Jesd209 4

JEDEC
STANDARD
Low Power Double Data Rate 4

(LPDDR4)
JESD209-4
AUGUST 2014
JEDEC SOLID STATE TECHNOLOGY ASSOCIATION

NOTICE
JEDEC standards and publications contain material that has been prepared, reviewed, and approved through the
JEDEC Board of Directors level and subsequently reviewed and approved by the JEDEC legal counsel.
JEDEC standards and publications are designed to serve the public interest through eliminating misunderstandings
between manufacturers and purchasers, facilitating interchangeability and improvement of products, and assisting the
purchaser in selecting and obtaining with minimum delay the proper product for use by those other than JEDEC
members, whether the standard is to be used either domestically or internationally.
JEDEC standards and publications are adopted without regard to whether or not their adoption may involve patents
or articles, materials, or processes. By such action JEDEC does not assume any liability to any patent owner, nor
does it assume any obligation whatever to parties adopting the JEDEC standards or publications.
The information included in JEDEC standards and publications represents a sound approach to product specification
and application, principally from the solid state device manufacturer viewpoint. Within the JEDEC organization
there are procedures whereby a JEDEC standard or publication may be further processed and ultimately become an
ANSI standard.
No claims to be in conformance with this standard may be made unless all requirements stated in the standard are
met.
Inquiries, comments, and suggestions relative to the content of this JEDEC standard or publication should be
addressed to JEDEC at the address below, or refer to www.jedec.org under Standards and Documents for alternative
contact information.
Published by
©JEDEC Solid State Technology Association 2014
3103 North 10th Street
Suite 240 South
Arlington, VA 22201-2107
This document may be downloaded free of charge; however JEDEC retains the
copyright on this material. By downloading this file the individual agrees not to
charge for or resell the resulting material.
PRICE: Contact JEDEC
PLEASE!
DON'T VIOLATE
THE
LAW!
This document is copyrighted by the JEDEC and may not be

reproduced without permission.
For information, contact:
JEDEC Solid State Technology Association

3103 North 10th Street, Suite 240 South
Arlington, Virginia 22201-2107
or refer to www.jedec.org under Standards-Documents/Copyright Information

JEDEC Standard No. 209-4
Page 1
LOW POWER DOUBLE DATA RATE 4 (LPDDR4)
(From JEDEC Board Ballot JCB-14-41, formulated under the cognizance of the JC-42.6 Subcommittee on
Low Power Memories.)
1 Scope
This document defines the LPDDR4 standard, including features, functionalities, AC and DC
characteristics, packages, and ball/signal assignments. The purpose of this specification is to define the
minimum set of requirements for JEDEC compliant 4 Gb through 32 Gb for x16x2channel SDRAM
devices. This document was created using aspects of the following standards: DDR2 (JESD79-2), DDR3
(JESD79-3), DDR4 (JESD79-4), LPDDR (JESD209), LPDDR2 (JESD209-2) and LPDDR3 (JESD209-3).
Each aspect of the standard was considered and approved by committee ballot(s). The accumulation of
these ballots was then incorporated to prepare the LPDDR4 standard.
Page 2
2 Package Ballout & Pin Definition
2.1 Pad Order

Ch. A Top Ch. B Top
1 VDD2 41 VDD2 101 VDD2 141 VDD2
2 VSS 42 CKE_A 102 VSS 142 CKE_B
3 VDD1 43 CS_A 103 VDD1 143 CS_B
4 VDD2 44 VSS 104 VDD2 144 VSS
5 VSS 45 CA1_A 105 VSS 145 CA1_B
6 VSSQ 46 CA0_A 106 VSSQ 146 CA0_B
7 DQ8_A 47 VDD2 107 DQ8_B 147 VDD2
8 VDDQ 48 ODT(ca)_A 108 VDDQ 148 ODT(ca)_B
9 DQ9_A 49 VSS 109 DQ9_B 149 VSS
10 VSSQ 50 VDD1 110 VSSQ 150 VDD1
11 DQ10_A 51 VSSQ 111 DQ10_B 151 VSSQ
12 VDDQ 52 DQ7_A 112 VDDQ 152 DQ7_B
13 DQ11_A 53 VDDQ Top 113 DQ11_B 153 VDDQ
14 VSSQ 54 DQ6_A 114 VSSQ 154 DQ6_B
15 DQS1_t_A 55 VSSQ 115 DQS1_t_B 155 VSSQ
16 DQS1_c_A 56 DQ5_A 116 DQS1_c_B 156 DQ5_B
17 VDDQ 57 VDDQ 117 VDDQ 157 VDDQ
18 DMI1_A 58 DQ4_A 118 DMI1_B 158 DQ4_B
Channel A
Channel B
19 VSSQ 59 VSSQ 119 VSSQ 159 VSSQ

20 DQ12_A 60 DMI0_A 120 DQ12_B 160 DMI0_B
21 VDDQ 61 VDDQ 121 VDDQ 161 VDDQ
22 DQ13_A 62 DQS0_c_A 122 DQ13_B 162 DQS0_c_B
23 VSSQ 63 DQS0_t_A 123 VSSQ 163 DQS0_t_B
24 DQ14_A 64 VSSQ 124 DQ14_B 164 VSSQ
26 DQ15_A 66 VDDQ Boom 126 DQ15_B 166 VDDQ
27 VSSQ 67 DQ2_A 127 VSSQ 167 DQ2_B
28 ZQ 68 VSSQ 128 RESET_n 168 VSSQ
30 VDD2 70 VDDQ 130 VDD2 170 VDDQ
31 VDD1 71 DQ0_A 131 VDD1 171 DQ0_B
32 VSS 72 VSSQ 132 VSS 172 VSSQ
33 CA5_A 73 VSS 133 CA5_B 173 VSS
34 CA4_A 74 VDD2 134 CA4_B 174 VDD2
35 VDD2 75 VDD1 135 VDD2 175 VDD1
36 CA3_A 76 VSS 136 CA3_B 176 VSS
37 CA2_A 77 VDD2 137 CA2_B 177 VDD2
38 VSS Ch. A Bottom 138 VSS Ch. B Bottom
39 CK_c_A 139 CK_c_B
40 CK_t_A 140 CK_t_B
NOTE 1 Applications are recommended to follow bit/byte assignments. Bit or Byte swapping at the application level
requires review of MR and calibration features assigned to specific data bits/bytes.
NOTE 2 Additional pads are allowed for DRAM mfg-specific pads (“DNU”), or additional power pads as long as the
extra pads are grouped with like-named pads.
Page 3
2.2 Package Ballout
2.2.1 272-ball 15mm x 15mm 0.4mm pitch, Quad-Channel POP FBGA (top view)
Using Variation VFFCDB for MO-273
NOTE 1 15mm x 15mm, 0.4mm ball pitch

NOTE 2 272 ball count, 36 rows
NOTE 3 Top View, A1 in top left corner
NOTE 4 ODT ca_[x] balls are wired to ODT(ca)_[x] pads of Rank 0 DRAM die. ODT(ca)_[x] pads for other ranks (if
present) are disabled in the package.
NOTE 5 Package Channel a and Channel c shall be assigned to die Channel A of different DRAM die.
NOTE 6 Die pad VSS and VSSQ signals are combined to VSS package balls.
Page 4
2.2.2 200-ball x32 Discrete Package, 0.80mm x 0.65mm using MO-311
0.80mm Pitch
1 2 3 4 5 6 7 8 9 10 11 12
A DNU DNU VSS VDD2 ZQ0 ZQ1 VDD2 VSS DNU DNU
B DNU DQ0_A VDDQ DQ7_A VDDQ VDDQ DQ15_A VDDQ DQ8_A DNU
C VSS DQ1_A DMI0_A DQ6_A VSS VSS DQ14_A DMI1_A DQ9_A VSS
D VDDQ VSS DQS0_T_A VSS VDDQ VDDQ VSS DQS1_T_A VSS VDDQ
DQS0_C_ DQS1_C_
E VSS DQ2_A DQ5_A VSS VSS DQ13_A DQ10_A VSS
A A
F VDD1 DQ3_A VDDQ DQ4_A VDD2 VDD2 DQ12_A VDDQ DQ11_A VDD1
ODT_CA_
G VSS VSS VDD1 VSS VSS VDD1 VSS ZQ2 VSS
A
H VDD2 CA0_A CS1_A CS0_A VDD2 VDD2 CA2_A CA3_A CA4_A VDD2
J VSS CA1_A VSS CKE0_A CKE1_A CK_t_A CK_c_A VSS CA5_A VSS
0.65mm Pitch
K VDD2 VSS VDD2 VSS CS2_A CKE2_A VSS VDD2 VSS VDD2
N VDD2 VSS VDD2 VSS CS2_B CKE2_B VSS VDD2 VSS VDD2
P VSS CA1_B VSS CKE0_B CKE1_B CK_T_B CK_C_B VSS CA5_B VSS
R VDD2 CA0_B CS1_B CS0_B VDD2 VDD2 CA2_B CA3_B CA4_B VDD2
ODT_CA_
T VSS VSS VDD1 VSS VSS VDD1 VSS RESET_N VSS
B
U VDD1 DQ3_B VDDQ DQ4_B VDD2 VDD2 DQ12_B VDDQ DQ11_B VDD1
DQS0_C_ DQS1_C_
V VSS DQ2_B DQ5_B VSS VSS DQ13_B DQ10_B VSS
B B
W VDDQ VSS DQS0_T_B VSS VDDQ VDDQ VSS DQS1_T_B VSS VDDQ
Y VSS DQ1_B DMI0_B DQ6_B VSS VSS DQ14_B DMI1_B DQ9_B VSS
AA DNU DQ0_B VDDQ DQ7_B VDDQ VDDQ DQ15_B VDDQ DQ8_B DNU
AB DNU DNU VSS VDD2 VSS VSS VDD2 VSS DNU DNU
NOTE 1 0.8mm pitch (X-axis), 0.65mm pitch (Y-axis), 22 rows.

NOTE 2 Top View, A1 in top left corner.
NOTE 3 ODT_CA_[x] balls are wired to ODT_CA)_[x] pads of Rank 0 DRAM die. ODT(ca)_[x] pads for other ranks
(if present) are disabled in the package.
NOTE 4 ZQ2, CKE2_A, CKE2_B, CS2_A, and CS2_B balls are reserved for 3-rank package. For 1-rank and
2-rank package those balls are NC.
NOTE 5 Die pad VSS and VSSQ signals are combined to VSS package balls.
2.2.3 432-ball x64 HDI Discrete Package, 0.50mm x 0.50mm (MO: TBD)
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
A VDDQ VDD1 VDDQ VDDQ VDDQ VDD2 VDD2 VDDQ VDDQ VDDQ VDD1 VDDQ VDDQ VDD1 VDDQ VDDQ VDDQ VDD2 VDD2 VDDQ VDDQ VDDQ VDD1 VDDQ A
B VDDQ VDD1 DQ0_A VSS CA0_A VDD2 VDD2 CA4_A VSS DQ8_A VDD1 VDDQ VDDQ VDD1 DQ0_C VSS CA0_C VDD2 VDD2 CA4_C VSS DQ8_C VDD1 VDDQ B
C VDDQ DQ1_A VSS DQ5_A VSS CA2_A CA3_A VSS DQ13_A VSS DQ9_A VDDQ VDDQ DQ1_C VSS DQ5_C VSS CA2_C CA3_C VSS DQ13_C VSS DQ9_C VDDQ C
D VDDQ VSS DQ4_A VSS CA1_A VDD2 VDD2 CA5_A VSS DQ12_A VSS VDDQ VDDQ VSS DQ4_C VSS CA1_C VDD2 VDD2 CA5_C VSS DQ12_C VSS VDDQ D
E VDDQ DQ2_A VSS DQ6_A VSS CLK_t_A CLK_c_A VSS DQ14_A VSS DQ10_A VDDQ VDDQ DQ2_C VSS DQ6_C VSS CLK_t_C CLK_c_C VSS DQ14_C VSS DQ10_C VDDQ E
DQS0_t_ DQS1_t_ DQS0_t_ DQS1_t_

F VDDQ VSS VSS CS1_A VDD2 VDD2 CKE0_A VSS VSS VDDQ VDDQ VSS VSS CS1_C VDD2 VDD2 CKE0_C VSS VSS VDDQ F
A A C C
DQS0_c_ DQS1_c_ DQS0_c_ DQS1_c_
G VDDQ DQ3_A VSS VSS CS0_A CKE1_A VSS VSS DQ11_A VDDQ VDDQ DQ3_C VSS VSS CS0_C CKE1_C VSS VSS DQ11_C VDDQ G
A A C C
H VDDQ VSS DMI0_A VSS DQ7_A VDD2 VDD2 DQ15_A VSS DMI1_A VSS VDDQ VDDQ VSS DMI0_C VSS DQ7_C VDD2 VDD2 DQ15_C VSS DMI1_C VSS VDDQ H
ODT ODT
J VDDQ ZQ3_A ZQ2_A CS3_A CS2_A CKE3_A CKE2_A ZQ0_A VSS ZQ1_A VDDQ VDDQ ZQ3_C ZQ2_C CS3_C CS2_C CKE3_C CKE2_C ZQ0_C VSS ZQ1_C VDDQ J
ca_A ca_C
K K
NOTE 1 0.5 mm ball pitch. MO-TBD

L L
NOTE 2 432 ball count
M NOTE 3 Top view, A1 in top left corner M
NOTE 4 ODT ca_[x] balls are wired to ODT(ca)_[x] pads of Rank 0 DRAM die. ODT(ca)_[x] pads for other ranks (if present) are disabled in the package
N N
NOTE 5 Package Channel A and Channel C shall be assigned to die Channel A of different DRAM die
P NOTE 6 ZQ2, CKE2_A, CKE2_B, CS2_A, and CS2_B balls are reserved for 3-rank package. ZQ3, CKE3_A, P
CKE3_B, CS3_A, and CS3_B balls are reserved for 4-rank package. For 1-rank and 2-rank package those balls are NC
R R
NOTE 7 Die pad VSS and VSSQ signals are combined to VSS package balls
T T
ODT ODT
V VDDQ VSS VSS CS3_B CS2_B CKE3_B CKE2_B VSS VSS RESET_n VDDQ VDDQ VSS VSS CS3_D CS2_D CKE3_D CKE2_D VSS VSS NC VDDQ V
ca_B ca_D
W VDDQ VSS DMI0_B VSS DQ7_B VDD2 VDD2 DQ15_B VSS DMI1_B VSS VDDQ VDDQ VSS DMI0_D VSS DQ7_D VDD2 VDD2 DQ15_D VSS DMI1_D VSS VDDQ W
DQS0_c_ DQS1_c_ DQS0_c_ DQS1_c_

Y VDDQ DQ3_B VSS VSS CS0_B CKE1_B VSS VSS DQ11_B VDDQ VDDQ DQ3_D VSS VSS CS0_D CKE1_D VSS VSS DQ11_D VDDQ Y
B B D D
DQS0_t_ DQS1_t_ DQS0_t_ DQS1_t_
JEDEC Standard No.209-4

AA VDDQ VSS VSS CS1_B VDD2 VDD2 CKE0_B VSS VSS VDDQ VDDQ VSS VSS CS1_D VDD2 VDD2 CKE0_D VSS VSS VDDQ AA
B B D D
AB VDDQ DQ2_B VSS DQ6_B VSS CLK_t_B CLK_c_B VSS DQ14_B VSS DQ10_B VDDQ VDDQ DQ2_D VSS DQ6_D VSS CLK_t_D CLK_c_D VSS DQ14_D VSS DQ10_D VDDQ AB
AC VDDQ VSS DQ4_B VSS CA1_B VDD2 VDD2 CA5_B VSS DQ12_B VSS VDDQ VDDQ VSS DQ4_D VSS CA1_D VDD2 VDD2 CA5_D VSS DQ12_D VSS VDDQ AC
AD VDDQ DQ1_B VSS DQ5_B VSS CA2_B CA3_B VSS DQ13_B VSS DQ9_B VDDQ VDDQ DQ1_D VSS DQ5_D VSS CA2_D CA3_D VSS DQ13_D VSS DQ9_D VDDQ AD
AE VDDQ VDD1 DQ0_B VSS CA0_B VDD2 VDD2 CA4_B VSS DQ8_B VDD1 VDDQ VDDQ VDD1 DQ0_D VSS CA0_D VDD2 VDD2 CA4_D VSS DQ8_D VDD1 VDDQ AE
AF VDDQ VDD1 VDDQ VDDQ VDDQ VDD2 VDD2 VDDQ VDDQ VDDQ VDD1 VDDQ VDDQ VDD1 VDDQ VDDQ VDDQ VDD2 VDD2 VDDQ VDDQ VDDQ VDD1 VDDQ AF
Page 5
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
Page 6
2.3 Pad Definition and Description
Table 1 — Pad Definition and Description

Symbol Type Description
CK_t_A, Clock: CK_t and CK_c are differential clock inputs. All address, command,
CK_c_A, and control input signals are sampled on the crossing of the positive edge of
Input
CK_t_B, CK_t and the negative edge of CK_c. AC timings for CA parameters are
CK_c_B referenced to CK. Each channel (A & B) has its own clock pair.
Clock Enable: CKE HIGH activates and CKE LOW deactivates the internal
CKE_A clock circuits, input buffers, and output drivers. Power-saving modes are
Input
CKE_B entered and exited via CKE transitions. CKE is part of the command code.
Each channel (A & B) has its own CKE signal.
CS_A Chip Select: CS is part of the command code. Each channel (A & B) has its
Input
CS_B own CS signal.
Command/Address Inputs: CA signals provide the Command and Address
CA[5:0]_A
Input inputs according to Table 63 — Command Truth Table. Each channel (A&B)
CA[5:0]_B
has its own CA signals.
ODT_CA_A CA ODT Control: The ODT_CA pin is used in conjunction with the Mode
Input Register to turn on/off the On-Die-Termination for CA pins.
ODT_CA_B
DQ[15:0]_A, Data Input/Output: Bi-direction data bus.
I/O
DQ[15:0]_B
Data Strobe: DQS_t and DQS_c are bi-directional differential output clock
DQS[1:0]_t_A, signals used to strobe data during a READ or WRITE. The Data Strobe is
DQS[1:0]_c_A, generated by the DRAM for a READ and is edge-aligned with Data. The Data
I/O Strobe is generated by the Memory Controller for a WRITE and must arrive
DQS[1:0]_t_B,
DQS[1:0]_c_B prior to Data. Each byte of data has a Data Strobe signal pair. Each channel
(A & B) has its own DQS strobes.
Data Mask Inversion: DMI is a bi-directional signal which is driven HIGH

when the data on the data bus is inverted, or driven LOW when the data is in
DMI[1:0]_A, its normal state. Data Inversion can be disabled via a mode register setting.
I/O
DMI[1:0]_B Each byte of data has a DMI signal. Each channel (A & B) has its own DMI
signals.
Calibration Reference: Used to calibrate the output drive strength and the
ZQ Reference termination resistance. There is one ZQ pin per die. The ZQ pin shall be
connected to VDDQ through a 240Ω ± 1% resistor.
VDDQ,
VDD1, Supply Power Supplies: Isolated on the die for improved noise immunity.
VDD2
VSS, VSSQ GND Ground Reference: Power supply ground reference
RESET: When asserted LOW, the RESET_n signal resets both channels of
RESET_n Input
the die.
Page 7
3 Functional Description
LPDDR4-SDRAM is a high-speed synchronous DRAM device internally configured as 2-channel and 8-

bank per channel memory that is up to 16Gb density. The configuration for the device density that is
greater than 16Gb is still TBD1.
These devices contain the following number of bits:
4Gb has 4,294,967,296 bits

6Gb has 6,442,450,944 bits
8Gb has 8,589,934,592 bits
12Gb has 12,884,901,888 bits
16Gb has 17,179,869,184 bits
24Gb has 25,769,803,776 bits
32Gb has 34,359,738,368 bits
LPDDR4 devices use a 2 or 4 clocks architecture on the Command/Address (CA) bus to reduce the
number of input pins in the system. The 6-bit CA bus contains command, address, and bank information.
Each command uses 1, 2 or 4 clock cycle, during which command information is transferred on the positive
edge of the clock. See Table 63 — Command Truth Table, for details.
These devices use a double data rate architecture on the DQ pins to achieve high speed operation. The
double data rate architecture is essentially an 16n prefetch architecture with an interface designed to
transfer two data bits per DQ every clock cycle at the I/O pins. A single read or write access for the
LPDDR4 SDRAM effectively consists of a single 16n-bit wide, one clock cycle data transfer at the internal
DRAM core and eight corresponding n-bit wide, one half-clock-cycle data transfers at the I/O pins. Read
and write accesses to the LPDDR4 SDRAMs are burst oriented; accesses start at a selected location and
continue for a programmed number of locations in a programmed sequence. Accesses begin with the
registration of an Activate command, which is then followed by a Read, Write or Mask Write command.
The address and BA bits registered coincident with the Activate command are used to select the row and
the bank to be accessed. The address bits registered coincident with the Read, Write or Mask Write
command are used to select the bank and the starting column location for the burst access.
Prior to normal operation, the LPDDR4 SDRAM must be initialized. The following provides detailed
information covering device initialization, register definition, command description and device operation.
1. As of publication of this document, under discussion by the formulating committee.

Page 8
3.1 LPDDR4 SDRAM Addressing

Table 2 — LPDDR4 SDRAM Addressing
Memory
Density 4Gb 6Gb 8Gb 12Gb 16Gb 24Gb 32Gb
(per Die)
Memory
Density 2Gb 3Gb 4Gb 6Gb 8Gb 12Gb 16Gb
(per channel)
16Mb x 16DQ 24Mb x 16DQ 32Mb x 16DQ 48Mb x 16DQ 64Mb x 16DQ TBD x 16DQ TBD x 16DQ
Configuration x 8 banks x 8 banks x 8 banks x 8 banks x 8 banks x TBD banks x TBD banks
x 2 channels x 2 channels x 2 channels x 2 channels x 2 channels x 2 channels x 2 channels
Number of
Channels 2 2 2 2 2 2 2
(per die)
Number of
Banks 8 8 8 8 8 TBD TBD
(per channel)
Array
Pre-Fetch
256 256 256 256 256 256 256
(bits,
per channel)
Number of
Rows 16,384 24,576 32,768 49,152 65,536 TBD TBD
(per channel)
Number of
Columns
64 64 64 64 64 TBD TBD
(fetch
boundaries)
Page Size
2048 2048 2048 2048 2048 TBD TBD
(Bytes)
Channel
Density
2,147,483,648 3,221,225,472 4,294,967,296 6,442,450,944 8,589,934,592 12,884,901,888 17,179,869,184
(Bits per
channel)
Total Density
4,294,967,296 6,442,450,944 8,589,934,592 12,884,901,888 17,179,869,184 25,769,803,776 34,359,738,368
(Bits per die)
Bank Address BA0 - BA2 BA0 - BA2 BA0 - BA2 BA0 - BA2 BA0 - BA2 TBD TBD
R0 - R14 R0 - R15
Row
R0 - R13 (R13=0 when R0 - R14 (R14=0 when R0 - R15 TBD TBD
Addresses
x16 R14=1) R15=1)
Column
C0 - C9 C0 - C9 C0 - C9 C0 - C9 C0 - C9 TBD TBD
Addresses
Burst Starting
Address 64 - bit 64 - bit 64 - bit 64 - bit 64 - bit 64 - Bit 64 - bit
Boundary
NOTE 1 The lower two column addresses (C0 - C1) are assumed to be “zero” and are not transmitted on the CA bus.
NOTE 2 Row and Column address values on the CA bus that are not used for a particular density be at valid logic levels.
NOTE 3 For non - binary memory densities,only half of the row address space is valid. When the MSB address bit is “HIGH”, then
the MSB - 1 address bit must be “LOW”.
NOTE 4 TBD, as of publication of this document, under discussion by the formulating committee.
Page 9
3.2 Simplified LPDDR4 State Diagram
LPDDR4-SDRAM state diagram provides a simplified illustration of allowed state transitions and the
related commands to control them. For a complete definition of the device behavior, the information
provided by the state diagram should be integrated with the truth tables and timing specification.
The truth tables provide complementary information to the state diagram, they clarify the device behavior
and the applied restrictions when considering the actual state of all the banks.
For the command definition, see clause 4, Command Definition and Timing Diagram.
Page 10
Power
On
Re = L
se
SR
t _n
Power Command Sequence
Down Automatic Sequence
MPC
Based
CK
Training
E=
Reset Per
Command
L
MRW Bank
Bus REF
MRW Refresh
Training
Re = H
MPC MRW
CK
se
MRW
E=
t_
n
H
MPC SRE REF All MPC MPC

Self
Based Idle Bank Based
Refresh
Training MPC SRX Refresh Training
PDX MRW
ACT
MRR MRW Command
PDE MRW
Idle MRR Bus
MRR Power Training
Down
MRR
MRW
Activating
MRR
MRW
Active
Power MRW
Down PDE MRR
MRR
PDX
MPC
Bank
Based
Active
Training REF
WR or Per
MWR RD Bank
Refresh
WR or RD
MWR
Write
or Read
MWR
WRA or
MWRA
RDA
WRA or
RDA
MWRA
PRE or PRE(A) = Precharge (All)
PREA ACT = Activate
WR(A) = Write (with Autoprecharge)
Write or MWR(A) = Mask-Write (with Autoprecharge)
PRE or PRE or Read
MWR RD(A) = Read (with Autoprecharge)
PREA PREA with Auto-
with Auto- MRW = Mode Register Write
Precharge
Precharge MRR = Mode Register Read
PDE = Enter Power Down
PDX = Exit Power Down
SRE = Enter Self Refresh
Pre- SRX = Exit Self Refresh
charging REF = Refresh
MPC = Multi-Purpose Command (w/NOP)
Figure 1 — LPDDR4: Simplified Bus Interface State Diagram-1

Page 11
A) FIFO Based Write / Read Timing
MPC MPC
MPC MPC
FIFO FIFO
WRTR RDTR
MPC MPC MRW

Based MPC
Training
MRW
WRTR
MRW
B) Read DQ Calibration
MPC
MPC DQ
Calibration
C) ZQ CAL Start
MPC ZQ
Calibration
Start
D) ZQ CAL Latch
MPC ZQ
Calibration
Latch
NOTE 1 From the Self-Refresh state the device can enter Power-Down, MRR, MRW, or MPC states. See 4.13, on
Self-Refresh for more information.
NOTE 2 In IDLE state, all banks are pre-charged.
NOTE 3 In the case of a MRW command to enter a training mode, the state machine will not automatically return to
the IDLE state at the conclusion of training. See 4.17, on Mode Register Write (MRW) for more information.
NOTE 4 In the case of a MPC command to enter a training mode, the state machine may not automatically return to
the IDLE state at the conclusion of training. See 4.28, Multi-Purpose Command (MPC) for more information.
NOTE 5 This simplified State Diagram is intended to provide an overview of the possible state transitions and the
commands to control them. In particular, situations involving more than one bank, the enabling or disabling of on-die
termination, and some other events are not captured in full detail.
NOTE 6 States that have an “automatic return” and can be accessed from more than one prior state (Ex. MRW from
either Idle or Active states) will return to the state from when they were initiated (Ex. MRW from Idle will return to Idle).
NOTE 7 The RESET_n pin can be asserted from any state, and will cause the SDRAM to go to the Reset State. The
diagram shows RESET applied from the Power-On as an example, but the Diagram should not be construed as a
restriction on RESET_n.
Figure 2 — LPDDR4: Simplified Bus Interface State Diagram-2

Page 12
3.3 Power-up, Initialization and Power-off Procedure
For power-up and reset initialization, in order to prevent DRAM from functioning improperly, default values
of the following MR settings are defined as Table 3.
Table 3 — MRS defaults settings

Default
Item MRS Description
Setting
FSP-OP/WR MR13 OP[7:6] 00B FSP-OP/WR[0] are enabled
WLS MR2 OP[6] 0B Write Latency Set 0 is selected
WL MR2 OP[5:3] 000B WL = 4
RL MR2 OP[2:0] 000B RL = 6, nRTP=8
nWR MR1 OP[6:4] 000B nWR = 6
DBI-WR/RD MR3 OP[7:6] 00B Write & Read DBI are disabled
CA ODT MR11 OP[6:4] 000B CA ODT is disabled
DQ ODT MR11 OP[2:0] 000B DQ ODT is disabled
VREF(CA) Setting MR12 OP[6] 1B VREF(CA) Range[1] enabled
VREF(CA) Value MR12 OP[5:0] 001101B Range1 : 27.2% of VDD2
VREF(DQ) Setting MR14 OP[6] 1B VREF(DQ) Range[1] enabled
VREF(DQ) Value MR14 OP[5:0] 001101B Range1 : 27.2% of VDDQ
3.3.1 Voltage Ramp and Device Initialization
The following sequence shall be used to power up the LPDDR4 device. Unless specified otherwise, these
steps are mandatory. Note that the power-up sequence of all channels must proceed simultaneously.
1. While applying power (after Ta), RESET_n is recommended to be LOW (≤0.2 x VDD2) and all other
inputs must be between VILmin and VIHmax. The device outputs remain at High-Z while RESET_n is
held LOW. Power supply voltage ramp requirements are provided in Table 4. VDD1 must ramp at the
same time or earlier than VDD2. VDD2 must ramp at the same time or earlier than VDDQ.
Table 4 — Voltage Ramp Conditions

After Applicable Conditions
VDD1 must be greater than VDD2
Ta is reached
VDD2 must be greater than VDDQ - 200mV
NOTE 1 Ta is the point when any power supply first reaches 300mV.
NOTE 2 Voltage ramp conditions in Table 4 apply between Ta and power-off (controlled or uncontrolled).
NOTE 3 Tb is the point at which all supply and reference voltages are within their defined ranges.
NOTE 4 Power ramp duration tINIT0 (Tb-Ta) must not exceed 20ms.
NOTE 5 The voltage difference between any of Vss and Vssq pins must not excess 100mV.
2. Following the completion of the voltage ramp (Tb), RESET_n must be maintained LOW. DQ, DMI,
DQS_t and DQS_c voltage levels must be between Vssq and Vddq during voltage ramp to avoid
latch-up. CKE, CK_t, CK_c, CS_n and CA input levels must be between Vss and VDD2 during voltage
ramp to avoid latch-up.
3. Beginning at Tb, RESET_n must remain LOW for at least tINIT1(Tc), after which RESET_n can be
de-asserted to HIGH(Tc). At least 10ns before RESET_n de-assertion, CKE is required to be set LOW.
All other input signals are "Don't Care".
Page 13
Ta Tb Tc Td Te Tf Tg Th Ti Tj Tk
Power Ramp Reset Initialization Training
tINIT4=5tCK(min)
CK_c
CK_t
tINIT0=20ms(max) tINIT1=200us(min)
Supplies
Reset_n
tINIT2=10ns(min) tINIT3=2ms(min)
CKE
tINIT5=2us(min) tZQCAL = 1us(min) tZQLAT = Max(30ns, 8tCK)(Min)
CA[5:0] MRW ZQ Cal ZQ Cal CA BUS Write DQ

Exit PD DES MRR DES Start DES Latch DES Training DES Leveling DES Training DES Valid
CS
DQs Valid Valid Valid
NOTES : 1. Training is optional and may be done at the system architects discretion. The training sequence after ZQ_CAL Latch(Th, Sequence7~9) in Figure 1 is simplified recommendation and actual training
sequence may vary depending on systems.
Figure 3 — Power Ramp and Initialization Sequence
4. After RESET_n is de-asserted(Tc), wait at least tINIT3 before activating CKE. Clock(CK_t,CK_c) is required to be started and stabilized for tINIT4
before CKE goes active(Td). CS is required to be maintained LOW when controller activates CKE.
5. After setting CKE high, wait minimum of tINIT5 to issue any MRR or MRW commands(Te). For both MRR and MRW commands, the clock frequency
must be within the range defined for tCKb. Some AC parameters (for example, tDQSCK) could have relaxed timings (such as tDQSCKb) before the
system is appropriately configured.
6. After completing all MRW commands to set the Pull-up, Pull-down and Rx termination values, the DRAM controller can issue ZQCAL Start command
to the memory(Tf). This command is used to calibrate VOH level and output impedance over process, voltage and temperature. In systems where
more than one LPDDR4 DRAM devices share one external ZQ resistor, the controller must not overlap the ZQ calibration sequence of each
LPDDR4 device. ZQ calibration sequence is completed after tZQCAL (Tg) and the ZQCAL Latch command must be issued to update the DQ drivers
and DQ+CA ODT to the calibrated values.
Page 14
3.3.1 Voltage Ramp and Device Initialization
7. After tZQLAT is satisfied (Th) the command bus (internal VREF(ca), CS, and CA) should be trained for
high-speed operation by issuing an MRW command (Command Bus Training Mode). This command is
used to calibrate the device's internal VREF and align CS/CA with CK for high-speed operation. The
LPDDR4 device will power-up with receivers configured for low-speed operations, and VREF(ca) set to a
default factory setting. Normal device operation at clock speeds higher than tCKb may not be possible
until command bus training has been completed.
NOTE The command bus training MRW command uses the CA bus as inputs for the calibration data stream, and
outputs the results asynchronously on the DQ bus. See 4.21, (item 1.), MRW for information on how to enter/exit the
training mode.
8. After command bus training, DRAM controller must perform write leveling. Write leveling mode is
enabled when MR2 OP[7] is high (Ti). See 4.23, Mode Register Write-WR Leveling Mode, for detailed
description of write leveling entry and exit sequence. In write leveling mode, the DRAM controller adjusts
write DQS_t/_c timing to the point where the LPDDR4 device recognizes the start of write DQ data burst
with desired write latency.
9. After write leveling, the DQ Bus (internal VREF(dq), DQS, and DQ) should be trained for high-speed
operation using the MPC training commands and by issuing MRW commands to adjust VREF(dq)(Tj).
The LPDDR4 device will power-up with receivers configured for low-speed operations and VREF(dq) set
to a default factory setting. Normal device operation at clock speeds higher than tCKb should not be
attempted until DQ Bus training has been completed. The MPC Read Calibration command is used
together with MPC FIFO Write/Read commands to train DQ bus without disturbing the memory array
contents. See 4.25, DQ Bus Training for detailed DQ Bus Training sequence.
10. At Tk the LPDDR4 device is ready for normal operation, and is ready to accept any valid command.
Any more registers that have not previously been set up for normal operation should be written at this
time.
Table 5 — Initialization Timing Parameters

Value
Parameter Unit Comment
Min Max
tINIT0 - 20 ms Maximum voltage-ramp time
Minimum RESET_n LOW time after completion of
tINIT1 200 - us
voltage ramp
tINIT2 10 - ns Minimum CKE low time before RESET_n high
tINIT3 2 - ms Minimum CKE low time after RESET_n high

tINIT4 5 - tCK Minimum stable clock before first CKE high
Minimum idle time before first MRW/MRR
tINIT5 2 - us
command
tZQCAL 1 - us ZQ calibration time
tZQLAT Max(30ns, 8tCK) - ns ZQCAL latch quiet time.
tCKb Note *1,2 Note *1,2 ns Clock cycle time during boot
NOTE 1 Min tCKb guaranteed by DRAM test is 18ns.
NOTE 2 The system may boot at a higher frequency than dictated by min tCKb. The higher boot frequency is system
dependent.
Page 15
3.3.2 Reset Initialization with Stable Power

The following sequence is required for RESET at no power interruption initialization.
1. Assert RESET_n below 0.2 x VDD2 anytime when reset is needed. RESET_n needs to be maintained
for minimum tPW_RESET. CKE must be pulled LOW at least 10 ns before de-asserting RESET_n.
2. Repeat steps 4 to 10 in 3.3.1, Voltage Ramp and Device Initialization.
Table 6 — Reset Timing Parameter

Value
Parameter Unit Comment
Min Max
Minimum RESET_n low Time for Reset
tPW_RESET 100 - ns
Initialization with stable power
3.3.3 Power-off Sequence

The following procedure is required to power off the device.
While powering off, CKE must be held LOW (≤0.2 X VDD2) and all other inputs must be between VILmin
and VIHmax. The device outputs remain at High-Z while CKE is held LOW. DQ, DMI, DQS_t and DQS_c
voltage levels must be between Vssq and Vddq during voltage ramp to avoid latch-up. RESET_n, CK_t,
CK_c, CS and CA input levels must be between VSS and VDD2 during voltage ramp to avoid latch-up.
Tx is the point where any power supply drops below the minimum value specified.
Tz is the point where all power supplies are below 300mV. After TZ, the device is powered off.
Table 7 — Power Supply Conditions

After Applicable Conditions
VDD1 must be greater than VDD2
Tx and Tz
VDD2 must be greater than VDDQ - 200mV
The voltage difference between any of VSS, VSSQ pins must not exceed 100mV.
3.3.4 Uncontrolled Power-Off Sequence

When an uncontrolled power-off occurs, the following conditions must be met:
At Tx, when the power supply drops below the minimum values specified, all power supplies must be
turned off and all power supply current capacity must be at zero, except any static charge remaining in the
system.
After Tz (the point at which all power supplies first reach 300mV), the device must power off. During this
period the relative voltage between power supplies is uncontrolled. VDD1 and VDD2 must decrease with a
slope lower than 0.5V/μs between Tx and Tz.
An uncontrolled power-off sequence can occur a maximum of 400 times over the life of the device.
Table 8 — Timing Parameters Power Off

Value
Symbol Unit Comment
Min Max
tPOFF - 2 s Maximum Power-off ramp item
Page 16
3.4 Mode Register Definition
Table 9 shows the mode registers for LPDDR4 SDRAM. Each register is denoted as "R" if it can be read
but not written, "W" if it can be written but not read, and "R/W" if it can be read and written. A Mode
Register Read command is used to read a mode register. A Mode Register Write command is used to write
a mode register.
Table 9 — Mode Register Assignment in LPDDR4 SDRAM
MR# OP[7] OP[6] OP[5] OP[4] OP[3] OP[2] OP[1] OP[0]
0 CATR RFU RFU RZQI RFU RFU Refresh mode
1 RPST nWR (for AP) RD-PRE WR-PRE BL
2 WR Lev WLS WL RL
3 DBI-WR DBI-RD PDDS PPRP WR PST PU-CAL
4 TUF Thermal Offset PPRE SR Abort Refresh Rate
5 LPDDR4 Manufacturer ID
6 Revision ID-1
7 Revision ID-2
8 IO Width Density Type
9 Vendor Specific Test Register
10 RFU ZQ-Reset
11 RFU CA ODT RFU DQ ODT
12 RFU VR-CA VREF(ca)
13 FSP-OP FSP-WR DMD RRO VRCG VRO RPT CBT
14 RFU VR(dq) VREF(dq)
15 Lower-Byte Invert Register for DQ Calibration
16 PASR Bank Mask
17 PASR Segment Mask
18 DQS Oscillator Count - LSB
19 DQS Oscillator Count - MSB
20 Upper-Byte Invert Register for DQ Calibration
21 RFU
22 RFU ODTD-CA ODTE-CS ODTE-CK CODT
23 DQS interval timer run time setting
Unlimited
24 TRR Mode TRR Mode BAn MAC Value
MAC
25 PPR Resource
26 RFU
27 RFU
28 RFU
29 RFU
30 RFU
31 RFU
32 DQ Calibration Pattern “A” (default = 5AH)
33 RFU
34 RFU
35 RFU
36 RFU
37 RFU
38 RFU
39 RFU
40 DQ Calibration Pattern “B” (default = 3CH)
Page 17
3.4.1 MR0 Register Information (MA [7:0] = 00H)
OP[7] OP[6] OP[5] OP[4] OP[3] OP[2] OP[1] OP[0]

Refresh
CATR RFU RFU RZQI RFU
mode
Register
Function Operand Data Notes
Type
0B : Both legacy & modified refresh
Refresh mode OP[0] mode supported
1B : Only modified refresh mode supported
00B: RZQ Self-Test Not Supported

01B: ZQ pin may connect to VSSQ or float
RZQI Read-only 10B: ZQ-pin may short to VDDQ
OP[4:3] 1,2,3,4
(Built-in Self-Test for RZQ) 11B: ZQ-pin Self-Test Completed, no error
condition detected (ZQ-pin may not connect to
VSSQ or float, nor short to VDDQ)
CATR 0B: CA for this rank is not terminated

OP[7] 5
(CA Terminating Rank) 1B: CA for this rank is terminated
NOTE 1 RZQI, if supported, will be set upon the completion of the MRW ZQ Initialization Calibration command.
NOTE 2 If the ZQ-pin is connected to VSSQ to set default calibration, OP[4:3] shall be set to 01B. If the ZQ-pin is not
connected to VSSQ, either OP[4:3] = 01B or OP[4:3] = 10B might indicate might indicate a ZQ-pin assembly error. It is
recommended that the assembly error is corrected.
NOTE 3 In the case of possible assembly error, the LPDDR4-SDRAM device will default to factory trim settings for
RON, and will ignore ZQ Calibration commands. In either case, the device may not function as intended.
NOTE 4 If ZQ Self-Test returns OP[4:3] = 11B, the device has detected a resistor connected to the ZQ-pin. However,
this result cannot be used to validate the ZQ resistor value or that the ZQ resistor tolerance meets the specified limits
(i.e., 240Ω ± 1%).
NOTE 5 OP[7] is set at power-up, according to the state of the CA-ODT pad on the die AND the state of MR11 OP[7].
If the CAODT pad is tied LOW, then the die will not terminate the CA bus and MR12 OP[7]=0B, regardless of the state
of ODTECA (MR11 OP[7]). If the CA-ODT pad is tied HIGH AND ODTE-CA is enabled (MR11 OP[7]=1B), then this bit
will be set (MR0 OP[7]=1B) and the die will terminate the CA bus.
Page 18
3.4.2 MR1 Register Information (MA[7:0] = 01H)

