarm interrupt is disabled.
• SECDIS: Second Event Interrupt Disable

0 = No effect.
1 = The second periodic interrupt is disabled.
• TIMDIS: Time Event Interrupt Disable

0 = No effect.
1 = The selected time event interrupt is disabled.
• CALDIS: Calendar Event Interrupt Disable

0 = No effect.
1 = The selected calendar event interrupt is disabled.
105 AT91SAM9G45
13.12.11 RTC Interrupt Mask Register
Name: RTC_IMR
Address: 0xFFFFFDD8
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – CAL TIM SEC ALR ACK
• ACK: Acknowledge Update Interrupt Mask

0 = The acknowledge for update interrupt is disabled.
1 = The acknowledge for update interrupt is enabled.
• ALR: Alarm Interrupt Mask

0 = The alarm interrupt is disabled.
1 = The alarm interrupt is enabled.
• SEC: Second Event Interrupt Mask

0 = The second periodic interrupt is disabled.
1 = The second periodic interrupt is enabled.
• TIM: Time Event Interrupt Mask

0 = The selected time event interrupt is disabled.
1 = The selected time event interrupt is enabled.
• CAL: Calendar Event Interrupt Mask

0 = The selected calendar event interrupt is disabled.
1 = The selected calendar event interrupt is enabled.
106 AT91SAM9G45
13.12.12 RTC Valid Entry Register
Name: RTC_VER
Address: 0xFFFFFDDC
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – NVCALALR NVTIMALR NVCAL NVTIM
• NVTIM: Non-valid Time

0 = No invalid data has been detected in RTC_TIMR (Time Register).
1 = RTC_TIMR has contained invalid data since it was last programmed.
• NVCAL: Non-valid Calendar

0 = No invalid data has been detected in RTC_CALR (Calendar Register).
1 = RTC_CALR has contained invalid data since it was last programmed.
• NVTIMALR: Non-valid Time Alarm

0 = No invalid data has been detected in RTC_TIMALR (Time Alarm Register).
1 = RTC_TIMALR has contained invalid data since it was last programmed.
• NVCALALR: Non-valid Calendar Alarm

0 = No invalid data has been detected in RTC_CALALR (Calendar Alarm Register).
1 = RTC_CALALR has contained invalid data since it was last programmed.
107 AT91SAM9G45
108 AT91SAM9G45
AT91SAM9G45
14. Periodic Interval Timer (PIT)
14.1 Description
The Periodic Interval Timer (PIT) provides the operating system’s scheduler interrupt. It is
designed to offer maximum accuracy and efficient management, even for systems with long
response time .

• Includes a 20-bit Periodic Counter, with less than 1μs accuracy
• Includes a 12-bit Interval Overlay Counter
• Real Time OS or Linux/WinCE compliant tick generator
14.3 Block Diagram
Figure 14-1. Periodic Interval Timer

PIT_MR
PIV
=? PIT_MR
PITIEN
set
0 pit_irq
PIT_SR PITS
reset
0 0 1
12-bit
0 1 Adder
read PIT_PIVR
MCK 20-bit
Counter
MCK/16
Prescaler CPIV PIT_PIVR PICNT
CPIV PIT_PIIR PICNT

The Periodic Interval Timer aims at providing periodic interrupts for use by operating systems.
The PIT provides a programmable overflow counter and a reset-on-read feature. It is built
around two counters: a 20-bit CPIV counter and a 12-bit PICNT counter. Both counters work at
Master Clock /16.
109
The first 20-bit CPIV counter increments from 0 up to a programmable overflow value set in the
field PIV of the Mode Register (PIT_MR). When the counter CPIV reaches this value, it resets to
0 and increments the Periodic Interval Counter, PICNT. The status bit PITS in the Status Regis-
ter (PIT_SR) rises and triggers an interrupt, provided the interrupt is enabled (PITIEN in
PIT_MR).
Writing a new PIV value in PIT_MR does not reset/restart the counters.
When CPIV and PICNT values are obtained by reading the Periodic Interval Value Register
(PIT_PIVR), the overflow counter (PICNT) is reset and the PITS is cleared, thus acknowledging
the interrupt. The value of PICNT gives the number of periodic intervals elapsed since the last
read of PIT_PIVR.
When CPIV and PICNT values are obtained by reading the Periodic Interval Image Register
(PIT_PIIR), there is no effect on the counters CPIV and PICNT, nor on the bit PITS. For exam-
ple, a profiler can read PIT_PIIR without clearing any pending interrupt, whereas a timer
interrupt clears the interrupt by reading PIT_PIVR.
The PIT may be enabled/disabled using the PITEN bit in the PIT_MR register (disabled on
reset). The PITEN bit only becomes effective when the CPIV value is 0. Figure 14-2 illustrates
the PIT counting. After the PIT Enable bit is reset (PITEN= 0), the CPIV goes on counting until
the PIV value is reached, and is then reset. PIT restarts counting, only if the PITEN is set again.
The PIT is stopped when the core enters debug state.
Figure 14-2. Enabling/Disabling PIT with PITEN

APB cycle APB cycle
MCK
15
restarts MCK Prescaler
MCK Prescaler 0
PITEN
CPIV 0 1 PIV - 1 PIV 0 1
PICNT 0 1 0
PITS (PIT_SR)
APB Interface
read PIT_PIVR
110 AT91SAM9G45
AT91SAM9G45
14.5 Periodic Interval Timer (PIT) User Interface

0x00 Mode Register PIT_MR Read-write 0x000F_FFFF
0x04 Status Register PIT_SR Read-only 0x0000_0000
0x08 Periodic Interval Value Register PIT_PIVR Read-only 0x0000_0000
0x0C Periodic Interval Image Register PIT_PIIR Read-only 0x0000_0000
111
14.5.1 Periodic Interval Timer Mode Register
Register Name: PIT_MR
Address: 0xFFFFFD30
31 30 29 28 27 26 25 24
– – – – – – PITIEN PITEN
23 22 21 20 19 18 17 16
– – – – PIV
15 14 13 12 11 10 9 8
PIV
7 6 5 4 3 2 1 0
PIV
• PIV: Periodic Interval Value

Defines the value compared with the primary 20-bit counter of the Periodic Interval Timer (CPIV). The period is equal to
(PIV + 1).
• PITEN: Period Interval Timer Enabled
0 = The Periodic Interval Timer is disabled when the PIV value is reached.
1 = The Periodic Interval Timer is enabled.
• PITIEN: Periodic Interval Timer Interrupt Enable
0 = The bit PITS in PIT_SR has no effect on interrupt.
1 = The bit PITS in PIT_SR asserts interrupt.
112 AT91SAM9G45
AT91SAM9G45
14.5.2 Periodic Interval Timer Status Register

Register Name: PIT_SR
Address: 0xFFFFFD34
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – PITS
• PITS: Periodic Interval Timer Status

0 = The Periodic Interval timer has not reached PIV since the last read of PIT_PIVR.
1 = The Periodic Interval timer has reached PIV since the last read of PIT_PIVR.
14.5.3 Periodic Interval Timer Value Register

Register Name: PIT_PIVR
Address: 0xFFFFFD38
31 30 29 28 27 26 25 24
PICNT
23 22 21 20 19 18 17 16
PICNT CPIV
15 14 13 12 11 10 9 8
CPIV
7 6 5 4 3 2 1 0
CPIV
Reading this register clears PITS in PIT_SR.

• CPIV: Current Periodic Interval Value
Returns the current value of the periodic interval timer.
• PICNT: Periodic Interval Counter
Returns the number of occurences of periodic intervals since the last read of PIT_PIVR.
113
AT91SAM9G45
14.5.4 Periodic Interval Timer Image Register

Register Name: PIT_PIIR
Address: 0xFFFFFD3C
31 30 29 28 27 26 25 24
PICNT
23 22 21 20 19 18 17 16
PICNT CPIV
15 14 13 12 11 10 9 8
CPIV
7 6 5 4 3 2 1 0
CPIV
• CPIV: Current Periodic Interval Value

Returns the current value of the periodic interval timer.
• PICNT: Periodic Interval Counter
Returns the number of occurences of periodic intervals since the last read of PIT_PIVR.
114
AT91SAM9G45
15. Watchdog Timer (WDT)
15.1 Description
The Watchdog Timer can be used to prevent system lock-up if the software becomes trapped in
a deadlock. It features a 12-bit down counter that allows a watchdog period of up to 16 seconds
(slow clock at 32.768 kHz). It can generate a general reset or a processor reset only. In addition,
it can be stopped while the processor is in debug mode or idle mode.

• 16-bit key-protected only-once-Programmable Counter
• Windowed, prevents the processor to be in a dead-lock on the watchdog access
15.3 Block Diagram
Figure 15-1. Watchdog Timer Block Diagram
write WDT_MR
WDT_MR
WDV
WDT_CR
WDRSTT reload
1 0
12-bit Down
Counter
WDT_MR
reload
WDD Current
1/128 SLCK
Value
<= WDD
WDT_MR
WDRSTEN
= 0
wdt_fault
(to Reset Controller)
set
WDUNF wdt_int
set reset
WDERR
read WDT_SR reset WDFIEN
or
reset WDT_MR
115
The Watchdog Timer can be used to prevent system lock-up if the software becomes trapped in
a deadlock. It is supplied with VDDCORE. It restarts with initial values on processor reset.
The Watchdog is built around a 12-bit down counter, which is loaded with the value defined in
the field WDV of the Mode Register (WDT_MR). The Watchdog Timer uses the Slow Clock
divided by 128 to establish the maximum Watchdog period to be 16 seconds (with a typical Slow
Clock of 32.768 kHz).
After a Processor Reset, the value of WDV is 0xFFF, corresponding to the maximum value of
the counter with the external reset generation enabled (field WDRSTEN at 1 after a Backup
Reset). This means that a default Watchdog is running at reset, i.e., at power-up. The user must
either disable it (by setting the WDDIS bit in WDT_MR) if he does not expect to use it or must
reprogram it to meet the maximum Watchdog period the application requires.
The Watchdog Mode Register (WDT_MR) can be written only once. Only a processor reset
resets it. Writing the WDT_MR register reloads the timer with the newly programmed mode
parameters.
In normal operation, the user reloads the Watchdog at regular intervals before the timer under-
flow occurs, by writing the Control Register (WDT_CR) with the bit WDRSTT to 1. The
Watchdog counter is then immediately reloaded from WDT_MR and restarted, and the Slow
Clock 128 divider is reset and restarted. The WDT_CR register is write-protected. As a result,
writing WDT_CR without the correct hard-coded key has no effect. If an underflow does occur,
the “wdt_fault” signal to the Reset Controller is asserted if the bit WDRSTEN is set in the Mode
Register (WDT_MR). Moreover, the bit WDUNF is set in the Watchdog Status Register
(WDT_SR).
To prevent a software deadlock that continuously triggers the Watchdog, the reload of the
Watchdog must occur while the Watchdog counter is within a window between 0 and WDD,
WDD is defined in the WatchDog Mode Register WDT_MR.
Any attempt to restart the Watchdog while the Watchdog counter is between WDV and WDD
results in a Watchdog error, even if the Watchdog is disabled. The bit WDERR is updated in the
WDT_SR and the “wdt_fault” signal to the Reset Controller is asserted.
Note that this feature can be disabled by programming a WDD value greater than or equal to the
WDV value. In such a configuration, restarting the Watchdog Timer is permitted in the whole
range [0; WDV] and does not generate an error. This is the default configuration on reset (the
WDD and WDV values are equal).
The status bits WDUNF (Watchdog Underflow) and WDERR (Watchdog Error) trigger an inter-
rupt, provided the bit WDFIEN is set in the mode register. The signal “wdt_fault” to the reset
controller causes a Watchdog reset if the WDRSTEN bit is set as already explained in the reset
controller programmer Datasheet. In that case, the processor and the Watchdog Timer are
reset, and the WDERR and WDUNF flags are reset.
If a reset is generated or if WDT_SR is read, the status bits are reset, the interrupt is cleared,
and the “wdt_fault” signal to the reset controller is deasserted.
Writing the WDT_MR reloads and restarts the down counter.
While the processor is in debug state or in idle mode, the counter may be stopped depending on
the value programmed for the bits WDIDLEHLT and WDDBGHLT in the WDT_MR.
116 AT91SAM9G45
Figure 15-2. Watchdog Behavior
Watchdog Error Watchdog Underflow
if WDRSTEN is 1
FFF
Normal behavior if WDRSTEN is 0

WDV
Forbidden
Window
WDD
Permitted
Window
WDT_CR = WDRSTT
Watchdog
Fault
117 AT91SAM9G45
15.5 Watchdog Timer (WDT) User Interface

0x00 Control Register WDT_CR Write-only -
0x04 Mode Register WDT_MR Read-write Once 0x3FFF_2FFF
0x08 Status Register WDT_SR Read-only 0x0000_0000
15.5.1 Watchdog Timer Control Register

Register Name: WDT_CR
Address: 0xFFFFFD40
31 30 29 28 27 26 25 24
KEY
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – WDRSTT
• WDRSTT: Watchdog Restart

0: No effect.
1: Restarts the Watchdog.
• KEY: Password
118 AT91SAM9G45
15.5.2 Watchdog Timer Mode Register
Register Name: WDT_MR
Address: 0xFFFFFD44
Access Type: Read-write Once
31 30 29 28 27 26 25 24
WDIDLEHLT WDDBGHLT WDD
23 22 21 20 19 18 17 16
WDD
15 14 13 12 11 10 9 8
WDDIS WDRPROC WDRSTEN WDFIEN WDV
7 6 5 4 3 2 1 0
WDV
• WDV: Watchdog Counter Value

Defines the value loaded in the 12-bit Watchdog Counter.
• WDFIEN: Watchdog Fault Interrupt Enable

0: A Watchdog fault (underflow or error) has no effect on interrupt.
1: A Watchdog fault (underflow or error) asserts interrupt.
• WDRSTEN: Watchdog Reset Enable

0: A Watchdog fault (underflow or error) has no effect on the resets.
1: A Watchdog fault (underflow or error) triggers a Watchdog reset.
• WDRPROC: Watchdog Reset Processor

0: If WDRSTEN is 1, a Watchdog fault (underflow or error) activates all resets.
1: If WDRSTEN is 1, a Watchdog fault (underflow or error) activates the processor reset.
• WDD: Watchdog Delta Value

Defines the permitted range for reloading the Watchdog Timer.
If the Watchdog Timer value is less than or equal to WDD, writing WDT_CR with WDRSTT = 1 restarts the timer.
If the Watchdog Timer value is greater than WDD, writing WDT_CR with WDRSTT = 1 causes a Watchdog error.
• WDDBGHLT: Watchdog Debug Halt

0: The Watchdog runs when the processor is in debug state.
1: The Watchdog stops when the processor is in debug state.
• WDIDLEHLT: Watchdog Idle Halt

0: The Watchdog runs when the system is in idle mode.
1: The Watchdog stops when the system is in idle state.
• WDDIS: Watchdog Disable

0: Enables the Watchdog Timer.
1: Disables the Watchdog Timer.
119 AT91SAM9G45
15.5.3 Watchdog Timer Status Register
Register Name: WDT_SR
Address: 0xFFFFFD48
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – WDERR WDUNF
• WDUNF: Watchdog Underflow

0: No Watchdog underflow occurred since the last read of WDT_SR.
1: At least one Watchdog underflow occurred since the last read of WDT_SR.
• WDERR: Watchdog Error

0: No Watchdog error occurred since the last read of WDT_SR.
1: At least one Watchdog error occurred since the last read of WDT_SR.
120 AT91SAM9G45
AT91SAM9G45
16. Shutdown Controller (SHDWC)
16.1 Description
The Shutdown Controller controls the power supplies VDDIO and VDDCORE and the wake-up
detection on debounced input lines.

The Shut Down Controller is supplied on VDDBU and allows a software-controllable shut down
of the system through the pin SHDN. An input change of the WKUP pin or an alarm releases the
SHDN pin, and thus wakes up the system power supply.
16.3 Block Diagram
Figure 16-1. Shutdown Controller Block Diagram

SLCK
Shutdown Controller
SHDW_MR read SHDW_SR
CPTWK0 reset
WKMODE0 WAKEUP0 SHDW_SR

set
WKUP0
read SHDW_SR
Wake-up
reset
RTTWKEN SHDW_MR RTTWK SHDW_SR Shutdown SHDN

Output
RTT Alarm set
Controller
SHDW_CR
SHDW Shutdown
121
Figure 16-2. Shutdown Controller Block Diagram
SLCK
Shutdown Controller
SHDW_MR read SHDW_SR
CPTWK0 reset
WKMODE0 WAKEUP0 SHDW_SR

set
WKUP0
read SHDW_SR
Wake-up
reset
RTTWKEN SHDW_MR RTTWK SHDW_SR Shutdown SHDN

Output
RTT Alarm set Controller
SHDW_CR
read SHDW_SR SHDW Shutdown
reset
RTCWKEN SHDW_MR RTCWK SHDW_SR

RTC Alarm set
16.4 I/O Lines Description
Table 16-1. I/O Lines Description

Name Description Type
WKUP0 Wake-up 0 input Input
SHDN Shutdown output Output

The Shutdown Controller is continuously clocked by Slow Clock. The Power Management Con-
troller has no effect on the behavior of the Shutdown Controller.
122 AT91SAM9G45
AT91SAM9G45

The Shutdown Controller manages the main power supply. To do so, it is supplied with VDDBU
and manages wake-up input pins and one output pin, SHDN.
A typical application connects the pin SHDN to the shutdown input of the DC/DC Converter pro-
viding the main power supplies of the system, and especially VDDCORE and/or VDDIO. The
wake-up inputs (WKUP0) connect to any push-buttons or signal that wake up the system.
The software is able to control the pin SHDN by writing the Shutdown Control Register
(SHDW_CR) with the bit SHDW at 1. The shutdown is taken into account only 2 slow clock
cycles after the write of SHDW_CR. This register is password-protected and so the value written
should contain the correct key for the command to be taken into account. As a result, the system
should be powered down.
A level change on WKUP0 is used as wake-up. Wake-up is configured in the Shutdown Mode
Register (SHDW_MR). The transition detector can be programmed to detect either a positive or
negative transition or any level change on WKUP0. The detection can also be disabled. Pro-
gramming is performed by defining WKMODE0.
Moreover, a debouncing circuit can be programmed for WKUP0. The debouncing circuit filters
pulses on WKUP0 shorter than the programmed number of 16 SLCK cycles in CPTWK0 of the
SHDW_MR register. If the programmed level change is detected on a pin, a counter starts.
When the counter reaches the value programmed in the corresponding field, CPTWK0, the
SHDN pin is released. If a new input change is detected before the counter reaches the corre-
sponding value, the counter is stopped and cleared. WAKEUP0 of the Status Register
(SHDW_SR) reports the detection of the programmed events on WKUP0 with a reset after the
read of SHDW_SR.
The Shutdown Controller can be programmed so as to activate the wake-up using the RTT
alarm (the detection of the rising edge of the RTT alarm is synchronized with SLCK). This is
done by writing the SHDW_MR register using the RTTWKEN fields. When enabled, the detec-
tion of the RTT alarm is reported in the RTTWK bit of the SHDW_SR Status register. It is reset
after the read of SHDW_SR. When using the RTT alarm to wake up the system, the user must
ensure that the RTT alarm status flag is cleared before shutting down the system. Otherwise, no
rising edge of the status flag may be detected and the wake-up fails.
The Shutdown Controller can be programmed so as to activate the wake-up using the RTC
alarm (the detection of the rising edge of the RTC alarm is synchronized with SLCK). This is
done by writing the SHDW_MR register using the RTCWKEN field. When enabled, the detection
of the RTC alarm is reported in the RTCWK bit of the SHDW_SR Status register. It is reset after
the read of SHDW_SR. When using the RTC alarm to wake up the system, the user must
ensure that the RTC alarm status flag is cleared before shutting down the system. Otherwise, no
rising edge of the status flag may be detected and the wake-up fails fail.
123
16.7 Shutdown Controller (SHDWC) User Interface

0x00 Shutdown Control Register SHDW_CR Write-only -
0x04 Shutdown Mode Register SHDW_MR Read-write 0x0000_0003
0x08 Shutdown Status Register SHDW_SR Read-only 0x0000_0000
16.7.1 Shutdown Control Register

Register Name: SHDW_CR
Address: 0xFFFFFD10
31 30 29 28 27 26 25 24
KEY
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – SHDW
• SHDW: Shutdown Command

0 = No effect.
1 = If KEY is correct, asserts the SHDN pin.
• KEY: Password
124 AT91SAM9G45
AT91SAM9G45
16.7.2 Shutdown Mode Register

Register Name: SHDW_MR
Address: 0xFFFFFD14
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – RTCWKEN RTTWKEN
15 14 13 12 11 10 9 8
– – – –
7 6 5 4 3 2 1 0
CPTWK0 – – WKMODE0
• WKMODE0: Wake-up Mode 0
WKMODE[1:0] Wake-up Input Transition Selection

0 0 None. No detection is performed on the wake-up input
0 1 Low to high level
1 0 High to low level
1 1 Both levels change
• CPTWK0: Counter on Wake-up 0

Defines the number of 16 Slow Clock cycles, the level detection on the corresponding input pin shall last before the wake-
up event occurs. Because of the internal synchronization of WKUP0, the SHDN pin is released
(CPTWK x 16 + 1) Slow Clock cycles after the event on WKUP.
• RTTWKEN: Real-time Timer Wake-up Enable

0 = The RTT Alarm signal has no effect on the Shutdown Controller.
1 = The RTT Alarm signal forces the de-assertion of the SHDN pin.
• RTCWKEN: Real-time Clock Wake-up Enable

0 = The RTC Alarm signal has no effect on the Shutdown Controller.
1 = The RTC Alarm signal forces the de-assertion of the SHDN pin.
125
16.7.3 Shutdown Status Register
Register Name: SHDW_SR
Address: 0xFFFFFD18
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – RTCWK RTTWK
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – WAKEUP0
• WAKEUP0: Wake-up 0 Status

0 = No wake-up event occurred on the corresponding wake-up input since the last read of SHDW_SR.
1 = At least one wake-up event occurred on the corresponding wake-up input since the last read of SHDW_SR.
• RTTWK: Real-time Timer Wake-up

0 = No wake-up alarm from the RTT occurred since the last read of SHDW_SR.
1 = At least one wake-up alarm from the RTT occurred since the last read of SHDW_SR.
• RTCWK: Real-time Clock Wake-up

0 = No wake-up alarm from the RTC occurred since the last read of SHDW_SR.
1 = At least one wake-up alarm from the RTC occurred since the last read of SHDW_SR.
126 AT91SAM9G45
AT91SAM9G45
17. General Purpose Backup Registers (GPBR)
17.1 Description
The System Controller embeds Four general-purpose backup registers.

• Four 32-bit general-purpose backup registers
17.3 General Purpose Backup Registers (GPBR) User Interface

0x0 General Purpose Backup Register 0 SYS_GPBR0 Read-write –
... ... ... ... ...
0xc General Purpose Backup Register 3 SYS_GPBR3 Read-write –
127
17.3.0.1 General Purpose Backup Register x
Name: SYS_GPBRx
Addresses: 0xFFFFFD60 [0], 0xFFFFFD64 [1], 0xFFFFFD68 [2], 0xFFFFFD6C [3]
Type: Read-write
31 30 29 28 27 26 25 24
GPBR_VALUEx
23 22 21 20 19 18 17 16
GPBR_VALUEx
15 14 13 12 11 10 9 8
GPBR_VALUEx
7 6 5 4 3 2 1 0
GPBR_VALUEx
• GPBR_VALUEx: Value of GPBR x
128 AT91SAM9G45
AT91SAM9G45
18. Bus Matrix (MATRIX)
18.1 Description
The Bus Matrix implements a multi-layer AHB, based on the AHB-Lite protocol, that enables par-
allel access paths between multiple AHB masters and slaves in a system, thus increasing the
overall bandwidth. The Bus Matrix interconnects up to 16 AHB masters to up to 16 AHB slaves.
The normal latency to connect a master to a slave is one cycle except for the default master of
the accessed slave which is connected directly (zero cycle latency).
The Bus Matrix user interface is compliant with ARM Advanced Peripheral Bus and provides a
Chip Configuration User Interface with Registers that allow the Bus Matrix to support application
specific features.

• 11-layer Matrix, handling requests from 11 masters
• Programmable Arbitration strategy
– Fixed-priority Arbitration
– Round-Robin Arbitration, either with no default master, last accessed default master
or fixed default master
• Burst Management
– Breaking with Slot Cycle Limit Support
– Undefined Burst Length Support
• One Address Decoder provided per Master
– Three different slaves may be assigned to each decoded memory area: one for
internal ROM boot, one for internal flash boot, one after remap
• Boot Mode Select
– Non-volatile Boot Memory can be internal ROM or external memory on EBI_NCS0
– Selection is made by General purpose NVM bit sampled at reset
• Remap Command
– Allows Remapping of an Internal SRAM in Place of the Boot Non-Volatile Memory
(ROM or External Flash)
– Allows Handling of Dynamic Exception Vectors
18.2.1 Matrix Masters

The Bus Matrix of the AT91SAM9G45 manages Masters, thus each master can perform an
access concurrently with others, depending on whether the slave it accesses is available.
Each Master has its own decoder, which can be defined specifically for each master. In order to
simplify the addressing, all the masters have the same decodings.
129
Table 18-1. List of Bus Matrix Masters
Master 0 ARM926™ Instruction
Master 1 ARM926 Data
Master 2 Peripheral DMA Controller (PDC)
Master 3 USB HOST OHCI
Master 4 DMA
Master 5 DMA
Master 6 ISI Controller DMA
Master 7 LCD DMA
Master 8 Ethernet MAC DMA
Master 9 USB Device High Speed DMA
Master 10 USB Host High Speed EHCI DMA
18.2.2 Matrix Slaves

Each Slave has its own arbiter, thus allowing a different arbitration per Slave to be programmed.
Table 18-2. List of Bus Matrix Slaves

Slave 0 Internal SRAM
Internal ROM
USB OHCI
Slave 1 USB EHCI
UDP High Speed RAM
LCD User Interface
Slave 2 DDR Port 0
Slave 3 DDR Port 1
Slave 4 DDR Port 2
Slave 5 DDR Port 3
Slave 6 External Bus Interface
Slave 7 Internal Peripherals
18.2.3 Masters to Slaves Access

All the Masters can normally access all the Slaves. However, some paths do not make sense,
such as allowing access from the Ethernet MAC to the internal peripherals. Thus, these paths
are forbidden or simply not wired, and shown as “-” in the following tables.
130 AT91SAM9G45
AT91SAM9G45
Table 18-3. AT91SAM9G45 Masters to Slaves Access

Master 0 1 2 3 4&5 6 7 8 9 10
ARM ARM USB USB USB

Slave 926 926 Host ISI LCD Ethernet Device Host
Instr. Data PDC OHCI DMA DMA DMA MAC HS EHCI
0 Internal SRAM 0 X X X X X X - X X X
Internal ROM X X X - - - - - X -
UHP OHCI X X - - - - - - - -
UHP EHCI X X - - - - - - - -
LCD User Int. X X - - - - - - - -
1 UDPHS RAM X X - - - - - - - -
2 DDR Port 0 X - - - - - - - - -
3 DDR Port 1 - X - - - - - - - -
4 DDR Port 2 - - X X X X - X X X
5 DDR Port 3 - - X X X X X X X X
6 EBI X X X X X X X X X X
7 Internal Periph. X X X - X - - - - -
Table 18-4 summarizes the Slave Memory Mapping for each connected Master, depending on
the Remap status (RCBx bit in Bus Matrix Master Remap Control Register MATRIX_MRCR) and
the BMS state at reset.
Table 18-4. Internal Memory Mapping

Master
Slave RCBx = 0
RCBx = 1
Base Address BMS = 1 BMS = 0
0x0000 0000 Internal ROM EBI NCS0 Internal SRAM
18.3 Memory Mapping

The Bus Matrix provides one decoder for every AHB master interface. The decoder offers each
AHB master several memory mappings. In fact, depending on the product, each memory area
may be assigned to several slaves. Booting at the same address while using different AHB
slaves (i.e. external RAM, internal ROM or internal Flash, etc.) becomes possible.
The Bus Matrix user interface provides Master Remap Control Register (MATRIX_MRCR) that
performs remap action for every master independently.
18.4 Special Bus Granting Mechanism

The Bus Matrix provides some speculative bus granting techniques in order to anticipate access
requests from some masters. This mechanism reduces latency at first access of a burst or single
transfer as long as the slave is free from any other master access, but does not provide any ben-
efit as soon as the slave is continuously accessed by more than one master, since arbitration is
pipelined and then has no negative effect on the slave bandwidth or access latency.
131
This bus granting mechanism sets a different default master for every slave.
At the end of the current access, if no other request is pending, the slave remains connected to
its associated default master. A slave can be associated with three kinds of default masters: no
default master, last access master and fixed default master.
To change from one kind of default master to another, the Bus Matrix user interface provides the
Slave Configuration Registers, one for each slave, that set a default master for each slave. The
Slave Configuration Register contains two fields: DEFMSTR_TYPE and FIXED_DEFMSTR. The
2-bit DEFMSTR_TYPE field selects the default master type (no default, last access master, fixed
default master), whereas the 4-bit FIXED_DEFMSTR field selects a fixed default master pro-
vided that DEFMSTR_TYPE is set to fixed default master. Please refer to Section 18.7.2 “Bus
Matrix Slave Configuration Registers” on page 140.
18.4.1 No Default Master

After the end of the current access, if no other request is pending, the slave is disconnected from
all masters. No Default Master suits low-power mode.
This configuration incurs one latency clock cycle for the first access of a burst after bus Idle.
Arbitration without default master may be used for masters that perform significant bursts or sev-
eral transfers with no Idle in between, or if the slave bus bandwidth is widely used by one or
more masters.
This configuration provides no benefit on access latency or bandwidth when reaching maximum
slave bus throughput whatever is the number of requesting masters.
18.4.2 Last Access Master

After the end of the current access, if no other request is pending, the slave remains connected
to the last master that performed an access request.
This allows the Bus Matrix to remove the one latency cycle for the last master that accessed the
slave. Other non privileged masters still get one latency clock cycle if they want to access the
same slave. This technique is useful for masters that mainly perform single accesses or short
bursts with some Idle cycles in between.
18.4.3 Fixed Default Master

After the end of the current access, if no other request is pending, the slave connects to its fixed
default master. Unlike last access master, the fixed master does not change unless the user
modifies it by a software action (field FIXED_DEFMSTR of the related MATRIX_SCFG).
This allows the Bus Matrix arbiters to remove the one latency clock cycle for the fixed default
master of the slave. Every request attempted by this fixed default master will not cause any arbi-
tration latency whereas other non privileged masters will still get one latency cycle. This
technique is useful for a master that mainly perform single accesses or short bursts with some
Idle cycles in between.
132 AT91SAM9G45
AT91SAM9G45
18.5 Arbitration
The Bus Matrix provides an arbitration mechanism that reduces latency when conflict cases
occur, i.e. when two or more masters try to access the same slave at the same time. One arbiter
per AHB slave is provided, thus arbitrating each slave differently.
The Bus Matrix provides the user with the possibility of choosing between 2 arbitration types or
mixing them for each slave:
1. Round-Robin Arbitration (default)
2. Fixed Priority Arbitration
The resulting algorithm may be complemented by selecting a default master configuration for
each slave.
When a re-arbitration must be done, specific conditions apply. See Section 18.5.1 “Arbitration
Scheduling” on page 133.
18.5.1 Arbitration Scheduling

Each arbiter has the ability to arbitrate between two or more different master requests. In order
to avoid burst breaking and also to provide the maximum throughput for slave interfaces, arbitra-
tion may only take place during the following cycles:
1. Idle Cycles: When a slave is not connected to any master or is connected to a master
which is not currently accessing it.
2. Single Cycles: When a slave is currently doing a single access.
3. End of Burst Cycles: When the current cycle is the last cycle of a burst transfer. For
defined length burst, predicted end of burst matches the size of the transfer but is man-
aged differently for undefined length burst. See “Undefined Length Burst Arbitration” on
page 133
4. Slot Cycle Limit: When the slot cycle counter has reached the limit value indicating that
the current master access is too long and must be broken. See “Slot Cycle Limit Arbi-
tration” on page 134
18.5.1.1 Undefined Length Burst Arbitration

In order to optimize AHB burst lengths and arbitration, it may be interesting to set a maximum for
undefined length bursts (INCR). The Bus Matrix provides specific logic in order to re-arbitrate
before the end of the INCR transfer. A predicted end of burst is used as a defined length burst
transfer and can be selected from among the following Undefined Length Burst Type (ULBT)
possibilities:
1. Unlimited: No predicted end of burst is generated and therefore INCR burst transfer will
not be broken by this way, but will be able to complete unless broken at the Slot Cycle
Limit. This is normally the default and should be let as is in order to be able to allow full
1 Kilobyte AHB intra-boundary 256-beat word bursts performed by some ATMEL AHB
masters.
2. 1-beat bursts: Predicted end of burst is generated at each single transfer inside the
INCR transfer.
3. 4-beat bursts: Predicted end of burst is generated at the end of each 4-beat boundary
inside INCR transfer.
4. 8-beat bursts: Predicted end of burst is generated at the end of each 8-beat boundary
inside INCR transfer.
5. 16-beat bursts: Predicted end of burst is generated at the end of each 16-beat bound-
ary inside INCR transfer.
133
AT91SAM9G45
8. 128-beat bursts: Predicted end of burst is generated at the end of each 128-beat
boundary inside INCR transfer.
Use of undefined length 16-beat bursts or less is discouraged since this generally decreases
significantly overall bus bandwidth due to arbitration and slave latencies at each first access of a
burst.
If the master does not permanently and continuously request the same slave or has an intrinsi-
cally limited average throughput, the ULBT should be let at its default unlimited value, knowing
that the AHB specification natively limits all word bursts to 256 beats and double-word bursts to
128 beats because of its 1 Kilobyte address boundaries.
Unless duly needed the ULBT should be let to its default 0 value for power saving.
This selection can be done through the field ULBT of the Master Configuration Registers
(MATRIX_MCFG).
18.5.1.2 Slot Cycle Limit Arbitration

The Bus Matrix contains specific logic to break long accesses, such as back to back undefined
length bursts or very long bursts on a very slow slave (e.g., an external low speed memory). At
each arbitration time a counter is loaded with the value previously written in the SLOT_CYCLE
field of the related Slave Configuration Register (MATRIX_SCFG) and decreased at each clock
cycle. When the counter elapses, the arbiter has the ability to re-arbitrate at the end of the cur-
rent AHB bus access cycle.
Unless some master has a very tight access latency constraint which could lead to data overflow
or underflow due to a badly undersized internal fifo with respect to its throughput, the Slot Cycle
Limit should be disabled (SLOT_CYCLE = 0) or let to its default maximum value in order not to
inefficiently break long bursts performed by some ATMEL masters.
However, the Slot Cycle Limit should not be disabled in the very particular case of a master
capable of accessing the slave by performing back to back undefined length bursts shorter than
the number of ULBT beats with no Idle cycle in between, since in this case the arbitration could
be frozen all along the bursts sequence.
In most cases this feature is not needed and should be disabled for power saving.
Warning: This feature cannot prevent any slave from locking its access indefinitely.
18.5.2 Arbitration Priority Scheme

The bus Matrix arbitration scheme is organized in priority pools.
Round-Robin priority is used inside the highest and lowest priority pools, whereas fix level prior-
ity is used between priority pools and inside the intermediate priority pools.
For each slave, each master x is assigned to one of the slave priority pools through the Priority
Registers for Slaves (MxPR fields of MATRIX_PRAS and MATRIX_PRBS). When evaluating
masters requests, this programmed priority level always takes precedence.
After reset, all the masters are belonging to the lowest priority pool (MxPR = 0) and so are
granted bus access in a true Round-Robin fashion.
134
AT91SAM9G45
The highest priority pool must be specifically reserved for masters requiring very low access
latency. If more than one master belong to this pool, these will be granted bus access in a
biased Round-Robin fashion which allow tight and deterministic maximum access latency from
AHB bus request. In fact, at worst, any currently high priority master request will be granted after
the current bus master access is ended and the other high priority pool masters, if any, have
been granted once each.
The lowest priority pool shares the remaining bus bandwidth between AHB Masters.
Intermediate priority pools allow fine priority tuning. Typically, a moderately latency critical mas-
ter or a bandwidth only critical master will use such a priority level. The higher the priority level
(MxPR value), the higher the master priority.
All combination of MxPR values are allowed for all masters and slaves. For example some mas-
ters might be assigned to the highest priority pool (round-robin) and the remaining masters to
the lowest priority pool (round-robin), with no master for intermediate fix priority levels.
If more than one master is requesting the slave bus, whatever are the respective masters priori-
ties, no master will be granted the slave bus for two consecutive runs. A master can only get
back to back grants as long as it is the only requesting master.
18.5.2.1 Fixed Priority Arbitration

This arbitration algorithm is the first and only applied between masters from distinct priority
pools. It is also used inside priority pools other than the highest and lowest ones (intermediate
priority pools).
It allows the Bus Matrix arbiters to dispatch the requests from different masters to the same
slave by using the fixed priority defined by the user in the MxPR field for each master inside the
MATRIX_PRAS and MATRIX_PRBS Priority Registers. If two or more master requests are
active at the same time, the master with the highest priority number MxPR is serviced first.
Inside intermediate priority pools, if two or more master requests with the same priority are
active at the same time, the master with the highest number is serviced first.
18.5.2.2 Round-Robin Arbitration

This algorithm is only used inside the highest and lowest priority pools. It allows the Bus Matrix
arbiters to dispatch the requests from different masters to the same slave in a fair way. If two or
more master requests are active at the same time inside the priority pool, they are serviced in a
round-robin increasing master number order.
135
AT91SAM9G45
18.6 Write Protect Registers

To prevent any single software error that may corrupt MATRIX behavior, the entire MATRIX
address space from address offset 0x000 to 0x1FC can be write-protected by setting the WPEN
bit in the MATRIX Write Protect Mode Register (MATRIX_WPMR).
If a write access to anywhere in the MATRIX address space from address offset 0x000 to 0x1FC
is detected, then the WPVS flag in the MATRIX Write Protect Status Register (MATRIX_WPSR)
is set and the field WPVSRC indicates in which register the write access has been attempted.
The WPVS flag is reset by writing the MATRIX Write Protect Mode Register (MATRIX_WPMR)
with the appropriate access key WPKEY.
The protected registers are:
“Bus Matrix Master Configuration Registers”
“Bus Matrix Slave Configuration Registers”
“Bus Matrix Priority Registers A For Slaves”
“Bus Matrix Priority Registers B For Slaves”
“Bus Matrix Master Remap Control Register”
136
AT91SAM9G45
18.7 Bus Matrix (MATRIX) User Interface

0x0000 Master Configuration Register 0 MATRIX_MCFG0 Read-write 0x00000001
0x000C Master Configuration Register 3 MATRIX_MCFG3 Read-write 0x00000000
0x001C Master Configuration Register 7 MATRIX_MCFG7 Read-write 0x00000000
0x002C - 0x003C Reserved – – –
0x0040 Slave Configuration Register 0 MATRIX_SCFG0 Read-write 0x000001FF
0x004C Slave Configuration Register 3 MATRIX_SCFG3 Read-write 0x000001FF
0x005C Slave Configuration Register 7 MATRIX_SCFG7 Read-write 0x000001FF
0x0060 - 0x007C Reserved – – –
0x0080 Priority Register A for Slave 0 MATRIX_PRAS0 Read-write 0x00000000
0x0084 Priority Register B for Slave 0 MATRIX_PRBS0 Read-write 0x00000000
0x008C Priority Register B for Slave 1 MATRIX_PRBS1 Read-write 0x00000000
0x0094 Priority Register B for Slave 2 MATRIX_PRBS2 Read-write 0x00000000
0x009C Priority Register B for Slave 3 MATRIX_PRBS3 Read-write 0x00000000
0x00A0 Priority Register A for Slave 4 MATRIX_PRAS4 Read-write 0x00000000
0x00A4 Priority Register B for Slave 4 MATRIX_PRBS4 Read-write 0x00000000
0x00A8 Priority Register A for Slave 5 MATRIX_PRAS5 Read-write 0x00000000
0x00AC Priority Register B for Slave 5 MATRIX_PRBS5 Read-write 0x00000000
0x00B0 Priority Register A for Slave 6 MATRIX_PRAS6 Read-write 0x00000000
0x00B4 Priority Register B for Slave 6 MATRIX_PRBS6 Read-write 0x00000000
137
AT91SAM9G45
Table 18-5. Register Mapping (Continued)

0x00B8 Priority Register A for Slave 7 MATRIX_PRAS7 Read-write 0x00000000
0x00BC Priority Register B for Slave 7 MATRIX_PRBS7 Read-write 0x00000000
0x00C0 - 0x00FC Reserved – – –
0x0100 Master Remap Control Register MATRIX_MRCR Read-write 0x00000000
0x0104 - 0x010C Reserved – – –
0x0110 - 0x01E0 Chip Configuration Registers – – –
0x01E4 Write Protect Mode Register MATRIX_WPMR Read-write 0x00000000
0x01E8 Write Protect Status Register MATRIX_WPSR Read-only 0x00000000
138
AT91SAM9G45
18.7.1 Bus Matrix Master Configuration Registers

Name: MATRIX_MCFG0...MATRIX_MCFG10
Address: 0xFFFFEA00
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – ULBT
This register can only be written if the WPEN bit is cleared in the “Write Protect Mode Register”.
• ULBT: Undefined Length Burst Type

0: Unlimited Length Burst
No predicted end of burst is generated and therefore INCR bursts coming from this master can only be broken if the Slave
Slot Cycle Limit is reached. If the Slot Cycle Limit is not reached, the burst is normally completed by the master, at the lat-
est, on the next AHB 1 KByte address boundary, allowing up to 256-beat word bursts or 128-beat double-word bursts.
1: Single Access
The undefined length burst is treated as a succession of single accesses, allowing re-arbitration at each beat of the INCR
burst.
2: 4-beat Burst
The undefined length burst is split into 4-beat bursts, allowing re-arbitration at each 4-beat burst end.
3: 8-beat Burst
4: 16-beat Burst
5: 32-beat Burst
6: 64-beat Burst
7: 128-beat Burst
Unless duly needed the ULBT should be let to its default 0 value for power saving.
139
AT91SAM9G45
18.7.2 Bus Matrix Slave Configuration Registers

Name: MATRIX_SCFG0...MATRIX_SCFG7
Address: 0xFFFFEA40
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – FIXED_DEFMSTR DEFMSTR_TYPE
15 14 13 12 11 10 9 8
– – – – – – – SLOT_CYCLE
7 6 5 4 3 2 1 0
SLOT_CYCLE
• SLOT_CYCLE: Maximum Bus Grant Duration for Masters

When SLOT_CYCLE AHB clock cycles have elapsed since the last arbitration, a new arbitration takes place so as to let an
other master access this slave. If an other master is requesting the slave bus, then the current master burst is broken.
If SLOT_CYCLE = 0, the Slot Cycle Limit feature is disabled and bursts always complete unless broken according to the
ULBT.
This limit has been placed in order to enforce arbitration so as to meet potential latency constraints of masters waiting for
slave access or in the particular case of a master performing back to back undefined length bursts indefinitely freezing the
arbitration.
This limit must not be small. Unreasonably small values break every burst and the Bus Matrix arbitrates without performing
any data transfer. The default maximum value is usually an optimal conservative choice.
In most cases this feature is not needed and should be disabled for power saving.
See “Slot Cycle Limit Arbitration” on page 134 for details.
• DEFMSTR_TYPE: Default Master Type

0: No Default Master
At the end of the current slave access, if no other master request is pending, the slave is disconnected from all masters.
This results in a one clock cycle latency for the first access of a burst transfer or for a single access.
1: Last Default Master
At the end of the current slave access, if no other master request is pending, the slave stays connected to the last master
having accessed it.
This results in not having one clock cycle latency when the last master tries to access the slave again.
2: Fixed Default Master
At the end of the current slave access, if no other master request is pending, the slave connects to the fixed master the
number that has been written in the FIXED_DEFMSTR field.
This results in not having one clock cycle latency when the fixed master tries to access the slave again.
140
AT91SAM9G45
• FIXED_DEFMSTR: Fixed Default Master

This is the number of the Default Master for this slave. Only used if DEFMSTR_TYPE is 2. Specifying the number of a mas-
ter which is not connected to the selected slave is equivalent to setting DEFMSTR_TYPE to 0.
141
AT91SAM9G45
18.7.3 Bus Matrix Priority Registers A For Slaves

Name: MATRIX_PRAS0...MATRIX_PRAS7
Addresses: 0xFFFFEA80 [0], 0xFFFFEA88 [1], 0xFFFFEA90 [2], 0xFFFFEA98 [3], 0xFFFFEAA0 [4], 0xFFFFEAA8 [5],
0xFFFFEAB0 [6], 0xFFFFEAB8 [7]
Access: Read-write
31 30 29 28 27 26 25 24
– – M7PR – – M6PR
23 22 21 20 19 18 17 16
– – M5PR – – M4PR
15 14 13 12 11 10 9 8
– – M3PR – – M2PR
7 6 5 4 3 2 1 0
– – M1PR – – M0PR
• MxPR: Master x Priority

Fixed priority of Master x for accessing the selected slave. The higher the number, the higher the priority.
All the masters programmed with the same MxPR value for the slave make up a priority pool.
Round-Robin arbitration is used inside the lowest (MxPR = 0) and highest (MxPR = 3) priority pools.
Fixed priority is used inside intermediate priority pools (MxPR = 1) and (MxPR = 2).
See Section 18.5.2 “Arbitration Priority Scheme” for details.
142
AT91SAM9G45
18.7.4 Bus Matrix Priority Registers B For Slaves

Name: MATRIX_PRBS0...MATRIX_PRBS7
Addresses: 0xFFFFEA84 [0], 0xFFFFEA8C [1], 0xFFFFEA94 [2], 0xFFFFEA9C [3], 0xFFFFEAA4 [4], 0xFFFFEAAC [5],
0xFFFFEAB4 [6], 0xFFFFEABC [7]
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – M10PR
7 6 5 4 3 2 1 0
– – M9PR – – M8PR
• MxPR: Master x Priority

Fixed priority of Master x for accessing the selected slave. The higher the number, the higher the priority.
All the masters programmed with the same MxPR value for the slave make up a priority pool.
Round-Robin arbitration is used inside the lowest (MxPR = 0) and highest (MxPR = 3) priority pools.
Fixed priority is used inside intermediate priority pools (MxPR = 1) and (MxPR = 2).
See Section 18.5.2 “Arbitration Priority Scheme” for details.
143
AT91SAM9G45
18.7.5 Bus Matrix Master Remap Control Register

Name: MATRIX_MRCR
Address: 0xFFFFEB00
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – RCB10 RCB9 RCB8
7 6 5 4 3 2 1 0
RCB7 RCB6 RCB5 RCB4 RCB3 RCB2 RCB1 RCB0
• RCB: Remap Command Bit for Master x

0: Disable remapped address decoding for the selected Master
1: Enable remapped address decoding for the selected Master
144
AT91SAM9G45
18.7.6 Chip Configuration User Interface
Table 18-6. Chip Configuration User Interface

0x0110 Bus Matrix TCM Configuration Register CCFG_TCMR Read-write 0x00000000
0x0114 Reserved – – –
0x0118 DDR Multi-Port Register CCFG_DDRMPR Read-write 0x00000001
0x011C - 0x0124 Reserved – – –
0x0128 EBI Chip Select Assignment Register CCFG_EBICSA Read-write 0x00010000
0x012C - 0x01FC Reserved – – –
145
AT91SAM9G45
18.7.6.1 Bus Matrix TCM Configuration Register

Name: CCFG_TCMR
Access: Read-write
Reset: 0x0000_0000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – TCM_NWS – – –
7 6 5 4 3 2 1 0
DTCM_SIZE ITCM_SIZE
• ITCM_SIZE: Size of ITCM enabled memory block

000: 0 KB (No ITCM Memory)
110: 32 KB
Others: Reserved
• DTCM_SIZE: Size of DTCM enabled memory block

000: 0 KB (No DTCM Memory)
110: 32 KB
111: 64 KB
Others: Reserved
• TCM_NWS: TCM Wait State

0: no TCM Wait State
1: 1 TCM Wait State (only for ration 3:1 or 4:1)
146
AT91SAM9G45
18.7.6.2 Bus Matrix DDR Multi-Port Register

Register Name: CCFG_DDRMPR
Reset: 0x0000_0001
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – DDRMP_DIS
• DDRMP_DIS: DDR Multi-Port Disable

0: Multi-Port is enabled
1: Multi-Port is disabled
147
18.7.6.3 EBI Chip Select Assignment Register
Name: CCFG_EBICSA
Access: Read-write
Reset: 0x0001_0000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – DDR_DRIVE EBI_DRIVE
15 14 13 12 11 10 9 8
– – – – – – – EBI_DBPUC
7 6 5 4 3 2 1 0
– – EBI_CS5A EBI_CS4A EBI_CS3A – EBI_CS1A –
• EBI_CS1A: EBI Chip Select 1 Assignment

0 = EBI Chip Select 1 is assigned to the Static Memory Controller.
1 = EBI Chip Select 1 is assigned to the SDRAM Controller.

0 = EBI Chip Select 3 is only assigned to the Static Memory Controller and EBI_NCS3 behaves as defined by the SMC.
1 = EBI Chip Select 3 is assigned to the Static Memory Controller and the SmartMedia Logic is activated.

1 = EBI Chip Select 4 is assigned to the Static Memory Controller and the Compact Flash Logic Slot 0 is activated.

1 = EBI Chip Select 5 is assigned to the Static Memory Controller and the Compact Flash Logic Slot 1 is activated.
• EBI_DBPUC: EBI Data Bus Pull-Up Configuration

0 = EBI D0 - D15 Data Bus bits are internally pulled-up to the VDDIOM1 power supply.
1 = EBI D0 - D15 Data Bus bits are not internally pulled-up.
• EBI_DRIVE: EBI I/O Drive Configuration

This allows to avoid overshoots and give the best performances according to the bus load and external memories.
Value Drive configuration Conditions

Operating frequency = 133 MHz
00 optimized for 1.8V powered memories with High Drive
data bus + memory load capacitance > TBD pF
01 optimized for 3.3V powered memories with High Drive
10 optimized for 1.8V powered memories with LowDrive
11 optimized for 3.3V powered memories with Low Drive
148 AT91SAM9G45
AT91SAM9G45
• DDR_DRIVE: DDR2 dedicated port I/O slew rate selection

This allows to avoid overshoots and give the best performances according to the bus load and external memories.
0 = High Drive, data bus + memory load capacitance > TBD pF.
1 = Low Drive, data bus + memory load capacitance < TBD pF,
Note: This concerns only stand-alone DDR controller.
149
18.7.7 Write Protect Mode Register
Name: MATRIX_WPMR
Address: 0xFFFFEBE4
Access: Read-write
31 30 29 28 27 26 25 24
WPKEY
23 22 21 20 19 18 17 16
WPKEY
15 14 13 12 11 10 9 8
WPKEY
7 6 5 4 3 2 1 0
– – – – – – – WPEN
For more details on MATRIX_WPMR, refer to Section 18.6 “Write Protect Registers” on page 136.
• WPEN: Write Protect ENable

0 = Disables the Write Protect if WPKEY corresponds to 0x4D4154 (“MAT” in ASCII).
1 = Enables the Write Protect if WPKEY corresponds to 0x4D4154 (“MAT” in ASCII).
Protects the entire MATRIX address space from address offset 0x000 to 0x1FC.
• WPKEY: Write Protect KEY (Write-only)

Should be written at value 0x4D4154 (“MAT”in ASCII). Writing any other value in this field aborts the write operation of the
WPEN bit. Always reads as 0.
150 AT91SAM9G45
AT91SAM9G45
18.7.8 Write Protect Status Register

Name: MATRIX_WPSR
Address: 0xFFFFEBE8
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
WPVSRC
15 14 13 12 11 10 9 8
WPVSRC
7 6 5 4 3 2 1 0
– – – – – – – WPVS
For more details on MATRIX_WPSR, refer to Section 18.6 “Write Protect Registers” on page 136.
• WPVS: Write Protect Violation Status

0: No Write Protect Violation has occurred since the last write of the MATRIX_WPMR.
1: At least one Write Protect Violation has occurred since the last write of the MATRIX_WPMR.
• WPVSRC: Write Protect Violation Source

When WPVS is active, this field indicates the register address offset in which a write access has been attempted.
Otherwise it reads as 0.
151
152 AT91SAM9G45
AT91SAM9G45
19. External Memories
19.1 Programmable I/O Lines Power Supplies and Drive Levels
19.1.1 External Bus interface

The power supply pin VDDIOM1 accepts two voltage ranges. This allows the device to reach its
maximum speed either out of 1.8V or 3.3V external memories.
The maximum speed is 133 MHz on the SDCK pin and #SDCK signals loaded with 10 pF. The
load on data/address and control signals are 30 pF for power supply at 1.8V and 50 pF for power
supply at 3.3V. The data lines frequency reaches 133 MHz in DDR2 mode. The other signals
(control and address) do not go over 66 MHz.
The EBI I/Os accept two drive levels , HIGH and LOW. This allows to avoid overshoots and give
the best performance according to the bus load and external memories. Refer to the EBI Chip
Select Assignment Register for more details.
The voltage ranges and the drive level are determined by programming EBI_DRIVE field in the
Chip Configuration registers located in the Matrix User Interface.
At reset the selected default drive level is High.
At reset, the selected voltage defaults to 3.3V nominal and power supply pins can accept either
1.8V or 3.3V. The user must make sure to program the EBI voltage range before getting the
device out of its Slow Clock Mode. The user must make sure to program the EBI voltage range
before getting the device out of its Slow Clock Mode.
19.2 DDR2 Controller
19.2.1 Description
The DDR2 Controller is dedicated to 4-port DDR2/LPDDR support. Data transfers are performed
through a 16-bit data bus on one chip select. The DDR2 Controller operates with 1.8V Power
Supply (VDDIOM0).
19.2.2 Embedded Characteristics
19.2.2.1 DDR2/LPDDR Controller

Four AHB Interfaces, Management of All Accesses Maximizes Memory Bandwidth and Mini-
mizes Transaction Latency.
• Supports AHB Transfers:
– Word, Half Word, Byte Access.
• Supports DDR-SDRAM 2, LPDDR
• Numerous Configurations Supported
– 2K, 4K, 8K, 16K Row Address Memory Parts
– DDR2 with Four Internal Banks
– DDR2/LPDDR with 16-bit Data Path
– One Chip Select for DDR2/LPDDR Device (256 Mbytes Address Space)
• Programming Facilities
153
– Multibank Ping-pong Access (Up to 4 Banks Opened at Same Time = Reduces
Average Latency of Transactions)
– Timing Parameters Specified by Software
– Automatic Refresh Operation, Refresh Rate is Programmable
– Automatic Update of DS, TCR and PASR Parameters
• Energy-saving Capabilities
– Self-refresh, Power-down and Deep Power Modes Supported
• Power-up Initialization by Software
• CAS Latency of 2, 3 Supported
• Reset function supported (DDR2)
• Auto Precharge Command Not Used
• On Die Termination not supported
• OCD mode not supported
154 AT91SAM9G45
AT91SAM9G45
19.2.3 DDR2 Controller Block Diagram
Figure 19-1. Organization of the DDR2
DDR2
DDR_A0-DDR_A13
DDR_D0-DDR_D15
Bus Matrix DDR_CS
DDR_CKE
DDR_RAS, DDR_CAS
DDR2
AHB LPDDR DDR_CLK,#DDR_CLK
Controller
DDR_DQS[0..1]
DDR_DQM[0..1]
DDR_WE
DDR_BA0, DDR_BA1
Address Decoders
DDR_VREF
User Interface
APB
155
19.2.4 I/O Lines Description
Table 19-1. DDR2 I/O Lines Description

Name Function Type Active Level
DDR2/LPDDR Controller
DDR_D0 - DDR_D15 Data Bus I/O
DDR_A0 - DDR_A13 Address Bus Output
DDR_DQM0 - DDR_DQM1 Data Mask Output
DDR_DQS0 - DDR_DQS1 Data Strobe Output
DDR_VREF Reference Voltage for DDR2 operations, typically 0.9V Input
DDR_CS Chip Select Output Low
DDR_CLK - DDR_CLK# DDR2 Differential Clock Output
DDR_CKE Clock enable Output High
DDR_RAS Row signal Output Low
DDR_CAS Column signal Output Low
DDR_WE Write enable Output Low
DDR_BA0 - DDR_BA1 Bank Select Output
19.2.5 Product Dependencies

The pins used for interfacing the DDR2 memory are not multiplexed with the PIO lines.
19.2.6 Implementation Example

The following hardware configuration is given for illustration only. The user should refer to the
memory manufacturer web site to check current device availability.
156 AT91SAM9G45
AT91SAM9G45
19.2.6.1 16-bit DDR2

Hardware Configuration
DDR_D[0..15]
DDR_A[0..13]
MN6 MN7
DDR_A0 H8 C8 DDR_D0 DDR_A0 H8 C8 DDR_D8
DDR_A1 A0 DDR2 SDRAM DQ0 DDR_D1 DDR_A1 A0 DDR2 SDRAM DQ0 DDR_D9
H3 A1 DQ1 C2 H3 A1 DQ1 C2
DDR_A2 H7 D7 DDR_D2 DDR_A2 H7 D7 DDR_D10
DDR_A3 A2 MT47H64M8CF-3 DQ2 DDR_D3 DDR_A3 A2 MT47H64M8CF-3 DQ2 DDR_D11
J2 A3 DQ3 D3 J2 A3 DQ3 D3
DDR_A4 J8 D1 DDR_D4 DDR_A4 J8 D1 DDR_D12
DDR_A5 A4 DQ4 DDR_D5 DDR_A5 A4 DQ4 DDR_D13
J3 A5 DQ5 D9 J3 A5 DQ5 D9
DDR_A6 J7 B1 DDR_D6 DDR_A6 J7 B1 DDR_D14
DDR_A7 A6 DQ6 DDR_D7 DDR_A7 A6 DQ6 DDR_D15
K2 A7 DQ7 B9 K2 A7 DQ7 B9
DDR_A8 K8 DDR_A8 K8
DDR_A9 A8 DDR_A9 A8
K3 A9 DQS B7 DDR_DQS0 K3 A9 DQS B7 DDR_DQS1
DDR_A10 H2 A8 DDR_A10 H2 A8
DDR_A11 A10 DQS DDR_A11 A10 DQS
K7 A11 K7 A11
DDR_A12 L2 B3 DDR_A12 L2 B3
A12 RDQS/DM DDR_DQM0 A12 RDQS/DM DDR_DQM1
DDR_A13 L8 A2 DDR_A13 L8 A2
A13 RDQS/NU 1V8 A13 RDQS/NU 1V8
BA0 G2 A1 C55 100nF BA0 G2 A1 C56 100nF
DDR_BA0 BA0 VDD BA0 VDD
BA1 G3 E9 C57 100nF BA1 G3 E9 C58 100nF
DDR_BA1 BA1 VDD BA1 VDD
VDD H9 C59 100nF VDD H9 C60 100nF
VDD L1 C61 100nF VDD L1 C62 100nF
F9 ODT F9 ODT
VDDL E1 C63 100nF VDDL E1 C64 100nF
CKE F2 A9 C65 100nF CKE F2 A9 C66 100nF
DDR_CKE CKE VDDQ CKE VDDQ
VDDQ C1 C67 100nF VDDQ C1 C68 100nF
CK E8 C3 C69 100nF CK E8 C3 C70 100nF
DDR_CLK CK VDDQ CK VDDQ
NCK F8 C7 C71 100nF NCK F8 C7 C72 100nF
DDR_NCLK CK VDDQ CK VDDQ
VDDQ C9 C73 100nF VDDQ C9 C74 100nF
CS G8 E2 DDR_VREF CS G8 E2 DDR_VREF
DDR_CS CS VREF CS VREF
CAS G7 A3 C75 CAS G7 A3 C76
DDR_CAS CAS VSS CAS VSS
RAS F7 E3 100nF RAS F7 E3 100nF
DDR_RAS RAS VSS RAS VSS
VSS J1 VSS J1
NWE F3 K9 NWE F3 K9
DDR_WE WE VSS WE VSS
VSSQ A7 VSSQ A7
VSSQ B2 VSSQ B2
G1 RFU1 VSSQ B8 G1 RFU1 VSSQ B8
L3 RFU2 VSSQ D2 L3 RFU2 VSSQ D2
L7 RFU3 VSSQ D8 L7 RFU3 VSSQ D8
VSSDL E7 VSSDL E7
Software Configuration
The following configuration has to be performed:
• Initialize the DDR2 Controller depending on the DDR2 device and system bus frequency.
The DDR2 initialization sequence is described in the sub-section “DDR2 Device Initialization” of
the DDRSDRC section.
157
19.3 External Bus Interface (EBI)
19.3.1 Description
The External Bus Interface (EBI) is designed to ensure the successful data transfer between
several external devices and the embedded Memory Controller of an ARM-based device.
The Static Memory, DDR, SDRAM and ECC Controllers are all featured external Memory Con-
trollers on the EBI. These external Memory Controllers are capable of handling several types of
external memory and peripheral devices, such as SRAM, PROM, EPROM, EEPROM, Flash,
DDR2 and SDRAM. The EBI operates with 1.8V or 3.3V Power Supply (VDDIOM1).
The EBI also supports the CompactFlash and the NAND Flash protocols via integrated circuitry
that greatly reduces the requirements for external components. Furthermore, the EBI handles
data transfers with up to six external devices, each assigned to six address spaces defined by
the embedded Memory Controller. Data transfers are performed through a 16-bit or 32-bit data
bus, an address bus of up to 26 bits, up to six chip select lines (NCS[5:0]) and several control
pins that are generally multiplexed between the different external Memory Controllers.
19.3.2 Embbeded Characteristics

The AT91SAM9G45 features an External Bus Interface to interface to a wide range of external
memories and to any parallel peripheral.
19.3.2.1 External Bus Interface

• Integrates Three External Memory Controllers:
– Static Memory Controller
– DDR2/SDRAM Controller
– SLC Nand Flash ECC Controller
• Additional logic for NAND Flash and CompactFlash
• Optional Full 32-bit External Data Bus
• Up to 26-bit Address Bus (up to 64 MBytes linear per chip select)
• Up to 6 chip selects, Configurable Assignment:
– Static Memory Controller on NCS0
– DDR2/SDRAM Controller (SDCS) or Static Memory Controller on NCS1
– Static Memory Controller on NCS2
– Static Memory Controller on NCS3, Optional NAND Flash support
– Static Memory Controller on NCS4 - NCS5, Optional CompactFlashM support
19.3.2.2 Static Memory Controller

• 8-, 16- or 32-bit Data Bus
• Multiple Access Modes supported
– Byte Write or Byte Select Lines
– Asynchronous read in Page Mode supported (4- up to 32-byte page size)
• Multiple device adaptability
– Control signals programmable setup, pulse and hold time for each Memory Bank
• Multiple Wait State Management
– Programmable Wait State Generation
– External Wait Request
158 AT91SAM9G45
AT91SAM9G45
– Programmable Data Float Time

• Slow Clock mode supported
19.3.2.3 DDR2/SDR Controller

• Supports DDR/LPDDR, SDR-SDRAM and LPSDR
• Numerous Configurations Supported
– 2K, 4K, 8K, 16K Row Address Memory Parts
– SDRAM with Four Internal Banks
– SDR-SDRAM with 16- or 32- bit Data Path
– DDR2/LPDDR with 16- bit Data Path
– One Chip Select for SDRAM Device (256 Mbyte Address Space)
• Programming Facilities
– Multibank Ping-pong Access (Up to 4 Banks Opened at Same Time = Reduces
Average Latency of Transactions)
– Timing Parameters Specified by Software
– Automatic Refresh Operation, Refresh Rate is Programmable
– Automatic Update of DS, TCR and PASR Parameters (LPSDR)
• Energy-saving Capabilities
– Self-refresh, Power-down and Deep Power Modes Supported
• SDRAM Power-up Initialization by Software
• CAS Latency of 2, 3 Supported
• Auto Precharge Command Not Used
• SDR-SDRAM with 16-bit Datapath and Eight Columns Not Supported
– Clock Frequency Change in Precharge Power-down Mode Not Supported
19.3.2.4 NAND Flash Error Corrected Code Controller

• Tracking the accesses to a NAND Flash device by triggering on the corresponding chip select
• Single bit error correction and 2-bit Random detection.
• Automatic Hamming Code Calculation while writing
– ECC value available in a register
• Automatic Hamming Code Calculation while reading
– Error Report, including error flag, correctable error flag and word address being
detected erroneous
– Support 8- or 16-bit NAND Flash devices with 512-, 1024-, 2048- or 4096-bytes
pages
159
19.3.3 EBI Block Diagram
Figure 19-2. Organization of the External Bus Interface
Bus Matrix External Bus Interface

D[15:0]
DDR2 A0/NBS0
AHB LPDDR
SDRAM A1/NWR2/NBS2
Controller A[15:2], A18
A16/BA0
MUX A17/BA1
Static Logic NCS0
Memory
Controller NCS1/SDCS
NRD/CFOE
NWR0/NWE/CFWE
NWR1/NBS1/CFIOR
NWR3/NBS3/CFIOW
SDCK, SDCK#, SDCKE
DQM[1:0]
CompactFlash DQS[1:0]
Logic
RAS, CAS
SDWE
SDA10
NAND Flash NANDOE

Logic NANDWE
A21/NANDALE
A22/NANDCLE
D[31:16]
ECC
Controller A[24:19]
PIO A25/CFRNW
NCS5/CFCS1
Chip Select
Address Decoders
Assignor NCS4/CFCS0
NCS3/NANDCS
NCS2
NWAIT
User Interface
CFCE1
CFCE2
DQM[3:2]
APB
160 AT91SAM9G45
AT91SAM9G45
Table 19-2. EBI I/O Lines Description

Name Function Type Active Level
EBI
EBI_D0 - EBI_D31 Data Bus I/O
EBI_A0 - EBI_A25 Address Bus Output
EBI_NWAIT External Wait Signal Input Low
SMC
EBI_NCS0 - EBI_NCS5 Chip Select Lines Output Low
EBI_NWR0 - EBI_NWR3 Write Signals Output Low
EBI_NRD Read Signal Output Low
EBI_NWE Write Enable Output Low
EBI_NBS0 - EBI_NBS3 Byte Mask Signals Output Low
EBI for NAND Flash Support
EBI_NANDCS NAND Flash Chip Select Line Output Low
EBI_NANDOE NAND Flash Output Enable Output Low
EBI_NANDWE NAND Flash Write Enable Output Low
DDR2/SDRAM Controller
EBI_SDCK, EBI_SDCK# DDR2/SDRAM Differential Clock Output
EBI_SDCKE DDR2/SDRAM Clock Enable Output High
EBI_SDCS DDR2/SDRAM Controller Chip Select Line Output Low
EBI_BA0 - EBI_BA1 Bank Select Output
EBI_SDWE DDR2/SDRAM Write Enable Output Low
EBI_RAS - EBI_CAS Row and Column Signal Output Low
EBI_SDA10 SDRAM Address 10 Line Output
The connection of some signals through the MUX logic is not direct and depends on the Memory
Controller in use at the moment.
Table 19-3 on page 161 details the connections between the two Memory Controllers and the
EBI pins.
Table 19-3. EBI Pins and Memory Controllers I/O Lines Connections
EBIx Pins SDRAM I/O Lines SMC I/O Lines
EBI_NWR1/NBS1/CFIOR NBS1 NWR1
EBI_A0/NBS0 Not Supported SMC_A0
EBI_A1/NBS2/NWR2 Not Supported SMC_A1
EBI_A[11:2] SDRAMC_A[9:0] SMC_A[11:2]
EBI_SDA10 SDRAMC_A10 Not Supported
EBI_A12 Not Supported SMC_A12
EBI_A[14:13] SDRAMC_A[12:11] SMC_A[14:13]
EBI_A[25:15] Not Supported SMC_A[25:15]
EBI_D[31:0] D[31:0] D[31:0]
161
19.3.5 Application Example
19.3.5.1 Hardware Interface

Table 19-4 on page 162 details the connections to be applied between the EBI pins and the
external devices for each Memory Controller.
Table 19-4. EBI Pins and External Static Devices Connections

Pins of the Interfaced Device
2 x 8-bit 4 x 8-bit 2 x 16-bit

Signals: 8-bit Static 16-bit Static 32-bit Static
Static Static Static
Device Device Device
EBI_ Devices Devices Devices
Controller SMC
D0 - D7 D0 - D7 D0 - D7 D0 - D7 D0 - D7 D0 - D7 D0 - D7
D8 - D15 – D8 - D15 D8 - D15 D8 - D15 D8 - 15 D8 - 15
D16 - D23 – – – D16 - D23 D16 - D23 D16 - D23

D24 - D31 – – – D24 - D31 D24 - D31 D24 - D31
A0/NBS0 A0 – NLB – NLB(3) BE0
A1/NWR2/NBS2 A1 A0 A0 WE(2) NLB(4) BE2
A2 - A22 A[2:22] A[1:21] A[1:21] A[0:20] A[0:20] A[0:20]

A23 - A25(5) A[23:25] A[22:24] A[22:24] A[21:23] A[21:23] A[21:23]
NCS0 CS CS CS CS CS CS
NCS1/DDRSDCS CS CS CS CS CS CS
NCS2 CS CS CS CS CS CS
NCS3/NANDCS CS CS CS CS CS CS
NCS4/CFCS0 CS CS CS CS CS CS
NCS5/CFCS1 CS CS CS CS CS CS
NRD/CFOE OE OE OE OE OE OE
NWR0/NWE WE WE(1) WE WE(2) WE WE

NWR1/NBS1 – WE(1) NUB WE(2) NUB(3) BE1
NWR3/NBS3 – – – WE(2) NUB(4) BE3

Notes: 1. NWR1 enables upper byte writes. NWR0 enables lower byte writes.
2. NWRx enables corresponding byte x writes. (x = 0,1,2 or 3)
3. NBS0 and NBS1 enable respectively lower and upper bytes of the lower 16-bit word.
4. NBS2 and NBS3 enable respectively lower and upper bytes of the upper 16-bit word.
162 AT91SAM9G45
AT91SAM9G45
Table 19-5. EBI Pins and External Device Connections

Signals: CompactFlash
DDR2/LPDDR SDRAM CompactFlash NAND Flash
EBI_ True IDE Mode
Controller DDRC SDRAMC SMC
D0 - D7 D0 - D7 D0 - D7 D0 - D7 D0 - D7 I/O0-I/O7
D8 - D15 D8 - D15 D8 - D15 D8 - 15 D8 - 15 I/O8-I/O15(4)
D16 - D31 – D16 - D31 – – –
A0/NBS0 – – A0 A0 –
A1/NWR2/NBS2 – – A1 A1 –
DQM0-DQM3 DQM0-DQM3 DQM0-DQM3 – – –
DQS0-DQM1 DQS0-DQS1 – – – –
A2 - A10 A[0:8] A[0:8] A[2:10] A[2:10] –
A11 A9 A9 – – –
SDA10 – A10 – – –
A12 – – – – –
A13 - A14 A[11:12] A[11:12] – – –
A15 – – – – –
A16/BA0 BA0 BA0 – – –
A17/BA1 BA1 BA1 – – –
A18 - A20 – – – – –
A21/NANDALE – – – – ALE
A22/NANDCLE – – REG REG CLE
A23 - A24 – – – – –
A25 – – CFRNW(1) CFRNW(1) –
NCS0 – – – – –
NCS1/DDRSDCS DDRCS SDCS – – –
NCS2 – – – – –
NCS3/NANDCS – – – – CE(3)
NCS4/CFCS0 – – CFCS0(1) CFCS0(1) –
NCS5/CFCS1 – – CFCS1(1) CFCS1(1) –
NANDOE – – – – OE
NANDWE – – – – WE
NRD/CFOE – – OE – –
NWR0/NWE/CFWE – – WE WE –
NWR1/NBS1/CFIOR – – IOR IOR –
NWR3/NBS3/CFIOW – – IOW IOW –
CFCE1 – – CE1 CS0 –
CFCE2 – – CE2 CS1 –
SDCK CK CLK – – –
SDCK# CK# – – – –
SDCKE CKE CKE – – –
RAS RAS RAS – – –
CAS CAS CAS – – –
163
Table 19-5. EBI Pins and External Device Connections (Continued)
Signals: CompactFlash
DDR2/LPDDR SDRAM CompactFlash NAND Flash
EBI_ True IDE Mode
Controller DDRC SDRAMC SMC
SDWE WE WE – – –
(5)
NWAIT – – WAIT WAIT –
Pxx(2) – – CD1 or CD2 CD1 or CD2 –
Pxx(2) – – – – CE(3)
Pxx(2) – – – – RDY
Notes: 1. Not directly connected to the CompactFlash slot. Permits the control of the bidirectional buffer between the EBI data bus and
the CompactFlash slot.
2. Any PIO line.
3. CE connection depends on the NAND Flash.
For standard NAND Flash devices, it must be connected to any free PIO line.
For "CE don't care" NAND Flash devices, it can be either connected to NCS3/NANDCS or to any free PIO line.
4. I/O8 - !/O15 pins used only for 16-bit NANDFlash device.
5. EBI_NWAIT signal is multiplexed with PC15.
19.3.5.2 Connection Examples

Figure 19-3 shows an example of connections between the EBI and external devices.
Figure 19-3. EBI Connections to Memory Devices

EBI
D0-D31
RAS
CAS 2M x 8 2M x 8
SDCK
SDCKE D0-D7
SDRAM D8-D15
SDRAM
D0-D7 D0-D7
SDWE
A0/NBS0 CS CS
NWR1/NBS1 CLK CLK
A1/NWR2/NBS2 A0-A9, A11 A2-A11, A13 A0-A9, A11 A2-A11, A13
CKE CKE
NWR3/NBS3 SDWE WE A10 SDA10 SDWE A10 SDA10
WE
NRD/NOE BA0 A16/BA0 BA0 A16/BA0
RAS RAS
NWR0/NWE BA1 A17/BA1 BA1 A17/BA1
CAS CAS
DQM DQM
NBS0 NBS1
SDA10
A2-A15
A16/BA0
A17/BA1
A18-A25
2M x 8 2M x 8
D16-D23 SDRAM D24-D31
SDRAM
D0-D7 D0-D7
NCS0
NCS1/SDCS CS CS
NCS2 CLK CLK
CKE A0-A9, A11 A2-A11, A13
NCS3 CKE A0-A9, A11
SDWE WE A10 SDA10 SDWE
NCS4 WE A10 A2-A11, A13
RAS BA0 A16/BA0
NCS5 RAS BA0 SDA10
CAS BA1 A17/BA1
CAS BA1 A16/BA0
DQM DQM A17/BA1
NBS3
NBS2
128K x 8 128K x 8
SRAM SRAM
A1-A17 A1-A17
D0-D7 D0-D7 A0-A16 D8-D15 D0-D7 A0-A16
CS CS
OE OE
NRD/NOE NRD/NOE
WE WE
A0/NWR0/NBS0 NWR1/NBS1
164 AT91SAM9G45
AT91SAM9G45
19.3.6 Product Dependencies
19.3.6.1 I/O Lines

The pins used for interfacing the External Bus Interface may be multiplexed with the PIO lines.
The programmer must first program the PIO controller to assign the External Bus Interface pins
to their peripheral function. If I/O lines of the External Bus Interface are not used by the applica-
tion, they can be used for other purposes by the PIO Controller.
19.3.7 Functional Description

The EBI transfers data between the internal AHB Bus (handled by the Bus Matrix) and the exter-
nal memories or peripheral devices. It controls the waveforms and the parameters of the
external address, data and control buses and is composed of the following elements:
• the Static Memory Controller (SMC)
• the DDR2/SDRAM Controller (DDR2SDRAMC)
• the ECC Controller (ECC)
• a chip select assignment feature that assigns an AHB address space to the external devices
• a multiplex controller circuit that shares the pins between the different Memory Controllers
• programmable CompactFlash support logic
• programmable NAND Flash support logic
19.3.7.1 Bus Multiplexing

The EBI offers a complete set of control signals that share the 32-bit data lines, the address
lines of up to 26 bits and the control signals through a multiplex logic operating in function of the
memory area requests.
Multiplexing is specifically organized in order to guarantee the maintenance of the address and
output control lines at a stable state while no external access is being performed. Multiplexing is
also designed to respect the data float times defined in the Memory Controllers. Furthermore,
refresh cycles of the DDR2 and SDRAM are executed independently by the DDR2SDRAM Con-
troller without delaying the other external Memory Controller accesses.
19.3.7.2 Pull-up Control

The EBI_CSA registers in the Chip Configuration User Interface permit enabling of on-chip pull-
up resistors on the data bus lines not multiplexed with the PIO Controller lines. The pull-up resis-
tors are enabled after reset. Setting the EBIx_DBPUC bit disables the pull-up resistors on the D0
to D15 lines. Enabling the pull-up resistor on the D16-D31 lines can be performed by program-
ming the appropriate PIO controller.
19.3.7.3 Static Memory Controller

For information on the Static Memory Controller, refer to the Static Memory Controller section.
19.3.7.4 DDR2SDRAM Controller

For information on the DDR2SDRAM Controller, refer to the DDR2SDRAMC section.
19.3.7.5 ECC Controller

For information on the ECC Controller, refer to the ECC section.
165
19.3.7.6 CompactFlash Support
The External Bus Interface 0 integrates circuitry that interfaces to CompactFlash devices.
The CompactFlash logic is driven by the Static Memory Controller (SMC) on the NCS4 and/or
NCS5 address space. Programming the EBI_CS4A and/or EBI_CS5A bit of the EBI_CSA Reg-
ister in the Chip Configuration User Interface to the appropriate value enables this logic. (For
details on this register, refer to the Chip Configuration User Interface in the Bus Matrix Section.)
Access to an external CompactFlash device is then made by accessing the address space
reserved to NCS4 and/or NCS5 (i.e., between 0x5000 0000 and 0x5FFF FFFF for NCS4 and
between 0x6000 0000 and 0x6FFF FFFF for NCS5).
All CompactFlash modes (Attribute Memory, Common Memory, I/O and True IDE) are sup-
ported but the signals _IOIS16 (I/O and True IDE modes) and _ATA SEL (True IDE mode) are
not handled.
I/O Mode, Common Memory Mode, Attribute Memory Mode and True IDE Mode
Within the NCS4 and/or NCS5 address space, the current transfer address is used to distinguish
I/O mode, common memory mode, attribute memory mode and True IDE mode.
The different modes are accessed through a specific memory mapping as illustrated on Figure
19-4. A[23:21] bits of the transfer address are used to select the desired mode as described in
Table 19-6 on page 166.
Figure 19-4. CompactFlash Memory Mapping
True IDE Alternate Mode Space

Offset 0x00E0 0000
True IDE Mode Space

Offset 0x00C0 0000
CF Address Space I/O Mode Space

Offset 0x0080 0000
Common Memory Mode Space

Offset 0x0040 0000
Attribute Memory Mode Space

Offset 0x0000 0000
Note: The A22 pin is used to drive the REG signal of the CompactFlash Device (except in True IDE
mode).
Table 19-6. CompactFlash Mode Selection

A[23:21] Mode Base Address
000 Attribute Memory
010 Common Memory
100 I/O Mode
110 True IDE Mode
111 Alternate True IDE Mode
166 AT91SAM9G45
AT91SAM9G45
CFCE1 and CFCE2 Signals

To cover all types of access, the SMC must be alternatively set to drive 8-bit data bus or 16-bit
data bus. The odd byte access on the D[7:0] bus is only possible when the SMC is configured to
drive 8-bit memory devices on the corresponding NCS pin (NCS4 or NCS5). The Chip Select
Register (DBW field in the corresponding Chip Select Register) of the NCS4 and/or NCS5
address space must be set as shown in Table 19-7 to enable the required access type.
NBS1 and NBS0 are the byte selection signals from SMC and are available when the SMC is set
in Byte Select mode on the corresponding Chip Select.
The CFCE1 and CFCE2 waveforms are identical to the corresponding NCSx waveform. For
details on these waveforms and timings, refer to the Static Memory Controller section.
Table 19-7. CFCE1 and CFCE2 Truth Table

Mode CFCE2 CFCE1 DBW Comment SMC Access Mode
Attribute Memory NBS1 NBS0 16 bits Access to Even Byte on D[7:0] Byte Select
Access to Even Byte on D[7:0]
NBS1 NBS0 16bits Byte Select
Common Memory Access to Odd Byte on D[15:8]
1 0 8 bits Access to Odd Byte on D[7:0]
NBS1 NBS0 16 bits Byte Select
I/O Mode Access to Odd Byte on D[15:8]
1 0 8 bits Access to Odd Byte on D[7:0]
True IDE Mode
Task File 1 0 8 bits
Access to Odd Byte on D[7:0]
Data Register 1 0 16 bits Byte Select
Access to Odd Byte on D[15:8]
Alternate True IDE Mode
Control Register Don’t
0 1 Access to Even Byte on D[7:0] Don’t Care
Alternate Status Read Care
Drive Address 0 1 8 bits Access to Odd Byte on D[7:0]

Standby Mode or
Address Space is not 1 1 – – –
assigned to CF
Read/Write Signals
In I/O mode and True IDE mode, the CompactFlash logic drives the read and write command
signals of the SMC on CFIOR and CFIOW signals, while the CFOE and CFWE signals are deac-
tivated. Likewise, in common memory mode and attribute memory mode, the SMC signals are
driven on the CFOE and CFWE signals, while the CFIOR and CFIOW are deactivated. Figure
19-5 on page 168 demonstrates a schematic representation of this logic.
Attribute memory mode, common memory mode and I/O mode are supported by setting the
address setup and hold time on the NCS4 (and/or NCS5) chip select to the appropriate values.
For details on these signal waveforms, please refer to the section: Setup and Hold Cycles of the
Static Memory Controller section.
167
Figure 19-5. CompactFlash Read/Write Control Signals
External Bus Interface
SMC CompactFlash Logic
A23
1
0
1
0
0 CFOE
1
1 1 CFWE
A22
NRD_NOE
0
NWR0_NWE CFIOR
1 CFIOW
1
1
Table 19-8. CompactFlash Mode Selection

Mode Base Address CFOE CFWE CFIOR CFIOW
Attribute Memory
NRD NWR0_NWE 1 1
Common Memory
I/O Mode 1 1 NRD NWR0_NWE
True IDE Mode 0 1 NRD NWR0_NWE
Multiplexing of CompactFlash Signals on EBI Pins

Table 19-9 on page 168 and Table 19-10 on page 169 illustrate the multiplexing of the Compact-
Flash logic signals with other EBI signals on the EBI pins. The EBI pins in Table 19-9 are strictly
dedicated to the CompactFlash interface as soon as the EBI_CS4A and/or EBI_CS5A field of
the EBI_CSA Register in the Chip Configuration User Interface is set. These pins must not be
used to drive any other memory devices.
The EBI pins in Table 19-10 on page 169 remain shared between all memory areas when the
corresponding CompactFlash interface is enabled (EBI_CS4A = 1 and/or EBI_CS5A = 1).
Table 19-9. Dedicated CompactFlash Interface Multiplexing

CompactFlash Signals EBI Signals
Pins
CS4A = 1 CS5A = 1 CS4A = 0 CS5A = 0
NCS4/CFCS0 CFCS0 NCS4
NCS5/CFCS1 CFCS1 NCS5
168 AT91SAM9G45
AT91SAM9G45
Table 19-10. Shared CompactFlash Interface Multiplexing

Access to CompactFlash Device Access to Other EBI Devices
Pins CompactFlash Signals EBI Signals
NRD/CFOE CFOE NRD
NWR0/NWE/CFWE CFWE NWR0/NWE
NWR1/NBS1/CFIOR CFIOR NWR1/NBS1
NWR3/NBS3/CFIOW CFIOW NWR3/NBS3
A25/CFRNW CFRNW A25
Application Example
Figure 19-6 on page 169 illustrates an example of a CompactFlash application. CFCS0 and
CFRNW signals are not directly connected to the CompactFlash slot 0, but do control the direc-
tion and the output enable of the buffers between the EBI and the CompactFlash Device. The
timing of the CFCS0 signal is identical to the NCS4 signal. Moreover, the CFRNW signal
remains valid throughout the transfer, as does the address bus. The CompactFlash _WAIT sig-
nal is connected to the NWAIT input of the Static Memory Controller. For details on these
waveforms and timings, refer to the Static Memory Controller Section.
Figure 19-6. CompactFlash Application Example

EBI CompactFlash Connector
D[15:0] D[15:0]
DIR /OE
A25/CFRNW
NCS4/CFCS0
_CD1
CD (PIO)
_CD2
/OE
A[10:0] A[10:0]
A22/REG _REG
NOE/CFOE _OE
NWE/CFWE _WE
NWR1/CFIOR _IORD
NWR3/CFIOW _IOWR
CFCE1 _CE1
CFCE2 _CE2
NWAIT _WAIT
169
19.3.7.7 NAND Flash Support
External Bus Interfaces 1 integrate circuitry that interfaces to NAND Flash devices.
External Bus Interface
The NAND Flash logic is driven by the Static Memory Controller on the NCS2 address space.
Programming the EBI_CS2A field in the EBI_CSA Register in the Chip Configuration User Inter-
face to the appropriate value enables the NAND Flash logic. For details on this register, refer to
the Bus Matrix Section. Access to an external NAND Flash device is then made by accessing
the address space reserved to NCS4 (i.e., between 0x4000 0000 and 0x4FFF FFFF).
The NAND Flash Logic drives the read and write command signals of the SMC on the NANDOE
and NANDWE signals when the NCS3 signal is active. NANDOE and NANDWE are invalidated
as soon as the transfer address fails to lie in the NCS3 address space. See Figure 19-7 on page
170 for more information. For details on these waveforms, refer to the Static Memory Controller
section.
NAND Flash Signals
The address latch enable and command latch enable signals on the NAND Flash device are
driven by address bits A22 and A21 of the EBI address bus. The command, address or data
words on the data bus of the NAND Flash device are distinguished by using their address within
the NCSx address space. The chip enable (CE) signal of the device and the ready/busy (R/B)
signals are connected to PIO lines. The CE signal then remains asserted even when NCSx is
not selected, preventing the device from returning to standby mode.
Figure 19-7. NAND Flash Application Example
D[7:0]
AD[7:0]
A[22:21]
ALE
CLE
NCSx/NANDCS
Not Connected
EBI
NAND Flash
NANDOE
NOE
NANDWE
NWE
PIO CE
PIO R/B
170 AT91SAM9G45
AT91SAM9G45
19.3.8 Implementation Examples

The following hardware configurations are given for illustration only. The user should refer to the
memory manufacturer web site to check current device availability.
19.3.8.1 2x8-bit DDR2 on EBI

• Assign EBI_CS1 to the DDR2 controller by setting the EBI_CS1A bit in the EBI Chip Select
Register located in the bus matrix memory space.
• Initialize the DDR2 Controller depending on the DDR2 device and system bus frequency.
The DDR2 initialization sequence is described in the sub-section “DDR2 Device Initialization” of
the DDRSDRC section.
171
19.3.8.2 16-bit SDRAM
• Assign the EBI CS1 to the SDRAM controller by setting the bit EBI_CS1A in the EBI Chip
Select Assignment Register located in the bus matrix memory space.
• Initialize the SDRAM Controller depending on the SDRAM device and system bus frequency.
The Data Bus Width is to be programmed to 16 bits.
The SDRAM initialization sequence is described in the section “SDRAM Device Initialization” in
“SDRAM Controller (SDRAMC)”.
172 AT91SAM9G45
AT91SAM9G45
19.3.8.3 2x16-bit SDRAM

A[1..14]
D[0..31]
SDRAM
MN1 MN2
A2 23 2 D0 A2 23 2 D16
A3 A0 MT48LC16M16A2 DQ0 D1 A3 A0 MT48LC16M16A2 DQ0 D17
24 A1 DQ1 4 24 A1 DQ1 4
A4 25 5 D2 A4 25 5 D18
A5 A2 DQ2 D3 A5 A2 DQ2 D19
26 A3 DQ3 7 26 A3 DQ3 7
A6 29 8 D4 A6 29 8 D20
A7 A4 DQ4 D5 A7 A4 DQ4 D21
30 A5 DQ5 10 30 A5 DQ5 10
A8 31 11 D6 A8 31 11 D22
A9 A6 DQ6 D7 A9 A6 DQ6 D23
32 A7 DQ7 13 32 A7 DQ7 13
A10 33 42 D8 A10 33 42 D24
A11 A8 DQ8 D9 A11 A8 DQ8 D25
34 A9 DQ9 44 34 A9 DQ9 44
SDA10 22 45 D10 SDA10 22 45 D26
A13 A10 DQ10 D11 A13 A10 DQ10 D27
35 A11 DQ11 47 35 A11 DQ11 47
48 D12 48 D28
BA0 DQ12 D13 BA0 DQ12 D29
20 BA0 DQ13 50 20 BA0 DQ13 50
BA1 21 51 D14 BA1 21 51 D30
BA1 DQ14 D15 BA1 DQ14 D31
DQ15 53 DQ15 53
A14 36 A14 36
A12 VDDIOM A12 VDDIOM
40 N.C1 VDD 1 40 N.C1 VDD 1
VDD 14 VDD 14
CKE 37 27 CKE 37 27
CKE VDD CKE VDD
VDDQ 3 VDDQ 3
CLK 38 9 CLK 38 9
CLK VDDQ CLK VDDQ
VDDQ 43 VDDQ 43
DQM0 15 49 DQM2 15 49
DQM1 DQML VDDQ DQM3 DQML VDDQ
39 DQMH 39 DQMH
VSS 28 VSS 28
CAS 17 41 CAS 17 41
VDDIOM RAS CAS VSS RAS CAS VSS
18 RAS VSS 54 18 RAS VSS 54
6 C1 C3 C5 C7 6 C8 C10 C12 C14
VSSQ 100NF 100NF 100NF 100NF VSSQ 100NF 100NF 100NF 100NF
VSSQ 12 VSSQ 12
WE 16 46 C2 C4 C6 VDDIOM WE 16 46 C9 C11 C13
R1 WE VSSQ 100NF 100NF 100NF WE VSSQ 100NF 100NF 100NF
19 CS VSSQ 52 19 CS VSSQ 52
470K
MT48LC16M16A2P-75IT
R3
SDCS
R2 0R 256 Mbits 470K
256 Mbits
R4 0R
• Assign the EBI CS1 to the SDRAM controller by setting the bit EBI_CS1A in the EBI Chip
Select Assignment Register located in the bus matrix memory space.
• Initialize the SDRAM Controller depending on the SDRAM device and system bus frequency.
The Data Bus Width is to be programmed to 32 bits. The data lines D[16..31] are multiplexed
with PIO lines and thus the dedicated PIOs must be programmed in peripheral mode in the PIO
controller.
The SDRAM initialization sequence is described in the section “SDRAM Device Initialization” in
“SDRAM Controller (SDRAMC)”.
173
19.3.8.4 8-bit NAND Flash
D[0..7]
U1 K9F2G08U0M
CLE 16 29 D0
CLE I/O0 D1
ALE 17 ALE I/O1 30
NANDOE 8 31 D2
RE I/O2 D3
NANDWE 18 WE I/O3 32
(ANY PIO) 9 41 D4
CE I/O4 D5
I/O5 42
7 43 D6
(ANY PIO) R/B I/O6
R1 10K 44 D7
I/O7
3V3 19 WP
R2 10K N.C 48
N.C 47
1 N.C N.C 46
2 N.C N.C 45
3 N.C N.C 40
4 N.C N.C 39
5 N.C PRE 38
6 N.C N.C 35
10 N.C N.C 34
11 N.C N.C 33
14 N.C N.C 28
15 N.C N.C 27 3V3
20 N.C
21 N.C VCC 37
22 N.C VCC 12
23 C2
N.C
24 N.C
25 36 100NF
N.C VSS
26 N.C VSS 13
C1
2 Gb 100NF
TSOP48 PACKAGE
• Assign the EBI CS3 to the NAND Flash by setting the bit EBI_CS3A in the EBI Chip Select
Assignment Register located in the bus matrix memory space
• Reserve A21 / A22 for ALE / CLE functions. Address and Command Latches are controlled
respectively by setting to 1 the address bit A21 and A22 during accesses.
• Configure a PIO line as an input to manage the Ready/Busy signal.
• Configure Static Memory Controller CS3 Setup, Pulse, Cycle and Mode accordingly to NAND
Flash timings, the data bus width and the system bus frequency.
174 AT91SAM9G45
AT91SAM9G45
19.3.8.5 16-bit NAND Flash

D[0..15]
U1 MT29F2G16AABWP-ET
16 D0
CLE CLE I/O0 26 D1
ALE 17 ALE I/O1 28
8 D2
NANDOE RE I/O2 30 D3
NANDWE 18 WE I/O3 32
9 D4
(ANY PIO) CE I/O4 40 D5
I/O5 42 D6
(ANY PIO) 7 R/B I/O6 44
R1 10K D7
I/O7 46 D8
3V3 19 WP I/O8 27 D9
R2 10K I/O9 29 D10
I/O10 31 D11
1 N.C I/O11 33
2 D12
N.C I/O12 41 D13
3 N.C I/O13 43
4 D14
N.C I/O14 45 D15
5 N.C I/O15 47
6 N.C
10 N.C N.C 39
11 N.C PRE 38
14 N.C N.C 36
15 3V3
N.C
20 N.C
21 N.C VCC 37
22 N.C VCC 12
23 N.C
24 48 C2
N.C VSS 100NF
34 N.C VSS 25
35 N.C VSS 13
C1
2 Gb 100NF
TSOP48 PACKAGE
The software configuration is the same as for an 8-bit NAND Flash except for the data bus width
programmed in the mode register of the Static Memory Controller.
175
19.3.8.6 NOR Flash on NCS0
D[0..15]
A[1..22]
U1
A1 25 29 D0
A2 A0 DQ0 D1
24 A1 DQ1 31
A3 23 33 D2
A4 A2 DQ2 D3
22 A3 DQ3 35
A5 21 38 D4
A6 A4 DQ4 D5
20 A5 DQ5 40
A7 19 42 D6
A8 A6 DQ6 D7
18 A7 DQ7 44
A9 8 30 D8
A10 A8 DQ8 D9
7 A9 DQ9 32
A11 6 34 D10
A12 A10 DQ10 D11
5 A11 DQ11 36
A13 4 39 D12
A14 A12 DQ12 D13
3 A13 DQ13 41
A15 2 43 D14
A16 A14 DQ14 D15
1 A15 DQ15 45
A17 48
A18 A16
17 A17
A19 16 AT49BV6416
A20 A18
15 A19
A21 10 3V3
A22 A20
9 A21
VCCQ 47
NRST 12 RESET
NWE 11 WE
14 WP VCC 37 C2
3V3 13 VPP 100NF
NCS0 26 CE
NRD 28 OE VSS 46
VSS 27
C1
100NF
TSOP48 PACKAGE
The default configuration for the Static Memory Controller, byte select mode, 16-bit data bus,
Read/Write controlled by Chip Select, allows boot on 16-bit non-volatile memory at slow clock.
For another configuration, configure the Static Memory Controller CS0 Setup, Pulse, Cycle and
Mode depending on Flash timings and system bus frequency.
176 AT91SAM9G45
AT91SAM9G45
19.3.8.7 CompactFlash
D[0..15] MEMORY & I/O MODE
3V3
MN1A J1
D15 A2 A5 CF_D15 CF_D15 31 38 C1
D14 1B1 1A1 CF_D14 CF_D14 D15 VCC 100NF
A1 1B2 1A2 A6 30 D14
D13 B2 B5 CF_D13 CF_D13 29 13 C2
B1 1B4 1A4 B6 28 D12
D11 C2 C5 CF_D11 CF_D11 27
D10 1B5 1A5 CF_D10 CF_D10 D11
C1 1B6 1A6 C6 49 D10 GND 50
D9 D2 D5 CF_D9 CF_D9 48 1
D8 1B7 1A7 CF_D8 CF_D8 D9 GND
D1 1B8 1A8 D6 47 D8
CF_D7 6
CF_D6 D7
A3 1DIR 5 D6
A4 CF_D5 4
1OE CF_D4 D5
3 D4
74ALVCH32245 CF_D3 2
MN1B CF_D2 D3
23 D2
D7 E2 E5 CF_D7 CF_D1 22
D6 2B1 2A1 CF_D6 CF_D0 D1
E1 2B2 2A2 E6 21 D0
D5 F2 F5 CF_D5
D4 2B3 2A3 CF_D4 CD2
F1 2B4 2A4 F6 25 CD2#
D3 G2 G5 CF_D3 CD1 26
D2 2B5 2A5 CF_D2 3V3 CD1#
G1 2B6 2A6 G6
D1 H2 H5 CF_D1 CF_A10 8
D0 2B7 2A7 CF_D0 CF_A9 A10
H1 2B8 2A8 H6 10 A9
CF_A8 11
CF_A7 A8
A25/CFRNW H3 2DIR 12 A7
H4 CF_A6 14
2OE R1 R2 CF_A5 A6
CFCSx 4 MN2A 15 A5
6 MN2B 74ALVCH32245 47K 47K CF_A4 16
SN74ALVC32 A4
(CFCS0 or CFCS1) 5 SN74ALVC32 CF_A3 17
CD2 CF_A2 A3
1 18 A2
3 CF_A1 19
(ANY PIO) A1
2 CD1 CF_A0 20 A0

REG 44
MN1C REG#
A[0..10]
A10 J5 J2 CF_A10 WE 36 39
A9 3A1 3B1 CF_A9 OE WE# CSEL#
J6 3A2 3B2 J1 9 OE#
A8 K5 K2 CF_A8 IOWR 35 43
A7 3A3 3B3 CF_A7 IORD IOWR# INPACK#
K6 3A4 3B4 K1 34 IORD#
A6 L5 L2 CF_A6 45
A5 3A5 3B5 CF_A5 CE2 BVD2
L6 3A6 3B6 L1 32 CE2# BVD1 46
A4 M5 M2 CF_A4 CE1 7
A3 3A7 3B7 CF_A3 CE1#
M6 3A8 3B8 M1
24 WP
3V3 J3 3DIR VS2# 40
J4 WAIT# 42 33
3OE WAIT# VS1#
74ALVCH32245 RESET 41 37 RDY/BSY
MN1D RESET RDY/BSY
A2 N5 N2 CF_A2 N7E50-7516VY-20
A1 4A1 4B1 CF_A1
N6 4A2 4B2 N1
A0 P5 P2 CF_A0
4A3 4B3 REG
A22/REG P6 4A4 4B4 P1
CFWE R5 R2 WE
4A5 4B5 OE
CFOE R6 4A6 4B6 R1
CFIOW T6 T1 IOWR
4A7 4B7 IORD
CFIOR T5 4A8 4B8 T2
T3 4DIR
T4 4OE
74ALVCH32245
1
MN3A
SN74ALVC125
CFCE2 2 3 CE2
4
MN3B
SN74ALVC125
CFCE1 5 6 CE1
10
MN3C
SN74ALVC125
(ANY PIO) CFRST 9 8 RESET
13
MN3D R3
SN74ALVC125 10K
(ANY PIO) CFIRQ 11 12 RDY/BSY 3V3
MN4
3V3 5 VCC 1
R4
10K
NWAIT 4 2 WAIT# 3V3
GND 3
SN74LVC1G125-Q1
177
• Assign the EBI CS4 and/or EBI_CS5 to the CompactFlash Slot 0 and/or Slot 1 by setting the
bit EBI_CS4A and/or EBI_CS5A in the EBI Chip Select Assignment Register located in the
bus matrix memory space.
• The address line A23 is to select I/O (A23=1) or Memory mode (A23=0) and the address line
A22 for REG function.
• A22, A23, CFRNW, CFS0, CFCS1, CFCE1 and CFCE2 signals are multiplexed with PIO
lines and thus the dedicated PIOs must be programmed in peripheral mode in the PIO
controller.
• Configure a PIO line as an output for CFRST and two others as an input for CFIRQ and
CARD DETECT functions respectively.
• Configure SMC CS4 and/or SMC_CS5 (for Slot 0 or 1) Setup, Pulse, Cycle and Mode
accordingly to CompactFlash timings and system bus frequency.
178 AT91SAM9G45
AT91SAM9G45
19.3.8.8 CompactFlash True IDE

D[0..15] TRUE IDE MODE

3V3
MN1A J1
D15 A2 A5 CF_D15 CF_D15 31 38 C1
A1 1B2 1A2 A6 30 D14
D13 B2 B5 CF_D13 CF_D13 29 13 C2
B1 1B4 1A4 B6 28 D12
D11 C2 C5 CF_D11 CF_D11 27
D10 1B5 1A5 CF_D10 CF_D10 D11
C1 1B6 1A6 C6 49 D10 GND 50
D9 D2 D5 CF_D9 CF_D9 48 1
D8 1B7 1A7 CF_D8 CF_D8 D9 GND
D1 1B8 1A8 D6 47 D8
CF_D7 6
CF_D6 D7
A3 1DIR 5 D6
A4 CF_D5 4
1OE CF_D4 D5
3 D4
74ALVCH32245 CF_D3 2
MN1B CF_D2 D3
23 D2
D7 E2 E5 CF_D7 CF_D1 22
D6 2B1 2A1 CF_D6 CF_D0 D1
E1 2B2 2A2 E6 21 D0
D5 F2 F5 CF_D5
D4 2B3 2A3 CF_D4 CD2
F1 2B4 2A4 F6 25 CD2#
D3 G2 G5 CF_D3 CD1 26
D2 2B5 2A5 CF_D2 3V3 CD1#
G1 2B6 2A6 G6
D1 H2 H5 CF_D1 8
D0 2B7 2A7 CF_D0 A10
H1 2B8 2A8 H6 10 A9
11 A8
A25/CFRNW H3 2DIR 12 A7
H4 2OE 14 A6
CFCSx 4 R1 R2 15
MN2A A5
6 MN2B 74ALVCH32245 47K 47K 16
(CFCS0 or CFCS1) SN74ALVC32 A4
5 SN74ALVC32 17 A3
1 CD2 CF_A2 18
CF_A1 A2
(ANY PIO) 3 19 A1
2 CD1 CF_A0 20
A0
A[0..10] 3V3 44 REG#

MN1C
A10 J5 J2 CF_A10 36 39
A9 3A1 3B1 CF_A9 WE# CSEL#
J6 3A2 3B2 J1 9 ATA SEL#
A8 K5 K2 CF_A8 IOWR 35 43
A7 3A3 3B3 CF_A7 IORD IOWR# INPACK#
K6 3A4 3B4 K1 34 IORD#
A6 L5 L2 CF_A6 45
A5 3A5 3B5 CF_A5 CE2 DASP#
L6 3A6 3B6 L1 32 CS1# PDIAG# 46
A4 M5 M2 CF_A4 CE1 7
A3 3A7 3B7 CF_A3 CS0#
M6 3A8 3B8 M1
24 IOIS16#
3V3 J3 3DIR VS2# 40
J4 IORDY 42 33
3OE IORDY VS1#
74ALVCH32245 RESET# 41 37 INTRQ
MN1D RESET# INTRQ
A2 N5 N2 CF_A2 N7E50-7516VY-20
A1 4A1 4B1 CF_A1
N6 4A2 4B2 N1
A0 P5 P2 CF_A0
4A3 4B3 REG
A22/REG P6 4A4 4B4 P1
CFWE R5 R2 WE
4A5 4B5 OE
CFOE R6 4A6 4B6 R1
CFIOW T6 T1 IOWR
4A7 4B7 IORD
CFIOR T5 4A8 4B8 T2
T3 4DIR
T4 4OE
74ALVCH32245
1
MN3A
SN74ALVC125
CFCE2 2 3 CE2
4
MN3B
SN74ALVC125
CFCE1 5 6 CE1
10
MN3C
SN74ALVC125
(ANY PIO) CFRST 9 8 RESET#
13
MN3D R3
SN74ALVC125 10K
(ANY PIO) CFIRQ 11 12 INTRQ 3V3
MN4
3V3 5 VCC 1
R4
10K
NWAIT 4 2 IORDY 3V3
GND 3
SN74LVC1G125-Q1
179
• Assign the EBI CS4 and/or EBI_CS5 to the CompactFlash Slot 0 and/or Slot 1 by setting the
bit EBI_CS4A and/or EBI_CS5A in the EBI Chip Select Assignment Register located in the
bus matrix memory space.
• The address line A21 is to select Alternate True IDE (A21=1) or True IDE (A21=0) modes.
• A21, CFRNW, CFS0, CFCS1, CFCE1 and CFCE2 signals are multiplexed with PIO lines and
thus the dedicated PIOs must be programmed in peripheral mode in the PIO controller.
• Configure a PIO line as an output for CFRST and two others as an input for CFIRQ and
CARD DETECT functions respectively.
• Configure SMC CS4 and/or SMC_CS5 (for Slot 0 or 1) Setup, Pulse, Cycle and Mode
accordingly to CompactFlash timings and system bus frequency.
180 AT91SAM9G45
AT91SAM9G45
20. Static Memory Controller (SMC)
20.1 Description
The Static Memory Controller (SMC) generates the signals that control the access to the exter-
nal memory devices or peripheral devices. It has 6 Chip Selects and a 26-bit address bus. The
32-bit data bus can be configured to interface with 8-, 16-, or 32-bit external devices. Separate
read and write control signals allow for direct memory and peripheral interfacing. Read and write
signal waveforms are fully parametrizable.
The SMC can manage wait requests from external devices to extend the current access. The
SMC is provided with an automatic slow clock mode. In slow clock mode, it switches from user-
programmed waveforms to slow-rate specific waveforms on read and write signals. The SMC
supports asynchronous burst read in page mode access for page size up to 32 bytes.

Table 20-1. I/O Line Description
Name Description Type Active Level
NCS[7:0] Static Memory Controller Chip Select Lines Output Low
NRD Read Signal Output Low
NWR0/NWE Write 0/Write Enable Signal Output Low
A0/NBS0 Address Bit 0/Byte 0 Select Signal Output Low
NWR1/NBS1 Write 1/Byte 1 Select Signal Output Low
A1/NWR2/NBS2 Address Bit 1/Write 2/Byte 2 Select Signal Output Low
NWR3/NBS3 Write 3/Byte 3 Select Signal Output Low
A[25:2] Address Bus Output
D[31:0] Data Bus I/O
NWAIT External Wait Signal Input Low
20.3 Multiplexed Signals

Table 20-2. Static Memory Controller (SMC) Multiplexed Signals
Multiplexed Signals Related Function
NWR0 NWE Byte-write or byte-select access, see “Byte Write or Byte Select Access” on page 183
A0 NBS0 8-bit or 16-/32-bit data bus, see “Data Bus Width” on page 183
NWR1 NBS1 Byte-write or byte-select access see “Byte Write or Byte Select Access” on page 183
8-/16-bit or 32-bit data bus, see “Data Bus Width” on page 183.
A1 NWR2 NBS2
Byte-write or byte-select access, see “Byte Write or Byte Select Access” on page 183
NWR3 NBS3 Byte-write or byte-select access see “Byte Write or Byte Select Access” on page 183
181
20.4 Application Example
20.4.1 Hardware Interface
Figure 20-1. SMC Connections to Static Memory Devices
D0-D31
A0/NBS0
NWR0/NWE
128K x 8 128K x 8
NWR1/NBS1
D0 - D7 SRAM D8-D15 SRAM
A1/NWR2/NBS2 D0 - D7 D0-D7
NWR3/NBS3
CS CS
A0 - A16 A2 - A18 A2 - A18

A0 - A16
NCS0 NRD NRD
NCS1 OE OE
NCS2 NWR0/NWE NWR1/NBS1
WE WE
NCS3
NCS4
NCS5
NCS6
NCS7
128K x 8 128K x 8
D16 - D23 SRAM D24-D31 SRAM
A2 - A25 D0 - D7 D0-D7
CS CS
A2 - A18
A2 - A18
A0 - A16 A0 - A16
NRD NRD
OE OE
A1/NWR2/NBS2 NWR3/NBS3
WE WE
Static Memory
Controller
20.5.1 I/O Lines

The pins used for interfacing the Static Memory Controller may be multiplexed with the PIO
lines. The programmer must first program the PIO controller to assign the Static Memory Con-
troller pins to their peripheral function. If I/O Lines of the SMC are not used by the application,
they can be used for other purposes by the PIO Controller.
182 AT91SAM9G45
AT91SAM9G45
20.6 External Memory Mapping

The SMC provides up to 26 address lines, A[25:0]. This allows each chip select line to address
up to 64 Mbytes of memory.
If the physical memory device connected on one chip select is smaller than 64 Mbytes, it wraps
around and appears to be repeated within this space. The SMC correctly handles any valid
access to the memory device within the page (see Figure 20-2).
A[25:0] is only significant for 8-bit memory, A[25:1] is used for 16-bit memory, A[25:2] is used for
32-bit memory.
Figure 20-2. Memory Connections for Eight External Devices
NCS[0] - NCS[7]
NCS7
Memory Enable
NRD NCS6
Memory Enable
SMC NWE NCS5
Memory Enable
A[25:0] NCS4
Memory Enable
D[31:0] NCS3
Memory Enable
NCS2
Memory Enable
NCS1
Memory Enable
NCS0
Memory Enable
Output Enable
Write Enable
A[25:0]
8 or 16 or 32
D[31:0] or D[15:0] or
D[7:0]
20.7 Connection to External Devices
20.7.1 Data Bus Width

A data bus width of 8, 16, or 32 bits can be selected for each chip select. This option is con-
trolled by the field DBW in SMC_MODE (Mode Register) for the corresponding chip select.
Figure 20-3 shows how to connect a 512K x 8-bit memory on NCS2. Figure 20-4 shows how to
connect a 512K x 16-bit memory on NCS2. Figure 20-5 shows two 16-bit memories connected
as a single 32-bit memory
20.7.2 Byte Write or Byte Select Access

Each chip select with a 16-bit or 32-bit data bus can operate with one of two different types of
write access: byte write or byte select access. This is controlled by the BAT field of the
SMC_MODE register for the corresponding chip select.
183
Figure 20-3. Memory Connection for an 8-bit Data Bus
D[7:0] D[7:0]
A[18:2] A[18:2]
A0 A0
SMC A1 A1
NWE Write Enable

NRD Output Enable
NCS[2] Memory Enable
Figure 20-4. Memory Connection for a 16-bit Data Bus
D[15:0] D[15:0]
A[19:2] A[18:1]
A1 A[0]
SMC NBS0 Low Byte Enable

NBS1 High Byte Enable
NWE Write Enable
NRD Output Enable
Figure 20-5. Memory Connection for a 32-bit Data Bus
D[31:16]
D[31:16]
D[15:0] D[15:0]
A[20:2] A[18:0]
SMC NBS0 Byte 0 Enable

NBS1 Byte 1 Enable
NBS2 Byte 2 Enable
NBS3 Byte 3 Enable
NWE Write Enable
NRD Output Enable
184 AT91SAM9G45
AT91SAM9G45
20.7.2.1 Byte Write Access

Byte write access supports one byte write signal per byte of the data bus and a single read
signal.
Note that the SMC does not allow boot in Byte Write Access mode.
• For 16-bit devices: the SMC provides NWR0 and NWR1 write signals for respectively byte0
(lower byte) and byte1 (upper byte) of a 16-bit bus. One single read signal (NRD) is provided.
Byte Write Access is used to connect 2 x 8-bit devices as a 16-bit memory.
• For 32-bit devices: NWR0, NWR1, NWR2 and NWR3, are the write signals of byte0 (lower
byte), byte1, byte2 and byte 3 (upper byte) respectively. One single read signal (NRD) is
provided.
Byte Write Access is used to connect 4 x 8-bit devices as a 32-bit memory.
Byte Write option is illustrated on Figure 20-6.
20.7.2.2 Byte Select Access

In this mode, read/write operations can be enabled/disabled at a byte level. One byte-select line
per byte of the data bus is provided. One NRD and one NWE signal control read and write.
• For 16-bit devices: the SMC provides NBS0 and NBS1 selection signals for respectively
byte0 (lower byte) and byte1 (upper byte) of a 16-bit bus.
Byte Select Access is used to connect one 16-bit device.
• For 32-bit devices: NBS0, NBS1, NBS2 and NBS3, are the selection signals of byte0 (lower
byte), byte1, byte2 and byte 3 (upper byte) respectively. Byte Select Access is used to
connect two 16-bit devices.
Figure 20-7 shows how to connect two 16-bit devices on a 32-bit data bus in Byte Select Access
mode, on NCS3 (BAT = Byte Select Access).
185
Figure 20-6. Connection of 2 x 8-bit Devices on a 16-bit Bus: Byte Write Option
D[7:0] D[7:0]
D[15:8]
A[24:2] A[23:1]
SMC A1 A[0]
NWR0 Write Enable
NWR1
NRD Read Enable
D[15:8]
A[23:1]
A[0]
Write Enable
Read Enable
Memory Enable
20.7.2.3 Signal Multiplexing

Depending on the BAT, only the write signals or the byte select signals are used. To save IOs at
the external bus interface, control signals at the SMC interface are multiplexed. Table 20-3
shows signal multiplexing depending on the data bus width and the byte access type.
For 32-bit devices, bits A0 and A1 are unused. For 16-bit devices, bit A0 of address is unused.
When Byte Select Option is selected, NWR1 to NWR3 are unused. When Byte Write option is
selected, NBS0 to NBS3 are unused.
186 AT91SAM9G45
AT91SAM9G45
Figure 20-7. Connection of 2x16-bit Data Bus on a 32-bit Data Bus (Byte Select Option)
D[15:0] D[15:0]
D[31:16]
A[25:2] A[23:0]
NWE Write Enable
NBS0 Low Byte Enable
NBS1 High Byte Enable
SMC NBS2
NBS3
NRD Read Enable
D[31:16]
A[23:0]
Write Enable
Low Byte Enable

High Byte Enable
Read Enable
Memory Enable
Table 20-3. SMC Multiplexed Signal Translation

Signal Name 32-bit Bus 16-bit Bus 8-bit Bus
Device Type 1x32-bit 2x16-bit 4 x 8-bit 1x16-bit 2 x 8-bit 1 x 8-bit
Byte Access Type (BAT) Byte Select Byte Select Byte Write Byte Select Byte Write
NBS0_A0 NBS0 NBS0 NBS0 A0
NWE_NWR0 NWE NWE NWR0 NWE NWR0 NWE
NBS1_NWR1 NBS1 NBS1 NWR1 NBS1 NWR1
NBS2_NWR2_A1 NBS2 NBS2 NWR2 A1 A1 A1
NBS3_NWR3 NBS3 NBS3 NWR3
187
20.8 Standard Read and Write Protocols
In the following sections, the byte access type is not considered. Byte select lines (NBS0 to
NBS3) always have the same timing as the A address bus. NWE represents either the NWE sig-
nal in byte select access type or one of the byte write lines (NWR0 to NWR3) in byte write
access type. NWR0 to NWR3 have the same timings and protocol as NWE. In the same way,
NCS represents one of the NCS[0..5] chip select lines.
20.8.1 Read Waveforms

The read cycle is shown on Figure 20-8.
The read cycle starts with the address setting on the memory address bus, i.e.:
{A[25:2], A1, A0} for 8-bit devices
{A[25:2], A1} for 16-bit devices
A[25:2] for 32-bit devices.
Figure 20-8. Standard Read Cycle
MCK
A[25:2]
NBS0,NBS1,
NBS2,NBS3,
A0, A1
NRD
NCS
D[31:0]
NRD_SETUP NRD_PULSE NRD_HOLD
NCS_RD_SETUP NCS_RD_PULSE NCS_RD_HOLD
NRD_CYCLE
20.8.1.1 NRD Waveform

The NRD signal is characterized by a setup timing, a pulse width and a hold timing.
1. NRD_SETUP: the NRD setup time is defined as the setup of address before the NRD
falling edge;
2. NRD_PULSE: the NRD pulse length is the time between NRD falling edge and NRD
rising edge;
3. NRD_HOLD: the NRD hold time is defined as the hold time of address after the NRD
rising edge.
188 AT91SAM9G45
AT91SAM9G45
20.8.1.2 NCS Waveform

Similarly, the NCS signal can be divided into a setup time, pulse length and hold time:
1. NCS_RD_SETUP: the NCS setup time is defined as the setup time of address before
the NCS falling edge.
2. NCS_RD_PULSE: the NCS pulse length is the time between NCS falling edge and
NCS rising edge;
3. NCS_RD_HOLD: the NCS hold time is defined as the hold time of address after the
NCS rising edge.
20.8.1.3 Read Cycle

The NRD_CYCLE time is defined as the total duration of the read cycle, i.e., from the time where
address is set on the address bus to the point where address may change. The total read cycle
time is equal to:
NRD_CYCLE = NRD_SETUP + NRD_PULSE + NRD_HOLD
= NCS_RD_SETUP + NCS_RD_PULSE + NCS_RD_HOLD
All NRD and NCS timings are defined separately for each chip select as an integer number of
Master Clock cycles. To ensure that the NRD and NCS timings are coherent, user must define
the total read cycle instead of the hold timing. NRD_CYCLE implicitly defines the NRD hold time
and NCS hold time as:
NRD_HOLD = NRD_CYCLE - NRD SETUP - NRD PULSE
NCS_RD_HOLD = NRD_CYCLE - NCS_RD_SETUP - NCS_RD_PULSE
20.8.1.4 Null Delay Setup and Hold

If null setup and hold parameters are programmed for NRD and/or NCS, NRD and NCS remain
active continuously in case of consecutive read cycles in the same memory (see Figure 20-9).
189
Figure 20-9. No Setup, No Hold On NRD and NCS Read Signals
MCK
A[25:2]
NBS0,NBS1,
NBS2,NBS3,
A0, A1
NRD
NCS
D[31:0]
NRD_PULSE NRD_PULSE NRD_PULSE
NCS_RD_PULSE NCS_RD_PULSE NCS_RD_PULSE
NRD_CYCLE NRD_CYCLE NRD_CYCLE
20.8.1.5 Null Pulse

Programming null pulse is not permitted. Pulse must be at least set to 1. A null value leads to
unpredictable behavior.
20.8.2 Read Mode

As NCS and NRD waveforms are defined independently of one other, the SMC needs to know
when the read data is available on the data bus. The SMC does not compare NCS and NRD tim-
ings to know which signal rises first. The READ_MODE parameter in the SMC_MODE register
of the corresponding chip select indicates which signal of NRD and NCS controls the read
operation.
20.8.2.1 Read is Controlled by NRD (READ_MODE = 1):

Figure 20-10 shows the waveforms of a read operation of a typical asynchronous RAM. The
read data is available tPACC after the falling edge of NRD, and turns to ‘Z’ after the rising edge of
NRD. In this case, the READ_MODE must be set to 1 (read is controlled by NRD), to indicate
that data is available with the rising edge of NRD. The SMC samples the read data internally on
the rising edge of Master Clock that generates the rising edge of NRD, whatever the pro-
grammed waveform of NCS may be.
190 AT91SAM9G45
AT91SAM9G45
Figure 20-10. READ_MODE = 1: Data is sampled by SMC before the rising edge of NRD
MCK
A[25:2]
NBS0,NBS1,
NBS2,NBS3,
A0, A1
NRD
NCS
tPACC
D[31:0]
Data Sampling
20.8.2.2 Read is Controlled by NCS (READ_MODE = 0)

Figure 20-11 shows the typical read cycle of an LCD module. The read data is valid tPACC after
the falling edge of the NCS signal and remains valid until the rising edge of NCS. Data must be
sampled when NCS is raised. In that case, the READ_MODE must be set to 0 (read is controlled
by NCS): the SMC internally samples the data on the rising edge of Master Clock that generates
the rising edge of NCS, whatever the programmed waveform of NRD may be.
Figure 20-11. READ_MODE = 0: Data is sampled by SMC before the rising edge of NCS
MCK
A[25:2]
NBS0,NBS1,
NBS2,NBS3,
A0, A1
NRD
NCS
tPACC
D[31:0]
Data Sampling
191
20.8.3 Write Waveforms
The write protocol is similar to the read protocol. It is depicted in Figure 20-12. The write cycle
starts with the address setting on the memory address bus.
20.8.3.1 NWE Waveforms

The NWE signal is characterized by a setup timing, a pulse width and a hold timing.
1. NWE_SETUP: the NWE setup time is defined as the setup of address and data before
the NWE falling edge;
2. NWE_PULSE: The NWE pulse length is the time between NWE falling edge and NWE
rising edge;
3. NWE_HOLD: The NWE hold time is defined as the hold time of address and data after
the NWE rising edge.
The NWE waveforms apply to all byte-write lines in Byte Write access mode: NWR0 to NWR3.
20.8.3.2 NCS Waveforms

The NCS signal waveforms in write operation are not the same that those applied in read opera-
tions, but are separately defined:
1. NCS_WR_SETUP: the NCS setup time is defined as the setup time of address before
the NCS falling edge.
2. NCS_WR_PULSE: the NCS pulse length is the time between NCS falling edge and
NCS rising edge;
3. NCS_WR_HOLD: the NCS hold time is defined as the hold time of address after the
NCS rising edge.
Figure 20-12. Write Cycle
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
NWE
NCS
NWE_SETUP NWE_PULSE NWE_HOLD
NCS_WR_SETUP NCS_WR_PULSE NCS_WR_HOLD
NWE_CYCLE
192 AT91SAM9G45
AT91SAM9G45
20.8.3.3 Write Cycle

The write_cycle time is defined as the total duration of the write cycle, that is, from the time
where address is set on the address bus to the point where address may change. The total write
cycle time is equal to:
NWE_CYCLE = NWE_SETUP + NWE_PULSE + NWE_HOLD
= NCS_WR_SETUP + NCS_WR_PULSE + NCS_WR_HOLD
All NWE and NCS (write) timings are defined separately for each chip select as an integer num-
ber of Master Clock cycles. To ensure that the NWE and NCS timings are coherent, the user
must define the total write cycle instead of the hold timing. This implicitly defines the NWE hold
time and NCS (write) hold times as:
NWE_HOLD = NWE_CYCLE - NWE_SETUP - NWE_PULSE
NCS_WR_HOLD = NWE_CYCLE - NCS_WR_SETUP - NCS_WR_PULSE
20.8.3.4 Null Delay Setup and Hold

If null setup parameters are programmed for NWE and/or NCS, NWE and/or NCS remain active
continuously in case of consecutive write cycles in the same memory (see Figure 20-13). How-
ever, for devices that perform write operations on the rising edge of NWE or NCS, such as
SRAM, either a setup or a hold must be programmed.
Figure 20-13. Null Setup and Hold Values of NCS and NWE in Write Cycle
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
NWE,
NWR0, NWR1,
NWR2, NWR3
NCS
D[31:0]
NWE_PULSE NWE_PULSE NWE_PULSE
NCS_WR_PULSE NCS_WR_PULSE NCS_WR_PULSE
NWE_CYCLE NWE_CYCLE NWE_CYCLE
20.8.3.5 Null Pulse

Programming null pulse is not permitted. Pulse must be at least set to 1. A null value leads to
193
20.8.4 Write Mode
The WRITE_MODE parameter in the SMC_MODE register of the corresponding chip select indi-
cates which signal controls the write operation.
20.8.4.1 Write is Controlled by NWE (WRITE_MODE = 1):

Figure 20-14 shows the waveforms of a write operation with WRITE_MODE set to 1. The data is
put on the bus during the pulse and hold steps of the NWE signal. The internal data buffers are
turned out after the NWE_SETUP time, and until the end of the write cycle, regardless of the
programmed waveform on NCS.
Figure 20-14. WRITE_MODE = 1. The write operation is controlled by NWE
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
NWE,
NWR0, NWR1,
NWR2, NWR3
NCS
D[31:0]
20.8.4.2 Write is Controlled by NCS (WRITE_MODE = 0)

Figure 20-15 shows the waveforms of a write operation with WRITE_MODE set to 0. The data is
put on the bus during the pulse and hold steps of the NCS signal. The internal data buffers are
turned out after the NCS_WR_SETUP time, and until the end of the write cycle, regardless of
the programmed waveform on NWE.
194 AT91SAM9G45
AT91SAM9G45
Figure 20-15. WRITE_MODE = 0. The write operation is controlled by NCS
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
NWE,
NWR0, NWR1,
NWR2, NWR3
NCS
D[31:0]
20.8.5 Coding Timing Parameters

All timing parameters are defined for one chip select and are grouped together in one
SMC_REGISTER according to their type.
The SMC_SETUP register groups the definition of all setup parameters:
• NRD_SETUP, NCS_RD_SETUP, NWE_SETUP, NCS_WR_SETUP
The SMC_PULSE register groups the definition of all pulse parameters:
• NRD_PULSE, NCS_RD_PULSE, NWE_PULSE, NCS_WR_PULSE
The SMC_CYCLE register groups the definition of all cycle parameters:
• NRD_CYCLE, NWE_CYCLE
Table 20-4 shows how the timing parameters are coded and their permitted range.
Table 20-4. Coding and Range of Timing Parameters

Permitted Range
Coded Value Number of Bits Effective Value Coded Value Effective Value
setup [5:0] 6 128 x setup[5] + setup[4:0] 0 ≤≤31 0 ≤≤128+31
pulse [6:0] 7 256 x pulse[6] + pulse[5:0] 0 ≤≤63 0 ≤≤256+63
0 ≤≤256+127
cycle [8:0] 9 256 x cycle[8:7] + cycle[6:0] 0 ≤≤127 0 ≤≤512+127
0 ≤≤768+127
195
20.8.6 Reset Values of Timing Parameters
Table 20-5 gives the default value of timing parameters at reset.
Table 20-5. Reset Values of Timing Parameters

Register Reset Value
SMC_SETUP 0x01010101 All setup timings are set to 1
SMC_PULSE 0x01010101 All pulse timings are set to 1
The read and write operation last 3 Master Clock
SMC_CYCLE 0x00030003
cycles and provide one hold cycle
WRITE_MODE 1 Write is controlled with NWE
READ_MODE 1 Read is controlled with NRD
20.8.7 Usage Restriction

The SMC does not check the validity of the user-programmed parameters. If the sum of SETUP
and PULSE parameters is larger than the corresponding CYCLE parameter, this leads to unpre-
dictable behavior of the SMC.
For read operations:
Null but positive setup and hold of address and NRD and/or NCS can not be guaranteed at the
memory interface because of the propagation delay of theses signals through external logic and
pads. If positive setup and hold values must be verified, then it is strictly recommended to pro-
gram non-null values so as to cover possible skews between address, NCS and NRD signals.
For write operations:
If a null hold value is programmed on NWE, the SMC can guarantee a positive hold of address,
byte select lines, and NCS signal after the rising edge of NWE. This is true for WRITE_MODE =
1 only. See “Early Read Wait State” on page 197.
For read and write operations: a null value for pulse parameters is forbidden and may lead to
In read and write cycles, the setup and hold time parameters are defined in reference to the
address bus. For external devices that require setup and hold time between NCS and NRD sig-
nals (read), or between NCS and NWE signals (write), these setup and hold times must be
converted into setup and hold times in reference to the address bus.
20.9 Automatic Wait States

Under certain circumstances, the SMC automatically inserts idle cycles between accesses to
avoid bus contention or operation conflict.
20.9.1 Chip Select Wait States

The SMC always inserts an idle cycle between 2 transfers on separate chip selects. This idle
cycle ensures that there is no bus contention between the de-activation of one device and the
activation of the next one.
During chip select wait state, all control lines are turned inactive: NBS0 to NBS3, NWR0 to
NWR3, NCS[0..5], NRD lines are all set to 1.
Figure 20-16 illustrates a chip select wait state between access on Chip Select 0 and Chip
Select 2.
196 AT91SAM9G45
AT91SAM9G45
Figure 20-16. Chip Select Wait State between a Read Access on NCS0 and a Write Access on NCS2
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1
NRD
NWE
NCS0
NCS2
NRD_CYCLE NWE_CYCLE
D[31:0]
Read to Write Chip Select

Wait State Wait State
20.9.2 Early Read Wait State

In some cases, the SMC inserts a wait state cycle between a write access and a read access to
allow time for the write cycle to end before the subsequent read cycle begins. This wait state is
not generated in addition to a chip select wait state. The early read cycle thus only occurs
between a write and read access to the same memory device (same chip select).
An early read wait state is automatically inserted if at least one of the following conditions is
valid:
• if the write controlling signal has no hold time and the read controlling signal has no setup
time (Figure 20-17).
• in NCS write controlled mode (WRITE_MODE = 0), if there is no hold timing on the NCS
signal and the NCS_RD_SETUP parameter is set to 0, regardless of the read mode (Figure
20-18). The write operation must end with a NCS rising edge. Without an Early Read Wait
State, the write operation could not complete properly.
• in NWE controlled mode (WRITE_MODE = 1) and if there is no hold timing (NWE_HOLD =
0), the feedback of the write control signal is used to control address, data, chip select and
byte select lines. If the external write control signal is not inactivated as expected due to load
capacitances, an Early Read Wait State is inserted and address, data and control signals are
maintained one more cycle. See Figure 20-19.
197
Figure 20-17. Early Read Wait State: Write with No Hold Followed by Read with No Setup
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
NWE
NRD
no hold
no setup
D[31:0]
write cycle Early Read read cycle

wait state
Figure 20-18. Early Read Wait State: NCS Controlled Write with No Hold Followed by a Read with No NCS Setup
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1
NCS
NRD
no hold no setup
D[31:0]

(WRITE_MODE = 0) wait state (READ_MODE = 0 or READ_MODE = 1)
198 AT91SAM9G45
AT91SAM9G45
Figure 20-19. Early Read Wait State: NWE-controlled Write with No Hold Followed by a Read with one Set-up Cycle
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
internal write controlling signal
external write controlling signal

(NWE)
no hold read setup = 1

NRD
D[31:0]

(WRITE_MODE = 1) wait state (READ_MODE = 0 or READ_MODE = 1)
20.9.3 Reload User Configuration Wait State

The user may change any of the configuration parameters by writing the SMC user interface.
When detecting that a new user configuration has been written in the user interface, the SMC
inserts a wait state before starting the next access. The so called “Reload User Configuration
Wait State” is used by the SMC to load the new set of parameters to apply to next accesses.
The Reload Configuration Wait State is not applied in addition to the Chip Select Wait State. If
accesses before and after re-programming the user interface are made to different devices
(Chip Selects), then one single Chip Select Wait State is applied.
On the other hand, if accesses before and after writing the user interface are made to the same
device, a Reload Configuration Wait State is inserted, even if the change does not concern the
current Chip Select.
20.9.3.1 User Procedure

To insert a Reload Configuration Wait State, the SMC detects a write access to any
SMC_MODE register of the user interface. If the user only modifies timing registers
(SMC_SETUP, SMC_PULSE, SMC_CYCLE registers) in the user interface, he must validate
the modification by writing the SMC_MODE, even if no change was made on the mode
parameters.
The user must not change the configuration parameters of an SMC Chip Select (Setup, Pulse,
Cycle, Mode) if accesses are performed on this CS during the modification. Any change of the
Chip Select parameters, while fetching the code from a memory connected on this CS, may lead
199
to unpredictable behavior. The instructions used to modify the parameters of an SMC Chip
Select can be executed from the internal RAM or from a memory connected to another CS.
20.9.3.2 Slow Clock Mode Transition

A Reload Configuration Wait State is also inserted when the Slow Clock Mode is entered or
exited, after the end of the current transfer (see “Slow Clock Mode” on page 211).
20.9.4 Read to Write Wait State

Due to an internal mechanism, a wait cycle is always inserted between consecutive read and
write SMC accesses.
This wait cycle is referred to as a read to write wait state in this document.
This wait cycle is applied in addition to chip select and reload user configuration wait states
when they are to be inserted. See Figure 20-16 on page 197.
200 AT91SAM9G45
AT91SAM9G45
20.10 Data Float Wait States

Some memory devices are slow to release the external bus. For such devices, it is necessary to
add wait states (data float wait states) after a read access:
• before starting a read access to a different external memory
• before starting a write access to the same device or to a different external one.
The Data Float Output Time (t DF ) for each external memory device is programmed in the
TDF_CYCLES field of the SMC_MODE register for the corresponding chip select. The value of
TDF_CYCLES indicates the number of data float wait cycles (between 0 and 15) before the
external device releases the bus, and represents the time allowed for the data output to go to
high impedance after the memory is disabled.
Data float wait states do not delay internal memory accesses. Hence, a single access to an
external memory with long t DF will not slow down the execution of a program from internal
memory.
The data float wait states management depends on the READ_MODE and the TDF_MODE
fields of the SMC_MODE register for the corresponding chip select.
20.10.1 READ_MODE
Setting the READ_MODE to 1 indicates to the SMC that the NRD signal is responsible for turn-
ing off the tri-state buffers of the external memory device. The Data Float Period then begins
after the rising edge of the NRD signal and lasts TDF_CYCLES MCK cycles.
When the read operation is controlled by the NCS signal (READ_MODE = 0), the TDF field gives
the number of MCK cycles during which the data bus remains busy after the rising edge of NCS.
Figure 20-20 illustrates the Data Float Period in NRD-controlled mode (READ_MODE =1),
assuming a data float period of 2 cycles (TDF_CYCLES = 2). Figure 20-21 shows the read oper-
ation when controlled by NCS (READ_MODE = 0) and the TDF_CYCLES parameter equals 3.
201
Figure 20-20. TDF Period in NRD Controlled Read Access (TDF = 2)
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
NRD
NCS
tpacc
D[31:0]
TDF = 2 clock cycles
NRD controlled read operation
Figure 20-21. TDF Period in NCS Controlled Read Operation (TDF = 3)
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1
NRD
NCS
tpacc
D[31:0]
TDF = 3 clock cycles
NCS controlled read operation
202 AT91SAM9G45
AT91SAM9G45
20.10.2 TDF Optimization Enabled (TDF_MODE = 1)

When the TDF_MODE of the SMC_MODE register is set to 1 (TDF optimization is enabled), the
SMC takes advantage of the setup period of the next access to optimize the number of wait
states cycle to insert.
Figure 20-22 shows a read access controlled by NRD, followed by a write access controlled by
NWE, on Chip Select 0. Chip Select 0 has been programmed with:
NRD_HOLD = 4; READ_MODE = 1 (NRD controlled)
NWE_SETUP = 3; WRITE_MODE = 1 (NWE controlled)
TDF_CYCLES = 6; TDF_MODE = 1 (optimization enabled).
Figure 20-22. TDF Optimization: No TDF wait states are inserted if the TDF period is over when the next access begins
MCK
A[25:2]
NRD
NRD_HOLD= 4
NWE
NWE_SETUP= 3
NCS0
TDF_CYCLES = 6
D[31:0]
read access on NCS0 (NRD controlled) Read to Write write access on NCS0 (NWE controlled)
Wait State
20.10.3 TDF Optimization Disabled (TDF_MODE = 0)

When optimization is disabled, tdf wait states are inserted at the end of the read transfer, so that
the data float period is ended when the second access begins. If the hold period of the read1
controlling signal overlaps the data float period, no additional tdf wait states will be inserted.
Figure 20-23, Figure 20-24 and Figure 20-25 illustrate the cases:
• read access followed by a read access on another chip select,
• read access followed by a write access on another chip select,
• read access followed by a write access on the same chip select,
with no TDF optimization.
203
Figure 20-23. TDF Optimization Disabled (TDF Mode = 0). TDF wait states between 2 read accesses on different chip
selects
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
read1 controlling signal

(NRD) read1 hold = 1 read2 setup = 1

(NRD) TDF_CYCLES = 6
D[31:0]
5 TDF WAIT STATES
read1 cycle read 2 cycle

TDF_CYCLES = 6 TDF_MODE = 0
(optimization disabled)
Chip Select Wait State
Figure 20-24. TDF Mode = 0: TDF wait states between a read and a write access on different chip selects
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1
(NRD) read1 hold = 1 write2 setup = 1
write2 controlling signal

(NWE) TDF_CYCLES = 4
D[31:0]
read1 cycle 2 TDF WAIT STATES write2 cycle

TDF_CYCLES = 4 TDF_MODE = 0
Read to Write Chip Select (optimization disabled)
Wait State Wait State
204 AT91SAM9G45
AT91SAM9G45
Figure 20-25. TDF Mode = 0: TDF wait states between read and write accesses on the same chip select
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0, A1

(NRD)
read1 hold = 1 write2 setup = 1
write2 controlling signal

(NWE) TDF_CYCLES = 5
D[31:0]
4 TDF WAIT STATES

read1 cycle
TDF_CYCLES = 5 write2 cycle
Read to Write
Wait State TDF_MODE = 0
(optimization disabled)
20.11 External Wait

Any access can be extended by an external device using the NWAIT input signal of the SMC.
The EXNW_MODE field of the SMC_MODE register on the corresponding chip select must be
set to either to “10” (frozen mode) or “11” (ready mode). When the EXNW_MODE is set to “00”
(disabled), the NWAIT signal is simply ignored on the corresponding chip select. The NWAIT
signal delays the read or write operation in regards to the read or write controlling signal,
depending on the read and write modes of the corresponding chip select.
20.11.1 Restriction
When one of the EXNW_MODE is enabled, it is mandatory to program at least one hold
cycle for the read/write controlling signal. For that reason, the NWAIT signal cannot be
used in Page Mode (“Asynchronous Page Mode” on page 214), or in Slow Clock Mode
(“Slow Clock Mode” on page 211).
The NWAIT signal is assumed to be a response of the external device to the read/write request
of the SMC. Then NWAIT is examined by the SMC only in the pulse state of the read or write
controlling signal. The assertion of the NWAIT signal outside the expected period has no impact
on SMC behavior.
205
20.11.2 Frozen Mode
When the external device asserts the NWAIT signal (active low), and after internal synchroniza-
tion of this signal, the SMC state is frozen, i.e., SMC internal counters are frozen, and all control
signals remain unchanged. When the resynchronized NWAIT signal is deasserted, the SMC
completes the access, resuming the access from the point where it was stopped. See Figure 20-
26. This mode must be selected when the external device uses the NWAIT signal to delay the
access and to freeze the SMC.
The assertion of the NWAIT signal outside the expected period is ignored as illustrated in Figure
20-27.
Figure 20-26. Write Access with NWAIT Assertion in Frozen Mode (EXNW_MODE = 10)
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1 FROZEN STATE
4 3 2 1 1 1 1 0
NWE
6 5 4 3 2 2 2 2 1 0
NCS
D[31:0]
NWAIT
internally synchronized
NWAIT signal
Write cycle
EXNW_MODE = 10 (Frozen)
WRITE_MODE = 1 (NWE_controlled)
NWE_PULSE = 5
NCS_WR_PULSE = 7
206 AT91SAM9G45
AT91SAM9G45
Figure 20-27. Read Access with NWAIT Assertion in Frozen Mode (EXNW_MODE = 10)
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1
FROZEN STATE
4 3 2 2 2 1 0
NCS 2 1 0
1 0
NRD 5 5 5 4 3 2 1 0
NWAIT
NWAIT signal
Read cycle
EXNW_MODE = 10 (Frozen)
READ_MODE = 0 (NCS_controlled)
Assertion is ignored
NRD_PULSE = 2, NRD_HOLD = 6
NCS_RD_PULSE =5, NCS_RD_HOLD =3
207
AT91SAM9G45
20.11.3 Ready Mode

In Ready mode (EXNW_MODE = 11), the SMC behaves differently. Normally, the SMC begins
the access by down counting the setup and pulse counters of the read/write controlling signal. In
the last cycle of the pulse phase, the resynchronized NWAIT signal is examined.
If asserted, the SMC suspends the access as shown in Figure 20-28 and Figure 20-29. After
deassertion, the access is completed: the hold step of the access is performed.
This mode must be selected when the external device uses deassertion of the NWAIT signal to
indicate its ability to complete the read or write operation.
If the NWAIT signal is deasserted before the end of the pulse, or asserted after the end of the
pulse of the controlling read/write signal, it has no impact on the access length as shown in Fig-
ure 20-29.
Figure 20-28. NWAIT Assertion in Write Access: Ready Mode (EXNW_MODE = 11)
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1
Wait STATE
4 3 2 1 0 0 0
NWE
6 5 4 3 2 1 1 1 0
NCS
D[31:0]
NWAIT
NWAIT signal
Write cycle
EXNW_MODE = 11 (Ready mode)

WRITE_MODE = 1 (NWE_controlled)
NWE_PULSE = 5
NCS_WR_PULSE = 7
208
AT91SAM9G45
Figure 20-29. NWAIT Assertion in Read Access: Ready Mode (EXNW_MODE = 11)
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1
Wait STATE
6 5 4 3 2 1 0 0
NCS
6 5 4 3 2 1 1 0
NRD
NWAIT
NWAIT signal
Read cycle
EXNW_MODE = 11(Ready mode)
READ_MODE = 0 (NCS_controlled)
Assertion is ignored Assertion is ignored
NRD_PULSE = 7
NCS_RD_PULSE =7
209
AT91SAM9G45
20.11.4 NWAIT Latency and Read/Write Timings

There may be a latency between the assertion of the read/write controlling signal and the asser-
tion of the NWAIT signal by the device. The programmed pulse length of the read/write
controlling signal must be at least equal to this latency plus the 2 cycles of resynchronization + 1
cycle. Otherwise, the SMC may enter the hold state of the access without detecting the NWAIT
signal assertion. This is true in frozen mode as well as in ready mode. This is illustrated on Fig-
ure 20-30.
When EXNW_MODE is enabled (ready or frozen), the user must program a pulse length of the
read and write controlling signal of at least:
minimal pulse length = NWAIT latency + 2 resynchronization cycles + 1 cycle
Figure 20-30. NWAIT Latency
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1 WAIT STATE
4 3 2 1 0 0 0
NRD
minimal pulse length
NWAIT
intenally synchronized NWAIT latency 2 cycle resynchronization

NWAIT signal
Read cycle
EXNW_MODE = 10 or 11
READ_MODE = 1 (NRD_controlled)
NRD_PULSE = 5
210
AT91SAM9G45
20.12 Slow Clock Mode

The SMC is able to automatically apply a set of “slow clock mode” read/write waveforms when
an internal signal driven by the Power Management Controller is asserted because MCK has
been turned to a very slow clock rate (typically 32kHz clock rate). In this mode, the user-pro-
grammed waveforms are ignored and the slow clock mode waveforms are applied. This mode is
provided so as to avoid reprogramming the User Interface with appropriate waveforms at very
slow clock rate. When activated, the slow mode is active on all chip selects.
20.12.1 Slow Clock Mode Waveforms

Figure 20-31 illustrates the read and write operations in slow clock mode. They are valid on all
chip selects. Table 20-6 indicates the value of read and write parameters in slow clock mode.
Figure 20-31. Read/write Cycles in Slow Clock Mode
MCK MCK
A[25:2] A[25:2]
NBS0, NBS1, NBS0, NBS1,

NBS2, NBS3, NBS2, NBS3,
A0,A1 A0,A1
NRD
NWE 1 1 1
1 1
NCS
NCS
NWE_CYCLE = 3 NRD_CYCLE = 2
SLOW CLOCK MODE WRITE SLOW CLOCK MODE READ
Table 20-6. Read and Write Timing Parameters in Slow Clock Mode
Read Parameters Duration (cycles) Write Parameters Duration (cycles)
NRD_SETUP 1 NWE_SETUP 1
NRD_PULSE 1 NWE_PULSE 1
NCS_RD_SETUP 0 NCS_WR_SETUP 0
NCS_RD_PULSE 2 NCS_WR_PULSE 3
NRD_CYCLE 2 NWE_CYCLE 3
211
AT91SAM9G45
20.12.2 Switching from (to) Slow Clock Mode to (from) Normal Mode
When switching from slow clock mode to the normal mode, the current slow clock mode transfer
is completed at high clock rate, with the set of slow clock mode parameters.See Figure 20-32 on
page 212. The external device may not be fast enough to support such timings.
Figure 20-33 illustrates the recommended procedure to properly switch from one mode to the
other.
Figure 20-32. Clock Rate Transition Occurs while the SMC is Performing a Write Operation
Slow Clock Mode

internal signal from PMC
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1
NWE
1 1 1 1 1 1 2 3 2
NCS
NWE_CYCLE = 3 NWE_CYCLE = 7
SLOW CLOCK MODE WRITE SLOW CLOCK MODE WRITE NORMAL MODE WRITE
This write cycle finishes with the slow clock mode set Slow clock mode transition is detected:
of parameters after the clock rate transition Reload Configuration Wait State
212
AT91SAM9G45
Figure 20-33. Recommended Procedure to Switch from Slow Clock Mode to Normal Mode or from Normal Mode to Slow
Clock Mode
Slow Clock Mode
internal signal from PMC
MCK
A[25:2]
NBS0, NBS1,
NBS2, NBS3,
A0,A1
NWE
1 1 1 2 3 2
NCS
SLOW CLOCK MODE WRITE IDLE STATE NORMAL MODE WRITE
Reload Configuration
Wait State
213
AT91SAM9G45
20.13 Asynchronous Page Mode

The SMC supports asynchronous burst reads in page mode, providing that the page mode is
enabled in the SMC_MODE register (PMEN field). The page size must be configured in the
SMC_MODE register (PS field) to 4, 8, 16 or 32 bytes.
The page defines a set of consecutive bytes into memory. A 4-byte page (resp. 8-, 16-, 32-byte
page) is always aligned to 4-byte boundaries (resp. 8-, 16-, 32-byte boundaries) of memory. The
MSB of data address defines the address of the page in memory, the LSB of address define the
address of the data in the page as detailed in Table 20-7.
With page mode memory devices, the first access to one page (tpa) takes longer than the subse-
quent accesses to the page (tsa ) as shown in Figure 20-34. When in page mode, the SMC
enables the user to define different read timings for the first access within one page, and next
accesses within the page.
Table 20-7. Page Address and Data Address within a Page

Page Size Page Address(1) Data Address in the Page(2)
4 bytes A[25:2] A[1:0]
8 bytes A[25:3] A[2:0]
16 bytes A[25:4] A[3:0]
32 bytes A[25:5] A[4:0]
Notes: 1. A denotes the address bus of the memory device

2. For 16-bit devices, the bit 0 of address is ignored. For 32-bit devices, bits [1:0] are ignored.
20.13.1 Protocol and Timings in Page Mode

Figure 20-34 shows the NRD and NCS timings in page mode access.
Figure 20-34. Page Mode Read Protocol (Address MSB and LSB are defined in Table 20-7)
MCK
A[MSB]
A[LSB]
NRD
NCS tpa tsa tsa
D[31:0]
NCS_RD_PULSE NRD_PULSE NRD_PULSE
The NRD and NCS signals are held low during all read transfers, whatever the programmed val-
ues of the setup and hold timings in the User Interface may be. Moreover, the NRD and NCS
214
AT91SAM9G45
timings are identical. The pulse length of the first access to the page is defined with the
NCS_RD_PULSE field of the SMC_PULSE register. The pulse length of subsequent accesses
within the page are defined using the NRD_PULSE parameter.
In page mode, the programming of the read timings is described in Table 20-8:
Table 20-8. Programming of Read Timings in Page Mode

Parameter Value Definition
READ_MODE ‘x’ No impact
NCS_RD_SETUP ‘x’ No impact
NCS_RD_PULSE tpa Access time of first access to the page
NRD_SETUP ‘x’ No impact
NRD_PULSE tsa Access time of subsequent accesses in the page
NRD_CYCLE ‘x’ No impact
The SMC does not check the coherency of timings. It will always apply the NCS_RD_PULSE
timings as page access timing (tpa) and the NRD_PULSE for accesses to the page (tsa), even if
the programmed value for tpa is shorter than the programmed value for tsa.
20.13.2 Byte Access Type in Page Mode

The Byte Access Type configuration remains active in page mode. For 16-bit or 32-bit page
mode devices that require byte selection signals, configure the BAT field of the
SMC_REGISTER to 0 (byte select access type).
20.13.3 Page Mode Restriction

The page mode is not compatible with the use of the NWAIT signal. Using the page mode and
the NWAIT signal may lead to unpredictable behavior.
20.13.4 Sequential and Non-sequential Accesses

If the chip select and the MSB of addresses as defined in Table 20-7 are identical, then the cur-
rent access lies in the same page as the previous one, and no page break occurs.
Using this information, all data within the same page, sequential or not sequential, are accessed
with a minimum access time (tsa). Figure 20-35 illustrates access to an 8-bit memory device in
page mode, with 8-byte pages. Access to D1 causes a page access with a long access time
(tpa). Accesses to D3 and D7, though they are not sequential accesses, only require a short
access time (tsa).
If the MSB of addresses are different, the SMC performs the access of a new page. In the same
way, if the chip select is different from the previous access, a page break occurs. If two sequen-
tial accesses are made to the page mode memory, but separated by an other internal or external
peripheral access, a page break occurs on the second access because the chip select of the
device was deasserted between both accesses.
215
AT91SAM9G45
Figure 20-35. Access to Non-sequential Data within the Same Page
MCK
A[25:3] Page address
A[2], A1, A0 A1 A3 A7
NRD
NCS
D[7:0] D1 D3 D7
NCS_RD_PULSE NRD_PULSE NRD_PULSE
216
AT91SAM9G45
20.14 Programmable IO Delays

The external bus interface consists of a data bus, an address bus and control signals. The simul-
taneous switching outputs on these busses may lead to a peak of current in the internal and
external power supply lines.
In order to reduce the peak of current in such cases, additional propagation delays can be
adjusted independently for pad buffers by means of configuration registers, SMC_DELAY1-8.
The additional programmable delays for each IO range from 0 to 4 ns (Worst Case PVT). The
delay can differ between IOs supporting this feature. Delay can be modified per programming for
each IO. The minimal additional delay that can be programmed on a PAD suppporting this fea-
ture is 1/16 of the maximum programmable delay.
When programming 0x0 in fields “Delay1 to Delay 8”, no delay is added (reset value) and the
propagation delay of the pad buffers is the inherent delay of the pad buffer. When programming
0xF in field “Delay1” the propagation delay of the corresponding pad is maximal.
SMC_DELAY1, SMC_DELAY2 allow to configure delay on D[15:0], SMC_DELAY1[3:0] corre-
sponds to D[0] and SMC_DELAY2[3:0] corresponds to D[8].
SMC_DELAY3, SMC_DELAY4 allow to configure delay on D[31:16], SMC_DELAY3[3:0] corre-
sponds to D[16] and SMC_DELAY4[3:0] corresponds to D[24]. In case of multiplexing through
the PIO controller, refer to the alternate function of D[31:16].
SMC_DELAY5, 6, 7 and 8 allow to configure delay on A[25:0], SMC_DELAY5[3:0] corresponds
to A[0]. In case of multiplexing through the PIO controller, refer to the alternate function of
A[25:0].
Figure 20-36. Programmable IO Delays
SMC D_in[0]
D_out[0] D[0]
Programmable Delay Line
DELAY1
D_in[1]
D_out[1] D[1]
DELAY2
PIO
D_in[n]
D_out[n] D[n]
DELAYx
PIO
A[m] A[m]
DELAYy
217
AT91SAM9G45
20.15 Static Memory Controller (SMC) User Interface

The SMC is programmed using the registers listed in Table 20-9. For each chip select, a set of 4 registers is used to pro-
gram the parameters of the external device connected on it. In Table 20-9, “CS_number” denotes the chip select number.
16 bytes (0x10) are required per chip select.
The user must complete writing the configuration by writing any one of the SMC_MODE registers.