RPST nWR (for AP) RD-PRE WR-PRE BL
Register
Type
00B: BL=16 Sequential (default)
BL 01B: BL=32 Sequential
OP[1:0] 1,5,6
(Burst Length) 10B: BL=16 or 32 Sequential (on-the-fly)
All Others: Reserved
WR-PRE 0B: Reserved
OP[2] 5,6
(WR Pre-amble Length) 1B: WR Pre-amble = 2*tCK
RD-PRE 0B: RD Pre-amble = Static (default)
OP[3] 3,5,6
(RD Pre-amble Type) 1B: RD Pre-amble = Toggle
000B: nWR = 6 (default)
Write-only
001B: nWR = 10
010B: nWR = 16
nWR 011B: nWR = 20
(Write-Recovery for Auto- OP[6:4] 2,5,6
100B: nWR = 24
Pre-charge commands)
101B: nWR = 30
110B: nWR = 34
111B: nWR = 40
RPST 0B: RD Post-amble = 0.5*tCK (default)
OP[7] 4,5,6
(RD Post-Amble Length) 1B: RD Post-amble = 1.5*tCK
NOTE 1 Burst length on-the-fly can be set to either BL=16 or BL=32 by setting the “BL” bit in the command operands.
See Table 63 — Command Truth Table.
NOTE 2 The programmed value of nWR is the number of clock cycles the LPDDR4-SDRAM device uses to
determine the starting point of an internal Pre-charge operation after a Write burst with AP (auto-pre-charge) enabled.
(Ref. See Latency Code Frequency Table for allowable Frequency Ranges for RL/WL/nWR, available in next revision
of this document)
NOTE 3 For Read operations this bit must be set to select between a "toggling" pre-amble and a "Non-toggling"
Pre-amble. See 4.4, Read Preamble and Postambil, for a drawing of each type of pre-amble.
NOTE 4 OP[7] provides an optional READ post-amble with an additional rising and falling edge of DQS_t. The
optional postamble cycle is provided for the benefit of certain memory controllers.
NOTE 5 There are two physical registers assigned to each bit of this MR parameter, designated set point 0 and set
point 1. Only the registers for the set point determined by the state of the FSP-WR bit (MR13 OP[6]) will be written to
with an MRW command to this MR address, or read from with an MRR command to this address.
point 1. The device will operate only according to the values stored in the registers for the active set point, i.e., the set
point determined by the state of the FSP-OP bit (MR13 OP[7]). The values in the registers for the inactive set point will
be ignored by the device, and may be changed without affecting device operation.
Page 19
Table 10 — Burst Sequence for READ
Burst Burst Burst Cycle Number and Burst Address Sequence
C4 C3 C2 C1 C0
Length Type 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 32
V 0 0 0 0 0 1 2 3 4 5 6 7 8 9 A B C D E F
V 0 1 0 0 4 5 6 7 8 9 A B C D E F 0 1 2 3
16 SEQ
V 1 0 0 0 8 9 A B C D E F 0 1 2 3 4 5 6 7
V 1 1 0 0 C D E F 0 1 2 3 4 5 6 7 8 9 A B
0 0 0 0 0 0 1 2 3 4 5 6 7 8 9 A B C D E F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
0 0 1 0 0 4 5 6 7 8 9 A B C D E F 0 1 2 3 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 10 11 12 13
0 1 0 0 0 8 9 A B C D E F 0 1 2 3 4 5 6 7 18 19 1A 1B 1C 1D 1E 1F 10 11 12 13 14 15 16 17
0 1 1 0 0 C D E F 0 1 2 3 4 5 6 7 8 9 A B 1C 1D 1E 1F 10 11 12 13 14 15 16 17 18 19 1A 1B
32 SEQ
1 0 0 0 0 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 0 1 2 3 4 5 6 7 8 9 A B C D E F
1 0 1 0 0 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 10 11 12 13 4 5 6 7 8 9 A B C D E F 0 1 2 3
1 1 0 0 0 18 19 1A 1B 1C 1D 1E 1F 10 11 12 13 14 15 16 17 8 9 A B C D E F 0 1 2 3 4 5 6 7
1 1 1 0 0 1C 1D 1E 1F 10 11 12 13 14 15 16 17 18 19 1A 1B C D E F 0 1 2 3 4 5 6 7 8 9 A B
NOTE 1 C0-C1 are assumed to be '0', and are not transmitted on the command bus.
NOTE 2 The starting burst address is on 64-bit (4n) boundaries.
Table 11 — Burst Sequence for Write

Burst Burst Burst Cycle Number and Burst Address Sequence
C4 C3 C2 C1 C0
Length Type 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 32
16 SEQ V 0 0 0 0 0 1 2 3 4 5 6 7 8 9 A B C D E F
32 SEQ 0 0 0 0 0 0 1 2 3 4 5 6 7 8 9 A B C D E F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
NOTE 1 C0-C1 are assumed to be '0', and are not transmitted on the command bus.
NOTE 2 The starting burst address is on 64-bit (4n) boundaries.
NOTE 3 C2-C3 shall be set to '0' for all Write operations.
Page 20

WR Lev WLS WL RL
Register
Type
RL & nRTP for DBI-RD Disabled (MR3 OP[6]=0B)
000B: RL=6, nRTP = 8 (Default)
001B: RL=10, nRTP = 8
010B: RL=14, nRTP = 8
011B: RL=20, nRTP = 8
100B: RL=24, nRTP = 10
101B: RL=28, nRTP = 12
110B: RL=32, nRTP = 14
RL 111B: RL=36, nRTP = 16
OP[2:0] 1,3,4
(Read latency) RL & nRTP for DBI-RD Enabled (MR3 OP[6]=1B)
000B: RL=6, nRTP = 8
001B: RL=12, nRTP = 8
010B: RL=16, nRTP = 8
011B: RL=22, nRTP = 8
100B: RL=28, nRTP = 10
101B: RL=32, nRTP = 12
110B: RL=36, nRTP = 14
111B: RL=40, nRTP = 16
WL Set "A” (MR2 OP[6]=0B)
Write-only 000B: WL=4 (Default)
001B: WL=6
010B: WL=8
011B: WL=10
100B: WL=12
101B: WL=14
110B: WL=16
WL 111B: WL=18
OP[5:3] 1,3,4
(Write latency) WL Set "B" (MR2 OP[6]=1B)
000B: WL=4
001B: WL=8
010B: WL=12
011B: WL=18
100B: WL=22
101B: WL=26
110B: WL=30
111B: WL=34
WLS 0B: WL Set "A" (default)
OP[6] 1,3,4
(Write Latency Set) 1B: WL Set "B"
WR LEV 0B: Disabled (default)
OP[7] 2
(Write Leveling) 1B: Enabled
NOTE 1 (Ref. See Latency Code Frequency Table for allowable Frequency Ranges for RL/WL/nWR/nRTP, available in next revision of this
document)
NOTE 2 See Latency Code Frequency Table for allowable frequency ranges for RL/WL/nWR/nRTP. (The next revision of this document should
contain a Table for "Latency Code Frequency")
NOTE 3 After a MRW to set the Write Leveling Enable bit (OP[7]=1B), the LPDDR4-SDRAM device remains in the MRW state until another MRW
command clears the bit (OP[7]=0B). No other commands are allowed until the Write Leveling Enable bit is cleared.
NOTE 4 There are two physical registers assigned to each bit of this MR parameter, designated set point 0 and set point 1. Only the registers for
the set point determined by the state of the FSP-WR bit (MR13 OP[6]) will be written to with an MRW command to this MR address, or read from
with an MRR command to this address.
NOTE 5 There are two physical registers assigned to each bit of this MR parameter, designated set point 0 and set point 1. The device will
operate only according to the values stored in the registers for the active set point, i.e., the set point determined by the state of the FSP-OP bit
(MR13 OP[7]). The values in the registers for the inactive set point will be ignored by the device, and may be changed without affecting device
operation.
Page 21

DBI-WR DBI-RD PDDS PPRP WR PST PU-CAL
Register
Type
PU-Cal 0B: VDDQ/2.5
OP[0] 1,4
(Pull-up Calibration Point) 1B: VDDQ/3 (default)
0B: WR Post-amble = 0.5*tCK (default)
WR PST(WR Post-Amble
OP[1] 1B: WR Post-amble = 1.5*tCK(Vendor specific 2,3,5
Length)
function)
Post Package Repair 0B: PPR protection disabled (default)

OP[2] 6
Protection 1B: PPR protection enabled
000B: RFU
Write-only 001B: RZQ/1
010B: RZQ/2
PDDS 011B: RZQ/3
OP[5:3] 1,2,3
(Pull-Down Drive Strength) 100B: RZQ/4
101B: RZQ/5
110B: RZQ/6 (default)
111B: Reserved
DBI-RD 0B: Disabled (default)

OP[6] 2,3
(DBI-Read Enable) 1B: Enabled
DBI-WR 0B: Disabled (default)
OP[7] 2,3
(DBI-Write Enable) 1B: Enabled
NOTE 1 All values are "typical". The actual value after calibration will be within the specified tolerance for a given
voltage and temperature. Re-calibration may be required as voltage and temperature vary.
NOTE 4 PU-CAL setting is required as the same value for both Ch.A and Ch.B before issuing ZQ Cal start command.
NOTE 5 Refer to the supplier data sheet for vender specific function. 1.5*tCK apply > 1.6GHz clock.
NOTE 6 If MR3 OP[2] is set to 1b then PPR protection mode is enabled. The PPR Protection bit is a sticky bit and can
only be set to 0b by a power on reset.
NOTE 7 MR4 OP[4] controls entry to PPR Mode. If PPR protection is enabled then DRAM will not allow writing of 1 to
MR4 OP[4].
Page 22
TUF Thermal Offset PPRE SR Abort Refresh Rate
Register
Type
000B: SDRAM Low temperature

operating limit exceeded
001B: 4x refresh
010B: 2x refresh
011B: 1x refresh (default) 1,2,3,4
Refresh Rate Read OP[2:0]
100B: 0.5x refresh 7,8,9
101B: 0.25x refresh, no derating
110B: 0.25x refresh, with derating
111B: SDRAM High temperature
operating limit exceeded
0B: Disable (default)

SR Abort (Self Refresh Abort) Write OP[3] 9,11
1B: Enable
PPRE 0B: Exit PPR mode (default)

(Post-package repair Write OP[4] 5,9
1B: Enter PPR mode
entry/exit)
00B: No offset, 0~5°C gradient (default)

Thermal Offset 01B: 5°C offset, 5~10°C gradient
Write OP[6:5] 10
(Vender Specific Function) 10B: 10°C offset, 10~15°C gradient
11B: Reserved
0B: No change in OP[2:0] since last

TUF MR4 read (default)
Read OP[7] 6,7,8
(Temperature Update Flag) 1B: Change in OP[2:0] since last
MR4 read
NOTE 1 The refresh rate for each MR4-OP[2:0] setting applies to tREFI, tREFIpb, and tREFW. OP[2:0]=011B corresponds to a
device temperature of 85°C. Other values require either a longer (2x, 4x) refresh interval at lower temperatures, or a shorter (0.5x,
0.25x) refresh interval at higher temperatures. If OP[2]=1B, the device temperature is greater than 85 °C.
NOTE 2 At higher temperatures (>85 °C), AC timing derating may be required. If derating is required the LPDDR4-SDRAM will set
OP[2:0]=110B. See derating timing requirements in 10.3, Table 90.
NOTE 3 DRAM vendors may or may not report all of the possible settings over the operating temperature range of the device. Each
vendor guarantees that their device will work at any temperature within the range using the refresh interval requested by their device.
NOTE 4 The device may not operate properly when OP[2:0]=000B or 111B.
NOTE 5 Post-package repair can be entered or exited by writing to OP[4].
NOTE 6 When OP[7]=1, the refresh rate reported in OP[2:0] has changed since the last MR4 read. A mode register read from MR4
will reset OP[7] to '0'.
NOTE 7 OP[7]=0 at power-up. OP[2:0] bits are undefined at power-up.
NOTE 8 See the 4.30, Temperature Sensor for information on the recommended frequency of reading MR4.
NOTE 9 OP[6:3] bits that can be written in this register. All other bits will be ignored by the DRAM during a MRW to this register.
NOTE 10 Refer to the supplier data sheet for vender specific function.
NOTE 11 Self refresh abort feature is available for higher density devices starting with 12Gb device.
Page 23

LPDDR4 Manufacturer ID
Register
Type
LPDDR4 Manufacturer ID Read-only OP[7:0] See JEP166, LPDDR4 Manufacturer ID Codes

Revision ID-1
Register
Type
00000000B: A-version
LPDDR4 Revision ID-1 Read-only OP[7:0] 1
00000001B: B-version
NOTE 1 MR6 is vendor specific.

Revision ID-2
Register Operan
Function Data Notes
Type d
00000000B: A-version
LPDDR4 Revision ID-2 Read-only OP[7:0] 1
00000001B: B-version
NOTE 1 MR7 is vendor specific.

Page 24

IO Width Density Type
Register
Type
00B: S16 SDRAM (16n pre-fetch)
Type OP[1:0]
0000B: 4Gb per die (2Gb per channel)

Density Read-only OP[5:2]
00B: x16 (per channel)

IO Width OP[7:6]

Vendor Specific Test Register
NOTE 1 Only 00H should be written to this register.
3.4.11 MR10 Register Information (MA[7:0] = 0AH)

RFU ZQ-Reset
Register
Type
0B: Normal Operation (Default)
ZQ-Reset Write-only OP[0] 1,2
1B: ZQ Reset
NOTE 1 See Table 51, ZQCal Timing Parameters for calibration latency and timing.
NOTE 2 If the ZQ-pin is connected to VDDQ through RZQ, either the ZQ calibration function or default calibration (via
ZQ-Reset) is supported. If the ZQ-pin is connected to VSS, the device operates with default calibration, and ZQ
calibration commands are ignored. In both cases, the ZQ connection shall not change after power is applied to the
device.
Page 25
3.4.12 MR11 Register Information (MA[7:0] = 0BH)

RFU CA ODT RFU DQ ODT
Register Operan
Function Data Notes
Type d
000B: Disable (Default)
001B: RZQ/1
010B: RZQ/2
DQ ODT 011B: RZQ/3
(DQ Bus Receiver On-Die- OP[2:0] 1,2,3
100B: RZQ/4
Termination)
101B: RZQ/5
110B: RZQ/6
111B: RFU
Write-only
001B: RZQ/1
010B: RZQ/2
CA ODT 011B: RZQ/3
(CA Bus Receiver On-Die- OP[6:4] 1,2,3
100B: RZQ/4
Termination)
101B: RZQ/5
110B: RZQ/6
111B: RFU
NOTE 1 All values are "typical". The actual value after calibration will be within the specified tolerance for a given
voltage and temperature. Re-calibration may be required as voltage and temperature vary.
Page 26
3.4.13 MR12 Register Information (MA[7:0] = 0CH)

RFU VR-CA VREF(ca)
Register
Type
000000B:
VREF(ca) -- Thru -- 1,2,3,
OP[5:0]
(VREF(ca) Setting) Read/ 110010B: See Table 12 5,6
Write All Others: Reserved
VR-CA 0B: VREF(ca) Range[0] enabled 1,2,4,
OP[6]
(VREF(ca) Range) 1B: VREF(ca) Range[1] enabled (default) 5,6
NOTE 1 This register controls the VREF(ca) levels for Frequency-Set-Point[1:0]. Values from either VR(ca)[0] or
VR(ca)[1] may be selected by setting OP[6] appropriately.
NOTE 2 A read to this register places the contents of OP[7:0] on DQ[7:0]. Any RFU bits and unused DQ's shall be set
to '0'. See 4.16, MRR Operation.
NOTE 3 A write to OP[5:0] sets the internal VREF(ca) level for FSP[0] when MR13 OP[6]=0B, or sets FSP[1] when
MR13 OP[6]=1B. The time required for VREF(ca) to reach the set level depends on the step size from the current level
to the new level. See 4.19, VREF(ca) training for more information.
NOTE 4 A write to OP[6] switches the LPDDR4-SDRAM between two internal VREF(ca) ranges. The range (Range[0]
or Range[1]) must be selected when setting the VREF(ca) register. The value, once set, will be retained until overwrit-
ten, or until the next power-on or RESET event.
Page 27
3.4.13 MR12 Register Information (MA[7:0] = 0CH) (cont’d)
Table 12 — VREF Settings for Range[0] and Range[1]

Function Operand Range[0] Values (% of VDD2) Range[1] Values (% of VDD2) Notes
000000B: 10.0% 011010B: 20.4% 000000B: 22.0% 011010B: 32.4%
000001B: 10.4% 011011B: 20.8% 000001B: 22.4% 011011B: 32.8%
000010B: 10.8% 011100B: 21.2% 000010B: 22.8% 011100B: 33.2%
000011B: 11.2% 011101B: 21.6% 000011B: 23.2% 011101B: 33.6%
000100B: 11.6% 011110B: 22.0% 000100B: 23.6% 011110B: 34.0%
000101B: 12.0% 011111B: 22.4% 000101B: 24.0% 011111B: 34.4%
000110B: 12.4% 100000B: 22.8% 000110B: 24.4% 100000B: 34.8%
000111B: 12.8% 100001B: 23.2% 000111B: 24.8% 100001B: 35.2%
001000B: 13.2% 100010B: 23.6% 001000B: 25.2% 100010B: 35.6%
001001B: 13.6% 100011B: 24.0% 001001B: 25.6% 100011B: 36.0%
001010B: 14.0% 100100B: 24.4% 001010B: 26.0% 100100B: 36.4%
001011B: 14.4% 100101B: 24.8% 001011B: 26.4% 100101B: 36.8%
VREF
Settings 001100B: 14.8% 100110B: 25.2% 001100B: 26.8% 100110B: 37.2%
OP[5:0] 1,2,3
for 001101B: 15.2% 100111B: 25.6% 001101B: 27.2% (Default) 100111B: 37.6%
MR12 001110B: 15.6% 101000B: 26.0% 001110B: 27.6% 101000B: 38.0%
001111B: 16.0% 101001B: 26.4% 001111B: 28.0% 101001B: 38.4%
010000B: 16.4% 101010B: 26.8% 010000B: 28.4% 101010B: 38.8%
010001B: 16.8% 101011B: 27.2% 010001B: 28.8% 101011B: 39.2%
010010B: 17.2% 101100B: 27.6% 010010B: 29.2% 101100B: 39.6%
010011B: 17.6% 101101B: 28.0% 010011B: 29.6% 101101B: 40.0%
010100B: 18.0% 101110B: 28.4% 010100B: 30.0% 101110B: 40.4%
010101B: 18.4% 101111B: 28.8% 010101B: 30.4% 101111B: 40.8%
010110B: 18.8% 110000B: 29.2% 010110B: 30.8% 110000B: 41.2%
010111B: 19.2% 110001B: 29.6% 010111B: 31.2% 110001B: 41.6%
011000B: 19.6% 110010B: 30.0% 011000B: 31.6% 110010B: 42.0%
011001B: 20.0% All Others: Reserved 011001B: 32.0% All Others: Reserved
NOTE 1 These values may be used for MR12 OP[5:0] to set the VREF(ca) levels in the LPDDR4-SDRAM.
NOTE 2 The range may be selected in the MR12 register by setting OP[6] appropriately.
NOTE 3 The MR12 registers represents either FSP[0] or FSP[1]. Two frequency-set-points each for CA and DQ are
provided to allow for faster switching between terminated and un-terminated operation, or between different high-fre-
quency setting which may use different terminations values.
Page 28
3.4.14 MR13 Register Information (MA [7:0] = 0DH)

FSP-OP FSP-WR DMD RFU VRCG VRO RPT CBT
Register
Type
CBT 0B: Normal Operation (default)
OP[0] 1
(Command Bus Training) 1B: Command Bus Training Mode Enabled
RPT 0B : Disable (default)
(Read Preamble Training OP[1]
1B : Enable
Mode)
0B: Normal operation (default)
VRO
OP[2] 1B: Output the VREF(ca) and VREF(dq) values 2
(VREF Output)
on DQ bits
VRCG 0B: Normal Operation (default)
OP[3] 3
(VREF Current Generator) 1B: VREF Fast Response (high current) mode
Write-only
RRO 0B: Disable codes 001 and 010 in MR4 OP[2:0]
OP[4] 4, 5
Refresh rate option 1B: Enable all codes in MR4 OP[2:0]
DMD 0B: Data Mask Operation Enabled (default)
OP[5] 6
(Data Mask Disable) 1B: Data Mask Operation Disabled
FSP-WR 0B: Frequency-Set-Point[0] (default)
(Frequency Set Point OP[6] 7
1B: Frequency-Set-Point [1]
Write Enable)
FSP-OP 0B: Frequency-Set-Point[0] (default)
(Frequency Set Point OP[7] 8
1B: Frequency-Set-Point [1]
Operation Mode)
NOTE 1 A write to set OP[0]=1 causes the LPDDR4-SDRAM to enter the Command Bus Training mode. When
OP[0]=1 and CKE goes LOW, commands are ignored and the contents of CA[5:0] are mapped to the DQ bus. CKE
must be brought HIGH before doing a MRW to clear this bit (OP[0]=0) and return to normal operation. See 4.21,
Command Bus Training, for more information.
NOTE 2 When set, the LPDDR4-SDRAM will output the VREF(ca) and VREF(dq) voltages on DQ pins. Only the
“active” frequency-set-point, as defined by MR13 OP[7], will be output on the DQ pins. This function allows an external
test system to measure the internal VREF levels. The DQ pins used for VREF output are vendor specific.
NOTE 3 When OP[3]=1, the VREF circuit uses a high-current mode to improve VREF settling time.
NOTE 4 MR13 OP4 RRO bit is valid only when MR0 OP0 = 1. For LPDDR4 devices with MR0 OP0 = 0, MR4 OP[2:0]
bits are not dependent on MR13 OP4.
NOTE 5 When OP[4] = 0, only 001b and 010b in MR4 OP[2:0] are disabled. LPDDR4 devices must report 011b
instead of 001b or 010b in this case. Controller should follow the refresh mode reported by MR4 OP[2:0], regardless of
RRO setting. TCSR function does not depend on RRO setting.
NOTE 6 When enabled (OP[5]=0B) data masking is enabled for the device. When disabled (OP[5]=1B), masked write
command is illegal. See 4.10, LPDDR4 Data Mask (DM) and Data Bus Inversion (DBIdc) Function.
NOTE 7 FSP-WR determines which frequency-set-point registers are accessed with MRW commands for the
following functions: VREF(CA) Setting, VREF(CA) Range, VREF(DQ) Setting, VREF(DQ) Range, CA ODT Enable, CA
ODT value, DQ ODT Enable, DQ ODT value, DQ Calibration Point, WL, RL, nWR, Read and Write Preamble, Read
postamble, and DBI Enables.
NOTE 8 FSP-OP determines which frequency-set-point register values are currently used to specify device operation
for the following functions: VREF(CA) Setting, VREF(CA) Range, VREF(DQ) Setting, VREF(DQ) Range, CA ODT Enable,
CA ODT value, DQ ODT Enable, DQ ODT value, DQ Calibration Point, WL, RL, nWR, Read and Write Preamble,
Read postamble, and DBI Enables.
Page 29
3.4.15 MR14 Register Information (MA[7:0] = 0EH)

RFU VR(dq) VREF(dq)
Register
Type
000000B:
VREF(dq) -- Thru -- 1,2,3,
OP[5:0]
(VREF(dq) Setting) Read/ 110010B: See Table 13 5,6
Write All Others: Reserved
VR(dq) 0B: VREF(dq) Range[0] enabled 1,2,4,
OP[6]
(VREF(dq) Range) 1B: VREF(dq) Range[1] enabled (default) 5,6
NOTE 1 This register controls the VREF(dq) levels for Frequency-Set-Point[1:0]. Values from either VR(dq)[0] or
VR(dq)[1] may be selected by setting OP[6] appropriately.
NOTE 2 A read (MRR) to this register places the contents of OP[7:0] on DQ[7:0]. Any RFU bits and unused DQ's shall
be set to‘0’. See 4.16, MRR Operation.
NOTE 3 A write to OP[5:0] sets the internal VREF(dq) level for FSP[0] when MR13 OP[6]=0B, or sets FSP[1] when
MR13 OP[6]=1B. The time required for VREF(dq) to reach the set level depends on the step size from the current level
to the new level. See 4.20, VREF(dq) training for more information.
NOTE 4 A write to OP[6] switches the LPDDR4-SDRAM between two internal VREF(dq) ranges. The range (Range[0]
or Range[1]) must be selected when setting the VREF(dq) register. The value, once set, will be retained until
overwritten, or until the next power-on or RESET event.
Page 30
3.4.15 MR14 Register Information (MA[7:0] = 0EH)
Table 13 — VREF Settings for Range[0] and Range[1]

FunctionOperand Range[0] Values (% of VDDQ) Range[1] Values (% of VDDQ) Notes
000000B: 10.0% 011010B: 20.4% 000000B: 22.0% 011010B: 32.4%
000001B: 10.4% 011011B: 20.8% 000001B: 22.4% 011011B: 32.8%
000010B: 10.8% 011100B: 21.2% 000010B: 22.8% 011100B: 33.2%
000011B: 11.2% 011101B: 21.6% 000011B: 23.2% 011101B: 33.6%
000100B: 11.6% 011110B: 22.0% 000100B: 23.6% 011110B: 34.0%
000101B: 12.0% 011111B: 22.4% 000101B: 24.0% 011111B: 34.4%
000110B: 12.4% 100000B: 22.8% 000110B: 24.4% 100000B: 34.8%
000111B: 12.8% 100001B: 23.2% 000111B: 24.8% 100001B: 35.2%
001000B: 13.2% 100010B: 23.6% 001000B: 25.2% 100010B: 35.6%
001001B: 13.6% 100011B: 24.0% 001001B: 25.6% 100011B: 36.0%
001010B: 14.0% 100100B: 24.4% 001010B: 26.0% 100100B: 36.4%
001011B: 14.4% 100101B: 24.8% 001011B: 26.4% 100101B: 36.8%
VREF
Settings 001100B: 14.8% 100110B: 25.2% 001100B: 26.8% 100110B: 37.2%
OP[5:0] 1,2,3
for 001101B: 15.2% 100111B: 25.6% 001101B: 27.2% (Default) 100111B: 37.6%
MR14 001110B: 15.6% 101000B: 26.0% 001110B: 27.6% 101000B: 38.0%
001111B: 16.0% 101001B: 26.4% 001111B: 28.0% 101001B: 38.4%
010000B: 16.4% 101010B: 26.8% 010000B: 28.4% 101010B: 38.8%
010001B: 16.8% 101011B: 27.2% 010001B: 28.8% 101011B: 39.2%
010010B: 17.2% 101100B: 27.6% 010010B: 29.2% 101100B: 39.6%
010011B: 17.6% 101101B: 28.0% 010011B: 29.6% 101101B: 40.0%
010100B: 18.0% 101110B: 28.4% 010100B: 30.0% 101110B: 40.4%
010101B: 18.4% 101111B: 28.8% 010101B: 30.4% 101111B: 40.8%
010110B: 18.8% 110000B: 29.2% 010110B: 30.8% 110000B: 41.2%
010111B: 19.2% 110001B: 29.6% 010111B: 31.2% 110001B: 41.6%
011000B: 19.6% 110010B: 30.0% 011000B: 31.6% 110010B: 42.0%
011001B: 20.0% All Others: Reserved 011001B: 32.0% All Others: Reserved
NOTE 1 These values may be used for MR14 OP[5:0] to set the VREF(dq) levels in the LPDDR4-SDRAM.
NOTE 2 The range may be selected in the MR14 register by setting OP[6] appropriately.
NOTE 3 The MR14 registers represents either FSP[0] or FSP[1]. Two frequency-set-points each for CA and DQ are
provided to allow for faster switching between terminated and un-terminated operation, or between different
high-frequency setting which may use different terminations values.
Page 31
3.4.16 MR15 Register Information (MA[7:0] = 0FH)

Lower-Byte Invert Register for DQ Calibration
Register
Type
The following values may be written for any operand
OP[7:0], and will be applied to the corresponding
DQ locations DQ[7:0] within a byte lane:
Lower-Byte Invert
Write-only OP[7:0] 0B: Do not invert 1,2,3
for DQ Calibration
1B: Invert the DQ Calibration patterns in MR32
and MR40
Default value for OP[7:0]=55H
NOTE 1 This register will invert the DQ Calibration pattern found in MR32 and MR40 for any single DQ, or any com-
bination of DQ's. Example: If MR15 OP[7:0]=00010101B, then the DQ Calibration patterns transmitted on
DQ[7,6,5,3,1] will not be inverted, but the DQ Calibration patterns transmitted on DQ[4,2,0] will be inverted.
NOTE 2 DMI[0] is not inverted, and always transmits the “true” data contained in MR32/MR40.
NOTE 3 No Data Bus Inversion (DBI) function is enacted during DQ Read Calibration, even if DBI is enabled in MR3-
OP[6].
Table 14 — MR15 Invert Register Pin Mapping

PIN DQ0 DQ1 DQ2 DQ3 DMI0 DQ4 DQ5 DQ6 DQ7
MR15 OP0 OP1 OP2 OP3 NO-Invert OP4 OP5 OP6 OP7
Page 32

PASR Bank Mask
Register
Type
0B: Bank Refresh enabled (default) : Unmasked
Bank[7:0] Mask Write-only OP[7:0] 1
1B: Bank Refresh disabled : Masked
OP[n] Bank Mask 8-Bank SDRAM

0 xxxxxxx1 Bank 0
1 xxxxxx1x Bank 1
2 xxxxx1xx Bank 2
3 xxxx1xxx Bank 3
4 xxx1xxxx Bank 4
5 xx1xxxxx Bank 5
6 x1xxxxxx Bank 6
7 1xxxxxxx Bank 7
NOTE 1 When a mask bit is asserted (OP[n]=1), refresh to that bank is disabled.
NOTE 2 PASR bank-masking is on a per-channel basis. The two channels on the die may have different bank
masking.
Page 33

PASR Segment Mask
Register
Type
0B: Segment Refresh enabled (default)
PASR Segment Mask Write-only OP[7:0]
1B: Segment Refresh disabled
Segment 4Gb 6Gb 8Gb 12Gb 16Gb 24Gb 32Gb

Segment OP[n]
Mask R13:R11 R14:R12 R14:R12 R15:R13 R15:R13 TBD TBD
0 0 xxxxxxx1 000B
1 1 xxxxxx1x 001B
2 2 xxxxx1xx 010B
3 3 xxxx1xxx 011B
4 4 xxx1xxxx 100B
5 5 xx1xxxxx 101B
6 6 x1xxxxxx 110B Not 110B Not 110B Not 110B
7 7 1xxxxxxx 111B Allowed 111B Allowed 111B Allowed 111B
NOTE 1 This table indicates the range of row addresses in each masked segment. "X" is don't care for a particular
segment.
NOTE 2 PASR segment-masking is on a per-channel basis. The two channels on the die may have different segment
masking.
NOTE 3 For 6Gb, 12Gb, and 24Gb densities, OP[7:6] must always be LOW (=00B).
Page 34

DQS Oscillator Count - LSB
Register
Type
DQS Oscillator
(WR Training DQS Read-only OP[7:0] 0 - 255 LSB DRAM DQS Oscillator Count 1,2,3
Oscillator)
NOTE 1 MR18 reports the LSB bits of the DRAM DQS Oscillator count. The DRAM DQS Oscillator count value is
used to train DQS to the DQ data valid window. The value reported by the DRAM in this mode register can be used by
the memory controller to periodically adjust the phase of DQS relative to DQ.
NOTE 2 Both MR18 and MR19 must be read (MRR) and combined to get the value of the DQS Oscillator count.
NOTE 3 A new MPC [Start DQS Oscillator] should be issued to reset the contents of MR18/MR19.
Page 35

DQS Oscillator Count - MSB
Register
Type
DQS Oscillator
(WR Training DQS Read-only OP[7:0] 0-255 MSB DRAM DQS Oscillator Count 1,2,3
Oscillator)
NOTE 1 MR19 reports the MSB bits of the DRAM DQS Oscillator count. The DRAM DQS Oscillator count value is
used to train DQS to the DQ data valid window. The value reported by the DRAM in this mode register can be used by
the memory controller to periodically adjust the phase of DQS relative to DQ.
NOTE 2 Both MR18 and MR19 must be read (MRR) and combined to get the value of the DQS Oscillator count.
NOTE 3 A new MPC [Start DQS Oscillator] should be issued to reset the contents of MR18/MR19.

Upper-Byte Invert Register for DQ Calibration
Register
Type
The following values may be written for any operand

OP[7:0], and will be applied to the corresponding
DQ locations DQ[15:8] within a byte lane:
Upper-Byte Invert
Write-only OP[7:0] 0B: Do not invert 1,2
for DQ Calibration
1B: Invert the DQ Calibration patterns in MR32
and MR40
Default value for OP[7:0] = 55H
NOTE 1 This register will invert the DQ Calibration pattern found in MR32 and MR40 for any single DQ, or any
combination of DQ's. Example: If MR20 OP[7:0]=00010101B, then the DQ Calibration patterns transmitted on
DQ[15,14,13,11,9] will not be inverted, but the DQ Calibration patterns transmitted on DQ[12,10,8] will be inverted.
NOTE 2 DMI[1] is not inverted, and always transmits the "true" data contained in MR32/MR40.
NOTE 3 No Data Bus Inversion (DBI) function is enacted during DQ Read Calibration, even if DBI is enabled in
MR3-OP[6].
Table 15 - MR20 Invert Register Pin Mapping

PIN DQ8 DQ9 DQ10 DQ11 DMI1 DQ12 DQ13 DQ14 DQ15
MR20 OP0 OP1 OP2 OP3 NO-Invert OP4 OP5 OP6 OP7
Page 36

RFU ODTD-CA ODTE-CS ODTE-CK CODT
Register
Type

001B: RZQ/1
010B: RZQ/2
SoC ODT 011B: RZQ/3
(Controller ODT Value for OP[2:0] 1,2,3
100B: RZQ/4
VOH calibration)
101B: RZQ/5
110B: RZQ/6
111B: RFU
ODTE-CK Write-only
(CK ODT enabled for 0B: ODT-CK Over-ride Disabled (Default) 2,3,4,
OP[3]
nonterminating 1B: ODT-CK Over-ride Enabled 6,8
rank)
ODTE-CS
0B: ODT-CS Over-ride Disabled (Default) 2,3,5,
(CS ODT enable for non OP[4]
1B: ODT-CS Over-ride Enabled 6,8
terminating rank)
ODTD-CA 0B: ODT-CA Obeys ODT_CA bond pad (default) 2,3,6,
(CA ODT termination OP[5]
1B: ODT-CA Disabled 7,8
disable)
NOTE 1 All values are “typical".