0x10 x CS_number + 0x00 SMC Setup Register SMC_SETUP Read-write 0x01010101
0x10 x CS_number + 0x04 SMC Pulse Register SMC_PULSE Read-write 0x01010101
0x10 x CS_number + 0x08 SMC Cycle Register SMC_CYCLE Read-write 0x00030003
0x10 x CS_number + 0x0C SMC Mode Register SMC_MODE Read-write 0x10001000
0xC0 SMC Delay on I/O SMC_DELAY1 Read-write 0x00000000
0xCC SMC Delay on I/O SMC_DELAY4 Read-write 0x00000000
0xD0 SMC Delay on I/O SMC_DELAY5 Read-write 0x00000000
0xDC SMC Delay on I/O SMC_DELAY8 Read-write 0x00000000
0xEC-0xFC Reserved - - -
218
AT91SAM9G45
20.15.1 SMC Setup Register

Register Name: SMC_SETUP[0..5]
Addresses: 0xFFFFE800 [0], 0xFFFFE810 [1], 0xFFFFE820 [2], 0xFFFFE830 [3], 0xFFFFE840 [4],
0xFFFFE850 [5]
31 30 29 28 27 26 25 24
– – NCS_RD_SETUP
23 22 21 20 19 18 17 16
– – NRD_SETUP
15 14 13 12 11 10 9 8
– – NCS_WR_SETUP
7 6 5 4 3 2 1 0
– – NWE_SETUP
• NWE_SETUP: NWE Setup Length

The NWE signal setup length is defined as:
NWE setup length = (128* NWE_SETUP[5] + NWE_SETUP[4:0]) clock cycles
• NCS_WR_SETUP: NCS Setup Length in WRITE Access

In write access, the NCS signal setup length is defined as:
NCS setup length = (128* NCS_WR_SETUP[5] + NCS_WR_SETUP[4:0]) clock cycles
• NRD_SETUP: NRD Setup Length

The NRD signal setup length is defined in clock cycles as:
NRD setup length = (128* NRD_SETUP[5] + NRD_SETUP[4:0]) clock cycles
• NCS_RD_SETUP: NCS Setup Length in READ Access

In read access, the NCS signal setup length is defined as:
NCS setup length = (128* NCS_RD_SETUP[5] + NCS_RD_SETUP[4:0]) clock cycles
219
AT91SAM9G45
20.15.2 SMC Pulse Register

Register Name: SMC_PULSE[0..5]
0xFFFFE854 [5]
31 30 29 28 27 26 25 24
– NCS_RD_PULSE
23 22 21 20 19 18 17 16
– NRD_PULSE
15 14 13 12 11 10 9 8
– NCS_WR_PULSE
7 6 5 4 3 2 1 0
– NWE_PULSE
• NWE_PULSE: NWE Pulse Length

The NWE signal pulse length is defined as:
NWE pulse length = (256* NWE_PULSE[6] + NWE_PULSE[5:0]) clock cycles
The NWE pulse length must be at least 1 clock cycle.
• NCS_WR_PULSE: NCS Pulse Length in WRITE Access

In write access, the NCS signal pulse length is defined as:
NCS pulse length = (256* NCS_WR_PULSE[6] + NCS_WR_PULSE[5:0]) clock cycles
The NCS pulse length must be at least 1 clock cycle.
• NRD_PULSE: NRD Pulse Length

In standard read access, the NRD signal pulse length is defined in clock cycles as:
NRD pulse length = (256* NRD_PULSE[6] + NRD_PULSE[5:0]) clock cycles
The NRD pulse length must be at least 1 clock cycle.
In page mode read access, the NRD_PULSE parameter defines the duration of the subsequent accesses in the page.
• NCS_RD_PULSE: NCS Pulse Length in READ Access

In standard read access, the NCS signal pulse length is defined as:
NCS pulse length = (256* NCS_RD_PULSE[6] + NCS_RD_PULSE[5:0]) clock cycles
The NCS pulse length must be at least 1 clock cycle.
In page mode read access, the NCS_RD_PULSE parameter defines the duration of the first access to one page.
220
AT91SAM9G45
20.15.3 SMC Cycle Register

Register Name: SMC_CYCLE[0..5]
0xFFFFE858 [5]
31 30 29 28 27 26 25 24
– – – – – – – NRD_CYCLE
23 22 21 20 19 18 17 16
NRD_CYCLE
15 14 13 12 11 10 9 8
– – – – – – – NWE_CYCLE
7 6 5 4 3 2 1 0
NWE_CYCLE
• NWE_CYCLE: Total Write Cycle Length

The total write cycle length is the total duration in clock cycles of the write cycle. It is equal to the sum of the setup, pulse
and hold steps of the NWE and NCS signals. It is defined as:
Write cycle length = (NWE_CYCLE[8:7]*256 + NWE_CYCLE[6:0]) clock cycles
• NRD_CYCLE: Total Read Cycle Length

The total read cycle length is the total duration in clock cycles of the read cycle. It is equal to the sum of the setup, pulse
and hold steps of the NRD and NCS signals. It is defined as:
Read cycle length = (NRD_CYCLE[8:7]*256 + NRD_CYCLE[6:0]) clock cycles
221
AT91SAM9G45
20.15.4 SMC MODE Register

Register Name: SMC_MODE[0..5]
Addresses: 0xFFFFE80C [0], 0xFFFFE81C [1], 0xFFFFE82C [2], 0xFFFFE83C [3], 0xFFFFE84C [4],
0xFFFFE85C [5]
31 30 29 28 27 26 25 24
– – PS – – – PMEN
23 22 21 20 19 18 17 16
– – – TDF_MODE TDF_CYCLES
15 14 13 12 11 10 9 8
– – DBW – – – BAT
7 6 5 4 3 2 1 0
– – EXNW_MODE – – WRITE_MODE READ_MODE
• READ_MODE:
1: The read operation is controlled by the NRD signal.
– If TDF cycles are programmed, the external bus is marked busy after the rising edge of NRD.
– If TDF optimization is enabled (TDF_MODE =1), TDF wait states are inserted after the setup of NRD.
0: The read operation is controlled by the NCS signal.
– If TDF cycles are programmed, the external bus is marked busy after the rising edge of NCS.
– If TDF optimization is enabled (TDF_MODE =1), TDF wait states are inserted after the setup of NCS.
• WRITE_MODE
1: The write operation is controlled by the NWE signal.
– If TDF optimization is enabled (TDF_MODE =1), TDF wait states will be inserted after the setup of NWE.
0: The write operation is controlled by the NCS signal.
– If TDF optimization is enabled (TDF_MODE =1), TDF wait states will be inserted after the setup of NCS.
• EXNW_MODE: NWAIT Mode

The NWAIT signal is used to extend the current read or write signal. It is only taken into account during the pulse phase of
the read and write controlling signal. When the use of NWAIT is enabled, at least one cycle hold duration must be pro-
grammed for the read and write controlling signal.
EXNW_MODE NWAIT Mode

0 0 Disabled
0 1 Reserved
1 0 Frozen Mode
1 1 Ready Mode
• Disabled Mode: The NWAIT input signal is ignored on the corresponding Chip Select.
• Frozen Mode: If asserted, the NWAIT signal freezes the current read or write cycle. After deassertion, the read/write
cycle is resumed from the point where it was stopped.
222
AT91SAM9G45
• Ready Mode: The NWAIT signal indicates the availability of the external device at the end of the pulse of the controlling
read or write signal, to complete the access. If high, the access normally completes. If low, the access is extended until
NWAIT returns high.
• BAT: Byte Access Type

This field is used only if DBW defines a 16- or 32-bit data bus.
• 1: Byte write access type:
– Write operation is controlled using NCS, NWR0, NWR1, NWR2, NWR3.
– Read operation is controlled using NCS and NRD.
• 0: Byte select access type:
– Write operation is controlled using NCS, NWE, NBS0, NBS1, NBS2 and NBS3
– Read operation is controlled using NCS, NRD, NBS0, NBS1, NBS2 and NBS3
• DBW: Data Bus Width

DBW Data Bus Width
0 0 8-bit bus
0 1 16-bit bus
1 0 32-bit bus
1 1 Reserved
• TDF_CYCLES: Data Float Time

This field gives the integer number of clock cycles required by the external device to release the data after the rising edge
of the read controlling signal. The SMC always provide one full cycle of bus turnaround after the TDF_CYCLES period. The
external bus cannot be used by another chip select during TDF_CYCLES + 1 cycles. From 0 up to 15 TDF_CYCLES can
be set.
• TDF_MODE: TDF Optimization

1: TDF optimization is enabled.
– The number of TDF wait states is optimized using the setup period of the next read/write access.
0: TDF optimization is disabled.
– The number of TDF wait states is inserted before the next access begins.
• PMEN: Page Mode Enabled

1: Asynchronous burst read in page mode is applied on the corresponding chip select.
0: Standard read is applied.
• PS: Page Size

If page mode is enabled, this field indicates the size of the page in bytes.
PS Page Size
0 0 4-byte page
0 1 8-byte page
1 0 16-byte page
1 1 32-byte page
223
AT91SAM9G45
20.15.5 SMC DELAY I/O Register

Register Name: SMC_DELAY 1-8
Addresses: 0xFFFFE8C0 [1], 0xFFFFE8C4 [2], 0xFFFFE8C8 [3], 0xFFFFE8CC [4], 0xFFFFE8D0 [5],
0xFFFFE8D4 [6], 0xFFFFE8D8 [7], 0xFFFFE8DC [8]
Reset Value: See Table 20-9
31 30 29 28 27 26 25 24
Delay8 Delay7
23 22 21 20 19 18 17 16
Delay6 Delay5
15 14 13 12 11 10 9 8
Delay4 Delay3
7 6 5 4 3 2 1 0
Delay2 Delay1
• Delay x:
Gives the number of elements in the delay line.
224
AT91SAM9G45
21. DDR/SDR SDRAM Controller (DDRSDRC)
21.1 Description
The DDR/SDR SDRAM Controller (DDRSDRC) is a multiport memory controller. It comprises
four slave AHB interfaces. All simultaneous accesses (four independent AHB ports) are inter-
leaved to maximize memory bandwidth and minimize transaction latency due to SDRAM
protocol.
The DDRSDRC extends the memory capabilities of a chip by providing the interface to an exter-
nal 16-bit or 32-bit SDR-SDRAM device and external 16-bit DDR-SDRAM device. The page size
supports ranges from 2048 to 16384 and the number of columns from 256 to 4096. It supports
byte (8-bit), half-word (16-bit) and word (32-bit) accesses.
The DDRSDRC supports a read or write burst length of 8 locations which frees the command
and address bus to anticipate the next command, thus reducing latency imposed by the SDRAM
protocol and improving the SDRAM bandwidth. Moreover it keeps track of the active row in each
bank, thus maximizing SDRAM performance, e.g., the application may be placed in one bank
and data in the other banks. So as to optimize performance, it is advisable to avoid accessing
different rows in the same bank. The DDRSDRC supports a CAS latency of 2 or 3 and optimizes
the read access depending on the frequency.
The features of self refresh, power-down and deep power-down modes minimize the consump-
tion of the SDRAM device.
The DDRSDRC user interface is compliant with ARM Advanced Peripheral Bus (APB rev2).
225
21.2 DDRSDRC Module Diagram
Figure 21-1. DDRSDRC Module Diagram
DDR-SDR Controller
AHB Slave Interface 0 Input
Stage Power Management
clk/nclk
AHB Slave Interface 1 Input ras,cas,we

Stage cke
Output Memory Controller Addr, DQM

Stage Finite State Machine DDR-SDR
SDRAM Signal Management Devices
Arbiter DQS
AHB Slave Interface 2 Input
Stage
Data
odt
Asynchronous Timing
AHB Slave Interface 3 Input Refresh Management
Stage
Interconnect Matrix
APB
Interface APB
DDRSDRC is partitioned in two blocks (see Figure 21-1):

• An Interconnect-Matrix that manages concurrent accesses on the AHB bus between four
AHB masters and integrates an arbiter.
• A controller that translates AHB requests (Read/Write) in the SDRAM protocol.
226 AT91SAM9G45
AT91SAM9G45

The addresses given are for example purposes only. The real address depends on implementa-
tion in the product.
21.3.1 SDR-SDRAM Initialization

The initialization sequence is generated by software. The SDR-SDRAM devices are initialized
by the following sequence:
1. Program the memory device type into the Memory Device Register (see Section 21.7.8
on page 264).
2. Program the features of the SDR-SDRAM device into the Timing Register (asynchro-
nous timing (trc, tras, etc.)), and into the Configuration Register (number of columns,
rows, banks, cas latency) (see Section 21.7.3 on page 255, Section 21.7.4 on page 258
and Section 21.7.5 on page 260).
3. For low-power SDRAM, temperature-compensated self refresh (TCSR), drive strength
(DS) and partial array self refresh (PASR) must be set in the Low-power Register (see
Section 21.7.7 on page 262).
A minimum pause of 200 μs is provided to precede any signal toggle.
4. A NOP command is issued to the SDR-SDRAM. Program NOP command into Mode
Register, the application must set Mode to 1 in the Mode Register (See Section 21.7.1
on page 253). Perform a write access to any SDR-SDRAM address to acknowledge
this command. Now the clock which drives SDR-SDRAM device is enabled.
5. An all banks precharge command is issued to the SDR-SDRAM. Program all banks
precharge command into Mode Register, the application must set Mode to 2 in the
Mode Register (See Section 21.7.1 on page 253). Perform a write access to any SDR-
SDRAM address to acknowledge this command.
6. Eight auto-refresh (CBR) cycles are provided. Program the auto refresh command
(CBR) into Mode Register, the application must set Mode to 4 in the Mode Register
(see Section 21.7.1 on page 253).Performs a write access to any SDR-SDRAM loca-
tion eight times to acknowledge these commands.
7. A Mode Register set (MRS) cycle is issued to program the parameters of the SDR-
SDRAM devices, in particular CAS latency and burst length. The application must set
Mode to 3 in the Mode Register (see Section 21.7.1 on page 253) and perform a write
access to the SDR-SDRAM to acknowledge this command. The write address must be
chosen so that BA[1:0] are set to 0. For example, with a 16-bit 128 MB SDR-SDRAM
(12 rows, 9 columns, 4 banks) bank address, the SDRAM write access should be done
at the address 0x20000000.
Note: This address is for example purposes only. The real address is dependent on implementation in
the product.
8. For low-power SDR-SDRAM initialization, an Extended Mode Register set (EMRS)
cycle is issued to program the SDR-SDRAM parameters (TCSR, PASR, DS). The appli-
cation must set Mode to 5 in the Mode Register (see Section 21.7.1 on page 253) and
perform a write access to the SDR-SDRAM to acknowledge this command. The write
address must be chosen so that BA[1] is set to 1 and BA[0] is set to 0. For example,
with a 16-bit 128 MB SDRAM, (12 rows, 9 columns, 4 banks) bank address the SDRAM
write access should be done at the address 0x20800000.
9. The application must go into Normal Mode, setting Mode to 0 in the Mode Register (see
Section 21.7.1 on page 253) and perform a write access at any location in the SDRAM
to acknowledge this command.
227
10. Write the refresh rate into the count field in the DDRSDRC Refresh Timer register (see
page 254). (Refresh rate = delay between refresh cycles). The SDR-SDRAM device
requires a refresh every 15.625 μs or 7.81 μs. With a 100 MHz frequency, the refresh
timer count register must to be set with (15.625 /100 MHz) = 1562 i.e. 0x061A or (7.81
/100 MHz) = 781 i.e. 0x030d
After initialization, the SDR-SDRAM device is fully functional.
21.3.2 Low-power DDR1-SDRAM Initialization

The initialization sequence is generated by software. The low-power DDR1-SDRAM devices are
initialized by the following sequence:
on page 264).
2. Program the features of the low-power DDR1-SDRAM device into the Configuration
Register: asynchronous timing (trc, tras, etc.), number of columns, rows, banks, cas
latency. See Section 21.7.3 on page 255, Section 21.7.4 on page 258 and Section
21.7.5 on page 260.
3. Program temperature compensated self refresh (tcr), Partial array self refresh (pasr)
and Drive strength (ds) into the Low-power Register. See Section 21.7.7 on page 262.
4. An NOP command will be issued to the low-power DDR1-SDRAM. Program NOP com-
mand into the Mode Register, the application must set Mode to 1 in the Mode Register
(see Section 21.7.1 on page 253). Perform a write access to any DDR1-SDRAM
address to acknowledge this command. Now clocks which drive DDR1-SDRAM device
are enabled.
A minimum pause of 200 μs will be provided to precede any signal toggle.
5. An all banks precharge command is issued to the low-power DDR1-SDRAM. Program
all banks precharge command into the Mode Register, the application must set Mode to
2 in the Mode Register (See Section 21.7.1 on page 253). Perform a write access to
any low-power DDR1-SDRAM address to acknowledge this command
6. Two auto-refresh (CBR) cycles are provided. Program the auto refresh command
(CBR) into the Mode Register, the application must set Mode to 4 in the Mode Register
(see Section 21.7.1 on page 253). Perform a write access to any low-power DDR1-
SDRAM location twice to acknowledge these commands.
7. An Extended Mode Register set (EMRS) cycle is issued to program the low-power
DDR1-SDRAM parameters (TCSR, PASR, DS). The application must set Mode to 5 in
the Mode Register (see Section 21.7.1 on page 253) and perform a write access to the
SDRAM to acknowledge this command. The write address must be chosen so that
BA[1] is set to 1 BA[0] is set to 0. For example, with a 16-bit 128 MB SDRAM (12 rows,
9 columns, 4 banks) bank address, the low-power DDR1-SDRAM write access should
be done at the address 0x20800000.
the product.
8. A Mode Register set (MRS) cycle is issued to program the parameters of the low-power
DDR1-SDRAM devices, in particular CAS latency, burst length. The application must
set Mode to 3 in the Mode Register (see Section 21.7.1 on page 253) and perform a
write access to the low-power DDR1-SDRAM to acknowledge this command. The write
address must be chosen so that BA[1:0] bits are set to 0. For example, with a 16-bit 128
MB low-power DDR1-SDRAM (12 rows, 9 columns, 4 banks) bank address, the
SDRAM write access should be done at the address 0x20000000
228 AT91SAM9G45
AT91SAM9G45
9. The application must go into Normal Mode, setting Mode to 0 in the Mode Register (see
Section 21.7.1 on page 253) and performing a write access at any location in the low-
power DDR1-SDRAM to acknowledge this command.
10. Perform a write access to any low-power DDR1-SDRAM address.
11. Write the refresh rate into the count field in the DDRSDRC Refresh Timer register (see
page 254). (Refresh rate = delay between refresh cycles). The low-power DDR1-
SDRAM device requires a refresh every 15.625 μs or 7.81 μs. With a 100 MHz fre-
quency, the refresh timer count register must to be set with (15.625 /100 MHz) = 1562
i.e. 0x061A or (7.81 /100 MHz) = 781 i.e. 0x030d
12. After initialization, the low-power DDR1-SDRAM device is fully functional.
21.3.3 DDR2-SDRAM Initialization

The initialization sequence is generated by software. The DDR2-SDRAM devices are initialized
by the following sequence:
on page 264).
2. Program the features of DDR2-SDRAM device into the Timing Register (asynchronous
timing (trc, tras, etc.)), and into the Configuration Register (number of columns, rows,
banks, cas latency and output drive strength) (see Section 21.7.3 on page 255, Section
21.7.4 on page 258 and Section 21.7.5 on page 260).
3. An NOP command is issued to the DDR2-SDRAM. Program the NOP command into
the Mode Register, the application must set Mode to 1 in the Mode Register (see Sec-
tion 21.7.1 on page 253). Perform a write access to any DDR2-SDRAM address to
acknowledge this command. Now clocks which drive DDR2-SDRAM device are
enabled.
A minimum pause of 200 μs is provided to precede any signal toggle.
4. An NOP command is issued to the DDR2-SDRAM. Program the NOP command into
the Mode Register, the application must set Mode to 1 in the Mode Register (see Sec-
tion 21.7.1 on page 253). Perform a write access to any DDR2-SDRAM address to
acknowledge this command. Now CKE is driven high.
5. An all banks precharge command is issued to the DDR2-SDRAM. Program all banks
precharge command into the Mode Register, the application must set Mode to 2 in the
Mode Register (See Section 21.7.1 on page 253). Perform a write access to any DDR2-
SDRAM address to acknowledge this command
6. An Extended Mode Register set (EMRS2) cycle is issued to chose between commer-
cial or high temperature operations. The application must set Mode to 5 in the Mode
Register (see Section 21.7.1 on page 253) and perform a write access to the DDR2-
SDRAM to acknowledge this command. The write address must be chosen so that
BA[1] is set to 1 and BA[0] are set to 0. For example, with a 16-bit 128 MB DDR2-
SDRAM (12 rows, 9 columns, 4 banks) bank address, the DDR2-SDRAM write access
should be done at the address 0x20800000.
the product.
7. An Extended Mode Register set (EMRS3) cycle is issued to set all registers to “0”. The
application must set Mode to 5 in the Mode Register (see Section 21.7.1 on page 253)
and perform a write access to the DDR2-SDRAM to acknowledge this command. The
write address must be chosen so that BA[1] is set to 1 and BA[0] is set to 1. For exam-
ple, with a 16-bit 128 MB DDR2-SDRAM (12 rows, 9 columns, 4 banks) bank address,
the DDR2-SDRAM write access should be done at the address 0x20C00000.
229
8. An Extended Mode Register set (EMRS1) cycle is issued to enable DLL. The applica-
tion must set Mode to 5 in the Mode Register (see Section 21.7.1 on page 253) and
perform a write access to the DDR2-SDRAM to acknowledge this command. The write
address must be chosen so that BA[1] is set to 0 and BA[0] is set to 1. For example,
with a 16-bit 128 MB DDR2-SDRAM (12 rows, 9 columns, 4 banks) bank address, the
DDR2-SDRAM write access should be done at the address 0x20400000.
An additional 200 cycles of clock are required for locking DLL
9. Program DLL field into the Configuration Register (see Section 21.7.3 on page 255) to
high (Enable DLL reset).
10. A Mode Register set (MRS) cycle is issued to reset DLL. The application must set
Mode to 3 in the Mode Register (see Section 21.7.1 on page 253) and perform a write
access to the DDR2-SDRAM to acknowledge this command. The write address must
be chosen so that BA[1:0] bits are set to 0. For example, with a 16-bit 128 MB DDR2-
SDRAM (12 rows, 9 columns, 4 banks) bank address, the SDRAM write access should
be done at the address 0x20000000.
11. An all banks precharge command is issued to the DDR2-SDRAM. Program all banks
precharge command into the Mode Register, the application must set Mode to 2 in the
Mode Register (See Section 21.7.1 on page 253). Perform a write access to any DDR2-
SDRAM address to acknowledge this command
12. Two auto-refresh (CBR) cycles are provided. Program the auto refresh command
(CBR) into the Mode Register, the application must set Mode to 4 in the Mode Register
(see Section 21.7.1 on page 253). Performs a write access to any DDR2-SDRAM loca-
tion twice to acknowledge these commands.
13. Program DLL field into the Configuration Register (see Section 21.7.3 on page 255) to
low (Disable DLL reset).
14. A Mode Register set (MRS) cycle is issued to program the parameters of the DDR2-
SDRAM devices, in particular CAS latency, burst length and to disable DLL reset. The
write address must be chosen so that BA[1:0] are set to 0. For example, with a 16-bit
128 MB SDRAM (12 rows, 9 columns, 4 banks) bank address, the SDRAM write
access should be done at the address 0x20000000
15. Program OCD field into the Configuration Register (see Section 21.7.3 on page 255) to
high (OCD calibration default).
16. An Extended Mode Register set (EMRS1) cycle is issued to OCD default value. The
the DDR2-SDRAM write access should be done at the address 0x20400000.
17. Program OCD field into the Configuration Register (see Section 21.7.3 on page 255) to
low (OCD calibration mode exit).
18. An Extended Mode Register set (EMRS1) cycle is issued to enable OCD exit. The
the DDR2-SDRAM write access should be done at the address 0x20400000.
230 AT91SAM9G45
AT91SAM9G45
19. A mode Normal command is provided. Program the Normal mode into Mode Register
(see Section 21.7.1 on page 253). Perform a write access to any DDR2-SDRAM
address to acknowledge this command.
20. Perform a write access to any DDR2-SDRAM address.
21. Write the refresh rate into the count field in the Refresh Timer register (see page 254).
(Refresh rate = delay between refresh cycles). The DDR2-SDRAM device requires a
refresh every 15.625 μs or 7.81 μs. With a 133 MHz frequency, the refresh timer count
register must to be set with (15.625 /133 MHz) = 1175 i.e. 0x0497 or (7.81 /133 MHz) =
587 i.e. 0x024B
After initialization, the DDR2-SDRAM devices are fully functional.
231
21.4.1 SDRAM Controller Write Cycle

The DDRSDRC allows burst access or single access in normal mode (mode = 000). Whatever
the access type, the DDRSDRC keeps track of the active row in each bank, thus maximizing
performance.
The SDRAM device is programmed with a burst length equal to 8. This determines the length of
a sequential data input by the write command that is set to 8. The latency from write command to
data input is fixed to 1 in the case of DDR-SDRAM devices. In the case of SDR-SDRAM
devices, there is no latency from write command to data input.
To initiate a single access, the DDRSDRC checks if the page access is already open. If
row/bank addresses match with the previous row/bank addresses, the controller generates a
write command. If the bank addresses are not identical or if bank addresses are identical but the
row addresses are not identical, the controller generates a precharge command, activates the
new row and initiates a write command. To comply with SDRAM timing parameters, additional
clock cycles are inserted between precharge/active (t RP) commands and active/write (t RCD)
command. As the burst length is fixed to 8, in the case of single access, it has to stop the burst,
otherwise seven invalid values may be written. In the case of SDR-SDRAM devices, a Burst
Stop command is generated to interrupt the write operation. In the case of DDR-SDRAM
devices, Burst Stop command is not supported for the burst write operation. In order to then
interrupt the write operation, Dm must be set to 1 to mask invalid data (see Figure 21-2 on page
233 and Figure 21-5 on page 234) and DQS must continue to toggle.
To initiate a burst access, the DDRSDRC uses the transfer type signal provided by the master
requesting the access. If the next access is a sequential write access, writing to the SDRAM
device is carried out. If the next access is a write non-sequential access, then an automatic
access break is inserted, the DDRSDRC generates a precharge command, activates the new
row and initiates a write command. To comply with SDRAM timing parameters, additional clock
cycles are inserted between precharge/active (tRP) commands and active/write (tRCD)
commands.
For a definition of timing parameters, refer to Section 21.7.4 “DDRSDRC Timing 0 Parameter
Register” on page 258.
Write accesses to the SDRAM devices are burst oriented and the burst length is programmed to
8. It determines the maximum number of column locations that can be accessed for a given write
command. When the write command is issued, 8 columns are selected. All accesses for that
burst take place within these eight columns, thus the burst wraps within these 8 columns if a
boundary is reached. These 8 columns are selected by addr[13:3]. addr[2:0] is used to select the
starting location within the block.
In the case of incrementing burst (INCR/INCR4/INCR8/INCR16), the addresses can cross the
16-byte boundary of the SDRAM device. For example, in the case of DDR-SDRAM devices,
when a transfer (INCR4) starts at address 0x0C, the next access is 0x10, but since the burst
length is programmed to 8, the next access is at 0x00. Since the boundary is reached, the burst
is wrapping. The DDRSDRC takes this feature of the SDRAM device into account. In the case of
transfer starting at address 0x04/0x08/0x0C (DDR-SDRAM devices) or starting at address
0x10/0x14/0x18/0x1C, two write commands are issued to avoid to wrap when the boundary is
reached. The last write command is subject to DM input logic level. If DM is registered high, the
corresponding data input is ignored and write access is not done. This avoids additional writing
being done.
232 AT91SAM9G45
AT91SAM9G45
Figure 21-2. Single Write Access, Row Closed, Low-power DDR1-SDRAM Device
SDCLK
A[12:0] Row a col a
COMMAND NOP PRCHG NOP ACT NOP WRITE NOP
BA[1:0] 00
DQS[1:0]
DM[1:0] 3 0 3
D[15:0] Da Db
Trp = 2 Trcd = 2
Figure 21-3. Single Write Access, Row Closed, DDR2-SDRAM Device
SDCLK
A[12:0] Row a col a
BA[1:0] 00
DQS[1:0]
DM[1:0] 3 0 3
D[15:0] Da Db
Trp = 2 Trcd = 2
233
Figure 21-4. Single Write Access, Row Closed, SDR-SDRAM Device
SDCLK
A[12:0] Row a Col a
COMMAND NOP PRCHG NOP ACT NOP WRITE BST NOP
BA[1:0] 00
DM[1:0] 3 0 3
D[31:0] DaDb
Trp = 2 Trcd = 2
Figure 21-5. Burst Write Access, Row Closed, Low-power DDR1-SDRAM Device
SDCLK
A[12:0] Row a col a
BA[1:0] 0
DQS[1:0]
DM[1:0] 3 0 3
D [15:0] Da Db Dc Dd De Df Dg Dh
Trp = 2 Trcd = 2
234 AT91SAM9G45
AT91SAM9G45
Figure 21-6. Burst Write Access, Row Closed, DDR2-SDRAM Device

SDCLK
A[12:0] Row a col a
BA[1:0] 0
DQS[1:0]
DM[1:0] 3 0 3
D [15:0] Da Db Dc Dd De Df Dg Dh
Trp = 2 Trcd = 2
Figure 21-7. Burst Write Access, Row Closed, SDR-SDRAM Device
SDCLK
A[12:0] Row a Col a
COMMAND NOP PRCHG NOP ACT NOP WRITE NOP BST NOP
BA[1:0] 0
DM[3:0] F 0 F
D[31:0] Da Db Dc Dd De Df Dg Dhs
Trp Trcd
A write command can be followed by a read command. To avoid breaking the current write
burst, Twtr/Twrd (bl/2 + 2 = 6 cycles) should be met. See Figure 21-8 on page 236.
235
Figure 21-8. Write Command Followed By a Read Command without Burst Write Interrupt, Low-power DDR1-SDRAM
Device
SDCLK
A[12:0] col a col a
COMMAND NOP WRITE NOP READ BST NOP
BA[1:0] 0
DQS[1:0]
DM[1:0] 3 0 3
D[15:0] Da Db Dc Dd De Df Dg Dh Da Db
Twrd = BL/2 +2 = 8/2 +2 = 6
Twr = 1
In the case of a single write access, write operation should be interrupted by a read access but
DM must be input 1 cycle prior to the read command to avoid writing invalid data. See Figure 21-
9 on page 236.
Figure 21-9. Single Write Access Followed By A Read Access Low-power DDR1-SDRAM Devices
SDCLK
A[12:0] Row a col a
COMMAND NOP PRCHG NOP ACT NOP WRITE NOP READ BST NOP
BA[1:0] 0
DQS[1:0]
DM[1:0] 3 0 3
D[15:0] Da Db Da Db
Data masked
236 AT91SAM9G45
AT91SAM9G45
Figure 21-10. SINGLE Write Access Followed By A Read Access, DDR2 -SDRAM Device
SDCLK
A[12:0] Row a col a
COMMAND NOP PRCHG NOP ACT NOP WRITE NOP READ NOP
BA[1:0] 0
DQS[1:0]
DM[1:0] 3 0 3
D[15:0] Da Db Da Db
Data masked
twtr
21.4.2 SDRAM Controller Read Cycle

The DDRSDRC allows burst access or single access in normal mode (mode =000). Whatever
access type, the DDRSDRC keeps track of the active row in each bank, thus maximizing perfor-
mance of the DDRSDRC.
The SDRAM devices are programmed with a burst length equal to 8 which determines the length
of a sequential data output by the read command that is set to 8. The latency from read com-
mand to data output is equal to 2 or 3. This value is programmed during the initialization phase
(see Section 21.3.1 “SDR-SDRAM Initialization” on page 227).
To initiate a single access, the DDRSDRC checks if the page access is already open. If
row/bank addresses match with the previous row/bank addresses, the controller generates a
read command. If the bank addresses are not identical or if bank addresses are identical but the
row addresses are not identical, the controller generates a precharge command, activates the
new row and initiates a read command. To comply with SDRAM timing parameters, additional
clock cycles are inserted between precharge/active (Trp) commands and active/read (Trcd)
command. After a read command, additional wait states are generated to comply with cas
latency. The DDRSDRC supports a cas latency of two, two and half, and three (2 or 3 clocks
delay). As the burst length is fixed to 8, in the case of single access or burst access inferior to 8
data requests, it has to stop the burst otherwise seven or X values could be read. Burst Stop
Command (BST) is used to stop output during a burst read.
To initiate a burst access, the DDRSDRC checks the transfer type signal. If the next accesses
are sequential read accesses, reading to the SDRAM device is carried out. If the next access is
a read non-sequential access, then an automatic page break can be inserted. If the bank
addresses are not identical or if bank addresses are identical but the row addresses are not
identical, the controller generates a precharge command, activates the new row and initiates a
read command. In the case where the page access is already open, a read command is
generated.
237
To comply with SDRAM timing parameters, additional clock cycles are inserted between pre-
charge/active (Trp) commands and active/read (Trcd) commands. The DDRSDRC supports a
cas latency of two, two and half, and three (2 or 3 clocks delay). During this delay, the controller
uses internal signals to anticipate the next access and improve the performance of the control-
ler. Depending on the latency(2/3), the DDRSDRC anticipates 2 or 3 read accesses. In the case
of burst of specified length, accesses are not anticipated, but if the burst is broken (border, busy
mode, etc.), the next access is treated as an incrementing burst of unspecified length, and in
function of the latency(2/3), the DDRSDRC anticipates 2 or 3 read accesses.
For a definition of timing parameters, refer to Section 21.7.3 “DDRSDRC Configuration Register”
on page 255.
Read accesses to the SDRAM are burst oriented and the burst length is programmed to 8. It
determines the maximum number of column locations that can be accessed for a given read
command. When the read command is issued, 8 columns are selected. All accesses for that
burst take place within these eight columns, meaning that the burst wraps within these 8 col-
umns if the boundary is reached. These 8 columns are selected by addr[13:3]; addr[2:0] is used
to select the starting location within the block.
In the case of incrementing burst (INCR/INCR4/INCR8/INCR16), the addresses can cross the
16-byte boundary of the SDRAM device. For example, when a transfer (INCR4) starts at
address 0x0C, the next access is 0x10, but since the burst length is programmed to 8, the next
access is 0x00. Since the boundary is reached, the burst wraps. The DDRSDRC takes into
account this feature of the SDRAM device. In the case of DDR-SDRAM devices, transfers start
at address 0x04/0x08/0x0C. In the case of SDR-SDRAM devices, transfers start at address
0x14/0x18/0x1C. Two read commands are issued to avoid wrapping when the boundary is
reached. The last read command may generate additional reading (1 read cmd = 4 DDR words
or 1 read cmd = 8 SDR words).
To avoid additional reading, it is possible to use the burst stop command to truncate the read
burst and to decrease power consumption.
238 AT91SAM9G45
AT91SAM9G45
Figure 21-11. Single Read Access, Row Close, Latency = 2,Low-power DDR1-SDRAM Device
SDCLK
A[12:0] Row a Col a
COMMAND NOP PRCHG NOP ACT NOP READ BST NOP
BA[1:0] 0
DQS[1]
DQS[0]
DM[1:0] 3
D[15:0] Da Db
Trp Trcd Latency = 2
Figure 21-12. Single Read Access, Row Close, Latency = 3, DDR2-SDRAM Device
SDCLK
A[12:0] Row a Col a
COMMAND NOP PRCHG NOP ACT NOP READ
BA[1:0] 0
DQS[1]
DQS[0]
DM[1:0] 3
D[15:0] Da Db
239
Figure 21-13. Single Read Access, Row Close, Latency = 2, SDR-SDRAM Device
SDCLK
A[12:0] Row a col a
BA[1:0] 0
DM[3:0] 3
D[31:0] DaDb
Figure 21-14. Burst Read Access, Latency = 2, Low-power DDR1-SDRAM Devices
SDCLK
A[12:0] Col a
COMMAND NOP READ NOP
BA[1:0] 0
DQS[1:0]
DM[1:0] 3
D[15:0] Da Db Dc Dd De Df Dg Dh
Latency = 2
240 AT91SAM9G45
AT91SAM9G45
Figure 21-15. Burst Read Access, Latency = 3, DDR2-SDRAM Devices
SDCLK
A[12:0] Col a
COMMAND NOP READ NOP
BA[1:0] 0
DQS[1:0]
DM[1:0] 3
D[15:0] Da Db Dc Dd De Df Dg Dh
Latency = 3
Figure 21-16. Burst Read Access, Latency = 2, SDR-SDRAM Devices
SDCLK
A[12:0] col a
COMMAND NOP READ NOP BST NOP
BA[1:0] 0
DQS[1:0]
DM[3:0] F
D[31:0] DaDb DcDd DeDf Dg Dh

Latency = 2
241
21.4.3 Refresh (Auto-refresh Command)
An auto-refresh command is used to refresh the DDRSDRC. Refresh addresses are generated
internally by the SDRAM device and incremented after each auto-refresh automatically. The
DDRSDRC generates these auto-refresh commands periodically. A timer is loaded with the
value in the register DDRSDRC_TR that indicates the number of clock cycles between refresh
cycles. When the DDRSDRC initiates a refresh of an SDRAM device, internal memory accesses
are not delayed. However, if the CPU tries to access the SDRAM device, the slave indicates that
the device is busy. A request of refresh does not interrupt a burst transfer in progress.
21.4.4.1 Self Refresh Mode

This mode is activated by setting low-power command bits [LPCB] to ‘01’ in the
DDRSDRC_LPR Register
Self refresh mode is used to reduce power consumption, i.e., when no access to the SDRAM
device is possible. In this case, power consumption is very low. In self refresh mode, the
SDRAM device retains data without external clocking and provides its own internal clocking,
thus performing its own auto-refresh cycles. All the inputs to the SDRAM device become don’t
care except CKE, which remains low. As soon as the SDRAM device is selected, the DDRSDRC
provides a sequence of commands and exits self refresh mode.
The DDRSDRC re-enables self refresh mode as soon as the SDRAM device is not selected. It is
possible to define when self refresh mode will be enabled by setting the register LPR (see Sec-
tion 21.7.7 “DDRSDRC Low-power Register” on page 262), timeout command bit:
• 00 = Self refresh mode is enabled as soon as the SDRAM device is not selected
• 01 = Self refresh mode is enabled 64 clock cycles after completion of the last access
• 10 = Self refresh mode is enabled 128 clock cycles after completion of the last access
As soon as the SDRAM device is no longer selected, PRECHARGE ALL BANKS command is
generated followed by a SELF-REFREFSH command. If beetwen these two commands an
SDRAM access is detected, SELF-REFREFSH command will be replaced by an AUTO-
REFRESH command . According to the application, more AUTO-REFRESH commands will be
performed when the self refresh mode is enabled during the application.
This controller also interfaces low-power SDRAM. These devices add a new feature: A single
quarter, one half quarter or all banks of the SDRAM array can be enabled in self refresh mode.
Disabled banks will be not refreshed in self refresh mode. This feature permits to reduce the self
refresh current. The extended mode register controls this new feature, it includes Temperature
Compensated Self Refresh (TSCR), Partial Array Self Refresh (PASR) parameters and drive
strength (DS). These parameters are set during the initialization phase. After initialization, as
soon as PASR/DS/TCSR fields are modified and self-refresh mode is activated, the Extended
Mode Register is accessed automatically and PASR/DS/TCSR bits are updated before any
entry into self refresh mode. The automatic update feature is not supported when the DDRS-
DRC shares an external bus with another controller.
The low-power SDR-SDRAM must remain in self refresh mode for a minimum period of TRAS
periods and may remain in self refresh mode for an indefinite period. (See Figure 21-17 on page
243)
The low-power DDR-SDRAM must remain in self refresh mode for a minimum of TRFC periods
and may remain in self refresh mode for an indefinite period.
242 AT91SAM9G45
AT91SAM9G45
The DDR2-SDRAM must remain in self refresh mode for a minimum of TCKE periods and may
remain in self refresh mode for an indefinite period.
Figure 21-17. Self Refresh Mode Entry, Timeout = 0

SDCLK
A[12:0]
COMMAND NOP READ BST NOP PRCHG NOP ARFSH NOP
CKE
BA[1:0] 0
DQS[0:1]
DM[1:0] 3
D[15:0] Da Db
Enter Self refresh
Mode
Trp
Figure 21-18. Self Refresh Mode Entry, Timeout = 1 or 2

SDCLK
A[12:0]
COMMAND NOP READ BST NOP PRCHG NOP ARFSH NOP
CKE
BA[1:0] 0
DQS[1:0]
DM[1:0] 3
D[15:0] Da Db
64 or 128 Trp Enter Self refresh

wait states Mode
243
Figure 21-19. Self Refresh Mode Exit
SDCLK
A[12:0]
COMMAND NOP VALID NOP
CKE
BA[1:0] 0
DQS[1:0]
DM[1:0] 3
D[15:0] DaDb
Exit Self Refresh mode

clock must be stable TXNRD/TXSRD (DDR device)
before exiting self refresh mode TXSR (Low-power DDR device)
TXSR (Low-power SDR, SDR-SDRAM device)
21.4.4.2 Power-down Mode

This mode is activated by setting the low-power command bits [LPCB] to ‘10’.
Power-down mode is used when no access to the SDRAM device is possible. In this mode,
power consumption is greater than in self refresh mode. This state is similar to normal mode (No
low-power mode/No self refresh mode), but the CKE pin is low and the input and output buffers
are deactivated as soon the SDRAM device is no longer accessible. In contrast to self refresh
mode, the SDRAM device cannot remain in low-power mode longer than the refresh period (64
ms). As no auto-refresh operations are performed in this mode, the DDRSDRC carries out the
refresh operation. In order to exit low-power mode, a NOP command is required in the case of
Low-power SDR-SDRAM and SDR-SDRAM devices. In the case of Low-power DDR-SDRAM
devices, the controller generates a NOP command during a delay of at least TXP. In addition,
Low-power DDR-SDRAM and DDR2-SDRAM must remain in power-down mode for a minimum
period of TCKE periods.
The exit procedure is faster than in self refresh mode. See Figure 21-20 on page 245. The
DDRSDRC returns to power-down mode as soon as the SDRAM device is not selected. It is
possible to define when power-down mode is enabled by setting the register LPR, timeout com-
mand bit.
• 00 = Power-down mode is enabled as soon as the SDRAM device is not selected
• 01 = Power-down mode is enabled 64 clock cycles after completion of the last access
• 10 = Power-down mode is enabled 128 clock cycles after completion of the last access
244 AT91SAM9G45
AT91SAM9G45
Figure 21-20. Power-down Entry/Exit, Timeout = 0
SDCLK
A[12:0]
COMMAND READ BST NOP READ
CKE
BA[1:0] 0
DQS[1:0]
DM[1:0] 3
D[15:0] Da Db
Entry power down mode Exit power down mode
21.4.4.3 Deep Power-down Mode

The deep power-down mode is a new feature of the Low-power SDRAM. When this mode is
activated, all internal voltage generators inside the device are stopped and all data is lost.
This mode is activated by setting the low-power command bits [LPCB] to ‘11’. When this mode is
enabled, the DDRSDRC leaves normal mode (mode == 000) and the controller is frozen. To exit
deep power-down mode, the low-power bits (LPCB) must be set to “00”, an initialization
sequence must be generated by software. See Section 21.3.2 “Low-power DDR1-SDRAM Ini-
tialization” on page 228.
Figure 21-21. Deep Power-down Mode Entry

SDCLK
A[12:0]
COMMAND NOP READ BST NOP PRCHG NOP DEEPOWER NOP
CKE
BA[1:0] 0
DQS[1:0]
DM[1:0] 3
D[15:0] Da Db
Enter Deep
Trp Power-down
Mode
245
21.4.4.4 Reset Mode
The reset mode is a feature of the DDR2-SDRAM. This mode is activated by setting the low-
power command bits (LPCB) to 11 and the clock frozen command bit (CLK_FR) to 1.
When this mode is enabled, the DDRSDRC leaves normal mode (mode == 000) and the control-
ler is frozen. Before enabling this mode, the end user must assume there is not an access in
progress.
To exit reset mode, the low-power command bits (LPCB) must be set to “00”, clock frozen com-
mand bit (CLK_FR) set to 0 and an initialization sequence must be generated by software. See,
Section 21.3.3 “DDR2-SDRAM Initialization” on page 229.
21.4.5 Multi-port Functionality

The SDRAM protocol imposes a check of timings prior to performing a read or a write access,
thus decreasing the performance of systems. An access to SDRAM is performed if banks and
rows are open (or active). To activate a row in a particular bank, it has to de-active the last open
row and open the new row. Two SDRAM commands must be performed to open a bank: Pre-
charge and Active command with respect to Trp timing. Before performing a read or write
command, Trcd timing must checked.
This operation represents a significative loss. (see Figure 21-22)
Figure 21-22. Trp and Trcd Timings
SDCLK
A[12:0]
BA[1:0] 0
DQS[1:0]
DM1:0] 3
D[15:0] Da Db
Trp Trcd Latency =2

4 cycles before performing a read command
The multi-port controller has been designed to mask these timings and thus improve the band-
width of the system.
DDRSDRC is a multi-port controller since four masters can simultaneously reach the controller.
This feature improves the bandwidth of the system because it can detect four requests on the
AHB slave inputs and thus anticipate the commands that follow, PRECHARGE and ACTIVE
commands in bank X during current access in bank Y. This allows Trp and Trcd timings to be
masked (see Figure 21-23). In the best case, all accesses are done as if the banks and rows
were already open. The best condition is met when the four masters work in different banks. In
246 AT91SAM9G45
AT91SAM9G45
the case of four simultaneous read accesses, when the four banks and associated rows are
open, the controller reads with a continuous flow and masks the cas latency for each different
access. To allow a continuous flow, the read command must be set at 2 or 3 cycles (cas latency)
before the end of current access. This requires that the scheme of arbitration changes since the
round-robin arbitration cannot be respected. If the controller anticipates a read access, and thus
before the end of current access a master with a high priority arises, then this master will not
serviced.
The arbitration mechanism reduces latency when conflicts occur, i.e., when two or more masters
try to access the SDRAM device at the same time.
The arbitration type is round-robin arbitration. This algorithm dispatches the requests from differ-
ent masters to the SDRAM device in a round-robin manner. If two or more master requests arise
at the same time, the master with the lowest number is serviced first, then the others are ser-
viced in a round-robin manner. To avoid burst breaking and to provide the maximum throughput
for the SDRAM device, arbitration may only take place during the following cycles:
1. Idle cycles: When no master is connected to the SDRAM device.
2. Single cycles: When a slave is currently doing a single access.
3. End of Burst cycles: When the current cycle is the last cycle of a burst transfer. For
bursts of defined length, predicted end of burst matches the size of the transfer. For
bursts of undefined length, predicted end of burst is generated at the end of each four
beat boundary inside the INCR transfer.
4. Anticipated Access: When an anticipate read access is done while current access is
not complete, the arbitration scheme can be changed if the anticipated access is not
the next access serviced by the arbitration scheme.
Figure 21-23. Anticipate Precharge/Active Command in Bank 2 during Read Access in Bank 1
SDClK
A[12:0]
COMMAND NOP READ PRECH NOP ACT READ NOP
BA[1:0] 0 1 2 1
DQS[1:0]
DM1:0] 3
D[15:0] Da Db Dc Dd De Df Dg Dh Di Dj Dk Dl
Trp
Anticipate command, Precharge/Active Bank 2
Read access in Bank 1
247
21.4.6 Write Protect Registers
To prevent any single software error that may corrupt DDRSDRC behavior, the registers listed
below can be write-protected by setting the WPEN bit in the DDRSDRC Write Protect Mode
Register (DDRSDRC_WPMR).
If a write access in a write-protected register is detected, then the WPVS flag in the DDRSDRC
Write Protect Status Register (DDRSDRC_WPSR) is set and the field WPVSRC indicates in
which register the write access has been attempted.
The WPVS flag is automatically reset after reading the DDRSDRC Write Protect Status Register
(DDRSDRC_WPSR).
List of the write-protectable registers:
• “DDRSDRC Mode Register” on page 253
• “DDRSDRC Refresh Timer Register” on page 254
• “DDRSDRC Configuration Register” on page 255
• “DDRSDRC Timing 0 Parameter Register” on page 258
• “DDRSDRC Memory Device Register” on page 264
• “DDRSDRC High Speed Register” on page 266
248 AT91SAM9G45
AT91SAM9G45
21.5 Software Interface/SDRAM Organization, Address Mapping

The SDRAM address space is organized into banks, rows and columns. The DDRSDRC maps
different memory types depending on the values set in the DDRSDRC Configuration Register.
See Section 21.7.3 “DDRSDRC Configuration Register” on page 255. The following figures illus-
trate the relation between CPU addresses and columns, rows and banks addresses for 16-bit
memory data bus widths and 32-bit memory data bus widths.
The DDRSDRC supports address mapping in linear mode.
Linear mode is a method for address mapping where banks alternate at each last SDRAM page
of current bank.
The DDRSDRC makes the SDRAM devices access protocol transparent to the user. Table 21-1
to Table 21-7 illustrate the SDRAM device memory mapping seen by the user in correlation with
the device structure. Various configurations are illustrated.
21.5.1 SDRAM Address Mapping for 16-bit Memory Data Bus Width(1)
Table 21-1. Linear Mapping for SDRAM Configuration, 2K Rows, 512/1024/2048/4096 Columns
CPU Address Line
27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Bk[1:0] Row[10:0] Column[8:0] M0

Bk[1:0] Row[10:0] Column[9:0] M0
Bk[1:0] Row[10:0] Column[10:0] M0
Bk[1:0] Row[10:0] Column[11:0] M0
Table 21-2. Linear Mapping for SDRAM Configuration: 4K Rows, 512/1024/2048/4096 Columns
CPU Address Line
27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Bk[1:0] Row[11:0] Column[8:0] M0

Bk[1:0] Row[11:0] Column[9:0] M0
Bk[1:0] Row[11:0] Column[10:0] M0
Bk[1:0] Row[11:0] Column[11:0] M0
Table 21-3. Linear Mapping for SDRAM Configuration: 8K Rows, 512/1024/2048/4096 Columns
CPU Address Line
27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Bk[1:0] Row[12:0] Column[8:0] M0

Bk[1:0] Row[12:0] Column[9:0] M0
Bk[1:0] Row[12:0] Column[10:0] M0
Bk[1:0] Row[12:0] Column[11:0] M0
249
Table 21-4. Linear Mapping for SDRAM Configuration: 16K Rows, 512/1024/2048 Columns
CPU Address Line
27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Bk[1:0] Row[13:0] Column[8:0] M0

Bk[1:0] Row[13:0] Column[9:0] M0
Bk[1:0] Row[13:0] Column[10:0] M0
Note: 1. SDR-SDRAM devices with eight columns in 16-bit mode are not supported.
21.5.2 SDR-SDRAM Address Mapping for 32-bit Memory Data Bus Width
Table 21-5. SDR-SDRAM Configuration Mapping: 2K Rows, 256/512/1024/2048 Columns

CPU Address Line
27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Bk[1:0] Row[10:0] Column[7:0] M[1:0]
Bk[1:0] Row[10:0] Column[8:0] M[1:0]
Bk[1:0] Row[10:0] Column[9:0] M[1:0]
Bk[1:0] Row[10:0] Column[10:0] M[1:0]

CPU Address Line
27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Bk[1:0] Row[11:0] Column[7:0] M[1:0]
Bk[1:0] Row[11:0] Column[8:0] M[1:0]
Bk[1:0] Row[11:0] Column[9:0] M[1:0]
Bk[1:0] Row[11:0] Column[10:0] M[1:0]

CPU Address Line
27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Bk[1:0] Row[12:0] Column[7:0] M[1:0]
Bk[1:0] Row[12:0] Column[8:0] M[1:0]
Bk[1:0] Row[12:0] Column[9:0] M[1:0]
Bk[1:0] Row[12:0] Column[10:0] M[1:0]
Notes: 1. M[1:0] is the byte address inside a 32-bit word.

2. Bk[1] = BA1, Bk[0] = BA0
250 AT91SAM9G45
AT91SAM9G45
21.6 Programmable IO Delays

The external bus interface consists of a data bus, an address bus and control signals. The simul-
taneous switching outputs on these busses may lead to a peak of current in the internal and
external power supply lines.
In order to reduce the peak of current in such cases, additional propagation delays can be
adjusted independently for pad buffers by means of configuration registers,
DDRSDRC_DELAY1-8.
The additional programmable delays for each IO range from 0 to 4 ns (Worst Case PVT). The
delay can differ between IOs supporting this feature. Delay can be modified per programming for
each IO. The minimal additional delay that can be programmed on a PAD supporting this feature
is 1/16 of the maximum programmable delay.
When programming 0x0 in fields “Delay1 to Delay 8”, no delay is added (reset value) and the
propagation delay of the pad buffers is the inherent delay of the pad buffer. When programming
0xF in field “Delay1” the propagation delay of the corresponding pad is maximal.
DDRSDRC_DELAY1, DDRSDRC_DELAY2 allow to configure delay on D[15:0],
DDRSDRC_DELAY1[3:0] corresponds to D[0] and DDRSDRC_DELAY2[3:0] corresponds to
D[8].
DDRSDRC_DELAY3, DDRSDRC_DELAY4 allow to configure delay on A13:0],
DDRSDRC_DELAY3[3:0] corresponds to A[0] and DDRSDRC_DELAY4[3:0] corresponds to
A[8].
Figure 21-24. Programmable IO Delays
SMC D_in[0]
D_out[0] D[0]
DELAY1
D_in[1]
D_out[1] D[1]
DELAY2
D_in[n]
D_out[n] D[n]
DELAYx
A[m] A[m]
DELAYy
251
21.7 DDR-SDRAM Controller (DDRSDRC) User Interface
The User Interface is connected to the APB bus.
The DDRSDRC is programmed using the registers listed in Table 21-8.

0x00 DDRSDRC Mode Register DDRSDRC_MR Read-write 0x00000000
0x04 DDRSDRC Refresh Timer Register DDRSDRC_RTR Read-write 0x00000000
0x08 DDRSDRC Configuration Register DDRSDRC_CR Read-write 0x7024
0x0C DDRSDRC Timing0 Register DDRSDRC_T0PR Read-write 0x20227225
0x10 DDRSDRC Timing1 Register DDRSDRC_T1PR Read-write 0x3c80808
0x14 DDRSDRC Timing2 Register DDRSDRC_T2PR Read-write 0x2062
0x1C DDRSDRC Low-power Register DDRSDRC_LPR Read-write 0x10000
0x20 DDRSDRC Memory Device Register DDRSDRC_MD Read-write 0x10
0x24 DDRSDRC DLL Information Register DDRSDRC_DLL Read-only 0x00000001
0x2C DDRSDRC High Speed Register DDRSDRC_HS Read-write 0x4
0x30 DDRSDRC Delay on I/O DDRSDRC_DELAY1 Read-write 0x00000000
0x3C DDRSDRC Delay on I/O DDRSDRC_DELAY4 Read-write 0x00000000
0x44-0x4C Reserved - - -
0x50 DDRSDRC Drive Strength DDRSDRC_DRIVESTRENGTH Read-write 0xFFFFFFFF
0x58-0xE0 Reserved – – –
0xE4 DDRSDRC Write Protect Mode Register DDRSDRC_WPMR Read-write 0x00000000
0xE8 DDRSDRC Write Protect Status Register DDRSDRC_WPSR Read-only 0x00000000
252 AT91SAM9G45
AT91SAM9G45
21.7.1 DDRSDRC Mode Register

Name: DDRSDRC_MR
Access: Read-write
Reset: See Table 21-8
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – MODE
This register can only be written if the bit WPEN is cleared in “DDRSDRC Write Protect Mode Register” on page 269.
• MODE: DDRSDRC Command Mode

This field defines the command issued by the DDRSDRC when the SDRAM device is accessed. This register is used to ini-
tialize the SDRAM device and to activate deep power-down mode.
MODE Description
Normal Mode. Any access to the DDRSDRC will be decoded normally. To activate this mode, command must be followed
000
by a write to the SDRAM.
The DDRSDRC issues a NOP command when the SDRAM device is accessed regardless of the cycle. To activate this
001
mode, command must be followed by a write to the SDRAM.
The DDRSDRC issues an “All Banks Precharge” command when the SDRAM device is accessed regardless of the cycle.
010
To activate this mode, command must be followed by a write to the SDRAM.
The DDRSDRC issues a “Load Mode Register” command when the SDRAM device is accessed regardless of the cycle.
011
To activate this mode, command must be followed by a write to the SDRAM.
The DDRSDRC issues an “Auto-Refresh” Command when the SDRAM device is accessed regardless of the cycle.
100 Previously, an “All Banks Precharge” command must be issued. To activate this mode, command must be followed by a
write to the SDRAM.
The DDRSDRC issues an “Extended Load Mode Register” command when the SDRAM device is accessed regardless of
101 the cycle. To activate this mode, the “Extended Load Mode Register” command must be followed by a write to the SDRAM.
The write in the SDRAM must be done in the appropriate bank.
110 Deep power mode: Access to deep power-down mode
111 Reserved
253
21.7.2 DDRSDRC Refresh Timer Register
Name: DDRSDRC_RTR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – COUNT
7 6 5 4 3 2 1 0
COUNT
• COUNT: DDRSDRC Refresh Timer Count

This 12-bit field is loaded into a timer which generates the refresh pulse. Each time the refresh pulse is generated, a refresh
sequence is initiated.
SDRAM devices require a refresh of all rows every 64 ms. The value to be loaded depends on the DDRSDRC clock fre-
quency (MCK: Master Clock) and the number of rows in the device.
For example, for an SDRAM with 8192 rows and a 100 MHz Master clock, the value of Refresh Timer Count bit is pro-
grammed: (((64 x 10-3)/8192) x100 x106 = 781 or 0x030D.
254 AT91SAM9G45
AT91SAM9G45
21.7.3 DDRSDRC Configuration Register

Name: DDRSDRC_CR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – ACTBST – DQMS
15 14 13 12 11 10 9 8
– OCD – – DIS_DLL DIC/DS
7 6 5 4 3 2 1 0
DLL CAS NR NC
• NC: Number of Column Bits

The reset value is 9 column bits.
SDR-SDRAM devices with eight columns in 16-bit mode (b16mode ==1) are not supported.
NC DDR - Column bits SDR - Column bits

00 9 8
01 10 9
10 11 10
11 12 11
• NR: Number of Row Bits

The reset value is 12 row bits.
NR Row bits
00 11
01 12
10 13
11 14
255
• CAS: CAS Latency
The reset value is 2 cycles.
CAS DDR2 CAS Latency SDR CAS Latency

000 Reserved Reserved
010 Reserved 2
011 3 3
• DLL: Reset DLL

Reset value is 0.
This field defines the value of Reset DLL.
0 = Disable DLL reset.
1 = Enable DLL reset.
This value is used during the power-up sequence.
Note: This field is found only in DDR1-SDRAM devices.
• DIC/DS: Output Driver Impedance Control:

Reset value is 0.
This field defines the output drive strength.
0 = Normal driver strength.
1 = Weak driver strength.
This value is used during the power-up sequence. This parameter is found in the datasheet as DIC or DS.
• DIS_DLL: Disable DLL

0 = Enable DLL
1 = Disable DLL
• OCD: Off-chip Driver

Reset value is 3’b111.
Note: OCD is NOT supported by the controller, but these values MUST be programmed during the initialization sequence.
OCD
000 OCD calibration mode exit, maintain setting
111 OCD calibration default
256 AT91SAM9G45
AT91SAM9G45
• DQMS: Mask Data is Shared

Reset value is 0.
0 = DQM is not shared with another controller.
1 = DQM is shared with another controller.
• ACTBST: ACTIVE Bank X to Burst Stop Read Access Bank Y

Reset value is 0.
0 = After an ACTIVE command in Bank X, BURST STOP command can be issued to another bank to stop current read
access.
1 = After an ACTIVE command in Bank X, BURST STOP command cannot be issued to another bank to stop current read
access.
This field is unique to SDR-SDRAM, Low-power SDR-SDRAM and Low-power DDR-SDRAM devices.
257
21.7.4 DDRSDRC Timing 0 Parameter Register
Name: DDRSDRC_T0PR
Access: Read-write
31 30 29 28 27 26 25 24
TMRD REDUCE_WRRD TWTR
23 22 21 20 19 18 17 16
TRRD TRP
15 14 13 12 11 10 9 8
TRC TWR
7 6 5 4 3 2 1 0
TRCD TRAS
• TRAS: Active to Precharge Delay

Reset Value is 5 cycles.
This field defines the delay between an Activate Command and a Precharge Command in number of cycles. Number of
cycles is between 0 and 15.
• TRCD: Row to Column Delay

This field defines the delay between an Activate Command and a Read/Write Command in number of cycles. Number of
cycles is between 0 and 15.
• TWR: Write Recovery Delay

Reset value is 2.
This field defines the Write Recovery Time in number of cycles. Number of cycles is between 1 and 15.
• TRC: Row Cycle Delay

Reset value is 7 cycles.
This field defines the delay between an Activate command and Refresh command in number of cycles. Number of cycles is
between 0 and 15
• TRP: Row Precharge Delay

This field defines the delay between a Precharge Command and another command in number of cycles. Number of cycles
is between 0 and 15.
• TRRD Active bankA to Active bankB

Reset value is 2.
This field defines the delay between an Active command in BankA and an active command in bankB in number of cycles.
Number of cycles is between 1 and 15.
258 AT91SAM9G45
AT91SAM9G45
• TWTR: Internal Write to Read Delay

Reset value is 0.
This field defines the internal write to read command time in number of cycles.
In the case of low power DDR-SDRAM only bit field 0 is used. Number of cycles is between 1 and 2.
bit[0] TWTR
0 1
1 1
In the case of DDR2-SDRAM, all bit foemds are ised. Number of cycles is between 0 and 7.
• REDUCE_WRRD: Reduce Write to Read Delay

Reset value is 0.
This field reduces the delay between write to read access for low-power DDR-SDRAM devices with a latency equal to 2. To
use this feature, TWTR field must be equal to 0. Important to note is that some devices do not support this feature.
• TMRD: Load Mode Register Command to Active or Refresh Command

This field defines the delay between a Load mode register command and an active or refresh command in number of
cycles. Number of cycles is between 0 and 15.
259
Name: DDRSDRC_T1PR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – TXP
23 22 21 20 19 18 17 16
TXSRD
15 14 13 12 11 10 9 8
TXSNR
7 6 5 4 3 2 1 0
– – – TRFC
• TRFC: Row Cycle Delay
This field defines the delay between a Refresh and an Activate command or Refresh command in number of cycles. Num-
ber of cycles is between 0 and 31
• TXSNR: Exit Self Refresh Delay to Non-read Command
This field defines the delay between cke set high and a non Read Command in number of cycles. Number of cycles is
between 0 and 15. This field is used for SDR-SDRAM and DDR-SDRAM devices. In the case of SDR-SDRAM devices and
Low-power DDR-SDRAM, this field is equivalent to TXSR timing.
• TXSRD: ExiT Self Refresh Delay to Read Command
Reset Value is C8.
This field defines the delay between cke set high and a Read Command in number of cycles. Number of cycles is between
0 and 255 cycles.This field is unique to DDR-SDRAM devices.
• TXP: Exit Power-down Delay to First Command
Reset Value is 3.
This field defines the delay between cke set high and a Valid Command in number of cycles. Number of cycles is between
0 and 15 cycles. This field is unique to Low-power DDR-SDRAM devices and DDR2-SDRAM devices.
260 AT91SAM9G45
AT91SAM9G45

Name: DDRSDRC_T2PR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
TRTP TRPA
7 6 5 4 3 2 1 0
TXARDS TXARD
• TXARD: Exit Active Power Down Delay to Read Command in Mode “Fast Exit”.
The Reset Value is 0 Cycle.
0 and 15.
• TXARDS: Exit Active Power Down Delay to Read Command in Mode “Slow Exit”.
0 and 15.
• TRPA: Row Precharge All Delay

This field defines the delay between a Precharge ALL banks Command and another command in number of cycles. Num-
ber of cycles is between 0 and 15.
• TRTP: Read to Precharge

This field defines the delay between Read Command and a Precharge command in number of cycle.
Number of cycles is between 0 and 15.
261
21.7.7 DDRSDRC Low-power Register
Name: DDRSDRC_LPR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – APDE
15 14 13 12 11 10 9 8
– – TIMEOUT DS TCR
7 6 5 4 3 2 1 0
– PASR CLK_FR LPCB
• LPCB: Low-power Command Bit

Reset value is “00”.
00 = Low-power Feature is inhibited: no power-down, self refresh and Deep power mode are issued to the SDRAM device.
01 = The DDRSDRC issues a Self Refresh Command to the SDRAM device, the clock(s) is/are de-activated and the CKE
signal is set low. The SDRAM device leaves the self refresh mode when accessed and enters it after the access.
10 = The DDRSDRC issues a Power-down Command to the SDRAM device after each access, the CKE signal is set low.
The SDRAM device leaves the power-down mode when accessed and enters it after the access.
11 = The DDRSDRC issues a Deep Power-down Command to the Low-power SDRAM device.This mode is unique to
Low-power SDRAM devices.
• CLK_FR: Clock Frozen Command Bit

This field sets the clock low during power-down mode or during deep power-down mode. Some SDRAM devices do not
support freezing the clock during power-down mode or during deep power-down mode. Refer to the SDRAM device
datasheet for details on this.
1 = Clock(s) is/are frozen.
0 = Clock(s) is/are not frozen.
• PASR: Partial Array Self Refresh

This field is unique to Low-power SDRAM. It is used to specify whether only one quarter, one half or all banks of the
SDRAM array are enabled. Disabled banks are not refreshed in self refresh mode.
The values of this field are dependant on Low-power SDRAM devices.
After the initialization sequence, as soon as PASR field is modified and self refresh mode is activated, Extended Mode
Register is accessed automatically and PASR bits are updated before entering self refresh mode. The automatic update
feature is not supported when the DDRSDRC shares an external bus with another controller.
262 AT91SAM9G45
AT91SAM9G45
• TCR: Temperature Compensated Self Refresh

This field is unique to Low-power SDRAM. It is used to program the refresh interval during self refresh mode, depending
on the case temperature of the low-power SDRAM.
The values of this field are dependent on Low-power SDRAM devices.
After the initialization sequence, as soon as TCR field is modified and self refresh mode is activated, Extended Mode Reg-
ister is accessed automatically and TCR bits are updated before entering in self refresh mode. The automatic update
• DS: Drive Strength

This field is unique to Low-power SDRAM. It selects the driver strength of SDRAM output.
After the initialization sequence, as soon as DS field is modified and self refresh mode is activated, Extended Mode Regis-
ter is accessed automatically and DS bits are updated before entering self refresh mode. The automatic update feature is
not supported when the DDRSDRC shares an external bus with another controller.
• TIMEOUT
This field defines when low-power mode is enabled.
00 The SDRAM controller activates the SDRAM low-power mode immediately after the end of the last transfer.
01 The SDRAM controller activates the SDRAM low-power mode 64 clock cycles after the end of the last transfer.
10 The SDRAM controller activates the SDRAM low-power mode 128 clock cycles after the end of the last transfer.
11 Reserved
• APDE: Active Power Down Exit Time

This mode is unique to DDR2-SDRAM devices. This mode allows to determine the active power-down mode, which
determines performance versus power saving.
0 = Fast Exit
1 = Slow Exit
After the initialization sequence, as soon as APDE field is modified and self refresh mode is activated, the Load Mode Reg-
ister is accessed automatically and APDE bits are updated before entering self refresh mode. The automatic update
263
AT91SAM9G45
21.7.8 DDRSDRC Memory Device Register

Name: DDRSDRC_MD
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – DBW – MD
• MD: Memory Device

Indicates the type of memory used.
Reset value is for SDR-SDRAM device.
000 = SDR-SDRAM
001 = Low-power SDR-SDRAM
010 = Reserved
011 = Low-power DDR1-SDRAM
110 = DDR2-SDRAM
• DBW: Data Bus Width

Reset value is 16 bits.
0 = Data bus width is 32 bits (reserved for SDR-SDRAM device).
1 = Data bus width is 16 bits.
264
AT91SAM9G45
21.7.9 DDRSDRC DLL Register

Name: DDRSDRC_DLL
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
MDVAL
7 6 5 4 3 2 1 0
– – – – – MDOVF MDDEC MDINC
The DLL logic is internally used by the controller in order to delay DQS inputs.This is necessary to center the strobe time
and the data valid window.
• MDINC: DLL Master Delay Increment

0 = The DLL is not incrementing the Master delay counter.
1 = The DLL is incrementing the Master delay counter.
• MDDEC: DLL Master Delay Decrement

0 = The DLL is not decrementing the Master delay counter.
1 = The DLL is decrementing the Master delay counter.
• MDOVF: DLL Master Delay Overflow Flag

0 = The Master delay counter has not reached its maximum value, or the Master is not locked yet.
1 = The Master delay counter has reached its maximum value, the Master delay counter increment is stopped and the DLL
forces the Master lock. If this flag is set, it means the DDRSDRC clock frequency is too low compared to Master delay line
number of elements.
• MDVAL: DLL Master Delay Value

Value of the Master delay counter.
265
AT91SAM9G45
21.7.10 DDRSDRC High Speed Register

Name: DDRSDRC_HS
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – reserved DIS_ANTICIP_READ NO_OPTI reserved
• NO_OPTI: No Optimization
0 = optimization is enabled.
1 = optimization is disabled.
• DIS_ANTICIP_READ
0 = anticip read access is enabled.
1 = anticip read access is disabled.
266
AT91SAM9G45
21.7.11 DDRSDRC DELAY I/O Register

Name: DDRSDRC_DELAY 1-5
Access: Read-write
31 30 29 28 27 26 25 24
Delay8 Delay7
23 22 21 20 19 18 17 16
Delay6 Delay5
15 14 13 12 11 10 9 8
Delay4 Delay3
7 6 5 4 3 2 1 0
Delay2 Delay1
• Delay x:
Gives the number of elements in the delay line.
267
AT91SAM9G45
21.7.12 DDRSDRC Drive Strength I/O Register

Name: DDRSDRC_DRIVESTRENGTH
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – DSC
• DSC: Drive Strength Configuration

1 = Full Drive.
0 = Low Drive.
268
AT91SAM9G45
21.7.13 DDRSDRC Write Protect Mode Register

Name: DDRSDRC_WPMR
Access: Read-write
31 30 29 28 27 26 25 24
WPKEY
23 22 21 20 19 18 17 16
WPKEY
15 14 13 12 11 10 9 8
WPKEY
7 6 5 4 3 2 1 0
— — — — — — — WPEN
• WPEN: Write Protect Enable

0 = Disables the Write Protect if WPKEY corresponds to 0x444452 (“DDR” in ASCII).
1 = Enables the Write Protect if WPKEY corresponds to 0x444452 (“DDR” in ASCII).
Protects the registers:
• “DDRSDRC Mode Register” on page 253
• “DDRSDRC Refresh Timer Register” on page 254
• “DDRSDRC Configuration Register” on page 255
• “DDRSDRC Memory Device Register” on page 264
• “DDRSDRC High Speed Register” on page 266
• WPKEY: Write Protect KEY

Should be written at value 0x444452 (“DDR” in ASCII). Writing any other value in this field aborts the write operation of the
269
AT91SAM9G45
21.7.14 DDRSDRC Write Protect Status Register

Name: DDRSDRC_WPSR
Access: Read-only
31 30 29 28 27 26 25 24
— — — — — — — —
23 22 21 20 19 18 17 16
WPVSRC
15 14 13 12 11 10 9 8
WPVSRC
7 6 5 4 3 2 1 0
— — — — — — — WPVS
• WPVS: Write Protect Violation Status

0 = No Write Protect Violation has occurred since the last read of the DDRSDRC_WPSR register.
1 = A Write Protect Violation has occurred since the last read of the DDRSDRC_WPSR register. If this violation is an unau-
thorized attempt to write a protected register, the associated violation is reported into field WPVSRC.

When WPVS is active, this field indicates the write-protected register (through address offset or code) in which a write
access has been attempted.
Note: Reading DDRSDRC_WPSR automatically clears all fields.
270
AT91SAM9G45
22. Error Corrected Code Controller (ECC)
22.1 Description
NAND Flash/SmartMedia devices contain by default invalid blocks which have one or more
invalid bits. Over the NAND Flash/SmartMedia lifetime, additional invalid blocks may occur
which can be detected/corrected by ECC code.
The ECC Controller is a mechanism that encodes data in a manner that makes possible the
identification and correction of certain errors in data. The ECC controller is capable of single bit
error correction and 2-bit random detection. When NAND Flash/SmartMedia have more than 2
bits of errors, the data cannot be corrected.
The ECC user interface is compliant with the ARM Advanced Peripheral Bus (APB rev2).
22.2 Block Diagram
Figure 22-1. Block Diagram
NAND Flash
Static
Memory
SmartMedia
Controller
Logic
ECC
Controller
Ctrl/ECC Algorithm
User Interface
APB

A page in NAND Flash and SmartMedia memories contains an area for main data and an addi-
tional area used for redundancy (ECC). The page is organized in 8-bit or 16-bit words. The page
size corresponds to the number of words in the main area plus the number of words in the extra
area used for redundancy.
Over time, some memory locations may fail to program or erase properly. In order to ensure that
data is stored properly over the life of the NAND Flash device, NAND Flash providers recom-
271
AT91SAM9G45
mend to utilize either 1 ECC per 256 bytes of data, 1 ECC per 512 bytes of data or 1 ECC for all
of the page.
The only configurations required for ECC are the NAND Flash or the SmartMedia page size
(528/2112/4224) and the type of correction wanted (1 ECC for all the page/1 ECC per 256 bytes
of data /1 ECC per 512 bytes of data). Page size is configured setting the PAGESIZE field in the
ECC Mode Register (ECC_MR). Type of correction is configured setting the TYPeCORRECT
field in the ECC Mode Register (ECC_MR).
ECC is automatically computed as soon as a read (00h)/write (80h) command to the NAND
Flash or the SmartMedia is detected. Read and write access must start at a page boundary.
ECC results are available as soon as the counter reaches the end of the main area. Values in
the ECC Parity Registers (ECC_PR0 to ECC_PR15) are then valid and locked until a new start
condition occurs (read/write command followed by address cycles).
22.3.1 Write Access

Once the Flash memory page is written, the computed ECC codes are available in the ECC Par-
ity (ECC_PR0 to ECC_PR15) registers. The ECC code values must be written by the software
application in the extra area used for redundancy. The number of write accesses in the extra
area is a function of the value of the type of correction field. For example, for 1 ECC per 256
bytes of data for a page of 512 bytes, only the values of ECC_PR0 and ECC_PR1 must be writ-
ten by the software application. Other registers are meaningless.
22.3.2 Read Access

After reading the whole data in the main area, the application must perform read accesses to the
extra area where ECC code has been previously stored. Error detection is automatically per-
formed by the ECC controller. Please note that it is mandatory to read consecutively the entire
main area and the locations where Parity and NParity values have been previously stored to let
the ECC controller perform error detection.
The application can check the ECC Status Registers (ECC_SR1/ECC_SR2) for any detected
errors. It is up to the application to correct any detected error. ECC computation can detect four
different circumstances:
• No error: XOR between the ECC computation and the ECC code stored at the end of the
NAND Flash or SmartMedia page is equal to 0. No error flags in the ECC Status Registers
(ECC_SR1/ECC_SR2).
• Recoverable error: Only the RECERR flags in the ECC Status registers
(ECC_SR1/ECC_SR2) are set. The corrupted word offset in the read page is defined by the
WORDADDR field in the ECC Parity Registers (ECC_PR0 to ECC_PR15). The corrupted bit
position in the concerned word is defined in the BITADDR field in the ECC Parity Registers
(ECC_PR0 to ECC_PR15).
• ECC error: The ECCERR flag in the ECC Status Registers (ECC_SR1/ECC_SR2) are set.
An error has been detected in the ECC code stored in the Flash memory. The position of the
corrupted bit can be found by the application performing an XOR between the Parity and the
NParity contained in the ECC code stored in the Flash memory.
• Non correctable error: The MULERR flag in the ECC Status Registers
(ECC_SR1/ECC_SR2) are set. Several unrecoverable errors have been detected in the
Flash memory page.
272
AT91SAM9G45
ECC Status Registers, ECC Parity Registers are cleared when a read/write command is
detected or a software reset is performed.
For Single-bit Error Correction and Double-bit Error Detection (SEC-DED) hsiao code is used.
24-bit ECC is generated in order to perform one bit correction per 256 or 512 bytes for pages of
512/2048/4096 8-bit words. 32-bit ECC is generated in order to perform one bit correction per
512/1024/2048/4096 8- or 16-bit words.They are generated according to the schemes shown in
Figure 22-2 and Figure 22-3.
Figure 22-2. Parity Generation for

Bit6512/1024/2048/4096
Bit5 Bit4 Bit3 8-bit
Bit2 Words
Bit1 Bit0
1st byte P16
2nd byte Bit7 Bit6 Bit5 Bit4 Bit3 Bit2 Bit1 Bit0 P8'
P32 PX
3rd byte Bit7 Bit6 Bit5 Bit4 Bit3 Bit2 Bit1 Bit0 P8
P16'
4 th byte Bit7 Bit6 Bit5 Bit4 Bit3 Bit2 Bit1 Bit0 P8'
(page size -3 )th byte Bit7 Bit6 Bit5 Bit4 Bit3 Bit2 Bit1 Bit0 P8
P16
(page size -2 )th byte Bit7 Bit6 Bit5 Bit4 Bit3 Bit2 Bit1 Bit0 P8' PX'
P32
(page size -1 )th byte Bit7 Bit6 Bit5 Bit4 Bit3 Bit2 Bit1 Bit0 P8
P16'
Page size th byte Bit7 Bit6 Bit5 Bit4 Bit3 Bit2 Bit1 Bit0 P8'
P1 P1' P1 P1' P1 P1' P1 P1'
P2 P2' P2 P2'
P4 P4'
Page size = 512 Px = 2048 P1=bit7(+)bit5(+)bit3(+)bit1(+)P1

Page size = 4096 Px = 16384 P1'=bit6(+)bit4(+)bit2(+)bit0(+)P1'
P2' bit5( )bit4( )bit1( )bit0( )P2'
To calculate P8’ to PX’ and P8 to PX, apply the algorithm that follows.
Page size = 2n
for i =0 to n
begin
for (j = 0 to page_size_byte)
begin
if(j[i] ==1)
P[2i+3]=bit7(+)bit6(+)bit5(+)bit4(+)bit3(+)
bit2(+)bit1(+)bit0(+)P[2i+3]
else
P[2i+3]’=bit7(+)bit6(+)bit5(+)bit4(+)bit3(+)
bit2(+)bit1(+)bit0(+)P[2i+3]'
end
end
273
1st word
2nd word
3rd word
4th word
(Page size -3 )th word

Page size th word (+)
Figure 22-3. Parity Generation for 512/1024/2048/4096 16-bit Words
274
AT91SAM9G45
AT91SAM9G45
To calculate P8’ to PX’ and P8 to PX, apply the algorithm that follows.
Page size = 2n
for i =0 to n
begin
for (j = 0 to page_size_word)
begin
if(j[i] ==1)
P[2i+3]= bit15(+)bit14(+)bit13(+)bit12(+)
bit11(+)bit10(+)bit9(+)bit8(+)
bit7(+)bit6(+)bit5(+)bit4(+)bit3(+)
bit2(+)bit1(+)bit0(+)P[2n+3]
else
P[2i+3]’=bit15(+)bit14(+)bit13(+)bit12(+)
bit11(+)bit10(+)bit9(+)bit8(+)
bit7(+)bit6(+)bit5(+)bit4(+)bit3(+)
bit2(+)bit1(+)bit0(+)P[2i+3]'
end
end
275
AT91SAM9G45
22.4 Error Corrected Code Controller (ECC) User Interface

0x00 ECC Control Register ECC_CR Write-only 0x0
0x04 ECC Mode Register ECC_MR Read-write 0x0
0x08 ECC Status1 Register ECC_SR1 Read-only 0x0
0x0C ECC Parity Register 0 ECC_PR0 Read-only 0x0
0x10 ECC Parity Register 1 ECC_PR1 Read-only 0x0
0x14 ECC Status2 Register ECC_SR2 Read-only 0x0
0x18 ECC Parity 2 ECC_PR2 Read-only 0x0
0x1C ECC Parity 3 ECC_PR3 Read-only 0x0
276
AT91SAM9G45
22.4.1 ECC Control Register

Name: ECC_CR
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – SRST RST
• RST: RESET Parity

Provides reset to current ECC by software.
1: Reset ECC Parity registers
0: No effect
• SRST: Soft Reset

Provides soft reset to ECC block
1: Resets all registers.
0: No effect.
277
AT91SAM9G45
22.4.2 ECC Mode Register

Register Name: ECC_MR
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – TYPECORRECT – – PAGESIZE
• PAGESIZE: Page Size

This field defines the page size of the NAND Flash device.
Page Size Description

00 528 words
01 1056 words
10 2112 words
11 4224 words
A word has a value of 8 bits or 16 bits, depending on the NAND Flash or SmartMedia memory organization.
• TYPECORRECT: Type of Correction

00: 1 bit correction for a page size of 512/1024/2048/4096 bytes.
01: 1 bit correction for 256 bytes of data for a page size of 512/2048/4096 bytes.
10: 1 bit correction for 512 bytes of data for a page size of 512/2048/4096 bytes.
278
AT91SAM9G45
22.4.3 ECC Status Register 1

Register Name: ECC_SR1
31 30 29 28 27 26 25 24
– MULERR7 ECCERR7 RECERR7 – MULERR6 ECCERR6 RECERR6
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• RECERR0: Recoverable Error

0 = No Errors Detected.
1 = Errors Detected. If MUL_ERROR is 0, a single correctable error was detected. Otherwise multiple uncorrected errors
were detected.
• ECCERR0: ECC Error

1 = A single bit error occurred in the ECC bytes.
If TYPECORRECT = 0, read both ECC Parity 0 and ECC Parity 1 registers, the error occurred at the location which con-
tains a 1 in the least significant 16 bits; else read ECC Parity 0 register, the error occurred at the location which contains a
1 in the least significant 24 bits.
• MULERR0: Multiple Error

0 = No Multiple Errors Detected.
1 = Multiple Errors Detected.
• RECERR1: Recoverable Error in the page between the 256th and the 511th bytes or the 512th and the 1023rd
bytes
Fixed to 0 if TYPECORREC = 0.
were detected.
• ECCERR1: ECC Error in the page between the 256th and the 511th bytes or the 512th and the 1023rd bytes
Fixed to 0 if TYPECORREC = 0
Read ECC Parity 1 register, the error occurred at the location which contains a 1 in the least significant 24 bits.
• MULERR1: Multiple Error in the page between the 256th and the 511th bytes or the 512th and the 1023rd bytes
279
AT91SAM9G45
• RECERR2: Recoverable Error in the page between the 512th and the 767th bytes or the 1024th and the 1535th
bytes
1 = Errors Detected. If MUL_ERROR is 0, a single correctable error was detected. Otherwise, multiple uncorrected errors
were detected.
• ECCERR2: ECC Error in the page between the 512th and the 767th bytes or the 1024th and the 1535th bytes
• MULERR2: Multiple Error in the page between the 512th and the 767th bytes or the 1024th and the 1535th bytes
• RECERR3: Recoverable Error in the page between the 768th and the 1023rd bytes or the 1536th and the 2047th
bytes
were detected.
• ECCERR3: ECC Error in the page between the 768th and the 1023rd bytes or the 1536th and the 2047th bytes
• MULERR3: Multiple Error in the page between the 768th and the 1023rd bytes or the 1536th and the 2047th bytes
• RECERR4: Recoverable Error in the page between the 1024th and the 1279th bytes or the 2048th and the 2559th
bytes
were detected.
280
AT91SAM9G45
• ECCERR4: ECC Error in the page between the 1024th and the 1279th bytes or the 2048th and the 2559th bytes
• MULERR4: Multiple Error in the page between the 1024th and the 1279th bytes or the 2048th and the 2559th
bytes
• RECERR5: Recoverable Error in the page between the 1280th and the 1535th bytes or the 2560th and the 3071st
bytes
were detected
• ECCERR5: ECC Error in the page between the 1280th and the 1535th bytes or the 2560th and the 3071st bytes
• MULERR5: Multiple Error in the page between the 1280th and the 1535th bytes or the 2560th and the 3071st
bytes
• RECERR6: Recoverable Error in the page between the 1536th and the 1791st bytes or the 3072nd and the 3583rd
bytes
were detected.
• ECCERR6: ECC Error in the page between the 1536th and the 1791st bytes or the 3072nd and the 3583rd bytes
281
AT91SAM9G45
• MULERR6: Multiple Error in the page between the 1536th and the 1791st bytes or the 3072nd and the 3583rd
bytes
• RECERR7: Recoverable Error in the page between the 1792nd and the 2047th bytes or the 3584th and the 4095th
bytes
were detected.
• ECCERR7: ECC Error in the page between the 1792nd and the 2047th bytes or the 3584th and the 4095th bytes
• MULERR7: Multiple Error in the page between the 1792nd and the 2047th bytes or the 3584th and the 4095th
bytes
282
AT91SAM9G45
22.4.4 ECC Status Register 2

Register Name: ECC_SR2
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• RECERR8: Recoverable Error in the page between the 2048th and the 2303rd bytes
were detected
• ECCERR8: ECC Error in the page between the 2048th and the 2303rd bytes
• MULERR8: Multiple Error in the page between the 2048th and the 2303rd bytes
• RECERR9: Recoverable Error in the page between the 2304th and the 2559th bytes
were detected.
• ECCERR9: ECC Error in the page between the 2304th and the 2559th bytes
• MULERR9: Multiple Error in the page between the 2304th and the 2559th bytes
283
AT91SAM9G45
were detected.
• RECERR11: Recoverable Error in the page between the 2816th and the 3071st bytes
1 = Errors Detected.. If MUL_ERROR is 0, a single correctable error was detected. Otherwise, multiple uncorrected errors
were detected
• ECCERR11: ECC Error in the page between the 2816th and the 3071st bytes
• MULERR11: Multiple Error in the page between the 2816th and the 3071st bytes
• RECERR12: Recoverable Error in the page between the 3072nd and the 3327th bytes
0 = No Errors Detected
were detected
• ECCERR12: ECC Error in the page between the 3072nd and the 3327th bytes
284
AT91SAM9G45
0 = No Errors Detected
• MULERR12: Multiple Error in the page between the 3072nd and the 3327th bytes
• RECERR13: Recoverable Error in the page between the 3328th and the 3583rd bytes
were detected.
• ECCERR13: ECC Error in the page between the 3328th and the 3583rd bytes
• MULERR13: Multiple Error in the page between the 3328th and the 3583rd bytes
were detected.
285
AT91SAM9G45
were detected
286
AT91SAM9G45
22.5 Registers for 1 ECC for a page of 512/1024/2048/4096 bytes
22.5.1 ECC Parity Register 0

Register Name: ECC_PR0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
WORDADDR
7 6 5 4 3 2 1 0
WORDADDR BITADDR
Once the entire main area of a page is written with data, the register content must be stored at any free location of the
spare area.
• BITADDR: Bit Address

During a page read, this value contains the corrupted bit offset where an error occurred, if a single error was detected. If
multiple errors were detected, this value is meaningless.
• WORDADDR: Word Address

During a page read, this value contains the word address (8-bit or 16-bit word depending on the memory plane organiza-
tion) where an error occurred, if a single error was detected. If multiple errors were detected, this value is meaningless.
287
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
NPARITY
7 6 5 4 3 2 1 0
NPARITY
• NPARITY:
Parity N
288
AT91SAM9G45
22.6 Registers for 1 ECC per 512 bytes for a page of 512/2048/4096 bytes, 8-bit word

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
NPARITY0
15 14 13 12 11 10 9 8
NPARITY0 WORDADD0
7 6 5 4 3 2 1 0
WORDADDR0 BITADDR0
spare area.
• BITADDR0: corrupted Bit Address in the page between the first byte and the 511th bytes
• WORDADDR0: corrupted Word Address in the page between the first byte and the 511th bytes
During a page read, this value contains the word address (8-bit word) where an error occurred, if a single error was
detected. If multiple errors were detected, this value is meaningless.
• NPARITY0:
Parity N
289
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
NPARITY1
15 14 13 12 11 10 9 8
NPARITY1 WORDADD1
7 6 5 4 3 2 1 0
WORDADDR1 BITADDR1
spare area.
• BITADDR1: corrupted Bit Address in the page between the 512th and the 1023rd bytes
• WORDADDR1: corrupted Word Address in the page between the 512th and the 1023rd bytes
• NPARITY1:
Parity N
290
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
NPARITY2
15 14 13 12 11 10 9 8
NPARITY2 WORDADDR2
7 6 5 4 3 2 1 0
WORDADDR2 BITADDR2
spare area.
• BITADDR2: corrupted Bit Address in the page between the 1023rd and the 1535th bytes
• WORDADDR2: corrupted Word Address in the page in the page between the 1023rd and the 1535th bytes
• NPARITY2:
Parity N
291
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
NPARITY3
15 14 13 12 11 10 9 8
NPARITY3 WORDADDR3
7 6 5 4 3 2 1 0
WORDADDR3 BITADDR3
spare area.
• BITADDR3: corrupted Bit Address in the page between the1536th and the 2047th bytes
• WORDADDR3 corrupted Word Address in the page between the 1536th and the 2047th bytes
• NPARITY3
Parity N
292
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
NPARITY4
15 14 13 12 11 10 9 8
NPARITY4 WORDADDR4
7 6 5 4 3 2 1 0
WORDADDR4 BITADDR4
spare area.
• BITADDR4: corrupted Bit Address in the page between the 2048th and the 2559th bytes
• WORDADDR4: corrupted Word Address in the page between the 2048th and the 2559th bytes
• NPARITY4:
Parity N
293
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
NPARITY5
15 14 13 12 11 10 9 8
NPARITY5 WORDADDR5
7 6 5 4 3 2 1 0
WORDADDR5 BITADDR5
spare area.
• BITADDR5: corrupted Bit Address in the page between the 2560th and the 3071st bytes
• WORDADDR5: corrupted Word Address in the page between the 2560th and the 3071st bytes
• NPARITY5:
Parity N
294
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
NPARITY6
15 14 13 12 11 10 9 8
NPARITY6 WORDADDR6
7 6 5 4 3 2 1 0
WORDADDR6 BITADDR6
spare area.
• BITADDR6: corrupted Bit Address in the page between the 3072nd and the 3583rd bytes
• WORDADDR6: corrupted Word Address in the page between the 3072nd and the 3583rd bytes
• NPARITY6:
Parity N
295
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
NPARITY7
15 14 13 12 11 10 9 8
NPARITY7 WORDADDR7
7 6 5 4 3 2 1 0
WORDADDR7 BITADDR7
spare area.
• BITADDR7: corrupted Bit Address in the page between the 3584h and the 4095th bytes
• NPARITY7:
Parity N
296
AT91SAM9G45
22.7 Registers for 1 ECC per 256 bytes for a page of 512/2048/4096 bytes, 8-bit word

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY0
15 14 13 12 11 10 9 8
NPARITY0 0 WORDADDR0
7 6 5 4 3 2 1 0
WORDADDR0 BITADDR0
spare area.
• BITADDR0: corrupted Bit Address in the page between the first byte and the 255th bytes
• WORDADDR0: corrupted Word Address in the page between the first byte and the 255th bytes
• NPARITY0:
Parity N
297
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY1
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR1 BITADDR1
spare area
• NPARITY1:
Parity N
298
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY2
15 14 13 12 11 10 9 8
NPARITY2 0 WORDADD2
7 6 5 4 3 2 1 0
WORDADDR2 BITADDR2
spare area.
• NPARITY2:
Parity N
299
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY3
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR3 BITADDR3
spare area.
detected. If multiple errors were detected, this value is meaningless
• NPARITY3:
Parity N
300
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY4
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR4 BITADDR4
spare area
• BITADDR4: corrupted bit address in the page between the 1024th and the 1279th bytes
• WORDADDR4: corrupted word address in the page between the 1024th and the 1279th bytes
• NPARITY4
Parity N
301
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY5
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR5 BITADDR5
spare area.
• NPARITY5:
Parity N
302
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY6
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR6 BITADDR6
spare area.
• BITADDR6: corrupted bit address in the page between the 1536th and the1791st bytes
• WORDADDR6: corrupted word address in the page between the 1536th and the1791st bytes
• NPARITY6:
Parity N
303
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY7
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR7 BITADDR7
spare area.
• BITADDR7: corrupted Bit Address in the page between the 1792nd and the 2047th bytes
• WORDADDR7: corrupted Word Address in the page between the 1792nd and the 2047th bytes
• NPARITY7:
Parity N
304
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY8
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR8 BITADDR8
spare area.
• BITADDR8: corrupted Bit Address in the page between the 2048th and the2303rd bytes
• NPARITY8:
Parity N.
305
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY9
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR9 BITADDR9
spare area
• BITADDR9: corrupted bit address in the page between the 2304th and the 2559th bytes
• WORDADDR9: corrupted word address in the page between the 2304th and the 2559th bytes
detected. If multiple errors were detected, this value is meaningless
• NPARITY9
Parity N
306
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY10
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
WORDADDR10 BITADDR10
spare area.
• BITADDR10: corrupted Bit Address in the page between the 2560th and the2815th bytes
• NPARITY10:
Parity N
307
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY11
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
spare area.
• BITADDR11: corrupted Bit Address in the page between the 2816th and the 3071st bytes
• WORDADDR11: corrupted Word Address in the page between the 2816th and the 3071st bytes
• NPARITY11:
Parity N
308
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY12
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
spare area.
• BITADDR12; corrupted Bit Address in the page between the 3072nd and the 3327th bytes
• WORDADDR12: corrupted Word Address in the page between the 3072nd and the 3327th bytes
• NPARITY12:
Parity N
309
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY13
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
spare area.
• NPARITY13:
Parity N
310
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY14
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
spare area.
• NPARITY14:
Parity N
311
AT91SAM9G45

31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
0 NPARITY15
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
spare area
• NPARITY15
Parity N
312
AT91SAM9G45
23. Peripheral DMA Controller (PDC)
23.1 Description
The Peripheral DMA Controller (PDC) transfers data between on-chip serial peripherals and the
on- and/or off-chip memories. The link between the PDC and a serial peripheral is operated by
the AHB to ABP bridge.
The user interface of each PDC channel is integrated into the user interface of the peripheral it
serves. The user interface of mono directional channels (receive only or transmit only), contains
two 32-bit memory pointers and two 16-bit counters, one set (pointer, counter) for current trans-
fer and one set (pointer, counter) for next transfer. The bi-directional channel user interface
contains four 32-bit memory pointers and four 16-bit counters. Each set (pointer, counter) is
used by current transmit, next transmit, current receive and next receive.
Using the PDC removes processor overhead by reducing its intervention during the transfer.
This significantly reduces the number of clock cycles required for a data transfer, which
improves microcontroller performance.
To launch a transfer, the peripheral triggers its associated PDC channels by using transmit and
receive signals. When the programmed data is transferred, an end of transfer interrupt is gener-
ated by the peripheral itself.

• Acting as one AHB Bus Matrix Master
• Allows data transfers from/to peripheral to/from any memory space without any intervention
of the processor.
• Next Pointer support, prevents strong real-time constraints on buffer management.
The Peripheral DMA Controller handles transfer requests from the channel according to the fol-
lowing priorities (Low to High priorities):
Table 23-1. Peripheral DMA Controller

Instance name Channel T/R
DBGU Transmit
USART3 Transmit
USART2 Transmit
USART1 Transmit
USART0 Transmit
AC97 Transmit
SPI1 Transmit
SPI0 Transmit
SSC1 Transmit
SSC0 Transmit
TSDAC Receive
DBGU Receive
USART3 Receive
313
Table 23-1. Peripheral DMA Controller
Instance name Channel T/R
USART2 Receive
USART1 Receive
USART0 Receive
AC97 Receive
SPI1 Receive
SPI0 Receive
SSC1 Receive
SSC0 Receive
314 AT91SAM9G45
AT91SAM9G45
23.3 Block Diagram
FULL DUPLEX PDC

PERIPHERAL
THR PDC Channel A
RHR PDC Channel B
Status & Control

Control
HALF DUPLEX
PERIPHERAL Control
THR
PDC Channel C
RHR
Status & Control

Control
RECEIVE or TRANSMIT
PERIPHERAL
RHR or THR PDC Channel D
Status & Control

Control
23.4.1 Configuration
The PDC channel user interface enables the user to configure and control data transfers for
each channel. The user interface of each PDC channel is integrated into the associated periph-
eral user interface.
The user interface of a serial peripheral, whether it is full or half duplex, contains four 32-bit
pointers (RPR, RNPR, TPR, TNPR) and four 16-bit counter registers (RCR, RNCR, TCR,
TNCR). However, the transmit and receive parts of each type are programmed differently: the
315
transmit and receive parts of a full duplex peripheral can be programmed at the same time,
whereas only one part (transmit or receive) of a half duplex peripheral can be programmed at a
time.
32-bit pointers define the access location in memory for current and next transfer, whether it is
for read (transmit) or write (receive). 16-bit counters define the size of current and next transfers.
It is possible, at any moment, to read the number of transfers left for each channel.
The PDC has dedicated status registers which indicate if the transfer is enabled or disabled for
each channel. The status for each channel is located in the associated peripheral status register.
Transfers can be enabled and/or disabled by setting TXTEN/TXTDIS and RXTEN/RXTDIS in
the peripheral’s Transfer Control Register.
At the end of a transfer, the PDC channel sends status flags to its associated peripheral. These
flags are visible in the peripheral status register (ENDRX, ENDTX, RXBUFF, and TXBUFE).
Refer to Section 23.4.3 and to the associated peripheral user interface.
23.4.2 Memory Pointers

Each full duplex peripheral is connected to the PDC by a receive channel and a transmit chan-
nel. Both channels have 32-bit memory pointers that point respectively to a receive area and to
a transmit area in on- and/or off-chip memory.
Each half duplex peripheral is connected to the PDC by a bidirectional channel. This channel
has two 32-bit memory pointers, one for current transfer and the other for next transfer. These
pointers point to transmit or receive data depending on the operating mode of the peripheral.
Depending on the type of transfer (byte, half-word or word), the memory pointer is incremented
respectively by 1, 2 or 4 bytes.
If a memory pointer address changes in the middle of a transfer, the PDC channel continues
operating using the new address.
23.4.3 Transfer Counters

Each channel has two 16-bit counters, one for current transfer and the other one for next trans-
fer. These counters define the size of data to be transferred by the channel. The current transfer
counter is decremented first as the data addressed by current memory pointer starts to be trans-
ferred. When the current transfer counter reaches zero, the channel checks its next transfer
counter. If the value of next counter is zero, the channel stops transferring data and sets the
appropriate flag. But if the next counter value is greater then zero, the values of the next
pointer/next counter are copied into the current pointer/current counter and the channel resumes
the transfer whereas next pointer/next counter get zero/zero as values. At the end of this trans-
fer the PDC channel sets the appropriate flags in the Peripheral Status Register.
The following list gives an overview of how status register flags behave depending on the coun-
ters’ values:
• ENDRX flag is set when the PERIPH_RCR register reaches zero.
• RXBUFF flag is set when both PERIPH_RCR and PERIPH_RNCR reach zero.
• ENDTX flag is set when the PERIPH_TCR register reaches zero.
• TXBUFE flag is set when both PERIPH_TCR and PERIPH_TNCR reach zero.
These status flags are described in the Peripheral Status Register.
316 AT91SAM9G45
AT91SAM9G45
23.4.4 Data Transfers

The serial peripheral triggers its associated PDC channels’ transfers using transmit enable
(TXEN) and receive enable (RXEN) flags in the transfer control register integrated in the periph-
eral’s user interface.
When the peripheral receives an external data, it sends a Receive Ready signal to its PDC
receive channel which then requests access to the Matrix. When access is granted, the PDC
receive channel starts reading the peripheral Receive Holding Register (RHR). The read data
are stored in an internal buffer and then written to memory.
When the peripheral is about to send data, it sends a Transmit Ready to its PDC transmit chan-
nel which then requests access to the Matrix. When access is granted, the PDC transmit
channel reads data from memory and puts them to Transmit Holding Register (THR) of its asso-
ciated peripheral. The same peripheral sends data according to its mechanism.
23.4.5 PDC Flags and Peripheral Status Register

Each peripheral connected to the PDC sends out receive ready and transmit ready flags and the
PDC sends back flags to the peripheral. All these flags are only visible in the Peripheral Status
Register.
Depending on the type of peripheral, half or full duplex, the flags belong to either one single
channel or two different channels.
23.4.5.1 Receive Transfer End

This flag is set when PERIPH_RCR register reaches zero and the last data has been transferred
to memory.
It is reset by writing a non zero value in PERIPH_RCR or PERIPH_RNCR.
23.4.5.2 Transmit Transfer End

This flag is set when PERIPH_TCR register reaches zero and the last data has been written into
peripheral THR.
It is reset by writing a non zero value in PERIPH_TCR or PERIPH_TNCR.
23.4.5.3 Receive Buffer Full

This flag is set when PERIPH_RCR register reaches zero with PERIPH_RNCR also set to zero
and the last data has been transferred to memory.
23.4.5.4 Transmit Buffer Empty

This flag is set when PERIPH_TCR register reaches zero with PERIPH_TNCR also set to zero
and the last data has been written into peripheral THR.
317
23.5 Peripheral DMA Controller (PDC) User Interface
(1)
0x100 Receive Pointer Register PERIPH _RPR Read-write 0
0x104 Receive Counter Register PERIPH_RCR Read-write 0
0x108 Transmit Pointer Register PERIPH_TPR Read-write 0
0x10C Transmit Counter Register PERIPH_TCR Read-write 0
0x110 Receive Next Pointer Register PERIPH_RNPR Read-write 0
0x114 Receive Next Counter Register PERIPH_RNCR Read-write 0
0x118 Transmit Next Pointer Register PERIPH_TNPR Read-write 0
0x11C Transmit Next Counter Register PERIPH_TNCR Read-write 0
0x120 Transfer Control Register PERIPH_PTCR Write-only 0
0x124 Transfer Status Register PERIPH_PTSR Read-only 0
Note: 1. PERIPH: Ten registers are mapped in the peripheral memory space at the same offset. These can be defined by the user
according to the function and the desired peripheral.)
318 AT91SAM9G45
AT91SAM9G45
23.5.1 Receive Pointer Register

Name: PERIPH_RPR
Access: Read-write
31 30 29 28 27 26 25 24
RXPTR
23 22 21 20 19 18 17 16
RXPTR
15 14 13 12 11 10 9 8
RXPTR
7 6 5 4 3 2 1 0
RXPTR
• RXPTR: Receive Pointer Register

RXPTR must be set to receive buffer address.
When a half duplex peripheral is connected to the PDC, RXPTR = TXPTR.
319
23.5.2 Receive Counter Register
Name: PERIPH_RCR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RXCTR
7 6 5 4 3 2 1 0
RXCTR
• RXCTR: Receive Counter Register

RXCTR must be set to receive buffer size.
When a half duplex peripheral is connected to the PDC, RXCTR = TXCTR.
0 = Stops peripheral data transfer to the receiver
1 - 65535 = Starts peripheral data transfer if corresponding channel is active
320 AT91SAM9G45
AT91SAM9G45
23.5.3 Transmit Pointer Register

Name: PERIPH_TPR
Access: Read-write
31 30 29 28 27 26 25 24
TXPTR
23 22 21 20 19 18 17 16
TXPTR
15 14 13 12 11 10 9 8
TXPTR
7 6 5 4 3 2 1 0
TXPTR
• TXPTR: Transmit Counter Register

TXPTR must be set to transmit buffer address.
When a half duplex peripheral is connected to the PDC, RXPTR = TXPTR.
321
AT91SAM9G45
23.5.4 Transmit Counter Register

Name: PERIPH_TCR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
TXCTR
7 6 5 4 3 2 1 0
TXCTR
• TXCTR: Transmit Counter Register

TXCTR must be set to transmit buffer size.
When a half duplex peripheral is connected to the PDC, RXCTR = TXCTR.
0 = Stops peripheral data transfer to the transmitter
1- 65535 = Starts peripheral data transfer if corresponding channel is active
322
AT91SAM9G45
23.5.5 Receive Next Pointer Register

Name: PERIPH_RNPR
Access: Read-write
31 30 29 28 27 26 25 24
RXNPTR
23 22 21 20 19 18 17 16
RXNPTR
15 14 13 12 11 10 9 8
RXNPTR
7 6 5 4 3 2 1 0
RXNPTR
• RXNPTR: Receive Next Pointer

RXNPTR contains next receive buffer address.
When a half duplex peripheral is connected to the PDC, RXNPTR = TXNPTR.
323
AT91SAM9G45
23.5.6 Receive Next Counter Register

Name: PERIPH_RNCR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RXNCTR
7 6 5 4 3 2 1 0
RXNCTR
• RXNCTR: Receive Next Counter

RXNCTR contains next receive buffer size.
When a half duplex peripheral is connected to the PDC, RXNCTR = TXNCTR.
324
AT91SAM9G45
23.5.7 Transmit Next Pointer Register

Name: PERIPH_TNPR
Access: Read-write
31 30 29 28 27 26 25 24
TXNPTR
23 22 21 20 19 18 17 16
TXNPTR
15 14 13 12 11 10 9 8
TXNPTR
7 6 5 4 3 2 1 0
TXNPTR
• TXNPTR: Transmit Next Pointer

TXNPTR contains next transmit buffer address.
When a half duplex peripheral is connected to the PDC, RXNPTR = TXNPTR.
325
23.5.8 Transmit Next Counter Register
Name: PERIPH_TNCR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
TXNCTR
7 6 5 4 3 2 1 0
TXNCTR
• TXNCTR: Transmit Counter Next

TXNCTR contains next transmit buffer size.
When a half duplex peripheral is connected to the PDC, RXNCTR = TXNCTR.
326 AT91SAM9G45
AT91SAM9G45
23.5.9 Transfer Control Register

Name: PERIPH_PTCR
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – TXTDIS TXTEN
7 6 5 4 3 2 1 0
– – – – – – RXTDIS RXTEN
• RXTEN: Receiver Transfer Enable

0 = No effect.
1 = Enables PDC receiver channel requests if RXTDIS is not set.
When a half duplex peripheral is connected to the PDC, enabling the receiver channel requests automatically disables the
transmitter channel requests. It is forbidden to set both TXTEN and RXTEN for a half duplex peripheral.
• RXTDIS: Receiver Transfer Disable

0 = No effect.
1 = Disables the PDC receiver channel requests.
When a half duplex peripheral is connected to the PDC, disabling the receiver channel requests also disables the transmit-
ter channel requests.
• TXTEN: Transmitter Transfer Enable

0 = No effect.
1 = Enables the PDC transmitter channel requests.
When a half duplex peripheral is connected to the PDC, it enables the transmitter channel requests only if RXTEN is not
set. It is forbidden to set both TXTEN and RXTEN for a half duplex peripheral.
• TXTDIS: Transmitter Transfer Disable

0 = No effect.
1 = Disables the PDC transmitter channel requests.
When a half duplex peripheral is connected to the PDC, disabling the transmitter channel requests disables the receiver
channel requests.
327
23.5.10 Transfer Status Register
Name: PERIPH_PTSR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – TXTEN
7 6 5 4 3 2 1 0
– – – – – – – RXTEN
• RXTEN: Receiver Transfer Enable

0 = PDC Receiver channel requests are disabled.
1 = PDC Receiver channel requests are enabled.
• TXTEN: Transmitter Transfer Enable

0 = PDC Transmitter channel requests are disabled.
1 = PDC Transmitter channel requests are enabled.
328 AT91SAM9G45
AT91SAM9G45
24. Clock Generator
24.1 Description
The Clock Generator User Interface is embedded within the Power Management Controller
Interface and is described in Section 25.11. However, the Clock Generator registers are named
CKGR_.

The Clock Generator is made up of:
• One Low Power 32768 Hz Slow Clock Oscillator with bypass mode
• One Low-Power RC oscillator
• One 12 MHz Main Oscillator, which can be bypassed
• One 400 to 800 MHz programmable PLLA, capable to provide the clock MCK to the
processor and to the peripherals. This PLL has an input divider to offer a wider range of
output frequencies from the 12 MHz input, the only limitation being the lowest input frequency
shall be higher or equal to 2 MHz.
The USB Device and Host HS Clocks are provided by a the dedicated UTMI PLL (UPLL)
embedded in the UTMI macro.
Figure 24-1. Clock Generator Block Diagram

Clock Generator
RCEN
On Chip
RC OSC
Slow Clock
SLCK
XIN32 Slow Clock
Oscillator
XOUT32 OSCSEL
OSC32EN
OSC32BYP
XIN
12M Main Main Clock
Oscillator MAINCK
XOUT
UPLL UPLLCK
PLLA and PLLA Clock

Divider PLLACK
Status Control
Power
Management
Controller
24.3 Slow Clock Crystal Oscillator

The Clock Generator integrates a 32,768 Hz low-power oscillator. The XIN32 and XOUT32 pins
must be connected to a 32,768 Hz crystal. Two external capacitors must be wired as shown in
Figure 24-2.
329
Figure 24-2. Typical Slow Clock Crystal Oscillator Connection
XIN32 XOUT32 GNDPLL
32,768 Hz
Crystal
24.4 Slow Clock RC Oscillator

The user has to take into account the possible drifts of the RC Oscillator. More details are given
in the section “DC Characteristics” of the product datasheet.
24.5 Slow Clock Selection

The AT91SAM9G45 slow clock can be generated either by an external 32,768 Hz crystal or by
the on-chip RC oscillator. The 32,768 Hz crystal oscillator can be bypassed by setting the bit
OSC32BYP to accept an external slow clock on XIN32.
The internal RC oscillator and the 32,768 Hz oscillator can be enabled by setting to 1, respec-
tively, RCEN bit and OSC32EN bit in the System Controller user interface. The OSCSEL
command selects the slow clock source.
Figure 24-3. Slow Clock Selection

Clock Generator
RCEN
On Chip
RC OSC
Slow Clock
SLCK
XIN32 Slow Clock

Oscillator
XOUT32
OSCSEL
OSC32EN
OSC32BYP
RCEN, OSC32EN,OSCSEL and OSC32BYP bits are located in the Slow Clock Control Register
(SCKCR) located at address 0xFFFFFD50 in the backed up part of the System Controller and so
are preserved while VDDBU is present.
After a VDDBU power on reset, the default configuration is RCEN=1, OSC32EN=0 and OSC-
SEL=0, allowing the system to start on the internal RC oscillator.
The programmer controls the slow clock switching by software and so must take precautions
during the switching phase.
330 AT91SAM9G45
AT91SAM9G45
24.5.1 Switch from Internal RC Oscillator to the 32768 Hz Crystal

To switch from internal RC oscillator to the 32768 Hz crystal, the programmer must execute the
following sequence:
• Switch the master clock to a source different from slow clock (PLLA or PLLB or Main
Oscillator) through the Power management Controller.
• Enable the 32768 Hz oscillator by setting the bit OSCEN to 1.
• Wait 32768 Hz Startup Time for clock stabilization (software loop)
• Switch from internal RC to 32768Hz by setting the bit OSCSEL to 1.
• Wait 5 slow clock cycles for internal resynchronization
• Disable the RC oscillator by setting the bit RCEN to 0.
24.5.2 Bypass the 32768 Hz Oscillator

The following step must be added to bypass the 32768Hz Oscillator.
• An external clock must be connected on XIN32.
• Enable the bypass path OSC32BYP bit set to 1.
• Disable the 32768 Hz oscillator by setting the bit OSC32EN to 0.
24.5.3 Switch from 32768 Hz Crystal to the Internal RC oscillator

The same procedure must be followed to switch from 32768Hz crystal to the internal RC
oscillator.
• Switch the master clock to a source different from slow clock (PLLA or PLLB or Main
Oscillator)
• Enable the internal RC oscillator by setting the bit RCEN to 1.
• Wait internal RC Startup Time for clock stabilization (software loop)
• Switch from 32768Hz oscillator to internal RC by setting the bit OSCSEL to 0
• Wait 5 slow clock cycles for internal resynchronization
• Disable the 32768Hz oscillator by setting the bit OSC32EN to 0
24.5.4 Slow Clock Configuration Register

Register Name:SCKCR
Address: 0xFFFFFD50
Access Type:Read/Write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – OSCSEL OSC32BYP OSC32EN RCEN
331
• RCEN: Internal RC
0: RC is disabled
1: RC is enabled
• OSC32EN: 32768 Hz oscillator

0: 32768Hz oscillator is disabled
1: 32768Hz oscillator is enabled
• OSC32BYP: 32768Hz oscillator bypass

0: 32768Hz oscillator is not bypassed
1: 32768Hz oscillator is bypassed, accept an external slow clock on XIN32
• OSCSEL: Slow clock selector

0: Slow clock is internal RC
1: Slow clock is 32768 Hz oscillator
332 AT91SAM9G45
AT91SAM9G45
24.6 Main Oscillator

The Main Oscillator is designed for a 12 MHz fundamental crystal. The 12 MHz is an input of the
PLLA and the UPLL used to generate the 480 MHz USB High Speed Clock (UPLLCK).
Figure 24-4 shows the Main Oscillator block diagram.
Figure 24-4. Main Oscillator Block Diagram
XIN
12M Main Main Clock
Oscillator MAINCK
XOUT
UPLL
UPLLCK
PLLA and PLLA Clock

Divider PLLACK
24.6.1 Main Oscillator Connections

The typical crystal connection is illustrated in Figure 24-5. For further details on the electrical
characteristics of the Main Oscillator, see the section “DC Characteristics” of the product
datasheet.
Figure 24-5. Typical Crystal Connection
XIN XOUT GND
24.6.2 Main Oscillator Startup Time

The startup time of the 12 MHz Main Oscillator is given in the section “DC Characteristics” of the
product datasheet.
24.6.3 Main Oscillator Control

To minimize the power required to start up the system, the main oscillator is disabled after reset
and slow clock is selected.
The software enables or disables the main oscillator so as to reduce power consumption by
clearing the MOSCEN bit in the Main Oscillator Register (CKGR_MOR).
When disabling the main oscillator by clearing the MOSCEN bit in CKGR_MOR, the MOSCS bit
in PMC_SR is automatically cleared, indicating the main clock is off.
When enabling the main oscillator, the user must initiate the main oscillator counter with a value
corresponding to the startup time of the oscillator. This startup time depends on the crystal fre-
quency connected to the main oscillator.
333
When the MOSCEN bit and the OSCOUNT are written in CKGR_MOR to enable the main oscil-
lator, the MOSCS bit in PMC_SR (Status Register) is cleared and the counter starts counting
down on the slow clock divided by 8 from the OSCOUNT value. Since the OSCOUNT value is
coded with 8 bits, the maximum startup time is about 62 ms.
When the counter reaches 0, the MOSCS bit is set, indicating that the main clock is valid. Set-
ting the MOSCS bit in PMC_IMR can trigger an interrupt to the processor.
24.6.4 Main Oscillator Bypass

The user can input a clock on the device instead of connecting a crystal. In this case, the user
has to provide the external clock signal on the XIN pin. The input characteristics of the XIN pin
under these conditions are given in the product electrical characteristics section. The program-
mer has to be sure to set the OSCBYPASS bit to 1 and the MOSCEN bit to 0 in the Main OSC
register (CKGR_MOR) for the external clock to operate properly.
24.7 Divider and PLLA Block

The PLLA embeds an input divider to increase the accuracy of the resulting clock signals. How-
ever, the user must respect the PLLA minimum input frequency when programming the divider.
The PLLA embeds also an output divisor by 2.
Figure 24-6 shows the block diagram of the divider and PLLA block.
Figure 24-6. Divider and PLLA Block Diagram

DIVA MULA OUTA PLLADIV2
MAINCK Divider /1 or /2
PLLA PLLACK
Divider
PLLACOUNT
PLLA
SLCK LOCKA
Counter
24.7.1 Divider and Phase Lock Loop Programming

The divider can be set between 1 and 255 in steps of 1. When a divider field (DIV) is set to 0, the
output of the corresponding divider and the PLL output is a continuous signal at level 0. On
reset, each DIV field is set to 0, thus the corresponding PLL input clock is set to 0.
The PLLA allows multiplication of the divider’s outputs. The PLLA clock signal has a frequency
that depends on the respective source signal frequency and on the parameters DIVA and
MULA. The factor applied to the source signal frequency is (MULA + 1)/DIVA. When MULA is
written to 0, the PLLA is disabled and its power consumption is saved. Re-enabling the PLLA
can be performed by writing a value higher than 0 in the MUL field.
Whenever the PLLA is re-enabled or one of its parameters is changed, the LOCKA bit in
PMC_SR is automatically cleared. The values written in the PLLACOUNT field in CKGR_PLLAR
are loaded in the PLLA counter. The PLLA counter then decrements at the speed of the Slow
334 AT91SAM9G45
AT91SAM9G45
Clock until it reaches 0. At this time, the LOCK bit is set in PMC_SR and can trigger an interrupt
to the processor. The user has to load the number of Slow Clock cycles required to cover the
PLLA transient time into the PLLACOUNT field.
The PLLA clock can be divided by 2 by writing the PLLADIV2 bit in PMC_MCKR register.
24.8 UTMI Bias and Phase Lock Loop Programming
The multiplier is built-in to 40 to obtain the USB High Speed 480 MHz.
UPLLEN
MAINCK UPLL UPLLCK
PLLCOUNT
UPLL
SLCK LOCKU
Counter
Whenever the UPLL is enabled by writing UPLLEN in CKGR_UCKR, the LOCKU bit in PMC_SR
is automatically cleared. The values written in the PLLCOUNT field in CKGR_UCKR are loaded
in the UPLL counter. The UPLL counter then decrements at the speed of the Slow Clock divided
by 8 until it reaches 0. At this time, the LOCKU bit is set in PMC_SR and can trigger an interrupt
to the processor. The user has to load the number of Slow Clock cycles required to cover the
UPLL transient time into the PLLCOUNT field. The BIAS, needed for High Speed operations, is
enabled by writing BIASEN in CKGR_UCKR once the PLL locked.
335
25. Power Management Controller (PMC)
25.1 Description
The Power Management Controller (PMC) optimizes power consumption by controlling all sys-
tem and user peripheral clocks. The PMC enables/disables the clock inputs to many of the
peripherals and the ARM Processor.

The Power Management Controller provides all the clock signals to the system.
PMC input clocks:
• UPLLCK: From UTMI PLL
• PLLACK From PLLA
• SLCK: slow clock from OSC32K or internal RC OSC
• MAINCK: from 12 MHz external oscillator
PMC output clocks
• Processor Clock PCK
• Master Clock MCK, in particular to the Matrix and the memory interfaces. The divider can be
1,2,3 or 4
• DDR system clock equal to 2xMCK
Note: DDR system clock is not available when Master Clock (MCK) equals Processor Clock (PCK).
• USB Host EHCI High speed clock (UPLLCK)
• USB OHCI clocks (UHP48M and UHP12M)
• Independent peripheral clocks, typically at the frequency of MCK
• Two programmable clock outputs: PCK0 and PCK1
This allows the software control of five flexible operating modes:
• Normal Mode, processor and peripherals running at a programmable frequency
• Idle Mode, processor stopped waiting for an interrupt
• Slow Clock Mode, processor and peripherals running at low frequency
• Standby Mode, mix of Idle and Backup Mode, peripheral running at low frequency, processor
stopped waiting for an interrupt
• Backup Mode, Main Power Supplies off, VDDBU powered by a battery
336 AT91SAM9G45
AT91SAM9G45
Figure 25-1. AT91SAM9G45 Power Management Controller Block Diagram

PLLACK
USBS
UHP48M USB
USBDIV+1 OHCI
UHP12M
/4
USB
EHCI
/1,/2
Processor PCK
UPLLCK Clock
Controller int
Divider
X /1 /1.5 /2 SysClk DDR

Prescaler /1 /2 /3 /4
/1,/2,/4,.../64 MCK
MAINCK
SLCK Peripherals
Master Clock Controller
Clock Controller periph_clk[..]
ON/OFF
SLCK ON/OFF
MAINCK Prescaler pck[..]
/1,/2,/4,...,/64
UPLLCK
Programmable Clock Controller
25.2.1 Main Application Modes

The Power Management Controller provides 3 main application modes.
25.2.1.1 Normal Mode

• PLLA and UPLL are running respectively at 400 MHz and 480 MHz
• USB Device High Speed and Host EHCI High Speed operations are allowed
• Full Speed OHCI input clock is UPLLCK, USBDIV is 9 (division by 10)
• System Input clock is PLLACK, PCK is 400 MHz
• MDIV is ‘11’, MCK is 133 MHz
• DDR2 can be used at up to 133 MHz
25.2.1.2 USB HS and LP-DDR Mode

• Only UPLL is running at 480 MHz, PLLA power consumption is saved
• USB Device High Speed and Host EHCI High Speed operations are allowed
• Full Speed OHCI input clock is UPLLCK, USBDIV is 9 (division by 10)
• System Input clock is UPLLCK, Prescaler is 2, PCK is 240 MHz
• Only LP-DDR can be used at up to 120 MHz
337
25.2.1.3 No UDP HS, UHP FS and DDR2 Mode
• Only PLLA is running at 384 MHz, UPLL power consumption is saved
• USB Device High Speed and Host EHCI High Speed operations are NOT allowed
• Full Speed OHCI input clock is PLLACK, USBDIV is 7 (division by 8)
• System Input clock is PLLACK, PCK is 384 MHz
• DDR2 can be used at up to 128 MHz
25.3 Master Clock Controller

The Master Clock Controller provides selection and division of the Master Clock (MCK). MCK is
the clock provided to all the peripherals and the memory controller.
The Master Clock is selected from one of the clocks provided by the Clock Generator. Selecting
the Slow Clock provides a Slow Clock signal to the whole device. Selecting the Main Clock
saves power consumption of the PLLA.
The Master Clock Controller is made up of a clock selector and a prescaler. It also contains a
Master Clock divider which allows the processor clock to be faster than the Master Clock.
The Master Clock selection is made by writing the CSS field (Clock Source Selection) in
PMC_MCKR (Master Clock Register). The prescaler supports the division by a power of 2 of the
selected clock between 1 and 64. The PRES field in PMC_MCKR programs the prescaler. The
Master Clock divider can be programmed through the MDIV field in PMC_MCKR.
Each time PMC_MCKR is written to define a new Master Clock, the MCKRDY bit is cleared in
PMC_SR. It reads 0 until the Master Clock is established. Then, the MCKRDY bit is set and can
trigger an interrupt to the processor. This feature is useful when switching from a high-speed
clock to a lower one to inform the software when the change is actually done.
Figure 25-2. Master Clock Controller

PMC_MCKR PMC_MCKR
CSS PRES PMC_MCKR
MDIV
SLCK
Master
MAINCK Master Clock
Clock MCK
Prescaler
PLLACK Divider
UPLLCK
Processor
To the Processor
Clock
Clock Controller (PCK)
Divider
25.4 Processor Clock Controller

The PMC features a Processor Clock Controller (PCK) that implements the Processor Idle
Mode. The Processor Clock can be disabled by writing the System Clock Disable Register
(PMC_SCDR). The status of this clock (at least for debug purpose) can be read in the System
Clock Status Register (PMC_SCSR).
338 AT91SAM9G45
AT91SAM9G45
The Processor Clock PCK is enabled after a reset and is automatically re-enabled by any
enabled interrupt. The Processor Idle Mode is achieved by disabling the Processor Clock, which
is automatically re-enabled by any enabled fast or normal interrupt, or by the reset of the
product.
When the Processor Clock is disabled, the current instruction is finished before the clock is
stopped, but this does not prevent data transfers from other masters of the system bus.
25.5 USB Device and Host clocks

The USB Device and Host High Speed ports clocks are controlled by the UDPHS and UHPHS
bits in PMC_PCER. To save power on this peripheral when they are is not used, the user can
set these bits in PMC_PCDR. The UDPHS and UHPHS bits PMC_PCSR gives the activity of
these clocks.
The PMC also provides the clocks UHP48M and UHP12M to the USB Host OHCI. The USB
Host OHCI clocks are controlled by the UHP bit in PMC_SCER. To save power on this periph-
eral when it is not used, the user can set the UHP bit in PMC_SCDR. The UHP bit in
PMC_SCSR gives the activity of this clock. The USB host OHCI requires both the 12/48 MHz
signal and the Master Clock. USBDIV field in PMC_USB register is to be programmed to 9 (divi-
sion by 10) for normal operations.
To save more power consumption user can stop UTMI PLL, in this case USB high-speed opera-
tions are not possible. Nevertheless, as the USB OHCI Input clock can be selected with USBS
bit (PLLA or UTMI PLL) in PMC_USB register, OHCI full-speed operation remain possible.
The user must program the USB OHCI Input Clock and the USBDIV divider in PMC_USB regis-
ter to generate a 48 MHz and a 12 MHz signal with an accuracy of ± 0.25%.
25.6 LP-DDR/DDR2 Clock

The Power Management Controller controls the clocks of the DDR memory. It provides SysClk
DDR internal clock. That clock is used by the DDR Controller to provide DDR control, data and
DDR clock signals.
The DDR clock can be enabled and disabled with DDRCK bit respectively in PMC_SCER and
PMC_SDER registers. At reset DDR clock is disabled to save power consumption.
The Input clock is the same as Master Clock. The Output SysClk DDR Clock is 2xMCK.
In the case MDIV = ‘00’, PCK = MCK and SysClk DDR and DDRCK clocks are not available.
If Input clock is PLLACK/PLLADIV2 the DDR Controller can drive DDR2 and LP-DDR at up to
133MHz with MDIV = ‘11’.
To save PLLA power consumption, the user can choose UPLLCK an Input clock for the system.
In this case the DDR Controller can drive LD-DDR at up to 120MHz.
25.7 Peripheral Clock Controller

The Power Management Controller controls the clocks of each embedded peripheral by the way
of the Peripheral Clock Controller. The user can individually enable and disable the Master
Clock on the peripherals by writing into the Peripheral Clock Enable (PMC_PCER) and Periph-
339
eral Clock Disable (PMC_PCDR) registers. The status of the peripheral clock activity can be
read in the Peripheral Clock Status Register (PMC_PCSR).
When a peripheral clock is disabled, the clock is immediately stopped. The peripheral clocks are
automatically disabled after a reset.
In order to stop a peripheral, it is recommended that the system software wait until the peripheral
has executed its last programmed operation before disabling the clock. This is to avoid data cor-
ruption or erroneous behavior of the system.
The bit number within the Peripheral Clock Control registers (PMC_PCER, PMC_PCDR, and
PMC_PCSR) is the Peripheral Identifier defined at the product level. Generally, the bit number
corresponds to the interrupt source number assigned to the peripheral.
25.8 Programmable Clock Output Controller

The PMC controls 2 signals to be output on external pins PCKx. Each signal can be indepen-
dently programmed via the PMC_PCKx registers.
PCKx can be independently selected between the Slow clock, the Master Clock, the
PLLACK/PLLADIV2, the UTMI PLL output and the main clock by writing the CSS and CSSMCK
fields in PMC_PCKx. Each output signal can also be divided by a power of 2 between 1 and 64
by writing the PRES (Prescaler) field in PMC_PCKx.
Each output signal can be enabled and disabled by writing 1 in the corresponding bit, PCKx of
PMC_SCER and PMC_SCDR, respectively. Status of the active programmable output clocks
are given in the PCKx bits of PMC_SCSR (System Clock Status Register).
Moreover, like the PCK, a status bit in PMC_SR indicates that the Programmable Clock is actu-
ally what has been programmed in the Programmable Clock registers.
As the Programmable Clock Controller does not manage with glitch prevention when switching
clocks, it is strongly recommended to disable the Programmable Clock before any configuration
change and to re-enable it after the change is actually performed.
25.9 Programming Sequence

1. Enabling the 12MHz Main Oscillator:
The main oscillator is enabled by setting the MOSCEN field in the CKGR_MOR register. In
some cases it may be advantageous to define a start-up time. This can be achieved by writ-
ing a value in the OSCOUNT field in the CKGR_MOR register.
Once this register has been correctly configured, the user must wait for MOSCS field in the
PMC_SR register to be set. This can be done either by polling the status register or by wait-
ing the interrupt line to be raised if the associated interrupt to MOSCS has been enabled in
the PMC_IER register.
2. Setting PLLA and divider:

All parameters needed to configure PLLA and the divider are located in the CKGR_PLLAR
register.
The DIVA field is used to control divider itself. A value between 0 and 255 can be pro-
grammed. Divider output is divider input divided by DIVA parameter. By default DIVA
parameter is set to 0 which means that divider is turned off.
340 AT91SAM9G45
AT91SAM9G45
The OUTA field is used to select the PLLA output frequency range.
The MULA field is the PLLA multiplier factor. This parameter can be programmed between 0
and 254. If MULA is set to 0, PLLA will be turned off, otherwise the PLLA output frequency is
PLLA input frequency multiplied by (MULA + 1).
The PLLACOUNT field specifies the number of slow clock cycles before LOCKA bit is set in
the PMC_SR register after CKGR_PLLAR register has been written.
Once the PMC_PLLAR register has been written, the user must wait for the LOCKA bit to be
set in the PMC_SR register. This can be done either by polling the status register or by wait-
ing the interrupt line to be raised if the associated interrupt to LOCKA has been enabled in
the PMC_IER register. All parameters in CKGR_PLLAR can be programmed in a single write
operation. If at some stage one of the following parameters, MULA, DIVA is modified,
LOCKA bit will go low to indicate that PLLA is not ready yet. When PLLA is locked, LOCKA
will be set again. The user is constrained to wait for LOCKA bit to be set before using the
PLLA output clock.
Code Example:
write_register(CKGR_PLLAR,0x00040805)
If PLLA and divider are enabled, the PLLA input clock is the main clock. PLLA output clock is
PLLA input clock multiplied by 5. Once CKGR_PLLAR has been written, LOCKA bit will be
set after eight slow clock cycles.
3. Setting Bias and High Speed PLL (UPLL) for UTMI

The UTMI PLL is enabled by setting the UPLLEN field in the CKGR_UCKR register. The
UTMI Bias must is enabled by setting the BIASEN field in the CKGR_UCKR register in the
same time. In some cases it may be advantageous to define a start-up time. This can be
achieved by writing a value in the PLLCOUNT field in the CKGR_UCKR register.
Once this register has been correctly configured, the user must wait for LOCKU field in the
PMC_SR register to be set. This can be done either by polling the status register or by wait-
ing the interrupt line to be raised if the associated interrupt to LOCKU has been enabled in
the PMC_IER register.
4. Selection of Master Clock and Processor Clock

The Master Clock and the Processor Clock are configurable via the PMC_MCKR register.
The CSS field is used to select the clock source of the Master Clock and Processor Clock
dividers. By default, the selected clock source is slow clock.
The PRES field is used to control the Master/Processor Clock prescaler. The user can
choose between different values (1, 2, 4, 8, 16, 32, 64). Prescaler output is the selected clock
source divided by PRES parameter. By default, PRES parameter is set to 1 which means
that the input clock of the Master Clock and Processor Clock dividers is equal to slow clock.
The MDIV field is used to control the Master Clock divider. It is possible to choose between
different values (0, 1, 2, 3). The Master Clock output is Master/Processor Clock Prescaler
output divided by 1, 2, 4 or 3, depending on the value programmed in MDIV.
341
The PLLADIV2 field is used to control the PLLA Clock divider. It is possible to choose
between different values (0, 1). The PMC PLLA Clock input is divided by 1 or 2, depending
on the value programmed in PLLADIV2.
By default, MDIV and PLLLADIV2 are set to 0, which indicates that Processor Clock is equal
to the Master Clock.
Once the PMC_MCKR register has been written, the user must wait for the MCKRDY bit to
be set in the PMC_SR register. This can be done either by polling the status register or by
waiting for the interrupt line to be raised if the associated interrupt to MCKRDY has been
enabled in the PMC_IER register.
The PMC_MCKR register must not be programmed in a single write operation. The preferred
programming sequence for the PMC_MCKR register is as follows:
• If a new value for CSS field corresponds to PLLA Clock,
– Program the PRES field in the PMC_MCKR register.
– Wait for the MCKRDY bit to be set in the PMC_SR register.
– Program the CSS field in the PMC_MCKR register.
• If a new value for CSS field corresponds to Main Clock or Slow Clock,
– Program the CSS field in the PMC_MCKR register.
– Program the PRES field in the PMC_MCKR register.
If at some stage one of the following parameters, CSS or PRES, is modified, the MCKRDY
bit will go low to indicate that the Master Clock and the Processor Clock are not ready yet.
The user must wait for MCKRDY bit to be set again before using the Master and Processor
Clocks.
Note: IF PLLA clock was selected as the Master Clock and the user decides to modify it by writing in
CKGR_PLLAR, the MCKRDY flag will go low while PLLA is unlocked. Once PLLA is locked again,
LOCK goes high and MCKRDY is set.
While PLLA is unlocked, the Master Clock selection is automatically changed to Main Clock. For
further information, see Section 25.10.2. “Clock Switching Waveforms” on page 345.
Code Example:
write_register(PMC_MCKR,0x00000001)
wait (MCKRDY=1)
write_register(PMC_MCKR,0x00000011)
wait (MCKRDY=1)
The Master Clock is main clock divided by 16.

The Processor Clock is the Master Clock.
5. Selection of Programmable clocks
Programmable clocks are controlled via registers; PMC_SCER, PMC_SCDR and
PMC_SCSR.
342 AT91SAM9G45
AT91SAM9G45
Programmable clocks can be enabled and/or disabled via the PMC_SCER and PMC_SCDR
registers. Depending on the system used, 2 programmable clocks can be enabled or dis-
abled. The PMC_SCSR provides a clear indication as to which Programmable clock is
enabled. By default all Programmable clocks are disabled.
PMC_PCKx registers are used to configure programmable clocks.
The CSS and CSSMCK fields are used to select the programmable clock divider source. Five
clock options are available: main clock, slow clock, master clock, PLLACK, UPLLCK. By
default, the clock source selected is slow clock.
The PRES field is used to control the programmable clock prescaler. It is possible to choose
between different values (1, 2, 4, 8, 16, 32, 64). Programmable clock output is prescaler input
divided by PRES parameter. By default, the PRES parameter is set to 1 which means that
master clock is equal to slow clock.
Once the PMC_PCKx register has been programmed, The corresponding programmable
clock must be enabled and the user is constrained to wait for the PCKRDYx bit to be set in
the PMC_SR register. This can be done either by polling the status register or by waiting the
interrupt line to be raised if the associated interrupt to PCKRDYx has been enabled in the
PMC_IER register. All parameters in PMC_PCKx can be programmed in a single write
operation.
If the CSS and PRES parameters are to be modified, the corresponding programmable clock
must be disabled first. The parameters can then be modified. Once this has been done, the
user must re-enable the programmable clock and wait for the PCKRDYx bit to be set.
Code Example:
write_register(PMC_PCK0,0x00000015)
Programmable clock 0 is main clock divided by 32.

6. Enabling Peripheral Clocks
Once all of the previous steps have been completed, the peripheral clocks can be enabled
and/or disabled via registers PMC_PCER and PMC_PCDR.
Depending on the system used, 19 peripheral clocks can be enabled or disabled. The
PMC_PCSR provides a clear view as to which peripheral clock is enabled.
Note: Each enabled peripheral clock corresponds to Master Clock.
Code Examples:
write_register(PMC_PCER,0x00000110)
Peripheral clocks 4 and 8 are enabled.
write_register(PMC_PCDR,0x00000010)
343
Peripheral clock 4 is disabled.
25.10 Clock Switching Details
25.10.1 Master Clock Switching Timings

Table 25-1 gives the worst case timings required for the Master Clock to switch from one
selected clock to another one. This is in the event that the prescaler is de-activated. When the
prescaler is activated, an additional time of 64 clock cycles of the new selected clock has to be
added.
Table 25-1. Clock Switching Timings (Worst Case)

From Main Clock SLCK PLLA Clock
To
3 x PLLA Clock +
4 x SLCK +
Main Clock – 4 x SLCK +
2.5 x Main Clock
1 x Main Clock
0.5 x Main Clock + 3 x PLLA Clock +
SLCK –
4.5 x SLCK 5 x SLCK
0.5 x Main Clock +
2.5 x PLLA Clock + 2.5 x PLLA Clock +
4 x SLCK +
PLLA Clock 5 x SLCK + 4 x SLCK +
PLLACOUNT x SLCK +
PLLACOUNT x SLCK PLLACOUNT x SLCK
2.5 x PLLAx Clock
344 AT91SAM9G45
AT91SAM9G45
25.10.2 Clock Switching Waveforms
Figure 25-3. Switch Master Clock from Slow Clock to PLLA Clock
Slow Clock
PLL Clock
LOCK
MCKRDY
Master Clock
Write PMC_MCKR
Figure 25-4. Switch Master Clock from Main Clock to Slow Clock
Slow Clock
Main Clock
MCKRDY
Master Clock
Write PMC_MCKR
345
Figure 25-5. Change PLLA Programming
Main Clock
PLL Clock
LOCK
MCKRDY
Master Clock
Main Clock
Write CKGR_PLLR
Figure 25-6. Programmable Clock Output Programming
PLL Clock
PCKRDY
PCKx Output
Write PMC_PCKx PLL Clock is selected
Write PMC_SCER
PCKx is enabled
Write PMC_SCDR PCKx is disabled
346 AT91SAM9G45
AT91SAM9G45
25.11 Power Management Controllerr (PMC) User Interface

Offset Register Name Access Reset Value
0x0000 System Clock Enable Register PMC_SCER Write-only –
0x0004 System Clock Disable Register PMC_SCDR Write-only –
0x0008 System Clock Status Register PMC _SCSR Read-only 0x01
0x000C Reserved – – –
0x0010 Peripheral Clock Enable Register PMC _PCER Write-only –
0x0014 Peripheral Clock Disable Register PMC_PCDR Write-only –
0x0018 Peripheral Clock Status Register PMC_PCSR Read-only 0x0
0x001C UTMI Clock Register CKGR_UCKR Read/Write 0x1020 0800
0x0020 Main Oscillator Register CKGR_MOR Read/Write 0x0
0x0024 Main Clock Frequency Register CKGR_MCFR Read-only 0x0
0x0028 PLLA Register CKGR_PLLAR Read/Write 0x3F00
0x0030 Master Clock Register PMC_MCKR Read/Write 0x0
0x0038 USB Clock Register PMC_USB Read/Write 0x0
0x0040 Programmable Clock 0 Register PMC_PCK0 Read/Write 0x0
0x0044 Programmable Clock 1 Register PMC_PCK1 Read/Write 0x0
0x0048-0x005C Reserved – – –
0x0060 Interrupt Enable Register PMC_IER Write-only --
0x0064 Interrupt Disable Register PMC_IDR Write-only --
0x0068 Status Register PMC_SR Read-only 0x08
0x006C Interrupt Mask Register PMC_IMR Read-only 0x0
0x0070 - 0x007C Reserved – – –
0x0080 PLL Charge Pump Current Register PMC_PLLICPR Write-only 0x0
347
25.11.1 PMC System Clock Enable Register
Register Name:PMC_SCER
Address: 0xFFFFFC00
Access Type:Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – PCK1 PCK0
7 6 5 4 3 2 1 0
– UHP – – – DDRCK – –
• DDRCK: DDR Clock Enable

0 = No effect.
1 = Enables the DDR clock.
• UHP: USB Host OHCI Clocks Enable
0 = No effect.
1 = Enables the UHP48M and UHP12M OHCI clocks.
• PCKx: Programmable Clock x Output Enable

0 = No effect.
1 = Enables the corresponding Programmable Clock output.
348 AT91SAM9G45
AT91SAM9G45
25.11.2 PMC System Clock Disable Register

Register Name:PMC_SCDR
Address: 0xFFFFFC04
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – PCK1 PCK0
7 6 5 4 3 2 1 0
– UHP – – – DDRCK – PCK
• PCK: Processor Clock Disable

0 = No effect.
1 = Disables the Processor clock. This is used to enter the processor in Idle Mode.
• DDRCK: DDR Clock Disable

0 = No effect.
1 = Disables the DDR clock.
• UHP: USB Host OHCI Clock Disable
0 = No effect.
1 = Disables the UHP48M and UHP12M OHCI clocks.
• PCKx: Programmable Clock x Output Disable

0 = No effect.
1 = Disables the corresponding Programmable Clock output.
349
25.11.3 PMC System Clock Status Register
Register Name:PMC_SCSR
Address: 0xFFFFFC08
Access Type:Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – PCK1 PCK0
7 6 5 4 3 2 1 0
– UHP – – – DDRCK – PCK
• PCK: Processor Clock Status

0 = The Processor clock is disabled.
1 = The Processor clock is enabled.
• DDRCK: DDR Clock Status

0 = The DDR clock is disabled.
1 = The DDR clock is enabled.
• UHP: USB Host Port Clock Status
0 = The UHP48M and UHP12M OHCI clocks are disabled.
1 = The UHP48M and UHP12M OHCI clocks are enabled.
• PCKx: Programmable Clock x Output Status

0 = The corresponding Programmable Clock output is disabled.
1 = The corresponding Programmable Clock output is enabled.
350 AT91SAM9G45
AT91SAM9G45
25.11.4 PMC Peripheral Clock Enable Register

Register Name:PMC_PCER
Address: 0xFFFFFC10
31 30 29 28 27 26 25 24
PID31 PID30 PID29 PID28 PID27 PID26 PID25 PID24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
PID7 PID6 PID5 PID4 PID3 PID2 - -
• PIDx: Peripheral Clock x Enable

0 = No effect.
1 = Enables the corresponding peripheral clock.
Note: PID2 to PID31 refer to identifiers as defined in the section “Peripheral Identifiers” in the product datasheet.
Note: Programming the control bits of the Peripheral ID that are not implemented has no effect on the behavior of the PMC.
351
25.11.5 PMC Peripheral Clock Disable Register
Register Name:PMC_PCDR
Address: 0xFFFFFC14
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
PID7 PID6 PID5 PID4 PID3 PID2 - -
• PIDx: Peripheral Clock x Disable

0 = No effect.
1 = Disables the corresponding peripheral clock.
352 AT91SAM9G45
AT91SAM9G45
25.11.6 PMC Peripheral Clock Status Register

Register Name:PMC_PCSR
Address: 0xFFFFFC18
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
PID7 PID6 PID5 PID4 PID3 PID2 – –
• PIDx: Peripheral Clock x Status

0 = The corresponding peripheral clock is disabled.
1 = The corresponding peripheral clock is enabled.
353
25.11.7 PMC UTMI Clock Configuration Register
Register Name:CKGR_UCKR
Address: 0xFFFFFC1C
31 30 29 28 27 26 25 24
BIASCOUNT – – – BIASEN
23 22 21 20 19 18 17 16
PLLCOUNT – – – UPLLEN
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – –
• UPLLEN: UTMI PLL Enable

0 = The UTMI PLL is disabled.
1 = The UTMI PLL is enabled.
When UPLLEN is set, the LOCKU flag is set once the UTMI PLL startup time is achieved.
• PLLCOUNT: UTMI PLL Start-up Time

Specifies the number of Slow Clock cycles multiplied by 8 for the UTMI PLL start-up time.
• BIASEN: UTMI BIAS Enable

0 = The UTMI BIAS is disabled.
1 = The UTMI BIAS is enabled.
• BIASCOUNT: UTMI BIAS Start-up Time
Specifies the number of Slow Clock cycles for the UTMI BIAS start-up time.
354 AT91SAM9G45
AT91SAM9G45
25.11.8 PMC Clock Generator Main Oscillator Register

Register Name:CKGR_MOR
Address: 0xFFFFFC20
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
OSCOUNT
7 6 5 4 3 2 1 0
– – – – – – OSCBYPASS MOSCEN
• MOSCEN: Main Oscillator Enable

A crystal must be connected between XIN and XOUT.
0 = The Main Oscillator is disabled.
1 = The Main Oscillator is enabled. OSCBYPASS must be set to 0.
When MOSCEN is set, the MOSCS flag is set once the Main Oscillator startup time is achieved.
• OSCBYPASS: Oscillator Bypass
0 = No effect.
1 = The Main Oscillator is bypassed. MOSCEN must be set to 0. An external clock must be connected on XIN.
When OSCBYPASS is set, the MOSCS flag in PMC_SR is automatically set.
Clearing MOSCEN and OSCBYPASS bits allows resetting the MOSCS flag.
• OSCOUNT: Main Oscillator Start-up Time
Specifies the number of Slow Clock cycles multiplied by 8 for the Main Oscillator start-up time.
355
25.11.9 PMC Clock Generator Main Clock Frequency Register
Register Name:CKGR_MCFR
Address: 0xFFFFFC24
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – MAINRDY
15 14 13 12 11 10 9 8
MAINF
7 6 5 4 3 2 1 0
MAINF
• MAINF: Main Clock Frequency

Gives the number of Main Clock cycles within 16 Slow Clock periods.
• MAINRDY: Main Clock Ready
0 = MAINF value is not valid or the Main Oscillator is disabled.
1 = The Main Oscillator has been enabled previously and MAINF value is available.
356 AT91SAM9G45
AT91SAM9G45
25.11.10 PMC Clock Generator PLLA Register

Register Name:CKGR_PLLAR
Address: 0xFFFFFC28
31 30 29 28 27 26 25 24
– – 1 – – – – –
23 22 21 20 19 18 17 16
MULA
15 14 13 12 11 10 9 8
OUTA PLLACOUNT
7 6 5 4 3 2 1 0
DIVA
Possible limitations on PLL input frequencies and multiplier factors should be checked before using the PMC.
Warning: Bit 29 must always be set to 1 when programming the CKGR_PLLAR register.
• DIVA: Divider A
DIVA Divider Selected

0 Divider output is 0
1 Divider is bypassed
2 - 255 Divider output is the selected clock divided by DIVA.
• PLLACOUNT: PLLA Counter

Specifies the number of slow clock cycles before the LOCKA bit is set in PMC_SR after CKGR_PLLAR is written.
• OUTA: PLLA Clock Frequency Range
To optimize clock performance, this field must be programmed as specified in “PLL Characteristics” in the Electrical Char-
acteristics section of the product datasheet.
• MULA: PLLA Multiplier
0 = The PLLA is deactivated.
1 up to 254 = The PLLA Clock frequency is the PLLA input frequency multiplied by MULA+ 1.
357
25.11.11 PMC USB Clock Register
Register Name:PMC_USB
Address: 0xFFFFFC38
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – USBDIV
7 6 5 4 3 2 1 0
– – – – – – – USBS
• USBS: USB OHCI Input clock selection

0 = USB Clock Input is PLLA
1 = USB Clock Input is UPLL
• USBDIV: Divider for USB OHCI Clock.
USB Clock is Input clock divided by USBDIV+1
358 AT91SAM9G45
AT91SAM9G45
25.11.12 PMC Master Clock Register

Register Name:PMC_MCKR
Address: 0xFFFFFC30
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – PLLADIV2 – – MDIV
7 6 5 4 3 2 1 0
– – – PRES CSS
• CSS: Master/Processor Clock Source Selection
CSS Clock Source Selection

0 0 Slow Clock is selected.
0 1 Main Clock is selected.
1 0 PLLA Output clock is selected.
1 1 UPLL Output clock is selected.
• PRES: Master/Processor Clock Prescaler
Master/Processor Clock Dividers

PRES Input Clock
0 0 0 Selected clock
0 0 1 Selected clock divided by 2
1 1 1 Reserved
359
• MDIV: Master Clock Division
MDIV Master Clock Division

Master Clock is Prescaler Output Clock divided by 1.
0 0
Warning: SysClk DDR and DDRCK are not available.
0 1 SysClk DDR is equal to 2 x MCK.
DDRCK is equal to MCK.
• PLLADIV2: PLLA divisor by 2
PLLADIV2 PLLA Clock Division

0 PLLA clock frequency is divided by 1.
1 PLLA clock frequency is divided by 2.
360 AT91SAM9G45
AT91SAM9G45
25.11.13 PMC Programmable Clock Register

Register Name:PMC_PCKx
Address: 0xFFFFFC40
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – SLCKMCK
7 6 5 4 3 2 1 0
– – – PRES CSS
• CSS: Master Clock Selection
CSS Clock Source Selection

0 0 Slow Clock is selected.
0 1 Main Clock is selected.
1 0 PLLACK/PLLADIV2 is selected.
1 1 UPLLCK is selected.
• PRES: Programmable Clock Prescaler
PRES Programmable Clock

0 0 0 Selected clock
1 1 1 Reserved
• SLCKMCK: Slow Clock or Master Clock Selection

0 = Slow clock is selected
1 = Master clock is selected
If CSS field is not programmed to ‘00’ no Output clock will be provided.
361
25.11.14 PMC Interrupt Enable Register
Register Name:PMC_IER
Address: 0xFFFFFC60
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – PCKRDY1 PCKRDY0
7 6 5 4 3 2 1 0
– LOCKU – – MCKRDY – LOCKA MOSCS
• MOSCS: Main Oscillator Status Interrupt Enable
• LOCKA: PLL Lock Interrupt Enable
• MCKRDY: Master Clock Ready Interrupt Enable
• LOCKU: UTMI PLL Lock Interrupt Enable
• PCKRDYx: Programmable Clock Ready x Interrupt Enable

0 = No effect.
1 = Enables the corresponding interrupt.
362 AT91SAM9G45
AT91SAM9G45
25.11.15 PMC Interrupt Disable Register

Register Name:PMC_IDR
Address: 0xFFFFFC64
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• MOSCS: Main Oscillator Status Interrupt Disable
• LOCKA: PLLA Lock Interrupt Disable
• MCKRDY: Master Clock Ready Interrupt Disable
• LOCKU: UTMI PLL Lock Interrupt Disable
• PCKRDYx: Programmable Clock Ready x Interrupt Disable

0 = No effect.
1 = Disables the corresponding interrupt.
363
25.11.16 PMC Status Register
Register Name:PMC_SR
Address: 0xFFFFFC68
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• MOSCS: MOSCS Flag Status

0 = Main oscillator is not stabilized.
1 = Main oscillator is stabilized.
• LOCKA: PLLA Lock Status

0 = PLLA is not locked
1 = PLLA is locked.
• MCKRDY: Master Clock Status

0 = Master Clock is not ready.
1 = Master Clock is ready.
• LOCKU: UPLL Lock Status

0 = UPLL is not locked
1 = UPLL is locked.
• PCKRDYx: Programmable Clock Ready Status

0 = Programmable Clock x is not ready.
1 = Programmable Clock x is ready.
364 AT91SAM9G45
AT91SAM9G45
25.11.17 PMC Interrupt Mask Register

Register Name:PMC_IMR
Address: 0xFFFFFC6C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• MOSCS: Main Oscillator Status Interrupt Mask
• LOCKA: PLLA Lock Interrupt Mask
• MCKRDY: Master Clock Ready Interrupt Mask
• LOCKU: UTMI PLL Lock Interrupt Mask
• PCKRDYx: Programmable Clock Ready x Interrupt Mask

0 = The corresponding interrupt is enabled.
1 = The corresponding interrupt is disabled.
365
25.11.18 PLL Charge Pump Current Register
Register Name:PMC_PLLICPR
Address: 0xFFFFFC80
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – ICPLLA
• ICPLLA: Charge Pump Current

To optimize clock performance, this field must be programmed as specified in “PLL A Characteristics” in the Electrical Char-
acteristics section of the product datasheet.
366 AT91SAM9G45
AT91SAM9G45
26. Advanced Interrupt Controller (AIC)
26.1 Description
The Advanced Interrupt Controller (AIC) is an 8-level priority, individually maskable, vectored
interrupt controller, providing handling of up to thirty-two interrupt sources. It is designed to sub-
stantially reduce the software and real-time overhead in handling internal and external
interrupts.
The AIC drives the nFIQ (fast interrupt request) and the nIRQ (standard interrupt request) inputs
of an ARM processor. Inputs of the AIC are either internal peripheral interrupts or external inter-
rupts coming from the product's pins.
The 8-level Priority Controller allows the user to define the priority for each interrupt source, thus
permitting higher priority interrupts to be serviced even if a lower priority interrupt is being
treated.
Internal interrupt sources can be programmed to be level sensitive or edge triggered. External
interrupt sources can be programmed to be positive-edge or negative-edge triggered or high-
level or low-level sensitive.
The fast forcing feature redirects any internal or external interrupt source to provide a fast inter-
rupt rather than a normal interrupt.