NOTE 4 When OP[3]=1, then the CK signals will be terminated to the value set by MR11-OP[6:4] regardless of the
state of the ODT_CA bond pad. This overrides the ODT_CA bond pad for configurations where CA is shared by two or
more DRAMs but CK is not, allowing CK to terminate on all DRAMs.
NOTE 5 When OP[4]=1, then the CS signal will be terminated to the value set by MR11-OP[6:4] regardless of the
state of the ODT_CA bond pad. This overrides the ODT_CA bond pad for configurations where CA is shared by two or
more DRAMs but CS is not, allowing CS to terminate on all DRAMs.
NOTE 6 For system configurations where the CK, CS, and CA signals are shared between packages, the package
design should provide for the ODT_CA ball to be bonded on the system board outside of the memory package. This
provides the necessary control of the ODT function for all die with shared Command Bus signals.
NOTE 7 When OP[5]=0, CA[5:0] will terminate when the ODT_CA bond pad is HIGH and MR11-OP[6:4] is VALID,
and disables termination when ODT_CA is LOW or MR11-OP[6:4] is disabled. When OP[5]=1, termination for CA[5:0]
is disabled, regardless of the state of the ODT_CA bond pad or MR11-OP[6:4].
NOTE 8 To ensure proper operation in a multi-rank configuration, when CA, CK or CS ODT is enabled via MR11
OP[6:4] and also via MR22 or CA-ODT pad setting, the rank providing ODT will continue to terminate the command
bus in all DRAM states including Active Self-refresh, Self-refresh Power-down, Active Power-down and Precharge
Power-down.
Page 37

DQS interval timer run time setting
Register
Type
00000000B: DQS interval timer stop via

MPC Command (Default)
00000001B: DQS timer stops automatically
at 16th clocks after timer start
at 32nd clocks after timer start
DQS interval timer run time Write-only OP[7:0] 1, 2
-------------- Thru --------------

at (63X16)th clocks after timer start
01XXXXXXB: DQS timer stops automatically
at 8192nd clocks after timer start
NOTE 1 MPC command with OP[6:0]=1001101B (Stop DQS Interval Oscillator) stops DQS interval timer in case of
MR23 OP[7:0] = 00000000B.
NOTE 2 MPC command with OP[6:0]=1001101B (Stop DQS Interval Oscillator) is illegal with non-zero values in
MR23 OP[7:0].
Page 38

TRR Mode TRR Mode BAn Unlimited MAC Value
Register
Type
000B: Unknown when bit OP3=0 (Note 1)

Unlimited when bit OP3=1 (Note 2)
001B: 700K
010B: 600K
MAC Value OP[2:0] 011B: 500K
Read-only 100B: 400K
101B: 300K
110B: 200K
111B: Reserved
0B: OP[2:0] define MAC value

Unlimited MAC OP[3]
1B: Unlimited MAC value (Note 2, Note 3)
000B: Bank 0
001B: Bank 1
010B: Bank 2
011B: Bank 3
TRR Mode BAn OP[6:4]
100B: Bank 4
Write-only 101B: Bank 5
110B: Bank 6
111B: Bank 7
0B: Disabled (default)

TRR Mode OP[7]
1B: Enabled
NOTE 1 Unknown means that the device is not tested for tMAC and pass/fail value in unknown.
NOTE 2 There is no restriction to number of activates.
NOTE 3 MR24 OP [2:0] is set to zero.
Page 39
Mode Register 25 contains one bit of readout per bank indicating that at least one resource is available for
Post Package Repair programming.
Bank7 Bank6 Bank5 Bank4 Bank3 Bank2 Bank1 Bank0
Register
Type
0B: PPR Resource is not available

PPR Resource Read-only OP[7:0]
1B: PPR Resource is available

DQ Calibration Pattern “A” (default = 5AH)
Register
Type
XB: An MPC command with OP[6:0]= 1000011B

causes the device to return the DQ Calibration
Pattern contained in this register and (followed by)
the contents of MR40. A default pattern “5AH”is
Return DQ Calibration
Write OP[7:0] loaded at power-up or RESET, or the pattern may
Pattern MR32 + MR40
be overwritten with a MRW to this register. The
contents of MR15 and MR20 will invert the data
pattern for a given DQ (See MR15 for more
information)
Page 40

DQ Calibration Pattern “B” (default = 3CH)
Register
Type
XB: A default pattern “3CH” is loaded at power-up or
Return DQ Calibration Write RESET, or the pattern may be overwritten with a
OP[7:0] 1,2,3
Pattern MR32 + MR40 only MRW to this register.
See MR32 for more information.
NOTE 1 The pattern contained in MR40 is concatenated to the end of MR32 and transmitted on DQ[15:0] and
DMI[1:0] when DQ Read Calibration is initiated via a MPC command. The pattern transmitted serially on each data
lane, organized “little endian” such that the low-order bit in a byte is transmitted first. If the data pattern in MR40 is 27H,
then the first bit transmitted with be a '1', followed by '1', '1', '0', '0', '1', '0', and '0'. The bit stream will be 00100111B.
NOTE 2 MR15 and MR20 may be used to invert the MR32/MR40 data patterns on the DQ pins. See MR15 and MR20
for more information. Data is never inverted on the DMI[1:0] pins.
NOTE 3 The data pattern is not transmitted on the DMI[1:0] pins if DBI-RD is disabled via MR3-OP[6].
NOTE 4 No Data Bus Inversion (DBI) function is enacted during DQ Read Calibration, even if DBI is enabled in
MR3-OP[6].
Page 41
4 Command Definitions and Timing Diagrams
4.1 Activate Command
The ACTIVATE command is composed of two consecutive commands, Activate-1 command and
Activate-2. Activate-1 command is issued by holding CS HIGH, CA0 HIGH and CA1 LOW at the first rising
edge of the clock and Activate-2 command issued by holding CS HIGH, CA0 HIGH and CA1 HIGH at the
first rising edge of the clock. The bank addresses BA0, BA1 and BA2 are used to select desired bank. Row
addresses are used to determine which row to activate in the selected bank. The ACTIVATE command
must be applied before any READ or WRITE operation can be executed. The device can accept a READ
or WRITE command at tRCD after the ACTIVATE command is issed. After a bank has been activated it
must be precharged before another ACTIVATE commnand can be applied to the same bank. The bank
active and precharge times are defined as tRAS and tRP respectively. The minimum time interval between
ACTIVATE commands to the same bank is determined by the RAS cycle time of the device(tRC). The
minimum time interval between ACTIVATE commands to different banks is tRRD.
T0 T1 T2 T3 Ta0 Ta1 Ta2 Ta3 Tb0 Tb1 Tb2 Tb3 Tc0 Tc1 Td0 Td1 Td2 Td3 Td4 Td5
CK_c
CK_t
CKE
CS
CA RA RA RA
RA BA0 RA RA RA BA1 RA RA Valid BA0 CA CA Valid BA0 RA BA0 RA RA
tRRD tRP
tRCD
COMMAND Precharge
Activate -1 Activate -2 DES Activate -1 Activate -2 DES Read-1 CAS-2 DES Per Bank DES Activate -1 Activate -2 DES DES
tRAS
tRC
NOTES : 1. A PRECHARGE command uses tRPab timing for all-bank PRECHARGE and tRPpb timing for single-bank PRECHARGE.
In this figure, tRP is used to denote either all-bank PRECHARGE or a single-bank PRECHARGE. DON'T CARE TIME BREAK
Figure 4 — ACTIVATE Command

Page 42
4.2 8-Bank Device Operation
Certain restrictions on operation of the 8-bank LPDDR4 devices must be observed. There are two rules:
One rule restricts the number of sequential ACTIVATE commands that can be issued; the other provides
more time for RAS precharge for a PRECHARGE ALL command. The rules are as follows:
8 bank device Sequential Bank Activation Restriction:

No more than 4 banks may be activated (or refreshed, in the case of REFpb) in a rolling tFAW window.
The number of clocks in a tFAW period is dependent upon the clock frequency, which may vary. If the
clock frequency is not changed over this period, converting clocks is done by dividing tFAW[ns] by
tCK[ns], and rounding up to the next integer value. As an example of the rolling window, if
RU(tFAW/tCK) is 10 clocks, and an ACTIVATE command is issued in clock n, no more than three fur-
ther ACTIVATE commands can be issued at or between clock n + 1 and n + 9. REFpb also counts as
bank activation for purposes of tFAW. If the clock frequency is changed during the tFAW period, the
rolling tFAW window may be calculated in clock cycles by adding up the time spent in each clock
period. The tFAW requirement is met when the previous n clock cycles exceeds the tFAW time.
The 8-Bank Device Precharge-All Allowance:

tRP for a PRECHRGE ALL command must equal tRPab, which is greater than tRPpb.
T0 T1 T2 T3 Ta0 Ta1 Ta2 Ta3 Tb0 Tb1 Tb2 Tb3 Tc0 Tc1 Tc2 Tc3 Tc4 Td0 Td1 Td2 Td3 Td4
CK_c
CK_t
CKE
CS
CA RA RA RA RA RA
RA BA0 RA RA RA BA1 RA RA RA BA2 RA RA RA BA3 RA RA RA BA4 RA RA
COMMAND Activate -1 Activate -2 DES Activate -1 Activate -2 DES Activate -1 Activate -2 DES Activate -1 Activate -2 DES DES Activate -1 Activate -2
tRRD tRRD tRRD

tFAW
DON'T CARE TIME BREAK
Figure 5 — tFAW Timing
4.3 Read and Write Access Operations
After a bank has been activated, a read or write command can be executed. This is accomplished by
asserting CKE asynchronously, with CS and CA[5:0] set to the proper state (see Table 63 — Command
Truth Table) at a rising edge of CK.
The LPDDR4-SDRAM provides a fast column access operation. A single Read or Write command will
initiate a burst read or write operation, where data is transferred to/from the DRAM on successive clock
cycles. Burst interrupts are not allowed, but the optimal burst length may be set on the fly (see Table 63 —
Command Truth Table).
Page 43
4.4 Read Preamble and Postamble
The DQS strobe for the LPDDR4-SDRAM requires a pre-amble prior to the first latching edge (the rising
edge of DQS_t with DATA "valid"), and it requires a post-amble after the last latching edge. The pre-amble
and post-amble lengths are set via mode register writes (MRW).
For READ operations the pre-amble is 2*tCK, but the pre-amble is static (no-toggle) or toggling, selectable
via mode register.
LPDDR4 will have a DQS Read post-amble of 0.5*tCK (or extended to 1.5*tCK). Standard DQS post-
amble will be 0.5*tCK driven by the DRAM for Reads. A mode register setting instructs the DRAM to drive
an additional (extended) one cycle DQS Read post-amble. Figure 6 and Figure 7 show examples of DQS
Read post-amble for both standard (tRPST) and extended (tRPSTE ) post-amble operation.
T0 T1 T2 T3 T4 Ta0 Ta1 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6 Tc0 Tc1 Tc2 Tc3 Tc4
CK_c
CK_t
COMMAND RD-1 RD-1 CAS-2 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES
RL tDQSCK
tRPRE
DQS_c
DQS_t
tDQSQ tRPST
Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout
DQ n0 n1 n2 n3 n4 n5 n10 n11 n12 n13 n14 m15
Note
1. BL = 16, Preamble = Toggling, Postamble = 0.5nCK
2. DQS and DQ terminated VSSQ.
3. DQS_t/DQS_c is "don't care" prior to the start of tRPRE. TIME BRAKE
No transition of DQS is implied, as DQS_t/DQS_c can be HIGH, LOW, or HI-Z prior to tRPRE.
Figure 6 — DQS Read Preamble and Postamble: Toggling Preamble and 0.5nCK Postamble
T0 T1 T2 T3 T4 Ta0 Ta1 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6 Tc0 Tc1 Tc2 Tc3 Tc4
CK_c
CK_t
COMMAND RD-1 RD-1 CAS-2 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES
RL tDQSCK
Extended tCK
tRPRE Postamble
DQS_c
DQS_t
tDQSQ tRPSTE
Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout
DQ n0 n1 n2 n3 n4 n5 n10 n11 n12 n13 n14 m15
Note
1. BL = 16, Preamble = Static, Postamble = 1.5nCK (Extended)
2. DQS and DQ terminated VSSQ. TIME BRAKE
3. DQS_t/DQS_c is "don't care" prior to the start of tRPRE.
No transition of DQS is implied, as DQS_t/DQS_c can be HIGH, LOW, or HI-Z prior to tRPRE.
Figure 7 — DQS Read Preamble and Postamble: Static Preamble and 1.5nCK Postamble
Page 44
4.5 Burst Read Operation
A burst Read command is initiated with CS, and CA[5:0] asserted to the proper state at the rising edge of
CK, as defined by Table 63 — Command Truth Table.
The command address bus inputs determine the starting column address for the burst. The two low-order
address bits are not transmitted on the CA bus and are implied to be “0”, so that the starting burst address
is always a multiple of four (ex. 0x0, 0x4, 0x8, 0xC).
The read latency (RL) is defined from the last rising edge of the clock that completes a read command (Ex:
the second rising edge of the CAS-2 command) to the rising edge of the clock from which the tDQSCK
delay is measured. The first valid data is available RL * tCK + tDQSCK + tDQSQ after the rising edge of
Clock that completes a read command.
The data strobe output is driven tRPRE before the first valid rising strobe edge. The first data-bit of the
burst is synchronized with the first valid (i.e., post-preamble) rising edge of the data strobe. Each
subsequent dataout appears on each DQ pin, edge-aligned with the data strobe. At the end of a burst the
DQS signals are driven for another half cycle post-amble, or for a 1.5-cycle postamble if the programmable
post-amble bit is set in the mode register.
The RL is programmed in the mode registers.
Pin timings for the data strobe are measured relative to the cross-point of DQS_t and DQS_c.
T0 T1 T2 T3 T4 T5 T6 T7 T15 T16 T17 T18 T19 T20 T21 T22 T23 T33 T34 T35 T36 T41 T42 T43 T44
CK_c
CK_t
CS
CA BA0, BA0,
BL CA, AP CAn CAn BL CA, AP CAm CAm
COMMAND Read-1 CAS-2 DES DES DES DES DES Read-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES
tCCD = 16 RL = 14 tDQSCK BL/2 = 8

RL = 14 tDQSCK BL/2 = 16
tRPRE
DQS_c
DQS_t
tDQSQ tDQSQ tRPST
Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout
DQ n0 n1 n2 n3 n4 n5 n6 n7 n26 n27 n28 n29 n30 n31 m0 m1 m10 m11 m12 m13 m14 m15
Note
1. BL = 32 for column n, BL = 16 for column m, RL = 14, Preamble = Toggle, Postamble = 0.5nCK, DQ/DQS: VSSQ termination DON'T CARE TIME BRAKE
2. Dout n/m = data-out from column n and column m.
3. DES commands are shown for ease of illustration; other commands may be valid at these times.
Figure 8 — Burst Read Timing

Page 45
Page 46
T0 T1 T2 T3 T4 T8 Ta0 Ta1 Ta2 Ta3 Ta4 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6 Tb7 Tc0 Tc1 Tc2 Tc3 Tc4 Tc5 Tc6 Tc7
CK_c
CK_t
CS
CA BA0, BA0,
BL CA, AP CA CA BL CA, AP CA CA
COMMAND Read-1 CAS-2 DES DES DES WR-1/MWR-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES
RL+RU(tDQSCK(max)/tCK)+BL/2+ RD(tRPST)-WL+tWPRE WL tDQSS
RL tDQSCK BL/2 = 8
tRPRE tWPRE
DQS_c
DQS_t
tDQSQ tRPST
tDQS2DQ
Dout Dout Dout Dout Dout Dout Dout Dout Din Din Din Din Din Din Din Din
DQ n0 n9 n10 n11 n12 n13 n14 n15 n0 n1 n2 n3 n4 n13 n14 n15
Note
1. BL=16, Read Preamble = Toggle, Read Postamble = 0.5nCK, Write Preamble = 2nCK, Write Postamble = 0.5nCK, DQ/DQS: VSSQ termination
2. Dout n = data-out from column n and Din n = data-in to columnm.n
Figure — 9 Burst Read followed by Burst Write or Burst Mask Write
The minimum time from a Burst Read command to a Write or MASK WRITE command is defined by the read latency (RL) and the burst length (BL).
Minimum READ-to-WRITE or MASK WRITE latency is RL+RU(tDQSCK(max)/tCK)+BL/2+ RD(tRPST)-WL+tWPRE.
T0 T1 T2 T3 Ta0 Ta1 Ta2 Ta3 Tb0 Tb1 Tb2 Tb3 Tb4 Tc0 Tc1 Tc2 Tc3 Td0 Td1 Td2 Td3 Te0 Te1 Te2 Te3
CK_c
CK_t
tCCD tCCD
CS
CA BA0, BA0, BA1,

BL CA, AP CAn CAn BL CA, AP CAm CAm BL CA, AP CAn CAn
COMMAND Read-1 CAS-2 DES Read-1 CAS-2 DES Read-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES
RL tDQSCK
RL tDQSCK
RL tDQSCK
tRPRE
DQS_c
DQS_t
tDQSQ tDQSQ tDQSQ tRPST
Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout Dout
DQ n0 n1 n10 n11 n12 n13 n14 n15 m0 m1 m10 m11 m12 m13 m14 m15 n0 n1 n10 n11 n12 n13 n14 n15
Bank 0 Bank 1
Note
1. BL = 16, tCCD = 8, Preamble = Toggle, Postamble = 0.5nCK, DQ/DQS: VSSQ termination DON'T CARE TIME BRAKE
2. Dout n/m = data-out from column n and column m.
Figure 10 — Seamless Burst Read
The seamless Burst READ operation is supported by placing a READ command at every tCCD(Min) interval for BL16 (or every 2 x tCCD(Min) for
BL32). The seamless Burst READ can access any open bank.

Page 47
Page 48
4.6 tDQSCK Timing Table
Table 16 — tDQSCK Timing Table

Parameter Symbol Min Max Unit Notes
DQS Output Access Time from CK_t/CK_c tDQSCK 1.5 3.5 ns 1
DQS Output Access Time from CK_t/CK_c -
tDQSCK_temp - 4 ps/°C 2
Temperature Variation
DQS Output Access Time from CK_t/CK_c -
tDQSCK_volt - 7 ps/mV 3
Voltage Variation
NOTE 1 Includes DRAM process, voltage and temperature variation. It includes the AC noise impact for frequencies
> 20 MHz and max voltage of 45 mV pk-pk from DC-20 MHz at a fixed temperature on the package. The volage supply
noise must comply to the component Min-Max DC Operating conditions.
NOTE 2 tDQSCK_temp max delay variation as a function of Temperature.
NOTE 3 tDQSCK_volt max delay variation as a function of DC voltage variation for VDDQ and VDD2. tDQSCK_volt
should be used to calculate timing variation due to VDDQ and VDD2 noise < 20 MHz. Host controller do not need to
account for any variation due to VDDQ and VDD2 noise > 20 MHz. The voltage supply noise must comply to the com-
ponent Min-Max DC Operating conditions. The voltage variation is defined as the Max[abs{tDQSCKmin@V1-tDQSCK-
max@V2}, abs{tDQSCKmax@V1-tDQSCKmin@V2}]/abs{V1-V2}. For tester measurement VDDQ = VDD2 is
assumed.
Page 49
4.7 Write Preamble and Postamble
The DQS strobe for the LPDDR4-SDRAM requires a pre-amble prior to the first latching edge (the rising
edge of DQS_t with DATA "valid"), and it requires a post-amble after the last latching edge. The pre-amble
and post-amble lengths are set via mode register writes (MRW).
For WRITE operations, a 2*tCK pre-amble is required at all operating frequencies.
LPDDR4 will have a DQS Write post-amble of 0.5*tCK or extended to 1.5*tCK. Standard DQS post-amble
will be 0.5*tCK driven by the memory controller for Writes. A mode register setting instructs the DRAM to
drive an additional (extended) one cycle DQS Write post-amble. Figure 11 and Figure 12 show examples
of DQS Write post-amble for both standard (tWPST) and extended (tWPSTE ) post-amble operation.
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6 Tb7
CK_c
CK_t
CKE
CS
CA Valid Valid Valid Valid
COMMAND Write-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES
WL tDQSS tWPST
tWPRE tDQS2DQ BL/2
Din Din Din Din Din Din Din Din Din Din Din Din
n0 n1 n2 n3 n8 n9 n10 n11 n12 n13 n14 n15

NOTES : 1. BL = 16, Postamble = 0.5nCK
2. DQS and DQ terminated VSSQ
3. DQS_t/DQS_c is “don’t care” prior to the start of tWPRE.
No transition of DQS is implied, as DQS_t/DQS_c can be HIGH, LOW, or HI-Z prior to tWPRE.
Figure 11 — DQS Write Preamble and Postamble: 0.5nCK Postamble

Page 50
4.7 Write Preamble and Postamble (cont’d)
CK_c
CK_t
CKE
CS
COMMAND Write-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES
WL tDQSS tWPST
tWPRE tDQS2DQ BL/2
Din Din Din Din Din Din Din Din Din Din Din Din
n0 n1 n2 n3 n8 n9 n10 n11 n12 n13 n14 n15

NOTES : 1. BL = 16, Postamble = 1.5nCK
2. DQS and DQ terminated VSSQ
3. DQS_t/DQS_c is “don’t care” prior to the start of tWPRE.
No transition of DQS is implied, as DQS_t/DQS_c can be HIGH, LOW, or HI-Z prior to tWPRE.
Figure 12 — DQS WriteRead Preamble and Postamble: 1.5nCK Postamble
4.8 Burst Write Operation
A burst WRITE command is initiated with CS, and CA[5:0] asserted to the proper state at the rising edge of
CK, as defined by Table 63 — Command Truth Table. Column addresses C[3:2] should be driven LOW for
Burst WRITE commands, and column addresses C[1:0] are not transmitted on the CA bus (and are
assumed to be zero), so that the starting column burst address is always aligned with a 32B boundary. The
write latency (WL) is defined from the last rising edge of the clock that completes a write command (Ex: the
second rising edge of the CAS-2 command) to the rising edge of the clock from which tDQSS is measured.
The first valid “latching” edge of DQS must be driven WL * tCK + tDQSS after the rising edge of Clock that
completes a write command.
The LPDDR4-SDRAM uses an un-matched DQS-DQ path for lower power, so the DQS-strobe must arrive
at the SDRAM ball prior to the DQ signal by the amount of tDQS2DQ. The DQS-strobe output is driven
tWPRE before the first valid rising strobe edge. The tWPRE pre-amble is required to be 1 x tCK at lower
speeds, and 2 x tCK at higher speeds (Frequency TBD1). The DQS-strobe must be trained to arrive at the
DQ pad center-aligned with the DQ-data. The DQ-data must be held for tDIVW (data input valid window)
and the DQS must be periodically trained to stay centered in the tDIVW window to compensate for timing
changes due to temperature and voltage variation. Burst data is captured by the SDRAM on successive
edges of DQS until the 16 or 32 bit data burst is complete. The DQS-strobe must remain active (toggling)
for tWPST (WRITE post-amble) after the completion of the burst WRITE. After a burst WRITE operation,
tWR must be satisfied before a PRECHARGE command to the same bank can be issued. Pin input timings
are measured relative to the crosspoint of DQS_t and DQS_c.

T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0 Tb1 Tb2 Tc0 Tc1 Tc2 Tc3 Tc4 Td0 Td1 Td2 Td3 Tb4 Td5
CK_c
CK_t
CS
CA BA0, BA0,
BL CA, AP CA CA Valid BA0 RA RA RA RA
COMMAND Write-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES Precharge DES DES DES DES ACT-1 ACT-2
WL BL/2 + 1 Clock tWR tRP
tDQSS (Min)
tDSS tDSH
tWPRE tDSS tDSH tWPST
DQS_c
DQS_t
tDQS2DQ
DQ Din Din Din Din Din Din Din Din Din Din
n0 n1 n2 n3 n4 n5 n12 n13 n14 n15
tDQSS (Max)
tWPRE
DQS_c
DQS_t
tDQS2DQ
n0 n1 n2 n3 n4 n11 n12 n13 n14 n15
Note
1. BL=16, Write Postamble = 0.5nCK, DQ/DQS: VSSQ termination DON'T CARE TIME BREAK
2. Din n = data-in to columnm.n
3. The minimum number of clock cycles from the burst write command to the burst read command for any bank is [WL + 1 + BL/2 + RU(tWR/tCK)].
4. tWR starts at the rising edge of CK after the last latching edge of DQS.

Figure 13 — Burst Write Operation
Page 51
Page 52
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0 Tb1 Tb2 Tc0 Tc1 Tc2 Tc3 Tc4 Tc5 Tc6 Tc7 Tc8 Tc9 Tc10
CK_c
CK_t
CS
CA BA0, BA0,
COMMAND Write-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES Read-1 CAS-2 DES DES DES DES
WL BL/2 + 1 Clock tWTR RL
tDQSS (Min)
tDSS tDSH
tWPRE tDSS tDSH tWPST
DQS_c
DQS_t
tDQS2DQ
n0 n1 n2 n3 n4 n5 n12 n13 n14 n15
Note
3. The minimum number of clock cycles from the burst write command to the burst read command for any bank is [WL + 1 + BL/2 + RU(tWTR/tCK)].
4. tWTR starts at the rising edge of CK after the last latching edge of DQS.
Figure 14 — Burst Write Followed by Burst Read

Page 53
4.9 Masked Write Operation
The LPDDR4-SDRAM requires that Write operations which include a byte mask anywhere in the burst
sequence must use the Masked Write command. This allows the DRAM to implement efficient data
protection schemes based on larger data blocks. The Masked Write-1 command is used to begin the
operation, followed by a CAS-2 command. A Masked Write command to the same bank cannot be issued
until tCCDMW later, to allow the LPDDR4-SDRAM to finish the internal Read-Modify-Write. One Data
Mask-Invert (DMI) pin is provided per byte lane, and the Data Mask-Invert timings match data bit (DQ)
timing. See 4.10, Data Mask Inversion for more information on the use of the DMI signal.
Page 54
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0 Tb1 Tb2 Tc0 Tc1 Tc2 Tc3 Tc4 Tc5 Tc6 Tc7 Tc8 Tc9 Tc10
CK_c
CK_t
CS
CA BA0, BA0,
COMMAND Mask Write-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES Mask Write-1 CAS-2 DES DES DES DES
WL tCCDMW WL
tDQSS (Min)
tWPRE tWPST
DQS_c
DQS_t
tDQS2DQ
DMI n0 n1 n2 n3 n4 n5 n12 n13 n14 n15
Note
3. Mask-Write supports only BL16 operations. For BL32 configuration, the system needs to insert only 16 bit wide data for masked write operation.
4. DES commands are shown for ease of illustration; other commands may be valid at these time.
Figure 15 — Masked Write Command - Same Bank

T0 T1 T2 T3 T8 T9 T10 T11 T16 T17 T18 T19 T24 T25 T26 T27 T32 T33 T34 T35 T36 T37 T38
CK_c
CK_t
CS
CA BA0, BA1 BA2 BA3 BA0,

BL CA, AP CA CA BL CA, AP CA CA BL CA, AP CA CA BL CA, AP CA CA BL CA, AP CA CA
COMMAND Mask Write-1 CAS-2 DES Mask Write-1 CAS-2 DES Mask Write-1 CAS-2 DES Mask Write-1 CAS-2 DES Mask Write-1 CAS-2 DES DES DES
tCCD tCCD tCCD tCCD
tCCDMW
WL tDQSS
tWPRE
DQS_c
DQS_t
tDQS2DQ
DQ Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din
DMI n0 n1 n2 n3 n10 n11 n12 n13 n14 n15 n0 n1 n2 n3 n10 n11 n12 n13 n14 n15 n0 n1 n2 n3 n10 n11 n12 n13 n14 n15 n0 n1 n2 n3 n4 n5 n6 n7 n8
Note
1. BL=16, DQ/DQS/DMI: VSSQ termination DON'T CARE TIME BREAK
3. Mask-Write supports only BL16 operations. For BL32 configuration, the system needs to insert only 16 bit wide data for masked write operation.
4. DES commands are shown for ease of illustration; other commands may be valid at these time.
Figure 16 — Masked Write Command - Different Bank

Page 55
4.9.1 Masked Write Timing constraints for BL16
Page 56
Table 17 — Timing constraints for Same bank
Next CMD Read Write
Active Masked Write Precharge
Current CMD (BL=16 or 32) (BL=16 or 32)
Active illegal RU(tRCD/tCK) RU(tRCD/tCK) RU(tRCD/tCK) RU(tRAS/tCK)
Read with RL+RU(tDQSCK(max)/tCK) RL+RU(tDQSCK(max)/tCK)
illegal 81) BL/2+max{(8,RU(tRTP/tCK)}-8
BL = 16 +BL/2-WL+tWPRE+tRPST +BL/2-WL+tWPRE+tRPST
illegal 162) BL/2+max{(8,RU(tRTP/tCK)}-8
Write with WL+1+BL/2
illegal 81) tCCDMW3) WL+ 1 + BL/2+RU(tWR/tCK)
BL = 16 +RU(tWTR/tCK)
illegal 162) tCCDMW +84) WL+ 1 + BL/2+RU(tWR/tCK)
WL+1+BL/2 WL+ 1 + BL/2
Masked Write illegal tCCD tCCDMW3)
+RU(tWTR/tCK) +RU(tWR/tCK)
RU(tRP/tCK),
Precharge illegal illegal illegal 4
RU(tRPab/tCK)
NOTE 1 In the case of BL = 16, tCCD is 8*tCK.
NOTE 3 tCCDMW = 32*tCK (4*tCCD at BL=16)
NOTE 4 Write with BL=32 operation has 8*tCK longer than BL =16.
4.9.1 Masked Write Timing constraints for BL16 (cont’d)
Table 18 — Timing constraints for Different bank

Next CMD Read (BL=16 or
Active Write (BL=16 or 32) Masked Write Precharge
Current CMD 32)
Active RU(tRRD/tCK) 4 4 4 4
4 81) 4
4 162) 4
4 81) 81) 4
4 162) 162) 4
WL+1+BL/2
Masked Write 4 81) 81) 4
+RU(tWTR/tCK)
Precharge 4 4 4 4 4

Page 57
Page 58
4.10 LPDDR4 Data Mask (DM) and Data Bus Inversion (DBIdc) Function
LPDDR4 SDRAM supports the function of Data Mask and Data Bus inversion. Its details are shown below.
• LPDDR4 device supports Data Mask (DM) function for Write operation.
• LPDDR4 device supports Data Bus Inversion (DBIdc) function for Write and Read operation.
• LPDDR4 supports DM and DBIdc function with a byte granularity.
• DBIdc function during Write or Masked Write can be enabled or disabled through MR3 OP[7].
• DBIdc function during Read can be enabled or disabled through MR3 OP[6].
• DM function during Masked Write can be enabled or disabled through MR13 OP[5].
• LPDDR4 device has one Data Mask Inversion (DMI) signal pin per byte; total of 2 DMI signals per
channel.
• DMI signal is a bi-directional DDR signal and is sampled along with the DQ signals for Read and Write or
Masked Write operation.
There are eight possible combinations for LPDDR4 device with DM and DBIdc function. Table 19 describes
the functional behavior for all combinations.
Page 59
4.10 LPDDR4 Data Mask (DM) and Data Bus Inversion (DBIdc) Function (cont’d)
Table 19 — Function Behavior of DMI Signal During Write, Masked Write and Read Operation
DMI DMI
Signal DMI DMI
Signal DMI Signal DMI
Write Read during Signal Signal
DM during Signal during Signal
DBIdc DBIdc Masked during during
Function Write During MPC DQ During
Function Function Write MPC WR MPC RD
Comman Read Read MRR
Command FIFO FIFO
d Training
Disable Disable Disable Notes: 1 Notes: 1, 3 Notes: 2 Note: 1 Note: 2 Note: 2 Notes: 2
Disable Enable Disable Notes: 4 Notes: 3 Notes: 2 Note: 9 Note: 10 Note: 11 Notes: 2
Disable Disable Enable Notes: 1 Notes: 3 Notes: 5 Note: 9 Note: 10 Note: 11 Notes: 12
Disable Enable Enable Notes: 4 Notes: 3 Notes: 5 Note: 9 Note: 10 Note: 11 Notes: 12
Enable Disable Disable Notes: 6 Notes: 7 Notes: 2 Note: 9 Note: 10 Note: 11 Notes: 2
Enable Enable Disable Notes: 4 Notes: 8 Notes: 2 Note: 9 Note: 10 Note: 11 Notes: 2
Enable Disable Enable Notes: 6 Notes: 7 Notes: 5 Note: 9 Note: 10 Note: 11 Notes: 12
Enable Enable Enable Notes: 4 Notes: 8 Notes: 5 Note: 9 Note: 10 Note: 11 Notes: 12
NOTE 1 DMI input signal is a don't care. DMI input receivers are turned OFF.
NOTE 2 DMI output drivers are turned OFF.
NOTE 3 Masked Write Command is not allowed and is considered an illegal command as DM function is disabled.
NOTE 4 DMI signal is treated as DBI signal and it indicates whether DRAM needs to invert the Write data received on
DQs within a byte. The LPDDR4 device inverts Write data received on the DQ inputs in case DMI was sampled HIGH,
or leaves the Write data non-inverted in case DMI was sampled LOW.
NOTE 5 The LPDDR4 DRAM inverts Read data on its DQ outputs associated within a byte and drives DMI signal
HIGH when the number of ‘1’ data bits within a given byte lane is greater than four; otherwise the DRAM does not
invert the read data and drives DMI signal LOW.
NOTE 6 The LPDDR4 DRAM does not perform any mask operation when it receives Write command. During the
Write burst associated with Write command, DMI signal must be driven LOW.
NOTE 7 The LPDDR4 DRAM requires an explicit Masked Write command for all masked write operations. DMI signal
is treated as DM signal and it indicates which bit time within the burst is to be masked. When DMI signal is HIGH,
DRAM masks that bit time across all DQs associated within a byte. All DQ input signals within a byte are don't care
(either HIGH or LOW) when DMI signal is HIGH. When DMI signal is LOW, the LPDDR4 DRAM does not perform
mask operation and data received on DQ input is written to the array.
NOTE 8 The LPDDR4 DRAM requires an explicit Masked Write command for all masked write operations. The
LPDDR4 device masks the Write data received on the DQ inputs if the total count of '1' data bits on DQ[2:7] or
DQ[10:15] (for Lower Byte or Upper Byte respectively) is equal to or greater than five and DMI signal is LOW. Other-
wise the LPDDR4 DRAM does not perform mask operation and treats it as a legal DBI pattern; DMI signal is treated as
DBI signal and data received on DQ input is written to the array.
NOTE 9 DMI signal is treated as a training pattern. The LPDDR4 DRAM does not perform any mask operation and
does not invert Write data received on the DQ inputs.
NOTE 10 DMI signal is treated as a training pattern. The LPDDR4 DRAM returns DMI pattern written in WR FIFO.
NOTE 11 DMI signal is treated as a training pattern. For more details, see 4.24, RD DQ Calibration.
NOTE 12 DBI may apply or may not apply during normal MRR. It's vendor specific.
If read DBI is enable with MRS and vendor cannot support the DBI during MRR, DBI pin status should be
low.
If read DBI is enable with MRS and vendor can support the DBI during MRR, the LPDDR4 DRAM inverts
Mode Register Read data on its DQ outputs associated within a byte and drives DMI signal HIGH when the
number of ‘1’ data bits within a given byte lane is greater than four; otherwise the DRAM does not invert
the read data and drives DMI signal LOW.
Page 60
T0 T1 T2 T3 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0

CK_c
CK_t
CKE
CS
COMMAND Mask Write-1 CAS-2 DES DES DES DES DES DES DES DES DES
WL tDQSS
DQS_c
DQS_t
tWPRE tDQS2DQ
DQ[7:0] Valid1 Valid1 Valid2 Valid2 Valid3 Valid1 Valid2 Valid1 Valid3 Valid1 Valid1 Valid1
N N I I M N I N M N N N
DMI[0]

Valid1 Input data is written to DRAM Cell.
N
Valid2 Input data is inverted, then written to DRAM Cell.

I
Valid3 Input data is masked. The total count of ‘1’ data bits on DQ[2:7] is equal to or greater than five.
M
NOTES : 1. Data Mask (DM) is Enable: MR13 OP [5] = 1, Data BUS Inversion (DBI) Write is Enable: MR3 OP[7] = 1
Figure 17 — Masked Write Command w/ Write DBI Enabled; DM Enabled

Page 61
T0 T1 T2 T3 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0

CK_c
CK_t
CKE
CS
COMMAND Write-1 CAS-2 DES DES DES DES DES DES DES DES DES
WL tDQSS
DQS_c
DQS_t
tWPRE tDQS2DQ
DQ[7:0] Valid1 Valid1 Valid2 Valid2 Valid1 Valid1 Valid2 Valid1 Valid1 Valid1 Valid1 Valid1
N N I I N N I N N N N N
DMI[0]

Valid1 Input data is written to DRAM Cell.
N
Valid2 Input data is inverted, then written to DRAM Cell.

I
NOTES : 1. Data Mask (DM) is Disable: MR13 OP [5] = 0, Data BUS Inversion (DBI) Write is Enable: MR3 OP[7] = 1
Figure 18 — Write Command w/ Write DBI Enabled; DM Enabled

Page 62
4.11 Pre-Charge Operation
The PRECHARGE command is used to precharge or close a bank that has been activated. The
PRECHARGE command is initiated with CS, and CA[5:0] in the proper state as defined by Table 63 —
Command Truth Table. The PRECHARGE command can be used to precharge each bank independently
or all banks simultaneously. The AB flag and the bank address bit are used to determine which bank(s) to
precharge. The precharged bank(s) will be available for subsequent row access tRPab after an all-bank
PRECHARGE command is issued, or tRPpb after a single-bank PRECHARGE command is issued.
To ensure that LPDDR4 devices can meet the instantaneous current demands, the row-precharge time for
an all-bank PRECHARGE (tRPab) is longer than the perbank precharge time (tRPpb).
Table 20 — Precharge Bank Selection

AB BA2 BA1 BA0 Precharged
(CA[5], R1) (CA[2], R2) (CA[1], R2) (CA[0], R2) Bank(s)
0 0 0 0 Bank 0 only
0 0 0 1 Bank 1 only
0 0 1 0 Bank 2 only
0 0 1 1 Bank 3 only
0 1 0 0 Bank 4 only
0 1 0 1 Bank 5 only
0 1 1 0 Bank 6 only
0 1 1 1 Bank 7 only
1 Valid Valid Valid All Banks
Page 63
4.11.1 Burst Read Operation Followed by a Precharge
The PRECHARGE command can be issued as early as BL/2 clock cycles after a READ command, but
PRECHARGE cannot be issued until after tRAS is satisfied. A new bank ACTIVATE command can be
issued to the same bank after the row PRECHARGE time (tRP) has elapsed. The minimum READ-to-
PRECHARGE time must also satisfy a minimum analog time from the 2nd rising clock edge of the CAS-2
command. tRTP begins BL/2 - 8 clock cycles after the READ command. For LPDDR4 READ-to-
PRECHARGE timings see Table 21.
t0 t1 t2 t3 t4 tx t x+1 t x+2 t x+3 t x+4 t x+5 t x+6 t x+7 t y t y+1 t y+2 t y+3 t y+4
CK_c
CK_t
CA[5:0] VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL
tRTP tRP
PRE-
CMD READ-1 CAS-2 VALID CHARGE VALID VALID VALID ACT-1 ACT-2
DQS_c
DQS_t
BL/2 = 8 Clocks
DQ[15:0] VAL VAL VAL VAL VAL VAL VAL VAL
DMI[1:0]
Figure 19 — Burst READ followed by PRECHARGE

(Shown with BL16, 2tCK pre-amble)
t0 t1 t2 t3 t4 t5 t 10 t 11 t 12 tx t x+1 t x+2 t x+3 t x+4 t x+5 t x+6 t x+7 t y t y+1 t y+2 t y+3 t y+4
CK_c
CK_t
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
8 Clocks tRTP tRP
CMD READ-1 CAS-2 VALID VALID PRE-

VALID CHARGE VALID VALID VALID ACT-1 ACT-2
DQS_c
DQS_t
BL/2 = 16 Clocks
DQ[15:0]
VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL
DMI[1:0]
Figure 20 — Burst READ followed by PRECHARGE

Page 64
4.11.2 Burst WRITE Followed by PRECHARGE
A Write Recovery time (tWR) must be provided before a PRECHARGE command may be issued. This
delay is referenced from the next rising edge of CK_t after the last latching DQS clock of the burst.
LPDDR4-SDRAM devices write data to the memory array in prefetch multiples (prefetch=16). An internal
WRITE operation can only begin after a prefetch group has been clocked, so tWR starts at the prefetch
boundaries. The minimum WRITE-to-PRECHARGE time for commands to the same bank is WL + BL/2 +
1 + RU(tWR/tCK) clock cycles.
t0 t1 t2 t3 t4 t x t x+1 t x+2 t x+3 t x+4 t x+5 t x+6 ta t a+1 t a+2 t n t n+1 t n+2 t n+3 t y t y+1 t y+2 t y+3 t y+4
CK_c
CK_t
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
WL BL/2 + 1 Clock
CMD WRITE-1 CAS-2 VALID VALID VALID VALID VALID PRE-

VALID CHARGE VALID ACT-1 ACT-2
tDQSS(max) tWR tRP
DQS_c
DQS_t
tDQS2DQ
DQ[15:0]
VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL
DMI[1:0]
Figure 21 — Burst WRITE Followed by PRECHARGE

4.11.3 Auto-PRECHARGE Operation
Before a new row can be opened in an active bank, the active bank must be precharged using either the
PRECHARGE command or the Auto-PRECHARGE function. When a READ, a WRITE or Masked Write
command is issued to the device, the AP bit (CA5) can be set to enable the active bank to automatically
begin precharge at the earliest possible moment during the burst READ, WRITE or Masked Write cycle.
If AP is LOW when the READ or WRITE command is issued, then the normal READ, WRITE or Masked
Write burst operation is executed and the bank remains active at the completion of the burst.
If AP is HIGH when the READ, WRITE or Masked Write command is issued, the Auto-PRECHARGE
function is engaged. This feature enables the PRECHARGE operation to be partially or completely hidden
during burst READ cycles (dependent upon READ or WRITE latency), thus improving system performance
for random data access.
Page 65
4.11.4 Burst READ with Auto-PRECHARGE
If AP is HIGH when a READ command is issued, the READ with Auto-PRECHARGE function is engaged.
An internal precharge procedure starts a following delay time after the READ command. And this delay
time depends on BL setting.
BL = 16: nRTP
BL = 32: 8nCK + nRTP
For LPDDR4 Auto-PRECHARGE calculations, see Table 21. Following an Auto-PRECHARGE operation,
an ACTIVATE command can be issued to the same bank if the following two conditions are both satisfied:
a. The RAS precharge time (tRP) has been satisfied from the clock at which the Auto-PRECHARGE
began, or
b. The RAS cycle time (tRC) from the previous bank activation has been satisfied.
t0 t1 t2 t3 t4 tx t x+1 t x+2 t x+3 t x+4 t x+5 t x+6 t x+7 t y t y+1 t y+2 t y+3 t y+4
CK_c
CK_t
CA[5:0] VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL
nRTP tRPpb
CMD READ-1 CAS-2 VALID VALID VALID VALID VALID ACT-1 ACT-2
DQS_c
DQS_t
BL/2 = 8 Clocks
DQ[15:0] VAL VAL VAL VAL VAL VAL VAL VAL
DMI[1:0]
Figure 22 — Burst READ with Auto-PRECHARGE

t0 t1 t2 t3 t4 t5 t 10 t 11 t 12 tx t x+1 t x+2 t x+3 t x+4 t x+5 t x+6 t x+7 t y t y+1 t y+2 t y+3 t y+4
CK_c
CK_t
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
8 Clocks nRTP tRPpb
CMD READ-1 CAS-2 VALID VALID VALID VALID VALID VALID VALID ACT-1 ACT-2
DQS_c
DQS_t
BL/2 = 16 Clocks
DQ[15:0]
VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL
DMI[1:0]
Figure 23 — Burst READ with Auto-PRECHARGE

Page 66
4.11.5 Burst WRITE with Auto-PRECHARGE
If AP is HIGH when a WRITE command is issued, the WRITE with Auto-PRECHARGE function is
engaged. The device starts an Auto-PRECHARGE on the rising edge tWR cycles after the completion of
the Burst WRITE.
Following a WRITE with Auto-PRECHARGE, an ACTIVATE command can be issued to the same bank if
the following conditions are met:
a. The RAS precharge time (tRP) has been satisfied from the clock at which the Auto-PRECHARGE
began, and
b. The RAS cycle time (tRC) from the previous bank activation has been satisfied.
t0 t1 t2 t3 t4 t x t x+1 t x+2 t x+3 t x+4 t x+5 t x+6 ta t a+1 t a+2 t n t n+1 t n+2 t n+3 t y t y+1 t y+2 t y+3 t y+4
CK_c
CK_t
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
WL BL/2 + 1 Clock
CMD WRITE-1 CAS-2 VALID VALID VALID VALID VALID VALID VALID VALID ACT-1 ACT-2
tDQSS(max) nWR tRPpb
DQS_c
DQS_t
tDQS2DQ
DQ[15:0]
VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL
DMI[1:0]
Figure 24 — Burst WRITE with Auto-PRECHARGE

Page 67
4.11.5 Burst WRITE with Auto-PRECHARGE (cont’d)
Table 21 — Timing Between Commands (PRECHARGE and Auto-PRECHARGE)(DQ ODT is Disable)

Minimum Delay between
From
To Command "From Command" and Unit Notes
Command
To Command"
PRECHARGE
tRTP tCK 1,6
READ (to same bank as Read)
BL=16
PRECHARGE All tRTP tCK 1,6
PRECHARGE
READ 8tCK + tRTP tCK 1,6
(to same bank as Read)
BL=32
PRECHARGE All 8tCK + tRTP tCK 1,6
PRECHARGE
nRTP tCK 1,10
(to same bank as READ w/AP)
PRECHARGE All nRTP tCK 1,10
Activate
nRTP + tRPpb tCK 1,8,10
WRITE or WRITE w/AP
Illegal -
(same bank)
READ w/AP MASK-WR or MASK-WR w/AP
Illegal -
BL=16 (same bank)
WRITE or WRITE w/AP RL+RU(tDQSCK(max)/tCK)
tCK 3,4,5
(different bank) +BL/2+RD(tRPST)-WL+tWPRE
MASK-WR or MASK-WR w/AP RL+RU(tDQSCK(max)/tCK)
tCK 3,4,5
READ or READ w/AP
Illegal -
(same bank)
READ or READ w/AP
BL/2 tCK 3
(different bank)
PRECHARGE
8tCK + nRTP tCK 1,10
PRECHARGE All 8tCK + nRTP tCK 1,10
Activate
8tCK + nRTP +tRPpb tCK 1,8,10
WRITE or WRITE w/AP
Illegal -
(same bank)
READ w/AP MASK-WR or MASK-WR w/AP
BL=32 Illegal -
(same bank)
WRITE or WRITE w/AP RL+RU(tDQSCK(max)/tCK)
tCK 3,4,5
MASK-WR or MASK-WR w/AP RL+RU(tDQSCK(max)/tCK)
tCK 3,4,5
READ or READ w/AP
Illegal -
(same bank)
READ or READ w/AP
BL/2 tCK 3
(different bank)
PRECHARGE
WL + BL/2 + tWR + 1 tCK 1,7
WRITE (to same bank as WRITE)
BL=16 & 32
PRECHARGE All WL + BL/2 + tWR + 1 tCK 1,7
Page 68
Table 21 — Timing Between Commands (PRECHARGE and Auto-PRECHARGE)(DQ ODT is Disable)

(cont’d)
From
Command
To Command"
PRECHARGE
WL + BL/2 + tWR + 1 tCK 1,7
MASK-WR (to same bank as MASK-WR)
BL=16
PRECHARGE All WL + BL/2 + tWR + 1 tCK 1,7
PRECHARGE
WL + BL/2 + nWR + 1 tCK 1,11
(to same bank as WRITE w/AP)
PRECHARGE All WL + BL/2 + nWR + 1 tCK 1,11
ACTIVATE
WL + BL/2 +nWR + 1 + tRPpb tCK 1,8,11
(to same bank as WRITE w/AP)
WRITE or WRITE w/AP
Illegal -
WRITE w/AP (same bank)
BL=16 & 32 READ or READ w/AP
Illegal -
(same bank)
WRITE or WRITE w/AP
BL/2 tCK 3
(different bank)
MASK-WR or MASK-WR w/AP
BL/2 tCK 3
(different bank)
READ or READ w/AP
WL + BL/2 + tWTR + 1 tCK 3,9
(different bank)
PRECHARGE
WL + BL/2 + nWR + 1 tCK 1,11
(to same bank as MASK-WR w/AP)
PRECHARGE All WL + BL/2 +nWR + 1 tCK 1,11
ACTIVATE
WL + BL/2 + nWR + 1 + tRPpb tCK 1,8,11
(to same bank as MASK-WR w/AP)
WRITE or WRITE w/AP
Illegal - 3
(same bank)
MASK-WR w/AP MASK-WR or MASK-WR w/AP
Illegal - 3
BL=16 (same bank)
WRITE or WRITE w/AP
BL/2 tCK 3
(different bank)
BL/2 tCK 3
(different bank)
READ or READ w/AP
Illegal - 3
(same bank)
READ or READ w/AP
WL + BL/2 + tWTR + 1 tCK 3,9
(different bank)
PRECHARGE
4 tCK 1
(to same bank as PRECHARGE)
PRECHARGE
PRECHARGE All 4 tCK 1
PRECHARGE 4 tCK 1
PRECHARGE
All
PRECHARGE All 4 tCK 1
Page 69
Table 21 Notes.
NOTE 1 For a given bank, the precharge period should be counted from the latest precharge command, whether per-
bank or all-bank, issued to that bank. The precharge period is satisfied tRP after that latest precharge command.
NOTE 2 Any command issued during the minimum delay time as specified in Table 21 is illegal.
NOTE 3 After READ w/AP, seamless read operations to different banks are supported. After WRITE w/AP or MASK-
WR w/AP, seamless write operations to different banks are supported. READ, WRITE, and MASK-WR operations may
not be truncated or interrupted.
NOTE 4 tRPST values depend on MR1-OP[7] respectively.
NOTE 5 tWPRE values depend on MR1-OP[2] respectively.
NOTE 6 Minimum Delay between "From Command" and "To Command" in clock cycle is calculated by dividing
tRTP(in ns) by tCK(in ns) and rounding up to the next integer: Minimum Delay[cycles] = Roundup(tRTP[ns] / tCK[ns])
tWR(in ns) by tCK(in ns) and rounding up to the next integer: Minimum Delay[cycles] = Roundup(tWR[ns] / tCK[ns])
tRPpb(in ns) by tCK(in ns) and rounding up to the next integer: Minimum Delay[cycles] = Roundup(tRPpb[ns] /
tCK[ns])
tWTR(in ns) by tCK(in ns) and rounding up to the next integer: Minimum Delay[cycles] = Roundup(tWTR[ns] / tCK[ns])
NOTE 10 For Read w/AP the value is nRTP which is defined in Mode Register 2.
NOTE 11 For Write w/AP the value is nWR which is defined in Mode Register 1.
Table 22 - Timing Between Commands (PRECHARGE and Auto-PRECHARGE): DQ ODT is Enable

From
Command
"To Command"
RL+RU(tDQSCK(max)/tCK)+BL/2+RD
WRITE or WRITE w/AP
(tRPST)-ODTLon- tCK 2, 3
(different bank)
READ w/AP RD(tODTon,min/tCK)
BL=16 RL+RU(tDQSCK(max)/tCK)+BL/2+RD
(different bank)
RD(tODTon,min/tCK)
RL+RU(tDQSCK(max)/tCK)+BL/2+RD
WRITE or WRITE w/AP
(different bank)
READ w/AP RD(tODTon,min/tCK)
BL=32 RL+RU(tDQSCK(max)/tCK)+BL/2+RD
(different bank)
RD(tODTon,min/tCK)
NOTE 1 The rest of the timing about prechage and Auto-Precharge is same as DQ ODT is Disable case.
NOTE 2 After READ w/AP, seamless read operations to different banks are supported. READ, WRITE, and MASK-
WR operations may not be truncated or interrupted.
NOTE 3 tRPST values depend on MR1-OP[7] respectively.
Page 70
4.12 Refresh command
The REFRESH command is initiated with CS HIGH, CA0 LOW, CA1 LOW, CA2 LOW, CA3 HIGH and CA4
LOW at the first rising edge of the clock. Per-bank REFRESH is initiated with CA5 LOW at the first rising
edge of the clock. All-bank REFRESH is initiated with CA5 HIGH at the first rising edge of the clock.
A per-bank REFRESH command (REFpb) is performed to the bank address as transferred on CA0, CA1
and CA2 at the second rising edge of the clock. Bank address BA0 is transferred on CA0, bank address
BA1 is transferred on CA1 and bank address BA2 is transferred on CA2. A per-bank REFRESH command
(REFpb) to the eight banks can be issued in any order. e.g. REFpb commands are issued in the following
order: 1-3-0-2-4-7-5-6. After the eight banks have been refreshed using the per-bank REFRESH command
the controller can send another set of per-bank REFRESH commands in the same order or a different
order. e.g., REFpb commands are issued in the following order that is different from the previous order: 7-
1-3-5-0-4-2-6. One of the possible order can also be a sequential round robin: 0-1-2-3-4-5-6-7. It is illegal
to send a per-bank REFRESH command to the same bank unless all eight banks have been refreshed
using the per-bank REFRESH command. The count of eight REFpb commands starts with the first REFpb
command after a synchronization event.
The bank count is synchronized between the controller and the SDRAM by resetting the bank count to
zero. Synchronization can occur upon issuing a RESET command or at every exit from self refresh. REFab
command also synchronizes the counter between the controller and SDRAM to zero. The SDRAM device
can be placed in self-refresh or a REFab command can be issued at any time without cycling through all
eight banks using per-bank REFRESH command. After the bank count is synchronized to zero the
controller can issue per-bank REFRESH commands in any order as described in the previous paragraph.
An All Bank Refresh command issued when the bank counter is not zero will not increment the SDRAM’s
refresh counter. The required number of REFRESH Commands in a tREFW window must be satisfied
excluding the above-mentioned All Bank Refresh Command.
A bank must be idle before it can be refreshed. The controller must track the bank being refreshed by the
per-bank REFRESH command.
The REFpb command must not be issued to the device until the following conditions are met:
- tRFCab has been satisfied after the prior REFab command.

- tRFCpb has been satisfied after the prior REFpb command.
- tRP has been satisfied after the prior PRECHARGE command to that bank.
- tRRD has been satisfied after the prior ACTIVATE command (if applicable, for example after activating a
row in a different bank than the one affected by the REFpb command).
The target bank is inaccessible during per-bank REFRESH cycle time (tRFCpb), however, other banks
within the device are accessible and can be addressed during the cycle. During the REFpb operation, any
of the banks other than the one being refreshed can be maintained in an active state or accessed by a
READ or a WRITE command. When the per-bank REFRESH cycle has completed, the affected bank will
be in the idle state.
After issuing REFpb, these conditions must be met:
- tRFCpb must be satisfied before issuing a REFab command.

- tRFCpb must be satisfied before issuing an ACTIVATE command to the same bank.
- tRRD must be satisfied before issuing an ACTIVATE command to a different bank.
- tRFCpb must be satisfied before issuing another REFpb command.
Page 71
4.12 Refresh command (cont’d)
An all-bank REFRESH command (REFab) issues a REFRESH command to all banks. All banks must be
idle when REFab is issued (for instance, by issuing a PRECHARGE-all command prior to issuing an all-
bank REFRESH command). REFab also synchronizes the bank count between the controller and the
SDRAM to zero. The REFab command must not be issued to the device until the following conditions have
been met:
- tRFCab has been satisfied following the prior REFab command.
- tRFCpb has been satisfied following the prior REFpb command.
- tRP has been satisfied following the prior PRECHARGE commands.
When an all-bank refresh cycle has completed, all banks will be idle. After issuing REFab:
- tRFCab latency must be satisfied before issuing an ACTIVATE command.
- tRFCab latency must be satisfied before issuing a REFab or REFpb command.
Table 23 — REFRESH Command Scheduling Separation requirements

Minimum Delay
Symbol To Notes
From
REFab
tRFCab REFab Activate command to any bank
REFpb
REFab
tRFCpb REFpb Activate command to same bank as REFpb
REFpb
REFpb Activate command to different bank than REFpb
tRRD REFpb 1
Activate Activate command to different bank than prior Activate
command
NOTE 1 A bank must be in the idle state before it is refreshed, so following an ACTIVATE command REFab is
prohibited; REFpb is supported only if it affects a bank that is in the idle state.
Page 72
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tc0 Tc1 Tc2 Tc3
CK_c
CK_t
CKE
CS
CA Valid Valid Valid Valid Valid Valid Valid Valid

tRPab tRFCab tRFCab
COMMAND Precharge All Bank All bank Any

All Bank DES DES DES DES Refresh DES DES DES DES Refresh DES DES DES DES Commnad DES

NOTES : 1. Only DES commands allowed after Refresh command registered untill tRFCab(min) expires.
Figure 25 — All-Bank Refresh Operation
T0 T1 T2 T3 Ta0 Ta1 Ta2 Ta3 Ta4 Tb0 Tb1 Tb2 Tb3 Tb4 Tc0 Tc1 Tc2 Tc3 Tc4 Tc5 Tc6
CK_c
CK_t
CKE
CS
CA Valid Valid Valid BA0 Valid BA1 Valid BA1 Valid Valid
tRPab tRFCpb tRFCpb
COMMAND Precharge Per Bank Per Bank

All Bank DES DES DES Refresh DES DES DES Refresh DES DES DES Activate -1 Activate -2 DES DES

NOTES : 1. Only DES commands allowed after Refresh command registered untill tRFCpb(min) expires.
2. In the beginning of this example, the REFpb bank is pointing to bank 0.
3. Operations to banks other than the bank being refreshed are supported during the tRFCpb period.
Figure 26 — Per-Bank Refresh Operation
In general, a Refresh command needs to be issued to the LPDDR4 SDRAM regularly every tREFI interval.
To allow for improved efficiency in scheduling and switching between tasks, some flexibility in the absolute
refresh interval is provided. A maximum of 8 Refresh commands can be postponed during operation of the
LPDDR4 SDRAM, meaning that at no point in time more than a total of 8 Refresh commands are allowed
to be postponed and maximum number of pulled-in or postponed REF command is dependent on refresh
rate. It is described in Table 24.
Page 73
In case that 8 Refresh commands are postponed in a row, the resulting maximum interval between the
surrounding Refresh commands is limited to 9 × tREFI. A maximum of 8 additional Refresh commands can
be issued in advance (“pulled in”), with each one reducing the number of regular Refresh commands
required later by one. Note that pulling in more than 8 Refresh commands in advance does not further
reduce the number of regular Refresh commands required later, so that the resulting maximum interval
between two surrounding Refresh commands is limited to 9 × tREFI. At any given time, a maximum of 16
REF commands can be issued within 2 x tREFI. Self-Refresh Mode may be entered with a maximum of
eight Refresh commands being postponed. After exiting Self-Refresh Mode with one or more Refresh
commands postponed, additional Refresh commands may be postponed to the extent that the total
number of postponed Refresh commands (before and after the Self-Refresh) will never exceed eight.
During Self-Refresh Mode, the number of postponed or pulled-in REFcommands does not change.
Table 24 — Legacy Refresh Command Timing Constraints

Max. No. of REFab
Max. No. of Max. Interval within max
MR4 Per-bank
Refresh rate pulled-in or between two (2xtREFI x refresh
OP[2:0] Refresh
postponed REFab REFab rate multiplier,
16xtRFC)
000B Low Temp. Limit N/A N/A N/A N/A
001B 4x tREFI 8 9 x 4 x tREFI 16 1/8 of REFab
011B 1x tREFI 8 9 x tREFI 16 1/8 of REFab
100B 0.5x tREFI 8 9 x 0.5 x tREFI 16 1/8 of REFab
111B High Temp. Limit N/A N/A N/A N/A
Table 25 — Modified REFRESH Command Timing Constraints

Max. No. of REFab
Max. No. of Max. Interval within max
MR4 Per-bank
Refresh rate pulled-in or between two (2xtREFI x refresh
OP[2:0] Refresh
postponed REFab REFab rate multiplier,
16xtRFC)
000B Low Temp. Limit N/A N/A N/A N/A
011B 1x tREFI 8 9 x tREFI 16 1/8 of REFab
111B High Temp. Limit N/A N/A N/A N/A
NOTE 1 For any thermal transition phase where Refresh mode is transitioned to either 2x tREFIor 4x tREFI, DRAM
will support the previous postponed refresh requirement provided the number of postponed refreshes is monotonically
reduced to meet the new requirement. However, the pulled-in refresh commands in previous thermal phase are not
applied in new thermal phase. Entering new thermal phase the controller must count the number of pulled-in refresh
commands as zero, regardless of remaining pulled-in refresh commands in previous thermal phase.
NOTE 2 LPDDR4 devices are refreshed properly if memory controller issues refresh commands with same or shorter
refresh period than reported by MR4 OP[2:0]. If shorter refresh period is applied, the corresponding requirements from
Table apply. For example, when MR4 OP[2:0]=001B, controller can be in any refresh rate from 4xtREFI to 0.25x tREFI.
When MR4 OP[2:0]=010B, the only prohibited refresh rate is 4x tREFI.
Page 74
tREFI
9 x tREFI
tRFC
8 REF-Commands postponed
Figure 27 — Postponing Refresh Commands (Example)
tREFI
9 x tREFI
tRFC
8 REF-Commands pulled-in
Figure 28 — Pulling-in Refresh Commands (Example)

Page 75
4.12.1 Refresh Requirement
Between SRX command and SRE command, at least one extra refresh command is required. After the
DRAM Self Refresh Exit command, in addition to the normal Refresh command at tREFI interval, the
LPDDR4 DRAM requires minimum of one extra Refresh command prior to Self Refresh Entry command.
Table 26 — Refresh Requirement Parameters per Die
Refresh Requirements Symbol 4Gb 6Gb 8Gb 12Gb 16Gb 24Gb 32Gb Units
Density per Channel 2Gb 3Gb 4Gb 6Gb 8Gb 12Gb 16Gb
Number of banks per channel 8 TBD TBD TBD TBD
Refresh Window (tREFW)
tREFW 32 TBD TBD TBD TBD ms
(TCASE ≤ 85°C)
(1/2 Rate Refresh)
(1/4 Rate Refresh)
Required Number of REFRESH
R 8192 TBD TBD TBD TBD -
Commands in a tREFW window
Average Refresh REFAB tREFI 3.904 TBD TBD TBD TBD us
Interval (TCASE ≤
REFPB tREFIpb 488 TBD TBD TBD TBD ns
85°C)
Interval
(1/2 Rate Refresh)
Interval (1/4 Rate
Refresh)
Refresh Cycle Time (All Banks) tRFCab 130 180 180 TBD TBD TBD ns
Refresh Cycle Time (Per Bank) tRFCpb 60 90 90 TBD TBD TBD ns
NOTE 1 Refresh is defined with a 32ms window, which refreshes 1/2 of the channel (or die). The entire channel (or
die) is refreshed every 64ms.
NOTE 2 Refresh for each channel is independent of the other channel on the die, or other channels in a package.
Power delivery in the user’s system should be verified to make sure the DC operating conditions are maintained when
multiple channels are refreshed simultaneously.
Page 76
4.13 Self Refresh Operation
4.13.1 Self Refresh Entry and Exit
The Self Refresh command can be used to retain data in the LPDDR4 SDRAM, the SDRAM retains data
without external Refresh command. The device has a built-in timer to accommodate Self Refresh
operation. The Self Refresh is entered by Self Refresh Entry Command defined by having CKE High, CS
High, CA0 Low, CA1 Low, CA2 Low; CA3 High; CA4 High, CA5 Valid (Valid that means it is Logic Level,
High or Low) for the first rising edge and CKE High, CS Low, CA0 Valid, CA1 Valid, CA2 Valid, CA3 Valid,
CA4 Valid, CA5 Valid at the second rising edge of the clock. Self Refresh command is only allowed when
SDRAM is idle state.
During Self Refresh mode, external clock input is needed and all input pin of SDRAM are activated.
SDRAM can accept the following commands, MRR-1, CAS-2, SRX, MPC, MRW-1, and MRW-2 except
PASR Bank/Segment setting.
LPDDR4 SDRAM can operate in Self Refresh in both the standard or elevated temperature ranges.
SDRAM will also manage Self Refresh power consumption when the operating temperature changes,
lower at low temperature and higher at high temperatures.
For proper Self Refresh operation, power supply pins (VDD1 and VDD2) must be at valid levels. VDDQ may
be turned off during Self-Refresh after tESCKE is satisfied (Refer to Figure 30 about tESCKE).
Prior to exiting Self-Refresh VDDQ must be within specified limits. The minimum time that the SDRAM must
remain in Self Refresh model is tSR,min.
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0 Tb1 Tb2
CK_c
CK_t
CKE
CS
COMMAND DES SRE1 SRE2 VALID1 VALID1 VALID1 VALID1 VALID1 VALID1 SRX1 SRX2 DES DES VALID VALID
tSR tXSV
ADDRESS VALID VALID VALID2 VALID2 VALID2 VALID2 VALID2 VALID2 VALID2 VALID2 VALID VALID
Enter Self Refresh Exit Self Refresh

TIME BREAK DON'T CARE
NOTES : 1. MRR-1, CAS-2, SRX, MPC, MRW-1 and MRW-2 except PASR Bank/Segment setting is allowed during Self Refresh.
2. Address input may be don’t care when input command is Deselect.
Figure 29 — Self Refresh Entry/Exit Timing

Page 77
4.13.2 Power Down Entry and Exit during Self Refresh

Entering/Exiting Power Down Mode is allowed during Self Refresh mode in SDRAM. The related timing
parameters between Self Refresh Entry/Exit and Power Down Entry/Exit are shown in Figure 30.
T0 T1 T2 Ta0 Tb0 Tc0 Td0 Te0 Tf0 Tg0 Tg1 Th0 Th1 Th2 Tk0
CK_c
CK_t
(6)
tISCKE
tCKCKEL tCKCKEH tISCKE
CKE
tESCKE tCKELPD
CS
(4)
tCMDCKE (4) tCKELCMD tCSCKEH(5) tCKEHCMD (5)
COMMAND SRE1 SRE2 VALID1 DES DES DES DES DES DES SRX1 SRX2 DES VALID
tSR tXSV
ADDRESS VALID VALID VALID 2 VALID VALID VALID
2. Address input may be don’t care when input command is Deselect. TIME BREAK DON'T CARE
3. CS input must be low when input command is Deselct.
4. Deselect is only allowed during tCMDCKE(min) and tCKELCMD(min)
5. Deselect is only allowed during tCSCKEH(min) and tCKEHCMD(min)
6. The input clock frequency can be changed after tCKCKEL(min) satisfied.
Figure 30 — Self Refresh Entry/Exit Timing with Power Down Entry/Exit

Page 78
4.13.3 Command input Timing after Power Down Exit
Command input timings after Power Down Exit during Self Refresh mode are shown in Figure 31.
T0 T1 T2 Ta0 Tb0 Tc0 Td0 Te0 Tf0 Tg0 Tg1 Th0 Th1 Th2 Tk0
CK_c
CK_t
(6)
tISCKE
tCKCKEL tCKCKEH tISCKE
CKE
tESCKE tCKELPD
CS
(4)
tCMDCKE (4) tCKELCMD tCSCKEH(5) tCKEHCMD (5)
COMMAND SRE1 SRE2 VALID1 DES DES DES DES DES DES VALID1 VALID1 DES VALID1
tSR
ADDRESS VALID VALID VALID2 VALID2 VALID2 VALID2
2. Address input may be don’t care when input command is Deselect. TIME BREAK DON'T CARE
3. CS input must be low when input command is Deselct.
4. Deselect is only allowed during tCMDCKE(min) and tCKELCMD(min)
5. Deselect is only allowed during tCSCKEH(min) and tCKEHCMD(min)
6. The input clock frequency can be changed after tCKCKEL(min) satisfied.
Figure 31 — Command input timings after Power Down Exit during Self Refresh
Page 79
4.13.4 Self Refresh AC Timing Table

Table 27 — Self Refresh AC Timing
Speed LPDDR4-3200 DDR4-4266
Units NOTE
Parameter Symbol MIN MAX MIN MAX
CKE input setup time tISCKE TBD - TBD - nCK
Valid Clock Requirement after max(7.5ns, max(7.5ns,
tCKCKEL - - nCK
CKE Input low 3tCK) 3tCK)
Valid Clock Requirement before
tCKCKEH 2tCK - 2tCK - nCK
CKE Input High
Delay from SRE command to
tESCKE 2 - 2 - nCK
CKE Input low
Delay from Valid command to
tCMDCKE 2 - 2 - nCK
CKE Input low
max(7.5ns, max(7.5ns,
CKE minimum pulse width tCKELPD - - nCK
3tCK) 3tCK)
Valid Command Requirement max(7.5ns, max(7.5ns,
tCKELCMD - - nCK
after CKE Input low 3tCK) 3tCK)
Valid Command Requirement max(7.5ns, max(7.5ns,
tCKEHCMD - - nCK
after CKE Input High 3tCK) 3tCK)
Valid CS low Requirement before
tCSCKEH TBD - TBD - nCK
CKE Input High
Minimum Self Refresh Time tSR TBD - TBD - nCK
max(tRFCab max(tRFCab
Exit Self Refresh to Valid
tXSV + 7.5ns, - + 7.5ns, - nCK 1
commands
2nCK) 2nCK)
NOTE 1 MRR-1, CAS-2, SRX, MPC, MRW-1 and MRW-2 except PASR Bank/Segment setting are only allowed
during this period.
4.14 Self Refresh Abort
If MR4 OP[3] is enabled then DRAM aborts any ongoing refresh during Self Refresh exit and does not
increment the internal refresh counter. Controller can issue a valid command after a delay of tXSR_abort
instead of tXSR.
The value of tXSR_abort(min) is defined as tRFCpb + TBD1ns.
Upon exit from Self Refresh mode, the LPDDR4 SDRAM requires a minimum of one extra refresh (8 per
bank or 1 all bank) before entry into a subsequent Self Refresh mode. This requirement remains the same
irrespective of the setting of the MR bit for self refresh abort.
Self refresh abort feature is available for higher density devices starting with12 Gb device.

Page 80
4.15 MRR, MRW, MPC Command during tXSR, tRFC
Mode Register Read (MRR), Mode Register Write (MRW) and Multi Purpose Command (MPC) can be
issued during tXSR period.
T0 T1 T2 T3 T4 T5 T6 T7 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5
CK_c
CK_t
CKE CS “H” for CAS-2
CS
CA Valid Valid Valid Valid Valid Valid Valid Valid Valid Valid
MPC
COMMAND DES SRX (2 clock command) DES DES DES DES MRW-1 MRW-2 DES DES DES DES Any Command*2
(Case-1)
tMRD
MPC
COMMAND DES SRX (4 clock command) CAS-2 DES DES MRW-1 MRW-2 DES DES DES DES Any Command*2
(Case-2)
tMRD
tXSR
NOTES : 1. MPC and MRW command are shown in figure at this time, Any combination of MRR, MRW and MPC is allowed during tXSR period.
2. Any command also includes MRR, MRW and all MPC command.
Figure 32 — MRR, MRW and MPC Commands Issuing Timing during tXSR
Mode Register Read (MRR), Mode Register Write (MRW) and Multi Purpose Command (MPC) can be
issued during tRFC period.
T0 T1 T2 T3 T4 T5 T6 T7 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5
CK_c
CK_t
CKE CS “H” for CAS-2
CS
CA Valid Valid Valid Valid Valid Valid Valid Valid Valid Valid
tMRD
COMMAND MPC
(Case-1) DES REF All Bank (2 clock command) DES DES DES DES MRW-1 MRW-2 DES DES DES DES Any Command*3
tMRD
COMMAND MPC
(Case-2) DES REF All Bank (4 clock command) CAS-2 DES DES MRW-1 MRW-2 DES DES DES DES Any Command*3
tRFCab*2
NOTES : 1. MPC and MRW command are shown in figure at this time, Any combination of MRR, MRW and MPC is allowed during tRFCab or tRFCpb period.
2. Refresh cycle time depends on Refresh command. In case of REF per Bank command issued, Refresh cycle time will be tRFCpb.
3. Any command also includes MRR, MRW and all MPC command.
Figure 33 — MRR, MRW and MPC Commands Issuing Timing during tRFC
Page 81
4.16 Mode Register Read (MRR)
The Mode Register Read (MRR) command is used to read configuration and status data from the
LPDDR4-SDRAM registers.The MRR command is initiated with CKE, CS and CA[5:0] in the proper state
as defined by Table 63 — Command Truth Table. The mode register address operands (MA[5:0]) allow the
user to select one of 64 registers. The mode register contents are available on the first 4UI's data bits of
DQ[7:0] after RL x tCK + tDQSCK + tDQSQ following the MRR command. Subsequent data bits contain
valid but undefined content. DQS is toggled for the duration of the Mode Register READ burst. The MRR
has a command burst length 16.
MRR operation must not be interrupted.
Table 28 — DQ output mapping

BL 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
DQ0 OP0 V
DQ1 OP1 V
DQ2 OP2 V
DQ3 OP3 V
DQ4 OP4 V
DQ5 OP5 V
DQ6 OP6 V
DQ7 OP7 V
DBI0 V
NOTE 1 MRR data are extended to first 4 UI’s for DRAM controller to sample data easily.
NOTE 2 DBI may apply or may not apply during normal MRR. It’s vendor specific. If read DBI is enable with MRS and
vendor cannot support the DBI during MRR, DBI pin status should be low.
NOTE 3 The read pre-amble and post-amble of MRR are same as normal read.
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tc0 Tc1
CK_c
CK_t
CS
CA Valid MA CAn CAn Valid Valid Valid Valid
COMMAND MRR-1 CAS-2 DES DES Any Command Any Command DES DES DES DES DES DES DES DES DES
tMRR
RL tDQSCK BL/2 = 8
tRPRE
DQS_c
DQS_t
tDQSQ tRPST
DQ7:0 OP Code out Valid Valid Valid Valid
DQ15:8 Valid Valid Valid Valid Valid Valid Valid Valid
Note
1. Only BL=16 is supported
2. Only DE-SELECT is allowed during tMRR period DON'T CARE TIME BRAKE
3. DQ/DQS: VSSQ termination
Figure 34 — Mode Register Read Operation

Page 82
4.17 Mode Register Write (MRW) Operation
The Mode Register Write (MRW) command is used to write configuration data to the mode registers. The
MRW command is initiated by setting CKE, CS, and CA[5:0] to valid levels at a rising edge of the clock
(see Table 63 — Command Truth Table). The mode register address and the data written to the mode
registers is contained in CA[5:0] according to Table 63 — Command Truth Table. The MRW command
period is defined by tMRW. Mode register Writes to read-only registers have no impact on the functionality
of the device.
CK_c
CK_t
CS
CA OPn MA OPn OPn OPn MA OPn OPn Valid Valid Valid Valid
COMMAND MRW-1 MRW2 DES DES MRW-1 MRW2 DES DES Any Command Any Command DES DES DES
tMRW tMRD
Note
1. Only Deselect command is allowed during tMRW and tMRD periods. DON'T CARE TIME BRAKE
Figure 35 — Mode Register Write Timing
4.17.1 Mode Register Write
MRW can be issued from either a Bank-Idle or Bank-Active state. Certain restrictions may apply for MRW
from an Active state.
Table 29 — Truth Table for Mode Register Read (MRR) and Mode Register Write (MRW)
Current State Intermediate State Next State
Command
SDRAM SDRAM SDRAM
Mode Register Reading
MRR All Banks Idle
(All Banks Idle)
All Banks Idle
Mode Register Writing
MRW All Banks Idle
(All Banks Idle)
MRR Mode Register Reading Bank(s) Active
Bank(s) Active
MRW Mode Register Writing Bank(s) Active
Page 83
4.18 VREF Current Generator (VRCG)
LPDDR4 SDRAM VREF current generators (VRCG) incorporate a high current mode to reduce the settling
time of the internal VREF(DQ) and VREF(CA) levels during training and when changing frequency set points
during operation. The high current mode is enabled by setting MR13[OP3] = 1. Only Deselect commands
may be issued until tVRCG_ENABLE is satisfied. tVRCG_ENABLE timing is shown in Figure 36.
T0 T1 T2 T3 T4 T5 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6
CK_c
CK_t
CKE
CS
CA DES MRW-1 MRW-1 MRW-2 MRW-2 DES DES Valid Valid Valid Valid DES DES Valid Valid Valid Valid DES DES
COMMAND DES VRCG Enable: MR13 [OP3] = 1 DES DES Valid DES DES Valid DES DES
tVRCG_ENABLE
TIME BREAK
Figure 36 — VRCG Enable timing
VRCG high current mode is disabled by setting MR13[OP3] = 0. Only Deselect commands may be issued
until tVRCG_DISABLE is satisfied. tVRCG_DISABLE timing is shown in Figure 37.
T0 T1 T2 T3 T4 T5 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6
CK_c
CK_t
CKE
CS
CA DES Valid Valid Valid Valid DES DES MRW-1 MRW-1 MRW-2 MRW-2 DES DES Valid Valid Valid Valid DES DES
COMMAND DES Valid DES DES VRCG Disable: MR13 [OP3] = 0 DES DES Valid DES DES
tVRCG_DISABLE
TIME BREAK
Figure 37 — VRCG Disable timing

Note that LPDDR4 SDRAM devices support VREF(CA) and VREF(DQ) range and value changes without
enabling VRCG high current mode.
Table 30 — VRCG Enable/Disable Timing
Speed 533, 1066, 1600, 2133, 2667, 3200, 3733, 4267Mbps
Units NOTE
Parameter Symbol MIN MAX
VREF high current mode
tVRCG_ENABLE - 200 ns
enable time
VREF high current mode
tVRCG_DISABLE - 100 ns
disable time
Page 84
4.19 CA VREF Training
The DRAM internal CA VREF specification parameters are voltage operating range, stepsize, VREF set
tolerance, VREF step time and VREF valid level.
The voltage operating range specifies the minimum required VREF setting range for LPDDR4 DRAM
devices. The minimum range is defined by VREFmax and VREFmin as depicted in Figure 38.
VDD2
Vin DC m ax
Vrefm ax
Vref
Range
Vrefm in
Vin DC Low
Vswing Sm all System Variance

Vswing Large
Total Range
Figure 38 — VREF operating range(VREFmin, VREFmax)

Page 85
4.19 CA VREF Training (cont’d)
The VREF stepsize is defined as the stepsize between adjacent steps. However, for a given design, DRAM
has one value for VREF step size that falls within the range.
The VREF set tolerance is the variation in the VREF voltage from the ideal setting. This accounts for
accumulated error over multiple steps. There are two ranges for VREF set tolerance uncertainity. The range
of VREF set tolerance uncertainity is a function of number of steps n.
The VREF set tolerance is measured with respect to the ideal line which is based on the two endpoints.
Where the endpoints are at the min and max VREF values for a specified range. An illustration depicting an
example of the stepsize and VREF set tolerance is snown in Figure 39.
Actual Vref
Output
Straight Line
Vref
(endpoint Fit)
Vref Set
Tolerance
Vref
Stepsize
Digital Code
Figure 39 — Example of VREF set tolerance(max case only shown) and stepsize
Page 86
The VREF increment/decrement step times are define by VREF_time-short, Middle and long. The
VREF_time-short, VREF_time-Middle and VREF_time-long is defined from TS to TE as shown in Figure 40
where TE is referenced to when the VREF voltage is at the final DC level within the VREF valid
tolerance(VREF_val_tol).
The VREF valid level is defined by VREF_val tolerance to qualify the step time TE as shown in Figure 40.
This parameter is used to insure an adequate RC time constant behavior of the voltage level change after
any VREF increment/decrement adjustment. This parameter is only applicable for DRAM component level
validation/characerization.
VREF_time-Short is for a single stepsize increment/decrement change in VREF voltage.
VREF_time-Middle is at least 2 stepsizes increment/decrement change within the same VREFCA range in
VREF voltage.
VREF_time-Long is the time including up to VREFmin to VREFmax or VREFmax to VREFmin change across
the VREFCA Range in VREF voltage.
TS - is referenced to MRS command clock

TE - is referenced to the VREF_val_tol
T0 T1 T2 T3 T4 T5 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Ta7 Ta8 Ta9 Ta10 Ta11 Ta12
CK_c
CK_t
CKE
CS
CA DES MRW-1 MRW-1 MRW-2 MRW-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES
OMMAND DES VRFF(CA) Value/Range Set DES DES DES DES DES DES DES DES DES DES DES DES DES DES
Vref_time-Short/Middle/Long
Old Vref Setting Updating VRFF(CA) Setting New Vref Setting
TS TE
Vref Setting TIME BREAK
Adjustment
Figure 40 — VREF_time for Short, Middle and Long Timing Diagram
The MRW command to the mode register bits are as follows.
MR12 OP[5:0] : VREF(CA) Setting

MR12 OP[6] : VREF(CA) Range
Page 87
The minimum time required between two VREF MRS commands is VREF_time-short for single step and
VREF_time-Middle for a full voltage range step.
Vref
Voltage
Vref DC
(VDD2 DC)
Stepsize
Vref_val_tol
t1
Time
Figure 41 — VREF step single stepsize increment case
Vref
Voltage
t1
Stepsize
Vref_val_tol
Vref DC
(VDD2 DC)
Time
Figure 42 — VREF step single stepsize decrement case
Page 88
Vref Vrefmax Vref DC

Voltage (VDD2 DC)
Vref_val_tol
Full t1
Range
Step
Vrefmin
Time
Figure 43 — VREF full step from VREFmin to VREFmax case
Vref Vrefmax
Voltage
Full
Range
Step
t1
Vref_val_tol
Vrefmin Vref DC
(VDD2 DC)
Time
Figure 44 — VREF full step from VREFmax to VREFmin case
Page 89

Table 31 contains the CA internal VREF specifications that will be characterized at the component level for
compliance. The component level characterization method is TBD1.
Table 31 — CA Internal VREF Specifications

Parameter Symbol Min Typ Max Unit Notes
VREF Max operating point Range0 VREF_max_R0 30% - - VDD2 1,11
VREF Min operating point Range0 VREF_min_R0 - - 10% VDD2 1,11
VREF Max operating point Range1 VREF_max_R1 42% - - VDD2 1,11
VREF Min operating point Range1 VREF_min_R1 - - 22% VDD2 1,11
VREF Stepsize VREF_step 0.30% 0.40% 0.50% VDD2 2
-1.00% 0.00% 1.00% VDD2 3,4,6
VREF Set Tolerance VREF_set_tol
-0.10% 0.00% 0.10% VDD2 3,5,7
VREF_time-Short - - 100 ns 8
VREF_time_Middle - - 200 ns 12
VREF Step Time
VREF_time-Long - - 250 ns 9
VREF_time_weak - - 1 ms 13,14
VREF Valid tolerance VREF_val_tol -0.10% 0.00% 0.10% VDD2 10
NOTE 1 VREF DC voltage referenced to VDD2_DC.
NOTE 2 VREF stepsize increment/decrement range. VREF at DC level.
NOTE 3 VREF_new = VREF_old + n*VREF_step; n= number of steps; if increment use "+"; If decrement use "-".
NOTE 4 The minimum value of VREF setting tolerance = VREF_new - 1.0%*VDD2. The maximum value of VREF setting
tolerance = VREF_new + 1.0%*VDD2. For n>4.
NOTE 5 The minimum value of VREF setting tolerance = VREF_new - 0.10%*VDD2. The maximum value of VREF
setting tolerance = VREF_new + 0.10%*VDD2. For n≤ 4.
NOTE 6 Measured by recording the min and max values of the VREF output over the range, drawing a straight line
between those points and comparing all other VREF output settings to that line.
NOTE 7 Measured by recording the min and max values of the VREF output across 4 consecutive steps(n=4), drawing
a straight line between those points and comparing all other VREF output settings to that line.
NOTE 8 Time from MRS command to increment or decrement one step size for VREF.
NOTE 9 Time from MRS command to increment or decrement VREFmin to VREFmax or VREFmax to VREFmin change
across the VREFCA Range in VREF voltage.
NOTE 10 Only applicable for DRAM component level test/characterization purpose. Not applicable for normal mode
of operation. VREF valid is to qualify the step times which will be characterized at the component level.
NOTE 11 DRAM range 0 or 1 set by MR12 OP[6].
NOTE 12 Time from MRS command to increment or decrement more than one step size up to a full range of VREF
voltage withiin the same VREFCA range.
NOTE 13 Applies when VRCG high current mode is not enabled, specified by MR13[OP3] = 0.
NOTE 14 VREF_time_weak covers all VREF(CA) Range and Value change conditions are applied to
VREF_time_Short/Middle/Long.

Page 90
4.20 DQ VREF Training
The DRAM internal DQ VREF specification parameters are voltage operating range, stepsize, VREF set
tolerance, VREF step time and VREF valid level.
The voltage operating range specifies the minimum required VREF setting range for LPDDR4 DRAM
devices. The minimum range is defined by VREFmax and VREFmin as depicted in Figure 45.
VDDQ
Vin DC m ax
Vrefm ax
Vref
Range
Vrefm in
Vin DC Low
Vswing Sm all System Variance

Vswing Large
Total Range
Figure 45 — VREF operating range(VREFmin, VREFmax)
Page 91
4.20 DQ VREF Training (cont’d)
The VREF stepsize is defined as the stepsize between adjacent steps. However, for a given design, DRAM
has one value for VREF step size that falls within the range.
The VREF set tolerance is the variation in the VREF voltage from the ideal setting. This accounts for
accumulated error over multiple steps. There are two ranges for VREF set tolerance
uncertainity. The range of VREF set tolerance uncertainity is a function of number of steps n.
The VREF set tolerance is measured with respect to the ideal line which is based on the two endpoints.
Where the endpoints are at the min and max VREF values for a specified range. An illustration depicting an
example of the stepsize and VREF set tolerance shown in Figure 46.
Actual Vref
Output
Straight Line
Vref
(endpoint Fit)
Vref Set
Tolerance
Vref
Stepsize
Digital Code
Figure 46 — Example of VREF set tolerance(max case only shown) and stepsize
Page 92
The VREF increment/decrement step times are define by VREF_time-short, Middle and long. The
VREF_time-short, VREF_time-Middle and VREF_time-long is defined from TS to TE as shown in Figure 47
where TE is referenced to when the VREF voltage is at the final DC level within the VREF valid
tolerance(VREF_val_tol).
The VREF valid level is defined by VREF_val tolerance to qualify the step time TE as shown in Figure 47.
This parameter is used to insure an adequate RC time constant behavior of the voltage level change after
any VREF increment/decrement adjustment. This parameter is only applicable for DRAM component level
validation/characerization.
VREF_time-Short is for a single stepsize increment/decrement change in VREF voltage.