• Controls the interrupt lines (nIRQ and nFIQ) of the ARM Processor
• Thirty-two individually maskable and vectored interrupt sources
– Source 0 is reserved for the Fast Interrupt Input (FIQ)
– Source 1 is reserved for system peripherals (PIT, RTT, PMC, DBGU, etc.)
– Programmable Edge-triggered or Level-sensitive Internal Sources
– Programmable Positive/Negative Edge-triggered or High/Low Level-sensitive
• One External Sources plus the Fast Interrupt signal
• 8-level Priority Controller
– Drives the Normal Interrupt of the processor
– Handles priority of the interrupt sources 1 to 31
– Higher priority interrupts can be served during service of lower priority interrupt
• Vectoring
– Optimizes Interrupt Service Routine Branch and Execution
– One 32-bit Vector Register per interrupt source
– Interrupt Vector Register reads the corresponding current Interrupt Vector
• Protect Mode
– Easy debugging by preventing automatic operations when protect modes are
enabled
• Fast Forcing
– Permits redirecting any normal interrupt source on the Fast Interrupt of the
processor
367
26.3 Block Diagram
FIQ AIC
ARM
IRQ0-IRQn Processor
Up to
Thirty-two nFIQ
Embedded Sources
PeripheralEE
Embedded nIRQ
Peripheral
Embedded
Peripheral
APB
26.4 Application Block Diagram
Figure 26-2. Description of the Application Block
OS-based Applications
Standalone
Applications OS Drivers RTOS Drivers
Hard Real Time Tasks
General OS Interrupt Handler
Advanced Interrupt Controller
External Peripherals
Embedded Peripherals
(External Interrupts)
26.5 AIC Detailed Block Diagram
Figure 26-3. AIC Detailed Block Diagram
Advanced Interrupt Controller ARM

FIQ Processor
Fast nFIQ
PIO External
Interrupt
Controller Source
Controller
Input
Stage
nIRQ
IRQ0-IRQn
PIOIRQ Fast Interrupt
Priority Processor
Forcing
Controller Clock
Internal
Source
Input Power
Embedded Stage Management
Peripherals Controller
User Interface Wake Up
APB
368 AT91SAM9G45
AT91SAM9G45
26.6 I/O Line Description

Pin Name Pin Description Type
FIQ Fast Interrupt Input
IRQ0 - IRQn Interrupt 0 - Interrupt n Input
26.7.1 I/O Lines

The interrupt signals FIQ and IRQ0 to IRQn are normally multiplexed through the PIO control-
lers. Depending on the features of the PIO controller used in the product, the pins must be
programmed in accordance with their assigned interrupt function. This is not applicable when
the PIO controller used in the product is transparent on the input path.
Table 26-2. I/O Lines

Instance Signal I/O Line Peripheral
AIC FIQ PD19 B
AIC IRQ PD18 B

The Advanced Interrupt Controller is continuously clocked. The Power Management Controller
has no effect on the Advanced Interrupt Controller behavior.
The assertion of the Advanced Interrupt Controller outputs, either nIRQ or nFIQ, wakes up the
ARM processor while it is in Idle Mode. The General Interrupt Mask feature enables the AIC to
wake up the processor without asserting the interrupt line of the processor, thus providing syn-
chronization of the processor on an event.
26.7.3 Interrupt Sources

The Interrupt Source 0 is always located at FIQ. If the product does not feature an FIQ pin, the
Interrupt Source 0 cannot be used.
The Interrupt Source 1 is always located at System Interrupt. This is the result of the OR-wiring
of the system peripheral interrupt lines. When a system interrupt occurs, the service routine
must first distinguish the cause of the interrupt. This is performed by reading successively the
status registers of the above mentioned system peripherals.
The interrupt sources 2 to 31 can either be connected to the interrupt outputs of an embedded
user peripheral or to external interrupt lines. The external interrupt lines can be connected
directly, or through the PIO Controller.
The PIO Controllers are considered as user peripherals in the scope of interrupt handling.
Accordingly, the PIO Controller interrupt lines are connected to the Interrupt Sources 2 to 31.
The peripheral identification defined at the product level corresponds to the interrupt source
number (as well as the bit number controlling the clock of the peripheral). Consequently, to sim-
plify the description of the functional operations and the user interface, the interrupt sources are
named FIQ, SYS, and PID2 to PID31.
369
26.8.1 Interrupt Source Control
26.8.1.1 Interrupt Source Mode

The Advanced Interrupt Controller independently programs each interrupt source. The SRC-
TYPE field of the corresponding AIC_SMR (Source Mode Register) selects the interrupt
condition of each source.
The internal interrupt sources wired on the interrupt outputs of the embedded peripherals can be
programmed either in level-sensitive mode or in edge-triggered mode. The active level of the
internal interrupts is not important for the user.
The external interrupt sources can be programmed either in high level-sensitive or low level-sen-
sitive modes, or in positive edge-triggered or negative edge-triggered modes.
26.8.1.2 Interrupt Source Enabling

Each interrupt source, including the FIQ in source 0, can be enabled or disabled by using the
command registers; AIC_IECR (Interrupt Enable Command Register) and AIC_IDCR (Interrupt
Disable Command Register). This set of registers conducts enabling or disabling in one instruc-
tion. The interrupt mask can be read in the AIC_IMR register. A disabled interrupt does not affect
servicing of other interrupts.
26.8.1.3 Interrupt Clearing and Setting

All interrupt sources programmed to be edge-triggered (including the FIQ in source 0) can be
individually set or cleared by writing respectively the AIC_ISCR and AIC_ICCR registers. Clear-
ing or setting interrupt sources programmed in level-sensitive mode has no effect.
The clear operation is perfunctory, as the software must perform an action to reinitialize the
“memorization” circuitry activated when the source is programmed in edge-triggered mode.
However, the set operation is available for auto-test or software debug purposes. It can also be
used to execute an AIC-implementation of a software interrupt.
The AIC features an automatic clear of the current interrupt when the AIC_IVR (Interrupt Vector
Register) is read. Only the interrupt source being detected by the AIC as the current interrupt is
affected by this operation. (See “Priority Controller” on page 373.) The automatic clear reduces
the operations required by the interrupt service routine entry code to reading the AIC_IVR. Note
that the automatic interrupt clear is disabled if the interrupt source has the Fast Forcing feature
enabled as it is considered uniquely as a FIQ source. (For further details, See “Fast Forcing” on
page 377.)
The automatic clear of the interrupt source 0 is performed when AIC_FVR is read.
26.8.1.4 Interrupt Status

For each interrupt, the AIC operation originates in AIC_IPR (Interrupt Pending Register) and its
mask in AIC_IMR (Interrupt Mask Register). AIC_IPR enables the actual activity of the sources,
whether masked or not.
The AIC_ISR register reads the number of the current interrupt (see “Priority Controller” on page
373) and the register AIC_CISR gives an image of the signals nIRQ and nFIQ driven on the
processor.
Each status referred to above can be used to optimize the interrupt handling of the systems.
370 AT91SAM9G45
AT91SAM9G45
26.8.1.5 Internal Interrupt Source Input Stage
Figure 26-4. Internal Interrupt Source Input Stage

AIC_SMRI
(SRCTYPE)
Level/ AIC_IPR
Source i Edge
AIC_IMR
Fast Interrupt Controller
or
Priority Controller
Edge
Detector AIC_IECR
Set Clear
FF
AIC_ISCR
AIC_ICCR
AIC_IDCR
26.8.1.6 External Interrupt Source Input Stage
Figure 26-5. External Interrupt Source Input Stage

AIC_SMRi
High/Low SRCTYPE
Level/ AIC_IPR
Edge
AIC_IMR
Source i
Fast Interrupt Controller
or
Priority Controller
Pos./Neg. AIC_IECR
Edge
Detector FF
Set Clear
AIC_ISCR AIC_IDCR
AIC_ICCR
371
26.8.2 Interrupt Latencies
Global interrupt latencies depend on several parameters, including:
• The time the software masks the interrupts.
• Occurrence, either at the processor level or at the AIC level.
• The execution time of the instruction in progress when the interrupt occurs.
• The treatment of higher priority interrupts and the resynchronization of the hardware signals.
This section addresses only the hardware resynchronizations. It gives details of the latency
times between the event on an external interrupt leading in a valid interrupt (edge or level) or the
assertion of an internal interrupt source and the assertion of the nIRQ or nFIQ line on the pro-
cessor. The resynchronization time depends on the programming of the interrupt source and on
its type (internal or external). For the standard interrupt, resynchronization times are given
assuming there is no higher priority in progress.
The PIO Controller multiplexing has no effect on the interrupt latencies of the external interrupt
sources.
26.8.2.1 External Interrupt Edge Triggered Source
Figure 26-6. External Interrupt Edge Triggered Source

MCK
IRQ or FIQ
(Positive Edge)
IRQ or FIQ
(Negative Edge)
nIRQ
Maximum IRQ Latency = 4 Cycles
nFIQ
Maximum FIQ Latency = 4 Cycles
26.8.2.2 External Interrupt Level Sensitive Source
Figure 26-7. External Interrupt Level Sensitive Source
MCK
IRQ or FIQ
(High Level)
IRQ or FIQ
(Low Level)
nIRQ
Maximum IRQ
Latency = 3 Cycles
nFIQ
Maximum FIQ
Latency = 3 cycles
372 AT91SAM9G45
AT91SAM9G45
26.8.2.3 Internal Interrupt Edge Triggered Source
Figure 26-8. Internal Interrupt Edge Triggered Source
MCK
nIRQ
Maximum IRQ Latency = 4.5 Cycles
Peripheral Interrupt
Becomes Active
26.8.2.4 Internal Interrupt Level Sensitive Source
Figure 26-9. Internal Interrupt Level Sensitive Source
MCK
nIRQ
Maximum IRQ Latency = 3.5 Cycles
Peripheral Interrupt
Becomes Active
26.8.3 Normal Interrupt
26.8.3.1 Priority Controller

An 8-level priority controller drives the nIRQ line of the processor, depending on the interrupt
conditions occurring on the interrupt sources 1 to 31 (except for those programmed in Fast
Forcing).
Each interrupt source has a programmable priority level of 7 to 0, which is user-definable by writ-
ing the PRIOR field of the corresponding AIC_SMR (Source Mode Register). Level 7 is the
highest priority and level 0 the lowest.
As soon as an interrupt condition occurs, as defined by the SRCTYPE field of the AIC_SMR
(Source Mode Register), the nIRQ line is asserted. As a new interrupt condition might have hap-
pened on other interrupt sources since the nIRQ has been asserted, the priority controller
determines the current interrupt at the time the AIC_IVR (Interrupt Vector Register) is read. The
read of AIC_IVR is the entry point of the interrupt handling which allows the AIC to consider
that the interrupt has been taken into account by the software.
The current priority level is defined as the priority level of the current interrupt.
If several interrupt sources of equal priority are pending and enabled when the AIC_IVR is read,
the interrupt with the lowest interrupt source number is serviced first.
373
The nIRQ line can be asserted only if an interrupt condition occurs on an interrupt source with a
higher priority. If an interrupt condition happens (or is pending) during the interrupt treatment in
progress, it is delayed until the software indicates to the AIC the end of the current service by
writing the AIC_EOICR (End of Interrupt Command Register). The write of AIC_EOICR is the
exit point of the interrupt handling.
26.8.3.2 Interrupt Nesting

The priority controller utilizes interrupt nesting in order for the high priority interrupt to be handled
during the service of lower priority interrupts. This requires the interrupt service routines of the
lower interrupts to re-enable the interrupt at the processor level.
When an interrupt of a higher priority happens during an already occurring interrupt service rou-
tine, the nIRQ line is re-asserted. If the interrupt is enabled at the core level, the current
execution is interrupted and the new interrupt service routine should read the AIC_IVR. At this
time, the current interrupt number and its priority level are pushed into an embedded hardware
stack, so that they are saved and restored when the higher priority interrupt servicing is finished
and the AIC_EOICR is written.
The AIC is equipped with an 8-level wide hardware stack in order to support up to eight interrupt
nestings pursuant to having eight priority levels.
26.8.3.3 Interrupt Vectoring

The interrupt handler addresses corresponding to each interrupt source can be stored in the reg-
isters AIC_SVR1 to AIC_SVR31 (Source Vector Register 1 to 31). When the processor reads
AIC_IVR (Interrupt Vector Register), the value written into AIC_SVR corresponding to the cur-
rent interrupt is returned.
This feature offers a way to branch in one single instruction to the handler corresponding to the
current interrupt, as AIC_IVR is mapped at the absolute address 0xFFFF F100 and thus acces-
sible from the ARM interrupt vector at address 0x0000 0018 through the following instruction:
LDR PC,[PC,# -&F20]
When the processor executes this instruction, it loads the read value in AIC_IVR in its program
counter, thus branching the execution on the correct interrupt handler.
This feature is often not used when the application is based on an operating system (either real
time or not). Operating systems often have a single entry point for all the interrupts and the first
task performed is to discern the source of the interrupt.
However, it is strongly recommended to port the operating system on AT91 products by support-
ing the interrupt vectoring. This can be performed by defining all the AIC_SVR of the interrupt
source to be handled by the operating system at the address of its interrupt handler. When doing
so, the interrupt vectoring permits a critical interrupt to transfer the execution on a specific very
fast handler and not onto the operating system’s general interrupt handler. This facilitates the
support of hard real-time tasks (input/outputs of voice/audio buffers and software peripheral han-
dling) to be handled efficiently and independently of the application running under an operating
system.
26.8.3.4 Interrupt Handlers

This section gives an overview of the fast interrupt handling sequence when using the AIC. It is
assumed that the programmer understands the architecture of the ARM processor, and espe-
cially the processor interrupt modes and the associated status bits.
374 AT91SAM9G45
AT91SAM9G45
It is assumed that:
1. The Advanced Interrupt Controller has been programmed, AIC_SVR registers are
loaded with corresponding interrupt service routine addresses and interrupts are
enabled.
2. The instruction at the ARM interrupt exception vector address is required to work with
the vectoring
LDR PC, [PC, # -&F20]
When nIRQ is asserted, if the bit “I” of CPSR is 0, the sequence is as follows:
1. The CPSR is stored in SPSR_irq, the current value of the Program Counter is loaded in
the Interrupt link register (R14_irq) and the Program Counter (R15) is loaded with 0x18.
In the following cycle during fetch at address 0x1C, the ARM core adjusts R14_irq, dec-
rementing it by four.
2. The ARM core enters Interrupt mode, if it has not already done so.
3. When the instruction loaded at address 0x18 is executed, the program counter is
loaded with the value read in AIC_IVR. Reading the AIC_IVR has the following effects:
– Sets the current interrupt to be the pending and enabled interrupt with the highest
priority. The current level is the priority level of the current interrupt.
– De-asserts the nIRQ line on the processor. Even if vectoring is not used, AIC_IVR
must be read in order to de-assert nIRQ.
– Automatically clears the interrupt, if it has been programmed to be edge-triggered.
– Pushes the current level and the current interrupt number on to the stack.
– Returns the value written in the AIC_SVR corresponding to the current interrupt.
4. The previous step has the effect of branching to the corresponding interrupt service
routine. This should start by saving the link register (R14_irq) and SPSR_IRQ. The link
register must be decremented by four when it is saved if it is to be restored directly into
the program counter at the end of the interrupt. For example, the instruction SUB PC,
LR, #4 may be used.
5. Further interrupts can then be unmasked by clearing the “I” bit in CPSR, allowing re-
assertion of the nIRQ to be taken into account by the core. This can happen if an inter-
rupt with a higher priority than the current interrupt occurs.
6. The interrupt handler can then proceed as required, saving the registers that will be
used and restoring them at the end. During this phase, an interrupt of higher priority
than the current level will restart the sequence from step 1.
Note: If the interrupt is programmed to be level sensitive, the source of the interrupt must be cleared dur-
ing this phase.
7. The “I” bit in CPSR must be set in order to mask interrupts before exiting to ensure that
the interrupt is completed in an orderly manner.
8. The End of Interrupt Command Register (AIC_EOICR) must be written in order to indi-
cate to the AIC that the current interrupt is finished. This causes the current level to be
popped from the stack, restoring the previous current level if one exists on the stack. If
another interrupt is pending, with lower or equal priority than the old current level but
with higher priority than the new current level, the nIRQ line is re-asserted, but the inter-
rupt sequence does not immediately start because the “I” bit is set in the core.
SPSR_irq is restored. Finally, the saved value of the link register is restored directly into
the PC. This has the effect of returning from the interrupt to whatever was being exe-
cuted before, and of loading the CPSR with the stored SPSR, masking or unmasking
the interrupts depending on the state saved in SPSR_irq.
375
Note: The “I” bit in SPSR is significant. If it is set, it indicates that the ARM core was on the verge of
masking an interrupt when the mask instruction was interrupted. Hence, when SPSR is restored,
the mask instruction is completed (interrupt is masked).
26.8.4 Fast Interrupt
26.8.4.1 Fast Interrupt Source

The interrupt source 0 is the only source which can raise a fast interrupt request to the processor
except if fast forcing is used. The interrupt source 0 is generally connected to a FIQ pin of the
product, either directly or through a PIO Controller.
26.8.4.2 Fast Interrupt Control

The fast interrupt logic of the AIC has no priority controller. The mode of interrupt source 0 is
programmed with the AIC_SMR0 and the field PRIOR of this register is not used even if it reads
what has been written. The field SRCTYPE of AIC_SMR0 enables programming the fast inter-
rupt source to be positive-edge triggered or negative-edge triggered or high-level sensitive or
low-level sensitive
Writing 0x1 in the AIC_IECR (Interrupt Enable Command Register) and AIC_IDCR (Interrupt
Disable Command Register) respectively enables and disables the fast interrupt. The bit 0 of
AIC_IMR (Interrupt Mask Register) indicates whether the fast interrupt is enabled or disabled.
26.8.4.3 Fast Interrupt Vectoring

The fast interrupt handler address can be stored in AIC_SVR0 (Source Vector Register 0). The
value written into this register is returned when the processor reads AIC_FVR (Fast Vector Reg-
ister). This offers a way to branch in one single instruction to the interrupt handler, as AIC_FVR
is mapped at the absolute address 0xFFFF F104 and thus accessible from the ARM fast inter-
rupt vector at address 0x0000 001C through the following instruction:
LDR PC,[PC,# -&F20]
When the processor executes this instruction it loads the value read in AIC_FVR in its program
counter, thus branching the execution on the fast interrupt handler. It also automatically per-
forms the clear of the fast interrupt source if it is programmed in edge-triggered mode.
26.8.4.4 Fast Interrupt Handlers

This section gives an overview of the fast interrupt handling sequence when using the AIC. It is
assumed that the programmer understands the architecture of the ARM processor, and espe-
cially the processor interrupt modes and associated status bits.
Assuming that:
1. The Advanced Interrupt Controller has been programmed, AIC_SVR0 is loaded with
the fast interrupt service routine address, and the interrupt source 0 is enabled.
2. The Instruction at address 0x1C (FIQ exception vector address) is required to vector
the fast interrupt:
LDR PC, [PC, # -&F20]
3. The user does not need nested fast interrupts.
When nFIQ is asserted, if the bit “F” of CPSR is 0, the sequence is:
1. The CPSR is stored in SPSR_fiq, the current value of the program counter is loaded in
the FIQ link register (R14_FIQ) and the program counter (R15) is loaded with 0x1C. In
376 AT91SAM9G45
AT91SAM9G45
the following cycle, during fetch at address 0x20, the ARM core adjusts R14_fiq, decre-
menting it by four.
2. The ARM core enters FIQ mode.
3. When the instruction loaded at address 0x1C is executed, the program counter is
loaded with the value read in AIC_FVR. Reading the AIC_FVR has effect of automati-
cally clearing the fast interrupt, if it has been programmed to be edge triggered. In this
case only, it de-asserts the nFIQ line on the processor.
4. The previous step enables branching to the corresponding interrupt service routine. It is
not necessary to save the link register R14_fiq and SPSR_fiq if nested fast interrupts
are not needed.
5. The Interrupt Handler can then proceed as required. It is not necessary to save regis-
ters R8 to R13 because FIQ mode has its own dedicated registers and the user R8 to
R13 are banked. The other registers, R0 to R7, must be saved before being used, and
restored at the end (before the next step). Note that if the fast interrupt is programmed
to be level sensitive, the source of the interrupt must be cleared during this phase in
order to de-assert the interrupt source 0.
6. Finally, the Link Register R14_fiq is restored into the PC after decrementing it by four
(with instruction SUB PC, LR, #4 for example). This has the effect of returning from
the interrupt to whatever was being executed before, loading the CPSR with the SPSR
and masking or unmasking the fast interrupt depending on the state saved in the
SPSR.
Note: The “F” bit in SPSR is significant. If it is set, it indicates that the ARM core was just about to mask
FIQ interrupts when the mask instruction was interrupted. Hence when the SPSR is restored, the
interrupted instruction is completed (FIQ is masked).
Another way to handle the fast interrupt is to map the interrupt service routine at the address of
the ARM vector 0x1C. This method does not use the vectoring, so that reading AIC_FVR must
be performed at the very beginning of the handler operation. However, this method saves the
execution of a branch instruction.
26.8.4.5 Fast Forcing

The Fast Forcing feature of the advanced interrupt controller provides redirection of any normal
Interrupt source on the fast interrupt controller.
Fast Forcing is enabled or disabled by writing to the Fast Forcing Enable Register (AIC_FFER)
and the Fast Forcing Disable Register (AIC_FFDR). Writing to these registers results in an
update of the Fast Forcing Status Register (AIC_FFSR) that controls the feature for each inter-
nal or external interrupt source.
When Fast Forcing is disabled, the interrupt sources are handled as described in the previous
pages.
When Fast Forcing is enabled, the edge/level programming and, in certain cases, edge detec-
tion of the interrupt source is still active but the source cannot trigger a normal interrupt to the
processor and is not seen by the priority handler.
If the interrupt source is programmed in level-sensitive mode and an active level is sampled,
Fast Forcing results in the assertion of the nFIQ line to the core.
If the interrupt source is programmed in edge-triggered mode and an active edge is detected,
Fast Forcing results in the assertion of the nFIQ line to the core.
The Fast Forcing feature does not affect the Source 0 pending bit in the Interrupt Pending Reg-
ister (AIC_IPR).
377
The FIQ Vector Register (AIC_FVR) reads the contents of the Source Vector Register 0
(AIC_SVR0), whatever the source of the fast interrupt may be. The read of the FVR does not
clear the Source 0 when the fast forcing feature is used and the interrupt source should be
cleared by writing to the Interrupt Clear Command Register (AIC_ICCR).
All enabled and pending interrupt sources that have the fast forcing feature enabled and that are
programmed in edge-triggered mode must be cleared by writing to the Interrupt Clear Command
Register. In doing so, they are cleared independently and thus lost interrupts are prevented.
The read of AIC_IVR does not clear the source that has the fast forcing feature enabled.
The source 0, reserved to the fast interrupt, continues operating normally and becomes one of
the Fast Interrupt sources.
Figure 26-10. Fast Forcing

Source 0 _ FIQ AIC_IPR
Input Stage
Automatic Clear AIC_IMR

nFIQ
Read FVR if Fast Forcing is

disabled on Sources 1 to 31.
AIC_FFSR
Source n AIC_IPR
Input Stage
Priority
Manager
Automatic Clear AIC_IMR nIRQ
Read IVR if Source n is the current interrupt

and if Fast Forcing is disabled on Source n.
26.8.5 Protect Mode

The Protect Mode permits reading the Interrupt Vector Register without performing the associ-
ated automatic operations. This is necessary when working with a debug system. When a
debugger, working either with a Debug Monitor or the ARM processor's ICE, stops the applica-
tions and updates the opened windows, it might read the AIC User Interface and thus the IVR.
This has undesirable consequences:
• If an enabled interrupt with a higher priority than the current one is pending, it is stacked.
• If there is no enabled pending interrupt, the spurious vector is returned.
In either case, an End of Interrupt command is necessary to acknowledge and to restore the
context of the AIC. This operation is generally not performed by the debug system as the debug
system would become strongly intrusive and cause the application to enter an undesired state.
This is avoided by using the Protect Mode. Writing PROT in AIC_DCR (Debug Control Register)
at 0x1 enables the Protect Mode.
When the Protect Mode is enabled, the AIC performs interrupt stacking only when a write access
is performed on the AIC_IVR. Therefore, the Interrupt Service Routines must write (arbitrary
data) to the AIC_IVR just after reading it. The new context of the AIC, including the value of the
Interrupt Status Register (AIC_ISR), is updated with the current interrupt only when AIC_IVR is
written.
378 AT91SAM9G45
AT91SAM9G45
An AIC_IVR read on its own (e.g., by a debugger), modifies neither the AIC context nor the
AIC_ISR. Extra AIC_IVR reads perform the same operations. However, it is recommended to
not stop the processor between the read and the write of AIC_IVR of the interrupt service routine
to make sure the debugger does not modify the AIC context.
To summarize, in normal operating mode, the read of AIC_IVR performs the following opera-
tions within the AIC:
1. Calculates active interrupt (higher than current or spurious).
2. Determines and returns the vector of the active interrupt.
3. Memorizes the interrupt.
4. Pushes the current priority level onto the internal stack.
5. Acknowledges the interrupt.
However, while the Protect Mode is activated, only operations 1 to 3 are performed when
AIC_IVR is read. Operations 4 and 5 are only performed by the AIC when AIC_IVR is written.
Software that has been written and debugged using the Protect Mode runs correctly in Normal
Mode without modification. However, in Normal Mode the AIC_IVR write has no effect and can
be removed to optimize the code.
26.8.6 Spurious Interrupt

The Advanced Interrupt Controller features protection against spurious interrupts. A spurious
interrupt is defined as being the assertion of an interrupt source long enough for the AIC to
assert the nIRQ, but no longer present when AIC_IVR is read. This is most prone to occur when:
• An external interrupt source is programmed in level-sensitive mode and an active level occurs
for only a short time.
• An internal interrupt source is programmed in level sensitive and the output signal of the
corresponding embedded peripheral is activated for a short time. (As in the case for the
Watchdog.)
• An interrupt occurs just a few cycles before the software begins to mask it, thus resulting in a
pulse on the interrupt source.
The AIC detects a spurious interrupt at the time the AIC_IVR is read while no enabled interrupt
source is pending. When this happens, the AIC returns the value stored by the programmer in
AIC_SPU (Spurious Vector Register). The programmer must store the address of a spurious
interrupt handler in AIC_SPU as part of the application, to enable an as fast as possible return to
the normal execution flow. This handler writes in AIC_EOICR and performs a return from
interrupt.
26.8.7 General Interrupt Mask

The AIC features a General Interrupt Mask bit to prevent interrupts from reaching the processor.
Both the nIRQ and the nFIQ lines are driven to their inactive state if the bit GMSK in AIC_DCR
(Debug Control Register) is set. However, this mask does not prevent waking up the processor if
it has entered Idle Mode. This function facilitates synchronizing the processor on a next event
and, as soon as the event occurs, performs subsequent operations without having to handle an
interrupt. It is strongly recommended to use this mask with caution.
379
26.9 Advanced Interrupt Controller (AIC) User Interface
26.9.1 Base Address

The AIC is mapped at the address 0xFFFF F000. It has a total 4-Kbyte addressing space. This permits the vectoring fea-
ture, as the PC-relative load/store instructions of the ARM processor support only a ± 4-Kbyte offset.

0x00 Source Mode Register 0 AIC_SMR0 Read-write 0x0
0x04 Source Mode Register 1 AIC_SMR1 Read-write 0x0
--- --- --- --- ---
0x7C Source Mode Register 31 AIC_SMR31 Read-write 0x0
0x80 Source Vector Register 0 AIC_SVR0 Read-write 0x0
0x84 Source Vector Register 1 AIC_SVR1 Read-write 0x0
--- --- --- --- ---
0xFC Source Vector Register 31 AIC_SVR31 Read-write 0x0
0x100 Interrupt Vector Register AIC_IVR Read-only 0x0
0x104 FIQ Interrupt Vector Register AIC_FVR Read-only 0x0
0x108 Interrupt Status Register AIC_ISR Read-only 0x0
(2)
0x10C Interrupt Pending Register AIC_IPR Read-only 0x0(1)
0x110 Interrupt Mask Register(2) AIC_IMR Read-only 0x0
0x114 Core Interrupt Status Register AIC_CISR Read-only 0x0
0x118 - 0x11C Reserved --- --- ---
(2)
0x120 Interrupt Enable Command Register AIC_IECR Write-only ---
(2)
0x124 Interrupt Disable Command Register AIC_IDCR Write-only ---
0x128 Interrupt Clear Command Register(2) AIC_ICCR Write-only ---
(2)
0x12C Interrupt Set Command Register AIC_ISCR Write-only ---
0x130 End of Interrupt Command Register AIC_EOICR Write-only ---
0x134 Spurious Interrupt Vector Register AIC_SPU Read-write 0x0
0x138 Debug Control Register AIC_DCR Read-write 0x0
0x13C Reserved --- --- ---
(2)
0x140 Fast Forcing Enable Register AIC_FFER Write-only ---
(2)
0x144 Fast Forcing Disable Register AIC_FFDR Write-only ---
0x148 Fast Forcing Status Register(2) AIC_FFSR Read-only 0x0
0x14C - 0x1E0 Reserved --- --- ---
0x1EC - 0x1FC Reserved
Notes: 1. The reset value of this register depends on the level of the external interrupt source. All other sources are cleared at reset,
thus not pending.
2. PID2...PID31 bit fields refer to the identifiers as defined in the Peripheral Identifiers Section of the product datasheet.
3. Values in the Version Register vary with the version of the IP block implementation.
380 AT91SAM9G45
AT91SAM9G45
26.9.2 AIC Source Mode Register

Register Name: AIC_SMR0..AIC_SMR31
Address: 0xFFFFF000
Reset Value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– SRCTYPE – – PRIOR
• PRIOR: Priority Level

Programs the priority level for all sources except FIQ source (source 0).
The priority level can be between 0 (lowest) and 7 (highest).
The priority level is not used for the FIQ in the related SMR register AIC_SMRx.
• SRCTYPE: Interrupt Source Type

The active level or edge is not programmable for the internal interrupt sources.
SRCTYPE Internal Interrupt Sources External Interrupt Sources

0 0 High level Sensitive Low level Sensitive
0 1 Positive edge triggered Negative edge triggered
1 0 High level Sensitive High level Sensitive
1 1 Positive edge triggered Positive edge triggered
381
26.9.3 AIC Source Vector Register
Register Name: AIC_SVR0..AIC_SVR31
Address: 0xFFFFF080
Reset Value: 0x0
31 30 29 28 27 26 25 24
VECTOR
23 22 21 20 19 18 17 16
VECTOR
15 14 13 12 11 10 9 8
VECTOR
7 6 5 4 3 2 1 0
VECTOR
• VECTOR: Source Vector

The user may store in these registers the addresses of the corresponding handler for each interrupt source.
26.9.4 AIC Interrupt Vector Register

Register Name: AIC_IVR
Address: 0xFFFFF100
Reset Value: 0x0
31 30 29 28 27 26 25 24
IRQV
23 22 21 20 19 18 17 16
IRQV
15 14 13 12 11 10 9 8
IRQV
7 6 5 4 3 2 1 0
IRQV
• IRQV: Interrupt Vector Register

The Interrupt Vector Register contains the vector programmed by the user in the Source Vector Register corresponding to
the current interrupt.
The Source Vector Register is indexed using the current interrupt number when the Interrupt Vector Register is read.
When there is no current interrupt, the Interrupt Vector Register reads the value stored in AIC_SPU.
382 AT91SAM9G45
AT91SAM9G45
26.9.5 AIC FIQ Vector Register

Register Name: AIC_FVR
Address: 0xFFFFF104
Reset Value: 0x0
31 30 29 28 27 26 25 24
FIQV
23 22 21 20 19 18 17 16
FIQV
15 14 13 12 11 10 9 8
FIQV
7 6 5 4 3 2 1 0
FIQV
• FIQV: FIQ Vector Register

The FIQ Vector Register contains the vector programmed by the user in the Source Vector Register 0. When there is no
fast interrupt, the FIQ Vector Register reads the value stored in AIC_SPU.
26.9.6 AIC Interrupt Status Register

Register Name: AIC_ISR
Address: 0xFFFFF108
Reset Value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – IRQID
• IRQID: Current Interrupt Identifier

The Interrupt Status Register returns the current interrupt source number.
383
26.9.7 AIC Interrupt Pending Register
Register Name: AIC_IPR
Address: 0xFFFFF10C
Reset Value: 0x0
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
PID7 PID6 PID5 PID4 PID3 PID2 SYS FIQ
• FIQ, SYS, PID2-PID31: Interrupt Pending

0 = Corresponding interrupt is not pending.
1 = Corresponding interrupt is pending.
26.9.8 AIC Interrupt Mask Register

Register Name: AIC_IMR
Address: 0xFFFFF110
Reset Value: 0x0
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• FIQ, SYS, PID2-PID31: Interrupt Mask

0 = Corresponding interrupt is disabled.
1 = Corresponding interrupt is enabled.
384 AT91SAM9G45
AT91SAM9G45
26.9.9 AIC Core Interrupt Status Register

Register Name: AIC_CISR
Address: 0xFFFFF114
Reset Value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – NIRQ NFIQ
• NFIQ: NFIQ Status

0 = nFIQ line is deactivated.
1 = nFIQ line is active.
• NIRQ: NIRQ Status

0 = nIRQ line is deactivated.
1 = nIRQ line is active.
26.9.10 AIC Interrupt Enable Command Register

Register Name: AIC_IECR
Address: 0xFFFFF120
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• FIQ, SYS, PID2-PID31: Interrupt Enable

0 = No effect.
1 = Enables corresponding interrupt.
385
26.9.11 AIC Interrupt Disable Command Register
Register Name: AIC_IDCR
Address: 0xFFFFF124
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• FIQ, SYS, PID2-PID31: Interrupt Disable

0 = No effect.
1 = Disables corresponding interrupt.
26.9.12 AIC Interrupt Clear Command Register

Register Name: AIC_ICCR
Address: 0xFFFFF128
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• FIQ, SYS, PID2-PID31: Interrupt Clear

0 = No effect.
1 = Clears corresponding interrupt.
386 AT91SAM9G45
AT91SAM9G45
26.9.13 AIC Interrupt Set Command Register

Register Name: AIC_ISCR
Address: 0xFFFFF12C
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• FIQ, SYS, PID2-PID31: Interrupt Set

0 = No effect.
1 = Sets corresponding interrupt.
26.9.14 AIC End of Interrupt Command Register

Register Name: AIC_EOICR
Address: 0xFFFFF130
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – –
The End of Interrupt Command Register is used by the interrupt routine to indicate that the interrupt treatment is complete.
Any value can be written because it is only necessary to make a write to this register location to signal the end of interrupt
treatment.
387
26.9.15 AIC Spurious Interrupt Vector Register
Register Name: AIC_SPU
Address: 0xFFFFF134
Reset Value: 0x0
31 30 29 28 27 26 25 24
SIVR
23 22 21 20 19 18 17 16
SIVR
15 14 13 12 11 10 9 8
SIVR
7 6 5 4 3 2 1 0
SIVR
• SIVR: Spurious Interrupt Vector Register

The user may store the address of a spurious interrupt handler in this register. The written value is returned in AIC_IVR in
case of a spurious interrupt and in AIC_FVR in case of a spurious fast interrupt.
26.9.16 AIC Debug Control Register

Register Name: AIC_DCR
Address: 0xFFFFF138
Reset Value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – GMSK PROT
• PROT: Protection Mode

0 = The Protection Mode is disabled.
1 = The Protection Mode is enabled.
• GMSK: General Mask

0 = The nIRQ and nFIQ lines are normally controlled by the AIC.
1 = The nIRQ and nFIQ lines are tied to their inactive state.
388 AT91SAM9G45
AT91SAM9G45
26.9.17 AIC Fast Forcing Enable Register

Register Name: AIC_FFER
Address: 0xFFFFF140
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
PID7 PID6 PID5 PID4 PID3 PID2 SYS –
• SYS, PID2-PID31: Fast Forcing Enable

0 = No effect.
1 = Enables the fast forcing feature on the corresponding interrupt.
26.9.18 AIC Fast Forcing Disable Register

Register Name: AIC_FFDR
Address: 0xFFFFF144
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• SYS, PID2-PID31: Fast Forcing Disable

0 = No effect.
1 = Disables the Fast Forcing feature on the corresponding interrupt.
389
AT91SAM9G45
26.9.19 AIC Fast Forcing Status Register

Register Name: AIC_FFSR
Address: 0xFFFFF148
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• SYS, PID2-PID31: Fast Forcing Status

0 = The Fast Forcing feature is disabled on the corresponding interrupt.
1 = The Fast Forcing feature is enabled on the corresponding interrupt.
390
AT91SAM9G45
27. Debug Unit (DBGU)
27.1 Description
The Debug Unit provides a single entry point from the processor for access to all the debug
capabilities of Atmel’s ARM-based systems.
The Debug Unit features a two-pin UART that can be used for several debug and trace purposes
and offers an ideal medium for in-situ programming solutions and debug monitor communica-
tions. The Debug Unit two-pin UART can be used stand-alone for general purpose serial
communication. Moreover, the association with two peripheral data controller channels permits
packet handling for these tasks with processor time reduced to a minimum.
The Debug Unit also makes the Debug Communication Channel (DCC) signals provided by the
In-circuit Emulator of the ARM processor visible to the software. These signals indicate the sta-
tus of the DCC read and write registers and generate an interrupt to the ARM processor, making
possible the handling of the DCC under interrupt control.
Chip Identifier registers permit recognition of the device and its revision. These registers inform
as to the sizes and types of the on-chip memories, as well as the set of embedded peripherals.
Finally, the Debug Unit features a Force NTRST capability that enables the software to decide
whether to prevent access to the system via the In-circuit Emulator. This permits protection of
the code, stored in ROM.

• Composed of two functions
– Two-pin UART
– Debug Communication Channel (DCC) support
• Two-pin UART
– Implemented features are 100% compatible with the standard Atmel USART
– Independent receiver and transmitter with a common programmable Baud Rate
Generator
– Even, Odd, Mark or Space Parity Generation
– Parity, Framing and Overrun Error Detection
– Automatic Echo, Local Loopback and Remote Loopback Channel Modes
– Support for two PDC channels with connection to receiver and transmitter
• Debug Communication Channel Support
– Offers visibility of an interrupt trigger from COMMRX and COMMTX signals from the
ARM Processor’s ICE Interface
391
27.3 Block Diagram
Figure 27-1. Debug Unit Functional Block Diagram
Peripheral
Bridge
Peripheral DMA Controller
APB Debug Unit

DTXD
Transmit
Power MCK Baud Rate Parallel
Management Generator Input/
Controller Output
Receive
DRXD
COMMRX
DCC
Chip ID
ARM COMMTX Handler
Processor
nTRST
ICE
Interrupt
Access dbgu_irq
Control
Handler
Power-on
Reset
force_ntrst
Table 27-1. Debug Unit Pin Description

Pin Name Description Type
Figure 27-2. Debug Unit Application Example
Boot Program Debug Monitor Trace Manager
Debug Unit
RS232 Drivers
Programming Tool Debug Console Trace Console
392 AT91SAM9G45
AT91SAM9G45
27.4.1 I/O Lines

Depending on product integration, the Debug Unit pins may be multiplexed with PIO lines. In this
case, the programmer must first configure the corresponding PIO Controller to enable I/O lines
operations of the Debug Unit.

DBGU DRXD PB12 A
DBGU DTXD PB13 A

Depending on product integration, the Debug Unit clock may be controllable through the Power
Management Controller. In this case, the programmer must first configure the PMC to enable the
Debug Unit clock. Usually, the peripheral identifier used for this purpose is 1.
27.4.3 Interrupt Source

Depending on product integration, the Debug Unit interrupt line is connected to one of the inter-
rupt sources of the Advanced Interrupt Controller. Interrupt handling requires programming of
the AIC before configuring the Debug Unit. Usually, the Debug Unit interrupt line connects to the
interrupt source 1 of the AIC, which may be shared with the real-time clock, the system timer
interrupt lines and other system peripheral interrupts, as shown in Figure 27-1. This sharing
requires the programmer to determine the source of the interrupt when the source 1 is triggered.
27.5 UART Operations

The Debug Unit operates as a UART, (asynchronous mode only) and supports only 8-bit charac-
ter handling (with parity). It has no clock pin.
The Debug Unit's UART is made up of a receiver and a transmitter that operate independently,
and a common baud rate generator. Receiver timeout and transmitter time guard are not imple-
mented. However, all the implemented features are compatible with those of a standard USART.
27.5.1 Baud Rate Generator

The baud rate generator provides the bit period clock named baud rate clock to both the receiver
and the transmitter.
The baud rate clock is the master clock divided by 16 times the value (CD) written in
DBGU_BRGR (Baud Rate Generator Register). If DBGU_BRGR is set to 0, the baud rate clock
is disabled and the Debug Unit's UART remains inactive. The maximum allowable baud rate is
Master Clock divided by 16. The minimum allowable baud rate is Master Clock divided by (16 x
65536).
MCK
Baud Rate = ----------------------
16 × CD
393
AT91SAM9G45
Figure 27-3. Baud Rate Generator

CD
CD
MCK 16-bit Counter

OUT
>1
1 Divide Baud Rate
by 16 Clock
0 0
Receiver
Sampling Clock
27.5.2 Receiver
27.5.2.1 Receiver Reset, Enable and Disable

After device reset, the Debug Unit receiver is disabled and must be enabled before being used.
The receiver can be enabled by writing the control register DBGU_CR with the bit RXEN at 1. At
this command, the receiver starts looking for a start bit.
The programmer can disable the receiver by writing DBGU_CR with the bit RXDIS at 1. If the
receiver is waiting for a start bit, it is immediately stopped. However, if the receiver has already
detected a start bit and is receiving the data, it waits for the stop bit before actually stopping its
operation.
The programmer can also put the receiver in its reset state by writing DBGU_CR with the bit
RSTRX at 1. In doing so, the receiver immediately stops its current operations and is disabled,
whatever its current state. If RSTRX is applied when data is being processed, this data is lost.
27.5.2.2 Start Detection and Data Sampling

The Debug Unit only supports asynchronous operations, and this affects only its receiver. The
Debug Unit receiver detects the start of a received character by sampling the DRXD signal until
it detects a valid start bit. A low level (space) on DRXD is interpreted as a valid start bit if it is
detected for more than 7 cycles of the sampling clock, which is 16 times the baud rate. Hence, a
space that is longer than 7/16 of the bit period is detected as a valid start bit. A space which is
7/16 of a bit period or shorter is ignored and the receiver continues to wait for a valid start bit.
When a valid start bit has been detected, the receiver samples the DRXD at the theoretical mid-
point of each bit. It is assumed that each bit lasts 16 cycles of the sampling clock (1-bit period)
so the bit sampling point is eight cycles (0.5-bit period) after the start of the bit. The first sampling
point is therefore 24 cycles (1.5-bit periods) after the falling edge of the start bit was detected.
Each subsequent bit is sampled 16 cycles (1-bit period) after the previous one.
394
AT91SAM9G45
Figure 27-4. Start Bit Detection

Sampling Clock
DRXD
True Start D0
Detection
Baud Rate
Clock
Figure 27-5. Character Reception

Example: 8-bit, parity enabled 1 stop
0.5 bit 1 bit

period period
DRXD
Sampling D0 D1 D2 D3 D4 D5 D6 D7 Stop Bit

True Start Detection Parity Bit
27.5.2.3 Receiver Ready

When a complete character is received, it is transferred to the DBGU_RHR and the RXRDY sta-
tus bit in DBGU_SR (Status Register) is set. The bit RXRDY is automatically cleared when the
receive holding register DBGU_RHR is read.
Figure 27-6. Receiver Ready

DRXD S D0 D1 D2 D3 D4 D5 D6 D7 P S D0 D1 D2 D3 D4 D5 D6 D7 P
RXRDY
Read DBGU_RHR
27.5.2.4 Receiver Overrun

If DBGU_RHR has not been read by the software (or the Peripheral Data Controller) since the
last transfer, the RXRDY bit is still set and a new character is received, the OVRE status bit in
DBGU_SR is set. OVRE is cleared when the software writes the control register DBGU_CR with
the bit RSTSTA (Reset Status) at 1.
Figure 27-7. Receiver Overrun

DRXD S D0 D1 D2 D3 D4 D5 D6 D7 P stop S D0 D1 D2 D3 D4 D5 D6 D7 P stop
RXRDY
OVRE
RSTSTA
27.5.2.5 Parity Error

Each time a character is received, the receiver calculates the parity of the received data bits, in
accordance with the field PAR in DBGU_MR. It then compares the result with the received parity
395
AT91SAM9G45
bit. If different, the parity error bit PARE in DBGU_SR is set at the same time the RXRDY is set.
The parity bit is cleared when the control register DBGU_CR is written with the bit RSTSTA
(Reset Status) at 1. If a new character is received before the reset status command is written,
the PARE bit remains at 1.
Figure 27-8. Parity Error

DRXD S D0 D1 D2 D3 D4 D5 D6 D7 P stop
RXRDY
PARE
Wrong Parity Bit RSTSTA
27.5.2.6 Receiver Framing Error

When a start bit is detected, it generates a character reception when all the data bits have been
sampled. The stop bit is also sampled and when it is detected at 0, the FRAME (Framing Error)
bit in DBGU_SR is set at the same time the RXRDY bit is set. The bit FRAME remains high until
the control register DBGU_CR is written with the bit RSTSTA at 1.
Figure 27-9. Receiver Framing Error

DRXD S D0 D1 D2 D3 D4 D5 D6 D7 P stop
RXRDY
FRAME
Stop Bit RSTSTA

Detected at 0
27.5.3 Transmitter
27.5.3.1 Transmitter Reset, Enable and Disable

After device reset, the Debug Unit transmitter is disabled and it must be enabled before being
used. The transmitter is enabled by writing the control register DBGU_CR with the bit TXEN at 1.
From this command, the transmitter waits for a character to be written in the Transmit Holding
Register DBGU_THR before actually starting the transmission.
The programmer can disable the transmitter by writing DBGU_CR with the bit TXDIS at 1. If the
transmitter is not operating, it is immediately stopped. However, if a character is being pro-
cessed into the Shift Register and/or a character has been written in the Transmit Holding
Register, the characters are completed before the transmitter is actually stopped.
The programmer can also put the transmitter in its reset state by writing the DBGU_CR with the
bit RSTTX at 1. This immediately stops the transmitter, whether or not it is processing
characters.
27.5.3.2 Transmit Format

The Debug Unit transmitter drives the pin DTXD at the baud rate clock speed. The line is driven
depending on the format defined in the Mode Register and the data stored in the Shift Register.
One start bit at level 0, then the 8 data bits, from the lowest to the highest bit, one optional parity
bit and one stop bit at 1 are consecutively shifted out as shown on the following figure. The field
396
AT91SAM9G45
PARE in the mode register DBGU_MR defines whether or not a parity bit is shifted out. When a
parity bit is enabled, it can be selected between an odd parity, an even parity, or a fixed space or
mark bit.
Figure 27-10. Character Transmission

Example: Parity enabled
Baud Rate
Clock
DTXD
Start D0 D1 D2 D3 D4 D5 D6 D7 Parity Stop

Bit Bit Bit
27.5.3.3 Transmitter Control

When the transmitter is enabled, the bit TXRDY (Transmitter Ready) is set in the status register
DBGU_SR. The transmission starts when the programmer writes in the Transmit Holding Regis-
ter DBGU_THR, and after the written character is transferred from DBGU_THR to the Shift
Register. The bit TXRDY remains high until a second character is written in DBGU_THR. As
soon as the first character is completed, the last character written in DBGU_THR is transferred
into the shift register and TXRDY rises again, showing that the holding register is empty.
When both the Shift Register and the DBGU_THR are empty, i.e., all the characters written in
DBGU_THR have been processed, the bit TXEMPTY rises after the last stop bit has been
completed.
Figure 27-11. Transmitter Control

DBGU_THR Data 0 Data 1
Shift Register Data 0 Data 1
DTXD S Data 0 P stop S Data 1 P stop
TXRDY
TXEMPTY
Write Data 0 Write Data 1

in DBGU_THR in DBGU_THR
27.5.4 Peripheral Data Controller

Both the receiver and the transmitter of the Debug Unit's UART are generally connected to a
Peripheral Data Controller (PDC) channel.
The peripheral data controller channels are programmed via registers that are mapped within
the Debug Unit user interface from the offset 0x100. The status bits are reported in the Debug
Unit status register DBGU_SR and can generate an interrupt.
397
AT91SAM9G45
The RXRDY bit triggers the PDC channel data transfer of the receiver. This results in a read of
the data in DBGU_RHR. The TXRDY bit triggers the PDC channel data transfer of the transmit-
ter. This results in a write of a data in DBGU_THR.
27.5.5 Test Modes

The Debug Unit supports three tests modes. These modes of operation are programmed by
using the field CHMODE (Channel Mode) in the mode register DBGU_MR.
The Automatic Echo mode allows bit-by-bit retransmission. When a bit is received on the DRXD
line, it is sent to the DTXD line. The transmitter operates normally, but has no effect on the
DTXD line.
The Local Loopback mode allows the transmitted characters to be received. DTXD and DRXD
pins are not used and the output of the transmitter is internally connected to the input of the
receiver. The DRXD pin level has no effect and the DTXD line is held high, as in idle state.
The Remote Loopback mode directly connects the DRXD pin to the DTXD line. The transmitter
and the receiver are disabled and have no effect. This mode allows a bit-by-bit retransmission.
Figure 27-12. Test Modes

Automatic Echo
Receiver RXD
Disabled
Transmitter TXD
Local Loopback
Disabled
Receiver RXD
VDD
Disabled
Transmitter TXD
Remote Loopback VDD

Disabled
Receiver RXD
Disabled
Transmitter TXD
27.5.6 Debug Communication Channel Support

The Debug Unit handles the signals COMMRX and COMMTX that come from the Debug Com-
munication Channel of the ARM Processor and are driven by the In-circuit Emulator.
398
AT91SAM9G45
The Debug Communication Channel contains two registers that are accessible through the ICE
Breaker on the JTAG side and through the coprocessor 0 on the ARM Processor side.
As a reminder, the following instructions are used to read and write the Debug Communication
Channel:
MRC p14, 0, Rd, c1, c0, 0
Returns the debug communication data read register into Rd
MCR p14, 0, Rd, c1, c0, 0

Writes the value in Rd to the debug communication data write register.
The bits COMMRX and COMMTX, which indicate, respectively, that the read register has been
written by the debugger but not yet read by the processor, and that the write register has been
written by the processor and not yet read by the debugger, are wired on the two highest bits of
the status register DBGU_SR. These bits can generate an interrupt. This feature permits han-
dling under interrupt a debug link between a debug monitor running on the target system and a
debugger.
27.5.7 Chip Identifier

The Debug Unit features two chip identifier registers, DBGU_CIDR (Chip ID Register) and
DBGU_EXID (Extension ID). Both registers contain a hard-wired value that is read-only. The first
register contains the following fields:
• EXT - shows the use of the extension identifier register
• NVPTYP and NVPSIZ - identifies the type of embedded non-volatile memory and its size
• ARCH - identifies the set of embedded peripherals
• SRAMSIZ - indicates the size of the embedded SRAM
• EPROC - indicates the embedded ARM processor
• VERSION - gives the revision of the silicon
The second register is device-dependent and reads 0 if the bit EXT is 0.
27.5.8 ICE Access Prevention

The Debug Unit allows blockage of access to the system through the ARM processor's ICE
interface. This feature is implemented via the register Force NTRST (DBGU_FNR), that allows
assertion of the NTRST signal of the ICE Interface. Writing the bit FNTRST (Force NTRST) to 1
in this register prevents any activity on the TAP controller.
On standard devices, the bit FNTRST resets to 0 and thus does not prevent ICE access.
This feature is especially useful on custom ROM devices for customers who do not want their
on-chip code to be visible.
399
AT91SAM9G45
27.6 Debug Unit (DBGU) User Interface

0x0000 Control Register DBGU_CR Write-only –
0x0004 Mode Register DBGU_MR Read-write 0x0
0x0008 Interrupt Enable Register DBGU_IER Write-only –
0x000C Interrupt Disable Register DBGU_IDR Write-only –
0x0010 Interrupt Mask Register DBGU_IMR Read-only 0x0
0x0014 Status Register DBGU_SR Read-only –
0x0018 Receive Holding Register DBGU_RHR Read-only 0x0
0x001C Transmit Holding Register DBGU_THR Write-only –
0x0020 Baud Rate Generator Register DBGU_BRGR Read-write 0x0
0x0024 - 0x003C Reserved – – –
0x0040 Chip ID Register DBGU_CIDR Read-only –
0x0044 Chip ID Extension Register DBGU_EXID Read-only –
0x0048 Force NTRST Register DBGU_FNR Read-write 0x0
0x0100 - 0x0124 PDC Area – – –
400
AT91SAM9G45
27.6.1 Debug Unit Control Register

Name: DBGU_CR
Address: 0xFFFFEE00
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – RSTSTA
7 6 5 4 3 2 1 0
TXDIS TXEN RXDIS RXEN RSTTX RSTRX – –
• RSTRX: Reset Receiver

0 = No effect.
1 = The receiver logic is reset and disabled. If a character is being received, the reception is aborted.
• RSTTX: Reset Transmitter
0 = No effect.
1 = The transmitter logic is reset and disabled. If a character is being transmitted, the transmission is aborted.
• RXEN: Receiver Enable
0 = No effect.
1 = The receiver is enabled if RXDIS is 0.
• RXDIS: Receiver Disable
0 = No effect.
1 = The receiver is disabled. If a character is being processed and RSTRX is not set, the character is completed before the
receiver is stopped.
• TXEN: Transmitter Enable
0 = No effect.
1 = The transmitter is enabled if TXDIS is 0.
• TXDIS: Transmitter Disable
0 = No effect.
1 = The transmitter is disabled. If a character is being processed and a character has been written the DBGU_THR and
RSTTX is not set, both characters are completed before the transmitter is stopped.
• RSTSTA: Reset Status Bits
0 = No effect.
1 = Resets the status bits PARE, FRAME and OVRE in the DBGU_SR.
401
AT91SAM9G45
27.6.2 Debug Unit Mode Register

Name: DBGU_MR
Address: 0xFFFFEE04
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
CHMODE – – PAR –
7 6 5 4 3 2 1 0
– – – – – – – –
• PAR: Parity Type
PAR Parity Type

0 0 0 Even parity
0 0 1 Odd parity
0 1 0 Space: parity forced to 0
0 1 1 Mark: parity forced to 1
1 x x No parity
• CHMODE: Channel Mode
CHMODE Mode Description

0 0 Normal Mode
0 1 Automatic Echo
1 0 Local Loopback
1 1 Remote Loopback
402
AT91SAM9G45
27.6.3 Debug Unit Interrupt Enable Register

Name: DBGU_IER
Address: 0xFFFFEE08
31 30 29 28 27 26 25 24
COMMRX COMMTX – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – RXBUFF TXBUFE – TXEMPTY –
7 6 5 4 3 2 1 0
PARE FRAME OVRE ENDTX ENDRX – TXRDY RXRDY
• RXRDY: Enable RXRDY Interrupt

• TXRDY: Enable TXRDY Interrupt
• ENDRX: Enable End of Receive Transfer Interrupt
• ENDTX: Enable End of Transmit Interrupt
• OVRE: Enable Overrun Error Interrupt
• FRAME: Enable Framing Error Interrupt
• PARE: Enable Parity Error Interrupt
• TXEMPTY: Enable TXEMPTY Interrupt
• TXBUFE: Enable Buffer Empty Interrupt
• RXBUFF: Enable Buffer Full Interrupt
• COMMTX: Enable COMMTX (from ARM) Interrupt
• COMMRX: Enable COMMRX (from ARM) Interrupt
0 = No effect.
403
AT91SAM9G45
27.6.4 Debug Unit Interrupt Disable Register

Name: DBGU_IDR
Address: 0xFFFFEE0C
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• RXRDY: Disable RXRDY Interrupt

• TXRDY: Disable TXRDY Interrupt
• ENDRX: Disable End of Receive Transfer Interrupt
• ENDTX: Disable End of Transmit Interrupt
• OVRE: Disable Overrun Error Interrupt
• FRAME: Disable Framing Error Interrupt
• PARE: Disable Parity Error Interrupt
• TXEMPTY: Disable TXEMPTY Interrupt
• TXBUFE: Disable Buffer Empty Interrupt
• RXBUFF: Disable Buffer Full Interrupt
• COMMTX: Disable COMMTX (from ARM) Interrupt
• COMMRX: Disable COMMRX (from ARM) Interrupt
0 = No effect.
404
AT91SAM9G45
27.6.5 Debug Unit Interrupt Mask Register

Name: DBGU_IMR
Address: 0xFFFFEE10
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• RXRDY: Mask RXRDY Interrupt

• TXRDY: Disable TXRDY Interrupt
• ENDRX: Mask End of Receive Transfer Interrupt
• ENDTX: Mask End of Transmit Interrupt
• OVRE: Mask Overrun Error Interrupt
• FRAME: Mask Framing Error Interrupt
• PARE: Mask Parity Error Interrupt
• TXEMPTY: Mask TXEMPTY Interrupt
• TXBUFE: Mask TXBUFE Interrupt
• RXBUFF: Mask RXBUFF Interrupt
• COMMTX: Mask COMMTX Interrupt
• COMMRX: Mask COMMRX Interrupt
405
AT91SAM9G45
27.6.6 Debug Unit Status Register

Name: DBGU_SR
Address: 0xFFFFEE14
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• RXRDY: Receiver Ready

0 = No character has been received since the last read of the DBGU_RHR or the receiver is disabled.
1 = At least one complete character has been received, transferred to DBGU_RHR and not yet read.
• TXRDY: Transmitter Ready
0 = A character has been written to DBGU_THR and not yet transferred to the Shift Register, or the transmitter is disabled.
1 = There is no character written to DBGU_THR not yet transferred to the Shift Register.
• ENDRX: End of Receiver Transfer
0 = The End of Transfer signal from the receiver Peripheral Data Controller channel is inactive.
1 = The End of Transfer signal from the receiver Peripheral Data Controller channel is active.
• ENDTX: End of Transmitter Transfer
0 = The End of Transfer signal from the transmitter Peripheral Data Controller channel is inactive.
1 = The End of Transfer signal from the transmitter Peripheral Data Controller channel is active.
• OVRE: Overrun Error
0 = No overrun error has occurred since the last RSTSTA.
1 = At least one overrun error has occurred since the last RSTSTA.
• FRAME: Framing Error
0 = No framing error has occurred since the last RSTSTA.
1 = At least one framing error has occurred since the last RSTSTA.
• PARE: Parity Error

0 = No parity error has occurred since the last RSTSTA.
1 = At least one parity error has occurred since the last RSTSTA.
• TXEMPTY: Transmitter Empty
0 = There are characters in DBGU_THR, or characters being processed by the transmitter, or the transmitter is disabled.
1 = There are no characters in DBGU_THR and there are no characters being processed by the transmitter.
406
AT91SAM9G45
• TXBUFE: Transmission Buffer Empty

0 = The buffer empty signal from the transmitter PDC channel is inactive.
1 = The buffer empty signal from the transmitter PDC channel is active.
• RXBUFF: Receive Buffer Full

0 = The buffer full signal from the receiver PDC channel is inactive.
1 = The buffer full signal from the receiver PDC channel is active.
• COMMTX: Debug Communication Channel Write Status
0 = COMMTX from the ARM processor is inactive.
1 = COMMTX from the ARM processor is active.
• COMMRX: Debug Communication Channel Read Status
0 = COMMRX from the ARM processor is inactive.
1 = COMMRX from the ARM processor is active.
407
AT91SAM9G45
27.6.7 Debug Unit Receiver Holding Register

Name: DBGU_RHR
Address: 0xFFFFEE18
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
RXCHR
• RXCHR: Received Character

Last received character if RXRDY is set.
27.6.8 Debug Unit Transmit Holding Register

Name: DBGU_THR
Address: 0xFFFFEE1C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
TXCHR
• TXCHR: Character to be Transmitted

Next character to be transmitted after the current character if TXRDY is not set.
408
AT91SAM9G45
27.6.9 Debug Unit Baud Rate Generator Register

Name: DBGU_BRGR
Address: 0xFFFFEE20
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
CD
7 6 5 4 3 2 1 0
CD
• CD: Clock Divisor
CD Baud Rate Clock

0 Disabled
1 MCK
2 to 65535 MCK / (CD x 16)
409
AT91SAM9G45
27.6.10 Debug Unit Chip ID Register

Name: DBGU_CIDR
Address: 0xFFFFEE40
31 30 29 28 27 26 25 24
EXT NVPTYP ARCH
23 22 21 20 19 18 17 16
ARCH SRAMSIZ
15 14 13 12 11 10 9 8
NVPSIZ2 NVPSIZ
7 6 5 4 3 2 1 0
EPROC VERSION
• VERSION: Version of the Device

Values depend upon the version of the device.
• EPROC: Embedded Processor
EPROC Processor
0 0 1 ARM946ES
0 1 0 ARM7TDMI
1 0 0 ARM920T
1 0 1 ARM926EJS
• NVPSIZ: Nonvolatile Program Memory Size
NVPSIZ Size
0 0 0 0 None
0 0 0 1 8K bytes
0 0 1 0 16K bytes
0 0 1 1 32K bytes
0 1 0 0 Reserved
0 1 0 1 64K bytes
0 1 1 0 Reserved
0 1 1 1 128K bytes
1 0 0 0 Reserved
1 0 0 1 256K bytes
1 0 1 0 512K bytes
1 0 1 1 Reserved
1 1 0 0 1024K bytes
410
AT91SAM9G45
NVPSIZ Size
1 1 0 1 Reserved
1 1 1 0 2048K bytes
1 1 1 1 Reserved
• NVPSIZ2 Second Nonvolatile Program Memory Size
NVPSIZ2 Size
0 0 0 0 None
0 0 0 1 8K bytes
0 0 1 0 16K bytes
0 0 1 1 32K bytes
0 1 0 0 Reserved
0 1 0 1 64K bytes
0 1 1 0 Reserved
0 1 1 1 128K bytes
1 0 0 0 Reserved
1 0 0 1 256K bytes
1 0 1 0 512K bytes
1 0 1 1 Reserved
1 1 0 0 1024K bytes
1 1 0 1 Reserved
1 1 1 0 2048K bytes
1 1 1 1 Reserved
• SRAMSIZ: Internal SRAM Size
SRAMSIZ Size
0 0 0 0 Reserved
0 0 0 1 1K bytes
0 0 1 0 2K bytes
0 0 1 1 6K bytes
0 1 0 0 112K bytes
0 1 0 1 4K bytes
0 1 1 0 80K bytes
0 1 1 1 160K bytes
1 0 0 0 8K bytes
1 0 0 1 16K bytes
1 0 1 0 32K bytes
1 0 1 1 64K bytes
411
AT91SAM9G45
SRAMSIZ Size
1 1 0 0 128K bytes
1 1 0 1 256K bytes
1 1 1 0 96K bytes
1 1 1 1 512K bytes
• ARCH: Architecture Identifier
ARCH
Hex Bin Architecture
0x19 0001 1001 AT91SAM9xx Series
0x29 0010 1001 AT91SAM9XExx Series
0x34 0011 0100 AT91x34 Series
0x37 0011 0111 CAP7 Series
0x39 0011 1001 CAP9 Series
0x3B 0011 1011 CAP11 Series
0x40 0100 0000 AT91x40 Series
0x42 0100 0010 AT91x42 Series
0x55 0101 0101 AT91x55 Series
0x60 0110 0000 AT91SAM7Axx Series
0x61 0110 0001 AT91SAM7AQxx Series
0x63 0110 0011 AT91x63 Series
0x70 0111 0000 AT91SAM7Sxx Series
0x71 0111 0001 AT91SAM7XCxx Series
0x72 0111 0010 AT91SAM7SExx Series
0x73 0111 0011 AT91SAM7Lxx Series
0x75 0111 0101 AT91SAM7Xxx Series
0x92 1001 0010 AT91x92 Series
0xF0 1111 0000 AT75Cxx Series
• NVPTYP: Nonvolatile Program Memory Type
NVPTYP Memory
0 0 0 ROM
0 0 1 ROMless or on-chip Flash
1 0 0 SRAM emulating ROM
0 1 0 Embedded Flash Memory
ROM and Embedded Flash Memory
0 1 1 NVPSIZ is ROM size
NVPSIZ2 is Flash size
412
AT91SAM9G45
• EXT: Extension Flag

0 = Chip ID has a single register definition without extension
1 = An extended Chip ID exists.
413
AT91SAM9G45
27.6.11 Debug Unit Chip ID Extension Register

Name: DBGU_EXID
Address: 0xFFFFEE44
31 30 29 28 27 26 25 24
EXID
23 22 21 20 19 18 17 16
EXID
15 14 13 12 11 10 9 8
EXID
7 6 5 4 3 2 1 0
EXID
• EXID: Chip ID Extension

Reads 0 if the bit EXT in DBGU_CIDR is 0.
27.6.12 Debug Unit Force NTRST Register

Name: DBGU_FNR
Address: 0xFFFFEE48
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – FNTRST
• FNTRST: Force NTRST

0 = NTRST of the ARM processor’s TAP controller is driven by the power_on_reset signal.
1 = NTRST of the ARM processor’s TAP controller is held low.
414
AT91SAM9G45
28. Parallel Input/Output Controller (PIO)
28.1 Description
The Parallel Input/Output Controller (PIO) manages up to 32 fully programmable input/output
lines. Each I/O line may be dedicated as a general-purpose I/O or be assigned to a function of
an embedded peripheral. This assures effective optimization of the pins of a product.
Each I/O line is associated with a bit number in all of the 32-bit registers of the 32-bit wide User
Interface.
Each I/O line of the PIO Controller features:
• An input change interrupt enabling level change detection on any I/O line.
• A glitch filter providing rejection of pulses lower than one-half of clock cycle.
• Multi-drive capability similar to an open drain I/O line.
• Control of the the pull-up of the I/O line.
• Input visibility and output control.
The PIO Controller also features a synchronous output providing up to 32 bits of data output in a
single write operation.

• 5 PIO Controllers, PIOA, PIOB, PIOC, PIOD and PIOE, controlling a maximum of 160 I/O
Lines
• Each PIO Controller controls up to 32 programmable I/O Lines
– PIOA has 32 I/O Lines
– PIOB has 32 I/O Lines
– PIOC has 32 I/O Lines
– PIOD has 32 I/O Lines
– PIOE has 32 I/O Lines
• Fully programmable through Set/Clear Registers
• Multiplexing of two peripheral functions per I/O Line
• For each I/O Line (whether assigned to a peripheral or used as general purpose I/O)
– Input change interrupt
– Glitch filter
– Multi-drive option enables driving in open drain
– Programmable pull up on each I/O line
– Pin data status register, supplies visibility of the level on the pin at any time
• Synchronous output, provides Set and Clear of several I/O lines in a single write
415
28.3 Block Diagram
PIO Controller
PIO Interrupt
AIC
PIO Clock
PMC
Data, Enable
Up to 32
peripheral IOs
Embedded
Peripheral
PIN 0
Data, Enable
PIN 1
Up to 32 pins
Up to 32
Embedded peripheral IOs
Peripheral PIN 31
APB
Figure 28-2. Application Block Diagram
On-Chip Peripheral Drivers

Keyboard Driver Control & Command
Driver
On-Chip Peripherals
PIO Controller
Keyboard Driver General Purpose I/Os External Devices
416 AT91SAM9G45
AT91SAM9G45
28.4.1 Pin Multiplexing

Each pin is configurable, according to product definition as either a general-purpose I/O line
only, or as an I/O line multiplexed with one or two peripheral I/Os. As the multiplexing is hard-
ware-defined and thus product-dependent, the hardware designer and programmer must
carefully determine the configuration of the PIO controllers required by their application. When
an I/O line is general-purpose only, i.e. not multiplexed with any peripheral I/O, programming of
the PIO Controller regarding the assignment to a peripheral has no effect and only the PIO Con-
troller can control how the pin is driven by the product.
28.4.2 External Interrupt Lines

The interrupt signals FIQ and IRQ0 to IRQn are most generally multiplexed through the PIO
Controllers. However, it is not necessary to assign the I/O line to the interrupt function as the
PIO Controller has no effect on inputs and the interrupt lines (FIQ or IRQs) are used only as
inputs.

The Power Management Controller controls the PIO Controller clock in order to save power.
Writing any of the registers of the user interface does not require the PIO Controller clock to be
enabled. This means that the configuration of the I/O lines does not require the PIO Controller
clock to be enabled.
However, when the clock is disabled, not all of the features of the PIO Controller are available.
Note that the Input Change Interrupt and the read of the pin level require the clock to be
validated.
After a hardware reset, the PIO clock is disabled by default.
The user must configure the Power Management Controller before any access to the input line
information.
28.4.4 Interrupt Generation

For interrupt handling, the PIO Controllers are considered as user peripherals. This means that
the PIO Controller interrupt lines are connected among the interrupt sources 2 to 31. Refer to the
PIO Controller peripheral identifier in the product description to identify the interrupt sources
dedicated to the PIO Controllers.
The PIO Controller interrupt can be generated only if the PIO Controller clock is enabled.
417
The PIO Controller features up to 32 fully-programmable I/O lines. Most of the control logic asso-
ciated to each I/O is represented in Figure 28-3. In this description each signal shown
represents but one of up to 32 possible indexes.
Figure 28-3. I/O Line Control Logic
PIO_OER[0]
PIO_OSR[0]
PIO_ODR[0] PIO_PUER[0]
PIO_PUSR[0]
1 PIO_PUDR[0]
Peripheral A
Output Enable 0
0
Peripheral B
0
Output Enable 1
PIO_ASR[0] PIO_PER[0]
PIO_ABSR[0] PIO_PSR[0] 1
PIO_BSR[0] PIO_PDR[0] PIO_MDER[0]

Peripheral A 0 PIO_MDSR[0]
Output
0 PIO_MDDR[0]
Peripheral B 1 0
PIO_SODR[0]
Output
PIO_ODSR[0] 1 Pad
PIO_CODR[0] 1
Peripheral A
Input
Peripheral B
PIO_PDSR[0] PIO_ISR[0] Input
0
(Up to 32 possible inputs)
Edge
Detector
Glitch 1 PIO Interrupt
Filter
PIO_IFER[0]
PIO_IFSR[0] PIO_IER[0]
PIO_IFDR[0] PIO_IMR[0]
PIO_IDR[0]
PIO_ISR[31]
PIO_IER[31]
PIO_IMR[31]
PIO_IDR[31]
418 AT91SAM9G45
AT91SAM9G45
28.5.1 Pull-up Resistor Control

Each I/O line is designed with an embedded pull-up resistor. The pull-up resistor can be enabled
or disabled by writing respectively PIO_PUER (Pull-up Enable Register) and PIO_PUDR (Pull-
up Disable Resistor). Writing in these registers results in setting or clearing the corresponding bit
in PIO_PUSR (Pull-up Status Register). Reading a 1 in PIO_PUSR means the pull-up is dis-
abled and reading a 0 means the pull-up is enabled.
Control of the pull-up resistor is possible regardless of the configuration of the I/O line.
After reset, all of the pull-ups are enabled, i.e. PIO_PUSR resets at the value 0x0.
28.5.2 I/O Line or Peripheral Function Selection

When a pin is multiplexed with one or two peripheral functions, the selection is controlled with
the registers PIO_PER (PIO Enable Register) and PIO_PDR (PIO Disable Register). The regis-
ter PIO_PSR (PIO Status Register) is the result of the set and clear registers and indicates
whether the pin is controlled by the corresponding peripheral or by the PIO Controller. A value of
0 indicates that the pin is controlled by the corresponding on-chip peripheral selected in the
PIO_ABSR (AB Select Status Register). A value of 1 indicates the pin is controlled by the PIO
controller.
If a pin is used as a general purpose I/O line (not multiplexed with an on-chip peripheral),
PIO_PER and PIO_PDR have no effect and PIO_PSR returns 1 for the corresponding bit.
After reset, most generally, the I/O lines are controlled by the PIO controller, i.e. PIO_PSR
resets at 1. However, in some events, it is important that PIO lines are controlled by the periph-
eral (as in the case of memory chip select lines that must be driven inactive after reset or for
address lines that must be driven low for booting out of an external memory). Thus, the reset
value of PIO_PSR is defined at the product level, depending on the multiplexing of the device.
28.5.3 Peripheral A or B Selection

The PIO Controller provides multiplexing of up to two peripheral functions on a single pin. The
selection is performed by writing PIO_ASR (A Select Register) and PIO_BSR (Select B Regis-
ter). PIO_ABSR (AB Select Status Register) indicates which peripheral line is currently selected.
For each pin, the corresponding bit at level 0 means peripheral A is selected whereas the corre-
sponding bit at level 1 indicates that peripheral B is selected.
Note that multiplexing of peripheral lines A and B only affects the output line. The peripheral
input lines are always connected to the pin input.
After reset, PIO_ABSR is 0, thus indicating that all the PIO lines are configured on peripheral A.
However, peripheral A generally does not drive the pin as the PIO Controller resets in I/O line
mode.
Writing in PIO_ASR and PIO_BSR manages PIO_ABSR regardless of the configuration of the
pin. However, assignment of a pin to a peripheral function requires a write in the corresponding
peripheral selection register (PIO_ASR or PIO_BSR) in addition to a write in PIO_PDR.
28.5.4 Output Control

When the I/0 line is assigned to a peripheral function, i.e. the corresponding bit in PIO_PSR is at
0, the drive of the I/O line is controlled by the peripheral. Peripheral A or B, depending on the
value in PIO_ABSR, determines whether the pin is driven or not.
When the I/O line is controlled by the PIO controller, the pin can be configured to be driven. This
is done by writing PIO_OER (Output Enable Register) and PIO_ODR (Output Disable Register).
419
The results of these write operations are detected in PIO_OSR (Output Status Register). When
a bit in this register is at 0, the corresponding I/O line is used as an input only. When the bit is at
1, the corresponding I/O line is driven by the PIO controller.
The level driven on an I/O line can be determined by writing in PIO_SODR (Set Output Data
Register) and PIO_CODR (Clear Output Data Register). These write operations respectively set
and clear PIO_ODSR (Output Data Status Register), which represents the data driven on the I/O
lines. Writing in PIO_OER and PIO_ODR manages PIO_OSR whether the pin is configured to
be controlled by the PIO controller or assigned to a peripheral function. This enables configura-
tion of the I/O line prior to setting it to be managed by the PIO Controller.
Similarly, writing in PIO_SODR and PIO_CODR effects PIO_ODSR. This is important as it
defines the first level driven on the I/O line.
28.5.5 Synchronous Data Output

Controlling all parallel busses using several PIOs requires two successive write operations in the
PIO_SODR and PIO_CODR registers. This may lead to unexpected transient values. The PIO
controller offers a direct control of PIO outputs by single write access to PIO_ODSR (Output
Data Status Register). Only bits unmasked by PIO_OWSR (Output Write Status Register) are
written. The mask bits in the PIO_OWSR are set by writing to PIO_OWER (Output Write Enable
Register) and cleared by writing to PIO_OWDR (Output Write Disable Register).
After reset, the synchronous data output is disabled on all the I/O lines as PIO_OWSR resets at
0x0.
28.5.6 Multi Drive Control (Open Drain)

Each I/O can be independently programmed in Open Drain by using the Multi Drive feature. This
feature permits several drivers to be connected on the I/O line which is driven low only by each
device. An external pull-up resistor (or enabling of the internal one) is generally required to guar-
antee a high level on the line.
The Multi Drive feature is controlled by PIO_MDER (Multi-driver Enable Register) and
PIO_MDDR (Multi-driver Disable Register). The Multi Drive can be selected whether the I/O line
is controlled by the PIO controller or assigned to a peripheral function. PIO_MDSR (Multi-driver
Status Register) indicates the pins that are configured to support external drivers.
After reset, the Multi Drive feature is disabled on all pins, i.e. PIO_MDSR resets at value 0x0.
28.5.7 Output Line Timings

Figure 28-4 shows how the outputs are driven either by writing PIO_SODR or PIO_CODR, or by
directly writing PIO_ODSR. This last case is valid only if the corresponding bit in PIO_OWSR is
set. Figure 28-4 also shows when the feedback in PIO_PDSR is available.
420 AT91SAM9G45
AT91SAM9G45
Figure 28-4. Output Line Timings
MCK
Write PIO_SODR APB Access

Write PIO_ODSR at 1
Write PIO_CODR APB Access

Write PIO_ODSR at 0
PIO_ODSR
2 cycles 2 cycles
PIO_PDSR
28.5.8 Inputs
The level on each I/O line can be read through PIO_PDSR (Pin Data Status Register). This reg-
ister indicates the level of the I/O lines regardless of their configuration, whether uniquely as an
input or driven by the PIO controller or driven by a peripheral.
Reading the I/O line levels requires the clock of the PIO controller to be enabled, otherwise
PIO_PDSR reads the levels present on the I/O line at the time the clock was disabled.
28.5.9 Input Glitch Filtering

Optional input glitch filters are independently programmable on each I/O line. When the glitch fil-
ter is enabled, a glitch with a duration of less than 1/2 Master Clock (MCK) cycle is automatically
rejected, while a pulse with a duration of 1 Master Clock cycle or more is accepted. For pulse
durations between 1/2 Master Clock cycle and 1 Master Clock cycle the pulse may or may not
be taken into account, depending on the precise timing of its occurrence. Thus for a pulse to be
visible it must exceed 1 Master Clock cycle, whereas for a glitch to be reliably filtered out, its
duration must not exceed 1/2 Master Clock cycle. The filter introduces one Master Clock cycle
latency if the pin level change occurs before a rising edge. However, this latency does not
appear if the pin level change occurs before a falling edge. This is illustrated in Figure 28-5.
The glitch filters are controlled by the register set; PIO_IFER (Input Filter Enable Register),
PIO_IFDR (Input Filter Disable Register) and PIO_IFSR (Input Filter Status Register). Writing
PIO_IFER and PIO_IFDR respectively sets and clears bits in PIO_IFSR. This last register
enables the glitch filter on the I/O lines.
When the glitch filter is enabled, it does not modify the behavior of the inputs on the peripherals.
It acts only on the value read in PIO_PDSR and on the input change interrupt detection. The
glitch filters require that the PIO Controller clock is enabled.
421
Figure 28-5. Input Glitch Filter Timing
MCK
up to 1.5 cycles
Pin Level
1 cycle 1 cycle 1 cycle 1 cycle
PIO_PDSR
if PIO_IFSR = 0
2 cycles 1 cycle
PIO_PDSR up to 2.5 cycles up to 2 cycles

if PIO_IFSR = 1
28.5.10 Input Change Interrupt

The PIO Controller can be programmed to generate an interrupt when it detects an input change
on an I/O line. The Input Change Interrupt is controlled by writing PIO_IER (Interrupt Enable
Register) and PIO_IDR (Interrupt Disable Register), which respectively enable and disable the
input change interrupt by setting and clearing the corresponding bit in PIO_IMR (Interrupt Mask
Register). As Input change detection is possible only by comparing two successive samplings of
the input of the I/O line, the PIO Controller clock must be enabled. The Input Change Interrupt is
available, regardless of the configuration of the I/O line, i.e. configured as an input only, con-
trolled by the PIO Controller or assigned to a peripheral function.
When an input change is detected on an I/O line, the corresponding bit in PIO_ISR (Interrupt
Status Register) is set. If the corresponding bit in PIO_IMR is set, the PIO Controller interrupt
line is asserted. The interrupt signals of the thirty-two channels are ORed-wired together to gen-
erate a single interrupt signal to the Advanced Interrupt Controller.
When the software reads PIO_ISR, all the interrupts are automatically cleared. This signifies that
all the interrupts that are pending when PIO_ISR is read must be handled.
Figure 28-6. Input Change Interrupt Timings
MCK
Pin Level
PIO_ISR
Read PIO_ISR APB Access APB Access
28.6 I/O Lines Programming Example

The programing example as shown in Table 28-1 below is used to define the following
configuration.
• 4-bit output port on I/O lines 0 to 3, (should be written in a single write operation), open-drain,
with pull-up resistor
422 AT91SAM9G45
AT91SAM9G45
• Four output signals on I/O lines 4 to 7 (to drive LEDs for example), driven high and low, no
pull-up resistor
• Four input signals on I/O lines 8 to 11 (to read push-button states for example), with pull-up
resistors, glitch filters and input change interrupts
• Four input signals on I/O line 12 to 15 to read an external device status (polled, thus no input
change interrupt), no pull-up resistor, no glitch filter
• I/O lines 16 to 19 assigned to peripheral A functions with pull-up resistor
• I/O lines 20 to 23 assigned to peripheral B functions, no pull-up resistor
• I/O line 24 to 27 assigned to peripheral A with Input Change Interrupt and pull-up resistor
Table 28-1. Programming Example

Register Value to be Written
PIO_PER 0x0000 FFFF
PIO_PDR 0x0FFF 0000
PIO_OER 0x0000 00FF
PIO_ODR 0x0FFF FF00
PIO_IFER 0x0000 0F00
PIO_IFDR 0x0FFF F0FF
PIO_SODR 0x0000 0000
PIO_CODR 0x0FFF FFFF
PIO_IER 0x0F00 0F00
PIO_IDR 0x00FF F0FF
PIO_MDER 0x0000 000F
PIO_MDDR 0x0FFF FFF0
PIO_PUDR 0x00F0 00F0
PIO_PUER 0x0F0F FF0F
PIO_ASR 0x0F0F 0000
PIO_BSR 0x00F0 0000
PIO_OWER 0x0000 000F
PIO_OWDR 0x0FFF FFF0
28.7 Parallel Input/Output Controller (PIO) User Interface

Each I/O line controlled by the PIO Controller is associated with a bit in each of the PIO Control-
ler User Interface registers. Each register is 32 bits wide. If a parallel I/O line is not defined,
writing to the corresponding bits has no effect. Undefined bits read zero. If the I/O line is not mul-
423
tiplexed with any peripheral, the I/O line is controlled by the PIO Controller and PIO_PSR returns
1 systematically.

0x0000 PIO Enable Register PIO_PER Write-only –
0x0004 PIO Disable Register PIO_PDR Write-only –
(1)
0x0008 PIO Status Register PIO_PSR Read-only
0x000C Reserved
0x0010 Output Enable Register PIO_OER Write-only –
0x0014 Output Disable Register PIO_ODR Write-only –
0x0018 Output Status Register PIO_OSR Read-only 0x0000 0000
0x001C Reserved
0x0020 Glitch Input Filter Enable Register PIO_IFER Write-only –
0x0024 Glitch Input Filter Disable Register PIO_IFDR Write-only –
0x0028 Glitch Input Filter Status Register PIO_IFSR Read-only 0x0000 0000
0x002C Reserved
0x0030 Set Output Data Register PIO_SODR Write-only –
0x0034 Clear Output Data Register PIO_CODR Write-only
Read-only
0x0038 Output Data Status Register PIO_ODSR or(2) –
Read/Write
(3)
0x003C Pin Data Status Register PIO_PDSR Read-only
0x0040 Interrupt Enable Register PIO_IER Write-only –
0x0044 Interrupt Disable Register PIO_IDR Write-only –
0x0048 Interrupt Mask Register PIO_IMR Read-only 0x00000000
0x004C Interrupt Status Register(4) PIO_ISR Read-only 0x00000000
0x0050 Multi-driver Enable Register PIO_MDER Write-only –
0x0054 Multi-driver Disable Register PIO_MDDR Write-only –
0x0058 Multi-driver Status Register PIO_MDSR Read-only 0x00000000
0x005C Reserved
0x0060 Pull-up Disable Register PIO_PUDR Write-only –
0x0064 Pull-up Enable Register PIO_PUER Write-only –
0x0068 Pad Pull-up Status Register PIO_PUSR Read-only 0x00000000
0x006C Reserved
424 AT91SAM9G45
AT91SAM9G45

(5)
0x0070 Peripheral A Select Register PIO_ASR Write-only –
(5)
0x0074 Peripheral B Select Register PIO_BSR Write-only –
(5)
0x0078 AB Status Register PIO_ABSR Read-only 0x00000000
0x007C
to Reserved
0x009C
0x00A0 Output Write Enable PIO_OWER Write-only –
0x00A4 Output Write Disable PIO_OWDR Write-only –
0x00A8 Output Write Status Register PIO_OWSR Read-only 0x00000000
0x00AC Reserved
Notes: 1. Reset value of PIO_PSR depends on the product implementation.
2. PIO_ODSR is Read-only or Read/Write depending on PIO_OWSR I/O lines.
3. Reset value of PIO_PDSR depends on the level of the I/O lines. Reading the I/O line levels requires the clock of the PIO
Controller to be enabled, otherwise PIO_PDSR reads the levels present on the I/O line at the time the clock was disabled.
4. PIO_ISR is reset at 0x0. However, the first read of the register may read a different value as input changes may have
occurred.
5. Only this set of registers clears the status by writing 1 in the first register and sets the status by writing 1 in the second
register.
425
28.7.1 PIO Controller PIO Enable Register
Name: PIO_PER
Addresses: 0xFFFFF200 (PIOA), 0xFFFFF400 (PIOB), 0xFFFFF600 (PIOC), 0xFFFFF800 (PIOD),
0xFFFFFA00 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: PIO Enable

0 = No effect.
1 = Enables the PIO to control the corresponding pin (disables peripheral control of the pin).
28.7.2 PIO Controller PIO Disable Register

Name: PIO_PDR
0xFFFFFA04 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: PIO Disable

0 = No effect.
1 = Disables the PIO from controlling the corresponding pin (enables peripheral control of the pin).
426 AT91SAM9G45
AT91SAM9G45
28.7.3 PIO Controller PIO Status Register

Name: PIO_PSR
0xFFFFFA08 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: PIO Status

0 = PIO is inactive on the corresponding I/O line (peripheral is active).
1 = PIO is active on the corresponding I/O line (peripheral is inactive).
28.7.4 PIO Controller Output Enable Register

Name: PIO_OER
0xFFFFFA10 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Output Enable

0 = No effect.
1 = Enables the output on the I/O line.
427
28.7.5 PIO Controller Output Disable Register
Name: PIO_ODR
0xFFFFFA14 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Output Disable

0 = No effect.
1 = Disables the output on the I/O line.
28.7.6 PIO Controller Output Status Register

Name: PIO_OSR
0xFFFFFA18 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Output Status

0 = The I/O line is a pure input.
1 = The I/O line is enabled in output.
428 AT91SAM9G45
AT91SAM9G45
28.7.7 PIO Controller Input Filter Enable Register

Name: PIO_IFER
0xFFFFFA20 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Input Filter Enable

0 = No effect.
1 = Enables the input glitch filter on the I/O line.
28.7.8 PIO Controller Input Filter Disable Register

Name: PIO_IFDR
0xFFFFFA24 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Input Filter Disable

0 = No effect.
1 = Disables the input glitch filter on the I/O line.
429
28.7.9 PIO Controller Input Filter Status Register
Name: PIO_IFSR
0xFFFFFA28 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Input Filer Status

0 = The input glitch filter is disabled on the I/O line.
1 = The input glitch filter is enabled on the I/O line.
28.7.10 PIO Controller Set Output Data Register

Name: PIO_SODR
0xFFFFFA30 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Set Output Data

0 = No effect.
1 = Sets the data to be driven on the I/O line.
430 AT91SAM9G45
AT91SAM9G45
28.7.11 PIO Controller Clear Output Data Register

Name: PIO_CODR
0xFFFFFA34 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Set Output Data

0 = No effect.
1 = Clears the data to be driven on the I/O line.
28.7.12 PIO Controller Output Data Status Register

Name: PIO_ODSR
0xFFFFFA38 (PIOE)
Access Type: Read-only or Read/Write
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Output Data Status

0 = The data to be driven on the I/O line is 0.
1 = The data to be driven on the I/O line is 1.
431
28.7.13 PIO Controller Pin Data Status Register
Name: PIO_PDSR
Addresses: 0xFFFFF23C (PIOA), 0xFFFFF43C (PIOB), 0xFFFFF63C (PIOC), 0xFFFFF83C (PIOD),
0xFFFFFA3C (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Output Data Status

0 = The I/O line is at level 0.
1 = The I/O line is at level 1.
28.7.14 PIO Controller Interrupt Enable Register

Name: PIO_IER
0xFFFFFA40 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Input Change Interrupt Enable

0 = No effect.
1 = Enables the Input Change Interrupt on the I/O line.
432 AT91SAM9G45
AT91SAM9G45
28.7.15 PIO Controller Interrupt Disable Register

Name: PIO_IDR
0xFFFFFA44 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Input Change Interrupt Disable

0 = No effect.
1 = Disables the Input Change Interrupt on the I/O line.
28.7.16 PIO Controller Interrupt Mask Register

Name: PIO_IMR
0xFFFFFA48 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Input Change Interrupt Mask

0 = Input Change Interrupt is disabled on the I/O line.
1 = Input Change Interrupt is enabled on the I/O line.
433
28.7.17 PIO Controller Interrupt Status Register
Name: PIO_ISR
Addresses: 0xFFFFF24C (PIOA), 0xFFFFF44C (PIOB), 0xFFFFF64C (PIOC), 0xFFFFF84C (PIOD),
0xFFFFFA4C (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Input Change Interrupt Status

0 = No Input Change has been detected on the I/O line since PIO_ISR was last read or since reset.
1 = At least one Input Change has been detected on the I/O line since PIO_ISR was last read or since reset.
28.7.18 PIO Multi-driver Enable Register

Name: PIO_MDER
0xFFFFFA50 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Multi Drive Enable.

0 = No effect.
1 = Enables Multi Drive on the I/O line.
434 AT91SAM9G45
AT91SAM9G45
28.7.19 PIO Multi-driver Disable Register

Name: PIO_MDDR
0xFFFFFA54 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Multi Drive Disable.

0 = No effect.
1 = Disables Multi Drive on the I/O line.
28.7.20 PIO Multi-driver Status Register

Name: PIO_MDSR
0xFFFFFA58 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Multi Drive Status.

0 = The Multi Drive is disabled on the I/O line. The pin is driven at high and low level.
1 = The Multi Drive is enabled on the I/O line. The pin is driven at low level only.
435
28.7.21 PIO Pull Up Disable Register
Name: PIO_PUDR
0xFFFFFA60 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Pull Up Disable.

0 = No effect.
1 = Disables the pull up resistor on the I/O line.
28.7.22 PIO Pull Up Enable Register

Name: PIO_PUER
0xFFFFFA64 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Pull Up Enable.

0 = No effect.
1 = Enables the pull up resistor on the I/O line.
436 AT91SAM9G45
AT91SAM9G45
28.7.23 PIO Pull Up Status Register

Name: PIO_PUSR
0xFFFFFA68 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Pull Up Status.

0 = Pull Up resistor is enabled on the I/O line.
1 = Pull Up resistor is disabled on the I/O line.
28.7.24 PIO Peripheral A Select Register

Name: PIO_ASR
0xFFFFFA70 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Peripheral A Select.

0 = No effect.
1 = Assigns the I/O line to the Peripheral A function.
437
AT91SAM9G45
28.7.25 PIO Peripheral B Select Register

Name: PIO_BSR
0xFFFFFA74 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Peripheral B Select.

0 = No effect.
1 = Assigns the I/O line to the peripheral B function.
28.7.26 PIO Peripheral A B Status Register

Name: PIO_ABSR
0xFFFFFA78 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Peripheral A B Status.

0 = The I/O line is assigned to the Peripheral A.
1 = The I/O line is assigned to the Peripheral B.
438
AT91SAM9G45
28.7.27 PIO Output Write Enable Register

Name: PIO_OWER
Addresses: 0xFFFFF2A0 (PIOA), 0xFFFFF4A0 (PIOB), 0xFFFFF6A0 (PIOC), 0xFFFFF8A0 (PIOD),
0xFFFFFAA0 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Output Write Enable.

0 = No effect.
1 = Enables writing PIO_ODSR for the I/O line.
28.7.28 PIO Output Write Disable Register

Name: PIO_OWDR
0xFFFFFAA4 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Output Write Disable.

0 = No effect.
1 = Disables writing PIO_ODSR for the I/O line.
439
AT91SAM9G45
28.7.29 PIO Output Write Status Register

Name: PIO_OWSR
0xFFFFFAA8 (PIOE)
31 30 29 28 27 26 25 24
P31 P30 P29 P28 P27 P26 P25 P24
23 22 21 20 19 18 17 16
P23 P22 P21 P20 P19 P18 P17 P16
15 14 13 12 11 10 9 8
P15 P14 P13 P12 P11 P10 P9 P8
7 6 5 4 3 2 1 0
P7 P6 P5 P4 P3 P2 P1 P0
• P0-P31: Output Write Status.

0 = Writing PIO_ODSR does not affect the I/O line.
1 = Writing PIO_ODSR affects the I/O line.
440
AT91SAM9G45
29. Serial Peripheral Interface (SPI)

29.1 Description
The Serial Peripheral Interface (SPI) circuit is a synchronous serial data link that provides com-
munication with external devices in Master or Slave Mode. It also enables communication
between processors if an external processor is connected to the system.
The Serial Peripheral Interface is essentially a shift register that serially transmits data bits to
other SPIs. During a data transfer, one SPI system acts as the “master”' which controls the data
flow, while the other devices act as “slaves'' which have data shifted into and out by the master.
Different CPUs can take turn being masters (Multiple Master Protocol opposite to Single Master
Protocol where one CPU is always the master while all of the others are always slaves) and one
master may simultaneously shift data into multiple slaves. However, only one slave may drive its
output to write data back to the master at any given time.
A slave device is selected when the master asserts its NSS signal. If multiple slave devices
exist, the master generates a separate slave select signal for each slave (NPCS).
The SPI system consists of two data lines and two control lines:
• Master Out Slave In (MOSI): This data line supplies the output data from the master shifted
into the input(s) of the slave(s).
• Master In Slave Out (MISO): This data line supplies the output data from a slave to the input
of the master. There may be no more than one slave transmitting data during any particular
transfer.
• Serial Clock (SPCK): This control line is driven by the master and regulates the flow of the
data bits. The master may transmit data at a variety of baud rates; the SPCK line cycles once
for each bit that is transmitted.
• Slave Select (NSS): This control line allows slaves to be turned on and off by hardware.

• Supports communication with serial external devices
– Four chip selects with external decoder support allow communication with up to 15
peripherals
– Serial memories, such as DataFlash and 3-wire EEPROMs
– Serial peripherals, such as ADCs, DACs, LCD Controllers, CAN Controllers and
Sensors
– External co-processors
• Master or slave serial peripheral bus interface
– 8- to 16-bit programmable data length per chip select
– Programmable phase and polarity per chip select
– Programmable transfer delays between consecutive transfers and between clock
and data per chip select
– Programmable delay between consecutive transfers
– Selectable mode fault detection
• Very fast transfers supported
– Transfers with baud rates up to MCK
– The chip select line may be left active to speed up transfers on the same device
441
29.3 Block Diagram
PDC
APB
SPCK
MISO
MCK MOSI
PMC
SPI Interface PIO NPCS0/NSS
NPCS1
NPCS2
Interrupt Control
NPCS3
SPI Interrupt
AHB Matrix DMA Ch.
Peripheral Bridge
APB
SPCK
MISO
MCK MOSI
PMC
SPI Interface PIO NPCS0/NSS
NPCS1
NPCS2
Interrupt Control
NPCS3
SPI Interrupt
442 AT91SAM9G45
AT91SAM9G45
Figure 29-3. Application Block Diagram: Single Master/Multiple Slave Implementation
SPCK SPCK
MISO MISO
Slave 0
MOSI MOSI
SPI Master NPCS0 NSS
SPCK
NPCS1
MISO
NPCS2 NC Slave 1
NPCS3 MOSI
NSS
SPCK
MISO
Slave 2
MOSI
NSS
443
29.5 Signal Description
Table 29-1. Signal Description
Type
Pin Name Pin Description Master Slave
MISO Master In Slave Out Input Output
MOSI Master Out Slave In Output Input
SPCK Serial Clock Output Input
NPCS1-NPCS3 Peripheral Chip Selects Output Unused
NPCS0/NSS Peripheral Chip Select/Slave Select Output Input
29.6.1 I/O Lines

The pins used for interfacing the compliant external devices may be multiplexed with PIO lines.
The programmer must first program the PIO controllers to assign the SPI pins to their peripheral
functions.

SPI0 SPI0_MISO PB0 A
SPI0 SPI0_MOSI PB1 A
SPI0 SPI0_NPCS0 PB3 A
SPI0 SPI0_NPCS1 PB18 B
SPI0 SPI0_NPCS1 PD24 A
SPI0 SPI0_NPCS2 PB19 B
SPI0 SPI0_NPCS3 PD27 B
SPI0 SPI0_SPCK PB2 A
SPI1 SPI1_MISO PB14 A
SPI1 SPI1_MOSI PB15 A
SPI1 SPI1_NPCS0 PB17 A
SPI1 SPI1_NPCS1 PD28 B
SPI1 SPI1_SPCK PB16 A

The SPI may be clocked through the Power Management Controller (PMC), thus the program-
mer must first configure the PMC to enable the SPI clock.
444 AT91SAM9G45
AT91SAM9G45
29.6.3 Interrupt
The SPI interface has an interrupt line connected to the Advanced Interrupt Controller
(AIC).Handling the SPI interrupt requires programming the AICbefore configuring the SPI.
Table 29-3. Peripheral IDs

Instance ID
SPI0 14
SPI1 15
29.6.4 Peripheral DMA Controller (PDMA) Direct Memory Access Controller (DMAC)
The SPI interface can be used in conjunction with the PDMA DMAC in order to reduce processor
overhead. For a full description of the PDMA DMAC, refer to the corresponding section in the full
datasheet.
29.7.1 Modes of Operation

The SPI operates in Master Mode or in Slave Mode.
Operation in Master Mode is programmed by writing at 1 the MSTR bit in the Mode Register.
The pins NPCS0 to NPCS3 are all configured as outputs, the SPCK pin is driven, the MISO line
is wired on the receiver input and the MOSI line driven as an output by the transmitter.
If the MSTR bit is written at 0, the SPI operates in Slave Mode. The MISO line is driven by the
transmitter output, the MOSI line is wired on the receiver input, the SPCK pin is driven by the
transmitter to synchronize the receiver. The NPCS0 pin becomes an input, and is used as a
Slave Select signal (NSS). The pins NPCS1 to NPCS3 are not driven and can be used for other
purposes.
The data transfers are identically programmable for both modes of operations. The baud rate
generator is activated only in Master Mode.
29.7.2 Data Transfer

Four combinations of polarity and phase are available for data transfers. The clock polarity is
programmed with the CPOL bit in the Chip Select Register. The clock phase is programmed with
the NCPHA bit. These two parameters determine the edges of the clock signal on which data is
driven and sampled. Each of the two parameters has two possible states, resulting in four possi-
ble combinations that are incompatible with one another. Thus, a master/slave pair must use the
same parameter pair values to communicate. If multiple slaves are used and fixed in different
configurations, the master must reconfigure itself each time it needs to communicate with a dif-
ferent slave.
445
Table 29-4 shows the four modes and corresponding parameter settings.
Table 29-4. SPI Bus Protocol Mode

SPI Mode CPOL NCPHA Shift SPCK Edge Capture SPCK Edge SPCK Inactive Level
0 0 1 Falling Rising Low
1 0 0 Rising Falling Low
2 1 1 Rising Falling High
3 1 0 Falling Rising High
Figure 29-4 and Figure 29-5 show examples of data transfers.
Figure 29-4. SPI Transfer Format (NCPHA = 1, 8 bits per transfer)

SPCK cycle (for reference) 1 2 3 4 5 6 7 8
SPCK
(CPOL = 0)
SPCK
(CPOL = 1)
MOSI
MSB 6 5 4 3 2 1 LSB
(from master)
MISO
(from slave)
MSB 6 5 4 3 2 1 LSB *
NSS
(to slave)
* Not defined, but normally MSB of previous character received.
446 AT91SAM9G45
AT91SAM9G45
Figure 29-5. SPI Transfer Format (NCPHA = 0, 8 bits per transfer)

SPCK cycle (for reference) 1 2 3 4 5 6 7 8
SPCK
(CPOL = 0)
SPCK
(CPOL = 1)
MOSI MSB 6 5 4 3 2 1 LSB

(from master)
MISO
(from slave) * MSB 6 5 4 3 2 1 LSB
NSS
(to slave)
* Not defined but normally LSB of previous character transmitted.
29.7.3 Master Mode Operations

When configured in Master Mode, the SPI operates on the clock generated by the internal pro-
grammable baud rate generator. It fully controls the data transfers to and from the slave(s)
connected to the SPI bus. The SPI drives the chip select line to the slave and the serial clock
signal (SPCK).
The SPI features two holding registers, the Transmit Data Register and the Receive Data Regis-
ter, and a single Shift Register. The holding registers maintain the data flow at a constant rate.
After enabling the SPI, a data transfer begins when the processor writes to the SPI_TDR (Trans-
mit Data Register). The written data is immediately transferred in the Shift Register and transfer
on the SPI bus starts. While the data in the Shift Register is shifted on the MOSI line, the MISO
line is sampled and shifted in the Shift Register. Receiving data cannot occur without transmit-
ting data. If receiving mode is not needed, for example when communicating with a slave
receiver only (such as an LCD), the receive status flags in the status register can be discarded.
Before writing the TDR, the PCS field in the SPI_MR register must be set in order to select a
slave.
After enabling the SPI, a data transfer begins when the processor writes to the SPI_TDR (Trans-
mit Data Register). The written data is immediately transferred in the Shift Register and transfer
on the SPI bus starts. While the data in the Shift Register is shifted on the MOSI line, the MISO
line is sampled and shifted in the Shift Register. Transmission cannot occur without reception.
Before writing the TDR, the PCS field must be set in order to select a slave.
If new data is written in SPI_TDR during the transfer, it stays in it until the current transfer is
completed. Then, the received data is transferred from the Shift Register to SPI_RDR, the data
in SPI_TDR is loaded in the Shift Register and a new transfer starts.
447
The transfer of a data written in SPI_TDR in the Shift Register is indicated by the TDRE bit
(Transmit Data Register Empty) in the Status Register (SPI_SR). When new data is written in
SPI_TDR, this bit is cleared. The TDRE bit is used to trigger the Transmit PDC channel.
The end of transfer is indicated by the TXEMPTY flag in the SPI_SR register. If a transfer delay
(DLYBCT) is greater than 0 for the last transfer, TXEMPTY is set after the completion of said
delay. The master clock (MCK) can be switched off at this time.
The transfer of received data from the Shift Register in SPI_RDR is indicated by the RDRF bit
(Receive Data Register Full) in the Status Register (SPI_SR). When the received data is read,
the RDRF bit is cleared.
If the SPI_RDR (Receive Data Register) has not been read before new data is received, the
Overrun Error bit (OVRES) in SPI_SR is set. As long as this flag is set, data is loaded in
SPI_RDR. The user has to read the status register to clear the OVRES bit.
Figure 29-6, shows a block diagram of the SPI when operating in Master Mode. Figure 29-7 on
page 450 shows a flow chart describing how transfers are handled.
448 AT91SAM9G45
AT91SAM9G45
29.7.3.1 Master Mode Block Diagram
Figure 29-6. Master Mode Block Diagram

SPI_CSR0..3
SCBR
MCK Baud Rate Generator SPCK
SPI
Clock
SPI_CSR0..3
BITS SPI_RDR RDRF
NCPHA RD OVRES
CPOL
LSB Shift Register MSB

MISO MOSI
SPI_TDR
TD TDRE
SPI_CSR0..3
SPI_RDR
CSAAT PCS
PS
NPCS3
SPI_MR PCSDEC
PCS Current NPCS2
0 Peripheral
NPCS1
SPI_TDR
PCS NPCS0
1
MSTR
MODF
NPCS0
MODFDIS
449
29.7.3.2 Master Mode Flow Diagram
Figure 29-7. Master Mode Flow Diagram
SPI Enable - NPCS defines the current Chip Select

- CSAAT, DLYBS, DLYBCT refer to the fields of the
Chip Select Register corresponding to the Current Chip Select
- When NPCS is 0xF, CSAAT is 0.
1
TDRE ?
Fixed
1 0 peripheral
CSAAT ? PS ?
Variable
0 1 peripheral
Fixed
0 peripheral yes
PS ? SPI_TDR(PCS) SPI_MR(PCS)
= NPCS ? = NPCS ?
Variable
1 peripheral no no
NPCS = SPI_TDR(PCS) NPCS = SPI_MR(PCS) NPCS = 0xF NPCS = 0xF
Delay DLYBCS Delay DLYBCS
NPCS = SPI_TDR(PCS) NPCS = SPI_MR(PCS),

SPI_TDR(PCS)
Delay DLYBS
Serializer = SPI_TDR(TD)
TDRE = 1
Data Transfer
SPI_RDR(RD) = Serializer
RDRF = 1
Delay DLYBCT
0
TDRE ?
1
CSAAT ?
NPCS = 0xF
Delay DLYBCS
450 AT91SAM9G45
AT91SAM9G45
Figure 29-8 shows Transmit Data Register Empty (TDRE), Receive Data Register (RDRF) and
Transmission Register Empty (TXEMPTY) status flags behavior within the SPI_SR (Status Reg-
ister) during an 8-bit data transfer in fixed mode and no Peripheral Data Controller involved.
Figure 29-8. Status Register Flags Behavior

1 2 3 4 5 6 7 8
SPCK
NPCS0
MOSI
MSB 6 5 4 3 2 1 LSB
(from master)
TDRE
RDR read
Write in
SPI_TDR
RDRF
MISO
MSB 6 5 4 3 2 1 LSB
(from slave)
TXEMPTY
shift register empty
Figure 29-9 shows Transmission Register Empty (TXEMPTY), End of RX buffer (ENDRX), End
of TX buffer (ENDTX), RX Buffer Full (RXBUFF) and TX Buffer Empty (TXBUFE) status flags
behavior within the SPI_SR (Status Register) during an 8-bit data transfer in fixed mode with the
Peripheral Data Controller involved. The PDC is programmed to transfer and receive three data.
The next pointer and counter are not used. The RDRF and TDRE are not shown because these
flags are managed by the PDC when using the PDC.
451
Figure 29-9. PDC Status Register Flags Behavior
1 2 3
SPCK
NPCS0
MOSI
MSB 6 5 4 3 2 1 LSB MSB 6 5 4 3 2 1 LSB MSB 6 5 4 3 2 1 LSB
(from master)
MISO
MSB 6 5 4 3 2 1 LSB MSB 6 5 4 3 2 1 LSB MSB 6 5 4 3 2 1 LSB
(from slave)
ENDTX
ENDRX
TXBUFE
RXBUFF
TXEMPTY
29.7.3.3 Clock Generation

The SPI Baud rate clock is generated by dividing the Master Clock (MCK), by a value between 1
and 255.
This allows a maximum operating baud rate at up to Master Clock and a minimum operating
baud rate of MCK divided by 255.
Programming the SCBR field at 0 is forbidden. Triggering a transfer while SCBR is at 0 can lead
to unpredictable results.
At reset, SCBR is 0 and the user has to program it at a valid value before performing the first
transfer.
The divisor can be defined independently for each chip select, as it has to be programmed in the
SCBR field of the Chip Select Registers. This allows the SPI to automatically adapt the baud
rate for each interfaced peripheral without reprogramming.
29.7.3.4 Transfer Delays

Figure 29-10 shows a chip select transfer change and consecutive transfers on the same chip
select. Three delays can be programmed to modify the transfer waveforms:
• The delay between chip selects, programmable only once for all the chip selects by writing
the DLYBCS field in the Mode Register. Allows insertion of a delay between release of one
chip select and before assertion of a new one.
• The delay before SPCK, independently programmable for each chip select by writing the field
DLYBS. Allows the start of SPCK to be delayed after the chip select has been asserted.
• The delay between consecutive transfers, independently programmable for each chip select
by writing the DLYBCT field. Allows insertion of a delay between two transfers occurring on
the same chip select
452 AT91SAM9G45
AT91SAM9G45
These delays allow the SPI to be adapted to the interfaced peripherals and their speed and bus
release time.
Figure 29-10. Programmable Delays
Chip Select 1
Chip Select 2
SPCK
DLYBCS DLYBS DLYBCT DLYBCT
29.7.3.5 Peripheral Selection

The serial peripherals are selected through the assertion of the NPCS0 to NPCS3 signals. By
default, all the NPCS signals are high before and after each transfer.
• Fixed Peripheral Select: SPI exchanges data with only one peripheral
Fixed Peripheral Select is activated by writing the PS bit to zero in SPI_MR (Mode Register). In
this case, the current peripheral is defined by the PCS field in SPI_MR and the PCS field in the
SPI_TDR has no effect.
• Variable Peripheral Select: Data can be exchanged with more than one peripheral without
having to reprogram the NPCS field in the SPI_MR register.
Variable Peripheral Select is activated by setting PS bit to one. The PCS field in SPI_TDR is
used to select the current peripheral. This means that the peripheral selection can be defined for
each new data. The value to write in the SPI_TDR register as the following format.
[xxxxxxx(7-bit) + LASTXFER(1-bit)(1)+ xxxx(4-bit) + PCS (4-bit) + DATA (8 to 16-bit)] with PCS
equals to the chip select to assert as defined in Section 29.8.4 (SPI Transmit Data Register) and
LASTXFER bit at 0 or 1 depending on CSAAT bit. CSAAT, LASTXFER and CSNAAT bit are dis-
cussed in the Peripheral Deselection in Section 29.7.3.11.
Note: 1. Optional.
29.7.3.6 SPI Peripheral DMA Controller (PDC)

In both fixed and variable mode the Peripheral DMA Controller (PDC) can be used to reduce
processor overhead.
The Fixed Peripheral Selection allows buffer transfers with a single peripheral. Using the PDC is
an optimal means, as the size of the data transfer between the memory and the SPI is either 8
bits or 16 bits. However, changing the peripheral selection requires the Mode Register to be
reprogrammed.
The Variable Peripheral Selection allows buffer transfers with multiple peripherals without repro-
gramming the Mode Register. Data written in SPI_TDR is 32 bits wide and defines the real data
to be transmitted and the peripheral it is destined to. Using the PDC in this mode requires 32-bit
wide buffers, with the data in the LSBs and the PCS and LASTXFER fields in the MSBs, how-
453
ever the SPI still controls the number of bits (8 to16) to be transferred through MISO and MOSI
lines with the chip select configuration registers. This is not the optimal means in term of mem-
ory size for the buffers, but it provides a very effective means to exchange data with several
peripherals without any intervention of the processor.
29.7.3.7 Transfer Size

Depending on the data size to transmit, from 8 to 16 bits, the PDC manages automatically the
type of pointer's size it has to point to. The PDC will perform the following transfer size depend-
ing on the mode and number of bits per data.
Fixed Mode:
• 8-bit Data:
Byte transfer,
PDC Pointer Address = Address + 1 byte,
PDC Counter = Counter - 1
• 8-bit to 16-bit Data:
2 bytes transfer. n-bit data transfer with don’t care data (MSB) filled with 0’s,
PDC Pointer Address = Address + 2 bytes,
PDC Counter = Counter - 1
Variable Mode:
In variable Mode, PDC Pointer Address = Address +4 bytes and PDC Counter = Counter - 1 for
8 to 16-bit transfer size. When using the PDC, the TDRE and RDRF flags are handled by the
PDC, thus the user’s application does not have to check those bits. Only End of RX Buffer
(ENDRX), End of TX Buffer (ENDTX), Buffer Full (RXBUFF), TX Buffer Empty (TXBUFE) are
significant. For further details about the Peripheral DMA Controller and user interface, refer to
the PDC section of the product datasheet.
29.7.3.8 SPI Direct Access Memory Controller (DMAC)

In both fixed and variable mode the Direct Memory Access Controller (DMAC) can be used to
reduce processor overhead.
The Fixed Peripheral Selection allows buffer transfers with a single peripheral. Using the DMAC
is an optimal means, as the size of the data transfer between the memory and the SPI is either 8
bits or 16 bits. However, changing the peripheral selection requires the Mode Register to be
reprogrammed.
The Variable Peripheral Selection allows buffer transfers with multiple peripherals without repro-
gramming the Mode Register. Data written in SPI_TDR is 32 bits wide and defines the real data
to be transmitted and the peripheral it is destined to. Using the DMAC in this mode requires 32-
bit wide buffers, with the data in the LSBs and the PCS and LASTXFER fields in the MSBs, how-
ever the SPI still controls the number of bits (8 to16) to be transferred through MISO and MOSI
lines with the chip select configuration registers. This is not the optimal means in term of mem-
ory size for the buffers, but it provides a very effective means to exchange data with several
peripherals without any intervention of the processor.
454 AT91SAM9G45
AT91SAM9G45
29.7.3.9 Peripheral Chip Select Decoding

The user can program the SPI to operate with up to 15 peripherals by decoding the four Chip
Select lines, NPCS0 to NPCS3 with 1 of up to 16 decoder/demultiplexer. This can be enabled by
writing the PCSDEC bit at 1 in the Mode Register (SPI_MR).
When operating without decoding, the SPI makes sure that in any case only one chip select line
is activated, i.e., one NPCS line driven low at a time. If two bits are defined low in a PCS field,
only the lowest numbered chip select is driven low.
When operating with decoding, the SPI directly outputs the value defined by the PCS field on
NPCS lines of either the Mode Register or the Transmit Data Register (depending on PS).
As the SPI sets a default value of 0xF on the chip select lines (i.e. all chip select lines at 1) when
not processing any transfer, only 15 peripherals can be decoded.
The SPI has only four Chip Select Registers, not 15. As a result, when decoding is activated,
each chip select defines the characteristics of up to four peripherals. As an example, SPI_CRS0
defines the characteristics of the externally decoded peripherals 0 to 3, corresponding to the
PCS values 0x0 to 0x3. Thus, the user has to make sure to connect compatible peripherals on
the decoded chip select lines 0 to 3, 4 to 7, 8 to 11 and 12 to 14. Figure 29-11 below shows such
an implementation.
If the CSAAT bit is used, with or without the PDC, the Mode Fault detection for NPCS0 line must
be disabled. This is not needed for all other chip select lines since Mode Fault Detection is only
on NPCS0.
If the CSAAT bit is used, with or without the DMAC, the Mode Fault detection for NPCS0 line
must be disabled. This is not needed for all other chip select lines since Mode Fault Detection is
only on NPCS0.
Figure 29-11. Chip Select Decoding Application Block Diagram: Single Master/Multiple Slave Implementation
SPCK
MISO
MOSI
SPCK MISO MOSI SPCK MISO MOSI SPCK MISO MOSI

Slave 0 Slave 1 Slave 14
SPI Master
NSS NSS NSS
NPCS0
NPCS1
NPCS2
NPCS3
1-of-n Decoder/Demultiplexer
455
29.7.3.10 Peripheral Deselection without PDCDMAC
During a transfer of more than one data on a Chip Select without the PDCDMAC, the SPI_TDR
is loaded by the processor, the flag TDRE rises as soon as the content of the SPI_TDR is trans-
ferred into the internal shift register. When this flag is detected high, the SPI_TDR can be
reloaded. If this reload by the processor occurs before the end of the current transfer and if the
next transfer is performed on the same chip select as the current transfer, the Chip Select is not
de-asserted between the two transfers. But depending on the application software handling the
SPI status register flags (by interrupt or polling method) or servicing other interrupts or other
tasks, the processor may not reload the SPI_TDR in time to keep the chip select active (low). A
null Delay Between Consecutive Transfer (DLYBCT) value in the SPI_CSR register, will give
even less time for the processor to reload the SPI_TDR. With some SPI slave peripherals,
requiring the chip select line to remain active (low) during a full set of transfers might lead to
communication errors.
To facilitate interfacing with such devices, the Chip Select Register [CSR0...CSR3] can be pro-
grammed with the CSAAT bit (Chip Select Active After Transfer) at 1. This allows the chip select
lines to remain in their current state (low = active) until transfer to another chip select is required.
Even if the SPI_TDR is not reloaded the chip select will remain active. To have the chip select
line to raise at the end of the transfer the Last transfer Bit (LASTXFER) in the SPI_MR register
must be set at 1 before writing the last data to transmit into the SPI_TDR.
29.7.3.11 Peripheral Deselection with PDC

When the Peripheral DMA Controller is used, the chip select line will remain low during the
whole transfer since the TDRE flag is managed by the PDC itself. The reloading of the SPI_TDR
by the PDC is done as soon as TDRE flag is set to one. In this case the use of CSAAT bit might
not be needed. However, it may happen that when other PDC channels connected to other
peripherals are in use as well, the SPI PDC might be delayed by another (PDC with a higher pri-
ority on the bus). Having PDC buffers in slower memories like flash memory or SDRAM
compared to fast internal SRAM, may lengthen the reload time of the SPI_TDR by the PDC as
well. This means that the SPI_TDR might not be reloaded in time to keep the chip select line
low. In this case the chip select line may toggle between data transfer and according to some
SPI Slave devices, the communication might get lost. The use of the CSAAT bit might be
needed.
29.7.3.12 Peripheral Deselection with DMAC

When the Direct Memory Access Controller is used, the chip select line will remain low during
the whole transfer since the TDRE flag is managed by the DMAC itself. The reloading of the
SPI_TDR by the DMAC is done as soon as TDRE flag is set to one. In this case the use of
CSAAT bit might not be needed. However, it may happen that when other DMAC channels con-
nected to other peripherals are in use as well, the SPI DMAC might be delayed by another
(DMAC with a higher priority on the bus). Having DMAC buffers in slower memories like flash
memory or SDRAM compared to fast internal SRAM, may lengthen the reload time of the
SPI_TDR by the DMAC as well. This means that the SPI_TDR might not be reloaded in time to
keep the chip select line low. In this case the chip select line may toggle between data transfer
and according to some SPI Slave devices, the communication might get lost. The use of the
CSAAT bit might be needed.
Figure 29-12 shows different peripheral deselction cases and the effect of the CSAAT bit.
456 AT91SAM9G45
AT91SAM9G45
Figure 29-12. Peripheral Deselection

CSAAT = 0 CSAAT = 1
TDRE
DLYBCT DLYBCT
NPCS[0..3] A A A A A
DLYBCS DLYBCS
PCS = A PCS = A
Write SPI_TDR
TDRE
DLYBCT DLYBCT
NPCS[0..3] A A A A A
DLYBCS DLYBCS
PCS=A PCS = A
Write SPI_TDR
TDRE
DLYBCT DLYBCT
NPCS[0..3] A B A B
DLYBCS DLYBCS
PCS = B PCS = B
Write SPI_TDR
29.7.3.13 Mode Fault Detection

A mode fault is detected when the SPI is programmed in Master Mode and a low level is driven
by an external master on the NPCS0/NSS signal. In this case, multi-master configuration,
NPCS0, MOSI, MISO and SPCK pins must be configured in open drain (through the PIO control-
ler). When a mode fault is detected, the MODF bit in the SPI_SR is set until the SPI_SR is read
and the SPI is automatically disabled until re-enabled by writing the SPIEN bit in the SPI_CR
(Control Register) at 1.
By default, the Mode Fault detection circuitry is enabled. The user can disable Mode Fault
detection by setting the MODFDIS bit in the SPI Mode Register (SPI_MR).
29.7.4 SPI Slave Mode

When operating in Slave Mode, the SPI processes data bits on the clock provided on the SPI
clock pin (SPCK).
The SPI waits for NSS to go active before receiving the serial clock from an external master.
When NSS falls, the clock is validated on the serializer, which processes the number of bits
defined by the BITS field of the Chip Select Register 0 (SPI_CSR0). These bits are processed
following a phase and a polarity defined respectively by the NCPHA and CPOL bits of the
SPI_CSR0. Note that BITS, CPOL and NCPHA of the other Chip Select Registers have no
effect when the SPI is programmed in Slave Mode.
457
The bits are shifted out on the MISO line and sampled on the MOSI line.
(Note:)
(For more information on BITS field, see also, the below the register table; Section 29.8.9
“SPI Chip Select Register” on page 470.)
When all the bits are processed, the received data is transferred in the Receive Data Register
and the RDRF bit rises. If the SPI_RDR (Receive Data Register) has not been read before new
data is received, the Overrun Error bit (OVRES) in SPI_SR is set. As long as this flag is set, data
is loaded in SPI_RDR. The user has to read the status register to clear the OVRES bit.
When a transfer starts, the data shifted out is the data present in the Shift Register. If no data
has been written in the Transmit Data Register (SPI_TDR), the last data received is transferred.
If no data has been received since the last reset, all bits are transmitted low, as the Shift Regis-
ter resets at 0.
When a first data is written in SPI_TDR, it is transferred immediately in the Shift Register and the
TDRE bit rises. If new data is written, it remains in SPI_TDR until a transfer occurs, i.e. NSS falls
and there is a valid clock on the SPCK pin. When the transfer occurs, the last data written in
SPI_TDR is transferred in the Shift Register and the TDRE bit rises. This enables frequent
updates of critical variables with single transfers.
Then, a new data is loaded in the Shift Register from the Transmit Data Register. In case no
character is ready to be transmitted, i.e. no character has been written in SPI_TDR since the last
load from SPI_TDR to the Shift Register, the Shift Register is not modified and the last received
character is retransmitted.
Figure 29-13 shows a block diagram of the SPI when operating in Slave Mode.
Figure 29-13. Slave Mode Functional Bloc Diagram

SPCK
NSS SPI
Clock
SPIEN
SPIENS
SPIDIS
SPI_CSR0
BITS SPI_RDR RDRF
NCPHA RD OVRES
CPOL
LSB Shift Register MSB

MOSI MISO
SPI_TDR
TD TDRE
458 AT91SAM9G45
AT91SAM9G45
29.8 Serial Peripheral Interface (SPI) User Interface

0x00 Control Register SPI_CR Write-only ---
0x04 Mode Register SPI_MR Read-write 0x0
0x08 Receive Data Register SPI_RDR Read-only 0x0
0x0C Transmit Data Register SPI_TDR Write-only ---
0x10 Status Register SPI_SR Read-only 0x000000F0
0x14 Interrupt Enable Register SPI_IER Write-only ---
0x18 Interrupt Disable Register SPI_IDR Write-only ---
0x1C Interrupt Mask Register SPI_IMR Read-only 0x0
0x20 - 0x2C Reserved
0x30 Chip Select Register 0 SPI_CSR0 Read-write 0x0
0x3C Chip Select Register 3 SPI_CSR3 Read-write 0x0
0x004C - 0x00F8 Reserved – – –
0x100 - 0x124 Reserved for the PDC – – –
459
29.8.1 SPI Control Register
Name: SPI_CR
Addresses: 0xFFFA4000 (0), 0xFFFA8000 (1)
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – LASTXFER
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
SWRST – – – – – SPIDIS SPIEN
• SPIEN: SPI Enable

0 = No effect.
1 = Enables the SPI to transfer and receive data.
• SPIDIS: SPI Disable

0 = No effect.
1 = Disables the SPI.
As soon as SPIDIS is set, SPI finishes its transfer.
All pins are set in input mode and no data is received or transmitted.
If a transfer is in progress, the transfer is finished before the SPI is disabled.
If both SPIEN and SPIDIS are equal to one when the control register is written, the SPI is disabled.
• SWRST: SPI Software Reset

0 = No effect.
1 = Reset the SPI. A software-triggered hardware reset of the SPI interface is performed.
The SPI is in slave mode after software reset.
PDC channels are not affected by software reset.
• LASTXFER: Last Transfer

0 = No effect.
1 = The current NPCS will be deasserted after the character written in TD has been transferred. When CSAAT is set, this
allows to close the communication with the current serial peripheral by raising the corresponding NPCS line as soon as TD
transfer has completed.
460 AT91SAM9G45
AT91SAM9G45
29.8.2 SPI Mode Register

Name: SPI_MR
Access: Read/Write
31 30 29 28 27 26 25 24
DLYBCS
23 22 21 20 19 18 17 16
– – – – PCS
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
LLB – – MODFDIS – PCSDEC PS MSTR
• MSTR: Master/Slave Mode

0 = SPI is in Slave mode.
1 = SPI is in Master mode.
• PS: Peripheral Select

0 = Fixed Peripheral Select.
1 = Variable Peripheral Select.
• PCSDEC: Chip Select Decode

0 = The chip selects are directly connected to a peripheral device.
1 = The four chip select lines are connected to a 4- to 16-bit decoder.
When PCSDEC equals one, up to 15 Chip Select signals can be generated with the four lines using an external 4- to 16-bit
decoder. The Chip Select Registers define the characteristics of the 15 chip selects according to the following rules:
SPI_CSR0 defines peripheral chip select signals 0 to 3.
• MODFDIS: Mode Fault Detection

0 = Mode fault detection is enabled.
1 = Mode fault detection is disabled.
• LLB: Local Loopback Enable

0 = Local loopback path disabled.
1 = Local loopback path enabled
LLB controls the local loopback on the data serializer for testing in Master Mode only. (MISO is internally connected on
MOSI.)
461
• PCS: Peripheral Chip Select
This field is only used if Fixed Peripheral Select is active (PS = 0).
If PCSDEC = 0:
PCS = xxx0 NPCS[3:0] = 1110
PCS = xx01 NPCS[3:0] = 1101
PCS = x011 NPCS[3:0] = 1011
PCS = 0111 NPCS[3:0] = 0111
PCS = 1111 forbidden (no peripheral is selected)
(x = don’t care)
If PCSDEC = 1:
NPCS[3:0] output signals = PCS.
• DLYBCS: Delay Between Chip Selects

This field defines the delay from NPCS inactive to the activation of another NPCS. The DLYBCS time guarantees non-over-
lapping chip selects and solves bus contentions in case of peripherals having long data float times.
If DLYBCS is less than or equal to six, six MCK periods will be inserted by default.
Otherwise, the following equation determines the delay:
Delay Between Chip Selects = DLYBCS

-----------------------
MCK
462 AT91SAM9G45
AT91SAM9G45
29.8.3 SPI Receive Data Register

Name: SPI_RDR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – PCS
15 14 13 12 11 10 9 8
RD
7 6 5 4 3 2 1 0
RD
• RD: Receive Data

Data received by the SPI Interface is stored in this register right-justified. Unused bits read zero.

In Master Mode only, these bits indicate the value on the NPCS pins at the end of a transfer. Otherwise, these bits read
zero.
463
AT91SAM9G45
29.8.4 SPI Transmit Data Register

Name: SPI_TDR
Addresses: 0xFFFA400C (0), 0xFFFA800C (1)
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – LASTXFER
23 22 21 20 19 18 17 16
– – – – PCS
15 14 13 12 11 10 9 8
TD
7 6 5 4 3 2 1 0
TD
• TD: Transmit Data

Data to be transmitted by the SPI Interface is stored in this register. Information to be transmitted must be written to the
transmit data register in a right-justified format.

This field is only used if Variable Peripheral Select is active (PS = 1).
If PCSDEC = 0:
PCS = xxx0 NPCS[3:0] = 1110
PCS = xx01 NPCS[3:0] = 1101
PCS = x011 NPCS[3:0] = 1011
PCS = 0111 NPCS[3:0] = 0111
PCS = 1111 forbidden (no peripheral is selected)
(x = don’t care)
If PCSDEC = 1:
NPCS[3:0] output signals = PCS
• LASTXFER: Last Transfer

0 = No effect.
1 = The current NPCS will be deasserted after the character written in TD has been transferred. When CSAAT is set, this
allows to close the communication with the current serial peripheral by raising the corresponding NPCS line as soon as TD
transfer has completed.
This field is only used if Variable Peripheral Select is active (PS = 1).
464
AT91SAM9G45
29.8.5 SPI Status Register

Name: SPI_SR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – SPIENS
15 14 13 12 11 10 9 8
– – – – – – TXEMPTY NSSR
7 6 5 4 3 2 1 0
– – – – OVRES MODF TDRE RDRF
• RDRF: Receive Data Register Full

0 = No data has been received since the last read of SPI_RDR
1 = Data has been received and the received data has been transferred from the serializer to SPI_RDR since the last read
of SPI_RDR.
• TDRE: Transmit Data Register Empty

0 = Data has been written to SPI_TDR and not yet transferred to the serializer.
1 = The last data written in the Transmit Data Register has been transferred to the serializer.
TDRE equals zero when the SPI is disabled or at reset. The SPI enable command sets this bit to one.
• MODF: Mode Fault Error

0 = No Mode Fault has been detected since the last read of SPI_SR.
1 = A Mode Fault occurred since the last read of the SPI_SR.
• OVRES: Overrun Error Status

0 = No overrun has been detected since the last read of SPI_SR.
1 = An overrun has occurred since the last read of SPI_SR.
An overrun occurs when SPI_RDR is loaded at least twice from the serializer since the last read of the SPI_RDR.
• ENDRX: End of RX buffer

0 = The Receive Counter Register has not reached 0 since the last write in SPI_RCR(1) or SPI_RNCR(1).
1 = The Receive Counter Register has reached 0 since the last write in SPI_RCR(1) or SPI_RNCR(1).
• ENDTX: End of TX buffer

0 = The Transmit Counter Register has not reached 0 since the last write in SPI_TCR(1) or SPI_TNCR(1).
1 = The Transmit Counter Register has reached 0 since the last write in SPI_TCR(1) or SPI_TNCR(1).
• RXBUFF: RX Buffer Full

0 = SPI_RCR(1) or SPI_RNCR(1) has a value other than 0.
1 = Both SPI_RCR(1) and SPI_RNCR(1) have a value of 0.
465
AT91SAM9G45
• TXBUFE: TX Buffer Empty

0 = SPI_TCR(1) or SPI_TNCR(1) has a value other than 0.
1 = Both SPI_TCR(1) and SPI_TNCR(1) have a value of 0.
• NSSR: NSS Rising

0 = No rising edge detected on NSS pin since last read.
1 = A rising edge occurred on NSS pin since last read.
• TXEMPTY: Transmission Registers Empty

0 = As soon as data is written in SPI_TDR.
1 = SPI_TDR and internal shifter are empty. If a transfer delay has been defined, TXEMPTY is set after the completion of
such delay.
• SPIENS: SPI Enable Status

0 = SPI is disabled.
1 = SPI is enabled.
Note: 1. SPI_RCR, SPI_RNCR, SPI_TCR, SPI_TNCR are physically located in the PDC.
466
AT91SAM9G45
29.8.6 SPI Interrupt Enable Register

Name: SPI_IER
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
0 = No effect.
• RDRF: Receive Data Register Full Interrupt Enable
• TDRE: SPI Transmit Data Register Empty Interrupt Enable
• MODF: Mode Fault Error Interrupt Enable
• OVRES: Overrun Error Interrupt Enable
• ENDRX: End of Receive Buffer Interrupt Enable
• ENDTX: End of Transmit Buffer Interrupt Enable
• RXBUFF: Receive Buffer Full Interrupt Enable
• TXBUFE: Transmit Buffer Empty Interrupt Enable
• NSSR: NSS Rising Interrupt Enable
• TXEMPTY: Transmission Registers Empty Enable
467
AT91SAM9G45
29.8.7 SPI Interrupt Disable Register

Name: SPI_IDR
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
0 = No effect.
• RDRF: Receive Data Register Full Interrupt Disable
• TDRE: SPI Transmit Data Register Empty Interrupt Disable
• MODF: Mode Fault Error Interrupt Disable
• OVRES: Overrun Error Interrupt Disable
• ENDRX: End of Receive Buffer Interrupt Disable
• ENDTX: End of Transmit Buffer Interrupt Disable
• RXBUFF: Receive Buffer Full Interrupt Disable
• TXBUFE: Transmit Buffer Empty Interrupt Disable
• NSSR: NSS Rising Interrupt Disable
• TXEMPTY: Transmission Registers Empty Disable
468
AT91SAM9G45
29.8.8 SPI Interrupt Mask Register

Name: SPI_IMR
Addresses: 0xFFFA401C (0), 0xFFFA801C (1)
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
0 = The corresponding interrupt is not enabled.

• RDRF: Receive Data Register Full Interrupt Mask
• TDRE: SPI Transmit Data Register Empty Interrupt Mask
• MODF: Mode Fault Error Interrupt Mask
• OVRES: Overrun Error Interrupt Mask
• ENDRX: End of Receive Buffer Interrupt Mask
• ENDTX: End of Transmit Buffer Interrupt Mask
• RXBUFF: Receive Buffer Full Interrupt Mask
• TXBUFE: Transmit Buffer Empty Interrupt Mask
• NSSR: NSS Rising Interrupt Mask
• TXEMPTY: Transmission Registers Empty Mask
469
AT91SAM9G45
29.8.9 SPI Chip Select Register

Name: SPI_CSR0... SPI_CSR3
Access: Read/Write
31 30 29 28 27 26 25 24
DLYBCT
23 22 21 20 19 18 17 16
DLYBS
15 14 13 12 11 10 9 8
SCBR
7 6 5 4 3 2 1 0
BITS CSAAT – NCPHA CPOL
Note: SPI_CSRx registers must be written even if the user wants to use the defaults. The BITS field will not be updated with the trans-
lated value unless the register is written.
• CPOL: Clock Polarity

0 = The inactive state value of SPCK is logic level zero.
1 = The inactive state value of SPCK is logic level one.
CPOL is used to determine the inactive state value of the serial clock (SPCK). It is used with NCPHA to produce the
required clock/data relationship between master and slave devices.
• NCPHA: Clock Phase

0 = Data is changed on the leading edge of SPCK and captured on the following edge of SPCK.
1 = Data is captured on the leading edge of SPCK and changed on the following edge of SPCK.
NCPHA determines which edge of SPCK causes data to change and which edge causes data to be captured. NCPHA is
used with CPOL to produce the required clock/data relationship between master and slave devices.
• CSAAT: Chip Select Active After Transfer

0 = The Peripheral Chip Select Line rises as soon as the last transfer is achieved.
1 = The Peripheral Chip Select does not rise after the last transfer is achieved. It remains active until a new transfer is
requested on a different chip select.
• BITS: Bits Per Transfer (See the (Note:) below the register table; Section 29.8.9 “SPI Chip Select Register” on page 470.)
The BITS field determines the number of data bits transferred. Reserved values should not be used.
BITS Bits Per Transfer

0000 8
0001 9
0010 10
0011 11
0100 12
0101 13
0110 14
0111 15
470
AT91SAM9G45
BITS Bits Per Transfer

1000 16
1001 Reserved
1010 Reserved
1011 Reserved
1100 Reserved
1101 Reserved
1110 Reserved
1111 Reserved
• SCBR: Serial Clock Baud Rate

In Master Mode, the SPI Interface uses a modulus counter to derive the SPCK baud rate from the Master Clock MCK. The
Baud rate is selected by writing a value from 1 to 255 in the SCBR field. The following equations determine the SPCK baud
rate:
MCK-
SPCK Baudrate = --------------
SCBR
Programming the SCBR field at 0 is forbidden. Triggering a transfer while SCBR is at 0 can lead to unpredictable results.
At reset, SCBR is 0 and the user has to program it at a valid value before performing the first transfer.
• DLYBS: Delay Before SPCK

This field defines the delay from NPCS valid to the first valid SPCK transition.
When DLYBS equals zero, the NPCS valid to SPCK transition is 1/2 the SPCK clock period.
Otherwise, the following equations determine the delay:
Delay Before SPCK = DLYBS

-------------------
MCK
• DLYBCT: Delay Between Consecutive Transfers

This field defines the delay between two consecutive transfers with the same peripheral without removing the chip select.
The delay is always inserted after each transfer and before removing the chip select if needed.
When DLYBCT equals zero, no delay between consecutive transfers is inserted and the clock keeps its duty cycle over the
character transfers.
Otherwise, the following equation determines the delay:
32 × DLYBCT
Delay Between Consecutive Transfers = -------------------------------------
MCK
471
472 AT91SAM9G45
AT91SAM9G45
30. Two-wire Interface (TWI)
30.1 Description
The Atmel Two-wire Interface (TWI) interconnects components on a unique two-wire bus, made
up of one clock line and one data line with speeds of up to 400 Kbits per second, based on a
byte-oriented transfer format. It can be used with any Atmel Two-wire Interface bus Serial
EEPROM and I²C compatible device such as Real Time Clock (RTC), Dot Matrix/Graphic LCD
Controllers and Temperature Sensor, to name but a few. The TWI is programmable as a master
or a slave with sequential or single-byte access. Multiple master capability is supported. 20
Arbitration of the bus is performed internally and puts the TWI in slave mode automatically if the
bus arbitration is lost.
A configurable baud rate generator permits the output data rate to be adapted to a wide range of
core clock frequencies.
Below, Table 30-1 lists the compatibility level of the Atmel Two-wire Interface in Master Mode and
a full I2C compatible device.
Table 30-1. Atmel TWI compatibility with i2C Standard

I2C Standard Atmel TWI
Standard Mode Speed (100 KHz) Supported
Fast Mode Speed (400 KHz) Supported
7 or 10 bits Slave Addressing Supported
(1)
START BYTE Not Supported
Repeated Start (Sr) Condition Supported
ACK and NACK Management Supported
Slope control and input filtering (Fast mode) Not Supported
Clock stretching Supported
Multi Master Capability Supported
Note: 1. START + b000000001 + Ack + Sr

• Compatibility with standard two-wire serial memory
• One, two or three bytes for slave address
• Sequential read/write operations
• Supports either master or slave modes
• Compatible with Standard Two-wire Serial Memories
• Master, Multi-master and Slave Mode Operation
• Bit Rate: Up to 400 Kbits
• General Call Supported in Slave mode
• Connection to Peripheral DMA Controller (PDC) Channel Capabilities Optimizes Data
Transfers in Master Mode Only
– One Channel for the Receiver, One Channel for the Transmitter
– Next Buffer Support
473
30.3 List of Abbreviations
Table 30-2. Abbreviations

Abbreviation Description
TWI Two-wire Interface
A Acknowledge
NA Non Acknowledge
P Stop
S Start
Sr Repeated Start
SADR Slave Address
ADR Any address except SADR
R Read
W Write
30.4 Block Diagram
APB Bridge
TWCK
PIO
Two-wire TWD
Interface
MCK
PMC
TWI
Interrupt
AIC
474 AT91SAM9G45

VDD
Rp Rp
TWD
Host with
TWI
Interface TWCK
Atmel TWI I²C LCD I²C Temp.

I²C RTC
Serial EEPROM Controller Sensor
Slave 1 Slave 2 Slave 3 Slave 4
Rp: Pull up value as given by the I²C Standard

TWD Two-wire Serial Data Input/Output
TWCK Two-wire Serial Clock Input/Output
30.6.1 I/O Lines

Both TWD and TWCK are bidirectional lines, connected to a positive supply voltage via a current
source or pull-up resistor (see Figure 30-2 on page 475). When the bus is free, both lines are
high. The output stages of devices connected to the bus must have an open-drain or open-col-
lector to perform the wired-AND function.
TWD and TWCK pins may be multiplexed with PIO lines. To enable the TWI, the programmer
must perform the following step:
• Program the PIO controller to dedicate TWD and TWCK as peripheral lines.
The user must not program TWD and TWCK as open-drain. It is already done by the hardware.

TWI0 TWCK0 PA21 A
TWI0 TWD0 PA20 A
TWI1 TWCK1 PB11 A
TWI1 TWD1 PB10 A
475 AT91SAM9G45
• Enable the peripheral clock.
The TWI interface may be clocked through the Power Management Controller (PMC), thus the
programmer must first configure the PMC to enable the TWI clock.
30.6.3 Interrupt
The TWI interface has an interrupt line connected to the Advanced Interrupt Controller (AIC). In
order to handle interrupts, the AIC must be programmed before configuring the TWI.

Instance ID
TWI0 12
TWI1 13
30.7.1 Transfer Format

The data put on the TWD line must be 8 bits long. Data is transferred MSB first; each byte must
be followed by an acknowledgement. The number of bytes per transfer is unlimited (see Figure
30-4).
Each transfer begins with a START condition and terminates with a STOP condition (see Figure
30-3).
• A high-to-low transition on the TWD line while TWCK is high defines the START condition.
• A low-to-high transition on the TWD line while TWCK is high defines a STOP condition.
Figure 30-3. START and STOP Conditions
TWD
TWCK
Start Stop
Figure 30-4. Transfer Format
TWD
TWCK
Start Address R/W Ack Data Ack Data Ack Stop
30.7.2 Modes of Operation

The TWI has six modes of operations:
• Master transmitter mode
• Master receiver mode
476 AT91SAM9G45
• Multi-master transmitter mode
• Multi-master receiver mode
• Slave transmitter mode
• Slave receiver mode
These modes are described in the following chapters.
477 AT91SAM9G45
30.8 Master Mode
30.8.1 Definition
The Master is the device that starts a transfer, generates a clock and stops it.
30.8.2 Application Block Diagram
Figure 30-5. Master Mode Typical Application Block Diagram

VDD
Rp Rp
TWD
Host with
TWI
Interface TWCK
Atmel TWI I²C LCD I²C Temp.

I²C RTC
Serial EEPROM Controller Sensor
Slave 1 Slave 2 Slave 3 Slave 4
Rp: Pull up value as given by the I²C Standard
30.8.3 Programming Master Mode

The following registers have to be programmed before entering Master mode:
1. DADR (+ IADRSZ + IADR if a 10 bit device is addressed): The device address is used
to access slave devices in read or write mode.
2. CKDIV + CHDIV + CLDIV: Clock Waveform.
3. SVDIS: Disable the slave mode.
4. MSEN: Enable the master mode.
30.8.4 Master Transmitter Mode

After the master initiates a Start condition when writing into the Transmit Holding Register,
TWI_THR, it sends a 7-bit slave address, configured in the Master Mode register (DADR in
TWI_MMR), to notify the slave device. The bit following the slave address indicates the transfer
direction, 0 in this case (MREAD = 0 in TWI_MMR).
The TWI transfers require the slave to acknowledge each received byte. During the acknowl-
edge clock pulse (9th pulse), the master releases the data line (HIGH), enabling the slave to pull
it down in order to generate the acknowledge. The master polls the data line during this clock
pulse and sets the Not Acknowledge bit (NACK) in the status register if the slave does not
acknowledge the byte. As with the other status bits, an interrupt can be generated if enabled in
the interrupt enable register (TWI_IER). If the slave acknowledges the byte, the data written in
the TWI_THR, is then shifted in the internal shifter and transferred. When an acknowledge is
detected, the TXRDY bit is set until a new write in the TWI_THR.
While no new data is written in the TWI_THR, the Serial Clock Line is tied low. When new data is
written in the TWI_THR, the SCL is released and the data is sent. To generate a STOP event,
the STOP command must be performed by writing in the STOP field of TWI_CR.
478 AT91SAM9G45
After a Master Write transfer, the Serial Clock line is stretched (tied low) while no new data is
written in the TWI_THR or until a STOP command is performed.
See Figure 30-6, Figure 30-7, and Figure 30-8.
Figure 30-6. Master Write with One Data Byte

STOP Command sent (write in TWI_CR)
TWD S DADR W A DATA A P
TXCOMP
TXRDY
Write THR (DATA)
Figure 30-7. Master Write with Multiple Data Bytes

STOP command performed
(by writing in the TWI_CR)
TWD S DADR W A DATA n A DATA n+1 A DATA n+2 A P
TWCK
TXCOMP
TXRDY
Write THR (Data n)

Write THR (Data n+1) Write THR (Data n+2)
Last data sent
479 AT91SAM9G45
Figure 30-8. Master Write with One Byte Internal Address and Multiple Data Bytes
STOP command performed
(by writing in the TWI_CR)
TWD S DADR W A IADR A DATA n A DATA n+1 A DATA n+2 A P
TWCK
TXCOMP
TXRDY
Write THR (Data n)

Write THR (Data n+1) Write THR (Data n+2)
Last data sent
30.8.5 Master Receiver Mode

The read sequence begins by setting the START bit. After the start condition has been sent, the
master sends a 7-bit slave address to notify the slave device. The bit following the slave address
indicates the transfer direction, 1 in this case (MREAD = 1 in TWI_MMR). During the acknowl-
edge clock pulse (9th pulse), the master releases the data line (HIGH), enabling the slave to pull
it down in order to generate the acknowledge. The master polls the data line during this clock
pulse and sets the NACK bit in the status register if the slave does not acknowledge the byte.
If an acknowledge is received, the master is then ready to receive data from the slave. After data
has been received, the master sends an acknowledge condition to notify the slave that the data
has been received except for the last data, after the stop condition. See Figure 30-9. When the
RXRDY bit is set in the status register, a character has been received in the receive-holding reg-
ister (TWI_RHR). The RXRDY bit is reset when reading the TWI_RHR.
When a single data byte read is performed, with or without internal address (IADR), the START
and STOP bits must be set at the same time. See Figure 30-9. When a multiple data byte read is
performed, with or without internal address (IADR), the STOP bit must be set after the next-to-
last data received. See Figure 30-10. For Internal Address usage see Section 30.8.6.
Figure 30-9. Master Read with One Data Byte
TWD S DADR R A DATA N P
TXCOMP
Write START &

STOP Bit
RXRDY
Read RHR
480 AT91SAM9G45
Figure 30-10. Master Read with Multiple Data Bytes
TWD S DADR R A DATA n A DATA (n+1) A DATA (n+m)-1 A DATA (n+m) N P
TXCOMP
Write START Bit
RXRDY
Read RHR Read RHR Read RHR Read RHR

DATA n DATA (n+1) DATA (n+m)-1 DATA (n+m)
Write STOP Bit
after next-to-last data read
30.8.6 Internal Address

The TWI interface can perform various transfer formats: Transfers with 7-bit slave address
devices and 10-bit slave address devices.
30.8.6.1 7-bit Slave Addressing

When Addressing 7-bit slave devices, the internal address bytes are used to perform random
address (read or write) accesses to reach one or more data bytes, within a memory page loca-
tion in a serial memory, for example. When performing read operations with an internal address,
the TWI performs a write operation to set the internal address into the slave device, and then
switch to Master Receiver mode. Note that the second start condition (after sending the IADR) is
sometimes called “repeated start” (Sr) in I2C fully-compatible devices. See Figure 30-12. See
Figure 30-11 and Figure 30-13 for Master Write operation with internal address.
The three internal address bytes are configurable through the Master Mode register
(TWI_MMR).
If the slave device supports only a 7-bit address, i.e. no internal address, IADRSZ must be set to
0.
In the figures below the following abbreviations are used:
•S Start
• Sr Repeated Start
•P Stop
•W Write
•R Read
•A Acknowledge
•N Not Acknowledge
• DADR Device Address
• IADR Internal Address
481 AT91SAM9G45
Figure 30-11. Master Write with One, Two or Three Bytes Internal Address and One Data Byte
Three bytes internal address
TWD S DADR W A IADR(23:16) A IADR(15:8) A IADR(7:0) A DATA A P
Two bytes internal address

TWD S DADR W A IADR(15:8) A IADR(7:0) A DATA A P
One byte internal address

TWD S DADR W A IADR(7:0) A DATA A P
Figure 30-12. Master Read with One, Two or Three Bytes Internal Address and One Data Byte
Three bytes internal address
TWD S DADR W A IADR(23:16) A IADR(15:8) A IADR(7:0) A Sr DADR R A
DATA N P
Two bytes internal address

TWD S DADR W A IADR(15:8) A IADR(7:0) A Sr DADR R A DATA N P
One byte internal address

TWD S DADR W A IADR(7:0) A Sr DADR R A DATA N P
30.8.6.2 10-bit Slave Addressing

For a slave address higher than 7 bits, the user must configure the address size (IADRSZ) and
set the other slave address bits in the internal address register (TWI_IADR). The two remaining
Internal address bytes, IADR[15:8] and IADR[23:16] can be used the same as in 7-bit Slave
Addressing.
Example: Address a 10-bit device (10-bit device address is b1 b2 b3 b4 b5 b6 b7 b8 b9 b10)
1. Program IADRSZ = 1,
2. Program DADR with 1 1 1 1 0 b1 b2 (b1 is the MSB of the 10-bit address, b2, etc.)
3. Program TWI_IADR with b3 b4 b5 b6 b7 b8 b9 b10 (b10 is the LSB of the 10-bit
address)
Figure 30-13 below shows a byte write to an Atmel AT24LC512 EEPROM. This demonstrates
the use of internal addresses to access the device.
Figure 30-13. Internal Address Usage

S W
T R S
A I T
R Device T FIRST SECOND O
T Address E WORD ADDRESS WORD ADDRESS DATA P
0
M LR A M A LA A
S S / C S C SC C
B BW K B K BK K
482 AT91SAM9G45
30.8.7 SMBUS Quick Command (Master Mode Only)
The TWI interface can perform a Quick Command:
1. Configure the master mode (DADR, CKDIV, etc.).
2. Write the MREAD bit in the TWI_MMR register at the value of the one-bit command to
be sent.
3. Start the transfer by setting the QUICK bit in the TWI_CR.
Figure 30-14. SMBUS Quick Command
TWD S DADR R/W A P
TXCOMP
TXRDY
Write QUICK command in TWI_CR
30.8.8 Read-write Flowcharts

The following flowcharts shown in Figure 30-16 on page 485, Figure 30-17 on page 486, Figure
30-18 on page 487, Figure 30-19 on page 488 and Figure 30-20 on page 489 give examples for
read and write operations. A polling or interrupt method can be used to check the status bits.
The interrupt method requires that the interrupt enable register (TWI_IER) be configured first.
483 AT91SAM9G45
Figure 30-15. TWI Write Operation with Single Data Byte without Internal Address
BEGIN
Set TWI clock

(CLDIV, CHDIV, CKDIV) in TWI_CWGR
(Needed only once)
Set the Control register:

- Master enable
TWI_CR = MSEN + SVDIS
Set the Master Mode register:

- Device slave address (DADR)
- Transfer direction bit
Write ==> bit MREAD = 0
Load Transmit register

TWI_THR = Data to send
Write STOP Command

TWI_CR = STOP
Read Status register
No
TXRDY = 1?
Yes
No
TXCOMP = 1?
Yes
Transfer finished
484 AT91SAM9G45
Figure 30-16. TWI Write Operation with Single Data Byte and Internal Address
BEGIN
Set TWI clock

(Needed only once)

- Master enable

- Device slave address (DADR)
- Internal address size (IADRSZ)
Set the internal address

TWI_IADR = address
Load transmit register

Write STOP command

TWI_CR = STOP
No
TXRDY = 1?
Yes
TXCOMP = 1?
No
Yes
Transfer finished
485 AT91SAM9G45
Figure 30-17. TWI Write Operation with Multiple Data Bytes with or without Internal Address
BEGIN
Set TWI clock

(Needed only once)

- Master enable

- Device slave address
- Internal address size (if IADR used)
No
Internal address size = 0?

TWI_IADR = address
Yes
Load Transmit register

TWI_THR = data to send No

TXRDY = 1?
Yes
Data to send?
Yes
Write STOP Command

TWI_CR = STOP
Yes
No
TXCOMP = 1?
END
486 AT91SAM9G45
Figure 30-18. TWI Read Operation with Single Data Byte without Internal Address
BEGIN
Set TWI clock

(Needed only once)

- Master enable

Read ==> bit MREAD = 1
Start the transfer

TWI_CR = START | STOP
Read status register
No
RXRDY = 1?
Yes
Read Receive Holding Register
No
TXCOMP = 1?
Yes
END
487 AT91SAM9G45
Figure 30-19. TWI Read Operation with Single Data Byte and Internal Address
BEGIN
Set TWI clock

(Needed only once)

- Master enable

- Internal address size (IADRSZ)

TWI_IADR = address
Start the transfer

TWI_CR = START | STOP
No
RXRDY = 1?
Yes
Read Receive Holding register
No
TXCOMP = 1?
Yes
END
488 AT91SAM9G45
Figure 30-20. TWI Read Operation with Multiple Data Bytes with or without Internal Address
BEGIN
Set TWI clock

(Needed only once)

- Master enable

- Internal address size (if IADR used)
Internal address size = 0?

TWI_IADR = address
Yes
Start the transfer

TWI_CR = START
No
RXRDY = 1?
Yes
Read Receive Holding register (TWI_RHR)
No Last data to read

but one?
Yes
Stop the transfer

TWI_CR = STOP
No
RXRDY = 1?
Yes
Read Receive Holding register (TWI_RHR)
Read status register
No
TXCOMP = 1?
Yes
END
489 AT91SAM9G45
30.9 Multi-master Mode
30.9.1 Definition
More than one master may handle the bus at the same time without data corruption by using
arbitration.
Arbitration starts as soon as two or more masters place information on the bus at the same time,
and stops (arbitration is lost) for the master that intends to send a logical one while the other
master sends a logical zero.
As soon as arbitration is lost by a master, it stops sending data and listens to the bus in order to
detect a stop. When the stop is detected, the master who has lost arbitration may put its data on
the bus by respecting arbitration.
Arbitration is illustrated in Figure 30-22 on page 491.
30.9.2 Different Multi-master Modes

Two multi-master modes may be distinguished:
1. TWI is considered as a Master only and will never be addressed.
2. TWI may be either a Master or a Slave and may be addressed.
Note: In both Multi-master modes arbitration is supported.
30.9.2.1 TWI as Master Only

In this mode, TWI is considered as a Master only (MSEN is always at one) and must be driven
like a Master with the ARBLST (ARBitration Lost) flag in addition.
If arbitration is lost (ARBLST = 1), the programmer must reinitiate the data transfer.
If the user starts a transfer (ex.: DADR + START + W + Write in THR) and if the bus is busy, the
TWI automatically waits for a STOP condition on the bus to initiate the transfer (see Figure 30-
21 on page 491).
Note: The state of the bus (busy or free) is not indicated in the user interface.
30.9.2.2 TWI as Master or Slave

The automatic reversal from Master to Slave is not supported in case of a lost arbitration.
Then, in the case where TWI may be either a Master or a Slave, the programmer must manage
the pseudo Multi-master mode described in the steps below.
1. Program TWI in Slave mode (SADR + MSDIS + SVEN) and perform Slave Access (if
TWI is addressed).
2. If TWI has to be set in Master mode, wait until TXCOMP flag is at 1.
3. Program Master mode (DADR + SVDIS + MSEN) and start the transfer (ex: START +
Write in THR).
4. As soon as the Master mode is enabled, TWI scans the bus in order to detect if it is
busy or free. When the bus is considered as free, TWI initiates the transfer.
5. As soon as the transfer is initiated and until a STOP condition is sent, the arbitration
becomes relevant and the user must monitor the ARBLST flag.
6. If the arbitration is lost (ARBLST is set to 1), the user must program the TWI in Slave
mode in the case where the Master that won the arbitration wanted to access the TWI.
7. If TWI has to be set in Slave mode, wait until TXCOMP flag is at 1 and then program the
Slave mode.
490 AT91SAM9G45
Note: In the case where the arbitration is lost and TWI is addressed, TWI will not acknowledge even if it
is programmed in Slave mode as soon as ARBLST is set to 1. Then, the Master must repeat
SADR.
Figure 30-21. Programmer Sends Data While the Bus is Busy
TWCK
STOP sent by the master START sent by the TWI
TWD DATA sent by a master DATA sent by the TWI

Bus is busy
Bus is free
TWI DATA transfer Transfer is kept
A transfer is programmed Bus is considered as free

(DADR + W + START + Write THR) Transfer is initiated
Figure 30-22. Arbitration Cases
TWCK
TWD
TWCK
Arbitration is lost
Data from a Master S 1 0 0 1 1 P S 1 0 1
The master stops sending data
Arbitration is lost
Data from TWI S 1 0 1 S 1 0 0 1 1
TWI stops sending data
TWD S 1 0 0 1 1 Data from the master P S 1 0 0 1 1 Data from the TWI
ARBLST
Bus is busy Bus is free
TWI DATA transfer Transfer is kept
A transfer is programmed Transfer is stopped Bus is considered as free

Transfer is programmed again Transfer is initiated
(DADR + W + START + Write THR)
(DADR + W + START + Write THR)
The flowchart shown in Figure 30-23 on page 492 gives an example of read and write operations
in Multi-master mode.
491 AT91SAM9G45
Figure 30-23. Multi-master Flowchart
START
Programm the SLAVE mode:

SADR + MSDIS + SVEN
Read Status Register
Yes No
SVACC = 1 ? GACC = 1 ?
No
No SVREAD = 0 ?
No No
EOSACC = 1 ? Yes TXRDY= 1 ?
Yes Yes
No Write in TWI_THR
TXCOMP = 1 ? No
RXRDY= 0 ?
Yes
Yes
No Need to perform Read TWI_RHR

a master access ?
GENERAL CALL TREATMENT
Yes
Decoding of the
programming sequence
Prog seq No
OK ?
Change SADR
Program the Master mode

DADR + SVDIS + MSEN + CLK + R / W
Yes No
ARBLST = 1 ?
Yes No
MREAD = 1 ?
Yes Yes
RXRDY= 0 ? TXRDY= 0 ?
No No
Yes Yes
Read TWI_RHR Data to read? Data to send ? Write in TWI_THR
No No
Stop Transfer
TWI_CR = STOP
Yes No
TXCOMP = 0 ?
492 AT91SAM9G45
30.10 Slave Mode
30.10.1 Definition
The Slave Mode is defined as a mode where the device receives the clock and the address from
another device called the master.
In this mode, the device never initiates and never completes the transmission (START,
REPEATED_START and STOP conditions are always provided by the master).
30.10.2 Application Block Diagram
Figure 30-24. Slave Mode Typical Application Block Diagram
VDD
R R
Master
TWD
Host with
TWI
Interface TWCK
Host with TWI Host with TWI

LCD Controller
Interface Interface
Slave 1 Slave 2 Slave 3
30.10.3 Programming Slave Mode

The following fields must be programmed before entering Slave mode:
1. SADR (TWI_SMR): The slave device address is used in order to be accessed by mas-
ter devices in read or write mode.
2. MSDIS (TWI_CR): Disable the master mode.
3. SVEN (TWI_CR): Enable the slave mode.
As the device receives the clock, values written in TWI_CWGR are not taken into account.
30.10.4 Receiving Data

After a Start or Repeated Start condition is detected and if the address sent by the Master
matches with the Slave address programmed in the SADR (Slave ADdress) field, SVACC (Slave
ACCess) flag is set and SVREAD (Slave READ) indicates the direction of the transfer.
SVACC remains high until a STOP condition or a repeated START is detected. When such a
condition is detected, EOSACC (End Of Slave ACCess) flag is set.
30.10.4.1 Read Sequence

In the case of a Read sequence (SVREAD is high), TWI transfers data written in the TWI_THR
(TWI Transmit Holding Register) until a STOP condition or a REPEATED_START + an address
different from SADR is detected. Note that at the end of the read sequence TXCOMP (Transmis-
sion Complete) flag is set and SVACC reset.
As soon as data is written in the TWI_THR, TXRDY (Transmit Holding Register Ready) flag is
reset, and it is set when the shift register is empty and the sent data acknowledged or not. If the
data is not acknowledged, the NACK flag is set.
493 AT91SAM9G45
Note that a STOP or a repeated START always follows a NACK.
See Figure 30-25 on page 495.
30.10.4.2 Write Sequence

In the case of a Write sequence (SVREAD is low), the RXRDY (Receive Holding Register
Ready) flag is set as soon as a character has been received in the TWI_RHR (TWI Receive
Holding Register). RXRDY is reset when reading the TWI_RHR.
TWI continues receiving data until a STOP condition or a REPEATED_START + an address dif-
ferent from SADR is detected. Note that at the end of the write sequence TXCOMP flag is set
and SVACC reset.
30.10.4.3 Clock Synchronization Sequence

In the case where TWI_THR or TWI_RHR is not written/read in time, TWI performs a clock
synchronization.
Clock stretching information is given by the SCLWS (Clock Wait state) bit.
See Figure 30-28 on page 497 and Figure 30-29 on page 498.
30.10.4.4 General Call

In the case where a GENERAL CALL is performed, GACC (General Call ACCess) flag is set.
After GACC is set, it is up to the programmer to interpret the meaning of the GENERAL CALL
and to decode the new address programming sequence.
30.10.4.5
30.10.5 Data Transfer
30.10.5.1 Read Operation

The read mode is defined as a data requirement from the master.
After a START or a REPEATED START condition is detected, the decoding of the address
starts. If the slave address (SADR) is decoded, SVACC is set and SVREAD indicates the direc-
tion of the transfer.
Until a STOP or REPEATED START condition is detected, TWI continues sending data loaded
in the TWI_THR register.
If a STOP condition or a REPEATED START + an address different from SADR is detected,
SVACC is reset.
Figure 30-25 on page 495 describes the write operation.
494 AT91SAM9G45
Figure 30-25. Read Access Ordered by a MASTER
SADR does not match, SADR matches,
TWI answers with a NACK TWI answers with an ACK
ACK/NACK from the Master
TWD S ADR R NA DATA NA P/S/Sr SADR R A DATA A A DATA NA S/Sr
TXRDY
Write THR Read RHR
NACK
SVACC
SVREAD SVREAD has to be taken into account only while SVACC is active
EOSVACC
Notes: 1. When SVACC is low, the state of SVREAD becomes irrelevant.

2. TXRDY is reset when data has been transmitted from TWI_THR to the shift register and set when this data has been
acknowledged or non acknowledged.
30.10.5.2 Write Operation

The write mode is defined as a data transmission from the master.
After a START or a REPEATED START, the decoding of the address starts. If the slave address
is decoded, SVACC is set and SVREAD indicates the direction of the transfer (SVREAD is low in
this case).
Until a STOP or REPEATED START condition is detected, TWI stores the received data in the
TWI_RHR register.
If a STOP condition or a REPEATED START + an address different from SADR is detected,
SVACC is reset.
Figure 30-26 on page 495 describes the Write operation.
Figure 30-26. Write Access Ordered by a Master

SADR does not match, SADR matches,
TWI answers with a NACK TWI answers with an ACK
Read RHR
TWD S ADR W NA DATA NA P/S/Sr SADR W A DATA A A DATA NA S/Sr
RXRDY
SVACC
SVREAD SVREAD has to be taken into account only while SVACC is active
EOSVACC
Notes: 1. When SVACC is low, the state of SVREAD becomes irrelevant.