VREF_time-Middle is at least 2 stepsizes increment/decrement change within the same VREFDQ range in
VREF voltage.
VREF_time-Long is the time including up to VREFmin to VREFmax or VREFmax to VREFmin change across
the VREFDQ Range in Vref voltage.
TS - is referenced to MRS command clock

TE - is referenced to the VREF_val_tol
T0 T1 T2 T3 T4 T5 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Ta7 Ta8 Ta9 Ta10 Ta11 Ta12
CK_c
CK_t
CKE
CS
CA DES MRW-1 MRW-1 MRW-2 MRW-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES
COMMAND DES VRFF(DQ) Value/Range Set DES DES DES DES DES DES DES DES DES DES DES DES DES DES
Vref_time-Short/Middle/Long
Old Vref Setting Updating VRFF(DQ) Setting New Vref Setting
TS TE
Vref Setting TIME BREAK
Adjustment
Figure 47 — VREF_time for Short, Middle and Long Timing Diagram
The MRW command to the mode register bits are as follows.
MR14 OP[5:0] : VREF(DQ) Setting

MR14 OP[6] : VREF(DQ) Range
Page 93
The minimum time required between two VREF MRS commands is VREF_time-short for single step and
VREF_time-Middle for a full voltage range step.
Vref
Voltage
Vref DC
(VDDQ DC)
Stepsize
Vref_val_tol
t1
Time
Figure 48 — VREF step single stepsize increment case
Vref
Voltage
t1
Stepsize
Vref_val_tol
Vref DC
(VDDQ DC)
Time
Figure 49 — VREF step single stepsize decrement case

Page 94
Vref Vrefmax Vref DC

Voltage (VDDQ DC)
Vref_val_tol
Full t1
Range
Step
Vrefmin
Time
Figure 50 — VREF full step from VREFmin to VREFmax case
Vref Vrefmax
Voltage
Full
Range
Step
t1
Vref_val_tol
Vrefmin Vref DC
(VDDQ DC)
Time
Figure 51 — VREF full step from VREFmax to VREFmin case

Page 95
Table 32 contains the DQ internal VREF specifications that will be characterized at the component level for
compliance. The component level characterization method is TBD1.
Table 32 - DQ Internal VREF Specifications

Parameter Symbol Min Typ Max Unit Notes
VREF Max operating point Range0 VREF_max_R0 30% - - VDDQ 1,11
VREF Min operating point Range0 VREF_min_R0 - - 10% VDDQ 1,11
VREF Max operating point Range1 VREF_max_R1 42% - - VDDQ 1,11
VREF Min operating point Range1 VREF_min_R1 - - 22% VDDQ 1,11
VREF Stepsize VREF_step 0.30% 0.40% 0.50% VDDQ 2
-1.00% 0.00% 1.00% VDDQ 3,4,6
VREF Set Tolerance VREF_set_tol
-0.10% 0.00% 0.10% VDDQ 3,5,7
VREF_time-Short - - 100 ns 8
VREF_time_Middle - - 200 ns 12
VREF Step Time
VREF_time-Long - - 250 ns 9
VREF_time_weak - - 1 ms 13,14
VREF Valid tolerance VREF_val_tol -0.10% 0.00% 0.10% VDDQ 10
NOTE 1 VREF DC voltage referenced to VDDQ_DC.

NOTE 2 VREF stepsize increment/decrement range. VREF at DC level.
NOTE 3 VREF_new = VREF_old + n*VREF_step; n= number of steps; if increment use "+"; If decrement use "-".
NOTE 4 The minimum value of VREF setting tolerance = VREF_new - 1.0%*VDDQ. The maximum value of VREF
setting tolerance = VREF_new + 1.0%*VDDQ. For n>4.
NOTE 5 The minimum value of VREF setting tolerance = VREF _new - 0.10%*VDDQ. The maximum value of VREF
setting tolerance = VREF_new + 0.10%*VDDQ. For n< 4.
NOTE 6 Measured by recording the min and max values of the VREF output over the range, drawing a straight line
between those points and comparing all other VREF output settings to that line.
NOTE 7 Measured by recording the min and max values of the VREF output across 4 consectuive steps(n=4), drawing
a straight line between those points and comparing all other VREF output settings to that line.
NOTE 8 Time from MRS command to increment or decrement one step size for VREF .
NOTE 9 Time from MRS command to increment or decrement VREFmin to VREFmax or VREFmax to VREFmin change
across the VREFDQ Range in VREF voltage.
NOTE 10 Only applicable for DRAM component level test/characterization purpose. Not applicable for normal mode
of operation. VREF valid is to qualify the step times which will be characterized at the component level.
NOTE 11 DRAM range 0 or 1 set by MR14 OP[6].
NOTE 12 Time from MRS command to increment or decrement more than one step size up to a full range of VREF
voltage withiin the same VREFDQ range.
NOTE 13 Applies when VRCG high current mode is not enabled, specified by MR13[OP3] = 0.
NOTE 14 VREF_time_weak covers all VREF(DQ) Range and Value change conditions are applied to
VREF_time_Short/Middle/Long.

Page 96
4.21 Command Bus Training
The LPDDR4-SDRAM command bus must be trained before enabling termination for high-frequency
operation. LPDDR4 provides an internal VREF(ca) that defaults to a level suitable for un-terminated, low-
frequency operation, but the VREF(ca) must be trained to achieve suitable receiver voltage margin for
terminated, high-frequency operation. The training mode described here centers the internal VREF(ca) in
the CAdata eye and at the same time allows for timing adjustments of the CS and CA signals to meet
setup/hold requirements. Because it can be difficult to capture commands prior to training the CA inputs,
the training mode described here uses a minimum of external commands to enter, train, and exit the
Command Bus Training mode.
NOTE It is up to the system designer to determine what constitutes “low-frequency” and “high-frequency” based on
the capabilities of the system. Low-frequency should then be defined as an operating frequency in which the system
can reliably communicate with the SDRAM before Command Bus Training is executed.
The LPDDR4-SDRAM die has a bond-pad (ODT-CA) for multi-rank operation. In a multi-rank system, the
terminating rank should be trained first, followed by the nonterminating rank(s). See 4.33.1, ODT Mode
Register, and Table 55 for more information.
The LPDDR4-SDRAM uses Frequency Set-Points to enable multiple operating settings for the die. The
LPDDR4-SDRAM defaults to FSP-OP[0] at power-up, which has the default settings to operate in un-
terminated, low-frequency environments. Prior to training, the mode register settings should be configured
by setting MR13 OP[6]=1B (FSP-WR[1]) and setting all other mode register bits for FSP-OP[1] to the
desired settings for high-frequency operation. Prior to entering Command Bus Training, the SDRAM will be
operating from FSP-OP[x]. Upon Command Bus Training entry when CKE is driven LOW, the LPDDR4-
SDRAM will automatically switch to the alternate FSP register set (FSP-OP[y]) and use the alternate
register settings during training (See Figure 52, Note 6 for more information on FSP-OP register sets).
Upon training exit when CKE is driven HIGH, the LPDDR4-SDRAM will automatically switch back to the
original FSP register set (FSP-OP[x]), returning to the “known-good” state that was operating prior to
training. The training values for VREF(ca) are not retained by the DRAM in FSP-OP[y] registers, and must
be written to the registers after training exit.
1. To enter Command Bus Training mode, issue a MRW-1 command followed by a MRW-2 command to
set MR13 OP[0]=1B (Command Bus Training Mode Enabled).
2. After time tMRD, CKE may be set LOW, causing the LPDDR4-SDRAM to switch from FSP-OP[x] to
FSP-OP[y], and completing the entry into Command Bus Training mode.
A status of DQS_t, DQS_c, DQ and DMI are as follows, and DQ ODT state will be followed Frequency Set
Point function except output pins.
- DQS_t[0], DQS_c[0] become input pins for capturing DQ[6:0] levels by its toggling.
- DQ[5:0] become input pins for setting VREF(ca) Level.
- DQ[6] becomes a input pin for setting VREF(ca) Range.
- DQ[7] and DMI[0] become input pins and their input level is Valid level or floating, either way is fine.
- DQ[13:8] become output pins to feedback its capturing value via command bus by CS signal.
- DQS_t[1], DQS_c[1],DMI[1] and DQ[15:14] become output pins or disable, it means that SDRAM may
drive to a valid level or left floating.
3. At time tCAENT later, LPDDR4 SDRAM can accept to change its VFREF(ca) Range and Value using
input signals of DQS_t[0], DQS_c[0] and DQ[6:0] from existing value that’s setting via MR12 OP[6:0]. The
mapping between MR12 OP code and DQs is shown in Table 33. At least one VREF CA setting is required
before proceeding to next training steps.
Page 97
4.21 Command Bus Training (cont’d)
Table 33 — Mapping of MR12 OP Code and DQ Numbers

Mapping
MR12 OP Code OP6 OP5 OP4 OP3 OP2 OP1 OP0
DQ Number DQ6 DQ5 DQ4 DQ3 DQ2 DQ1 DQ0
4. The new VREF(ca) value must “settle” for time tVREF_LONG before attempting to latch CA information.
5. To verify that the receiver has the correct VREF(ca) setting and to further train the CA eye relative to
clock (CK), values latched at the receiver on the CA bus are asynchronously output to the DQ bus.
6. To exit Command Bus Training mode, drive CKE HIGH, and after time tVREF_LONG issue the MRW-1
command followed by the MRW-2 command to set MR13 OP[0]=0B. After time tMRW the LPDDR4-
SDRAM is ready for normal operation. After training exit the LPDDR4-SDRAM will automatically switch
back to the FSP-OP registers that were in use prior to training.
Command Bus Training may executed from IDLE or Self Refresh states. When executing CBT within the
Self Refresh state, the SDRAM must not be a power down state (i.e., CKE must be HIGH prior to training
entry). Command Bus Training entry and exit is the same, regardless of the SDRAM state from which CBT
is initiated.
4.21.1 Training Sequence for single-rank systems
NOTE An example shown here is assuming an initial low-frequency, no-terminating operating point,
training a high-frequency, terminating operating point. The green text is low-frequency, the magenta text is
high-frequency. Any operating point may be trained from any known good operating point.
1. Set MR13 OP[6]=1B to enable writing to Frequency Set Point ‘y’ (FSP-WR[y]) (or FSP-OP[x], See note).
2. Write FSP-WR[y] (or FSP-WR[x]) registers for all channels to set up high-frequency operating
parameters.
3. Issue MRW-1 and MRW-2 commands to enter Command Bus Training mode.
4. Drive CKE LOW, and change CK frequency to the high-frequency operating point.
5. Perform Command Bus Training (VREFca, CS, and CA).
6. Exit training, a change CK frequency to the low-frequency operating point prior to driving CKE HIGH,
then issue MRW-1 and MRW-2 commands. When CKE is driven HIGH, the SDRAM will automatically
switch back to the FSP-OP registers that were in use prior to training (i.e. trained values are not retained
by the SDRAM).
7. Write the trained values to FSP-WR[y] (or FSP-WR[x]) by issuing MRW-1 and MRW-2 commands to the
SDRAM and setting all applicable mode register parameters.
8. Issue MRW-1 and MRW-2 commands to switch to FSP-OP[y] (or FSP-OP[x]), to turn on termination,
and change CK frequency to the high frequency operating point. At this point the Command Bus is
trained and you may proceed to other training or normal operation.
Page 98
4.21.2 Training Sequence for multi-rank systems
NOTE An example shown here is assuming an initial low-frequency operating point, training a high-fre-
quency operating point. The green text is low-frequency, the magenta text is high-frequency. Any operating
point may be trained from any known good operating point.
1. Set MR13 OP[6]=1B to enable writing to Frequency Set Point ‘y’ (FSP-WR[y]) (or FSP-WR[x], See
Note).
2. Write FSP-WR[y] (or FSP-WR[x]) registers for all channels and ranks to set up high frequency operating
parameters.
3. Read MR0 OP[7] on all channels and ranks to determine which die are terminating, signified by
MR0 OP[7]=1B.
4. Issue MRW-1 and MRW-2 commands to enter Command Bus Training mode on the terminating rank.
5. Drive CKE LOW on the terminating rank (or all ranks), and change CK frequency to the high-frequency
operating point.
6. Perform Command Bus Training on the terminating rank (VREFca, CS, and CA).
7. Exit training by driving CKE HIGH, change CK frequency to the low-frequency operating point, and issue
MRW-1 and MRW-2 commands to write the trained values to FSP-WR[y] (or FSP-WR[x]). When CKE is
driven HIGH, the SDRAM will automatically switch back to the FSP-OP registers that were in use prior to
training (i.e. trained values are not retained by the SDRAM).
8. Issue MRW-1 and MRW-2 command to enter training mode on the non-terminating rank (but keep CKE
HIGH)
9. Issue MRW-1 and MRW-2 commands to switch the terminating rank to FSP-OP[y] (or FSP-OP[x]), to
turn on termination, and change CK frequency to the high frequency operating point.
10. Drive CKE LOW on the non-terminating (or all) ranks. The non-terminating rank(s) will now be using
FSP-OP[y] (or FSP-OP[x]).
11. Perform Command Bus Training on the non-terminating rank (VREFca, CS, and CA).
12. Issue MRW-1 and MRW-2 commands to switch the terminating rank to FSP-OP[x] (or FSP-OP[y]) to
turn off termination.
13. Exit training by driving CKE HIGH on the non-terminating rank, change CK frequency to the low-
frequency operating point, and issue MRW-1 and MRW-2 commands. When CKE is driven HIGH, the
SDRAM will automatically switch back to the FSP-OP registers that were in use prior to training (i.e.
trained values are not retained by the SDRAM).
14. Write the trained values to FSP-WR[y] (or FSP-WR[x]) by issuing MRW-1 and MRW-2 commands to
the SDRAM and setting all applicable mode register parameters.
15. Issue MRW-1 and MRW-2 commands to switch the terminating rank to FSP-OP[y] (or FSP-OP[x]), to
turn on termination, and change CK frequency to the high frequency operating point. At this point the
Command Bus is trained for both ranks and you may proceed to other training or normal operation.
Page 99
4.21.3 Relation between CA input pin DQ output pin.
The relation between CA input pin DQ output pin is shown in Table 34.
Table 34 - Mapping of CA Input pin and DQ Output pin

Mapping
CA Number CA5 CA4 CA3 CA2 CA1 CA0
DQ Number DQ13 DQ12 DQ11 DQ10 DQ9 DQ8
4.21.4 Timing Diagram
The basic Timing diagrams of Command Bus Training are shown in Figure 52 through Figure 55.
T0 T1 T2 T3 T4 T5 Ta0 Tb0 Tb1 Tc0 Td0 Te0 Te1 Te2 Tf0 Tg0 Th0 Th1 Th2
CK_c
CK_t
*1 tCKPRECS tCKPSTCS
*7
CKE *2
tMRD tCKELCK * 3 tCACD
CS
CA DES MRW-1 MRW-1 MRW-2 MRW-2 DES DES DES DES DES Valid Valid
COMMAND DES Enter Command Bus Training mode DES DES DES DES DES CA training CA
Pattern A Pattern B
tDQSCKE tCAENT tVREFca_Long * 5 tADR
*4
DQS_t[0]
DQS_c[0]
tDStrain tDHtrain
DQ[6:0] Valid
DQ[7]
DMI[0]
DQ[13:8] Pattern A
DQ[15:14]
DMI[1]
DQS_t[1]
DQS_c[1]
VrefCA Setting Value of MR X (Y) Updating Setting from FSP Switching Updating Setting Temporary Setting Value
(Reference)
tCKELODTon * 6
ODT_CA Mode Register X (Y) Switching MR Mode Register Y (X)

(Reference)

NOTES : 1. After tCKELCK clock can be stopped or frequency changed any time.
2. The input clock condition should be satisfied tCKPRECS and tCKPSTCS. DON'T CARE TIME BREAK
3. Continue to Drive CK and Hold CA & CS pins low until tCKELCK after CKE is low (which disables command decoding).
4. DRAM may or may not capture first rising/falling edge of DQS_t/c due to an unstable first rising edge. Hence provide at least consecutive 2 pulses of DQS signal input is required in every DQS input signal at capturing DQ6:0 signals.
The captured value of DQ6:0 signal level by each DQS edges are overwritten at any time and the DRAM updates its VREFca setting of MR12 temporary after time tVREFca_Long.
5. tVREF_LONG may be reduced to tVREF_SHORT if the following conditions are met: 1) The new Vref setting is a single step above or below the old Vref setting,
and 2) The DQS pulses a single time, or the new Vref setting value on DQ[6:0] is static and meets tDSTRAIN/tDHTRAIN for every DQS pulse applied.
6. When CKE is driven LOW, the SDRAM will switch its FSP-OP registers to use the alternate (i.e. non-active) set. Example: If the SDRAM is currently using FSP-OP[0],
Page 100
then it will switch to FSP-OP[1] when CKE is driven LOW. All operating parameters should be written to the alternate mode registers before entering Command Bus Training to ensure that ODT settings,
RL/WL/nWR setting, etc., are set to the correct values. If the alternate FSP-OP has ODT_CA disabled then termination will not enable in CA Bus Training mode. If the ODT_CA pad is bonded to Vss or floating,
ODT_CA termination will never enable for that die.
7. When CKE is driven low in Command Bus Training mode, the LPDDR4-SDRAM will change operation to the alternate FSP, i.e. the inverse of the FSP programmed in the FSP-OP mode register.
Figure 52 — Entering Command Bus Training Mode and CA Training Pattern Input and Output with VREFCA Value Update
Page 101
T0 Ta0 Tb0 Tb1 Tc0 Td0 Te0 Te1 Te2 Te3 Te4 Te5 Te6 Te7 Te8 Te9 Te10 Tf0 Tf1 Tf2 Tf3
CK_c
CK_t
*1 tCKPRECS *2
*7
CKE
tCKELCK * 3
CS
CA DES DES DES DES DES Valid
COMMAND DES DES DES DES DES CA training

Pattern A
tDQSCKE tCAENT tVREFca_Long * 5 tCS_VREF tVREFca_Long * 5
*4 *4
DQS_t[0]
DQS_c[0]
tDStrain tDHtrain tDStrain tDHtrain
DQ[6:0] Valid Valid
DQ[7]
DMI[0]
tADR
DQ[13:8] Pattern A
DQ[15:14]
DMI[1]
DQS_t[1]
DQS_c[1]
VrefCA Setting Value of MR X(Y) Updating Setting from FSP Switching Updating Setting Temporary Setting Value Updating Setting
(Reference)
tCKELODTon * 6
ODT_CA Mode Register X (Y) Switching MR Mode Register Y (X)

(Reference)
NOTES : 1. After tCKELCK clock can be stopped or frequency changed any time.
2. The input clock condition should be satisfied tCKPRECS. DON'T CARE TIME BREAK
3. Continue to Drive CK and Hold CA & CS pins low until tCKELCK after CKE is low (which disables command decoding).
4. DRAM may or may not capture first rising/falling edge of DQS_t/c due to an unstable first rising edge. Hence provide at least consecutive 2 pulses of DQS signal input is required in every DQS input signal at capturing DQ6:0 signals.
The captured value of DQ6:0 signal level by each DQS edges are overwritten at any time and the DRAM updates its VREFca setting of MR12 temporary after time tVREFca_Long.
5. tVREF_LONG may be reduced to tVREF_SHORT if the following conditions are met: 1) The new Vref setting is a single step above or below the old Vref setting,
and 2) The DQS pulses a single time, or the new Vref setting value on DQ[6:0] is static and meets tDSTRAIN/tDHTRAIN for every DQS pulse applied.
6. When CKE is driven LOW, the SDRAM will switch its FSP-OP registers to use the alternate (i.e. non-active) set. Example: If the SDRAM is currently using FSP-OP[0],
then it will switch to FSP-OP[1] when CKE is driven LOW. All operating parameters should be written to the alternate mode registers before entering Command Bus Training to ensure that ODT settings,
RL/WL/nWR setting, etc., are set to the correct values. If the alternate FSP-OP has ODT_CA disabled then termination will not enable in CA Bus Training mode. If the ODT_CA pad is bonded to Vss or floating,
ODT_CA termination will never enable for that die.
7. When CKE is driven low in Command Bus Training mode, the LPDDR4-SDRAM will change operation to the alternate FSP, i.e. the inverse of the FSP programmed in the FSP-OP mode register.
Figure 53 — Consecutive VREFCA Value Update

T0 T1 T2 Ta0 Ta1 Ta2 Tb0 Tc0 Td0 Td1 Te0 Te1 Tf0 Tf1 Tf2 Tf3 Tf4 Tg0 Tg1 Tg2 Tg3 Tg4 Tg5
CK_c
CK_t
tCKPSTCS
CKE
*5
tCACD tCKCKEH * 1 tFC tMRW
CS
*2
CA Valid Valid DES DES DES DES DES DES MRW-1 MRW-1 MRW-2 MRW-2 DES Valid Valid Valid Valid DES
COMMAND CA CA
Pattern B Pattern C DES DES DES DES DES DES Exiting Command Bus Training mode DES Valid DES
tADR tADR tCKEHDQS
DQS_t[0]
DQS_c[0]
DQ[6:0]
DQ[7]
DMI[0]
tMRZ
DQ[13:8] Pattern A Pattern B Pattern C
DQ[15:14]
DMI[1]
DQS_t[1]
DQS_c[1]
*4
VrefCA Temporary Setting Value Switching MR Setting Value of MR X (Y)
(Reference)
tCKELODToff * 3
ODT_CA Mode Register Y (X) Switching MR Mode Register X (Y)

(Reference)
NOTES : 1. Clock can be stopped or frequency changed any time before tCKCKEH. CK must meet tCKCKEH before CKE is driven high.
When CKE is driven high the clock frequency must be returned to the original frequency (the frequency corresponding to the FSP at which Command Bus Training mode was entered)
2. CS and CA[5:0] must be Deselect (all low) tCKCKEH before CKE is driven high.
3. When CKE is driven high, the SDRAM’s ODT_CA will revert to the state/value defined by FSP-OP prior to Command Bus Training mode entry, i.e. the original frequency set point (FSP-OP, MR13-OP[7]).
Example: If the SDRAM was using FSP-OP[1] for training, then it will switch to FSP-OP[0] when CKE is driven HIGH.
4. Training values are not retained by the SDRAM, and must be written to the FSP-OP register set before returning to operation at the trained frequency.
Example: VREF(ca) will return to the value programmed in the original set point.
Page 102
5. When CKE is driven high the LPDDR4-SDRAM will revert to the FSP in operation when Command Bus Training mode was entered.
Figure 54 — Exiting Command Bus Training Mode with Valid Command

Page 103
T0 T1 T2 Ta0 Ta1 Ta2 Tb0 Tc0 Td0 Td1 Te0 Te1 Tf0 Tf1 Tf2 Tf3 Tf4 Tg0 Tg1 Tg2 Tg3 Th0 Tk1
CK_c
CK_t
tCKPSTCS tCKCKEL
CKE
*5
tCACD tCKCKEH * 1 tFC tMRD tCKELCMD
CS
*2
CA Valid Valid DES DES DES DES DES DES MRW-1 MRW-1 MRW-2 MRW-2 DES Valid Valid DES DES
COMMAND CA CA
Pattern B Pattern C DES DES DES DES DES DES Exiting Command Bus Training mode DES Power Down Entry DES DES
tADR tADR tCKEHDQS
DQS_t[0]
DQS_c[0]
DQ[6:0]
DQ[7]
DMI[0]
tMRZ
DQ[13:8] Pattern A Pattern B Pattern C
DQ[15:14]
DMI[1]
DQS_t[1]
DQS_c[1]
*4
VrefCA Temporary Setting Value Switching MR Setting Value of MR X (Y)
(Reference)
tCKELODToff * 3
ODT_CA Mode Register Y (X) Switching MR Mode Register X (Y)

(Reference)
NOTES : 1. Clock can be stopped or frequency changed any time before tCKCKEH. CK must meet tCKCKEH before CKE is driven high.
When CKE is driven high the clock frequency must be returned to the original frequency (the frequency corresponding to the FSP at which Command Bus Training mode was entered)
2. CS and CA[5:0] must be Deselect (all low) tCKCKEH before CKE is driven high.
3. When CKE is driven high, the SDRAM’s ODT_CA will revert to the state/value defined by FSP-OP prior to Command Bus Training mode entry, i.e. the original frequency set point (FSP-OP, MR13-OP[7]).
Example: If the SDRAM was using FSP-OP[1] for training, then it will switch to FSP-OP[0] when CKE is driven HIGH.
4. Training values are not retained by the SDRAM, and must be written to the FSP-OP register set before returning to operation at the trained frequency.
Example: VREF(ca) will return to the value programmed in the original set point.
5. When CKE is driven high the LPDDR4-SDRAM will revert to the FSP in operation when Command Bus Training mode was entered.
Figure 55 — Exiting Command Bus Training Mode with Power Down Entry
Page 104
4.21.4 Timing Diagram (cont’d)
Table 35 — Command Bus Training AC Timing

Min/ Data Rate
Parameter Symbol Unit Note
Max
533 1066 1600 2133 2667 3200 3733 4267
Command Bus Training Timing
Clock and Command
tCKELCK Min max(7.5ns, 3nCK) tCK
Valid after CKE Low
Data Setup for
tDStrain Min 2 ns
VREF Training Mode
Data Hold for
tDHtrain Min 2 ns
VREF Training Mode
Asynchronous Data Read tADR Max 20 ns
CA Bus Training
Command
tCACD Min RU(tADR/tCK ) tCK 2
to CA Bus Training
Command Delay
Valid Strobe Requirement
tDQSCKE Min 10 ns 1
before CKE Low
First CA Bus Training
Command Following tCAENT Min 250 ns
CKE Low
VREF Step Time
tVREFca_LONG Max 250 ns
– multiple steps
VREF Step Time
tVREFca_SHORT Max 80 ns
– one step
Valid Clock Requirement
tCKPRECS Min 2tck + tXP (tXP = max(7.5ns, 3nCK)) -
before CS High
tCKPSTCS Min max(7.5ns, 5nCK)) -
after CS High
Minimum delay from
CS to DQS toggle in tCS_VREF Min 2 tCK
command bus training
Minimum delay from
CKE High to tCKEHDQS 10 ns
Strobe High Impedance
Clock and Command
tCKCKEH Min 2 tCK
Valid before CKE High
CA Bus Training CKE
tMRZ Min 1.5 ns
High to DQ Tri-state
ODT turn-on Latency
tCKELODTon Min 20 ns
from CKE
ODT turn-off Latency
tCKELODToff Min 20 ns
from CKE
NOTE 1 DQS_t has to retain a low level during tDQSCKE period, as well as DQS_c has to retain a high level.
NOTE 2 If tCACD is violated, the data for samples which violate tCACD will not be available, except for the last
sample (where tCACD after this sample is met). Valid data for the last sample will be available after tADR.
Page 105
4.22 Frequency Set Point
Frequency Set-Points allow the LPDDR4-SDRAM CA Bus to be switched between two differing operating
frequencies, with changes in voltage swings and termination values, without ever being in an un-trained
state which could result in a loss of communication to the DRAM. This is accomplished by duplicating all
CA Bus mode register parameters, as well as other mode register parameters commonly changed with
operating frequency. These duplicated registers form two sets that use the same mode register addresses,
with read/write access controlled by MR bit FSP-WR (Frequency Set-Point Write/Read) and the DRAM
operating point controlled by another MR bit FSP-OP (Frequency Set-Point Operation). Changing the FSP-
WR bit allows MR parameters to be changed for an alternate Frequency Set-Point without affecting the
LPDDR4-SDRAM's current operation. Once all necessary parameters have been written to the alternate
Set-Point, changing the FSP-OP bit will switch operation to use all of the new parameters simultaneously
(within tFC), eliminating the possibility of a loss of communication that could be caused by a partial
configuration change.
Parameters which have two physical registers controlled by FSP-WR and FSP-OP include:
Table 36 — Mode Register Function with two physical registers

MR# Operand Function Note
OP[2] WR-PRE (WR Pre-amble Length)
OP[3] RD-PRE (RD Pre-amble Type)
MR1
OP[6:4] nWR (Write-Recovery for Auto-Pre-charge commands)
OP[7] PST (RD Post-Amble Length)
OP[2:0] RL (Read latency)
MR2 OP[5:3] WL (Write latency)
OP[6] WLS (Write Latency Set)
OP[0] PU-Cal (Pull-up Calibration Point) 1
OP[1] WR PST(WR Post-Amble Length)
MR3 OP[5:3] PDDS (Pull-Down Drive Strength)
OP[6] DBI-RD (DBI-Read Enable)
OP[7] DBI-WR (DBI-Write Enable)
OP[2:0] DQ ODT (DQ Bus Receiver On-Die-Termination)
MR11
OP[6:4] CA ODT (CA Bus Receiver On-Die-Termination)
OP[5:0] VREF(ca) (VREF(ca) Setting)
MR12
OP[6] VR-CA (VREF(ca) Range)
OP[5:0] VREF(dq) (VREF(dq) Setting)
MR14
OP[6] VR(dq) (VREF(dq) Range)
OP[2:0] SoC ODT (Controller ODT Value for VOH calibration)
OP[3] ODTE-CK (CK ODT enabled for nonterminating rank)
MR22
OP[4] ODTE-CS (CS ODT enable for non terminating rank)
OP[5] ODTD-CA (CA ODT termination disable)
NOTE 1 PU-CAL setting is required as the same value for both Ch.A and Ch.B before issuing ZQ Cal start command.
See 3.4, Mode Register Definition, for more details.
Page 106
4.22 Frequency Set Point (cont’d)
Table 37 shows how the two mode registers for each of the parameters above can be modified by setting
the appropriate FSP-WR value, and how device operation can be switched between operating points by
setting the appropriate FSP-OP value. The FSP-WR and FSP-OP functions operate completely
independently.
Table 37 — Relation between MR Setting and DRAM Operation
MR# &
Function Data Operation Note
Operand
Data write to Mode Register N for FSP-OP[0]
0 (Default)
by MRW Command
FSP-WR MR13 OP[6] 1
Data write to Mode Register N for FSP-OP[1]
1
by MRW Command
DRAM operates with Mode Register N for FSP-OP[0]
0 (Default)
setting.
FSP-OP MR13 OP[7] 2
DRAM operates with Mode Register N for FSP-OP[1]
1
setting.
NOTE 1 FSP-WR stands for Frequency Set Point Write/Read.
NOTE 2 FSP-OP stands for Frequency Set Point Operating Point.
4.22.1 Frequency set point update Timing
The Frequency set point update timing is shown in Figure 56. When changing the frequency set point via
MR13 OP[7], the VRCG setting: MR13 OP[3] have to be changed into VREF Fast Response (high current)
mode at the same time. After Frequency change time(tFC) is satisfied. VRCG can be changed into Normal
Operation mode via MR13 OP[3].
T0 T1 T2 T3 T4 T5 Ta0 Ta1 Tb0 Tb1 Tc0 Tc1 Tc2 Tc3 Tc4 Tc5 Tc6
CK_c
CK_t
tCKFSPE Frequency tCKFSPX tVRCG_DISABLE
Change
CKE
CS
CA DES MRW-1 MRW-1 MRW-2 MRW-2 DES DES DES DES DES MRW-1 MRW-1 MRW-2 MRW-2 DES
COMMAND DES FSP changes from 0 to 1 DES DES DES DES DES VRCG changes DES
VRCG changes from normal to High current from High current to normal
tFC
Applicable
Mode Mode Register for FSP-OP[0] Switching Mode Register Mode Register for FSP-OP[1]
Register
Figure 56 — Frequency Set Point Switching Timing

Page 107
4.22.1 Frequency set point update Timing (cont’d)
Table 38 — Frequency Set Point AC Timing

Min/ Data Rate
Max
533 1066 1600 2133 2667 3200 3733 4267
Frequency Set Point parameters
tFC_Short MIN TBD ns 1

Frequency Set Point
tFC_Middle MIN TBD ns 1
Switching Time
tFC_Long MIN 250 ns 1
tCKFSPE MIN max(7.5ns, 4nCK) -
after Entering FSP Change
before 1st Valid Command tCKFSPX MIN max(7.5ns, 4nCK) -
after FSP change
NOTE 1 Frequency Set Point Switching Time depends on value of VREF(ca) setting: MR12 OP[5:0] and VREF(ca)
Range: MR12 OP[6] of FSP-OP 0 and 1. The details are shown in Table 39. Additionally change of Frequency Set
Point may affect VREF(dq) setting. Settling time of VREF(dq) level is same as VREF(ca) level.
Table 39 — tFC value mapping

Step Size Range
Application From To From To
FSP -OP0 FSP-OP1 FSP -OP0 FSP-OP1
tFC_Short Base A single step size increment/decrement Base No Change
tFC_Middle Base Two or more step size increment/decrement Base No Change
tFC_Long - - Base Change
NOTE 1 As well as change from FSP-OP1 to FSP-OP0.
Page 108
The LPDDR4-SDRAM defaults to FSP-OP[0] at power-up. Both Set-Points default to settings needed to
operate in un-terminated, low-frequency environments. To enable the LPDDR4-SDRAM to operate at
higher frequencies, Command Bus Training mode should be utilized to train the alternate Frequency Set-
Point (Figure 57). See 4.21, Command Bus Training, for more details on this training mode.
Prepare for CA Bus

Powerup/ Training of FSP[1] for CA Bus Training,
Initialization High Frequency FSP-OP[1]
FSP-OP=0 FSP-OP=0 FSP-WR=1
FSP-WR=0 FSP-WR=1 CKE high->low
Freq = high
Freq = Boot
Freq = Boot
CKE low->high
Prepare for CA Bus

Exit CA Bus Switch to High-
Training of FSP[0]
Training Speed Mode for Med. Frequency
FSP-OP=0 FSP-OP=1
FSP-OP=1
FSP-WR=1 FSP-WR=1
FSP-WR=0
Freq = Boot Freq = High
Freq = high
CKE high->low
Exit CA Bus
CA Bus Training,
Training Operate at
FSP-OP[0] FSP-OP=1
FSP-WR=0 CKE low->high FSP-WR=0
High Speed
Freq = Med
Freq = High
Figure 57 — Training Two Frequency Set-Points
Once both Frequency Set-Points have been trained, switching between points can be performed by a
single MRW followed by waiting for tFC (Figure 58).
State n-1 : FSP-OP=1

Operate at Operate at
MWR Command
High Speed tFC Medium Speed
State n : FSP-OP=0
State n-1 : FSP-OP=0

MWR Command Operate at
State n : FSP-OP=1 tFC High Speed
Figure 58 — Switching Between Two Trained Frequency Set-Points

Page 109
Switching to a third (or more) Set-Point can be accomplished if the memory controller has stored the
previously-trained values (in particular the VREF-CA calibration value) and re-writes these to the alternate
Set-Point before switching FSP-OP (Figure 59).
MRW Vref-CA
Operate at State n-1: FSP-WR=1 CA-ODT, DQ-ODT,
High Speed State n: FSP-WR =0 tFC
RL, WL, Vref-
DQ,...etc.
Operate at
State n-1: FSP-OP=1
State n: FSP-OP =0
Third
tFC Speed
Figure 59 — Switching to a Third Trained Frequency Set-Point

Page 110
4.23 Mode Reister Write-WR Leveling Mode
To improve signal-integrity performance, the LPDDR4 SDRAM provides a write-leveling feature to

compensate CK-to-DQS timing skew affecting timing parameters such as tDQSS, tDSS, and tDSH. The
DRAM samples the clock state with the rising edge of DQS signals, and asynchronously feeds back to the
memory controller. The memory controller references this feedback to adjust the clock-to-data strobe
signal relationship for each DQS_t/DQS_c signal pair.
All data bits (DQ[7:0] for DQS_t/DQS_c[0], and DQ[15:8] for DQS_t/DQS_c[1]) carry the training feedback
to the controller. Both DQS signals in each channel must be leveled independently. Write-leveling entry/exit
is independent between channels.
The LPDDR4 SDRAM enters into write-leveling mode when mode register MR2-OP[7] is set HIGH. When
entering write-leveling mode, the state of the DQ pins is undefined. During write-leveling mode, only
DESELECT commands are allowed, or a MRW command to exit the write-leveling operation. Depending
on the absolute values of tDQSL and tDQSH in the application, the value of tDQSS may have to be better
than the limits provided in the Write AC Timing Table1 in order to satisfy the tDSS and tDSH specifications.
Upon completion of the write-leveling operation, the DRAM exits from write-leveling mode when MR2-
OP[7] is reset LOW.
Write Leveling should be performed before Write Training (DQS2DQ Training).
4.23.1 Write Leveling Procedure
The following steps should be folloed for Write Leveling:
1. Enter into Write-leveling mode by setting MR2-OP[7]=1.

2. Once entered into Write-leveling mode, DQS_t must be driven LOW and DQS_c HIGH after a delay of
tWLDQSEN.
3. Wait for a time tWLMRD before providing the first DQS signal input. The delay time tWLMRD(MAX) is
controllerdependent.
4. DRAM may or may not capture first rising edge of DQS_t due to an unstable first risign edge. Hence
provide at least consecutive 2 pulses of DQS signal input is required in every DQS input signal during
Write Training Mode.
The captured clock level by each DQS edges are overwritten at any time and the DRAM provides
asynchronous feedback on all the DQ bits after time tWLO.
5. The feedback provided by the DRAM is referenced by the controller to increment or decrement the
DQS_t and/or DQS_c delay settings.
6. Repeat step 4 through step 5 until the proper DQS_t/DQS_c delay is established.
7. Exit from Write-leveling mode by setting MR2-OP[7]=0.