2. RXRDY is set when data has been transmitted from the shift register to the TWI_RHR and reset when this data is read.
495 AT91SAM9G45
30.10.5.3 General Call
The general call is performed in order to change the address of the slave.
If a GENERAL CALL is detected, GACC is set.
After the detection of General Call, it is up to the programmer to decode the commands which
come afterwards.
In case of a WRITE command, the programmer has to decode the programming sequence and
program a new SADR if the programming sequence matches.
Figure 30-27 on page 496 describes the General Call access.
Figure 30-27. Master Performs a General Call

RESET command = 00000110X
0000000 + W WRITE command = 00000100X
TXD S GENERAL CALL A Reset or write DADD A DATA1 A DATA2 A New SADR A P
New SADR
Programming sequence
GCACC
Reset after read
SVACC
Note: This method allows the user to create an own programming sequence by choosing the program-
ming bytes and the number of them. The programming sequence has to be provided to the
master.
496 AT91SAM9G45
30.10.5.4 Clock Synchronization
In both read and write modes, it may happen that TWI_THR/TWI_RHR buffer is not filled /emp-
tied before the emission/reception of a new character. In this case, to avoid sending/receiving
undesired data, a clock stretching mechanism is implemented.
30.10.5.5 Clock Synchronization in Read Mode

The clock is tied low if the shift register is empty and if a STOP or REPEATED START condition
was not detected. It is tied low until the shift register is loaded.
Figure 30-28 on page 497 describes the clock synchronization in Read mode.
Figure 30-28. Clock Synchronization in Read Mode
TWI_THR DATA0 1 DATA1 DATA2
S SADR R A DATA0 A DATA1 A XXXXXXX DATA2 NA S
TWCK
CLOCK is tied low by the TWI
as long as THR is empty
Write THR
SCLWS
TXRDY
SVACC
SVREAD
As soon as a START is detected
TXCOMP
TWI_THR is transmitted to the shift register Ack or Nack from the master
1 The data is memorized in TWI_THR until a new value is written

2 The clock is stretched after the ACK, the state of TWD is undefined during clock stretching
Notes: 1. TXRDY is reset when data has been written in the TWI_THR to the shift register and set when this data has been acknowl-
edged or non acknowledged.
2. At the end of the read sequence, TXCOMP is set after a STOP or after a REPEATED_START + an address different from
SADR.
3. SCLWS is automatically set when the clock synchronization mechanism is started.
497 AT91SAM9G45
30.10.5.6 Clock Synchronization in Write Mode
The c lock is tied lo w if the shift register and the TWI_RHR is full. If a STOP or
REPEATED_START condition was not detected, it is tied low until TWI_RHR is read.
Figure 30-29 on page 498 describes the clock synchronization in Read mode.
Figure 30-29. Clock Synchronization in Write Mode
TWCK
CLOCK is tied low by the TWI as long as RHR is full
TWD S SADR W A DATA0 A DATA1 A DATA2 NA S ADR
TWI_RHR DATA0 is not read in the RHR DATA1 DATA2
SCLWS
SCL is stretched on the last bit of DATA1
RXRDY
Rd DATA0 Rd DATA1 Rd DATA2
SVACC
SVREAD
As soon as a START is detected
TXCOMP
Notes: 1. At the end of the read sequence, TXCOMP is set after a STOP or after a REPEATED_START + an address different from
SADR.
2. SCLWS is automatically set when the clock synchronization mechanism is started and automatically reset when the mecha-
nism is finished.
498 AT91SAM9G45
30.10.5.7 Reversal after a Repeated Start
30.10.5.8 Reversal of Read to Write

The master initiates the communication by a read command and finishes it by a write command.
Figure 30-30 on page 499 describes the repeated start + reversal from Read to Write mode.
Figure 30-30. Repeated Start + Reversal from Read to Write Mode
TWI_THR
DATA0 DATA1
TWD S SADR R A DATA0 A DATA1 NA Sr SADR W A DATA2 A DATA3 A P
TWI_RHR DATA2 DATA3
SVACC
SVREAD
TXRDY
RXRDY
EOSACC Cleared after read
TXCOMP As soon as a START is detected
1. TXCOMP is only set at the end of the transmission because after the repeated start, SADR is detected again.
30.10.5.9 Reversal of Write to Read

The master initiates the communication by a write command and finishes it by a read com-
mand.Figure 30-31 on page 499 describes the repeated start + reversal from Write to Read
mode.
Figure 30-31. Repeated Start + Reversal from Write to Read Mode
TWI_THR DATA2 DATA3
TWD S SADR W A DATA0 A DATA1 A Sr SADR R A DATA2 A DATA3 NA P
TWI_RHR DATA0 DATA1
SVACC
SVREAD
TXRDY
RXRDY
Read TWI_RHR
EOSACC Cleared after read
TXCOMP As soon as a START is detected
Notes: 1. In this case, if TWI_THR has not been written at the end of the read command, the clock is automatically stretched before
the ACK.
2. TXCOMP is only set at the end of the transmission because after the repeated start, SADR is detected again.
499 AT91SAM9G45
30.10.6 Read Write Flowcharts
The flowchart shown in Figure 30-32 on page 500 gives an example of read and write operations
in Slave mode. A polling or interrupt method can be used to check the status bits. The interrupt
method requires that the interrupt enable register (TWI_IER) be configured first.
Figure 30-32. Read Write Flowchart in Slave Mode
Set the SLAVE mode:

SADR + MSDIS + SVEN
No
SVACC = 1 ? GACC = 1 ?
No SVREAD = 0 ?
No No
EOSACC = 1 ? No TXRDY= 1 ?
No Write in TWI_THR
TXCOMP = 1 ?
No
RXRDY= 0 ?
END
Read TWI_RHR
GENERAL CALL TREATMENT
Decoding of the
programming sequence
Prog seq No
OK ?
Change SADR
500 AT91SAM9G45
30.11 Two-wire Interface (TWI) User Interface

0x00 Control Register TWI_CR Write-only N/A
0x04 Master Mode Register TWI_MMR Read-write 0x00000000
0x08 Slave Mode Register TWI_SMR Read-write 0x00000000
0x0C Internal Address Register TWI_IADR Read-write 0x00000000
0x10 Clock Waveform Generator Register TWI_CWGR Read-write 0x00000000
0x20 Status Register TWI_SR Read-only 0x0000F009
0x24 Interrupt Enable Register TWI_IER Write-only N/A
0x28 Interrupt Disable Register TWI_IDR Write-only N/A
0x2C Interrupt Mask Register TWI_IMR Read-only 0x00000000
0x30 Receive Holding Register TWI_RHR Read-only 0x00000000
0x34 Transmit Holding Register TWI_THR Write-only 0x00000000
0x38 - 0xFC Reserved – – –
– – –
501 AT91SAM9G45
30.11.1 TWI Control Register
Name: TWI_CR
Addresses: 0xFFF84000 (0), 0xFFF88000 (1)
Access: Write-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
SWRST QUICK SVDIS SVEN MSDIS MSEN STOP START
• START: Send a START Condition

0 = No effect.
1 = A frame beginning with a START bit is transmitted according to the features defined in the mode register.
This action is necessary when the TWI peripheral wants to read data from a slave. When configured in Master Mode with a
write operation, a frame is sent as soon as the user writes a character in the Transmit Holding Register (TWI_THR).
• STOP: Send a STOP Condition

0 = No effect.
1 = STOP Condition is sent just after completing the current byte transmission in master read mode.
– In single data byte master read, the START and STOP must both be set.
– In multiple data bytes master read, the STOP must be set after the last data received but one.
– In master read mode, if a NACK bit is received, the STOP is automatically performed.
– In master data write operation, a STOP condition will be sent after the transmission of the current data is
finished.
• MSEN: TWI Master Mode Enabled

0 = No effect.
1 = If MSDIS = 0, the master mode is enabled.
Note: Switching from Slave to Master mode is only permitted when TXCOMP = 1.
• MSDIS: TWI Master Mode Disabled

0 = No effect.
1 = The master mode is disabled, all pending data is transmitted. The shifter and holding characters (if it contains data) are
transmitted in case of write operation. In read operation, the character being transferred must be completely received
before disabling.
502 AT91SAM9G45
• SVEN: TWI Slave Mode Enabled
0 = No effect.
1 = If SVDIS = 0, the slave mode is enabled.
Note: Switching from Master to Slave mode is only permitted when TXCOMP = 1.
• SVDIS: TWI Slave Mode Disabled

0 = No effect.
1 = The slave mode is disabled. The shifter and holding characters (if it contains data) are transmitted in case of read oper-
ation. In write operation, the character being transferred must be completely received before disabling.
• QUICK: SMBUS Quick Command

0 = No effect.
1 = If Master mode is enabled, a SMBUS Quick Command is sent.
• SWRST: Software Reset

0 = No effect.
1 = Equivalent to a system reset.
503 AT91SAM9G45
30.11.2 TWI Master Mode Register
Name: TWI_MMR
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– DADR
15 14 13 12 11 10 9 8
– – – MREAD – – IADRSZ
7 6 5 4 3 2 1 0
– – – – – – – –
• IADRSZ: Internal Device Address Size
IADRSZ[9:8]
0 0 No internal device address
0 1 One-byte internal device address
1 0 Two-byte internal device address
1 1 Three-byte internal device address
• MREAD: Master Read Direction

0 = Master write direction.
1 = Master read direction.
• DADR: Device Address

The device address is used to access slave devices in read or write mode. Those bits are only used in Master mode.
504 AT91SAM9G45
30.11.3 TWI Slave Mode Register
Name: TWI_SMR
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– SADR
15 14 13 12 11 10 9 8
– – – – – –
7 6 5 4 3 2 1 0
– – – – – – – –
• SADR: Slave Address

The slave device address is used in Slave mode in order to be accessed by master devices in read or write mode.
SADR must be programmed before enabling the Slave mode or after a general call. Writes at other times have no effect.
505 AT91SAM9G45
30.11.4 TWI Internal Address Register
Name: TWI_IADR
Addresses: 0xFFF8400C (0), 0xFFF8800C (1)
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
IADR
15 14 13 12 11 10 9 8
IADR
7 6 5 4 3 2 1 0
IADR
• IADR: Internal Address

0, 1, 2 or 3 bytes depending on IADRSZ.
506 AT91SAM9G45
30.11.5 TWI Clock Waveform Generator Register
Name: TWI_CWGR
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
CKDIV
15 14 13 12 11 10 9 8
CHDIV
7 6 5 4 3 2 1 0
CLDIV
TWI_CWGR is only used in Master mode.
• CLDIV: Clock Low Divider

The SCL low period is defined as follows:
CKDIV
T low = ( ( CLDIV × 2 ) + 4 ) × T MCK
• CHDIV: Clock High Divider

The SCL high period is defined as follows:
CKDIV
T high = ( ( CHDIV × 2 ) + 4 ) × T MCK
• CKDIV: Clock Divider

The CKDIV is used to increase both SCL high and low periods.
507 AT91SAM9G45
30.11.6 TWI Status Register
Name: TWI_SR
Access: Read-only
Reset: 0x0000F009
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
EOSACC SCLWS ARBLST NACK
7 6 5 4 3 2 1 0
– OVRE GACC SVACC SVREAD TXRDY RXRDY TXCOMP
• TXCOMP: Transmission Completed (automatically set / reset)

TXCOMP used in Master mode:
0 = During the length of the current frame.
1 = When both holding and shifter registers are empty and STOP condition has been sent.
TXCOMP behavior in Master mode can be seen in Figure 30-8 on page 480 and in Figure 30-10 on page 481.
TXCOMP used in Slave mode:
0 = As soon as a Start is detected.
1 = After a Stop or a Repeated Start + an address different from SADR is detected.
TXCOMP behavior in Slave mode can be seen in Figure 30-28 on page 497, Figure 30-29 on page 498, Figure 30-30 on
page 499 and Figure 30-31 on page 499.
• RXRDY: Receive Holding Register Ready (automatically set / reset)

0 = No character has been received since the last TWI_RHR read operation.
1 = A byte has been received in the TWI_RHR since the last read.
RXRDY behavior in Master mode can be seen in Figure 30-10 on page 481.
RXRDY behavior in Slave mode can be seen in Figure 30-26 on page 495, Figure 30-29 on page 498, Figure 30-30 on
• TXRDY: Transmit Holding Register Ready (automatically set / reset)

TXRDY used in Master mode:
0 = The transmit holding register has not been transferred into shift register. Set to 0 when writing into TWI_THR register.
1 = As soon as a data byte is transferred from TWI_THR to internal shifter or if a NACK error is detected, TXRDY is set at
the same time as TXCOMP and NACK. TXRDY is also set when MSEN is set (enable TWI).
TXRDY behavior in Master mode can be seen in Figure 30-8 on page 480.
508 AT91SAM9G45
TXRDY used in Slave mode:
0 = As soon as data is written in the TWI_THR, until this data has been transmitted and acknowledged (ACK or NACK).
1 = It indicates that the TWI_THR is empty and that data has been transmitted and acknowledged.
If TXRDY is high and if a NACK has been detected, the transmission will be stopped. Thus when TRDY = NACK = 1, the
programmer must not fill TWI_THR to avoid losing it.
TXRDY behavior in Slave mode can be seen in Figure 30-25 on page 495, Figure 30-28 on page 497, Figure 30-30 on
• SVREAD: Slave Read (automatically set / reset)

This bit is only used in Slave mode. When SVACC is low (no Slave access has been detected) SVREAD is irrelevant.
0 = Indicates that a write access is performed by a Master.
1 = Indicates that a read access is performed by a Master.
SVREAD behavior can be seen in Figure 30-25 on page 495, Figure 30-26 on page 495, Figure 30-30 on page 499 and
Figure 30-31 on page 499.
• SVACC: Slave Access (automatically set / reset)

This bit is only used in Slave mode.
0 = TWI is not addressed. SVACC is automatically cleared after a NACK or a STOP condition is detected.
1 = Indicates that the address decoding sequence has matched (A Master has sent SADR). SVACC remains high until a
NACK or a STOP condition is detected.
SVACC behavior can be seen in Figure 30-25 on page 495, Figure 30-26 on page 495, Figure 30-30 on page 499 and Fig-
ure 30-31 on page 499.
• GACC: General Call Access (clear on read)

0 = No General Call has been detected.
1 = A General Call has been detected. After the detection of General Call, if need be, the programmer may acknowledge
this access and decode the following bytes and respond according to the value of the bytes.
GACC behavior can be seen in Figure 30-27 on page 496.
• OVRE: Overrun Error (clear on read)

This bit is only used in Master mode.
0 = TWI_RHR has not been loaded while RXRDY was set
1 = TWI_RHR has been loaded while RXRDY was set. Reset by read in TWI_SR when TXCOMP is set.
• NACK: Not Acknowledged (clear on read)

NACK used in Master mode:
0 = Each data byte has been correctly received by the far-end side TWI slave component.
1 = A data byte has not been acknowledged by the slave component. Set at the same time as TXCOMP.
509 AT91SAM9G45
NACK used in Slave Read mode:
0 = Each data byte has been correctly received by the Master.
1 = In read mode, a data byte has not been acknowledged by the Master. When NACK is set the programmer must not fill
TWI_THR even if TXRDY is set, because it means that the Master will stop the data transfer or re initiate it.
Note that in Slave Write mode all data are acknowledged by the TWI.
• ARBLST: Arbitration Lost (clear on read)

This bit is only used in Master mode.
0: Arbitration won.
1: Arbitration lost. Another master of the TWI bus has won the multi-master arbitration. TXCOMP is set at the same time.
• SCLWS: Clock Wait State (automatically set / reset)

0 = The clock is not stretched.
1 = The clock is stretched. TWI_THR / TWI_RHR buffer is not filled / emptied before the emission / reception of a new
character.
SCLWS behavior can be seen in Figure 30-28 on page 497 and Figure 30-29 on page 498.
• EOSACC: End Of Slave Access (clear on read)

0 = A slave access is being performing.
1 = The Slave Access is finished. End Of Slave Access is automatically set as soon as SVACC is reset.
EOSACC behavior can be seen in Figure 30-30 on page 499 and Figure 30-31 on page 499
510 AT91SAM9G45
30.11.7 TWI Interrupt Enable Register
Name: TWI_IER
Access: Write-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
EOSACC SCL_WS ARBLST NACK
7 6 5 4 3 2 1 0
– OVRE GACC SVACC – TXRDY RXRDY TXCOMP
• TXCOMP: Transmission Completed Interrupt Enable
• RXRDY: Receive Holding Register Ready Interrupt Enable
• TXRDY: Transmit Holding Register Ready Interrupt Enable
• SVACC: Slave Access Interrupt Enable
• GACC: General Call Access Interrupt Enable
• OVRE: Overrun Error Interrupt Enable
• NACK: Not Acknowledge Interrupt Enable
• ARBLST: Arbitration Lost Interrupt Enable
• SCL_WS: Clock Wait State Interrupt Enable
• EOSACC: End Of Slave Access Interrupt Enable

0 = No effect.
511 AT91SAM9G45
30.11.8 TWI Interrupt Disable Register
Name: TWI_IDR
Access: Write-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• TXCOMP: Transmission Completed Interrupt Disable
• RXRDY: Receive Holding Register Ready Interrupt Disable
• TXRDY: Transmit Holding Register Ready Interrupt Disable
• SVACC: Slave Access Interrupt Disable
• GACC: General Call Access Interrupt Disable
• OVRE: Overrun Error Interrupt Disable
• NACK: Not Acknowledge Interrupt Disable
• ARBLST: Arbitration Lost Interrupt Disable
• SCL_WS: Clock Wait State Interrupt Disable
• EOSACC: End Of Slave Access Interrupt Disable

0 = No effect.
512 AT91SAM9G45
30.11.9 TWI Interrupt Mask Register
Name: TWI_IMR
Addresses: 0xFFF8402C (0), 0xFFF8802C (1)
Access: Read-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• TXCOMP: Transmission Completed Interrupt Mask
• RXRDY: Receive Holding Register Ready Interrupt Mask
• TXRDY: Transmit Holding Register Ready Interrupt Mask
• SVACC: Slave Access Interrupt Mask
• GACC: General Call Access Interrupt Mask
• OVRE: Overrun Error Interrupt Mask
• NACK: Not Acknowledge Interrupt Mask
• ARBLST: Arbitration Lost Interrupt Mask
• SCL_WS: Clock Wait State Interrupt Mask
• EOSACC: End Of Slave Access Interrupt Mask

513 AT91SAM9G45
30.11.10 TWI Receive Holding Register
Name: TWI_RHR
Access: Read-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
RXDATA
• RXDATA: Master or Slave Receive Holding Data
30.11.11 TWI Transmit Holding Register

Name: TWI_THR
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
TXDATA
• TXDATA: Master or Slave Transmit Holding Data
514 AT91SAM9G45
AT91SAM9G45
31. Universal Synchronous Asynchronous Receiver Transeiver (USART)
31.1 Description
The Universal Synchronous Asynchronous Receiver Transceiver (USART) provides one full
duplex universal synchronous asynchronous serial link. Data frame format is widely programma-
ble (data length, parity, number of stop bits) to support a maximum of standards. The receiver
implements parity error, framing error and overrun error detection. The receiver time-out enables
handling variable-length frames and the transmitter timeguard facilitates communications with
slow remote devices. Multidrop communications are also supported through address bit han-
dling in reception and transmission.
The USART features three test modes: remote loopback, local loopback and automatic echo.
The USART supports specific operating modes providing interfaces on RS485 buses, with
ISO7816 T = 0 or T = 1 smart card slots and infrared transceivers. The hardware handshaking
feature enables an out-of-band flow control by automatic management of the pins RTS and
CTS.
The USART supports the connection to the Peripheral DMA Controller, which enables data
transfers to the transmitter and from the receiver. The PDC provides chained buffer manage-
ment without any intervention of the processor.

• Programmable Baud Rate Generator
• 5- to 9-bit full-duplex synchronous or asynchronous serial communications
– 1, 1.5 or 2 stop bits in Asynchronous Mode or 1 or 2 stop bits in Synchronous Mode
– Parity generation and error detection
– Framing error detection, overrun error detection
– MSB- or LSB-first
– Optional break generation and detection
– By 8 or by-16 over-sampling receiver frequency
– Hardware handshaking RTS-CTS
– Receiver time-out and transmitter timeguard
– Optional Multi-drop Mode with address generation and detection
– Optional Manchester Encoding
• RS485 with driver control signal
• ISO7816, T = 0 or T = 1 Protocols for interfacing with smart cards
– NACK handling, error counter with repetition and iteration limit
• IrDA modulation and demodulation
– Communication at up to 115.2 Kbps
• Test Modes
– Remote Loopback, Local Loopback, Automatic Echo
515
31.3 Block Diagram
Figure 31-1. USART Block Diagram
Peripheral DMA
Controller
Channel Channel
PIO
USART Controller
RXD
Receiver
RTS
AIC USART TXD

Interrupt Transmitter
CTS
PMC
MCK SCK
Baud Rate
Generator
MCK/DIV
DIV
User Interface
SLCK
APB
516 AT91SAM9G45
AT91SAM9G45
PPP IrLAP
Field Bus EMV
Serial Driver Driver IrDA
Driver Driver
USART
RS232 RS485 Smart IrDA

Drivers Drivers Card Transceivers
Slot
Serial Differential
Port Bus
517
AT91SAM9G45

Name Description Type Active Level
SCK Serial Clock I/O
TXD Transmit Serial Data I/O
RXD Receive Serial Data Input
CTS Clear to Send Input Low
RTS Request to Send Output Low
518
AT91SAM9G45
31.6.1 I/O Lines

The pins used for interfacing the USART may be multiplexed with the PIO lines. The program-
mer must first program the PIO controller to assign the desired USART pins to their peripheral
function. If I/O lines of the USART are not used by the application, they can be used for other
purposes by the PIO Controller.
To prevent the TXD line from falling when the USART is disabled, the use of an internal pull up
is mandatory. If the hardware handshaking feature is used, the internal pull up on TXD must
also be enabled.

USART0 CTS0 PB15 B
USART0 RTS0 PB17 B
USART0 RXD0 PB18 A
USART0 SCK0 PB16 B
USART0 TXD0 PB19 A
USART1 CTS1 PD17 A
USART1 RTS1 PD16 A
USART1 RXD1 PB5 A
USART1 SCK1 PD29 B
USART1 TXD1 PB4 A
USART2 CTS2 PC11 B
USART2 RTS2 PC9 B
USART2 RXD2 PB7 A
USART2 SCK2 PD30 B
USART2 TXD2 PB6 A
USART3 CTS3 PA24 B
USART3 RTS3 PA23 B
USART3 RXD3 PB9 A
USART3 SCK3 PA22 B
USART3 TXD3 PB8 A

The USART is not continuously clocked. The programmer must first enable the USART Clock in
the Power Management Controller (PMC) before using the USART. However, if the application
does not require USART operations, the USART clock can be stopped when not needed and be
restarted later. In this case, the USART will resume its operations where it left off.
Configuring the USART does not require the USART clock to be enabled.
519
AT91SAM9G45
31.6.3 Interrupt
The USART interrupt line is connected on one of the internal sources of the Advanced Inter-rupt
Controller. Using the USART interrupt requires the AIC to be programmed first. Note that it is not
recommended to use the USART interrupt line in edge sensitive mode.

Instance ID
USART0 7
USART1 8
USART2 9
USART3 10

The USART is capable of managing several types of serial synchronous or asynchronous
communications.
It supports the following communication modes:
• 5- to 9-bit full-duplex asynchronous serial communication
– 1, 1.5 or 2 stop bits
– Parity even, odd, marked, space or none
– By 8 or by 16 over-sampling receiver frequency
– Optional hardware handshaking
– Optional break management
– Optional multidrop serial communication
• High-speed 5- to 9-bit full-duplex synchronous serial communication
– 1 or 2 stop bits
– Parity even, odd, marked, space or none
– By 8 or by 16 over-sampling frequency
– Optional hardware handshaking
– Optional break management
– Optional multidrop serial communication
• RS485 with driver control signal
• ISO7816, T0 or T1 protocols for interfacing with smart cards
– NACK handling, error counter with repetition and iteration limit
• InfraRed IrDA Modulation and Demodulation
• Test modes
– Remote loopback, local loopback, automatic echo
520
AT91SAM9G45
31.7.1 Baud Rate Generator

The Baud Rate Generator provides the bit period clock named the Baud Rate Clock to both the
receiver and the transmitter.
The Baud Rate Generator clock source can be selected by setting the USCLKS field in the Mode
Register (US_MR) between:
• the Master Clock MCK
• a division of the Master Clock, the divider being product dependent, but generally set to 8
• the external clock, available on the SCK pin
The Baud Rate Generator is based upon a 16-bit divider, which is programmed with the CD field
of the Baud Rate Generator Register (US_BRGR). If CD is programmed at 0, the Baud Rate
Generator does not generate any clock. If CD is programmed at 1, the divider is bypassed and
becomes inactive.
If the external SCK clock is selected, the duration of the low and high levels of the signal pro-
vided on the SCK pin must be longer than a Master Clock (MCK) period. The frequency of the
signal provided on SCK must be at least 4.5 times lower than MCK.
Figure 31-3. Baud Rate Generator
USCLKS CD
MCK CD
0 SCK
MCK/DIV
1
Reserved 16-bit Counter
SCK 2 FIDI
>1 SYNC
3 OVER
1 0
0 0 Sampling 0
Divider
Baud Rate
1 Clock
1
SYNC
Sampling
USCLKS = 3 Clock
31.7.1.1 Baud Rate in Asynchronous Mode

If the USART is programmed to operate in asynchronous mode, the selected clock is first
divided by CD, which is field programmed in the Baud Rate Generator Register (US_BRGR).
The resulting clock is provided to the receiver as a sampling clock and then divided by 16 or 8,
depending on the programming of the OVER bit in US_MR.
If OVER is set to 1, the receiver sampling is 8 times higher than the baud rate clock. If OVER is
cleared, the sampling is performed at 16 times the baud rate clock.
The following formula performs the calculation of the Baud Rate.
SelectedClock
Baudrate = --------------------------------------------
( 8 ( 2 – Over )CD )
This gives a maximum baud rate of MCK divided by 8, assuming that MCK is the highest possi-
ble clock and that OVER is programmed at 1.
521
AT91SAM9G45
31.7.1.2 Baud Rate Calculation Example

Table 31-4 shows calculations of CD to obtain a baud rate at 38400 bauds for different source
clock frequencies. This table also shows the actual resulting baud rate and the error.
Table 31-4. Baud Rate Example (OVER = 0)

Expected Baud
Source Clock Rate Calculation Result CD Actual Baud Rate Error
MHz Bit/s Bit/s
3 686 400 38 400 6.00 6 38 400.00 0.00%
4 915 200 38 400 8.00 8 38 400.00 0.00%
5 000 000 38 400 8.14 8 39 062.50 1.70%
7 372 800 38 400 12.00 12 38 400.00 0.00%
8 000 000 38 400 13.02 13 38 461.54 0.16%
12 000 000 38 400 19.53 20 37 500.00 2.40%
12 288 000 38 400 20.00 20 38 400.00 0.00%
14 318 180 38 400 23.30 23 38 908.10 1.31%
14 745 600 38 400 24.00 24 38 400.00 0.00%
18 432 000 38 400 30.00 30 38 400.00 0.00%
24 000 000 38 400 39.06 39 38 461.54 0.16%
24 576 000 38 400 40.00 40 38 400.00 0.00%
25 000 000 38 400 40.69 40 38 109.76 0.76%
32 000 000 38 400 52.08 52 38 461.54 0.16%
32 768 000 38 400 53.33 53 38 641.51 0.63%
33 000 000 38 400 53.71 54 38 194.44 0.54%
40 000 000 38 400 65.10 65 38 461.54 0.16%
50 000 000 38 400 81.38 81 38 580.25 0.47%
The baud rate is calculated with the following formula:

BaudRate = MCK ⁄ CD × 16
The baud rate error is calculated with the following formula. It is not recommended to work with
an error higher than 5%.
ExpectedBaudRate
Error = 1 – ⎛⎝ ---------------------------------------------------⎞⎠
ActualBaudRate
31.7.1.3 Fractional Baud Rate in Asynchronous Mode

The Baud Rate generator previously defined is subject to the following limitation: the output fre-
quency changes by only integer multiples of the reference frequency. An approach to this
problem is to integrate a fractional N clock generator that has a high resolution. The generator
architecture is modified to obtain Baud Rate changes by a fraction of the reference source clock.
This fractional part is programmed with the FP field in the Baud Rate Generator Register
(US_BRGR). If FP is not 0, the fractional part is activated. The resolution is one eighth of the
522
AT91SAM9G45
clock divider. This feature is only available when using USART normal mode. The fractional
Baud Rate is calculated using the following formula:
SelectedClock
Baudrate = ----------------------------------------------------------------
-
⎛ 8 ( 2 – Over ) ⎛ CD + FP ------- ⎞⎞
⎝ ⎝ 8 ⎠⎠
The modified architecture is presented below:
Figure 31-4. Fractional Baud Rate Generator

FP
USCLKS Modulus
CD
Control
FP
MCK CD
0 SCK
MCK/DIV
1
Reserved 16-bit Counter
SCK 2 glitch-free FIDI
logic >1 SYNC
3 OVER
1 0
0 0 Sampling 0
Divider
Baud Rate
1 Clock
1
SYNC Sampling
USCLKS = 3 Clock
31.7.1.4 Baud Rate in Synchronous Mode

If the USART is programmed to operate in synchronous mode, the selected clock is simply
divided by the field CD in US_BRGR.
BaudRate = SelectedClock
--------------------------------------
CD
In synchronous mode, if the external clock is selected (USCLKS = 3), the clock is provided
directly by the signal on the USART SCK pin. No division is active. The value written in
US_BRGR has no effect. The external clock frequency must be at least 4.5 times lower than the
system clock. In synchronous mode master (USCLKS = 0 or 1, CLK0 set to 1), the receive part
limits the SCK maximum frequency to MCK/4.5,
When either the external clock SCK or the internal clock divided (MCK/DIV) is selected, the
value programmed in CD must be even if the user has to ensure a 50:50 mark/space ratio on the
SCK pin. If the internal clock MCK is selected, the Baud Rate Generator ensures a 50:50 duty
cycle on the SCK pin, even if the value programmed in CD is odd.
523
AT91SAM9G45
31.7.1.5 Baud Rate in ISO 7816 Mode

The ISO7816 specification defines the bit rate with the following formula:
Di
B = ------ × f
Fi
where:
• B is the bit rate
• Di is the bit-rate adjustment factor
• Fi is the clock frequency division factor
• f is the ISO7816 clock frequency (Hz)
Di is a binary value encoded on a 4-bit field, named DI, as represented in Table 31-5.
Table 31-5. Binary and Decimal Values for Di

DI field 0001 0010 0011 0100 0101 0110 1000 1001
Di (decimal) 1 2 4 8 16 32 12 20
Fi is a binary value encoded on a 4-bit field, named FI, as represented in Table 31-6.
Table 31-6. Binary and Decimal Values for Fi

FI field 0000 0001 0010 0011 0100 0101 0110 1001 1010 1011 1100 1101
Fi (decimal 372 372 558 744 1116 1488 1860 512 768 1024 1536 2048
Table 31-7 shows the resulting Fi/Di Ratio, which is the ratio between the ISO7816 clock and the
baud rate clock.
Table 31-7. Possible Values for the Fi/Di Ratio

Fi/Di 372 558 774 1116 1488 1806 512 768 1024 1536 2048
1 372 558 744 1116 1488 1860 512 768 1024 1536 2048
2 186 279 372 558 744 930 256 384 512 768 1024
4 93 139.5 186 279 372 465 128 192 256 384 512
8 46.5 69.75 93 139.5 186 232.5 64 96 128 192 256
16 23.25 34.87 46.5 69.75 93 116.2 32 48 64 96 128
32 11.62 17.43 23.25 34.87 46.5 58.13 16 24 32 48 64
12 31 46.5 62 93 124 155 42.66 64 85.33 128 170.6
20 18.6 27.9 37.2 55.8 74.4 93 25.6 38.4 51.2 76.8 102.4
If the USART is configured in ISO7816 Mode, the clock selected by the USCLKS field in the
Mode Register (US_MR) is first divided by the value programmed in the field CD in the Baud
Rate Generator Register (US_BRGR). The resulting clock can be provided to the SCK pin to
feed the smart card clock inputs. This means that the CLKO bit can be set in US_MR.
This clock is then divided by the value programmed in the FI_DI_RATIO field in the FI_DI_Ratio
register (US_FIDI). This is performed by the Sampling Divider, which performs a division by up
to 2047 in ISO7816 Mode. The non-integer values of the Fi/Di Ratio are not supported and the
user must program the FI_DI_RATIO field to a value as close as possible to the expected value.
The FI_DI_RATIO field resets to the value 0x174 (372 in decimal) and is the most common
divider between the ISO7816 clock and the bit rate (Fi = 372, Di = 1).
524
AT91SAM9G45
Figure 31-5 shows the relation between the Elementary Time Unit, corresponding to a bit time,
and the ISO 7816 clock.
Figure 31-5. Elementary Time Unit (ETU)

FI_DI_RATIO
ISO7816 Clock Cycles
ISO7816 Clock
on SCK
ISO7816 I/O Line

on TXD
1 ETU
31.7.2 Receiver and Transmitter Control

After reset, the receiver is disabled. The user must enable the receiver by setting the RXEN bit
in the Control Register (US_CR). However, the receiver registers can be programmed before the
receiver clock is enabled.
After reset, the transmitter is disabled. The user must enable it by setting the TXEN bit in the
Control Register (US_CR). However, the transmitter registers can be programmed before being
enabled.
The Receiver and the Transmitter can be enabled together or independently.
At any time, the software can perform a reset on the receiver or the transmitter of the USART by
setting the corresponding bit, RSTRX and RSTTX respectively, in the Control Register
(US_CR). The software resets clear the status flag and reset internal state machines but the
user interface configuration registers hold the value configured prior to software reset. Regard-
less of what the receiver or the transmitter is performing, the communication is immediately
stopped.
The user can also independently disable the receiver or the transmitter by setting RXDIS and
TXDIS respectively in US_CR. If the receiver is disabled during a character reception, the
USART waits until the end of reception of the current character, then the reception is stopped. If
the transmitter is disabled while it is operating, the USART waits the end of transmission of both
the current character and character being stored in the Transmit Holding Register (US_THR). If
a timeguard is programmed, it is handled normally.
31.7.3 Synchronous and Asynchronous Modes
31.7.3.1 Transmitter Operations

The transmitter performs the same in both synchronous and asynchronous operating modes
(SYNC = 0 or SYNC = 1). One start bit, up to 9 data bits, one optional parity bit and up to two
stop bits are successively shifted out on the TXD pin at each falling edge of the programmed
serial clock.
The number of data bits is selected by the CHRL field and the MODE 9 bit in the Mode Register
(US_MR). Nine bits are selected by setting the MODE 9 bit regardless of the CHRL field. The
parity bit is set according to the PAR field in US_MR. The even, odd, space, marked or none
parity bit can be configured. The MSBF field in US_MR configures which data bit is sent first. If
written at 1, the most significant bit is sent first. At 0, the less significant bit is sent first. The num-
525
AT91SAM9G45
ber of stop bits is selected by the NBSTOP field in US_MR. The 1.5 stop bit is supported in
asynchronous mode only.
Figure 31-6. Character Transmit

Example: 8-bit, Parity Enabled One Stop
Baud Rate
Clock
TXD
Start D0 D1 D2 D3 D4 D5 D6 D7 Parity Stop

Bit Bit Bit
The characters are sent by writing in the Transmit Holding Register (US_THR). The transmitter
reports two status bits in the Channel Status Register (US_CSR): TXRDY (Transmitter Ready),
which indicates that US_THR is empty and TXEMPTY, which indicates that all the characters
written in US_THR have been processed. When the current character processing is completed,
the last character written in US_THR is transferred into the Shift Register of the transmitter and
US_THR becomes empty, thus TXRDY rises.
Both TXRDY and TXEMPTY bits are low when the transmitter is disabled. Writing a character in
US_THR while TXRDY is low has no effect and the written character is lost.
Figure 31-7. Transmitter Status

Baud Rate
Clock
TXD
Start Parity Stop Start Parity Stop
D0 D1 D2 D3 D4 D5 D6 D7 D0 D1 D2 D3 D4 D5 D6 D7
Bit Bit Bit Bit Bit Bit
Write
US_THR
TXRDY
TXEMPTY
31.7.3.2 Manchester Encoder

When the Manchester encoder is in use, characters transmitted through the USART are
encoded based on biphase Manchester II format. To enable this mode, set the MAN field in the
US_MR register to 1. Depending on polarity configuration, a logic level (zero or one), is transmit-
ted as a coded signal one-to-zero or zero-to-one. Thus, a transition always occurs at the
midpoint of each bit time. It consumes more bandwidth than the original NRZ signal (2x) but the
receiver has more error control since the expected input must show a change at the center of a
bit cell. An example of Manchester encoded sequence is: the byte 0xB1 or 10110001 encodes
to 10 01 10 10 01 01 01 10, assuming the default polarity of the encoder. Figure 31-8 illustrates
this coding scheme.
526
AT91SAM9G45
Figure 31-8. NRZ to Manchester Encoding

NRZ 1 0 1 1 0 0 0 1
encoded
data
Manchester
encoded Txd
data
The Manchester encoded character can also be encapsulated by adding both a configurable
preamble and a start frame delimiter pattern. Depending on the configuration, the preamble is a
training sequence, composed of a pre-defined pattern with a programmable length from 1 to 15
bit times. If the preamble length is set to 0, the preamble waveform is not generated prior to any
character. The preamble pattern is chosen among the following sequences: ALL_ONE,
ALL_ZERO, ONE_ZERO or ZERO_ONE, writing the field TX_PP in the US_MAN register, the
field TX_PL is used to configure the preamble length. Figure 31-9 illustrates and defines the
valid patterns. To improve flexibility, the encoding scheme can be configured using the
TX_MPOL field in the US_MAN register. If the TX_MPOL field is set to zero (default), a logic
zero is encoded with a zero-to-one transition and a logic one is encoded with a one-to-zero tran-
sition. If the TX_MPOL field is set to one, a logic one is encoded with a one-to-zero transition
and a logic zero is encoded with a zero-to-one transition.
Figure 31-9. Preamble Patterns, Default Polarity Assumed
Manchester
encoded SFD DATA
data Txd
8 bit width "ALL_ONE" Preamble
Manchester
encoded SFD DATA
data Txd
8 bit width "ALL_ZERO" Preamble
Manchester
encoded SFD
Txd DATA
data
8 bit width "ZERO_ONE" Preamble
Manchester
encoded SFD DATA
data Txd
8 bit width "ONE_ZERO" Preamble
A start frame delimiter is to be configured using the ONEBIT field in the US_MR register. It con-
sists of a user-defined pattern that indicates the beginning of a valid data. Figure 31-10
illustrates these patterns. If the start frame delimiter, also known as start bit, is one bit, (ONEBIT
at 1), a logic zero is Manchester encoded and indicates that a new character is being sent seri-
ally on the line. If the start frame delimiter is a synchronization pattern also referred to as sync
(ONEBIT at 0), a sequence of 3 bit times is sent serially on the line to indicate the start of a new
character. The sync waveform is in itself an invalid Manchester waveform as the transition
527
AT91SAM9G45
occurs at the middle of the second bit time. Two distinct sync patterns are used: the command
sync and the data sync. The command sync has a logic one level for one and a half bit times,
then a transition to logic zero for the second one and a half bit times. If the MODSYNC field in
the US_MR register is set to 1, the next character is a command. If it is set to 0, the next charac-
ter is a data. When direct memory access is used, the MODSYNC field can be immediately
updated with a modified character located in memory. To enable this mode, VAR_SYNC field in
US_MR register must be set to 1. In this case, the MODSYNC field in US_MR is bypassed and
the sync configuration is held in the TXSYNH in the US_THR register. The USART character for-
mat is modified and includes sync information.
Figure 31-10. Start Frame Delimiter

Preamble Length
is set to 0
SFD
Manchester
encoded DATA
data Txd
One bit start frame delimiter

SFD
Manchester
encoded DATA
Txd
data
Command Sync
start frame delimiter
SFD
Manchester
encoded DATA
data Txd
Data Sync
start frame delimiter
31.7.3.3 Drift Compensation

Drift compensation is available only in 16X oversampling mode. An hardware recovery system
allows a larger clock drift. To enable the hardware system, the bit in the USART_MAN register
must be set. If the RXD edge is one 16X clock cycle from the expected edge, this is considered
as normal jitter and no corrective actions is taken. If the RXD event is between 4 and 2 clock
cycles before the expected edge, then the current period is shortened by one clock cycle. If the
RXD event is between 2 and 3 clock cycles after the expected edge, then the current period is
lengthened by one clock cycle. These intervals are considered to be drift and so corrective
actions are automatically taken.
528
AT91SAM9G45
Figure 31-11. Bit Resynchronization
Oversampling
16x Clock
RXD
Sampling
point
Expected edge
Synchro. Tolerance Sync Synchro.
Synchro. Jump Jump Error
Error
31.7.3.4 Asynchronous Receiver

If the USART is programmed in asynchronous operating mode (SYNC = 0), the receiver over-
samples the RXD input line. The oversampling is either 16 or 8 times the Baud Rate clock,
depending on the OVER bit in the Mode Register (US_MR).
The receiver samples the RXD line. If the line is sampled during one half of a bit time at 0, a start
bit is detected and data, parity and stop bits are successively sampled on the bit rate clock.
If the oversampling is 16, (OVER at 0), a start is detected at the eighth sample at 0. Then, data
bits, parity bit and stop bit are sampled on each 16 sampling clock cycle. If the oversampling is 8
(OVER at 1), a start bit is detected at the fourth sample at 0. Then, data bits, parity bit and stop
bit are sampled on each 8 sampling clock cycle.
The number of data bits, first bit sent and parity mode are selected by the same fields and bits
as the transmitter, i.e. respectively CHRL, MODE9, MSBF and PAR. For the synchronization
mechanism only, the number of stop bits has no effect on the receiver as it considers only one
stop bit, regardless of the field NBSTOP, so that resynchronization between the receiver and the
transmitter can occur. Moreover, as soon as the stop bit is sampled, the receiver starts looking
for a new start bit so that resynchronization can also be accomplished when the transmitter is
operating with one stop bit.
Figure 31-12 and Figure 31-13 illustrate start detection and character reception when USART
operates in asynchronous mode.
529
AT91SAM9G45
Figure 31-12. Asynchronous Start Detection

Baud Rate
Clock
Sampling
Clock (x16)
RXD
Sampling
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
D0
Start Sampling
Detection
RXD
Sampling
1 2 3 4 5 6 7 0 1 2 3 4
Start
Rejection
Figure 31-13. Asynchronous Character Reception

Example: 8-bit, Parity Enabled
Baud Rate
Clock
RXD
Start 16 16 16 16 16 16 16 16 16 16
Detection samples samples samples samples samples samples samples samples samples samples
D0 D1 D2 D3 D4 D5 D6 D7 Parity Stop
Bit Bit
31.7.3.5 Manchester Decoder

When the MAN field in US_MR register is set to 1, the Manchester decoder is enabled. The
decoder performs both preamble and start frame delimiter detection. One input line is dedicated
to Manchester encoded input data.
An optional preamble sequence can be defined, its length is user-defined and totally indepen-
dent of the emitter side. Use RX_PL in US_MAN register to configure the length of the preamble
sequence. If the length is set to 0, no preamble is detected and the function is disabled. In addi-
tion, the polarity of the input stream is programmable with RX_MPOL field in US_MAN register.
Depending on the desired application the preamble pattern matching is to be defined via the
RX_PP field in US_MAN. See Figure 31-9 for available preamble patterns.
Unlike preamble, the start frame delimiter is shared between Manchester Encoder and Decoder.
So, if ONEBIT field is set to 1, only a zero encoded Manchester can be detected as a valid start
frame delimiter. If ONEBIT is set to 0, only a sync pattern is detected as a valid start frame
delimiter. Decoder operates by detecting transition on incoming stream. If RXD is sampled dur-
ing one quarter of a bit time at zero, a start bit is detected. See Figure 31-14. The sample pulse
rejection mechanism applies.
530
AT91SAM9G45
Figure 31-14. Asynchronous Start Bit Detection

Sampling
Clock
(16 x)
Manchester
encoded
data Txd
Start
Detection
1 2 3 4
The receiver is activated and starts Preamble and Frame Delimiter detection, sampling the data
at one quarter and then three quarters. If a valid preamble pattern or start frame delimiter is
detected, the receiver continues decoding with the same synchronization. If the stream does not
match a valid pattern or a valid start frame delimiter, the receiver re-synchronizes on the next
valid edge.The minimum time threshold to estimate the bit value is three quarters of a bit time.
If a valid preamble (if used) followed with a valid start frame delimiter is detected, the incoming
stream is decoded into NRZ data and passed to USART for processing. Figure 31-15 illustrates
Manchester pattern mismatch. When incoming data stream is passed to the USART, the
receiver is also able to detect Manchester code violation. A code violation is a lack of transition
in the middle of a bit cell. In this case, MANE flag in US_CSR register is raised. It is cleared by
writing the Control Register (US_CR) with the RSTSTA bit at 1. See Figure 31-16 for an exam-
ple of Manchester error detection during data phase.
Figure 31-15. Preamble Pattern Mismatch

Preamble Mismatch Preamble Mismatch
Manchester coding error invalid pattern
Manchester
encoded SFD DATA
data Txd
Preamble Length is set to 8
Figure 31-16. Manchester Error Flag

Preamble Length
is set to 4
Elementary character bit time
SFD
Manchester
encoded
Txd
data
Entering USART character area
sampling points
Preamble subpacket Manchester

and Start Frame Delimiter Coding Error
were successfully detected
decoded
When the start frame delimiter is a sync pattern (ONEBIT field at 0), both command and data
delimiter are supported. If a valid sync is detected, the received character is written as RXCHR
531
AT91SAM9G45
field in the US_RHR register and the RXSYNH is updated. RXCHR is set to 1 when the received
character is a command, and it is set to 0 if the received character is a data. This mechanism
alleviates and simplifies the direct memory access as the character contains its own sync field in
the same register.
As the decoder is setup to be used in unipolar mode, the first bit of the frame has to be a zero-to-
one transition.
31.7.3.6 Radio Interface: Manchester Encoded USART Application

This section describes low data rate RF transmission systems and their integration with a Man-
chester encoded USART. These systems are based on transmitter and receiver ICs that support
ASK and FSK modulation schemes.
The goal is to perform full duplex radio transmission of characters using two different frequency
carriers. See the configuration in Figure 31-17.
Figure 31-17. Manchester Encoded Characters RF Transmission
Fup frequency Carrier
ASK/FSK
Upstream Receiver
Upstream
LNA Serial
Emitter VCO Configuration
RF filter Interface
Demod
control Manchester USART

Fdown frequency Carrier bi-dir decoder Receiver
line
ASK/FSK
downstream transmitter
Manchester USART
Downstream encoder Emitter
Receiver PA
RF filter
Mod
VCO
control
The USART module is configured as a Manchester encoder/decoder. Looking at the down-

stream communication channel, Manchester encoded characters are serially sent to the RF
emitter. This may also include a user defined preamble and a start frame delimiter. Mostly, pre-
amble is used in the RF receiver to distinguish between a valid data from a transmitter and
signals due to noise. The Manchester stream is then modulated. See Figure 31-18 for an exam-
ple of ASK modulation scheme. When a logic one is sent to the ASK modulator, the power
amplifier, referred to as PA, is enabled and transmits an RF signal at downstream frequency.
When a logic zero is transmitted, the RF signal is turned off. If the FSK modulator is activated,
two different frequencies are used to transmit data. When a logic 1 is sent, the modulator out-
puts an RF signal at frequency F0 and switches to F1 if the data sent is a 0. See Figure 31-19.
From the receiver side, another carrier frequency is used. The RF receiver performs a bit check
operation examining demodulated data stream. If a valid pattern is detected, the receiver
532
AT91SAM9G45
switches to receiving mode. The demodulated stream is sent to the Manchester decoder.
Because of bit checking inside RF IC, the data transferred to the microcontroller is reduced by a
user-defined number of bits. The Manchester preamble length is to be defined in accordance
with the RF IC configuration.
Figure 31-18. ASK Modulator Output

1 0 0 1
NRZ stream
Manchester
encoded
data
default polarity Txd
unipolar output
ASK Modulator
Output
Uptstream Frequency F0
Figure 31-19. FSK Modulator Output

1 0 0 1
NRZ stream
Manchester
encoded
data Txd
default polarity
unipolar output
FSK Modulator
Output
Uptstream Frequencies
[F0, F0+offset]
31.7.3.7 Synchronous Receiver

In synchronous mode (SYNC = 1), the receiver samples the RXD signal on each rising edge of
the Baud Rate Clock. If a low level is detected, it is considered as a start. All data bits, the parity
bit and the stop bits are sampled and the receiver waits for the next start bit. Synchronous mode
operations provide a high speed transfer capability.
Configuration fields and bits are the same as in asynchronous mode.
Figure 31-20 illustrates a character reception in synchronous mode.
Figure 31-20. Synchronous Mode Character Reception

Example: 8-bit, Parity Enabled 1 Stop
Baud Rate
Clock
RXD
Sampling
Start D0 D1 D2 D3 D4 D5 D6 D7 Stop Bit
Parity Bit
533
AT91SAM9G45
31.7.3.8 Receiver Operations

When a character reception is completed, it is transferred to the Receive Holding Register
(US_RHR) and the RXRDY bit in the Status Register (US_CSR) rises. If a character is com-
pleted while the RXRDY is set, the OVRE (Overrun Error) bit is set. The last character is
transferred into US_RHR and overwrites the previous one. The OVRE bit is cleared by writing
the Control Register (US_CR) with the RSTSTA (Reset Status) bit at 1.
Figure 31-21. Receiver Status
Baud Rate
Clock
RXD
RSTSTA = 1
Write
US_CR
Read
US_RHR
RXRDY
OVRE
534
AT91SAM9G45
31.7.3.9 Parity
The USART supports five parity modes selected by programming the PAR field in the Mode
Register (US_MR). The PAR field also enables the Multidrop mode, see “Multidrop Mode” on
page 536. Even and odd parity bit generation and error detection are supported.
If even parity is selected, the parity generator of the transmitter drives the parity bit at 0 if a num-
ber of 1s in the character data bit is even, and at 1 if the number of 1s is odd. Accordingly, the
receiver parity checker counts the number of received 1s and reports a parity error if the sam-
pled parity bit does not correspond. If odd parity is selected, the parity generator of the
transmitter drives the parity bit at 1 if a number of 1s in the character data bit is even, and at 0 if
the number of 1s is odd. Accordingly, the receiver parity checker counts the number of received
1s and reports a parity error if the sampled parity bit does not correspond. If the mark parity is
used, the parity generator of the transmitter drives the parity bit at 1 for all characters. The
receiver parity checker reports an error if the parity bit is sampled at 0. If the space parity is
used, the parity generator of the transmitter drives the parity bit at 0 for all characters. The
receiver parity checker reports an error if the parity bit is sampled at 1. If parity is disabled, the
transmitter does not generate any parity bit and the receiver does not report any parity error.
Table 31-8 shows an example of the parity bit for the character 0x41 (character ASCII “A”)
depending on the configuration of the USART. Because there are two bits at 1, 1 bit is added
when a parity is odd, or 0 is added when a parity is even.
Table 31-8. Parity Bit Examples

Character Hexa Binary Parity Bit Parity Mode
A 0x41 0100 0001 1 Odd
A 0x41 0100 0001 0 Even
A 0x41 0100 0001 1 Mark
A 0x41 0100 0001 0 Space
A 0x41 0100 0001 None None
When the receiver detects a parity error, it sets the PARE (Parity Error) bit in the Channel Status
Register (US_CSR). The PARE bit can be cleared by writing the Control Register (US_CR) with
the RSTSTA bit at 1. Figure 31-22 illustrates the parity bit status setting and clearing.
535
AT91SAM9G45
Figure 31-22. Parity Error
Baud Rate
Clock
RXD
Start Bad Stop
D0 D1 D2 D3 D4 D5 D6 D7
Bit Parity Bit
Bit RSTSTA = 1
Write
US_CR
PARE
RXRDY
31.7.3.10 Multidrop Mode

If the PAR field in the Mode Register (US_MR) is programmed to the value 0x6 or 0x07, the
USART runs in Multidrop Mode. This mode differentiates the data characters and the address
characters. Data is transmitted with the parity bit at 0 and addresses are transmitted with the
parity bit at 1.
If the USART is configured in multidrop mode, the receiver sets the PARE parity error bit when
the parity bit is high and the transmitter is able to send a character with the parity bit high when
the Control Register is written with the SENDA bit at 1.
To handle parity error, the PARE bit is cleared when the Control Register is written with the bit
RSTSTA at 1.
The transmitter sends an address byte (parity bit set) when SENDA is written to US_CR. In this
case, the next byte written to US_THR is transmitted as an address. Any character written in
US_THR without having written the command SENDA is transmitted normally with the parity at
0.
31.7.3.11 Transmitter Timeguard

The timeguard feature enables the USART interface with slow remote devices.
The timeguard function enables the transmitter to insert an idle state on the TXD line between
two characters. This idle state actually acts as a long stop bit.
The duration of the idle state is programmed in the TG field of the Transmitter Timeguard Regis-
ter (US_TTGR). When this field is programmed at zero no timeguard is generated. Otherwise,
the transmitter holds a high level on TXD after each transmitted byte during the number of bit
periods programmed in TG in addition to the number of stop bits.
As illustrated in Figure 31-23, the behavior of TXRDY and TXEMPTY status bits is modified by
the programming of a timeguard. TXRDY rises only when the start bit of the next character is
sent, and thus remains at 0 during the timeguard transmission if a character has been written in
US_THR. TXEMPTY remains low until the timeguard transmission is completed as the time-
guard is part of the current character being transmitted.
536
AT91SAM9G45
Figure 31-23. Timeguard Operations

TG = 4 TG = 4
Baud Rate
Clock
TXD
Write
US_THR
TXRDY
TXEMPTY
Table 31-9 indicates the maximum length of a timeguard period that the transmitter can handle
in relation to the function of the Baud Rate.
Table 31-9. Maximum Timeguard Length Depending on Baud Rate

Baud Rate Bit time Timeguard
Bit/sec μs ms
1 200 833 212.50
9 600 104 26.56
14400 69.4 17.71
19200 52.1 13.28
28800 34.7 8.85
33400 29.9 7.63
56000 17.9 4.55
57600 17.4 4.43
115200 8.7 2.21
31.7.3.12 Receiver Time-out

The Receiver Time-out provides support in handling variable-length frames. This feature detects
an idle condition on the RXD line. When a time-out is detected, the bit TIMEOUT in the Channel
Status Register (US_CSR) rises and can generate an interrupt, thus indicating to the driver an
end of frame.
The time-out delay period (during which the receiver waits for a new character) is programmed
in the TO field of the Receiver Time-out Register (US_RTOR). If the TO field is programmed at
0, the Receiver Time-out is disabled and no time-out is detected. The TIMEOUT bit in US_CSR
remains at 0. Otherwise, the receiver loads a 16-bit counter with the value programmed in TO.
This counter is decremented at each bit period and reloaded each time a new character is
received. If the counter reaches 0, the TIMEOUT bit in the Status Register rises. Then, the user
can either:
• Stop the counter clock until a new character is received. This is performed by writing the
Control Register (US_CR) with the STTTO (Start Time-out) bit at 1. In this case, the idle state
537
AT91SAM9G45
on RXD before a new character is received will not provide a time-out. This prevents having
to handle an interrupt before a character is received and allows waiting for the next idle state
on RXD after a frame is received.
• Obtain an interrupt while no character is received. This is performed by writing US_CR with
the RETTO (Reload and Start Time-out) bit at 1. If RETTO is performed, the counter starts
counting down immediately from the value TO. This enables generation of a periodic interrupt
so that a user time-out can be handled, for example when no key is pressed on a keyboard.
If STTTO is performed, the counter clock is stopped until a first character is received. The idle
state on RXD before the start of the frame does not provide a time-out. This prevents having to
obtain a periodic interrupt and enables a wait of the end of frame when the idle state on RXD is
detected.
If RETTO is performed, the counter starts counting down immediately from the value TO. This
enables generation of a periodic interrupt so that a user time-out can be handled, for example
when no key is pressed on a keyboard.
Figure 31-24 shows the block diagram of the Receiver Time-out feature.
Figure 31-24. Receiver Time-out Block Diagram
Baud Rate TO
Clock
16-bit
Value
1 D Q Clock 16-bit Time-out
Counter
STTTO = TIMEOUT
Load 0
Clear
Character
Received
RETTO
Table 31-10 gives the maximum time-out period for some standard baud rates.
Table 31-10. Maximum Time-out Period

Baud Rate Bit Time Time-out
bit/sec μs ms
600 1 667 109 225
1 200 833 54 613
2 400 417 27 306
4 800 208 13 653
9 600 104 6 827
14400 69 4 551
19200 52 3 413
28800 35 2 276
33400 30 1 962
538
AT91SAM9G45
Table 31-10. Maximum Time-out Period (Continued)

Baud Rate Bit Time Time-out
56000 18 1 170
57600 17 1 138
200000 5 328
31.7.3.13 Framing Error

The receiver is capable of detecting framing errors. A framing error happens when the stop bit of
a received character is detected at level 0. This can occur if the receiver and the transmitter are
fully desynchronized.
A framing error is reported on the FRAME bit of the Channel Status Register (US_CSR). The
FRAME bit is asserted in the middle of the stop bit as soon as the framing error is detected. It is
cleared by writing the Control Register (US_CR) with the RSTSTA bit at 1.
Figure 31-25. Framing Error Status
Baud Rate
Clock
RXD
Start Parity Stop
D0 D1 D2 D3 D4 D5 D6 D7
Bit Bit Bit
RSTSTA = 1
Write
US_CR
FRAME
RXRDY
31.7.3.14 Transmit Break

The user can request the transmitter to generate a break condition on the TXD line. A break con-
dition drives the TXD line low during at least one complete character. It appears the same as a
0x00 character sent with the parity and the stop bits at 0. However, the transmitter holds the
TXD line at least during one character until the user requests the break condition to be removed.
A break is transmitted by writing the Control Register (US_CR) with the STTBRK bit at 1. This
can be performed at any time, either while the transmitter is empty (no character in either the
Shift Register or in US_THR) or when a character is being transmitted. If a break is requested
while a character is being shifted out, the character is first completed before the TXD line is held
low.
Once STTBRK command is requested further STTBRK commands are ignored until the end of
the break is completed.
The break condition is removed by writing US_CR with the STPBRK bit at 1. If the STPBRK is
requested before the end of the minimum break duration (one character, including start, data,
parity and stop bits), the transmitter ensures that the break condition completes.
539
AT91SAM9G45
The transmitter considers the break as though it is a character, i.e. the STTBRK and STPBRK
commands are taken into account only if the TXRDY bit in US_CSR is at 1 and the start of the
break condition clears the TXRDY and TXEMPTY bits as if a character is processed.
Writing US_CR with the both STTBRK and STPBRK bits at 1 can lead to an unpredictable
result. All STPBRK commands requested without a previous STTBRK command are ignored. A
byte written into the Transmit Holding Register while a break is pending, but not started, is
ignored.
After the break condition, the transmitter returns the TXD line to 1 for a minimum of 12 bit times.
Thus, the transmitter ensures that the remote receiver detects correctly the end of break and the
start of the next character. If the timeguard is programmed with a value higher than 12, the TXD
line is held high for the timeguard period.
After holding the TXD line for this period, the transmitter resumes normal operations.
Figure 31-26 illustrates the effect of both the Start Break (STTBRK) and Stop Break (STPBRK)
commands on the TXD line.
Figure 31-26. Break Transmission
Baud Rate
Clock
TXD
Start Parity Stop
Bit
D0 D1 D2 D3 D4 D5 D6 D7
Bit Bit
Break Transmission End of Break
STTBRK = 1 STPBRK = 1
Write
US_CR
TXRDY
TXEMPTY
31.7.3.15 Receive Break

The receiver detects a break condition when all data, parity and stop bits are low. This corre-
sponds to detecting a framing error with data at 0x00, but FRAME remains low.
When the low stop bit is detected, the receiver asserts the RXBRK bit in US_CSR. This bit may
be cleared by writing the Control Register (US_CR) with the bit RSTSTA at 1.
An end of receive break is detected by a high level for at least 2/16 of a bit period in asynchro-
nous operating mode or one sample at high level in synchronous operating mode. The end of
break detection also asserts the RXBRK bit.
31.7.3.16 Hardware Handshaking

The USART features a hardware handshaking out-of-band flow control. The RTS and CTS pins
are used to connect with the remote device, as shown in Figure 31-27.
540
AT91SAM9G45
Figure 31-27. Connection with a Remote Device for Hardware Handshaking
USART Remote
Device
TXD RXD
RXD TXD
CTS RTS
RTS CTS
Setting the USART to operate with hardware handshaking is performed by writing the
USART_MODE field in the Mode Register (US_MR) to the value 0x2.
The USART behavior when hardware handshaking is enabled is the same as the behavior in
standard synchronous or asynchronous mode, except that the receiver drives the RTS pin as
described below and the level on the CTS pin modifies the behavior of the transmitter as
described below. Using this mode requires using the PDC channel for reception. The transmitter
can handle hardware handshaking in any case.
Figure 31-28 shows how the receiver operates if hardware handshaking is enabled. The RTS
pin is driven high if the receiver is disabled and if the status RXBUFF (Receive Buffer Full) com-
ing from the PDC channel is high. Normally, the remote device does not start transmitting while
its CTS pin (driven by RTS) is high. As soon as the Receiver is enabled, the RTS falls, indicating
to the remote device that it can start transmitting. Defining a new buffer to the PDC clears the
status bit RXBUFF and, as a result, asserts the pin RTS low.
Figure 31-28. Receiver Behavior when Operating with Hardware Handshaking
RXD
RXEN = 1 RXDIS = 1
Write
US_CR
RTS
RXBUFF
Figure 31-29 shows how the transmitter operates if hardware handshaking is enabled. The CTS
pin disables the transmitter. If a character is being processing, the transmitter is disabled only
after the completion of the current character and transmission of the next character happens as
soon as the pin CTS falls.
Figure 31-29. Transmitter Behavior when Operating with Hardware Handshaking
CTS
TXD
541
AT91SAM9G45
31.7.4 ISO7816 Mode

The USART features an ISO7816-compatible operating mode. This mode permits interfacing
with smart cards and Security Access Modules (SAM) communicating through an ISO7816 link.
Both T = 0 and T = 1 protocols defined by the ISO7816 specification are supported.
Setting the USART in ISO7816 mode is performed by writing the USART_MODE field in the
Mode Register (US_MR) to the value 0x4 for protocol T = 0 and to the value 0x5 for protocol T =
1.
31.7.4.1 ISO7816 Mode Overview

The ISO7816 is a half duplex communication on only one bidirectional line. The baud rate is
determined by a division of the clock provided to the remote device (see “Baud Rate Generator”
on page 521).
The USART connects to a smart card as shown in Figure 31-30. The TXD line becomes bidirec-
tional and the Baud Rate Generator feeds the ISO7816 clock on the SCK pin. As the TXD pin
becomes bidirectional, its output remains driven by the output of the transmitter but only when
the transmitter is active while its input is directed to the input of the receiver. The USART is con-
sidered as the master of the communication as it generates the clock.
Figure 31-30. Connection of a Smart Card to the USART
USART
CLK
SCK Smart
Card
I/O
TXD
When operating in ISO7816, either in T = 0 or T = 1 modes, the character format is fixed. The
configuration is 8 data bits, even parity and 1 or 2 stop bits, regardless of the values pro-
grammed in the CHRL, MODE9, PAR and CHMODE fields. MSBF can be used to transmit LSB
or MSB first. Parity Bit (PAR) can be used to transmit in normal or inverse mode. Refer to
“USART Mode Register” on page 553 and “PAR: Parity Type” on page 554.
The USART cannot operate concurrently in both receiver and transmitter modes as the commu-
nication is unidirectional at a time. It has to be configured according to the required mode by
enabling or disabling either the receiver or the transmitter as desired. Enabling both the receiver
and the transmitter at the same time in ISO7816 mode may lead to unpredictable results.
The ISO7816 specification defines an inverse transmission format. Data bits of the character
must be transmitted on the I/O line at their negative value. The USART does not support this for-
mat and the user has to perform an exclusive OR on the data before writing it in the Transmit
Holding Register (US_THR) or after reading it in the Receive Holding Register (US_RHR).
31.7.4.2 Protocol T = 0
In T = 0 protocol, a character is made up of one start bit, eight data bits, one parity bit and one
guard time, which lasts two bit times. The transmitter shifts out the bits and does not drive the
I/O line during the guard time.
If no parity error is detected, the I/O line remains at 1 during the guard time and the transmitter
can continue with the transmission of the next character, as shown in Figure 31-31.
542
AT91SAM9G45
If a parity error is detected by the receiver, it drives the I/O line at 0 during the guard time, as
shown in Figure 31-32. This error bit is also named NACK, for Non Acknowledge. In this case,
the character lasts 1 bit time more, as the guard time length is the same and is added to the
error bit time which lasts 1 bit time.
When the USART is the receiver and it detects an error, it does not load the erroneous character
in the Receive Holding Register (US_RHR). It appropriately sets the PARE bit in the Status Reg-
ister (US_SR) so that the software can handle the error.
Figure 31-31. T = 0 Protocol without Parity Error

Baud Rate
Clock
RXD
Start D0 D1 D2 D3 D4 D5 D6 D7 Parity Guard Guard Next

Bit Bit Time 1 Time 2 Start
Bit
Figure 31-32. T = 0 Protocol with Parity Error

Baud Rate
Clock
I/O Error
Start D0 D1 D2 D3 D4 D5 D6 D7 Parity Guard Guard Start D0 D1

Bit Bit Time 1 Time 2 Bit
Repetition
31.7.4.3 Receive Error Counter

The USART receiver also records the total number of errors. This can be read in the Number of
Error (US_NER) register. The NB_ERRORS field can record up to 255 errors. Reading US_NER
automatically clears the NB_ERRORS field.
31.7.4.4 Receive NACK Inhibit

The USART can also be configured to inhibit an error. This can be achieved by setting the
INACK bit in the Mode Register (US_MR). If INACK is at 1, no error signal is driven on the I/O
line even if a parity bit is detected, but the INACK bit is set in the Status Register (US_SR). The
INACK bit can be cleared by writing the Control Register (US_CR) with the RSTNACK bit at 1.
Moreover, if INACK is set, the erroneous received character is stored in the Receive Holding
Register, as if no error occurred. However, the RXRDY bit does not raise.
31.7.4.5 Transmit Character Repetition

When the USART is transmitting a character and gets a NACK, it can automatically repeat the
character before moving on to the next one. Repetition is enabled by writing the
MAX_ITERATION field in the Mode Register (US_MR) at a value higher than 0. Each character
can be transmitted up to eight times; the first transmission plus seven repetitions.
If MAX_ITERATION does not equal zero, the USART repeats the character as many times as
the value loaded in MAX_ITERATION.
543
AT91SAM9G45
When the USART repetition number reaches MAX_ITERATION, the ITERATION bit is set in the
Channel Status Register (US_CSR). If the repetition of the character is acknowledged by the
receiver, the repetitions are stopped and the iteration counter is cleared.
The ITERATION bit in US_CSR can be cleared by writing the Control Register with the RSIT bit
at 1.
31.7.4.6 Disable Successive Receive NACK

The receiver can limit the number of successive NACKs sent back to the remote transmitter.
This is programmed by setting the bit DSNACK in the Mode Register (US_MR). The maximum
number of NACK transmitted is programmed in the MAX_ITERATION field. As soon as
MAX_ITERATION is reached, the character is considered as correct, an acknowledge is sent on
the line and the ITERATION bit in the Channel Status Register is set.
31.7.4.7 Protocol T = 1
When operating in ISO7816 protocol T = 1, the transmission is similar to an asynchronous for-
mat with only one stop bit. The parity is generated when transmitting and checked when
receiving. Parity error detection sets the PARE bit in the Channel Status Register (US_CSR).
31.7.5 IrDA Mode

The USART features an IrDA mode supplying half-duplex point-to-point wireless communica-
tion. It embeds the modulator and demodulator which allows a glueless connection to the
infrared transceivers, as shown in Figure 31-33. The modulator and demodulator are compliant
with the IrDA specification version 1.1 and support data transfer speeds ranging from 2.4 Kb/s to
115.2 Kb/s.
The USART IrDA mode is enabled by setting the USART_MODE field in the Mode Register
(US_MR) to the value 0x8. The IrDA Filter Register (US_IF) allows configuring the demodulator
filter. The USART transmitter and receiver operate in a normal asynchronous mode and all
parameters are accessible. Note that the modulator and the demodulator are activated.
Figure 31-33. Connection to IrDA Transceivers
USART IrDA
Transceivers
Receiver Demodulator RXD RX
TX
Transmitter Modulator TXD
The receiver and the transmitter must be enabled or disabled according to the direction of the
transmission to be managed.
To receive IrDA signals, the following needs to be done:
• Disable TX and Enable RX
544
AT91SAM9G45
• Configure the TXD pin as PIO and set it as an output at 0 (to avoid LED emission). Disable
the internal pull-up (better for power consumption).
• Receive data
31.7.5.1 IrDA Modulation

For baud rates up to and including 115.2 Kbits/sec, the RZI modulation scheme is used. “0” is
represented by a light pulse of 3/16th of a bit time. Some examples of signal pulse duration are
shown in Table 31-11.
Table 31-11. IrDA Pulse Duration

Baud Rate Pulse Duration (3/16)
2.4 Kb/s 78.13 μs
9.6 Kb/s 19.53 μs
19.2 Kb/s 9.77 μs
38.4 Kb/s 4.88 μs
57.6 Kb/s 3.26 μs
115.2 Kb/s 1.63 μs
Figure 31-34 shows an example of character transmission.
Figure 31-34. IrDA Modulation
Start Data Bits Stop

Bit Bit
Transmitter
0 1 0 1 0 0 1 1 0 1
Output
TXD
Bit Period 3
16 Bit Period
31.7.5.2 IrDA Baud Rate

Table 31-12 gives some examples of CD values, baud rate error and pulse duration. Note that
the requirement on the maximum acceptable error of ±1.87% must be met.
Table 31-12. IrDA Baud Rate Error

Peripheral Clock Baud Rate CD Baud Rate Error Pulse Time
3 686 400 115 200 2 0.00% 1.63
20 000 000 115 200 11 1.38% 1.63
32 768 000 115 200 18 1.25% 1.63
40 000 000 115 200 22 1.38% 1.63
3 686 400 57 600 4 0.00% 3.26
20 000 000 57 600 22 1.38% 3.26
32 768 000 57 600 36 1.25% 3.26
545
AT91SAM9G45
Table 31-12. IrDA Baud Rate Error (Continued)

Peripheral Clock Baud Rate CD Baud Rate Error Pulse Time
40 000 000 57 600 43 0.93% 3.26
3 686 400 38 400 6 0.00% 4.88
20 000 000 38 400 33 1.38% 4.88
32 768 000 38 400 53 0.63% 4.88
40 000 000 38 400 65 0.16% 4.88
3 686 400 19 200 12 0.00% 9.77
20 000 000 19 200 65 0.16% 9.77
32 768 000 19 200 107 0.31% 9.77
40 000 000 19 200 130 0.16% 9.77
3 686 400 9 600 24 0.00% 19.53
20 000 000 9 600 130 0.16% 19.53
32 768 000 9 600 213 0.16% 19.53
40 000 000 9 600 260 0.16% 19.53
3 686 400 2 400 96 0.00% 78.13
20 000 000 2 400 521 0.03% 78.13
32 768 000 2 400 853 0.04% 78.13
31.7.5.3 IrDA Demodulator

The demodulator is based on the IrDA Receive filter comprised of an 8-bit down counter which is
loaded with the value programmed in US_IF. When a falling edge is detected on the RXD pin,
the Filter Counter starts counting down at the Master Clock (MCK) speed. If a rising edge is
detected on the RXD pin, the counter stops and is reloaded with US_IF. If no rising edge is
detected when the counter reaches 0, the input of the receiver is driven low during one bit time.
Figure 31-35 illustrates the operations of the IrDA demodulator.
Figure 31-35. IrDA Demodulator Operations

MCK
RXD
Counter
Value 6 5 4 3 2 6 6 5 4 3 2 1 0
Pulse Pulse
Rejected Accepted
Receiver
Input
As the IrDA mode uses the same logic as the ISO7816, note that the FI_DI_RATIO field in
US_FIDI must be set to a value higher than 0 in order to assure IrDA communications operate
correctly.
546
AT91SAM9G45
31.7.6 RS485 Mode

The USART features the RS485 mode to enable line driver control. While operating in RS485
mode, the USART behaves as though in asynchronous or synchronous mode and configuration
of all the parameters is possible. The difference is that the RTS pin is driven high when the
transmitter is operating. The behavior of the RTS pin is controlled by the TXEMPTY bit. A typical
connection of the USART to a RS485 bus is shown in Figure 31-36.
Figure 31-36. Typical Connection to a RS485 Bus
USART
RXD
Differential
TXD Bus
RTS
The USART is set in RS485 mode by programming the USART_MODE field in the Mode Regis-
ter (US_MR) to the value 0x1.
The RTS pin is at a level inverse to the TXEMPTY bit. Significantly, the RTS pin remains high
when a timeguard is programmed so that the line can remain driven after the last character com-
pletion. Figure 31-37 gives an example of the RTS waveform during a character transmission
when the timeguard is enabled.
Figure 31-37. Example of RTS Drive with Timeguard

TG = 4
Baud Rate
Clock
TXD
Start Parity Stop
D0 D1 D2 D3 D4 D5 D6 D7
Bit Bit Bit
Write
US_THR
TXRDY
TXEMPTY
RTS
547
AT91SAM9G45
31.7.7 Test Modes

The USART can be programmed to operate in three different test modes. The internal loopback
capability allows on-board diagnostics. In the loopback mode the USART interface pins are dis-
connected or not and reconfigured for loopback internally or externally.
31.7.7.1 Normal Mode

Normal mode connects the RXD pin on the receiver input and the transmitter output on the TXD
pin.
Figure 31-38. Normal Mode Configuration

RXD
Receiver
TXD
Transmitter
31.7.7.2 Automatic Echo Mode

Automatic echo mode allows bit-by-bit retransmission. When a bit is received on the RXD pin, it
is sent to the TXD pin, as shown in Figure 31-39. Programming the transmitter has no effect on
the TXD pin. The RXD pin is still connected to the receiver input, thus the receiver remains
active.
Figure 31-39. Automatic Echo Mode Configuration

RXD
Receiver
TXD
Transmitter
31.7.7.3 Local Loopback Mode

Local loopback mode connects the output of the transmitter directly to the input of the receiver,
as shown in Figure 31-40. The TXD and RXD pins are not used. The RXD pin has no effect on
the receiver and the TXD pin is continuously driven high, as in idle state.
Figure 31-40. Local Loopback Mode Configuration

RXD
Receiver
TXD
Transmitter 1
548
AT91SAM9G45
31.7.7.4 Remote Loopback Mode

Remote loopback mode directly connects the RXD pin to the TXD pin, as shown in Figure 31-41.
The transmitter and the receiver are disabled and have no effect. This mode allows bit-by-bit
retransmission.
Figure 31-41. Remote Loopback Mode Configuration

RXD
Receiver 1
TXD
Transmitter
549
AT91SAM9G45
31.8 Universal Synchronous Asynchronous Receiver Transmitter (USART) User Interface

0x0000 Control Register US_CR Write-only –
0x0004 Mode Register US_MR Read-write –
0x0008 Interrupt Enable Register US_IER Write-only –
0x000C Interrupt Disable Register US_IDR Write-only –
0x0010 Interrupt Mask Register US_IMR Read-only 0x0
0x0014 Channel Status Register US_CSR Read-only –
0x0018 Receiver Holding Register US_RHR Read-only 0x0
0x001C Transmitter Holding Register US_THR Write-only –
0x0020 Baud Rate Generator Register US_BRGR Read-write 0x0
0x0024 Receiver Time-out Register US_RTOR Read-write 0x0
0x0028 Transmitter Timeguard Register US_TTGR Read-write 0x0
0x2C - 0x3C Reserved – – –
0x0040 FI DI Ratio Register US_FIDI Read-write 0x174
0x0044 Number of Errors Register US_NER Read-only –
0x004C IrDA Filter Register US_IF Read-write 0x0
0x0050 Manchester Encoder Decoder Register US_MAN Read-write 0x30011004
0x5C - 0xFC Reserved – – –
0x100 - 0x128 Reserved for PDC Registers – – –
550
AT91SAM9G45
31.8.1 USART Control Register

Name: US_CR
Addresses: 0xFFF8C000 (0), 0xFFF90000 (1), 0xFFF94000 (2), 0xFFF98000 (3)
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – RTSDIS RTSEN – –
15 14 13 12 11 10 9 8
RETTO RSTNACK RSTIT SENDA STTTO STPBRK STTBRK RSTSTA
7 6 5 4 3 2 1 0
TXDIS TXEN RXDIS RXEN RSTTX RSTRX – –
• RSTRX: Reset Receiver

0: No effect.
1: Resets the receiver.
• RSTTX: Reset Transmitter

0: No effect.
1: Resets the transmitter.
• RXEN: Receiver Enable

0: No effect.
1: Enables the receiver, if RXDIS is 0.
• RXDIS: Receiver Disable

0: No effect.
1: Disables the receiver.
• TXEN: Transmitter Enable

0: No effect.
1: Enables the transmitter if TXDIS is 0.
• TXDIS: Transmitter Disable

0: No effect.
1: Disables the transmitter.
551
AT91SAM9G45
• RSTSTA: Reset Status Bits

0: No effect.
1: Resets the status bits PARE, FRAME, OVRE, MANERR and RXBRK in US_CSR.
• STTBRK: Start Break

0: No effect.
1: Starts transmission of a break after the characters present in US_THR and the Transmit Shift Register have been trans-
mitted. No effect if a break is already being transmitted.
• STPBRK: Stop Break

0: No effect.
1: Stops transmission of the break after a minimum of one character length and transmits a high level during 12-bit periods.
No effect if no break is being transmitted.
• STTTO: Start Time-out

0: No effect.
1: Starts waiting for a character before clocking the time-out counter. Resets the status bit TIMEOUT in US_CSR.
• SENDA: Send Address

0: No effect.
1: In Multidrop Mode only, the next character written to the US_THR is sent with the address bit set.
• RSTIT: Reset Iterations

0: No effect.
1: Resets ITERATION in US_CSR. No effect if the ISO7816 is not enabled.
• RSTNACK: Reset Non Acknowledge

0: No effect
1: Resets NACK in US_CSR.
• RETTO: Rearm Time-out

0: No effect
1: Restart Time-out
• RTSEN: Request to Send Enable

0: No effect.
1: Drives the pin RTS to 0.
• RTSDIS: Request to Send Disable

0: No effect.
1: Drives the pin RTS to 1.
552
AT91SAM9G45
31.8.2 USART Mode Register

Name: US_MR
Access: Read-write
31 30 29 28 27 26 25 24
ONEBIT MODSYNC– MAN FILTER – MAX_ITERATION
23 22 21 20 19 18 17 16
VAR_SYNC DSNACK INACK OVER CLKO MODE9 MSBF
15 14 13 12 11 10 9 8
CHMODE NBSTOP PAR SYNC
7 6 5 4 3 2 1 0
CHRL USCLKS USART_MODE
• USART_MODE
USART_MODE Mode of the USART

0 0 0 0 Normal
0 0 0 1 RS485
0 0 1 0 Hardware Handshaking
0 1 0 0 IS07816 Protocol: T = 0
0 1 1 0 IS07816 Protocol: T = 1
1 0 0 0 IrDA
Others Reserved
• USCLKS: Clock Selection
USCLKS Selected Clock

0 0 MCK
0 1 MCK/DIV (DIV = 8)
1 0 Reserved
1 1 SCK
• CHRL: Character Length.
CHRL Character Length

0 0 5 bits
0 1 6 bits
1 0 7 bits
1 1 8 bits
553
AT91SAM9G45
• SYNC: Synchronous Mode Select

0: USART operates in Asynchronous Mode.
1: USART operates in Synchronous Mode.
• PAR: Parity Type
PAR Parity Type

0 0 0 Even parity
0 0 1 Odd parity
0 1 0 Parity forced to 0 (Space)
0 1 1 Parity forced to 1 (Mark)
1 0 x No parity
1 1 x Multidrop mode
• NBSTOP: Number of Stop Bits
NBSTOP Asynchronous (SYNC = 0) Synchronous (SYNC = 1)

0 0 1 stop bit 1 stop bit
0 1 1.5 stop bits Reserved
1 0 2 stop bits 2 stop bits
1 1 Reserved Reserved
• CHMODE: Channel Mode
CHMODE Mode Description

0 0 Normal Mode
0 1 Automatic Echo. Receiver input is connected to the TXD pin.
1 0 Local Loopback. Transmitter output is connected to the Receiver Input..
1 1 Remote Loopback. RXD pin is internally connected to the TXD pin.
• MSBF: Bit Order

0: Least Significant Bit is sent/received first.
1: Most Significant Bit is sent/received first.
• MODE9: 9-bit Character Length

0: CHRL defines character length.
1: 9-bit character length.
• CLKO: Clock Output Select

0: The USART does not drive the SCK pin.
1: The USART drives the SCK pin if USCLKS does not select the external clock SCK.
554
AT91SAM9G45
• OVER: Oversampling Mode

0: 16x Oversampling.
1: 8x Oversampling.
• INACK: Inhibit Non Acknowledge

0: The NACK is generated.
1: The NACK is not generated.
• DSNACK: Disable Successive NACK

0: NACK is sent on the ISO line as soon as a parity error occurs in the received character (unless INACK is set).
1: Successive parity errors are counted up to the value specified in the MAX_ITERATION field. These parity errors gener-
ate a NACK on the ISO line. As soon as this value is reached, no additional NACK is sent on the ISO line. The flag
ITERATION is asserted.
• VAR_SYNC: Variable Synchronization of Command/Data Sync Start Frame Delimiter

0: User defined configuration of command or data sync field depending on SYNC value.
1: The sync field is updated when a character is written into US_THR register.
• MAX_ITERATION
Defines the maximum number of iterations in mode ISO7816, protocol T= 0.
• FILTER: Infrared Receive Line Filter

0: The USART does not filter the receive line.
1: The USART filters the receive line using a three-sample filter (1/16-bit clock) (2 over 3 majority).
• MAN: Manchester Encoder/Decoder Enable

0: Manchester Encoder/Decoder are disabled.
1: Manchester Encoder/Decoder are enabled.
• MODSYNC: Manchester Synchronization Mode

0:The Manchester Start bit is a 0 to 1 transition
1: The Manchester Start bit is a 1 to 0 transition.
• ONEBIT: Start Frame Delimiter Selector

0: Start Frame delimiter is COMMAND or DATA SYNC.
1: Start Frame delimiter is One Bit.
555
AT91SAM9G45
31.8.3 USART Interrupt Enable Register

Name: US_IER
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – MANE
23 22 21 20 19 18 17 16
– – – – CTSIC – – –
15 14 13 12 11 10 9 8
– – NACK RXBUFF TXBUFE ITER TXEMPTY TIMEOUT
7 6 5 4 3 2 1 0
PARE FRAME OVRE ENDTX ENDRX RXBRK TXRDY RXRDY
• RXRDY: RXRDY Interrupt Enable
• TXRDY: TXRDY Interrupt Enable
• RXBRK: Receiver Break Interrupt Enable
• ENDRX: End of Receive Transfer Interrupt Enable
• ENDTX: End of Transmit Interrupt Enable
• OVRE: Overrun Error Interrupt Enable
• FRAME: Framing Error Interrupt Enable
• PARE: Parity Error Interrupt Enable
• TIMEOUT: Time-out Interrupt Enable
• TXEMPTY: TXEMPTY Interrupt Enable
• ITER: Iteration Interrupt Enable
• TXBUFE: Buffer Empty Interrupt Enable
• RXBUFF: Buffer Full Interrupt Enable
• NACK: Non Acknowledge Interrupt Enable
• CTSIC: Clear to Send Input Change Interrupt Enable
• MANE: Manchester Error Interrupt Enable
556
AT91SAM9G45
31.8.4 USART Interrupt Disable Register

Name: US_IDR
Addresses: 0xFFF8C00C (0), 0xFFF9000C (1), 0xFFF9400C (2), 0xFFF9800C (3)
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – MANE
23 22 21 20 19 18 17 16
– – – – CTSIC – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• RXRDY: RXRDY Interrupt Disable
• TXRDY: TXRDY Interrupt Disable
• RXBRK: Receiver Break Interrupt Disable
• ENDRX: End of Receive Transfer Interrupt Disable
• ENDTX: End of Transmit Interrupt Disable
• OVRE: Overrun Error Interrupt Disable
• FRAME: Framing Error Interrupt Disable
• PARE: Parity Error Interrupt Disable
• TIMEOUT: Time-out Interrupt Disable
• TXEMPTY: TXEMPTY Interrupt Disable
• TXBUFE: Buffer Empty Interrupt Disable
• RXBUFF: Buffer Full Interrupt Disable
• NACK: Non Acknowledge Interrupt Disable
• CTSIC: Clear to Send Input Change Interrupt Disable
• MANE: Manchester Error Interrupt Disable
557
AT91SAM9G45
31.8.5 USART Interrupt Mask Register

Name: US_IMR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – MANE
23 22 21 20 19 18 17 16
– – – – CTSIC – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• RXRDY: RXRDY Interrupt Mask
• TXRDY: TXRDY Interrupt Mask
• RXBRK: Receiver Break Interrupt Mask
• ENDRX: End of Receive Transfer Interrupt Mask
• ENDTX: End of Transmit Interrupt Mask
• OVRE: Overrun Error Interrupt Mask
• FRAME: Framing Error Interrupt Mask
• PARE: Parity Error Interrupt Mask
• TIMEOUT: Time-out Interrupt Mask
• TXEMPTY: TXEMPTY Interrupt Mask
• TXBUFE: Buffer Empty Interrupt Mask
• RXBUFF: Buffer Full Interrupt Mask
• NACK: Non Acknowledge Interrupt Mask
• CTSIC: Clear to Send Input Change Interrupt Mask
• MANE: Manchester Error Interrupt Mask
558
AT91SAM9G45
31.8.6 USART Channel Status Register

Name: US_CSR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – MANERR
23 22 21 20 19 18 17 16
CTS – – – CTSIC – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0

0: No complete character has been received since the last read of US_RHR or the receiver is disabled. If characters were
being received when the receiver was disabled, RXRDY changes to 1 when the receiver is enabled.
1: At least one complete character has been received and US_RHR has not yet been read.
• TXRDY: Transmitter Ready

0: A character is in the US_THR waiting to be transferred to the Transmit Shift Register, or an STTBRK command has been
requested, or the transmitter is disabled. As soon as the transmitter is enabled, TXRDY becomes 1.
1: There is no character in the US_THR.
• RXBRK: Break Received/End of Break

0: No Break received or End of Break detected since the last RSTSTA.
1: Break Received or End of Break detected since the last RSTSTA.
• ENDRX: End of Receiver Transfer

0: The End of Transfer signal from the Receive PDC channel is inactive.
1: The End of Transfer signal from the Receive PDC channel is active.
• ENDTX: End of Transmitter Transfer

0: The End of Transfer signal from the Transmit PDC channel is inactive.
1: The End of Transfer signal from the Transmit PDC channel is active.
• OVRE: Overrun Error

0: No overrun error has occurred since the last RSTSTA.
1: At least one overrun error has occurred since the last RSTSTA.
559
AT91SAM9G45
• FRAME: Framing Error

0: No stop bit has been detected low since the last RSTSTA.
1: At least one stop bit has been detected low since the last RSTSTA.
• PARE: Parity Error

0: No parity error has been detected since the last RSTSTA.
1: At least one parity error has been detected since the last RSTSTA.
• TIMEOUT: Receiver Time-out

0: There has not been a time-out since the last Start Time-out command (STTTO in US_CR) or the Time-out Register is 0.
1: There has been a time-out since the last Start Time-out command (STTTO in US_CR).
• TXEMPTY: Transmitter Empty

0: There are characters in either US_THR or the Transmit Shift Register, or the transmitter is disabled.
1: There are no characters in US_THR, nor in the Transmit Shift Register.
• ITER: Max number of Repetitions Reached

0: Maximum number of repetitions has not been reached since the last RSTSTA.
1: Maximum number of repetitions has been reached since the last RSTSTA.
• TXBUFE: Transmission Buffer Empty

0: The signal Buffer Empty from the Transmit PDC channel is inactive.
1: The signal Buffer Empty from the Transmit PDC channel is active.
• RXBUFF: Reception Buffer Full

0: The signal Buffer Full from the Receive PDC channel is inactive.
1: The signal Buffer Full from the Receive PDC channel is active.
• NACKNon Acknowledge
0: No Non Acknowledge has not been detected since the last RSTNACK.
1: At least one Non Acknowledge has been detected since the last RSTNACK.
• CTSIC: Clear to Send Input Change Flag

0: No input change has been detected on the CTS pin since the last read of US_CSR.
1: At least one input change has been detected on the CTS pin since the last read of US_CSR.
• CTS: Image of CTS Input

0: CTS is at 0.
1: CTS is at 1.
• MANERR: Manchester Error

0: No Manchester error has been detected since the last RSTSTA.
1: At least one Manchester error has been detected since the last RSTSTA.
560
AT91SAM9G45
31.8.7 USART Receive Holding Register

Name: US_RHR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RXSYNH – – – – – – RXCHR
7 6 5 4 3 2 1 0
RXCHR
• RXCHR: Received Character

Last character received if RXRDY is set.
• RXSYNH: Received Sync

0: Last Character received is a Data.
1: Last Character received is a Command.
561
AT91SAM9G45
31.8.8 USART Transmit Holding Register

Name: US_THR
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
TXSYNH – – – – – – TXCHR
7 6 5 4 3 2 1 0
TXCHR
• TXCHR: Character to be Transmitted

Next character to be transmitted after the current character if TXRDY is not set.
• TXSYNH: Sync Field to be transmitted

0: The next character sent is encoded as a data. Start Frame Delimiter is DATA SYNC.
1: The next character sent is encoded as a command. Start Frame Delimiter is COMMAND SYNC.
562
AT91SAM9G45
31.8.9 USART Baud Rate Generator Register

Name: US_BRGR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – FP
15 14 13 12 11 10 9 8
CD
7 6 5 4 3 2 1 0
CD
• CD: Clock Divider
USART_MODE ≠ ISO7816
SYNC = 0 SYNC = 1
USART_MODE =
CD OVER = 0 OVER = 1 ISO7816
0 Baud Rate Clock Disabled
Baud Rate = Baud Rate = Baud Rate = Baud Rate = Selected
1 to 65535
Selected Clock/16/CD Selected Clock/8/CD Selected Clock /CD Clock/CD/FI_DI_RATIO
• FP: Fractional Part

0: Fractional divider is disabled.
1 - 7: Baudrate resolution, defined by FP x 1/8.
563
AT91SAM9G45
31.8.10 USART Receiver Time-out Register

Name: US_RTOR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
TO
7 6 5 4 3 2 1 0
TO
• TO: Time-out Value

0: The Receiver Time-out is disabled.
1 - 65535: The Receiver Time-out is enabled and the Time-out delay is TO x Bit Period.
564
AT91SAM9G45
31.8.11 USART Transmitter Timeguard Register

Name: US_TTGR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
TG
• TG: Timeguard Value

0: The Transmitter Timeguard is disabled.
1 - 255: The Transmitter timeguard is enabled and the timeguard delay is TG x Bit Period.
565
AT91SAM9G45
31.8.12 USART FI DI RATIO Register

Name: US_FIDI
Access: Read-write
Reset Value: 0x174
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – FI_DI_RATIO
7 6 5 4 3 2 1 0
FI_DI_RATIO
• FI_DI_RATIO: FI Over DI Ratio Value

0: If ISO7816 mode is selected, the Baud Rate Generator generates no signal.
1 - 2047: If ISO7816 mode is selected, the Baud Rate is the clock provided on SCK divided by FI_DI_RATIO.
566
AT91SAM9G45
31.8.13 USART Number of Errors Register

Name: US_NER
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
NB_ERRORS
• NB_ERRORS: Number of Errors

Total number of errors that occurred during an ISO7816 transfer. This register automatically clears when read.
567
AT91SAM9G45
31.8.14 USART IrDA FILTER Register

Name: US_IF
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
IRDA_FILTER
• IRDA_FILTER: IrDA Filter

Sets the filter of the IrDA demodulator.
568
AT91SAM9G45
31.8.15 USART Manchester Configuration Register

Name: US_MAN
Access: Read-write
31 30 29 28 27 26 25 24
– DRIFT 1 RX_MPOL – – RX_PP
23 22 21 20 19 18 17 16
– – – – RX_PL
15 14 13 12 11 10 9 8
– – – TX_MPOL – – TX_PP
7 6 5 4 3 2 1 0
– – – – TX_PL
• TX_PL: Transmitter Preamble Length

0: The Transmitter Preamble pattern generation is disabled
1 - 15: The Preamble Length is TX_PL x Bit Period
• TX_PP: Transmitter Preamble Pattern
TX_PP Preamble Pattern default polarity assumed (TX_MPOL field not set)
0 0 ALL_ONE
0 1 ALL_ZERO
1 0 ZERO_ONE
1 1 ONE_ZERO
• TX_MPOL: Transmitter Manchester Polarity

0: Logic Zero is coded as a zero-to-one transition, Logic One is coded as a one-to-zero transition.
1: Logic Zero is coded as a one-to-zero transition, Logic One is coded as a zero-to-one transition.
• RX_PL: Receiver Preamble Length

0: The receiver preamble pattern detection is disabled
1 - 15: The detected preamble length is RX_PL x Bit Period
• RX_PP: Receiver Preamble Pattern detected
RX_PP Preamble Pattern default polarity assumed (RX_MPOL field not set)
0 0 ALL_ONE
0 1 ALL_ZERO
1 0 ZERO_ONE
1 1 ONE_ZERO
569
AT91SAM9G45
• RX_MPOL: Receiver Manchester Polarity

0: Logic Zero is coded as a zero-to-one transition, Logic One is coded as a one-to-zero transition.
1: Logic Zero is coded as a one-to-zero transition, Logic One is coded as a zero-to-one transition.
• DRIFT: Drift compensation

0: The USART can not recover from an important clock drift
1: The USART can recover from clock drift. The 16X clock mode must be enabled.
570
AT91SAM9G45
32. Synchronous Serial Controller (SSC)
32.1 Description
The Atmel Synchronous Serial Controller (SSC) provides a synchronous communication link
with external devices. It supports many serial synchronous communication protocols generally
used in audio and telecom applications such as I2S, Short Frame Sync, Long Frame Sync, etc.
The SSC contains an independent receiver and transmitter and a common clock divider. The
receiver and the transmitter each interface with three signals: the TD/RD signal for data, the
TK/RK signal for the clock and the TF/RF signal for the Frame Sync. The transfers can be pro-
grammed to start automatically or on different events detected on the Frame Sync signal.
The SSC’s high-level of programmability and its two dedicated PDC channels of up to 32 bits
permit a continuous high bit rate data transfer without processor intervention.
The SSC’s high-level of programmability and its use of DMA permit a continuous high bit rate
data transfer without processor intervention.
Featuring connection to two PDC channels and connection to the DMA, the SSC permits inter-
facing with low processor overhead to the following:
• CODEC’s in master or slave mode
• DAC through dedicated serial interface, particularly I2S
• Magnetic card reader

• Provides serial synchronous communication links used in audio and telecom applications
(with CODECs in Master or Slave Modes, I2S, TDM Buses, Magnetic Card Reader,...)
• Contains an independent receiver and transmitter and a common clock divider
• Offers a configurable frame sync and data length
• Receiver and transmitter can be programmed to start automatically or on detection of
different event on the frame sync signal
• Receiver and transmitter include a data signal, a clock signal and a frame synchronization
signal
571
32.3 Block Diagram
System
Bus
APB Bridge
PDC
Peripheral
Bus
TF
TK
TD
MCK
PMC
SSC Interface PIO
RF
RK
Interrupt Control
RD
SSC Interrupt
572 AT91SAM9G45
AT91SAM9G45
System
Bus
APB Bridge
DMA
Peripheral
Bus
TF
TK
TD
MCK
PMC
SSC Interface PIO
RF
RK
Interrupt Control
RD
SSC Interrupt
Power Interrupt Test

OS or RTOS Driver
Management Management Management
SSC
Time Slot Frame

Serial AUDIO Codec Line Interface
Management Management
573
32.5 Pin Name List
RF Receiver Frame Synchro Input/Output
RK Receiver Clock Input/Output
RD Receiver Data Input
TF Transmitter Frame Synchro Input/Output
TK Transmitter Clock Input/Output
TD Transmitter Data Output
32.6.1 I/O Lines

Before using the SSC receiver, the PIO controller must be configured to dedicate the SSC
receiver I/O lines to the SSC peripheral mode.
Before using the SSC transmitter, the PIO controller must be configured to dedicate the SSC
transmitter I/O lines to the SSC peripheral mode.

SSC0 RD0 PD3 A
SSC0 RF0 PD5 A
SSC0 RK0 PD4 A
SSC0 TD0 PD2 A
SSC0 TF0 PD1 A
SSC0 TK0 PD0 A
SSC1 RD1 PD11 A
SSC1 RF1 PD15 A
SSC1 RK1 PD13 A
SSC1 TD1 PD10 A
SSC1 TF1 PD14 A
SSC1 TK1 PD12 A

The SSC is not continuously clocked. The SSC interface may be clocked through the Power
Management Controller (PMC), therefore the programmer must first configure the PMC to
enable the SSC clock.
32.6.3 Interrupt
The SSC interface has an interrupt line connected to the Advanced Interrupt Controller (AIC).
Handling interrupts requires programming the AICbefore configuring the SSC.
574 AT91SAM9G45
AT91SAM9G45
All SSC interrupts can be enabled/disabled configuring the SSC Interrupt mask register. Each

Instance ID
SSC0 16
SSC1 17
pending and unmasked SSC interrupt will assert the SSC interrupt line. The SSC interrupt ser-
vice routine can get the interrupt origin by reading the SSC interrupt status register.
575
This chapter contains the functional description of the following: SSC Functional Block, Clock
Management, Data format, Start, Transmitter, Receiver and Frame Sync.
The receiver and transmitter operate separately. However, they can work synchronously by pro-
gramming the receiver to use the transmit clock and/or to start a data transfer when transmission
starts. Alternatively, this can be done by programming the transmitter to use the receive clock
and/or to start a data transfer when reception starts. The transmitter and the receiver can be pro-
grammed to operate with the clock signals provided on either the TK or RK pins. This allows the
SSC to support many slave-mode data transfers. The maximum clock speed allowed on the TK
and RK pins is the master clock divided by 2.
Figure 32-4. SSC Functional Block Diagram
Transmitter
Clock Output
TK
Controller
TK Input
MCK Clock Transmit Clock TX clock Frame Sync TF
Divider Controller Controller
RX clock
TXEN
TX Start Data
RX Start Start TD
Selector Controller
TF Transmit Shift Register
APB Transmit Holding Transmit Sync

Register Holding Register
User
Interface
Receiver Clock Output

Controller RK
RK Input
Receive Clock RX Clock Frame Sync
Controller RF
Controller
TX Clock
RXEN
RX Start
TX Start Start Data
RF RD
Selector Receive Shift Register Controller
RC0R
Receive Holding Receive Sync

Interrupt Control Register Holding Register
AIC
576 AT91SAM9G45
AT91SAM9G45
32.7.1 Clock Management
Transmitter
Clock Output
TK
Controller
TK Input
MCK Clock Transmit Clock TX clock Frame Sync
Controller TF
Divider Controller
RX clock
TXEN
TX Start Data
RX Start Start TD
TF Selector Controller
Transmit Shift Register
APB Transmit Holding Transmit Sync

Register Holding Register
User
Interface
Receiver Clock Output

Controller RK
RK Input
Receive Clock RX Clock Frame Sync
Controller RF
Controller
TX Clock
RXEN
TX Start RX Start
Start Data
RF RD
Selector Controller
Receive Shift Register
RC0R
Receive Holding Receive Sync

Interrupt Control Register Holding Register
NVIC
The transmitter clock can be generated by:

• an external clock received on the TK I/O pad
• the receiver clock
• the internal clock divider
The receiver clock can be generated by:
• an external clock received on the RK I/O pad
• the transmitter clock
• the internal clock divider
Furthermore, the transmitter block can generate an external clock on the TK I/O pad, and the
receiver block can generate an external clock on the RK I/O pad.
This allows the SSC to support many Master and Slave Mode data transfers.
577
32.7.1.1 Clock Divider
Figure 32-5. Divided Clock Block Diagram

Clock Divider
SSC_CMR
MCK Divided Clock

/2 12-bit Counter
The Master Clock divider is determined by the 12-bit field DIV counter and comparator (so its
maximal value is 4095) in the Clock Mode Register SSC_CMR, allowing a Master Clock division
by up to 8190. The Divided Clock is provided to both the Receiver and Transmitter. When this
field is programmed to 0, the Clock Divider is not used and remains inactive.
When DIV is set to a value equal to or greater than 1, the Divided Clock has a frequency of Mas-
ter Clock divided by 2 times DIV. Each level of the Divided Clock has a duration of the Master
Clock multiplied by DIV. This ensures a 50% duty cycle for the Divided Clock regardless of
whether the DIV value is even or odd.
Figure 32-6. Divided Clock Generation

Master Clock
Divided Clock
DIV = 1
Divided Clock Frequency = MCK/2
Master Clock
Divided Clock
DIV = 3
Divided Clock Frequency = MCK/6
Table 32-4.
Maximum Minimum
MCK / 2 MCK / 8190
32.7.1.2 Transmitter Clock Management

The transmitter clock is generated from the receiver clock or the divider clock or an external
clock scanned on the TK I/O pad. The transmitter clock is selected by the CKS field in
SSC_TCMR (Transmit Clock Mode Register). Transmit Clock can be inverted independently by
the CKI bits in SSC_TCMR.
The transmitter can also drive the TK I/O pad continuously or be limited to the actual data trans-
fer. The clock output is configured by the SSC_TCMR register. The Transmit Clock Inversion
(CKI) bits have no effect on the clock outputs. Programming the TCMR register to select TK pin
578 AT91SAM9G45
AT91SAM9G45
(CKS field) and at the same time Continuous Transmit Clock (CKO field) might lead to unpredict-
able results.
Figure 32-7. Transmitter Clock Management

TK (pin)
MUX Tri_state Clock

Controller Output
Receiver
Clock
Divider
Clock
CKO Data Transfer
CKS INV Tri-state Transmitter

MUX Controller Clock
CKI CKG
32.7.1.3 Receiver Clock Management

The receiver clock is generated from the transmitter clock or the divider clock or an external
clock scanned on the RK I/O pad. The Receive Clock is selected by the CKS field in
SSC_RCMR (Receive Clock Mode Register). Receive Clocks can be inverted independently by
the CKI bits in SSC_RCMR.
The receiver can also drive the RK I/O pad continuously or be limited to the actual data transfer.
The clock output is configured by the SSC_RCMR register. The Receive Clock Inversion (CKI)
bits have no effect on the clock outputs. Programming the RCMR register to select RK pin (CKS
field) and at the same time Continuous Receive Clock (CKO field) can lead to unpredictable
results.
Figure 32-8. Receiver Clock Management

RK (pin)
MUX Tri-state Clock

Controller Output
Transmitter
Clock
Divider
Clock
CKO Data Transfer
CKS INV Tri-state

MUX Controller Receiver
Clock
CKI CKG
579
32.7.1.4 Serial Clock Ratio Considerations
The Transmitter and the Receiver can be programmed to operate with the clock signals provided
on either the TK or RK pins. This allows the SSC to support many slave-mode data transfers. In
this case, the maximum clock speed allowed on the RK pin is:
– Master Clock divided by 2 if Receiver Frame Synchro is input
– Master Clock divided by 3 if Receiver Frame Synchro is output
In addition, the maximum clock speed allowed on the TK pin is:
– Master Clock divided by 6 if Transmit Frame Synchro is input
– Master Clock divided by 2 if Transmit Frame Synchro is output
32.7.2 Transmitter Operations

A transmitted frame is triggered by a start event and can be followed by synchronization data
before data transmission.
The start event is configured by setting the Transmit Clock Mode Register (SSC_TCMR). See
“Start” on page 581.
The frame synchronization is configured setting the Transmit Frame Mode Register
(SSC_TFMR). See “Frame Sync” on page 583.
To transmit data, the transmitter uses a shift register clocked by the transmitter clock signal and
the start mode selected in the SSC_TCMR. Data is written by the application to the SSC_THR
register then transferred to the shift register according to the data format selected.
When both the SSC_THR and the transmit shift register are empty, the status flag TXEMPTY is
set in SSC_SR. When the Transmit Holding register is transferred in the Transmit shift register,
the status flag TXRDY is set in SSC_SR and additional data can be loaded in the holding
register.
Figure 32-9. Transmitter Block Diagram
SSC_CRTXEN
TXEN
SSC_SRTXEN
SSC_CRTXDIS
SSC_TCMR.STTDLY
SSC_TFMR.FSDEN
SSC_RCMR.START SSC_TCMR.START SSC_TFMR.DATNB
SSC_TFMR.DATDEF
RXEN SSC_TFMR.MSBF TX Controller
TXEN
TX Start RX Start Start TX Start
Start TD
RF Selector Selector
RC0R RF
SSC_TFMR.FSDEN
SSC_TCMR.STTDLY != 0 0 1 Transmitter Clock
SSC_TFMR.DATLEN SSC_THR SSC_TSHR SSC_TFMR.FSLEN
TX Controller counter reached STTDLY
580 AT91SAM9G45
AT91SAM9G45
32.7.3 Receiver Operations

A received frame is triggered by a start event and can be followed by synchronization data
before data transmission.
The start event is configured setting the Receive Clock Mode Register (SSC_RCMR). See
“Start” on page 581.
The frame synchronization is configured setting the Receive Frame Mode Register
(SSC_RFMR). See “Frame Sync” on page 583.
The receiver uses a shift register clocked by the receiver clock signal and the start mode
selected in the SSC_RCMR. The data is transferred from the shift register depending on the
data format selected.
When the receiver shift register is full, the SSC transfers this data in the holding register, the sta-
tus flag RXRDY is set in SSC_SR and the data can be read in the receiver holding register. If
another transfer occurs before read of the RHR register, the status flag OVERUN is set in
SSC_SR and the receiver shift register is transferred in the RHR register.
Figure 32-10. Receiver Block Diagram
SSC_CR.RXEN
SSC_SR.RXEN
SSC_CR.RXDIS
SSC_TCMR.START
SSC_RCMR.START SSC_RFMR.MSBF
TXEN SSC_RFMR.DATNB
RXEN
RX Start Start RX Start
Selector Start RX Controller
RF RF Selector
RC0R RD
SSC_RCMR.STTDLY != 0
load SSC_RSHR load SSC_RHR Receiver Clock
SSC_RFMR.FSLEN SSC_RFMR.DATLEN
RX Controller counter reached STTDLY
32.7.4 Start
The transmitter and receiver can both be programmed to start their operations when an event
occurs, respectively in the Transmit Start Selection (START) field of SSC_TCMR and in the
Receive Start Selection (START) field of SSC_RCMR.
Under the following conditions the start event is independently programmable:
• Continuous. In this case, the transmission starts as soon as a word is written in SSC_THR
and the reception starts as soon as the Receiver is enabled.
• Synchronously with the transmitter/receiver
• On detection of a falling/rising edge on TF/RF
• On detection of a low level/high level on TF/RF
• On detection of a level change or an edge on TF/RF
581
A start can be programmed in the same manner on either side of the Transmit/Receive Clock
Register (RCMR/TCMR). Thus, the start could be on TF (Transmit) or RF (Receive).
Moreover, the Receiver can start when data is detected in the bit stream with the Compare
Functions.
Detection on TF/RF input/output is done by the field FSOS of the Transmit/Receive Frame Mode
Register (TFMR/RFMR).
Figure 32-11. Transmit Start Mode

TK
TF
(Input)
TD
Start = Low Level on TF X BO B1
(Output)
STTDLY
Start = Falling Edge on TF TD BO B1

X
(Output) STTDLY
Start = High Level on TF TD X BO B1

(Output) STTDLY
Start = Rising Edge on TF TD X BO B1

(Output) STTDLY
TD
Start = Level Change on TF X BO B1 BO B1
(Output)
STTDLY
TD
Start = Any Edge on TF (Output) X BO B1 BO B1
STTDLY
Figure 32-12. Receive Pulse/Edge Start Modes

RK
RF
(Input)
RD
Start = Low Level on RF X BO B1
(Input)
STTDLY
Start = Falling Edge on RF RD BO B1

X
(Input) STTDLY
Start = High Level on RF RD X BO B1

(Input) STTDLY
Start = Rising Edge on RF RD X BO B1

(Input) STTDLY
RD
Start = Level Change on RF X BO B1 BO B1
(Input)
STTDLY
RD
Start = Any Edge on RF (Input) X BO B1 BO B1
STTDLY
582 AT91SAM9G45
AT91SAM9G45
32.7.5 Frame Sync

The Transmitter and Receiver Frame Sync pins, TF and RF, can be programmed to generate
different kinds of frame synchronization signals. The Frame Sync Output Selection (FSOS) field
in the Receive Frame Mode Register (SSC_RFMR) and in the Transmit Frame Mode Register
(SSC_TFMR) are used to select the required waveform.
• Programmable low or high levels during data transfer are supported.
• Programmable high levels before the start of data transfers or toggling are also supported.
If a pulse waveform is selected, the Frame Sync Length (FSLEN) field in SSC_RFMR and
SSC_TFMR programs the length of the pulse, from 1 bit time up to 256 bit time.
The periodicity of the Receive and Transmit Frame Sync pulse output can be programmed
through the Period Divider Selection (PERIOD) field in SSC_RCMR and SSC_TCMR.
32.7.5.1 Frame Sync Data

Frame Sync Data transmits or receives a specific tag during the Frame Sync signal.
During the Frame Sync signal, the Receiver can sample the RD line and store the data in the
Receive Sync Holding Register and the transmitter can transfer Transmit Sync Holding Register
in the Shifter Register. The data length to be sampled/shifted out during the Frame Sync signal
is programmed by the FSLEN field in SSC_RFMR/SSC_TFMR and has a maximum value of 16.
Concerning the Receive Frame Sync Data operation, if the Frame Sync Length is equal to or
lower than the delay between the start event and the actual data reception, the data sampling
operation is performed in the Receive Sync Holding Register through the Receive Shift Register.
The Transmit Frame Sync Operation is performed by the transmitter only if the bit Frame Sync
Data Enable (FSDEN) in SSC_TFMR is set. If the Frame Sync length is equal to or lower than
the delay between the start event and the actual data transmission, the normal transmission has
priority and the data contained in the Transmit Sync Holding Register is transferred in the Trans-
mit Register, then shifted out.
32.7.5.2 Frame Sync Edge Detection

The Frame Sync Edge detection is programmed by the FSEDGE field in
SSC_RFMR/SSC_TFMR. This sets the corresponding flags RXSYN/TXSYN in the SSC Status
Register (SSC_SR) on frame synchro edge detection (signals RF/TF).
32.7.6 Receive Compare Modes
Figure 32-13. Receive Compare Modes
RK
RD CMP0 CMP1 CMP2 CMP3 Ignored B0 B1 B2

(Input)
Start
FSLEN STDLY DATLEN

Up to 16 Bits
(4 in This Example)
583
32.7.6.1 Compare Functions
Length of the comparison patterns (Compare 0, Compare 1) and thus the number of bits they
are compared to is defined by FSLEN, but with a maximum value of 16 bits. Comparison is
always done by comparing the last bits received with the comparison pattern. Compare 0 can be
one start event of the Receiver. In this case, the receiver compares at each new sample the last
bits received at the Compare 0 pattern contained in the Compare 0 Register (SSC_RC0R).
When this start event is selected, the user can program the Receiver to start a new data transfer
either by writing a new Compare 0, or by receiving continuously until Compare 1 occurs. This
selection is done with the bit (STOP) in SSC_RCMR.
32.7.7 Data Format

The data framing format of both the transmitter and the receiver are programmable through the
Transmitter Frame Mode Register (SSC_TFMR) and the Receiver Frame Mode Register
(SSC_RFMR). In either case, the user can independently select:
• the event that starts the data transfer (START)
• the delay in number of bit periods between the start event and the first data bit (STTDLY)
• the length of the data (DATLEN)
• the number of data to be transferred for each start event (DATNB).
• the length of synchronization transferred for each start event (FSLEN)
• the bit sense: most or lowest significant bit first (MSBF)
Additionally, the transmitter can be used to transfer synchronization and select the level driven
on the TD pin while not in data transfer operation. This is done respectively by the Frame Sync
Data Enable (FSDEN) and by the Data Default Value (DATDEF) bits in SSC_TFMR.
584 AT91SAM9G45
AT91SAM9G45
Table 32-5. Data Frame Registers

Transmitter Receiver Field Length Comment
SSC_TFMR SSC_RFMR DATLEN Up to 32 Size of word
SSC_TFMR SSC_RFMR DATNB Up to 16 Number of words transmitted in frame
SSC_TFMR SSC_RFMR MSBF Most significant bit first
SSC_TFMR SSC_RFMR FSLEN Up to 16 Size of Synchro data register
SSC_TFMR DATDEF 0 or 1 Data default value ended
SSC_TFMR FSDEN Enable send SSC_TSHR
SSC_TCMR SSC_RCMR PERIOD Up to 512 Frame size
SSC_TCMR SSC_RCMR STTDLY Up to 255 Size of transmit start delay
Figure 32-14. Transmit and Receive Frame Format in Edge/Pulse Start Modes
Start Start
PERIOD
(1)
TF/RF
FSLEN
TD Sync Data Default Data Data Default Sync Data

(If FSDEN = 1) From SSC_TSHR FromDATDEF From SSC_THR From SSC_THR FromDATDEF
Default Data Data Default

TD
(If FSDEN = 0) From DATDEF From SSC_THR From SSC_THR From DATDEF
RD Sync Data Ignored Data Data Ignored Sync Data

To SSC_RSHR To SSC_RHR To SSC_RHR
STTDLY DATLEN DATLEN
DATNB
Note: 1. Example of input on falling edge of TF/RF.
Figure 32-15. Transmit Frame Format in Continuous Mode
Start
TD Data Data Default

From SSC_THR From SSC_THR
DATLEN DATLEN
Start: 1. TXEMPTY set to 1

2. Write into the SSC_THR
585
AT91SAM9G45
Note: 1. STTDLY is set to 0. In this example, SSC_THR is loaded twice. FSDEN value has no effect on
the transmission. SyncData cannot be output in continuous mode.
Figure 32-16. Receive Frame Format in Continuous Mode
Start = Enable Receiver
RD Data Data
To SSC_RHR To SSC_RHR
DATLEN DATLEN
Note: 1. STTDLY is set to 0.
32.7.8 Loop Mode

The receiver can be programmed to receive transmissions from the transmitter. This is done by
setting the Loop Mode (LOOP) bit in SSC_RFMR. In this case, RD is connected to TD, RF is
connected to TF and RK is connected to TK.
32.7.9 Interrupt
Most bits in SSC_SR have a corresponding bit in interrupt management registers.
The SSC can be programmed to generate an interrupt when it detects an event. The interrupt is
controlled by writing SSC_IER (Interrupt Enable Register) and SSC_IDR (Interrupt Disable Reg-
ister) These registers enable and disable, respectively, the corresponding interrupt by setting
and clearing the corresponding bit in SSC_IMR (Interrupt Mask Register), which controls the
generation of interrupts by asserting the SSC interrupt line connected to the AIC.
Figure 32-17. Interrupt Block Diagram
SSC_IMR
SSC_IER SSC_IDR
PDC Set Clear
TXBUFE
ENDTX
Transmitter
TXRDY
TXEMPTY
TXSYNC
Interrupt SSC Interrupt
RXBUFF Control
ENDRX
Receiver
RXRDY
OVRUN
RXSYNC
586
AT91SAM9G45
Figure 32-18. Interrupt Block Diagram
SSC_IMR
SSC_IER SSC_IDR
Set Clear
Transmitter
TXRDY
TXEMPTY
TXSYNC
Interrupt SSC Interrupt
Control
Receiver
RXRDY
OVRUN
RXSYNC
587
AT91SAM9G45
32.8 SSC Application Examples

The SSC can support several serial communication modes used in audio or high speed serial
links. Some standard applications are shown in the following figures. All serial link applications
supported by the SSC are not listed here.
Figure 32-19. Audio Application Block Diagram
Clock SCK
TK
Word Select WS
TF I2S
RECEIVER
Data SD
TD
SSC
RD Clock SCK
RF Word Select WS
RK
Data SD MSB LSB MSB
Left Channel Right Channel
Figure 32-20. Codec Application Block Diagram

Serial Data Clock (SCLK)
TK
Frame sync (FSYNC)
TF
CODEC
Serial Data Out
TD
SSC
Serial Data In
RD
RF
RK Frame sync (FSYNC) First Time Slot

Dstart Dend
Serial Data Out
Serial Data In
588
AT91SAM9G45
Figure 32-21. Time Slot Application Block Diagram

SCLK
TK
FSYNC
TF CODEC
First
Data Out
TD Time Slot
SSC
Data in
RD
RF
RK
CODEC
Second
Time Slot
Frame sync (FSYNC) First Time Slot Second Time Slot

Dstart Dend
Serial Data Out
Serial Data in
589
AT91SAM9G45
32.9 Synchronous Serial Controller (SSC) User Interface

0x0 Control Register SSC_CR Write-only –
0x4 Clock Mode Register SSC_CMR Read-write 0x0
0xC Reserved – – –
0x10 Receive Clock Mode Register SSC_RCMR Read-write 0x0
0x14 Receive Frame Mode Register SSC_RFMR Read-write 0x0
0x18 Transmit Clock Mode Register SSC_TCMR Read-write 0x0
0x1C Transmit Frame Mode Register SSC_TFMR Read-write 0x0
0x20 Receive Holding Register SSC_RHR Read-only 0x0
0x24 Transmit Holding Register SSC_THR Write-only –
0x30 Receive Sync. Holding Register SSC_RSHR Read-only 0x0
0x34 Transmit Sync. Holding Register SSC_TSHR Read-write 0x0
0x38 Receive Compare 0 Register SSC_RC0R Read-write 0x0
0x3C Receive Compare 1 Register SSC_RC1R Read-write 0x0
0x40 Status Register SSC_SR Read-only 0x000000CC
0x44 Interrupt Enable Register SSC_IER Write-only –
0x48 Interrupt Disable Register SSC_IDR Write-only –
0x4C Interrupt Mask Register SSC_IMR Read-only 0x0
0x50-0xFC Reserved – – –
0x100- 0x124 Reserved for Peripheral Data Controller (PDC) – – –
590
AT91SAM9G45
32.9.1 SSC Control Register

Name: SSC_CR:
Addresses: 0xFFF9C000 (0), 0xFFFA0000 (1)
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
SWRST – – – – – TXDIS TXEN
7 6 5 4 3 2 1 0
– – – – – – RXDIS RXEN
• RXEN: Receive Enable

0 = No effect.
1 = Enables Receive if RXDIS is not set.
• RXDIS: Receive Disable

0 = No effect.
1 = Disables Receive. If a character is currently being received, disables at end of current character reception.
• TXEN: Transmit Enable

0 = No effect.
1 = Enables Transmit if TXDIS is not set.
• TXDIS: Transmit Disable

0 = No effect.
1 = Disables Transmit. If a character is currently being transmitted, disables at end of current character transmission.

0 = No effect.
1 = Performs a software reset. Has priority on any other bit in SSC_CR.
591
AT91SAM9G45
32.9.2 SSC Clock Mode Register

Name: SSC_CMR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – DIV
7 6 5 4 3 2 1 0
DIV
• DIV: Clock Divider

0 = The Clock Divider is not active.
Any Other Value: The Divided Clock equals the Master Clock divided by 2 times DIV. The maximum bit rate is MCK/2. The
minimum bit rate is MCK/2 x 4095 = MCK/8190.
592
AT91SAM9G45
32.9.3 SSC Receive Clock Mode Register

Name: SSC_RCMR
Access: Read-write
31 30 29 28 27 26 25 24
PERIOD
23 22 21 20 19 18 17 16
STTDLY
15 14 13 12 11 10 9 8
– – – STOP START
7 6 5 4 3 2 1 0
CKG CKI CKO CKS
• CKS: Receive Clock Selection

CKS Selected Receive Clock
0x0 Divided Clock
0x1 TK Clock signal
0x2 RK pin
0x3 Reserved
• CKO: Receive Clock Output Mode Selection

CKO Receive Clock Output Mode RK pin
0x0 None Input-only
0x1 Continuous Receive Clock Output
0x2 Receive Clock only during data transfers Output
0x3-0x7 Reserved
• CKI: Receive Clock Inversion

0 = The data inputs (Data and Frame Sync signals) are sampled on Receive Clock falling edge. The Frame Sync signal
output is shifted out on Receive Clock rising edge.
1 = The data inputs (Data and Frame Sync signals) are sampled on Receive Clock rising edge. The Frame Sync signal out-
put is shifted out on Receive Clock falling edge.
CKI affects only the Receive Clock and not the output clock signal.
593
AT91SAM9G45
• CKG: Receive Clock Gating Selection

CKG Receive Clock Gating
0x0 None, continuous clock
0x1 Receive Clock enabled only if RF Low
0x2 Receive Clock enabled only if RF High
0x3 Reserved
• START: Receive Start Selection

START Receive Start
Continuous, as soon as the receiver is enabled, and immediately after the end of
0x0
transfer of the previous data.
0x1 Transmit start
0x2 Detection of a low level on RF signal
0x3 Detection of a high level on RF signal
0x4 Detection of a falling edge on RF signal
0x5 Detection of a rising edge on RF signal
0x6 Detection of any level change on RF signal
0x7 Detection of any edge on RF signal
0x8 Compare 0
0x9-0xF Reserved
• STOP: Receive Stop Selection

0 = After completion of a data transfer when starting with a Compare 0, the receiver stops the data transfer and waits for a
new compare 0.
1 = After starting a receive with a Compare 0, the receiver operates in a continuous mode until a Compare 1 is detected.
• STTDLY: Receive Start Delay

If STTDLY is not 0, a delay of STTDLY clock cycles is inserted between the start event and the actual start of reception.
When the Receiver is programmed to start synchronously with the Transmitter, the delay is also applied.
Note: It is very important that STTDLY be set carefully. If STTDLY must be set, it should be done in relation to TAG
(Receive Sync Data) reception.
• PERIOD: Receive Period Divider Selection

This field selects the divider to apply to the selected Receive Clock in order to generate a new Frame Sync Signal. If 0, no
PERIOD signal is generated. If not 0, a PERIOD signal is generated each 2 x (PERIOD+1) Receive Clock.
594
AT91SAM9G45
32.9.4 SSC Receive Frame Mode Register

Name: SSC_RFMR
Access: Read-write
31 30 29 28 27 26 25 24
FSLEN_EXT FSLEN_EXT FSLEN_EXT FSLEN_EXT – – – FSEDGE
23 22 21 20 19 18 17 16
– FSOS FSLEN
15 14 13 12 11 10 9 8
– – – – DATNB
7 6 5 4 3 2 1 0
MSBF – LOOP DATLEN
• DATLEN: Data Length

0 = Forbidden value (1-bit data length not supported).
Any other value: The bit stream contains DATLEN + 1 data bits. Moreover, it defines the transfer size performed by the
PDC assigned to the Receiver. If DATLEN is lower or equal to 7, data transfers are in bytes. If DATLEN is between 8 and
15 (included), half-words are transferred, and for any other value, 32-bit words are transferred.
• LOOP: Loop Mode

0 = Normal operating mode.
1 = RD is driven by TD, RF is driven by TF and TK drives RK.
• MSBF: Most Significant Bit First

0 = The lowest significant bit of the data register is sampled first in the bit stream.
1 = The most significant bit of the data register is sampled first in the bit stream.
• DATNB: Data Number per Frame

This field defines the number of data words to be received after each transfer start, which is equal to (DATNB + 1).
• FSLEN: Receive Frame Sync Length

This field defines the number of bits sampled and stored in the Receive Sync Data Register. When this mode is selected by
the START field in the Receive Clock Mode Register, it also determines the length of the sampled data to be compared to
the Compare 0 or Compare 1 register.
This field is used with FSLEN_EXT to determine the pulse length of the Receive Frame Sync signal.
Pulse length is equal to FSLEN + (FSLEN_EXT * 16) + 1 Receive Clock periods.
595
AT91SAM9G45
• FSOS: Receive Frame Sync Output Selection

FSOS Selected Receive Frame Sync Signal RF Pin
0x0 None Input-only
0x1 Negative Pulse Output
0x2 Positive Pulse Output
0x3 Driven Low during data transfer Output
0x4 Driven High during data transfer Output
0x5 Toggling at each start of data transfer Output
0x6-0x7 Reserved Undefined
• FSEDGE: Frame Sync Edge Detection

Determines which edge on Frame Sync will generate the interrupt RXSYN in the SSC Status Register.
FSEDGE Frame Sync Edge Detection

0x0 Positive Edge Detection
0x1 Negative Edge Detection
• FSLEN_EXT: FSLEN Field Extension

Extends FSLEN field. For details, refer to FSLEN bit description on page 595.
596
AT91SAM9G45
32.9.5 SSC Transmit Clock Mode Register

Name: SSC_TCMR
Access: Read-write
31 30 29 28 27 26 25 24
PERIOD
23 22 21 20 19 18 17 16
STTDLY
15 14 13 12 11 10 9 8
– – – – START
7 6 5 4 3 2 1 0
CKG CKI CKO CKS
• CKS: Transmit Clock Selection

CKS Selected Transmit Clock
0x0 Divided Clock
0x1 RK Clock signal
0x2 TK Pin
0x3 Reserved
• CKO: Transmit Clock Output Mode Selection

CKO Transmit Clock Output Mode TK pin
0x0 None Input-only
0x1 Continuous Transmit Clock Output
0x2 Transmit Clock only during data transfers Output
0x3-0x7 Reserved
• CKI: Transmit Clock Inversion

0 = The data outputs (Data and Frame Sync signals) are shifted out on Transmit Clock falling edge. The Frame sync signal
input is sampled on Transmit clock rising edge.
1 = The data outputs (Data and Frame Sync signals) are shifted out on Transmit Clock rising edge. The Frame sync signal
input is sampled on Transmit clock falling edge.
CKI affects only the Transmit Clock and not the output clock signal.
597
AT91SAM9G45
• CKG: Transmit Clock Gating Selection

CKG Transmit Clock Gating
0x0 None, continuous clock
0x1 Transmit Clock enabled only if TF Low
0x2 Transmit Clock enabled only if TF High
0x3 Reserved
• START: Transmit Start Selection

START Transmit Start
Continuous, as soon as a word is written in the SSC_THR Register (if Transmit is enabled), and
0x0
immediately after the end of transfer of the previous data.
0x1 Receive start
0x2 Detection of a low level on TF signal
0x3 Detection of a high level on TF signal
0x4 Detection of a falling edge on TF signal
0x5 Detection of a rising edge on TF signal
0x6 Detection of any level change on TF signal
0x7 Detection of any edge on TF signal
0x8 - 0xF Reserved
• STTDLY: Transmit Start Delay

If STTDLY is not 0, a delay of STTDLY clock cycles is inserted between the start event and the actual start of transmission
of data. When the Transmitter is programmed to start synchronously with the Receiver, the delay is also applied.
Note: STTDLY must be set carefully. If STTDLY is too short in respect to TAG (Transmit Sync Data) emission, data is emit-
ted instead of the end of TAG.
• PERIOD: Transmit Period Divider Selection

This field selects the divider to apply to the selected Transmit Clock to generate a new Frame Sync Signal. If 0, no period
signal is generated. If not 0, a period signal is generated at each 2 x (PERIOD+1) Transmit Clock.
598
AT91SAM9G45
32.9.6 SSC Transmit Frame Mode Register

Name: SSC_TFMR
Addresses: 0xFFF9C01C (0), 0xFFFA001C (1)
Access: Read-write
31 30 29 28 27 26 25 24
FSLEN_EXT FSLEN_EXT FSLEN_EXT FSLEN_EXT – – – FSEDGE
23 22 21 20 19 18 17 16
FSDEN FSOS FSLEN
15 14 13 12 11 10 9 8
– – – – DATNB
7 6 5 4 3 2 1 0
MSBF – DATDEF DATLEN
• DATLEN: Data Length

0 = Forbidden value (1-bit data length not supported).
Any other value: The bit stream contains DATLEN + 1 data bits. Moreover, it defines the transfer size performed by the
PDC assigned to the Transmit. If DATLEN is lower or equal to 7, data transfers are bytes, if DATLEN is between 8 and 15
(included), half-words are transferred, and for any other value, 32-bit words are transferred.
• DATDEF: Data Default Value

This bit defines the level driven on the TD pin while out of transmission. Note that if the pin is defined as multi-drive by the
PIO Controller, the pin is enabled only if the SCC TD output is 1.
• MSBF: Most Significant Bit First

0 = The lowest significant bit of the data register is shifted out first in the bit stream.
1 = The most significant bit of the data register is shifted out first in the bit stream.
• DATNB: Data Number per frame

This field defines the number of data words to be transferred after each transfer start, which is equal to (DATNB +1).
• FSLEN: Transmit Frame Syn Length

This field defines the length of the Transmit Frame Sync signal and the number of bits shifted out from the Transmit Sync
Data Register if FSDEN is 1.
This field is used with FSLEN_EXT to determine the pulse length of the Transmit Frame Sync signal.
Pulse length is equal to FSLEN + (FSLEN_EXT * 16) + 1 Transmit Clock period.
599
AT91SAM9G45
• FSOS: Transmit Frame Sync Output Selection

FSOS Selected Transmit Frame Sync Signal TF Pin
0x0 None Input-only
0x1 Negative Pulse Output
0x2 Positive Pulse Output
0x3 Driven Low during data transfer Output
0x4 Driven High during data transfer Output
0x5 Toggling at each start of data transfer Output
0x6-0x7 Reserved Undefined
• FSDEN: Frame Sync Data Enable

0 = The TD line is driven with the default value during the Transmit Frame Sync signal.
1 = SSC_TSHR value is shifted out during the transmission of the Transmit Frame Sync signal.
• FSEDGE: Frame Sync Edge Detection

Determines which edge on frame sync will generate the interrupt TXSYN (Status Register).
FSEDGE Frame Sync Edge Detection

0x0 Positive Edge Detection
0x1 Negative Edge Detection
• FSLEN_EXT: FSLEN Field Extension

Extends FSLEN field. For details, refer to FSLEN bit description on page 599.
600
AT91SAM9G45
32.9.7 SSC Receive Holding Register

Name: SSC_RHR
Access: Read-only
31 30 29 28 27 26 25 24
RDAT
23 22 21 20 19 18 17 16
RDAT
15 14 13 12 11 10 9 8
RDAT
7 6 5 4 3 2 1 0
RDAT
• RDAT: Receive Data

Right aligned regardless of the number of data bits defined by DATLEN in SSC_RFMR.
32.9.8 SSC Transmit Holding Register

Name: SSC_THR
Access: Write-only
31 30 29 28 27 26 25 24
TDAT
23 22 21 20 19 18 17 16
TDAT
15 14 13 12 11 10 9 8
TDAT
7 6 5 4 3 2 1 0
TDAT
• TDAT: Transmit Data

Right aligned regardless of the number of data bits defined by DATLEN in SSC_TFMR.
601
AT91SAM9G45
32.9.9 SSC Receive Synchronization Holding Register

Name: SSC_RSHR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RSDAT
7 6 5 4 3 2 1 0
RSDAT
• RSDAT: Receive Synchronization Data
32.9.10 SSC Transmit Synchronization Holding Register

Name: SSC_TSHR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
TSDAT
7 6 5 4 3 2 1 0
TSDAT
• TSDAT: Transmit Synchronization Data
602
AT91SAM9G45
32.9.11 SSC Receive Compare 0 Register

Name: SSC_RC0R
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
CP0
7 6 5 4 3 2 1 0
CP0
• CP0: Receive Compare Data 0
603
AT91SAM9G45
32.9.12 SSC Receive Compare 1 Register

Name: SSC_RC1R
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
CP1
7 6 5 4 3 2 1 0
CP1
• CP1: Receive Compare Data 1
604
AT91SAM9G45
32.9.13 SSC Status Register

Name: SSC_SR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – RXEN TXEN
15 14 13 12 11 10 9 8
– – – – RXSYN TXSYN CP1 CP0
7 6 5 4 3 2 1 0
RXBUFF ENDRX OVRUN RXRDY TXBUFE ENDTX TXEMPTY TXRDY
• TXRDY: Transmit Ready

0 = Data has been loaded in SSC_THR and is waiting to be loaded in the Transmit Shift Register (TSR).
1 = SSC_THR is empty.
• TXEMPTY: Transmit Empty

0 = Data remains in SSC_THR or is currently transmitted from TSR.
1 = Last data written in SSC_THR has been loaded in TSR and last data loaded in TSR has been transmitted.
• ENDTX: End of Transmission

0 = The register SSC_TCR has not reached 0 since the last write in SSC_TCR or SSC_TNCR.
1 = The register SSC_TCR has reached 0 since the last write in SSC_TCR or SSC_TNCR.
• TXBUFE: Transmit Buffer Empty

0 = SSC_TCR or SSC_TNCR have a value other than 0.
1 = Both SSC_TCR and SSC_TNCR have a value of 0.
• RXRDY: Receive Ready

0 = SSC_RHR is empty.
1 = Data has been received and loaded in SSC_RHR.
• OVRUN: Receive Overrun

0 = No data has been loaded in SSC_RHR while previous data has not been read since the last read of the Status
Register.
1 = Data has been loaded in SSC_RHR while previous data has not yet been read since the last read of the Status
Register.
• ENDRX: End of Reception

0 = Data is written on the Receive Counter Register or Receive Next Counter Register.
1 = End of PDCDMAC transfer when Receive Counter Register has arrived at zero.
605
AT91SAM9G45
• RXBUFF: Receive Buffer Full

0 = SSC_RCR or SSC_RNCR have a value other than 0.
1 = Both SSC_RCR and SSC_RNCR have a value of 0.
• CP0: Compare 0
0 = A compare 0 has not occurred since the last read of the Status Register.
1 = A compare 0 has occurred since the last read of the Status Register.
• CP1: Compare 1
0 = A compare 1 has not occurred since the last read of the Status Register.
1 = A compare 1 has occurred since the last read of the Status Register.
• TXSYN: Transmit Sync

0 = A Tx Sync has not occurred since the last read of the Status Register.
1 = A Tx Sync has occurred since the last read of the Status Register.
• RXSYN: Receive Sync

0 = An Rx Sync has not occurred since the last read of the Status Register.
1 = An Rx Sync has occurred since the last read of the Status Register.
• TXEN: Transmit Enable

0 = Transmit is disabled.
1 = Transmit is enabled.
• RXEN: Receive Enable

0 = Receive is disabled.
1 = Receive is enabled.
606
AT91SAM9G45
32.9.14 SSC Interrupt Enable Register

Name: SSC_IER
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• TXRDY: Transmit Ready Interrupt Enable

0 = 0 = No effect.
1 = Enables the Transmit Ready Interrupt.
• TXEMPTY: Transmit Empty Interrupt Enable

0 = No effect.
1 = Enables the Transmit Empty Interrupt.
• ENDTX: End of Transmission Interrupt Enable

0 = No effect.
1 = Enables the End of Transmission Interrupt.
• TXBUFE: Transmit Buffer Empty Interrupt Enable

0 = No effect.
1 = Enables the Transmit Buffer Empty Interrupt
• RXRDY: Receive Ready Interrupt Enable

0 = No effect.
1 = Enables the Receive Ready Interrupt.
• OVRUN: Receive Overrun Interrupt Enable

0 = No effect.
1 = Enables the Receive Overrun Interrupt.
• ENDRX: End of Reception Interrupt Enable

0 = No effect.
1 = Enables the End of Reception Interrupt.
607
AT91SAM9G45
0 = No effect.
1 = Enables the Receive Buffer Full Interrupt.
• CP0: Compare 0 Interrupt Enable

0 = No effect.
1 = Enables the Compare 0 Interrupt.
• CP1: Compare 1 Interrupt Enable

0 = No effect.
1 = Enables the Compare 1 Interrupt.
• TXSYN: Tx Sync Interrupt Enable

0 = No effect.
1 = Enables the Tx Sync Interrupt.
• RXSYN: Rx Sync Interrupt Enable

0 = No effect.
1 = Enables the Rx Sync Interrupt.
608
AT91SAM9G45
32.9.15 SSC Interrupt Disable Register

Name: SSC_IDR
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• TXRDY: Transmit Ready Interrupt Disable

0 = No effect.
1 = Disables the Transmit Ready Interrupt.
• TXEMPTY: Transmit Empty Interrupt Disable

0 = No effect.
1 = Disables the Transmit Empty Interrupt.
• ENDTX: End of Transmission Interrupt Disable

0 = No effect.
1 = Disables the End of Transmission Interrupt.
• TXBUFE: Transmit Buffer Empty Interrupt Disable

0 = No effect.
1 = Disables the Transmit Buffer Empty Interrupt.
• RXRDY: Receive Ready Interrupt Disable

0 = No effect.
1 = Disables the Receive Ready Interrupt.
• OVRUN: Receive Overrun Interrupt Disable

0 = No effect.
1 = Disables the Receive Overrun Interrupt.
• ENDRX: End of Reception Interrupt Disable

0 = No effect.
1 = Disables the End of Reception Interrupt.
609
AT91SAM9G45
0 = No effect.
1 = Disables the Receive Buffer Full Interrupt.
• CP0: Compare 0 Interrupt Disable

0 = No effect.
1 = Disables the Compare 0 Interrupt.
• CP1: Compare 1 Interrupt Disable

0 = No effect.
1 = Disables the Compare 1 Interrupt.
• TXSYN: Tx Sync Interrupt Enable

0 = No effect.
1 = Disables the Tx Sync Interrupt.
• RXSYN: Rx Sync Interrupt Enable

0 = No effect.
1 = Disables the Rx Sync Interrupt.
610
AT91SAM9G45
32.9.16 SSC Interrupt Mask Register

Name: SSC_IMR
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• TXRDY: Transmit Ready Interrupt Mask

0 = The Transmit Ready Interrupt is disabled.
1 = The Transmit Ready Interrupt is enabled.
• TXEMPTY: Transmit Empty Interrupt Mask

0 = The Transmit Empty Interrupt is disabled.
1 = The Transmit Empty Interrupt is enabled.
• ENDTX: End of Transmission Interrupt Mask

0 = The End of Transmission Interrupt is disabled.
1 = The End of Transmission Interrupt is enabled.
• TXBUFE: Transmit Buffer Empty Interrupt Mask

0 = The Transmit Buffer Empty Interrupt is disabled.
1 = The Transmit Buffer Empty Interrupt is enabled.
• RXRDY: Receive Ready Interrupt Mask

0 = The Receive Ready Interrupt is disabled.
1 = The Receive Ready Interrupt is enabled.
• OVRUN: Receive Overrun Interrupt Mask

0 = The Receive Overrun Interrupt is disabled.
1 = The Receive Overrun Interrupt is enabled.
• ENDRX: End of Reception Interrupt Mask

0 = The End of Reception Interrupt is disabled.
1 = The End of Reception Interrupt is enabled.
611
AT91SAM9G45

0 = The Receive Buffer Full Interrupt is disabled.
1 = The Receive Buffer Full Interrupt is enabled.
• CP0: Compare 0 Interrupt Mask

0 = The Compare 0 Interrupt is disabled.
1 = The Compare 0 Interrupt is enabled.
• CP1: Compare 1 Interrupt Mask

0 = The Compare 1 Interrupt is disabled.
1 = The Compare 1 Interrupt is enabled.
• TXSYN: Tx Sync Interrupt Mask

0 = The Tx Sync Interrupt is disabled.
1 = The Tx Sync Interrupt is enabled.
• RXSYN: Rx Sync Interrupt Mask

0 = The Rx Sync Interrupt is disabled.
1 = The Rx Sync Interrupt is enabled.
612
AT91SAM9G45
33. Timer Counter (TC)
33.1 Description
The Timer Counter (TC) includes three identical 16-bit Timer Counter channels.
Each channel can be independently programmed to perform a wide range of functions including
frequency measurement, event counting, interval measurement, pulse generation, delay timing
and pulse width modulation.
Each channel has three external clock inputs, five internal clock inputs and two multi-purpose
input/output signals which can be configured by the user. Each channel drives an internal inter-
rupt signal which can be programmed to generate processor interrupts.
The Timer Counter block has two global registers which act upon all three TC channels.
The Block Control Register allows the three channels to be started simultaneously with the same
instruction.
The Block Mode Register defines the external clock inputs for each channel, allowing them to be
chained.
Table 33-1 gives the assignment of the device Timer Counter clock inputs common to Timer
Counter 0 to 2.
Table 33-1. Timer Counter Clock Assignment

Name Definition
TIMER_CLOCK1 MCK/2
TIMER_CLOCK2 MCK/8
TIMER_CLOCK3 MCK/32
TIMER_CLOCK4 MCK/128
(1)
TIMER_CLOCK5 SLCK
Note: 1. When Slow Clock is selected for Master Clock (CSS = 0 in PMC Master CLock Register),
TIMER_CLOCK5 input is Master Clock, i.e., Slow CLock modified by PRES and MDIV fields.

• Three 16-bit Timer Counter Channels
• Wide range of functions including:
– Frequency Measurement
– Event Counting
– Interval Measurement
– Pulse Generation
– Delay Timing
– Pulse Width Modulation
– Up/down Capabilities
• Each channel is user-configurable and contains:
– Three external clock inputs
– Five internal clock inputs
613
– Two multi-purpose input/output signals
• Two global registers that act on all three TC Channels
33.3 Block Diagram
Figure 33-1. Timer Counter Block Diagram
Parallel I/O
TIMER_CLOCK1 Controller
TCLK0
TCLK0
TIMER_CLOCK2 TCLK1
TIOA1 TCLK2
TIMER_CLOCK3 TIOA2 XC0 Timer/Counter

TIOA
TCLK1 XC1 Channel 0 TIOA0 TIOA0
TIMER_CLOCK4 TIOB TIOB0
TCLK2 XC2 TIOB0
TIMER_CLOCK5 TC0XC0S SYNC
INT0
TCLK0
TCLK1 XC0 Timer/Counter

TIOA
TIOA0 XC1 Channel 1 TIOA1 TIOA1
TIOB TIOB1
TIOA2 XC2 TIOB1
TCLK2 SYNC
INT1
TC1XC1S
TCLK0 XC0 Timer/Counter

TIOA
TCLK1 XC1 Channel 2 TIOA2 TIOA2
TIOB TIOB2
TCLK2 XC2 TIOB2
TIOA0 SYNC
TC2XC2S INT2
TIOA1
Timer Counter
Interrupt
Controller
Table 33-2. Signal Name Description

Block/Channel Signal Name Description
XC0, XC1, XC2 External Clock Inputs
Capture Mode: Timer Counter Input
TIOA
Waveform Mode: Timer Counter Output
Channel Signal Capture Mode: Timer Counter Input
TIOB
Waveform Mode: Timer Counter Input/Output
INT Interrupt Signal Output
SYNC Synchronization Input Signal
614 AT91SAM9G45
33.4 Pin Name List
Table 33-3. TC pin list

Pin Name Description Type
TCLK0-TCLK2 External Clock Input Input
TIOA0-TIOA2 I/O Line A I/O
TIOB0-TIOB2 I/O Line B I/O
33.5.1 I/O Lines

The programmer must first program the PIO controllers to assign the TC pins to their peripheral
functions.