Page 111
4.23.1 Write Leveling Procedure (cont’d)
Write Leveling timing examples are shown in Figure 60 and Figure 61.
T0 T1 T2 T3 T4 Ta0 Ta1 Tb0 Tb1 Tc0 Tc1 Td0 Td1 Td2 Td3 Te0 Te1 Tf0 Tf1 Tf2 Tf3 Tf4 Tg0 Tg1 Tg2 Tg3
CK_t
CK_c
CS
CKE
MRW MRW MRW MRW MRW MRW MRW MRW
CA MA MA OP OP
DES DES DES DES DES DES DES DES DES DES DES DES DES DES MA MA OP OP
DES DES Valid Valid Valid
MRW-1 MRW-2 MRW-1 MRW-2

CMD WR Leveling WR Leveling
Deselect Deselect Deselect Deselect Deselect Deselect Deselect
WR Lev Exit WR Lev Exit
Deselect Valid
tWLDQSEN tDQSH
tWLWPRE tDQSL
DQS_c
DQS_t
tWLMRD tWLO tWLO tMRD
tWLO
tWLO
DQ
Figure 60 — Write Leveling Timing, tDQSL(max)
T0 T1 T2 T3 T4 Ta0 Ta1 Tb0 Tb1 Tc0 Tc1 Td0 Td1 Td2 Td3 Te0 Te1 Tf0 Tf1 Tf2 Tf3 Tf4 Tg0 Tg1 Tg
CK_t
CK_c
CS
CKE
MRW MRW MRW MRW MRW MRW MRW MRW
CA MA MA OP OP
DES DES DES DES DES DES DES DES DES DES DES DES DES DES MA MA OP OP
DES DES Valid Valid Va
MRW-1 MRW-2 MRW-1 MRW-2

Deselect Deselect Deselect Deselect Deselect Deselect Deselect
WR Lev Exit WR Lev Exit
Deselect Valid
tWLDQSEN tDQSH tDQSL

tWLWPRE
DQS_c
DQS_t
tWLMRD tWLO tMRD
tWLO
tWLO
DQ
Figure 61 — Write Leveling Timing, tDQSL(min)

Page 112
4.23.2 Input Clock Frequency Stop and Change
The input clock frequency can be stopped or changed from one stable clock rate to another stable clock
rate during Write Leveling mode.
The Frequency stop or change timing is shown in Figure 62.
T0 T1 T2 T3 T4 Ta0 Ta1 Tb0 Tb1 Tb2 Tc0 Td0 Te0 Te1 Te2 Te3 Te4 Tf0 Tf1 Tf2 Tf3
CK_t
CK_c
tCKPSTDQS tCKPRDQS
CS
CKE
MRW MRW MRW MRW
CA MA MA OP OP
DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES
MRW-1 MRW-2
Deselect Deselect Deselect Deselect Deselect Deselect Deselect Deselect
tDQSH
tWLDQSEN
tWLWPRE tDQSL
DQS_c
DQS_t
tWLMRD tWLO tWLO
tWLO tWLO
DQ
NOTES : 1. CK_t is held LOW and CK_c is held HIGH during clock stop.
2. CS shall be held LOW during clock clock stop
Figure 62 — Clock Stop and Timing during Write Leveling
Table 40 — Write Leveling Timing Parameters

Parameter Symbol Min/Max Value Units Notes
DQS_t/DQS_c delay Min 20
tWLDQSEN tCK
after write leveling mode is programmed Max -
Min 20
Write preamble for Write Leveling tWLWPRE tCK
Max -
First DQS_t/DQS_c edge Min 40
tWLMRD tCK
after write leveling mode is programmed Max -
Min 0
Write leveling output delay tWLO ns
Max 20
Min max(14ns, 10nCK)
Mode register set command delay tMRD ns
Max -
Min max(7.5ns, 4nCK)
Valid Clock Requirement before DQS Toggle tCKPRDQS -
Max -
Min max(7.5ns, 4nCK)
Valid Clock Requirement after DQS Toggle tCKPSTDQS -
Max -
Page 113
4.23.3 Write Leveling Setup and Hold Time
Table 41 — Write Leveling Setup and Hold Time

Data Rate
Parameter Symbol Min/Max Unit
1600 2400 3200 4266
Write Leveling Parameters
Write leveling hold time tWLH MIN 150 100 75 50 ps
Write leveling setup time tWLS MIN 150 100 75 50 ps
Write leveling invalid window tWLIVW_Total MIN 240 160 120 90 ps
NOTE 1 In addition to the traditional setup and hold time specifications above, there is value in a invalid window
based specification for write-leveling training. As the training is based on each device, worst case process skews for
setup and hold do not make sense to close timing between CK and DQS.
NOTE 2 tWLIVW_Total is defined in a similar manner to tdIVW_Total, except that here it is a DQS invalid window with
respect to CK. This would need to account for all VT (voltage and temperature) drift terms between CK and DQS within
the DRAM that affect the write-leveling invalid window.
The DQS input mask for timing with respect to CK is shown in Figure 63. The "total" mask (TdiVW_total)
defines the time the input signal must not encroach in order for the DQS input to be successfully captured
by CK with a BER of lower than tbd. The mask is a receiver property and it is not the valid data-eye.
DQS_t/DQS_c and CK_t/CK_c at DRAM Latch
Internal Componsite DQS Eye

Center aligned to CK
CK_c
CK_t
DQS_diff =
DQS_t-DQS_c
TWLIVW
Figure 63 — DQS_t/DQS_c to CK_t/CK_t timings at the DRAM pins

referenced from the internal latch
Page 114
4.24 RD DQ Calibration
LPDDR4 devices feature a RD DQ Calibration training function that outputs a 16-bit user-defined pattern
on the DQ pins. RD DQ Calibration is initiated by issuing a MPC-1 [RD DQ Calibration] command followed
by a CAS-2 command, cause the LPDDR4-SDRAM to drive the contents of MR32 followed by the contents
of MR40 on each of DQ[15:0] and DMI[1:0]. The pattern can be inverted on selected DQ pins according to
user-defined invert masks written to MR15 and MR20.
4.24.1 RD DQ Calibration Training Procedure
The procedure for executing RD DQ Calibration is:
• Issue MRW commands to write MR32 (first eight bits), MR40 (second eight bits), MR15 (eight-bit invert
mask for byte 0), and MR20 (eight-bit invert mask for byte 1).
• Optionally this step could be skipped to use the default patterns
- MR32 default = 5Ah
- MR40 default = 3Ch
- MR15 default = 55h
- MR20 default = 55h
• Issue an MPC-1 [RD DQ Calibration] command followed immediately by a CAS-2 command.
• Each time an MPC-1 [RD DQ Calibration] command followed by a CAS-2 is received by the LPDDR4
SDRAM, a 16-bit data burst will, after the currently set RL, drive the eight bits programmed in MR32
followed by the eight bits programmed in MR40 on all I/O pins.
• The data pattern will be inverted for I/O pins with a '1' programmed in the corresponding invert mask
mode register bit (see Table 42).
• Note that the pattern is driven on the DMI pins, but no data bus inversion function is enabled, even if
Read DBI is enabled in the DRAM mode register.
• This command can be issued every tCCD seamlessly, and can be issued seamlessly with array Read
commands.
• The operands received with the CAS-2 command must be driven LOW.
• DQ Read Training can be performed with any or no banks active, during Refresh, or during SREF with
CKE high.
Table 42 — Invert Mask Assignments
DQ Pin 0 1 2 3 DMI0 4 5 6 7
MR15 bit 0 1 2 3 N/A 4 5 6 7
DQ Pin 8 9 10 11 DMI1 12 13 14 15
MR20 bit 0 1 2 3 N/A 4 5 6 7
T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 T24 T25 T26 T27 T28 T29 T30 T31
CK_t,
CK_c
tRTRRD tCCD tRTRRD
CA READ READ CAS-2 CAS-2 MPC MPC CAS-2 CAS-2 MPC MPC CAS-2 CAS-2 READ READ CAS-2 CAS-2
CMD READ CAS-2 DES DES MPC CAS-2 DES DES MPC CAS-2 DES DES READ CAS-2 DES DES
bank m col a Read Training Dummy Read Training Dummy Bank n Col b
RL tDQSCK
DQS_t
DQS_c
tRPRE tDQSQ
DQ[15:0] D0 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D11 D12 D13 D14 D15 D0 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D11 D12 D13 D14 D15 D0 D1 D2 D3 D4 D5
DMI[1:0] m m m m m m m m m m m m m m m m c al c al c al c al cal c al cal c al cal c al c al c al c al cal c al cal c al cal c al c al cal c al
Note
1. Array READ commands before and after MPC-Read Training commands are shown for illustration only and are not required DON'T CARE
Figure 64 — DQ Read Training Timing

Page 115
4.24.2 DQ Read Training Example
An example of DQ Read Training output is shown in Table 43. This shows the 16-bit data pattern that will
be driven on each DQ in byte 0 when one DQ Read Training command is executed. This output assumes
the following mode register values are used:
- MR32 = 1CH
- MR40 = 59H
- MR15 = 55H
- MR20 = 55H
Table 43 — DQ Read Calibration Bit Ordering and Inversion Example

Bit Sequence →
Pin
Invert 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
DQ0 Yes 1 0 1 0 0 1 1 0 1 1 1 0 0 0 1 1
DQ1 No 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DQ2 Yes 1 0 1 0 0 1 1 0 1 1 1 0 0 0 1 1
DQ3 No 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DMI0 Never 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DQ4 Yes 1 0 1 0 0 1 1 0 1 1 1 0 0 0 1 1
DQ5 No 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DQ6 Yes 1 0 1 0 0 1 1 0 1 1 1 0 0 0 1 1
DQ7 No 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DQ8 Yes 1 0 1 0 0 1 1 0 1 1 1 0 0 0 1 1
DQ9 No 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DQ10 Yes 1 0 1 0 0 1 1 0 1 1 1 0 0 0 1 1
DQ11 No 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DMI1 Never 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DQ12 Yes 1 0 1 0 0 1 1 0 1 1 1 0 0 0 1 1
DQ13 No 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
DQ14 Yes 1 0 1 0 0 1 1 0 1 1 1 0 0 0 1 1
DQ15 No 0 1 0 1 1 0 0 1 0 0 0 1 1 1 0 0
NOTE 1 The patterns contained in MR32 and MR40 are transmitted on DQ[15:0] and DMI[1:0] when RD DQ Calibra-
tion is initiated via a MPC-1 [RD DQ Calibration] command. The pattern transmitted serially on each data lane, orga-
nized "little endian" such that the low-order bit in a byte is transmitted first. If the data pattern is 27H, then the first bit
transmitted with be a '1', followed by '1', '1', '0', '0', '1', '0', and '0'. The bit stream will be 00100111 →.
NOTE 2 MR15 and MR22 may be used to invert the MR32/MR40 data pattern on the DQ pins. See MR15 and MR20
for more information. Data is never inverted on the DMI[1:0] pins.
NOTE 3 The data pattern is not transmitted on the DMI[1:0] pins if DBI-RD is disabled via MR3-OP[6].
NOTE 4 No Data Bus Inversion (DBI) function is enacted during RD DQ Calibration, even if DBI is enabled in MR3-
OP[6].
Page 116
4.25 DQS-DQ Training
The LPDDR4-SDRAM uses an un-matched DQS-DQ path to enable high speed performance and save
power in the DRAM. As a result, the DQS strobe must be trained to arrive at the DQ latch center-aligned
with the Data eye. The SDRAM DQ receiver is located at the DQ pad, and has a shorter internal delay in
the SDRAM than does the DQS signal. The SDRAM DQ receiver will latch the data present on the DQ bus
when DQS reaches the latch, and training is accomplished by delaying the DQ signals relative to DQS
such that the Data eye arrives at the receiver latch centered on the DQS transition.
Two modes of training are available in LPDDR4:
• Command-based FIFO WR/RD with user patterns

• A internal DQS clock-tree oscillator, to determine the need for, and the magnitude of required training.
The command-based FIFO WR/RD uses the MPC command with operands to enable this special mode of
operation. When issuing the MPC command, if OP6 is set LOW then the DRAM will perform a NOP
command. When OP6 is set HIGH, then OP5:0 enable training functions or are reserved for future use
(RFU). MPC commands that initiate a Read FIFO, READ DQ Calibration or Write FIFO to the SDRAM
must be followed immediately by a CAS-2 command. See 4.28, Multi Purpose Command (MPC)
Definition” for more information.
To perform Write Training, the controller can issue a MPC [Write DQ FIFO] command with OP[6:0] set as
described in 4.28, followed immediately by a CAS-2 command (CAS-2 operands should be driven LOW) to
initiate a Write DQ FIFO. Timings for MPC [Write DQ FIFO] are identical to a Write command, with WL
(Write Latency) timed from the 2nd rising clock edge of the CAS-2 command. Up to 5 consecutive MPC
[Write DQ FIFO] commands with user defined patterns may be issued to the SDRAM to store up to 80
values (BL16 x5) per pin that can be read back via the MPC [Read DQ FIFO] command. Write/Read FIFO
Pointer operation is described in 4.25.1.
After writing data to the SDRAM with the MPC [Write DQ FIFO] command, the data can be read back with
the MPC [Read DQ FIFO] command and results compared with “expect” data to see if further training (DQ
delay) is needed. MPC [Read DQ FIFO] is initiated by issuing a MPC command with OP[6:0] set as
described in 4.28, followed immediately by a CAS-2 command (CAS-2 operands must be driven LOW).
Timings for the MPC [Read DQ FIFO] command are identical to a Read command, with RL (Read Latency)
timed from the 2nd rising clock edge of the CAS-2 command.
Read DQ FIFO is non-destructive to the data captured in the FIFO, so data may be read continuously until
it is either overwritten by a Write DQ FIFO command or disturbed by CKE LOW or any of the following
commands; Write, Masked Write, Read, Read DQ Calibration and a MRR. If fewer than 5 Write DQ FIFO
commands were executed, then unwritten registers will have un-defined (but valid) data when read back.
The following command about MRW is only allowed from MPC [Write DQ FIFO] command to MPC [Read
DQ FIFO].
Allowing MRW command is for OP[7]:FSP-OP, OP[6]:FSP-WR and OP[3]:VRCG of MR13 and MR14. And
the rest of MRW command is prohibited.
For example: If 5 Write DQ FIFO commands are executed sequentially, then a series of Read DQ FIFO
commands will read valid data from FIFO[0], FIFO[1]….FIFO[4], and will then wrap back to FIFO[0] on the
next Read DQ FIFO.
On the other hand, if fewer than 5 Write DQ FIFO commands are executed sequentially (example=3), then
a series of Read DQ FIFO commands will return valid data for FIFO[0], FIFO[1], and FIFO[2], but the next
two Read DQ FIFO commands will return un-defined data for FIFO[3] and FIFO[4] before wrapping back to
the valid data in FIFO[0].
Page 117
4.25.1 FIFO Pointer Reset and Synchronism
The Read and Write DQ FIFO pointers are reset under the following conditions:
• Power-up initialization
• RESET_n asserted
• Power-down entry
• Self Refresh Power-Down entry
The MPC [Write DQ FIFO] command advances the WR-FIFO pointer, and the MPC [Read DQ FIFO]
advances the RD-FIFO pointer. Also any normal (non-FIFO) Read Operation (RD, RDA) advances both
WR-FIFO pointer and RD-FIFO pointer. Issuing (non-FIFO) Read Operation command is inhibited during
Write training period. To keep the pointers aligned, the SoC memory controller must adhere to the following
restriction at the end of Write training period:
• b = a + (n x c)
Where:
‘a’ is the number of MPC [Write DQ FIFO] commands

‘b’ is the number of MPC [Read DQ FIFO] commands
‘c’ is the FIFO depth (=5 for LPDDR4)
‘n’ is a positive integer, ≥ 0
t0 t1 t2 t3 t4 ta ta+1 ta+2 ta+3 ta+4 ta+5 tb tb+1 tb+2 tb+3 tb+4 tb+5 tc tc+1 tc+2 tc+3 tc+4 td td+1 td+2 te te+1 te+2
CK_t,
CK_c
tWRWTR tCCD
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
WL
WL
CMD WRITE-1 CAS-2 VALID MPC CAS-2 VALID MPC CAS-2 VALID VALID VALID VALID VALID VALID
WR FIFO WR FIFO
tDQSS tDQSS
DQS_t
DQS_c
tWPRE tDQS2DQ tDQS2DQ
DQ[15:0] D0 D1 D2 D3 D8 D9 D10 D11 D12 D13 D14 D15 D0 D1 D12 D13 D14 D15 D0 D1 D12 D13 D14 D15
DMI[1:0]
NOTES : 1. MPC [WR FIFO] can be executed with a single bank or multiple banks active, during Refresh, or during SREF with CKE HIGH.
2. Write-1 to MPC-1 is shown as an example of command-to-command timing for MPC.
Timing from Write-1 to MPC [WR-FIFO] is specified in the command-to-command timing table.
3. Seamless MPC [WR-FIFO] commands may be executed by repeating the command every tCCD time.
4. MPC [WR-FIFO] uses the same command-to-data timing relationship (WL, tDQSS, tDQS2DQ) as a Write-1 command.
5. A maximum of 5 MPC [WR-FIFO] commands may be executed consecutively without corrupting FIFO data. The 6th MPC [WR-FIFO] command will overwrite
the FIFO data from the first command. If fewer than 5 MPC [WR-FIFO] commands are executed, then the remaining FIFO locations will contain undefined data.
6. For the CAS-2 command following a MPC command, the CAS-2 operands must be driven “LOW.”
7. To avoid corrupting the FIFO contents, MPC [RD-FIFO] must immediately follow MPC [WR-FIFO]/CAS-2 without any other commands disturbing FIFO pointers
in-between. FIFO pointers are disturbed by CKE Low, Write, Masked Write, Read, Read DQ Calibration and MRR. See Write Training section for more information
on FIFO pointer behavior.
Figure 65 — MPC [Write FIFO] Operation (shown with tWPRE=2nCK, tWPST=0.5nCK)

Page 118
4.25.1 FIFO Pointer Reset and Synchronism (cont’d)
2. MPC [WR-FIFO] to MPC [RD-FIFO] is shown as an example of command-to-command timing for MPC.
Timing from MPC [WR-FIFO] to MPC [RD-FIFO] is specified in the command-to-command timing table.
3. Seamless MPC [RD-FIFO] commands may be executed by repeating the command every tCCD time.
4. MPC [RD-FIFO] uses the same command-to-data timing relationship (RL, tDQSCK) as a Read-1 command.
5. Data may be continuously read from the FIFO without any data corruption. After 5 MPC [RD‐FIFO] commands the FIFO pointer will wrap back to
the 1st FIFO and continue advancing. If fewer than 5 MPC [WR‐FIFO] commands were executed, then the MPC [RD-FIFO] commands to those
FIFO locations will return undefined data. See the Write Training section for more information on the FIFO pointer behavior.
6. For the CAS-2 command immediately following a MPC-1 command, the CAS-2 operands must be driven “LOW.”
7. DMI[1:0] signals will be driven if any of WR-DBI, RD-DBI, or DM is enabled in the mode registers. See Write Training section for more information on DMI behavior.
Figure 66 — MPC [Read FIFO] Operation

(shown with tWPRE=2nCK, tWPST=0.5nCK, tRPRE=toggling, tRPST=1.5nCK)
t0 t1 t2 t3 t4 ta ta+1 ta+2 ta+3 ta+4 ta+5 tb tb+1 tb+2 tb+3 tb+4 tc tc+1 tc+2 td td+1 td+2 td+3 te te+1 te+2 te+3 te+4 te+5
CK_t,
CK_c
tRTRRD
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
RL
RL
CMD MPC CAS-2 VALID Read-1 CAS-2 VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
Read FIFO
tDQSCK tDQSCK
DQS_t
DQS_c
tRPRE tRPST tRPRE tRPST
DQ[15:0] D0 D1 D2 D13 D14 D15 D0 D1 D10 D11 D12 D13 D14 D15
DMI[1:0]
2. MPC [RD-FIFO] to Read-1 Operation is shown as an example of command-to-command timing for MPC.
Timing from MPC [RD-FIFO] command to Read is specified in the command-to-command timing table.
5. Data may be continuously read from the FIFO without any data corruption. After 5 MPC [RD-FIFO] commands the FIFO pointer will wrap back to
the 1st FIFO and continue advancing. If fewer than 5 MPC [WR-FIFO] commands were executed, then the MPC [RD-FIFO] commands to those
6. For the CAS-2 command immediately following a MPC command, the CAS-2 operands must be driven “LOW.”
Figure 67 — FIFO Read to Read Operation (shown with tRPRE=toggling, tRPST=1.5nCK)

Page 119
4.26 DQS Interval Oscillator
As voltage and temperature change on the SDRAM die, the DQS clock tree delay will shift and may require
re-training. The LPDDR4-SDRAM includes an internal DQS clock-tree oscillator to measure the amount of
delay over a given time interval (determined by the controller), allowing the controller to compare the
trained delay value to the delay value seen at a later time. The DQS Oscillator will provide the controller
with important information regarding the need to re-train, and the magnitude of potential error.
The DQS Interval Oscillator is started by issuing a MPC [Start DQS Osc] command with OP[6:0] set as
described in 4.28, which will start an internal ring oscillator that counts the number of time a signal
propagates through a copy of the DQS clock tree.
The DQS Oscillator may be stopped by issuing a MPC [Stop DQS Osc] command with OP[6:0] set as
described in 4.28, or the controller may instruct the SDRAM to count for a specific number of clocks and
then stop automatically (See MR23 for more information). If MR23 is set to automatically stop the DQS
Oscillator, then the MPC [Stop DQS Osc] command should not be used (illegal). When the DQS Oscillator
is stopped by either method, the result of the oscillator counter is automatically stored in MR18 and MR19.
The controller may adjust the accuracy of the result by running the DQS Interval Oscillator for shorter (less
accurate) or longer (more accurate) duration. The accuracy of the result for a given temperature and
voltage is determined by the following equation:
2 * (DQS delay)
DQS Oscillator Granularity Error =
Run Time
Where:
Run Time = total time between start and stop commands

DQS delay = the value of the DQS clock tree delay (tDQS2DQ min/max)
Additional matching error must be included, which is the difference between DQS training circuit and the
actual DQS clock tree across voltage and temperature. The matching error is vendor specific.
Therefore, the total accuracy of the DQS Oscillator counter is given by:
DQS Oscillator Accuracy = 1 - Granularity Error - Matching Error

Page 120
4.26 DQS Interval Oscillator (cont’d)
Example: If the total time between start and stop commands is 100ns, and the maximum DQS clock tree
delay is 800ps (tDQS2DQ max), then the DQS Oscillator Granularity Error is:
2 * (0.8ns)
DQS Oscillator Granularity Error = = 1.6%
100ns
This equates to a granularity timing error of 12.8ps.
Assuming a circuit Matching Error of 5.5ps across voltage and temperature, then the accuracy is:
12.8 + 5.5
DQS Oscillator Accuracy = 1- = 97.7%
800
Example: Running the DQS Oscillator for a longer period improves the accuracy. If the total time between
start and stop commands is 500ns, and the maximum DQS clock tree delay is 800ps (tDQS2DQ max),
then the DQS Oscillator Granularity Error is:
2 * (0.8ns)
DQS Oscillator Granularity Error = = 0.32%
500ns
This equates to a granularity timing error or 2.56ps.
Assuming a circuit Matching Error of 5.5ps across voltage and temperature, then the accuracy is:
2.56 + 5.5
DQS Oscillator Accuracy = 1- = 99.0%
800
The result of the DQS Interval Oscillator is defined as the number of DQS Clock Tree Delays that can be
counted within the “run time,” determined by the controller. The result is stored in MR18-OP[7:0] and
MR19-OP[7:0]. MR18 contains the least significant bits (LSB) of the result, and MR19 contains the most
significant bits (MSB) of the result. MR18 and MR19 are overwritten by the SDRAM when a MPC-1 [Stop
DQS Osc] command is received. The SDRAM counter will count to its maximum value (=2^16) and stop. If
the maximum value is read from the mode registers, then the memory controller must assume that the
counter overflowed the register and discard the result. The longest “run time” for the oscillator that will not
overflow the counter registers can be calculated as follows:
Longest Run Time Interval = 216 * tDQS2DQ(min) = 216 * 0.2ns = 13.1us

Page 121
4.26.1 Interval Oscillator matching error
The interval oscillator matching error is defined as the difference between the DQS training ckt(interval
oscillator) and the actual DQS clock tree across voltage and temperature.
• Parameters:
- tDQS2DQ: Actual DQS clock tree delay
- tDQSOSC: Training ckt(interval oscillator) delay
- OSCOffset: Average delay difference over voltage and temp(shown in Figure 68)
- OSCMatch: DQS oscillator matching error
Offset 2
tDQS2DQ
tDQS osc
Time
(ps)
OSC offset = avg(offset1,offset2)
Offset 1(at end point) = tDQS2DQ(V,T) - tDQS osc (V,T)

Offset 2(at end point) = tDQS2DQ(V,T) - tDQS osc (V,T)
Offset 1
Temp(T)/Voltage(V)
Figure 68 — Interval oscillator offset OSCoffset
• OSCMatch :
OSCMatch = [tDQS2DQ(V,T) - tDQSOSC(V,T) - OSCoffset ]
• tDQSOSC:
Runtime
tDQSOSC(V,T) =
2 * Count
Page 122
4.26.1 Interval Oscillator matching error (cont’d)
Table 44 — DQS Oscillator Matching Error Specification

Parameter Symbol Min Max Units Notes
DQS Oscillator Matching Error OSCMatch -20 20 ps 1,2,3,4,5,6,7
DQS Oscillator Offset OSCoffset -100 100 ps 2,4,7
NOTE 1 The OSCMatch is the matching error per between the actual DQS and DQS interval oscillator over voltage
and temp.
NOTE 2 This parameter will be characterized or guaranteed by design.
NOTE 3 The OSCMatch is defined as the following:
OSCMatch = [tDQS2DQ(V,T) - tDQSOSC(V,T) - OSCoffset ]
Where tDQS2DQ(V,T) and tDQSOSC(V,T) are determined over the same voltage and temp conditions.
NOTE 4 The runtime of the oscillator must be at least 200ns for determining tDQSOSC(V,T)
Runtime
tDQSOSC(V,T) =
2 * Count
NOTE 5 The input stimulus for tDQS2DQ will be consistent over voltage and temp conditions.
NOTE 6 The OSCoffset is the average difference of the endpoints across voltage and temp.
NOTE 7 These parameters are defined per channel.
Page 123
4.26.2 DQS Interval Oscillator Readout Timing
OSC Stop to its counting value readout timing is shown in Figure 69 and Figure 70.
T0 T1 T2 T3 T4 T5 Ta0 Ta1 Ta2 Ta3 Ta4 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6
CK_c
CK_t
CKE
CS
CA Valid Valid Valid Valid Valid Valid Valid Valid
COMMAND MPC :Start MPC :Stop MRR-1

DES MR Write-2
DQS Oscillator DES DES DES DES DQS Oscillator DES DES DES DES MR18/MR19 CAS-2
tOSCO
NOTES : 1. DQS interval timer run time setting : MR23 OP[7:0] = 00000000 DON'T CARE TIME BREAK
Figure 69 — In case of DQS Interval Oscillator is stopped by MPC Command
T0 T1 T2 T3 T4 T5 Ta0 Ta1 Ta2 Ta3 Ta4 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6
CK_c
CK_t
CKE
CS
CA Valid Valid Valid Valid Valid Valid
COMMAND MPC :Start MRR-1

DES MR Write-2
DQS Oscillator DES DES DES DES DES DES DES DES DES DES MR18/MR19 CAS-2
See Note 2 tOSCO
NOTES : 1. DQS interval timer run time setting : MR23 OP[7:0] 00000000 DON'T CARE TIME BREAK
2. Setting counts of MR23
Figure 70 — In case of DQS Interval Oscillator is stopped by DQS interval timer
Table 45 —DQS Interval Oscillator AC Timing

Parameter Symbol Min/Max Value Units Notes
Delay time from OSC stop to
tOSCO Min Max(40ns,8nCK) ns
Mode Register Readout
Page 124
4.27 READ Preamble Training
LPDDR4 READ Preamble Training is supported through the MPC function.
This mode can be used to train or read level the DQS receivers. Once READ Preamble Training is enabled
by MR13[OP1] = 1, the LPDDR4 DRAM will drive DQS_t LOW, DQS_c HIGH within tSDO and remain at
these levels until an MPC DQ READ Training command is issued.
During READ Preamble Training the DQS preamble provided during normal operation will not be driven by
the DRAM. Once the MPC DQ READ Training command is issued, the DRAM will drive DQS_t/DQD_c like
a normal READ burst after RL. The DRAM may or may not drive DQ[15:0] in this mode.
While in READ Preamble Training Mode, only READ DQ Training commands may be issued.
READ Preamble Training is exited within tSDO after setting MR13[OP1] = 0.
The READ Preamble Training Feature is optional in the 8Gb device but required for other densities.
Enter Read Preamble Training Exit Read Preamble Training
CK_c
CK_t
MRW-2 MPC-1 MRW-2

CMD MRW-1 DESELECT READ Train CAS-2 DESELECT MRW-1
RPT=1 RPT=0
tSDO RL tSDO
DQS_c
DQS_t
Figure 71 — Read Preamble Training
Table 46 — Timing Parameters

Parameter Symbol Min Max
Delay from MRW command to DQS Driven Out tSDO - min(24 nCK, 15ns)
Page 125
4.28 Multi-Purpose Command (MPC)
LPDDR4-SDRAMs use the MPC command to issue a NOP and to access various training modes. The
MPC command is initiated with CS, and CA[5:0] asserted to the proper state at the rising edge of CK, as
defined by Table 63 — Command Truth Table. The MPC command has seven operands (OP[6:0]) that are
decoded to execute specific commands in the SDRAM. OP[6] is a special bit that is decoded on the first
rising CK edge of the MPC command. When OP[6]=0 then the SDRAM executes a NOP (no operation)
command, and when OP[6]=1 then the SDRAM further decodes one of several training commands.
When OP[6]=1 and when the training command includes a Read or Write operation, the MPC command
must be followed immediately by a CAS-2 command. For training commands that Read or Write the
SDRAM, read latency (RL) and write latency (WL) are counted from the second rising CK edge of the
CAS-2 command with the same timing relationship as any normal Read or Write command. The operands
of the CAS-2 command following a MPC Read/Write command must be driven LOW.
The following MPC commands must be followed by a CAS-2 command:
- Write FIFO
- Read FIFO
- Read DQ Calibration
All other MPC-1 commands do not require a CAS-2 command, including:
- NOP
- Start DQS Interval Oscillator
- Stop DQS Interval Oscillator
- Start ZQ Calibration
- Latch ZQ Calibration
Table 47 — MPC Command Definition

SDR Command Pins SDR CA Pins
SDRAM CK_t
CKE Notes
Command CS CA0 CA1 CA2 CA3 CA4 CA5 EDGE
CK_t(n-1) CK_t(n)
MPC H L L L L L OP6 R1
H H 1, 2
(Train, NOP) L OP0 OP1 OP2 OP3 OP4 OP5 R2
Page 126
4.28 Multi-Purpose Command (MPC) (cont’d)
Table 48 — MPC Command Definition for OP[6:0]

0XXXXXXB: NOP
1000001B: RD FIFO
1000011B: RD DQ Calibration (MR32/MR40)
1000101B: RFU
1000111B: WR FIFO
Training Modes OP[6:0] 1001001B: RFU 1, 2
1001011B: Start DQS Osc
1001101B: Stop DQS Osc
1001111B: ZQCal Start
1010001B: ZQCal Latch
NOTE 1 See Table 63 — Command Truth Table for more information.

NOTE 2 MPC commands for Read or Write training operations must be immediately followed by CAS-2 command
consecutively without any other commands in-between. MPC command must be issued first before issuing the CAS-2
command.
t0 t1 t2 t3 t4 ta t a+1 t a+2 t a+3 t a+4 t a+5 tb t b+1 t b+2 t b+3 t b+4 t b+5 tc t c+1 t c+2 t c+3 t c+4 td t d+1 t d+2 te t e+1 t e+
CK_t,
CK_c
tWRWTR tCCD
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALI
WL
WL
CMD WRITE-1 CAS-2 VALID MPC CAS-2 VALID MPC CAS-2 VALID VALID VALID VALID VALID VALID
WR FIFO WR FIFO
tDQSS tDQSS
DQS_t
DQS_c
tWPRE tDQS2DQ tDQS2DQ
DQ[15:0] D0 D1 D2 D3 D8 D9 D10 D11 D12 D13 D14 D15 D0 D1 D12 D13 D14 D15 D0 D1 D12 D13 D14 D15
DMI[1:0]
2. Write-1 to MPC is shown as an example of command-to-command timing for MPC. Timing from Write-1 to MPC [WR-FIFO] is tWRWTR.
3. Seamless MPC [WR-FIFO] commands may be executed by repeating the command every tCCD time.
4. MPC [WR-FIFO] uses the same command-to-data timing relationship (WL, tDQSS, tDQS2DQ) as a Write-1 command.
5. A maximum of 5 MPC [WR-FIFO] commands may be executed consecutively without corrupting FIFO data.
The 6th MPC [WR-FIFO] command will overwrite the FIFO data from the first command. If fewer than 5 MPC [WR-FIFO] commands are executed,
then the remaining FIFO locations will contain undefined data.
6. For the CAS-2 command following a MPC command, the CAS-2 operands must be driven “LOW.”
7. To avoid corrupting the FIFO contents, MPC-1 [RD-FIFO] must immediately follow MPC-1 [WR-FIFO]/CAS-2 without any other command disturbing
FIFO pointers in-between. FIFO pointers are disturbed by CKE Low, Write, Masked Write, Read, Read DQ Calibration and MRR.
See Write Training session for more information on FIFO pointer behavior.
Figure 72 — MPC [WRITE FIFO] Operation: tWPRE=2nCK, tWPST=0.5nCK

Page 127
t0 t1 t2 t3 t4 ta ta+1 ta+2 t a+3 t b t b+1 tb+2 t b+3 tb+4 t b+5 tc+2 t c+3 tc+4 t c+5 tc+6 tc+7 td td+1 t d+2 td+3 td+4 td+5 te te+1 tf tf+1 t f+2 tf+3 t f+4
CK_t,
CK_c
tCCD
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
WL RL
CMD MPC CAS-2 VALID VALID VALID MPC CAS-2 VALID MPC CAS-2 VALID VALID VALID VALID VALID VALID VALID
Write FIFO Read FIFO Read FIFO
tDQSS tWTR tDQSCK

DQS_t
DQS_c
tDQS2DQ tRPRE tRPST
DQ[15:0] D0 D1 D12 D13 D14 D15 D0 D1 D2 D13 D14 D15 D0 D11 D12 D13 D14 D15
DMI[1:0]
2. Write-1 to MPC is shown as an example of command-to-command timing for MPC. Timing from Write-1 to MPC-1 [WR-FIFO] is tWRWTR.
the 1st FIFO and continue advancing. If fewer than 5 MPC [WR-FIFO] commands were executed, then the MPC [RD-FIFO] commands to
those FIFO locations will return undefined data. See the Write Training section for more information on the FIFO pointer behavior.
Figure 73 — MPC [RD FIFO] Read Operation:

tWPRE=2nCK, tWPST=0.5nCK, tRPRE=toggling, tRPST=1.5nCK
t0 t1 t2 t3 t4 ta t a+1 t a+2 t a+3 t a+4 t a+5 tb t b+1 t b+2 t b+3 t b+4 tc t c+1 t c+2 td t d+1 t d+2 t d+3 te t e+1 t e+2 t e+3 t e+4 t e+5
CK_t,
CK_c
tRTRRD
CA[5:0] VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
RL
RL
CMD MPC CAS-2 VALID Read-1 CAS-2 VALID VALID VALID VALID VALID VALID VALID VALID VALID VALID
Read FIFO
tDQSCK tDQSCK
DQS_t
DQS_c
tRPRE tRPST tRPRE tRPST
DQ[15:0] D0 D1 D2 D13 D14 D15 D0 D1 D10 D11 D12 D13 D14 D15
DMI[1:0]
2. MPC [RD-FIFO] to Read-1 Operation is shown as an example of command-to-command timing for MPC. Timing from MPC-1 [RD-FIFO] command
to Read is tRTRRD.
the 1st FIFO and continue advancing. If fewer than 5 MPC [WR-FIFO] commands were executed, then the MPC [RD-FIFO] commands to those
Figure 74 — MPC [RD FIFO] Operation: tRPRE=toggling, tRPST=1.5nCK

Page 128
Table 49 — Timing Constraints for Training Commands

Previous
Next Command Minimum Delay Unit Notes
Command
MPC
tWRWTR nCK 1
[WR FIFO]
MPC
WR/MWR Not Allowed - 2
[RD FIFO]
MPC
WL+RU(tDQSS(max)/tCK) +BL/2+RU(tWTR/tCK) nCK
[RD DQ Calibration]
MPC
tRTW nCK 4
[WR FIFO]
MPC
RD/MRR Not Allowed - 2
[RD FIFO]
MPC
tRTRRD nCK 3
[RD DQ Calibration]
WR/MWR Not Allowed - 2
MPC
tCCD nCK
[WR FIFO]
MPC RD/MRR Not Allowed - 2
[WR FIFO] MPC
WL+RU(tDQSS(max)/tCK) +BL/2+RU(tWTR/tCK) nCK
[RD FIFO]
MPC
Not Allowed - 2
[RD DQ Calibration]
WR/MWR tRTW nCK 4
MPC
tRTW nCK 4
[WR FIFO]
MPC RD/MRR tRTRRD nCK 3
[RD FIFO] MPC
tCCD nCK
[RD FIFO]
MPC
tRTRRD nCK 3
[RD DQ Calibration]
WR/MWR tRTW nCK 4
MPC
tRTW nCK 4
[WR FIFO]
MPC
RD/MRR tRTRRD nCK 3
[RD DQ
MPC
Calibration] Not Allowed - 2
[RD FIFO]
MPC
tCCD nCK
[RD DQ Calibration]
NOTE 1 tWRWTR = WL + BL/2 + RU(tDQSS(max)/tCK) + max(RU(7.5ns/tCK),8nCK)
NOTE 2 No commands are allowed between MPC [WR FIFO] and MPC-1 [RD FIFO] except MRW commands related
to training parameters.
NOTE 3 tRTRRD = RL + RU(tDQSCK(max)/tCK) + BL/2 + RD(tRPST) + max(RU(7.5ns/tCK),8nCK)
NOTE 4 tRTW :
In Case of DQ ODT Disable MR11 OP[2:0] = 000B:
RL+RU(tDQSCK(max)/ tCK) +BL/2- WL+tWPRE+RD(tRPST)
In Case of DQ ODT Enable MR11 OP[2:0] ≠ 000B:
RL + RU(tDQSCK(max)/tCK) + BL/2 + RD(tRPST) - ODTLon - RD(tODTon,min/tCK) + 1
Page 129
4.29 Thermal Offset
Because of their tight thermal coupling with the LPDDR4 device, hot spots on an SOC can induce thermal
gradients across the LPDDR4 device. As these hot spots may not be located near the device thermal
sensor, the devices’ temperature compensated self-refresh circuit may not generate enough refresh cycles
to guarantee memory retention. To address this shortcoming, the controller can provide a thermal offset
that the memory uses to adjust its TCSR circuit to ensure reliable operation.
This offset is provided through MR4(6:5) to either or to both the channels. This temperature offset may
modify refresh behavior for the channel to which the offset is provided. It will take a max of 200us to have
the change reflected in MR4(2:0) for the channel to which the offset is provided. If the induced thermal
gradient from the device temperature sensor location to the hot spot location of the controller is larger than
15 degrees C, then self-refresh mode will not reliably maintain memory contents.
To accurately determine the temperature gradient between the memory thermal sensor and the induced
hot spot, the memory thermal sensor location must be provided to the LPDDR4 memory controller.
Support of thermal offset function is optional. Please refer to vendor datasheet to figure out if the function
is supported or not.
4.30 Temperature Sensor
LPDDR4 devices feature a temperature sensor whose status can be read from MR4. This sensor can be
used to determine an appropriate refresh rate, determine whether AC timing derating is required in the
elevated temperature range, and/or monitor the operating temperature. Either the temperature sensor or
the device TOPER may be used to determine whether operating temperature requirements are being met.
LPDDR4 devices shall monitor device temperature and update MR4 according to tTSI. Upon assertion of
CKE (Low to High transition), the device temperature status bits shall be no older than tTSI. MR4 will be
updated even when device is in self refresh state with CKE HIGH.
When using the temperature sensor, the actual device case temperature may be higher than the TOPER
specificat ion that applies for the standard or elevated temperature ranges. For example, TCASE may be
above 85°C when MR4[2:0] equals ‘b011. LPDDR4 devices shall allow for 2°C temperature margin
between the point at which the device updates the MR4 value and the point at which the controller
reconfigures the system accordingly. In the case of tight thermal coupling of the memory device to external
hot spots, the maximum device temperature might be higher than what is indicated by MR4.
Page 130
4.30 Temperature Sensor (cont’d)
To assure proper operation using the temperature sensor, applications should consider the following
factors:
• TempGradient is the maximum temperature gradient experienced by the memory device at the
temperature of interest over a range of 2 °C.
• ReadInterval is the time period between MR4 reads from the system.
• TempSensorInterval (tTSI) is maximum delay between internal updates of MR4.
• SysRespDelay is the maximum time between a read of MR4 and the response by the system.
In order to determine the required frequency of polling MR4, the system shall use the maximum
TempGradient and the maximum response time of the system using the following equation:
TempGradient x (ReadInterval + tTSI + SysRespDelay) ≤ 2 °C
Table 50 — Temperature Sensor

Parameter Symbol Max/Min Value Unit Notes
System Temperature Gradient TempGradient Max System Dependent °C/s
MR4 Read Interval ReadInterval Max System Dependent ms
Temperature Sensor Interval tTSI Max 32 ms
System Response Delay SysRespDelay Max System Dependent ms
Device Temperature Margin TempMargin Max 2 °C
For example, if TempGradient is 10°C/s and the SysRespDelay is 1 ms:
(10°C/s) x (ReadInterval + 32ms + 1ms) ≤ 2 °C
In this case, ReadInterval shall be no greater than 167 ms.
Temp
< (t TSI + ReadInterval + SysRespDelay)
Device
Temp
Margin
nt
G r ad i e
2oC Tem p
MR4
Trip Level
t TSI
MR4 = 0x03 MR4 = 0x86 MR4 = 0x86 MR4 = 0x86 MR4 = 0x86 MR4 = 0x06
Time
Temperature
Sensor
Update
SysRespDelay
ReadInterval
Host
MR4 Read MRR MR4 = 0x 03 MRR MR4 = 0x 86
Figure 75 — Temp Sensor Timing
Page 131
4.31 ZQ Calibration
The MPC command is used to initiate ZQ Calibration, which calibrates the output driver impedance across
process, temperature, and voltage. ZQ Calibration occurs in the background of device operation, and is
designed to eliminate any need for coordination between channels (i.e. it allows for channel
independence).
There are two ZQ Calibration modes initiated with the MPC command: ZQCal Start, and ZQCal Latch.
ZQCal Start initiates the SDRAM’s calibration procedure, and ZQCal Latch captures the result and loads it
into the SDRAM's drivers.
A ZQCal Start command may be issued anytime the LPDDR4-SDRAM is not in a power-down state. A
ZQCal Latch Command may be issued anytime outside of power-down after tZQCAL has expired and all
DQ bus operations have completed. The CA Bus must maintain a Deselect state during tZQLAT to allow
CA ODT calibration settings to be updated. The following mode register fields that modify I/O parameters
cannot be changed following a ZQCal Start command and before tZQCAL has expired:
• PU-Cal (Pull-up Calibration VOH Point)

• PDDS (Pull Down Drive Strength and Rx Termination)
• DQ-ODT (DQ ODT Value)
• CA-ODT (CA ODT Value)
4.31.1 ZQCal Reset
The ZQCal Reset command resets the output impedance calibration to a default accuracy of +/- 30%
across process, voltage, and temperature. This command is used to ensure output impedance accuracy to
+/- 30% when ZQCal Start and ZQCal Latch commands are not used.
The ZQCal Reset command is executed by writing MR10 OP[0]=1B.
Table 51 — ZQCal Timing Parameters

Parameter Symbol Min/Max Value Unit
ZQ Calibration Time tZQCAL MIN 1 us
ZQ Calibration Latch Time tZQLAT MIN max(30ns,8nCK) ns
ZQ Calibration Reset Time tZQRESET MIN max(50ns,3nCK) ns
T0 T1 T2 T3 T4 T5 T6 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Ta7 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6 Tb7 Tb8 Tb9 Tc0 Tc1 Tc2 Tc3 Tc4 Tc5 Tc6
CK_t,
CK_c
tZQCAL tZQLAT
ZQ ZQ ZQ ZQ
CA CAL CAL WR WR CAS CAS CAL CAL Val Val
Start Start Latch Latch
WL
CMD MPC Write CAS-2 Deselect Deselect Deselect Deselect Deselect Deselect Deselect Deselect MPC Deselect Deselect Precharge Deselect
(Train/Cal) (Train/Cal)
tDQSS
DQS_t
DQS_c tWPRE
tDQS2DQ tWPST
DQ[15:0] VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL VAL
Note DON'T CARE TIME BREAK

1. Write and Precharge operations shown for illustrative purposes.
Any single or multiple valid commands may be executed within the tZQCAL time and prior to latching the results.
2. Before the ZQ-Latch command can be executed, any prior commands utilizing the DQ bus must have completed.
Write commands with DQ Termination must be given enough time to turn off the DQ-ODT before issuing the ZQ-Latch command.
See the ODT section for ODT timing.
Figure 76 — ZQCal Timing
Page 132
4.31.2 Multi-Channel Considerations
The LPDDR4-SDRAM includes a single ZQ pin and associated ZQ Calibration circuitry. Calibration values
from this circuit will be used by both channels according to the following protocol:
1. ZQCal Start commands may be issued to either or both channels.