TC0 TCLK0 PD23 A
TC0 TCLK1 PD29 A
TC0 TCLK2 PC10 B
TC0 TIOA0 PD20 A
TC0 TIOA1 PD21 A
TC0 TIOA2 PD22 A
TC0 TIOB0 PD30 A
TC0 TIOB1 PD31 A
TC0 TIOB2 PA26 B
TC1 TCLK3 PA0 B
TC1 TCLK4 PA3 B
TC1 TCLK5 PD9 B
TC1 TIOA3 PA1 B
TC1 TIOA4 PA4 B
TC1 TIOA5 PD7 B
TC1 TIOB3 PA2 B
TC1 TIOB4 PA5 B
TC1 TIOB5 PD8 B

The TC is clocked through the Power Management Controller (PMC), thus the programmer must
first configure the PMC to enable the Timer Counter clock.
615 AT91SAM9G45
33.5.3 Interrupt
The TC has an interrupt line connected to the Interrupt Controller (IC). Handling the TC interrupt
requires programming the IC before configuring the TC.
33.6.1 TC Description
The three channels of the Timer Counter are independent and identical in operation . The regis-
ters for channel programming are listed in Table 33-5 on page 629.
33.6.2 16-bit Counter

Each channel is organized around a 16-bit counter. The value of the counter is incremented at
each positive edge of the selected clock. When the counter has reached the value 0xFFFF and
passes to 0x0000, an overflow occurs and the COVFS bit in TC_SR (Status Register) is set.
The current value of the counter is accessible in real time by reading the Counter Value Regis-
ter, TC_CV. The counter can be reset by a trigger. In this case, the counter value passes to
0x0000 on the next valid edge of the selected clock.
33.6.3 Clock Selection

At block level, input clock signals of each channel can either be connected to the external inputs
TCLK0, TCLK1 or TCLK2, or be connected to the internal I/O signals TIOA0, TIOA1 or TIOA2
for chaining by programming the TC_BMR (Block Mode). See Figure 33-2 ”Clock Chaining
Selection”.
Each channel can independently select an internal or external clock source for its counter:
• Internal clock signals: TIMER_CLOCK1, TIMER_CLOCK2, TIMER_CLOCK3,
TIMER_CLOCK4, TIMER_CLOCK5
• External clock signals: XC0, XC1 or XC2
This selection is made by the TCCLKS bits in the TC Channel Mode Register.
The selected clock can be inverted with the CLKI bit in TC_CMR. This allows counting on the
opposite edges of the clock.
The burst function allows the clock to be validated when an external signal is high. The BURST
parameter in the Mode Register defines this signal (none, XC0, XC1, XC2). See Figure 33-3
”Clock Selection”
Note: In all cases, if an external clock is used, the duration of each of its levels must be longer than the
master clock period. The external clock frequency must be at least 2.5 times lower than the mas-
ter clock
616 AT91SAM9G45
Figure 33-2. Clock Chaining Selection
TC0XC0S
Timer/Counter
TCLK0 Channel 0
TIOA1
XC0 TIOA0
TIOA2
XC1 = TCLK1
XC2 = TCLK2 TIOB0
SYNC
TC1XC1S
Timer/Counter
Channel 1
TCLK1 XC0 = TCLK2 TIOA1
TIOA0
XC1
TIOA2
XC2 = TCLK2 TIOB1
SYNC
Timer/Counter
TC2XC2S Channel 2
XC0 = TCLK0 TIOA2
TCLK2 XC1 = TCLK1
TIOA0
XC2 TIOB2
TIOA1
SYNC
Figure 33-3. Clock Selection
TCCLKS
TIMER_CLOCK1
TIMER_CLOCK2 CLKI
TIMER_CLOCK3
TIMER_CLOCK4
TIMER_CLOCK5
Selected
XC0 Clock
XC1
XC2
BURST
617 AT91SAM9G45
33.6.4 Clock Control
The clock of each counter can be controlled in two different ways: it can be enabled/disabled
and started/stopped. See Figure 33-4.
• The clock can be enabled or disabled by the user with the CLKEN and the CLKDIS
commands in the Control Register. In Capture Mode it can be disabled by an RB load event
if LDBDIS is set to 1 in TC_CMR. In Waveform Mode, it can be disabled by an RC Compare
event if CPCDIS is set to 1 in TC_CMR. When disabled, the start or the stop actions have no
effect: only a CLKEN command in the Control Register can re-enable the clock. When the
clock is enabled, the CLKSTA bit is set in the Status Register.
• The clock can also be started or stopped: a trigger (software, synchro, external or compare)
always starts the clock. The clock can be stopped by an RB load event in Capture Mode
(LDBSTOP = 1 in TC_CMR) or a RC compare event in Waveform Mode (CPCSTOP = 1 in
TC_CMR). The start and the stop commands have effect only if the clock is enabled.
Figure 33-4. Clock Control

Selected
Clock Trigger
CLKSTA CLKEN CLKDIS
Q S
R
Q S
R
Stop Disable
Counter Event Event
Clock
33.6.5 TC Operating Modes

Each channel can independently operate in two different modes:
• Capture Mode provides measurement on signals.
• Waveform Mode provides wave generation.
The TC Operating Mode is programmed with the WAVE bit in the TC Channel Mode Register.
In Capture Mode, TIOA and TIOB are configured as inputs.
In Waveform Mode, TIOA is always configured to be an output and TIOB is an output if it is not
selected to be the external trigger.
33.6.6 Trigger
A trigger resets the counter and starts the counter clock. Three types of triggers are common to
both modes, and a fourth external trigger is available to each mode.
618 AT91SAM9G45
Regardless of the trigger used, it will be taken into account at the following active edge of the
selected clock. This means that the counter value can be read differently from zero just after a
trigger, especially when a low frequency signal is selected as the clock.
The following triggers are common to both modes:
• Software Trigger: Each channel has a software trigger, available by setting SWTRG in
TC_CCR.
• SYNC: Each channel has a synchronization signal SYNC. When asserted, this signal has
the same effect as a software trigger. The SYNC signals of all channels are asserted
simultaneously by writing TC_BCR (Block Control) with SYNC set.
• Compare RC Trigger: RC is implemented in each channel and can provide a trigger when
the counter value matches the RC value if CPCTRG is set in TC_CMR.
The channel can also be configured to have an external trigger. In Capture Mode, the external
trigger signal can be selected between TIOA and TIOB. In Waveform Mode, an external event
can be programmed on one of the following signals: TIOB, XC0, XC1 or XC2. This external
event can then be programmed to perform a trigger by setting ENETRG in TC_CMR.
If an external trigger is used, the duration of the pulses must be longer than the master clock
period in order to be detected.
33.6.7 Capture Operating Mode

This mode is entered by clearing the WAVE parameter in TC_CMR (Channel Mode Register).
Capture Mode allows the TC channel to perform measurements such as pulse timing, fre-
quency, period, duty cycle and phase on TIOA and TIOB signals which are considered as
inputs.
Figure 33-5 shows the configuration of the TC channel when programmed in Capture Mode.
33.6.8 Capture Registers A and B

Registers A and B (RA and RB) are used as capture registers. This means that they can be
loaded with the counter value when a programmable event occurs on the signal TIOA.
The LDRA parameter in TC_CMR defines the TIOA edge for the loading of register A, and the
LDRB parameter defines the TIOA edge for the loading of Register B.
RA is loaded only if it has not been loaded since the last trigger or if RB has been loaded since
the last loading of RA.
RB is loaded only if RA has been loaded since the last trigger or the last loading of RB.
Loading RA or RB before the read of the last value loaded sets the Overrun Error Flag (LOVRS)
in TC_SR (Status Register). In this case, the old value is overwritten.
33.6.9 Trigger Conditions

In addition to the SYNC signal, the software trigger and the RC compare trigger, an external trig-
ger can be defined.
The ABETRG bit in TC_CMR selects TIOA or TIOB input signal as an external trigger. The
ETRGEDG parameter defines the edge (rising, falling or both) detected to generate an external
trigger. If ETRGEDG = 0 (none), the external trigger is disabled.
619 AT91SAM9G45
620
TCCLKS
CLKSTA CLKEN CLKDIS
CLKI
TIMER_CLOCK1
TIMER_CLOCK2
TIMER_CLOCK3
Q S
TIMER_CLOCK4
TIMER_CLOCK5 R
Q S
Figure 33-5. Capture Mode
XC0
R
XC1
XC2
LDBSTOP LDBDIS
AT91SAM9G45
BURST
Register C
Capture Capture
1 Register A Register B Compare RC =
16-bit Counter
SWTRG
CLK
OVF
RESET
SYNC
Trig
ABETRG
ETRGEDG CPCTRG
MTIOB Edge
Detector
TIOB
LDRA LDRB
CPCS
LDRAS
LDRBS
LOVRS
ETRGS
COVFS
TC1_SR
MTIOA Edge Edge

Detector Detector
If RA is not loaded
or RB is Loaded If RA is Loaded
TIOA
TC1_IMR
Timer/Counter Channel
INT
33.6.10 Waveform Operating Mode
Waveform operating mode is entered by setting the WAVE parameter in TC_CMR (Channel
Mode Register).
In Waveform Operating Mode the TC channel generates 1 or 2 PWM signals with the same fre-
quency and independently programmable duty cycles, or generates different types of one-shot
or repetitive pulses.
In this mode, TIOA is configured as an output and TIOB is defined as an output if it is not used
as an external event (EEVT parameter in TC_CMR).
Figure 33-6 shows the configuration of the TC channel when programmed in Waveform Operat-
ing Mode.
33.6.11 Waveform Selection

Depending on the WAVSEL parameter in TC_CMR (Channel Mode Register), the behavior of
TC_CV varies.
With any selection, RA, RB and RC can all be used as compare registers.
RA Compare is used to control the TIOA output, RB Compare is used to control the TIOB output
(if correctly configured) and RC Compare is used to control TIOA and/or TIOB outputs.
621 AT91SAM9G45
622
TCCLKS
CLKSTA CLKEN CLKDIS
TIMER_CLOCK1 ACPC
CLKI
TIMER_CLOCK2
TIMER_CLOCK3
Figure 33-6. Waveform Mode
Q S
TIMER_CLOCK4 CPCDIS MTIOA
TIMER_CLOCK5 R ACPA
Q S
XC0
R
AT91SAM9G45
XC1
XC2 CPCSTOP TIOA
AEEVT
Output Controller
BURST Register A Register B Register C

WAVSEL
ASWTRG
1 Compare RA = Compare RB = Compare RC =
16-bit Counter
CLK
OVF
RESET
SWTRG
BCPC
SYNC
Trig
BCPB MTIOB
WAVSEL
EEVT
TIOB
BEEVT
EEVTEDG
ENETRG
Output Controller
CPAS
CPBS
CPCS
ETRGS
COVFS
Edge
TC1_SR
Detector BSWTRG
TIOB
TC1_IMR
Timer/Counter Channel
INT
33.6.11.1 WAVSEL = 00
When WAVSEL = 00, the value of TC_CV is incremented from 0 to 0xFFFF. Once 0xFFFF has
been reached, the value of TC_CV is reset. Incrementation of TC_CV starts again and the cycle
continues. See Figure 33-7.
An external event trigger or a software trigger can reset the value of TC_CV. It is important to
note that the trigger may occur at any time. See Figure 33-8.
RC Compare cannot be programmed to generate a trigger in this configuration. At the same
time, RC Compare can stop the counter clock (CPCSTOP = 1 in TC_CMR) and/or disable the
counter clock (CPCDIS = 1 in TC_CMR).
Figure 33-7. WAVSEL= 00 without trigger

Counter Value Counter cleared by compare match with 0xFFFF
0xFFFF
RC
RB
RA
Waveform Examples Time
TIOB
TIOA
623 AT91SAM9G45
Figure 33-8. WAVSEL= 00 with trigger
Counter Value Counter cleared by compare match with 0xFFFF
0xFFFF
RC Counter cleared by trigger
RB
RA
Time
Waveform Examples
TIOB
TIOA
33.6.11.2 WAVSEL = 10
When WAVSEL = 10, the value of TC_CV is incremented from 0 to the value of RC, then auto-
matically reset on a RC Compare. Once the value of TC_CV has been reset, it is then
incremented and so on. See Figure 33-9.
It is important to note that TC_CV can be reset at any time by an external event or a software
trigger if both are programmed correctly. See Figure 33-10.
In addition, RC Compare can stop the counter clock (CPCSTOP = 1 in TC_CMR) and/or disable
the counter clock (CPCDIS = 1 in TC_CMR).
Figure 33-9. WAVSEL = 10 Without Trigger

Counter Value
0xFFFF
Counter cleared by compare match with RC
RC
RB
RA
TIOB
TIOA
624 AT91SAM9G45
Figure 33-10. WAVSEL = 10 With Trigger
Counter Value
0xFFFF
Counter cleared by compare match with RC Counter cleared by trigger
RC
RB
RA
TIOB
TIOA
33.6.11.3 WAVSEL = 01
When WAVSEL = 01, the value of TC_CV is incremented from 0 to 0xFFFF. Once 0xFFFF is
reached, the value of TC_CV is decremented to 0, then re-incremented to 0xFFFF and so on.
See Figure 33-11.
A trigger such as an external event or a software trigger can modify TC_CV at any time. If a trig-
ger occurs while TC_CV is incrementing, TC_CV then decrements. If a trigger is received while
TC_CV is decrementing, TC_CV then increments. See Figure 33-12.
RC Compare cannot be programmed to generate a trigger in this configuration.
At the same time, RC Compare can stop the counter clock (CPCSTOP = 1) and/or disable the
counter clock (CPCDIS = 1).
625 AT91SAM9G45
Counter Value Counter decremented by compare match with 0xFFFF
0xFFFF
RC
RB
RA
TIOB
TIOA

Counter Value Counter decremented by compare match with 0xFFFF
0xFFFF
Counter decremented
by trigger
RC
RB
Counter incremented
by trigger
RA
TIOB
TIOA
33.6.11.4 WAVSEL = 11
When WAVSEL = 11, the value of TC_CV is incremented from 0 to RC. Once RC is reached, the
value of TC_CV is decremented to 0, then re-incremented to RC and so on. See Figure 33-13.
A trigger such as an external event or a software trigger can modify TC_CV at any time. If a trig-
ger occurs while TC_CV is incrementing, TC_CV then decrements. If a trigger is received while
TC_CV is decrementing, TC_CV then increments. See Figure 33-14.
RC Compare can stop the counter clock (CPCSTOP = 1) and/or disable the counter clock
(CPCDIS = 1).
626 AT91SAM9G45
Counter Value
0xFFFF
Counter decremented by compare match with RC
RC
RB
RA
TIOB
TIOA

Counter Value
0xFFFF
Counter decremented by compare match with RC
RC
Counter decremented
by trigger
RB
Counter incremented
by trigger
RA
TIOB
TIOA
627 AT91SAM9G45
33.6.12 External Event/Trigger Conditions
An external event can be programmed to be detected on one of the clock sources (XC0, XC1,
XC2) or TIOB. The external event selected can then be used as a trigger.
The EEVT parameter in TC_CMR selects the external trigger. The EEVTEDG parameter defines
the trigger edge for each of the possible external triggers (rising, falling or both). If EEVTEDG is
cleared (none), no external event is defined.
If TIOB is defined as an external event signal (EEVT = 0), TIOB is no longer used as an output
and the compare register B is not used to generate waveforms and subsequently no IRQs. In
this case the TC channel can only generate a waveform on TIOA.
When an external event is defined, it can be used as a trigger by setting bit ENETRG in
TC_CMR.
As in Capture Mode, the SYNC signal and the software trigger are also available as triggers. RC
Compare can also be used as a trigger depending on the parameter WAVSEL.
33.6.13 Output Controller

The output controller defines the output level changes on TIOA and TIOB following an event.
TIOB control is used only if TIOB is defined as output (not as an external event).
The following events control TIOA and TIOB: software trigger, external event and RC compare.
RA compare controls TIOA and RB compare controls TIOB. Each of these events can be pro-
grammed to set, clear or toggle the output as defined in the corresponding parameter in
TC_CMR.
628 AT91SAM9G45
33.7 Timer Counter (TC) User Interface

Offset(1) Register Name Access Reset
0x00 + channel * 0x40 + 0x00 Channel Control Register TC_CCR Write-only –
0x00 + channel * 0x40 + 0x04 Channel Mode Register TC_CMR Read-write 0
0x00 + channel * 0x40 + 0x08 Reserved
0x00 + channel * 0x40 + 0x0C Reserved
0x00 + channel * 0x40 + 0x10 Counter Value TC_CV Read-only 0
0x00 + channel * 0x40 + 0x14 Register A TC_RA Read-write(2) 0
0x00 + channel * 0x40 + 0x18 Register B TC_RB Read-write(2) 0
0x00 + channel * 0x40 + 0x1C Register C TC_RC Read-write 0
0x00 + channel * 0x40 + 0x20 Status Register TC_SR Read-only 0
0x00 + channel * 0x40 + 0x24 Interrupt Enable Register TC_IER Write-only –
0x00 + channel * 0x40 + 0x28 Interrupt Disable Register TC_IDR Write-only –
0x00 + channel * 0x40 + 0x2C Interrupt Mask Register TC_IMR Read-only 0
0xC0 Block Control Register TC_BCR Write-only –
0xC4 Block Mode Register TC_BMR Read-write 0
0xD8 Reserved
0xE4 Reserved
0xFC Reserved – – –
Notes: 1. Channel index ranges from 0 to 2.
2. Read-only if WAVE = 0
629 AT91SAM9G45
33.7.1 TC Block Control Register
Name: TC_BCR
Addresses: 0xFFF7C0C0 (0), 0xFFFD40C0 (1)
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – SYNC
• SYNC: Synchro Command

0 = no effect.
1 = asserts the SYNC signal which generates a software trigger simultaneously for each of the channels.
630 AT91SAM9G45
33.7.2 TC Block Mode Register
Name: TC_BMR
Addresses: 0xFFF7C0C4 (0), 0xFFFD40C4 (1)
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – TC2XC2S TC1XC1S TC0XC0S
• TC0XC0S: External Clock Signal 0 Selection
TC0XC0S Signal Connected to XC0

0 0 TCLK0
0 1 none
1 0 TIOA1
1 1 TIOA2

0 0 TCLK1
0 1 none
1 0 TIOA0
1 1 TIOA2

0 0 TCLK2
0 1 none
1 0 TIOA0
1 1 TIOA1
631 AT91SAM9G45
33.7.3 TC Channel Control Register
Name: TC_CCRx [x=0..2]
Addresses: 0xFFF7C000 (0)[0], 0xFFF7C040 (0)[1], 0xFFF7C080 (0)[2], 0xFFFD4000 (1)[0], 0xFFFD4040
(1)[1], 0xFFFD4080 (1)[2]
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – SWTRG CLKDIS CLKEN
• CLKEN: Counter Clock Enable Command

0 = no effect.
1 = enables the clock if CLKDIS is not 1.
• CLKDIS: Counter Clock Disable Command

0 = no effect.
1 = disables the clock.
• SWTRG: Software Trigger Command

0 = no effect.
1 = a software trigger is performed: the counter is reset and the clock is started.
632 AT91SAM9G45
33.7.4 TC Channel Mode Register: Capture Mode
Name: TC_CMRx [x=0..2] (WAVE = 0)
Addresses: 0xFFF7C004 (0)[0], 0xFFF7C044 (0)[1], 0xFFF7C084 (0)[2], 0xFFFD4004 (1)[0],
0xFFFD4044 (1)[1], 0xFFFD4084 (1)[2]
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – LDRB LDRA
15 14 13 12 11 10 9 8
WAVE CPCTRG – – – ABETRG ETRGEDG
7 6 5 4 3 2 1 0
LDBDIS LDBSTOP BURST CLKI TCCLKS
• TCCLKS: Clock Selection
TCCLKS Clock Selected

0 0 0 TIMER_CLOCK1
0 0 1 TIMER_CLOCK2
0 1 0 TIMER_CLOCK3
0 1 1 TIMER_CLOCK4
1 0 0 TIMER_CLOCK5
1 0 1 XC0
1 1 0 XC1
1 1 1 XC2
• CLKI: Clock Invert

0 = counter is incremented on rising edge of the clock.
1 = counter is incremented on falling edge of the clock.
• BURST: Burst Signal Selection
BURST
0 0 The clock is not gated by an external signal.
0 1 XC0 is ANDed with the selected clock.
• LDBSTOP: Counter Clock Stopped with RB Loading

0 = counter clock is not stopped when RB loading occurs.
1 = counter clock is stopped when RB loading occurs.
633 AT91SAM9G45
• LDBDIS: Counter Clock Disable with RB Loading
0 = counter clock is not disabled when RB loading occurs.
1 = counter clock is disabled when RB loading occurs.
• ETRGEDG: External Trigger Edge Selection
ETRGEDG Edge
0 0 none
0 1 rising edge
1 0 falling edge
1 1 each edge
• ABETRG: TIOA or TIOB External Trigger Selection

0 = TIOB is used as an external trigger.
1 = TIOA is used as an external trigger.
• CPCTRG: RC Compare Trigger Enable

0 = RC Compare has no effect on the counter and its clock.
1 = RC Compare resets the counter and starts the counter clock.
• WAVE
0 = Capture Mode is enabled.
1 = Capture Mode is disabled (Waveform Mode is enabled).
• LDRA: RA Loading Selection
LDRA Edge
0 0 none
0 1 rising edge of TIOA
1 0 falling edge of TIOA
1 1 each edge of TIOA
• LDRB: RB Loading Selection
LDRB Edge
0 0 none
0 1 rising edge of TIOA
1 0 falling edge of TIOA
1 1 each edge of TIOA
634 AT91SAM9G45
33.7.5 TC Channel Mode Register: Waveform Mode
Name: TC_CMRx [x=0..2] (WAVE = 1)
0xFFFD4044 (1)[1], 0xFFFD4084 (1)[2]
Access: Read-write
31 30 29 28 27 26 25 24
BSWTRG BEEVT BCPC BCPB
23 22 21 20 19 18 17 16
ASWTRG AEEVT ACPC ACPA
15 14 13 12 11 10 9 8
WAVE WAVSEL ENETRG EEVT EEVTEDG
7 6 5 4 3 2 1 0
CPCDIS CPCSTOP BURST CLKI TCCLKS
• TCCLKS: Clock Selection
TCCLKS Clock Selected

0 0 0 TIMER_CLOCK1
0 0 1 TIMER_CLOCK2
0 1 0 TIMER_CLOCK3
0 1 1 TIMER_CLOCK4
1 0 0 TIMER_CLOCK5
1 0 1 XC0
1 1 0 XC1
1 1 1 XC2
• CLKI: Clock Invert

0 = counter is incremented on rising edge of the clock.
1 = counter is incremented on falling edge of the clock.
• BURST: Burst Signal Selection
BURST
0 0 The clock is not gated by an external signal.
• CPCSTOP: Counter Clock Stopped with RC Compare

0 = counter clock is not stopped when counter reaches RC.
1 = counter clock is stopped when counter reaches RC.
635 AT91SAM9G45
• CPCDIS: Counter Clock Disable with RC Compare
0 = counter clock is not disabled when counter reaches RC.
1 = counter clock is disabled when counter reaches RC.
• EEVTEDG: External Event Edge Selection
EEVTEDG Edge
0 0 none
0 1 rising edge
1 0 falling edge
1 1 each edge
• EEVT: External Event Selection
EEVT Signal selected as external event TIOB Direction

0 0 TIOB input (1)
0 1 XC0 output
1 0 XC1 output
1 1 XC2 output
Note: 1. If TIOB is chosen as the external event signal, it is configured as an input and no longer generates waveforms and subse-
quently no IRQs.
• ENETRG: External Event Trigger Enable

0 = the external event has no effect on the counter and its clock. In this case, the selected external event only controls the
TIOA output.
1 = the external event resets the counter and starts the counter clock.
• WAVSEL: Waveform Selection
WAVSEL Effect
0 0 UP mode without automatic trigger on RC Compare
1 0 UP mode with automatic trigger on RC Compare
0 1 UPDOWN mode without automatic trigger on RC Compare
1 1 UPDOWN mode with automatic trigger on RC Compare
• WAVE
0 = Waveform Mode is disabled (Capture Mode is enabled).
1 = Waveform Mode is enabled.
636 AT91SAM9G45
• ACPA: RA Compare Effect on TIOA
ACPA Effect
0 0 none
0 1 set
1 0 clear
1 1 toggle
• ACPC: RC Compare Effect on TIOA
ACPC Effect
0 0 none
0 1 set
1 0 clear
1 1 toggle
• AEEVT: External Event Effect on TIOA
AEEVT Effect
0 0 none
0 1 set
1 0 clear
1 1 toggle
• ASWTRG: Software Trigger Effect on TIOA
ASWTRG Effect
0 0 none
0 1 set
1 0 clear
1 1 toggle
• BCPB: RB Compare Effect on TIOB
BCPB Effect
0 0 none
0 1 set
1 0 clear
1 1 toggle
637 AT91SAM9G45
• BCPC: RC Compare Effect on TIOB
BCPC Effect
0 0 none
0 1 set
1 0 clear
1 1 toggle
• BEEVT: External Event Effect on TIOB
BEEVT Effect
0 0 none
0 1 set
1 0 clear
1 1 toggle
• BSWTRG: Software Trigger Effect on TIOB
BSWTRG Effect
0 0 none
0 1 set
1 0 clear
1 1 toggle
638 AT91SAM9G45
33.7.6 TC Counter Value Register
Name: TC_CVx [x=0..2]
Addresses: 0xFFF7C010 (0)[0], 0xFFF7C050 (0)[1], 0xFFF7C090 (0)[2], 0xFFFD4010 (1)[0]
0xFFFD4050 (1)[1], 0xFFFD4090 (1)[2]
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
CV
7 6 5 4 3 2 1 0
CV
• CV: Counter Value

CV contains the counter value in real time.
639 AT91SAM9G45
33.7.7 TC Register A
Name: TC_RAx [x=0..2]
0xFFFD4054 (1)[1], 0xFFFD4094 (1)[2]
Access: Read-only if WAVE = 0, Read-write if WAVE = 1
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RA
7 6 5 4 3 2 1 0
RA
• RA: Register A
RA contains the Register A value in real time.
33.7.8 TC Register B
Name: TC_RBx [x=0..2]
0xFFFD4058 (1)[1], 0xFFFD4098 (1)[2]
Access: Read-only if WAVE = 0, Read-write if WAVE = 1
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RB
7 6 5 4 3 2 1 0
RB
• RB: Register B
RB contains the Register B value in real time.
640 AT91SAM9G45
33.7.9 TC Register C
Name: TC_RCx [x=0..2]
Addresses: 0xFFF7C01C (0)[0], 0xFFF7C05C (0)[1], 0xFFF7C09C (0)[2], 0xFFFD401C (1)[0],
0xFFFD405C (1)[1], 0xFFFD409C (1)[2]
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RC
7 6 5 4 3 2 1 0
RC
• RC: Register C
RC contains the Register C value in real time.
641 AT91SAM9G45
33.7.10 TC Status Register
Name: TC_SRx [x=0..2]
Addresses: 0xFFF7C020 (0)[0], 0xFFF7C060 (0)[1], 0xFFF7C0A0 (0)[2], 0xFFFD4020 (1)[0],
0xFFFD4060 (1)[1], 0xFFFD40A0 (1)[2]
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – MTIOB MTIOA CLKSTA
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
ETRGS LDRBS LDRAS CPCS CPBS CPAS LOVRS COVFS
• COVFS: Counter Overflow Status

0 = no counter overflow has occurred since the last read of the Status Register.
1 = a counter overflow has occurred since the last read of the Status Register.
• LOVRS: Load Overrun Status

0 = Load overrun has not occurred since the last read of the Status Register or WAVE = 1.
1 = RA or RB have been loaded at least twice without any read of the corresponding register since the last read of the Sta-
tus Register, if WAVE = 0.
• CPAS: RA Compare Status

0 = RA Compare has not occurred since the last read of the Status Register or WAVE = 0.
1 = RA Compare has occurred since the last read of the Status Register, if WAVE = 1.
• CPBS: RB Compare Status

0 = RB Compare has not occurred since the last read of the Status Register or WAVE = 0.
1 = RB Compare has occurred since the last read of the Status Register, if WAVE = 1.
• CPCS: RC Compare Status

0 = RC Compare has not occurred since the last read of the Status Register.
1 = RC Compare has occurred since the last read of the Status Register.
• LDRAS: RA Loading Status

0 = RA Load has not occurred since the last read of the Status Register or WAVE = 1.
1 = RA Load has occurred since the last read of the Status Register, if WAVE = 0.
• LDRBS: RB Loading Status

0 = RB Load has not occurred since the last read of the Status Register or WAVE = 1.
1 = RB Load has occurred since the last read of the Status Register, if WAVE = 0.
642 AT91SAM9G45
• ETRGS: External Trigger Status
0 = external trigger has not occurred since the last read of the Status Register.
1 = external trigger has occurred since the last read of the Status Register.
• CLKSTA: Clock Enabling Status

0 = clock is disabled.
1 = clock is enabled.
• MTIOA: TIOA Mirror

0 = TIOA is low. If WAVE = 0, this means that TIOA pin is low. If WAVE = 1, this means that TIOA is driven low.
1 = TIOA is high. If WAVE = 0, this means that TIOA pin is high. If WAVE = 1, this means that TIOA is driven high.
• MTIOB: TIOB Mirror

0 = TIOB is low. If WAVE = 0, this means that TIOB pin is low. If WAVE = 1, this means that TIOB is driven low.
1 = TIOB is high. If WAVE = 0, this means that TIOB pin is high. If WAVE = 1, this means that TIOB is driven high.
643 AT91SAM9G45
33.7.11 TC Interrupt Enable Register
Name: TC_IERx [x=0..2]
0xFFFD4064 (1)[1], 0xFFFD40A4 (1)[2]
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
• COVFS: Counter Overflow

0 = no effect.
1 = enables the Counter Overflow Interrupt.
• LOVRS: Load Overrun

0 = no effect.
1 = enables the Load Overrun Interrupt.
• CPAS: RA Compare
0 = no effect.
1 = enables the RA Compare Interrupt.
• CPBS: RB Compare
0 = no effect.
1 = enables the RB Compare Interrupt.
• CPCS: RC Compare
0 = no effect.
1 = enables the RC Compare Interrupt.
• LDRAS: RA Loading
0 = no effect.
1 = enables the RA Load Interrupt.
• LDRBS: RB Loading
0 = no effect.
1 = enables the RB Load Interrupt.
• ETRGS: External Trigger

0 = no effect.
1 = enables the External Trigger Interrupt.
644 AT91SAM9G45
33.7.12 TC Interrupt Disable Register
Name: TC_IDRx [x=0..2]
0xFFFD4068 (1)[1], 0xFFFD40A8 (1)[2]
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0

0 = no effect.
1 = disables the Counter Overflow Interrupt.

0 = no effect.
1 = disables the Load Overrun Interrupt (if WAVE = 0).
0 = no effect.
1 = disables the RA Compare Interrupt (if WAVE = 1).
0 = no effect.
1 = disables the RB Compare Interrupt (if WAVE = 1).
0 = no effect.
1 = disables the RC Compare Interrupt.
0 = no effect.
1 = disables the RA Load Interrupt (if WAVE = 0).
0 = no effect.
1 = disables the RB Load Interrupt (if WAVE = 0).

0 = no effect.
1 = disables the External Trigger Interrupt.
645 AT91SAM9G45
33.7.13 TC Interrupt Mask Register
Name: TC_IMRx [x=0..2]
Addresses: 0xFFF7C02C (0)[0], 0xFFF7C06C (0)[1], 0xFFF7C0AC (0)[2], 0xFFFD402C (1)[0],
0xFFFD406C (1)[1], 0xFFFD40AC (1)[2]
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0

0 = the Counter Overflow Interrupt is disabled.
1 = the Counter Overflow Interrupt is enabled.

0 = the Load Overrun Interrupt is disabled.
1 = the Load Overrun Interrupt is enabled.
0 = the RA Compare Interrupt is disabled.
1 = the RA Compare Interrupt is enabled.
0 = the RB Compare Interrupt is disabled.
1 = the RB Compare Interrupt is enabled.
0 = the RC Compare Interrupt is disabled.
1 = the RC Compare Interrupt is enabled.
0 = the Load RA Interrupt is disabled.
1 = the Load RA Interrupt is enabled.
0 = the Load RB Interrupt is disabled.
1 = the Load RB Interrupt is enabled.

0 = the External Trigger Interrupt is disabled.
1 = the External Trigger Interrupt is enabled.
646 AT91SAM9G45
AT91SAM9G45
34. High Speed MultiMedia Card Interface (HSMCI)
34.1 Description
The High Speed Multimedia Card Interface (HSMCI) supports the MultiMedia Card (MMC)
Specification V4.3, the SD Memory Card Specification V2.0, the SDIO V1.1 specification and
CE-ATA V1.1.
The HSMCI includes a command register, response registers, data registers, timeout counters
and error detection logic that automatically handle the transmission of commands and, when
required, the reception of the associated responses and data with a limited processor overhead.
The HSMCI supports stream, block and multi block data read and write, and is compatible with
the DMA Controller, minimizing processor intervention for large buffers transfers.
The HSMCI operates at a rate of up to Master Clock divided by 2 and supports the interfacing of
1 slot(s). Each slot may be used to interface with a High Speed MultiMediaCard bus (up to 30
Cards) or with an SD Memory Card. Only one slot can be selected at a time (slots are multi-
plexed). A bit field in the SD Card Register performs this selection.
The SD Memory Card communication is based on a 9-pin interface (clock, command, four data
and three power lines) and the High Speed MultiMedia Card on a 7-pin interface (clock, com-
mand, one data, three power lines and one reserved for future use).
The SD Memory Card interface also supports High Speed MultiMedia Card operations. The
main differences between SD and High Speed MultiMedia Cards are the initialization process
and the bus topology.
HSMCI fully supports CE-ATA Revision 1.1, built on the MMC System Specification v4.0. The
module includes dedicated hardware to issue the command completion signal and capture the
host command completion signal disable.

• Compatibility with MultiMedia Card Specification Version 4.3
• Compatibility with SD Memory Card Specification Version 2.0
• Compatibility with SDIO Specification Version V2.0.
• Compatibility with Memory Stick PRO
• Compatibility with CE ATA
647
34.3 Block Diagram
APB Bridge
DMAC
APB
(1)
MCCK
(1)
MCCDA
HSMCI Interface MCDA0

(1)
PMC MCK
PIO
MCDA1 (1)
(1)
MCDA2
(1)
MCDA3
(1)
MCDA4
(1)
MCDA5
MCDA6 (1)
Interrupt Control (1)

MCDA7
HSMCI Interrupt
Note: 1. When several HSMCI (x HSMCI) are embedded in a product, MCCK refers to HSMCIx_CK, MCCDA to HSMCIx_CDA,
MCDAy to HSMCIx_DAy.
648 AT91SAM9G45
AT91SAM9G45
Application Layer
ex: File System, Audio, Security, etc.
Physical Layer
HSMCI Interface
1 2 3 4 5 6 7
1 2 3 4 5 6 78
9
9 1011 1213 8
MMC SDCard
34.5 Pin Name List

Pin Name(2) Pin Description Type(1) Comments
MCCDA Command/response I/O/PP/OD CMD of an MMC or SDCard/SDIO
MCCK Clock I/O CLK of an MMC or SD Card/SDIO
MCDA0 - MCDA7 Data 0..7 of Slot A I/O/PP DAT[0..7] of an MMC
DAT[0..3] of an SD Card/SDIO
Notes: 1. I: Input, O: Output, PP: Push/Pull, OD: Open Drain.
2. When several HSMCI (x HSMCI) are embedded in a product, MCCK refers to HSMCIx_CK, MCCDA to HSMCIx_CDA,
MCDAy to HSMCIx_DAy.
649
AT91SAM9G45
34.6.1 I/O Lines

The pins used for interfacing the High Speed MultiMedia Cards or SD Cards are multiplexed with
PIO lines. The programmer must first program the PIO controllers to assign the peripheral func-
tions to HSMCI pins.

HSMCI0 MCI0_CDA PA1 A
HSMCI0 MCI0_CK PA0 A
HSMCI0 MCI0_DA0 PA2 A
HSMCI1 MCI1_CDA PA22 A
HSMCI1 MCI1_CK PA31 A

The HSMCI is clocked through the Power Management Controller (PMC), so the programmer
must first configure the PMC to enable the HSMCI clock.
34.6.3 Interrupt
The HSMCI interface has an interrupt line connected to the Advanced Interrupt Controller (AIC).
Handling the HSMCI interrupt requires programming the AIC before configuring the HSMCI.

Instance ID
HSMCI0 11
HSMCI1 29
650
AT91SAM9G45
34.7 Bus Topology
Figure 34-3. High Speed MultiMedia Memory Card Bus Topology
1 2 3 4 5 6 7
9 1011 1213 8
MMC
The High Speed MultiMedia Card communication is based on a 13-pin serial bus interface. It has
three communication lines and four supply lines.
Table 34-4. Bus Topology

HSMCI Pin
Pin Name((2)
Number Name Type(1) Description (Slot z)
1 DAT[3] I/O/PP Data MCDz3
2 CMD I/O/PP/OD Command/response MCCDz
3 VSS1 S Supply voltage ground VSS
4 VDD S Supply voltage VDD
5 CLK I/O Clock MCCK
7 DAT[0] I/O/PP Data 0 MCDz0
Notes: 1. I: Input, O: Output, PP: Push/Pull, OD: Open Drain.

2. When several HSMCI (x HSMCI) are embedded in a product, MCCK refers to HSMCIx_CK,
MCCDA to HSMCIx_CDA, MCDAy to HSMCIx_DAy.
651
AT91SAM9G45
Figure 34-4. MMC Bus Connections (One Slot)
HSMCI
MCDA0
MCCDA
MCCK
1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7
9 1011 1213 8 9 1011 1213 8 9 1011 1213 8
MMC1 MMC2 MMC3
Note: When several HSMCI (x HSMCI) are embedded in a product, MCCK refers to HSMCIx_CK, MCCDA to HSMCIx_CDA MCDAy
to HSMCIx_DAy.
Figure 34-5. SD Memory Card Bus Topology
1 2 3 4 5 6 78
9
SD CARD
The SD Memory Card bus includes the signals listed in Table 34-5.
Table 34-5. SD Memory Card Bus Signals

HSMCI Pin
Pin Name(2)
Number Name Type(1) Description (Slot z)
1 CD/DAT[3] I/O/PP Card detect/ Data line Bit 3 MCDz3
2 CMD PP Command/response MCCDz
4 VDD S Supply voltage VDD
5 CLK I/O Clock MCCK
7 DAT[0] I/O/PP Data line Bit 0 MCDz0
8 DAT[1] I/O/PP Data line Bit 1 or Interrupt MCDz1
9 DAT[2] I/O/PP Data line Bit 2 MCDz2
Notes: 1. I: input, O: output, PP: Push Pull, OD: Open Drain.

2. When several HSMCI (x HSMCI) are embedded in a product, MCCK refers to HSMCIx_CK,
MCCDA to HSMCIx_CDA, MCDAy to HSMCIx_DAy.
652
AT91SAM9G45
Figure 34-6. SD Card Bus Connections with One Slot
1 2 3 4 5 6 78
MCDA0 - MCDA3
MCCK SD CARD
MCCDA
9
Note: When several HSMCI (x HSMCI) are embedded in a product, MCCK refers to HSMCIx_CK, MCCDA to HSMCIx_CDA MCDAy
to HSMCIx_DAy.
When the HSMCI is configured to operate with SD memory cards, the width of the data bus can
be selected in the HSMCI_SDCR register. Clearing the SDCBUS bit in this register means that
the width is one bit; setting it means that the width is four bits. In the case of High Speed Multi-
Media cards, only the data line 0 is used. The other data lines can be used as independent
PIOs.
34.8 High Speed MultiMedia Card Operations

After a power-on reset, the cards are initialized by a special message-based High Speed Multi-
Media Card bus protocol. Each message is represented by one of the following tokens:
• Command: A command is a token that starts an operation. A command is sent from the host
either to a single card (addressed command) or to all connected cards (broadcast
command). A command is transferred serially on the CMD line.
• Response: A response is a token which is sent from an addressed card or (synchronously)
from all connected cards to the host as an answer to a previously received command. A
response is transferred serially on the CMD line.
• Data: Data can be transferred from the card to the host or vice versa. Data is transferred via
the data line.
Card addressing is implemented using a session address assigned during the initialization
phase by the bus controller to all currently connected cards. Their unique CID number identifies
individual cards.
The structure of commands, responses and data blocks is described in the High Speed MultiMe-
dia-Card System Specification. See also Table 34-6 on page 654.
High Speed MultiMediaCard bus data transfers are composed of these tokens.
There are different types of operations. Addressed operations always contain a command and a
response token. In addition, some operations have a data token; the others transfer their infor-
mation directly within the command or response structure. In this case, no data token is present
in an operation. The bits on the DAT and the CMD lines are transferred synchronous to the clock
HSMCI Clock.
Two types of data transfer commands are defined:
• Sequential commands: These commands initiate a continuous data stream. They are
terminated only when a stop command follows on the CMD line. This mode reduces the
command overhead to an absolute minimum.
653
AT91SAM9G45
• Block-oriented commands: These commands send a data block succeeded by CRC bits.
Both read and write operations allow either single or multiple block transmission. A multiple
block transmission is terminated when a stop command follows on the CMD line similarly to the
sequential read or when a multiple block transmission has a pre-defined block count (See “Data
Transfer Operation” on page 656.).
The HSMCI provides a set of registers to perform the entire range of High Speed MultiMedia
Card operations.
34.8.1 Command - Response Operation

After reset, the HSMCI is disabled and becomes valid after setting the MCIEN bit in the
HSMCI_CR Control Register.
The PWSEN bit saves power by dividing the HSMCI clock by 2PWSDIV + 1 when the bus is
inactive.
The two bits, RDPROOF and WRPROOF in the HSMCI Mode Register (HSMCI_MR) allow
stopping the HSMCI Clock during read or write access if the internal FIFO is full. This will guar-
antee data integrity, not bandwidth.
All the timings for High Speed MultiMedia Card are defined in the High Speed MultiMediaCard
System Specification.
The two bus modes (open drain and push/pull) needed to process all the operations are defined
in the HSMCI command register. The HSMCI_CMDR allows a command to be carried out.
For example, to perform an ALL_SEND_CID command:
Host Command NID Cycles CID

CMD S T Content CRC E Z ****** Z S T Content Z Z Z
The command ALL_SEND_CID and the fields and values for the HSMCI_CMDR Control Regis-
ter are described in Table 34-6 and Table 34-7.
Table 34-6. ALL_SEND_CID Command Description

Command
CMD Index Type Argument Resp Abbreviation Description
Asks all cards to send
CMD2 bcr [31:0] stuff bits R2 ALL_SEND_CID their CID numbers on
the CMD line
Note: bcr means broadcast command with response.
Table 34-7. Fields and Values for HSMCI_CMDR Command Register

Field Value
CMDNB (command number) 2 (CMD2)
RSPTYP (response type) 2 (R2: 136 bits response)
SPCMD (special command) 0 (not a special command)
OPCMD (open drain command) 1
MAXLAT (max latency for command to response) 0 (NID cycles ==> 5 cycles)
654
AT91SAM9G45
Table 34-7. Fields and Values for HSMCI_CMDR Command Register

Field Value
TRCMD (transfer command) 0 (No transfer)
TRDIR (transfer direction) X (available only in transfer command)
TRTYP (transfer type) X (available only in transfer command)
IOSPCMD (SDIO special command) 0 (not a special command)
The HSMCI_ARGR contains the argument field of the command.

To send a command, the user must perform the following steps:
• Fill the argument register (HSMCI_ARGR) with the command argument.
• Set the command register (HSMCI_CMDR) (see Table 34-7).
The command is sent immediately after writing the command register.
As soon as the command register is written, then the status bit CMDRDY in the status register
(HSMCI_SR) is cleared.
It is released and the end of the card response.
If the command requires a response, it can be read in the HSMCI response register
(HSMCI_RSPR). The response size can be from 48 bits up to 136 bits depending on the com-
mand. The HSMCI embeds an error detection to prevent any corrupted data during the transfer.
The following flowchart shows how to send a command to the card and read the response if
needed. In this example, the status register bits are polled but setting the appropriate bits in the
interrupt enable register (HSMCI_IER) allows using an interrupt method.
655
AT91SAM9G45
Figure 34-7. Command/Response Functional Flow Diagram
Set the command argument

HSMCI_ARGR = Argument(1)
Set the command

HSMCI_CMDR = Command
Read HSMCI_SR
Wait for command 0

ready status flag CMDRDY
Check error bits in the Yes

status register (1) Status error flags?
Read response if required
RETURN ERROR(1)
RETURN OK
Note: 1. If the command is SEND_OP_COND, the CRC error flag is always present (refer to R3 response in the High Speed MultiMe-
dia Card specification).
34.8.2 Data Transfer Operation

The High Speed MultiMedia Card allows several read/write operations (single block, multiple
blocks, stream, etc.). These kinds of transfer can be selected setting the Transfer Type (TRTYP)
field in the HSMCI Command Register (HSMCI_CMDR).
These operations can be done using the features of the DMA Controller.
In all cases, the block length (BLKLEN field) must be defined either in the mode register
HSMCI_MR, or in the Block Register HSMCI_BLKR. This field determines the size of the data
block.
Consequent to MMC Specification 3.1, two types of multiple block read (or write) transactions
are defined (the host can use either one at any time):
• Open-ended/Infinite Multiple block read (or write):
656
AT91SAM9G45
The number of blocks for the read (or write) multiple block operation is not defined. The card
will continuously transfer (or program) data blocks until a stop transmission command is
received.
• Multiple block read (or write) with pre-defined block count (since version 3.1 and higher):
The card will transfer (or program) the requested number of data blocks and terminate the
transaction. The stop command is not required at the end of this type of multiple block read
(or write), unless terminated with an error. In order to start a multiple block read (or write)
with pre-defined block count, the host must correctly program the HSMCI Block Register
(HSMCI_BLKR). Otherwise the card will start an open-ended multiple block read. The BCNT
field of the Block Register defines the number of blocks to transfer (from 1 to 65535 blocks).
Programming the value 0 in the BCNT field corresponds to an infinite block transfer.
34.8.3 Read Operation

The following flowchart (Figure 34-8) shows how to read a single block with or without use of
DMAC facilities. In this example, a polling method is used to wait for the end of read. Similarly,
the user can configure the interrupt enable register (HSMCI_IER) to trigger an interrupt at the
end of read.
657
AT91SAM9G45
Figure 34-8. Read Functional Flow Diagram

Send SELECT/DESELECT_CARD
(1)
command to select the card
Send SET_BLOCKLEN command(1)
No Yes
Read with DMAC
Reset the DMAEN bit

MCI_DMA &= ~DMAEN Set the DMAEN bit
Set the block length (in bytes) HSMCI_DMA |= DMAEN
HSMCI_MR l= (BlockLength<<16) (2) Set the block length (in bytes)
(2)
Set the block count (if neccessary) HSMCI_BLKR |= (BlockLength << 16)
HSMCI_BLKR l= (BlockCount<<0)
Send READ_SINGLE_BLOCK Configure the DMA channel X

command(1) DMAC_SADDRX = Data Address
DMAC_BTSIZE = BlockLength/4
DMACHEN[X] = TRUE
Number of words to read = BlockLength/4

Send READ_SINGLE_BLOCK
command(1)
Yes
Number of words to read = 0 ?
Read status register HSMCI_SR
No

Poll the bit Yes
XFRDONE = 0?
Poll the bit Yes

RXRDY = 0? No
No
RETURN
Read data = HSMCI_RDR
Number of words to read =

Number of words to read -1
RETURN
Note: 1. It is assumed that this command has been correctly sent (see Figure 34-7).
2. This field is also accessible in the HSMCI Block Register (HSMCI_BLKR).
658
AT91SAM9G45
34.8.4 Write Operation

In write operation, the HSMCI Mode Register (HSMCI_MR) is used to define the padding value
when writing non-multiple block size. If the bit PADV is 0, then 0x00 value is used when padding
data, otherwise 0xFF is used.
If set, the bit DMAEN in the HSMCI_DMA register enables DMA transfer.
The following flowchart (Figure 34-9) shows how to write a single block with or without use of
DMA facilities. Polling or interrupt method can be used to wait for the end of write according to
the contents of the Interrupt Mask Register (HSMCI_IMR).
659
AT91SAM9G45
Figure 34-9. Write Functional Flow Diagram
command(1) to select the card
No Yes
Write using DMAC
Reset theDMAEN bit

HSMCI_DMA &= ~DMAEN Set the DMAEN bit
Set the block length (in bytes) HSMCI_DMA |= DMAEN
HSMCI_MR |= (BlockLength) <<16)(2) Set the block length (in bytes)
Set the block count (if necessary) HSMCI_BLKR |= (BlockLength << 16)(2)
HSMCI_BLKR |= (BlockCount << 0)
Send WRITE_SINGLE_BLOCK Send WRITE_SINGLE_BLOCK

command(1) command(1)
Configure the DMA channel X

Number of words to write = BlockLength/4 DMAC_DADDRX = Data Address to write
DMAC_CHEN[X] = TRUE
Yes
Number of words to write = 0 ?
Read status register MCI_SR

No
Poll the bit Yes

XFRDONE = 0?
Poll the bit Yes

TXRDY = 0?
No
No
HSMCI_TDR = Data to write RETURN
Number of words to write =

Number of words to write -1
RETURN
Note: 1. It is assumed that this command has been correctly sent (see Figure 34-7).
2. This field is also accessible in the HSMCI Block Register (HSMCI_BLKR).
660
AT91SAM9G45
The following flowchart (Figure 34-10) shows how to manage read multiple block and write mul-
tiple block transfers with the DMA Controller. Polling or interrupt method can be used to wait for
the end of write according to the contents of the Interrupt Mask Register (HSMCI_IMR).
Figure 34-10. Read Multiple Block and Write Multiple Block
command(1) to select the card
Set the block length

HSMCI_MR |= (BlockLength << 16)
Set the DMAEN bit
HSMCI_DMA |= DMAEN
Send WRITE_MULTIPLE_BLOCK or
READ_MULTIPLE_BLOCK command(1)
Configure the HDMA channel X

DMAC_SADDRX and DMAC_DADDRX
DMAC_CHEN[X] = TRUE
Read status register DMAC_EBCISR

and Poll Bit CBTC[X]
Yes
New Buffer ?(2)
No
and Poll Bit FIFOEMPTY
Send STOP_TRANSMISSION
command(1)
Poll the bit No

XFRDONE = 1
Yes
RETURN
Notes: 1. It is assumed that this command has been correctly sent (see Figure 34-7).
2. Handle errors reported in HSMCI_SR.
661
AT91SAM9G45
34.8.5 WRITE_SINGLE_BLOCK Operation using DMA Controller

1. Wait until the current command execution has successfully terminated.
a. Check that CMDRDY and NOTBUSY fields are asserted in HSMCI_SR
2. Program the block length in the card. This value defines the value block_length.
3. Program the block length in the HSMCI configuration register with block_length value.
4. Program HSMCI_DMA register with the following fields:
– OFFSET field with dma_offset.
– CHKSIZE is user defined and set according to DMAC_DCSIZE.
– DMAEN is set to true to enable DMA hardware handshaking in the HSMCI. This bit
was previously set to false.
5. Issue a WRITE_SINGLE_BLOCK command writing HSMCI_ARG then HSMCI_CMDR.
6. Program the DMA Controller.
a. Read the channel Register to choose an available (disabled) channel.
b. Clear any pending interrupts on the channel from the previous DMAC transfer by
reading the DMAC_EBCISR register.
c. Program the channel registers.
d. The DMAC_SADDRx register for channel x must be set to the location of the
source data. When the first data location is not word aligned, the two LSB bits
define the temporary value called dma_offset. The two LSB bits of
DMAC_SADDRx must be set to 0.
e. The DMAC_DADDRx register for channel x must be set with the starting address of
the HSMCI_FIFO address.
f. Program DMAC_CTRLAx register of channel x with the following field’s values:
–DST_WIDTH is set to WORD.
–SRC_WIDTH is set to WORD.
–DCSIZE must be set according to the value of HSMCI_DMA, CHKSIZE field.
–BTSIZE is programmed with CEILING((block_length + dma_offset) / 4), where
the ceiling function is the function that returns the smallest integer not less than
x.
g. Program DMAC_CTRLBx register for channel x with the following field’s values:
–DST_INCR is set to INCR, the block_length value must not be larger than the
HSMCI_FIFO aperture.
–SRC_INCR is set to INCR.
–FC field is programmed with memory to peripheral flow control mode.
–both DST_DSCR and SRC_DSCR are set to 1 (descriptor fetch is disabled).
–DIF and SIF are set with their respective layer ID. If SIF is different from DIF, the
DMA controller is able to prefetch data and write HSMCI simultaneously.
h. Program DMAC_CFGx register for channel x with the following field’s values:
–FIFOCFG defines the watermark of the DMAC channel FIFO.
–DST_H2SEL is set to true to enable hardware handshaking on the destination.
–DST_PER is programmed with the hardware handshaking ID of the targeted
HSMCI Host Controller.
662
AT91SAM9G45
i. Enable Channel x, writing one to DMAC_CHER[x]. The DMAC is ready and waiting
for request.
7. Wait for XFRDONE in HSMCI_SR register.
34.8.6 READ_SINGLE_BLOCK Operation using DMA Controller
34.8.6.1 Block Length is Multiple of 4

1. Wait until the current command execution has successfully completed.
a. Check that CMDRDY and NOTBUSY are asserted in HSMCI_SR.
4. Set RDPROOF bit in HSMCI_MR to avoid overflow.
– ROPT field is set to 0.
– OFFSET field is set to 0.
– CHKSIZE is user defined.
– DMAEN is set to true to enable DMAC hardware handshaking in the HSMCI. This bit
6. Issue a READ_SINGLE_BLOCK command.
7. Program the DMA controller.
b. Clear any pending interrupts on the channel from the previous DMA transfer by
d. The DMAC_SADDRx register for channel x must be set with the starting address of
e. The DMAC_DADDRx register for channel x must be word aligned.
–SCSIZE must be set according to the value of HSMCI_DMA, CHKSIZE field.
–BTSIZE is programmed with block_length/4.
– DST_INCR is set to INCR.
– SRC_INCR is set to INCR.
– FC field is programmed with peripheral to memory flow control mode.
– both DST_DSCR and SRC_DSCR are set to 1 (descriptor fetch is disabled).
– DIF and SIF are set with their respective layer ID. If SIF is different from DIF, the
DMA controller is able to prefetch data and write HSMCI simultaneously.
–FIFOCFG defines the watermark of the DMA channel FIFO.
–SRC_H2SEL is set to true to enable hardware handshaking on the destination.
–SRC_PER is programmed with the hardware handshaking ID of the targeted
663
AT91SAM9G45
–Enable Channel x, writing one to DMAC_CHER[x]. The DMAC is ready and

waiting for request.
34.8.6.2 Block Length is Not Multiple of 4 and Padding Not Used (ROPT field in HSMCI_DMA register set to 0)
In the previous DMA transfer flow (block length multiple of 4), the DMA controller is configured to
use only WORD AHB access. When the block length is no longer a multiple of 4 this is no longer
true. The DMA controller is programmed to copy exactly the block length number of bytes using
2 transfer descriptors.
1. Use the previous step until READ_SINGLE_BLOCK then
2. Program the DMA controller to use a two descriptors linked list.
c. Program the channel registers in the Memory for the first descriptor. This descriptor
will be word oriented. This descriptor is referred to as LLI_W, standing for LLI word
oriented transfer.
d. The LLI_W.DMAC_SADDRx field in memory must be set with the starting address
of the HSMCI_FIFO address.
e. The LLI_W.DMAC_DADDRx field in the memory must be word aligned.
f. Program LLI_W.DMAC_CTRLAx with the following field’s values:
–BTSIZE is programmed with block_length/4. If BTSIZE is zero, this descriptor is
skipped later.
g. Program LLI_W.DMAC_CTRLBx with the following field’s values:
–DST_INCR is set to INCR
–SRC_INCR is set to INCR
–FC field is programmed with peripheral to memory flow control mode.
–SRC_DSCR is set to zero. (descriptor fetch is enabled for the SRC)
–DST_DSCR is set to one. (descriptor fetch is disabled for the DST)
–DIF and SIF are set with their respective layer ID. If SIF is different from DIF, DMA
controller is able to prefetch data and write HSMCI simultaneously.
h. Program LLI_W.DMAC_CFGx register for channel x with the following field’s
values:
–DST_REP is set to zero meaning that address are contiguous.
i. Program LLI_W.DMAC_DSCRx with the address of LLI_B descriptor. And set
DSCRx_IF to the AHB Layer ID. This operation actually links the Word oriented
664
AT91SAM9G45
descriptor on the second byte oriented descriptor. When block_length[1:0] is equal

to 0 (multiple of 4) LLI_W.DMAC_DSCRx points to 0, only LLI_W is relevant.
j. Program the channel registers in the Memory for the second descriptor. This
descriptor will be byte oriented. This descriptor is referred to as LLI_B, standing for
LLI Byte oriented.
k. The LLI_B.DMAC_SADDRx field in memory must be set with the starting address
of the HSMCI_FIFO address.
l. The LLI_B.DMAC_DADDRx is not relevant if previous word aligned descriptor was
enabled. If 1, 2 or 3 bytes are transferred that address is user defined and not word
aligned.
m. Program LLI_B.DMAC_CTRLAx with the following field’s values:
–DST_WIDTH is set to BYTE.
–SRC_WIDTH is set to BYTE.
–BTSIZE is programmed with block_length[1:0]. (last 1, 2, or 3 bytes of the buffer).
n. Program LLI_B.DMAC_CTRLBx with the following field’s values:
–Both SRC_DSCR and DST_DSCR are set to 1 (descriptor fetch is disabled) or
Next descriptor location points to 0.
Controller is able to prefetch data and write HSMCI simultaneously.
o. Program LLI_B.DMAC_CFGx memory location for channel x with the following
field’s values:
– FIFOCFG defines the watermark of the DMA channel FIFO.
– SRC_H2SEL is set to true to enable hardware handshaking on the destination.
– SRC_PER is programmed with the hardware handshaking ID of the targeted HSMCI
Host Controller.
p. Program LLI_B.DMAC_DSCR with 0.
q. Program DMAC_CTRLBx register for channel x with 0. its content is updated with
the LLI fetch operation.
r. Program DMAC_DSCRx with the address of LLI_W if block_length greater than 4
else with address of LLI_B.
s. Enable Channel x writing one to DMAC_CHER[x]. The DMAC is ready and waiting
for request.
34.8.6.3 Block Length is Not Multiple of 4, with Padding Value (ROPT field in HSMCI_DMA register set to 1)
When the ROPT field is set to one, The DMA Controller performs only WORD access on the bus
to transfer a non-multiple of 4 block length. Unlike previous flow, iin which the transfer size is
rounded to the nearest multiple of 4.
1. Program the HSMCI Interface, see previous flow.
2. Program the DMA Controller
665
AT91SAM9G45

d. The DMAC_SADDRx register for channel x must be set with the starting address of
e. The DMAC_DADDRx register for channel x must be word aligned.
–DST_WIDTH is set to WORD
–SRC_WIDTH is set to WORD
–SCSIZE must be set according to the value of HSMCI_DMA.CHKSIZE Field.
–BTSIZE is programmed with CEILING(block_length/4).
–both DST_DSCR and SRC_DSCR are set to 1. (descriptor fetch is disabled)
DMA Controller is able to prefetch data and write HSMCI simultaneously.
–Enable Channel x writing one to DMAC_CHER[x]. The DMAC is ready and
waiting for request.
34.8.7 WRITE_MULTIPLE_BLOCK
34.8.7.1 One Block per Descriptor

– OFFSET field with dma_offset.
5. Issue a WRITE_MULTIPLE_BLOCK command.
6. Program the DMA Controller to use a list of descriptors. Each descriptor transfers one
block of data. Block n of data is transferred with descriptor LLI(n).
666
AT91SAM9G45

c. Program a List of descriptors.
d. The LLI(n).DMAC_SADDRx memory location for channel x must be set to the loca-
tion of the source data. When the first data location is not word aligned, the two
LSB bits define the temporary value called dma_offset. The two LSB bits of
LLI(n).DMAC_SADDRx must be set to 0.
e. The LLI(n).DMAC_DADDRx register for channel x must be set with the starting
address of the HSMCI_FIFO address.
f. Program LLI(n).DMAC_CTRLAx register of channel x with the following field’s
values:
–DCSIZE must be set according to the value of HSMCI_DMA, CHKSIZE field.
–BTSIZE is programmed with CEILING((block_length + dma_offset)/4).
g. Program LLI(n).DMAC_CTRLBx register for channel x with the following field’s
values:
–DST_INCR is set to INCR.
–DST_DSCR is set to 0 (fetch operation is enabled for the destination).
–SRC_DSCR is set to 1 (source address is contiguous).
–FC field is programmed with memory to peripheral flow control mode.
–Both DST_DSCR and SRC_DSCR are set to 1 (descriptor fetch is disabled).
Controller is able to prefetch data and write HSMCI simultaneously.
h. Program LLI(n).DMAC_CFGx register for channel x with the following field’s values:
–DST_H2SEL is set to true to enable hardware handshaking on the destination.
–SRC_REP is set to 0. (contiguous memory access at block boundary)
–DST_PER is programmed with the hardware handshaking ID of the targeted
i. If LLI(n) is the last descriptor, then LLI(n).DSCR points to 0 else LLI(n) points to the
start address of LLI(n+1).
j. Program DMAC_CTRLBx for channel register x with 0. Its content is updated with
the LLI fetch operation.
k. Program DMAC_DSCRx for channel register x with the address of the first descrip-
tor LLI(0).
l. Enable Channel x writing one to DMAC_CHER[x]. The DMA is ready and waiting
for request.
7. Poll CBTC[x] bit in the DMAC_EBCISR Register.
8. If a new list of buffers shall be transferred, repeat step 6. Check and handle HSMCI
errors.
9. Poll FIFOEMPTY field in the HSMCI_SR.
667
AT91SAM9G45
10. Send The STOP_TRANSMISSION command writing HSMCI_ARG then

HSMCI_CMDR.
34.8.8 READ_MULTIPLE_BLOCK
34.8.8.1 Block Length is a Multiple of 4

4. Set RDPROOF bit in HSMCI_MR to avoid overflow.
– OFFSET field is set to 0.
6. Issue a READ_MULTIPLE_BLOCK command.
7. Program the DMA Controller to use a list of descriptors:
c. Program the channel registers in the Memory with the first descriptor. This descrip-
tor will be word oriented. This descriptor is referred to as LLI_W(n), standing for LLI
word oriented transfer for block n.
d. The LLI_W(n).DMAC_SADDRx field in memory must be set with the starting
e. The LLI_W(n).DMAC_DADDRx field in the memory must be word aligned.
f. Program LLI_W(n).DMAC_CTRLAx with the following field’s values:
–DST_WIDTH is set to WORD
–SRC_WIDTH is set to WORD
–BTSIZE is programmed with block_length/4.
g. Program LLI_W(n).DMAC_CTRLBx with the following field’s values:
–SRC_DSCR is set to 0 (descriptor fetch is enabled for the SRC).
–DST_DSCR is set to TRUE (descriptor fetch is disabled for the DST).
668
AT91SAM9G45
h. Program LLI_W(n).DMAC_CFGx register for channel x with the following field’s

values:
–DST_REP is set to zero. Addresses are contiguous.
i. Program LLI_W(n).DMAC_DSCRx with the address of LLI_W(n+1) descriptor. And
set the DSCRx_IF to the AHB Layer ID. This operation actually links descriptors
together. If LLI_W(n) is the last descriptor then LLI_W(n).DMAC_DSCRx points to
0.
j. Program DMAC_CTRLBx register for channel x with 0. its content is updated with
the LLI Fetch operation.
k. Program DMAC_DSCRx register for channel x with the address of LLI_W(0).
l. Enable Channel x writing one to DMAC_CHER[x]. The DMA is ready and waiting
for request.
9. If a new list of buffer shall be transferred repeat step 6. Check and handle HSMCI
errors.
11. Send The STOP_TRANSMISSION command writing the HSMCI_ARG then the
HSMCI_CMDR.
34.8.8.2 Block Length is Not Multiple of 4. (ROPT field in HSMCI_DMA register set to 0)
Two DMA Transfer descriptors are used to perform the HSMCI block transfer.
1. Use the previous step to configure the HSMCI to perform a READ_MULTIPLE_BLOCK
command.
3. Program the DMA Controller to use a list of descriptors.
a. Read the channel register to choose an available (disabled) channel.
c. For every block of data repeat the following procedure:
d. Program the channel registers in the Memory for the first descriptor. This descriptor
will be word oriented. This descriptor is referred to as LLI_W(n) standing for LLI
e. The LLI_W(n).DMAC_SADDRx field in memory must be set with the starting
f. The LLI_W(n).DMAC_DADDRx field in the memory must be word aligned.
g. Program LLI_W(n).DMAC_CTRLAx with the following field’s values:
–BTSIZE is programmed with block_length/4. If BTSIZE is zero, this descriptor is
skipped later.
669
AT91SAM9G45
h. Program LLI_W(n).DMAC_CTRLBx with the following field’s values:

–SRC_DSCR is set to 0 (descriptor fetch is enabled for the SRC).
–DST_DSCR is set to TRUE (descriptor fetch is disabled for the DST).
i. Program LLI_W(n).DMAC_CFGx register for channel x with the following field’s
values:
–DST_REP is set to zero. Address are contiguous.
j. Program LLI_W(n).DMAC_DSCRx with the address of LLI_B(n) descriptor. And set
the DSCRx_IF to the AHB Layer ID. This operation actually links the Word oriented
descriptor on the second byte oriented descriptor. When block_length[1:0] is equal
to 0 (multiple of 4) LLI_W(n).DMAC_DSCRx points to 0, only LLI_W(n) is relevant.
k. Program the channel registers in the Memory for the second descriptor. This
descriptor will be byte oriented. This descriptor is referred to as LLI_B(n), standing
for LLI Byte oriented.
l. The LLI_B(n).DMAC_SADDRx field in memory must be set with the starting
m. The LLI_B(n).DMAC_DADDRx is not relevant if previous word aligned descriptor
was enabled. If 1, 2 or 3 bytes are transferred, that address is user defined and not
word aligned.
n. Program LLI_B(n).DMAC_CTRLAx with the following field’s values:
–DST_WIDTH is set to BYTE.
–SRC_WIDTH is set to BYTE.
–BTSIZE is programmed with block_length[1:0]. (last 1, 2, or 3 bytes of the buffer).
o. Program LLI_B(n).DMAC_CTRLBx with the following field’s values:
– DST_INCR is set to INCR.
– SRC_INCR is set to INCR.
– FC field is programmed with peripheral to memory flow control mode.
– Both SRC_DSCR and DST_DSCR are set to 1 (descriptor fetch is disabled) or Next
descriptor location points to 0.
– DIF and SIF are set with their respective layer ID. If SIF is different from DIF, the
p. Program LLI_B(n).DMAC_CFGx memory location for channel x with the following
field’s values:
– FIFOCFG defines the watermark of the DMAC channel FIFO.
– SRC_H2SEL is set to true to enable hardware handshaking on the destination.
670
AT91SAM9G45
– SRC_PER is programmed with the hardware handshaking ID of the targeted HSMCI

Host Controller
q. Program LLI_B(n).DMAC_DSCR with address of descriptor LLI_W(n+1). If
LLI_B(n) is the last descriptor, then program LLI_B(n).DMAC_DSCR with 0.
r. Program DMAC_CTRLBx register for channel x with 0, its content is updated with
s. Program DMAC_DSCRx with the address of LLI_W(0) if block_length is greater
than 4 else with address of LLI_B(0).
t. Enable Channel x writing one to DMAC_CHER[x]. The DMAC is ready and waiting
for request.
4. Enable DMADONE interrupt in the HSMCI_IER register.
6. If a new list of buffers shall be transferred, repeat step 7. Check and handle HSMCI
errors.
8. Send The STOP_TRANSMISSION command writing HSMCI_ARG then
HSMCI_CMDR.
34.8.8.3 Block Length is Not a Multiple of 4. (ROPT field in HSMCI_DMA register set to 1)
One DMA Transfer descriptor is used to perform the HSMCI block transfer, the DMA writes a
rounded up value to the nearest multiple of 4.
1. Use the previous step to configure the HSMCI to perform a READ_MULTIPLE_BLOCK.
2. Set the ROPT field to 1 in the HSMCI_DMA register.
4. Program the DMA controller to use a list of descriptors:
c. Program the channel registers in the Memory with the first descriptor. This descrip-
tor will be word oriented. This descriptor is referred to as LLI_W(n), standing for LLI
d. The LLI_W(n).DMAC_SADDRx field in memory must be set with the starting
e. The LLI_W(n).DMAC_DADDRx field in the memory must be word aligned.
f. Program LLI_W(n).DMAC_CTRLAx with the following field’s values:
–BTSIZE is programmed with Ceiling(block_length/4).
g. Program LLI_W(n).DMAC_CTRLBx with the following field’s values:
–SRC_DSCR is set to 0. (descriptor fetch is enabled for the SRC)
671
AT91SAM9G45
–DST_DSCR is set to TRUE. (descriptor fetch is disabled for the DST)

h. Program LLI_W(n).DMAC_CFGx register for channel x with the following field’s
values:
–DST_REP is set to zero. Address are contiguous.
i. Program LLI_W(n).DMAC_DSCRx with the address of LLI_W(n+1) descriptor. And
set the DSCRx_IF to the AHB Layer ID. This operation actually links descriptors
together. If LLI_W(n) is the last descriptor then LLI_W(n).DMAC_DSCRx points to
0.
j. Program DMAC_CTRLBx register for channel x with 0. its content is updated with
k. Program DMAC_DSCRx register for channel x with the address of LLI_W(0).
l. Enable Channel x writing one to DMAC_CHER[x]. The DMAC is ready and waiting
for request.
6. If a new list of buffers shall be transferred repeat step 7. Check and handle HSMCI
errors.
8. Send The STOP_TRANSMISSION command writing the HSMCI_ARG then the
HSMCI_CMDR.
34.9 SD/SDIO Card Operation

The High Speed MultiMedia Card Interface allows processing of SD Memory (Secure Digital
Memory Card) and SDIO (SD Input Output) Card commands.
SD/SDIO cards are based on the Multi Media Card (MMC) format, but are physically slightly
thicker and feature higher data transfer rates, a lock switch on the side to prevent accidental
overwriting and security features. The physical form factor, pin assignment and data transfer
protocol are forward-compatible with the High Speed MultiMedia Card with some additions. SD
slots can actually be used for more than flash memory cards. Devices that support SDIO can
use small devices designed for the SD form factor, such as GPS receivers, Wi-Fi or Bluetooth
adapters, modems, barcode readers, IrDA adapters, FM radio tuners, RFID readers, digital cam-
eras and more.
SD/SDIO is covered by numerous patents and trademarks, and licensing is only available
through the Secure Digital Card Association.
The SD/SDIO Card communication is based on a 9-pin interface (Clock, Command, 4 x Data
and 3 x Power lines). The communication protocol is defined as a part of this specification. The
main difference between the SD/SDIO Card and the High Speed MultiMedia Card is the initial-
ization process.
672
AT91SAM9G45
The SD/SDIO Card Register (HSMCI_SDCR) allows selection of the Card Slot and the data bus
width.
The SD/SDIO Card bus allows dynamic configuration of the number of data lines. After power
up, by default, the SD/SDIO Card uses only DAT0 for data transfer. After initialization, the host
can change the bus width (number of active data lines).
34.9.1 SDIO Data Transfer Type

SDIO cards may transfer data in either a multi-byte (1 to 512 bytes) or an optional block format
(1 to 511 blocks), while the SD memory cards are fixed in the block transfer mode. The TRTYP
field in the HSMCI Command Register (HSMCI_CMDR) allows to choose between SDIO Byte or
SDIO Block transfer.
The number of bytes/blocks to transfer is set through the BCNT field in the HSMCI Block Regis-
ter (HSMCI_BLKR). In SDIO Block mode, the field BLKLEN must be set to the data block size
while this field is not used in SDIO Byte mode.
An SDIO Card can have multiple I/O or combined I/O and memory (called Combo Card). Within
a multi-function SDIO or a Combo card, there are multiple devices (I/O and memory) that share
access to the SD bus. In order to allow the sharing of access to the host among multiple devices,
SDIO and combo cards can implement the optional concept of suspend/resume (Refer to the
SDIO Specification for more details). To send a suspend or a resume command, the host must
set the SDIO Special Command field (IOSPCMD) in the HSMCI Command Register.
34.9.2 SDIO Interrupts

Each function within an SDIO or Combo card may implement interrupts (Refer to the SDIO
Specification for more details). In order to allow the SDIO card to interrupt the host, an interrupt
function is added to a pin on the DAT[1] line to signal the card’s interrupt to the host. An SDIO
interrupt on each slot can be enabled through the HSMCI Interrupt Enable Register. The SDIO
interrupt is sampled regardless of the currently selected slot.
34.10 CE-ATA Operation

CE-ATA maps the streamlined ATA command set onto the MMC interface. The ATA task file is
mapped onto MMC register space.
CE-ATA utilizes five MMC commands:
• GO_IDLE_STATE (CMD0): used for hard reset.
• STOP_TRANSMISSION (CMD12): causes the ATA command currently executing to be
aborted.
• FAST_IO (CMD39): Used for single register access to the ATA taskfile registers, 8 bit access
only.
• RW_MULTIPLE_REGISTERS (CMD60): used to issue an ATA command or to access the
control/status registers.
• RW_MULTIPLE_BLOCK (CMD61): used to transfer data for an ATA command.
CE-ATA utilizes the same MMC command sequences for initialization as traditional MMC
devices.
34.10.1 Executing an ATA Polling Command

1. Issue READ_DMA_EXT with RW_MULTIPLE_REGISTER (CMD60) for 8kB of DATA.
2. Read the ATA status register until DRQ is set.
673
AT91SAM9G45
3. Issue RW_MULTIPLE_BLOCK (CMD61) to transfer DATA.

4. Read the ATA status register until DRQ && BSY are set to 0.
34.10.2 Executing an ATA Interrupt Command

1. Issue READ_DMA_EXT with RW_MULTIPLE_REGISTER (CMD60) for 8kB of DATA
with nIEN field set to zero to enable the command completion signal in the device.
2. Issue RW_MULTIPLE_BLOCK (CMD61) to transfer DATA.
3. Wait for Completion Signal Received Interrupt.
34.10.3 Aborting an ATA Command

If the host needs to abort an ATA command prior to the completion signal it must send a special
command to avoid potential collision on the command line. The SPCMD field of the
HSMCI_CMDR must be set to 3 to issue the CE-ATA completion Signal Disable Command.
34.10.4 CE-ATA Error Recovery

Several methods of ATA command failure may occur, including:
• No response to an MMC command, such as RW_MULTIPLE_REGISTER (CMD60).
• CRC is invalid for an MMC command or response.
• CRC16 is invalid for an MMC data packet.
• ATA Status register reflects an error by setting the ERR bit to one.
• The command completion signal does not arrive within a host specified time out period.
Error conditions are expected to happen infrequently. Thus, a robust error recovery mechanism
may be used for each error event. The recommended error recovery procedure after a timeout
is:
• Issue the command completion signal disable if nIEN was cleared to zero and the
RW_MULTIPLE_BLOCK (CMD61) response has been received.
• Issue STOP_TRANSMISSION (CMD12) and successfully receive the R1 response.
• Issue a software reset to the CE-ATA device using FAST_IO (CMD39).
If STOP_TRANMISSION (CMD12) is successful, then the device is again ready for ATA com-
mands. However, if the error recovery procedure does not work as expected or there is another
timeout, the next step is to issue GO_IDLE_STATE (CMD0) to the device. GO_IDLE_STATE
(CMD0) is a hard reset to the device and completely resets all device states.
Note that after issuing GO_IDLE_STATE (CMD0), all device initialization needs to be completed
again. If the CE-ATA device completes all MMC commands correctly but fails the ATA command
with the ERR bit set in the ATA Status register, no error recovery action is required. The ATA
command itself failed implying that the device could not complete the action requested, how-
ever, there was no communication or protocol failure. After the device signals an error by setting
the ERR bit to one in the ATA Status register, the host may attempt to retry the command.
34.11 HSMCI Boot Operation Mode

In boot operation mode, the processor can read boot data from the slave (MMC device) by keep-
ing the CMD line low after power-on before issuing CMD1. The data can be read from either the
boot area or user area, depending on register setting.
674
AT91SAM9G45
34.11.1 Boot Procedure, Processor Mode

1. Configure the HSMCI data bus width programming SDCBUS Field in the
HSMCI_SDCR register. The BOOT_BUS_WIDTH field located in the device Extended
CSD register must be set accordingly.
2. Set the byte count to 512 bytes and the block count to the desired number of blocks,
writing BLKLEN and BCNT fields of the HSMCI_BLKR Register.
3. Issue the Boot Operation Request command by writing to the HSMCI_CMDR register
with SPCMD field set to BOOTREQ, TRDIR set to READ and TRCMD set to “start data
transfer”.
4. The BOOT_ACK field located in the HSMCI_CMDR register must be set to one, if the
BOOT_ACK field of the MMC device located in the Extended CSD register is set to one.
5. Host processor can copy boot data sequentially as soon as the RXRDY flag is
asserted.
6. When Data transfer is completed, host processor shall terminate the boot stream by
writing the HSMCI_CMDR register with SPCMD field set to BOOTEND.
34.11.2 Boot Procedure, DMA Mode

1. Configure the HSMCI data bus width by programming SDCBUS Field in the
HSMCI_SDCR register. The BOOT_BUS_WIDTH field in the device Extended CSD
register must be set accordingly.
2. Set the byte count to 512 bytes and the block count to the desired number of blocks by
writing BLKLEN and BCNT fields of the HSMCI_BLKR Register.
3. Enable DMA transfer in the HSMCI_DMA register.
4. Configure DMA controller, program the total amount of data to be transferred and
enable the relevant channel.
5. Issue the Boot Operation Request command by writing to the HSMCI_CMDR register
with SPCND set to BOOTREQ, TRDIR set to READ and TRCMD set to “start data
transfer”.
6. DMA controller copies the boot partition to the memory.
7. When DMA transfer is completed, host processor shall terminate the boot stream by
writing the HSMCI_CMDR register with SPCMD field set to BOOTEND.
675
AT91SAM9G45
34.12 HSMCI Transfer Done Timings
34.12.1 Definition
The XFRDONE flag in the HSMCI_SR indicates exactly when the read or write sequence is
finished.
34.12.2 Read Access

During a read access, the XFRDONE flag behaves as shown in Figure 34-11.
Figure 34-11. XFRDONE During a Read Access

CMD line
MCI read CMD Card response
The CMDRDY flag is released 8 tbit after the end of the card response.
CMDRDY flag
Data
1st Block Last Block
Not busy flag
XFRDONE flag
676
AT91SAM9G45
34.12.3 Write Access

During a write access, the XFRDONE flag behaves as shown in Figure 34-12.
Figure 34-12. XFRDONE During a Write Access

CMD line
MCI writeCMD Card response
CMDRDY flag The CMDRDY flag is released 8 tbit after the end of the card response.
D0 D0 is tied by the card

D0 is released
Data bus - D0
Not busy flag
XFRDONE flag
677
AT91SAM9G45
34.13 MultiMedia Card Interface (MCI) User Interface

0x00 Control Register HSMCI_CR Write –
0x04 Mode Register HSMCI_MR Read-write 0x0
0x08 Data Timeout Register HSMCI_DTOR Read-write 0x0
0x0C SD/SDIO Card Register HSMCI_SDCR Read-write 0x0
0x10 Argument Register HSMCI_ARGR Read-write 0x0
0x14 Command Register HSMCI_CMDR Write –
0x18 Block Register HSMCI_BLKR Read-write 0x0
0x1C Completion Signal Timeout Register HSMCI_CSTOR Read-write 0x0
(1)
0x20 Response Register HSMCI_RSPR Read 0x0
(1)
(1)
0x2C Response Register(1) HSMCI_RSPR Read 0x0
0x30 Receive Data Register HSMCI_RDR Read 0x0
0x34 Transmit Data Register HSMCI_TDR Write –
0x38 - 0x3C Reserved – – –
0x40 Status Register HSMCI_SR Read 0xC0E5
0x44 Interrupt Enable Register HSMCI_IER Write –
0x48 Interrupt Disable Register HSMCI_IDR Write –
0x4C Interrupt Mask Register HSMCI_IMR Read 0x0
0x50 DMA Configuration Register HSMCI_DMA Read-write 0x00
0x54 Configuration Register HSMCI_CFG Read-write 0x00
0x58-0xE0 Reserved – – –
0xE4 Write Protection Mode Register HSMCI_WPMR Read-write –
0xE8 Write Protection Status Register HSMCI_WPSR Read-only –
0xEC - 0xFC Reserved – – –
0x100-0x124 Reserved – – –
0x200-0x3FFC FIFO Memory Aperture HSMCI_FIFO Read-write 0x0
Note: 1. The response register can be read by N accesses at the same HSMCI_RSPR or at consecutive addresses (0x20 to 0x2C).
N depends on the size of the response.
678
AT91SAM9G45
34.13.1 HSMCI Control Register

Name: HSMCI_CR
Addresses: 0xFFF80000 (0), 0xFFFD0000 (1)
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
SWRST – – – PWSDIS PWSEN HSMCIDIS MCIEN
• MCIEN: Multi-Media Interface Enable

0 = No effect.
1 = Enables the Multi-Media Interface if MCDIS is 0.
• MCIDIS: Multi-Media Interface Disable

0 = No effect.
1 = Disables the Multi-Media Interface.
• PWSEN: Power Save Mode Enable

0 = No effect.
1 = Enables the Power Saving Mode if PWSDIS is 0.
Warning: Before enabling this mode, the user must set a value different from 0 in the PWSDIV field (Mode Register,
HSMCI_MR).
• PWSDIS: Power Save Mode Disable

0 = No effect.
1 = Disables the Power Saving Mode.

0 = No effect.
1 = Resets the HSMCI. A software triggered hardware reset of the HSMCI interface is performed.
679
AT91SAM9G45
34.13.2 HSMCI Mode Register

Name: HSMCI_MR
Access: Read-write
31 30 29 28 27 26 25 24
BLKLEN
23 22 21 20 19 18 17 16
BLKLEN
15 14 13 12 11 10 9 8
– PADV FBYTE WRPROOF RDPROOF PWSDIV
7 6 5 4 3 2 1 0
CLKDIV
• CLKDIV: Clock Divider

High Speed MultiMedia Card Interface clock (MCCK or HSMCI_CK) is Master Clock (MCK) divided by (2*(CLKDIV+1)).
• PWSDIV: Power Saving Divider

High Speed MultiMedia Card Interface clock is divided by 2(PWSDIV) + 1 when entering Power Saving Mode.
Warning: This value must be different from 0 before enabling the Power Save Mode in the HSMCI_CR (HSMCI_PWSEN bit).
• RDPROOF Read Proof Enable

Enabling Read Proof allows to stop the HSMCI Clock during read access if the internal FIFO is full. This will guarantee data
integrity, not bandwidth.
0 = Disables Read Proof.
1 = Enables Read Proof.
• WRPROOF Write Proof Enable

Enabling Write Proof allows to stop the HSMCI Clock during write access if the internal FIFO is full. This will guarantee data
integrity, not bandwidth.
0 = Disables Write Proof.
1 = Enables Write Proof.
• FBYTE: Force Byte Transfer

Enabling Force Byte Transfer allow byte transfers, so that transfer of blocks with a size different from modulo 4 can be
supported.
Warning: BLKLEN value depends on FBYTE.
0 = Disables Force Byte Transfer.
1 = Enables Force Byte Transfer.
• PADV: Padding Value

0 = 0x00 value is used when padding data in write transfer.
1 = 0xFF value is used when padding data in write transfer.
PADV may be only in manual transfer.
680
AT91SAM9G45
• BLKLEN: Data Block Length

This field determines the size of the data block.
This field is also accessible in the HSMCI Block Register (HSMCI_BLKR).
Bits 16 and 17 must be set to 0 if FBYTE is disabled.
Note: In SDIO Byte mode, BLKLEN field is not used.
681
AT91SAM9G45
34.13.3 HSMCI Data Timeout Register

Name: HSMCI_DTOR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– DTOMUL DTOCYC
• DTOCYC: Data Timeout Cycle Number
• DTOMUL: Data Timeout Multiplier

These fields determine the maximum number of Master Clock cycles that the HSMCI waits between two data block trans-
fers. It equals (DTOCYC x Multiplier).
Multiplier is defined by DTOMUL as shown in the following table:
DTOMUL Multiplier
0 0 0 1
0 0 1 16
0 1 0 128
0 1 1 256
1 0 0 1024
1 0 1 4096
1 1 0 65536
1 1 1 1048576
If the data time-out set by DTOCYC and DTOMUL has been exceeded, the Data Time-out Error flag (DTOE) in the HSMCI
Status Register (HSMCI_SR) raises.
682
AT91SAM9G45
34.13.4 HSMCI SDCard/SDIO Register

Name: HSMCI_SDCR
Addresses: 0xFFF8000C (0), 0xFFFD000C (1)
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
SDCBUS – – – – SDCSEL
• SDCSEL: SDCard/SDIO Slot
SDCSEL SDCard/SDIO Slot

0 0 Slot A is selected.
0 1 –
1 0 –
1 1 –
• SDCBUS: SDCard/SDIO Bus Width
SDCBUS BUS WIDTH

0 0 1 bit
1 0 4 bit
1 1 8 bit
683
AT91SAM9G45
34.13.5 HSMCI Argument Register

Name: HSMCI_ARGR
Access: Read-write
31 30 29 28 27 26 25 24
ARG
23 22 21 20 19 18 17 16
ARG
15 14 13 12 11 10 9 8
ARG
7 6 5 4 3 2 1 0
ARG
• ARG: Command Argument
684
AT91SAM9G45
34.13.6 HSMCI Command Register

Name: HSMCI_CMDR
Access: Write-only
31 30 29 28 27 26 25 24
– – – – BOOT_ACK ATACS IOSPCMD
23 22 21 20 19 18 17 16
– – TRTYP TRDIR TRCMD
15 14 13 12 11 10 9 8
– – – MAXLAT OPDCMD SPCMD
7 6 5 4 3 2 1 0
RSPTYP CMDNB
This register is write-protected while CMDRDY is 0 in HSMCI_SR. If an Interrupt command is sent, this register is only
writeable by an interrupt response (field SPCMD). This means that the current command execution cannot be interrupted
or modified.
• CMDNB: Command Number
• RSPTYP: Response Type
RSP Response Type

0 0 No response.
0 1 48-bit response.
1 0 136-bit response.
1 1 R1b response type
• SPCMD: Special Command
SPCMD Command
0 0 0 Not a special CMD.
Initialization CMD:
0 0 1
74 clock cycles for initialization sequence.
Synchronized CMD:
0 1 0 Wait for the end of the current data block transfer before sending the
pending command.
CE-ATA Completion Signal disable Command.
0 1 1 The host cancels the ability for the device to return a command
completion signal on the command line.
Interrupt command:
1 0 0
Corresponds to the Interrupt Mode (CMD40).
685
AT91SAM9G45
SPCMD Command
Interrupt response:
1 0 1
Corresponds to the Interrupt Mode (CMD40).
Boot Operation Request.
1 1 0 Start a boot operation mode, the host processor can read boot data from
the MMC device directly.
End Boot Operation.
1 1 1 This command allows the host processor to terminate the boot operation
mode.
• OPDCMD: Open Drain Command

0 = Push pull command
1 = Open drain command
• MAXLAT: Max Latency for Command to Response

0 = 5-cycle max latency
1 = 64-cycle max latency
• TRCMD: Transfer Command
TRCMD Transfer Type

0 0 No data transfer
0 1 Start data transfer
1 0 Stop data transfer
1 1 Reserved
• TRDIR: Transfer Direction

0 = Write
1 = Read
• TRTYP: Transfer Type
TRTYP Transfer Type

0 0 0 MMC/SDCard Single Block
0 0 1 MMC/SDCard Multiple Block
0 1 0 MMC Stream
0 1 1 Reserved
1 0 0 SDIO Byte
1 0 1 SDIO Block
1 1 0 Reserved
1 1 1 Reserved
686
AT91SAM9G45
• IOSPCMD: SDIO Special Command
IOSPCMD SDIO Special Command Type

0 0 Not a SDIO Special Command
0 1 SDIO Suspend Command
1 0 SDIO Resume Command
1 1 Reserved
• ATACS: ATA with Command Completion Signal

0 = Normal operation mode.
1 = This bit indicates that a completion signal is expected within a programmed amount of time (HSMCI_CSTOR).
• BOOT_ACK: Boot Operation Acknowledge.

The master can choose to receive the boot acknowledge from the slave when a Boot Request command is issued. When
set to one this field indicates that a Boot acknowledge is expected within a programmable amount of time defined with
DTOMUL and DTOCYC fields located in the HSMCI_DTOR register. If the acknowledge pattern is not received then an
acknowledge timeout error is raised. If the acknowledge pattern is corrupted then an acknowledge pattern error is set.
687
AT91SAM9G45
34.13.7 HSMCI Block Register

Name: HSMCI_BLKR
Access: Read-write
31 30 29 28 27 26 25 24
BLKLEN
23 22 21 20 19 18 17 16
BLKLEN
15 14 13 12 11 10 9 8
BCNT
7 6 5 4 3 2 1 0
BCNT
• BCNT: MMC/SDIO Block Count - SDIO Byte Count

This field determines the number of data byte(s) or block(s) to transfer.
The transfer data type and the authorized values for BCNT field are determined by the TRTYP field in the HSMCI Com-
mand Register (HSMCI_CMDR):
TRTYP Type of Transfer BCNT Authorized Values

0 0 1 MMC/SDCard Multiple Block From 1 to 65635: Value 0 corresponds to an infinite block transfer.
From 1 to 512 bytes: Value 0 corresponds to a 512-byte transfer.
1 0 0 SDIO Byte
Values from 0x200 to 0xFFFF are forbidden.
From 1 to 511 blocks: Value 0 corresponds to an infinite block transfer.
1 0 1 SDIO Block
Values from 0x200 to 0xFFFF are forbidden.
Other values - Reserved.
Warning: In SDIO Byte and Block modes, writing to the 7 last bits of BCNT field, is forbidden and may lead to unpredict-
able results.
• BLKLEN: Data Block Length

This field determines the size of the data block.
This field is also accessible in the HSMCI Mode Register (HSMCI_MR).
Bits 16 and 17 must be set to 0 if FBYTE is disabled.
Note: In SDIO Byte mode, BLKLEN field is not used.
688
AT91SAM9G45
34.13.8 HSMCI Completion Signal Timeout Register

Name: HSMCI_CSTOR
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– CSTOMUL CSTOCYC
• CSTOCYC: Completion Signal Timeout Cycle Number
• CSTOMUL: Completion Signal Timeout Multiplier

These fields determine the maximum number of Master Clock cycles that the HSMCI waits between two data block trans-
fers. Its value is calculated by (CSTOCYC x Multiplier).
These fields determine the maximum number of Master Clock cycles that the HSMCI waits between the end of the data
transfer and the assertion of the completion signal. The data transfer comprises data phase and the optional busy phase. If
a non-DATA ATA command is issued, the HSMCI starts waiting immediately after the end of the response until the comple-
tion signal.
Multiplier is defined by CSTOMUL as shown in the following table:
CSTOMUL Multiplier
0 0 0 1
0 0 1 16
0 1 0 128
0 1 1 256
1 0 0 1024
1 0 1 4096
1 1 0 65536
1 1 1 1048576
If the data time-out set by CSTOCYC and CSTOMUL has been exceeded, the Completion Signal Time-out Error flag
(CSTOE) in the HSMCI Status Register (HSMCI_SR) raises.
689
AT91SAM9G45
34.13.9 HSMCI Response Register

Name: HSMCI_RSPR
Access: Read-only
31 30 29 28 27 26 25 24
RSP
23 22 21 20 19 18 17 16
RSP
15 14 13 12 11 10 9 8
RSP
7 6 5 4 3 2 1 0
RSP
• RSP: Response
Note: 1. The response register can be read by N accesses at the same HSMCI_RSPR or at consecutive addresses (0x20 to 0x2C).
N depends on the size of the response.
34.13.10 HSMCI Receive Data Register

Name: HSMCI_RDR
Access: Read-only
31 30 29 28 27 26 25 24
DATA
23 22 21 20 19 18 17 16
DATA
15 14 13 12 11 10 9 8
DATA
7 6 5 4 3 2 1 0
DATA
• DATA: Data to Read
690
AT91SAM9G45
34.13.11 HSMCI Transmit Data Register

Name: HSMCI_TDR
Access: Write-only
31 30 29 28 27 26 25 24
DATA
23 22 21 20 19 18 17 16
DATA
15 14 13 12 11 10 9 8
DATA
7 6 5 4 3 2 1 0
DATA
• DATA: Data to Write
691
AT91SAM9G45
34.13.12 HSMCI Status Register

Name: HSMCI_SR
Access: Read-only
31 30 29 28 27 26 25 24
UNRE OVRE ACKRCVE ACKRCV XFRDONE FIFOEMPTY DMADONE BLKOVRE
23 22 21 20 19 18 17 16
CSTOE DTOE DCRCE RTOE RENDE RCRCE RDIRE RINDE
15 14 13 12 11 10 9 8
MCI_SDIOIR
– – CSRCV SDIOWAIT – – –
QA
7 6 5 4 3 2 1 0
– – NOTBUSY DTIP BLKE TXRDY RXRDY CMDRDY
• CMDRDY: Command Ready

0 = A command is in progress.
1 = The last command has been sent. Cleared when writing in the HSMCI_CMDR.

0 = Data has not yet been received since the last read of HSMCI_RDR.
1 = Data has been received since the last read of HSMCI_RDR.
• TXRDY: Transmit Ready

0 = The last data written in HSMCI_TDR has not yet been transferred in the Shift Register.
1 = The last data written in HSMCI_TDR has been transferred in the Shift Register.
• BLKE: Data Block Ended

This flag must be used only for Write Operations.
0 = A data block transfer is not yet finished. Cleared when reading the HSMCI_SR.
1 = A data block transfer has ended, including the CRC16 Status transmission.
the flag is set for each transmitted CRC Status.
Refer to the MMC or SD Specification for more details concerning the CRC Status.
• DTIP: Data Transfer in Progress

0 = No data transfer in progress.
1 = The current data transfer is still in progress, including CRC16 calculation. Cleared at the end of the CRC16 calculation.
• NOTBUSY: HSMCI Not Busy

This flag must be used only for Write Operations.
A block write operation uses a simple busy signalling of the write operation duration on the data (DAT0) line: during a data
transfer block, if the card does not have a free data receive buffer, the card indicates this condition by pulling down the data
line (DAT0) to LOW. The card stops pulling down the data line as soon as at least one receive buffer for the defined data
transfer block length becomes free.
The NOTBUSY flag allows to deal with these different states.
692
AT91SAM9G45
0 = The HSMCI is not ready for new data transfer. Cleared at the end of the card response.
1 = The HSMCI is ready for new data transfer. Set when the busy state on the data line has ended. This corresponds to a
free internal data receive buffer of the card.
Refer to the MMC or SD Specification for more details concerning the busy behavior.
For all the read operations, the NOTBUSY flag is cleared at the end of the host command.
For the Infinite Read Multiple Blocks, the NOTBUSY flag is set at the end of the STOP_TRANSMISSION host command
(CMD12).
For the Single Block Reads, the NOTBUSY flag is set at the end of the data read block.
For the Multiple Block Reads with pre-defined block count, the NOTBUSY flag is set at the end of the last received data
block.
• SDIOIRQA: SDIO Interrupt for Slot A

0 = No interrupt detected on SDIO Slot A.
1 = An SDIO Interrupt on Slot A occurred. Cleared when reading the HSMCI_SR.
• SDIOWAIT: SDIO Read Wait Operation Status

0 = Normal Bus operation.
1 = The data bus has entered IO wait state.
• CSRCV: CE-ATA Completion Signal Received

0 = No completion signal received since last status read operation.
1 = The device has issued a command completion signal on the command line. Cleared by reading in the HSMCI_SR
register.
• RINDE: Response Index Error

0 = No error.
1 = A mismatch is detected between the command index sent and the response index received. Cleared when writing in
the HSMCI_CMDR.
• RDIRE: Response Direction Error

0 = No error.
1 = The direction bit from card to host in the response has not been detected.
• RCRCE: Response CRC Error

0 = No error.
1 = A CRC7 error has been detected in the response. Cleared when writing in the HSMCI_CMDR.
• RENDE: Response End Bit Error

0 = No error.
1 = The end bit of the response has not been detected. Cleared when writing in the HSMCI_CMDR.
• RTOE: Response Time-out Error

0 = No error.
1 = The response time-out set by MAXLAT in the HSMCI_CMDR has been exceeded. Cleared when writing in the
HSMCI_CMDR.
693
AT91SAM9G45
• DCRCE: Data CRC Error

0 = No error.
1 = A CRC16 error has been detected in the last data block. Cleared by reading in the HSMCI_SR register.
• DTOE: Data Time-out Error

0 = No error.
1 = The data time-out set by DTOCYC and DTOMUL in HSMCI_DTOR has been exceeded. Cleared by reading in the
HSMCI_SR register.
• CSTOE: Completion Signal Time-out Error

0 = No error.
1 = The completion signal time-out set by CSTOCYC and CSTOMUL in HSMCI_CSTOR has been exceeded. Cleared by
reading in the HSMCI_SR register. Cleared by reading in the HSMCI_SR register.
• BLKOVRE: DMA Block Overrun Error

0 = No error.
1 = A new block of data is received and the DMA controller has not started to move the current pending block, a block over-
run is raised. Cleared by reading in the HSMCI_SR register.
• DMADONE: DMA Transfer done

0 = DMA buffer transfer has not completed since the last read of HSMCI_SR register.
1 = DMA buffer transfer has completed.
• FIFOEMPTY: FIFO empty flag

0 = FIFO contains at least one byte.
1 = FIFO is empty.
• XFRDONE: Transfer Done flag

0 = A transfer is in progress.
1 = Command register is ready to operate and the data bus is in the idle state.
• ACKRCV: Boot Operation Acknowledge Received

0 = No Boot acknowledge received since the last read of the status register.
1 = A Boot acknowledge signal has been received. Cleared by reading the HSMCI_SR register.
• ACKRCVE: Boot Operation Acknowledge Error

0 = No error
1 = Corrupted Boot Acknowledge signal received.
• OVRE: Overrun
0 = No error.
1 = At least one 8-bit received data has been lost (not read). Cleared when sending a new data transfer command.
When FERRCTRL in HSMCI_CFG is set to 1, OVRE becomes reset after read.
694
AT91SAM9G45
• UNRE: Underrun
0 = No error.
1 = At least one 8-bit data has been sent without valid information (not written). Cleared when sending a new data transfer
command or when setting FERRCTRL in HSMCI_CFG to 1.
When FERRCTRL in HSMCI_CFG is set to 1, UNRE becomes reset after read.
695
AT91SAM9G45
34.13.13 HSMCI Interrupt Enable Register

Name: HSMCI_IER
Access: Write-only
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
– – CSRCV SDIOWAIT – – – MCI_SDIOIRQA
7 6 5 4 3 2 1 0
• CMDRDY: Command Ready Interrupt Enable
• RXRDY: Receiver Ready Interrupt Enable
• TXRDY: Transmit Ready Interrupt Enable
• BLKE: Data Block Ended Interrupt Enable
• DTIP: Data Transfer in Progress Interrupt Enable
• NOTBUSY: Data Not Busy Interrupt Enable
• SDIOIRQA: SDIO Interrupt for Slot A Interrupt Enable
• SDIOWAIT: SDIO Read Wait Operation Status Interrupt Enable
• CSRCV: Completion Signal Received Interrupt Enable
• RINDE: Response Index Error Interrupt Enable
• RDIRE: Response Direction Error Interrupt Enable
• RCRCE: Response CRC Error Interrupt Enable
• RENDE: Response End Bit Error Interrupt Enable
• RTOE: Response Time-out Error Interrupt Enable
• DCRCE: Data CRC Error Interrupt Enable
• DTOE: Data Time-out Error Interrupt Enable
• CSTOE: Completion Signal Timeout Error Interrupt Enable
• BLKOVRE: DMA Block Overrun Error Interrupt Enable
• DMADONE: DMA Transfer completed Interrupt Enable
• FIFOEMPTY: FIFO empty Interrupt enable
• XFRDONE: Transfer Done Interrupt enable
696
AT91SAM9G45
• ACKRCV: Boot Acknowledge Interrupt Enable
• ACKRCVE: Boot Acknowledge Error Interrupt Enable
• OVRE: Overrun Interrupt Enable
• UNRE: Underrun Interrupt Enable

0 = No effect.
697
AT91SAM9G45
34.13.14 HSMCI Interrupt Disable Register

Name: HSMCI_IDR
Access: Write-only
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• CMDRDY: Command Ready Interrupt Disable
• RXRDY: Receiver Ready Interrupt Disable
• TXRDY: Transmit Ready Interrupt Disable
• BLKE: Data Block Ended Interrupt Disable
• DTIP: Data Transfer in Progress Interrupt Disable
• NOTBUSY: Data Not Busy Interrupt Disable
• SDIOIRQA: SDIO Interrupt for Slot A Interrupt Disable
• SDIOWAIT: SDIO Read Wait Operation Status Interrupt Disable
• CSRCV: Completion Signal received interrupt disable
• RINDE: Response Index Error Interrupt Disable
• RDIRE: Response Direction Error Interrupt Disable
• RCRCE: Response CRC Error Interrupt Disable
• RENDE: Response End Bit Error Interrupt Disable
• RTOE: Response Time-out Error Interrupt Disable
• DCRCE: Data CRC Error Interrupt Disable
• DTOE: Data Time-out Error Interrupt Disable
• CSTOE: Completion Signal Time out Error Interrupt Disable
• BLKOVRE: DMA Block Overrun Error Interrupt Disable
• DMADONE: DMA Transfer completed Interrupt Disable
• FIFOEMPTY: FIFO empty Interrupt Disable
• XFRDONE: Transfer Done Interrupt Disable
698
AT91SAM9G45
• ACKRCV: Boot Acknowledge Interrupt Disable
• ACKRCVE: Boot Acknowledge Error Interrupt Disable
• OVRE: Overrun Interrupt Disable
• UNRE: Underrun Interrupt Disable

0 = No effect.
699
AT91SAM9G45
34.13.15 HSMCI Interrupt Mask Register

Name: HSMCI_IMR
Access: Read-only
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• CMDRDY: Command Ready Interrupt Mask
• RXRDY: Receiver Ready Interrupt Mask
• TXRDY: Transmit Ready Interrupt Mask
• BLKE: Data Block Ended Interrupt Mask
• DTIP: Data Transfer in Progress Interrupt Mask
• NOTBUSY: Data Not Busy Interrupt Mask
• SDIOIRQA: SDIO Interrupt for Slot A Interrupt Mask
• SDIOWAIT: SDIO Read Wait Operation Status Interrupt Mask
• CSRCV: Completion Signal Received Interrupt Mask
• RINDE: Response Index Error Interrupt Mask
• RDIRE: Response Direction Error Interrupt Mask
• RCRCE: Response CRC Error Interrupt Mask
• RENDE: Response End Bit Error Interrupt Mask
• RTOE: Response Time-out Error Interrupt Mask
• DCRCE: Data CRC Error Interrupt Mask
• DTOE: Data Time-out Error Interrupt Mask
• CSTOE: Completion Signal Time-out Error Interrupt Mask
• BLKOVRE: DMA Block Overrun Error Interrupt Mask
• DMADONE: DMA Transfer Completed Interrupt Mask
• FIFOEMPTY: FIFO Empty Interrupt Mask
• XFRDONE: Transfer Done Interrupt Mask
700
AT91SAM9G45
• ACKRCV: Boot Operation Acknowledge Received Interrupt Mask
• ACKRCVE: Boot Operation Acknowledge Error Interrupt Mask
• OVRE: Overrun Interrupt Mask
• UNRE: Underrun Interrupt Mask

701
AT91SAM9G45
34.13.16 HSMCI DMA Configuration Register

Name: HSMCI_DMA
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – ROPT – – – DMAEN
7 6 5 4 3 2 1 0
– – CHKSIZE – – OFFSET
• OFFSET: DMA Write Buffer Offset

This field indicates the number of discarded bytes when the DMA writes the first word of the transfer.
• CHKSIZE: DMA Channel Read and Write Chunk Size

The CHKSIZE field indicates the number of data available when the DMA chunk transfer request is asserted.
CHKSIZE value Number of data transferred

00 1
01 4
10 8
11 16
• DMAEN: DMA Hardware Handshaking Enable

0 = DMA interface is disabled.
1 = DMA Interface is enabled.
Note: To avoid unpredictable behavior, DMA hardware handshaking must be disabled when CPU transfers are performed.
• ROPT: Read Optimization with padding

0: BLKLEN bytes are moved from the Memory Card to the system memory, two DMA descriptors are used when the trans-
fer size is not a multiple of 4.
1: Ceiling(BLKLEN/4) * 4 bytes are moved from the Memory Card to the system memory, only one DMA descriptor is used.
702
AT91SAM9G45
34.13.17 HSMCI Configuration Register

Name: HSMCI_CFG
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – LSYNC – – – HSMODE
7 6 5 4 3 2 1 0
– – – FERRCTRL – – – FIFOMODE
• FIFOMODE: HSMCI Internal FIFO control mode

0 = A write transfer starts when a sufficient amount of data is written into the FIFO.
When the block length is greater than or equal to 3/4 of the HSMCI internal FIFO size, then the write transfer starts as soon
as half the FIFO is filled. When the block length is greater than or equal to half the internal FIFO size, then the write transfer
starts as soon as one quarter of the FIFO is filled. In other cases, the transfer starts as soon as the total amount of data is
written in the internal FIFO.
1 = A write transfer starts as soon as one data is written into the FIFO.
• FERRCTRL: Flow Error flag reset control mode

0 = When an underflow/overflow condition flag is set, a new Write/Read command is needed to reset the flag.
1 = When an underflow/overflow condition flag is set, a read status resets the flag.
• HSMODE: High Speed Mode

0 = Default bus timing mode.
1 = If set to one, the host controller outputs command line and data lines on the rising edge of the card clock. The Host
driver shall check the high speed support in the card registers.
• LSYNC: Synchronize on the last block

0 = The pending command is sent at the end of the current data block.
1 = The pending command is sent at the end of the block transfer when the transfer length is not infinite (block count shall
be different from zero).
703
AT91SAM9G45
34.13.18 HSMCI Write Protect Mode Register

Name: HSMCI_WPMR
Addresses: 0xFFF800E4 (0), 0xFFFD00E4 (1)
Access: Read-write
31 30 29 28 27 26 25 24
WP_KEY (0x4D => “M”)
23 22 21 20 19 18 17 16
WP_KEY (0x43 => C”)
15 14 13 12 11 10 9 8
WP_KEY (0x49 => “I”)
7 6 5 4 3 2 1 0
WP_EN
• WP_EN: Write Protection Enable

0 = Disables the Write Protection if WP_KEY corresponds.
1 = Enables the Write Protection if WP_KEY corresponds.
• WP_KEY: Write Protection Key password

Should be written at value 0x4D4349 (ASCII code for “HSMCI”). Writing any other value in this field has no effect.
704
AT91SAM9G45
34.13.19 HSMCI Write Protect Status Register

Name: HSMCI_WPSR
Addresses: 0xFFF800E8 (0), 0xFFFD00E8 (1)
Access: Read-only
31 30 29 28 27 26 25 24
- - - - - - - -
23 22 21 20 19 18 17 16
WP_VSRC
15 14 13 12 11 10 9 8
WP_VSRC
7 6 5 4 3 2 1 0
- - - - WP_VS
• WP_VSRC: Write Protection Violation Status
WP_VS
No Write Protection Violation occurred since the last read of this
0 0 0 0
register (WP_SR)
Write Protection detected unauthorized attempt to write a control
0 0 0 1
register had occurred (since the last read.)
Software reset had been performed while Write Protection was
0 0 1 0
enabled (since the last read).
Both Write Protection violation and software reset with Write
0 0 1 1
Protection enabled had occurred since the last read.
Other value Reserved
• WP_VSRC: Write Protection Violation SouRCe
WP_VSRC
0 0 0 0
register (WP_SR)
Write access in HSMCI_MR while Write Protection was enabled
0 0 0 1
(since the last read).
Write access in HSMCI_DTOR while Write Protection was enabled
0 0 1 0
(since the last read)
Write access in HSMCI_SDCR while Write Protection was enabled
0 0 1 1
Write access in HSMCI_CSTOR while Write Protection was
0 1 0 0
enabled (since the last read)
Write access in HSMCI_DMA while Write Protection was enabled
0 1 0 1
Write access in HSMCI_CFG while Write Protection was enabled
0 1 1 0
705
AT91SAM9G45
34.13.20 HSMCI FIFO Memory Aperture

Name: HSMCI_FIFO
Access: Read-write
31 30 29 28 27 26 25 24
DATA
23 22 21 20 19 18 17 16
DATA
15 14 13 12 11 10 9 8
DATA
7 6 5 4 3 2 1 0
DATA
• DATA: Data to Read or Data to Write
706
AT91SAM9G45
35. Ethernet MAC 10/100 (EMAC)
35.1 Description
The EMAC module implements a 10/100 Ethernet MAC compatible with the IEEE 802.3 stan-
dard using an address checker, statistics and control registers, receive and transmit blocks, and
a DMA interface.
The address checker recognizes four specific 48-bit addresses and contains a 64-bit hash regis-
ter for matching multicast and unicast addresses. It can recognize the broadcast address of all
ones, copy all frames, and act on an external address match signal.
The statistics register block contains registers for counting various types of event associated
with transmit and receive operations. These registers, along with the status words stored in the
receive buffer list, enable software to generate network management statistics compatible with
IEEE 802.3.

• Compatibility with IEEE Standard 802.3
• 10 and 100 MBits per second data throughput capability
• Full- and half-duplex operations
• MII or RMII interface to the physical layer
• Register Interface to address, data, status and control registers
• DMA Interface, operating as a master on the Memory Controller
• Interrupt generation to signal receive and transmit completion
• 128-byte transmit and 128-byte receive FIFOs
• Automatic pad and CRC generation on transmitted frames
• Address checking logic to recognize four 48-bit addresses
• Supports promiscuous mode where all valid frames are copied to memory
• Supports physical layer management through MDIO interface
• Supports Wake On Lan. The receiver supports Wake on LAN by detecting the following
events on incoming receive frames:
– Magic packet
– ARP request to the device IP address
– Specific address 1 filter match
– Multicast hash filter match
707
35.3 Block Diagram
Figure 35-1. EMAC Block Diagram
Address Checker
APB
Slave
Register Interface
Statistics Registers
MDIO
Control Registers
DMA Interface
RX FIFO TX FIFO
Ethernet Receive
MII/RMII
AHB
Master
Ethernet Transmit
708 AT91SAM9G45
AT91SAM9G45

The MACB has several clock domains:
• System bus clock (AHB and APB): DMA and register blocks
• Transmit clock: transmit block
• Receive clock: receive and address checker block
The system bus clock must run at least as fast as the receive clock and transmit clock (25 MHz
at 100 Mbps, and 2.5 MHZ at 10 Mbps).
Figure 35-1 illustrates the different blocks of the EMAC module.
The control registers drive the MDIO interface, setup up DMA activity, start frame transmission
and select modes of operation such as full- or half-duplex.
The receive block checks for valid preamble, FCS, alignment and length, and presents received
frames to the address checking block and DMA interface.
The transmit block takes data from the DMA interface, adds preamble and, if necessary, pad
and FCS, and transmits data according to the CSMA/CD (carrier sense multiple access with col-
lision detect) protocol. The start of transmission is deferred if CRS (carrier sense) is active.
If COL (collision) becomes active during transmission, a jam sequence is asserted and the
transmission is retried after a random back off. CRS and COL have no effect in full duplex mode.
The DMA block connects to external memory through its AHB bus interface. It contains receive
and transmit FIFOs for buffering frame data. It loads the transmit FIFO and empties the receive
FIFO using AHB bus master operations. Receive data is not sent to memory until the address
checking logic has determined that the frame should be copied. Receive or transmit frames are
stored in one or more buffers. Receive buffers have a fixed length of 128 bytes. Transmit buffers
range in length between 0 and 2047 bytes, and up to 128 buffers are permitted per frame. The
DMA block manages the transmit and receive framebuffer queues. These queues can hold mul-
tiple frames.
35.4.1 Clock
Synchronization module in the EMAC requires that the bus clock (hclk) runs at the speed of the
macb_tx/rx_clk at least, which is 25 MHz at 100 Mbps, and 2.5 MHz at 10 Mbps.
35.4.2 Memory Interface

Frame data is transferred to and from the EMAC through the DMA interface. All transfers are 32-
bit words and may be single accesses or bursts of 2, 3 or 4 words. Burst accesses do not cross
sixteen-byte boundaries. Bursts of 4 words are the default data transfer; single accesses or
bursts of less than four words may be used to transfer data at the beginning or the end of a
buffer.
The DMA controller performs six types of operation on the bus. In order of priority, these are:
1. Receive buffer manager write
2. Receive buffer manager read
3. Transmit data DMA read
4. Receive data DMA write
5. Transmit buffer manager read
6. Transmit buffer manager write
709
35.4.2.1 FIFO
The FIFO depths are 128 bytes for receive and 128 bytes for transmit and are a function of the
system clock speed, memory latency and network speed.
Data is typically transferred into and out of the FIFOs in bursts of four words. For receive, a bus
request is asserted when the FIFO contains four words and has space for 28 more. For transmit,
a bus request is generated when there is space for four words, or when there is space for 27
words if the next transfer is to be only one or two words.
Thus the bus latency must be less than the time it takes to load the FIFO and transmit or receive
three words (112 bytes) of data.
At 100 Mbit/s, it takes 8960 ns to transmit or receive 112 bytes of data. In addition, six master
clock cycles should be allowed for data to be loaded from the bus and to propagate through the
FIFOs. For a 133 MHz master clock this takes 45 ns, making the bus latency requirement 8915
ns.
35.4.2.2 Receive Buffers

Received frames, including CRC/FCS optionally, are written to receive buffers stored in mem-
ory. Each receive buffer is 128 bytes long. The start location for each receive buffer is stored in
memory in a list of receive buffer descriptors at a location pointed to by the receive buffer queue
pointer register. The receive buffer start location is a word address. For the first buffer of a
frame, the start location can be offset by up to three bytes depending on the value written to bits
14 and 15 of the network configuration register. If the start location of the buffer is offset the
available length of the first buffer of a frame is reduced by the corresponding number of bytes.
Each list entry consists of two words, the first being the address of the receive buffer and the
second being the receive status. If the length of a receive frame exceeds the buffer length, the
status word for the used buffer is written with zeroes except for the “start of frame” bit and the
offset bits, if appropriate. Bit zero of the address field is written to one to show the buffer has
been used. The receive buffer manager then reads the location of the next receive buffer and
fills that with receive frame data. The final buffer descriptor status word contains the complete
frame status. Refer to Table 35-1 for details of the receive buffer descriptor list.
Table 35-1. Receive Buffer Descriptor Entry

Bit Function
Word 0
31:2 Address of beginning of buffer
1 Wrap - marks last descriptor in receive buffer descriptor list.
Ownership - needs to be zero for the EMAC to write data to the receive buffer. The EMAC sets this to one once it has
0 successfully written a frame to memory.
Software has to clear this bit before the buffer can be used again.
Word 1
31 Global all ones broadcast address detected
30 Multicast hash match
29 Unicast hash match
28 External address match
27 Reserved for future use
710 AT91SAM9G45
AT91SAM9G45
Table 35-1. Receive Buffer Descriptor Entry (Continued)

Bit Function
26 Specific address register 1 match
22 Type ID match
21 VLAN tag detected (i.e., type id of 0x8100)
20 Priority tag detected (i.e., type id of 0x8100 and null VLAN identifier)
19:17 VLAN priority (only valid if bit 21 is set)
16 Concatenation format indicator (CFI) bit (only valid if bit 21 is set)
End of frame - when set the buffer contains the end of a frame. If end of frame is not set, then the only other valid status
15
are bits 12, 13 and 14.
Start of frame - when set the buffer contains the start of a frame. If both bits 15 and 14 are set, then the buffer contains a
14
whole frame.
Receive buffer offset - indicates the number of bytes by which the data in the first buffer is offset from the word address.
Updated with the current values of the network configuration register. If jumbo frame mode is enabled through bit 3 of the
13:12
network configuration register, then bits 13:12 of the receive buffer descriptor entry are used to indicate bits 13:12 of the
frame length.
11:0 Length of frame including FCS (if selected). Bits 13:12 are also used if jumbo frame mode is selected.
To receive frames, the buffer descriptors must be initialized by writing an appropriate address to
bits 31 to 2 in the first word of each list entry. Bit zero must be written with zero. Bit one is the
wrap bit and indicates the last entry in the list.
The start location of the receive buffer descriptor list must be written to the receive buffer queue
pointer register before setting the receive enable bit in the network control register to enable
receive. As soon as the receive block starts writing received frame data to the receive FIFO, the
receive buffer manager reads the first receive buffer location pointed to by the receive buffer
queue pointer register.
If the filter block then indicates that the frame should be copied to memory, the receive data
DMA operation starts writing data into the receive buffer. If an error occurs, the buffer is recov-
ered. If the current buffer pointer has its wrap bit set or is the 1024th descriptor, the next receive
buffer location is read from the beginning of the receive descriptor list. Otherwise, the next
receive buffer location is read from the next word in memory.
There is an 11-bit counter to count out the 2048 word locations of a maximum length, receive
buffer descriptor list. This is added with the value originally written to the receive buffer queue
pointer register to produce a pointer into the list. A read of the receive buffer queue pointer reg-
ister returns the pointer value, which is the queue entry currently being accessed. The counter is
reset after receive status is written to a descriptor that has its wrap bit set or rolls over to zero
after 1024 descriptors have been accessed. The value written to the receive buffer pointer regis-
ter may be any word-aligned address, provided that there are at least 2048 word locations
available between the pointer and the top of the memory.
Section 3.6 of the AMBA 2.0 specification states that bursts should not cross 1K boundaries. As
receive buffer manager writes are bursts of two words, to ensure that this does not occur, it is
711
best to write the pointer register with the least three significant bits set to zero. As receive buffers
are used, the receive buffer manager sets bit zero of the first word of the descriptor to indicate
used. If a receive error is detected the receive buffer currently being written is recovered. Previ-
ous buffers are not recovered. Software should search through the used bits in the buffer
descriptors to find out how many frames have been received. It should be checking the start-of-
frame and end-of-frame bits, and not rely on the value returned by the receive buffer queue
pointer register which changes continuously as more buffers are used.
For CRC errored frames, excessive length frames or length field mismatched frames, all of
which are counted in the statistics registers, it is possible that a frame fragment might be stored
in a sequence of receive buffers. Software can detect this by looking for start of frame bit set in a
buffer following a buffer with no end of frame bit set.
For a properly working Ethernet system, there should be no excessively long frames or frames
greater than 128 bytes with CRC/FCS errors. Collision fragments are less than 128 bytes long.
Therefore, it is a rare occurrence to find a frame fragment in a receive buffer.
If bit zero is set when the receive buffer manager reads the location of the receive buffer, then
the buffer has already been used and cannot be used again until software has processed the
frame and cleared bit zero. In this case, the DMA block sets the buffer not available bit in the
receive status register and triggers an interrupt.
If bit zero is set when the receive buffer manager reads the location of the receive buffer and a
frame is being received, the frame is discarded and the receive resource error statistics register
is incremented.
A receive overrun condition occurs when bus was not granted in time or because HRESP was
not OK (bus error). In a receive overrun condition, the receive overrun interrupt is asserted and
the buffer currently being written is recovered. The next frame received with an address that is
recognized reuses the buffer.
If bit 17 of the network configuration register is set, the FCS of received frames shall not be cop-
ied to memory. The frame length indicated in the receive status field shall be reduced by four
bytes in this case.
35.4.2.3 Transmit Buffer

Frames to be transmitted are stored in one or more transmit buffers. Transmit buffers can be
between 0 and 2047 bytes long, so it is possible to transmit frames longer than the maximum
length specified in IEEE Standard 802.3. Zero length buffers are allowed. The maximum number
of buffers permitted for each transmit frame is 128.
The start location for each transmit buffer is stored in memory in a list of transmit buffer descrip-
tors at a location pointed to by the transmit buffer queue pointer register. Each list entry consists
of two words, the first being the byte address of the transmit buffer and the second containing
the transmit control and status. Frames can be transmitted with or without automatic CRC gen-
eration. If CRC is automatically generated, pad is also automatically generated to take frames to
a minimum length of 64 bytes. Table 35-2 on page 713 defines an entry in the transmit buffer
descriptor list. To transmit frames, the buffer descriptors must be initialized by writing an appro-
priate byte address to bits 31 to 0 in the first word of each list entry. The second transmit buffer
descriptor is initialized with control information that indicates the length of the buffer, whether or
not it is to be transmitted with CRC and whether the buffer is the last buffer in the frame.
After transmission, the control bits are written back to the second word of the first buffer along
with the “used” bit and other status information. Bit 31 is the “used” bit which must be zero when
712 AT91SAM9G45
AT91SAM9G45
the control word is read if transmission is to happen. It is written to one when a frame has been
transmitted. Bits 27, 28 and 29 indicate various transmit error conditions. Bit 30 is the “wrap” bit
which can be set for any buffer within a frame. If no wrap bit is encountered after 1024 descrip-
tors, the queue pointer rolls over to the start in a similar fashion to the receive queue.
The transmit buffer queue pointer register must not be written while transmit is active. If a new
value is written to the transmit buffer queue pointer register, the queue pointer resets itself to
point to the beginning of the new queue. If transmit is disabled by writing to bit 3 of the network
control, the transmit buffer queue pointer register resets to point to the beginning of the transmit
queue. Note that disabling receive does not have the same effect on the receive queue pointer.
Once the transmit queue is initialized, transmit is activated by writing to bit 9, the Transmit Start
bit of the network control register. Transmit is halted when a buffer descriptor with its used bit set
is read, or if a transmit error occurs, or by writing to the transmit halt bit of the network control
register. (Transmission is suspended if a pause frame is received while the pause enable bit is
set in the network configuration register.) Rewriting the start bit while transmission is active is
allowed.
Transmission control is implemented with a Tx_go variable which is readable in the transmit sta-
tus register at bit location 3. The Tx_go variable is reset when:
– transmit is disabled
– a buffer descriptor with its ownership bit set is read
– a new value is written to the transmit buffer queue pointer register
– bit 10, tx_halt, of the network control register is written
– there is a transmit error such as too many retries or a transmit underrun.
To set tx_go, write to bit 9, tx_start, of the network control register. Transmit halt does not take
effect until any ongoing transmit finishes. If a collision occurs during transmission of a multi-buf-
fer frame, transmission automatically restarts from the first buffer of the frame. If a “used” bit is
read midway through transmission of a multi-buffer frame, this is treated as a transmit error.
Transmission stops, tx_er is asserted and the FCS is bad.
If transmission stops due to a transmit error, the transmit queue pointer resets to point to the
beginning of the transmit queue. Software needs to re-initialize the transmit queue after a trans-
mit error.
If transmission stops due to a “used” bit being read at the start of the frame, the transmission
queue pointer is not reset and transmit starts from the same transmit buffer descriptor when the
transmit start bit is written
Table 35-2. Transmit Buffer Descriptor Entry

Bit Function
Word 0
31:0 Byte Address of buffer
Word 1
Used. Needs to be zero for the EMAC to read data from the transmit buffer. The EMAC sets this to one for the first buffer
of a frame once it has been successfully transmitted.
31
Software has to clear this bit before the buffer can be used again.
Note: This bit is only set for the first buffer in a frame unlike receive where all buffers have the Used bit set once used.
30 Wrap. Marks last descriptor in transmit buffer descriptor list.
713
Table 35-2. Transmit Buffer Descriptor Entry
Bit Function
29 Retry limit exceeded, transmit error detected
Transmit underrun, occurs either when hresp is not OK (bus error) or the transmit data could not be fetched in time or
28
when buffers are exhausted in mid frame.
27 Buffers exhausted in mid frame
26:17 Reserved
16 No CRC. When set, no CRC is appended to the current frame. This bit only needs to be set for the last buffer of a frame.
15 Last buffer. When set, this bit indicates the last buffer in the current frame has been reached.
14:11 Reserved
10:0 Length of buffer
35.4.3 Transmit Block

This block transmits frames in accordance with the Ethernet IEEE 802.3 CSMA/CD protocol.
Frame assembly starts by adding preamble and the start frame delimiter. Data is taken from the
transmit FIFO a word at a time. Data is transmitted least significant nibble first. If necessary,
padding is added to increase the frame length to 60 bytes. CRC is calculated as a 32-bit polyno-
mial. This is inverted and appended to the end of the frame, taking the frame length to a
minimum of 64 bytes. If the No CRC bit is set in the second word of the last buffer descriptor of a
transmit frame, neither pad nor CRC are appended.
In full-duplex mode, frames are transmitted immediately. Back-to-back frames are transmitted at
least 96 bit times apart to guarantee the interframe gap.
In half-duplex mode, the transmitter checks carrier sense. If asserted, it waits for it to de-assert
and then starts transmission after the interframe gap of 96 bit times. If the collision signal is
asserted during transmission, the transmitter transmits a jam sequence of 32 bits taken from the
data register and then retry transmission after the back off time has elapsed.
The back-off time is based on an XOR of the 10 least significant bits of the data coming from the
transmit FIFO and a 10-bit pseudo random number generator. The number of bits used depends
on the number of collisions seen. After the first collision, 1 bit is used, after the second 2, and so
on up to 10. Above 10, all 10 bits are used. An error is indicated and no further attempts are
made if 16 attempts cause collisions.
If transmit DMA underruns, bad CRC is automatically appended using the same mechanism as
jam insertion and the tx_er signal is asserted. For a properly configured system, this should
never happen.
If the back pressure bit is set in the network control register in half duplex mode, the transmit
block transmits 64 bits of data, which can consist of 16 nibbles of 1011 or in bit-rate mode 64 1s,
whenever it sees an incoming frame to force a collision. This provides a way of implementing
flow control in half-duplex mode.
714 AT91SAM9G45
AT91SAM9G45
35.4.4 Pause Frame Support

The start of an 802.3 pause frame is as follows:
Table 35-3. Start of an 802.3 Pause Frame

Source Type Pause
Destination Address Address (Mac Control Frame) Opcode Pause Time
0x0180C2000001 6 bytes 0x8808 0x0001 2 bytes
The network configuration register contains a receive pause enable bit (13). If a valid pause
frame is received, the pause time register is updated with the frame’s pause time, regardless of
its current contents and regardless of the state of the configuration register bit 13. An interrupt
(12) is triggered when a pause frame is received, assuming it is enabled in the interrupt mask
register. If bit 13 is set in the network configuration register and the value of the pause time reg-
ister is non-zero, no new frame is transmitted until the pause time register has decremented to
zero.
The loading of a new pause time, and hence the pausing of transmission, only occurs when the
EMAC is configured for full-duplex operation. If the EMAC is configured for half-duplex, there is
no transmission pause, but the pause frame received interrupt is still triggered.
A valid pause frame is defined as having a destination address that matches either the address
stored in specific address register 1 or matches 0x0180C2000001 and has the MAC control
frame type ID of 0x8808 and the pause opcode of 0x0001. Pause frames that have FCS or other
errors are treated as invalid and are discarded. Valid pause frames received increment the
Pause Frame Received statistic register.
The pause time register decrements every 512 bit times (i.e., 128 rx_clks in nibble mode)
once transmission has stopped. For test purposes, the register decrements every rx_clk cycle
once transmission has stopped if bit 12 (retry test) is set in the network configuration register. If
the pause enable bit (13) is not set in the network configuration register, then the decrementing
occurs regardless of whether transmission has stopped or not.
An interrupt (13) is asserted whenever the pause time register decrements to zero (assuming it
is enabled in the interrupt mask register).
35.4.5 Receive Block

The receive block checks for valid preamble, FCS, alignment and length, presents received
frames to the DMA block and stores the frames destination address for use by the address
checking block. If, during frame reception, the frame is found to be too long or rx_er is asserted,
a bad frame indication is sent to the DMA block. The DMA block then ceases sending data to
memory. At the end of frame reception, the receive block indicates to the DMA block whether the
frame is good or bad. The DMA block recovers the current receive buffer if the frame was bad.
The receive block signals the register block to increment the alignment error, the CRC (FCS)
error, the short frame, long frame, jabber error, the receive symbol error statistics and the length
field mismatch statistics.
The enable bit for jumbo frames in the network configuration register allows the EMAC to receive
jumbo frames of up to 10240 bytes in size. This operation does not form part of the IEEE802.3
specification and is disabled by default. When jumbo frames are enabled, frames received with a
frame size greater than 10240 bytes are discarded.
715
35.4.6 Address Checking Block
The address checking (or filter) block indicates to the DMA block which receive frames should
be copied to memory. Whether a frame is copied depends on what is enabled in the network
configuration register, the state of the external match pin, the contents of the specific address
and hash registers and the frame’s destination address. In this implementation of the EMAC, the
frame’s source address is not checked. Provided that bit 18 of the Network Configuration regis-
ter is not set, a frame is not copied to memory if the EMAC is transmitting in half duplex mode at
the time a destination address is received. If bit 18 of the Network Configuration register is set,
frames can be received while transmitting in half-duplex mode.
Ethernet frames are transmitted a byte at a time, least significant bit first. The first six bytes (48
bits) of an Ethernet frame make up the destination address. The first bit of the destination
address, the LSB of the first byte of the frame, is the group/individual bit: this is One for multicast
addresses and Zero for unicast. The All Ones address is the broadcast address, and a special
case of multicast.
The EMAC supports recognition of four specific addresses. Each specific address requires two
registers, specific address register bottom and specific address register top. Specific address
register bottom stores the first four bytes of the destination address and specific address register
top contains the last two bytes. The addresses stored can be specific, group, local or universal.
The destination address of received frames is compared against the data stored in the specific
address registers once they have been activated. The addresses are deactivated at reset or
when their corresponding specific address register bottom is written. They are activated when
specific address register top is written. If a receive frame address matches an active address,
the frame is copied to memory.
The following example illustrates the use of the address match registers for a MAC address of
21:43:65:87:A9:CB.
Preamble 55
SFD D5
DA (Octet0 - LSB) 21
DA(Octet 1) 43
DA(Octet 2) 65
DA(Octet 3) 87
DA(Octet 4) A9
DA (Octet5 - MSB) CB
SA (LSB) 00
SA 00
SA 00
SA 00
SA 00
SA (MSB) 43
SA (LSB) 21
716 AT91SAM9G45
AT91SAM9G45
The sequence above shows the beginning of an Ethernet frame. Byte order of transmission is
from top to bottom as shown. For a successful match to specific address 1, the following
address matching registers must be set up:
• Base address + 0x98 0x87654321 (Bottom)
• Base address + 0x9C 0x0000CBA9 (Top)
And for a successful match to the Type ID register, the following should be set up:
• Base address + 0xB8 0x00004321
35.4.7 Broadcast Address

The broadcast address of 0xFFFFFFFFFFFF is recognized if the ‘no broadcast’ bit in the net-
work configuration register is zero.
35.4.8 Hash Addressing

The hash address register is 64 bits long and takes up two locations in the memory map. The
least significant bits are stored in hash register bottom and the most significant bits in hash reg-
ister top.
The unicast hash enable and the multicast hash enable bits in the network configuration register
enable the reception of hash matched frames. The destination address is reduced to a 6-bit
index into the 64-bit hash register using the following hash function. The hash function is an
exclusive or of every sixth bit of the destination address.
hash_index[5] = da[5] ^ da[11] ^ da[17] ^ da[23] ^ da[29] ^ da[35] ^ da[41] ^ da[47]
da[0] represents the least significant bit of the first byte received, that is, the multicast/unicast
indicator, and da[47] represents the most significant bit of the last byte received.
If the hash index points to a bit that is set in the hash register, then the frame is matched accord-
ing to whether the frame is multicast or unicast.
A multicast match is signalled if the multicast hash enable bit is set. da[0] is 1 and the hash index
points to a bit set in the hash register.
A unicast match is signalled if the unicast hash enable bit is set. da[0] is 0 and the hash index
points to a bit set in the hash register.
To receive all multicast frames, the hash register should be set with all ones and the multicast
hash enable bit should be set in the network configuration register.
35.4.9 Copy All Frames (or Promiscuous Mode)

If the copy all frames bit is set in the network configuration register, then all non-errored frames
are copied to memory. For example, frames that are too long, too short, or have FCS errors or
rx_er asserted during reception are discarded and all others are received. Frames with FCS
errors are copied to memory if bit 19 in the network configuration register is set.
717
35.4.10 Type ID Checking
The contents of the type_id register are compared against the length/type ID of received frames
(i.e., bytes 13 and 14). Bit 22 in the receive buffer descriptor status is set if there is a match. The
reset state of this register is zero which is unlikely to match the length/type ID of any valid Ether-
net frame.
Note: A type ID match does not affect whether a frame is copied to memory.
35.4.11 VLAN Support

An Ethernet encoded 802.1Q VLAN tag looks like this:
Table 35-4. 802.1Q VLAN Tag

TPID (Tag Protocol Identifier) 16 bits TCI (Tag Control Information) 16 bits
0x8100 First 3 bits priority, then CFI bit, last 12 bits VID
The VLAN tag is inserted at the 13th byte of the frame, adding an extra four bytes to the frame. If
the VID (VLAN identifier) is null (0x000), this indicates a priority-tagged frame. The MAC can
support frame lengths up to 1536 bytes, 18 bytes more than the original Ethernet maximum
frame length of 1518 bytes. This is achieved by setting bit 8 in the network configuration register.
The following bits in the receive buffer descriptor status word give information about VLAN
tagged frames:
• Bit 21 set if receive frame is VLAN tagged (i.e. type id of 0x8100)
• Bit 20 set if receive frame is priority tagged (i.e. type id of 0x8100 and null VID). (If bit 20 is
set bit 21 is set also.)
• Bit 19, 18 and 17 set to priority if bit 21 is set
• Bit 16 set to CFI if bit 21 is set
35.4.12 PHY Maintenance

The register EMAC_MAN enables the EMAC to communicate with a PHY by means of the MDIO
interface. It is used during auto-negotiation to ensure that the EMAC and the PHY are config-
ured for the same speed and duplex configuration.
The PHY maintenance register is implemented as a shift register. Writing to the register starts a
shift operation which is signalled as complete when bit two is set in the network status register
(about 2000 MCK cycles later when bit ten is set to zero, and bit eleven is set to one in the net-
work configuration register). An interrupt is generated as this bit is set. During this time, the MSB
of the register is output on the MDIO pin and the LSB updated from the MDIO pin with each
MDC cycle. This causes transmission of a PHY management frame on MDIO.
Reading during the shift operation returns the current contents of the shift register. At the end of
management operation, the bits have shifted back to their original locations. For a read opera-
tion, the data bits are updated with data read from the PHY. It is important to write the correct
values to the register to ensure a valid PHY management frame is produced.
The MDIO interface can read IEEE 802.3 clause 45 PHYs as well as clause 22 PHYs. To read
clause 45 PHYs, bits[31:28] should be written as 0x0011. For a description of MDC generation,
see the network configuration register in the “Network Control Register” on page 725.
718 AT91SAM9G45
AT91SAM9G45
35.4.13 Media Independent Interface

The Ethernet MAC is capable of interfacing to both RMII and MII Interfaces. The RMII bit in the
EMAC_USRIO register controls the interface that is selected. When this bit is set, the RMII inter-
face is selected, else the MII interface is selected.
The MII and RMII interface are capable of both 10Mb/s and 100Mb/s data rates as described in
the IEEE 802.3u standard. The signals used by the MII and RMII interfaces are described in
Table 35-5.
Table 35-5. Pin Configuration

Pin Name MII RMII
ETXCK_EREFCK ETXCK: Transmit Clock EREFCK: Reference Clock
ECRS ECRS: Carrier Sense
ECOL ECOL: Collision Detect
ERXDV ERXDV: Data Valid ECRSDV: Carrier Sense/Data Valid
ERX0 - ERX3 ERX0 - ERX3: 4-bit Receive Data ERX0 - ERX1: 2-bit Receive Data
ERXER ERXER: Receive Error ERXER: Receive Error
ERXCK ERXCK: Receive Clock
ETXEN ETXEN: Transmit Enable ETXEN: Transmit Enable
ETX0-ETX3 ETX0 - ETX3: 4-bit Transmit Data ETX0 - ETX1: 2-bit Transmit Data
ETXER ETXER: Transmit Error
The intent of the RMII is to provide a reduced pin count alternative to the IEEE 802.3u MII. It
uses 2 bits for transmit (ETX0 and ETX1) and two bits for receive (ERX0 and ERX1). There is a
Transmit Enable (ETXEN), a Receive Error (ERXER), a Carrier Sense (ECRS_DV), and a 50
MHz Reference Clock (ETXCK_EREFCK) for 100Mb/s data rate.
35.4.13.1 RMII Transmit and Receive Operation

The same signals are used internally for both the RMII and the MII operations. The RMII maps
these signals in a more pin-efficient manner. The transmit and receive bits are converted from a
4-bit parallel format to a 2-bit parallel scheme that is clocked at twice the rate. The carrier sense
and data valid signals are combined into the ECRSDV signal. This signal contains information
on carrier sense, FIFO status, and validity of the data. Transmit error bit (ETXER) and collision
detect (ECOL) are not used in RMII mode.
719
35.5 Programming Interface
35.5.1 Initialization
35.5.1.1 Configuration
Initialization of the EMAC configuration (e.g., loop-back mode, frequency ratios) must be done
while the transmit and receive circuits are disabled. See the description of the network control
register and network configuration register earlier in this document.
To change loop-back mode, the following sequence of operations must be followed:
1. Write to network control register to disable transmit and receive circuits.
2. Write to network control register to change loop-back mode.
3. Write to network control register to re-enable transmit or receive circuits.
Note: These writes to network control register cannot be combined in any way.
35.5.1.2 Receive Buffer List

Receive data is written to areas of data (i.e., buffers) in system memory. These buffers are listed
in another data structure that also resides in main memory. This data structure (receive buffer
queue) is a sequence of descriptor entries as defined in “Receive Buffer Descriptor Entry” on
page 710. It points to this data structure.
Figure 35-2. Receive Buffer List
Receive Buffer 0
Receive Buffer Queue Pointer
(MAC Register)
Receive Buffer 1
Receive Buffer N
Receive Buffer Descriptor List

(In memory)
(In memory)
To create the list of buffers:

1. Allocate a number (n) of buffers of 128 bytes in system memory.
2. Allocate an area 2n words for the receive buffer descriptor entry in system memory and
create n entries in this list. Mark all entries in this list as owned by EMAC, i.e., bit 0 of
word 0 set to 0.
3. If less than 1024 buffers are defined, the last descriptor must be marked with the wrap
bit (bit 1 in word 0 set to 1).
4. Write address of receive buffer descriptor entry to EMAC register receive_buffer
queue pointer.
5. The receive circuits can then be enabled by writing to the address recognition registers
and then to the network control register.
720 AT91SAM9G45
AT91SAM9G45
35.5.1.3 Transmit Buffer List

Transmit data is read from areas of data (the buffers) in system memory These buffers are listed
in another data structure that also resides in main memory. This data structure (Transmit Buffer
Queue) is a sequence of descriptor entries (as defined in Table 35-2 on page 713) that points to
this data structure.
To create this list of buffers:
1. Allocate a number (n) of buffers of between 1 and 2047 bytes of data to be transmitted
in system memory. Up to 128 buffers per frame are allowed.
2. Allocate an area 2n words for the transmit buffer descriptor entry in system memory
and create N entries in this list. Mark all entries in this list as owned by EMAC, i.e. bit 31
of word 1 set to 0.
3. If fewer than 1024 buffers are defined, the last descriptor must be marked with the wrap
bit — bit 30 in word 1 set to 1.
4. Write address of transmit buffer descriptor entry to EMAC register transmit_buffer
queue pointer.
5. The transmit circuits can then be enabled by writing to the network control register.
35.5.1.4 Address Matching

The EMAC register-pair hash address and the four specific address register-pairs must be writ-
ten with the required values. Each register-pair comprises a bottom register and top register,
with the bottom register being written first. The address matching is disabled for a particular reg-
ister-pair after the bottom-register has been written and re-enabled when the top register is
written. See “Address Checking Block” on page 716. for details of address matching. Each reg-
ister-pair may be written at any time, regardless of whether the receive circuits are enabled or
disabled.
35.5.1.5 Interrupts
There are 14 interrupt conditions that are detected within the EMAC. These are ORed to make a
single interrupt. Depending on the overall system design, this may be passed through a further
level of interrupt collection (interrupt controller). On receipt of the interrupt signal, the CPU
enters the interrupt handler (Refer to the AIC programmer datasheet). To ascertain which inter-
rupt has been generated, read the interrupt status register. Note that this register clears itself
when read. At reset, all interrupts are disabled. To enable an interrupt, write to interrupt enable
register with the pertinent interrupt bit set to 1. To disable an interrupt, write to interrupt disable
register with the pertinent interrupt bit set to 1. To check whether an interrupt is enabled or dis-
abled, read interrupt mask register: if the bit is set to 1, the interrupt is disabled.
35.5.1.6 Transmitting Frames

To set up a frame for transmission:
1. Enable transmit in the network control register.
2. Allocate an area of system memory for transmit data. This does not have to be contigu-
ous, varying byte lengths can be used as long as they conclude on byte borders.
3. Set-up the transmit buffer list.
4. Set the network control register to enable transmission and enable interrupts.
5. Write data for transmission into these buffers.
6. Write the address to transmit buffer descriptor queue pointer.
7. Write control and length to word one of the transmit buffer descriptor entry.
721
8. Write to the transmit start bit in the network control register.
35.5.1.7 Receiving Frames

When a frame is received and the receive circuits are enabled, the EMAC checks the address
and, in the following cases, the frame is written to system memory:
• if it matches one of the four specific address registers.
• if it matches the hash address function.
• if it is a broadcast address (0xFFFFFFFFFFFF) and broadcasts are allowed.
• if the EMAC is configured to copy all frames.
The register receive buffer queue pointer points to the next entry (see Table 35-1 on page 710)
and the EMAC uses this as the address in system memory to write the frame to. Once the frame
has been completely and successfully received and written to system memory, the EMAC then
updates the receive buffer descriptor entry with the reason for the address match and marks the
area as being owned by software. Once this is complete an interrupt receive complete is set.
Software is then responsible for handling the data in the buffer and then releasing the buffer by
writing the ownership bit back to 0.
If the EMAC is unable to write the data at a rate to match the incoming frame, then an interrupt
receive overrun is set. If there is no receive buffer available, i.e., the next buffer is still owned by
software, the interrupt receive buffer not available is set. If the frame is not successfully
received, a statistic register is incremented and the frame is discarded without informing
software.
722 AT91SAM9G45
AT91SAM9G45
35.6 Ethernet MAC 10/100 (EMAC) User Interface

0x00 Network Control Register EMAC_NCR Read-write 0
0x04 Network Configuration Register EMAC_NCFG Read-write 0x800
0x08 Network Status Register EMAC_NSR Read-only -
0x0C Reserved
0x10 Reserved
0x14 Transmit Status Register EMAC_TSR Read-write 0x0000_0000
0x18 Receive Buffer Queue Pointer Register EMAC_RBQP Read-write 0x0000_0000
0x1C Transmit Buffer Queue Pointer Register EMAC_TBQP Read-write 0x0000_0000
0x20 Receive Status Register EMAC_RSR Read-write 0x0000_0000
0x24 Interrupt Status Register EMAC_ISR Read-write 0x0000_0000
0x28 Interrupt Enable Register EMAC_IER Write-only -
0x2C Interrupt Disable Register EMAC_IDR Write-only -
0x30 Interrupt Mask Register EMAC_IMR Read-only 0x0000_3FFF
0x34 Phy Maintenance Register EMAC_MAN Read-write 0x0000_0000
0x38 Pause Time Register EMAC_PTR Read-write 0x0000_0000
0x3C Pause Frames Received Register EMAC_PFR Read-write 0x0000_0000
0x40 Frames Transmitted Ok Register EMAC_FTO Read-write 0x0000_0000
0x44 Single Collision Frames Register EMAC_SCF Read-write 0x0000_0000
0x48 Multiple Collision Frames Register EMAC_MCF Read-write 0x0000_0000
0x4C Frames Received Ok Register EMAC_FRO Read-write 0x0000_0000
0x50 Frame Check Sequence Errors Register EMAC_FCSE Read-write 0x0000_0000
0x54 Alignment Errors Register EMAC_ALE Read-write 0x0000_0000
0x58 Deferred Transmission Frames Register EMAC_DTF Read-write 0x0000_0000
0x5C Late Collisions Register EMAC_LCOL Read-write 0x0000_0000
0x60 Excessive Collisions Register EMAC_ECOL Read-write 0x0000_0000
0x64 Transmit Underrun Errors Register EMAC_TUND Read-write 0x0000_0000
0x68 Carrier Sense Errors Register EMAC_CSE Read-write 0x0000_0000
0x6C Receive Resource Errors Register EMAC_RRE Read-write 0x0000_0000
0x70 Receive Overrun Errors Register EMAC_ROV Read-write 0x0000_0000
0x74 Receive Symbol Errors Register EMAC_RSE Read-write 0x0000_0000
0x78 Excessive Length Errors Register EMAC_ELE Read-write 0x0000_0000
0x7C Receive Jabbers Register EMAC_RJA Read-write 0x0000_0000
0x80 Undersize Frames Register EMAC_USF Read-write 0x0000_0000
0x84 SQE Test Errors Register EMAC_STE Read-write 0x0000_0000
0x88 Received Length Field Mismatch Register EMAC_RLE Read-write 0x0000_0000
723
0x90 Hash Register Bottom [31:0] Register EMAC_HRB Read-write 0x0000_0000
0x94 Hash Register Top [63:32] Register EMAC_HRT Read-write 0x0000_0000
0x98 Specific Address 1 Bottom Register EMAC_SA1B Read-write 0x0000_0000
0x9C Specific Address 1 Top Register EMAC_SA1T Read-write 0x0000_0000
0xA0 Specific Address 2 Bottom Register EMAC_SA2B Read-write 0x0000_0000
0xA4 Specific Address 2 Top Register EMAC_SA2T Read-write 0x0000_0000
0xA8 Specific Address 3 Bottom Register EMAC_SA3B Read-write 0x0000_0000
0xAC Specific Address 3 Top Register EMAC_SA3T Read-write 0x0000_0000
0xB0 Specific Address 4 Bottom Register EMAC_SA4B Read-write 0x0000_0000
0xB4 Specific Address 4 Top Register EMAC_SA4T Read-write 0x0000_0000
0xB8 Type ID Checking Register EMAC_TID Read-write 0x0000_0000
0xC0 User Input/Output Register EMAC_USRIO Read-write 0x0000_0000
0xC8 - 0xFC Reserved – – –
724 AT91SAM9G45
AT91SAM9G45
35.6.1 Network Control Register

Name: EMAC_NCR
Address: 0xFFFBC000
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – THALT TSTART BP
7 6 5 4 3 2 1 0
WESTAT INCSTAT CLRSTAT MPE TE RE LLB LB
• LB: LoopBack
Asserts the loopback signal to the PHY.
• LLB: Loopback local

Connects txd to rxd, tx_en to rx_dv, forces full duplex and drives rx_clk and tx_clk with pclk divided by 4.
rx_clk and tx_clk may glitch as the EMAC is switched into and out of internal loop back. It is important that receive
and transmit circuits have already been disabled when making the switch into and out of internal loop back.
• RE: Receive enable

When set, enables the EMAC to receive data. When reset, frame reception stops immediately and the receive FIFO is
cleared. The receive queue pointer register is unaffected.
• TE: Transmit enable

When set, enables the Ethernet transmitter to send data. When reset transmission, stops immediately, the transmit FIFO
and control registers are cleared and the transmit queue pointer register resets to point to the start of the transmit descrip-
tor list.
• MPE: Management port enable

Set to one to enable the management port. When zero, forces MDIO to high impedance state and MDC low.
• CLRSTAT: Clear statistics registers

This bit is write only. Writing a one clears the statistics registers.
• INCSTAT: Increment statistics registers

This bit is write only. Writing a one increments all the statistics registers by one for test purposes.
• WESTAT: Write enable for statistics registers

Setting this bit to one makes the statistics registers writable for functional test purposes.
• BP: Back pressure

If set in half duplex mode, forces collisions on all received frames.
725
• TSTART: Start transmission
Writing one to this bit starts transmission.
• THALT: Transmit halt

Writing one to this bit halts transmission as soon as any ongoing frame transmission ends.
726 AT91SAM9G45
AT91SAM9G45
35.6.2 Network Configuration Register

Name: EMAC_NCFG
Address: 0xFFFBC004
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – IRXFCS EFRHD DRFCS RLCE
15 14 13 12 11 10 9 8
RBOF PAE RTY CLK – BIG
7 6 5 4 3 2 1 0
UNI MTI NBC CAF JFRAME – FD SPD
• SPD: Speed
Set to 1 to indicate 100 Mbit/s operation, 0 for 10 Mbit/s. The value of this pin is reflected on the speed pin.
• FD: Full Duplex

If set to 1, the transmit block ignores the state of collision and carrier sense and allows receive while transmitting. Also con-
trols the half_duplex pin.
• CAF: Copy All Frames

When set to 1, all valid frames are received.
• JFRAME: Jumbo Frames

Set to one to enable jumbo frames of up to 10240 bytes to be accepted.
• NBC: No Broadcast
When set to 1, frames addressed to the broadcast address of all ones are not received.
• MTI: Multicast Hash Enable

When set, multicast frames are received when the 6-bit hash function of the destination address points to a bit that is set in
the hash register.
• UNI: Unicast Hash Enable

When set, unicast frames are received when the 6-bit hash function of the destination address points to a bit that is set in
the hash register.
• BIG: Receive 1536 bytes frames

Setting this bit means the EMAC receives frames up to 1536 bytes in length. Normally, the EMAC would reject any frame
above 1518 bytes.
• CLK: MDC clock divider

Set according to system clock speed. This determines by what number system clock is divided to generate MDC. For con-
formance with 802.3, MDC must not exceed 2.5MHz (MDC is only active during MDIO read and write operations)
727
.
CLK MDC
00 MCK divided by 8 (MCK up to 20 MHz)
• RTY: Retry test

Must be set to zero for normal operation. If set to one, the back off between collisions is always one slot time. Setting this
bit to one helps testing the too many retries condition. Also used in the pause frame tests to reduce the pause counters
decrement time from 512 bit times, to every rx_clk cycle.
• PAE: Pause Enable

When set, transmission pauses when a valid pause frame is received.
• RBOF: Receive Buffer Offset

Indicates the number of bytes by which the received data is offset from the start of the first receive buffer.
RBOF Offset
00 No offset from start of receive buffer
01 One-byte offset from start of receive buffer
10 Two-byte offset from start of receive buffer
11 Three-byte offset from start of receive buffer
• RLCE: Receive Length field Checking Enable

When set, frames with measured lengths shorter than their length fields are discarded. Frames containing a type ID in
bytes 13 and 14 — length/type ID = 0600 — are not be counted as length errors.
• DRFCS: Discard Receive FCS

When set, the FCS field of received frames are not be copied to memory.
• EFRHD:
Enable Frames to be received in half-duplex mode while transmitting.
• IRXFCS: Ignore RX FCS

When set, frames with FCS/CRC errors are not rejected and no FCS error statistics are counted. For normal operation, this
bit must be set to 0.
728 AT91SAM9G45
AT91SAM9G45
35.6.3 Network Status Register

Name: EMAC_NSR
Address: 0xFFFBC008
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – IDLE MDIO –
• MDIO
Returns status of the mdio_in pin. Use the PHY maintenance register for reading managed frames rather than this bit.
• IDLE
0 = The PHY logic is running.
1 = The PHY management logic is idle (i.e., has completed).
729
35.6.4 Transmit Status Register
Name: EMAC_TSR
Address: 0xFFFBC014
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– UND COMP BEX TGO RLE COL UBR
This register, when read, provides details of the status of a transmit. Once read, individual bits may be cleared by writing 1
to them. It is not possible to set a bit to 1 by writing to the register.
• UBR: Used Bit Read

Set when a transmit buffer descriptor is read with its used bit set. Cleared by writing a one to this bit.
• COL: Collision Occurred

Set by the assertion of collision. Cleared by writing a one to this bit.
• RLE: Retry Limit exceeded

Cleared by writing a one to this bit.
• TGO: Transmit Go
If high transmit is active.
• BEX: Buffers exhausted mid frame

If the buffers run out during transmission of a frame, then transmission stops, FCS shall be bad and tx_er asserted. Cleared
by writing a one to this bit.
• COMP: Transmit Complete

Set when a frame has been transmitted. Cleared by writing a one to this bit.
• UND: Transmit Underrun

Set when transmit DMA was not able to read data from memory, either because the bus was not granted in time, because
a not OK hresp(bus error) was returned or because a used bit was read midway through frame transmission. If this
occurs, the transmitter forces bad CRC. Cleared by writing a one to this bit.
730 AT91SAM9G45
AT91SAM9G45
35.6.5 Receive Buffer Queue Pointer Register

Name: EMAC_RBQP
Address: 0xFFFBC018
Access: Read-write
31 30 29 28 27 26 25 24
ADDR
23 22 21 20 19 18 17 16
ADDR
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR – –
This register points to the entry in the receive buffer queue (descriptor list) currently being used. It is written with the start
location of the receive buffer descriptor list. The lower order bits increment as buffers are used up and wrap to their original
values after either 1024 buffers or when the wrap bit of the entry is set.
Reading this register returns the location of the descriptor currently being accessed. This value increments as buffers are
used. Software should not use this register for determining where to remove received frames from the queue as it con-
stantly changes as new frames are received. Software should instead work its way through the buffer descriptor queue
checking the used bits.
Receive buffer writes also comprise bursts of two words and, as with transmit buffer reads, it is recommended that bit 2 is
always written with zero to prevent a burst crossing a 1K boundary, in violation of section 3.6 of the AMBA specification.
• ADDR: Receive buffer queue pointer address

Written with the address of the start of the receive queue, reads as a pointer to the current buffer being used.
731
35.6.6 Transmit Buffer Queue Pointer Register
Name: EMAC_TBQP
Address: 0xFFFBC01C
Access: Read-write
31 30 29 28 27 26 25 24
ADDR
23 22 21 20 19 18 17 16
ADDR
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR – –
This register points to the entry in the transmit buffer queue (descriptor list) currently being used. It is written with the start
location of the transmit buffer descriptor list. The lower order bits increment as buffers are used up and wrap to their original
values after either 1024 buffers or when the wrap bit of the entry is set. This register can only be written when bit 3 in the
transmit status register is low.
As transmit buffer reads consist of bursts of two words, it is recommended that bit 2 is always written with zero to prevent a
burst crossing a 1K boundary, in violation of section 3.6 of the AMBA specification.
• ADDR: Transmit buffer queue pointer address

Written with the address of the start of the transmit queue, reads as a pointer to the first buffer of the frame being transmit-
ted or about to be transmitted.
732 AT91SAM9G45
AT91SAM9G45
35.6.7 Receive Status Register

Name: EMAC_RSR
Address: 0xFFFBC020
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – OVR REC BNA
This register, when read, provides details of the status of a receive. Once read, individual bits may be cleared by writing 1
to them. It is not possible to set a bit to 1 by writing to the register.
• BNA: Buffer Not Available

An attempt was made to get a new buffer and the pointer indicated that it was owned by the processor. The DMA rereads
the pointer each time a new frame starts until a valid pointer is found. This bit is set at each attempt that fails even if it has
not had a successful pointer read since it has been cleared.
• REC: Frame Received

One or more frames have been received and placed in memory. Cleared by writing a one to this bit.
• OVR: Receive Overrun

The DMA block was unable to store the receive frame to memory, either because the bus was not granted in time or
because a not OK hresp(bus error) was returned. The buffer is recovered if this happens.
733
35.6.8 Interrupt Status Register
Name: EMAC_ISR
Address: 0xFFFBC024
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – PTZ PFR HRESP ROVR – –
7 6 5 4 3 2 1 0
TCOMP TXERR RLE TUND TXUBR RXUBR RCOMP MFD
• MFD: Management Frame Done

The PHY maintenance register has completed its operation. Cleared on read.
• RCOMP: Receive Complete

A frame has been stored in memory. Cleared on read.
• RXUBR: Receive Used Bit Read

Set when a receive buffer descriptor is read with its used bit set. Cleared on read.
• TXUBR: Transmit Used Bit Read

Set when a transmit buffer descriptor is read with its used bit set. Cleared on read.
• TUND: Ethernet Transmit Buffer Underrun

The transmit DMA did not fetch frame data in time for it to be transmitted or hresp returned not OK. Also set if a used bit
is read mid-frame or when a new transmit queue pointer is written. Cleared on read.
• RLE: Retry Limit Exceeded

Cleared on read.
• TXERR: Transmit Error

Transmit buffers exhausted in mid-frame - transmit error. Cleared on read.
• TCOMP: Transmit Complete

Set when a frame has been transmitted. Cleared on read.
• ROVR: Receive Overrun

Set when the receive overrun status bit gets set. Cleared on read.
• HRESP: Hresp not OK

Set when the DMA block sees a bus error. Cleared on read.
• PFR: Pause Frame Received

Indicates a valid pause has been received. Cleared on a read.
• PTZ: Pause Time Zero

Set when the pause time register, 0x38 decrements to zero. Cleared on a read.
734 AT91SAM9G45
AT91SAM9G45
35.6.9 Interrupt Enable Register

Name: EMAC_IER
Address: 0xFFFBC028
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• MFD: Management Frame sent

Enable management done interrupt.

Enable receive complete interrupt.

Enable receive used bit read interrupt.

Enable transmit used bit read interrupt.

Enable transmit underrun interrupt.

Enable retry limit exceeded interrupt.
• TXERR
Enable transmit buffers exhausted in mid-frame interrupt.

Enable transmit complete interrupt.

Enable receive overrun interrupt.

Enable Hresp not OK interrupt.

Enable pause frame received interrupt.

Enable pause time zero interrupt.
735
35.6.10 Interrupt Disable Register
Name: EMAC_IDR
Address: 0xFFFBC02C
Access: Write-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0

Disable management done interrupt.

Disable receive complete interrupt.

Disable receive used bit read interrupt.

Disable transmit used bit read interrupt.

Disable transmit underrun interrupt.

Disable retry limit exceeded interrupt.
• TXERR
Disable transmit buffers exhausted in mid-frame interrupt.

Disable transmit complete interrupt.

Disable receive overrun interrupt.

Disable Hresp not OK interrupt.

Disable pause frame received interrupt.

Disable pause time zero interrupt.
736 AT91SAM9G45
AT91SAM9G45
35.6.11 Interrupt Mask Register

Name: EMAC_IMR
Address: 0xFFFBC030
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0

Management done interrupt masked.

Receive complete interrupt masked.

Receive used bit read interrupt masked.

Transmit used bit read interrupt masked.

Transmit underrun interrupt masked.

Retry limit exceeded interrupt masked.
• TXERR
Transmit buffers exhausted in mid-frame interrupt masked.

Transmit complete interrupt masked.

Receive overrun interrupt masked.

Hresp not OK interrupt masked.

Pause frame received interrupt masked.

Pause time zero interrupt masked.
737
35.6.12 PHY Maintenance Register
Name: EMAC_MAN
Address: 0xFFFBC034
Access: Read-write
31 30 29 28 27 26 25 24
SOF RW PHYA
23 22 21 20 19 18 17 16
PHYA REGA CODE
15 14 13 12 11 10 9 8
DATA
7 6 5 4 3 2 1 0
DATA
• DATA
For a write operation this is written with the data to be written to the PHY.
After a read operation this contains the data read from the PHY.
• CODE:
Must be written to 10. Reads as written.
• REGA: Register Address

Specifies the register in the PHY to access.
• PHYA: PHY Address
• RW: Read-write
10 is read; 01 is write. Any other value is an invalid PHY management frame
• SOF: Start of frame

Must be written 01 for a valid frame.
738 AT91SAM9G45
AT91SAM9G45
35.6.13 Pause Time Register

Name: EMAC_PTR
Address: 0xFFFBC038
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
PTIME
7 6 5 4 3 2 1 0
PTIME
• PTIME: Pause Time

Stores the current value of the pause time register which is decremented every 512 bit times.
739
35.6.14 Hash Register Bottom
Name: EMAC_HRB
Address: 0xFFFBC090
Access: Read-write
31 30 29 28 27 26 25 24
ADDR
23 22 21 20 19 18 17 16
ADDR
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR:
Bits 31:0 of the hash address register. See “Hash Addressing” on page 717.
35.6.15 Hash Register Top

Name: EMAC_HRT
Address: 0xFFFBC094
Access: Read-write
31 30 29 28 27 26 25 24
ADDR
23 22 21 20 19 18 17 16
ADDR
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR:
Bits 63:32 of the hash address register. See “Hash Addressing” on page 717.
740 AT91SAM9G45
AT91SAM9G45
35.6.16 Specific Address 1 Bottom Register

Name: EMAC_SA1B
Address: 0xFFFBC098
Access: Read-write
31 30 29 28 27 26 25 24
ADDR
23 22 21 20 19 18 17 16
ADDR
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR
Least significant bits of the destination address. Bit zero indicates whether the address is multicast or unicast and corre-
sponds to the least significant bit of the first byte received.
35.6.17 Specific Address 1 Top Register

Name: EMAC_SA1T
Address: 0xFFFBC09C
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR
The most significant bits of the destination address, that is bits 47 to 32.
741
Name: EMAC_SA2B
Address: 0xFFFBC0A0
Access: Read-write
31 30 29 28 27 26 25 24
ADDR
23 22 21 20 19 18 17 16
ADDR
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR

Name: EMAC_SA2T
Address: 0xFFFBC0A4
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR
742 AT91SAM9G45
AT91SAM9G45

Name: EMAC_SA3B
Address: 0xFFFBC0A8
Access: Read-write
31 30 29 28 27 26 25 24
ADDR
23 22 21 20 19 18 17 16
ADDR
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR

Name: EMAC_SA3T
Address: 0xFFFBC0AC
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR
743
Name: EMAC_SA4B
Address: 0xFFFBC0B0
Access: Read-write
31 30 29 28 27 26 25 24
ADDR
23 22 21 20 19 18 17 16
ADDR
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR

Name: EMAC_SA4T
Address: 0xFFFBC0B4
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
ADDR
7 6 5 4 3 2 1 0
ADDR
• ADDR
744 AT91SAM9G45
AT91SAM9G45
35.6.24 Type ID Checking Register

Name: EMAC_TID
Address: 0xFFFBC0B8
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
TID
7 6 5 4 3 2 1 0
TID
• TID: Type ID checking

For use in comparisons with received frames TypeID/Length field.
745
35.6.25 User Input/Output Register
Name: EMAC_USRIO
Address: 0xFFFBC0C0
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – CLKEN RMII
• RMII
When set, this bit enables the RMII operation mode. When reset, it selects the MII mode.
• CLKEN
When set, this bit enables the transceiver input clock.
Setting this bit to 0 reduces power consumption when the treasurer is not used.
746 AT91SAM9G45
AT91SAM9G45
35.6.26 EMAC Statistic Registers

These registers reset to zero on a read and stick at all ones when they count to their maximum value. They should be read
frequently enough to prevent loss of data. The receive statistics registers are only incremented when the receive enable bit
is set in the network control register. To write to these registers, bit 7 must be set in the network control register. The statis-
tics register block contains the following registers.
35.6.26.1 Pause Frames Received Register

Name: EMAC_PFR
Address: 0xFFFBC03C
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
FROK
7 6 5 4 3 2 1 0
FROK
• FROK: Pause Frames received OK

A 16-bit register counting the number of good pause frames received. A good frame has a length of 64 to 1518 (1536 if bit
8 set in network configuration register) and has no FCS, alignment or receive symbol errors.
35.6.26.2 Frames Transmitted OK Register

Name: EMAC_FTO
Address: 0xFFFBC040
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
FTOK
15 14 13 12 11 10 9 8
FTOK
7 6 5 4 3 2 1 0
FTOK
• FTOK: Frames Transmitted OK

A 24-bit register counting the number of frames successfully transmitted, i.e., no underrun and not too many retries.
747
35.6.26.3 Single Collision Frames Register
Name: EMAC_SCF
Address: 0xFFFBC044
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
SCF
7 6 5 4 3 2 1 0
SCF
• SCF: Single Collision Frames

A 16-bit register counting the number of frames experiencing a single collision before being successfully transmitted, i.e.,
no underrun.
35.6.26.4 Multicollision Frames Register

Name: EMAC_MCF
Address: 0xFFFBC048
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
MCF
7 6 5 4 3 2 1 0
MCF
• MCF: Multicollision Frames

A 16-bit register counting the number of frames experiencing between two and fifteen collisions prior to being successfully
transmitted, i.e., no underrun and not too many retries.
748 AT91SAM9G45
AT91SAM9G45
35.6.26.5 Frames Received OK Register

Name: EMAC_FRO
Address: 0xFFFBC04C
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
FROK
15 14 13 12 11 10 9 8
FROK
7 6 5 4 3 2 1 0
FROK
• FROK: Frames Received OK

A 24-bit register counting the number of good frames received, i.e., address recognized and successfully copied to mem-
ory. A good frame is of length 64 to 1518 bytes (1536 if bit 8 set in network configuration register) and has no FCS,
alignment or receive symbol errors.
35.6.26.6 Frames Check Sequence Errors Register

Name: EMAC_FCSE
Address: 0xFFFBC050
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
FCSE
• FCSE: Frame Check Sequence Errors

An 8-bit register counting frames that are an integral number of bytes, have bad CRC and are between 64 and 1518 bytes
in length (1536 if bit 8 set in network configuration register). This register is also incremented if a symbol error is detected
and the frame is of valid length and has an integral number of bytes.
749
35.6.26.7 Alignment Errors Register
Name: EMAC_ALE
Address: 0xFFFBC054
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
ALE
• ALE: Alignment Errors

An 8-bit register counting frames that are not an integral number of bytes long and have bad CRC when their length is trun-
cated to an integral number of bytes and are between 64 and 1518 bytes in length (1536 if bit 8 set in network configuration
register). This register is also incremented if a symbol error is detected and the frame is of valid length and does not have
an integral number of bytes.
35.6.26.8 Deferred Transmission Frames Register

Name: EMAC_DTF
Address: 0xFFFBC058
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
DTF
7 6 5 4 3 2 1 0
DTF
• DTF: Deferred Transmission Frames

A 16-bit register counting the number of frames experiencing deferral due to carrier sense being active on their first attempt
at transmission. Frames involved in any collision are not counted nor are frames that experienced a transmit underrun.
750 AT91SAM9G45
AT91SAM9G45
35.6.26.9 Late Collisions Register

Name: EMAC_LCOL
Address: 0xFFFBC05C
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
LCOL
• LCOL: Late Collisions

An 8-bit register counting the number of frames that experience a collision after the slot time (512 bits) has expired. A late
collision is counted twice; i.e., both as a collision and a late collision.
35.6.26.10 Excessive Collisions Register

Name: EMAC_ECOL
Address: 0xFFFBC060
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
EXCOL
• EXCOL: Excessive Collisions

An 8-bit register counting the number of frames that failed to be transmitted because they experienced 16 collisions.
751
35.6.26.11 Transmit Underrun Errors Register
Name: EMAC_TUND
Address: 0xFFFBC064
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
TUND
• TUND: Transmit Underruns

An 8-bit register counting the number of frames not transmitted due to a transmit DMA underrun. If this register is incre-
mented, then no other statistics register is incremented.
35.6.26.12 Carrier Sense Errors Register

Name: EMAC_CSE
Address: 0xFFFBC068
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
CSE
• CSE: Carrier Sense Errors

An 8-bit register counting the number of frames transmitted where carrier sense was not seen during transmission or where
carrier sense was deasserted after being asserted in a transmit frame without collision (no underrun). Only incremented in
half-duplex mode. The only effect of a carrier sense error is to increment this register. The behavior of the other statistics
registers is unaffected by the detection of a carrier sense error.
752 AT91SAM9G45
AT91SAM9G45
35.6.26.13 Receive Resource Errors Register

Name: EMAC_RRE
Address: 0xFFFBC06C
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RRE
7 6 5 4 3 2 1 0
RRE
• RRE: Receive Resource Errors

A 16-bit register counting the number of frames that were address matched but could not be copied to memory because no
receive buffer was available.
35.6.26.14 Receive Overrun Errors Register

Name: EMAC_ROV
Address: 0xFFFBC070
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
ROVR

An 8-bit register counting the number of frames that are address recognized but were not copied to memory due to a
receive DMA overrun.
753
35.6.26.15 Receive Symbol Errors Register
Name: EMAC_RSE
Address: 0xFFFBC074
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
RSE
• RSE: Receive Symbol Errors

An 8-bit register counting the number of frames that had rx_er asserted during reception. Receive symbol errors are also
counted as an FCS or alignment error if the frame is between 64 and 1518 bytes in length (1536 if bit 8 is set in the network
configuration register). If the frame is larger, it is recorded as a jabber error.
35.6.26.16 Excessive Length Errors Register

Name: EMAC_ELE
Address: 0xFFFBC078
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
EXL
• EXL: Excessive Length Errors

An 8-bit register counting the number of frames received exceeding 1518 bytes (1536 if bit 8 set in network configuration
register) in length but do not have either a CRC error, an alignment error nor a receive symbol error.
754 AT91SAM9G45
AT91SAM9G45
35.6.26.17 Receive Jabbers Register

Name: EMAC_RJA
Address: 0xFFFBC07C
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
RJB
• RJB: Receive Jabbers

An 8-bit register counting the number of frames received exceeding 1518 bytes (1536 if bit 8 set in network configuration
register) in length and have either a CRC error, an alignment error or a receive symbol error.
35.6.26.18 Undersize Frames Register

Name: EMAC_USF
Address: 0xFFFBC080
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
USF
• USF: Undersize frames

An 8-bit register counting the number of frames received less than 64 bytes in length but do not have either a CRC error, an
alignment error or a receive symbol error.
755
35.6.26.19 SQE Test Errors Register
Name: EMAC_STE
Address: 0xFFFBC084
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
SQER
• SQER: SQE test errors

An 8-bit register counting the number of frames where col was not asserted within 96 bit times (an interframe gap) of
tx_en being deasserted in half duplex mode.
35.6.26.20 Received Length Field Mismatch Register

Name: EMAC_RLE
Address: 0xFFFBC088
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
RLFM
• RLFM: Receive Length Field Mismatch

An 8-bit register counting the number of frames received that have a measured length shorter than that extracted from its
length field. Checking is enabled through bit 16 of the network configuration register. Frames containing a type ID in bytes
13 and 14 (i.e., length/type ID ≥ 0x0600) are not counted as length field errors, neither are excessive length frames.
756 AT91SAM9G45
AT91SAM9G45
36. USB Host Port (UHP)
36.1 Description
The USB Host Port (UHP) interfaces the USB with the host application. It handles Open HCI
protocol (Open Host Controller Interface) as well as Enhanced HCI protocol (Enhanced Host
Controller Interface).