2. ZQCal Start commands may be issued when either or both channels are executing other commands
and other commands may be issued during tZQCAL.
3. ZQCal Start commands may be issued to both channels simultaneously.
4. The ZQCal Start command will begin the calibration unless a previously requested ZQ calibration is in
progress.
5. If a ZQCal Start command is received while a ZQ calibration is in progress on the SDRAM, the ZQCal
Start command will be ignored and the in-progress calibration will not be interrupted.
6. ZQCal Latch commands are required for each channel.
7. ZQCal Latch commands may be issued to both channels simultaneously.
8. ZQCal Latch commands will latch results of the most recent ZQCal Start command provided tZQCAL
has been met.
9. ZQCal Latch commands which do not meet tZQCAL will latch the results of the most recently
completed ZQ calibration.
10. ZQ Reset MRW commands will only reset the calibration values for the channel issuing the command.
In compliance with complete channel independence, either channel may issue ZQCal Start and ZQCal
Latch commands as needed without regard to the state of the other channel.
4.31.3 ZQ External Resistor, Tolerance, and Capacitive Loading
To use the ZQ calibration function, a 240 ohm +/- 1% tolerance external resistor must be connected
between the ZQ pin and VDDQ.
If the system configuration shares the CA bus to form a x32 (or wider) channel, the ZQ pin of each die’s
x16 channel shall use a separate ZQCal resistor.
If the system configuration has more than one rank, and if the ZQ pins of both ranks are attached to a
single resistor, then the SDRAM controller must ensure that the ZQCal’s don’t overlap.
The total capacitive loading on the ZQ pin must be limited to 25pF.
Example: If a system configuration shares a CA bus between ‘n’ channels to form a n * 16 wide bus, and
no means are available to control the ZQCal separately for each channel (i.e. separate CS, CKE, or CK),
then each x16 channel must have a separate ZQCal resistor.
Example: For a x32, two rank system, each x16 channel must have its own ZQCal resistor, but the ZQCal
resistor can be shared between ranks on each x16 channel. In this configuration, the CS signal can be
used to ensure that the ZQCal commands for Rank[0] and Rank[1] don't overlap.
Page 133
4.32 Pull Up/Pull Down Driver Characteristics and Calibration
Table 52 — Pull-down Driver Characteristics, with ZQ Calibration

RONPD,nom Resistor Min Nom Max Unit
40 Ohm RON40PD 0.9 1 1.1 RZQ/6
48 Ohm RON48PD 0.9 1 1.1 RZQ/5
60 Ohm RON60PD 0.9 1 1.1 RZQ/4
80 Ohm RON80PD 0.9 1 1.1 RZQ/3
120 Ohm RON120PD 0.9 1 1.1 RZQ/2
240 Ohm RON240PD 0.9 1 1.1 RZQ/1
NOTE 1 All value are after ZQ Calibration. Without ZQ Calibration RONPD values are ± 30%.
Table 53 — Pull-Up Characteristics, with ZQ Calibration

VOHPU,nom VOH,nom(mV) Min Nom Max Unit
VDDQ/2.5 440 0.9 1 1.1 VOH,nom
VDDQ/3 367 0.9 1 1.1 VOH,nom
NOTE 1 All values are after ZQ Calibration. Without ZQ Calibration VOH(nom) values are ± 30%.
NOTE 2 VOH,nom (mV) values are based on a nominal VDDQ = 1.1V.
Table 54 — Valid Calibration Points

ODT Value
VOHPU,nom
240 120 80 60 48 40
VDDQ/2.5 VALID VALID VALID DNU DNU DNU
VDDQ/3 VALID VALID VALID VALID VALID VALID
NOTE 1 Once the output is calibrated for a given VOH(nom) calibration point, the ODT value may be changed without
recalibration.
NOTE 2 If the VOH(nom) calibration point is changed, then re-calibration is required.
NOTE 3 DNU = Do Not Use
Page 134
4.33 On Die Termination (ODT) for Command/Address Bus
ODT (On-Die Termination) is a feature of the LPDDR4 SDRAM that allows the SDRAM to turn on/off
termination resistance for CK_t, CK_c, CS and CA[5:0] signals without the ODT control pin.
The ODT feature is designed to improve signal integrity of the memory channel by allowing the DRAM
controller to turn on and off termination resistance for any target DRAM devices via Mode Register setting.
A simple functional representation of the DRAM ODT feature is shown in Figure 77.
To other CK_t, CK_c, CS and CA[5:0]

circuitry RTT
like RCV, ...
Switch
VSS
Figure 77 — Functional Representation of CA ODT
4.33.1 ODT Mode Register and ODT State Table
ODT termination values are set and enabled via MR11. The CA bus (CK_t, CK_C, CS, CA[5:0]) ODT
resistance values are set by MR11 OP[6:4]. The default state for the CA is ODT disabled.
ODT is applied on the CA bus to the CK_t, CK_c, CS and CA[5:0] signals. The CA ODT of the device is
designed to enable one rank to terminate the entire command bus in a multirank system, so only one
termination load will be present even if multiple devices are sharing the command signals. For this reason,
CA ODT remains on even when the device is in the power-down or self-refresh power-down states.
The die has a bond-pad (ODT_CA) for multirank operations. When the ODT_CA pad is LOW, the die will
not terminate the CA bus regardless of the state of the mode register CA ODT bits (MR11 OP[6:4]). If,
however, the ODT_CA bond-pad is HIGH, and the mode register CA ODT bits are enabled, the die will
terminate the CA bus with the ODT values found in MR11 OP[6:4]. In a multirank system, the terminating
rank should be trained first, followed by the non-terminating rank(s).
Table 55 — Command Bus ODT State

ODTE-CA ODT_CA ODTD-CA ODTF-CK ODTF-CS ODT State ODT State ODT State
MR11[6:4] bond pad MR22[5] MR22[3] MR22[4] for CA for CK_t/CK_c for CS
Disabled1 Valid2 Valid3 Valid3 Valid3 Off Off Off
Valid3 0 Valid3 0 0 Off Off Off
Valid3 0 Valid3 0 1 Off Off On
Valid3 0 Valid3 1 0 Off On Off
Valid3 0 Valid3 1 1 Off On On
Valid3 1 0 Valid3 Valid3 On On On
Valid3 1 1 Valid3 Valid3 Off On On
NOTE 1 Default Value
NOTE 2 “Valid” means “H or L (but a defined logic level)”
NOTE 3 “Valid” means “0 or 1”
NOTE 4 The state of ODT_CA is not changed when the DRAM enters power-down mode. This maintains termination
for alternate ranks in multi-rank systems.
Page 135
4.33.2 ODT Mode Register and ODT Characteristics
Chip In Termination Mode
Vout
RTT = I out VDD2
To other
circuitry ODT
like RCV, ...
CA
IOUT
RTT VOUT
VSS
Figure 78 — On Die Termination for CA

Page 136
4.33.2 ODT Mode Register and ODT Characteristics (cont’d)
Table 56 — ODT DC Electrical Characteristics, assuming RZQ = 240Ω +/-1% over the entire
operating temperature range after a proper ZQ calibration
MR11[6:4] RTT Vout Min Nom Max Unit Note
VOLdc= 0.1 * VDD2 0.8 1.0 1.1 RZQ 1,2,3
001 240Ω VOMdc= 0.33 * VDD2 0.9 1.0 1.1 RZQ 1,2,3
VOHdc= 0.5 * VDD2 0.9 1.0 1.2 RZQ 1,2,3
VOLdc= 0.1 * VDD2 0.8 1.0 1.1 RZQ/2 1,2,3
010 120Ω VOMdc= 0.33 * VDD2 0.9 1.0 1.1 RZQ/2 1,2,3
VOHdc= 0.5 * VDD2 0.9 1.0 1.2 RZQ/2 1,2,3
VOLdc= 0.1 * VDD2 0.8 1.0 1.1 RZQ/3 1,2,3
011 80Ω VOMdc= 0.33 * VDD2 0.9 1.0 1.1 RZQ/3 1,2,3
VOHdc= 0.5 * VDD2 0.9 1.0 1.2 RZQ/3 1,2,3
VOLdc= 0.1 * VDD2 0.8 1.0 1.1 RZQ/4 1,2,3
100 60Ω VOMdc= 0.33 * VDD2 0.9 1.0 1.1 RZQ/4 1,2,3
VOHdc= 0.5 * VDD2 0.9 1.0 1.2 RZQ/4 1,2,3
VOLdc= 0.1 * VDD2 0.8 1.0 1.1 RZQ/5 1,2,3
101 48Ω VOMdc= 0.33 * VDD2 0.9 1.0 1.1 RZQ/5 1,2,3
VOHdc= 0.5 * VDD2 0.9 1.0 1.2 RZQ/5 1,2,3
VOLdc= 0.1 * VDD2 0.8 1.0 1.1 RZQ/6 1,2,3
110 40Ω VOMdc= 0.33 * VDD2 0.9 1.0 1.1 RZQ/6 1,2,3
VOHdc= 0.5 * VDD2 0.9 1.0 1.2 RZQ/6 1,2,3
Mismatch CA-CA
0.33* VDD2 - TBD % 1,2,4
within clk group
NOTE 1 The tolerance limits are specified after calibration with stable voltage and temperature. For the behavior of
the tolerance limits if temperature or voltage changes after calibration, see “voltage and temperature sensitivity”1.
NOTE 2 Pull-dn ODT resistors are recommended to be calibrated at 0.33*VDD2. Other calibration schemes may be
used to achieve the linearity spec shown above, e.g., calibration at 0.5*VDD2 and 0.1*VDD2.
NOTE 3 Measurement definition for RTT: TBD1

NOTE 4 CA to CA mismatch within clock group (CA,CS) variation for a given component including CK_t and CK_c
(characterized).
RODT(max) - RODT(min)
CA - CA Mismatch =
RODT(avg)

Page 137
4.33.3 ODT for Command/Address Update Time

ODT for Command/Address update time after Mode Register set are shown in Figure 79.
T0 T1 T2 T3 T4 T5 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Ta7 Ta8
CK_c
CK_t
CKE
CS
COMMAND DES MRW1 MRW1 MRW2 MRW2 DES DES DES DES DES VALID1 VALID1 VALID1 VALID1 VALID
ADDRESS VALID VALID VALID VALID VALID VALID VALID VALID VALID
CA ODT Old Setting Value Updating Setting New Setting Value
tODTUP
NOTE 1 : 4 Clock Cycle Command TIME BREAK DON'T CARE
Figure 79 — ODT for Command/Address setting update timing in 4 Clock Cycle Command
Table 57 — ODT CA AC Timing

Speed LPDDR4-1600/1866/2133/2400/3200/4266
Units NOTE
Parameter Symbol MIN MAX
ODT CA Value Update Time tODTUP RU(TBDns/tCK(avg)) -
4.34 On-Die Termination (ODT)
ODT (On-Die Termination) is a feature of the LPDDR4 SDRAM that allows the DRAM to turn on/off
termination resistance for each DQ, DQS_t, DQS_c and DMI signals without the ODT control pin. The
ODT feature is designed to improve signal integrity of the memory channel by allowing the DRAM
controller to turn on and off termination resistance for any target DRAM devices during Write or Mask Write
operation.
The ODT feature is off and cannot be supported in Power Down and Self-Refresh modes.
A simple functional representation of the DRAM ODT feature is shown in Figure 80.
To other DQ, DQS, DMI

circuitry RTT
like RCV, ...
Switch
VSSQ
Figure 80 — Functional Representation of ODT
The switch is enabled by the internal ODT control logic, which uses the Write-1 or Mask Write-1 command
and other mode register control information. The value of RTT is determined by the settings of Mode
Register bits.
Page 138
4.34.1 ODT Mode Register
The ODT Mode is enabled if MR11 OP[3:0] are non zero. In this case, the value of RTT is determined by
the settings of those bits. The ODT Mode is disabled if MR11 OP[3] = 0.
4.34.2 Asynchronous ODT
When ODT Mode is enabled in MR11 OP[3:0], DRAM ODT is always Hi-Z. DRAM ODT feature is
automatically turned ON asynchronously based on the Write-1 or Mask Write-1 command that DRAM
samples. After the write burst is complete, DRAM ODT featured is automatically turned OFF
asynchronously.
Following timing parameters apply when DRAM ODT mode is enabled:
• ODTLon, tODTon,min, tODTon,max

• ODTLoff, tODToff,min, tODToff,max
ODTLon is a synchronous parameter and it is the latency from CAS-2 command to tODTon reference.
ODTLon latency is a fixed latency value for each speed bin. Each speed bin has a different ODTLon
latency.
Minimum RTT turn-on time (tODTon,min) is the point in time when the device termination circuit leaves
high impedance state and ODT resistance begins to turn on.
Maximum RTT turn on time (tODTon,max) is the point in time when the ODT resistance is fully on.
tODTon,min and tODTon,max are measured once ODTLon latency is satisfied from CAS-2 command.
ODTLoff is a synchronous parameter and it is the latency from CAS-2 command to tODToff reference.
ODTLoff latency is a fixed latency value for each speed bin. Each speed bin has a different ODTLoff
latency.
Minimum RTT turn-off time (tODToff,min) is the point in time when the device termination circuit starts to
turn off the ODT resistance.
Maximum ODT turn off time (tODToff,max) is the point in time when the on-die termination has reached
high impedance.
tODToff,min and tODToff,max are measured once ODTLoff latency is satisfied from CAS-2 command.
Page 139
4.34.2 Asynchronous ODT (cont’d)
Table 58 — ODTLon and ODTLoff Latency Values

ODTLon Latency1 Lower Clock Upper Clock
ODTLoff Latency2
tWPRE = 2 tCK Frequency Limit Frequency Limit
WL Set "A" WL Set "B" WL Set "A" WL Set "B" (>) (≤)
N/A N/A N/A N/A 10 266
N/A N/A N/A N/A 266 533
N/A 6 N/A 22 533 800
4 12 20 28 800 1066
4 14 22 32 1066 1333
6 18 24 36 1333 1600
6 20 26 40 1600 1866
8 24 28 44 1866 2133
nCK nCK nCK nCK MHz MHz
NOTE 1 ODTLon is referenced from CAS-2 command. See Figure 81.
NOTE 2 ODTLoff as shown assumes BL=16. For BL32, 8 tCK should be added.
Table 59 — Asynchronous ODT Turn On and Turn Off Timing

Parameter 800 - 2133 MHz Unit
tODTon, min 1.5 ns
tODTon, max 3.5 ns
tODToff, min 1.5 ns
tODToff, max 3.5 ns
Page 140
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Ta7 Ta8 Ta9 Ta10 Ta11 Ta12 Ta13 Ta14 Ta15 Ta16 Ta17 Ta18 Ta19 Ta20 Ta21
CK_c
CK_t
CS
CA BA0,
BL CA, AP CA CA
COMMAND Write-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES DES
WL
tDQSS (Min)
tWPRE tWPST
DQS_c
DQS_t
tDQS2DQ
DQ Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din
n0 n1 n2 n3 n4 n5 n6 n7 n8 n9 n10 n11 n12 n13 n14 n15
tDQSS (Max)
tWPRE tWPST
DQS_c
DQS_t
tDQS2DQ
DQ Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din Din
n0 n1 n2 n3 n4 n5 n6 n7 n8 n9 n10 n11 n12 n13 n14 n15
ODTLon tODTon.Max
tODTon.Min
DRAM RTT ODT Hi-Z Transition ODT On Transition ODT Hi-Z

ODTLoff
tODToff.Min
Note
tODToff.Max
1. BL=16, Write Postamble = 0.5nCK, DQ/DQS: VSSQ termination
3. DES commands are shown for ease of illustration; other commands may be valid at these times. DON'T CARE TIME BREAK
Figure 81 — Asynchronous ODTon/ODToff Timing

Page 141
4.34.3 ODT during Write Leveling
If ODT is enabled in MR11 OP[3:0], in Write Leveling mode, DRAM always provides the termination on
DQS_t/DQS_c signals. DQ termination is always off in Write Leveling mode regardless.
Table 60 — DRAM Termination Function in Write Leveling Mode

ODT Enabled in MR11 DQS_t/DQS_c termination DQ termination
Disabled OFF OFF
Enabled ON OFF
4.35 On-Die Termination (ODT) for DQ, DQS, and DMI
On-Die Termination (ODT) effective resistance RTT is defined by MR11 OP[2:0].
ODT is applied to the DQ, DMI, DQS_t and DQS_c pins.
A functional representation of the on-die termination is shown in Figure 82.
Chip In Termination Mode

Vout
RTT = I out
VDDQ
To other
circuitry ODT
like RCV, ...
DQ
IOUT
RTT VOUT
VSSQ
Figure 82 — On Die
Terminiation
Page 142
4.35 On-Die Termination (ODT) for DQ, DQS, and DMI (cont’d)
Table 61 — ODT DC Electrical Characteristics, assuming RZQ = 240Ω +/-1% over the entire
operating temperature range after a proper ZQ calibration.
MR11
RTT Vout Min Nom Max Unit Notes
OP[2:0]
VOLdc= 0.1* VDDQ 0.8 1 1.1 RZQ 1,2,3
001 240Ω VOMdc= 0.33* V DDQ 0.9 1 1.1 RZQ 1,2,3
VOHdc= 0.5* VDDQ 0.9 1 1.2 RZQ 1,2,3
VOLdc= 0.1* VDDQ 0.8 1 1.1 RZQ/2 1,2,3
010 120Ω VOMdc= 0.33* VDDQ 0.9 1 1.1 RZQ/2 1,2,3
VOHdc= 0.5* VDDQ 0.9 1 1.2 RZQ/2 1,2,3
VOLdc= 0.1* VDDQ 0.8 1 1.1 RZQ/3 1,2,3
011 80Ω VOMdc= 0.33* VDDQ 0.9 1 1.1 RZQ/3 1,2,3
VOHdc= 0.5* VDDQ 0.9 1 1.2 RZQ/3 1,2,3
VOLdc= 0.1* VDDQ 0.8 1 1.1 RZQ/4 1,2,3
100 60Ω VOMdc= 0.33* VDDQ 0.9 1 1.1 RZQ/4 1,2,3
VOHdc= 0.5* VDDQ 0.9 1 1.2 RZQ/4 1,2,3
VOLdc= 0.1* VDDQ 0.8 1 1.1 RZQ/5 1,2,3
101 48Ω VOMdc= 0.33* VDDQ 0.9 1 1.1 RZQ/5 1,2,3
VOHdc= 0.5* VDDQ 0.9 1 1.2 RZQ/5 1,2,3
VOLdc= 0.1* VDDQ 0.8 1 1.1 RZQ/6 1,2,3
110 40Ω VOMdc= 0.33* VDDQ 0.9 1 1.1 RZQ/6 1,2,3
VOHdc= 0.5* VDDQ 0.9 1 1.2 RZQ/6 1,2,3
Mismatch DQ-DQ
0.33* VDDQ - 2 % 1,2,4
within byte
NOTE 1 The tolerance limits are specified after calibration with stable voltage and temperature. For the behavior of
the tolerance limits if temperature or voltage changes after calibration, see “voltage and temperature sensitivity”1.
NOTE 2 Pull-dn ODT resistors are recommended to be calibrated at 0.33*VDDQ. Other calibration schemes may be
used to achieve the linearity spec shown above, e.g., calibration at 0.5*VDDQ and 0.1*VDDQ.
NOTE 3 Measurement definition for RTT: TBD1

NOTE 4 DQ to DQ mismatch within byte variation for a given component including DQS_t and DQS_c
(characterized).
RODT(max) - RODT(min)
DQ - DQ Mismatch =
RODT(avg)

Page 143
4.36 Power-Down Mode
4.36.1 Power-Down Entry and Exit
Power-down is asynchronously entered when CKE is driven LOW. CKE must not go LOW while the
following operations are in progress:
• Mode Register Read

• Mode Register Write
• Read
• Write
• VREF(CA) Range and Value setting via MRW
• VREF(DQ) Range and Value setting via MRW
• Command Bus Training mode Entering/Exiting via MRW
• VRCG High Current mode Entering/Exiting via MRW
And the LPDDR4 DRAM cannot be placed in power-down state during “Start DQS Interval Oscillator”
operation.
CKE can go LOW while any other operations such as row activation, Precharge, Auto Precharge, or
Refresh are in progress. The power-down IDD specification will not be applied until such operations are
complete. Power-down entry and exit are shown in Figure 83.
Entering power-down deactivates the input and output buffers, excluding CKE and Reset_n. To ensure that
there is enough time to account for internal delay on the CKE signal path, CS input is required stable Low
level and CA input level is don’t care after CKE is driven LOW, this timing period is defined as tCKELCS.
Clock input is required after CKE is driven LOW, this timing period is defined as tCKELCK. CKE LOW will
result in deactivation of all input receivers except Reset_n after tCKELCK has expired. In power-down
mode, CKE must be held LOW; all other input signals except Reset_n are "Don't Care". CKE LOW must be
maintained until tCKE,min is satisfied.
VDDQ can be turned off during power-down. Prior to exiting power-down, VDDQ must be within its
minimum/maximum operating range.
No refresh operations are performed in power-down mode except Self-Refresh power-down. The
maximum duration in non-Self-Refresh power-down mode is only limited by the refresh requirements
outlined in 4.12, Refresh command.
The power-down state is asynchronously exited when CKE is driven HIGH. CKE HIGH must be maintained
until tCKE,min is satisfied. A valid, executable command can be applied with power-down exit latency tXP
after CKE goes HIGH. Power-down exit latency is defined in Table 62.
Clock frequency change or Clock Stop is inhibited during tCMDCKE, tCKELCK, tCKCKEH, tXP,
tMRWCKEL and tZQCKE periods.
If power-down occurs when all banks are idle, this mode is referred to as idle power-down. if power-down
occurs when there is a row active in any bank, this mode is referred to as active power-down. And If
power-down occurs when Self Refresh is in progress, this mode is referred to as Self Refresh power-down
in which the internal refresh is continuing in the same way as Self Refresh mode.
When CA, CK and/or CS ODT is enabled via MR11 OP[6:4] and also via MR22 or CA-ODT pad setting, the
rank providing ODT will continue to terminate the command bus in all DRAM states including power-down.
Page 144
4.36.1 Power-Down Entry and Exit (cont’d)
T0 T1 Ta0 Tb0 Tb1 Tc0 Tc1 Td0 Te0 Te1 Tf0 Tf1 Tg0 Th0 Th1 Th2 Th3 Tk0 Tk1 Tk2
CK_c
*1
CK_t
tCKE tCKE
tCMDCKE tCKELCK tCKCKEH tXP
CKE
tCSCKE tCKELCS tCSCKEH tCKEHCS
CS
COMMAND Valid DES DES DES Valid DES DES

NOTES : 1. Input clock frequency can be changed or the input clock can be stopped or floated during power-down,
provided that upon exiting power-down, the clock is stable and within specified limits for a minimum of
tCKCKEH of stable clock prior to power-down exit and the clock frequency is between the minimum and
maximum specified frequency for the speed grade in use.
Figure 83 — Basic Power-Down Entry and Exit Timing
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Tb0 Tb1 Tb2 Tb3 Tc0 Tc1 Tc2 Tc3 Tc4 Td0 Td1
CK_c
CK_t
CKE
See Note 2
CS
COMMAND Read-1 CAS-2 DES DES DES DES DES DES DES DES DES DES
RL tDQSCK
DQS_c
DQS_t
tRPST
tRPRE
Dout Dout Dout Dout Dout Dout
DQ n0 n1 n2 n13 n14 n15

NOTES : 1. CKE must be held HIGH until the end of the burst operation.
2. Minimum Delay time from Read Command or Read with Auto Precharge Command to falling edge of CKE signal is as follows.
Read Post-amble = 0.5nCK : MR1 OP[7]=[0] : (RL x tCK) + tDQSCK(Max) + ((BL/2) x tCK) + 1tCK
Figure 84 — Read and Read with Auto Precharge to Power-Down Entry

Page 145

T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Tb0 Tb1 Tb2 Tc0 Tc1 Tc2 Tc3 Td0 Td1 Td2
CK_c
CK_t
CKE
See Note 2
CS
COMMAND Write-1
Mask Write-1 CAS-2 DES DES DES DES DES DES DES DES DES
WL tDQSS tWPST
tWPRE tDQS2DQ BL/2 tWR
Din Din Din Din Din Din

n0 n1 n2 n13 n14 n15

2. Minimum Delay time from Write Command or Mask Write Command to falling edge of CKE signal is as follows.
(WL x tCK) + tDQSS(Max) + tDQS2DQ(Max) + ((BL/2) x tCK) + tWR
3. This timing is applied regardless of DQ ODT Disable/Enable setting: MR11[OP2:0].
4. This timing diagram only applies to the Write and Mask Write Commands without Auto Precharge.
Figure 85 — Write and Mask Write to Power-Down Entry
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Tb0 Tb1 Tb2 Tc0 Tc1 Tc2 Tc3 Tc4 Tc5 Td0
CK_c
CK_t
CKE
See Note 2
CS
COMMAND Write-1
Mask Write-1 CAS-2 DES DES DES DES DES DES DES DES DES DES DES DES
WL tDQSS tWPST *3
Precharge
tWPRE tDQS2DQ BL/2 nWR
Din Din Din Din Din Din

n0 n1 n2 n13 n14 n15

2. Delay time from Write with Auto Precharge Command or Mask Write with Auto Precharge Command to falling edge of CKE signal is more than
(WL x tCK) + tDQSS(Max) + tDQS2DQ(Max) + ((BL/2) x tCK) + (nWR x tCK) + (2 x tCK)
3. Internal Precharge Command
4. This timing is applied regardless of DQ ODT Disable/Enable setting: MR11[OP2:0].
Figure 86 — Write with Auto Precharge and Mask Write with Auto Precharge to Power-Down Entry
Page 146
T0 T1 T2 Ta0 Ta1 Ta2 Ta3 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Ta7 Ta8 Ta9 Ta10
CK_c
CK_t
CKE tCMDCKE
CS
CA Valid Valid
COMMAND Refresh DES DES
NOTES : 1. CKE must be held HIGH until tCMDCKE is satisfied.

Figure 87 — Refresh entry to Power-Down Entry
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Tb0 Tb1 Tb2 Tb4 Tb5 Tb6 Tb7 Tb8 Tb9
CK_c
CK_t
CKE
tCMDCKE
CS
COMMAND Activate-1 Activate-2 DES DES

Figure 88 — Activate Command to Power-Down Entry

Page 147
T0 T1 T2 Ta0 Ta1 Ta2 Ta3 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6 Tb7 Tb8 Tb9 Tb10 Tb11
CK_c
CK_t
CKE tCMDCKE
CS
CA Valid Valid
COMMAND Precharge DES DES

Figure 89 — Precharge Command to Power-Down Entry
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Tb0 Tb1 Tb2 Tb3 Tc0 Tc1 Tc2 Tc3 Tc4 Td0 Td1
CK_c
CK_t
CKE
See Note 2
CS
COMMAND MR Read-1 CAS-2 DES DES DES DES DES DES DES DES DES DES
RL tDQSCK
DQS_c
DQS_t
tRPST
tRPRE
Dout Dout Dout Dout Dout Dout
DQ n0 n1 n2 n13 n14 n15

2. Minimum Delay time from Mode Register Read Command to falling edge of CKE signal is as follows:
Figure 90 — Mode Register Read to Power-Down Entry

Page 148
T0 T1 T2 T3 T4 Ta0 Ta1 Ta2 Ta3 Ta4 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6 Tb7
CK_c
CK_t
CKE tMRWCKEL
CS
COMMAND MR Write-1 MR Write-2 DES DES DES
NOTES : 1. CKE must be held HIGH until tMRWCKEL is satisfied.

2. This timing is the general definition for Power Down Entry after Mode Register Write Command.
When a Mode Register Write Command changes a parameter or starts an operation that requires special timing
longer than tMRWCKEL, that timing must be satisfied before CKE is driven low.
Changing the Vref(DQ) value is one example, in this case the appropriate Vref_time-Short/Middle/Long must be satisfied.
Figure 91 — Mode Register Write to Power-Down Entry
T0 T1 T2 Ta0 Ta1 Ta2 Tb0 Tb1 Tb2 Tb3 Tb4 Tb5 Tb6 Tb7 Tb8 Tb9 Tb10 Tb11
CK_c
CK_t
CKE
tZQCKE
CS
CA Valid Valid
COMMAND MPC
Start ZQ Cal DES DES
ZQ Cal Status ZQ Calibration progresses
tZQCAL
NOTES : 1. ZQ Calibration continues if CKE goes low after tZQCKE is satisfied. DON'T CARE TIME BREAK
Figure 92 — Multi purpose Command for Start ZQ Calibration to Power-Down Entry

Page 149
Table 62 — Power-Down AC Timing
Min/
Parameter Symbol Data Rate Unit Note
Max
Power Down Timing

CKE minimum pulse width Max(7.5ns,
tCKE Min -
(HIGH and LOW pulse width) 4nCK)
Max(1.75ns,
Delay from valid command to CKE input LOW tCMDCKE Min ns 1
3nCK)
Max(5ns,
Valid Clock Requirement after CKE Input low tCKELCK Min ns 1
5nCK)
Valid CS Requirement before CKE Input Low tCSCKE Min 1.75 ns
Max(5ns,
Valid CS Requirement after CKE Input low tCKELCS Min ns
5nCK)
Max(1.75ns,
Valid Clock Requirement before CKE Input High tCKCKEH Min ns 1
3nCK)
Max(7.5ns,
Exit power- down to next valid command delay tXP Min ns 1
5nCK)
Valid CS Requirement before CKE Input High tCSCKEH Min 1.75 ns
Max(7.5ns,
Valid CS Requirement after CKE Input High tCKEHCS Min ns
5nCK)
Valid Clock and CS Requirement after CKE Input low Max(14ns,
tMRWCKEL Min ns 1
after MRW Command 10nCK)
Valid Clock and CS Requirement after CKE Input low Max(1.75ns,
tZQCKE Min ns 1
after ZQ Calibration Start Command 3nCK)
NOTE 1 Delay time has to satisfy both analog time(ns) and clock count(nCK).
For example, tCMDCKE will not expire until CK has toggled through at least 3 full cycles (3 *tCK) and
1.75ns has transpired.
The case which 3nCK is applied to is shown in Figure 93.

T-1 T0 T1 T2 T3 T4
CK_c
CK_t
tCMDCKE
CKE
CS
CA Valid Valid
COMMAND Valid DES
DON'T CARE
Figure 93 — tCMDCKE Timing

Page 150
4.37 Input Clock Stop and Frequency Change
LPDDR4 SDRAMs support input clock frequency change during CKE LOW under the following conditions:
• tCK(abs)min is met for each clock cycle;

• Refresh requirements apply during clock frequency change;
• During clock frequency change, only REFab or REFpb commands may be executing;
• Any Activate or Precharge commands have executed to completion prior to changing the frequency;
• The related timing conditions (tRCD, tRP) have been met prior to changing the frequency;
• The initial clock frequency shall be maintained for a minimum of 4 clock cycles after CKE goes LOW;
• The clock satisfies tCH(abs) and tCL(abs) for a minimum of 2 clock cycles prior to CKE going HIGH
After the input clock frequency is changed and CKE is held HIGH, additional MRW commands may be
required to set the WR, RL etc. These settings may need to be adjusted to meet minimum timing
requirements at the target clock frequency.
LPDDR4 devices support clock stop during CKE LOW under the following conditions:
• CK_t is held LOW and CK_c is held HIGH or both are floated during clock stop;
• Refresh requirements apply during clock stop;
• During clock stop, only REFab or REFpb commands may be executing;
• Any Activate or Precharge commands have executed to completion prior to stopping the clock;
• The related timing conditions (tRCD, tRP) have been met prior to stopping the clock;
• The initial clock frequency shall be maintained for a minimum of 4 clock cycles after CKE goes LOW;
• The clock satisfies tCH(abs) and tCL(abs) for a minimum of 2 clock cycles prior to CKE going HIGH
LPDDR4 devices support input clock frequency change during CKE HIGH under the following conditions:
• tCK(abs)min is met for each clock cycle;

• Refresh requirements apply during clock frequency change;
• Any Activate, Read, Write, Precharge, Mode Register Write, or Mode Register Read commands must
have executed to completion, including any associated data bursts prior to changing the frequency;
• The related timing conditions (tRCD, tWR, tWRA, tRP, tMRW, tMRR, etc.) have been met prior to
changing the frequency;
• CS shall be held LOW during clock frequency change;
• During clock frequency change, only REFab or REFpb commands may be executing;
• The LPDDR4 SDRAM is ready for normal operation after the clock satisfies tCH(abs) and tCL(abs) for a
minimum of 2*tCK+tXP.
After the input clock frequency is changed, additional MRW commands may be required to set the WR, RL
etc. These settings may need to be adjusted to meet minimum timing requirements at the target clock
frequency.
LPDDR4 devices support clock stop during CKE HIGH under the following conditions:
• CK_t is held LOW and CK_c is held HIGH during clock stop;
• CS shall be held LOW during clock clock stop;
• Refresh requirements apply during clock stop;
• During clock stop, only REFab or REFpb commands may be executing;
• Any Activate, Read, Write, Precharge, Mode Register Write, or Mode Register Read commands must
have executed to completion, including any associated data bursts prior to stopping the clock;
•The related timing conditions (tRCD, tWR, tWRA, tRP, tMRW, tMRR, etc.) have been met prior to stopping
the clock;
•The LPDDR4 SDRAM is ready for normal operation after the clock is restarted and satisfies tCH(abs) and
tCL(abs) for a minimum of 2*tCK+tXP.
Page 151
4.38 Truth Tables
Operation or timing that is not specified is illegal, and after such an event, in order to guarantee proper
operation, the LPDDR4 device must be reset or power-cycled and then restarted through the specified
initialization sequence before normal operation can continue.
CKE signal has to be held High when the commands listed in Table 63 — Command Truth Table input.
4.38.1 Command Truth Table
Table 63 — Command Truth Table

SDR
Command SDR CA Pins (6)
Pins
CK_t
SDRAM Command CS CA0 CA1 CA2 CA3 CA4 CA5 Notes
edge
Deselect (DES) L X R1 1,2
Multi Purpose Command H L L L L L OP6 R1
1,2,9
(MPC) L OP0 OP1 OP2 OP3 OP4 OP5 R2
Precharge (PRE) H L L L L H AB R1
1,2,3,4
(Per Bank, All Bank) L BA0 BA1 BA2 V V V R2
Refresh (REF) H L L L H L AB R1
1,2,3,4
(Per Bank, All Bank) L BA0 BA1 BA2 V V V R2
Self Refresh Entry H L L L H H V R1
1,2
(SRE) L V R2
H L L H L L BL R1
Write -1 (WR-1) 1,2,3,6,7,9
L BA0 BA1 BA2 V C9 AP R2
H L L H L H V R1
Self Refresh Exit (SRX) 1,2
L V R2
Mask Write -1 H L L H H L L R1
1,2,3,5,6,9
(MWR-1) L BA0 BA1 BA2 V C9 AP R2
H L L H H H V R1
RFU 1,2
L V R2
H L H L L L BL R1
Read -1 (RD-1) 1,2,3,6,7,9
L BA0 BA1 BA2 V C9 AP R2
CAS-2 (Write-2, Mask H L H L L H C8 R1

Write -2, Read-2, MRR- 1,8,9
2, MPC) L C2 C3 C4 C5 C6 C7 R2
H L H L H L V R1
RFU 1,2
L V R2
H L H L H H V R1
RFU 1,2
L V R2
Mode Register Write -1 H L H H L L OP7 R1
1,2,11
(MRW-1) L MA0 MA1 MA2 MA3 MA4 MA5 R2
Mode Register Write-2 H L H H L H OP6 R1
1,2,11
(MRW-2) L OP0 OP1 OP2 OP3 OP4 OP5 R2
Mode Register Read-1 H L H H H L V R1
1,2,12
(MRR-1) L MA0 MA1 MA2 MA3 MA4 MA5 R2
Page 152
4.38.1 Command Truth Table (cont’d)
Table 63 — Command Truth Table (cont’d)

SDR
Command SDR CA Pins (6)
Pins
CK_t
SDRAM Command CS CA0 CA1 CA2 CA3 CA4 CA5 Notes
edge
H L H H H H V R1
RFU 1,2
L V R2
H H L R12 R13 R14 R15 R1
Activate -1 (ACT-1) 1,2,3,10
L BA0 BA1 BA2 V R10 R11 R2
H H H R6 R7 R8 R9 R1
Activate -2 (ACT-2) 1,10
L R0 R1 R2 R3 R4 R5 R2
NOTE 1 All LPDDR4 commands except for Deselect are 2 clock cycle long and defined by states of CS and CA[5:0]
at the first rising edge of clock. Deselect command is 1 clock cycle long.
NOTE 2 "V" means "H" or "L" (a defined logic level). "X" means don’t care in which case CS, CK_t, CK_c and CA[5:0]
can be floated.
NOTE 3 Bank addresses BA[2:0] determine which bank is to be operated upon.
NOTE 4 AB "HIGH" during Precharge or Refresh command indicates that command must be applied to all banks and
bank address is a don’t care.
NOTE 5 Mask Write-1 command supports only BL 16. For Mark Write-1 comamnd, CA5 must be driven LOW on first
rising clock cycle (R1).
NOTE 6 AP "HIGH" during Write-1, Mask Write-1 or Read-1 commands indicates that an auto-precharge will occur to
the bank associated with the Write, Mask Write or Read command.
NOTE 7 If Burst Length on-the-fly is enabled, BL "HIGH" during Write-1 or Read-1 command indicates that Burst
Length should be set on-the-Fly to BL=32. BL "LOW" during Write-1 or Read-1 command indicates that Burst Length
should be set on-the-fly to BL=16. If Burst Length on-the-fly is disabled, then BL must be driven to defined logic level
"H" or "L".
NOTE 8 For CAS-2 commands (Write-2 or Mask Write-2 or Read-2 or MRR-2 or MPC (Only Write FIFO, Read FIFO
and Read DQ Calibration), C[1:0] are not transmitted on the CA[5:0] bus and are assumed to be zero. Note that for
CAS-2 Write-2 or CAS-2 Mask Write-2 command, C[3:2] must be driven LOW.
NOTE 9 Write-1 or Mask Write-1 or Read-1 or Mode Register Read-1 or MPC (Only Write FIFO, Read FIFO and
Read DQ Calibration) command must be immediately followed by CAS-2 command consecutively without any other
command in between. Write-1 or Mask Write-1 or Read-1 or Mode Register Read-1 or MPC (Only Write FIFO, Read
FIFO & Read DQ Calibration) command must be issued first before issuing CAS-2 command. MPC (Only Start & Stop
DQS Oscillator, Start & Latch ZQ Calibration) commands do not require CAS-2 command; they require two additional
DES or NOP commands consecutively before issuing any other commands.
NOTE 10 Activate-1 command must be immediately followed by Activate-2 command consecutively without any other
command in between. Activate-1 command must be issued first before issuing Activate-2 command. Once Activate-1
command is issued, Activate-2 command must be issued before issuing another Activate-1 command.
NOTE 11 MRW-1 command must be immediately followed by MRW-2 command consecutively without any other
command in between. MRW-1 command must be issued first before issuing MRW-2 command.
NOTE 12 MRR-1 command must be immediately followed by CAS-2 command consecutively without any other com-
mand in between. MRR-1 command must be issued first before issuing CAS-2 command.
Page 153
4.39 TRR Mode - Target Row Refresh
A LPDDR4 SDRAM's row has a limited number of times a given row can be accessed within a refresh
period (tREFW * 2) prior to requiring adjacent rows to be refreshed. The Maximum Activate Count (MAC) is
the maximum number of activates that a single row can sustain within a refresh period before the adjacent
rows need to be refreshed. The row receiving the excessive actives is the Target Row (TRn), the adjacent
rows to be refreshed are the victim rows. When the MAC limit is reached on TRn, either the LPDRR4
SDRAM receive all (R * 2) Refresh Commands before another row activate is issued, or the LPDRR4
SDRAM should be placed into Targeted Row Refresh (TRR) mode. The TRR Mode will re-fresh the rows
adjacent to the TRn that encountered tMAC limit.
There could be a maximum of two target rows to a victim row in a bank. The cumulative value of the ac-
tivates from the two target rows on a victim row in a bank should not exceed MAC value as well.
MR24 fields required to support the new TRR settings. Setting MR24 [OP7=1] enables TRR Mode and
setting MR24 [OP7=0] disables TRR Mode. MR24 [OP6:OP4] defines which bank (BAn) the target row is
located in (See 3.4.24, MR24 table for details).
The TRR mode must be disabled during initialization as well as any other LPDRR4 SDRAM calibration
modes. The TRR mode is entered from a DRAM Idle State, once TRR mode has been entered, no other
Mode Register commands are allowed until TRR mode is completed, except setting MR24 [OP7=0] to
interrupt and reissue the TRR mode is allowed.
When enabled; TRR Mode is self-clearing; the mode will be disabled automatically after the completion of
defined TRR flow; after the 3rd BAn precharge has completed plus tMRD. Optionally the TRR mode can
also be exited via another MRS command at the completion of TRR by setting MR24 [OP7=0]; if the TRR
is exited via another MRS command, the value written to MR24 [OP6:OP4] are don’t cares.
4.39.1 TRR Mode Operation
1. The timing diagram in Figure 94 depicts TRR mode. The following steps must be performed when TRR
mode is enabled. This mode requires all three ACT (ACT1, ACT2 and ACT3) and three cor-responding
PRE commands (PRE1, PRE2 and PRE3) to complete TRR mode. A Precharge All (PREA) commands
issued while LPDDR4 SDRAM is in TRR mode will also perform precharge to BAn and counts towards a
PREn command.
2. Prior to issuing the MRW command to enter TRR mode, the SDRAM should be in the idle state. A MRW
command must be issued with MR24 [OP7=1] and MR24 [OP6:4] defining the bank in which the
targeted row is located. All other MR24 bits should remain unchanged.
3. No activity is to occur in the DRAM until tMRD has been satisfied. Once tMRD has been satisfied, the
only commands to BAn allowed are ACT and PRE until the TRR mode has been completed.
4. The first ACT to the BAn with the TRn address can now be applied, no other command is allowed at this
point. All other banks must remain inactive from when the first BAn ACT command is issued until [(1.5 *
tRAS) + tRP] is satisfied.
5. After the first ACT to the BAn with the TRn address is issued, a PRE to BAn is to be issued (1.5 * tRAS)
later; and then followed tRP later by the second ACT to the BAn with the TRn address. Once the 2nd
activate to the BAn is issued, nonBAn banks are allowed to have activity.
6. After the second ACT to the BAn with the TRn address is issued, a PRE to BAn is to be issued tRAS
later and then followed tRP later by the third ACT to the BAn with the TRn address.
Page 154
4.39.1 TRR Mode Operation (cont’d)
7. After the third ACT to the BAn with the TRn address is issued, a PRE to BAn would be issued tRAS
later; and once the third PRE has been issued, nonBAn banks are not allowed to have activity until TRR
mode is exited. The TRR mode is completed once tRP plus tMRD is satisfied.
8. TRR mode must be completed as specified to guarantee that adjacent rows are refreshed. Any-time the
TRR mode is interrupted and not completed, the interrupted TRR Mode must be cleared and then
subsequently performed again. To clear an interrupted TRR mode, an MR24 change is required with
setting MR24 [OP7=0], MR24 [OP6:4] are don’t care, followed by three PRE to BAn, tRP time in
between each PRE command. The complete TRR sequence (Steps 2-7) must be then re-issued and
completed to guarantee that the adjacent rows are refreshed.
9. Refresh command to the LPDRR4 SDRAM or entering Self-Refresh mode is not allowed while the
DRAM is in TRR mode.
T0 T1 T2 T3 Ta0 Ta1 Ta2 Ta3 Tb0 Tb1 Tc0 Tc1 Tc2 Tc3 Td0 Td1 Td2 Td3 Te0 Te1 Tf0 Tf1 Tf2 Tf3 Tg0 Tg1 Tg2 Tg3 Th0 Th1 Th2 Th3 Tk0 Tk1 Tk2 Th0 Th1 Th2 Th3 Th4 Th5
CK_c
CK_t
CKE
CS
ACT 1 PRE1 ACT 2 PRE2 ACT 3 PRE3