The AT91SAM9G45 features USB communication ports as follows:
• 2 Ports USB Host full speed OHCI and High speed EHCI
• 1 Device High speed
USB Host Port A is directly connected to the first UTMI transceiver.
The Host Port B is multiplexed with the USB device High speed and connected to the second
UTMI port. The selection between Host Port B and USB device high speed is controlled by a the
UDPHS enable bit located in the UDPHS_CTRL control register.
Figure 36-1. USB Selection

HS HS
Transceiver Transceiver
EN_UDPHS
0 1
PA PB
HS
HS EHCI USB
FS OHCI
DMA DMA
• Compliant with Enhanced HCI Rev 1.0 Specification

– Compliant with USB V2.0 High-speed and Full-speed Specification
– Supports Both High-speed 480Mbps and Full-speed 12 Mbps USB devices
• Compliant with Open HCI Rev 1.0 Specification
– Compliant with USB V2.0 Full-speed and Low-speed Specification
– Supports Both Low-speed 1.5 Mbps and Full-speed 12 Mbps USB devices
• Root Hub Integrated with 2 Downstream USB Ports
• Shared Embedded USB Transceivers
36.2.1 EHCI
The USB Host Port controller is fully compliant with the Enhanced HCI specification. The USB
Host Port User Interface (registers description) can be found in the Enhanced HCI Rev 1.0
Specification available on http://www.intel.com/technology/usb/ehcispec.htm. The standard
EHCI USB stack driver can be easily ported to Atmel’s architecture in the same way all existing
class drivers run, without hardware specialization.
757
36.2.2 OHCI
The USB Host Port integrates a root hub and transceivers on downstream ports. It provides sev-
eral Full-speed half-duplex serial communication ports at a baud rate of 12 Mbit/s. Up to 127
USB devices (printer, camera, mouse, keyboard, disk, etc.) and the USB hub can be connected
to the USB host in the USB “tiered star” topology.
The USB Host Port controller is fully compliant with the Open HCI specification. The USB Host
Port User Interface (registers description) can be found in the Open HCI Rev 1.0 Specification
available on http://h18000.www1.hp.com/productinfo/development/openhci.html. The standard
OHCI USB stack driver can be easily ported to Atmel’s architecture, in the same way all existing
class drivers run without hardware specialization.
This means that all standard class devices are automatically detected and available to the user’s
application. As an example, integrating an HID (Human Interface Device) class driver provides a
plug & play feature for all USB keyboards and mouses.
36.3 Block Diagram
Root
HCI List Processor Hub Registers
Slave Block Block
OHCI Control ED & TD

AHB
Slave Registers Regsisters PORT S/M 1
Root Hub
and
Host SIE
PORT S/M 0 Embedded USB
HCI Data
AHB FIFO 64 x 8 v2.0 High-speed Transceiver
Master Block
Master HFSDPB
HFSDMB
USB transceiver
HHSDPB
HHSDMB
HFSDPA
HFSDMA
USB transceiver
SOF HHSDPA
generator HHSDMA
HCI
Slave Block Packet
Buffer
EHCI Control FIFO
AHB
Slave Registers
List
Processor
HCI Data
AHB
Master Block
Master
758 AT91SAM9G45
AT91SAM9G45
Access to the USB host operational registers is achieved through the AHB bus slave interface.
The Open HCI host controller and Enhanced HCI host controller initialize master DMA transfers
through the AHB bus master interface as follows:
• Fetches endpoint descriptors and transfer descriptors
• Access to endpoint data from system memory
• Access to the HC communication area
• Write status and retire transfer descriptor
Memory access errors (abort, misalignment) lead to an “Unrecoverable Error” indicated by the
corresponding flag in the host controller operational registers.
The USB root hub is integrated in the USB host. Several USB downstream ports are available.
The number of downstream ports can be determined by the software driver reading the root
hub’s operational registers. Device connection is automatically detected by the USB host port
logic.
USB physical transceivers are integrated in the product and driven by the root hub’s ports.
Over current protection on ports can be activated by the USB host controller. Atmel’s standard
product does not dedicate pads to external over current protection.
36.4.1 I/O Lines

HFSDPs, HFSDMs, HHSDPs and HHSDMs are not controlled by any PIO controllers. The
embedded USB High Speed physical transceivers are controlled by the USB host controller.
36.5 I/O Lines

HFSDPs, HFSDMs, HHSDPs and HHSDMs are not controlled by any PIO controllers. The
embedded USB High Speed physical transceivers are controlled by the USB host controller.
One transceiver is shared with USB Device (UDP) High Speed. In this case USB Host High
Speed Controller uses only Port A, ie, the signals HFSDPA, HFSDMA, HHSDPA and HHSDMA.
The port B is driven by the UDP High Speed, the output signals are DFSDP, DFSDM, DHSDP
and DHSDM.
The transceiver is automatically selected for Device operation once the UDP High Speed is
enabled.

The USB Host High Speed requires a 48 MHz clock for the embedded High-speed transceivers.
This clock is provided by the UTMI PLL, it is UPLLCK.
In case power consumption is saved by stopping the UTMI PLL, high-speed operations are not
possible. Nevertheless, OHCI Full-speed operations remain possible by selecting PLLACK as
the input clock of OHCI.
The High-speed transceiver returns a 30 MHz clock to the USB Host controller.
The USB Host controller requires 48 MHz and 12 MHz clocks for OHCI full-speed operations.
These clocks must be generated by a PLL with a correct accuracy of ± 0.25% thanks to USBDIV
field.
759
Thus the USB Host peripheral receives three clocks from the Power Management Controller
(PMC): the Peripheral Clock (MCK domain), the UHP48M and the UHP12M (built-in UHP48M
divided by four) used by the OHCI to interface with the bus USB signals (Recovered 12 MHz
domain) in Full-speed operations.
For High-speed operations, the user has to perform the following:
• Enable UHP peripheral clock, bit (1 << AT91C_ID_UHPHS) in PMC_PCER register.
• Write CKGR_PLLCOUNT field in PMC_UCKR register.
• Enable UPLL, bit AT91C_CKGR_UPLLEN in PMC_UCKR register.
• Wait until UTMI_PLL is locked. LOCKU bit in PMC_SR register
• Enable BIAS, bit AT91C_CKGR_BIASEN in PMC_UCKR register.
• Select UPLLCK as Input clock of OHCI part, USBS bit in PMC_USB register.
• Program the OHCI clocks (UHP48M and UHP12M) with USBDIV field in PMC_USB register.
USBDIV must be 9 (division by 10) if UPLLCK is selected.
• Enable OHCI clocks, UHP bit in PMC_SCER register.
For OHCI Full-speed operations only, the user has to perform the following:
• Enable UHP peripheral clock, bit (1 << AT91C_ID_UHPHS) in PMC_PCER register.
• Select PLLACK as Input clock of OHCI part, USBS bit in PMC_USB register.
• Program the OHCI clocks (UHP48M and UHP12M) with USBDIV field in PMC_USB register.
USBDIV value is to calculated regarding the PLLACK value and USB Full-speed accuracy.
• Enable the OHCI clocks, UHP bit in PMC_SCER register.
760 AT91SAM9G45
AT91SAM9G45
Figure 36-3. UHP Clock trees

UPLL (480 MHz)
EHCI 30 MHz
AHB Master UTMI transceiver
Interface
USB 2.0 EHCI Port
Host Controller Router
EHCI 30 MHz
User UTMI transceiver
Interface
MCK
OHCI
Master
Interface Root Hub
USB 1.1 OHCI and
Host Controller Host SIE
OHCI
User
Interface
UHP12M
UHP48M
OHCI clocks
36.5.2 Interrupt
The USB host interface has an interrupt line connected to the Advanced Interrupt Controller
(AIC).
Handling USB host interrupts requires programming the AIC before configuring the UHP HS.
761
36.6 Typical Connection
Figure 36-4. Board Schematic to Interface UHP High-speed Device Controller
PIO (VBUS DETECT)

"A" Receptacle
1 = VBUS HHSDM
2 = D-
3 = D+ 39 ± 5% Ω
(1)
15k Ω 4 = GND HFSDM
3 4
Shell = Shield
(1) HHSDP
22k Ω CRPB
1 2
39 ± 5% Ω
HFSDP
CRPB: 1µF to 10µF
6K8 ± 1% Ω
VBG
10 pF
GND
Note: 1. The values shown on the 22k Ω and 15k Ω resistors are only valid for 3v3 supplied PIOs.
762 AT91SAM9G45
AT91SAM9G45
37. USB High Speed Device Port (UDPHS)
37.1 Description
The USB High Speed Device Port (UDPHS) is compliant with the Universal Serial Bus (USB),
rev 2.0 High Speed device specification.
Each endpoint can be configured in one of several USB transfer types. It can be associated with
one, two or three banks of a dual-port RAM used to store the current data payload. If two or
three banks are used, one DPR bank is read or written by the processor, while the other is read
or written by the USB device peripheral. This feature is mandatory for isochronous endpoints.

The AT91SAM9G45 features USB communication ports as follows:
• 2 Ports USB Host full speed OHCI and High speed EHCI
• 1 Device High speed
USB Host Port A is directly connected to the first UTMI transceiver.
The Host Port B is multiplexed with the USB device High speed and connected to the second
UTMI port. The selection between Host Port B and USB device high speed is controlled by a the
UDPHS enable bit located in the UDPHS_CTRL control register.
Figure 37-1. USB Selection

HS HS
Transceiver Transceiver
EN_UDPHS
0 1
PA PB
HS
HS EHCI USB
FS OHCI
DMA DMA
• USB V2.0 high-speed compliant, 480 MBits per second

• Embedded USB V2.0 UTMI+ high-speed transceiver shared with UHP HS.
• Embedded 4-Kbyte dual-port RAM for endpoints
• Embedded 6 channels DMA controller
• Suspend/Resume logic
• Up to 2 or 3 banks for isochronous and bulk endpoints
• Seven endpoints:
– Endpoint 0: 64 bytes, 1 bank mode
– Endpoint 1 & 2: 1024 bytes, 2 banks mode, High Bandwidth, DMA
– Endpoint 3 & 4: 1024 bytes, 3 banks mode, DMA
– Endpoint 5 & 6: 1024 bytes, 3 banks mode, High Bandwidth, DMA
763
Table 37-1. UDPHS Endpoint Description
High
Endpoint # Mnemonic Nb Bank DMA BandWidth Max. Endpoint Size Endpoint Type
0 EPT_0 1 N N 64 Control
1 EPT_1 2 Y Y 1024 Ctrl/Bulk/Iso(1)/Interrupt
3 EPT_3 3 Y N 1024 Ctrl/Bulk/Iso(1)/Interrupt
4 EPT_4 3 Y N 1024 Ctrl/Bulk/Iso(1)/Interrupt
Note: 1. In Isochronous Mode (Iso), it is preferable that High Band Width capability is available.
The size of internal DPRAM is 4 KB.
Suspend and resume are automatically detected by the UDPHS device, which notifies the pro-
cessor by raising an interrupt.
764 AT91SAM9G45
37.3 Block Diagram
APB bus APB

Interface
ctrl
DHSDP
status
DHSDM
AHB1 Rd/Wr/Ready
USB2.0 UTMI DFSDP DP
AHB bus DMA
AHB0 CORE DFSDM DM
Master
APB bus
AHB
Multiplexeur
Slave
Local
AHB
Slave EPT
interface Alloc
32 bits 16/8 bits

DPRAM
System Clock USB Clock

Domain Domain
PMC
Notes: 1. System clock, bit (1 << AT91C_ID_UDPHS) in PMC_PCER register.

2. Enable UDPHS clock (peripheral clock) bit AT91C_CKGR_UPLLEN in PMC_UCKR register.
3. Enable BIAS bit AT91C_CKGR_BIASEN in PMC_UCKR register.
765 AT91SAM9G45
37.4 Typical Connection
Figure 37-3. Board Schematic
PIO (VBUS DETECT)

"B" Receptacle
1 = VBUS DHSDM
2 = D-
3 = D+ 39 ± 5% Ω
(1)
15k Ω 4 = GND DFSDM
1 2
Shell = Shield
(1) DHSDP
22k Ω CRPB 3 4
39 ± 5% Ω
DFSDP
CRPB:1µF to 10µF
6K8 ± 1% Ω
VBG
10 pF
GND
Notes: 1. The values shown on the 22kΩ and 15kΩ resistors are only valid with 3V3 supplied PIOs.
766 AT91SAM9G45
37.5.1 USB V2.0 High Speed Device Port Introduction

The USB V2.0 High Speed Device Port provides communication services between host and
attached USB devices. Each device is offered with a collection of communication flows (pipes)
associated with each endpoint. Software on the host communicates with a USB Device through
a set of communication flows.
37.5.2 USB V2.0 High Speed Transfer Types

A communication flow is carried over one of four transfer types defined by the USB device.
A device provides several logical communication pipes with the host. To each logical pipe is
associated an endpoint. Transfer through a pipe belongs to one of the four transfer types:
• Control Transfers: Used to configure a device at attach time and can be used for other device-
specific purposes, including control of other pipes on the device.
• Bulk Data Transfers: Generated or consumed in relatively large burst quantities and have
wide dynamic latitude in transmission constraints.
• Interrupt Data Transfers: Used for timely but reliable delivery of data, for example, characters
or coordinates with human-perceptible echo or feedback response characteristics.
• Isochronous Data Transfers: Occupy a prenegotiated amount of USB bandwidth with a
prenegotiated delivery latency. (Also called streaming real time transfers.)
As indicated below, transfers are sequential events carried out on the USB bus.
Endpoints must be configured according to the transfer type they handle.
Table 37-2. USB Communication Flow

Transfer Direction Bandwidth Endpoint Size Error Detection Retrying
Control Bidirectional Not guaranteed 8,16,32,64 Yes Automatic
Isochronous Unidirectional Guaranteed 8-1024 Yes No
Interrupt Unidirectional Not guaranteed 8-1024 Yes Yes
Bulk Unidirectional Not guaranteed 8-512 Yes Yes
37.5.3 USB Transfer Event Definitions

A transfer is composed of one or several transactions;
Table 37-3. USB Transfer Events
CONTROL Control Transfers (1) • Setup transaction →Data IN transactions →Status OUT transaction
(bidirectional) • Setup transaction →Data OUT transactions →Status IN transaction
• Setup transaction →Status IN transaction
Bulk IN Transfer • Data IN transaction →Data IN transaction
IN
Interrupt IN Transfer • Data IN transaction →Data IN transaction
(device toward host)
Isochronous IN Transfer (2) • Data IN transaction →Data IN transaction
767 AT91SAM9G45
Table 37-3. USB Transfer Events (Continued)
CONTROL Control Transfers (1) • Setup transaction →Data IN transactions →Status OUT transaction
(bidirectional) • Setup transaction →Data OUT transactions →Status IN transaction
• Setup transaction →Status IN transaction
Bulk OUT Transfer • Data OUT transaction →Data OUT transaction
OUT
Interrupt OUT Transfer • Data OUT transaction →Data OUT transaction
(host toward device)
Isochronous OUT Transfer (2) • Data OUT transaction →Data OUT transaction
Notes: 1. Control transfer must use endpoints with one bank and can be aborted using a stall handshake.
2. Isochronous transfers must use endpoints configured with two or three banks.
An endpoint handles all transactions related to the type of transfer for which it has been
configured.
37.5.4 USB V2.0 High Speed BUS Transactions

Each transfer results in one or more transactions over the USB bus.
There are five kinds of transactions flowing across the bus in packets:
1. Setup Transaction
2. Data IN Transaction
3. Data OUT Transaction
4. Status IN Transaction
5. Status OUT Transaction
Figure 37-4. Control Read and Write Sequences

Setup Stage Data Stage Status Stage
Control Write Setup TX Data OUT TX Data OUT TX Status IN TX
Setup Stage Data Stage Status Stage
Control Read Setup TX Data IN TX Data IN TX Status OUT TX
Setup Stage Status Stage
No Data
Control Setup TX Status IN TX
A status IN or OUT transaction is identical to a data IN or OUT transaction.
37.5.5 Endpoint Configuration

The endpoint 0 is always a control endpoint, it must be programmed and active in order to be
enabled when the End Of Reset interrupt occurs.
768 AT91SAM9G45
To configure the endpoints:
• Fill the configuration register (UDPHS_EPTCFG) with the endpoint size, direction (IN or
OUT), type (CTRL, Bulk, IT, ISO) and the number of banks.
• Fill the number of transactions (NB_TRANS) for isochronous endpoints.
Note: For control endpoints the direction has no effect.
• Verify that the EPT_MAPD flag is set. This flag is set if the endpoint size and the number of
banks are correct compared to the FIFO maximum capacity and the maximum number of
allowed banks.
• Configure control flags of the endpoint and enable it in UDPHS_EPTCTLENBx according to
“UDPHS Endpoint Control Register” on page 814.
Control endpoints can generate interrupts and use only 1 bank.
All endpoints (except endpoint 0) can be configured either as Bulk, Interrupt or Isochronous. See
Table 37-1. UDPHS Endpoint Description.
The maximum packet size they can accept corresponds to the maximum endpoint size.
Note: The endpoint size of 1024 is reserved for isochronous endpoints.
The size of the DPRAM is 4 KB. The DPR is shared by all active endpoints. The memory size
required by the active endpoints must not exceed the size of the DPRAM.
SIZE_DPRAM = SIZE _EPT0

+ NB_BANK_EPT1 x SIZE_EPT1
+... (refer to 37.6.11 UDPHS Endpoint Configuration Register)
If a user tries to configure endpoints with a size the sum of which is greater than the DPRAM,
then the EPT_MAPD is not set.
The application has access to the physical block of DPR reserved for the endpoint through a 64
KB logical address space.
The physical block of DPR allocated for the endpoint is remapped all along the 64 KB logical
address space. The application can write a 64 KB buffer linearly.
769 AT91SAM9G45
Figure 37-5. Logical Address Space for DPR Access:
DPR
x banks
Logical address 8 to 64 B
8 to 64 B
64 KB 8 to 64 B 8 to1024 B y banks
EP0 8 to 64 B
...
8 to1024 B z banks
64 KB 8 to1024 B 8 to1024 B
EP1 8 to1024 B
...
8 to1024 B
64 KB
EP2
64 KB
EP3
...
Configuration examples of UDPHS_EPTCTLx (UDPHS Endpoint Control Register) for Bulk IN

endpoint type follow below.
• With DMA
– AUTO_VALID: Automatically validate the packet and switch to the next bank.
– EPT_ENABL: Enable endpoint.
• Without DMA:
– TX_BK_RDY: An interrupt is generated after each transmission.
Configuration examples of Bulk OUT endpoint type follow below.
• With DMA
– AUTO_VALID: Automatically validate the packet and switch to the next bank.
• Without DMA
– RX_BK_RDY: An interrupt is sent after a new packet has been stored in the endpoint
FIFO.
770 AT91SAM9G45
37.5.6 Transfer With DMA
USB packets of any length may be transferred when required by the UDPHS Device. These
transfers always feature sequential addressing.
Packet data AHB bursts may be locked on a DMA buffer basis for drastic overall AHB bus band-
width performance boost with paged memories. These clock-cycle consuming memory row (or
bank) changes will then likely not occur, or occur only once instead of dozens times, during a
single big USB packet DMA transfer in case another AHB master addresses the memory. This
means up to 128-word single-cycle unbroken AHB bursts for Bulk endpoints and 256-word sin-
gle-cycle unbroken bursts for isochronous endpoints. This maximum burst length is then
controlled by the lowest programmed USB endpoint size (EPT_SIZE bit in the
UDPHS_EPTCFGx register) and DMA Size (BUFF_LENGTH bit in the
UDPHS_DMACONTROLx register).
The USB 2.0 device average throughput may be up to nearly 60 MBytes. Its internal slave aver-
age access latency decreases as burst length increases due to the 0 wait-state side effect of
unchanged endpoints. If at least 0 wait-state word burst capability is also provided by the exter-
nal DMA AHB bus slaves, each of both DMA AHB busses need less than 50% bandwidth
allocation for full USB 2.0 bandwidth usage at 30 MHz, and less than 25% at 60 MHz.
The UDPHS DMA Channel Transfer Descriptor is described in “UDPHS DMA Channel Transfer
Descriptor” on page 825.
Note: In case of debug, be careful to address the DMA to an SRAM address even if a remap is
done.
Figure 37-6. Example of DMA Chained List:

Transfer Descriptor
UDPHS Registers
(Current Transfer Descriptor) Next Descriptor Address
DMA Channel Address Transfer Descriptor

UDPHS Next Descriptor
DMA Channel Control Next Descriptor Address
DMA Channel Address DMA Channel Address Transfer Descriptor
DMA Channel Control Next Descriptor Address

DMA Channel Control
DMA Channel Address
DMA Channel Control

Null
Memory Area
Data Buff 1
Data Buff 2
Data Buff 3
771 AT91SAM9G45
37.5.7 Transfer Without DMA
Important. If the DMA is not to be used, it is necessary that it be disabled because otherwise it
can be enabled by previous versions of software without warning. If this should occur, the DMA
can process data before an interrupt without knowledge of the user.
The recommended means to disable DMA is as follows:
// Reset IP UDPHS
AT91C_BASE_UDPHS->UDPHS_CTRL &= ~AT91C_UDPHS_EN_UDPHS;
AT91C_BASE_UDPHS->UDPHS_CTRL |= AT91C_UDPHS_EN_UDPHS;
// With OR without DMA !!!
for( i=1; i<=((AT91C_BASE_UDPHS->UDPHS_IPFEATURES &
AT91C_UDPHS_DMA_CHANNEL_NBR)>>4); i++ ) {
// RESET endpoint canal DMA:
// DMA stop channel command
AT91C_BASE_UDPHS->UDPHS_DMA[i].UDPHS_DMACONTROL = 0; // STOP
command
// Disable endpoint
AT91C_BASE_UDPHS->UDPHS_EPT[i].UDPHS_EPTCTLDIS |= 0XFFFFFFFF;
// Reset endpoint config
AT91C_BASE_UDPHS->UDPHS_EPT[i].UDPHS_EPTCTLCFG = 0;
// Reset DMA channel (Buff count and Control field)
AT91C_BASE_UDPHS->UDPHS_DMA[i].UDPHS_DMACONTROL = 0x02; // NON
STOP command
// Reset DMA channel 0 (STOP)
AT91C_BASE_UDPHS->UDPHS_DMA[i].UDPHS_DMACONTROL = 0; // STOP
command
// Clear DMA channel status (read the register for clear it)
AT91C_BASE_UDPHS->UDPHS_DMA[i].UDPHS_DMASTATUS =
AT91C_BASE_UDPHS->UDPHS_DMA[i].UDPHS_DMASTATUS;
}
37.5.8 Handling Transactions with USB V2.0 Device Peripheral
37.5.8.1 Setup Transaction

The setup packet is valid in the DPR while RX_SETUP is set. Once RX_SETUP is cleared by
the application, the UDPHS accepts the next packets sent over the device endpoint.
When a valid setup packet is accepted by the UDPHS:
• the UDPHS device automatically acknowledges the setup packet (sends an ACK response)
• payload data is written in the endpoint
• sets the RX_SETUP interrupt
• the BYTE_COUNT field in the UDPHS_EPTSTAx register is updated
An endpoint interrupt is generated while RX_SETUP in the UDPHS_EPTSTAx register is not
cleared. This interrupt is carried out to the microcontroller if interrupts are enabled for this
endpoint.
772 AT91SAM9G45
Thus, firmware must detect RX_SETUP polling UDPHS_EPTSTAx or catching an interrupt, read
the setup packet in the FIFO, then clear the RX_SETUP bit in the UDPHS_EPTCLRSTA register
to acknowledge the setup stage.
If STALL_SNT was set to 1, then this bit is automatically reset when a setup token is detected by
the device. Then, the device still accepts the setup stage. (See Section 37.5.8.15 “STALL” on
page 784).
37.5.8.2 NYET
NYET is a High Speed only handshake. It is returned by a High Speed endpoint as part of the
PING protocol.
High Speed devices must support an improved NAK mechanism for Bulk OUT and control end-
points (except setup stage). This mechanism allows the device to tell the host whether it has
sufficient endpoint space for the next OUT transfer (see USB 2.0 spec 8.5.1 NAK Limiting via
Ping Flow Control).
The NYET/ACK response to a High Speed Bulk OUT transfer and the PING response are auto-
matically handled by hardware in the UDPHS_EPTCTLx register (except when the user wants to
force a NAK response by using the NYET_DIS bit).
If the endpoint responds instead to the OUT/DATA transaction with an NYET handshake, this
means that the endpoint accepted the data but does not have room for another data payload.
The host controller must return to using a PING token until the endpoint indicates it has space
available.
Figure 37-7. NYET Example with Two Endpoint Banks
data 0 ACK data 1 NYET PING ACK data 0 NYET PING NACK PING ACK
t=0 t = 125 μs t = 250 μs t = 375 μs t = 500 μs t = 625 μs E: empty

E': begin to empty
F: full
Bank 1 E Bank 1 F Bank 1 F Bank 1 F Bank 1 F Bank 1 E' Bank 1 E
Bank 0 F Bank 0 E' Bank 0 E Bank 0 E Bank 0 F Bank 0 F Bank 0 F
37.5.8.3 Data IN
37.5.8.4 Bulk IN or Interrupt IN

Data IN packets are sent by the device during the data or the status stage of a control transfer or
during an (interrupt/bulk/isochronous) IN transfer. Data buffers are sent packet by packet under
the control of the application or under the control of the DMA channel.
There are three ways for an application to transfer a buffer in several packets over the USB:
• packet by packet (see 37.5.8.5 below)
• 64 KB (see 37.5.8.5 below)
• DMA (see 37.5.8.6 below)
37.5.8.5 Bulk IN or Interrupt IN: Sending a Packet Under Application Control (Device to Host)
The application can write one or several banks.
773 AT91SAM9G45
A simple algorithm can be used by the application to send packets regardless of the number of
banks associated to the endpoint.
Algorithm Description for Each Packet:
• The application waits for TX_PK_RDY flag to be cleared in the UDPHS_EPTSTAx register
before it can perform a write access to the DPR.
• The application writes one USB packet of data in the DPR through the 64 KB endpoint logical
memory window.
• The application sets TX_PK_RDY flag in the UDPHS_EPTSETSTAx register.
The application is notified that it is possible to write a new packet to the DPR by the
TX_PK_RDY interrupt. This interrupt can be enabled or masked by setting the TX_PK_RDY bit
in the UDPHS_EPTCTLENB/UDPHS_EPTCTLDIS register.
Algorithm Description to Fill Several Packets:
Using the previous algorithm, the application is interrupted for each packet. It is possible to
reduce the application overhead by writing linearly several banks at the same time. The
AUTO_VALID bit in the UDPHS_EPTCTLx must be set by writing the AUTO_VALID bit in the
UDPHS_EPTCTLENBx register.
The auto-valid-bank mechanism allows the transfer of data (IN and OUT) without the interven-
tion of the CPU. This means that bank validation (set TX_PK_RDY or clear the RX_BK_RDY bit)
is done by hardware.
• The application checks the BUSY_BANK_STA field in the UDPHS_EPTSTAx register. The
application must wait that at least one bank is free.
• The application writes a number of bytes inferior to the number of free DPR banks for the
endpoint. Each time the application writes the last byte of a bank, the TX_PK_RDY signal is
automatically set by the UDPHS.
• If the last packet is incomplete (i.e., the last byte of the bank has not been written) the
application must set the TX_PK_RDY bit in the UDPHS_EPTSETSTAx register.
The application is notified that all banks are free, so that it is possible to write another burst of
packets by the BUSY_BANK interrupt. This interrupt can be enabled or masked by setting the
BUSY_BANK flag in the UDPHS_EPTCTLENB and UDPHS_EPTCTLDIS registers.
This algorithm must not be used for isochronous transfer. In this case, the ping-pong mechanism
does not operate.
A Zero Length Packet can be sent by setting just the TX_PKTRDY flag in the
UDPHS_EPTSETSTAx register.
37.5.8.6 Bulk IN or Interrupt IN: Sending a Buffer Using DMA (Device to Host)
The UDPHS integrates a DMA host controller. This DMA controller can be used to transfer a buf-
fer from the memory to the DPR or from the DPR to the processor memory under the UDPHS
control. The DMA can be used for all transfer types except control transfer.
Example DMA configuration:
1. Program UDPHS_DMAADDRESS x with the address of the buffer that should be
transferred.
2. Enable the interrupt of the DMA in UDPHS_IEN
3. Program UDPHS_ DMACONTROLx:
774 AT91SAM9G45
– Size of buffer to send: size of the buffer to be sent to the host.
– END_B_EN: The endpoint can validate the packet (according to the values
programmed in the AUTO_VALID and SHRT_PCKT fields of UDPHS_EPTCTLx.)
(See “UDPHS Endpoint Control Register” on page 814 and Figure 37-12. Autovalid
with DMA)
– END_BUFFIT: generate an interrupt when the BUFF_COUNT in
UDPHS_DMASTATUSx reaches 0.
– CHANN_ENB: Run and stop at end of buffer
The auto-valid-bank mechanism allows the transfer of data (IN & OUT) without the intervention
of the CPU. This means that bank validation (set TX_PK_RDY or clear the RX_BK_RDY bit) is
done by hardware.
A transfer descriptor can be used. Instead of programming the register directly, a descriptor
should be programmed and the address of this descriptor is then given to
UDPHS_DMANXTDSC to be processed after setting the LDNXT_DSC field (Load Next Descrip-
tor Now) in UDPHS_DMACONTROLx register.
The structure that defines this transfer descriptor must be aligned.
Each buffer to be transferred must be described by a DMA Transfer descriptor (see “UDPHS
DMA Channel Transfer Descriptor” on page 825). Transfer descriptors are chained. Before exe-
cuting transfer of the buffer, the UDPHS may fetch a new transfer descriptor from the memory
address pointed by the UDPHS_DMANXTDSCx register. Once the transfer is complete, the
transfer status is updated in the UDPHS_DMASTATUSx register.
To chain a new transfer descriptor with the current DMA transfer, the DMA channel must be
stopped. To do so, INTDIS_DMA and TX_BK_RDY may be set in the UDPHS_EPTCTLENBx
register. It is also possible for the application to wait for the completion of all transfers. In this
case the LDNXT_DSC field in the last transfer descriptor UDPHS_DMACONTROLx register
must be set to 0 and CHANN_ENB set to 1.
Then the application can chain a new transfer descriptor.
The INTDIS_DMA can be used to stop the current DMA transfer if an enabled interrupt is trig-
gered. This can be used to stop DMA transfers in case of errors.
The application can be notified at the end of any buffer transfer (ENB_BUFFIT bit in the
UDPHS_DMACONTROLx register).
775 AT91SAM9G45
Figure 37-8. Data IN Transfer for Endpoint with One Bank
Prevous Data IN TX Microcontroller Loads Data in FIFO Data is Sent on USB Bus
USB Bus
Packets Token IN Data IN 1 ACK Token IN NAK Token IN Data IN 2 ACK
TX_PK_RDY
Flag
(UDPHS_EPTSTAx) Set by firmware Cleared by hardware Set by the firmware Cleared by hardware
Interrupt Pending Interrupt Pending
TX_COMPLT Flag Payload in FIFO

(UDPHS_EPTSTAx)
Set by hardware Cleared by firmware Cleared by firmware
DPR access by firmware DPR access by hardware

FIFO
Data IN 1 Load in progress Data IN 2
Content
Figure 37-9. Data IN Transfer for Endpoint with Two Banks

Microcontroller Microcontroller Load Data IN Bank 1 Microcontroller Load Data IN Bank 0
Load Data IN Bank 0 UDPHS Device Send Bank 0 UDPHS Device Send Bank 1
USB Bus
Token IN Data IN ACK Token IN Data IN ACK
Packets
Set by Firmware, Cleared by Hardware

Data Payload Written switch to next bank
in FIFO Bank 0
Virtual TX_PK_RDY
bank 0
Cleared by Hardware
(UDPHS_EPTSTAx)
Data Payload Fully Transmitted
Virtual TX_PK_RDY
bank 1
(UDPHS_EPTSTAx) Set by Firmware,
Data Payload Written in FIFO Bank 1
Interrupt Pending
TX_COMPLT Set by Hardware Set by Hardware

Flag
(UDPHS_EPTSTAx) Interrupt Cleared by Firmware
FIFO
Written by Read by USB Device Written by
(DPR) Microcontroller Microcontroller
Bank 0
FIFO Written by
(DPR) Microcontroller Read by UDPHS Device
Bank1
776 AT91SAM9G45
Figure 37-10. Data IN Followed By Status OUT Transfer at the End of a Control Transfer
Device Sends the Last Device Sends a
Data Payload to Host Status OUT to Host
USB Bus
Packets Token IN Data IN ACK Token OUT Data OUT (ZLP) ACK Token OUT Data OUT (ZLP) ACK
Interrupt
Pending
RX_BK_RDY
(UDPHS_EPTSTAx) Set by Hardware
Cleared by Firmware
TX_COMPLT
(UDPHS_EPTSTAx)
Set by Hardware Cleared by Firmware
Note: A NAK handshake is always generated at the first status stage token.
Figure 37-11. Data OUT Followed by Status IN Transfer

Host Sends the Last Device Sends a Status IN
Data Payload to the Device to the Host
USB Bus
Token OUT Data OUT ACK Token IN Data IN ACK
Packets
Interrupt Pending
RX_BK_RDY
Cleared by Firmware
(UDPHS_EPTSTAx)
Set by Hardware
TX_PK_RDY
(UDPHS_EPTSTAx)
Set by Firmware Clear by Hardware
Note: Before proceeding to the status stage, the software should determine that there is no risk
of extra data from the host (data stage). If not certain (non-predictable data stage length), then
the software should wait for a NAK-IN interrupt before proceeding to the status stage. This pre-
caution should be taken to avoid collision in the FIFO.
777 AT91SAM9G45
Figure 37-12. Autovalid with DMA
Bank (system) Bank 0 Bank 1 Bank 0 Bank 1
Write write bank 0 write bank 1 write bank 0
bank 0 is full bank 1 is full bank 0 is full
Bank 1 Bank 0
IN data 0 IN data 1 IN data 0
Bank (usb) Bank 0 Bank 1 Bank 0 Bank 1
Virtual TX_PK_RDY Bank 0
Virtual TX_PK_RDY Bank 1
TX_PK_RDY
(Virtual 0 & Virtual 1)
Note: In the illustration above Autovalid validates a bank as full, although this might not be the case, in order to continue processing
data and to send to DMA.
37.5.8.7 Isochronous IN
Isochronous-IN is used to transmit a stream of data whose timing is implied by the delivery rate.
Isochronous transfer provides periodic, continuous communication between host and device.
It guarantees bandwidth and low latencies appropriate for telephony, audio, video, etc.
If the endpoint is not available (TX_PK_RDY = 0), then the device does not answer to the host.
An ERR_FL_ISO interrupt is generated in the UDPHS_EPTSTAx register and once enabled,
then sent to the CPU.
The STALL_SNT command bit is not used for an ISO-IN endpoint.
37.5.8.8 High Bandwidth Isochronous Endpoint Handling: IN Example

For high bandwidth isochronous endpoints, the DMA can be programmed with the number of
transactions (BUFF_LENGTH field in UDPHS_DMACONTROLx) and the system should provide
778 AT91SAM9G45
the required number of packets per microframe, otherwise, the host will notice a sequencing
problem.
A response should be made to the first token IN recognized inside a microframe under the fol-
lowing conditions:
• If at least one bank has been validated, the correct DATAx corresponding to the programmed
Number Of Transactions per Microframe (NB_TRANS) should be answered. In case of a
subsequent missed or corrupted token IN inside the microframe, the USB 2.0 Core available
data bank(s) that should normally have been transmitted during that microframe shall be
flushed at its end. If this flush occurs, an error condition is flagged (ERR_FLUSH is set in
UDPHS_EPTSTAx).
• If no bank is validated yet, the default DATA0 ZLP is answered and underflow is flagged
(ERR_FL_ISO is set in UDPHS_EPTSTAx). Then, no data bank is flushed at microframe
end.
• If no data bank has been validated at the time when a response should be made for the
second transaction of NB_TRANS = 3 transactions microframe, a DATA1 ZLP is answered
and underflow is flagged (ERR_FL_ISO is set in UDPHS_EPTSTAx). If and only if remaining
untransmitted banks for that microframe are available at its end, they are flushed and an error
condition is flagged (ERR_FLUSH is set in UDPHS_EPTSTAx).
• If no data bank has been validated at the time when a response should be made for the last
programmed transaction of a microframe, a DATA0 ZLP is answered and underflow is flagged
(ERR_FL_ISO is set in UDPHS_EPTSTAx). If and only if the remaining untransmitted data
bank for that microframe is available at its end, it is flushed and an error condition is flagged
(ERR_FLUSH is set in UDPHS_EPTSTAx).
• If at the end of a microframe no valid token IN has been recognized, no data bank is flushed
and no error condition is reported.
At the end of a microframe in which at least one data bank has been transmitted, if less than
NB_TRANS banks have been validated for that microframe, an error condition is flagged
(ERR_TRANS is set in UDPHS_EPTSTAx).
Cases of Error (in UDPHS_EPTSTAx)
• ERR_FL_ISO: There was no data to transmit inside a microframe, so a ZLP is answered by
default.
• ERR_FLUSH: At least one packet has been sent inside the microframe, but the number of
token IN received is lesser than the number of transactions actually validated (TX_BK_RDY)
and likewise with the NB_TRANS programmed.
• ERR_TRANS: At least one packet has been sent inside the microframe, but the number of
token IN received is lesser than the number of programmed NB_TRANS transactions and the
packets not requested were not validated.
• ERR_FL_ISO + ERR_FLUSH: At least one packet has been sent inside the microframe, but
the data has not been validated in time to answer one of the following token IN.
• ERR_FL_ISO + ERR_TRANS: At least one packet has been sent inside the microframe, but
the data has not been validated in time to answer one of the following token IN and the data
can be discarded at the microframe end.
• ERR_FLUSH + ERR_TRANS: The first token IN has been answered and it was the only one
received, a second bank has been validated but not the third, whereas NB_TRANS was
waiting for three transactions.
779 AT91SAM9G45
• ERR_FL_ISO + ERR_FLUSH + ERR_TRANS: The first token IN has been treated, the data
for the second Token IN was not available in time, but the second bank has been validated
before the end of the microframe. The third bank has not been validated, but three
transactions have been set in NB_TRANS.
37.5.8.9 Data OUT
37.5.8.10 Bulk OUT or Interrupt OUT

Like data IN, data OUT packets are sent by the host during the data or the status stage of con-
trol transfer or during an interrupt/bulk/isochronous OUT transfer. Data buffers are sent packet
by packet under the control of the application or under the control of the DMA channel.
37.5.8.11 Bulk OUT or Interrupt OUT: Receiving a Packet Under Application Control (Host to Device)
Algorithm Description for Each Packet:

• The application enables an interrupt on RX_BK_RDY.
• When an interrupt on RX_BK_RDY is received, the application knows that UDPHS_EPTSTAx
register BYTE_COUNT bytes have been received.
• The application reads the BYTE_COUNT bytes from the endpoint.
• The application clears RX_BK_RDY.
Note: If the application does not know the size of the transfer, it may not be a good option to use
AUTO_VALID. Because if a zero-length-packet is received, the RX_BK_RDY is automatically
cleared by the AUTO_VALID hardware and if the endpoint interrupt is triggered, the software will
not find its originating flag when reading the UDPHS_EPTSTAx register.
Algorithm to Fill Several Packets:
• The application enables the interrupts of BUSY_BANK and AUTO_VALID.
• When a BUSY_BANK interrupt is received, the application knows that all banks available for
the endpoint have been filled. Thus, the application can read all banks available.
If the application doesn’t know the size of the receive buffer, instead of using the BUSY_BANK
interrupt, the application must use RX_BK_RDY.
37.5.8.12 Bulk OUT or Interrupt OUT: Sending a Buffer Using DMA (Host To Device)
To use the DMA setting, the AUTO_VALID field is mandatory.
See 37.5.8.6 Bulk IN or Interrupt IN: Sending a Buffer Using DMA (Device to Host) for more
information.
DMA Configuration Example:
1. First program UDPHS_DMAADDRESSx with the address of the buffer that should be
transferred.
2. Enable the interrupt of the DMA in UDPHS_IEN
3. Program the DMA Channelx Control Register:
– Size of buffer to be sent.
– END_B_EN: Can be used for OUT packet truncation (discarding of unbuffered
packet data) at the end of DMA buffer.
780 AT91SAM9G45
– END_BUFFIT: Generate an interrupt when BUFF_COUNT in the
UDPHS_DMASTATUSx register reaches 0.
– END_TR_EN: End of transfer enable, the UDPHS device can put an end to the
current DMA transfer, in case of a short packet.
– END_TR_IT: End of transfer interrupt enable, an interrupt is sent after the last USB
packet has been transferred by the DMA, if the USB transfer ended with a short
packet. (Beneficial when the receive size is unknown.)
– CHANN_ENB: Run and stop at end of buffer.
For OUT transfer, the bank will be automatically cleared by hardware when the application has
read all the bytes in the bank (the bank is empty).
Note: When a zero-length-packet is received, RX_BK_RDY bit in UDPHS_EPTSTAx is cleared
automatically by AUTO_VALID, and the application knows of the end of buffer by the presence
of the END_TR_IT.
Note: If the host sends a zero-length packet, and the endpoint is free, then the device sends an
ACK. No data is written in the endpoint, the RX_BY_RDY interrupt is generated, and the
BYTE_COUNT field in UDPHS_EPTSTAx is null.
Figure 37-13. Data OUT Transfer for Endpoint with One Bank
Microcontroller Transfers Data
Host Sends Data Payload Host Sends the Next Data Payload Host Resends the Next Data Payload
USB Bus
Token OUT Data OUT 1 ACK Token OUT Data OUT 2 NAK Token OUT Data OUT 2 ACK
Packets
RX_BK_RDY Interrupt Pending

(UDPHS_EPTSTAx)
Set by Hardware Cleared by Firmware,
Data Payload Written in FIFO
FIFO (DPR)
Data OUT 1 Data OUT 1 Data OUT 2
Content
Written by UDPHS Device Microcontroller Read Written by UDPHS Device
781 AT91SAM9G45
Figure 37-14. Data OUT Transfer for an Endpoint with Two Banks
Microcontroller reads Data 1 in bank 0, Microcontroller reads Data 2 in bank 1,
Host sends first data payload Host sends second data payload Host sends third data payload
USB Bus
Packets Token OUT Data OUT 1 ACK Token OUT Data OUT 2 ACK Token OUT Data OUT 3
Virtual RX_BK_RDY Interrupt pending Cleared by Firmware

Bank 0
Set by Hardware,
Data payload written Set by Hardware
in FIFO endpoint bank 0 Cleared by Firmware
Data Payload written
Virtual RX_BK_RDY in FIFO endpoint bank 1 Interrupt pending
Bank 1
RX_BK_RDY = (virtual bank 0 | virtual bank 1)

(UDPHS_EPTSTAx)
FIFO (DPR)
Bank 0 Data OUT 1 Data OUT 1 Data OUT 3
Write by UDPHS Device Read by Microcontroller Write in progress
FIFO (DPR)
Bank 1 Data OUT 2 Data OUT 2
Write by Hardware Read by Microcontroller
37.5.8.13 High Bandwidth Isochronous Endpoint OUT
Figure 37-15. Bank Management, Example of Three Transactions per Microframe
USB bus
MDATA0 MDATA1 DATA2 MDATA0 MDATA1 DATA2
Transactions
t = 52.5 μs t = 125 μs USB line

t=0
(40% of 125 μs) RX_BK_RDY
RX_BK_RDY
Microcontroller FIFO
(DPR) Access Read Bank 1 Read Bank 2 Read Bank 3 Read Bank 1
USB 2.0 supports individual High Speed isochronous endpoints that require data rates up to 192
Mb/s (24 MB/s): 3x1024 data bytes per microframe.
To support such a rate, two or three banks may be used to buffer the three consecutive data
packets. The microcontroller (or the DMA) should be able to empty the banks very rapidly (at
least 24 MB/s on average).
NB_TRANS field in UDPHS_EPTCFGx register = Number Of Transactions per Microframe.
If NB_TRANS > 1 then it is High Bandwidth.
782 AT91SAM9G45
Example:
• If NB_TRANS = 3, the sequence should be either
– MData0
– MData0/Data1
– MData0/Data1/Data2
• If NB_TRANS = 2, the sequence should be either
– MData0
– MData0/Data1
• If NB_TRANS = 1, the sequence should be
– Data0
37.5.8.14 Isochronous Endpoint Handling: OUT Example

The user can ascertain the bank status (free or busy), and the toggle sequencing of the data
packet for each bank with the UDPHS_EPTSTAx register in the three bit fields as follows:
• TOGGLESQ_STA: PID of the data stored in the current bank
• CURRENT_BANK: Number of the bank currently being accessed by the microcontroller.
• BUSY_BANK_STA: Number of busy bank
This is particularly useful in case of a missing data packet.
If the inter-packet delay between the OUT token and the Data is greater than the USB standard,
then the ISO-OUT transaction is ignored. (Payload data is not written, no interrupt is generated
to the CPU.)
If there is a data CRC (Cyclic Redundancy Check) error, the payload is, none the less, written in
the endpoint. The ERR_CRISO flag is set in UDPHS_EPTSTAx register.
If the endpoint is already full, the packet is not written in the DPRAM. The ERR_FL_ISO flag is
set in UDPHS_EPTSTAx.
If the payload data is greater than the maximum size of the endpoint, then the ERR_OVFLW flag
is set. It is the task of the CPU to manage this error. The data packet is written in the endpoint
(except the extra data).
If the host sends a Zero Length Packet, and the endpoint is free, no data is written in the end-
point, the RX_BK_RDY flag is set, and the BYTE_COUNT field in UDPHS_EPTSTAx register is
null.
The FRCESTALL command bit is unused for an isochonous endpoint.
Otherwise, payload data is written in the endpoint, the RX_BK_RDY interrupt is generated and
the BYTE_COUNT in UDPHS_EPTSTAx register is updated.
783 AT91SAM9G45
37.5.8.15 STALL
STALL is returned by a function in response to an IN token or after the data phase of an OUT or
in response to a PING transaction. STALL indicates that a function is unable to transmit or
receive data, or that a control pipe request is not supported.
• OUT
To stall an endpoint, set the FRCESTALL bit in UDPHS_EPTSETSTAx register and after the
STALL_SNT flag has been set, set the TOGGLE_SEG bit in the UDPHS_EPTCLRSTAx
register.
• IN
Set the FRCESTALL bit in UDPHS_EPTSETSTAx register.
Figure 37-16. Stall Handshake Data OUT Transfer
USB Bus
Packets Token OUT Data OUT Stall PID
FRCESTALL
Set by Firmware Cleared by Firmware
Interrupt Pending
STALL_SNT
Figure 37-17. Stall Handshake Data IN Transfer
USB Bus Token IN Stall PID

Packets
FRCESTALL
Set by Firmware Cleared by Firmware
Interrupt Pending
STALL_SNT
784 AT91SAM9G45
37.5.9 Speed Identification
The high speed reset is managed by the hardware.
At the connection, the host makes a reset which could be a classic reset (full speed) or a high
speed reset.
At the end of the reset process (full or high), the ENDRESET interrupt is generated.
Then the CPU should read the SPEED bit in UDPHS_INTSTAx to ascertain the speed mode of
the device.
37.5.10 USB V2.0 High Speed Global Interrupt

Interrupts are defined in Section 37.6.3 ”UDPHS Interrupt Enable Register” (UDPHS_IEN) and
in Section 37.6.4 ”UDPHS Interrupt Status Register” (UDPHS_INTSTA).
37.5.11 Endpoint Interrupts

Interrupts are enabled in UDPHS_IEN (see Section 37.6.3 ”UDPHS Interrupt Enable Register”)
and individually masked in UDPHS_EPTCTLENBx (see Section 37.6.12 ”UDPHS Endpoint
Control Enable Register”).
.
Table 37-4. Endpoint Interrupt Source Masks

SHRT_PCKT Short Packet Interrupt
BUSY_BANK Busy Bank Interrupt
NAK_OUT NAKOUT Interrupt
NAK_IN/ERR_FLUSH NAKIN/Error Flush Interrupt
Stall Sent/CRC error/Number of Transaction
STALL_SNT/ERR_CRISO/ERR_NB_TRA
Error Interrupt
RX_SETUP/ERR_FL_ISO Received SETUP/Error Flow Interrupt
TX_PK_RD /ERR_TRANS TX Packet Read/Transaction Error Interrupt
TX_COMPLT Transmitted IN Data Complete Interrupt
RX_BK_RDY Received OUT Data Interrupt
ERR_OVFLW Overflow Error Interrupt
MDATA_RX MDATA Interrupt
DATAX_RX DATAx Interrupt
785 AT91SAM9G45
Figure 37-18. UDPHS Interrupt Control Interface
(UDPHS_IEN)
Global IT mask
DET_SUSPD Global IT sources
MICRO_SOF
INT_SOF
USB Global
IT Sources ENDRESET
WAKE_UP
ENDOFRSM
UPSTR_RES
(UDPHS_EPTCTLENBx)
SHRT_PCKT
EP mask (UDPHS_IEN)
BUSY_BANK EPT_0
EP sources
NAK_OUT
husb2dev
NAK_IN/ERR_FLUSH interrupt
STALL_SNT/ERR_CRISO/ERR_NBTRA
EPT0 IT RX_SETUP/ERR_FL_ISO
Sources
TX_BK_RDY/ERR_TRANS
TX_COMPLT
RX_BK_RDY
ERR_OVFLW
MDATA_RX
DATAX_RX
EP mask (UDPHS_IEN)
EPT_x
EP sources
EPT1-6 IT (UDPHS_EPTCTLx)
Sources INTDIS_DMA
disable DMA
channelx request
mask
(UDPHS_DMACONTROLx)
(UDPHS_IEN)
EN_BUFFIT DMA_x
mask
DMA CH x END_TR_IT
mask
DESC_LD_IT
786 AT91SAM9G45
37.5.12 Power Modes
37.5.12.1 Controlling Device States

A USB device has several possible states. Refer to Chapter 9 (USB Device Framework) of the
Universal Serial Bus Specification, Rev 2.0.
Figure 37-19. UDPHS Device State Diagram
Attached
Hub Reset Hub

or Configured
Deconfigured
Bus Inactive
Powered Suspended
Bus Activity
Power
Interruption Reset
Bus Inactive
Default Suspended
Reset Bus Activity

Address
Assigned
Bus Inactive
Address Suspended
Bus Activity
Device Device
Deconfigured Configured
Bus Inactive
Configured Suspended
Bus Activity
Movement from one state to another depends on the USB bus state or on standard requests
sent through control transactions via the default endpoint (endpoint 0).
After a period of bus inactivity, the USB device enters Suspend Mode. Accepting Sus-
pend/Resume requests from the USB host is mandatory. Constraints in Suspend Mode are very
strict for bus-powered applications; devices may not consume more than 500 μA on the USB
bus.
While in Suspend Mode, the host may wake up a device by sending a resume signal (bus activ-
ity) or a USB device may send a wake-up request to the host, e.g., waking up a PC by moving a
USB mouse.
The wake-up feature is not mandatory for all devices and must be negotiated with the host.
787 AT91SAM9G45
37.5.12.2 Not Powered State
Self powered devices can detect 5V VBUS using a PIO. When the device is not connected to a
host, device power consumption can be reduced by the DETACH bit in UDPHS_CTRL. Dis-
abling the transceiver is automatically done. HSDM, HSDP, FSDP and FSDP lines are tied to
GND pull-downs integrated in the hub downstream ports.
37.5.12.3 Entering Attached State

When no device is connected, the USB FSDP and FSDM signals are tied to GND by 15 KΩ pull-
downs integrated in the hub downstream ports. When a device is attached to an hub down-
stream port, the device connects a 1.5 KΩ pull-up on FSDP. The USB bus line goes into IDLE
state, FSDP is pulled-up by the device 1.5 KΩ resistor to 3.3V and FSDM is pulled-down by the
15 KΩ resistor to GND of the host.
After pull-up connection, the device enters the powered state. The transceiver remains disabled
until bus activity is detected.
In case of low power consumption need, the device can be stopped. When the device detects
the VBUS, the software must enable the USB transceiver by enabling the EN_UDPHS bit in
UDPHS_CTRL register.
The software can detach the pull-up by setting DETACH bit in UDPHS_CTRL register.
37.5.12.4 From Powered State to Default State (Reset)

After its connection to a USB host, the USB device waits for an end-of-bus reset. The unmasked
flag ENDRESET is set in the UDPHS_IEN register and an interrupt is triggered.
Once the ENDRESET interrupt has been triggered, the device enters Default State. In this state,
the UDPHS software must:
• Enable the default endpoint, setting the EPT_ENABL flag in the UDPHS_EPTCTLENB[0]
register and, optionally, enabling the interrupt for endpoint 0 by writing 1 in EPT_0 of the
UDPHS_IEN register. The enumeration then begins by a control transfer.
• Configure the Interrupt Mask Register which has been reset by the USB reset detection
• Enable the transceiver.
In this state, the EN_UDPHS bit in UDPHS_CTRL register must be enabled.
37.5.12.5 From Default State to Address State (Address Assigned)

After a Set Address standard device request, the USB host peripheral enters the address state.
Warning: before the device enters address state, it must achieve the Status IN transaction of
the control transfer, i.e., the UDPHS device sets its new address once the TX_COMPLT flag in
the UDPHS_EPTCTL[0] register has been received and cleared.
To move to address state, the driver software sets the DEV_ADDR field and the FADDR_EN
flag in the UDPHS_CTRL register.
37.5.12.6 From Address State to Configured State (Device Configured)

Once a valid Set Configuration standard request has been received and acknowledged, the
device enables endpoints corresponding to the current configuration. This is done by setting the
BK_NUMBER, EPT_TYPE, EPT_DIR and EPT_SIZE fields in the UDPHS_EPTCFGx registers
and enabling them by setting the EPT_ENABL flag in the UDPHS_EPTCTLENBx registers, and,
optionally, enabling corresponding interrupts in the UDPHS_IEN register.
788 AT91SAM9G45
37.5.12.7 Entering Suspend State (Bus Activity)
When a Suspend (no bus activity on the USB bus) is detected, the DET_SUSPD signal in the
UDPHS_STA register is set. This triggers an interrupt if the corresponding bit is set in the
UDPHS_IEN register. This flag is cleared by writing to the UDPHS_CLRINT register. Then the
device enters Suspend Mode.
In this state bus powered devices must drain less than 500 μA from the 5V VBUS. As an exam-
ple, the microcontroller switches to slow clock, disables the PLL and main oscillator, and goes
into Idle Mode. It may also switch off other devices on the board.
The UDPHS device peripheral clocks can be switched off. Resume event is asynchronously
detected.
37.5.12.8 Receiving a Host Resume

In Suspend mode, a resume event on the USB bus line is detected asynchronously, transceiver
and clocks disabled (however the pull-up should not be removed).
Once the resume is detected on the bus, the signal WAKE_UP in the UDPHS_INTSTA is set. It
may generate an interrupt if the corresponding bit in the UDPHS_IEN register is set. This inter-
rupt may be used to wake-up the core, enable PLL and main oscillators and configure clocks.
37.5.12.9 Sending an External Resume

In Suspend State it is possible to wake-up the host by sending an external resume.
The device waits at least 5 ms after being entered in Suspend State before sending an external
resume.
The device must force a K state from 1 to 15 ms to resume the host.
789 AT91SAM9G45
37.5.13 Test Mode
A device must support the TEST_MODE feature when in the Default, Address or Configured
High Speed device states.
TEST_MODE can be:

• Test_J
• Test_K
• Test_Packet
• Test_SEO_NAK
(See Section 37.6.7 “UDPHS Test Register” on page 802 for definitions of each test mode.)
const char test_packet_buffer[] = {
0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, // JKJKJKJK *
9
0xAA,0xAA,0xAA,0xAA,0xAA,0xAA,0xAA,0xAA, // JJKKJJKK *
8
0xEE,0xEE,0xEE,0xEE,0xEE,0xEE,0xEE,0xEE, // JJKKJJKK *
8
0xFE,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, //
JJJJJJJKKKKKKK * 8
0x7F,0xBF,0xDF,0xEF,0xF7,0xFB,0xFD, // JJJJJJJK * 8
0xFC,0x7E,0xBF,0xDF,0xEF,0xF7,0xFB,0xFD,0x7E // {JKKKKKKK
* 10}, JK
};
790 AT91SAM9G45
37.6 USB High Speed Device Port (UDPHS) User Interface
0x00 UDPHS Control Register UDPHS_CTRL Read-write 0x0000_0200
0x04 UDPHS Frame Number Register UDPHS_FNUM Read 0x0000_0000
0x08 - 0x0C Reserved – – –
0x10 UDPHS Interrupt Enable Register UDPHS_IEN Read-write 0x0000_0010
0x14 UDPHS Interrupt Status Register UDPHS_INTSTA Read 0x0000_0000
0x18 UDPHS Clear Interrupt Register UDPHS_CLRINT Write –
0x1C UDPHS Endpoints Reset Register UDPHS_EPTRST Write –
0x20 - 0xCC Reserved – – –
0xE0 UDPHS Test Register UDPHS_TST Read-write 0x0000_0000
0xE4 - 0xE8 Reserved – – –
0xF0 UDPHS Name1 Register UDPHS_IPNAME1 Read 0x4855_5342
0xF4 UDPHS Name2 Register UDPHS_IPNAME2 Read 0x3244_4556
0xF8 UDPHS Features Register UDPHS_IPFEATURES Read
0x100 + endpoint * 0x20 + 0x00 UDPHS Endpoint Configuration Register UDPHS_EPTCFG Read-write 0x0000_0000
0x100 + endpoint * 0x20 + 0x04 UDPHS Endpoint Control Enable Register UDPHS_EPTCTLENB Write –
0x100 + endpoint * 0x20 + 0x08 UDPHS Endpoint Control Disable Register UDPHS_EPTCTLDIS Write –
0x100 + endpoint * 0x20 + 0x0C UDPHS Endpoint Control Register UDPHS_EPTCTL Read 0x0000_0000(1)
0x100 + endpoint * 0x20 + 0x10 Reserved (for endpoint) – – –
0x100 + endpoint * 0x20 + 0x14 UDPHS Endpoint Set Status Register UDPHS_EPTSETSTA Write –
0x100 + endpoint * 0x20 + 0x18 UDPHS Endpoint Clear Status Register UDPHS_EPTCLRSTA Write –
0x100 + endpoint * 0x20 + 0x1C UDPHS Endpoint Status Register UDPHS_EPTSTA Read 0x0000_0040
0x120 - 0x1DC UDPHS Endpoint1 to 6 (2) Registers
0x1E0 - 0x300 Reserved
0x300 - 0x30C Reserved – – –
0x310 + channel * 0x10 + 0x00 UDPHS DMA Next Descriptor Address Register UDPHS_DMANXTDSC Read-write 0x0000_0000
0x310 + channel * 0x10 + 0x04 UDPHS DMA Channel Address Register UDPHS_DMAADDRESS Read-write 0x0000_0000
0x310 + channel * 0x10 + 0x08 UDPHS DMA Channel Control Register UDPHS_DMACONTROL Read-write 0x0000_0000
0x310 + channel * 0x10 + 0x0C UDPHS DMA Channel Status Register UDPHS_DMASTATUS Read-write 0x0000_0000
(3)
0x320 - 0x370 DMA Channel2 to 5 Registers
Notes: 1. The reset value for UDPHS_EPTCTL0 is 0x0000_0001.
2. The addresses for the UDPHS Endpoint registers shown here are for UDPHS Endpoint0. The structure of this group of reg-
isters is repeated successively for each endpoint according to the consecution of endpoint registers located between 0x120
and 0x1DC.
3. The addresses for the UDPHS DMA registers shown here are for UDPHS DMA Channel1. (There is no Channel0) The
structure of this group of registers is repeated successively for each DMA channel according to the consecution of DMA reg-
isters located between 0x320 and 0x370.
791 AT91SAM9G45
37.6.1 UDPHS Control Register
Name: UDPHS_CTRL
Address: 0xFFF78000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – PULLD_DIS REWAKEUP DETACH EN_UDPHS
7 6 5 4 3 2 1 0
FADDR_EN DEV_ADDR
• DEV_ADDR: UDPHS Address

Read:
This field contains the default address (0) after power-up or UDPHS bus reset.
Write:
This field is written with the value set by a SET_ADDRESS request received by the device firmware.
• FADDR_EN: Function Address Enable

Read:
0 = Device is not in address state.
1 = Device is in address state.
Write:
0 = only the default function address is used (0).
1 = this bit is set by the device firmware after a successful status phase of a SET_ADDRESS transaction. When set, the
only address accepted by the UDPHS controller is the one stored in the UDPHS Address field. It will not be cleared after-
wards by the device firmware. It is cleared by hardware on hardware reset, or when UDPHS bus reset is received (see
above).
• EN_UDPHS: UDPHS Enable

Read:
0 = UDPHS is disabled.
1 = UDPHS is enabled.
Write:
0 = disable and reset the UDPHS controller. Switch the host to UTMI.
1 = enables the UDPHS controller. Switch the host to UTMI.
792 AT91SAM9G45
• DETACH: Detach Command
Read:
0 = UDPHS is attached.
1 = UDPHS is detached, UTMI transceiver is suspended.
Write:
0 = pull up the DP line (attach command).
1 = simulate a detach on the UDPHS line and force the UTMI transceiver into suspend state (Suspend M = 0).
(See PULLD_DIS description below.)
• REWAKEUP: Send Remote Wake Up

Read:
0 = Remote Wake Up is disabled.
1 = Remote Wake Up is enabled.
Write:
0 = no effect.
1 = force an external interrupt on the UDPHS controller for Remote Wake UP purposes.
An Upstream Resume is sent only after the UDPHS bus has been in SUSPEND state for at least 5 ms.
This bit is automatically cleared by hardware at the end of the Upstream Resume.
• PULLD_DIS: Pull-Down Disable

When set, there is no pull-down on DP & DM. (DM Pull-Down = DP Pull-Down = 0).
Note: If the DETACH bit is also set, device DP & DM are left in high impedance state.
(See DETACH description above.)
DETACH PULLD_DIS DP DM Condition

0 0 Pull up Pull down not recommended
High impedence
0 1 Pull up VBUS present
state
1 0 Pull down Pull down No VBUS
High impedence High impedence VBUS present &
1 1
state state software disconnect
793 AT91SAM9G45
37.6.2 UDPHS Frame Number Register
Name: UDPHS_FNUM
Address: 0xFFF78004
Access Type: Read
31 30 29 28 27 26 25 24
FNUM_ERR – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – FRAME_NUMBER
7 6 5 4 3 2 1 0
FRAME_NUMBER MICRO_FRAME_NUM
• MICRO_FRAME_NUM: Microframe Number

Number of the received microframe (0 to 7) in one frame.This field is reset at the beginning of each new frame (1 ms).
One microframe is received each 125 microseconds (1 ms/8).
• FRAME_NUMBER: Frame Number as defined in the Packet Field Formats

This field is provided in the last received SOF packet (see INT_SOF in the UDPHS Interrupt Status Register).
• FNUM_ERR: Frame Number CRC Error

This bit is set by hardware when a corrupted Frame Number in Start of Frame packet (or Micro SOF) is received.
This bit and the INT_SOF (or MICRO_SOF) interrupt are updated at the same time.
794 AT91SAM9G45
37.6.3 UDPHS Interrupt Enable Register
Name: UDPHS_IEN
Address: 0xFFF78010
31 30 29 28 27 26 25 24
– DMA_6 DMA_5 DMA_4 DMA_3 DMA_2 DMA_1 –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– EPT_6 EPT_5 EPT_4 EPT_3 EPT_2 EPT_1 EPT_0
7 6 5 4 3 2 1 0
UPSTR_RES ENDOFRSM WAKE_UP ENDRESET INT_SOF MICRO_SOF DET_SUSPD –
• DET_SUSPD: Suspend Interrupt Enable

Read:
0 = Suspend Interrupt is disabled.
1 = Suspend Interrupt is enabled.
Write:
0 = disable Suspend Interrupt.
1 = enable Suspend Interrupt.
• MICRO_SOF: Micro-SOF Interrupt Enable

Read:
0 = Micro-SOF Interrupt is disabled.
1 = Micro-SOF Interrupt is enabled.
Write:
0 = disable Micro-SOF Interrupt.
1 = enable Micro-SOF Interrupt.
• INT_SOF: SOF Interrupt Enable

Read:
0 = SOF Interrupt is disabled.
1 = SOF Interrupt is enabled.
Write:
0 = disable SOF Interrupt.
1 = enable SOF Interrupt.
795 AT91SAM9G45
• ENDRESET: End Of Reset Interrupt Enable
Read:
0 = End Of Reset Interrupt is disabled.
1 = End Of Reset Interrupt is enabled.
Write:
0 = disable End Of Reset Interrupt.
1 = enable End Of Reset Interrupt. Automatically enabled after USB reset.
• WAKE_UP: Wake Up CPU Interrupt Enable

Read:
0 = Wake Up CPU Interrupt is disabled.
1 = Wake Up CPU Interrupt is enabled.
Write
0 = disable Wake Up CPU Interrupt.
1 = enable Wake Up CPU Interrupt.
• ENDOFRSM: End Of Resume Interrupt Enable

Read:
0 = Resume Interrupt is disabled.
1 = Resume Interrupt is enabled.
Write:
0 = disable Resume Interrupt.
1 = enable Resume Interrupt.
• UPSTR_RES: Upstream Resume Interrupt Enable

Read:
0 = Upstream Resume Interrupt is disabled.
1 = Upstream Resume Interrupt is enabled.
Write:
0 = disable Upstream Resume Interrupt.
1 = enable Upstream Resume Interrupt.
• EPT_x: Endpoint x Interrupt Enable

Read:
0 = the interrupts for this endpoint are disabled.
1 = the interrupts for this endpoint are enabled.
796 AT91SAM9G45
Write:
0 = disable the interrupts for this endpoint.
1 = enable the interrupts for this endpoint.
• DMA_x: DMA Channel x Interrupt Enable

Read:
0 = the interrupts for this channel are disabled.
1 = the interrupts for this channel are enabled.
Write:
0 = disable the interrupts for this channel.
1 = enable the interrupts for this channel.
797 AT91SAM9G45
37.6.4 UDPHS Interrupt Status Register
Name: UDPHS_INTSTA
Address: 0xFFF78014
31 30 29 28 27 26 25 24
– DMA_6 DMA_5 DMA_4 DMA_3 DMA_2 DMA_1 –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
UPSTR_RES ENDOFRSM WAKE_UP ENDRESET INT_SOF MICRO_SOF DET_SUSPD SPEED
• SPEED: Speed Status

0 = reset by hardware when the hardware is in Full Speed mode.
1 = set by hardware when the hardware is in High Speed mode
• DET_SUSPD: Suspend Interrupt

0 = cleared by setting the DET_SUSPD bit in UDPHS_CLRINT register
1 = set by hardware when a UDPHS Suspend (Idle bus for three frame periods, a J state for 3 ms) is detected. This triggers
a UDPHS interrupt when the DET_SUSPD bit is set in UDPHS_IEN register.
• MICRO_SOF: Micro Start Of Frame Interrupt

0 = cleared by setting the MICRO_SOF bit in UDPHS_CLRINT register.
1 = set by hardware when an UDPHS micro start of frame PID (SOF) has been detected (every 125 us) or synthesized by
the macro. This triggers a UDPHS interrupt when the MICRO_SOF bit is set in UDPHS_IEN. In case of detected SOF, the
MICRO_FRAME_NUM field in UDPHS_FNUM register is incremented and the FRAME_NUMBER field doesn’t change.
Note: The Micro Start Of Frame Interrupt (MICRO_SOF), and the Start Of Frame Interrupt (INT_SOF) are not generated at the same
time.
• INT_SOF: Start Of Frame Interrupt

0 = cleared by setting the INT_SOF bit in UDPHS_CLRINT.
1 = set by hardware when an UDPHS Start Of Frame PID (SOF) has been detected (every 1 ms) or synthesized by the
macro. This triggers a UDPHS interrupt when the INT_SOF bit is set in UDPHS_IEN register. In case of detected SOF, in
High Speed mode, the MICRO_FRAME_NUMBER field is cleared in UDPHS_FNUM register and the FRAME_NUMBER
field is updated.
• ENDRESET: End Of Reset Interrupt

0 = cleared by setting the ENDRESET bit in UDPHS_CLRINT.
1 = set by hardware when an End Of Reset has been detected by the UDPHS controller. This triggers a UDPHS interrupt
when the ENDRESET bit is set in UDPHS_IEN.
798 AT91SAM9G45
• WAKE_UP: Wake Up CPU Interrupt
0 = cleared by setting the WAKE_UP bit in UDPHS_CLRINT.
1 = set by hardware when the UDPHS controller is in SUSPEND state and is re-activated by a filtered non-idle signal from
the UDPHS line (not by an upstream resume). This triggers a UDPHS interrupt when the WAKE_UP bit is set in
UDPHS_IEN register. When receiving this interrupt, the user has to enable the device controller clock prior to operation.
Note: this interrupt is generated even if the device controller clock is disabled.
• ENDOFRSM: End Of Resume Interrupt

0 = cleared by setting the ENDOFRSM bit in UDPHS_CLRINT.
1 = set by hardware when the UDPHS controller detects a good end of resume signal initiated by the host. This triggers a
UDPHS interrupt when the ENDOFRSM bit is set in UDPHS_IEN.
• UPSTR_RES: Upstream Resume Interrupt

0 = cleared by setting the UPSTR_RES bit in UDPHS_CLRINT.
1 = set by hardware when the UDPHS controller is sending a resume signal called “upstream resume”. This triggers a
UDPHS interrupt when the UPSTR_RES bit is set in UDPHS_IEN.
• EPT_x: Endpoint x Interrupt

0 = reset when the UDPHS_EPTSTAx interrupt source is cleared.
1 = set by hardware when an interrupt is triggered by the UDPHS_EPTSTAx register and this endpoint interrupt is enabled
by the EPT_x bit in UDPHS_IEN.
• DMA_x: DMA Channel x Interrupt

0 = reset when the UDPHS_DMASTATUSx interrupt source is cleared.
1 = set by hardware when an interrupt is triggered by the DMA Channelx and this endpoint interrupt is enabled by the
DMA_x bit in UDPHS_IEN.
799 AT91SAM9G45
37.6.5 UDPHS Clear Interrupt Register
Name: UDPHS_CLRINT
Address: 0xFFF78018
Access Type: Write only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
UPSTR_RES ENDOFRSM WAKE_UP ENDRESET INT_SOF MICRO_SOF DET_SUSPD –
• DET_SUSPD: Suspend Interrupt Clear

0 = no effect.
1 = clear the DET_SUSPD bit in UDPHS_INTSTA.
• MICRO_SOF: Micro Start Of Frame Interrupt Clear

0 = no effect.
1 = clear the MICRO_SOF bit in UDPHS_INTSTA.
• INT_SOF: Start Of Frame Interrupt Clear

0 = no effect.
1 = clear the INT_SOF bit in UDPHS_INTSTA.
• ENDRESET: End Of Reset Interrupt Clear

0 = no effect.
1 = clear the ENDRESET bit in UDPHS_INTSTA.
• WAKE_UP: Wake Up CPU Interrupt Clear

0 = no effect.
1 = clear the WAKE_UP bit in UDPHS_INTSTA.
• ENDOFRSM: End Of Resume Interrupt Clear

0 = no effect.
1 = clear the ENDOFRSM bit in UDPHS_INTSTA.
• UPSTR_RES: Upstream Resume Interrupt Clear

0 = no effect.
1 = clear the UPSTR_RES bit in UDPHS_INTSTA.
800 AT91SAM9G45
37.6.6 UDPHS Endpoints Reset Register
Name: UDPHS_EPTRST
Address: 0xFFF7801C
Access Type: Write only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
• EPT_x: Endpoint x Reset

0 = no effect.
1 = reset the Endpointx state.
Setting this bit clears the Endpoint status UDPHS_EPTSTAx register, except for the TOGGLESQ_STA field.
801 AT91SAM9G45
37.6.7 UDPHS Test Register
Name: UDPHS_TST
Address: 0xFFF780E0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – OPMODE2 TST_PKT TST_K TST_J SPEED_CFG
• SPEED_CFG: Speed Configuration

Read-write:
Speed Configuration:
Normal Mode: The macro is in Full Speed mode, ready to make a High Speed identification, if the host supports it and then
00
to automatically switch to High Speed mode
01 Reserved
10 Force High Speed: Set this value to force the hardware to work in High Speed mode. Only for debug or test purpose.
Force Full Speed: Set this value to force the hardware to work only in Full Speed mode. In this configuration, the macro will
11
not respond to a High Speed reset handshake
• TST_J: Test J Mode

Read and write:
0 = no effect.
1 = set to send the J state on the UDPHS line. This enables the testing of the high output drive level on the D+ line.
• TST_K: Test K Mode

Read and write:
0 = no effect.
1 = set to send the K state on the UDPHS line. This enables the testing of the high output drive level on the D- line.
• TST_PKT: Test Packet Mode

Read and write:
0 = no effect.
1 = set to repetitively transmit the packet stored in the current bank. This enables the testing of rise and fall times, eye pat-
terns, jitter, and any other dynamic waveform specifications.
802 AT91SAM9G45
• OPMODE2: OpMode2
Read and write:
0 = no effect.
1 = set to force the OpMode signal (UTMI interface) to “10”, to disable the bit-stuffing and the NRZI encoding.
Note: For the Test mode, Test_SE0_NAK (see Universal Serial Bus Specification, Revision 2.0: 7.1.20, Test Mode Sup-
port). Force the device in High Speed mode, and configure a bulk-type endpoint. Do not fill this endpoint for sending NAK to
the host.
Upon command, a port’s transceiver must enter the High Speed receive mode and remain in that mode until the exit action
is taken. This enables the testing of output impedance, low level output voltage and loading characteristics. In addition,
while in this mode, upstream facing ports (and only upstream facing ports) must respond to any IN token packet with a NAK
handshake (only if the packet CRC is determined to be correct) within the normal allowed device response time. This
enables testing of the device squelch level circuitry and, additionally, provides a general purpose stimulus/response test for
basic functional testing.
803 AT91SAM9G45
37.6.8 UDPHS Name1 Register
Name: UDPHS_IPNAME1
Address: 0xFFF780F0
31 30 29 28 27 26 25 24
IP_NAME1
23 22 21 20 19 18 17 16
IP_NAME1
15 14 13 12 11 10 9 8
IP_NAME1
7 6 5 4 3 2 1 0
IP_NAME1
• IP_NAME1
ASCII string “HUSB”
804 AT91SAM9G45
37.6.9 UDPHS Name2 Register
Name: UDPHS_IPNAME2
Address: 0xFFF780F4
31 30 29 28 27 26 25 24
IP_NAME2
23 22 21 20 19 18 17 16
IP_NAME2
15 14 13 12 11 10 9 8
IP_NAME2
7 6 5 4 3 2 1 0
IP_NAME2
• IP_NAME2
ASCII string “2DEV”
805 AT91SAM9G45
37.6.10 UDPHS Features Register
Name: UDPHS_IPFEATURES
Address: 0xFFF780F8
31 30 29 28 27 26 25 24
ISO_EPT_15 ISO_EPT_14 ISO_EPT_13 ISO_EPT_12 ISO_EPT_11 ISO_EPT_10 ISO_EPT_9 ISO_EPT_8
23 22 21 20 19 18 17 16
ISO_EPT_7 ISO_EPT_6 ISO_EPT_5 ISO_EPT_4 ISO_EPT_3 ISO_EPT_2 ISO_EPT_1 DATAB16_8
15 14 13 12 11 10 9 8
BW_DPRAM FIFO_MAX_SIZE DMA_FIFO_WORD_DEPTH
7 6 5 4 3 2 1 0
DMA_B_SIZ DMA_CHANNEL_NBR EPT_NBR_MAX
• EPT_NBR_MAX: Max Number of Endpoints

Give the max number of endpoints.
0 = if 16 endpoints are hardware implemented.
1 = if 1 endpoint is hardware implemented.
...
• DMA_CHANNEL_NBR: Number of DMA Channels

Give the number of DMA channels.
1 = if 1 DMA channel is hardware implemented.
2 = if 2 DMA channels are hardware implemented.
...
7 = if 7 DMA channels are hardware implemented.
• DMA_B_SIZ: DMA Buffer Size

0 = if the DMA Buffer size is 16 bits.
1 = if the DMA Buffer size is 24 bits.
• DMA_FIFO_WORD_DEPTH: DMA FIFO Depth in Words

0 = if FIFO is 16 words deep.
1 = if FIFO is 1 word deep.
...
806 AT91SAM9G45
• FIFO_MAX_SIZE: DPRAM Size
0 = if DPRAM is 128 bytes deep.
• BW_DPRAM: DPRAM Byte Write Capability

0 = if DPRAM Write Data Shadow logic is implemented.
1 = if DPRAM is byte write capable.
• DATAB16_8: UTMI DataBus16_8

0 = if the UTMI uses an 8-bit parallel data interface (60 MHz, unidirectional).
1 = if the UTMI uses a 16-bit parallel data interface (30 MHz, bidirectional).
• ISO_EPT_x: Endpointx High Bandwidth Isochronous Capability

0 = if the endpoint does not have isochronous High Bandwidth Capability.
1 = if the endpoint has isochronous High Bandwidth Capability.
807 AT91SAM9G45
37.6.11 UDPHS Endpoint Configuration Register
Name: UDPHS_EPTCFGx [x=0..6]
Addresses: 0xFFF78100 [0], 0xFFF78120 [1], 0xFFF78140 [2], 0xFFF78160 [3], 0xFFF78180 [4],
0xFFF781A0 [5], 0xFFF781C0 [6]
31 30 29 28 27 26 25 24
EPT_MAPD – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – NB_TRANS
7 6 5 4 3 2 1 0
BK_NUMBER EPT_TYPE EPT_DIR EPT_SIZE
• EPT_SIZE: Endpoint Size

Read and write:
Set this field according to the endpoint size in bytes (see Section 37.5.5 ”Endpoint Configuration”).
Endpoint Size
000 8 bytes
001 16 bytes
010 32 bytes
011 64 bytes
100 128 bytes
101 256 bytes
110 512 bytes
111 1024 bytes(1)
Note: 1. 1024 bytes is only for isochronous endpoint.
• EPT_DIR: Endpoint Direction

Read and write:
0 = Clear this bit to configure OUT direction for Bulk, Interrupt and Isochronous endpoints.
1 = set this bit to configure IN direction for Bulk, Interrupt and Isochronous endpoints.
For Control endpoints this bit has no effect and should be left at zero.
• EPT_TYPE: Endpoint Type

Read and write:
Set this field according to the endpoint type (see Section 37.5.5 ”Endpoint Configuration”).
(Endpoint 0 should always be configured as control)
808 AT91SAM9G45
:Endpoint Type
00 Control endpoint
01 Isochronous endpoint
10 Bulk endpoint
11 Interrupt endpoint
• BK_NUMBER: Number of Banks

Read and write:
Set this field according to the endpoint’s number of banks (see Section 37.5.5 ”Endpoint Configuration”).
Number of Banks
00 Zero bank, the endpoint is not mapped in memory

01 One bank (bank 0)
10 Double bank (Ping-Pong: bank 0/bank 1)
11 Triple bank (bank 0/bank 1/bank 2)
• NB_TRANS: Number Of Transaction per Microframe

Read and Write:
The Number of transactions per microframe is set by software.
Note: Meaningful for high bandwidth isochronous endpoint only.
• EPT_MAPD: Endpoint Mapped

Read-only:
0 = the user should reprogram the register with correct values.
1 = set by hardware when the endpoint size (EPT_SIZE) and the number of banks (BK_NUMBER) are correct regarding:
– the fifo max capacity (FIFO_MAX_SIZE in UDPHS_IPFEATURES register)
– the number of endpoints/banks already allocated
– the number of allowed banks for this endpoint
809 AT91SAM9G45
37.6.12 UDPHS Endpoint Control Enable Register
Name: UDPHS_EPTCTLENBx [x=0..6]
0xFFF781A4 [5], 0xFFF781C4 [6]
31 30 29 28 27 26 25 24
SHRT_PCKT – – – – – – –
23 22 21 20 19 18 17 16
– – – – – BUSY_BANK – –
15 14 13 12 11 10 9 8
STALL_SNT/
NAK_IN/ RX_SETUP/ TX_PK_RDY/
NAK_OUT ERR_CRISO/ TX_COMPLT RX_BK_RDY ERR_OVFLW
ERR_FLUSH ERR_FL_ISO ERR_TRANS
ERR_NBTRA
7 6 5 4 3 2 1 0
MDATA_RX DATAX_RX – NYET_DIS INTDIS_DMA – AUTO_VALID EPT_ENABL
For additional Information, see “UDPHS Endpoint Control Register” on page 814.
• EPT_ENABL: Endpoint Enable

0 = no effect.
1 = enable endpoint according to the device configuration.
• AUTO_VALID: Packet Auto-Valid Enable

0 = no effect.
1 = enable this bit to automatically validate the current packet and switch to the next bank for both IN and OUT transfers.
• INTDIS_DMA: Interrupts Disable DMA

0 = no effect.
1 = If set, when an enabled endpoint-originated interrupt is triggered, the DMA request is disabled.
• NYET_DIS: NYET Disable (Only for High Speed Bulk OUT endpoints)
0 = no effect.
1 = forces an ACK response to the next High Speed Bulk OUT transfer instead of a NYET response.
• DATAX_RX: DATAx Interrupt Enable (Only for high bandwidth Isochronous OUT endpoints)
0 = no effect.
1 = enable DATAx Interrupt.
• MDATA_RX: MDATA Interrupt Enable (Only for high bandwidth Isochronous OUT endpoints)
0 = no effect.
1 = enable MDATA Interrupt.
810 AT91SAM9G45
• ERR_OVFLW: Overflow Error Interrupt Enable
0 = no effect.
1 = enable Overflow Error Interrupt.
• RX_BK_RDY: Received OUT Data Interrupt Enable

0 = no effect.
1 = enable Received OUT Data Interrupt.
• TX_COMPLT: Transmitted IN Data Complete Interrupt Enable

0 = no effect.
1 = enable Transmitted IN Data Complete Interrupt.
• TX_PK_RDY/ERR_TRANS: TX Packet Ready/Transaction Error Interrupt Enable

0 = no effect.
1 = enable TX Packet Ready/Transaction Error Interrupt.
• RX_SETUP/ERR_FL_ISO: Received SETUP/Error Flow Interrupt Enable

0 = no effect.
1 = enable RX_SETUP/Error Flow ISO Interrupt.
• STALL_SNT/ERR_CRISO/ERR_NBTRA: Stall Sent /ISO CRC Error/Number of Transaction Error Interrupt Enable
0 = no effect.
1 = enable Stall Sent/Error CRC ISO/Error Number of Transaction Interrupt.
• NAK_IN/ERR_FLUSH: NAKIN/Bank Flush Error Interrupt Enable

0 = no effect.
1 = enable NAKIN/Bank Flush Error Interrupt.
• NAK_OUT: NAKOUT Interrupt Enable

0 = no effect.
1 = enable NAKOUT Interrupt.
• BUSY_BANK: Busy Bank Interrupt Enable

0 = no effect.
1 = enable Busy Bank Interrupt.
• SHRT_PCKT: Short Packet Send/Short Packet Interrupt Enable

For OUT endpoints:
0 = no effect.
1 = enable Short Packet Interrupt.
For IN endpoints:
Guarantees short packet at end of DMA Transfer if the UDPHS_DMACONTROLx register END_B_EN and
UDPHS_EPTCTLx register AUTOVALID bits are also set.
811 AT91SAM9G45
37.6.13 UDPHS Endpoint Control Disable Register
Name: UDPHS_EPTCTLDISx [x=0..6]
0xFFF781A8 [5], 0xFFF781C8 [6]
31 30 29 28 27 26 25 24
SHRT_PCKT – – – – – – –
23 22 21 20 19 18 17 16
– – – – – BUSY_BANK – –
15 14 13 12 11 10 9 8
STALL_SNT/
ERR_NBTRA
7 6 5 4 3 2 1 0
MDATA_RX DATAX_RX – NYET_DIS INTDIS_DMA – AUTO_VALID EPT_DISABL
For additional Information, see “UDPHS Endpoint Control Register” on page 814.
• EPT_DISABL: Endpoint Disable

0 = no effect.
1 = disable endpoint.
• AUTO_VALID: Packet Auto-Valid Disable

0 = no effect.
1 = disable this bit to not automatically validate the current packet.
• INTDIS_DMA: Interrupts Disable DMA

0 = no effect.
1 = disable the “Interrupts Disable DMA”.
• NYET_DIS: NYET Enable (Only for High Speed Bulk OUT endpoints)
0 = no effect.
1 = let the hardware handle the handshake response for the High Speed Bulk OUT transfer.
• DATAX_RX: DATAx Interrupt Disable (Only for High Bandwidth Isochronous OUT endpoints)
0 = no effect.
1 = disable DATAx Interrupt.
• MDATA_RX: MDATA Interrupt Disable (Only for High Bandwidth Isochronous OUT endpoints)
0 = no effect.
1 = disable MDATA Interrupt.
812 AT91SAM9G45
• ERR_OVFLW: Overflow Error Interrupt Disable
0 = no effect.
1 = disable Overflow Error Interrupt.
• RX_BK_RDY: Received OUT Data Interrupt Disable

0 = no effect.
1 = disable Received OUT Data Interrupt.
• TX_COMPLT: Transmitted IN Data Complete Interrupt Disable

0 = no effect.
1 = disable Transmitted IN Data Complete Interrupt.
• TX_PK_RDY/ERR_TRANS: TX Packet Ready/Transaction Error Interrupt Disable

0 = no effect.
1 = disable TX Packet Ready/Transaction Error Interrupt.
• RX_SETUP/ERR_FL_ISO: Received SETUP/Error Flow Interrupt Disable

0 = no effect.
1 = disable RX_SETUP/Error Flow ISO Interrupt.
• STALL_SNT/ERR_CRISO/ERR_NBTRA: Stall Sent/ISO CRC Error/Number of Transaction Error Interrupt Disable

0 = no effect.
1 = disable Stall Sent/Error CRC ISO/Error Number of Transaction Interrupt.
• NAK_IN/ERR_FLUSH: NAKIN/bank flush error Interrupt Disable

0 = no effect.
1 = disable NAKIN/ Bank Flush Error Interrupt.
• NAK_OUT: NAKOUT Interrupt Disable

0 = no effect.
1 = disable NAKOUT Interrupt.
• BUSY_BANK: Busy Bank Interrupt Disable

0 = no effect.
1 = disable Busy Bank Interrupt.
• SHRT_PCKT: Short Packet Interrupt Disable

For OUT endpoints:
0 = no effect.
1 = disable Short Packet Interrupt.
For IN endpoints:
Never automatically add a zero length packet at end of DMA transfer.
813 AT91SAM9G45
37.6.14 UDPHS Endpoint Control Register
Name: UDPHS_EPTCTLx [x=0..6]
Addresses: 0xFFF7810C [0], 0xFFF7812C [1], 0xFFF7814C [2], 0xFFF7816C [3], 0xFFF7818C [4],
0xFFF781AC [5], 0xFFF781CC [6]
31 30 29 28 27 26 25 24
SHRT_PCKT – – – – – – –
23 22 21 20 19 18 17 16
– – – – – BUSY_BANK – –
15 14 13 12 11 10 9 8
STALL_SNT/
ERR_NBTRA
7 6 5 4 3 2 1 0
MDATA_RX DATAX_RX – NYET_DIS INTDIS_DMA – AUTO_VALID EPT_ENABL
• EPT_ENABL: Endpoint Enable

0 = If cleared, the endpoint is disabled according to the device configuration. Endpoint 0 should always be enabled after a
hardware or UDPHS bus reset and participate in the device configuration.
1 = If set, the endpoint is enabled according to the device configuration.
• AUTO_VALID: Packet Auto-Valid Enabled (Not for CONTROL Endpoints)

Set this bit to automatically validate the current packet and switch to the next bank for both IN and OUT endpoints.
For IN Transfer:
If this bit is set, then the UDPHS_EPTSTAx register TX_PK_RDY bit is set automatically when the current bank is full
and at the end of DMA buffer if the UDPHS_DMACONTROLx register END_B_EN bit is set.
The user may still set the UDPHS_EPTSTAx register TX_PK_RDY bit if the current bank is not full, unless the user
wants to send a Zero Length Packet by software.
For OUT Transfer:
If this bit is set, then the UDPHS_EPTSTAx register RX_BK_RDY bit is automatically reset for the current bank when
the last packet byte has been read from the bank FIFO or at the end of DMA buffer if the UDPHS_DMACONTROLx
register END_B_EN bit is set. For example, to truncate a padded data packet when the actual data transfer size is
reached.
The user may still clear the UDPHS_EPTSTAx register RX_BK_RDY bit, for example, after completing a DMA buffer
by software if UDPHS_DMACONTROLx register END_B_EN bit was disabled or in order to cancel the read of the
remaining data bank(s).
• INTDIS_DMA: Interrupt Disables DMA

If set, when an enabled endpoint-originated interrupt is triggered, the DMA request is disabled regardless of the
UDPHS_IEN register EPT_x bit for this endpoint. Then, the firmware will have to clear or disable the interrupt source or
clear this bit if transfer completion is needed.
814 AT91SAM9G45
If the exception raised is associated with the new system bank packet, then the previous DMA packet transfer is normally
completed, but the new DMA packet transfer is not started (not requested).
If the exception raised is not associated to a new system bank packet (NAK_IN, NAK_OUT, ERR_FL_ISO...), then the
request cancellation may happen at any time and may immediately stop the current DMA transfer.
This may be used, for example, to identify or prevent an erroneous packet to be transferred into a buffer or to complete a
DMA buffer by software after reception of a short packet, or to perform buffer truncation on ERR_FL_ISO interrupt for
adaptive rate.
• NYET_DIS: NYET Disable (Only for High Speed Bulk OUT endpoints)
0 = If clear, this bit lets the hardware handle the handshake response for the High Speed Bulk OUT transfer.
1 = If set, this bit forces an ACK response to the next High Speed Bulk OUT transfer instead of a NYET response.
Note: According to the Universal Serial Bus Specification, Rev 2.0 (8.5.1.1 NAK Responses to OUT/DATA During PING Protocol), a
NAK response to an HS Bulk OUT transfer is expected to be an unusual occurrence.
• DATAX_RX: DATAx Interrupt Enabled (Only for High Bandwidth Isochronous OUT endpoints)
0 = no effect.
1 = send an interrupt when a DATA2, DATA1 or DATA0 packet has been received meaning the whole microframe data
payload has been received.
• MDATA_RX: MDATA Interrupt Enabled (Only for High Bandwidth Isochronous OUT endpoints)
0 = no effect.
1 = send an interrupt when an MDATA packet has been received and so at least one packet of the microframe data pay-
load has been received.
• ERR_OVFLW: Overflow Error Interrupt Enabled

0 = Overflow Error Interrupt is masked.
1 = Overflow Error Interrupt is enabled.
• RX_BK_RDY: Received OUT Data Interrupt Enabled

0 = Received OUT Data Interrupt is masked.
1 = Received OUT Data Interrupt is enabled.
• TX_COMPLT: Transmitted IN Data Complete Interrupt Enabled

0 = Transmitted IN Data Complete Interrupt is masked.
1 = Transmitted IN Data Complete Interrupt is enabled.
• TX_PK_RDY/ERR_TRANS: TX Packet Ready/Transaction Error Interrupt Enabled

0 = TX Packet Ready/Transaction Error Interrupt is masked.
1 = TX Packet Ready/Transaction Error Interrupt is enabled.
Caution: Interrupt source is active as long as the corresponding UDPHS_EPTSTAx register TX_PK_RDY flag remains
low. If there are no more banks available for transmitting after the software has set UDPHS_EPTSTAx/TX_PK_RDY
for the last transmit packet, then the interrupt source remains inactive until the first bank becomes free again to transmit
at UDPHS_EPTSTAx/TX_PK_RDY hardware clear.
815 AT91SAM9G45
• RX_SETUP/ERR_FL_ISO: Received SETUP/Error Flow Interrupt Enabled
0 = Received SETUP/Error Flow Interrupt is masked.
1 = Received SETUP/Error Flow Interrupt is enabled.
• STALL_SNT/ERR_CRISO/ERR_NBTRA: Stall Sent/ISO CRC Error/Number of Transaction Error Interrupt Enabled

0 = Stall Sent/ISO CRC error/number of Transaction Error Interrupt is masked.
1 = Stall Sent /ISO CRC error/number of Transaction Error Interrupt is enabled.
• NAK_IN/ERR_FLUSH: NAKIN/Bank Flush Error Interrupt Enabled

0 = NAKIN Interrupt is masked.
1 = NAKIN/Bank Flush Error Interrupt is enabled.
• NAK_OUT: NAKOUT Interrupt Enabled

0 = NAKOUT Interrupt is masked.
1 = NAKOUT Interrupt is enabled.
• BUSY_BANK: Busy Bank Interrupt Enabled

0 = BUSY_BANK Interrupt is masked.
1 = BUSY_BANK Interrupt is enabled.
For OUT endpoints: an interrupt is sent when all banks are busy.
For IN endpoints: an interrupt is sent when all banks are free.
• SHRT_PCKT: Short Packet Interrupt Enabled

For OUT endpoints: send an Interrupt when a Short Packet has been received.
0 = Short Packet Interrupt is masked.
1 = Short Packet Interrupt is enabled.
For IN endpoints: a Short Packet transmission is guaranteed upon end of the DMA Transfer, thus signaling a BULK or
INTERRUPT end of transfer or an end of isochronous (micro-)frame data, but only if the UDPHS_DMACONTROLx
register END_B_EN and UDPHS_EPTCTLx register AUTO_VALID bits are also set.
816 AT91SAM9G45
37.6.15 UDPHS Endpoint Set Status Register
Name: UDPHS_EPTSETSTAx [x=0..6]
0xFFF781B4 [5], 0xFFF781D4 [6]
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – TX_PK_RDY – KILL_BANK –
7 6 5 4 3 2 1 0
– – FRCESTALL – – – – –
• FRCESTALL: Stall Handshake Request Set

0 = no effect.
1 = set this bit to request a STALL answer to the host for the next handshake
Refer to chapters 8.4.5 (Handshake Packets) and 9.4.5 (Get Status) of the Universal Serial Bus Specification, Rev 2.0 for
more information on the STALL handshake.
• KILL_BANK: KILL Bank Set (for IN Endpoint)

0 = no effect.
1 = kill the last written bank.
• TX_PK_RDY: TX Packet Ready Set

0 = no effect.
1 = set this bit after a packet has been written into the endpoint FIFO for IN data transfers
– This flag is used to generate a Data IN transaction (device to host).
– Device firmware checks that it can write a data payload in the FIFO, checking that TX_PK_RDY is cleared.
– Transfer to the FIFO is done by writing in the “Buffer Address” register.
– Once the data payload has been transferred to the FIFO, the firmware notifies the UDPHS device setting
TX_PK_RDY to one.
– UDPHS bus transactions can start.
– TXCOMP is set once the data payload has been received by the host.
– Data should be written into the endpoint FIFO only after this bit has been cleared.
– Set this bit without writing data to the endpoint FIFO to send a Zero Length Packet.
817 AT91SAM9G45
37.6.16 UDPHS Endpoint Clear Status Register
Name: UDPHS_EPTCLRSTAx [x=0..6]
Addresses: 0xFFF78118 [0], 0xFFF78138 [1], 0xFFF78158 [2], 0xFFF78178 [3], 0xFFF78198 [4]
0xFFF781B8 [5], 0xFFF781D8 [6]
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
NAK_IN/ STALL_SNT/ RX_SETUP/
NAK_OUT – TX_COMPLT RX_BK_RDY –
ERR_FLUSH ERR_NBTRA ERR_FL_ISO
7 6 5 4 3 2 1 0
– TOGGLESQ FRCESTALL – – – – –
• FRCESTALL: Stall Handshake Request Clear

0 = no effect.
1 = clear the STALL request. The next packets from host will not be STALLed.
• TOGGLESQ: Data Toggle Clear

0 = no effect.
1 = clear the PID data of the current bank
For OUT endpoints, the next received packet should be a DATA0.
For IN endpoints, the next packet will be sent with a DATA0 PID.
• RX_BK_RDY: Received OUT Data Clear

0 = no effect.
1 = clear the RX_BK_RDY flag of UDPHS_EPTSTAx.
• TX_COMPLT: Transmitted IN Data Complete Clear

0 = no effect.
1 = clear the TX_COMPLT flag of UDPHS_EPTSTAx.
• RX_SETUP/ERR_FL_ISO: Received SETUP/Error Flow Clear

0 = no effect.
1 = clear the RX_SETUP/ERR_FL_ISO flags of UDPHS_EPTSTAx.
• STALL_SNT/ERR_NBTRA: Stall Sent/Number of Transaction Error Clear

0 = no effect.
1 = clear the STALL_SNT/ERR_NBTRA flags of UDPHS_EPTSTAx.
818 AT91SAM9G45
• NAK_IN/ERR_FLUSH: NAKIN/Bank Flush Error Clear
0 = no effect.
1 = clear the NAK_IN/ERR_FLUSH flags of UDPHS_EPTSTAx.
• NAK_OUT: NAKOUT Clear

0 = no effect.
1 = clear the NAK_OUT flag of UDPHS_EPTSTAx.
819 AT91SAM9G45
37.6.17 UDPHS Endpoint Status Register
Name: UDPHS_EPTSTAx [x=0..6]
Addresses: 0xFFF7811C [0], 0xFFF7813C [1], 0xFFF7815C [2], 0xFFF7817C [3], 0xFFF7819C [4],
0xFFF781BC [5], 0xFFF781DC [6]
31 30 29 28 27 26 25 24
SHRT_PCKT BYTE_COUNT
23 22 21 20 19 18 17 16
CURRENT_BANK/
BYTE_COUNT BUSY_BANK_STA
CONTROL_DIR
15 14 13 12 11 10 9 8
STALL_SNT/
NAK_IN/ RX_SETUP/ TX_PK_RDY/ RX_BK_RDY/
NAK_OUT ERR_CRISO/ TX_COMPLT ERR_OVFLW
ERR_FLUSH ERR_FL_ISO ERR_TRANS KILL_BANK
ERR_NBTRA
7 6 5 4 3 2 1 0
TOGGLESQ_STA FRCESTALL – – – – –
• FRCESTALL: Stall Handshake Request

0 = no effect.
1= If set a STALL answer will be done to the host for the next handshake.
This bit is reset by hardware upon received SETUP.
• TOGGLESQ_STA: Toggle Sequencing

Toggle Sequencing:
– IN endpoint: it indicates the PID Data Toggle that will be used for the next packet sent. This is not relative to
the current bank.
– CONTROL and OUT endpoint:
These bits are set by hardware to indicate the PID data of the current bank:
00 Data0
01 Data1
10 Data2 (only for High Bandwidth Isochronous Endpoint)
11 MData (only for High Bandwidth Isochronous Endpoint)
Note 1: In OUT transfer, the Toggle information is meaningful only when the current bank is busy (Received OUT
Data = 1).
Note 2: These bits are updated for OUT transfer:
– a new data has been written into the current bank.
– the user has just cleared the Received OUT Data bit to switch to the next bank.
Note 3: For High Bandwidth Isochronous Out endpoint, it is recommended to check the UDPHS_EPTSTAx/ERR_TRANS
bit to know if the toggle sequencing is correct or not.
820 AT91SAM9G45
Note 4: This field is reset to DATA1 by the UDPHS_EPTCLRSTAx register TOGGLESQ bit, and by UDPHS_EPTCTLDISx
(disable endpoint).
• ERR_OVFLW: Overflow Error

This bit is set by hardware when a new too-long packet is received.
Example: If the user programs an endpoint 64 bytes wide and the host sends 128 bytes in an OUT transfer, then the Over-
flow Error bit is set.
This bit is updated at the same time as the BYTE_COUNT field.
This bit is reset by UDPHS_EPTRST register EPT_x (reset endpoint) and by UDPHS_EPTCTLDISx (disable endpoint).
• RX_BK_RDY/KILL_BANK: Received OUT Data/KILL Bank

– Received OUT Data: (For OUT endpoint or Control endpoint)
This bit is set by hardware after a new packet has been stored in the endpoint FIFO.
This bit is cleared by the device firmware after reading the OUT data from the endpoint.
For multi-bank endpoints, this bit may remain active even when cleared by the device firmware, this if an other packet has
been received meanwhile.
Hardware assertion of this bit may generate an interrupt if enabled by the UDPHS_EPTCTLx register RX_BK_RDY bit.
– KILL Bank: (For IN endpoint)
– the bank is really cleared or the bank is sent, BUSY_BANK_STA is decremented.
– the bank is not cleared but sent on the IN transfer, TX_COMPLT
– the bank is not cleared because it was empty. The user should wait that this bit is cleared before trying to clear
another packet.
Note: “Kill a packet” may be refused if at the same time, an IN token is coming and the current packet is sent on the
UDPHS line. In this case, the TX_COMPLT bit is set. Take notice however, that if at least two banks are ready to be sent,
there is no problem to kill a packet even if an IN token is coming. In fact, in that case, the current bank is sent (IN transfer)
and the last bank is killed.
• TX_COMPLT: Transmitted IN Data Complete

This bit is set by hardware after an IN packet has been transmitted for isochronous endpoints and after it has been
accepted (ACK’ed) by the host for Control, Bulk and Interrupt endpoints.
This bit is reset by UDPHS_EPTRST register EPT_x (reset endpoint), and by UDPHS_EPTCTLDISx (disable endpoint).
• TX_PK_RDY/ERR_TRANS: TX Packet Ready/Transaction Error

– TX Packet Ready:
This bit is cleared by hardware, as soon as the packet has been sent for isochronous endpoints, or after the host has
acknowledged the packet for Control, Bulk and Interrupt endpoints.
For Multi-bank endpoints, this bit may remain clear even after software is set if another bank is available to transmit.
Hardware clear of this bit may generate an interrupt if enabled by the UDPHS_EPTCTLx register TX_PK_RDY bit.
– Transaction Error: (For high bandwidth isochronous OUT endpoints) (Read-Only)
This bit is set by hardware when a transaction error occurs inside one microframe.
821 AT91SAM9G45
If one toggle sequencing problem occurs among the n-transactions (n = 1, 2 or 3) inside a microframe, then this bit is still
set as long as the current bank contains one “bad” n-transaction. (see “CURRENT_BANK/CONTROL_DIR: Current
Bank/Control Direction” on page 823) As soon as the current bank is relative to a new “good” n-transactions, then this bit is
reset.
Note1: A transaction error occurs when the toggle sequencing does not respect the Universal Serial Bus Specification, Rev
2.0 (5.9.2 High Bandwidth Isochronous endpoints) (Bad PID, missing data....)
Note2: When a transaction error occurs, the user may empty all the “bad” transactions by clearing the Received OUT Data
flag (RX_BK_RDY).
If this bit is reset, then the user should consider that a new n-transaction is coming.
• RX_SETUP/ERR_FL_ISO: Received SETUP/Error Flow

– Received SETUP: (for Control endpoint only)
This bit is set by hardware when a valid SETUP packet has been received from the host.
It is cleared by the device firmware after reading the SETUP data from the endpoint FIFO.
– Error Flow: (for isochronous endpoint only)
This bit is set by hardware when a transaction error occurs.
– Isochronous IN transaction is missed, the micro has no time to fill the endpoint (underflow).
– Isochronous OUT data is dropped because the bank is busy (overflow).
• STALL_SNT/ERR_CRISO/ERR_NBTRA: Stall Sent/CRC ISO Error/Number of Transaction Error

– STALL_SNT: (for Control, Bulk and Interrupt endpoints)
This bit is set by hardware after a STALL handshake has been sent as requested by the UDPHS_EPTSTAx register
FRCESTALL bit.
– ERR_CRISO: (for Isochronous OUT endpoints) (Read-only)
This bit is set by hardware if the last received data is corrupted (CRC error on data).
This bit is updated by hardware when new data is received (Received OUT Data bit).
– ERR_NBTRA: (for High Bandwidth Isochronous IN endpoints)
This bit is set at the end of a microframe in which at least one data bank has been transmitted, if less than the number of
transactions per micro-frame banks (UDPHS_EPTCFGx register NB_TRANS) have been validated for transmission inside
this microframe.
• NAK_IN/ERR_FLUSH: NAK IN/Bank Flush Error

– NAK_IN:
This bit is set by hardware when a NAK handshake has been sent in response to an IN request from the Host.
This bit is cleared by software.
822 AT91SAM9G45
– ERR_FLUSH: (for High Bandwidth Isochronous IN endpoints)
This bit is set when flushing unsent banks at the end of a microframe.
This bit is reset by UDPHS_EPTRST register EPT_x (reset endpoint) and by EPT_CTL_DISx (disable endpoint).
• NAK_OUT: NAK OUT

This bit is set by hardware when a NAK handshake has been sent in response to an OUT or PING request from the Host.
This bit is reset by UDPHS_EPTRST register EPT_x (reset endpoint) and by EPT_CTL_DISx (disable endpoint).
• CURRENT_BANK/CONTROL_DIR: Current Bank/Control Direction

– Current Bank: (all endpoints except Control endpoint)
These bits are set by hardware to indicate the number of the current bank.
00 Bank 0 (or single bank)

01 Bank 1
10 Bank 2
11 Invalid
Note: the current bank is updated each time the user:

– Sets the TX Packet Ready bit to prepare the next IN transfer and to switch to the next bank.
– Clears the received OUT data bit to access the next bank.
– Control Direction: (for Control endpoint only)
0 = a Control Write is requested by the Host.
1 = a Control Read is requested by the Host.
Note1: This bit corresponds with the 7th bit of the bmRequestType (Byte 0 of the Setup Data).
Note2: This bit is updated after receiving new setup data.
• BUSY_BANK_STA: Busy Bank Number

These bits are set by hardware to indicate the number of busy banks.
IN endpoint: it indicates the number of busy banks filled by the user, ready for IN transfer.
OUT endpoint: it indicates the number of busy banks filled by OUT transaction from the Host.
00 All banks are free

01 1 busy bank
10 2 busy banks
11 3 busy banks
823 AT91SAM9G45
• BYTE_COUNT: UDPHS Byte Count
Byte count of a received data packet.
This field is incremented after each write into the endpoint (to prepare an IN transfer).
This field is decremented after each reading into the endpoint (OUT transfer).
This field is also updated at RX_BK_RDY flag clear with the next bank.
This field is also updated at TX_PK_RDY flag set with the next bank.
This field is reset by EPT_x of UDPHS_EPTRST register.
• SHRT_PCKT: Short Packet

An OUT Short Packet is detected when the receive byte count is less than the configured UDPHS_EPTCFGx register
EPT_Size.
This bit is updated at the same time as the BYTE_COUNT field.
824 AT91SAM9G45
37.6.18 UDPHS DMA Channel Transfer Descriptor
The DMA channel transfer descriptor is loaded from the memory.
Be careful with the alignment of this buffer.
The structure of the DMA channel transfer descriptor is defined by three parameters as
described below:
Offset 0:
The address must be aligned: 0xXXXX0
Next Descriptor Address Register: UDPHS_DMANXTDSCx
Offset 4:
DMA Channelx Address Register: UDPHS_DMAADDRESSx
Offset 8:
DMA Channelx Control Register: UDPHS_DMACONTROLx
To use the DMA channel transfer descriptor, fill the structures with the correct value (as
described in the following pages).
Then write directly in UDPHS_DMANXTDSCx the address of the descriptor to be used first.
Then write 1 in the LDNXT_DSC bit of UDPHS_DMACONTROLx (load next channel transfer
descriptor). The descriptor is automatically loaded upon Endpointx request for packet transfer.
825 AT91SAM9G45
37.6.19 UDPHS DMA Next Descriptor Address Register
Name: UDPHS_DMANXTDSCx [x = 1..5]
31 30 29 28 27 26 25 24
NXT_DSC_ADD
23 22 21 20 19 18 17 16
NXT_DSC_ADD
15 14 13 12 11 10 9 8
NXT_DSC_ADD
7 6 5 4 3 2 1 0
NXT_DSC_ADD
• NXT_DSC_ADD
This field points to the next channel descriptor to be processed. This channel descriptor must be aligned, so bits 0 to 3 of
the address must be equal to zero.
826 AT91SAM9G45
37.6.20 UDPHS DMA Channel Address Register
Name: UDPHS_DMAADDRESSx [x = 1..5]
31 30 29 28 27 26 25 24
BUFF_ADD
23 22 21 20 19 18 17 16
BUFF_ADD
15 14 13 12 11 10 9 8
BUFF_ADD
7 6 5 4 3 2 1 0
BUFF_ADD
• BUFF_ADD
This field determines the AHB bus starting address of a DMA channel transfer.
Channel start and end addresses may be aligned on any byte boundary.
The firmware may write this field only when the UDPHS_DMASTATUS register CHANN_ENB bit is clear.
This field is updated at the end of the address phase of the current access to the AHB bus. It is incrementing of the access
byte width. The access width is 4 bytes (or less) at packet start or end, if the start or end address is not aligned on a word
boundary.
The packet start address is either the channel start address or the next channel address to be accessed in the channel
buffer.
The packet end address is either the channel end address or the latest channel address accessed in the channel buffer.
The channel start address is written by software or loaded from the descriptor, whereas the channel end address is either
determined by the end of buffer or the UDPHS device, USB end of transfer if the UDPHS_DMACONTROLx register
END_TR_EN bit is set.
827 AT91SAM9G45
37.6.21 UDPHS DMA Channel Control Register
Name: UDPHS_DMACONTROLx [x = 1..5]
31 30 29 28 27 26 25 24
BUFF_LENGTH
23 22 21 20 19 18 17 16
BUFF_LENGTH
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
BURST_LCK DESC_LD_IT END_BUFFIT END_TR_IT END_B_EN END_TR_EN LDNXT_DSC CHANN_ENB
• CHANN_ENB (Channel Enable Command)

0 = DMA channel is disabled at and no transfer will occur upon request. This bit is also cleared by hardware when the chan-
nel source bus is disabled at end of buffer.
If the UDPHS_DMACONTROL register LDNXT_DSC bit has been cleared by descriptor loading, the firmware will have to
set the corresponding CHANN_ENB bit to start the described transfer, if needed.
If the UDPHS_DMACONTROL register LDNXT_DSC bit is cleared, the channel is frozen and the channel registers may
then be read and/or written reliably as soon as both UDPHS_DMASTATUS register CHANN_ENB and CHANN_ACT flags
read as 0.
If a channel request is currently serviced when this bit is cleared, the DMA FIFO buffer is drained until it is empty, then the
UDPHS_DMASTATUS register CHANN_ENB bit is cleared.
If the LDNXT_DSC bit is set at or after this bit clearing, then the currently loaded descriptor is skipped (no data transfer
occurs) and the next descriptor is immediately loaded.
1 = UDPHS_DMASTATUS register CHANN_ENB bit will be set, thus enabling DMA channel data transfer. Then any pend-
ing request will start the transfer. This may be used to start or resume any requested transfer.
• LDNXT_DSC: Load Next Channel Transfer Descriptor Enable (Command)

0 = no channel register is loaded after the end of the channel transfer.
1 = the channel controller loads the next descriptor after the end of the current transfer, i.e. when the
UDPHS_DMASTATUS/CHANN_ENB bit is reset.
If the UDPHS_DMA CONTROL/CHANN_ENB bit is cleared, the next descriptor is immediately loaded upon transfer
request.
DMA Channel Control Command Summary
LDNXT_DSC CHANN_ENB Description

0 0 Stop now
0 1 Run and stop at end of buffer
1 0 Load next descriptor now
1 1 Run and link at end of buffer
828 AT91SAM9G45
• END_TR_EN: End of Transfer Enable (Control)
Used for OUT transfers only.
0 = USB end of transfer is ignored.
1 = UDPHS device can put an end to the current buffer transfer.
When set, a BULK or INTERRUPT short packet or the last packet of an ISOCHRONOUS (micro) frame (DATAX) will close
the current buffer and the UDPHS_DMASTATUSx register END_TR_ST flag will be raised.
This is intended for UDPHS non-prenegotiated end of transfer (BULK or INTERRUPT) or ISOCHRONOUS microframe
data buffer closure.
• END_B_EN: End of Buffer Enable (Control)

0 = DMA Buffer End has no impact on USB packet transfer.
1 = endpoint can validate the packet (according to the values programmed in the UDPHS_EPTCTLx register AUTO_VALID
and SHRT_PCKT fields) at DMA Buffer End, i.e. when the UDPHS_DMASTATUS register BUFF_COUNT reaches 0.
This is mainly for short packet IN validation initiated by the DMA reaching end of buffer, but could be used for OUT packet
truncation (discarding of unwanted packet data) at the end of DMA buffer.
• END_TR_IT: End of Transfer Interrupt Enable

0 = UDPHS device initiated buffer transfer completion will not trigger any interrupt at UDPHS_STATUSx/END_TR_ST
rising.
1 = an interrupt is sent after the buffer transfer is complete, if the UDPHS device has ended the buffer transfer.
Use when the receive size is unknown.
• END_BUFFIT: End of Buffer Interrupt Enable

0 = UDPHS_DMA_STATUSx/END_BF_ST rising will not trigger any interrupt.
1 = an interrupt is generated when the UDPHS_DMASTATUSx register BUFF_COUNT reaches zero.
• DESC_LD_IT: Descriptor Loaded Interrupt Enable

0 = UDPHS_DMASTATUSx/DESC_LDST rising will not trigger any interrupt.
1 = an interrupt is generated when a descriptor has been loaded from the bus.
• BURST_LCK: Burst Lock Enable

0 = the DMA never locks bus access.
1 = USB packets AHB data bursts are locked for maximum optimization of the bus bandwidth usage and maximization of
fly-by AHB burst duration.
• BUFF_LENGTH: Buffer Byte Length (Write-only)

This field determines the number of bytes to be transferred until end of buffer. The maximum channel transfer size (64
KBytes) is reached when this field is 0 (default value). If the transfer size is unknown, this field should be set to 0, but the
transfer end may occur earlier under UDPHS device control.
When this field is written, The UDPHS_DMASTATUSx register BUFF_COUNT field is updated with the write value.
Note: Bits [31:2] are only writable when issuing a channel Control Command other than “Stop Now”.
Note: For reliability it is highly recommended to wait for both UDPHS_DMASTATUSx register CHAN_ACT and CHAN_ENB flags are
at 0, thus ensuring the channel has been stopped before issuing a command other than “Stop Now”.
829 AT91SAM9G45
37.6.22 UDPHS DMA Channel Status Register
Name: UDPHS_DMASTATUSx [x = 1..5]
Addresses: 0xFFF7832C [1], 0xFFF7833C [2], 0xFFF7834C [3], 0xFFF7835C [4], 0xFFF7836C [5]
31 30 29 28 27 26 25 24
BUFF_COUNT
23 22 21 20 19 18 17 16
BUFF_COUNT
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– DESC_LDST END_BF_ST END_TR_ST – – CHANN_ACT CHANN_ENB
• CHANN_ENB: Channel Enable Status

0 = if cleared, the DMA channel no longer transfers data, and may load the next descriptor if the UDPHS_DMACONTROLx
register LDNXT_DSC bit is set.
When any transfer is ended either due to an elapsed byte count or a UDPHS device initiated transfer end, this bit is auto-
matically reset.
1 = if set, the DMA channel is currently enabled and transfers data upon request.
This bit is normally set or cleared by writing into the UDPHS_DMACONTROLx register CHANN_ENB bit field either by soft-
ware or descriptor loading.
If a channel request is currently serviced when the UDPHS_DMACONTROLx register CHANN_ENB bit is cleared, the
DMA FIFO buffer is drained until it is empty, then this status bit is cleared.
• CHANN_ACT: Channel Active Status

0 = the DMA channel is no longer trying to source the packet data.
When a packet transfer is ended this bit is automatically reset.
1 = the DMA channel is currently trying to source packet data, i.e. selected as the highest-priority requesting channel.
When a packet transfer cannot be completed due to an END_BF_ST, this flag stays set during the next channel descriptor
load (if any) and potentially until UDPHS packet transfer completion, if allowed by the new descriptor.
• END_TR_ST: End of Channel Transfer Status

0 = cleared automatically when read by software.
1 = set by hardware when the last packet transfer is complete, if the UDPHS device has ended the transfer.
Valid until the CHANN_ENB flag is cleared at the end of the next buffer transfer.
• END_BF_ST: End of Channel Buffer Status

1 = set by hardware when the BUFF_COUNT downcount reach zero.
830 AT91SAM9G45
AT91SAM9G45
• DESC_LDST: Descriptor Loaded Status

1 = set by hardware when a descriptor has been loaded from the system bus.
• BUFF_COUNT: Buffer Byte Count

This field determines the current number of bytes still to be transferred for this buffer.
This field is decremented from the AHB source bus access byte width at the end of this bus address phase.
The access byte width is 4 by default, or less, at DMA start or end, if the start or end address is not aligned on a word
boundary.
At the end of buffer, the DMA accesses the UDPHS device only for the number of bytes needed to complete it.
This field value is reliable (stable) only if the channel has been stopped or frozen (UDPHS_EPTCTLx register
NT_DIS_DMA bit is used to disable the channel request) and the channel is no longer active CHANN_ACT flag is 0.
Note: For OUT endpoints, if the receive buffer byte length (BUFF_LENGTH) has been defaulted to zero because the USB transfer
length is unknown, the actual buffer byte length received will be 0x10000-BUFF_COUNT.
831
832 AT91SAM9G45
AT91SAM9G45
38. Image Sensor Interface (ISI)
38.1 Description
The Image Sensor Interface (ISI) connects a CMOS-type image sensor to the processor and
provides image capture in various formats. It does data conversion, if necessary, before the stor-
age in memory through DMA.
The ISI supports color CMOS image sensor and grayscale image sensors with a reduced set of
functionalities.
In grayscale mode, the data stream is stored in memory without any processing and so is not
compatible with the LCD controller.
Internal FIFOs on the preview and codec paths are used to store the incoming data. The RGB
output on the preview path is compatible with the LCD controller. This module outputs the data
in RGB format (LCD compatible) and has scaling capabilities to make it compliant to the LCD
display resolution (See Table 38-3 on page 836).
Several input formats such as preprocessed RGB or YCbCr are supported through the data bus
interface.
It supports two modes of synchronization:
1. The hardware with ISI_VSYNC and ISI_HSYNC signals
2. The International Telecommunication Union Recommendation ITU-R BT.656-4 Start-of-
Active-Video (SAV) and End-of-Active-Video (EAV) synchronization sequence.
Using EAV/SAV for synchronization reduces the pin count (ISI_VSYNC, ISI_HSYNC not used).
The polarity of the synchronization pulse is programmable to comply with the sensor signals.
Table 38-1. I/O Description

Signal Dir Description
ISI_VSYNC IN Vertical Synchronization
ISI_HSYNC IN Horizontal Synchronization
ISI_DATA[11..0] IN Sensor Pixel Data
ISI_MCK OUT Master Clock Provided to the Image Sensor
ISI_PCK IN Pixel Clock Provided by the Image Sensor

• ITU-R BT. 601/656 8-bit mode external interface support
• Support for ITU-R BT.656-4 SAV and EAV synchronization
• Vertical and horizontal resolutions up to 2048 x 2048
• Preview Path up to 640*480
• Support for packed data formatting for YCbCr 4:2:2 formats
• Preview scaler to generate smaller size image
833
Figure 38-1. ISI Connection Example
Image Sensor Image Sensor Interface
data[11..0] ISI_DATA[11..0]
CLK ISI_MCK
PCLK ISI_PCK
VSYNC ISI_VSYNC
HSYNC ISI_HSYNC
38.3 Block Diagram
Figure 38-2. Image Sensor Interface Block Diagram
APB bus
Hsync/Len Timing Signals
Vsync/Fen Interface APB
Config
Camera Interface
Registers
Interrupt
Controller Camera
Interrupt Request Line
APB
CCIR-656
From Clock Domain
Embedded Timing
Decoder(SAV/EAV) Rx buffers
Pixel AHB
CMOS Clock Domain Clock Domain
sensor Frame Rate
Pixel input Clipping + Color Rx Direct
up to 12 bit 2-D Image Pixel Camera
Conversion Display
AHB bus
Scaler Formatter AHB
YCbCr 4:2:2 YCC to RGB FIFO Core
Pixel Sampling Master
8:8:8 Module Video Interface
RGB 5:6:5 Arbiter Scatter
Clipping + Color Packed Rx Direct Mode
Conversion Formatter Capture Support
RGB to YCC FIFO
CMOS
sensor
pixel clock codec_on
input
834 AT91SAM9G45
AT91SAM9G45

The Image Sensor Interface (ISI) supports direct connection to the ITU-R BT. 601/656 8-bit
mode compliant sensors and up to 12-bit grayscale sensors. It receives the image data stream
from the image sensor on the 12-bit data bus.
This module receives up to 12 bits for data, the horizontal and vertical synchronizations and the
pixel clock. The reduced pin count alternative for synchronization is supported for sensors that
embed SAV (start of active video) and EAV (end of active video) delimiters in the data stream.
The Image Sensor Interface interrupt line is connected to the Advanced Interrupt Controller and
can trigger an interrupt at the beginning of each frame and at the end of a DMA frame transfer. If
the SAV/EAV synchronization is used, an interrupt can be triggered on each delimiter event.
For 8-bit color sensors, the data stream received can be in several possible formats: YCbCr
4:2:2, RGB 8:8:8, RGB 5:6:5 and may be processed before the storage in memory. The data
stream may be sent on both preview path and codec path if the bit ISI_CDC in the ISI_CTRL is
one. To optimize the bandwidth, the codec path should be enabled only when a capture is
required.
In grayscale mode, the input data stream is stored in memory without any processing. The 12-bit
data, which represent the grayscale level for the pixel, is stored in memory one or two pixels per
word, depending on the GS_MODE bit in the ISI_CFG2 register. The codec datapath is not
available when grayscale image is selected.
A frame rate counter allows users to capture all frames or 1 out of every 2 to 8 frames.
38.4.1 Data Timing

The two data timings using horizontal and vertical synchronization and EAV/SAV sequence syn-
chronization are shown in Figure 38-3 and Figure 38-4.
In the VSYNC/HSYNC synchronization, the valid data is captured with the active edge of the
pixel clock (ISI_PCK), after SFD lines of vertical blanking and SLD pixel clock periods delay pro-
grammed in the control register.
The ITU-RBT.656-4 defines the functional timing for an 8-bit wide interface.
There are two timing reference signals, one at the beginning of each video data block SAV
(0xFF000080) and one at the end of each video data block EAV(0xFF00009D). Only data sent
between EAV and SAV is captured. Horizontal blanking and vertical blanking are ignored. Use of
the SAV and EAV synchronization eliminates the ISI_VSYNC and ISI_HSYNC signals from the
interface, thereby reducing the pin count. In order to retrieve both frame and line synchronization
properly, at least one line of vertical blanking is mandatory.
835
AT91SAM9G45
Figure 38-3. HSYNC and VSYNC Synchronization
Frame
ISI_VSYNC
1 line
ISI_HSYNC
ISI_PCK
DATA[7..0] Y Cb Y Cr Y Cb Y Cr Y Cb Y Cr
Figure 38-4. SAV and EAV Sequence Synchronization
ISII_PCK
DATA[7..0] FF 00 00 80 Y Cb Y Cr Y Cb Y Cr Y Y Cr Y Cb FF 00 00 9D
SAV Active Video EAV
38.4.2 Data Ordering

The RGB color space format is required for viewing images on a display screen preview, and the
YCbCr color space format is required for encoding.
All the sensors do not output the YCbCr or RGB components in the same order. The ISI allows
the user to program the same component order as the sensor, reducing software treatments to
restore the right format.
Table 38-2. Data Ordering in YCbCr Mode

Mode Byte 0 Byte 1 Byte 2 Byte 3
Default Cb(i) Y(i) Cr(i) Y(i+1)
Mode1 Cr(i) Y(i) Cb(i) Y(i+1)
Mode2 Y(i) Cb(i) Y(i+1) Cr(i)
Mode3 Y(i) Cr(i) Y(i+1) Cb(i)
Table 38-3. RGB Format in Default Mode, RGB_CFG = 00, No Swap

Mode Byte D7 D6 D5 D4 D3 D2 D1 D0
Byte 0 R7(i) R6(i) R5(i) R4(i) R3(i) R2(i) R1(i) R0(i)
Byte 1 G7(i) G6(i) G5(i) G4(i) G3(i) G2(i) G1(i) G0(i)
RGB 8:8:8
Byte 2 B7(i) B6(i) B5(i) B4(i) B3(i) B2(i) B1(i) B0(i)
Byte 3 R7(i+1) R6(i+1) R5(i+1) R4(i+1) R3(i+1) R2(i+1) R1(i+1) R0(i+1)
836
AT91SAM9G45
Table 38-3. RGB Format in Default Mode, RGB_CFG = 00, No Swap

Byte 0 R4(i) R3(i) R2(i) R1(i) R0(i) G5(i) G4(i) G3(i)
Byte 1 G2(i) G1(i) G0(i) B4(i) B3(i) B2(i) B1(i) B0(i)
RGB 5:6:5
Byte 2 R4(i+1) R3(i+1) R2(i+1) R1(i+1) R0(i+1) G5(i+1) G4(i+1) G3(i+1)
Byte 3 G2(i+1) G1(i+1) G0(i+1) B4(i+1) B3(i+1) B2(i+1) B1(i+1) B0(i+1)
Table 38-4. RGB Format, RGB_CFG = 10 (Mode 2), No Swap

Byte 0 G2(i) G1(i) G0(i) R4(i) R3(i) R2(i) R1(i) R0(i)
Byte 1 B4(i) B3(i) B2(i) B1(i) B0(i) G5(i) G4(i) G3(i)
RGB 5:6:5
Byte 2 G2(i+1) G1(i+1) G0(i+1) R4(i+1) R3(i+1) R2(i+1) R1(i+1) R0(i+1)
Byte 3 B4(i+1) B3(i+1) B2(i+1) B1(i+1) B0(i+1) G5(i+1) G4(i+1) G3(i+1)
Table 38-5. RGB Format in Default Mode, RGB_CFG = 00, Swap Activated
Byte 0 R0(i) R1(i) R2(i) R3(i) R4(i) R5(i) R6(i) R7(i)
Byte 1 G0(i) G1(i) G2(i) G3(i) G4(i) G5(i) G6(i) G7(i)
RGB 8:8:8
Byte 2 B0(i) B1(i) B2(i) B3(i) B4(i) B5(i) B6(i) B7(i)
Byte 3 R0(i+1) R1(i+1) R2(i+1) R3(i+1) R4(i+1) R5(i+1) R6(i+1) R7(i+1)
Byte 0 G3(i) G4(i) G5(i) R0(i) R1(i) R2(i) R3(i) R4(i)
Byte 1 B0(i) B1(i) B2(i) B3(i) B4(i) G0(i) G1(i) G2(i)
RGB 5:6:5
Byte 2 G3(i+1) G4(i+1) G5(i+1) R0(i+1) R1(i+1) R2(i+1) R3(i+1) R4(i+1)
Byte 3 B0(i+1) B1(i+1) B2(i+1) B3(i+1) B4(i+1) G0(i+1) G1(i+1) G2(i+1)
The RGB 5:6:5 input format is processed to be displayed as RGB 5:6:5 format, compliant with
the 16-bit mode of the LCD controller.
38.4.3 Clocks
The sensor master clock (ISI_MCK) can be generated either by the Advanced Power Manage-
ment Controller (APMC) through a Programmable Clock output or by an external oscillator
connected to the sensor.
None of the sensors embed a power management controller, so providing the clock by the
APMC is a simple and efficient way to control power consumption of the system.
Care must be taken when programming the system clock. The ISI has two clock domains, the
system bus clock and the pixel clock provided by sensor. The two clock domains are not syn-
chronized, but the system clock must be faster than pixel clock.
837
AT91SAM9G45
38.4.4 Preview Path
38.4.4.1 Scaling, Decimation (Subsampling)

This module resizes captured 8-bit color sensor images to fit the LCD display format. The resize
module performs only downscaling. The same ratio is applied for both horizontal and vertical
resize, then a fractional decimation algorithm is applied.
The decimation factor is a multiple of 1/16 and values 0 to 15 are forbidden.
Table 38-6. Decimation Factor

Dec value 0->15 16 17 18 19 ... 124 125 126 127
Dec Factor X 1 1.063 1.125 1.188 ... 7.750 7.813 7.875 7.938
Table 38-7. Decimation and Scaler Offset Values

INPUT
352*288 640*480 800*600 1280*1024 1600*1200 2048*1536
OUTPUT
VGA
F NA 16 20 32 40 51
640*480
QVGA
F 16 32 40 64 80 102
320*240
CIF
F 16 26 33 56 66 85
352*288
QCIF
F 32 53 66 113 133 170
176*144
Example:
Input 1280*1024 Output = 640*480
Hratio = 1280/640 = 2
Vratio = 1024/480 = 2.1333
The decimation factor is 2 so 32/16.
838
AT91SAM9G45
Figure 38-5. Resize Examples

1280 32/16 decimation
640
1024 480
1280 56/16 decimation
352
1024 288
38.4.4.2 Color Space Conversion

This module converts YCrCb or YUV pixels to RGB color space. Clipping is performed to ensure
that the samples value do not exceed the allowable range. The conversion matrix is defined
below and is fully programmable:
R C0 0 C1 Y – Y off
G = C 0 – C 2 – C 3 × C b – C boff
B C0 C4 0 C r – C roff
Example of programmable value to convert YCrCb to RGB:
R = 1,164 ⋅ ( Y – 16 ) + 1,596 ⋅ ( C r – 128 )

G = 1,164 ⋅ ( Y – 16 ) – 0,813 ⋅ ( C r – 128 ) – 0,392 ⋅ ( C b – 128
B = 1,164 ⋅ ( Y – 16 ) + 2,107 ⋅ ( C b – 128 )
An example of programmable value to convert from YUV to RGB:
⎧ R = Y + 1,596 ⋅ V
G = Y – 0,394 ⋅ U – 0,436 ⋅ V
⎩ B = Y + 2,032 ⋅ U
839
AT91SAM9G45
38.4.4.3 Memory Interface

Preview datapath contains a data formatter that converts 8:8:8 pixel to RGB 5:6:5 format compli-
ant with 16-bit format of the LCD controller. In general, when converting from a color channel
with more bits to one with fewer bits, formatter module discards the lower-order bits. Example:
Converting from RGB 8:8:8 to RGB 5:6:5, it discards the three LSBs from the red and blue chan-
nels, and two LSBs from the green channel. When grayscale mode is enabled, two memory
formats are supported. One mode supports 2 pixels per word, and the other mode supports 1
pixel per word.
Table 38-8. Grayscale Memory Mapping Configuration for 12-bit Data

GS_MODE DATA[31:24] DATA[23:16] DATA[15:8] DATA[7:0]
0 P_0[11:4] P_0[3:0], 0000 P_1[11:4] P_1[3:0], 0000
1 P_0[11:4] P_0[3:0], 0000 0 0
38.4.4.4 FIFO and DMA Features

Both preview and Codec datapaths contain FIFOs. These asynchronous buffers are used to
safely transfer formatted pixels from Pixel clock domain to AHB clock domain. A video arbiter is
used to manage FIFO thresholds and triggers a relevant DMA request through the AHB master
interface. Thus, depending on FIFO state, a specified length burst is asserted. Regarding AHB
master interface, it supports Scatter DMA mode through linked list operation. This mode of oper-
ation improves flexibility of image buffer location and allows the user to allocate two or more
frame buffers. The destination frame buffers are defined by a series of Frame Buffer Descriptors
(FBD). Each FBD controls the transfer of one entire frame and then optionally loads a further
FBD to switch the DMA operation at another frame buffer address. The FBD is defined by a
series of three words. The first one defines the current frame buffer address (named
DMA_X_ADDR register), the second defines control information (named DMA_X_CTRL regis-
ter) and the third defines the next descriptor address (named DMA_X_DSCR). DMA transfer
mode with linked list support is available for both codec and preview datapath. The data to be
transferred described by an FBD requires several burst accesses. In the example below, the use
of 2 ping-pong frame buffers is described.
38.4.4.5 Example
The first FBD, stored at address 0x00030000, defines the location of the first frame buffer. This
address is programmed in the ISI user interface DMA_P_DSCR. To enable Descriptor fetch
operation DMA_P_CTRL register must be set to 0x00000001. LLI_0 and LLI_1 are the two
descriptors of the Linked list.
Destination Address: frame buffer ID0 0x02A000 (LLI_0.DMA_P_ADDR)
Transfer 0 Control Information, fetch and writeback: 0x00000003 (LLI_0.DMA_P_CTRL)
Next FBD address: 0x00030010 (LLI_0.DMA_P_DSCR)
Second FBD, stored at address 0x00030010, defines the location of the second frame buffer.
Destination Address: frame buffer ID1 0x0003A000 (LLI_1.DMA_P_ADDR
Transfer 1 Control information fetch and writeback: 0x00000003 (LLI_1.DMA_P_CTRL)
Next FBD address: 0x00030000, wrapping to first FBD (LLI_1.DMA_P_DSCR)
Using this technique, several frame buffers can be configured through the linked list. Figure 38-6
illustrates a typical three frame buffer application. Frame n is mapped to frame buffer 0, frame
840
AT91SAM9G45
n+1 is mapped to frame buffer 1, frame n+2 is mapped to Frame buffer 2, further frames wrap. A
codec request occurs, and the full-size 4:2:2 encoded frame is stored in a dedicated memory
space.
Figure 38-6. Three Frame Buffers Application and Memory Mapping

Codec Done
Codec Request
frame n-1 frame n frame n+1 frame n+2 frame n+3 frame n+4
Memory Space
Frame Buffer 3
Frame Buffer 0
LCD
Frame Buffer 1
ISI config Space
4:2:2 Image
Full ROI
38.4.5 Codec Path
38.4.5.1 Color Space Conversion

Depending on user selection, this module can be bypassed so that input YCrCb stream is
directly connected to the format converter module. If the RGB input stream is selected, this mod-
ule converts RGB to YCrCb color space with the formulas given below:
Y C0 C1 C2 R Y off
Cr = C3 –C4 –C5 × G + Cr off
Cb –C6 –C7 C8 B Cb off
An example of coefficients is given below:
Y = 0,257 ⋅ R + 0,504 ⋅ G + 0,098 ⋅ B + 16

C r = 0,439 ⋅ R – 0,368 ⋅ G – 0,071 ⋅ B + 128
C b = – 0,148 ⋅ R – 0,291 ⋅ G + 0,439 ⋅ B + 128
841
AT91SAM9G45
38.4.5.2 Memory Interface

Dedicated FIFOs are used to support packed memory mapping. YCrCb pixel components are
sent in a single 32-bit word in a contiguous space (packed). Data is stored in the order of natural
scan lines. Planar mode is not supported.
38.4.5.3 DMA Features

Like preview datapath, codec datapath DMA mode uses linked list operation.
842
AT91SAM9G45
38.5 Image Sensor Interface (ISI) User Interface

Table 38-9. I Register Mapping
Offset Register Name Register Access Reset Value
0x00 ISI Configuration 1 Register ISI_CFG1 Read-write 0x00000000
0x04 ISI Configuration 2 Register ISI_CFG2 Read-write 0x00000000
0x08 ISI Preview Size Register ISI_PSIZE Read-write 0x00000000
0x0C ISI Preview Decimation Factor Register ISI_PDECF Read-write 0x00000010
0x10 ISI CSC YCrCb To RGB Set 0 Register ISI_Y2R_SET0 Read-write 0x6832cc95
0x14 ISI CSC YCrCb To RGB Set 1 Register ISI_Y2R_SET1 Read-write 0x00007102
0x18 ISI CSC RGB To YCrCb Set 0 Register ISI_R2Y_SET0 Read-write 0x01324145
0x1C ISI CSC RGB To YCrCb Set 1 Register ISI_R2Y_SET1 Read-write 0x01245e38
0x20 ISI CSC RGB To YCrCb Set 2 Register ISI_R2Y_SET2 Read-write 0x01384a4b
0x24 ISI Control Register ISI_CTRL Write 0x00000000
0x28 ISI Status Register ISI_STATUS Read 0x00000000
0x2C ISI Interrupt Enable Register ISI_INTEN Write 0x00000000
0x30 ISI Interrupt Disable Register ISI_INTDIS Write 0x00000000
0x34 ISI Interrupt Mask Register ISI_INTMASK Read 0x00000000
0x38 DMA Channel Enable Register DMA_CHER Write 0x00000000
0x3C DMA Channel Disable Register DMA_CHDR Write 0x00000000
0x40 DMA Channel Status Register DMA_CHSR Read 0x00000000
0x44 DMA Preview Base Address Register DMA_P_ADDR Read-write 0x00000000
0x48 DMA Preview Control Register DMA_P_CTRL Read-write 0x00000000
0x4C DMA Preview Descriptor Address Register DMA_P_DSCR Read-write 0x00000000
0x50 DMA Codec Base Address Register DMA_C_ADDR Read-write 0x00000000
0x54 DMA Codec Control Register DMA_C_CTRL Read-write 0x00000000
0x58 DMA Codec Descriptor Address Register DMA_C_DSCR Read-write 0x00000000
0xE4 Write Protection Control Register ISI_WPCR Read-write 0x00000000

0xE8 Write Protection Status Register ISI_WPSR Read 0x00000000
0x44-0xF8 Reserved – – –
0xFC Reserved – – –
Note: Several parts of the ISI controller use the pixel clock provided by the image sensor (ISI_PCK). Thus the user must first program
the image sensor to provide this clock (ISI_PCK) before programming the Image Sensor Controller.
843
AT91SAM9G45
38.5.1 ISI Configuration 1 Register

Register Name: ISI_CFG1
Reset Value: 0x00000000
31 30 29 28 27 26 25 24
SFD
23 22 21 20 19 18 17 16
SLD
15 14 13 12 11 10 9 8
– THMASK FULL DISCR FRATE
7 6 5 4 3 2 1 0
CRC_SYNC EMB_SYNC – PIXCLK_POL VSYNC_POL HSYNC_POL – –
• HSYNC_POL: Horizontal Synchronization Polarity

0: HSYNC active high.
1: HSYNC active low.
• VSYNC_POL: Vertical Synchronization Polarity

0: VSYNC active high.
1: VSYNC active low.
• PIXCLK_POL: Pixel Clock Polarity

0: Data is sampled on rising edge of pixel clock.
1: Data is sampled on falling edge of pixel clock.
• EMB_SYNC: Embedded Synchronization

0: Synchronization by HSYNC, VSYNC.
1: Synchronization by embedded synchronization sequence SAV/EAV.
• CRC_SYNC: Embedded Synchronization Correction

0: No CRC correction is performed on embedded synchronization.
1: CRC correction is performed. if the correction is not possible, the current frame is discarded and the CRC_ERR is set in
the status register.
• FRATE: Frame Rate [0..7]

0: All the frames are captured, else one frame every FRATE+1 is captured.
• DISCR: Disable Codec Request

0 = Codec datapath DMA interface requires a request to restart.
1 = Codec datapath DMA automatically restarts.
844
AT91SAM9G45
• FULL: Full Mode is Allowed

1: Both codec and preview datapaths are working simultaneously.
• THMASK: Threshold Mask

0: Only 4 beats AHB burst are allowed.
1: Only 4 and 8 beats AHB burst are allowed.
2: 4, 8 and 16 beats AHB burst are allowed.
• SLD: Start of Line Delay

SLD pixel clock periods to wait before the beginning of a line.
• SFD: Start of Frame Delay

SFD lines are skipped at the beginning of the frame.
845
AT91SAM9G45
38.5.2 ISI Configuration 2 Register

Register Name: ISI_CFG2
31 30 29 28 27 26 25 24
RGB_CFG YCC_SWAP - IM_HSIZE
23 22 21 20 19 18 17 16
IM_HSIZE
15 14 13 12 11 10 9 8
COL_SPACE RGB_SWAP GRAYSCALE RGB_MODE GS_MODE IM_VSIZE
7 6 5 4 3 2 1 0
IM_VSIZE
• IM_VSIZE: Vertical Size of the Image Sensor [0..2047]:

Vertical size = IM_VSIZE + 1.
• GS_MODE:
0: 2 pixels per word.
1: 1 pixel per word.
• RGB_MODE: RGB Input Mode:

0: RGB 8:8:8 24 bits.
1: RGB 5:6:5 16 bits.
• GRAYSCALE:
0: Grayscale mode is disabled.
1: Input image is assumed to be grayscale coded.
• RGB_SWAP:
0: D7 -> R7.
1: D0 -> R7.
The RGB_SWAP has no effect when the grayscale mode is enabled.
• COL_SPACE: Color Space for the Image Data

0: YCbCr.
1: RGB.
• IM_HSIZE: Horizontal Size of the Image Sensor [0..2047]

Horizontal size = IM_HSIZE + 1.
846
AT91SAM9G45
• YCC_SWAP: Defines the YCC Image Data
YCC_SWAP Byte 0 Byte 1 Byte 2 Byte 3

00: Default Cb(i) Y(i) Cr(i) Y(i+1)
01: Mode1 Cr(i) Y(i) Cb(i) Y(i+1)
10: Mode2 Y(i) Cb(i) Y(i+1) Cr(i)
11: Mode3 Y(i) Cr(i) Y(i+1) Cb(i)
• RGB_CFG: Defines RGB Pattern when RGB_MODE is set to 1
RGB_CFG Byte 0 Byte 1 Byte 2 Byte 3

00: Default R/G(MSB) G(LSB)/B R/G(MSB) G(LSB)/B
01: Mode1 B/G(MSB) G(LSB)/R B/G(MSB) G(LSB)/R
10: Mode2 G(LSB)/R B/G(MSB) G(LSB)/R B/G(MSB)
11: Mode3 G(LSB)/B R/G(MSB) G(LSB)/B R/G(MSB)
If RGB_MODE is set to RGB 8:8:8, then RGB_CFG = 0 implies RGB color sequence, else it implies BGR color sequence.
847
AT91SAM9G45
38.5.3 ISI Preview Register

Register Name: ISI_PSIZE
31 30 29 28 27 26 25 24
– – – – – – PREV_HSIZE
23 22 21 20 19 18 17 16
PREV_HSIZE
15 14 13 12 11 10 9 8
– – – – – – PREV_VSIZE
7 6 5 4 3 2 1 0
PREV_VSIZE
• PREV_VSIZE: Vertical Size for the Preview Path

Vertical Preview size = PREV_VSIZE + 1 (480 max only in RGB mode).
• PREV_HSIZE: Horizontal Size for the Preview Path

Horizontal Preview size = PREV_HSIZE + 1 (640 max only in RGB mode).
848
AT91SAM9G45
38.5.4 ISI Preview Decimation Factor Register

Register Name: ISI_PDECF
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
DEC_FACTOR
• DEC_FACTOR: Decimation Factor

DEC_FACTOR is 8-bit width, range is from 16 to 255. Values from 0 to 16 do not perform any decimation.
849
AT91SAM9G45
38.5.5 ISI Color Space Conversion YCrCb to RGB Set 0 Register

Register Name: ISI_Y2R_SET0
Reset Value: 0x6832cc95
31 30 29 28 27 26 25 24
C3
23 22 21 20 19 18 17 16
C2
15 14 13 12 11 10 9 8
C1
7 6 5 4 3 2 1 0
C0
• C0: Color Space Conversion Matrix Coefficient C0

C0 element default step is 1/128, ranges from 0 to 1.9921875.



850
AT91SAM9G45
38.5.6 ISI Color Space Conversion YCrCb to RGB Set 1 Register

Register Name: ISI_Y2R_SET1
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– Cboff Croff Yoff – – – C4
C4

• Yoff: Color Space Conversion Luminance Default Offset

0: No offset.
1: Offset = 128.
• Croff: Color Space Conversion Red Chrominance Default Offset

0: No offset.
1: Offset = 16.
• Cboff: Color Space Conversion Blue Chrominance Default Offset

0: No offset.
1: Offset = 16.
851
AT91SAM9G45
38.5.7 ISI Color Space Conversion RGB to YCrCb Set 0 Register

Register Name: ISI_R2Y_SET0
31 30 29 28 27 26 25 24
– – – – – – – Roff
23 22 21 20 19 18 17 16
C2
15 14 13 12 11 10 9 8
C1
7 6 5 4 3 2 1 0
C0

C0 element default step is 1/256, from 0 to 0.49609375.


• Roff: Color Space Conversion Red Component Offset

0: No offset.
1: Offset = 16.
852
AT91SAM9G45

Reset Value: 0x01245e38
31 30 29 28 27 26 25 24
– – – – – – – Goff
23 22 21 20 19 18 17 16
C5
15 14 13 12 11 10 9 8
C4
7 6 5 4 3 2 1 0
C3



• Goff: Color Space Conversion Green Component Offset

0: No offset.
1: Offset = 128.
853
AT91SAM9G45

Reset Value: 0x01384a4b
31 30 29 28 27 26 25 24
– – – – – – – Boff
23 22 21 20 19 18 17 16
C8
15 14 13 12 11 10 9 8
C7
7 6 5 4 3 2 1 0
C6



• Boff: Color Space Conversion Blue Component Offset

0: No offset.
1: Offset = 128.
854
AT91SAM9G45
38.5.10 ISI Control Register

Register Name: ISI_CTRL
Access Type: Write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – ISI_CDC
7 6 5 4 3 2 1 0
– – – – – ISI_SRST ISI_DIS ISI_EN
• ISI_EN: ISI Module Enable Request

Write one to this field to enable the module. Software must poll ENABLE field in the ISI_STATUS register to verify that the
command has successfully completed.
• ISI_DIS: ISI Module Disable Request

Write one to this field to disable the module. If both ISI_EN and ISI_DIS are asserted at the same time, the disable request
is not taken into account. Software must poll DIS_DONE field in the ISI_STATUS register to verify that the command has
successfully completed.
• ISI_SRST: ISI Software Reset Request

Write one to this field to request a software reset of the module. Software must poll SRST field in the ISI_STATUS register
to verify that the software request command has terminated.
• ISI_CDC: ISI Codec Request

Write one to this field to enable the codec datapath and capture a Full resolution Frame. A new request cannot be taken
into account while CDC_PND bit is active in the ISI_STATUS register.
855
AT91SAM9G45
38.5.11 ISI Status Register

Register Name: ISI_SR
Access Type: Read
31 30 29 28 27 26 25 24
– – – – FR_OVR CRC_ERR C_OVR P_OVR
23 22 21 20 19 18 17 16
– – – – SIP – CXFR_DONE PXFR_DONE
15 14 13 12 11 10 9 8
– – – – – VSYNC – CDC_PND
7 6 5 4 3 2 1 0
– – – – – SRST DIS_DONE ENABLE
• ENABLE (this bit is a status bit)

0: Module is enabled.
1: Module is disabled.
• DIS_DONE: Module Disable Request has Terminated

1: Disable request has completed. This flag is reset after a read operation.
• SRST: Module Software Reset Request has Terminated

1: Software reset request has completed. This flag is reset after a read operation.
• CDC_PND: Pending Codec Request (this bit is a status bit)

0: Indicates that no Codec request is pending.
1: Indicates that the request has been taken into account but cannot be serviced within the current frame. The operation is
postponed to the next frame.
• VSYNC: Vertical Synchronization

1: Indicates that a Vertical synchronization has been detected since the last read of the status register.
• PXFR_DONE: Preview DMA Transfer has Terminated.

When set to one, this bit indicates that the DATA transfer on the preview channel has completed. This flag is reset after a
read operation.
• CXFR_DONE: Codec DMA Transfer has Terminated.

When set to one, this bit indicates that the DATA transfer on the codec channel has completed. This flag is reset after a
read operation.
• SIP: Synchronization in Progress (this is a status bit)

When the status of the preview or codec DMA channel is modified, a minimum amount of time is required to perform the
clock domain synchronization. This bit is set when this operation occurs. No modification of the channel status is allowed
when this bit is set, to guarantee data integrity.
856
AT91SAM9G45
• P_OVR: Preview Datapath Overflow

0: No overflow
1: An overrun condition has occurred in input FIFO on the preview path. The overrun happens when the FIFO is full and an
attempt is made to write a new sample to the FIFO. This flag is reset after a read operation.
• C_OVR: Codec Datapath Overflow

0: No overflow
1: An overrun condition has occurred in input FIFO on the codec path. The overrun happens when the FIFO is full and an
attempt is made to write a new sample to the FIFO. This flag is reset after a read operation.
• CRC_ERR: CRC Synchronization Error

0: No CRC error in the embedded synchronization frame (SAV/EAV)
1: The CRC_SYNC is enabled in the control register and an error has been detected and not corrected. The frame is dis-
carded and the ISI waits for a new one. This flag is reset after a read operation.
• FR_OVR: Frame Rate Overrun

0: No frame overrun.
1: Frame overrun, the current frame is being skipped because a vsync signal has been detected while flushing FIFOs. This
flag is reset after a read operation.
857
AT91SAM9G45
38.5.12 ISI Interrupt Enable Register

Register Name: ISI_IER
Reset Value: 0x0
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
– – – – – – CXFR_DONE PXFR_DONE
15 14 13 12 11 10 9 8
– – – – – VSYNC – –
7 6 5 4 3 2 1 0
– – – – – SRST DIS_DONE –
• DIS_DONE: Disable Done Interrupt Enable
• SRST: Software Reset Interrupt Enable
• VSYNC: Vertical Synchronization Interrupt Enable
• PXFR_DONE: Preview DMA Transfer Done Interrupt Enable
• CXFR_DONE: Codec DMA Transfer Done Interrupt Enable
• P_OVR: Preview Datapath Overflow Interrupt Enable
• C_OVR: Codec Datapath Overflow Interrupt Enable
• CRC_ERR: Embedded Synchronization CRC Error Interrupt Enable
• FR_OVR: Frame Rate Overflow Interrupt Enable
858
AT91SAM9G45
38.5.13 ISI Interrupt Disable Register

Register Name: ISI_IDR
Reset Value: 0x0
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
– – – – – VSYNC – –
7 6 5 4 3 2 1 0
• DIS_DONE: Disable Done Interrupt Disable
• SRST: Software Reset Interrupt Disable
• VSYNC: Vertical Synchronization Interrupt Disable
• PXFR_DONE: Preview DMA Transfer Done Interrupt Disable
• CXFR_DONE: Codec DMA Transfer Done Interrupt Disable
• P_OVR: Preview Datapath Overflow Interrupt Disable
• C_OVR: Codec Datapath Overflow Interrupt Disable
• CRC_ERR: Embedded Synchronization CRC Error Interrupt Disable
• FR_OVR: Frame Rate Overflow Interrupt Disable
859
AT91SAM9G45
38.5.14 ISI Interrupt Mask Register

Register Name: ISI_IMR
Reset Value: 0x0
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
– – – – – VSYNC – –
7 6 5 4 3 2 1 0
• DIS_DONE: Module Disable Operation Completed

0: The disable completed interrupt is disabled.
1: The disable completed interrupt is enabled.
• SRST: Software Reset Completed

0: The software reset completed interrupt is disabled.
1: The software reset completed interrupt is enabled.
• VSYNC: Vertical Synchronization

0: The vertical synchronization interrupt is enabled.
1: The vertical synchronization interrupt is disabled.
• PXFR_DONE: Preview DMA Transfer Interrupt

0: The Preview DMA transfer completed interrupt is enabled
1: The Preview DMA transfer completed interrupt is disabled
• CXFR_DONE: Codec DMA Transfer Interrupt

0: The Codec DMA transfer completed interrupt is enabled
1: The Codec DMA transfer completed interrupt
• P_OVR: FIFO Preview Overflow

0: The preview FIFO overflow interrupt is disabled.
1: The preview FIFO overflow interrupt is enabled.
• P_OVR: FIFO Codec Overflow

0: The codec FIFO overflow interrupt is disabled.
1: The codec FIFO overflow interrupt is enabled.
860
AT91SAM9G45
• CRC_ERR: CRC Synchronization Error

0: The crc error interrupt is disabled.
1: The crc error interrupt is enabled.
• FR_OVR: Frame Rate Overrun
0: The frame overrun interrupt is disabled.
1: The frame overrun interrupt is enabled.
861
AT91SAM9G45
38.5.15 DMA Channel Enable Register

Register Name: DMA_CHER
Access Type: Write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – C_CH_EN P_CH_EN
• P_CH_EN: Preview Channel Enable

Write one to this field to enable the preview DMA channel.
• C_CH_EN: Codec Channel Enable

Write one to this field to enable the codec DMA channel.
862
AT91SAM9G45
38.5.16 DMA Channel Disable Register

Register Name: DMA_CHDR
Access Type: Write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – C_CH_DIS P_CH_DIS
• P_CH_DIS
Write one to this field to disable the channel. Poll P_CH_S in DMA_CHSR to verify that the preview channel status has
been successfully modified.
• C_CH_DIS
Write one to this field to disabled the channel. Poll C_CH_S in DMA_CHSR to verify that the codec channel status has
been successfully modified.
863
AT91SAM9G45
38.5.17 DMA Channel Status Register

Register Name: DMA_CHSR
Access Type: Read
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – C_CH_S P_CH_S
• P_CH_S:
0: indicates that the Preview DMA channel is disabled
1: indicates that the Preview DMA channel is enabled.
• C_CH_S:
0: indicates that the Codec DMA channel is disabled.
1: indicates that the Codec DMA channel is enabled.
864
AT91SAM9G45
38.5.18 DMA Preview Base Address Register

Register Name: DMA_P_ADDR
31 30 29 28 27 26 25 24
P_ADDR
23 22 21 20 19 18 17 16
P_ADDR
15 14 13 12 11 10 9 8
P_ADDR
7 6 5 4 3 2 1 0
P_ADDR – –
• P_ADDR: Preview Image Base Address. (This address is word aligned.)
865
AT91SAM9G45
38.5.19 DMA Preview Control Register

Register Name: DMA_P_CTRL
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – P_DONE P_IEN P_WB P_FETCH
• P_FETCH: Descriptor Fetch Control Field

0: Preview channel fetch operation is disabled.
1: Preview channel fetch operation is enabled.
• P_WB: Descriptor Writeback Control Field

0: Preview channel writeback operation is disabled.
1: Preview channel writeback operation is enabled.
• P_IEN: Transfer Done Flag Control

0: Preview Transfer done flag generation is enabled.
1: Preview Transfer done flag generation is disabled.
• P_DONE: (This field is only updated in the memory.)

0: The transfer related to this descriptor has not been performed.
1: The transfer related to this descriptor has completed. This field is updated in memory at the end of the transfer, when
writeback operation is enabled.
866
AT91SAM9G45
38.5.20 DMA Preview Descriptor Address Register

Register Name: DMA_P_DSCR
31 30 29 28 27 26 25 24
P_DSCR
23 22 21 20 19 18 17 16
P_DSCR
15 14 13 12 11 10 9 8
P_DSCR
7 6 5 4 3 2 1 0
P_DSCR – –
• P_DSCR: Preview Descriptor Base Address (This address is word aligned.)
867
AT91SAM9G45
38.5.21 DMA Codec Base Address Register

Register Name: DMA_C_ADDR
31 30 29 28 27 26 25 24
C_ADDR
23 22 21 20 19 18 17 16
C_ADDR
15 14 13 12 11 10 9 8
C_ADDR
7 6 5 4 3 2 1 0
C_ADDR – –
• C_ADDR: Codec Image Base Address (This address is word aligned.)
868
AT91SAM9G45
38.5.22 DMA Codec Control Register

Register Name: DMA_C_CTRL
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – C_DONE C_IEN C_WB C_FETCH
• C_FETCH: Descriptor Fetch Control Field

0: Codec channel fetch operation is disabled.
1: Codec channel fetch operation is enabled.
• C_WB: Descriptor Writeback Control Field

0: Codec channel writeback operation is disabled.
1: Codec channel writeback operation is enabled.
• C_IEN: Transfer Done flag control

0: Codec Transfer done flag generation is enabled.
1: Codec Transfer done flag generation is disabled.
• C_DONE: (This field is only updated in the memory.)

0: The transfer related to this descriptor has not been performed.
1: The transfer related to this descriptor has completed. This field is updated in memory at the end of the transfer, when.
writeback operation is enabled.
869
AT91SAM9G45
38.5.23 DMA Codec Descriptor Address Register

Register Name: DMA_C_DSCR
31 30 29 28 27 26 25 24
C_DSCR
23 22 21 20 19 18 17 16
C_DSCR
15 14 13 12 11 10 9 8
C_DSCR
7 6 5 4 3 2 1 0
C_DSCR – –
• C_DSCR: Codec Descriptor Base Address (This address is word aligned.)
870
AT91SAM9G45
38.5.24 ISI Write Protection Control

Register Name: ISI_WPCR
31 30 29 28 27 26 25 24
WP_KEY (0x49 => “I”)
23 22 21 20 19 18 17 16
WP_KEY (0x53 => “S”)
15 14 13 12 11 10 9 8
WP_KEY (0x49 => “I”)
7 6 5 4 3 2 1 0
WP_EN
• WP_PEN: Write Protection Enable

0 = Disables the Write Protection if WP_KEY corresponds.
1 = Enables the Write Protection if WP_KEY corresponds.
• WP_KEY: Write Protection KEY Password

Should be written at value 0x495349 (ASCII code for “ISI”). Writing any other value in this field has no effect.
871
AT91SAM9G45
38.5.25 ISI Write Protection Status

Register Name: ISI_WPSR
31 30 29 28 27 26 25 24
- - - - - - - -
23 22 21 20 19 18 17 16
WP_VSRC
15 14 13 12 11 10 9 8
WP_VSRC
7 6 5 4 3 2 1 0
- - - - WP_VS
• WP_VSRC: Write Protection Violation Status

WP_VS
0 0 0 0
register (WP_SR).
Write Protection detected unauthorized attempt to write a control
0 0 0 1
register had occurred (since the last read).
Software reset had been performed while Write Protection was
0 0 1 0
Both Write Protection violation and software reset with Write
0 0 1 1
Protection enabled had occurred since the last read.
• WP_VSRC: Write Protection Violation Source

WP_VSRC
0 0 0 0
register (WP_SR).
Write access in ISI_CFG1 while Write Protection was enabled
0 0 0 1
Write access in ISI_CFG2 while Write Protection was enabled
0 0 1 0
Write access in ISI_PSIZE while Write Protection was enabled
0 0 1 1
Write access in ISI_PDECF while Write Protection was enabled
0 1 0 0
Write access in ISI_Y2R_SET0 while Write Protection was
0 1 0 1
Write access in ISI_Y2R_SET1 while Write Protection was
0 1 1 0
Write access in ISI_R2Y_SET0 while Write Protection was
0 1 1 1
1 0 0 0
1 0 0 1
872
AT91SAM9G45
39. Touch Screen ADC Controller (TSADCC)
39.1 Description
The Touch Screen ADC Controller is based on a Successive Approximation Register (SAR) 10-
bit Analog-to-Digital Converter (ADC). It also integrates:
• a 8-to-1 analog multiplexer for analog-to-digital conversions of up to 8 analog lines
• 4 power switches that measure both axis positions on the resistive touch screen panel
• 1 additional power switch and an embedded resistor that detects pen-interrupt and pen loss
The conversions extend from 0V to TSADVREF.
The TSADCC supports an 8-bit or 10-bit resolution mode, and conversion results are reported in
a common register for all channels, as well as in a channel-dedicated register.
Conversions can be started for all enabled channels, either by a software trigger, by detection of
a rising edge on the external trigger pin TSADTRG or by an integrated programmable timer.
When the Touch Screen is enabled, a timer-triggered sequencer automatically configures the
power switches, performs the conversions and stores the results in dedicated registers.
The TSADCC also integrates a Sleep Mode and a Pen-Detect Mode and connects with one
PDC channel. These features reduce both power consumption and processor intervention.
The TSADCC timings, like the Startup Time and Sample and Hold Time, are fully configurable.

• 8-channel ADC
• Support 4-wire resistive Touch Screen
• 10-bit 384 Ksamples/sec. Successive Approximation Register ADC
• -3/+3 LSB Integral Non Linearity, -2/+2 LSB Differential Non Linearity
• Integrated 8-to-1 multiplexer, offering eight independent 3.3V analog inputs
• External voltage reference for better accuracy on low voltage inputs
• Individual enable and disable of each channel
• Multiple trigger sources
– Hardware or software trigger
– External trigger pin
• Sleep Mode and conversion sequencer
– Automatic wakeup on trigger and back to sleep mode after conversions of all
enabled channels
873
39.3 Block Diagram
Figure 39-1. TSADCC Block Diagram
TSADC
VDDANA
TSADCC
Memory
Controller
Trigger
ADTRG Selection Timer PDC
PIO Touch Screen ADC Control

Sequencer Logic
User
Interface Peripheral
Bridge
APB
AD0XP
AD1XM Touch
Successive
Analog Multiplexer
Screen
Approximation
Switches
AD2YP Register TSADC
Analog-to-Digital Clock
AD3YM Converter PMC
GPAD4
....
GPADx
TSADC
Interrupt
ADVREF AIC
GND
GPADx: last general-purpose ADC channel defined by the number of channels
874 AT91SAM9G45
AT91SAM9G45
39.4 Signal Description

Table 39-1. TSADCC Pin Description
Pin Name Description
VDDANA Analog power supply
TSADVREF Reference voltage
AD0XP Analog input channel 0 or Touch Screen Top channel
AD1XM Analog input channel 1 or Touch Screen Bottom channel
AD2YP Analog input channel 2 or Touch Screen Right channel
AD3YM Analog input channel 3 or Touch Screen Left channel
GPAD4 - GPAD7 General-purpose analog input channels 4 to 7
TSADTRG External trigger

The TSADC controller is not continuously clocked. The programmer must first enable the
TSADC controller Clock in the Power Management Controller (PMC) before using the TSADC
controller. However, if the application does not require TSADC controller operations, the TSADC
controller clock can be stopped when not needed and be restarted later.
Configuring the TSADC controller does not require the TSADC controller clock to be enabled.

The TSADCC interrupt line is connected on one of the internal sources of the Advanced Inter-
rupt Controller. Using the TSADCC interrupt requires the AIC to be programmed first.

Instance ID
TSADCC 20
39.5.3 Analog Inputs

The analog input pins can be multiplexed with PIO lines. In this case, the assignment of the
TSADCC input is automatically done as soon as the corresponding channel is enabled by writing
the register TSADCC_CHER. By default, after reset, the PIO lines are configured as input with
its pull-up enabled and the TSADCC inputs are connected to the GND.
39.5.4 I/O Lines

The pin TSADTRG may be shared with other peripheral functions through the PIO Controller. In
this case, the PIO Controller should be set accordingly to assign the pin TSADTRG to the
TSADCC function.

TSADCC TSADTRG PD28 A
875
39.5.5 Conversion Performances
For performance and electrical characteristics of the TSADCC, see the section “Electrical Char-
acteristics” of the full datasheet.
39.6 Analog-to-digital Converter Functional Description

The TSADCC embeds a Successive Approximation Register (SAR) Analog-to-Digital Converter
(ADC). The ADC supports 8-bit or 10-bit resolutions.
The conversion is performed on a full range between 0V and the reference voltage pin TSAD-
VREF. Analog inputs between these voltages convert to values based on a linear conversion.
39.6.1 ADC Resolution

The ADC supports 8-bit or 10-bit resolutions. The 8-bit selection is performed by setting the bit
LOWRES in the TSADCC Mode Register. See Section 39.11.2 “TSADCC Mode Register” on
page 894.
By default, after a reset, the resolution is the highest and the DATA field in the “TSADCC Chan-
nel Data Register x (x = 0..7)” are fully used.
By setting the bit LOWRES, the ADC switches in the lowest resolution and the conversion
results can be read in the eight lowest significant bits of the data registers. The two highest bits
of the DATA field in the corresponding TSADCC_CDR register and of the LDATA field in the
TSADCC_LCDR register read 0.
Moreover, when a PDC channel is connected to the TSADCC, 10-bit resolution sets the transfer
request sizes to 16-bit. Setting the bit LOWRES automatically switches to 8-bit data transfers. In
this case, the destination buffers are optimized.
All the conversions for the Touch Screen forces the ADC in 10-bit resolution, regardless of the
LOWRES setting. Further details are given in the section “Operating Modes” on page 884.
39.6.2 ADC Clock

The TSADCC uses the ADC Clock to perform conversions. Converting a single analog value to
a 10-bit digital data requires Sample and Hold Clock cycles as defined in the field SHTIM of the
“TSADCC Mode Register” and 10 ADC Clock cycles. The ADC Clock frequency is selected in
the PRESCAL field of the “TSADCC Mode Register”.
The ADC clock range is between MCK/2, if PRESCAL is 0, and MCK/128, if PRESCAL is set to
63 (0x3F). PRESCAL must be programmed in order to provide an ADC clock frequency accord-
ing to the maximum sampling rate parameter given in the Electrical Characteristics section.
39.6.3 Sleep Mode

The TSADCC Sleep Mode maximizes power saving by automatically deactivating the Analog-to-
Digital Converter cell when it is not being used for conversions. Sleep Mode is enabled by set-
ting the bit SLEEP in “TSADCC Mode Register”.
The SLEEP of the ADC is automatically managed by the conversion sequencer, which can auto-
matically process the conversions of all channels at lowest power consumption.
When a trigger occurs, the Analog-to-Digital Converter cell is automatically activated. As the
analog cell requires a start-up time, the logic waits during this time and then starts the conver-
sion on the enabled channels. When all conversions are complete, the ADC is deactivated until
the next trigger.
876 AT91SAM9G45
AT91SAM9G45
39.6.4 Startup Time

The Touch Screen ADC has a minimal Startup Time when it exits the Sleep Mode. As the ADC
Clock depends on the application, the user has to program the field STARTUP in the “TSADCC
Mode Register”, which defines how many ADC Clock cycles to wait before performing the first
conversion of the sequence.
The field STARTUP can define a Startup Time between 8 and 1024 ADC Clock cycles by steps
of 8.
The user must assure that ADC Startup Time given in the section “Electrical Characteristics” is
covered by this wait time.
39.6.5 Sample and Hold Time

In the same way, a minimal Sample and Hold Time is necessary for the TSADCC to guarantee
the best converted final value between selection of two channels. This time depends on the input
impedance of the analog input, but also on the output impedance of the driver providing the sig-
nal to the analog input, as there is no input buffer amplifier.
The Sample and Hold time has to be programmed through the bitfields SHTIM in the “TSADCC
Mode Register” and TSSHTIM in the “TSADCC Touch Screen Register”.
The field SHTIM defines the number of ADC Clock cycles for an analog input, while the field
TSSHTIM defines the number of ADC Clock cycles for a Touch Screen input.
These both fields can define a Sample and Hold time between 1 and 16 ADC Clock cycles.
The field TSSHTIM defines also the time the power switches of the Touch Screen are closed
when the TSADCC performs a conversion for the Touch Screen.
39.7 Touch Screen
39.7.1 Resistive Touch Screen Principles

A resistive touch screen is based on two resistive films, each one being fitted with a pair of elec-
trodes, placed at the top and bottom on one film, and on the right and left on the other. Between
the two, there is a layer that acts as an insulator, but also enables contact when you press the
screen. This is illustrated in Figure 39-2.
The TSADC controller has the ability to perform without external components:
• Position Measurement
• Pressure Measurement
• Pen Detection
877
Figure 39-2. Touch Screen Position Measurement
Pen
Contact
XP
YM YP
XM
VDD VDD
XP YP
YP XP
Volt Volt
XM YM
GND GND
Vertical Position Detection Horizontal Position Detection
39.7.2 Position Measurement Method

As shown in Figure 39-2, to detect the position of a contact, a supply is first applied from top to
bottom. Due to the linear resistance of the film, there is a voltage gradient from top to bottom.
When a contact is performed on the screen, the voltage propagates at the point the two surfaces
come into contact with the second film. If the input impedance on the right and left electrodes
sense is high enough, the film does not affect this voltage, despite its resistive nature.
For the horizontal direction, the same method is used, but by applying supply from left to right.
The range depends on the supply voltage and on the loss in the switches that connect to the top
and bottom electrodes.
In an ideal world (linear, with no loss through switches), the horizontal position is equal to:
VYM / VDD or VYP / VDD.
The proposed implementation with on-chip power switches is shown in Figure 39-3. The voltage
measurement at the output of the switch compensates for the switches loss.
It is possible to correct for the switch loss by performing the operation:
[VYP - VXM] / [VXP - VXM].
This requires additional measurements, as shown in Figure 39-3.
878 AT91SAM9G45
AT91SAM9G45
Figure 39-3. Touch Screen Switches Implementation
XP VDDANA
XM GND
To the ADC
YP VDDANA
YM GND
VDDANA VDDANA
Switch Switch
Resistor Resistor
XP YP
YP XP
XM YM
Switch Switch
Resistor Resistor
GND GND
Horizontal Position Detection Vertical Position Detection
39.7.3 Pressure Measurement Method

The method to measure the pressure (Rp) applied to the touch screen is based on the knowl-
edge of the X-Panel resistance (Rxp).
Three conversions (Xpos,Z1,Z2) are necessary to determine the value of Rp (Zaxis resistance).
Rp = Rxp*(Xpos/1024)*[(Z2/Z1)-1]
879
VDDANA VDDANA VDDANA
Switch Switch Switch

Resistor Resistor Resistor
Open
circuit
XP YP XP YP XP YP
Rp Rp Rp
XM YM XM YM XM YM
Open Open
Switch circuit Switch circuit Switch
Resistor Resistor Resistor
GND GND GND

XPos Measure(Yp) Z1 Measure(Xp) Z2 Measure(Xp)
39.7.4 Pen Detect Method

When there is no contact, it is not necessary to perform conversion. However, it is important to
detect a contact by keeping the power consumption as low as possible.
The proposed implementation polarizes the vertical panel by closing the switch on XP and ties
the horizontal panel by an embedded resistor connected to YM. This resistor is enabled by a fifth
switch. Since there is no contact, no current is flowing and there is no related power consump-
tion. As soon as a contact occurs, a current is flowing in the touch screen and a schmitt trigger
detects the voltage in the resistor.
The Touch Screen Interrupt configuration is entered by programming the bit PENDET in the
“TSADCC Mode Register”. If this bit is written at 1, the switch on XP and the switch on the resis-
tor are both closed, except when a touch screen conversion is in progress.
To complete the circuit, a programmable debouncer is placed at the output of the schmitt trigger.
This debouncer is programmable at 1 ADC Clock period, useful when the system is running at
Slow Clock, or at up to 2 15 ADC Clock periods, but better used to filter noise on the Touch
Screen panel when the system is running at high speed. The debouncer length can be selected
by programming the field PENDBC in “TSADCC Mode Register”.
880 AT91SAM9G45
AT91SAM9G45
Figure 39-4. Touch Screen Pen Detect
XP VDDANA
XM GND
To the ADC
YP VDDANA
YM GND
PENDBC
Debouncer Pen Interrupt
GND
The Touch Screen Pen Detect can be used to generate a TSADCC interrupt to wake up the sys-
tem or it can be programmed to trig a conversion, so that a position can be measured as soon as
a contact is detected if the TSADCC is programmed for an operating mode involving the Touch
Screen.
The Pen Detect generates two types of status, reported in the “TSADCC Status Register”:
• the bit PENCNT is set as soon as a current flows for a time over the debouncing time as
defined by PENDBC and remains set until TSADCC_SR is read.
• the bit NOCNT is set as soon as no current flows for a time over the debouncing time as
defined by PENDBC and remains set until TSADCC_SR is read.
Both bits are automatically cleared as soon as the Status Register TSADCC_SR is read, and
can generate an interrupt by writing accordingly the “TSADCC Interrupt Enable Register”.
881
39.8 Conversion Results
When a conversion is completed, the resulting 8-bit or 10-bit digital value is right-aligned and
stored in the “TSADCC Channel Data Register x (x = 0..7)” of the current channel and in the
“TSADCC Last Converted Data Register”.
The channel EOC bit and the bit DRDY in the “TSADCC Status Register” are both set. If the
PDC channel is enabled, DRDY rising triggers a data transfer. In any case, either EOC and
DRDY can trigger an interrupt.
Reading one of the “TSADCC Channel Data Register x (x = 0..7)” registers clears the corre-
sponding EOC bit.
Reading “TSADCC Last Converted Data Register” clears the DRDY bit and the EOC bit corre-
sponding to the last converted channel.
Figure 39-5. EOCx and DRDY Flag Behavior

Write the ADC_CR Write the ADC_CR
with START = 1 Read the ADC_CDRx Read the ADC_LCDR
with START = 1
CHx
(ADC_CHSR)
EOCx
(ADC_SR)
SHTIM Conversion SHTIM Conversion

Time Time
DRDY
(ADC_SR)
If the “TSADCC Channel Data Register x (x = 0..7)” is not read before further incoming data is
converted, the corresponding Overrun Error (OVRE) flag is set in the “TSADCC Status
Register”.
In the same way, new data converted when DRDY is high sets the bit GOVRE (General Overrun
Error) in the “TSADCC Status Register”.
The OVRE and GOVRE flags are automatically cleared when the “TSADCC Status Register”is
read.
882 AT91SAM9G45
AT91SAM9G45
Figure 39-6. GOVRE and OVREx Flag Behavior
Read ADC_SR
ADTRG
CH0
(ADC_CHSR)
CH1
(ADC_CHSR)
ADC_LCDR Undefined Data Data A Data B Data C
ADC_CDR0 Undefined Data Data A Data C
ADC_CDR1 Undefined Data Data B
EOC0 SHTIM Conversion

(ADC_SR) Read ADC_CDR0
SHTIM Conversion
EOC1 SHTIM Conversion Read ADC_CDR1

(ADC_SR)
GOVRE
(ADC_SR)
DRDY
(ADC_SR)
OVRE0
(ADC_SR)
Warning: If the corresponding channel is disabled during a conversion or if it is disabled and

then reenabled during a conversion, its associated data and its corresponding EOC and OVRE
flags in TSADCC_SR are unpredictable.
883
AT91SAM9G45
39.9 Conversion Triggers

Conversions of the active analog channels are started with a software or a hardware trigger.
The software trigger is provided by writing the “TSADCC Control Register” with the bit START at
1.
The hardware trigger can be selected by the filed TRGMOD in the TSADCC Trigger Register
(TSADCC_TRGR) between:
• an edge, either rising or falling or any, detected on the external trigger pin TSADTRG
• the Pen Detect, depending on how the PENDET bit is set in the “TSADCC Mode Register”
• a continuous trigger, meaning the TSADCC restarts the next sequence as soon as it finishes
the current one, in this case, only one software trigger is required at the beginning
• a periodic trigger, which is defined by programming the field TRGPER in the “TSADCC
Trigger Register”
Enabling hardware triggers does not disable the software trigger functionality. Thus, if a hard-
ware trigger is selected, the start of a conversion can still be initiated by the software trigger.
39.10 Operating Modes

The Touch Screen ADC Controller features several operating modes, each defining a conver-
sion sequence:
• The ADC Mode: at each trigger, all the enabled channels are converted
• The Touch Screen Mode: at each trigger, the touch screen inputs are converted with the
switches accordingly set and the results are processed and stored in the corresponding data
registers
• The Interleaved Mode: at each trigger, the 8 conversions for the touch screen and the analog
inputs conversions are performed. Only the analog inputs results are managed by the PDC
and the touch screen conversions can be performed less often than the analog inputs.
The Operating Mode of the TSADCC is programmed in the field TSAMOD in the “TSADCC
Mode Register”.
The conversion sequences for each Operating Mode are described in the following paragraphs.
The conversion sequencer, combined with the Sleep Modes, allows automatic processing with
minimum processor intervention and optimized power consumption. In any case, the sequence
starts with a trigger event.
Note: The reference voltage pins always remain connected in normal mode as in sleep mode.
39.10.1 ADC Mode

In the ADC Mode, the active channels are defined by the “TSADCC Channel Status Register”,
which is defined by writing the “TSADCC Channel Enable Register” and “TSADCC Channel Dis-
able Register”. The results are stored in the “TSADCC Channel Data Register x (x = 0..7)” and in
the “TSADCC Last Converted Data Register”, so that data transfers by using the PDC are
possible.
At each trigger, the following sequence is performed:
3. If SLEEP is set, wake up the ADC cell and wait for the Startup Time.
4. If Channel 0 is enabled, convert Channel 0 and store result in both TSADCC_CDR0
and TSADCC_LCDR.
884
AT91SAM9G45

and TSADCC_LCDR.
and TSADCC_LCDR.
and TSADCC_LCDR.
8. If Channel 4 to Channel 7 are enabled, convert the Channels and store result in the cor-
responding TSADCC_CDRx and TSADCC_LCDR.
9. If SLEEP is set, sleep down the ADC cell.
If the PDC is enabled, all the converted data are transferred contiguously in the memory buffer.
The bit LOWRES defines which resolution is used, either 8-bit or 10-bit, and thus the width of the
PDC memory buffer.
39.10.2 Touch Screen Mode

Writing TSAMOD to “Touch Screen Only Mode” automatically enables the touch screen pins as
analog inputs, and thus disables the digital function of the corresponding pins.
In Touch Screen Mode, the channels 0 to 3 corresponding to the Touch Screen inputs are auto-
matically activated and the bits CH0 to CH3 are automatically set in the “TSADCC Channel
Status Register”.
The remaining channels can be either enabled or disabled by the user and their conversions are
performed at the end of each touch screen sequence.
The resolution is forced to 10 bits, regardless of the LOWRES bit setting.
At each trigger, if the bit PRES in “TSADCC Mode Register” is disabled, the following sequence
is performed to measure only position.
2. Close the switches on the inputs XP and XM during the Sample and Hold Time.
3. Convert Channel XM and store the result in TSADCC_CDR1.
5. Convert Channel XP, subtract TSADCC_CDR1 from the result and store the subtraction
result in both TSADCC_CDR0 and TSADCC_LCDR.
7. Convert Channel YP, subtract TSADCC_CDR1 from the result and store the subtraction
8. Close the switches on the inputs YP and YM during the Sample and Hold Time.
9. Convert Channel YM and store the result in TSADCC_CDR3.
14. If Channel 4 to Channel 7 are enabled, convert the Channels and store result in the cor-
responding TSADCC_CDRx and TSADCC_LCDR.
The resulting buffer is 16 bits wide and its structure stored in memory is:
885
AT91SAM9G45
1. XP - XM
2. YP - XM
3. YP - YM
4. XP - YM
5. AD4 to AD7 if enabled.
The vertical position can be easily calculated by dividing the data at offset 0 (XP - XM) by the data
at offset 1 (YP - XM).
The horizontal position can be easily calculated by dividing the data at offset 2 (YP - YM) by the
data at offset 3 (XP - YM).
if the bit PRES in “TSADCC Mode Register” is enabled, the following sequence is performed to
measure both position and pressure.
2. Close the switches on the inputs XM and Yp during the Sample and Hold Time.
3. Convert Channel Xp and store the result in both TSADCC_Z1DR and TSADCC_LCDR.
4. Close the switches on the inputs XM and YP during the Sample and Hold Time.
5. Convert Channel YM and store the result in both TSADCC_Z2DR and TSADCC_LCDR.
9. Convert Channel Xp, subtract TSADCC_CDR1 from the result and store the subtraction
11. Convert Channel YP and store the result in TSADCC_XPDR, subtract TSADCC_CDR1
from the result and store the subtraction result in both TSADCC_CDR1 and
TSADCC_LCDR.
13. Convert Channel YM and store the result in TSADCC_CDR3 while storing content of
TSADCC_XPDR in TSADCC_LCDR.
15. Convert Channel Yp subtract TSADCC_CDR3 from the result and store the subtraction
17. Convert Channel Xp subtract TSADCC_CDR3 from the result and store the subtraction
18. if Channel 4 to channel 7are enabled, convert the channels and store result in the cor-
responding TSADCC_CDRx and TSADCC_LCDR
The resulting buffer is 16 bits wide and its structure stored in memory is:
1. Z1
2. Z2
3. XP - XM
4. YP - XM
5. Xpos
6. YP - YM
886
AT91SAM9G45
7. XP - YM
8. AD4 to AD7 if enabled.
The vertical position can be easily calculated by dividing the data at offset 2 (XP - XM) by the data
at offset 3(YP - XM).
The horizontal position can be easily calculated by dividing the data at offset 5 (YP - YM) by the
data at offset 7 (XP - YM).
The Pressure measure can be calculated using the following formula
Rp = Rxp*(Xpos/1024)*[(Z2/Z1)-1]
39.10.3 Interleaved Mode

In the Interleaved Mode, the conversion of the touch screen channels are made in parallel to
each channel. In addition to interleaving, the analog channels 4 and 5 can be converted more
often than the touch screen channels depending on the TSFREQ field in the register
TSADCC_MR. In the interleaved mode at least one ADC channel must be enabled.
In the Interleaved Mode, the channels 0 to 3 corresponding to the Touch Screen inputs are auto-
matically activated and the bits CH0 to CH3 are automatically set in the “TSADCC Channel
Status Register”.
This mode allows periodic conversion of the remaining channels at high sampling rate and con-
verted data transferred in memory with the PDC while the touch screen conversions are
performed at low rate. The PDC transfers only analog channel data and touch screen data must
be read in the “TSADCC Channel Data Register x (x = 0..7)”.
The resolution can be configured for the channel 4 to 7 only, through the LOWRES bit. The res-
olution for the conversion made on channels 0 to 3 is forced to 10 bits.
At each trigger, the sequence performed depends on a Trigger Counter, which is compared at
the end to the Touch Screen Frequency, as defined by the field TSFREQ in the register
TSADCC_MR:
Touch Screen Frequency = Trigger Frequency / (2TSFREQ+1)
unless TSFREQ is programmed at 0 or 1. In such cases, the Touch Screen Frequency is one-
sixth of the Trigger Frequency.
As TSFREQ varies between 0 and 15, this results in the ADC channels being converted
between 6 to 65536 less often than the Touch Screen channels.
If the bit PRES in “TSADCC Mode Register” is disabled (measure only position), the sequences
are as follow:
• For Trigger Counter at 0:
3. If Channel 4 to Channel 7 are enabled, convert Channels and store result in the corre-
sponding TSADCC_CDRx and TSADCC_LCDR.
4. Set Trigger Counter to 1.
887
AT91SAM9G45

1. Close the switches on the inputs XP and XM during the Sample and Hold Time..
result in TSADCC_CDR0 (and also in TSADCC_LCDR if PDCEN is enabled).

2. Convert Channel YP,, subtract TSADCC_CDR1 from the result and store the subtrac-
tion result in TSADCC_CDR1 (and also in TSADCC_LCDR if PDCEN is enabled).

2. Convert Channel YM and store the result in TSADCC_CDR3.


1. Close the switches on the inputs YP and YM during the Sample and Hold Time
• For Trigger Counter between 6 and (2TSFREQ+1):

1. Increment Trigger Counter.
2. If Trigger Counter equals (2TSFREQ+1), then set Trigger Counter to 0.
888
AT91SAM9G45
If the bit PRES in “TSADCC Mode Register” is enabled (measure both position and pressure),
the sequences are as follow:
1. Close the switches on the inputs XP and YM during the Sample and Hold Time.
2. Convert Channel XP and store the result in TSADCC_Z1DR (and also in
TSADCC_LCDR if PDCEN is enabled).

1. Close the switches on the inputs XP and YM during the Sample and Hold Time..
2. Convert Channel YM and store the result in TSADCC_Z2DR (and also in


1. Close the switches on the inputs XP and XM during the Sample and Hold Time..

2. Convert Channel YP and store the result in TSADCC_XPDR, subtract TSADCC_CDR1
from the result and store the subtraction result in TSADCC_CDR1 (and also in
889
AT91SAM9G45

2. Convert Channel YM and store the result in TSADCC_CDR3 (and store content of
TSADCC_XPDR in TSADCC_LCDR if PDCEN is enabled).


1. Close the switches on the inputs YP and YM during the Sample and Hold Time
• For Trigger Counter between 8 and (2TSFREQ+1):

1. Increment Trigger Counter.
2. If Trigger Counter equals (2TSFREQ+1), then set Trigger Counter to 0.
The Trigger Counter is cleared when TSAMOD is written to define the Interleaved Mode, then it
simply rolls over.
39.10.4 Manual Mode

The TSADCC features a manual mode allowing to control the state (open/close) of the four
switches.
Writing TSAMOD to “Manual Mode” automatically enables the ADC pins as analog inputs.The
switches positions are controlled through the “TSADCC Manual Switch Command Register”. In
this mode, the “Sample and Hold Time” used is the one defined for the Touchscreen mode
(TSSHTIM).
To perform a measurement, the following sequence must be followed
1. Select the switch (switches) to close.
890
AT91SAM9G45
3. Enable the Channel to convert and start a conversion. If SLEEP is set, wake up the
ADC cell and wait for the Startup Time are performed before the conversion.The result
is stored in TSADCC_CDRx (and TSADCC_LCDR if PDCEN is enabled).
5. Open the switches to reduce power consumption.
891
AT91SAM9G45
39.11 Touch Screen ADC Controller (TSADCC) User Interface

0x00 Control Register TSADCC_CR Write-only –
0x04 Mode Register TSADCC_MR Read-write 0x0000_0000
0x08 Trigger Register TSADCC_TRGR Read-write 0x0000_0000
0x0C Touch Screen Register TSADCC_TSR Read-write 0x0000_0000
0x10 Channel Enable Register TSADCC_CHER Write-only –
0x14 Channel Disable Register TSADCC_CHDR Write-only –
0x18 Channel Status Register TSADCC_CHSR Read-only 0x0000_0000
0x1C Status Register TSADCC_SR Read-only 0x000C_0000
0x20 Last Converted Data Register TSADCC_LCDR Read-only 0x0000_0000
0x24 Interrupt Enable Register TSADCC_IER Write-only –
0x28 Interrupt Disable Register TSADCC_IDR Write-only –
0x2C Interrupt Mask Register TSADCC_IMR Read-only 0x0000_0000
0x30 Channel Data Register 0 TSADCC_CDR0 Read-only 0x0000_0000
0x3C Channel Data Register 3 TSADCC_CDR3 Read-only 0x0000_0000
0x4C Channel Data Register 7 TSADCC_CDR7 Read-only 0x0000_0000
0x50 X Position Data Register TSADCC_XPDR Read-only 0x0000_0000
0x54 Z1 Data Register TSADCC_Z1DR Read-only 0x0000_0000
0x58 Z2 Data Register TSADCC_Z2DR Read-only 0x0000_0000
0X5C Reserved – – –
0X60 Manual Switch Command Register TSADCC_MSCR Read-write –
Reserved – – –
0xE4 Write Protection Mode Register TSADCC_WPMR Read-write –
0xE8 Write Protection Status Register TSADCC_WPSR Write-only –
892
AT91SAM9G45
39.11.1 TSADCC Control Register

Register Name: TSADCC_CR
Address: 0xFFFB0000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – START SWRST

0 = No effect.
1 = Resets the TSADCC simulating a hardware reset.
• START: Start Conversion

0 = No effect.
1 = Begins analog-to-digital conversion.
893
AT91SAM9G45
39.11.2 TSADCC Mode Register

Register Name: TSADCC_MR
Address: 0xFFFB0004
31 30 29 28 27 26 25 24
PENDBC SHTIM
23 22 21 20 19 18 17 16
– STARTUP
15 14 13 12 11 10 9 8
PRESCAL
7 6 5 4 3 2 1 0
PRES PENDET SLEEP LOWRES PDCEN - TSAMOD
• TSAMOD: Touch Screen ADC Mode
TSAMOD Touch Screen ADC Operating Mode

0 ADC Mode
1 Touch Screen Only Mode
2 Interleaved Mode
3 Manual Mode
• PDCEN: PDC transfer in Touchscreen/Interleaved mode or Manual mode

0: Disable the PDC transfer in Touchscreen/Interleaved mode or Manual mode
1: In Touchscreen/Interleaved mode or Manual mode, the data conversion is transferred in the TSADCC_LCDR register
allowing PDC transfer to memory.
• LOWRES: Resolution Selection
LOWRES Selected Resolution

0 10-bit resolution
1 8-bit resolution
This option is only valid in ADC mode.
• SLEEP: Sleep Mode
SLEEP Selected Mode

0 Normal Mode
1 Sleep Mode
• PRESCAL: Prescaler Rate Selection

ADCCLK = MCK / ( (PRESCAL+1) * 2 )
894
AT91SAM9G45
• PENDET: Pen Detect Selection

0: Disable the Touch screen pins as analog inputs
1: enable the Touch screen pins as analog inputs
• PRES: Pressure Measurement Selection

0: Disable the pressure measurement function
1: enable the pressure measurement function
• STARTUP: Start Up Time

Startup Time = (STARTUP+1) * 8/ADCCLK
• SHTIM: Sample & Hold Time for ADC Channels

Programming 0 for SHTIM gives a Sample & Hold Time equal to 1/ADCCLK.
Sample & Hold Time = SHTIM/ADCCLK
• PENDBC: Pen Detect debouncing period

Period = 2PENDBC/ADCCLK
895
AT91SAM9G45
39.11.3 TSADCC Trigger Register

Register Name: TSADCC_TRGR
Address: 0xFFFB0008
31 30 29 28 27 26 25 24
TRGPER
23 22 21 20 19 18 17 16
TRGPER
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – TRGMOD
• TRGMOD: Trigger Mode
TRGMOD Selected Trigger Mode

0 0 0 No trigger, only software trigger can start conversions
0 0 1 External Trigger Rising Edge
0 1 0 External Trigger Falling Edge
0 1 1 External Trigger Any Edge
1 0 0 Pen Detect Trigger (shall be selected only if PENDET is set and TSAMOD = Touch
Screen only mode)
1 0 1 Periodic Trigger (TRGPER shall be initiated appropriately)
1 1 0 Continuous Mode
1 1 1 Reserved
• TRGPER: Trigger Period

Effective only if TRGMOD defines a Periodic Trigger
Defines the periodic trigger period, with the following equations:
Trigger Period = (TRGPER+1) /ADCCLK
896
AT91SAM9G45
39.11.4 TSADCC Touch Screen Register

Register Name: TSADCC_TSR
Address: 0xFFFB000C
31 30 29 28 27 26 25 24
– – – – TSSHTIM
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – TSFREQ
• TSFREQ: Touch Screen Frequency in Interleaved Mode

Effective only if the Touch Screen Interleaved Mode is selected. Defines the Touch Screen Frequency compared to the
Trigger Frequency.
If TSFREQ is 0 or 1, the Touch Screen Frequency is a sixth of the Trigger Frequency.
Otherwise:
Touch Screen Frequency = Trigger Frequency / (2TSFREQ+1)
• TSSHTIM: Sample & Hold Time for Touch Screen Channels

Programming 0 for TSSHTIM gives a Touch Screen Sample & Hold Time equal to 1/ADCCLK.
Touch Screen Sample & Hold Time = TSSHTIM/ADCCLK
897
AT91SAM9G45
39.11.5 TSADCC Channel Enable Register

Register Name: TSADCC_CHER
Address: 0xFFFB0010
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
CH7 CH6 CH5 CH4 CH3 CH2 CH1 CH0
• CHx: Channel x Enable

0 = No effect.
1 = Enables the corresponding channel.
898
AT91SAM9G45
39.11.6 TSADCC Channel Disable Register

Register Name: TSADCC_CHDR
Address: 0xFFFB0014
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
• CHx: Channel x Disable

0 = No effect.
1 = Disables the corresponding channel.
Warning: If the corresponding channel is disabled during a conversion or if it is disabled then reenabled during a conver-
sion, its associated data and its corresponding EOC and OVRE flags in TSADCC_SR are unpredictable.
899
AT91SAM9G45
39.11.7 TSADCC Channel Status Register

Register Name: TSADCC_CHSR
Address: 0xFFFB0018
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
• CHx: Channel x Status

0 = Corresponding channel is disabled.
1 = Corresponding channel is enabled.
900
AT91SAM9G45
39.11.8 TSADCC Status Register

Register Name: TSADCC_SR
Address: 0xFFFB001C
31 30 29 28 27 26 25 24
– OVREZ2 OVREZ1 OVREXP – EOCZ2 EOCZ1 EOCXP
23 22 21 20 19 18 17 16
– – NOCNT PENCNT RXBUFF ENDRX GOVRE DRDY
15 14 13 12 11 10 9 8
OVRE7 OVRE6 OVRE5 OVRE4 OVRE3 OVRE2 OVRE1 OVRE0
7 6 5 4 3 2 1 0
EOC7 EOC6 EOC5 EOC4 EOC3 EOC2 EOC1 EOC0
• EOCx: End of Conversion x

0 = Corresponding analog channel is disabled, or the conversion is not finished.
1 = Corresponding analog channel is enabled and conversion is complete.
• OVREx: Overrun Error x

0 = No overrun error on the corresponding channel since the last read of TSADCC_SR.
1 = There has been an overrun error on the corresponding channel since the last read of TSADCC_SR.
• DRDY: Data Ready

0 = No data has been converted since the last read of TSADCC_LCDR.
1 = At least one data has been converted and is available in TSADCC_LCDR.
• GOVRE: General Overrun Error

0 = No General Overrun Error occurred since the last read of TSADCC_SR.
1 = At least one General Overrun Error has occurred since the last read of TSADCC_SR.
• ENDRX: End of RX Buffer

0 = The Receive Counter Register has not reached 0 since the last write in TSADCC_RCR or TSADCC_RNCR.
1 = The Receive Counter Register has reached 0 since the last write in TSADCC_RCR or TSADCC_RNCR.
• RXBUFF: RX Buffer Full

0 = TSADCC_RCR or TSADCC_RNCR have a value other than 0.
1 = Both TSADCC_RCR and TSADCC_RNCR have a value of 0.
• PENCNT: Pen Contact

0 = No contact has been detected since the last read of TSADCC_SR or PENDET is at 0.
1 = At least one contact has been detected since the last read of TSADCC_SR.
• NOCNT: No Contact
0 = No contact loss has been detected since the last read of TSADCC_SR or PENDET is at 0.
1 = At least one contact loss has been detected since the last read of TSADCC_SR.
901
AT91SAM9G45
• EOCXp: End of Conversion X Position

0 = The pressure measurement is disabled or the Xp conversion is not finished.
1 = The pressure measurement is enabled and the Xp conversion is complete
• EOCZ1: End of Conversion Z1 Measure

0 = The pressure measurement is disabled or the Z1 conversion is not finished.
1 = The pressure measurement is enabled and the Z1 conversion is complete

0 = The pressure measurement is disabled or the Z2 conversion is not finished.
1 = The pressure measurement is enabled and the Z2 conversion is complete
• OVREXP: Overrun Error on X Position

0 = No overrun error on the XP measure channel.
1 = There has been an overrun error on the XP measure channel.
• OVREZ1: Overrun Error on Z1 Measure

0 = No overrun error on the Z1 measure channel.
1 = There has been an overrun error on the Z1 measure channel.
• OVREZ2: Overrun Error on Z2 Measure

0 = No overrun error on the Z2 measure channel.
1 = There has been an overrun error on the Z2 measure channel.
902
AT91SAM9G45
39.11.9 TSADCC Channel Data Register x (x = 0..7)

Register Name: TSADCC_CDR0..TSADCC_CDR7
Address: 0xFFFB0030
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – DATA
7 6 5 4 3 2 1 0
DATA
• DATA: Channel Data

The analog-to-digital conversion data is placed into this register at the end of a conversion of the corresponding channel
and remains until a new conversion on the same channel is completed.
903
AT91SAM9G45
39.11.10 TSADCC Last Converted Data Register

Register Name: TSADCC_LCDR
Address: 0xFFFB0020
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – LDATA
7 6 5 4 3 2 1 0
LDATA
• LDATA: Last Data Converted

The analog-to-digital conversion data is placed into this register at the end of a conversion on any analog channel and
remains until a new conversion on any analog channel is completed.
904
AT91SAM9G45
39.11.11 TSADCC Interrupt Enable Register

Register Name: TSADCC_IER
Address: 0xFFFB0024
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• EOCx: End of Conversion Interrupt Enable x
• OVREx: Overrun Error Interrupt Enable x
• DRDY: Data Ready Interrupt Enable
• GOVRE: General Overrun Error Interrupt Enable
• ENDRX: End of Receive Buffer Interrupt Enable
• OVREXP: Overrun Error Interrupt Enable X Position
• OVREZ1: Overrun Error Interrupt Enable Z1 Measure
• OVREZ2: Overrun Error Interrupt Enable Z2 Measure

0 = No effect.
905
AT91SAM9G45
39.11.12 TSADCC Interrupt Disable Register

Register Name: TSADCC_IDR
Address: 0xFFFB0028
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• EOCx: End of Conversion Interrupt Disable x
• OVREx: Overrun Error Interrupt Disable x
• DRDY: Data Ready Interrupt Disable
• GOVRE: General Overrun Error Interrupt Disable
• ENDRX: End of Receive Buffer Interrupt Disable
• OVREXP: Overrun Error Interrupt Disable X Position
• OVREZ1: Overrun Error Interrupt Disable Z1 Measure
• OVREZ2: Overrun Error Interrupt Disable Z2 Measure
0 = No effect.
906
AT91SAM9G45
39.11.13 TSADCC Interrupt Mask Register

Register Name: TSADCC_IMR
Address: 0xFFFB002C
31 30 29 28 27 26 25 24
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• EOCx: End of Conversion Interrupt Mask x
• OVREx: Overrun Error Interrupt Mask x
• DRDY: Data Ready Interrupt Mask
• GOVRE: General Overrun Error Interrupt Mask
• ENDRX: End of Receive Buffer Interrupt Mask
• OVREXP: Overrun Error Interrupt Mask X Position
• OVREZ1: Overrun Error Interrupt Mask Z1 Measure
• OVREZ2: Overrun Error Interrupt Mask Z2 Measure

907
39.11.14 TSADCC X Position Data Register
Register Name: TSADCC_XPDR.
Address: 0xFFFB0050
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – DATA
7 6 5 4 3 2 1 0
DATA
• DATA: X Position Data
39.11.15 TSADCC Z1 Data Register

Register Name: TSADCC_Z1DR.
Address: 0xFFFB0054
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – DATA
7 6 5 4 3 2 1 0
DATA
• DATA: Z1 Measurement Data
908 AT91SAM9G45
AT91SAM9G45
39.11.16 TSADCC Z2 Data Register

Register Name: TSADCC_Z2DR.
Address: 0xFFFB0058
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – Z2
7 6 5 4 3 2 1 0
Z2
• DATA: Z2 Measurement Data
39.11.17 TSADCC Manual Switch Command Register

Register Name: TSADCC_MSCR.
Address: 0xFFFB0060
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – –
7 6 5 4 3 2 1 0
– – – – YM YP XM XP
• YM,YP,XM,XP: Switch Command

If bit is set the related switch is closed
If bit is cleared the related switch is open
909
39.11.18 TSADCC Write Protection Mode Register
Register Name: TSADCC_WPMR
Address: 0xFFFB00E4
Access Type: Read-Write
31 30 29 28 27 26 25 24
KEY
23 22 21 20 19 18 17 16
KEY
15 14 13 12 11 10 9 8
KEY
7 6 5 4 3 2 1 0
– – – – – – – WPEN
• WPEN: Write Protection of TSADCC_MR, TSADCC_TRGR and TSADCC_TSR

0 and KEY= 0x545341 Write protection is disabled.
1 and KEY = 0x545341, Write Protection is enabled.
39.11.19 TSADCC Write Protection Status Register

Register Name: TSADCC_WPSR
Address: 0xFFFB00E8
31 30 29 28 27 26 25 24
OFFSET_ERR
23 22 21 20 19 18 17 16
OFFSET_ERR
15 14 13 12 11 10 9 8
OFFSET_ERR
7 6 5 4 3 2 1 0
– – – – – – – WPS
• WPS: Write Protection Status

0: Write protection is disabled.
1:Write Protection is enabled.
• OFFSET_ERR: Offset error

Offset where the last unauthorized access occurred.
910 AT91SAM9G45
AT91SAM9G45
40. DMA Controller (DMAC)
40.1 Description
The DMA Controller (DMAC) is an AHB-central DMA controller core that transfers data from a
source peripheral to a destination peripheral over one or more AMBA buses. One channel is
required for each source/destination pair. In the most basic configuration, the DMAC has one
master interface and one channel. The master interface reads the data from a source and writes
it to a destination. Two AMBA transfers are required for each DMAC data transfer. This is also
known as a dual-access transfer.
The DMAC is programmed via the APB interface.

• Two Masters
• Embeds 8 channels
• 64 bytes/FIFO for Channel Buffering
• Linked List support with Status Write Back operation at End of Transfer
• Word, HalfWord, Byte transfer support.
• memory to memory transfer
• Peripheral to memory
• Memory to peripheral
The DMA controller can handle the transfer between peripherals and memory and so receives
the triggers from the peripherals below. The hardware interface numbers are also given below in
Table 40-1
Table 40-1. DMA Channel Definition

DMA Channel HW
Instance Name T/R interface Number
MCI0 TX/RX 0
SPI0 TX 1
SPI0 RX 2
SPI1 TX 3
SPI1 RX 4
SSC0 TX 5
SSC0 RX 6
SSC1 TX 7
SSC1 RX 8
AC97 TX 9
AC97 RX 10
MCI1 TX/RX 13
• Acting as two Matrix Masters

• Embeds 8 unidirectional channels with programmable priority
911
• Address Generation
– Source/Destination address programming
– Address increment, decrement or no change
– DMA chaining support for multiple non-contiguous data blocks through use of linked
lists
– Scatter support for placing fields into a system memory area from a contiguous
transfer. Writing a stream of data into non-contiguous fields in system memory
– Gather support for extracting fields from a system memory area into a contiguous
transfer
– User enabled auto-reloading of source, destination and control registers from initially
programmed values at the end of a block transfer
– Auto-loading of source, destination and control registers from system memory at end
of block transfer in block chaining mode
– Unaligned system address to data transfer width supported in hardware
• Channel Buffering
– 16-word FIFO
– Automatic packing/unpacking of data to fit FIFO width
• Channel Control
– Programmable multiple transaction size for each channel
– Support for cleanly disabling a channel without data loss
– Suspend DMA operation
– Programmable DMA lock transfer support
• Transfer Initiation
– Support for Software handshaking interface. Memory mapped registers can be used
to control the flow of a DMA transfer in place of a hardware handshaking interface
• Interrupt
– Programmable Interrupt generation on DMA Transfer completion Block Transfer
completion, Single/Multiple transaction completion or Error condition
912 AT91SAM9G45
AT91SAM9G45
40.3 Block Diagram
Figure 40-1. DMA Controller (DMAC) Block Diagram

AMBA AHB Layer 1
DMA AHB Lite Master Interface 1
DMA Global Control DMA Global

and Data Mux Request Arbiter
DMA Write DMA Destination

Datapath Bundles Requests Pool
DMA Channel n
Atmel APB rev2 Interface DMA
DMA Destination
Atmel
DMA Channel 2 APB
Status Interface
DMA Channel 1 Registers
DMA Channel 0
Configuration
DMA Destination Registers
DMA Channel 0
Write data path Control State Machine
to destination Destination Pointer
Management DMA Interrupt DMA Interrupt
Controller
DMA FIFO Controller
DMA FIFO Trigger Manager DMA

Hardware
Up to 64 bytes
External Handshaking
Triggers DMA Interface
REQ/ACK
Soft Interface
Triggers
DMA Source
DMA Channel 0 Control State Machine
Read data path Source Pointer
from source Management
DMA Read DMA Source

Datapath Bundles Requests Pool
DMA Global Control DMA Global

and Data Mux Request Arbiter
DMA AHB Lite Master Interface 0
AMBA AHB Layer 0
913
AT91SAM9G45
40.4.1 Basic Definitions

Source peripheral: Device on an AMBA layer from where the DMAC reads data, which is then
stored in the channel FIFO. The source peripheral teams up with a destination peripheral to form
a channel.
Destination peripheral: Device to which the DMAC writes the stored data from the FIFO (previ-
ously read from the source peripheral).
Memory: Source or destination that is always “ready” for a DMAC transfer and does not require
a handshaking interface to interact with the DMAC.
Channel: Read/write datapath between a source peripheral on one configured AMBA layer and
a destination peripheral on the same or different AMBA layer that occurs through the channel
FIFO. If the source peripheral is not memory, then a source handshaking interface is assigned to
the channel. If the destination peripheral is not memory, then a destination handshaking inter-
face is assigned to the channel. Source and destination handshaking interfaces can be assigned
dynamically by programming the channel registers.
Master interface: DMAC is a master on the AHB bus reading data from the source and writing it
to the destination over the AHB bus.
Slave interface: The APB interface over which the DMAC is programmed. The slave interface
in practice could be on the same layer as any of the master interfaces or on a separate layer.
Handshaking interface: A set of signal registers that conform to a protocol and handshake
between the DMAC and source or destination peripheral to control the transfer of a single or
chunk transfer between them. This interface is used to request, acknowledge, and control a
DMAC transaction. A channel can receive a request through one of two types of handshaking
interface: hardware or software.
Hardware handshaking interface: Uses hardware signals to control the transfer of a single or
chunk transfer between the DMAC and the source or destination peripheral.
Software handshaking interface: Uses software registers to contr5ol the transfer of a single or
chunk transfer between the DMAC and the source or destination peripheral. No special DMAC
handshaking signals are needed on the I/O of the peripheral. This mode is useful for interfacing
an existing peripheral to the DMAC without modifying it.
Flow controller: The device (either the DMAC or source/destination peripheral) that determines
the length of and terminates a DMAC buffer transfer. If the length of a buffer is known before
enabling the channel, then the DMAC should be programmed as the flow controller. If the length
of a buffer is not known prior to enabling the channel, the source or destination peripheral needs
to terminate a buffer transfer. In this mode, the peripheral is the flow controller.
Transfer hierarchy: Figure 40-2 on page 915 illustrates the hierarchy between DMAC transfers,
buffer transfers, chunk or single, and AMBA transfers (single or burst) for non-memory peripher-
als. Figure 40-3 on page 915 shows the transfer hierarchy for memory.
914
AT91SAM9G45
Figure 40-2. DMAC Transfer Hierarchy for Non-Memory Peripheral
HDMA Transfer DMA Transfer

Level
Buffer Transfer
Buffer Buffer Buffer
Level
Chunk Chunk Chunk Single DMA Transaction

Transfer Transfer Transfer Transfer Level
AMBA AMBA AMBA AMBA AMBA

Burst Single Single AMBA Transfer
Burst Burst
Transfer Transfer Transfer Transfer Transfer Level
Figure 40-3. DMAC Transfer Hierarchy for Memory
HDMA Transfer DMA Transfer

Level
Buffer Transfer
Buffer Buffer Buffer
Level
AMBA AMBA AMBA AMBA

Burst Single AMBA Transfer
Burst Burst
Transfer Transfer Transfer Transfer Level
Buffer: A buffer of DMAC data. The amount of data (length) is determined by the flow controller.
For transfers between the DMAC and memory, a buffer is broken directly into a sequence of
AMBA bursts and AMBA single transfers.
For transfers between the DMAC and a non-memory peripheral, a buffer is broken into a
sequence of DMAC transactions (single and chunks). These are in turn broken into a sequence
of AMBA transfers.
Transaction: A basic unit of a DMAC transfer as determined by either the hardware or software
handshaking interface. A transaction is only relevant for transfers between the DMAC and a
source or destination peripheral if the source or destination peripheral is a non-memory device.
There are two types of transactions: single transfer and chunk transfer.
– Single transfer: The length of a single transaction is always 1 and is converted to a
single AMBA access.
– Chunk transfer: The length of a chunk is programmed into the DMAC. The chunk is
then converted into a sequence of AHB access.DMAC executes each AMBA burst
transfer by performing incremental bursts that are no longer than 16 beats.
DMAC transfer: Software controls the number of buffers in a DMAC transfer. Once the DMAC
transfer has completed, then hardware within the DMAC disables the channel and can generate
915
AT91SAM9G45
an interrupt to signal the completion of the DMAC transfer. You can then re-program the channel
for a new DMAC transfer.
Single-buffer DMAC transfer: Consists of a single buffer.
Multi-buffer DMAC transfer: A DMAC transfer may consist of multiple DMAC buffers. Multi-buf-
fer DMAC transfers are supported through buffer chaining (linked list pointers), auto-reloading of
channel registers, and contiguous buffers. The source and destination can independently select
which method to use.
– Linked lists (buffer chaining) – A descriptor pointer (DSCR) points to the location
in system memory where the next linked list item (LLI) exists. The LLI is a set of
registers that describe the next buffer (buffer descriptor) and a descriptor pointer
register. The DMAC fetches the LLI at the beginning of every buffer when buffer
chaining is enabled.
– Replay – The DMAC automatically reloads the channel registers at the end of each
buffers to the value when the channel was first enabled.
– Contiguous buffers – Where the address of the next buffer is selected to be a
continuation from the end of the previous buffer.
Picture-in-Picture Mode: DMAC contains a picture-in-picture mode support. When this mode is
enabled, addresses are automatically incremented by a programmable value when the DMAC
channel transfer count reaches a user defined boundary.
Figure 40-4 on page 916 illustrates a memory mapped image 4:2:2 encoded located at
image_base_address in memory. A user defined start address is defined at
Picture_start_address. The incremented value is set to memory_hole_size = image_width -
picture_width, and the boundary is set to picture_width.
Figure 40-4. Picture-In-Picture Mode Support
Channel locking: Software can program a channel to keep the AHB master interface by locking
the arbitration for the master bus interface for the duration of a DMAC transfer, buffer, or chunk.
916
AT91SAM9G45
Bus locking: Software can program a channel to maintain control of the AMBA bus by asserting
hmastlock for the duration of a DMAC transfer, buffer, or transaction (single or chunk). Channel
locking is asserted for the duration of bus locking at a minimum.
40.4.2 Memory Peripherals

Figure 40-3 on page 915 shows the DMAC transfer hierarchy of the DMAC for a memory periph-
eral. There is no handshaking interface with the DMAC, and therefore the memory peripheral
can never be a flow controller. Once the channel is enabled, the transfer proceeds immediately
without waiting for a transaction request. The alternative to not having a transaction-level hand-
shaking interface is to allow the DMAC to attempt AMBA transfers to the peripheral once the
channel is enabled. If the peripheral slave cannot accept these AMBA transfers, it inserts wait
states onto the bus until it is ready; it is not recommended that more than 16 wait states be
inserted onto the bus. By using the handshaking interface, the peripheral can signal to the
DMAC that it is ready to transmit/receive data, and then the DMAC can access the peripheral
without the peripheral inserting wait states onto the bus.
40.4.3 Handshaking Interface

Handshaking interfaces are used at the transaction level to control the flow of single or chunk
transfers. The operation of the handshaking interface is different and depends on whether the
peripheral or the DMAC is the flow controller.
The peripheral uses the handshaking interface to indicate to the DMAC that it is ready to trans-
fer/accept data over the AMBA bus. A non-memory peripheral can request a DMAC transfer
through the DMAC using one of two handshaking interfaces:
• Hardware handshaking
• Software handshaking
Software selects between the hardware or software handshaking interface on a per-channel
basis. Software handshaking is accomplished through memory-mapped registers, while hard-
ware handshaking is accomplished using a dedicated handshaking interface.
40.4.3.1 Software Handshaking

When the slave peripheral requires the DMAC to perform a DMAC transaction, it communicates
this request by sending an interrupt to the CPU or interrupt controller.
The interrupt service routine then uses the software registers to initiate and control a DMAC
transaction. These software registers are used to implement the software handshaking
interface.
The SRC_H2SEL/DST_H2SEL bit in the DMAC_CFGx channel configuration register must be
set to zero to enable software handshaking.
When the peripheral is not the flow controller, then the last transaction register DMAC_LAST is
not used, and the values in these registers are ignored.
40.4.3.2 Chunk Transactions

Writing a 1 to the DMAC_CREQ[2x] register starts a source chunk transaction request, where x
is the channel number. Writing a 1 to the DMAC_CREQ[2x+1] register starts a destination chunk
transfer request, where x is the channel number.
Upon completion of the chunk transaction, the hardware clears the DMAC_CREQ[2x] or
DMAC_CREQ[2x+1].
917
AT91SAM9G45
40.4.3.3 Single Transactions

Writing a 1 to the DMAC_SREQ[2x] register starts a source single transaction request, where x
is the channel number. Writing a 1 to the DMAC_SREQ[2x+1] register starts a destination single
transfer request, where x is the channel number.
Upon completion of the chunk transaction, the hardware clears the DMAC_SREQ[x] or
DMAC_SREQ[2x+1].
Software can poll the relevant channel bit in the DMAC_CREQ[2x]/DMAC_CREQ[2x+1] and
DMAC_SREQ[x]/DMAC_SREQ[2x+1] registers. When both are 0, then either the requested
chunk or single transaction has completed.
40.4.4 DMAC Transfer Types

A DMAC transfer may consist of single or multi-buffers transfers. On successive buffers of a
multi-buffer transfer, the DMAC_SADDRx/DMAC_DADDRx registers in the DMAC are repro-
grammed using either of the following methods:
• Buffer chaining using linked lists
• Replay mode
• Contiguous address between buffers
On successive buffers of a multi-buffer transfer, the DMAC_CTRLAx and DMAC_CTRLBx regis-
ters in the DMAC are re-programmed using either of the following methods:
• Replay mode
When buffer chaining, using linked lists is the multi-buffer method of choice, and on successive
buffers, the DMAC_DSCRx register in the DMAC is re-programmed using the following method:
A buffer descriptor (LLI) consists of following registers, DMAC_SADDRx, DMAC_DADDRx,
DMAC_DSCRx, DMAC_CTRLAx, DMAC_CTRLBx.These registers, along with the
DMAC_CFGx register, are used by the DMAC to set up and describe the buffer transfer.
40.4.4.1 Multi-buffer Transfers
40.4.4.2 Buffer Chaining Using Linked Lists

In this case, the DMAC re-programs the channel registers prior to the start of each buffer by
fetching the buffer descriptor for that buffer from system memory. This is known as an LLI
update.
DMAC buffer chaining is supported by using a Descriptor Pointer register (DMAC_DSCRx) that
stores the address in memory of the next buffer descriptor. Each buffer descriptor contains the
corresponding buffer descriptor (DMAC_SADDRx, DMAC_DADDRx, DMAC_DSCRx,
DMAC_CTRLAx DMAC_CTRLBx).
To set up buffer chaining, a sequence of linked lists must be programmed in memory.
The DMAC_SADDRx, DMAC_DADDRx, DMAC_DSCRx, DMAC_CTRLAx and DMAC_CTRLBx
registers are fetched from system memory on an LLI update. The updated content of the
DMAC_CTRLAx register is written back to memory on buffer completion. Figure 40-5 on page
919 shows how to use chained linked lists in memory to define multi-buffer transfers using buffer
chaining.
918
AT91SAM9G45
The Linked List multi-buffer transfer is initiated by programming DMAC_DSCRx with DSCRx(0)
(LLI(0) base address) and DMAC_CTRLBx register with both SRC_DSCR and DST_DSCR set
to 0. Other fields and registers are ignored and overwritten when the descriptor is retrieved from
memory.
The last transfer descriptor must be written to memory with its next descriptor address set to 0.
Figure 40-5. Multi Buffer Transfer Using Linked List
System Memory
LLI(0) LLI(1)
DSCRx(1)= DSCRx(0) + 0x10 DSCRx(2)= DSCRx(1) + 0x10
CTRLBx= DSCRx(0) + 0xC CTRLBx= DSCRx(1) + 0xC
CTRLAx= DSCRx(0) + 0x8 CTRLBx= DSCRx(1) + 0x8
DADDRx= DSCRx(0) + 0x4 DADDRx= DSCRx(1) + 0x4
SADDRx= DSCRx(0) + 0x0 SADDRx= DSCRx(1) + 0x0

DSCRx(0) DSCRx(1) DSCRx(2)
(points to 0 if
LLI(1) is the last
transfer descriptor
919
AT91SAM9G45
40.4.4.3 Programming DMAC for Multiple Buffer Transfers
Table 40-2. Multiple Buffers Transfer Management Table

Other
Transfer Type AUTO SRC_REP DST_REP SRC_DSCR DST_DSCR BTSIZE SADDR DADDR Fields
1) Single Buffer or Last

buffer of a multiple buffer 0 – – 1 1 USR USR USR USR
transfer
2) Multi Buffer transfer

0 – 0 0 1 LLI LLI CONT LLI
with contiguous DADDR

0 0 – 1 0 LLI CONT LLI LLI
with contiguous SADDR

0 – – 0 0 LLI LLI LLI LLI
with LLI support

0 – 1 0 1 LLI LLI REP LLI
with DADDR reloaded

0 1 – 1 0 LLI REP LLI LLI
with SADDR reloaded
with BTSIZE reloaded and 1 – 0 0 1 REP LLI CONT LLI
contiguous DADDR

with BTSIZE reloaded and 1 0 – 1 0 REP CONT LLI LLI
contiguous SADDR
9) Automatic mode
channel is stalling 1 0 0 1 1 REP CONT CONT REP
BTsize is reloaded
10) Automatic mode

BTSIZE, SADDR and 1 1 1 1 1 REP REP REP REP
DADDR reloaded
11) Automatic mode

BTSIZE, SADDR reloaded 1 1 0 1 1 REP REP CONT REP
and DADDR contiguous
Notes: 1. USR means that the register field is manually programmed by the user.
2. CONT means that address are contiguous.
3. REP means that the register field is updated with its previous value. If the transfer is the first one, then the user must manu-
ally program the value.
4. Channel stalled is true if the relevant BTC interrupt is not masked.
5. LLI means that the register field is updated with the content of the linked list item.
40.4.4.4 Replay Mode of Channel Registers

During automatic replay mode, the channel registers are reloaded with their initial values at the
completion of each buffer and the new values used for the new buffer. Depending on the row
number in Table 40-2 on page 920, some or all of the DMAC_SADDRx, DMAC_DADDRx,
DMAC_CTRLAx and DMAC_CTRLBx channel registers are reloaded from their initial value at
the start of a buffer transfer.
40.4.4.5 Contiguous Address Between Buffers

In this case, the address between successive buffers is selected to be a continuation from the
end of the previous buffer. Enabling the source or destination address to be contiguous between
920
AT91SAM9G45
buffers is a function of DMAC_CTRLAx.SRC_DSCR, DMAC_CFGx.SRC_REP,

DMAC_CTRLAx.DST_DSCR and DMAC_CFGx.DST_REP registers.
40.4.4.6 Suspension of Transfers Between buffers

At the end of every buffer transfer, an end of buffer interrupt is asserted if:
• the channel buffer interrupt is unmasked, DMAC_EBCIMR.BTC[n] = ‘1’, where n is the
channel number.
Note: The buffer complete interrupt is generated at the completion of the buffer transfer to the
destination.
At the end of a chain of multiple buffers, an end of linked list interrupt is asserted if:
• the channel end of chained buffer interrupt is unmasked, DMAC_EBCIMR.CBTC[n] = ‘1’,
when n is the channel number.
40.4.4.7 Ending Multi-buffer Transfers

All multi-buffer transfers must end as shown in Row 1 of Table 40-2 on page 920. At the end of
every buffer transfer, the DMAC samples the row number, and if the DMAC is in Row 1 state,
then the previous buffer transferred was the last buffer and the DMAC transfer is terminated.
For rows 9, 10 and 11 of Table 40-2 on page 920, (DMAC_DSCRx = 0 and
DMAC_CTRLBx.AUTO is set), multi-buffer DMAC transfers continue until the automatic mode is
disabled by writing a ‘1’ in DMAC_CTRLBx.AUTO bit. This bit should be programmed to zero in
the end of buffer interrupt service routine that services the next-to-last buffer transfer. This puts
the DMAC into Row 1 state.
For rows 2, 3, 4, 5, and 6 (DMAC_CRTLBx.AUTO cleared) the user must setup the last buffer
descriptor in memory such that both LLI.DMAC_CTRLBx.SRC_DSCR and
LLI.DMAC_CTRLBx.DST_DSCR are one and LLI.DMAC_DSCRx is set to 0.
921
AT91SAM9G45
40.4.5 Programming a Channel

Four registers, the DMAC_DSCRx, the DMAC_CTRLAx, the DMAC_CTRLBx and
DMAC_CFGx, need to be programmed to set up whether single or multi-buffer transfers take
place, and which type of multi-buffer transfer is used. The different transfer types are shown in
The “BTSIZE, SADDR and DADDR” columns indicate where the values of DMAC_SARx,
DMAC_DARx, DMAC_CTLx, and DMAC_LLPx are obtained for the next buffer transfer when
multi-buffer DMAC transfers are enabled.
40.4.5.1 Programming Examples
40.4.5.2 Single-buffer Transfer (Row 1)

1. Read the Channel Handler Status Register DMAC_CHSR.ENABLE Field to choose a
free (disabled) channel.
2. Clear any pending interrupts on the channel from the previous DMAC transfer by read-
ing the interrupt status register, DMAC_EBCISR.
3. Program the following channel registers:
a. Write the starting source address in the DMAC_SADDRx register for channel x.
b. Write the starting destination address in the DMAC_DADDRx register for channel
x.
c. Program DMAC_CTRLAx, DMAC_CTRLBx and DMAC_CFGx according to Row 1
as shown in Table 40-2 on page 920. Program the DMAC_CTRLBx register with
both DST_DSCR and SRC_DSCR fields set to one and AUTO field set to 0.
d. Write the control information for the DMAC transfer in the DMAC_CTRLAx and
DMAC_CTRLBx registers for channel x. For example, in the register, you can pro-
gram the following:
– i. Set up the transfer type (memory or non-memory peripheral for source and
destination) and flow control device by programming the FC of the DMAC_CTRLBx
register.
– ii. Set up the transfer characteristics, such as:
– Transfer width for the source in the SRC_WIDTH field.
– Transfer width for the destination in the DST_WIDTH field.
– Source AHB Master interface layer in the SIF field where source resides.
– Destination AHB Master Interface layer in the DIF field where destination resides.
– Incrementing/decrementing or fixed address for source in SRC_INC field.
– Incrementing/decrementing or fixed address for destination in DST_INC field.
e. Write the channel configuration information into the DMAC_CFGx register for chan-
nel x.
– i. Designate the handshaking interface type (hardware or software) for the source
and destination peripherals. This is not required for memory. This step requires
programming the SRC_H2SEL/DST_H2SEL bits, respectively. Writing a ‘1’ activates
the hardware handshaking interface to handle source/destination requests. Writing a
‘0’ activates the software handshaking interface to handle source/destination
requests.
922
AT91SAM9G45
– ii. If the hardware handshaking interface is activated for the source or destination
peripheral, assign a handshaking interface to the source and destination peripheral.
This requires programming the SRC_PER and DST_PER bits, respectively.
f. If source picture-in-picture mode is enabled (DMAC_CTRLBx.SRC_PIP is
enabled), program the DMAC_SPIPx register for channel x.
g. If destination picture-in-picture mode is enabled (DMAC_CTRLBx.DST_PIP is
enabled), program the DMAC_DPIPx register for channel x.
4. After the DMAC selected channel has been programmed, enable the channel by writing
a ‘1’ to the DMAC_CHER.ENABLE[n] bit, where n is the channel number. Make sure
that bit 0 of DMAC_EN.ENABLE register is enabled.
5. Source and destination request single and chunk DMAC transactions to transfer the
buffer of data (assuming non-memory peripherals). The DMAC acknowledges at the
completion of every transaction (chunk and single) in the buffer and carry out the buffer
transfer.
6. Once the transfer completes, hardware sets the interrupts and disables the channel. At
this time you can either respond to the buffer Complete or Transfer Complete interrupts,
or poll for the Channel Handler Status Register (DMAC_CHSR.ENABLE[n]) bit until it is
cleared by hardware, to detect when the transfer is complete.
40.4.5.3 Multi-buffer Transfer with Linked List for Source and Linked List for Destination (Row 4)
1. Read the Channel Enable register to choose a free (disabled) channel.
2. Set up the chain of Linked List Items (otherwise known as buffer descriptors) in mem-
ory. Write the control information in the LLI.DMAC_CTRLAx and LLI.DMAC_CTRLBx
registers location of the buffer descriptor for each LLI in memory (see Figure 40-6 on
page 925) for channel x. For example, in the register, you can program the following:
a. Set up the transfer type (memory or non-memory peripheral for source and desti-
nation) and flow control device by programming the FC of the DMAC_CTRLBx
register.
b. Set up the transfer characteristics, such as:
– i. Transfer width for the source in the SRC_WIDTH field.
– ii. Transfer width for the destination in the DST_WIDTH field.
– iii. Source AHB master interface layer in the SIF field where source resides.
– iv. Destination AHB master interface layer in the DIF field where destination resides.
– v. Incrementing/decrementing or fixed address for source in SRC_INCR field.
– vi. Incrementing/decrementing or fixed address for destination DST_INCR field.
3. Write the channel configuration information into the DMAC_CFGx register for channel
x.
a. Designate the handshaking interface type (hardware or software) for the source
and destination peripherals. This is not required for memory. This step requires pro-
gramming the SRC_H2SEL/DST_H2SEL bits, respectively. Writing a ‘1’ activates
the hardware handshaking interface to handle source/destination requests for the
specific channel. Writing a ‘0’ activates the software handshaking interface to han-
dle source/destination requests.
b. If the hardware handshaking interface is activated for the source or destination
peripheral, assign the handshaking interface to the source and destination periph-
eral. This requires programming the SRC_PER and DST_PER bits, respectively.
4. Make sure that the LLI.DMAC_CTRLBx register locations of all LLI entries in memory
(except the last) are set as shown in Row 4 of Table 40-2 on page 920. The
923
AT91SAM9G45
LLI.DMAC_CTRLBx register of the last Linked List Item must be set as described in
Row 1 of Table 40-2. Figure 40-5 on page 919 shows a Linked List example with two
list items.
5. Make sure that the LLI.DMAC_DSCRx register locations of all LLI entries in memory
(except the last) are non-zero and point to the base address of the next Linked List
Item.
6. Make sure that the LLI.DMAC_SADDRx/LLI.DMAC_DADDRx register locations of all
LLI entries in memory point to the start source/destination buffer address preceding
that LLI fetch.
7. Make sure that the LLI.DMAC_CTRLAx.DONE field of the LLI.DMAC_CTRLAx register
locations of all LLI entries in memory are cleared.
8. If source picture-picture mode is enabled (DMAC_CTRLBx.SRC_PIP is enabled), pro-
gram the DMAC_SPIPx register for channel x.
9. If destination picture-in-picture is enabled (DMAC_CTRLBx.DST_PIP is enabled), pro-
gram the DMAC_DPIPx register for channel x.
ing the status register: DMAC_EBCISR.
11. Program the DMAC_CTRLBx, DMAC_CFGx registers according to Row 4 as shown in
12. Program the DMAC_DSCRx register with DMAC_DSCRx(0), the pointer to the first
Linked List item.
13. Finally, enable the channel by writing a ‘1’ to the DMAC_CHER.ENABLE[n] bit, where n
is the channel number. The transfer is performed.
14. The DMAC fetches the first LLI from the location pointed to by DMAC_DSCRx(0).
Note: The LLI.DMAC_SADDRx, LLI. DMAC_DADDRx, LLI.DMAC_DSCRx, LLI.DMAC_CTRLAx and
LLI.DMAC_CTRLBx registers are fetched. The DMAC automatically reprograms the
DMAC_SADDRx, DMAC_DADDRx, DMAC_DSCRx, DMAC_CTRLBx and DMAC_CTRLAx chan-
nel registers from the DMAC_DSCRx(0).
buffer of data (assuming non-memory peripheral). The DMAC acknowledges at the
transfer.
16. Once the buffer of data is transferred, the DMAC_CTRLAx register is written out to sys-
tem memory at the same location and on the same layer (DMAC_DSCRx.DSCR_IF)
where it was originally fetched, that is, the location of the DMAC_CTRLAx register of
the linked list item fetched prior to the start of the buffer transfer. Only DMAC_CTRLAx
register is written out because only the DMAC_CTRLAx.BTSIZE and
DMAC_CTRLAX.DONE bits have been updated by DMAC hardware. Additionally, the
DMAC_CTRLAx.DONE bit is asserted when the buffer transfer has completed.
Note: Do not poll the DMAC_CTRLAx.DONE bit in the DMAC memory map. Instead, poll the
LLI.DMAC_CTRLAx.DONE bit in the LLI for that buffer. If the poll LLI.DMAC_CTRLAx.DONE bit is
asserted, then this buffer transfer has completed. This LLI.DMAC_CTRLAx.DONE bit was cleared
at the start of the transfer.
17. The DMAC does not wait for the buffer interrupt to be cleared, but continues fetching
the next LLI from the memory location pointed to by current DMAC_DSCRx register
and automatically reprograms the DMAC_SADDRx, DMAC_DADDRx, DMAC_DSCRx,
DMAC_CTRLAx and DMAC_CTRLBx channel registers. The DMAC transfer continues
until the DMAC determines that the DMAC_CTRLBx and DMAC_DSCRx registers at
the end of a buffer transfer match described in Row 1 of Table 40-2 on page 920. The
DMAC then knows that the previous buffer transferred was the last buffer in the DMAC
transfer. The DMAC transfer might look like that shown in Figure 40-6 on page 925.
924
AT91SAM9G45
Figure 40-6. Multi-buffer with Linked List Address for Source and Destination
Address of Address of
Source Layer Destination Layer
Buffer 2 Buffer 2
SADDR(2) DADDR(2)
Buffer 1 Buffer 1
SADDR(1) DADDR(1)
Buffer 0 Buffer 0
SADDR(0) DADDR(0)
Source Buffers Destination Buffers
If the user needs to execute a DMAC transfer where the source and destination address are
contiguous but the amount of data to be transferred is greater than the maximum buffer size
DMAC_CTRLAx.BTSIZE, then this can be achieved using the type of multi-buffer transfer as
shown in Figure 40-7 on page 926.
925
AT91SAM9G45
Figure 40-7. Multi-buffer with Linked Address for Source and Destination Buffers are Contiguous
Buffer 2
DADDR(3)
Buffer 2
Buffer 2
SADDR(3) DADDR(2)
Buffer 2 Buffer 1
SADDR(2) DADDR(1)
Buffer 1
Buffer 0
SADDR(1)
DADDR(0)
Buffer 0
SADDR(0)
The DMAC transfer flow is shown in Figure 40-8 on page 927.
926
AT91SAM9G45
Figure 40-8. DMAC Transfer Flow for Source and Destination Linked List Address
Channel enabled by
software
LLI Fetch
Hardware reprograms
SADDRx, DADDRx, CTRLA/Bx, DSCRx
DMAC buffer transfer
Writeback of HDMA_CTRLAx
register in system memory
Buffer Complete interrupt

generated here
Is HDMA in
Row1 of no
HDMA State Machine Table?
HDMA Transfer Complete yes

interrupt generated here
Channel Disabled by
hardware
40.4.5.4 Multi-buffer Transfer with Source Address Auto-reloaded and Destination Address Auto-reloaded (Row 10)
1. Read the Channel Enable register to choose an available (disabled) channel.
ing the interrupt status register. Program the following channel registers:
927
AT91SAM9G45
x.
c. Program DMAC_CTRLAx, DMAC_CTRLBx and DMAC_CFGx according to Row
10 as shown in Table 40-2 on page 920. Program the DMAC_DSCRx register with
‘0’.
d. Write the control information for the DMAC transfer in the DMAC_CTRLAx and
DMAC_CTRLBx register for channel x. For example, in the register, you can pro-
gram the following:
register.
– Source AHB master interface layer in the SIF field where source resides.
– Destination AHB master interface layer in the DIF field where destination resides.
– Incrementing/decrementing or fixed address for source in SRC_INCR field.
– Incrementing/decrementing or fixed address for destination in DST_INCR field.
e. If source picture-in-picture mode is enabled (DMAC_CTRLBx.SPIP is enabled),
program the DMAC_SPIPx register for channel x.
f. If destination picture-in-picture is enabled (DMAC_CTRLBx.DPIP), program the
DMAC_DPIPx register for channel x.
g. Write the channel configuration information into the DMAC_CFGx register for chan-
nel x. Ensure that the reload bits, DMAC_CFGx.SRC_REP,
DMAC_CFGx.DST_REP and DMAC_CTRLBx.AUTO are enabled.
programming the SRC_H2SEL/DST_h2SEL bits, respectively. Writing a ‘1’ activates
specific channel. Writing a ‘0’ activates the software handshaking interface to handle
source/destination requests.
peripheral, assign handshaking interface to the source and destination peripheral.
3. After the DMAC selected channel has been programmed, enable the channel by writing
a ‘1’ to the DMAC_CHER.ENABLE[n] bit where is the channel number. Make sure that
bit 0 of the DMAC_EN register is enabled.
buffer of data (assuming non-memory peripherals). The DMAC acknowledges on com-
pletion of each chunk/single transaction and carry out the buffer transfer.
5. When the buffer transfer has completed, the DMAC reloads the DMAC_SADDRx,
DMAC_DADDRx and DMAC_CTRLAx registers. Hardware sets the buffer Complete
interrupt. The DMAC then samples the row number as shown in Table 40-2 on page
920. If the DMAC is in Row 1, then the DMAC transfer has completed. Hardware sets
the transfer complete interrupt and disables the channel. So you can either respond to
the Buffer Complete or Chained buffer transfer Complete interrupts, or poll for the
928
AT91SAM9G45
Channel Enable in the Channel Status Register (DMAC_CHSR.ENABLE[n]) until it is

disabled, to detect when the transfer is complete. If the DMAC is not in Row 1, the next
step is performed.
6. The DMAC transfer proceeds as follows:
a. If interrupts is un-masked (DMAC_EBCIMR.BTC[x] = ‘1’, where x is the channel
number) hardware sets the buffer complete interrupt when the buffer transfer has
completed. It then stalls until the STALLED[n] bit of DMAC_CHSR register is
cleared by software, writing ‘1’ to DMAC_CHER.KEEPON[n] bit where n is the
channel number. If the next buffer is to be the last buffer in the DMAC transfer, then
the buffer complete ISR (interrupt service routine) should clear the automatic mode
bit in the DMAC_CTRLBx.AUTO bit. This put the DMAC into Row 1 as shown in
Table 40-2 on page 920. If the next buffer is not the last buffer in the DMAC trans-
fer, then the reload bits should remain enabled to keep the DMAC in Row 4.
b. If the buffer complete interrupt is masked (DMAC_EBCIMR.BTC[x] = ‘1’, where x is
the channel number), then hardware does not stall until it detects a write to the buf-
fer complete interrupt enable register DMAC_EBCIER register but starts the next
buffer transfer immediately. In this case software must clear the automatic mode bit
in the DMAC_CTRLB to put the DMAC into ROW 1 of Table 40-2 on page 920
before the last buffer of the DMAC transfer has completed. The transfer is similar to
that shown in Figure 40-9 on page 929. The DMAC transfer flow is shown in Figure
40-10 on page 930.
Figure 40-9. Multi-buffer DMAC Transfer with Source and Destination Address Auto-reloaded
Block0
Block1
Block2
SADDR DADDR
BlockN
929
AT91SAM9G45
Figure 40-10. DMAC Transfer Flow for Source and Destination Address Auto-reloaded
Channel Enabled by
software
Buffer Transfer
Replay mode for SADDRx,

DADDRx, CTRLAx, CTRLBx

generated here
HDMA Transfer Complete

Interrupt generated here yes Is HDMA in Row1 of
HDMA State Machine table?
Channel Disabled by
hardware no
no
EBCIMR[x]=1?
yes
Stall until STALLED is cleared

by writing to KEEPON field
40.4.5.5 Multi-buffer Transfer with Source Address Auto-reloaded and Linked List Destination Address (Row 6)
2. Set up the chain of linked list items (otherwise known as buffer descriptors) in memory.
Write the control information in the LLI.DMAC_CTRLAx and DMAC_CTRLBx registers
location of the buffer descriptor for each LLI in memory for channel x. For example, in
the register you can program the following:
nation) and flow control peripheral by programming the FC of the DMAC_CTRLBx
register.
930
AT91SAM9G45
3. Write the starting source address in the DMAC_SADDRx register for channel x.
Note: The values in the LLI.DMAC_SADDRx register locations of each of the Linked List Items (LLIs)
setup up in memory, although fetched during a LLI fetch, are not used.
x.
specific channel. Writing a ‘0’ activates the software handshaking interface
5. Make sure that the LLI.DMAC_CTRLBx register locations of all LLIs in memory (except
the last) are set as shown in Row 6 of Table 40-2 on page 920 while the
LLI.DMAC_CTRLBx register of the last Linked List item must be set as described in
Row 1 of Table 40-2. Figure 40-5 on page 919 shows a Linked List example with two
list items.
6. Make sure that the LLI.DMAC_DSCRx register locations of all LLIs in memory (except
the last) are non-zero and point to the next Linked List Item.
7. Make sure that the LLI.DMAC_DADDRx register location of all LLIs in memory point to
the start destination buffer address proceeding that LLI fetch.
8. Make sure that the LLI.DMAC_CTLx.DONE field of the LLI.DMAC_CTRLA register
locations of all LLIs in memory is cleared.
9. If source picture-in-picture is enabled (DMAC_CTRLBx.SPIP is enabled), program the
DMAC_SPIPx register for channel x.
10. If destination picture-in-picture is enabled (DMAC_CTRLBx.DPIP is enabled), program
the DMAC_DPIPx register for channel x.
ing to the DMAC_EBCISR register.
12. Program the DMAC_CTLx, DMAC_CFGx registers according to Row 6 as shown in
Linked List item.
14. Finally, enable the channel by writing a ‘1’ to the DMAC_CHER.ENABLE[n] bit where n
is the channel number. The transfer is performed. Make sure that bit 0 of the DMAC_EN
register is enabled.
Note: The LLI.DMAC_SADDRx, LLI.DMAC_DADDRx, LLI. DMAC_LLPx LLI.DMAC_CTRLAx and
LLI.DMAC_CTRLBx registers are fetched. The LLI.DMAC_SADDRx register although fetched is
not used.
buffer of data (assuming non-memory peripherals). DMAC acknowledges at the com-
pletion of every transaction (chunk and single) in the buffer and carry out the buffer
transfer.
17. The DMAC_CTRLAx register is written out to system memory. The DMAC_CTRLAx
register is written out to the same location on the same layer
(DMAC_DSCRx.DSCR_IF) where it was originally fetched, that is the location of the
DMAC_CTRLAx register of the linked list item fetched prior to the start of the buffer
931
AT91SAM9G45
transfer. Only DMAC_CTRLAx register is written out, because only the

DMAC_CTRLAx.BTSIZE and DMAC_CTRLAx.DONE fields have been updated by
hardware within the DMAC. The LLI.DMAC_CTRLAx.DONE bit is asserted to indicate
buffer completion Therefore, software can poll the LLI.DMAC_CTRLAx.DONE field of
the DMAC_CTRLAx register in the LLi to ascertain when a buffer transfer has
completed.
Note: Do not poll the DMAC_CTRLAx.DONE bit in the DMAC memory map. Instead poll the
LLI.DMAC_CTRLAx.DONE bit in the LLI for that buffer. If the polled LLI.DMAC_CTRLAx.DONE bit
is asserted, then this buffer transfer has completed. This LLI.DMAC_CTRLA.DONE bit was
cleared at the start of the transfer.
18. The DMAC reloads the DMAC_SADDRx register from the initial value. Hardware sets
the buffer complete interrupt. The DMAC samples the row number as shown in Table
40-2 on page 920. If the DMAC is in Row 1, then the DMAC transfer has completed.
Hardware sets the transfer complete interrupt and disables the channel. You can either
respond to the Buffer Complete or Chained buffer Transfer Complete interrupts, or poll
for the Channel Enable (DMAC_CHSR.ENABLE) bit until it is cleared by hardware, to
detect when the transfer is complete. If the DMAC is not in Row 1 as shown in Table 40-
2 on page 920, the following step is performed.
19. The DMAC fetches the next LLI from memory location pointed to by the current
DMAC_DSCRx register, and automatically reprograms the DMAC_DADDRx,
DMAC_CTRLAx, DMAC_CTRLBx and DMAC_DSCRx channel registers. Note that the
DMAC_SADDRx is not re-programmed as the reloaded value is used for the next
DMAC buffer transfer. If the next buffer is the last buffer of the DMAC transfer then the
DMAC_CTRLBx and DMAC_DSCRx registers just fetched from the LLI should match
Row 1 of Table 40-2 on page 920. The DMAC transfer might look like that shown in Fig-
ure 40-11 on page 932.
Figure 40-11. Multi-buffer DMAC Transfer with Source Address Auto-reloaded and Linked List Destination Address
Buffer0
DADDR(0)
Buffer1
DADDR(1)
SADDR
Buffer2
DADDR(2)
BufferN
DADDR(N)
The DMAC Transfer flow is shown in Figure 40-12 on page 933.
932
AT91SAM9G45
Figure 40-12. DMAC Transfer Flow for Replay Mode at Source and Linked List Destination Address
Channel Enabled by
software
LLI Fetch
Hardware reprograms
DADDRx, CTRLAx, CTRLBx, DSCRx
DMA buffer transfer
Writeback of control
status information in LLI
Reload SADDRx

generated here
Is HDMA in
yes Row1 of
HDMA Transfer Complete
interrupt generated here HDMA State Machine Table?
Channel Disabled by
no
hardware
40.4.5.6 Multi-buffer Transfer with Source Address Auto-reloaded and Contiguous Destination Address (Row 11)
ing to the Interrupt Status Register.
3. Program the following channel registers:
x.
c. Program DMAC_CTRLAx, DMAC_CTRLBx and DMAC_CFGx according to Row
11 as shown in Table 40-2 on page 920. Program the DMAC_DSCRx register with
‘0’. DMAC_CTRLBx.AUTO field is set to ‘1’ to enable automatic mode support.
d. Write the control information for the DMAC transfer in the DMAC_CTRLBx and
DMAC_CTRLAx register for channel x. For example, in this register, you can pro-
gram the following:
register.
933
AT91SAM9G45

– Source AHB master interface layer in the SIF field where source resides.
– Destination AHB master interface master layer in the DIF field where destination
resides.
– Incrementing/decrementing or fixed address for source in SRC_INCR field.
– Incrementing/decrementing or fixed address for destination in DST_INCR field.
e. If source picture-in-picture is enabled (DMAC_CTRLBx.SPIP is enabled), program
the DMAC_SPIPx register for channel x.
f. If destination picture-in-picture is enabled (DMAC_CTRLBx.DPIP), program the
DMAC_DPIPx register for channel x.
g. Write the channel configuration information into the DMAC_CFGx register for chan-
nel x.
programming the SRC_H2SEL/DST_H2SEL bits, respectively. Writing a ‘1’ activates
specific channel. Writing a ‘0’ activates the software handshaking interface to handle
4. After the DMAC channel has been programmed, enable the channel by writing a ‘1’ to
the DMAC_CHER.ENABLE[n] bit where n is the channel number. Make sure that bit 0
of the DMAC_EN.ENABLE register is enabled.
completion of every transaction (chunk and single) in the buffer and carries out the buf-
fer transfer.
6. When the buffer transfer has completed, the DMAC reloads the DMAC_SADDRx regis-
ter. The DMAC_DADDRx register remains unchanged. Hardware sets the buffer
complete interrupt. The DMAC then samples the row number as shown in Table 40-2
on page 920. If the DMAC is in Row 1, then the DMAC transfer has completed. Hard-
ware sets the transfer complete interrupt and disables the channel. So you can either
respond to the Buffer Complete or Transfer Complete interrupts, or poll for ENABLE
field in the Channel Status Register (DMAC_CHSR.ENABLE[n] bit) until it is cleared by
hardware, to detect when the transfer is complete. If the DMAC is not in Row 1, the next
step is performed.
7. The DMAC transfer proceeds as follows:
a. If the buffer complete interrupt is un-masked (DMAC_EBCIMR.BTC[x] = ‘1’, where
x is the channel number) hardware sets the buffer complete interrupt when the buf-
fer transfer has completed. It then stalls until STALLED[n] bit of DMAC_CHSR is
cleared by writing in the KEEPON[n] field of DMAC_CHER register where n is the
channel number. If the next buffer is to be the last buffer in the DMAC transfer, then
the buffer complete ISR (interrupt service routine) should clear the automatic mode
bit, DMAC_CTRLBx.AUTO. This puts the DMAC into Row 1 as shown in Table 40-
2 on page 920. If the next buffer is not the last buffer in the DMAC transfer then the
934
AT91SAM9G45
automatic transfer mode bit should remain enabled to keep the DMAC in Row 11 as
shown in Table 40-2 on page 920.
b. If the buffer complete interrupt is masked (DMAC_EBCIMR.BTC[x] = ‘1’, where x is
the channel number) then hardware does not stall until it detects a write to the buf-
fer transfer completed interrupt enable register but starts the next buffer transfer
immediately. In this case software must clear the automatic mode bit,
DMAC_CTRLBx.AUTO, to put the device into ROW 1 of Table 40-2 on page 920
before the last buffer of the DMAC transfer has completed.
The transfer is similar to that shown in Figure 40-13 on page 935.
The DMAC Transfer flow is shown in Figure 40-14 on page 936.
Figure 40-13. Multi-buffer Transfer with Source Address Auto-reloaded and Contiguous Destination Address
Buffer2
DADDR(2)
Buffer1
DADDR(1)
Buffer0
SADDR
DADDR(0)
935
AT91SAM9G45
Figure 40-14. DMAC Transfer Replay Mode is Enabled for the Source and Contiguous Destination Address
Channel Enabled by
software
Buffer Transfer
Replay mode for SADDRx,

Contiguous mode for DADDRx
CTRLAx, CTRLBx

generated here
Buffer Transfer Complete

interrupt generated here yes Is HDMA in Row1of
HDMA State Machine Table?
Channel Disabled by no
hardware
no
DMA_EBCIMR[x]=1?
yes
Stall until STALLED field is

cleared by software writing
KEEPON Field
40.4.5.7 Multi-buffer DMAC Transfer with Linked List for Source and Contiguous Destination Address (Row 2)
2. Set up the linked list in memory. Write the control information in the
LLI.DMAC_CTRLAx and LLI.DMAC_CTRLBx register location of the buffer descriptor
for each LLI in memory for channel x. For example, in the register, you can program the
following:
nation) and flow control device by programming the FC of the DMAC_CTRLBx
register.
936
AT91SAM9G45

3. Write the starting destination address in the DMAC_DADDRx register for channel x.
Note: The values in the LLI.DMAC_DADDRx register location of each Linked List Item (LLI) in memory,
although fetched during an LLI fetch, are not used.
x.
specific channel. Writing a ‘0’ activates the software handshaking interface to han-
dle source/destination requests.
peripheral, assign handshaking interface to the source and destination peripherals.
5. Make sure that all LLI.DMAC_CTRLBx register locations of the LLI (except the last) are
set as shown in Row 2 of Table 40-2 on page 920, while the LLI.DMAC_CTRLBx regis-
ter of the last Linked List item must be set as described in Row 1 of Table 40-2. Figure
40-5 on page 919 shows a Linked List example with two list items.
6. Make sure that the LLI.DMAC_DSCRx register locations of all LLIs in memory (except
the last) are non-zero and point to the next Linked List Item.
7. Make sure that the LLI.DMAC_SADDRx register location of all LLIs in memory point to
the start source buffer address proceeding that LLI fetch.
8. Make sure that the LLI.DMAC_CTRLAx.DONE field of the LLI.DMAC_CTRLAx register
locations of all LLIs in memory is cleared.
9. If source picture-in-picture is enabled (DMAC_CTRLBx.SPIP is enabled), program the
DMAC_SPIPx register for channel x.
10. If destination picture-in-picture is enabled (DMAC_CTRLBx.DPIP is enabled), program
the DMAC_DPIPx register for channel x.
ing the interrupt status register.
12. Program the DMAC_CTRLAx, DMAC_CTRLBx and DMAC_CFGx registers according
to Row 2 as shown in Table 40-2 on page 920
Linked List item.
14. Finally, enable the channel by writing a ‘1’ to the DMAC_CHER.ENABLE[n] bit. The
transfer is performed. Make sure that bit 0 of the DMAC_EN register is enabled.
Note: The LLI.DMAC_SADDRx, LLI.DMAC_DADDRx, LLI.DMAC_DSCRx and LLI.DMAC_CTRLA/Bx
registers are fetched. The LLI.DMAC_DADDRx register location of the LLI although fetched is not
used. The DMAC_DADDRx register in the DMAC remains unchanged.
16. Source and destination requests single and chunk DMAC transactions to transfer the
transfer
17. Once the buffer of data is transferred, the DMAC_CTRLAx register is written out to sys-
tem memory at the same location and on the same layer (DMAC_DSCRx.DSCR_IF)
where it was originally fetched, that is, the location of the DMAC_CTRLAx register of
937
AT91SAM9G45
the linked list item fetched prior to the start of the buffer transfer. Only DMAC_CTRLAx
register is written out because only the DMAC_CTRLAx.BTSIZE and
DMAC_CTRLAX.DONE fields have been updated by DMAC hardware. Additionally, the
DMAC_CTRLAx.DONE bit is asserted when the buffer transfer has completed.
Note: Do not poll the DMAC_CTRLAx.DONE bit in the DMAC memory map. Instead, poll the
LLI.DMAC_CTRLAx.DONE bit in the LLI for that buffer. If the poll LLI.DMAC_CTRLAx.DONE bit is
asserted, then this buffer transfer has completed. This LLI.DMAC_CTRLAx.DONE bit was cleared
at the start of the transfer.
18. The DMAC does not wait for the buffer interrupt to be cleared, but continues and
fetches the next LLI from the memory location pointed to by current DMAC_DSCRx
register and automatically reprograms the DMAC_SADDRx, DMAC_CTRLAx,
DMAC_CTRLBx and DMAC_DSCRx channel registers. The DMAC_DADDRx register
is left unchanged. The DMAC transfer continues until the DMAC samples the
DMAC_CTRLAx, DMAC_CTRLBx and DMAC_DSCRx registers at the end of a buffer
transfer match that described in Row 1 of Table 40-2 on page 920. The DMAC then
knows that the previous buffer transferred was the last buffer in the DMAC transfer.
The DMAC transfer might look like that shown in Figure 40-15 on page 938 Note that the desti-
nation address is decrementing.
Figure 40-15. DMAC Transfer with Linked List Source Address and Contiguous Destination Address
Buffer 2
SADDR(2)
Buffer 2
DADDR(2)
Buffer 1 Buffer 1
SADDR(1) DADDR(1)
Buffer 0
DADDR(0)
Buffer 0
SADDR(0)
The DMAC transfer flow is shown in Figure 40-16 on page 939.
938
AT91SAM9G45
Figure 40-16. DMAC Transfer Flow for Linked List Source Address and Contiguous Destination Address
Channel Enabled by
software
LLI Fetch
Hardware reprograms
SADDRx, CTRLAx,CTRLBx, DSCRx
HDMA buffer transfer
Writeback of control
information of LLI

generated here
Is HDMA in no
Row 1 ?
HDMA Transfer Complete yes

interrupt generated here
Channel Disabled by
hardware
40.4.6 Disabling a Channel Prior to Transfer Completion

Under normal operation, software enables a channel by writing a ‘1’ to the Channel Handler
Enable Register, DMAC_CHER.ENABLE[n], and hardware disables a channel on transfer com-
pletion by clearing the DMAC_CHSR.ENABLE[n] register bit.
The recommended way for software to disable a channel without losing data is to use the SUS-
PEND[n] bit in conjunction with the EMPTY[n] bit in the Channel Handler Status Register.
939
AT91SAM9G45
1. If software wishes to disable a channel n prior to the DMAC transfer completion, then it
can set the DMAC_CHER.SUSPEND[n] bit to tell the DMAC to halt all transfers from
the source peripheral. Therefore, the channel FIFO receives no new data.
2. Software can now poll the DMAC_CHSR.EMPTY[n] bit until it indicates that the channel
n FIFO is empty, where n is the channel number.
3. The DMAC_CHER.ENABLE[n] bit can then be cleared by software once the channel n
FIFO is empty, where n is the channel number.
When DMAC_CTRLAx.SRC_WIDTH is less than DMAC_CTRLAx.DST_WIDTH and the
DMAC_CHSRx.SUSPEND[n] bit is high, the DMAC_CHSRx.EMPTY[n] is asserted once the
contents of the FIFO do not permit a single word of DMAC_CTRLAx.DST_WIDTH to be formed.
However, there may still be data in the channel FIFO but not enough to form a single transfer of
DMAC_CTLx.DST_WIDTH width. In this configuration, once the channel is disabled, the remain-
ing data in the channel FIFO are not transferred to the destination peripheral. It is permitted to
remove the channel from the suspension state by writing a ‘1’ to the DMAC_CHER.RESUME[n]
field register. The DMAC transfer completes in the normal manner. n defines the channel
number.
Note: If a channel is disabled by software, an active single or chunk transaction is not guaranteed to
receive an acknowledgement.
40.4.6.1 Abnormal Transfer Termination

A DMAC transfer may be terminated abruptly by software by clearing the channel enable bit,
DMAC_CHDR.ENABLE[n] where n is the channel number. This does not mean that the channel
is disabled immediately after the DMAC_CHSR.ENABLE[n] bit is cleared over the APB inter-
face. Consider this as a request to disable the channel. The DMAC_CHSR.ENABLE[n] must be
polled and then it must be confirmed that the channel is disabled by reading back 0.
Software may terminate all channels abruptly by clearing the global enable bit in the DMAC Con-
figuration Register (DMAC_EN.ENABLE bit). Again, this does not mean that all channels are
disabled immediately after the DMAC_EN.ENABLE is cleared over the APB slave interface.
Consider this as a request to disable all channels. The DMAC_CHSR.ENABLE must be polled
and then it must be confirmed that all channels are disabled by reading back ‘0’.
Note: If the channel enable bit is cleared while there is data in the channel FIFO, this data is not sent to
the destination peripheral and is not present when the channel is re-enabled. For read sensitive
source peripherals, such as a source FIFO, this data is therefore lost. When the source is not a
read sensitive device (i.e., memory), disabling a channel without waiting for the channel FIFO to
empty may be acceptable as the data is available from the source peripheral upon request and is
not lost.
Note: If a channel is disabled by software, an active single or chunk transaction is not guaranteed to
receive an acknowledgement.
40.5 DMAC Software Requirements

• There must not be any write operation to Channel registers in an active channel after the
channel enable is made HIGH. If any channel parameters must be reprogrammed, this can
only be done after disabling the DMAC channel.
• When destination peripheral is defined as the flow controller, source single transfer request
are not serviced until Destination Peripheral has asserted its Last Transfer Flag.
• When Source Peripheral is flow controller, destination single transfer request are not serviced
until Source Peripheral has asserted its Last Transfer Flag.
940
AT91SAM9G45
• When destination peripheral is defined as the flow controller, if the destination width is
smaller than the source width, then a data loss may occur, and the loss is equal to Source
Single Transfer size in bytes- destination Single Transfer size in bytes.
• When a Memory to Peripheral transfer occurs if the destination peripheral is flow controller,
then a prefetch operation is performed. It means that data are extracted from memory before
any request from the peripheral is generated.
• You must program the DMAC_SADDRx and DMAC_DADDRx channel registers with a byte,
half-word and word aligned address depending on the source width and destination width.
• After the software disables a channel by writing into the channel disable register, it must re-
enable the channel only after it has polled a 0 in the corresponding channel enable status
register. This is because the current AHB Burst must terminate properly.
• If you program the BTSIZE field in the DMAC_CTRLA, as zero, and the DMAC is defined as
the flow controller, then the channel is automatically disabled.
• When hardware handshaking interface protocol is fully implemented, a peripheral is expected
to deassert any sreq or breq signals on receiving the ack signal irrespective of the request
the ack was asserted in response to.
• Multiple Transfers involving the same peripheral must not be programmed and enabled on
different channel, unless this peripheral integrates several hardware handshaking interface.
• When a Peripheral is flow controller, the targeted DMAC Channel must be enabled before the
Peripheral. If you do not ensure this the DMAC Channel might miss a Last Transfer Flag, if
the First DMAC request is also the last transfer.
• When AUTO Field is set to TRUE, then the BTSIZE Field is automatically reloaded from its
previous value. BTSIZE must be initialized to a non zero value if the first transfer is initiated
with AUTO field set to TRUE even if LLI mode is enabled because the LLI fetch operation will
not update this field.
941
AT91SAM9G45
40.6 DMA Controller (DMAC) User Interface

0x000 DMAC Global Configuration Register DMAC_GCFG Read-write 0x10
0x004 DMAC Enable Register DMAC_EN Read-write 0x0
0x008 DMAC Software Single Request Register DMAC_SREQ Read-write 0x0

0x00C DMAC Software Chunk Transfer Request Register DMAC_CREQ Read-write 0x0
0x010 DMAC Software Last Transfer Flag Register DMAC_LAST Read-write 0x0
DMAC Error, Chained Buffer transfer completed and Buffer

0x018 DMAC_EBCIER Write-only –
transfer completed Interrupt Enable register.

0x01C DMAC_EBCIDR Write-only –
transfer completed Interrupt Disable register.

0x020 DMAC_EBCIMR Read-only 0x0
transfer completed Mask Register.
0x024 DMAC_EBCISR Read-only 0x0
transfer completed Status Register.
0x028 DMAC Channel Handler Enable Register DMAC_CHER Write-only –
0x02C DMAC Channel Handler Disable Register DMAC_CHDR Write-only –
0x030 DMAC Channel Handler Status Register DMAC_CHSR Read-only 0x00FF0000
0x03C+ch_num*(0x28)+(0x0) DMAC Channel Source Address Register DMAC_SADDR Read-write 0x0
0x03C+ch_num*(0x28)+(0x4) DMAC Channel Destination Address Register DMAC_DADDR Read-write 0x0
0x03C+ch_num*(0x28)+(0x8) DMAC Channel Descriptor Address Register DMAC_DSCR Read-write 0x0

0x03C+ch_num*(0x28)+(0xC) DMAC Channel Control A Register DMAC_CTRLA Read-write 0x0
0x03C+ch_num*(0x28)+(0x10) DMAC Channel Control B Register DMAC_CTRLB Read-write 0x0
0x03C+ch_num*(0x28)+(0x14) DMAC Channel Configuration Register DMAC_CFG Read-write 0x01000000
DMAC Channel Source Picture in Picture Configuration

0x03C+ch_num*(0x28)+(0x18) DMAC_SPIP Read-write 0x0
Register
DMAC Channel Destination Picture in Picture Configuration

0x03C+ch_num*(0x28)+(0x1C) DMAC_DPIP Read-write 0x0
Register
0x03C+ch_num*(0x28)+(0x20) Reserved – – –
0x03C+ch_num*(0x28)+(0x24) Reserved – – –
0x064 - 0x140 DMAC Channel 1 to 7 Register(1) Read-write 0x0
0x017C- 0x1FC Reserved – – –

Note: 1. The addresses for the DMAC registers shown here are for DMA Channel 0. This sequence of registers is repeated succes-
sively for each DMA channel located between 0x064 and 0x140.
942
AT91SAM9G45
40.6.1 DMAC Global Configuration Register

Name: DMAC_GCFG
Address: 0xFFFFEC00
Access: Read-write
Reset: 0x00000010
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – ARB_CFG – – – –
Note: Bit fields 0, 1, 2, 3, have a default value of 0. This should not be changed.
• ARB_CFG
0: Fixed priority arbiter.
1: Modified round robin arbiter.
943
AT91SAM9G45
40.6.2 DMAC Enable Register

Name:: DMAC_EN
Address: 0xFFFFEC04
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – ENABLE
• ENABLE
0: DMA Controller is disabled.
1: DMA Controller is enabled.
944
AT91SAM9G45
40.6.3 DMAC Software Single Request Register

Name: DMAC_SREQ
Address: 0xFFFFEC08
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
DSREQ7 SSREQ7 DSREQ6 SSREQ6 DSREQ5 SSREQ5 DSREQ4 SSREQ4
7 6 5 4 3 2 1 0
DSREQ3 SSREQ3 DSREQ2 SSREQ2 DSREQ1 SSREQ1 DSREQ0 SSREQ0
• DSREQx
Request a destination single transfer on channel i.
• SSREQx
Request a source single transfer on channel i.
945
AT91SAM9G45
40.6.4 DMAC Software Chunk Transfer Request Register

Name: DMAC_CREQ
Address: 0xFFFFEC0C
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
DCREQ7 SCREQ7 DCREQ6 SCREQ6 DCREQ5 SCREQ5 DCREQ4 SCREQ4
7 6 5 4 3 2 1 0
DCREQ3 SCREQ3 DCREQ2 SCREQ2 DCREQ1 SCREQ1 DCREQ0 SCREQ0
• DCREQx
Request a destination chunk transfer on channel i.
• SCREQx
Request a source chunk transfer on channel i.
946
AT91SAM9G45
40.6.5 DMAC Software Last Transfer Flag Register

Name: DMAC_LAST
Address: 0xFFFFEC10
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
DLAST7 SLAST7 DLAST6 SLAST6 DLAST5 SLAST5 DLAST4 SLAST4
7 6 5 4 3 2 1 0
DLAST3 SLAST3 DLAST2 SLAST2 DLAST1 SLAST1 DLAST0 SLAST0
• DLASTx
Writing one to DLASTx prior to writing one to DSREQx or DCREQx indicates that this destination request is the last transfer
of the buffer.
• SLASTx
Writing one to SLASTx prior to writing one to SSREQx or SCREQx indicates that this source request is the last transfer of
the buffer.
947
AT91SAM9G45
40.6.6 DMAC Error, Buffer Transfer and Chained Buffer Transfer Interrupt Enable Register
Name: DMAC_EBCIER
Address: 0xFFFFEC18
Access: Write-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
ERR7 ERR6 ERR5 ERR4 ERR3 ERR2 ERR1 ERR0
15 14 13 12 11 10 9 8
CBTC7 CBTC6 CBTC5 CBTC4 CBTC3 CBTC2 CBTC1 CBTC0
7 6 5 4 3 2 1 0
BTC7 BTC6 BTC5 BTC4 BTC3 BTC2 BTC1 BTC0
• BTC[7:0]
Buffer Transfer Completed Interrupt Enable Register. Set the relevant bit in the BTC field to enable the interrupt for channel
i.
• CBTC[7:0]
Chained Buffer Transfer Completed Interrupt Enable Register. Set the relevant bit in the CBTC field to enable the interrupt
for channel i.
• ERR[7:0]
Access Error Interrupt Enable Register. Set the relevant bit in the ERR field to enable the interrupt for channel i.
948
AT91SAM9G45
40.6.7 DMAC Error, Buffer Transfer and Chained Buffer Transfer Interrupt Disable Register
Name: DMAC_EBCIDR
Address: 0xFFFFEC1C
Access: Write-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• BTC[7:0]
Buffer transfer completed Disable Interrupt Register. When set, a bit of the BTC field disables the interrupt from the rele-
vant DMAC channel.
• CBTC[7:0]
Chained Buffer transfer completed Disable Register. When set, a bit of the CBTC field disables the interrupt from the rele-
vant DMAC channel.
• ERR[7:0]
Access Error Interrupt Disable Register. When set, a bit of the ERR field disables the interrupt from the relevant DMAC
channel.
949
AT91SAM9G45
40.6.8 DMAC Error, Buffer Transfer and Chained Buffer Transfer Interrupt Mask Register
Name: DMAC_EBCIMR
Address: 0xFFFFEC20
Access: Read-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• BTC[7:0]
0: Buffer Transfer completed interrupt is disabled for channel i.
1: Buffer Transfer completed interrupt is enabled for channel i.
• CBTC[7:0]
0: Chained Buffer Transfer interrupt is disabled for channel i.
1: Chained Buffer Transfer interrupt is enabled for channel i.
• ERR[7:0]
0: Transfer Error Interrupt is disabled for channel i.
1: Transfer Error Interrupt is enabled for channel i.
950
AT91SAM9G45
40.6.9 DMAC Error, Buffer Transfer and Chained Buffer Transfer Status Register
Name: DMAC_EBCISR
Address: 0xFFFFEC24
Access: Read-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• BTC[7:0]
When BTC[i] is set, Channel i buffer transfer has terminated.
• CBTC[7:0]
When CBTC[i] is set, Channel i Chained buffer has terminated. LLI Fetch operation is disabled.
• ERR[7:0]
When ERR[i] is set, Channel i has detected an AHB Read or Write Error Access.
951
AT91SAM9G45
40.6.10 DMAC Channel Handler Enable Register

Name: DMAC_CHER
Address: 0xFFFFEC28
Access: Write-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
KEEP7 KEEP6 KEEP5 KEEP4 KEEP3 KEEP2 KEEP1 KEEP0
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
SUSP7 SUSP6 SUSP5 SUSP4 SUSP3 SUSP2 SUSP1 SUSP0
7 6 5 4 3 2 1 0
ENA7 ENA6 ENA5 ENA4 ENA3 ENA2 ENA1 ENA0
• ENA[7:0]
When set, a bit of the ENA field enables the relevant channel.
• SUSP[7:0]
When set, a bit of the SUSPfield freezes the relevant channel and its current context.
• KEEP[7:0]
When set, a bit of the KEEP field resumes the current channel from an automatic stall state.
952
AT91SAM9G45
40.6.11 DMAC Channel Handler Disable Register

Name: DMAC_CHDR
Address: 0xFFFFEC2C
Access: Write-only
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RES7 RES6 RES5 RES4 RES3 RES2 RES1 RES0
7 6 5 4 3 2 1 0
DIS7 DIS6 DIS5 DIS4 DIS3 DIS2 DIS1 DIS0
• DIS[7:0]
Write one to this field to disable the relevant DMAC Channel. The content of the FIFO is lost and the current AHB access is
terminated. Software must poll DIS[7:0] field in the DMAC_CHSR register to be sure that the channel is disabled.
• RES[7:0]
Write one to this field to resume the channel transfer restoring its context.
953
AT91SAM9G45
40.6.12 DMAC Channel Handler Status Register

Name: DMAC_CHSR
Address: 0xFFFFEC30
Access: Read-only
Reset: 0x00FF0000
31 30 29 28 27 26 25 24
STAL7 STAL6 STAL5 STAL4 STAL3 STAL2 STAL1 STAL0
23 22 21 20 19 18 17 16
EMPT7 EMPT6 EMPT5 EMPT4 EMPT3 EMPT2 EMPT1 EMPT0
15 14 13 12 11 10 9 8
SUSP7 SUSP6 SUSP5 SUSP4 SUSP3 SUSP2 SUSP1 SUSP0
7 6 5 4 3 2 1 0
ENA7 ENA6 ENA5 ENA4 ENA3 ENA2 ENA1 ENA0
• ENA[7:0]
A one in any position of this field indicates that the relevant channel is enabled.
• SUSP[7:0]
A one in any position of this field indicates that the channel transfer is suspended.
• EMPT[7:0]
A one in any position of this field indicates that the relevant channel is empty.
• STAL[7:0]
A one in any position of this field indicates that the relevant channel is stalling.
954
AT91SAM9G45
40.6.13 DMAC Channel x [x = 0..7] Source Address Register

Name: DMAC_SADDRx [x = 0..7]
Addresses: 0xFFFFEC3C [0], 0xFFFFEC64 [1], 0xFFFFEC8C [2], 0xFFFFECB4 [3], 0xFFFFECDC [4],
0xFFFFED04 [5], 0xFFFFED2C [6], 0xFFFFED54 [7]
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
SADDRx
23 22 21 20 19 18 17 16
SADDRx
15 14 13 12 11 10 9 8
SADDRx
7 6 5 4 3 2 1 0
SADDRx
• SADDRx
Channel x source address. This register must be aligned with the source transfer width.
40.6.14 DMAC Channel x [x = 0..7] Destination Address Register

Name: DMAC_DADDRx [x = 0..7]
Addresses: 0xFFFFEC40 [0], 0xFFFFEC68 [1], 0xFFFFEC90 [2], 0xFFFFECB8 [3], 0xFFFFECE0 [4], 0xFFFFED08 [5],
0xFFFFED30 [6], 0xFFFFED58 [7]
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
DADDRx
23 22 21 20 19 18 17 16
DADDRx
15 14 13 12 11 10 9 8
DADDRx
7 6 5 4 3 2 1 0
DADDRx
• DADDRx
Channel x destination address. This register must be aligned with the destination transfer width.
955
AT91SAM9G45
40.6.15 DMAC Channel x [x = 0..7] Descriptor Address Register

Name: DMAC_DSCRx [x = 0..7]
Addresses: 0xFFFFEC44 [0], 0xFFFFEC6C [1], 0xFFFFEC94 [2], 0xFFFFECBC [3], 0xFFFFECE4 [4], 0xFFFFED0[5]
0xFFFFED34 [6], 0xFFFFED5C [7]
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
DSCRx
23 22 21 20 19 18 17 16
DSCRx
15 14 13 12 11 10 9 8
DSCRx
7 6 5 4 3 2 1 0
DSCRx DSCRx_IF
• DSCRx_IF
00: The Buffer Transfer descriptor is fetched via AHB-Lite Interface 0.
01: The Buffer Transfer descriptor is fetched via AHB-Lite Interface 1.
10: Reserved.
11: Reserved.
• DSCRx
Buffer Transfer descriptor address. This address is word aligned.
956
AT91SAM9G45
40.6.16 DMAC Channel x [x = 0..7] Control A Register

Name: DMAC_CTRLAx [x = 0..7]
Addresses: 0xFFFFEC48 [0], 0xFFFFEC70 [1], 0xFFFFEC98 [2], 0xFFFFECC0 [3], 0xFFFFECE8 [4], 0xFFFFED10 [5],
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
DONE – DST_WIDTH – – SRC_WIDTH
23 22 21 20 19 18 17 16
– DCSIZE – SCSIZE
15 14 13 12 11 10 9 8
BTSIZE
7 6 5 4 3 2 1 0
BTSIZE
• BTSIZE
Buffer Transfer Size. The transfer size relates to the number of transfers to be performed, that is, for writes it refers to the
number of source width transfers to perform when DMAC is flow controller. For Reads, BTSIZE refers to the number of
transfers completed on the Source Interface. When this field is set to 0, the DMAC module is automatically disabled when
the relevant channel is enabled.
• SCSIZE
Source Chunk Transfer Size.
SCSIZE value Number of data transferred

000 1
001 4
010 8
011 16
100 32
101 64
110 128
111 256
• DCSIZE
Destination Chunk Transfer size.
DCSIZE Number of data transferred

000 1
001 4
010 8
957
AT91SAM9G45
DCSIZE Number of data transferred

011 16
100 32
101 64
110 128
111 256
• SRC_WIDTH
SRC_WIDTH Single Transfer Size

00 BYTE
01 HALF-WORD
1X WORD
• DST_WIDTH
DST_WIDTH Single Transfer Size

00 BYTE
01 HALF-WORD
1X WORD
• DONE
0: The transfer is performed.
1: If SOD field of DMAC_CFG register is set to true, then the DMAC is automatically disabled when an LLI updates the con-
tent of this register.
The DONE field is written back to memory at the end of the transfer.
958
AT91SAM9G45
40.6.17 DMAC Channel x [x = 0..7] Control B Register

Name: DMAC_CTRLBx [x = 0..7]
Addresses: 0xFFFFEC4C [0], 0xFFFFEC74 [1], 0xFFFFEC9C [2], 0xFFFFECC4 [3], 0xFFFFECEC [4],
0xFFFFED14 [5], 0xFFFFED3C [6], 0xFFFFED64 [7]
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
AUTO IEN DST_INCR – – SRC_INCR
23 22 21 20 19 18 17 16
FC DST_DSCR – – – SRC_DSCR
15 14 13 12 11 10 9 8
– – DST_PIP – – – SRC_PIP
7 6 5 4 3 2 1 0
– – DIF – – SIF
• SIF: Source Interface Selection Field

00: The source transfer is done via AHB-Lite Interface 0.
01: The source transfer is done via AHB-Lite Interface 1.
10: Reserved.
11: Reserved.
• DIF: Destination Interface Selection Field

00: The destination transfer is done via AHB-Lite Interface 0.
01: The destination transfer is done via AHB-Lite Interface 1.
10: Reserved.
11: Reserved.
• SRC_PIP
0: Picture-in-Picture mode is disabled. The source data area is contiguous.
1: Picture-in-Picture mode is enabled. When the source PIP counter reaches the programmable boundary, the address is
automatically increment of a user defined amount.
• DST_PIP
0: Picture-in-Picture mode is disabled. The Destination data area is contiguous.
1: Picture-in-Picture mode is enabled. When the Destination PIP counter reaches the programmable boundary the address
is automatically incremented by a user-defined amount.
• SRC_DSCR
0: Source address is updated when the descriptor is fetched from the memory.
1: Buffer Descriptor Fetch operation is disabled for the source.
959
AT91SAM9G45
• DST_DSCR
0: Destination address is updated when the descriptor is fetched from the memory.
1: Buffer Descriptor Fetch operation is disabled for the destination.
• FC
This field defines which device controls the size of the buffer transfer, also referred as to the Flow Controller.
FC Type of transfer Flow Controller

000 Memory-to-Memory DMA Controller
001 Memory-to-Peripheral DMA Controller
010 Peripheral-to-Memory DMA Controller
011 Peripheral-to-Peripheral DMA Controller
100 Peripheral-to-Memory Peripheral
101 Memory-to-Peripheral Peripheral
110 Peripheral-to-Peripheral Source Peripheral
111 Peripheral-to-Peripheral Destination Peripheral
• SRC_INCR
SRC_INCR Type of addressing mode

00 INCREMENTING
01 DECREMENTING
10 FIXED
• DST_INCR
DST_INCR Type of addressing scheme

00 INCREMENTING
01 DECREMENTING
10 FIXED
• IEN
If this bit is cleared, when the buffer transfer is completed, the BTC[x] flag is set in the EBCISR status register. This bit is
active low.
• AUTO
Automatic multiple buffer transfer is enabled. When set, this bit enables replay mode or contiguous mode when several buf-
fers are transferred.
960
AT91SAM9G45
40.6.18 DMAC Channel x [x = 0..7] Configuration Register

Name: DMAC_CFGx [x = 0..7]
Addresses: 0xFFFFEC50 [0], 0xFFFFEC78 [1], 0xFFFFECA0 [2], 0xFFFFECC8 [3], 0xFFFFECF0 [4], 0xFFFFED18 [5],
Access: Read-write
Reset: 0x0100000000
31 30 29 28 27 26 25 24
– – FIFOCFG – AHB_PROT
23 22 21 20 19 18 17 16
– LOCK_IF_L LOCK_B LOCK_IF – – – SOD
15 14 13 12 11 10 9 8
– – DST_H2SEL DST_REP – – SRC_H2SEL SRC_REP
7 6 5 4 3 2 1 0
DST_PER SRC_PER
• SRC_PER
Channel x Source Request is associated with peripheral identifier coded SRC_PER handshaking interface.
• DST_PER
Channel x Destination Request is associated with peripheral identifier coded DST_PER handshaking interface.
• SRC_REP
0: When automatic mode is activated, source address is contiguous between two buffers.
1: When automatic mode is activated, the source address and the control register are reloaded from previous transfer.
• SRC_H2SEL
0: Software handshaking interface is used to trigger a transfer request.
1: Hardware handshaking interface is used to trigger a transfer request.
• DST_REP
0: When automatic mode is activated, destination address is contiguous between two buffers.
1: When automatic mode is activated, the destination and the control register are reloaded from the previous transfer.
• DST_H2SEL
0: Software handshaking interface is used to trigger a transfer request.
1: Hardware handshaking interface is used to trigger a transfer request.
• SOD
0: STOP ON DONE disabled, the descriptor fetch operation ignores DONE Field of CTRLA register.
1: STOP ON DONE activated, the DMAC module is automatically disabled if DONE FIELD is set to 1.
961
AT91SAM9G45
• LOCK_IF
0: Interface Lock capability is disabled
1: Interface Lock capability is enabled
• LOCK_B
0: AHB Bus Locking capability is disabled.
1: AHB Bus Locking capability is enabled.
• LOCK_IF_L
0: The Master Interface Arbiter is locked by the channel x for a chunk transfer.
1: The Master Interface Arbiter is locked by the channel x for a buffer transfer.
• AHB_PROT
AHB_PROT field provides additional information about a bus access and is primarily used to implement some level of
protection.
HPROT[3] HPROT[2] HPROT[1] HPROT[0] Description

1 Data access
0: User Access
AHB_PROT[0]
1: Privileged Access
0: Not Bufferable
AHB_PROT[1]
1: Bufferable
0: Not cacheable
AHB_PROT[2]
1: Cacheable
• FIFOCFG
FIFOCFG FIFO request

00 The largest defined length AHB burst is performed on the destination AHB interface.
01 When half FIFO size is available/filled, a source/destination request is serviced.
10 When there is enough space/data available to perfom a single AHB access, then the request is serviced.
962
AT91SAM9G45
40.6.19 DMAC Channel x [x = 0..7] Source Picture in Picture Configuration Register

Name: DMAC_SPIPx [x = 0..7]
Addresses: 0xFFFFEC54 [0], 0xFFFFEC7C [1], 0xFFFFECA4 [2], 0xFFFFECCC [3], 0xFFFFECF4 [4],
0xFFFFED1C [5], 0xFFFFED44 [6], 0xFFFFED6C [7]
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – SPIP_BOUNDARY
23 22 21 20 19 18 17 16
SPIP_BOUNDARY
15 14 13 12 11 10 9 8
SPIP_HOLE
7 6 5 4 3 2 1 0
SPIP_HOLE
• SPIP_HOLE
This field indicates the value to add to the address when the programmable boundary has been reached.
• SPIP_BOUNDARY
This field indicates the number of source transfers to perform before the automatic address increment operation.
963
40.6.20 DMAC Channel x [x = 0..7] Destination Picture in Picture Configuration Register
Name: DMAC_DPIPx [x = 0..7]
Addresses: 0xFFFFEC58 [0], 0xFFFFEC80 [1], 0xFFFFECA8 [2], 0xFFFFECD0 [3], 0xFFFFECF8 [4], 0xFFFFED20 [5],
Access: Read-write
Reset: 0x00000000
31 30 29 28 27 26 25 24
– – – – – – DPIP_BOUNDARY
23 22 21 20 19 18 17 16
DPIP_BOUNDARY
15 14 13 12 11 10 9 8
DPIP_HOLE
7 6 5 4 3 2 1 0
DPIP_HOLE
• DPIP_HOLE
This field indicates the value to add to the address when the programmable boundary has been reached.
• DPIP_BOUNDARY
This field indicates the number of source transfers to perform before the automatic address increment operation.
964 AT91SAM9G45
AT91SAM9G45
41. Pulse Width Modulation Controller (PWM)
41.1 Description
The PWM macrocell controls several channels independently. Each channel controls one
square output waveform. Characteristics of the output waveform such as period, duty-cycle and
polarity are configurable through the user interface. Each channel selects and uses one of the
clocks provided by the clock generator. The clock generator provides several clocks resulting
from the division of the PWM macrocell master clock.
All PWM macrocell accesses are made through APB mapped registers.
Channels can be synchronized, to generate non overlapped waveforms. All channels integrate a
double buffering system in order to prevent an unexpected output waveform while modifying the
period or the duty-cycle.

• Four channels, one 16-bit counter per channel
• Common clock generator, providing Thirteen Different Clocks
– A Modulo n counter providing eleven clocks
– Two independent Linear Dividers working on modulo n counter outputs
• Independent channel programming
– Independent Enable Disable Commands
– Independent Clock Selection
– Independent Period and Duty Cycle, with Double Buffering
– Programmable selection of the output waveform polarity
– Programmable center or left aligned output waveform
965
41.3 Block Diagram
Figure 41-1. Pulse Width Modulation Controller Block Diagram
PWM
Controller
PWMx
Channel Period
Update PWMx
Comparator
Duty Cycle PWMx
Clock
Selector Counter
PIO
PWM0 Period
Channel
Update PWM0
Comparator
Duty Cycle PWM0
Clock
Selector Counter
MCK
PMC Clock Generator APB Interface Interrupt Generator AIC
APB

Each channel outputs one waveform on one external I/O line.

PWMx PWM Waveform Output for channel x Output
41.5.1 I/O Lines

The pins used for interfacing the PWM may be multiplexed with PIO lines. The programmer must
first program the PIO controller to assign the desired PWM pins to their peripheral function. If I/O
lines of the PWM are not used by the application, they can be used for other purposes by the
PIO controller.
966 AT91SAM9G45
AT91SAM9G45
All of the PWM outputs may or may not be enabled. If an application requires only four channels,
then only four PIO lines will be assigned to PWM outputs.

PWM PWM0 PD24 B
PWM PWM1 PD25 B
PWM PWM1 PD31 B
PWM PWM2 PD26 B
PWM PWM2 PE31 A
PWM PWM3 PA25 B
PWM PWM3 PD0 B

The PWM is not continuously clocked. The programmer must first enable the PWM clock in the
Power Management Controller (PMC) before using the PWM. However, if the application does
not require PWM operations, the PWM clock can be stopped when not needed and be restarted
later. In this case, the PWM will resume its operations where it left off.
Configuring the PWM does not require the PWM clock to be enabled.

The PWM interrupt line is connected on one of the internal sources of the Advanced Interrupt
Controller. Using the PWM interrupt requires the AIC to be programmed first. Note that it is not
recommended to use the PWM interrupt line in edge sensitive mode.

Instance ID
PWM 19

The PWM macrocell is primarily composed of a clock generator module and 4 channels.
– Clocked by the system clock, MCK, the clock generator module provides 13 clocks.
– Each channel can independently choose one of the clock generator outputs.
– Each channel generates an output waveform with attributes that can be defined
independently for each channel through the user interface registers.
967
41.6.1 PWM Clock Generator
Figure 41-2. Functional View of the Clock Generator Block Diagram
modulo n counter
MCK
MCK
MCK/2
MCK/4
MCK/8
MCK/16
MCK/32
MCK/64
MCK/128
MCK/256
MCK/512
MCK/1024
Divider A clkA
PREA DIVA
PWM_MR
Divider B clkB
PREB DIVB
PWM_MR
Caution: Before using the PWM macrocell, the programmer must first enable the PWM clock in
the Power Management Controller (PMC).
The PWM macrocell master clock, MCK, is divided in the clock generator module to provide dif-
ferent clocks available for all channels. Each channel can independently select one of the
divided clocks.
The clock generator is divided in three blocks:

– a modulo n counter which provides 11 clocks: FMCK, FMCK/2, FMCK/4, FMCK/8,
FMCK/16, FMCK/32, FMCK/64, FMCK/128, FMCK/256, FMCK/512, FMCK/1024
– two linear dividers (1, 1/2, 1/3, ... 1/255) that provide two separate clocks: clkA and
clkB
Each linear divider can independently divide one of the clocks of the modulo n counter. The
selection of the clock to be divided is made according to the PREA (PREB) field of the PWM
Mode register (PWM_MR). The resulting clock clkA (clkB) is the clock selected divided by DIVA
(DIVB) field value in the PWM Mode register (PWM_MR).
968 AT91SAM9G45
AT91SAM9G45
After a reset of the PWM controller, DIVA (DIVB) and PREA (PREB) in the PWM Mode register
are set to 0. This implies that after reset clkA (clkB) are turned off.
At reset, all clocks provided by the modulo n counter are turned off except clock “clk”. This situa-
tion is also true when the PWM master clock is turned off through the Power Management
Controller.
41.6.2 PWM Channel
41.6.2.1 Block Diagram
Figure 41-3. Functional View of the Channel Block Diagram

inputs
from clock
Channel
generator
Clock Internal PWMx

Comparator
Selector Counter output waveform
inputs from
APB bus
Each of the 4 channels is composed of three blocks:

• A clock selector which selects one of the clocks provided by the clock generator described in
Section 41.6.1 “PWM Clock Generator” on page 968.
• An internal counter clocked by the output of the clock selector. This internal counter is
incremented or decremented according to the channel configuration and comparators events.
The size of the internal counter is 16 bits.
• A comparator used to generate events according to the internal counter value. It also
computes the PWMx output waveform according to the configuration.
41.6.2.2 Waveform Properties

The different properties of output waveforms are:
• the internal clock selection. The internal channel counter is clocked by one of the clocks
provided by the clock generator described in the previous section. This channel parameter is
defined in the CPRE field of the PWM_CMRx register. This field is reset at 0.
• the waveform period. This channel parameter is defined in the CPRD field of the
PWM_CPRDx register.
- If the waveform is left aligned, then the output waveform period depends on the counter
source clock and can be calculated:
By using the Master Clock (MCK) divided by an X given prescaler value
(with X being 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, or 1024), the resulting period formula
will be:
(-------------------------------
X × CPRD )-
MCK
By using a Master Clock divided by one of both DIVA or DIVB divider, the formula becomes,
respectively:
969
(---------------------------------------------
X*CPRD*DIVA -) ( X*CPRD*DIVB )
or ----------------------------------------------
MCK MCK
If the waveform is center aligned then the output waveform period depends on the counter
source clock and can be calculated:
By using the Master Clock (MCK) divided by an X given prescaler value
(with X being 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, or 1024). The resulting period formula will
be:
(------------------------------------------
2 × X × CPRD )-
MCK
By using a Master Clock divided by one of both DIVA or DIVB divider, the formula becomes,
respectively:
(---------------------------------------------------
2*X*CPRD*DIVA -) ( 2*X*CPRD*DIVB )
or ----------------------------------------------------
MCK MCK
• the waveform duty cycle. This channel parameter is defined in the CDTY field of the
PWM_CDTYx register.
If the waveform is left aligned then:
duty cycle = (-----------------------------------------------------------------------------------------------------------

period – 1 ⁄ fchannel_x_clock × CDTY -)
period
If the waveform is center aligned, then:
duty cycle = (-----------------------------------------------------------------------------------------------------------------------------

( period ⁄ 2 ) – 1 ⁄ fchannel_x_clock × CDTY ) )-
( period ⁄ 2 )
• the waveform polarity. At the beginning of the period, the signal can be at high or low level.
This property is defined in the CPOL field of the PWM_CMRx register. By default the signal
starts by a low level.
• the waveform alignment. The output waveform can be left or center aligned. Center aligned
waveforms can be used to generate non overlapped waveforms. This property is defined in
the CALG field of the PWM_CMRx register. The default mode is left aligned.
Figure 41-4. Non Overlapped Center Aligned Waveforms

No overlap
PWM0
PWM1
Period
Note: 1. See Figure 41-5 on page 972 for a detailed description of center aligned waveforms.
970 AT91SAM9G45
AT91SAM9G45
When center aligned, the internal channel counter increases up to CPRD and.decreases down
to 0. This ends the period.
When left aligned, the internal channel counter increases up to CPRD and is reset. This ends
the period.
Thus, for the same CPRD value, the period for a center aligned channel is twice the period for a
left aligned channel.
Waveforms are fixed at 0 when:
• CDTY = CPRD and CPOL = 0
• CDTY = 0 and CPOL = 1
Waveforms are fixed at 1 (once the channel is enabled) when:
• CDTY = 0 and CPOL = 0
• CDTY = CPRD and CPOL = 1
The waveform polarity must be set before enabling the channel. This immediately affects the
channel output level. Changes on channel polarity are not taken into account while the channel
is enabled.
971
Figure 41-5. Waveform Properties
PWM_MCKx
CHIDx(PWM_SR)
CHIDx(PWM_ENA)
CHIDx(PWM_DIS)
Center Aligned
CALG(PWM_CMRx) = 1
PWM_CCNTx
CPRD(PWM_CPRDx)
CDTY(PWM_CDTYx)
Period
Output Waveform PWMx

CPOL(PWM_CMRx) = 0

CPOL(PWM_CMRx) = 1
CHIDx(PWM_ISR)
Left Aligned
CALG(PWM_CMRx) = 0
PWM_CCNTx
CPRD(PWM_CPRDx)
CDTY(PWM_CDTYx)
Period

CPOL(PWM_CMRx) = 0

CPOL(PWM_CMRx) = 1
CHIDx(PWM_ISR)
972 AT91SAM9G45
AT91SAM9G45
41.6.3 PWM Controller Operations
41.6.3.1 Initialization
Before enabling the output channel, this channel must have been configured by the software
application:
• Configuration of the clock generator if DIVA and DIVB are required
• Selection of the clock for each channel (CPRE field in the PWM_CMRx register)
• Configuration of the waveform alignment for each channel (CALG field in the PWM_CMRx
register)
• Configuration of the period for each channel (CPRD in the PWM_CPRDx register). Writing in
PWM_CPRDx Register is possible while the channel is disabled. After validation of the
channel, the user must use PWM_CUPDx Register to update PWM_CPRDx as explained
below.
• Configuration of the duty cycle for each channel (CDTY in the PWM_CDTYx register).
Writing in PWM_CDTYx Register is possible while the channel is disabled. After validation of
the channel, the user must use PWM_CUPDx Register to update PWM_CDTYx as explained
below.
• Configuration of the output waveform polarity for each channel (CPOL in the PWM_CMRx
register)
• Enable Interrupts (Writing CHIDx in the PWM_IER register)
• Enable the PWM channel (Writing CHIDx in the PWM_ENA register)
It is possible to synchronize different channels by enabling them at the same time by means of
writing simultaneously several CHIDx bits in the PWM_ENA register.
• In such a situation, all channels may have the same clock selector configuration and the
same period specified.
41.6.3.2 Source Clock Selection Criteria

The large number of source clocks can make selection difficult. The relationship between the
value in the Period Register (PWM_CPRDx) and the Duty Cycle Register (PWM_CDTYx) can
help the user in choosing. The event number written in the Period Register gives the PWM accu-
racy. The Duty Cycle quantum cannot be lower than 1/PWM_CPRDx value. The higher the value
of PWM_CPRDx, the greater the PWM accuracy.
For example, if the user sets 15 (in decimal) in PWM_CPRDx, the user is able to set a value
between 1 up to 14 in PWM_CDTYx Register. The resulting duty cycle quantum cannot be lower
than 1/15 of the PWM period.
41.6.3.3 Changing the Duty Cycle or the Period

It is possible to modulate the output waveform duty cycle or period.
To prevent unexpected output waveform, the user must use the update register (PWM_CUPDx)
to change waveform parameters while the channel is still enabled. The user can write a new
period value or duty cycle value in the update register (PWM_CUPDx). This register holds the
new value until the end of the current cycle and updates the value for the next cycle. Depending
on the CPD field in the PWM_CMRx register, PWM_CUPDx either updates PWM_CPRDx or
PWM_CDTYx. Note that even if the update register is used, the period must not be smaller than
the duty cycle.
973
Figure 41-6. Synchronized Period or Duty Cycle Update
User's Writing
PWM_CUPDx Value
1 0
PWM_CMRx. CPD
PWM_CPRDx PWM_CDTYx
End of Cycle
To prevent overwriting the PWM_CUPDx by software, the user can use status events in order to
synchronize his software. Two methods are possible. In both, the user must enable the dedi-
cated interrupt in PWM_IER at PWM Controller level.
The first method (polling method) consists of reading the relevant status bit in PWM_ISR Regis-
ter according to the enabled channel(s). See Figure 41-7.
The second method uses an Interrupt Service Routine associated with the PWM channel.
Note: Reading the PWM_ISR register automatically clears CHIDx flags.
Figure 41-7. Polling Method
PWM_ISR Read
Acknowledgement and clear previous register state
Writing in CPD field

Update of the Period or Duty Cycle
CHIDx = 1
YES
Writing in PWM_CUPDx
The last write has been taken into account
Note: Polarity and alignment can be modified only when the channel is disabled.
974 AT91SAM9G45
AT91SAM9G45
41.6.3.4 Interrupts
Depending on the interrupt mask in the PWM_IMR register, an interrupt is generated at the end
of the corresponding channel period. The interrupt remains active until a read operation in the
PWM_ISR register occurs.
A channel interrupt is enabled by setting the corresponding bit in the PWM_IER register. A chan-
nel interrupt is disabled by setting the corresponding bit in the PWM_IDR register.
975
41.7 Pulse Width Modulation Controller (PWM) User Interface
Table 41-4. Register Mapping()

0x00 PWM Mode Register PWM_MR Read-write 0
0x04 PWM Enable Register PWM_ENA Write-only -
0x08 PWM Disable Register PWM_DIS Write-only -
0x0C PWM Status Register PWM_SR Read-only 0
0x10 PWM Interrupt Enable Register PWM_IER Write-only -
0x14 PWM Interrupt Disable Register PWM_IDR Write-only -
0x18 PWM Interrupt Mask Register PWM_IMR Read-only 0
0x1C PWM Interrupt Status Register PWM_ISR Read-only 0
0x20 - 0xFC Reserved – – –
0x100 - 0x1FC Reserved
0x200 + ch_num * 0x20 + 0x00 PWM Channel Mode Register PWM_CMR Read-write 0x0
0x200 + ch_num * 0x20 + 0x04 PWM Channel Duty Cycle Register PWM_CDTY Read-write 0x0
0x200 + ch_num * 0x20 + 0x08 PWM Channel Period Register PWM_CPRD Read-write 0x0
0x200 + ch_num * 0x20 + 0x0C PWM Channel Counter Register PWM_CCNT Read-only 0x0
0x200 + ch_num * 0x20 + 0x10 PWM Channel Update Register PWM_CUPD Write-only -
2. Some registers are indexed with “ch_num” index ranging from 0 to 3.
976 AT91SAM9G45
AT91SAM9G45
41.7.1 PWM Mode Register

Register Name: PWM_MR
Address: 0xFFFB8000
31 30 29 28 27 26 25 24
– – – – PREB
23 22 21 20 19 18 17 16
DIVB
15 14 13 12 11 10 9 8
– – – – PREA
7 6 5 4 3 2 1 0
DIVA
• DIVA, DIVB: CLKA, CLKB Divide Factor
DIVA, DIVB CLKA, CLKB

0 CLKA, CLKB clock is turned off
1 CLKA, CLKB clock is clock selected by PREA, PREB
2-255 CLKA, CLKB clock is clock selected by PREA, PREB divided by DIVA, DIVB factor.
• PREA, PREB
PREA, PREB Divider Input Clock

0 0 0 0 MCK.
0 0 0 1 MCK/2
0 0 1 0 MCK/4
0 0 1 1 MCK/8
0 1 0 0 MCK/16
0 1 0 1 MCK/32
0 1 1 0 MCK/64
0 1 1 1 MCK/128
1 0 0 0 MCK/256
1 0 0 1 MCK/512
1 0 1 0 MCK/1024
Other Reserved
977
41.7.2 PWM Enable Register
Register Name: PWM_ENA
Address: 0xFFFB8004
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – CHID3 CHID2 CHID1 CHID0
• CHIDx: Channel ID
0 = No effect.
1 = Enable PWM output for channel x.
41.7.3 PWM Disable Register

Register Name: PWM_DIS
Address: 0xFFFB8008
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
0 = No effect.
1 = Disable PWM output for channel x.
978 AT91SAM9G45
AT91SAM9G45
41.7.4 PWM Status Register

Register Name: PWM_SR
Address: 0xFFFB800C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
0 = PWM output for channel x is disabled.
1 = PWM output for channel x is enabled.
979
41.7.5 PWM Interrupt Enable Register
Register Name: PWM_IER
Address: 0xFFFB8010
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
• CHIDx: Channel ID.

0 = No effect.
1 = Enable interrupt for PWM channel x.
980 AT91SAM9G45
AT91SAM9G45
41.7.6 PWM Interrupt Disable Register

Register Name: PWM_IDR
Address: 0xFFFB8014
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0

0 = No effect.
1 = Disable interrupt for PWM channel x.
981
41.7.7 PWM Interrupt Mask Register
Register Name: PWM_IMR
Address: 0xFFFB8018
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0

0 = Interrupt for PWM channel x is disabled.
1 = Interrupt for PWM channel x is enabled.
982 AT91SAM9G45
AT91SAM9G45
41.7.8 PWM Interrupt Status Register

Register Name: PWM_ISR
Address: 0xFFFB801C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
0 = No new channel period has been achieved since the last read of the PWM_ISR register.
1 = At least one new channel period has been achieved since the last read of the PWM_ISR register.
Note: Reading PWM_ISR automatically clears CHIDx flags.
983
41.7.9 PWM Channel Mode Register
Register Name: PWM_CMR[0..3]
Addresses: 0xFFFB8200 [0], 0xFFFB8220 [1], 0xFFFB8240 [2], 0xFFFB8260 [3]
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – CPD CPOL CALG
7 6 5 4 3 2 1 0
– – – – CPRE
• CPRE: Channel Pre-scaler
CPRE Channel Pre-scaler

0 0 0 0 MCK
0 0 0 1 MCK/2
0 0 1 0 MCK/4
0 0 1 1 MCK/8
0 1 0 0 MCK/16
0 1 0 1 MCK/32
0 1 1 0 MCK/64
0 1 1 1 MCK/128
1 0 0 0 MCK/256
1 0 0 1 MCK/512
1 0 1 0 MCK/1024
1 0 1 1 CLKA
1 1 0 0 CLKB
Other Reserved
• CALG: Channel Alignment

0 = The period is left aligned.
1 = The period is center aligned.
• CPOL: Channel Polarity

0 = The output waveform starts at a low level.
1 = The output waveform starts at a high level.
• CPD: Channel Update Period

0 = Writing to the PWM_CUPDx will modify the duty cycle at the next period start event.
1 = Writing to the PWM_CUPDx will modify the period at the next period start event.
984 AT91SAM9G45
AT91SAM9G45
41.7.10 PWM Channel Duty Cycle Register

Register Name: PWM_CDTY[0..3]
31 30 29 28 27 26 25 24
CDTY
23 22 21 20 19 18 17 16
CDTY
15 14 13 12 11 10 9 8
CDTY
7 6 5 4 3 2 1 0
CDTY
Only the first 16 bits (internal channel counter size) are significant.
• CDTY: Channel Duty Cycle

Defines the waveform duty cycle. This value must be defined between 0 and CPRD (PWM_CPRx).
985
41.7.11 PWM Channel Period Register
Register Name: PWM_CPRD[0..3]
31 30 29 28 27 26 25 24
CPRD
23 22 21 20 19 18 17 16
CPRD
15 14 13 12 11 10 9 8
CPRD
7 6 5 4 3 2 1 0
CPRD
• CPRD: Channel Period

If the waveform is left-aligned, then the output waveform period depends on the counter source clock and can be
calculated:
– By using the Master Clock (MCK) divided by an X given prescaler value (with X being 1, 2, 4, 8, 16, 32, 64, 128,
256, 512, or 1024). The resulting period formula will be:
(-------------------------------
X × CPRD )-
MCK
– By using a Master Clock divided by one of both DIVA or DIVB divider, the formula becomes, respectively:
(------------------------------------------
CRPD × DIVA )- ( CRPD × DIVAB )
or -----------------------------------------------
MCK MCK
If the waveform is center-aligned, then the output waveform period depends on the counter source clock and can be
calculated:
– By using the Master Clock (MCK) divided by an X given prescaler value (with X being 1, 2, 4, 8, 16, 32, 64, 128,
256, 512, or 1024). The resulting period formula will be:
(------------------------------------------
2 × X × CPRD )-
MCK
– By using a Master Clock divided by one of both DIVA or DIVB divider, the formula becomes, respectively:
(-----------------------------------------------------
2 × CPRD × DIVA )- ( 2 × CPRD × DIVB )
or ------------------------------------------------------
MCK MCK
986 AT91SAM9G45
AT91SAM9G45
41.7.12 PWM Channel Counter Register

Register Name: PWM_CCNT[0..3]
Addresses: 0xFFFB820C [0], 0xFFFB822C [1], 0xFFFB824C [2], 0xFFFB826C [3]
31 30 29 28 27 26 25 24
CNT
23 22 21 20 19 18 17 16
CNT
15 14 13 12 11 10 9 8
CNT
7 6 5 4 3 2 1 0
CNT
• CNT: Channel Counter Register

Internal counter value. This register is reset when:
• the channel is enabled (writing CHIDx in the PWM_ENA register).
• the counter reaches CPRD value defined in the PWM_CPRDx register if the waveform is left aligned.
987
41.7.13 PWM Channel Update Register
Register Name: PWM_CUPD[0..3]
31 30 29 28 27 26 25 24
CUPD
23 22 21 20 19 18 17 16
CUPD
15 14 13 12 11 10 9 8
CUPD
7 6 5 4 3 2 1 0
CUPD
This register acts as a double buffer for the period or the duty cycle. This prevents an unexpected waveform when modify-
ing the waveform period or duty-cycle.
CPD (PWM_CMRx Register)

The duty-cycle (CDTY in the PWM_CDTYx register) is updated with the CUPD value at the
0
beginning of the next period.
The period (CPRD in the PWM_CPRDx register) is updated with the CUPD value at the beginning
1
of the next period.
988 AT91SAM9G45
AT91SAM9G45
42. AC97 Controller (AC97C)
42.1 Description
The AC97 Controller is the hardware implementation of the AC97 digital controller (DC’97) com-
pliant with AC97 Component Specification 2.2. The AC97 Controller communicates with an
audio codec (AC97) or a modem codec (MC’97) via the AC-link digital serial interface. All digital
audio, modem and handset data streams, as well as control (command/status) informations are
transferred in accordance to the AC-link protocol.
The AC97 Controller features a Peripheral DMA Controller (PDC) for audio streaming transfers.
It also supports variable sampling rate and four Pulse Code Modulation (PCM) sample resolu-
tions of 10, 16, 18 and 20 bits.

• Compatible with AC97 Component Specification V2.2
• Capable to Interface with a Single Analog Front end
• Three independent RX Channels and three independent TX Channels
– One RX and one TX channel dedicated to the AC97 Analog Front end control
– One RX and one TX channel for data transfers, associated with a PDC
– One RX and one TX channel for data transfers with no PDC
• Time Slot Assigner allowing to assign up to 12 time slots to a channel
• Channels support mono or stereo up to 20 bit sample length
– Variable sampling rate AC97 Codec Interface (48KHz and below)
989
42.3 Block Diagram
Figure 42-1. Functional Block Diagram
MCK Clock Domain

Slot Number SYNC
AC97 Slot Controller
Slot Number
16/20 bits
Slot #0
AC97 Tag Controller M
Slot #0,1 U
AC97 CODEC Channel SDATA_OUT
X
Slot #1,2 Transmit Shift Register
AC97C_COTHR
AC97C_CORHR Receive Shift Register

Slot #2
SDATA_IN
AC97 Channel A

D
AC97C_CATHR
Slot #3...12 E
AC97C_CARHR Receive Shift Register
BITCLK
AC97C Interrupt
M
AC97 Channel B

U
AC97C_CBTHR
Slot #3...12
AC97C_CBRHR
X
MCK
User Interface
Bit Clock Domain
APB Interface
990 AT91SAM9G45
42.4 Pin Name List
AC97CK 12.288-MHz bit-rate clock Input
AC97RX Receiver Data (Referred as SDATA_IN in AC-link spec) Input
AC97FS 48-KHz frame indicator and synchronizer Output
AC97TX Transmitter Data (Referred as SDATA_OUT in AC-link spec) Output
The AC97 reset signal provided to the primary codec can be generated by a PIO.
Figure 42-2. Application Block diagram
AC 97 Controller AC-link AC'97 Primary Codec
AC97_RESET
PIOx
AC97_SYNC
AC97FS
AC97_BITCLK
AC97CK
AC97_SDATA_OUT
AC97TX
AC97_SDATA_IN
AC97RX
991 AT91SAM9G45
42.6.1 I/O Lines

Before using the AC97 Controller receiver, the PIO controller must be configured in order for the
AC97C receiver I/O lines to be in AC97 Controller peripheral mode.
Before using the AC97 Controller transmitter, the PIO controller must be configured in order for
the AC97C transmitter I/O lines to be in AC97 Controller peripheral mode.

AC97C AC97CK PD9 A
AC97C AC97FS PD8 A
AC97C AC97RX PD6 A
AC97C AC97TX PD7 A

The AC97 Controller is not continuously clocked. Its interface may be clocked through the Power
Management Controller (PMC), therefore the programmer must first configure the PMC to
enable the AC97 Controller clock.
The AC97 Controller has two clock domains. The first one is supplied by PMC and is equal to
MCK. The second one is AC97CK which is sent by the AC97 Codec (Bit clock).
Signals that cross the two clock domains are re-synchronized. MCK clock frequency must be
higher than the AC97CK (Bit Clock) clock frequency.
42.6.3 Interrupt
The AC97 Controller interface has an interrupt line connected to the Advanced Interrupt Control-
ler (AIC). Handling interrupts requires programming the AIC before configuring the AC97C.
All AC97 Controller interrupts can be enabled/disabled by writing to the AC97 Controller Inter-
rupt Enable/Disable Registers. Each pending and unmasked AC97 Controller interrupt will
assert the interrupt line. The AC97 Controller interrupt service routine can get the interrupt
source in two steps:
• Reading and ANDing AC97 Controller Interrupt Mask Register (AC97C_IMR) and AC97
Controller Status Register (AC97C_SR).
• Reading AC97 Controller Channel x Status Register (AC97C_CxSR).

Instance ID
AC97C 24
992 AT91SAM9G45
42.7.1 Protocol overview

AC-link protocol is a bidirectional, fixed clock rate, serial digital stream. AC-link handles multiple
input and output Pulse Code Modulation PCM audio streams, as well as control register
accesses employing a Time Division Multiplexed (TDM) scheme that divides each audio frame
in 12 outgoing and 12 incoming 20-bit wide data slots.
Figure 42-3. Bidirectional AC-link Frame with Slot Assignment

Slot # 0 1 2 3 4 5 6 7 8 9 10 11 12
AC97FS
CMD CMD PCM PCM LINE 1 PCM PCM PCM PCM LINE 2 HSET IO
AC97TX TAG
ADDR DATA L Front R Front DAC Center L SURR R SURR LFE DAC DAC CTRL
(Controller Output)
STATUS STATUS PCM PCM LINE 1 PCM LINE 2 HSET IO

AC97RX TAG RSVED RSVED RSVED
ADDR DATA LEFT RIGHT DAC MIC ADC ADC STATUS
(Codec output)
Table 42-4. AC-link Output Slots Transmitted from the AC97C Controller
Slot # Pin Description
0 TAG
1 Command Address Port
2 Command Data Port
3,4 PCM playback Left/Right Channel
5 Modem Line 1 Output Channel
6, 7, 8 PCM Center/Left Surround/Right Surround
9 PCM LFE DAC
10 Modem Line 2 Output Channel
11 Modem Handset Output Channel
12 Modem GPIO Control Channel
Table 42-5. AC-link Input Slots Transmitted from the AC97C Controller
Slot # Pin Description
0 TAG
1 Status Address Port
2 Status Data Port
3,4 PCM playback Left/Right Channel
5 Modem Line 1 ADC
6 Dedicated Microphone ADC
7, 8, 9 Vendor Reserved
10 Modem Line 2 ADC
11 Modem Handset Input ADC
12 Modem IO Status
993 AT91SAM9G45
42.7.2 Slot Description
42.7.2.1 Tag Slot

The tag slot, or slot 0, is a 16-bit wide slot that always goes at the beginning of an outgoing or
incoming frame. Within tag slot, the first bit is a global bit that flags the entire frame validity. The
next 12 bit positions sampled by the AC97 Controller indicate which of the corresponding 12
time slots contain valid data. The slot’s last two bits (combined) called Codec ID, are used to dis-
tinguish primary and secondary codec.
The 16-bit wide tag slot of the output frame is automatically generated by the AC97 Controller
according to the transmit request of each channel and to the SLOTREQ from the previous input
frame, sent by the AC97 Codec, in Variable Sample Rate mode.
42.7.2.2 Codec Slot 1

The command/status slot is a 20-bit wide slot used to control features, and monitors status for
AC97 Codec functions.
The control interface architecture supports up to sixty-four 16-bit wide read-write registers. Only
the even registers are currently defined and addressed.
Slot 1’s bitmap is the following:
• Bit 19 is for read-write command, 1= read, 0 = write.
• Bits [18:12] are for control register index.
• Bits [11:0] are reserved.
42.7.2.3 Codec Slot 2

Slot 2 is a 20-bit wide slot used to carry 16-bit wide AC97 Codec control register data. If the cur-
rent command port operation is a read, the entire slot time is stuffed with zeros. Its bitmap is the
following:
• Bits [19:4] are the control register data
• Bits [3:0] are reserved and stuffed with zeros.
42.7.2.4 Data Slots [3:12]

Slots [3:12] are 20-bit wide data slots, they usually carry audio PCM and/or modem I/O data.
994 AT91SAM9G45
42.7.3 AC97 Controller Channel Organization
The AC97 Controller features a Codec channel and 2 logical channels: Channel A, Channel B.
The Codec channel controls AC97 Codec registers, it enables write and read configuration val-
ues in order to bring the AC97 Codec to an operating state. The Codec channel always runs slot
1 and slot 2 exclusively, in both input and output directions.
Channel A, Channel B transfer data to/from AC97 codec. All audio samples and modem data
must transit by these 2 channels. However, Channels A and B are connected to PDC channels
thus making it suitable for audio streaming applications.
Each slot of the input or the output frame that belongs to this range [3 to 12] can be operated by
Channel A or Channel B . The slot to channel assignment is configured by two registers:
• AC97 Controller Input Channel Assignment Register (AC97C_ICA)
• AC97 Controller Output Channel Assignment Register (AC97C_OCA)
The AC97 Controller Input Channel Assignment Register (AC97C_ICA) configures the input slot
to channel assignment. The AC97 Controller Output Channel Assignment Register
(AC97C_OCA) configures the output slot to channel assignment.
A slot can be left unassigned to a channel by the AC97 Controller. Slots 0, 1,and 2 cannot be
assigned to Channel A,or to Channel Bthrough the AC97C_OCA and AC97C_ICA Registers.
The width of sample data, that transit via the Channel varies and can take one of these values;
10, 16, 18 or 20 bits.
Figure 42-4. Logical Channel Assignment

Slot # 0 1 2 3 4 5 6 7 8 9 10 11 12
AC97FS
AC97TX CMD CMD PCM PCM LINE 1 PCM PCM PCM PCM LINE 2 HSET IO
TAG
(Controller Output) ADDR DATA L Front R Front DAC Center L SURR R SURR LFE DAC DAC CTRL
Codec Channel Channel A
AC97C_OCA = 0x0000_0209
AC97RX STATUS STATUS PCM PCM LINE 1 PCM LINE 2 HSET IO

TAG RSVED RSVED RSVED
(Codec output) ADDR DATA LEFT RIGHT DAC MIC ADC ADC STATUS
Codec Channel Channel A
AC97C_ICA = 0x0000_0009
995 AT91SAM9G45
42.7.3.1 AC97 Controller Setup
The following operations must be performed in order to bring the AC97 Controller into an operat-
ing state:
1. Enable the AC97 Controller clock in the PMC controller.
2. Turn on AC97 function by enabling the ENA bit in AC97 Controller Mode Register
(AC97C_MR).
3. Configure the input channel assignment by controlling the AC97 Controller Input
Assignment Register (AC97C_ICA).
4. Configure the output channel assignment by controlling the AC97 Controller Input
Assignment Register (AC97C_OCA).
5. Configure sample width for Channel A, Channel Bby writing the SIZE bit field in AC97C
Channel x Mode Register (AC97C_CAMR) , (AC97C_CBMR). The application can
write 10, 16, 18,or 20-bit wide PCM samples through the AC97 interface and they will
be transferred into 20-bit wide slots.
6. Configure data Endianness for Channel A, Channel B by writing CEM bit field in
(AC97C_CAMR) , (AC97C_CBMR) register. Data on the AC-link are shifted MSB first.
The application can write little- or big-endian data to the AC97 Controller interface.
7. Configure the PIO controller to drive the RESET signal of the external Codec. The
RESET signal must fulfill external AC97 Codec timing requirements.
8. Enable Channel A and/or Channel B by writing CEN bit field in AC97C_CxMR register.
42.7.3.2 Transmit Operation

The application must perform the following steps in order to send data via a channel to the AC97
Codec:
• Check if previous data has been sent by polling TXRDY flag in the AC97C Channel x Status
Register (AC97_CxSR). x being one of the 2 channels.
• Write data to the AC97 Controller Channel x Transmit Holding Register (AC97C_CxTHR).
Once data has been transferred to the Channel x Shift Register, the TXRDY flag is automatically
set by the AC97 Controller which allows the application to start a new write action. The applica-
tion can also wait for an interrupt notice associated with TXRDY in order to send data. The
interrupt remains active until TXRDY flag is cleared.
996 AT91SAM9G45
Figure 42-5. Audio Transfer (PCM L Front, PCM R Front) on Channel x
Slot # 0 1 2 3 4 5 6 7 8 9 10 11 12
AC97FS
AC97TX CMD CMD PCM PCM LINE 1 PCM PCM PCM PCM LINE 2 HSET IO
TAG
(Controller Output) ADDR DATA L Front R Front DAC Center L SURR R SURR LFE DAC DAC CTRL
TXRDYCx
(AC97C_SR)
TXEMPTY
(AC97C_SR)
Write access to
AC97C_THRx
PCM L Front
transfered to the shift register
PCM R Front
transfered to the shift register
The TXEMPTY flag in the AC97 Controller Channel x Status Register (AC97C_CxSR) is set
when all requested transmissions for a channel have been shifted on the AC-link. The applica-
tion can either poll TXEMPTY flag in AC97C_CxSR or wait for an interrupt notice associated
with the same flag.
In most cases, the AC97 Controller is embedded in chips that target audio player devices. In
such cases, the AC97 Controller is exposed to heavy audio transfers. Using the polling tech-
nique increases processor overhead and may fail to keep the required pace under an operating
system. In order to avoid these polling drawbacks, the application can perform audio streams by
using PDC connected to channel A, which reduces processor overhead and increases perfor-
mance especially under an operating system.
The PDC transmit counter values must be equal to the number of PCM samples to be transmit-
ted, each sample goes in one slot.
42.7.3.3 AC97 Output Frame

The AC97 Controller outputs a thirteen-slot frame on the AC-Link. The first slot (tag slot or slot 0)
flags the validity of the entire frame and the validity of each slot; whether a slot carries valid data
or not. Slots 1 and 2 are used if the application performs control and status monitoring actions
on AC97 Codec control/status registers. Slots [3:12] are used according to the content of the
AC97 Controller Output Channel Assignment Register (AC97C_OCA). If the application per-
forms many transmit requests on a channel, some of the slots associated to this channel or all of
them will carry valid data.
42.7.3.4 Receive Operation

The AC97 Controller can also receive data from AC97 Codec. Data is received in the channel’s
shift register and then transferred to the AC97 Controller Channel x Read Holding Register. To
read the newly received data, the application must perform the following steps:
• Poll RXRDY flag in AC97 Controller Channel x Status Register (AC97C_CxSR). x being one
of the 2 channels.
• Read data from AC97 Controller Channel x Read Holding Register.
997 AT91SAM9G45
The application can also wait for an interrupt notice in order to read data from AC97C_CxRHR.
The interrupt remains active until RXRDY is cleared by reading AC97C_CxSR.
The RXRDY flag in AC97C_CxSR is set automatically when data is received in the Channel x
shift register. Data is then shifted to AC97C_CxRHR.
Figure 42-6. Audio Transfer (PCM L Front, PCM R Front) on Channel x
Slot # 0 1 2 3 4 5 6 7 8 9 10 11 12
AC97FS
STATUS STATUS PCM PCM LINE 1 PCM LINE 2 HSET IO

AC97RX TAG RSVED RSVED RSVED
ADDR DATA LEFT RIGHT DAC MIC ADC ADC STATUS
(Codec output)
RXRDYCx
(AC97C_SR)
Read access to
AC97C_RHRx
If the previously received data has not been read by the application, the new data overwrites the
data already waiting in AC97C_CxRHR, therefore the OVRUN flag in AC97C_CxSR is raised.
The application can either poll the OVRUN flag in AC97C_CxSR or wait for an interrupt notice.
The interrupt remains active until the OVRUN flag in AC97C_CxSR is set.
The AC97 Controller can also be used in sound recording devices in association with an AC97
Codec. The AC97 Controller may also be exposed to heavy PCM transfers. The application can
usethe PDC connected to channel A in order to reduce processor overhead and increase perfor-
mance especially under an operating system.
The PDC receive counter values must be equal to the number of PCM samples to be received,
each sample goes in one slot.
42.7.3.5 AC97 Input Frame

The AC97 Controller receives a thirteen slot frame on the AC-Link sent by the AC97 Codec. The
first slot (tag slot or slot 0) flags the validity of the entire frame and the validity of each slot;
whether a slot carries valid data or not. Slots 1 and 2 are used if the application requires status
informations from AC97 Codec. Slots [3:12] are used according to AC97 Controller Output
Channel Assignment Register (AC97C_ICA) content. The AC97 Controller will not receive any
data from any slot if AC97C_ICA is not assigned to a channel in input.
42.7.3.6 Configuring and Using Interrupts

Instead of polling flags in AC97 Controller Global Status Register (AC97C_SR) and in AC97
Controller Channel x Status Register (AC97C_CxSR), the application can wait for an interrupt
notice. The following steps show how to configure and use interrupts correctly:
• Set the interruptible flag in AC97 Controller Channel x Mode Register (AC97C_CxMR).
• Set the interruptible event and channel event in AC97 Controller Interrupt Enable Register
(AC97C_IER).
The interrupt handler must read both AC97 Controller Global Status Register (AC97C_SR) and
AC97 Controller Interrupt Mask Register (AC97C_IMR) and AND them to get the real interrupt
source. Furthermore, to get which event was activated, the interrupt handler has to read AC97
Controller Channel x Status Register (AC97C_CxSR), x being the channel whose event triggers
the interrupt.
998 AT91SAM9G45
The application can disable event interrupts by writing in AC97 Controller Interrupt Disable Reg-
ister (AC97C_IDR). The AC97 Controller Interrupt Mask Register (AC97C_IMR) shows which
event can trigger an interrupt and which one cannot.
42.7.3.7 Endianness
Endianness can be managed automatically for each channel, except for the Codec channel, by
writing to Channel Endianness Mode (CEM) in AC97C_CxMR. This enables transferring data on
AC-link in Big Endian format without any additional operation.
42.7.3.8 To Transmit a Word Stored in Big Endian Format on AC-link

Word to be written in AC97 Controller Channel x Transmit Holding Register (AC97C_CxTHR)
(as it is stored in memory or microprocessor register).
31 24 23 16 15 8 7 0
Byte0[7:0] Byte1[7:0] Byte2[7:0] Byte3[7:0]
Word stored in Channel x Transmit Holding Register (AC97C_CxTHR) (data to transmit).
31 24 23 20 19 16 15 8 7 0
– – Byte2[3:0] Byte1[7:0] Byte0[7:0]
Data transmitted on appropriate slot: data[19:0] = {Byte2[3:0], Byte1[7:0], Byte0[7:0]}.
42.7.3.9 To Transmit A Halfword Stored in Big Indian Format on AC-link

Halfw ord to be written in AC97 Controlle r Channel x Transmit Holding Register
(AC97C_CxTHR).
31 24 23 16 15 8 7 0
– – Byte0[7:0] Byte1[7:0]
Halfword stored in AC97 Controller Channel x Transmit Holding Register (AC97C_CxTHR) (data
to transmit).
31 24 23 16 15 8 7 0
– – Byte1[7:0] Byte0[7:0]
Data emitted on related slot: data[19:0] = {0x0, Byte1[7:0], Byte0[7:0]}.
42.7.3.10 To Transmit a10-bit Sample Stored in Big Endian Format on AC-link

Halfw ord to be written in AC97 Controlle r Channel x Transmit Holding Register
(AC97C_CxTHR).
31 24 23 16 15 8 7 0
– – Byte0[7:0] {0x00, Byte1[1:0]}
Halfword stored in AC97 Controller Channel x Transmit Holding Register (AC97C_CxTHR) (data
to transmit).
31 24 23 16 15 10 9 8 7 0
Byte1
– – – Byte0[7:0]
[1:0]
Data emitted on related slot: data[19:0] = {0x000, Byte1[1:0], Byte0[7:0]}.
999 AT91SAM9G45
42.7.3.11 To Receive Word transfers
Data received on appropriate slot: data[19:0] = {Byte2[3:0], Byte1[7:0], Byte0[7:0]}.
Word stored in AC97 Controller Channel x Receive Holding Register (AC97C_CxRHR)
(Received Data).
31 24 23 20 19 16 15 8 7 0
– – Byte2[3:0] Byte1[7:0] Byte0[7:0]
Data is read from AC97 Controller Channel x Receive Holding Register (AC97C_CxRHR) when
Channel x data size is greater than 16 bits and when big-endian mode is enabled (data written to
memory).
31 24 23 16 15 8 7 0
Byte0[7:0] Byte1[7:0] {0x0, Byte2[3:0]} 0x00
42.7.3.12 To Receive Halfword Transfers

Data received on appropriate slot: data[19:0] = {0x0, Byte1[7:0], Byte0[7:0]}.
Halfword stored in AC97 Controller Channel x Receive Holding Register (AC97C_CxRHR)
(Received Data).
31 24 23 16 15 8 7 0
– – Byte1[7:0] Byte0[7:0]
Data is read from AC97 Controller Channel x Receive Holding Register (AC97C_CxRHR) when
data size is equal to 16 bits and when big-endian mode is enabled.
31 24 23 16 15 8 7 0
– – Byte0[7:0] Byte1[7:0]
42.7.3.13 To Receive 10-bit Samples

Data received on appropriate slot: data[19:0] = {0x000, Byte1[1:0], Byte0[7:0]}.Halfword stored
in AC97 Controller Channel x Receive Holding Register (AC97C_CxRHR) (Received Data)
31 24 23 16 15 10 9 8 7 0
Byte1
– – – Byte0[7:0]
[1:0]
Data read from AC97 Controller Channel x Receive Holding Register (AC97C_CxRHR) when
data size is equal to 10 bits and when big-endian mode is enabled.
31 24 23 16 15 8 7 3 1 0
Byte1
– – Byte0[7:0] 0x00
[1:0]
1000 AT91SAM9G45
42.7.4 Variable Sample Rate
The problem of variable sample rate can be summarized by a simple example. When passing a
44.1 kHz stream across the AC-link, for every 480 audio output frames that are sent across, 441
of them must contain valid sample data. The new AC97 standard approach calls for the addition
of “on-demand” slot request flags. The AC97 Codec examines its sample rate control register,
the state of its FIFOs, and the incoming SDATA_OUT tag bits (slot 0) of each output frame and
then determines which SLOTREQ bits to set active (low). These bits are passed from the AC97
Codec to the AC97 Controller in slot 1/SLOTREQ in every audio input frame. Each time the
AC97 controller sees one or more of the newly defined slot request flags set active (low) in a
given audio input frame, it must pass along the next PCM sample for the corresponding slot(s) in
the AC-link output frame that immediately follows.
The variable Sample Rate mode is enabled by performing the following steps:
• Setting the VRA bit in the AC97 Controller Mode Register (AC97C_MR).
• Enable Variable Rate mode in the AC97 Codec by performing a transfer on the Codec
channel.
Slot 1 of the input frame is automatically interpreted as SLOTREQ signaling bits. The AC97 Con-
troller will automatically fill the active slots according to both SLOTREQ and AC97C_OCA
register in the next transmitted frame.
42.7.5.1 Powering Down the AC-Link

The AC97 Codecs can be placed in low power mode. The application can bring AC97 Codec to
a power down state by performing sequential writes to AC97 Codec powerdown register . Both
the bit clock (clock delivered by AC97 Codec, AC97CK) and the input line (AC97RX) are held at
a logic low voltage level. This puts AC97 Codec in power down state while all its registers are
still holding current values. Without the bit clock, the AC-link is completely in a power down
state.
The AC97 Controller should not attempt to play or capture audio data until it has awakened
AC97 Codec.
To set the AC97 Codec in low power mode, the PR4 bit in the AC97 Codec powerdown register
(Codec address 0x26) must be set to 1. Then the primary Codec drives both AC97CK and
AC97RX to a low logic voltage level.
The following operations must be done to put AC97 Codec in low power mode:
• Disable Channel A clearing CEN field in the AC97C_CAMR register.
• Disable Channel B clearing CEN field in the AC97C_CBMR register.
• Write 0x2680 value in the AC97C_COTHR register.
• Poll the TXEMPTY flag in AC97C_CxSR registers for the 2 channels.
At this point AC97 Codec is in low power mode.
42.7.5.2 Waking up the AC-link

There are two methods to bring the AC-link out of low power mode. Regardless of the method, it
is always the AC97 Controller that performs the wake-up.
42.7.5.3 Wake-up Tiggered by the AC97 Controller

The AC97 Controller can wake up the AC97 Codec by issuing either a cold or a warm reset.
1001 AT91SAM9G45
The AC97 Controller can also wake up the AC97 Codec by asserting AC97FS signal, however
this action should not be performed for a minimum period of four audio frames following the
frame in which the powerdown was issued.
42.7.5.4 Wake-up Triggered by the AC97 Codec

This feature is implemented in AC97 modem codecs that need to report events such as Caller-
ID and wake-up on ring.
The AC97 Codec can drive AC97RX signal from low to high level and holding it high until the
controller issues either a cold or a worm reset. The AC97RX rising edge is asynchronously
(regarding AC97FS) detected by the AC97 Controller. If WKUP bit is enabled in AC97C_IMR
register, an interrupt is triggered that wakes up the AC97 Controller which should then immedi-
ately issue a cold or a warm reset.
If the processor needs to be awakened by an external event, the AC97RX signal must be exter-
nally connected to the WAKEUP entry of the system controller.
Figure 42-7. AC97 Power-Down/Up Sequence

Wake Event
Power Down Frame Sleep State Warm Reset New Audio Frame
AC97CK
AC97FS
AC97TX Write to Data

TAG TAG Slot1 Slot2
0x26 PR4
Write to Data
AC97RX TAG TAG Slot1 Slot2
0x26 PR4
42.7.5.5 AC97 Codec Reset

There are three ways to reset an AC97 Codec.
42.7.5.6 Cold AC97 Reset

A cold reset is generated by asserting the RESET signal low for the minimum specified time
(depending on the AC97 Codec) and then by de-asserting RESET high. AC97CK and AC97FS
is reactivated and all AC97 Codec registers are set to their default power-on values. Transfers
on AC-link can resume.
The RESET signal will be controlled via a PIO line. This is how an application should perform a
cold reset:
• Clear and set ENA flag in the AC97C_MR register to reset the AC97 Controller
• Clear PIO line output controlling the AC97 RESET signal
• Wait for the minimum specified time
• Set PIO line output controlling the AC97 RESET signal
AC97CK, the clock provided by AC97 Codec, is detected by the controller.
42.7.5.7 Warm AC97 Reset

A warm reset reactivates the AC-link without altering AC97 Codec registers. A warm reset is sig-
naled by driving AC97FX signal high for a minimum of 1us in the absence of AC97CK. In the
1002 AT91SAM9G45
absence of AC97CK, AC97FX is treated as an asynchronous (regarding AC97FX) input used to
signal a warm reset to AC97 Codec.
This is the right way to perform a warm reset:
• Set WRST in the AC97C_MR register.
• Wait for at least 1us
• Clear WRST in the AC97C_MR register.
The application can check that operations have resumed by checking SOF flag in the
AC97C_SR register or wait for an interrupt notice if SOF is enabled in AC97C_IMR.
1003 AT91SAM9G45
42.8 AC97 Controller (AC97C) User Interface
0x0-0x4 Reserved – – –
0x8 Mode Register AC97C_MR Read-write 0x0
0xC Reserved – – –
0x10 Input Channel Assignment Register AC97C_ICA Read-write 0x0
0x14 Output Channel Assignment Register AC97C_OCA Read-write 0x0
0x18-0x1C Reserved – – –
0x20 Channel A Receive Holding Register AC97C_CARHR Read 0x0
0x24 Channel A Transmit Holding Register AC97C_CATHR Write –
0x28 Channel A Status Register AC97C_CASR Read 0x0
0x2C Channel A Mode Register AC97C_CAMR Read-write 0x0
0x30 Channel B Receive Holding Register AC97C_CBRHR Read 0x0
0x34 Channel B Transmit Holding Register AC97C_CBTHR Write –
0x38 Channel B Status Register AC97C_CBSR Read 0x0
0x3C Channel B Mode Register AC97C_CBMR Read-write 0x0
0x40 Codec Channel Receive Holding Register AC97C_CORHR Read 0x0
0x44 Codec Channel Transmit Holding Register AC97C_COTHR Write –
0x48 Codec Status Register AC97C_COSR Read 0x0
0x4C Codec Mode Register AC97C_COMR Read-write 0x0
0x50 Status Register AC97C_SR Read 0x0
0x54 Interrupt Enable Register AC97C_IER Write –
0x58 Interrupt Disable Register AC97C_IDR Write –
0x5C Interrupt Mask Register AC97C_IMR Read 0x0
0x60-0xFB Reserved – – –
Reserved for Peripheral DMA Controller (PDC)
0x100-0x124 – – –
registers related to channel transfers
1004 AT91SAM9G45
42.8.1 AC97 Controller Mode Register
Register Name:AC97C_MR
Address: 0xFFFAC008
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – VRA WRST ENA
• VRA: Variable Rate (for Data Slots 3-12)

0: Variable Rate is inactive. (48 KHz only)
1: Variable Rate is active.
• WRST: Warm Reset

0: Warm Reset is inactive.
1: Warm Reset is active.
• ENA: AC97 Controller Global Enable

0: No effect. AC97 function as well as access to other AC97 Controller registers are disabled.
1: Activates the AC97 function.
1005 AT91SAM9G45
42.8.2 AC97 Controller Input Channel Assignment Register
Register Name:AC97C_ICA
Address: 0xFFFAC010
31 30 29 28 27 26 25 24
– – CHID12 CHID11
23 22 21 20 19 18 17 16
CHID10 CHID9 CHID8
15 14 13 12 11 10 9 8
CHID8 CHID7 CHID6 CHID5
7 6 5 4 3 2 1 0
CHID5 CHID4 CHID3
• CHIDx: Channel ID for the input slot x

CHIDx Selected Receive Channel
0x0 None. No data will be received during this slot time
0x1 Channel A data will be received during this slot time.
0x2 Channel B data will be received during this slot time
1006 AT91SAM9G45
42.8.3 AC97 Controller Output Channel Assignment Register
Register Name:AC97C_OCA
Address: 0xFFFAC014
31 30 29 28 27 26 25 24
– – CHID12 CHID11
23 22 21 20 19 18 17 16
CHID10 CHID9 CHID8
15 14 13 12 11 10 9 8
CHID8 CHID7 CHID6 CHID5
7 6 5 4 3 2 1 0
CHID5 CHID4 CHID3
• CHIDx: Channel ID for the output slot x

CHIDx Selected Transmit Channel
0x0 None. No data will be transmitted during this slot time
0x1 Channel A data will be transferred during this slot time.
0x2 Channel B data will be transferred during this slot time
1007 AT91SAM9G45
42.8.4 AC97 Controller Codec Channel Receive Holding Register
Register Name:AC97C_CORHR
Address: 0xFFFAC040
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
SDATA
7 6 5 4 3 2 1 0
SDATA
• SDATA: Status Data

Data sent by the CODEC in the third AC97 input frame slot (Slot 2).
1008 AT91SAM9G45
42.8.5 AC97 Controller Codec Channel Transmit Holding Register
Register Name:AC97C_COTHR
Address: 0xFFFAC044
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
READ CADDR
15 14 13 12 11 10 9 8
CDATA
7 6 5 4 3 2 1 0
CDATA
• READ: Read-write command

0: Write operation to the CODEC register indexed by the CADDR address.
1: Read operation to the CODEC register indexed by the CADDR address.
This flag is sent during the second AC97 frame slot
• CADDR: CODEC control register index

Data sent to the CODEC in the second AC97 frame slot.
• CDATA: Command Data

Data sent to the CODEC in the third AC97 frame slot (Slot 2).
1009 AT91SAM9G45
42.8.6 AC97 Controller Channel A, Channel B, Receive Holding Register
Register Name:AC97C_CARHR, AC97C_CBRHR
Address: 0xFFFAC020
Address: 0xFFFAC030
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – RDATA
15 14 13 12 11 10 9 8
RDATA
7 6 5 4 3 2 1 0
RDATA
• RDATA: Receive Data

Received Data on channel x.
42.8.7 AC97 Controller Channel A, Channel B, Transmit Holding Register

Register Name:AC97C_CATHR, AC97C_CBTHR
Address: 0xFFFAC024
Address: 0xFFFAC034
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – TDATA
15 14 13 12 11 10 9 8
TDATA
7 6 5 4 3 2 1 0
TDATA
• TDATA: Transmit Data

Data to be sent on channel x.
1010 AT91SAM9G45
42.8.8 AC97 Controller Channel A Status Register
Register Name:AC97C_CASR
Address: 0xFFFAC028
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RXBUFF ENDRX – – TXBUFE ENDTX – –
7 6 5 4 3 2 1 0
– – OVRUN RXRDY – UNRUN TXEMPTY TXRDY
• TXRDY: Channel Transmit Ready

0: Data has been loaded in Channel Transmit Register and is waiting to be loaded in the Channel Transmit Shift Register.
1: Channel Transmit Register is empty.
• TXEMPTY: Channel Transmit Empty

0: Data remains in the Channel Transmit Register or is currently transmitted from the Channel Transmit Shift Register.
1: Data in the Channel Transmit Register have been loaded in the Channel Transmit Shift Register and sent to the codec.
• UNRUN: Transmit Underrun

Active only when Variable Rate Mode is enabled (VRA bit set in the AC97C_MR register). Automatically cleared by a pro-
cessor read operation.
0: No data has been requested from the channel since the last read of the Status Register, or data has been available each
time the CODEC requested new data from the channel since the last read of the Status Register.
1: Data has been emitted while no valid data to send has been previously loaded into the Channel Transmit Shift Register
since the last read of the Status Register.
• RXRDY: Channel Receive Ready

0: Channel Receive Holding Register is empty.
1: Data has been received and loaded in Channel Receive Holding Register.

Automatically cleared by a processor read operation.
0: No data has been loaded in the Channel Receive Holding Register while previous data has not been read since the last
read of the Status Register.
1: Data has been loaded in the Channel Receive Holding Register while previous data has not yet been read since the last
• ENDTX: End of Transmission for Channel A

0: The register AC97C_CATCR has not reached 0 since the last write in AC97C_CATCR or AC97C_CANCR.
1: The register AC97C_CATCR has reached 0 since the last write in AC97C_CATCR or AC97C_CATNCR.
1011 AT91SAM9G45
• TXBUFE: Transmit Buffer Empty for Channel A
0: AC97C_CATCR or AC97C_CATNCR have a value other than 0.
1: Both AC97C_CATCR and AC97C_CATNCR have a value of 0.
• ENDRX: End of Reception for Channel A

0: The register AC97C_CARCR has not reached 0 since the last write in AC97C_CARCR or AC97C_CARNCR.
1: The register AC97C_CARCR has reached 0 since the last write in AC97C_CARCR or AC97C_CARNCR.
• RXBUFF: Receive Buffer Full for Channel A

0: AC97C_CARCR or AC97C_CARNCR have a value other than 0.
1: Both AC97C_CARCR and AC97C_CARNCR have a value of 0.
1012 AT91SAM9G45
42.8.9 AC97 Controller Channel B Status Register
Register Name:AC97C_CBSR
Address: 0xFFFAC038
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
RXBUFF ENDRX – – – TXBUFE ENDTX –
7 6 5 4 3 2 1 0





• ENDTX: End of Transmission for Channel B

0: The register AC97C_CBTCR has not reached 0 since the last write in AC97C_CBTCR or AC97C_CBNCR.
1: The register AC97C_CBTCR has reached 0 since the last write in AC97C_CBTCR or AC97C_CBTNCR.
1013 AT91SAM9G45
• TXBUFE: Transmit Buffer Empty for Channel B
0: AC97C_CBTCR or AC97C_CBTNCR have a value other than 0.
1: Both AC97C_CBTCR and AC97C_CBTNCR have a value of 0.
• ENDRX: End of Reception for Channel B

0: The register AC97C_CBRCR has not reached 0 since the last write in AC97C_CBRCR or AC97C_CBRNCR.
1: The register AC97C_CBRCR has reached 0 since the last write in AC97C_CBRCR or AC97C_CBRNCR.
• RXBUFF: Receive Buffer Full for Channel B

0: AC97C_CBRCR or AC97C_CBRNCR have a value other than 0.
1: Both AC97C_CBRCR and AC97C_CBRNCR have a value of 0.
1014 AT91SAM9G45
42.8.10 AC97 Controller Codec Status Register
Register Name: AC97C_COSR
Address: 0xFFFAC048
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0





1015 AT91SAM9G45
42.8.11 AC97 Controller Channel A Mode Register
Register Name:AC97C_CAMR
Address: 0xFFFAC02C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– PDCEN CEN – – CEM SIZE
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• TXRDY: Channel Transmit Ready Interrupt Enable
• TXEMPTY: Channel Transmit Empty Interrupt Enable
• UNRUN: Transmit Underrun Interrupt Enable
• RXRDY: Channel Receive Ready Interrupt Enable
• ENDTX: End of Transmission for Channel A Interrupt Enable
• TXBUFE: Transmit Buffer Empty for Channel A Interrupt Enable

0: Read: the corresponding interrupt is disabled. Write: disables the corresponding interrupt.
1: Read: the corresponding interrupt is enabled. Write: enables the corresponding interrupt.
• ENDRX: End of Reception for Channel A Interrupt Enable

• RXBUFF: Receive Buffer Full for Channel A Interrupt Enable

• SIZE: Channel A Data Size

SIZE Encoding
SIZE Selected Data Size
0x0 20 bits
0x1 18 bits
0x2 16 bits
0x3 10 bits
Note: Each time slot in the data phase is 20 bit long. For example, if a 16-bit sample stream is being played to an AC 97 DAC, the first
16 bit positions are presented to the DAC MSB-justified. They are followed by the next four bit positions that the AC97 Controller
1016 AT91SAM9G45
fills with zeroes. This process ensures that the least significant bits do not introduce any DC biasing, regardless of the imple-
mented DAC’s resolution (16-, 18-, or 20-bit)
• CEM: Channel A Endian Mode

0: Transferring Data through Channel A is straight forward (Little-Endian).
1: Transferring Data through Channel A from/to a memory is performed with from/to Big-Endian format translation.
• CEN: Channel A Enable

0: Data transfer is disabled on Channel A.
1: Data transfer is enabled on Channel A.
• PDCEN: Peripheral Data Controller Channel Enable

0: Channel A is not transferred through a Peripheral Data Controller Channel. Related PDC flags are ignored or not
generated.
1: Channel A is transferred through a Peripheral Data Controller Channel. Related PDC flags are taken into account or
generated.
1017 AT91SAM9G45
42.8.12 AC97 Controller Channel B Mode Register
Register Name:AC97C_CBMR
Address: 0xFFFAC03C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– PDCEN CEN – – CEM SIZE
15 14 13 12 11 10 9 8
7 6 5 4 3 2 1 0
• ENDTX: End of Transmission for Channel B Interrupt Enable
• TXBUFE: Transmit Buffer Empty for Channel B Interrupt Enable

• ENDRX: End of Reception for Channel B Interrupt Enable

• RXBUFF: Receive Buffer Full for Channel B Interrupt Enable

• SIZE: Channel B Data Size

SIZE Encoding
SIZE Selected Data Size
0x0 20 bits
0x1 18 bits
0x2 16 bits
0x3 10 bits
Note: Each time slot in the data phase is 20 bit long. For example, if a 16-bit sample stream is being played to an AC 97 DAC, the first
16 bit positions are presented to the DAC MSB-justified. They are followed by the next four bit positions that the AC97 Controller
1018 AT91SAM9G45
fills with zeroes. This process ensures that the least significant bits do not introduce any DC biasing, regardless of the imple-
mented DAC’s resolution (16-, 18-, or 20-bit)
• CEM: Channel B Endian Mode

0: Transferring Data through Channel B is straight forward (Little-Endian).
1: Transferring Data through Channel B from/to a memory is performed with from/to Big-Endian format translation.
• CEN: Channel B Enable

0: Data transfer is disabled on Channel B.
1: Data transfer is enabled on Channel B.
• PDCEN: Peripheral Data Controller Channel Enable

0: Channel B is not transferred through a Peripheral Data Controller Channel. Related PDC flags are ignored or not
generated.
1: Channel B is transferred through a Peripheral Data Controller Channel. Related PDC flags are taken into account or
generated.
1019 AT91SAM9G45
42.8.13 AC97 Controller Codec Mode Register
Register Name: AC97C_COMR
Address: 0xFFFAC04C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
1020 AT91SAM9G45
42.8.14 AC97 Controller Status Register
Register Name:AC97C_SR
Address: 0xFFFAC050
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – CBEVT CAEVT COEVT WKUP SOF
WKUP and SOF flags in AC97C_SR register are automatically cleared by a processor read operation.
• SOF: Start Of Frame

0: No Start of Frame has been detected since the last read of the Status Register.
1: At least one Start of frame has been detected since the last read of the Status Register.
• WKUP: Wake Up detection

0: No Wake-up has been detected.
1: At least one rising edge on SDATA_IN has been asynchronously detected. That means AC97 Codec has notified a
wake-up.
• COEVT: CODEC Channel Event

A Codec channel event occurs when AC97C_COSR AND AC97C_COMR is not 0. COEVT flag is automatically cleared
when the channel event condition is cleared.
0: No event on the CODEC channel has been detected since the last read of the Status Register.
1: At least one event on the CODEC channel is active.
• CAEVT: Channel A Event

A channel A event occurs when AC97C_CASR AND AC97C_CAMR is not 0. CAEVT flag is automatically cleared when the
channel event condition is cleared.
0: No event on the channel A has been detected since the last read of the Status Register.
1: At least one event on the channel A is active.
• CBEVT: Channel B Event

A channel B event occurs when AC97C_CBSR AND AC97C_CBMR is not 0. CBEVT flag is automatically cleared when the
channel event condition is cleared.
0: No event on the channel B has been detected since the last read of the Status Register.
1: At least one event on the channel B is active.
1021 AT91SAM9G45
42.8.15 AC97 Codec Controller Interrupt Enable Register
Register Name:AC97C_IER
Address: 0xFFFAC054
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
• WKUP: Wake Up
• COEVT: Codec Event

0: No Effect.
1: Enables the corresponding interrupt.
1022 AT91SAM9G45
42.8.16 AC97 Controller Interrupt Disable Register
Register Name:AC97C_IDR
Address: 0xFFFAC058
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
• WKUP: Wake Up

0: No Effect.
1: Disables the corresponding interrupt.
1023 AT91SAM9G45
42.8.17 AC97 Controller Interrupt Mask Register
Register Name:AC97C_IMR
Address: 0xFFFAC05C
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
• WKUP: Wake Up

0: The corresponding interrupt is disabled.
1: The corresponding interrupt is enabled.
1024 AT91SAM9G45
AT91SAM9G45
43. True Random Number Generator (TRNG)
43.1 Description
The True Random Number Generator (TRNG) passes the American NIST Special Publication
800-22 and Diehard Random Tests Suites.
As soon as the TRNG is enabled (TRNG_CTRL register), the generator provides one 32-bit
value every 84 clock cycles. Interrupt trng_int can be enabled through the TRNG_IER register
(respectively disabled in TRNG_IDR). This interrupt is set when a new random value is available
and is cleared when the status register is read (TRNG_SR register). The flag DATRDY of the
status register (TRNG_ISR) is set when the random data is ready to be read out on the 32-bit
output data register (TRNG_ODATA).
The normal mode of operation checks that the status register flag equals 1 before reading the
output data register when a 32-bit random value is required by the software application.
Figure 43-1. TRNG Data Generation Sequence
clock
trng_cr
enable
84 clock cycles 84 clock cycles 84 clock cycles
trng_int
Read TRNG_ISR Read TRNG_ISR

Read TRNG_ODATA Read TRNG_ODATA
1025
43.2 True Random Number Generator (TRNG) User Interface
0x00 Control Register TRNG_CR Write-only –
0x10 Interrupt Enable Register TRNG_IER Write-only –
0x14 Interrupt Disable Register TRNG_IDR Write-only –
0x18 Interrupt Mask Register TRNG_IMR Read-only 0x0000
0x1C Interrupt Status Register TRNG_ISR Read-only 0x0000
0x50 Output Data Register TRNG_ODATA Read-only 0x0000
1026 AT91SAM9G45
AT91SAM9G45
43.2.1 TRNG Control Register

Name: TRNG_CR
Address: 0xFFFCC000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – ENABLE
• ENABLE: Enables the TRNG to provide random values

0 = Disables the TRNG.
1 = Enables the TRNG.
1027
43.2.2 TRNG Interrupt Enable Register
Name: TRNG_IER
Address: 0xFFFCC010
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – DATRDY
• DATRDY: Data Ready Interrupt Enable

0 = No effect.
1028 AT91SAM9G45
AT91SAM9G45
43.2.3 TRNG Interrupt Disable Register

Name: TRNG_IDR
Address: 0xFFFCC014
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – DATRDY
• DATRDY: Data Ready Interrupt Disable

0 = No effect.
1029
43.2.4 TRNG Interrupt Mask Register
Name: TRNG_IMR
Address: 0xFFFCC018
Reset: 0x0000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – – – – DATRDY
• DATRDY: Data Ready Interrupt Mask

1030 AT91SAM9G45
AT91SAM9G45
43.2.5 TRNG Interrupt Status Register

Name: TRNG_ISR
Address: 0xFFFCC01C
Reset: 0x0000
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – –
7 6 5 4 3 2 1 0
– – – – – – – DATRDY
• DATRDY: Data Ready

0 = Output data is not valid or TRNG is disabled.
1 = New Random value is completed.
DATRDY is cleared when this register is read.
1031
43.2.6 TRNG Output Data Register
Name: TRNG_ODATA
Address: 0xFFFCC050
Reset: 0x0000
31 30 29 28 27 26 25 24
ODATA
23 22 21 20 19 18 17 16
ODATA
15 14 13 12 11 10 9 8
ODATA
7 6 5 4 3 2 1 0
ODATA
• ODATA: Output Data

The 32-bit Output Data register contains the 32-bit random data.
1032 AT91SAM9G45
AT91SAM9G45
44. LCD Controller (LCDC)
44.1 Description
The LCD Controller (LCDC) consists of logic for transferring LCD image data from an external
display buffer to an LCD module with integrated common and segment drivers.
The LCD Controller supports single and double scan monochrome and color passive STN LCD
modules and single scan active TFT LCD modules. On monochrome STN displays, up to 16
gray shades are supported using a time-based dithering algorithm and Frame Rate Control
(FRC) method. This method is also used in color STN displays to generate up to 4096 colors.
The LCD Controller has a display input buffer (FIFO) to allow a flexible connection of the exter-
nal AHB master interface, and a lookup table to allow palletized display configurations.
The LCD Controller is programmable in order to support many different requirements such as
resolutions up to 2048 x 2048; pixel depth (1, 2, 4, 8, 16, 24 bits per pixel); data line width (4, 8,
16 or 24 bits) and interface timing.
The LCD Controller is connected to the ARM Advanced High Performance Bus (AHB) as a mas-
ter for reading pixel data. However, the LCD Controller interfaces with the AHB as a slave in
order to configure its registers.

• Single and Dual scan color and monochrome passive STN LCD panels supported
• Single scan active TFT LCD panels supported.
• 4-bit single scan, 8-bit single or dual scan, 16-bit dual scan STN interfaces supported
• Up to 24-bit single scan TFT interfaces supported
• Up to 16 gray levels for mono STN and up to 4096 colors for color STN displays
• 1, 2 bits per pixel (palletized), 4 bits per pixel (non-palletized) for mono STN
• 1, 2, 4, 8 bits per pixel (palletized), 16 bits per pixel (non-palletized) for color STN
• 1, 2, 4, 8 bits per pixel (palletized), 16, 24 bits per pixel (non-palletized) for TFT
• Single clock domain architecture
• Resolution supported up to 2048 x 2048
1033
44.3 Block Diagram
Figure 44-1. LCD Macrocell Block Diagram
AHB MASTER AHB SLAVE
DMA Controller SPLIT
AHB IF
CFG
AHB SLAVE
DMA Data
Dvalid
Dvalid
CH-U CH-L CTRL
Upper Push Lower Push
Input Interface LCD Controller Core
FIFO Configuration IF CFG

AHB SLAVE
SERIALIZER
DATAPATH
LUT Mem Interface

PALETTE LUT Mem Interface
FIFO Mem Interface
Control Interface
DITHERING
FIFO LUT
MEM MEM
OUTPUT
SHIFTER Timegen
DISPLAY IF
LCDD
Control signals
Display PWM
DISPLAY IF
1034 AT91SAM9G45
AT91SAM9G45

LCDCC Contrast control signal Output
LCDHSYNC Line synchronous signal (STN) or Horizontal synchronous signal (TFT) Output
LCDDOTCK LCD clock signal (STN/TFT) Output
LCDVSYNC Frame synchronous signal (STN) or Vertical synchronization signal (TFT) Output
LCDDEN Data enable signal Output
LCDD[23:0] LCD Data Bus output Output
44.5.1 I/O Lines

The pins used for interfacing the LCD Controller may be multiplexed with PIO lines. The pro-
grammer must first program the PIO Controller to assign the pins to their peripheral function. If
I/O lines of the LCD Controller are not used by the application, they can be used for other pur-
poses by the PIO Controller.

LCDC LCDCC PE2 A
LCDC LCDDEN PE6 A
LCDC LCDDOTCK PE5 A
LCDC LCDD0 PE7 A
LCDC LCDD1 PE8 A
LCDC LCDD2 PE7 B
LCDC LCDD2 PE9 A
LCDC LCDD3 PE8 B
LCDC LCDD3 PE10 A
LCDC LCDD4 PE9 B
LCDC LCDD4 PE11 A
LCDC LCDD5 PE10 B
LCDC LCDD5 PE12 A
LCDC LCDD6 PE11 B
LCDC LCDD6 PE13 A
LCDC LCDD7 PE12 B
LCDC LCDD7 PE14 A
LCDC LCDD8 PE15 A
LCDC LCDD9 PE16 A
LCDC LCDD10 PE13 B
1035
LCDC LCDD10 PE17 A
LCDC LCDD11 PE14 B
LCDC LCDD11 PE18 A
LCDC LCDD12 PE15 B
LCDC LCDD12 PE19 A
LCDC LCDD13 PE16 B
LCDC LCDD13 PE20 A
LCDC LCDD14 PE17 B
LCDC LCDD14 PE21 A
LCDC LCDD15 PE18 B
LCDC LCDD15 PE22 A
LCDC LCDD16 PE23 A
LCDC LCDD17 PE24 A
LCDC LCDD18 PE19 B
LCDC LCDD18 PE25 A
LCDC LCDD19 PE20 B
LCDC LCDD19 PE26 A
LCDC LCDD20 PE21 B
LCDC LCDD20 PE27 A
LCDC LCDD21 PE22 B
LCDC LCDD21 PE28 A
LCDC LCDD22 PE23 B
LCDC LCDD22 PE29 A
LCDC LCDD23 PE24 B
LCDC LCDD23 PE30 A
LCDC LCDHSYNC PE4 A
LCDC LCDMOD PE1 A
LCDC LCDPWR PE0 A
LCDC LCDVSYNC PE3 A

The LCD Controller is not continuously clocked. The user must first enable the LCD Controller
clock in the Power Management Controller before using it (PMC_PCER).
1036 AT91SAM9G45
AT91SAM9G45

The LCD Controller interrupt line is connected to one of the internal sources of the Advanced
Interrupt Controller. Using the LCD Controller interrupt requires prior programming of the AIC.

Instance ID
LCDC 23

The LCD Controller consists of two main blocks (Figure 44-1 on page 1034), the DMA controller
and the LCD controller core (LCDC core). The DMA controller reads the display data from an
external memory through a AHB master interface. The LCD controller core formats the display
data. The LCD controller core continuously pumps the pixel data into the LCD module via the
LCD data bus (LCDD[23:0]); this bus is timed by the LCDDOTCK, LCDDEN, LCDHSYNC, and
LCDVSYNC signals.
44.6.1 DMA Controller
44.6.1.1 Configuration Block

The configuration block is a set of programmable registers that are used to configure the DMA
controller operation. These registers are written via the AHB slave interface. Only word access is
allowed.
For details on the configuration registers, see “LCD Controller (LCDC) User Interface” on page
1063.
44.6.1.2 AHB Interface

This block generates the AHB transactions. It generates undefined-length incrementing bursts
as well as 4-, 8- or 16-beat incrementing bursts. The size of the transfer can be configured in the
BRSTLN field of the DMAFRMCFG register. For details on this register, see “DMA Frame Con-
figuration Register” on page 1068.
44.6.1.3 Channel-U
This block stores the base address and the number of words transferred for this channel (frame
in single scan mode and Upper Panel in dual scan mode) since the beginning of the frame. It
also generates the end of frame signal.
It has two pointers, the base address and the number of words to transfer. When the module
receives a new_frame signal, it reloads the number of words to transfer pointer with the size of
the frame/panel. When the module receives the new_frame signal, it also reloads the base
address with the base address programmed by the host.
The size of the frame/panel can be programmed in the FRMSIZE field of the DMAFRMCFG
Register. This size is calculated as follows:
X_size*Y_size
Frame_size = --------------------------------------
32
where:
X_size = ((LINESIZE+1)*Bpp+PIXELOFF)/32
1037
Y_size = (LINEVAL+1)
• LINESIZE is the horizontal size of the display in pixels, minus 1, as programmed in the
LINESIZE field of the LCDFRMCFG register of the LCD Controller.
• Bpp is the number of bits per pixel configured.
• PIXELOFF is the pixel offset for 2D addressing, as programmed in the DMA2DCFG register.
Applicable only if 2D addressing is being used.
• LINEVAL is the vertical size of the display in pixels, minus 1, as programmed in the LINEVAL
field of the LCDFRMCFG register of the LCD Controller.
Note: X_size is calculated as an up-rounding of a division by 32. (This can also be done adding 31 to the
dividend before using an integer division by 32). When using the 2D-addressing mode (see “2D
Memory Addressing” on page 1060), it is important to note that the above calculation must be exe-
cuted and the FRMSIZE field programmed with every movement of the displaying window, since a
change in the PIXELOFF field can change the resulting FRMSIZE value.
44.6.1.4 Channel-L
This block has the same functionality as Channel-U, but for the Lower Panel in dual scan mode
only.
44.6.1.5 Control
This block receives the request signals from the LCDC core and generates the requests for the
channels.
44.6.2 LCD Controller Core
44.6.2.1 Configuration Block

The configuration block is a set of programmable registers that are used to configure the LCDC
core operation. These registers are written via the AHB slave interface. Only word access is
allowed.
The description of the configuration registers can be found in “LCD Controller (LCDC) User
Interface” on page 1063.
44.6.2.2 Datapath
The datapath block contains five submodules: FIFO, Serializer, Palette, Dithering and Shifter.
The structure of the datapath is shown in Figure 44-2.
1038 AT91SAM9G45
AT91SAM9G45
Figure 44-2. Datapath Structure

Input Interface
FIFO
Serializer
Configuration IF
Palette
Control Interface
Dithering
Output
Shifter
Output Interface
This module transforms the data read from the memory into a format according to the LCD mod-
ule used. It has four different interfaces: the input interface, the output interface, the
configuration interface and the control interface.
• The input interface connects the datapath with the DMA controller. It is a dual FIFO interface
with a data bus and two push lines that are used by the DMA controller to fill the FIFOs.
• The output interface is a 24-bit data bus. The configuration of this interface depends on the
type of LCD used (TFT or STN, Single or Dual Scan, 4-bit, 8-bit, 16-bit or 24-bit interface).
• The configuration interface connects the datapath with the configuration block. It is used to
select between the different datapath configurations.
• The control interface connects the datapath with the timing generation block. The main
control signal is the data-request signal, used by the timing generation module to request
new data from the datapath.
The datapath can be characterized by two parameters: initial_latency and cycles_per_data. The
parameter initial_latency is defined as the number of LCDC Core Clock cycles until the first data
is available at the output of the datapath. The parameter cycles_per_data is the minimum num-
ber of LCDC Core clock cycles between two consecutive data at the output interface.
1039
These parameters are different for the different configurations of the LCD Controller and are
shown in Table 44-4.
Table 44-4. Datapath Parameters

Configuration
DISTYPE SCAN IFWIDTH initial_latency cycles_per_data
TFT 9 1
STN Mono Single 4 13 4
STN Mono Single 8 17 8
STN Mono Dual 8 17 8
STN Mono Dual 16 25 16
STN Color Single 4 11 2
STN Color Single 8 12 3
STN Color Dual 8 14 4
STN Color Dual 16 15 6
44.6.2.3 FIFO
The FIFO block buffers the input data read by the DMA module. It contains two input FIFOs to
be used in Dual Scan configuration that are configured as a single FIFO when used in single
scan configuration.
The size of the FIFOs allows a wide range of architectures to be supported.
The upper threshold of the FIFOs can be configured in the FIFOTH field of the LCDFIFO regis-
ter. The LCDC core will request a DMA transfer when the number of words in each FIFO is less
than FIFOTH words. To avoid overwriting in the FIFO and to maximize the FIFO utilization, the
FIFOTH should be programmed with:
FIFOTH (in Words) = 512 - (2 x DMA_BURST_LENGTH + 3)
where:
• 512 is the effective size of the FIFO in Words. It is the total FIFO memory size in single scan
mode and half that size in dual scan mode.
• DMA_burst_length is the burst length of the transfers made by the DMA in Words.
44.6.2.4 Serializer
This block serializes the data read from memory. It reads words from the FIFO and outputs pix-
els (1 bit, 2 bits, 4 bits, 8 bits, 16 bits or 24 bits wide) depending on the format specified in the
PIXELSIZE field of the LCDCON2 register. It also adapts the memory-ordering format. Both big-
endian and little-endian formats are supported. They are configured in the MEMOR field of the
LCDCON2 register.
The organization of the pixel data in the memory depends on the configuration and is shown in
Table 44-5 and Table 44-7.
Note: For a color depth of 24 bits per pixel there are two different formats supported: packed and
unpacked. The packed format needs less memory but has some limitations when working in 2D
addressing mode (See “2D Memory Addressing” on page 1060.).
1040 AT91SAM9G45
AT91SAM9G45
Table 44-6. Little Endian Memory Organization

Mem Addr 0x3 0x2 0x1 0x0
Bit 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Pixel 1bpp 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Pixel 2bpp 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Pixel 4bpp 7 6 5 4 3 2 1 0
Pixel 8bpp 3 2 1 0
Pixel
1 0
16bpp
Pixel
24bpp 1 0
packed
Pixel
24bpp 2 1
packed
Pixel
24bpp 3 2
packed
Pixel
24bpp not used 0
unpacked
Table 44-7. Big Endian Memory Organization

Bit 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Pixel 1bpp 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Pixel 2bpp 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Pixel 4bpp 0 1 2 3 4 5 6 7
Pixel 8bpp 0 1 2 3
Pixel
0 1
16bpp
Pixel
24bpp 0 1
packed
Pixel
24bpp 1 2
packed
1041
Table 44-7. Big Endian Memory Organization (Continued)
Pixel
24bpp 2 3
packed
Pixel
24bpp 4 5
packed
Pixel
24bpp not used 0
unpacked
Table 44-8. WinCE Pixel Memory Organization

Bit 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Pixel 1bpp 24 25 26 27 28 29 30 31 16 17 18 19 20 21 22 23 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7
Pixel 2bpp 12 13 14 15 8 9 10 11 4 5 6 7 0 1 3 3
Pixel 4bpp 6 7 4 5 2 3 0 1
Pixel 8bpp 3 2 1 0
Pixel
1 0
16bpp
Pixel
24bpp 1 0
packed
Pixel
24bpp 2 1
packed
Pixel
24bpp 3 2
packed
Pixel
24bpp not used 0
unpacked
44.6.2.5 Palette
This block is used to generate the pixel gray or color information in palletized configurations. The
different modes with the palletized/non-palletized configuration can be found in Table 44-9. In
these modes, 1, 2, 4 or 8 input bits index an entry in the lookup table. The corresponding entry in
the lookup table contains the color or gray shade information for the pixel.
Table 44-9. Palette Configurations

Configuration
DISTYPE PIXELSIZE Palette
TFT 1, 2, 4, 8 Palletized
TFT 16, 24 Non-palletized
STN Mono 1, 2 Palletized
1042 AT91SAM9G45
AT91SAM9G45
Table 44-9. Palette Configurations

Configuration
DISTYPE PIXELSIZE Palette
STN Mono 4 Non-palletized
STN Color 1, 2, 4, 8 Palletized
STN Color 16 Non-palletized
The lookup table can be accessed by the host in R/W mode to allow the host to program and
check the values stored in the palette. It is mapped in the LCD controller configuration memory
map. The LUT is mapped as 16-bit half-words aligned at word boundaries, only word write
access is allowed (the 16 MSB of the bus are not used). For the detailed memory map, see
The lookup table contains 256 16-bit wide entries. The 256 entries are chosen by the program-
mer from the 216 possible combinations.
For the structure of each LUT entry, see Table 44-10.
Table 44-10. Lookup Table Structure in the Memory

Address Data Output [15:0]
00 Red_value_0[4:0] Green_value_0[5:0] Blue_value_0[4:0]
01 Red_value_1[4:0] Green_value_1[5:0] Blue_value_1[4:0]
...
FE Red_value_254[4:0] Green_value_254[5:0] Blue_value_254[4:0]
FF Red_value_255[4:0] Green_value_255[5:0] Blue_value_255[4:0]
In STN Monochrome, only the four most significant bits of the red value are used (16 gray
shades). In STN Color, only the four most significant bits of the blue, green and red value are
used (4096 colors).
In TFT mode, all the bits in the blue, green and red values are used. The LCDD unused bits are
tied to 0 when TFT palletized configurations are used (LCDD[18:16], LCDD[9:8], LCDD[2:0]).
44.6.2.6 Dithering
The dithering block is used to generate the shades of gray or color when the LCD Controller is
used with an STN LCD Module. It uses a time-based dithering algorithm and Frame Rate Con-
trol method.
The Frame Rate Control varies the duty cycle for which a given pixel is turned on, giving the dis-
play an appearance of multiple shades. In order to reduce the flicker noise caused by turning on
and off adjacent pixels at the same time, a time-based dithering algorithm is used to vary the
pattern of adjacent pixels every frame. This algorithm is expressed in terms of Dithering Pattern
registers (DP_i) and considers not only the pixel gray level number, but also its horizontal
coordinate.
1043
Table 44-11 shows the correspondences between the gray levels and the duty cycle.
Table 44-11. Dithering Duty Cycle

Gray Level Duty Cycle Pattern Register
15 1 -
14 6/7 DP6_7
13 4/5 DP4_5
12 3/4 DP3_4
11 5/7 DP5_7
10 2/3 DP2_3
9 3/5 DP3_5
8 4/7 DP4_7
7 1/2 ~DP1_2
6 3/7 ~DP4_7
5 2/5 ~DP3_5
4 1/3 ~DP2_3
3 1/4 ~DP3_4
2 1/5 ~DP4_5
1 1/7 ~DP6_7
0 0 -
The duty cycles for gray levels 0 and 15 are 0 and 1, respectively.
The same DP_i register can be used for the pairs for which the sum of duty cycles is 1 (e.g., 1/7
and 6/7). The dithering pattern for the first pair member is the inversion of the one for the
second.
The DP_i registers contain a series of 4-bit patterns. The (3-m)th bit of the pattern determines if a
pixel with horizontal coordinate x = 4n + m (n is an integer and m ranges from 0 to 3) should be
turned on or off in the current frame. The operation is shown by the examples below.
Consider the pixels a, b, c and d with the horizontal coordinates 4*n+0, 4*n+1, 4*n+2 and 4*n+3,
respectively. The four pixels should be displayed in gray level 9 (duty cycle 3/5) so the register
used is DP3_5 =”1010 0101 1010 0101 1111”.
The output sequence obtained in the data output for monochrome mode is shown in Table 44-
12.
Table 44-12. Dithering Algorithm for Monochrome Mode

Frame
Number Pattern Pixel a Pixel b Pixel c Pixel d
N 1010 ON OFF ON OFF
N+1 0101 OFF ON OFF ON
N+2 1010 ON OFF ON OFF
1044 AT91SAM9G45
AT91SAM9G45
Table 44-12. Dithering Algorithm for Monochrome Mode

Frame
Number Pattern Pixel a Pixel b Pixel c Pixel d
N+4 1111 ON ON ON ON
... ... ... ... ... ...
Consider now color display mode and two pixels p0 and p1 with the horizontal coordinates
4*n+0, and 4*n+1. A color pixel is composed of three components: {R, G, B}. Pixel p0 will be dis-
played sending the color components {R0, G0, B0} to the display. Pixel p1 will be displayed
sending the color components {R1, G1, B1}. Suppose that the data read from memory and
mapped to the lookup tables corresponds to shade level 10 for the three color components of
both pixels, with the dithering pattern to apply to all of them being DP2_3 = “1101 1011 0110”.
Table 44-13 shows the output sequence in the data output bus for single scan configurations. (In
Dual Scan Configuration, each panel data bus acts like in the equivalent single scan
configuration.)
Table 44-13. Dithering Algorithm for Color Mode

Frame Signal Shadow Level Bit used Dithering Pattern 4-bit LCDD 8-bit LCDD Output
N red_data_0 1010 3 1101 LCDD[3] LCDD[7] R0
N green_data_0 1010 2 1101 LCDD[2] LCDD[6] G0
N blue_data_0 1010 1 1101 LCDD[1] LCDD[5] b0
N red_data_1 1010 0 1101 LCDD[0] LCDD[4] R1
N green_data_1 1010 3 1101 LCDD[3] LCDD[3] G1
N blue_data_1 1010 2 1101 LCDD[2] LCDD[2] B1
… … … … … … … …
N+1 red_data_0 1010 3 1011 LCDD[3] LCDD[7] R0
N+1 green_data_0 1010 2 1011 LCDD[2] LCDD[6] g0
N+1 blue_data_0 1010 1 1011 LCDD[1] LCDD[5] B0
N+1 red_data_1 1010 0 1011 LCDD[0] LCDD[4] R1
N+1 green_data_1 1010 3 1011 LCDD[3] LCDD[3] G1
N+1 blue_data_1 1010 2 1011 LCDD[2] LCDD[2] b1
… … … … … … … …
N+2 red_data_0 1010 3 0110 LCDD[3] LCDD[7] r0
N+2 green_data_0 1010 2 0110 LCDD[2] LCDD[6] G0
N+2 red_data_1 1010 0 0110 LCDD[0] LCDD[4] r1
1045
Table 44-13. Dithering Algorithm for Color Mode (Continued)
Frame Signal Shadow Level Bit used Dithering Pattern 4-bit LCDD 8-bit LCDD Output
N+2 green_data_1 1010 3 0110 LCDD[3] LCDD[3] g1
… … … … … … … …
Note: Ri = red pixel component ON. Gi = green pixel component ON. Bi = blue pixel component ON. ri = red pixel component OFF.
gi = green pixel component OFF. bi = blue pixel component OFF.
44.6.2.7 Shifter
The FIFO, Serializer, Palette and Dithering modules process one pixel at a time in monochrome
mode and three sub-pixels at a time in color mode (R,G,B components). This module packs the
data according to the output interface. This interface can be programmed in the DISTYPE,
SCANMOD, and IFWIDTH fields of the LDCCON3 register.
The DISTYPE field selects between TFT, STN monochrome and STN color display. The SCAN-
MODE field selects between single and dual scan modes; in TFT mode, only single scan is
supported. The IFWIDTH field configures the width of the interface in STN mode: 4-bit (in single
scan mode only), 8-bit and 16-bit (in dual scan mode only).
For a more detailed description of the fields, see “LCD Controller (LCDC) User Interface” on
page 1063.
For a more detailed description of the LCD Interface, see “LCD Interface” on page 1051.
44.6.2.8 Timegen
The time generator block generates the control signals LCDDOTCK, LCDHSYNC, LCDVSYNC,
LCDDEN, used by the LCD module. This block is programmable in order to support different
types of LCD modules and obtain the output clock signals, which are derived from the LCDC
Core clock.
The LCDDOTCK signal is used to clock the data into the LCD drivers' shift register. The data is
sent through LCDD[23:0] synchronized by default with LCDDOTCK falling edge (rising edge can
be selected). The CLKVAL field of LCDCON1 register controls the rate of this signal. The divisor
can also be bypassed with the BYPASS bit in the LCDCON1 register. In this case, the rate of
LCDDOTCK is equal to the frequency of the LCDC Core clock. The minimum period of the LCD-
DOTCK signal depends on the configuration. This information can be found in Table 44-14.
f LCDC_clock
f LCDDOTCK = -------------------------------
-
CLKVAL + 1
The LCDDOTCK signal has two different timings that are selected with the CLKMOD field of the
LCDCON2 register:
• Always Active (used with TFT LCD Modules)
• Active only when data is available (used with STN LCD Modules)
1046 AT91SAM9G45
AT91SAM9G45
Table 44-14. Minimum LCDDOTCK Period in LCDC Core Clock Cycles

Configuration
DISTYPE SCAN IFWIDTH LCDDOTCK Period
TFT 1
STN Mono Single 4 4
STN Mono Single 8 8
STN Mono Dual 8 8
STN Mono Dual 16 16
STN Color Single 4 2
STN Color Single 8 2
STN Color Dual 8 4
STN Color Dual 16 6
The LCDDEN signal indicates valid data in the LCD Interface.

After each horizontal line of data has been shifted into the LCD, the LCDHSYNC is asserted to
cause the line to be displayed on the panel.
The following timing parameters can be configured:
• Vertical to Horizontal Delay (VHDLY): The delay between the falling edge of LCDVSYNC and
the generation of LCDHSYNC is configurable in the VHDLY field of the LCDTIM1 register.
The delay is equal to (VHDLY+1) LCDDOTCK cycles.
• Horizontal Pulse Width (HPW): The LCDHSYNC pulse width is configurable in HPW field of
LCDTIM2 register. The width is equal to (HPW + 1) LCDDOTCK cycles.
• Horizontal Back Porch (HBP): The delay between the LCDHSYNC falling edge and the first
LCDDOTCK rising edge with valid data at the LCD Interface is configurable in the HBP field
of the LCDTIM2 register. The delay is equal to (HBP+1) LCDDOTCK cycles.
• Horizontal Front Porch (HFP): The delay between end of valid data and the generation of the
next LCDHSYNC is configurable in the HFP field of the LCDTIM2 register. The delay is equal
to (HFP+VHDLY+2) LCDDOTCK cycles.
There is a limitation in the minimum values of VHDLY, HPW and HBP parameters imposed by
the initial latency of the datapath. The total delay in LCDC clock cycles must be higher than or
equal to the latency column in Table 44-4 on page 1040. This limitation is given by the following
formula:
44.6.2.9 Equation 1
( VHDLY + HPW + HBP + 3 ) × PCLK_PERIOD ≥ DPATH_LATENCY
where:
• VHDLY, HPW, HBP are the value of the fields of LCDTIM1 and LCDTIM2 registers
• PCLK_PERIOD is the period of LCDDOTCK signal measured in LCDC Clock cycles
• DPATH_LATENCY is the datapath latency of the configuration, given in Table 44-4 on page
1040
1047
The LCDVSYNC is asserted once per frame. This signal is asserted to cause the LCD's line
pointer to start over at the top of the display. The timing of this signal depends on the type of
LCD: STN or TFT LCD.
In STN mode, the high phase corresponds to the complete first line of the frame. In STN mode,
this signal is synchronized with the first active LCDDOTCK rising edge in a line.
In TFT mode, the high phase of this signal starts at the beginning of the first line. The following
timing parameters can be selected:
• Vertical Pulse Width (VPW): LCDVSYNC pulse width is configurable in VPW field of the
LCDTIM1 register. The pulse width is equal to (VPW+1) lines.
• Vertical Back Porch: Number of inactive lines at the beginning of the frame is configurable in
VBP field of LCDTIM1 register. The number of inactive lines is equal to VBP. This field should
be programmed with 0 in STN Mode.
• Vertical Front Porch: Number of inactive lines at the end of the frame is configurable in VFP
field of LCDTIM2 register. The number of inactive lines is equal to VFP. This field should be
programmed with 0 in STN mode.
There are two other parameters to configure in this module, the HOZVAL and the LINEVAL
fields of the LCDFRMCFG:
• HOZVAL configures the number of active LCDDOTCK cycles in each line. The number of
active cycles in each line is equal to (HOZVAL+1) cycles. The minimum value of this
parameter is 1.
• LINEVAL configures the number of active lines per frame. This number is equal to
(LINEVAL+1) lines. The minimum value of this parameter is 1.
Figure 44-3, Figure 44-4 and Figure 44-5 show the timing of LCDDOTCK, LCDDEN, LCDH-
SYNC and LCDVSYNC signals:
1048 AT91SAM9G45
AT91SAM9G45
Figure 44-3. STN Panel Timing, CLKMOD 0

Frame Period
LCDVSYNC
LCDHSYNC
LCDDEN
LCDDOTCK
LCDD
Line Period
VHDLY+ HPW+1 HBP+1 HOZVAL+1 HFP+VHDLY+2

LCDVSYNC
LCDHSYNC
LCDDEN
LCDDOTCK
LCDD
1 PCLK 1/2 PCLK 1/2 PCLK
Figure 44-4. TFT Panel Timing, CLKMOD = 0, VPW = 2, VBP = 2, VFP = 1

Frame Period
(VPW+1) Lines
LCDVSYNC
Vertical Back Porch = VBP Lines Vertical Fron t Porch = VFP Lines
VHDLY+1
LCDHSYNC
LCDDEN
LCDDOTCK
LCDD
Line Period
VHDLY+1 HPW+1 HBP+1 HOZVAL+1 HFP+VHDLY+2

LCDVSYNC
LCDHSYNC
LCDDEN
LCDDOTCK
LCDD
1049
Figure 44-5. TFT Panel Timing (Line Expanded View), CLKMOD = 1
Line Period
VHDLY+1 HPW+1 HBP+1 HOZVAL+1 HFP+VHDLY+2

LCDVSYNC
LCDHSYNC
LCDDEN
LCDDOTCK
LCDD
Usually the LCD_FRM rate is about 70 Hz to 75 Hz. It is given by the following equation:
VHDLY + HPW + HBP + HOZVAL + HFP + ( 4 )

1 - = ⎛ --------------------------------------------------------------------------------------------------------------------------⎞ ( VBP + LINEVAL + VFP + 1 )
---------------------------
f LCDVSYNC ⎝ f LCDDOTCK ⎠
where:
• HOZVAL determines de number of LCDDOTCK cycles per line
• LINEVAL determines the number of LCDHSYNC cycles per frame, according to the
expressions shown below:
In STN Mode:
HOZVAL = Horizontal_display_size
--------------------------------------------------------------- – 1
Number_data_lines
LINEVAL = Vertical_display_size – 1
In monochrome mode, Horizontal_display_size is equal to the number of horizontal pixels. The

number_data_lines is equal to the number of bits of the interface in single scan mode;
number_data_lines is equal to half the bits of the interface in dual scan mode.
In color mode, Horizontal_display_size equals three times the number of horizontal pixels.
In TFT Mode:
HOZVAL = Horizontal_display_size – 1
LINEVAL = Vertical_display_size – 1
The frame rate equation is used first without considering the clock periods added at the end
beginning or at the end of each line to determine, approximately, the LCDDOTCK rate:
f lcd_pclk = ( HOZVAL + 5 ) × ( f lcd_vsync × ( LINEVAL + 1 ) )
With this value, the CLKVAL is fixed, as well as the corresponding LCDDOTCK rate.
Then select VHDLY, HPW and HBP according to the type of LCD used and “Equation 1” on
page 1047.
Finally, the frame rate is adjusted to 70 Hz - 75 Hz with the HFP value:
1050 AT91SAM9G45
AT91SAM9G45
1
HFP = f LCDDOTCK × --------------------------------------------------------------------------------------------------------------- – ( VHDLY + HPW + HPB + HOZVAL + 4 )
f LCDVSYNC × ( LINEVAL + VBP + VFP + 1 )
The line counting is controlled by the read-only field LINECNT of LCDCON1 register. The LINE-
CNT field decreases by one unit at each falling edge of LCDHSYNC.
44.6.2.10 Display
This block is used to configure the polarity of the data and control signals. The polarity of all
clock signals can be configured by LCDCON2[12:8] register setting.
This block also generates the lcd_pwr signal internally used to control the state of the LCD pins
and to turn on and off by software the LCD module.
This signal is controlled by the PWRCON register and respects the number of frames configured
in the GUARD_TIME field of PWRCON register (PWRCON[7:1]) between the write access to
LCD_PWR field (PWRCON[0]) and the activation/deactivation of lcd_pwr signal.
The minimum value for the GUARD_TIME field is one frame. This gives the DMA Controller
enough time to fill the FIFOs before the start of data transfer to the LCD.
44.6.2.11 PWM
This block generates the LCD contrast control signal (LCDCC) to make possible the control of
the display's contrast by software. This is an 8-bit PWM (Pulse Width Modulation) signal that can
be converted to an analog voltage with a simple passive filter.
The PWM module has a free-running counter whose value is compared against a compare reg-
ister (CONSTRAST_VAL register). If the value in the counter is less than that in the register, the
output brings the value of the polarity (POL) bit in the PWM control register: CONTRAST_CTR.
Otherwise, the opposite value is output. Thus, a periodic waveform with a pulse width propor-
tional to the value in the compare register is generated.
Due to the comparison mechanism, the output pulse has a width between zero and 255 PWM
counter cycles. Thus by adding a simple passive filter outside the chip, an analog voltage
between 0 and (255/256) × VDD can be obtained (for the positive polarity case, or between
(1/256) × VDD and VDD for the negative polarity case). Other voltage values can be obtained by
adding active external circuitry.
For PWM mode, the frequency of the counter can be adjusted to four different values using field
PS of CONTRAST_CTR register.
44.6.3 LCD Interface

The LCD Controller interfaces with the LCD Module through the LCD Interface (Table 44-15 on
page 1057). The Controller supports the following interface configurations: 24-bit TFT single
scan, 16-bit STN Dual Scan Mono (Color), 8-bit STN Dual (Single) Scan Mono (Color), 4-bit sin-
gle scan Mono (Color).
A 4-bit single scan STN display uses 4 parallel data lines to shift data to successive single hori-
zontal lines one at a time until the entire frame has been shifted and transferred. The 4 LSB pins
of LCD Data Bus (LCDD [3:0]) can be directly connected to the LCD driver; the 20 MSB pins
(LCDD [23:4]) are not used.
1051
An 8-bit single scan STN display uses 8 parallel data lines to shift data to successive single hor-
izontal lines one at a time until the entire frame has been shifted and transferred. The 8 LSB pins
of LCD Data Bus (LCDD [7:0]) can be directly connected to the LCD driver; the 16 MSB pins
(LCDD [23:8]) are not used.
An 8-bit Dual Scan STN display uses two sets of 4 parallel data lines to shift data to successive
upper and lower panel horizontal lines one at a time until the entire frame has been shifted and
transferred. The bus LCDD[3:0] is connected to the upper panel data lines and the bus
LCDD[7:4] is connected to the lower panel data lines. The rest of the LCD Data Bus lines
(LCDD[23:8]) are not used.
A 16-bit Dual Scan STN display uses two sets of 8 parallel data lines to shift data to successive
upper and lower panel horizontal lines one at a time until the entire frame has been shifted and
transferred. The bus LCDD[7:0] is connected to the upper panel data lines and the bus
LCDD[15:8] is connected to the lower panel data lines. The rest of the LCD Data Bus lines
(LCDD[23:16]) are not used.
STN Mono displays require one bit of image data per pixel. STN Color displays require three bits
(Red, Green and Blue) of image data per pixel, resulting in a horizontal shift register of length
three times the number of pixels per horizontal line. This RGB or Monochrome data is shifted to
the LCD driver as consecutive bits via the parallel data lines.
A TFT single scan display uses up to 24 parallel data lines to shift data to successive horizontal
lines one at a time until the entire frame has been shifted and transferred. The 24 data lines are
divided in three bytes that define the color shade of each color component of each pixel. The
LCDD bus is split as LCDD[23:16] for the blue component, LCDD[15:8] for the green component
and LCDD[7:0] for the red component. If the LCD Module has lower color resolution (fewer bits
per color component), only the most significant bits of each component are used.
All these interfaces are shown in Figure 44-6 to Figure 44-10. Figure 44-6 on page 1052 shows
the 24-bit single scan TFT display timing; Figure 44-7 on page 1053 shows the 4-bit single scan
STN display timing for monochrome and color modes; Figure 44-8 on page 1054 shows the 8-bit
single scan STN display timing for monochrome and color modes; Figure 44-9 on page 1055
shows the 8-bit Dual Scan STN display timing for monochrome and color modes; Figure 44-10
on page 1056 shows the 16-bit Dual Scan STN display timing for monochrome and color modes.
Figure 44-6. TFT Timing (First Line Expanded View)

LCDVSYNC
LCDDEN
LCDHSYNC
LCDDOTCK
LCDD [24:16] B0 B1
LCDD [15:8] G0 G1
LCDD [7:0] R0 R1
1052 AT91SAM9G45
AT91SAM9G45
Figure 44-7. Single Scan Monochrome and Color 4-bit Panel Timing (First Line Expanded View)
LCDVSYNC
LCDDEN
LCDHSYNC
LCDDOTCK
LCDD [3] P0 P4
LCDD [2] P1 P5
LCDD [1] P2 P6
LCDD [0] P3 P7
LCDVSYNC
LCDDEN
LCDHSYNC
LCDDOTCK
LCDD [3] R0 G1
LCDD [2] G0 B1
LCDD [1] B0 R2
LCDD [0] R1 G2
1053
Figure 44-8. Single Scan Monochrome and Color 8-bit Panel Timing (First Line Expanded View)
LCDVSYNC
LCDDEN
LCDHSYNC
LCDDOTCK
LCDD [7] P0 P8
LCDD [6] P1 P9
LCDD [5] P2 P10
LCDD [4] P3 P11
LCDD [3] P4 P12
LCDD [2] P5 P13
LCDD [1] P6 P14
LCDD [0] P7 P15
LCDVSYNC
LCDDEN
LCDHSYNC
LCDDOTCK
LCDD [7] R0 B2
LCDD [6] G0 R3
LCDD [5] B0 G3
LCDD [4] R1 B3
LCDD [3] G1 R4
LCDD [2] B1 G4
LCDD [1] R2 B4
LCDD [0] G2 R5
1054 AT91SAM9G45
AT91SAM9G45
Figure 44-9. Dual Scan Monochrome and Color 8-bit Panel Timing (First Line Expanded View)
LCDVSYNC
LCDDEN
LCDHSYNC
LCDDOTCK
Lower Pane
LCDD [7] LP0 LP4
LCDD [6] LP1 LP5
LCDD [5] L2 LP6
LCDD [4] L3 LP7

Upper Pane
LCDD [3] UP0 UP4
LCDD [2] UP1 UP5
LCDD [1] UP2 UP6
LCDD [0] UP3 UP7
LCDVSYNC
LCDDEN
LCDHSYNC
LCDDOTCK
Lower Pane
LCDD [7] LR0 LG1
LCDD [6] LG0 LB1
LCDD [5] LB0 LR2
LCDD [4] LR1 LG2

Upper Pane
LCDD [3] UR0 UG1
LCDD [2] UG0 UB1
LCDD [1] UB0 UR2
LCDD [0] UR1 UG2
1055
Figure 44-10. Dual Scan Monochrome and Color 16-bit Panel Timing (First Line Expanded View)
LCDVSYNC
LCDDEN
LCDHSYNC
LCDDOTCK
Lower Panel
LCDD [15] LP0 LP8
LCDD [14] LP1 LP9
LCDD [13] LP2 LP10
LCDD [12] LP3 LP11
LCDD [11] LP4 LP12
LCDD [10] LP5 LP13
LCDD [9] LP6 LP14
LCDD [8] LP7 LP15

Upper Panel
LC DD [7] UP0 UP8
LCDD [6] UP1 UP9
LCDD [5] UP2 UP10
LCDD [4] UP3 UP11
LCDD [3] UP4 UP12
LCDD [2] UP5 UP13
LCDD [1] UP6 UP14
LCDD [0] UP7 UP15
LCDVSYNC
LCDDEN
LC DHSYNC
LCDDOTCK
Lower Panel
LCDD [15] LR0 LB2
LCDD [14] LG0 LR3
LCDD [13] LB0 LG3
LCDD [12] LR1 LB3
LCDD [11] LG1 LR4
LCDD [10] LB1 LG4
LCDD [9] LR2 LB4
LCDD [8] LG2 LR5

Upper Panel
LCDD [7] UR0 UB2
LCDD [6] UG0 UR3
LCDD [5] UB0 UG3
LCDD [4] UR1 UB3
LCDD [3] UG1 UR4
LCDD [2] UB1 UG4
LCDD [1] UR2 UB4
LCDD [0] UG2 UR5
1056 AT91SAM9G45
AT91SAM9G45
Table 44-15. LCD Signal Multiplexing

4-bit STN Single 8-bit STN Single 8-bit STN Dual 16-bit STN Dual
LCD Data Scan Scan Scan Scan
Bus (mono, color) (mono, color) (mono, color) (mono, color) 24-bit TFT 16-bit TFT
LCDD[23] LCD_BLUE7 LCD_BLUE4
LCDD[18] LCD_BLUE2
LCDD[17] LCD_BLUE1
LCDD[16] LCD_BLUE0
LCDD[15] LCDLP7 LCD_GREEN7 LCD_GREEN5
LCDD[9] LCDLP1 LCD_GREEN1
LCDD[8] LCDLP0 LCD_GREEN0
LCDD[7] LCD7 LCDLP3 LCDUP7 LCD_RED7 LCD_RED4
LCDD[3] LCD3 LCD3 LCDUP3 LCDUP3 LCD_RED3 LCD_RED0
LCDD[2] LCD2 LCD2 LCDUP2 LCDUP2 LCD_RED2
1057
44.7 Interrupts
The LCD Controller generates six different IRQs. All the IRQs are synchronized with the internal
LCD Core Clock. The IRQs are:
• DMA Memory error IRQ. Generated when the DMA receives an error response from an AHB
slave while it is doing a data transfer.
• FIFO underflow IRQ. Generated when the Serializer tries to read a word from the FIFO when
the FIFO is empty.
• FIFO overwrite IRQ. Generated when the DMA Controller tries to write a word in the FIFO
while the FIFO is full.
• DMA end of frame IRQ. Generated when the DMA controller updates the Frame Base
Address pointers. This IRQ can be used to implement a double-buffer technique. For more
information, see “Double-buffer Technique” on page 1059.
• End of Line IRQ. This IRQ is generated when the LINEBLANK period of each line is reached
and the DMA Controller is in inactive state.
• End of Last Line IRQ. This IRQ is generated when the LINEBLANK period of the last line of
the current frame is reached and the DMA Controller is in inactive state.
Each IRQ can be individually enabled, disabled or cleared, in the LCD_IER (Interrupt Enable
Register), LCD_IDR (Interrupt Disable Register) and LCD_ICR (Interrupt Clear Register) regis-
ters. The LCD_IMR register contains the mask value for each IRQ source and the LDC_ISR
contains the status of each IRQ source. A more detailed description of these registers can be
found in “LCD Controller (LCDC) User Interface” on page 1063.
44.8 Configuration Sequence

The DMA Controller starts to transfer image data when the LCDC Core is activated (Write to
LCD_PWR field of PWRCON register). Thus, the user should configure the LCDC Core and
configure and enable the DMA Controller prior to activation of the LCD Controller. In addition,
the image data to be shows should be available when the LCDC Core is activated, regardless of
the value programmed in the GUARD_TIME field of the PWRCON register.
To disable the LCD Controller, the user should disable the LCDC Core and then disable the
DMA Controller. The user should not enable LIP again until the LCDC Core is in IDLE state. This
is checked by reading the LCD_BUSY bit in the PWRCON register.
The initialization sequence that the user should follow to make the LCDC work is:
• Create or copy the first image to show in the display buffer memory.
• If a palletized mode is used, create and store a palette in the internal LCD Palette
memory(See “Palette” on page 1042.
• Configure the LCD Controller Core without enabling it:
– LCDCON1 register: Program the CLKVAL and BYPASS fields: these fields control the
pixel clock divisor that is used to generate the pixel clock LCDDOTCK. The value to
program depends on the LCD Core clock and on the type and size of the LCD
Module used. There is a minimum value of the LCDDOTCK clock period that
depends on the LCD Controller Configuration, this minimum value can be found in
Table 44-14 on page 1047. The equations that are used to calculate the value of the
pixel clock divisor can be found at the end of the section “Timegen” on page 1046
1058 AT91SAM9G45
AT91SAM9G45
– LCDCON2 register: Program its fields following their descriptions in the LCD
Controller User Interface section below and considering the type of LCD module
used and the desired working mode. Consider that not all combinations are possible.
– LCDTIM1 and LCDTIM2 registers: Program their fields according to the datasheet of
the LCD module used and with the help of the Timegen section in page 10. Note that
some fields are not applicable to STN modules and must be programmed with 0
values. Note also that there is a limitation on the minimum value of VHDLY, HPW,
HBP that depends on the configuration of the LCDC.
– LCDFRMCFG register: program the dimensions of the LCD module used.
– LCDFIFO register: To program it, use the formula in section “FIFO” on page 1040
– DP1_2 to DP6_7 registers: they are only used for STN displays. They contain the
dithering patterns used to generate gray shades or colors in these modules. They
are loaded with recommended patterns at reset, so it is not necessary to write
anything on them. They can be used to improve the image quality in the display by
tuning the patterns in each application.
– PWRCON Register: this register controls the power-up sequence of the LCD, so
take care to use it properly. Do not enable the LCD (writing a 1 in LCD_PWR field)
until the previous steps and the configuration of the DMA have been finished.
– CONTRAST_CTR and CONTRAST_VAL: use this registers to adjust the contrast of
the display, when the LCDCC line is used.
• Configure the DMA Controller. The user should configure the base address of the display
buffer memory, the size of the AHB transaction and the size of the display image in memory.
When the DMA is configured the user should enable the DMA. To do so the user should
configure the following registers:
– DMABADDR1 and DMABADDR2 registers: In single scan mode only DMABADDR1
register must be configured with the base address of the display buffer in memory. In
dual scan mode DMABADDR1 should be configured with the base address of the
Upper Panel display buffer and DMABADDR2 should be configured with the base
address of the Lower Panel display buffer.
– DMAFRMCFG register: Program the FRMSIZE field. Note that in dual scan mode
the vertical size to use in the calculation is that of each panel. Respect to the
BRSTLN field, a recommended value is a 4-word burst.
– DMACON register: Once both the LCD Controller Core and the DMA Controller have
been configured, enable the DMA Controller by writing a “1” to the DMAEN field of
this register. If using a dual scan module or the 2D addressing feature, do not forget
to write the DMAUPDT bit after every change to the set of DMA configuration values.
– DMA2DCFG register: Required only in 2D memory addressing mode (see “2D
Memory Addressing” on page 1060).
• Finally, enable the LCD Controller Core by writing a “1” in the LCD_PWR field of the
PWRCON register and do any other action that may be required to turn the LCD module on.
44.9 Double-buffer Technique

The double-buffer technique is used to avoid flickering while the frame being displayed is
updated. Instead of using a single buffer, there are two different buffers, the backbuffer (back-
ground buffer) and the primary buffer (the buffer being displayed).
1059
The host updates the backbuffer while the LCD Controller is displaying the primary buffer. When
the backbuffer has been updated the host updates the DMA Base Address registers.
When using a Dual Panel LCD Module, both base address pointers should be updated in the
same frame. There are two possibilities:
• Check the DMAFRMPTx register to ensure that there is enough time to update the DMA
Base Address registers before the end of frame.
• Update the Frame Base Address Registers when the End Of Frame IRQ is generated.
Once the host has updated the Frame Base Address Registers and the next DMA end of frame
IRQ arrives, the backbuffer and the primary buffer are swapped and the host can work with the
new backbuffer.
When using a dual-panel LCD module, both base address pointers should be updated in the
same frame. In order to achieve this, the DMAUPDT bit in DMACON register must be used to
validate the new base address.
44.10 2D Memory Addressing

The LCDC can be configured to work on a frame buffer larger than the actual screen size. By
changing the values in a few registers, it is easy to move the displayed area along the frame buf-
fer width and height.
Figure 44-11. Frame Buffer Addressing

Frame Buffer
Displayed Image
Base word address &

pixel offset
Line-to-line
address increment
In order to locate the displayed window within a larger frame buffer, the software must:
• Program the DMABADDR1 (DMABADDR2) register(s) to make them point to the word
containing the first pixel of the area of interest.
• Program the PIXELOFF field of DMA2DCFG register to specify the offset of this first pixel
within the 32-bit memory word that contains it.
• Define the width of the complete frame buffer by programming in the field ADDRINC of
DMA2DCFG register the address increment between the last word of a line and the first word
of the next line (in number of 32-bit words).
• Enable the 2D addressing mode by writing the DMA2DEN bit in DMACON register. If this bit
is not activated, the values in the DMA2DCFG register are not considered and the controller
assumes that the displayed area occupies a continuous portion of the memory.
1060 AT91SAM9G45
AT91SAM9G45
The above configuration can be changed frame to frame, so the displayed window can be
moved rapidly. Note that the FRMSIZE field of DMAFRMCFG register must be updated with any
movement of the displaying window. Note also that the software must write bit DMAUPDT in
DMACON register after each configuration for it to be accepted by LCDC.
Note: In 24 bpp packed mode, the DMA base address must point to a word containing a complete pixel
(possible values of PIXELOFF are 0 and 8). This means that the horizontal origin of the displaying
window must be a multiple of 4 pixels or a multiple of 4 pixels minus 1 (x = 4n or x = 4n-1, valid ori-
gins are pixel 0,3,4,7,8,11,12, etc.).
1061
44.11 Register Configuration Guide
Program the PIO Controller to enable LCD signals.
Enable the LCD controller clock in the Power Management Controller.
44.11.1 STN Mode Example

STN color(R,G,B) 320*240, 8-bit single scan, 70 frames/sec, Master clock = 60 Mhz
Data rate: 320*240*70*3/8 = 2.016 MHz
HOZVAL= ((3*320)/8) - 1
LINEVAL= 240 -1
CLKVAL = (60 MHz/2.016 MHz) - 1 = 29
LCDCON1= CLKVAL << 12
LCDCON2 = LITTLEENDIAN | SINGLESCAN | STNCOLOR | DISP8BIT| PS8BPP;
LCDTIM1 = 0;
LCDTIM2 = 10 | (10 << 21);
LCDFRMCFG = (HOZVAL << 21) | LINEVAL;
DMAFRMCFG = (7 << 24) + (320 * 240 * 8) / 32;
44.11.2 TFT Mode Example

This example is based on the NEC TFT color LCD module NL6448BC20-08.
TFT 640*480, 16-bit single scan, 60 frames/sec, pixel clock frequency = [21MHz..29MHz] with a
typical value = 25.175 MHz.
The Master clock must be (n + 1)*pixel clock frequency
HOZVAL = 640 - 1
LINEVAL = 480 - 1
If Master clock is 100 MHz
CLKVAL = (100 MHz/ 25.175 MHz) - 1= 3
VFP = (12 -1), VBP = (31-1), VPW = (2-1), VHDLY= (2-1)
HFP = (16-1), HBP = (48 -1), HPW= (96-1)
LCDCON1= CLKVAL << 12

LCDCON2 = LITTLEENDIAN | CLKMOD | INVERT_CLK | INVERT_LINE | INVERT_FRM |
PS16BPP | SINGLESCAN | TFT
LCDTIM1 = VFP | (VBP << 8) | (VPW << 16) | (VHDLY << 24)
LCDTIM2 = HBP | (HPW << 8) | (HFP << 21)
LCDFRMCFG = (HOZVAL << 21) | LINEVAL
DMAFRMCFG = (7 << 24) + (640 * 480* 16) / 32;
1062 AT91SAM9G45
AT91SAM9G45
44.12 LCD Controller (LCDC) User Interface

0x0 DMA Base Address Register 1 DMABADDR1 Read-write 0x00000000
0x4 DMA Base Address Register 2 DMABADDR2 Read-write 0x00000000
0x8 DMA Frame Pointer Register 1 DMAFRMPT1 Read-only 0x00000000
0xC DMA Frame Pointer Register 2 DMAFRMPT2 Read-only 0x00000000
0x10 DMA Frame Address Register 1 DMAFRMADD1 Read-only 0x00000000
0x14 DMA Frame Address Register 2 DMAFRMADD2 Read-only 0x00000000
0x18 DMA Frame Configuration Register DMAFRMCFG Read-write 0x00000000
0x1C DMA Control Register DMACON Read-write 0x00000000
0x20 DMA Control Register DMA2DCFG Read-write 0x00000000
0x800 LCD Control Register 1 LCDCON1 Read-write 0x00002000
0x804 LCD Control Register 2 LCDCON2 Read-write 0x00000000
0x808 LCD Timing Register 1 LCDTIM1 Read-write 0x00000000
0x80C LCD Timing Register 2 LCDTIM2 Read-write 0x00000000
0x810 LCD Frame Configuration Register LCDFRMCFG Read-write 0x00000000
0x814 LCD FIFO Register LCDFIFO Read-write 0x00000000
0x81C Dithering Pattern DP1_2 DP1_2 Read-write 0xA5
0x820 Dithering Pattern DP4_7 DP4_7 Read-write 0x5AF0FA5
0x824 Dithering Pattern DP3_5 DP3_5 Read-write 0xA5A5F
0x828 Dithering Pattern DP2_3 DP2_3 Read-write 0xA5F
0x82C Dithering Pattern DP5_7 DP5_7 Read-write 0xFAF5FA5
0x830 Dithering Pattern DP3_4 DP3_4 Read-write 0xFAF5
0x834 Dithering Pattern DP4_5 DP4_5 Read-write 0xFAF5F
0x838 Dithering Pattern DP6_7 DP6_7 Read-write 0xF5FFAFF
0x83C Power Control Register PWRCON Read-write 0x0000000e
0x840 Contrast Control Register CONTRAST_CTR Read-write 0x00000000
0x844 Contrast Value Register CONTRAST_VAL Read-write 0x00000000
0x848 LCD Interrupt Enable Register LCD_IER Write-only 0x0
0x84C LCD Interrupt Disable Register LCD_IDR Write-only 0x0
0x850 LCD Interrupt Mask Register LCD_IMR Read-only 0x0
0x854 LCD Interrupt Status Register LCD_ISR Read-only 0x0
0x858 LCD Interrupt Clear Register LCD_ICR Write-only 0x0
0x860 LCD Interrupt Test Register LCD_ITR Write-only 0
0x864 LCD Interrupt Raw Status Register LCD_IRR Read-only 0
0x8E4 Write Protection Control Register LCD_WPCR Read-write 0
1063
0x8E8 Write Protection Status Register LCD_WPSR Read-only 0
0xC00 Palette entry 0 LUT ENTRY 0 Read-write
0xC0C Palette entry 3 LUT ENTRY 3 Read-write
… …
0xFFC Palette entry 255 LUT ENTRY 255 Read-write
1064 AT91SAM9G45
AT91SAM9G45
44.12.1 DMA Base Address Register 1

Name: DMABADDR1
Address:0x00500000
Access: Read-write
Reset value: 0x00000000
31 30 29 28 27 26 25 24
BADDR-U
23 22 21 20 19 18 17 16
BADDR-U
15 14 13 12 11 10 9 8
BADDR-U
7 6 5 4 3 2 1 0
BADDR-U 0 0
• BADDR-U
Base Address for the upper panel in dual scan mode. Base Address for the complete frame in single scan mode.
If a dual scan configuration is selected in LCDCON2 register or bit DMA2DEN in register DMACON is set, the bit
DMAUPDT in that same register must be written after writing any new value to this field in order to make the DMA controller
use this new value.
44.12.2 DMA Base Address Register 2

Name: DMABADDR2
Address:0x00500004
Access: Read-write
31 30 29 28 27 26 25 24
BADDR-L
23 22 21 20 19 18 17 16
BADDR-L
15 14 13 12 11 10 9 8
BADDR-L
7 6 5 4 3 2 1 0
BADDR-L
• BADDR-L
Base Address for the lower panel in dual scan mode only.
use this new value.
1065
44.12.3 DMA Frame Pointer Register 1
Name: DMAFRMPT1
Address:0x00500008
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– FRMPT-U
15 14 13 12 11 10 9 8
FRMPT-U
7 6 5 4 3 2 1 0
FRMPT-U
• FRMPT-U
Current value of frame pointer for the upper panel in dual scan mode. Current value of frame pointer for the complete frame
in single scan mode. Down count from FRMSIZE to 0.
Note: This register is read-only and contains the current value of the frame pointer (number of words to the end of the frame). It can be
used as an estimation of the number of words transferred from memory for the current frame.
44.12.4 DMA Frame Pointer Register 2

Name: DMAFRMPT2
Address:0x0050000C
Access: Read-only
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
- FRMPT-L
15 14 13 12 11 10 9 8
FRMPT-L
7 6 5 4 3 2 1 0
FRMPT-L
• FRMPT-L
Current value of frame pointer for the Lower panel in dual scan mode only. Down count from FRMSIZE to 0.
Note: This register is read-only and contains the current value of the frame pointer (number of words to the end of the frame). It can be
used as an estimation of the number of words transferred from memory for the current frame.
1066 AT91SAM9G45
AT91SAM9G45
44.12.5 DMA Frame Address Register 1

Name: DMAFRMADD1
Address:0x00500010
Access: Read-only
31 30 29 28 27 26 25 24
FRMADD-U
23 22 21 20 19 18 17 16
FRMADD-U
15 14 13 12 11 10 9 8
FRMADD-U
7 6 5 4 3 2 1 0
FRMADD-U
• FRMADD-U
Current value of frame address for the upper panel in dual scan mode. Current value of frame address for the complete
frame in single scan.
Note: This register is read-only and contains the current value of the last DMA transaction in the bus for the panel/frame.
44.12.6 DMA Frame Address Register 2

Name: DMAFRMADD2
Address:0x00500014
Access: Read-only
31 30 29 28 27 26 25 24
FRMADD-L
23 22 21 20 19 18 17 16
FRMADD-L
15 14 13 12 11 10 9 8
FRMADD-L
7 6 5 4 3 2 1 0
FRMADD-L
• FRMADD-L
Current value of frame address for the lower panel in single scan mode only.
Note: This register is read-only and contains the current value of the last DMA transaction in the bus for the panel.
1067
44.12.7 DMA Frame Configuration Register
Name: DMAFRMCFG
Address:0x00500018
Access: Read-write
31 30 29 28 27 26 25 24
– BRSTLN
23 22 21 20 19 18 17 16
– FRMSIZE
15 14 13 12 11 10 9 8
FRMSIZE
7 6 5 4 3 2 1 0
FRMSIZE
• FRMSIZE: Frame Size

In single scan mode, this is the frame size in words. In dual scan mode, this is the size of each panel.
use this new value.
• BRSTLN: Burst Length in Words

Program with the desired burst length - 1
1068 AT91SAM9G45
AT91SAM9G45
44.12.8 DMA Control Register

Name: DMACON
Address:0x0050001C
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – DMA2DEN DMAUPDT DMABUSY DMARST DMAEN
• DMAEN: DMA Enable

0: DMA is disabled.
1: DMA is enabled.
• DMARST: DMA Reset (Write-only)

0: No effect.
1: Reset DMA module. DMA Module should be reset only when disabled and in idle state.
• DMABUSY: DMA Busy

0: DMA module is idle.
1: DMA module is busy (doing a transaction on the AHB bus).
• DMAUPDT: DMA Configuration Update

0: No effect
1: Update DMA Configuration. Used for simultaneous updating of DMA parameters in dual scan mode or when using 2D
addressing. The values written in the registers DMABADDR1, DMABADDR2 and DMA2DCFG, and in the field FRMSIZE of
register DMAFRMCFG, are accepted by the DMA controller and are applied at the next frame. This bit is used only if a dual
scan configuration is selected (bit SCANMOD of LCDCON2 register) or 2D addressing is enabled (bit DMA2DEN in this
register). Otherwise, the LCD controller accepts immediately the values written in the registers referred to above.
• DMA2DEN: DMA 2D Addressing Enable

0: 2D addressing is disabled (values in register DMA2DCFG are “don’t care”).
1: 2D addressing is enabled.
1069
44.12.9 LCD DMA 2D Addressing Register
Name: DMA2DCFG
Address:0x00500020
Access: Read-write
31 30 29 28 27 26 25 24
– – – PIXELOFF
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
ADDRINC
7 6 5 4 3 2 1 0
ADDRINC
• ADDRINC: DMA 2D Addressing Address increment

When 2-D DMA addressing is enabled (bit DMA2DEN is set in register DMACON), this field specifies the number of bytes
that the DMA controller must jump between screen lines. Itb must be programmed as: [({address of first 32-bit word in a
screen line} - {address of last 32-bit word in previous line})]. In other words, it is equal to 4*[number of 32-bit words occu-
pied by each line in the complete frame buffer minus the number of 32-bit words occupied by each displayed line]. Bit
DMAUPDT in register DMACON must be written after writing any new value to this field in order to make the DMA control-
ler use this new value.
• PIXELOFF: DAM2D Addressing Pixel offset

When 2D DMA addressing is enabled (bit DMA2DEN is set in register DMACON), this field specifies the offset of the first
pixel in each line within the memory word that contains this pixel. The offset is specified in number of bits in the range 0-31,
so for example a value of 4 indicates that the first pixel in the screen starts at bit 4 of the 32-bit word pointed by register
DMABADDR1. Bits 0 to 3 of that word are not used. This example is valid for little endian memory organization. When
using big endian memory organization, this offset is considered from bit 31 downwards, or equivalently, a given value of
this field always selects the pixel in the same relative position within the word, independently of the memory ordering con-
figuration. Bit DMAUPDT in register DMACON must be written after writing any new value to this field in order to make the
DMA controller use this new value.
1070 AT91SAM9G45
AT91SAM9G45
44.12.10 LCD Control Register 1

Name: LCDCON1
Address:0x00500800
Access: Read-write, except LINECNT: Read-only
31 30 29 28 27 26 25 24
LINECNT
23 22 21 20 19 18 17 16
LINECNT CLKVAL
15 14 13 12 11 10 9 8
CLKVAL – – – –
7 6 5 4 3 2 1 0
– – – – – – – BYPASS
• BYPASS: Bypass LCDDOTCK Divider

0: The divider is not bypassed. LCDDOTCK frequency defined by the CLKVAL field.
1: The LCDDOTCK divider is bypassed. LCDDOTCK frequency is equal to the LCDC Clock frequency.
• CLKVAL: Clock Divider

9-bit divider for pixel clock (LCDDOTCK) frequency.
Pixel_clock = system_clock ⁄ ( CLKVAL + 1 )
• LINECNT: Line Counter (Read-only)

Current Value of 11-bit line counter. Down count from LINEVAL to 0.
1071
44.12.11 LCD Control Register 2
Name: LCDCON2
Address:0x00500804
Access: Read-write
31 30 29 28 27 26 25 24
MEMOR – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
CLKMOD – – INVDVAL INVCLK INVLINE INVFRAME INVVD
7 6 5 4 3 2 1 0
PIXELSIZE IFWIDTH SCANMOD DISTYPE
• DISTYPE: Display Type
DISTYPE
0 0 STN Monochrome
0 1 STN Color
1 0 TFT
1 1 Reserved
• SCANMOD: Scan Mode

0: Single Scan
1: Dual Scan
• IFWIDTH: Interface width (STN)
IFWIDTH
0 0 4-bit (Only valid in single scan STN mono or color)
0 1 8-bit (Only valid in STN mono or Color)
1 0 16-bit (Only valid in dual scan STN mono or color)
1 1 Reserved
1072 AT91SAM9G45
AT91SAM9G45
• PIXELSIZE: Bits per pixel
PIXELSIZE
0 0 0 1 bit per pixel
0 0 1 2 bits per pixel
1 0 1 24 bits per pixel, packed (Only valid in TFT mode)
1 1 0 24 bits per pixel, unpacked (Only valid in TFT mode)
1 1 1 Reserved
• INVVD: LCDD polarity

0: Normal
1: Inverted
• INVFRAME: LCDVSYNC polarity

0: Normal (active high)
1: Inverted (active low)
• INVLINE: LCDHSYNC polarity

• INVCLK: LCDDOTCK polarity

0: Normal (LCDD fetched at LCDDOTCK falling edge)
1: Inverted (LCDD fetched at LCDDOTCK rising edge)
• INVDVAL: LCDDEN polarity

• CLKMOD: LCDDOTCK mode

0: LCDDOTCK only active during active display period
1: LCDDOTCK always active
• MEMOR: Memory Ordering Format

00: Big Endian
10: Little Endian
11: WinCE format
1073
44.12.12 LCD Timing Configuration Register 1
Name: LCDTIM1
Address:0x00500808
Access: Read-write
31 30 29 28 27 26 25 24
– – – – VHDLY
23 22 21 20 19 18 17 16
– – VPW
15 14 13 12 11 10 9 8
VBP
7 6 5 4 3 2 1 0
VFP
• VFP: Vertical Front Porch

In TFT mode, these bits equal the number of idle lines at the end of the frame.
In STN mode, these bits should be set to 0.
• VBP: Vertical Back Porch

In TFT mode, these bits equal the number of idle lines at the beginning of the frame.
• VPW: Vertical Synchronization pulse width

In TFT mode, these bits equal the vertical synchronization pulse width, given in number of lines. LCDVSYNC width is equal
to (VPW+1) lines.
• VHDLY: Vertical to horizontal delay

In TFT mode, this is the delay between LCDVSYNC rising or falling edge and LCDHSYNC rising edge. Delay is
(VHDLY+1) LCDDOTCK cycles.
1074 AT91SAM9G45
AT91SAM9G45
44.12.13 LCD Timing Configuration Register 2

Name: LCDTIM2
Address:0x0050080C
Access: Read-write
31 30 29 28 27 26 25 24
HFP
23 22 21 20 19 18 17 16
HFP – – – – –
15 14 13 12 11 10 9 8
– – HPW
7 6 5 4 3 2 1 0
HBP
• HBP: Horizontal Back Porch

Number of idle LCDDOTCK cycles at the beginning of the line. Idle period is (HBP+1) LCDDOTCK cycles.
• HPW: Horizontal synchronization pulse width

Width of the LCDHSYNC pulse, given in LCDDOTCK cycles. Width is (HPW+1) LCDDOTCK cycles.
• HFP: Horizontal Front Porch

Number of idle LCDDOTCK cycles at the end of the line. Idle period is (HFP+1) LCDDOTCK cycles.
44.12.14 LCD Frame Configuration Register

Name: LCDFRMCFG
Address:0x00500810
Access: Read-write
31 30 29 28 27 26 25 24
LINESIZE
23 22 21 20 19 18 17 16
LINESIZE – – – – –
15 14 13 12 11 10 9 8
– – – – – LINEVAL
7 6 5 4 3 2 1 0
LINEVAL
• LINEVAL: Vertical size of LCD module

In single scan mode: vertical size of LCD Module, in pixels, minus 1
In dual scan mode: vertical display size of each LCD panel, in pixels, minus 1
• LINESIZE: Horizontal size of LCD module, in pixels, minus 1
1075
44.12.15 LCD FIFO Register
Name: LCDFIFO
Address:0x00500814
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
FIFOTH
7 6 5 4 3 2 1 0
FIFOTH
• FIFOTH: FIFO Threshold

Must be programmed with:
FIFOTH (in Words) = 512 - (2 x DMA_BURST_LENGTH + 3)
where:
• 512 is the effective size of the FIFO in Words. It is the total FIFO memory size in single scan mode and half that size in
dual scan mode.
• DMA_burst_length is the burst length of the transfers made by the DMA (in Words). Refer to “BRSTLN: Burst Length in
Words” on page 1068.
1076 AT91SAM9G45
AT91SAM9G45
44.12.16 Dithering Pattern DP1_2 Register

Name: DP1_2
Address:0x0050081C
Access: Read-write
Reset value: 0xA5
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
DP1_2
• DP1_2: Pattern value for ½ duty cycle

Name: DP4_7
Address:0x00500820
Access: Read-write
Reset value: 0x5AF0FA5
31 30 29 28 27 26 25 24
– – – – DP4_7
23 22 21 20 19 18 17 16
DP4_7
15 14 13 12 11 10 9 8
DP4_7
7 6 5 4 3 2 1 0
DP4_7
• DP4_7: Pattern value for 4/7 duty cycle
1077
Name: DP3_5
Address:0x00500824
Access: Read-write
Reset value: 0xA5A5F
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – DP3_5
15 14 13 12 11 10 9 8
DP3_5
7 6 5 4 3 2 1 0
DP3_5

Name: DP2_3: Dithering Pattern DP2_3 Register
Address:0x00500828
Access: Read-write
Reset value: 0xA5F
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – DP2_3
7 6 5 4 3 2 1 0
DP2_3
1078 AT91SAM9G45
AT91SAM9G45

Name: DP5_7:
Address:0x0050082C
Access: Read-write
Reset value: 0xFAF5FA5
31 30 29 28 27 26 25 24
– – – – DP5_7
23 22 21 20 19 18 17 16
DP5_7
15 14 13 12 11 10 9 8
DP5_7
7 6 5 4 3 2 1 0
DP5_7

Name: DP3_4
Address:0x00500830
Access: Read-write
Reset value: 0xFAF5
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
DP3_4
7 6 5 4 3 2 1 0
DP3_4
1079
AT91SAM9G45

Name: DP4_5
Address:0x00500834
Access: Read-write
Reset value: 0xFAF5F
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – DP4_5
15 14 13 12 11 10 9 8
DP4_5
7 6 5 4 3 2 1 0
DP4_5

Name: DP6_7
Address:0x00500838
Access: Read-write
Reset value: 0xF5FFAFF
31 30 29 28 27 26 25 24
– – – – DP6_7
23 22 21 20 19 18 17 16
DP6_7
15 14 13 12 11 10 9 8
DP6_7
7 6 5 4 3 2 1 0
DP6_7
1080
AT91SAM9G45
44.12.24 Power Control Register

Name: PWRCON
Address:0x0050083C
Access: Read-write
Reset value: 0x0000000e
31 30 29 28 27 26 25 24
LCD_BUSY – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
GUARD_TIME LCD_PWR
• LCD_PWR: LCD module power control

0 = lcd_pwr signal is low, other lcd_* signals are low.
0->1 = lcd_* signals activated, lcd_pwr is set high with the delay of GUARD_TIME frame periods.
1 = lcd_pwr signal is high, other lcd_* signals are active.
1->0 = lcd_pwr signal is low, other lcd_* signals are active, but are set low after GUARD_TIME frame periods.
• GUARD_TIME
Delay in frame periods between applying control signals to the LCD module and setting LCD_PWR high, and between set-
ting LCD_PWR low and removing control signals from LCD module
• LCD_BUSY
Read-only field. If 1, it indicates that the LCD is busy (active and displaying data, in power on sequence or in power off
sequence).
1081
AT91SAM9G45
44.12.25 Contrast Control Register

Name: CONTRAST_CTR
Address:0x00500840
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– – – – ENA POL PS
• PS
This 2-bit value selects the configuration of a counter prescaler. The meaning of each combination is as follows:
PS
0 0 The counter advances at a rate of fCOUNTER = fLCDC_CLOCK.
0 1 The counter advances at a rate of fCOUNTER = fLCDC_CLOCK/2.
• POL
This bit defines the polarity of the output. If 1, the output pulses are high level (the output will be high whenever the value in
the counter is less than the value in the compare register CONSTRAST_VAL). If 0, the output pulses are low level.
• ENA
When 1, this bit enables the operation of the PWM generator. When 0, the PWM counter is stopped.
1082
AT91SAM9G45
44.12.26 Contrast Value Register

Name: CONSTRAST_VAL
Access: Read-write
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
CVAL
• CVAL
PWM compare value. Used to adjust the analog value obtained after an external filter to control the contrast of the display.
1083
AT91SAM9G45
44.12.27 LCD Interrupt Enable Register

Name: LCD_IER
Address:0x00500848
Access: Write-only
Reset value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– MERIE OWRIE UFLWIE - EOFIE LSTLNIE LNIE
• LNIE: Line interrupt enable

0: No effect
1: Enable each line interrupt
• LSTLNIE: Last line interrupt enable

0: No effect
1: Enable last line interrupt
• EOFIE: DMA End of frame interrupt enable

0: No effect
1: Enable End Of Frame interrupt
• UFLWIE: FIFO underflow interrupt enable

0: No effect
1: Enable FIFO underflow interrupt
• OWRIE: FIFO overwrite interrupt enable

0: No effect
1: Enable FIFO overwrite interrupt
• MERIE: DMA memory error interrupt enable

0: No effect
1: Enable DMA memory error interrupt
1084
AT91SAM9G45
44.12.28 LCD Interrupt Disable Register

Name: LCD_IDR
Address:0x0050084C
Access: Write-only
Reset value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– MERID OWRID UFLWID – EOFID LSTLNID LNID
• LNID: Line interrupt disable

0: No effect
1: Disable each line interrupt
• LSTLNID: Last line interrupt disable

0: No effect
1: Disable last line interrupt
• EOFID: DMA End of frame interrupt disable

0: No effect
1: Disable End Of Frame interrupt
• UFLWID: FIFO underflow interrupt disable

0: No effect
1: Disable FIFO underflow interrupt
• OWRID: FIFO overwrite interrupt disable

0: No effect
1: Disable FIFO overwrite interrupt
• MERID: DMA Memory error interrupt disable

0: No effect
1: Disable DMA Memory error interrupt
1085
AT91SAM9G45
44.12.29 LCD Interrupt Mask Register

Name: LCD_IMR
Address:0x00500850
Access: Read-only
Reset value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– MERIM OWRIM UFLWIM – EOFIM LSTLNIM LNIM
• LNIM: Line interrupt mask

0: Line Interrupt disabled
1: Line interrupt enabled
• LSTLNIM: Last line interrupt mask

0: Last Line Interrupt disabled
1: Last Line Interrupt enabled
• EOFIM: DMA End of frame interrupt mask

0: End Of Frame interrupt disabled
1: End Of Frame interrupt enabled
• UFLWIM: FIFO underflow interrupt mask

0: FIFO underflow interrupt disabled
1: FIFO underflow interrupt enabled
• OWRIM: FIFO overwrite interrupt mask

0: FIFO overwrite interrupt disabled
1: FIFO overwrite interrupt enabled
• MERIM: DMA Memory error interrupt mask

0: DMA Memory error interrupt disabled
1: DMA Memory error interrupt enabled
1086
AT91SAM9G45
44.12.30 LCD Interrupt Status Register

Name: LCD_ISR
Address:0x00500854
Access: Read-only
Reset value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– MERIS OWRIS UFLWIS – EOFIS LSTLNIS LNIS
• LNIS: Line interrupt status

0: Line Interrupt not active
1: Line Interrupt active
• LSTLNIS: Last line interrupt status

0: Last Line Interrupt not active
1: Last Line Interrupt active
• EOFIS: DMA End of frame interrupt status

0: End Of Frame interrupt not active
1: End Of Frame interrupt active
• UFLWIS: FIFO underflow interrupt status

0: FIFO underflow interrupt not active
1: FIFO underflow interrupt active
• OWRIS: FIFO overwrite interrupt status

0: FIFO overwrite interrupt not active
1: FIFO overwrite interrupt active
• MERIS: DMA Memory error interrupt status

0: DMA Memory error interrupt not active
1: DMA Memory error interrupt active
1087
AT91SAM9G45
44.12.31 LCD Interrupt Clear Register

Name: LCD_ICR
Address:0x00500858
Access: Write-only
Reset value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– MERIC OWRIC UFLWIC – EOFIC LSTLNIC LNIC
• LNIC: Line interrupt clear

0: No effect
1: Clear each line interrupt
• LSTLNIC: Last line interrupt clear

0: No effect
1: Clear Last line Interrupt
• EOFIC: DMA End of frame interrupt clear

0: No effect
1: Clear End Of Frame interrupt
• UFLWIC: FIFO underflow interrupt clear

0: No effect
1: Clear FIFO underflow interrupt
• OWRIC: FIFO overwrite interrupt clear

0: No effect
1: Clear FIFO overwrite interrupt
• MERIC: DMA Memory error interrupt clear

0: No effect
1: Clear DMA Memory error interrupt
1088
AT91SAM9G45
44.12.32 LCD Interrupt Test Register

Name: LCD_ITR
Address:0x00500860
Access: Write-only
Reset value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– MERIT OWRIT UFLWIT – EOFIT LSTLNIT LNIT
• LNIT: Line interrupt test

0: No effect
1: Set each line interrupt
• LSTLNIT: Last line interrupt test

0: No effect
1: Set Last line interrupt
• EOFIT: DMA End of frame interrupt test

0: No effect
1: Set End Of Frame interrupt
• UFLWIT: FIFO underflow interrupt test

0: No effect
1: Set FIFO underflow interrupt
• OWRIT: FIFO overwrite interrupt test

0: No effect
1: Set FIFO overwrite interrupt
• MERIT: DMA Memory error interrupt test

0: No effect
1: Set DMA Memory error interrupt
1089
AT91SAM9G45
44.12.33 LCD Interrupt Raw Status Register

Name: LCD_IRR
Address:0x00500864
Access: Write-only
Reset value: 0x0
31 30 29 28 27 26 25 24
– – – – – – – –
23 22 21 20 19 18 17 16
– – – – – – – –
15 14 13 12 11 10 9 8
– – – – – – – –
7 6 5 4 3 2 1 0
– MERIR OWRIR UFLWIR – EOFIR LSTLNIR LNIR
• LNIR: Line interrupt raw status

0: No effect
1: Line interrupt condition present
• LSTLNIR: Last line interrupt raw status

0: No effect
1: Last line Interrupt condition present
• EOFIR: DMA End of frame interrupt raw status

0: No effect
1: End Of Frame interrupt condition present
• UFLWIR: FIFO underflow interrupt raw status

0: No effect
1: FIFO underflow interrupt condition present
• OWRIR: FIFO overwrite interrupt raw status

0: No effect
1: FIFO overwrite interrupt condition present
• MERIR: DMA Memory error interrupt raw status

0: No effect
1: DMA Memory error interrupt condition present
1090
AT91SAM9G45
44.12.34 LCD Write Protect Mode Register

Register Name: LCD_WPMR
31 30 29 28 27 26 25 24
WPKEY
23 22 21 20 19 18 17 16
WPKEY
15 14 13 12 11 10 9 8
WPKEY
7 6 5 4 3 2 1 0
— — — — — — — WPEN
• WPEN: Write Protect Enable

0 = Disables the Write Protect if WPKEY corresponds to0x4C4344 ("LCD" in ASCII).
1 = Enables the Write Protect if WPKEY corresponds to0x4C4344 ("LCD" in ASCII).
Protects the registers:
• “LCD Control Register 1” on page 1071
• “LCD Control Register 2” on page 1072
• “LCD Timing Configuration Register 1” on page 1074
• “LCD Timing Configuration Register 2” on page 1075
• “LCD Frame Configuration Register” on page 1075
• “LCD FIFO Register” on page 1076
• “Dithering Pattern DP1_2 Register” on page 1077
• WPKEY: Write Protect KEY

Should be written at value 0x4C4344 ("LCD" in ASCII). Writing any other value in this field aborts the write operation of the
1091
AT91SAM9G45
44.12.35 LCD Write Protect Status Register

Register Name: LCD_WPSR
Address: 0x005008E8
31 30 29 28 27 26 25 24
— — — — — — — —
23 22 21 20 19 18 17 16
WPVSRC
15 14 13 12 11 10 9 8
WPVSRC
7 6 5 4 3 2 1 0
— — — — — — — WPVS
• WPVS: Write Protect Enable

0 = No Write Protect Violation has occurred since the last read of the LCD_WPSR register.
1 = A Write Protect Violation occurred since the last read of the LCD_WPSR register. If this violation is an unauthorized
attempt to write a protected register, the associated violation is reported into field WPVSRC.

When WPVS is active, this field indicates the write-protected register (through address offset or code) in which a write
access has been attempted.
Note: Reading LCD_WPSR automatically clears all fields
1092
AT91SAM9G45
45. AT91SAM9G45 Electrical Characteristics
45.1 Absolute Maximum Ratings
Table 45-1. Absolute Maximum Ratings*

Operating Temperature (Industrial)...............-40° C to + 85° C *NOTICE: Stresses beyond those listed under “Absolute Maximum
Ratings” may cause permanent damage to the device.
Junction Temperature-..................................................125°C This is a stress rating only and functional operation of
the device at these or other conditions beyond those
Storage Temperature-...................................60°C to + 150°C indicated in the operational sections of this specification
is not implied. Exposure to absolute maximum rating
Voltage on Input Pins conditions for extended periods may affect device reli-
with Respect to Ground......-0.3V to VDDIO+0.3V(+ 4V max) ability.
Maximum Operating Voltage

(VDDCORE, VDDPL).....................................................2.0V
Maximum Operating Voltage

(VDDIOM, VDDIOP and VDDBU)...................................4.0V
Total DC Output Current on all I/O lines....................350 mA
45.2 DC Characteristics
The following characteristics are applicable to the operating temperature range: TA = -40°C to +85°C, unless otherwise
specified.
Table 45-2. DC Characteristics

Symbol Parameter Conditions Min Typ Max Units
VVDDCORE DC Supply Core 0.9 1.0 1.1 V
DC Supply UDPHS and
VVDDUTMIC 0.9 1.0 1.1 V
UHPHS UTMI+ Core
DC Supply UDPHS and
VVDDUTMII 3.0 3.3 3.6 V
UHPHS UTMI+ Interface
VVDDBU DC Supply Backup 1.8 3.6 V
VVDDPLLA DC Supply PLLA 0.9 1.0 1.1 V
VVDDPLLUTMI DC Supply PLLUTMI 0.9 1.0 1.1 V
VVDDOSC DC Supply Oscillator 1.65 3.6 V
VVDDIOM0 DC Supply DDR I/Os 1.65 1.8 1.95 V
VVDDIOM1 DC Supply EBI I/Os 1.65/3.0 1.8/3.3 1.95/3.6 V
VVDDIOP0 DC Supply Peripheral I/Os 1.65 3.6 V
VVDDIOP1 DC Supply Peripheral I/Os 1.65 3.6 V
VVDDIOP2 DC Supply ISI 1.65 3.6 V
VVDDANA DC Supply Analog 3.0 3.3 3.6 V
VVDDIO from 3.0V to 3.6V -0.3 0.8 V
VIL Input Low-level Voltage
VVDDIO from 1.65V to 1.95V -0.3 0.3 x VVDDIO V
1093
Table 45-2. DC Characteristics
VVDDIO from 3.0V to 3.6V 2 VVDDIO + 0.3 V
VIH Input High-level Voltage
VVDDIO from 1.65V to 1.95V 0.7 x VVDDIO VVDDIO + 0.3 V
IO Max, VVDDIO from 3.0V to 3.6V 0.4 V
CMOS (IO <0.3 mA), VVDDIO from 1.65V
0.1 V
VOL Output Low-level Voltage to 1.95V
TTL (IO Max), VVDDIO from 1.65V to
0.4 V
1.95V
IO Max, VVDDIO from 3.0V to 3.6V VVDDIO - 0.4 V
VVDDIO - 0.1 V
VOH Output High-level Voltage to 1.95V
VVDDIO - 0.4 V
1.95V
IO Max, VVDDIO from 3.0V to 3.6V 0.8 1.1 V
Schmitt trigger Negative TBD TBD V
VT- to 1.95V
going threshold Voltage
TBD TBD V
1.95V
IO Max, VVDDIO from 3.0V to 3.6V 1.6 2.0 V
Schmitt trigger Positive TBD TBD V
VT+ to 1.95V
going threshold Voltage
TBD TBD V
1.95V
PA0-PA31 PB0-PB31 PD0-PD31 PE0-
PE31 40 75 190
NTRST and NRST
RPULLUP Pull-up Resistance kOhms
PC0-PC31 VVDDIOM1 In 1.8V range 240 1000
PC0-PC31 VVDDIOM1 In 3.3V range 120 350
PA0-PA31 PB0-PB31 PD0-PD31 PE0-
8
PE31
IO Output Current mA
PC0-PC31 VVDDIOM1 In 1.8 range 2
PC0-PC31 VVDDIOM1 In 1.8 range 4
On VVDDCORE = 1.0V,
MCK = 0 Hz, excluding TA = 25°C 11
POR mA
All inputs driven TMS, TDI,
TA = 85°C 55
ISC Static Current TCK, NRST = 1
On VVDDBU = 3.3V,
Logic cells consumption, TA = 25°C 8
excluding POR μA
All inputs driven WKUP = 0 TA = 85°C 15
1094 AT91SAM9G45
AT91SAM9G45
45.3 Power Consumption

• Typical power consumption of PLLs, Slow Clock and Main Oscillator.
• Power consumption of power supply in four different modes: Active, Idle, Ultra Low-power
and Backup.
• Power consumption by peripheral: calculated as the difference in current measurement after
having enabled then disabled the corresponding clock.
45.3.1 Power Consumption versus Modes

The values in Table 45-3 and Table 45-4 on page 1096 are estimated values of the power con-
sumption with operating conditions as follows:
• VDDIOM0 = 1.8V
• VDDIOP0 and 1= 3.3V
• VDDPLLA = 1.0V
• VDDCORE = VDDBU = 1.0V
• TA = 25° C
• There is no consumption on the I/Os of the device
Figure 45-1. Measures Schematics
VDDBU
AMP1
VDDCORE
AMP2
These figures represent the power consumption estimated on the power supplies.
Table 45-3. Power Consumption for Different Modes(1)

Mode Conditions Consumption Unit
ARM Core clock is 400 MHz.
MCK is 133 MHz.
Active
All peripheral clocks activated. 130 mA
onto AMP2
Idle state, waiting an interrupt.
Idle All peripheral clocks de-activated. 55 mA
onto AMP2
ARM Core clock is 500 Hz.
Ultra low
All peripheral clocks de-activated. 12 mA
power
onto AMP2
Device only VDDBU powered

Backup 8 μA
onto AMP1
1095
Table 45-4. Power Consumption by Peripheral in Active Mode
Peripheral Consumption Unit
PIO Controller 2.2
USART 7.2
UHPHS 53.2
UDPHS 21.7
TSADC 0.1
TWI 1.3
SPI 4.7
PWM 3.8
HSMCI 25.6 μA/MHz
AC97 5.3
SSC 6.6
Timer Counter Channels 6.9
ISI 4.8
LCD 20.4
DMA 0.2
EMAC 34.8
RNG 0.9
1096 AT91SAM9G45
AT91SAM9G45
45.4 Clock Characteristics
45.4.1 Processor Clock Characteristics
Table 45-5. Processor Clock Waveform Parameters

Symbol Parameter Conditions Min Max Units
VDDCORE = 0.9V
1/(tCPPCK) Processor Clock Frequency 400 MHz
T = 85°C
45.4.2 Master Clock Characteristics
Table 45-6. Master Clock Waveform Parameters

VDDCORE = 0.9V
1/(tCPMCK) Master Clock Frequency 133 MHz
T = 85°C
Note: 1. The system clock is the maximum clock at which the system is able to run. It is given by the smallest value the of
processor clock, internal bus clock, EBI clock.
45.5 Main Oscillator Characteristics
Table 45-7. Main Oscillator Characteristics

Symbol Parameter Conditions Min Typ Max Unit
1/(tCPMAIN) Crystal Oscillator Frequency 8 12 16 MHz
(1)
CCRYSTAL Crystal Load Capacitance 15 20 pF
CINT(1) Internal Load Capacitance 4 pF
CCRYSTAL = 15 pF(1) 22 pF
CLEXT External Load Capacitance
(1)
CCRYSTAL = 20 pF 32 Pf
Duty Cycle 40 50 60 %
tST Startup Time 2 ms
IDDST Standby Current Consumption Standby mode 1 μA
PON Drive Level 150 μW
@ 8 MHz 0.35 0.55 mA
IDD ON Current Dissipation
@ 16 MHz 0.7 1.1 mA
Note: 1. The CCRYSTAL value is specified by the crystal manufacturer. In our case, CCRYSTAL must be between 15 pf and 20 pF. All par-
asitic capacitance, package and board, must be calculated in order to reach 15 pF (minimum targeted load for the
oscillator) by taking into account the internal load CINT. So, to target the minimum oscillator load of 15 pF, external capaci-
tance must be: 15 pF - 4 pF = 11 pF which means that 22 pF is the target value (22 pF from xin to gnd and 22 pF from xout
to gnd) If 20 pF load is targeted, the sum of pad, package, board and external capacitances must be 20 pF - 4 pF = 16 pF
which means 32 pF (32 pF from xin to gnd and 32 pF from xout to gnd).
1097
AT91SAM9G45
XIN XOUT
GNDPLL
1K
CCRYSTAL
CLEXT CLEXT
45.5.1 Crystal Oscillator Characteristics

The following characteristics are applicable to the operating temperature range: TA = -40°C to 85°C and worst case of
power supply, unless otherwise specified.
Table 45-8. Crystal Characteristics

ESR Equivalent Series Resistor Rs 80 Ω
CM Motional Capacitance 9 fF
CS Shunt Capacitance 7 pF
45.5.2 XIN Clock Characteristics
Table 45-9. XIN Clock Electrical Characteristics

1/(tCPXIN) XIN Clock Frequency 50 MHz
tCPXIN XIN Clock Period 20 ns
tCHXIN XIN Clock High Half-period 0.4 x tCPXIN 0.6 x tCPXIN ns
tCLXIN XIN Clock Low Half-period 0.4 x tCPXIN 0.6 x tCPXIN ns
(1)
CIN XIN Input Capacitance 25 pF
(1)
RIN XIN Pulldown Resistor 500 kΩ
(1)
VIN XIN Voltage VDDOSC VDDOSC V
Note: 1. These characteristics apply only when the Main Oscillator is in bypass mode (i.e. when MOSCEN = 0 and OSCBYPASS = 1)
in the CKGR_MOR register. See “PMC Clock Generator Main Oscillator Register” in the PMC section.
1098 AT91SAM9G45
AT91SAM9G45
45.6 32 kHz Oscillator Characteristics

Table 45-10. 32 kHz Oscillator Characteristics
1/(tCP32KHz) Crystal Oscillator Frequency 32 768 kHz
CCRYSTAL32 Load Capacitance Crystal @ 32.768 kHz 6 12.5 pF
CCRYSTAL32 = 6 pF 6 pF
CLEXT32(2) External Load Capacitance
CCRYSTAL32 = 12.5 pF 19 pF
Duty Cycle 40 60 %
CCRYSTAL32 = 6 pF 400 ms
RS = 50 kΩ(1)
CCRYSTAL32 = 12.5 pF 900 ms
tST Startup Time
CCRYSTAL32 = 6 pF 600 ms
RS = 100 kΩ(1)
CCRYSTAL32 = 12.5 pF 1200 ms
Notes: 1. RS is the equivalent series resistance.

2. CLEXT32 is determined by taking into account internal, parasitic and package load capacitance.
AT91SAM9G45
XIN32 XOUT32 GNDBU
CCRYSTAL32
CLEXT32 CLEXT32
45.6.1 32 kHz Crystal Characteristics
Table 45-11. 32 kHz Crystal Characteristics

ESR Equivalent Series Resistor Rs Crystal @ 32.768 kHz 50 100 kΩ
CM Motional Capacitance Crystal @ 32.768 kHz 3 fF
CS Shunt Capacitance Crystal @ 32.768 kHz 2 pF
1099
45.6.2 XIN32 Clock Characteristics
Table 45-12. XIN32 Clock Electrical Characteristics

1/(tCPXIN32) XIN32 Clock Frequency 44 kHz
tCPXIN32 XIN32 Clock Period 22 μs
tCHXIN32 XIN32 Clock High Half-period 11 μs
tCLXIN32 XIN32 Clock Low Half-period 11 μs
tCLCH32 XIN32 Clock Rise time 400 ns
tCLCL32 XIN32 Clock Fall time 400 ns
(1)
CIN32 XIN32 Input Capacitance 6 pF
(1)
RIN32 XIN32 Pulldown Resistor 4 MΩ
(1)
VIN32 XIN32 Voltage VDDBU VDDBU V
(1)
VINIL32 XIN32 Input Low Level Voltage -0.3 0.3 x VVDDBU V
(1)
VINIH32 XIN32 Input High Level Voltage 0.7 x VVDDBU VVDDBU + 0.3 V
Note: 1. These characteristics apply only when the 32.768KHz Oscillator is in bypass mode (i.e. when RCEN = 0, OSC32EN = 0,
OSCSEL = 1 and OSC32BYP = 1) in the SCKCR register. See “Slow Clock Selection” in the PMC section.
45.7 32 kHz RC Oscillator Characteristics

Table 45-13. RC Oscillator Characteristics
1/(tCPRCz) Crystal Oscillator Frequency 22 42 kHz
Duty Cycle 45 55 %
tST Startup Time 75 μs
1100 AT91SAM9G45
AT91SAM9G45
45.8 PLL Characteristics

Table 45-14. PLLA Characteristics
FOUT Output Frequency Refer to following table 400 800 MHz
FIN Input Frequency 2 32 MHz
active mode 3.6 4.5 mA
IPLL Current Consumption
standby mode 1 μA
The following configuration of ICPLLA and OUTA must be done for each PLLA frequency range.
Table 45-15. PLLA Frequency Regarding ICPLLA and OUTA

PLL Frequancy Range (MHz) ICPLLA OUTA
745 - 800 0 0 0
695 - 750 0 0 1
645 - 700 0 1 0
595 - 650 0 1 1
545 - 600 1 0 0
495 - 550 1 0 1
445 - 500 1 1 0
400 - 450 1 1 1
45.8.1 UTMI PLL Characteristics
Table 45-16. Phase Lock Loop Characteristics

FIN Input Frequency 4 12 32 MHz
FOUT Output Frequency 450 480 600 MHz
active mode 5 8 mA
IPLL Current Consumption
standby mode 1.5 μA
1101
45.9 I/Os
Criteria used to define the maximum frequency of the I/Os:
• output duty cycle (40%-60%)
• minimum output swing: 100 mV to VDDIO - 100 mV
• Addition of rising and falling time inferior to 75% of the period
Table 45-17. I/O Characteristics

(1)
3.3V domain MHz
(2)
FreqMax VDDIOP powered Pins frequency 2.5V domain MHz
1.8V domain(3) MHz
Notes: 1. 3.3V domain: VVDDIOP from 3.0V to 3.6V, maximum external capacitor = 40 pF
2. 2.5V domain: VVDDIOP from 2.3V to 2.7V, maximum external capacitor = 30 pF
3. 1.8V domain: VVDDIOP from 1.65V to 1.95V, maximum external capacitor = 20 pF
45.10 I/O Drive Level

The EBI_DRIVE field in the EBI_CSA register allows the user to program the rising and falling
times of DDR2 signals. Rising/Falling time is given between 30% and 70%.
Table 45-18. SDRAM Clock

Value Conditions Rising time (ns) Falling time (ns)
00 1.8V range, HIGH drive, CLOAD = 10 pF 0.8 1.0
01 1.8V range, LOW drive, CLOAD = 10 pF 1.6 1.5
10 3.3V range, High drive, CLOAD = 10 pF 0.8 0.8
Table 45-19. SDRAM Data, Address and Control

Value Conditions Rising time (ns) Falling time (ns)
00 1.8V range,HIGH drivee, CLOAD = 30 pF 1.5 1.5
10 3.3V range, HIGH drive, CLOAD = 50 pF 2.8 3.1
1102 AT91SAM9G45
AT91SAM9G45
45.11 USB HS Characteristics
45.11.1 Electrical Characteristics
Table 45-20. Electrical Parameters

Bus Pull-up Resistor on
RPUI in LS or FS Mode 1.5 kOhms
Upstream Port (idle bus)
Bus Pull-up Resistor on
RPUA Upstream Port (upstream port in LS or FS Mode 15 kOhms
receiving)
Settlng time
TBIAS Bias settling time 20 μs
TOSC Oscillator settling time With Crystal 12MHz 2 ms
TSETTLING Settling time FIN = 12 MHz 0.3 0.5 ms
45.11.2 Static Power Consumption
Table 45-21. Static Power Consumption

Bias current consumption on
IBIAS 1 μA
VBG
HS Transceiver and I/O current
8 μA
consumption
IVDDUTMII
LS / FS Transceiver and I/O
no connection(1) 3 μA
current consumption
Core, PLL, and Oscillator
IVDDUTMIC 2 μA
current consumption
Note: 1. If cable is connected add 200 μA (Typical) due to Pull-up/Pull-down current consumption.
1103
45.11.3 Dynamic Power Consumption
Table 45-22. Dynamic Power Consumption

Bias current consumption on
IBIAS 0.7 0.8 mA
VBG
HS Transceiver current
HS transmission 47 60 mA
consumption
HS Transceiver current
HS reception 18 27 mA
consumption
LS / FS Transceiver current
IVDDUTMII FS transmission 0m cable(1) 4 6 mA
consumption
FS transmission 5m cable(1) 26 30 mA
consumption
FS reception(1) 3 4.5 mA
consumption
PLL, Core and Oscillator current
IVDDUTMIC 5.5 9 mA
consumption
Note: 1. Including 1mA due to Pull-up/Pull-down current consumption.
1104 AT91SAM9G45
AT91SAM9G45
45.12 Touch Screen ADC (TSADC)

Table 45-23. Channel Conversion Time and ADC Clock
Parameter Conditions Min Typ Max Units
ADC Clock Frequency 10-bit resolution mode 13.2 MHz
Startup Time Return from Idle Mode 40 μs
(1)
Track and Hold Acquisition Time ADC Clock = 13.2 MHz 0.5 μs
Conversion Time ADC Clock = 13.2 MHz(1) 1.75 μs
Throughput Rate ADC Clock = 13.2 MHz(1) 440 kSPS
Note: 1. The Track-and-Hold Acquisition Time is given by:
TTH (ns) = 500 + ( 12 × Z IN ) ( Ω)
The ADC internal clock is divided by 2 in order to generate a clock with a duty cycle of 75%. So the maximum conversion
time is give by:
23
TCT ( μs ) = ----------- ( MHz )
Fclk
The full speed is obtained for an input source impedance of < 500 Ohms maximum, or TTH = 500 ns..
Table 45-24. External Voltage Reference Input

Parameter Conditions Min Typ Max Units
ADVREF Input Voltage Range 2.4 VDDANA V
ADVREF Average Current 600 μA
Current Consumption on VDDANA 300 μA
Table 45-25. Analog Inputs

Parameter Min Typ Max Units
Input Voltage Range 0 ADVREF V
Input Leakage Current 1 μA
Input Capacitance 7 10 pF
Input Impedance 50 Ohms
Table 45-26. Transfer Characteristics

Parameter Min Typ Max Units
Resolution 10 bit
Integral Non-linearity ±2 LSB
Differential Non-linearity -0.9 +0.9 LSB
Offset Error -1.5 0.5 ±10 mV
Gain Error ±2 LSB
1105
46. AT91SAM9G45 Mechanical Characteristics
46.1 Package Drawings
Figure 46-1. 324-ball TFBGA Package Drawing
Table 46-1. Soldering Information

Ball Land 0.4 mm +/- 0.05
Soldering Mask Opening 0.275 mm +/- 0.03
Table 46-2. Device and 324-ball TFBGA Package Maximum Weight

572 mg
Table 46-3. 324-ball TFBGA Package Characteristics

Moisture Sensitivity Level 3
Table 46-4. Package Reference

JEDEC Drawing Reference MO-210
JESD97 Classification e1
This package respects the recommendations of the NEMI User Group.
1106 AT91SAM9G45
AT91SAM9G45
46.2 Soldering Profile

Table 46-5 gives the recommended soldering profile from J-STD-020C.
Table 46-5. Soldering Profile

Profile Feature Green Package
Average Ramp-up Rate (217°C to Peak) 3° C/sec. max.
Preheat Temperature 175°C ±25°C 180 sec. max.
Temperature Maintained Above 217°C 60 sec. to 150 sec.
Time within 5° C of Actual Peak Temperature 20 sec. to 40 sec.
Peak Temperature Range 260 +0 ° C
Ramp-down Rate 6° C/sec. max.
Time 25° C to Peak Temperature 8 min. max.
Note: It is recommended to apply a soldering temperature higher than 250°C
A maximum of three reflow passes is allowed per component.
1107
47. AT91SAM9G45 Ordering Information
Table 47-1. AT91SAM9G45 Ordering Information

Ordering Code Package Package Type Temperature Operating Range
Industrial
AT91SAM9G45-CU TFBGA324 Green
-40°C to 85°C
1108 AT91SAM9G45
AT91SAM9G45
48. AT91SAM9G45 Errata
48.1 Marking
All devices are marked with the Atmel logo and the ordering code.
Additional marking may be in one of the following formats:
YYWW V
XXXXXXXXX ARM
where
• “YY”: manufactory year
• “WW”: manufactory week
• “V”: revision
• “XXXXXXXXX”: lot number
Errata on engineering samples of the AT91SAM9G45 is available at

http://www.atmel.com/dyn/resources/prod_documents/doc6485.pdf.
1109
48.2 Errata
48.2.1 Error Corrected Code Controller (ECC)
48.2.1.1 ECC: Computation with a 1 clock cycle long NRD/NWE pulse

If the SMC is programmed with NRD/NWE pulse length equal to 1 clock cycle, HECC can't com-
pute the parity.
Problem/Fix Workaround
It is recommanded to program SMC with a value superior to 1.
48.2.2 Pulse Width Modulation Controller (PWM)
48.2.2.1 PWM: Zero Period

It is impossible to update a period equal to 0 by using the PWM_CUPD register.
None
48.2.3 Serial Synchronous Controller (SSC)
48.2.3.1 SSC: Data sent without any frame synchro

When SSC is configured with the following conditions:
• RF is in input,
• TD is synchronized on a receive START (any condition: START field = 2 to 7)
• TF toggles at each start of data transfer
• Transmit STTDLY = 0
• Check TD and TF after a receive START,
The data is sent but there is not any toggle of the TF line
Transmit STTDLY must be different from 0.
48.2.3.2 SSC: Unexpected delay on TD output

When SSC is configured with the following conditions:
• TCMR.STTDLY more than 0
• RCMR.START = Start on falling edge/Start on Rising edge/Start on any edge
• RFMR.FSOS = None (input)
• TCMR.START = Receive Start
Unexpected delay of 2 or 3 system clock cycles is added to TD output.
None
1110 AT91SAM9G45
AT91SAM9G45
Revision History
In the tables that follow, the most recent version appears first.
Change
Doc. Rev Comments Request Ref.
DDRSDRC:
Section 21.7.3 “DDRSDRC Configuration Register”, bit named ENRDM removed from register. 6606
6438B
Section 21.7.9 “DDRSDRC DLL Register” bits named, SDCOVF, SDCUDF, SDERF, SDVAL, SDCVAL
removed from register.
Change
Doc. Rev Comments Request Ref.
6438A First issue
1111
1112 AT91SAM9G45
AT91SAM9G45
Table of Contents
Features ..................................................................................................... 1
1 Description ............................................................................................... 2
2 Block Diagram .......................................................................................... 3
3 Signal Description ................................................................................... 4
4 Package and Pinout ............................................................................... 11

4.1 Mechanical Overview of the 324-ball LFBGA Package ...................................11
4.2 324-ball TFBGA Package Pinout .....................................................................12
5 Power Considerations ........................................................................... 14

5.1 Power Supplies ................................................................................................14
6 Memories ................................................................................................ 15
6.1 Memory Mapping .............................................................................................16
6.2 Embedded Memories ......................................................................................16
7 System Controller .................................................................................. 18

7.1 System Controller Mapping .............................................................................18
7.2 System Controller Block Diagram ....................................................................19
7.3 Chip Identification ............................................................................................20
7.4 Backup Section ................................................................................................20
8 Peripherals ............................................................................................. 21
8.1 Peripheral Mapping .........................................................................................21
8.2 Peripheral Identifiers ........................................................................................21
8.3 Peripheral Interrupts and Clock Control ..........................................................22
8.4 Peripheral Signals Multiplexing on I/O Lines ...................................................22
9 ARM926EJ-S Processor Overview ....................................................... 29

9.1 Description .......................................................................................................29
9.2 Embedded Characteristics ..............................................................................30
9.3 Block Diagram .................................................................................................31
9.4 ARM9EJ-S Processor ......................................................................................32
9.5 CP15 Coprocessor ..........................................................................................40
9.6 Memory Management Unit (MMU) ..................................................................42
9.7 Caches and Write Buffer .................................................................................44
9.8 Tightly-Coupled Memory Interface ..................................................................46
i
9.9 Bus Interface Unit ............................................................................................47
10 AT91SAM9G45 Debug and Test ........................................................... 49

10.1 Description .......................................................................................................49
10.3 Block Diagram .................................................................................................50
10.4 Application Examples ......................................................................................51
10.5 Debug and Test Pin Description ......................................................................52
10.6 Functional Description .....................................................................................53
11 Boot Strategies ...................................................................................... 55

11.1 Boot Program ..................................................................................................55
11.2 Flow Diagram ..................................................................................................56
11.3 Device Initialization ..........................................................................................57
11.4 NVM Boot ........................................................................................................58
11.5 SAM-BA Monitor ..............................................................................................64
12 Reset Controller (RSTC) ........................................................................ 69

12.1 Description .......................................................................................................69
12.3 Block Diagram .................................................................................................69
12.5 Reset Controller (RSTC) User Interface ..........................................................78
13 Real-time Timer (RTT) ............................................................................ 83

13.1 Description .......................................................................................................83
13.3 Block Diagram .................................................................................................83
13.5 Real-time Timer (RTT) User Interface .............................................................86
13.6 Real-time Clock (RTC) ....................................................................................91
13.7 Description .......................................................................................................91
13.9 Block Diagram .................................................................................................91
13.10 Product Dependencies ....................................................................................92
13.12 Real Time Clock (RTC) User Interface ............................................................95
14 Periodic Interval Timer (PIT) ............................................................... 109
ii AT91SAM9G45
AT91SAM9G45
14.1 Description .....................................................................................................109

14.2 Embedded Characteristics ............................................................................109
14.3 Block Diagram ...............................................................................................109
14.4 Functional Description ...................................................................................109
14.5 Periodic Interval Timer (PIT) User Interface ..................................................111
15 Watchdog Timer (WDT) ....................................................................... 115

15.1 Description .....................................................................................................115
15.3 Block Diagram ...............................................................................................115
15.5 Watchdog Timer (WDT) User Interface .........................................................118
16 Shutdown Controller (SHDWC) .......................................................... 121

16.1 Description .....................................................................................................121
16.3 Block Diagram ...............................................................................................121
16.4 I/O Lines Description .....................................................................................122
16.5 Product Dependencies ..................................................................................122
16.7 Shutdown Controller (SHDWC) User Interface .............................................124
17 General Purpose Backup Registers (GPBR) ..................................... 127

17.1 Description .....................................................................................................127
17.3 General Purpose Backup Registers (GPBR) User Interface ........................127
18 Bus Matrix (MATRIX) ........................................................................... 129

18.1 Description .....................................................................................................129
18.3 Memory Mapping ...........................................................................................131
18.4 Special Bus Granting Mechanism .................................................................131
18.5 Arbitration ......................................................................................................133
18.6 Write Protect Registers ..................................................................................136
18.7 Bus Matrix (MATRIX) User Interface .............................................................137
19 External Memories ............................................................................... 153

19.1 Programmable I/O Lines Power Supplies and Drive Levels ..........................153
19.2 DDR2 Controller ............................................................................................153
iii
19.3 External Bus Interface (EBI) ..........................................................................158
20 Static Memory Controller (SMC) ......................................................... 181

20.1 Description .....................................................................................................181
20.3 Multiplexed Signals ........................................................................................181
20.4 Application Example ......................................................................................182
20.6 External Memory Mapping .............................................................................183
20.7 Connection to External Devices ....................................................................183
20.8 Standard Read and Write Protocols ..............................................................188
20.9 Automatic Wait States ...................................................................................196
20.10 Data Float Wait States ...................................................................................201
20.11 External Wait .................................................................................................205
20.12 Slow Clock Mode ...........................................................................................211
20.13 Asynchronous Page Mode ............................................................................214
20.14 Programmable IO Delays ..............................................................................217
20.15 Static Memory Controller (SMC) User Interface ............................................218
21 DDR/SDR SDRAM Controller (DDRSDRC) ......................................... 225

21.1 Description .....................................................................................................225
21.2 DDRSDRC Module Diagram .........................................................................226
21.5 Software Interface/SDRAM Organization, Address Mapping ........................249
21.6 Programmable IO Delays ..............................................................................251
21.7 DDR-SDRAM Controller (DDRSDRC) User Interface ...................................252
22 Error Corrected Code Controller (ECC) ............................................. 271

22.1 Description .....................................................................................................271
22.2 Block Diagram ...............................................................................................271
22.4 Error Corrected Code Controller (ECC) User Interface .................................276
22.5 Registers for 1 ECC for a page of 512/1024/2048/4096 bytes ......................287
22.6 Registers for 1 ECC per 512 bytes for a page of 512/2048/4096 bytes,
8-bit word .......................................................................................................289
22.7 Registers for 1 ECC per 256 bytes for a page of 512/2048/4096 bytes,
8-bit word .......................................................................................................297
iv AT91SAM9G45
AT91SAM9G45
23 Peripheral DMA Controller (PDC) ....................................................... 313

23.1 Description .....................................................................................................313
23.3 Block Diagram ...............................................................................................315
23.5 Peripheral DMA Controller (PDC) User Interface ..........................................318
24 Clock Generator ................................................................................... 329

24.1 Description .....................................................................................................329
24.3 Slow Clock Crystal Oscillator .........................................................................329
24.4 Slow Clock RC Oscillator ...............................................................................330
24.5 Slow Clock Selection .....................................................................................330
24.6 Main Oscillator ...............................................................................................333
24.7 Divider and PLLA Block .................................................................................334
24.8 UTMI Bias and Phase Lock Loop Programming ...........................................335
25 Power Management Controller (PMC) ................................................ 336

25.1 Description .....................................................................................................336
25.3 Master Clock Controller .................................................................................338
25.4 Processor Clock Controller ............................................................................338
25.5 USB Device and Host clocks .........................................................................339
25.6 LP-DDR/DDR2 Clock ....................................................................................339
25.7 Peripheral Clock Controller ............................................................................339
25.8 Programmable Clock Output Controller .........................................................340
25.9 Programming Sequence ................................................................................340
25.10 Clock Switching Details .................................................................................344
25.11 Power Management Controllerr (PMC) User Interface .................................347
26 Advanced Interrupt Controller (AIC) .................................................. 367

26.1 Description .....................................................................................................367
26.3 Block Diagram ...............................................................................................368
26.4 Application Block Diagram .............................................................................368
26.5 AIC Detailed Block Diagram ..........................................................................368
26.6 I/O Line Description .......................................................................................369
v
26.9 Advanced Interrupt Controller (AIC) User Interface .......................................380
27 Debug Unit (DBGU) .............................................................................. 391

27.1 Description .....................................................................................................391
27.3 Block Diagram ...............................................................................................392
27.5 UART Operations ..........................................................................................393
27.6 Debug Unit (DBGU) User Interface ...............................................................400
28 Parallel Input/Output Controller (PIO) ................................................ 415

28.1 Description .....................................................................................................415
28.3 Block Diagram ...............................................................................................416
28.6 I/O Lines Programming Example ...................................................................422
28.7 Parallel Input/Output Controller (PIO) User Interface ....................................423
29 Serial Peripheral Interface (SPI) ......................................................... 441

29.1 Description .....................................................................................................441
29.3 Block Diagram ...............................................................................................442
29.5 Signal Description .........................................................................................444
29.8 Serial Peripheral Interface (SPI) User Interface ............................................459
30 Two-wire Interface (TWI) ..................................................................... 473

30.1 Description .....................................................................................................473
30.3 List of Abbreviations ......................................................................................474
30.4 Block Diagram ...............................................................................................474
30.8 Master Mode ..................................................................................................478
vi AT91SAM9G45
AT91SAM9G45
30.9 Multi-master Mode .........................................................................................490

30.10 Slave Mode ....................................................................................................493
30.11 Two-wire Interface (TWI) User Interface .......................................................501
31 Universal Synchronous Asynchronous Receiver Transeiver

(USART) ................................................................................................ 515
31.1 Description .....................................................................................................515
31.3 Block Diagram ...............................................................................................516
31.5 I/O Lines Description ....................................................................................518
31.8 Universal Synchronous Asynchronous Receiver Transmitter
(USART) User Interface ...............................................................................550
32 Synchronous Serial Controller (SSC) ................................................ 571

32.1 Description .....................................................................................................571
32.3 Block Diagram ...............................................................................................572
32.5 Pin Name List ................................................................................................574
32.8 SSC Application Examples ............................................................................588
32.9 Synchronous Serial Controller (SSC) User Interface ....................................590
33 Timer Counter (TC) .............................................................................. 613

33.1 Description .....................................................................................................613
33.3 Block Diagram ...............................................................................................614
33.4 Pin Name List ................................................................................................615
33.7 Timer Counter (TC) User Interface ................................................................629
34 High Speed MultiMedia Card Interface (HSMCI) ................................ 647

34.1 Description .....................................................................................................647
vii
34.3 Block Diagram ...............................................................................................648
34.5 Pin Name List ...............................................................................................649
34.7 Bus Topology .................................................................................................651
34.8 High Speed MultiMedia Card Operations ......................................................653
34.9 SD/SDIO Card Operation ..............................................................................672
34.10 CE-ATA Operation .........................................................................................673
34.11 HSMCI Boot Operation Mode ........................................................................674
34.12 HSMCI Transfer Done Timings .....................................................................676
34.13 MultiMedia Card Interface (MCI) User Interface ............................................678
35 Ethernet MAC 10/100 (EMAC) ............................................................. 707

35.1 Description .....................................................................................................707
35.3 Block Diagram ...............................................................................................708
35.5 Programming Interface ..................................................................................720
35.6 Ethernet MAC 10/100 (EMAC) User Interface ...............................................723
36 USB Host Port (UHP) ........................................................................... 757

36.1 Description .....................................................................................................757
36.3 Block Diagram ...............................................................................................758
36.5 I/O Lines ........................................................................................................759
36.6 Typical Connection ........................................................................................762
37 USB High Speed Device Port (UDPHS) .............................................. 763

37.1 Description .....................................................................................................763
37.3 Block Diagram ...............................................................................................765
37.4 Typical Connection ........................................................................................766
37.6 USB High Speed Device Port (UDPHS) User Interface ................................791
38 Image Sensor Interface (ISI) ................................................................ 833

38.1 Description .....................................................................................................833
viii AT91SAM9G45
AT91SAM9G45
38.3 Block Diagram ...............................................................................................834

38.5 Image Sensor Interface (ISI) User Interface ..................................................843
39 Touch Screen ADC Controller (TSADCC) .......................................... 873

39.1 Description .....................................................................................................873
39.3 Block Diagram ...............................................................................................874
39.4 Signal Description ..........................................................................................875
39.6 Analog-to-digital Converter Functional Description .......................................876
39.7 Touch Screen ................................................................................................877
39.8 Conversion Results .......................................................................................882
39.9 Conversion Triggers ......................................................................................884
39.10 Operating Modes ...........................................................................................884
39.11 Touch Screen ADC Controller (TSADCC) User Interface .............................892
40 DMA Controller (DMAC) ...................................................................... 911

40.1 Description .....................................................................................................911
40.3 Block Diagram ...............................................................................................913
40.5 DMAC Software Requirements .....................................................................940
40.6 DMA Controller (DMAC) User Interface ........................................................942
41 Pulse Width Modulation Controller (PWM) ........................................ 965

41.1 Description .....................................................................................................965
41.3 Block Diagram ...............................................................................................966
41.7 Pulse Width Modulation Controller (PWM) User Interface ............................976
42 AC97 Controller (AC97C) .................................................................... 989

42.1 Description .....................................................................................................989
42.3 Block Diagram ...............................................................................................990
42.4 Pin Name List ................................................................................................991
ix
42.8 AC97 Controller (AC97C) User Interface ....................................................1004
43 True Random Number Generator (TRNG) ........................................ 1025

43.1 Description ...................................................................................................1025
43.2 True Random Number Generator (TRNG) User Interface ..........................1026
44 LCD Controller (LCDC) ...................................................................... 1033

44.1 Description ...................................................................................................1033
44.2 Embedded Characteristics ..........................................................................1033
44.3 Block Diagram .............................................................................................1034
44.4 I/O Lines Description ...................................................................................1035
44.5 Product Dependencies ................................................................................1035
44.6 Functional Description .................................................................................1037
44.7 Interrupts .....................................................................................................1058
44.8 Configuration Sequence ..............................................................................1058
44.9 Double-buffer Technique .............................................................................1059
44.10 2D Memory Addressing ...............................................................................1060
44.11 Register Configuration Guide ......................................................................1062
44.12 LCD Controller (LCDC) User Interface ........................................................1063
45 AT91SAM9G45 Electrical Characteristics ....................................... 1093

45.1 Absolute Maximum Ratings .........................................................................1093
45.2 DC Characteristics .......................................................................................1093
45.3 Power Consumption ....................................................................................1095
45.4 Clock Characteristics ...................................................................................1097
45.5 Main Oscillator Characteristics ....................................................................1097
45.6 32 kHz Oscillator Characteristics .................................................................1099
45.7 32 kHz RC Oscillator Characteristics ..........................................................1100
45.8 PLL Characteristics .....................................................................................1101
45.9 I/Os ..............................................................................................................1102
45.10 I/O Drive Level .............................................................................................1102
45.11 USB HS Characteristics ..............................................................................1103
45.12 Touch Screen ADC (TSADC) ......................................................................1105
46 AT91SAM9G45 Mechanical Characteristics .................................... 1106

46.1 Package Drawings .......................................................................................1106
x AT91SAM9G45
AT91SAM9G45
46.2 Soldering Profile ..........................................................................................1107
47 AT91SAM9G45 Ordering Information .............................................. 1108
48 AT91SAM9G45 Errata ........................................................................ 1109

48.1 Marking ........................................................................................................1109
48.2 Errata ...........................................................................................................1110
Revision History.................................................................................. 1111
Table of Contents....................................................................................... i
xi
Headquarters International
Atmel Corporation Atmel Asia Atmel Europe Atmel Japan

2325 Orchard Parkway Unit 1-5 & 16, 19/F Le Krebs 9F, Tonetsu Shinkawa Bldg.
San Jose, CA 95131 BEA Tower, Millennium City 5 8, Rue Jean-Pierre Timbaud 1-24-8 Shinkawa
USA 418 Kwun Tong Road BP 309 Chuo-ku, Tokyo 104-0033
Tel: 1(408) 441-0311 Kwun Tong, Kowloon 78054 Saint-Quentin-en- Japan
Fax: 1(408) 487-2600 Hong Kong Yvelines Cedex Tel: (81) 3-3523-3551
Tel: (852) 2245-6100 France Fax: (81) 3-3523-7581
Fax: (852) 2722-1369 Tel: (33) 1-30-60-70-00
Fax: (33) 1-30-60-71-11
Product Contact
Web Site Technical Support Sales Contacts

www.atmel.com AT91SAM Support www.atmel.com/contacts/
www.atmel.com/AT91SAM Atmel techincal support
Literature Requests
www.atmel.com/literature
Disclaimer: The information in this document is provided in connection with Atmel products. No license, express or implied, by estoppel or otherwise, to any
intellectual property right is granted by this document or in connection with the sale of Atmel products. EXCEPT AS SET FORTH IN ATMEL’S TERMS AND CONDI-
TIONS OF SALE LOCATED ON ATMEL’S WEB SITE, ATMEL ASSUMES NO LIABILITY WHATSOEVER AND DISCLAIMS ANY EXPRESS, IMPLIED OR STATUTORY
WARRANTY RELATING TO ITS PRODUCTS INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTY OF MERCHANTABILITY, FITNESS FOR A PARTICULAR
PURPOSE, OR NON-INFRINGEMENT. IN NO EVENT SHALL ATMEL BE LIABLE FOR ANY DIRECT, INDIRECT, CONSEQUENTIAL, PUNITIVE, SPECIAL OR INCIDEN-
TAL DAMAGES (INCLUDING, WITHOUT LIMITATION, DAMAGES FOR LOSS OF PROFITS, BUSINESS INTERRUPTION, OR LOSS OF INFORMATION) ARISING OUT OF
THE USE OR INABILITY TO USE THIS DOCUMENT, EVEN IF ATMEL HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. Atmel makes no
representations or warranties with respect to the accuracy or completeness of the contents of this document and reserves the right to make changes to specifications
and product descriptions at any time without notice. Atmel does not make any commitment to update the information contained herein. Unless specifically provided
otherwise, Atmel products are not suitable for, and shall not be used in, automotive applications. Atmel’s products are not intended, authorized, or warranted for use
as components in applications intended to support or sustain life.
© 2009 Atmel Corporation. All rights reserved. Atmel ®, Atmel logo and combinations thereof, DataFlash®, SAM-BA ® and others, are regis-
tered trademarks or trademarks of Atmel Corporation or its subsidiaries. ARM ®, Thumb® and the ARMPowered logo® and others, are registered
trademarks or trademarks of ARM Ltd. Windows ® and others, are registered trademarks or trademarks of Microsoft Corporation in the US and/or
other countries. Other terms and product names may be trademarks of others.

At91sam9g45 Ds

Uploaded by

Copyright:

Available Formats

At91sam9g45 Ds

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

At91sam9g45 Ds

Uploaded by

Copyright:

Available Formats

Features

• 400 MHz ARM926EJ-S™ ARM® Thumb® Processor

JTAG / Boundary Scan HS HS PIO

PLLA DMA DMA DMA DMA DMA DDR_CLK,#DDR_CLK

Table 3-1. Signal Description List

Table 3-1. Signal Description List (Continued)

(1) Pulled-up input at reset

(1) Pulled-up input at reset

(1) Pulled-up input at reset

(1) Pulled-up input at reset

(1) Pulled-up input at reset

Table 3-1. Signal Description List (Continued)

Table 3-1. Signal Description List (Continued)

4. Package and Pinout

4.1 Mechanical Overview of the 324-ball LFBGA Package

Figure 4-1. Orientation of the 324-ball LFBGA Package

Table 4-1. AT91SAM9G45 Pinout for 324-ball BGA Package

Table 4-1. AT91SAM9G45 Pinout for 324-ball BGA Package

5.1 Power Supplies

6.2 Embedded Memories

6.2.1 Internal SRAM

Figure 6-2. Internal SRAM Reset

6.2.2 TCM Interface

Figure 6-3. RAM Mapping

Table 6-1. ITCM and DTCM Memory Configuration

7.1 System Controller Mapping

7.2 System Controller Block Diagram

Figure 7-1. AT91SAM9G45 System Controller Block Diagram

PLLA PLLACK idle

7.4 Backup Section

8.1 Peripheral Mapping

8.2 Peripheral Identifiers

Table 8-1. AT91SAM9G45 Peripheral Identifiers

8.3.1 System Interrupt

8.3.2 External Interrupts

8.4 Peripheral Signals Multiplexing on I/O Lines

8.4.1 PIO Controller A Multiplexing

Table 8-2. Multiplexing on PIO Controller A (PIOA)

Table 8-3. Multiplexing on PIO Controller B (PIOB)

8.4.3 PIO Controller C Multiplexing

Table 8-4. Multiplexing on PIO Controller C (PIOC)

Table 8-5. Multiplexing on PIO Controller D (PIOD)

8.4.5 PIO Controller E Multiplexing

Table 8-6. Multiplexing on PIO Controller E (PIOE)

9. ARM926EJ-S Processor Overview

9.3 Block Diagram

Figure 9-1. ARM926EJ-S Internal Functional Block Diagram

CP15 System External Trace Port

Data TCM Instruction TCM

9.4.1 ARM9EJ-S Operating States

9.4.2 Switching State

9.4.3 Instruction Pipelines

9.4.4 Memory Access

9.4.5 Jazelle Technology

9.4.6 ARM9EJ-S Operating Modes

9.4.7 ARM9EJ-S Registers

Table 9-1. ARM9TDMI Modes and Registers Layout

CPSR CPSR CPSR CPSR CPSR CPSR

Mode-specific banked registers