COMMAND MRW-1 MRW-2 DES ACT-1 ACT-2 DES PRE DES ACT-1 ACT-2 DES CMD-1 CMD-2 DES PRE DES CMD-1 CMD-2 DES ACT-1 ACT-2 DES CMD-1 CMD-2 DES PRE DES DES CMD-1 CMD-2 DES
TRR Entry 1st ACT 1st PRE 2nd ACT 2nd PRE 3rd ACT 3rd PRE
Bank *2 N/A N/A N/A N/A N/A BAn N/A N/A N/A BAn N/A BAn N/A N/A
Non
V BAn V V N/A BAn
Non
V BAn V V N/A BAn N/A N/A
Non
V BAn V V N/A BAn V
Any
V V
Address BA
Address OP MA OP OP TRn
O TRn
M TRn
O TRn
O N/A N/A TRn
O TRn
M TRn
O TRn
O V MV OV OV N/A N/A V MV OV OV TRn
O TRn
M TRn
O TRn
O V MV OV OV N/A N/A V MV OV OV
tMRD 1.5xtRAS tRP tRAS tRP tRAS tRP+tMRD
Non BAn Non BAn

in Idle No Activity allowed in other banks (banks closed) Activity allowed No Activity allowed
(may have bank(s) open)
Activity
allowed
BAn inBAn
Idle No Activity allowed (bank closed)
*9
BAn TRR operation allowed.
Activity
allowed
Note
1. TRn is targeted row. DON'T CARE TIME BRAKE
2. Bank BAn represents the bank in which the targeted row is located.
3. TRR mode self-clears after tMRD + tRP measured from 3rd BAn precharge PRE3 at clock edge Th4.
4. TRR mode or any other activity can be re-engaged after tRP + tMRD from 3rd BAn precharge PRE3.
PRE_ALL also counts if issued instead of PREn. TRR mode is cleared by DRAM after PRE3 to the BAn bank.
5. Activate commands to BAn during TRR mode do not provide refreshing support, i.e. the Refresh counter is unaffected.
6. The DRAM must restore the degraded row(s) caused by excessive activation of the targeted row (TRn) neccessary to meet refresh requirements.
7. A new TRR mode must wait tMRD+tRP time after the third precharge.
8. BAn may not be used with any other command.
9. ACT and PRE are the only allowed commands to BAn during TRR Mode.
10. Refresh commands are not allowed during TRR mode.
11. All DRAM timings are to be met by DRAM during TRR mode such as tFAW. Issuing of ACT1, ACT2 and ACT3 counts towards tFAW budget.
Figure 94 — TRR Mode
4.40 Post Package Repair (PPR)
LPDDR4 supports Fail Row address repair as optional feature and it is readable through MR25 OP[7:0]
PPR provides simple and easy repair method in the system and Fail Row address can be repaired by the
electrical programming of Electrical-fuse scheme.
With PPR, LPDDR4 can correct 1Row per Bank.
Electrical-fuse cannot be switched back to un-fused states once it is programmed. The controller should
prevent unintended the PPR mode entry and repair.
Page 155
4.40.1 Fail Row Address Repair
The following is procedure of PPR.
1. Before entering 'PPR' mode, All banks must be Precharged

2. Enable PPR using MR4 bit "OP4=1" and wait tMRD
3. Issue ACT command with Fail Row address
4. Wait tPGM to allow DRAM repair target Row Address internally then issue PRE
5. Wait tPGM_Exit after PRE which allow DRAM to recognize repaired Row address RAn
6. Exit PPR with setting MR4 bit "OP4=0"
7. LPDDR4 will accept any valid command after tPGMPST
8. In More than one fail address repair case, Repeat Step 2 to 7
Once PPR mode is exited, to confirm if target row is repaired correctly, host can verify by writing data into
the target row and reading it back after PPR exit with MR4 [OP4=0] and tPGMPST.
The following Timing diagram shows PPR operation.

T0 T1 T2 T3 T4 T5 Ta0 Ta1 Ta2 Ta3 Ta4 Ta5 Ta6 Tb0 Tb1 Tb2 Tb3 Tb4 Tc0 Tc1 Tc2 Tc3 Tc4 Tc5 Td0 Td1 Td2 Td3 Td4 Td5
CK_c
CK_t
CKE
CS
COMMAND DES MRW-1 MRW-2 Any Any

DES ACT-1 ACT-2 DES DES PRE DES MRW-1 MRW-2 DES Command Command
BA N/A N/A N/A N/A Valid BA Valid Valid Valid Valid N/A N/A N/A N/A Valid Valid Valid Valid
Address OP MA OP OP RAn RAn RAn RAn Valid Valid OP MA OP OP Valid Valid Valid Valid
PPR Normal Mode

(All banks must be Idle) Move to PPR Mode PPR Repair PPR Recognition Move to Normal Mode Normal
Mode
Status
tMRD tPGM tPGM_Exit tPGMPST
NOTE 1 During tPGM, any other commands (including refresh) are not allowed on each die.
NOTE 2 With one PPR command, only one row can be repaired at one time per die.
NOTE 3 When PPR procedure is done, reset command is required before normal operation.
NOTE 4 During PPR, memory contents is not refreshed and may be lost.
Figure 95 — PPR Timing
Table 64 — PPR Timing Parameters

Parameter Symbol min max Unit Note
PPR Programming Time tPGM 1000 - ms
PPR Exit Time tPGM_Exit 15 - ns
New Address Setting time tPGMPST 50 - us
Page 156
5 Absolute Maximum DC Ratings
Stresses greater than those listed may cause permanent damage to the device.
This is a stress rating only, and functional operation of the device at these or any other conditions above
those indicated in the operational sections of this standard are not implied. Exposure to absolute maximum
rating conditions for extended periods may affect reliability.
Table 65 — Absolute Maximum DC Ratings

VDD1 supply voltage relative to VSS VDD1 -0.4 2.1 V 1
VDDQ supply voltage relative to VSSQ VDDQ -0.4 1.5 V 1
Voltage on any ball except VDD1 relative to VSS VIN, VOUT -0.4 1.5 V
Storage Temperature TSTG -55 125 °C 2
NOTE 1 See "Power-Ramp" in 3.3 for relationships between power supplies.
NOTE 2 Storage Temperature is the case surface temperature on the center/top side of the LPDDR4 device. For the
measurement conditions, please refer to JESD51-2.
Page 157
6 AC and DC Operating Conditions
6.1 Recommended DC Operating Conditions
Table 66 — Recommended DC Operating Conditions

VDDQ supply voltage relative to VSSQ VDDQ -0.4 1.5 V 1
Voltage on any ball except VDD1 relative to VSS VIN, VOUT -0.4 1.5 V
Storage Temperature TSTG -55 125 °C 2
NOTE 1 VDD1 uses significantly less current than VDD2
NOTE 2 The voltage range is for DC voltage only. DC is defined as the voltage supplied at the DRAM and is inclusive
of all noise up to 20MHz at the DRAM package ball.
NOTE 3 The voltage noise tolerance from DC to 20MHz exceeding a pk-pk tolerance of 45mv at the DRAM ball is not
included in the TdIVW.
Page 158
7 AC and DC Input/Output Measurement Levels
7.1 V High speed LVCMOS (HS_LLVCMOS)
This standard defines power supply voltage range, dc interface, switching parameter and
overshoot/undershoot for high speed lower low-voltage CMOS family of non terminated digital circuits. The
specifications in this standard represent a minimum set of interface specifications for CMOS compatible
circuits.
The purpose of this standard is to provide a standard of specification for uniformity, multiplicity of sources,
elimination of confusion, and ease of device specification and design by users.
7.1.1 Standard specifications
All voltages are referenced to ground except where noted.
7.1.2 Recommended DC Operating conditions
Table 67 — Recommended DC operating conditions

Min. Typ. Max. Unit
VDD 1.06 1.1 1.17 V Input Buffer Power
VDDQ 1.06 1.1 1.17 V I/O Buffer Power
7.1.3 DC electrical characteristics
7.1.3.1 Class-1 LPDDR4 Input Level
Table 68 — Class-1 LPDDR4 Input Level

Parameter Symbol Min Max Unit Note
Input high level (AC) VIH(AC) 0.75*VDD(or VDDQ) VDD( or VDDQ)+0.2 V 1
Input low level (AC) VIL(AC) -0.2 0.25*VDD(or VDDQ) V 1
Input high level (DC) VIH(DC) 0.65*VDD(or VDDQ) VDD( or VDDQ)+0.2 V
Input low level (DC) VIL(DC) -0.2 0.35*VDD(or VDDQ) V
NOTE 1 Refer 7.1.4, LPDDR4 AC Over/Undershoot.
Input VIH(AC) VIH(DC)

VIL(AC)
Level VIL(DC)
= Don't Care
Note
1. AC level is guaranteed transition point.
2. DC level is hysteresis.
Figure 96 — Class-1 LPDDR4 Input AC timing definition
7.1.3.2 Class-1 LPDDR4 Output measurement Level
Table 69 — Class-1 LPDDR4 Output measurement Level

Parameter Symbol Min Max Unit
Output high voltage VOH 0.80 * VDDQ - V
Output low voltage VOL - 0.20 * VDDQ V
Page 159
7.1.4 AC Over/Undershoot
7.1.4.1 Class-1 LPDDR4 AC Over/Undershoot
Table 70 — Class 1 LPDDR4 AC Over/Undershoot

Parameter Specification
Maximum peak Amplitude allowed for overshoot area 0.35V
Maximum peak Amplitude allowed for undershoot area 0.35V
Maximum overshoot area above VDD/VDDQ 0.8V-ns
Maximum undershoot area below VSS/VSSQ 0.8V-ns
Maximum Amplitude
Overshoot Area
VDD
Volts
(V)
VSS
Undershoot Area
Maximum Amplitude
Time (ns)
Figure 97 — AC Overshoot and Undershoot Definition for Address and Control Pins
Page 160
7.2 Differential Input Cross Point Voltage

VDDQ
DQS_c
Vix_max
Vswing / 2
Vix_min
DQS_t
VSSQ
Figure 98 — DQS input Crosspoint voltage (Vix)
Table 71 —QS input voltage crosspoint(Vix) ratio

LPDDR4-2133 LPDDR4-3200 LPDDR4-4266
Parameter Symbol Units Notes
Min Max Min Max Min Max
DQS Differential input
Vix_DQS_ratio - 20 - 20 - 20 % 1,2
crosspoint voltage ratio
NOTE 1 Vix voltage is referenced to Vswing/2(avg)= 0.5(VDQS_t + VDQS_c) where the average is over (TBD1) UI.
NOTE 2 Ratio of the Vix pk voltage divided by Vdiff_DQS : Vix_DQS_Ratio = 100* (Vix_DQS/Vdiff DQS pkpk) where VdiffDQS pk-
pk = 2*|VDQS_t - VDQS_c|
VDD2
CK_c
Vix_max
Vswing / 2
Vix_min
CK_t
VSS
Figure 99 — CK input Crosspoint voltage (Vix)
Table 72 — CK input voltage crosspoint(Vix) ratio

LPDDR4-2133 LPDDR4-3200 LPDDR4-4266
Min Max Min Max Min Max
CK Differential input
Vix_CK_ratio - 25 - 25 - 25 % 1,2
crosspoint voltage ratio
NOTE 1 Vix voltage is referenced to Vswing/2(avg)= 0.5(VCK_t + VCK_c) where the average is over (TBD1) UI.
NOTE 2 Ratio of the Vix pk voltage divided by Vdiff_CK : Vix_CK_Ratio = 100* (Vix_CK/Vdiff CK pk-pk) where VdiffCK pk-pk =
2*|VCK_t - VCK_c|

Page 161
7.3 AC/DC Input level for ODT input
Table 73 — LPDDR4 Input Level for ODT

Symbol Min Max Unit Note
VIHODT(AC) ODT Input High Level (AC) 0.75*VDD VDD+0.2 V 1
VILODT(AC) ODT Input Low Level (AC) -0.2 0.25*VDD V 1
VIHODT(DC) ODT Input High Level (DC) 0.65*VDD VDD+0.2 V
VILODT(DC) ODT Input Low Level (DC) -0.2 0.35*VDD V
NOTE 1 See Table 76, Overshoot and Undershoot Specifications.
7.4 Single Ended Output Slew Rate
Delta TRse
VOH(AC)
Vcent
VOL(AC)
Delta TFse
Figure 100 — Single Ended Output Slew Rate Definition
Table 74 — Output Slew Rate (single-ended)

Value
Parameter Symbol 1
Units
Min Max2
Single-ended Output Slew Rate (VOH = VDDQ/3) SRQse* 3.5 9 V/ns
Output slew-rate matching Ratio (Rise to Fall) - 0.8 1.2 -
* SR = Slew Rate, Q = Query Output (like in DQ, which stands for Data-in, Query Output) se = Single-ended signals
NOTE 1 Measured with output reference load.

NOTE 2 The ratio of pull-up to pull-down slew rate is specified for the same temperature and voltage, over the entire
temperature and voltage range. For a given output, it represents the maximum difference between pull-up and pull-
down drivers due to process variation.
NOTE 3 The output slew rate for falling and rising edges is defined and measured between VOL(AC)=0.2*VOH(DC)
and VOH(AC)= 0.8*VOH(DC).
NOTE 4 Slew rates are measured under average SSO conditions, with 50% of DQ signals per data byte switching.
Page 162
7.4 Single Ended Output Slew Rate (cont’d)
Delta TRdiff
Delta TFdiff
Figure 101 — Differential Output Slew Rate Definition
Table 75 — Differential Output Slew Rate

Value
Parameter Symbol Units
Min Max
Differential Output Slew Rate (VOH=VDDQ/3) SRQdiff* 7 18 V/ns
* SR = Slew Rate, Q = Query Output (like in DQ, which stands for Data-in, Query Output) diff = Differential signals
NOTE 1 Measured with output reference load.

NOTE 2 The output slew rate for falling and rising edges is defined and measured between VOL(AC)=0.2*VOH(DC)
and VOH(AC)= 0.8*VOH(DC).
NOTE 3 Slew rates are measured under average SSO conditions, with 50% of DQ signals per data byte switching.
Page 163
7.5 Overshoot and Undershoot for LVSTL
Table 76 — AC Overshoot/Undershoot Specification

Data Rate
Parameter Units
1600 1866 3200 4266
Maximum peak amplitude allowed for overshoot area.
Max 0.3 0.3 0.3 TBD V
(See Figure 102)
Maximum peak amplitude allowed for undershoot area.
Max 0.3 0.3 0.3 TBD V
(See Figure 102)
Maximum area above VDD.
Max 0.1 0.1 0.1 TBD V-ns
(See Figure 102)
Maximum area below VSS.
Max 0.1 0.1 0.1 TBD V-ns
(See Figure 102)
NOTE 1 VDD2 stands for VDD for CA[5:0], CK_t, CK_c, CS_n, CKE and ODT. VDD stands for VDDQ for DQ, DMI,
DQS_t and DQS_c.
NOTE 2 VSS stands for VSS for CA[5:0], CK_t, CK_c, CS_n, CKE and ODT. VSS stands for VSSQ for DQ, DMI,
DQS_t and DQS_c.
NOTE 3 Maximum peak amplitude values are referenced from actual VDD and VSS values.
NOTE 4 Maximum area values are referenced from maximum operating VDD and VSS values.
Maximum Amplitude
Overshoot Area
VDD
Volts
(V)
VSS
Undershoot Area
Maximum Amplitude
Time (ns)
Figure 102 — Overshoot and Undershoot Definition

Page 164
7.6 LPDDR4 Driver Output Timing Reference load
These 'Timing Reference Loads’ are not intended as a precise representation of any particular system
environment or a depiction of the actual load presented by a production tester. System designers should
use IBIS or other simulation tools to correlate the timing reference load to a system environment.
Manufacturers correlate to their production test conditions, generally one or more coaxial transmission
lines terminated at the tester electronics.
DRAM
50
Note
1. All output timing parameter values are reported with respect to this reference load.
This reference load is also used to report slew rate.
Figure 103 — Driver Output Reference Load for Timing and Slew Rate
Page 165
7.7 LVSTL(Low Voltage Swing Terminated Logic) IO System
LVSTL I/O cell is comprised of pull-up, pull-down dirver and a terminator. The basic cell is shown in
Figure104.
VDDQ
PULL-UP
DQ
ODT
PULL-DOWN Enabled when receiving
VSSQ VSSQ
Figure 104 — LVSTL I/O Cell
To ensure that the target impedance is achived the LVSTL I/O cell is designed to be calibrated per the
following procedure:
1. Calibrate the pull-down device against a 240 Ohm resister to VDDQ via the ZQ pin.
• Set Strength Control to minimum setting.

• Increase drive strength until comparator detects data bit is less than VDDQ/2.
• NMOS pull-down device is calibrated to 240 Ohms.
VDDQ
Comparator
240
VDDQ/2
N
Strength control [N-1:0]
VSSQ
Figure 105 — Pull-down Calibration

Page 166
7.7 LVSTL(Low Voltage Swing Terminated Logic) IO System (cont’d)
2. Calibrate the pull-up device against the calibrated pull-down device.
• Set VOH target and NMOS controller ODT replica via MRS (VOH can be automatically controlled by ODT
MRS).
• Set Strength Control to minimum setting.
• Increase drive strength until comparator detects data bit is grater than VOH target.
• NMOS pull-up device is now calibrated to VOH target.
VDDQ
N
Strength control [N-1:0]
Comparator
VOH target
Calibrated NMOS PD Controller ODT replica could be

control + ODT information 60, 120 via MRS setting.
VSSQ
Figure 106 — Pull-up Calibration

Page 167
8 Input/Output Capacitance
Table 77 — Input/output capacitance

LPDDR4 LPDDR4
3200-533 4266-3733
Input capacitance, Min 0.5 TBD pF 1,2
CCK
CK_t and CK_c Max 0.9 TBD pF 1,2
Input capacitance Min 0.0 TBD pF 1,2,3
delta, CDCK
CK_t and CK_c Max 0.09 TBD pF 1,2,3
Input capacitance, Min 0.5 TBD pF 1,2,4
CI
All other input-only pins Max 0.9 TBD pF 1,2,4
Input capacitance Min -0.1 TBD pF 1,2,5
delta, CDI
All other input-only pins Max 0.1 TBD pF 1,2,5
Input/output
Min 0.7 TBD pF 1,2,6
capacitance,
CIO
DQ, DMI, DQS_t,
Max 1.3 TBD pF 1,2,6
DQS_c
Input/output Min 0.0 TBD pF 1,2,7
capacitance delta, CDDQS
DQS_t, DQS_c Max 0.1 TBD pF 1,2,7
Input/output Min -0.1 TBD pF 1,2,8
capacitance delta, CDIO
DQ, DMI Max 0.1 TBD pF 1,2,8
Input/output Min 0.0 TBD pF 1,2
capacitance, CZQ
ZQ pin Max 5.0 TBD pF 1,2
NOTE 1 This parameter applies to die device only (does not include package capacitance).
NOTE 2 This parameter is not subject to production test. It is verified by design and characterization. The capacitance
is measured according to JEP147 (Procedure for measuring input capacitance using a vector network analyzer (VNA)
with VDD1, VDD2, VDDQ, VSS, VSSQ applied and all other pins floating.
NOTE 3 Absolute value of CCK_t . CCK_c.
NOTE 4 CI applieds to CS_n, CKE, CA0~CA5.
NOTE 5 CDI = CI . 0.5 * (CCK_t + CCK_c)
NOTE 6 DMI loading matches DQ and DQS.
NOTE 7 Absolute value of CDQS_t and CDQS_c.
NOTE 8 CDIO = CIO . 0.5 * (CDQS_t + CDQS_c) in byte-lane.
Page 168
9 IDD Specification Parameters and Test Conditions
9.1 IDD Measurement Conditions
The following definitions are used within the IDD measurement tables unless stated otherwise:
LOW: VIN ≤ VIL(DC) MAX

HIGH: VIN ≥ VIH(DC) MIN
STABLE: Inputs are stable at a HIGH or LOW level
SWITCHING: See Table 78 and Table 79.
Table 78 — Definition of Switching for CA Input Signals

Switching for CA
CK_t edge R1 R2 R3 R4 R5 R6 R7 R8
CKE HIGH HIGH HIGH HIGH HIGH HIGH HIGH HIGH
CS LOW LOW LOW LOW LOW LOW LOW LOW
CA0 HIGH LOW LOW LOW LOW HIGH HIGH HIGH
CA1 HIGH HIGH HIGH LOW LOW LOW LOW HIGH
NOTE 1 CS must always be driven LOW.
NOTE 2 50% of CA bus is changing between HIGH and LOW once per clock for the CA bus.
NOTE 3 The above pattern is used continuously during IDD measurement for IDD values that require switching on the
CA bus.
Table 79 — CA pattern for IDD4R

Clock Cycle
CKE CS Command CA0 CA1 CA2 CA3 CA4 CA5
Number
N HIGH HIGH Read-1 L H L L L L
N+1 HIGH LOW L H L L L L
N+2 HIGH HIGH CAS-2 L H L L H L
N+3 HIGH LOW L L L L L L
N+4 HIGH LOW DES L L L L L L
N+8 HIGH HIGH Read-1 L H L L L L
N+9 HIGH LOW L H L L H L
N+10 HIGH HIGH CAS-2 L H L L H H
N+11 HIGH LOW H H H H H H
NOTE 1 BA[2:0] = 010, CA[9:4] = 000000 or 111111, Burst Order CA[3:2] = 00 or 11 (Same as JESD209-3, Table 57)
NOTE 2 Difference from JESD209-3: CA pins are kept low with DES CMD to reduce ODT current.
Page 169
9.1 IDD Measurement Conditions (cont’d)
Table 80 — CA pattern for IDD4W

Clock Cycle
CKE CS Command CA0 CA1 CA2 CA3 CA4 CA5
Number
N HIGH HIGH L L H L L L
Write-1
N+1 HIGH LOW L H L L L L
N+2 HIGH HIGH L H L L H L
CAS-2
N+3 HIGH LOW L L L L L L
N+8 HIGH HIGH L L H L L L
Write-1
N+9 HIGH LOW L H L L H L
N+10 HIGH HIGH L H L L H H
CAS-2
N+11 HIGH LOW L L H H H H
NOTE 1 BA[2:0] = 010, CA[9:4] = 000000 or 111111 (Same as JESD209-3, Table 58)
NOTE 2 Difference from JESD209-3: a) No burst ordering, and b) CA pins are kept low with DES CMD to reduce
ODT current.
Page 170
Table 81 - Data Pattern for IDD4W (DBI off)

DBI OFF Case No. of
DQ[7] DQ[6] DQ[5] DQ[4] DQ[3] DQ[2] DQ[1] DQ[0] DBI 1's
BL0 1 1 1 1 1 1 1 1 0 8
BL1 1 1 1 1 0 0 0 0 0 4
BL2 0 0 0 0 0 0 0 0 0 0
BL3 0 0 0 0 1 1 1 1 0 4
BL4 0 0 0 0 0 0 1 1 0 2
BL5 0 0 0 0 1 1 1 1 0 4
BL6 1 1 1 1 1 1 0 0 0 6
BL7 1 1 1 1 0 0 0 0 0 4
BL8 1 1 1 1 1 1 1 1 0 8
BL9 1 1 1 1 0 0 0 0 0 4
BL10 0 0 0 0 0 0 0 0 0 0
BL11 0 0 0 0 1 1 1 1 0 4
BL12 0 0 0 0 0 0 1 1 0 2
BL13 0 0 0 0 1 1 1 1 0 4
BL14 1 1 1 1 1 1 0 0 0 6
BL15 1 1 1 1 0 0 0 0 0 4
BL16 1 1 1 1 1 1 1 1 0 8
BL17 1 1 1 1 0 0 0 0 0 4
BL18 0 0 0 0 0 0 0 0 0 0
BL19 0 0 0 0 1 1 1 1 0 4
BL20 0 0 0 0 0 0 1 1 0 2
BL21 0 0 0 0 1 1 1 1 0 4
BL22 1 1 1 1 1 1 0 0 0 6
BL23 1 1 1 1 0 0 0 0 0 4
BL24 0 0 0 0 0 0 0 0 0 0
BL25 0 0 0 0 1 1 1 1 0 4
BL26 1 1 1 1 1 1 1 1 0 8
BL27 1 1 1 1 0 0 0 0 0 4
BL28 1 1 1 1 1 1 0 0 0 6
BL29 1 1 1 1 0 0 0 0 0 4
BL30 0 0 0 0 0 0 1 1 0 2
BL31 0 0 0 0 1 1 1 1 0 4
No. of
1's 16 16 16 16 16 16 16 16
NOTE 1 Same data pattern was applied to DQ[4], DQ[5], DQ[6], DQ[7] for reducing complexity for IDD4W/R pattern
programming.
Page 171
Table 82 — Data Pattern for IDD4R (DBI off)

DBI OFF Case No. of
BL0 1 1 1 1 1 1 1 1 0 8
BL1 1 1 1 1 0 0 0 0 0 4
BL2 0 0 0 0 0 0 0 0 0 0
BL3 0 0 0 0 1 1 1 1 0 4
BL4 0 0 0 0 0 0 1 1 0 2
BL5 0 0 0 0 1 1 1 1 0 4
BL6 1 1 1 1 1 1 0 0 0 6
BL7 1 1 1 1 0 0 0 0 0 4
BL8 1 1 1 1 1 1 1 1 0 8
BL9 1 1 1 1 0 0 0 0 0 4
BL10 0 0 0 0 0 0 0 0 0 0
BL11 0 0 0 0 1 1 1 1 0 4
BL12 0 0 0 0 0 0 1 1 0 2
BL13 0 0 0 0 1 1 1 1 0 4
BL14 1 1 1 1 1 1 0 0 0 6
BL15 1 1 1 1 0 0 0 0 0 4
BL16 1 1 1 1 1 1 1 1 0 8
BL17 1 1 1 1 0 0 0 0 0 4
BL18 0 0 0 0 0 0 0 0 0 0
BL19 0 0 0 0 1 1 1 1 0 4
BL20 1 1 1 1 1 1 0 0 0 6
BL21 1 1 1 1 0 0 0 0 0 4
BL22 0 0 0 0 0 0 1 1 0 2
BL23 0 0 0 0 1 1 1 1 0 4
BL24 0 0 0 0 0 0 0 0 0 0
BL25 0 0 0 0 1 1 1 1 0 4
BL26 1 1 1 1 1 1 1 1 0 8
BL27 1 1 1 1 0 0 0 0 0 4
BL28 0 0 0 0 0 0 1 1 0 2
BL29 0 0 0 0 1 1 1 1 0 4
BL30 1 1 1 1 1 1 0 0 0 6
BL31 1 1 1 1 0 0 0 0 0 4
No. of
1's 16 16 16 16 16 16 16 16
NOTE 1 Same data pattern was applied to DQ[4], DQ[5], DQ[6], DQ[7] for reducing complexity for IDD4W/R pattern
programming.
Page 172
Table 83 — Data Pattern for IDD4W (DBI on)

DBI ON Case No. of
BL0 0 0 0 0 0 0 0 0 1 1
BL1 1 1 1 1 0 0 0 0 0 4
BL2 0 0 0 0 0 0 0 0 0 0
BL3 0 0 0 0 1 1 1 1 0 4
BL4 0 0 0 0 0 0 1 1 0 2
BL5 0 0 0 0 1 1 1 1 0 4
BL6 0 0 0 0 0 0 1 1 1 3
BL7 1 1 1 1 0 0 0 0 0 4
BL8 0 0 0 0 0 0 0 0 1 1
BL9 1 1 1 1 0 0 0 0 0 4
BL10 0 0 0 0 0 0 0 0 0 0
BL11 0 0 0 0 1 1 1 1 0 4
BL12 0 0 0 0 0 0 1 1 0 2
BL13 0 0 0 0 1 1 1 1 0 4
BL14 0 0 0 0 0 0 1 1 1 3
BL15 1 1 1 1 0 0 0 0 0 4
BL16 0 0 0 0 0 0 0 0 1 1
BL17 1 1 1 1 0 0 0 0 0 4
BL18 0 0 0 0 0 0 0 0 0 0
BL19 0 0 0 0 1 1 1 1 0 4
BL20 0 0 0 0 0 0 1 1 0 2
BL21 0 0 0 0 1 1 1 1 0 4
BL22 0 0 0 0 0 0 1 1 1 3
BL23 1 1 1 1 0 0 0 0 0 4
BL24 0 0 0 0 0 0 0 0 0 0
BL25 0 0 0 0 1 1 1 1 0 4
BL26 0 0 0 0 0 0 0 0 1 1
BL27 1 1 1 1 0 0 0 0 0 4
BL28 0 0 0 0 0 0 1 1 1 3
BL29 1 1 1 1 0 0 0 0 0 4
BL30 0 0 0 0 0 0 1 1 0 2
BL31 0 0 0 0 1 1 1 1 0 4
No. of
1's 8 8 8 8 8 8 16 16 8
DBI enabled burst

Page 173
Table 84 — Data Pattern for IDD4R (DBI on)

DBI ON Case No. of
BL0 0 0 0 0 0 0 0 0 1 1
BL1 1 1 1 1 0 0 0 0 0 4
BL2 0 0 0 0 0 0 0 0 0 0
BL3 0 0 0 0 1 1 1 1 0 4
BL4 0 0 0 0 0 0 1 1 0 2
BL5 0 0 0 0 1 1 1 1 0 4
BL6 0 0 0 0 0 0 1 1 1 3
BL7 1 1 1 1 0 0 0 0 0 4
BL8 0 0 0 0 0 0 0 0 1 1
BL9 1 1 1 1 0 0 0 0 0 4
BL10 0 0 0 0 0 0 0 0 0 0
BL11 0 0 0 0 1 1 1 1 0 4
BL12 0 0 0 0 0 0 1 1 0 2
BL13 0 0 0 0 1 1 1 1 0 4
BL14 0 0 0 0 0 0 1 1 1 3
BL15 1 1 1 1 0 0 0 0 0 4
BL16 0 0 0 0 0 0 0 0 1 1
BL17 1 1 1 1 0 0 0 0 0 4
BL18 0 0 0 0 0 0 0 0 0 0
BL19 0 0 0 0 1 1 1 1 0 4
BL20 0 0 0 0 0 0 1 1 1 3
BL21 1 1 1 1 0 0 0 0 0 4
BL22 0 0 0 0 0 0 1 1 0 2
BL23 0 0 0 0 1 1 1 1 0 4
BL24 0 0 0 0 0 0 0 0 0 0
BL25 0 0 0 0 1 1 1 1 0 4
BL26 0 0 0 0 0 0 0 0 1 1
BL27 1 1 1 1 0 0 0 0 0 4
BL28 0 0 0 0 0 0 1 1 0 2
BL29 0 0 0 0 1 1 1 1 0 4
BL30 0 0 0 0 0 0 1 1 1 3
BL31 1 1 1 1 0 0 0 0 0 4
No. of
1's 8 8 8 8 8 8 16 16 8
DBI enabled burst

Page 174
9.2 IDD Specifications
IDD values are for the entire operating voltage range, and all of them are for the entire standard range, with
the exception of IDD6ET which is for the entire extended temperature range.
Table 85 — LPDDR4 IDD Specification Parameters and Operating Conditions

Parameter/Condition Symbol Power Supply Notes
Operating one bank active-precharge current: IDD01 VDD1

tCK = tCKmin; tRC = tRCmin;
CKE is HIGH;
CS is LOW between valid commands; IDD02 VDD2
CA bus inputs are switching;
Data bus inputs are stable
ODT disabled IDD0Q VDDQ 3
Idle power-down standby current:

IDD2P1 VDD1
tCK = tCKmin;
CKE is LOW;
CS is LOW; IDD2P2 VDD2
All banks are idle;
Data bus inputs are stable IDD2PQ VDDQ 3
ODT disabled
Idle power-down standby current with clock stop:

IDD2PS1 VDD1
CK_t =LOW, CK_c =HIGH;
CKE is LOW;
CS is LOW; IDD2PS2 VDD2
All banks are idle;
CA bus inputs are stable;
Data bus inputs are stable IDD2PSQ VDDQ 3
ODT disabled
Idle non power-down standby current:

IDD2N1 VDD1
tCK = tCKmin;
CKE is HIGH;
CS is LOW; IDD2N2 VDD2
All banks are idle;
Data bus inputs are stable IDD2NQ VDDQ 3
ODT disabled
Idle non power-down standby current with clock stopped:

IDD2NS1 VDD1
CK_t=LOW; CK_c=HIGH;
CKE is HIGH;
CS is LOW; IDD2NS2 VDD2
All banks are idle;
Data bus inputs are stable IDD2NSQ VDDQ 3
ODT disabled
Page 175
9.2 IDD Specifications (cont’d)
Table 85 — LPDDR4 IDD Specification Parameters and Operating Conditions (cont'd)

Active power-down standby current:
IDD3P1 VDD1
tCK = tCKmin;
CKE is LOW;
CS is LOW; IDD3P2 VDD2
One bank is active;
Data bus inputs are stable IDD3PQ VDDQ 3
ODT disabled
Active power-down standby current with clock stop:

IDD3PS1 VDD1
CK_t=LOW, CK_c=HIGH;
CKE is LOW;
CS is LOW; IDD3PS2 VDD2
One bank is active;
Data bus inputs are stable IDD3PSQ VDDQ 4
ODT disabled
Active non-power-down standby current:

IDD3N1 VDD1
tCK = tCKmin;
CKE is HIGH;
CS is LOW; IDD3N2 VDD2
One bank is active;
Data bus inputs are stable IDD3NQ VDDQ 4
ODT disabled
Active non-power-down standby current with clock stopped:

IDD3NS1 VDD1
CKE is HIGH;
CS is LOW; IDD3NS2 VDD2
One bank is active;
Data bus inputs are stable IDD3NSQ VDDQ 4
ODT disabled
Operating burst READ current:

IDD4R1 VDD1
tCK = tCKmin;
CS is LOW between valid commands;
One bank is active; IDD4R2 VDD2
BL = 8; RL = RL(MIN);
50% data change each burst transfer IDD4RQ VDDQ 5
ODT disabled
Page 176
9.2 IDD Specifications (cont’d)
Table 85 — LPDDR4 IDD Specification Parameters and Operating Conditions (cont'd)

Operating burst WRITE current:
IDD4W1 VDD1
tCK = tCKmin;
CS is LOW between valid commands;
One bank is active; IDD4W2 VDD2
BL = 8; WL = WLmin;
50% data change each burst transfer IDD4WQ VDDQ 4
ODT disabled
All-bank REFRESH Burst current:

IDD51 VDD1
tCK = tCKmin;
CKE is HIGH between valid commands;
tRC = tRFCabmin; IDD52 VDD2
Burst refresh;
Data bus inputs are stable; IDD5Q VDDQ 4
ODT disabled
All-bank REFRESH Average current: IDD5AB1 VDD1

tCK = tCKmin;
tRC = tREFI; IDD5AB2 VDD2
Data bus inputs are stable;
IDD5ABQ VDDQ 4
ODT disabled
Per-bank REFRESH Average current: IDD5PB1 VDD1

tCK = tCKmin;
tRC = tREFI/8; IDD5PB2 VDD2
IDD5PBQ VDDQ 4
ODT disabled
Self refresh current (-25°C to +85°C): IDD61 VDD1 6,7,8,10

CKE is LOW;
CA bus inputs are stable; IDD62 VDD2 6,7,8,10
Maximum 1x Self-Refresh Rate;
IDD6Q VDDQ 4,6,7,8,10
ODT disabled
NOTE 1 Published IDD values are the maximum of the distribution of the arithmetic mean.
NOTE 2 ODT disabled: MR11[2:0] = 000B.
NOTE 3 IDD current specifications are tested after the device is properly initialized.
NOTE 4 Measured currents are the summation of VDDQ and VDD2.
NOTE 5 Guaranteed by design with output load = 5pF and RON = 40 ohm.
NOTE 6 The 1x Self-Refresh Rate is the rate at which the LPDDR4 device is refreshed internally during Self-Refresh, before going
into the elevated Temperature range.
NOTE 7 This is the general definition that applies to full array Self Refresh.
NOTE 8 Supplier datasheets may contain additional Self Refresh IDD values for temperature subranges within the Standard or
elevated Temperature Ranges.
NOTE 9 For all IDD measurements, VIHCKE = 0.8 x VDD2, VILCKE = 0.2 x VDD2.
NOTE 10 IDD6 85°C is guaranteed, IDD6 45°C is typical of the distribution of the arithmetic mean.
Page 177
10 Electrical Characteristics and AC Timing
10.1 Clock Timing
Table 88 — Clock AC Timings

LPDDR4-1600 LPDDR4-2400 LPDDR4-3200 LPDDR4-4266
Min Max Min Max Min Max Min Max
Clock Timing
Average clock period tCK(avg) 1.25 100 0.833 100 0.625 100 0.467 100 ns
Average High pulse width tCH(avg) 0.46 0.54 0.46 0.54 0.46 0.54 TBD TBD tCK(avg)
Average Low pulse width tCL(avg) 0.46 0.54 0.46 0.54 0.46 0.54 TBD TBD tCK(avg)
tCK(avg) tCK(avg) tCK(avg) tCK(avg)
MIN + MIN + MIN + MIN +
Absolute clock period tCK(abs) - - - - ns
tJIT(per) tJIT(per) tJIT(per) tJIT(per)
MIN MIN MIN MIN
Absolute High clock
tCH(abs) 0.43 0.57 0.43 0.57 0.43 0.57 TBD TBD tCK(avg)
pulse width
Absolute Low clock
tCL(abs) 0.43 0.57 0.43 0.57 0.43 0.57 TBD TBD tCK(avg)
pulse width
Clock period jitter tJIT(per) -70 70 -50 50 -40 40 - TBD ps
Maximum Clock Jitter
between consecutive tJIT(cc) - 140 - 100 - 80 - TBD ps
cycles
Page 178
10.2 Core Timing

Table 89 — Core AC Timing
Parameter Symbol Min/Max Data Rate
Unit
Core Parameters 533 1066 1600 2133 2667 3200 3733 4267
tRAS + tRPab
ACTIVATE-to-ACTIVATE
(with all-bank precharge)
command period tRC MIN ns
tRAS + tRPpb
(same bank)
(with per-bank precharge)
Minimum Self-Refresh
tSR MIN max(15ns, 3nCK) ns
Time (Entry to Exit)
SELF REFRESH exit to
next valid command tXSR MIN max(tRFCab + 7.5ns, 2nCK) ns
delay
Exit Power-Down to next
tXP MIN max(7.5ns, 3nCK) ns
valid command delay
CAS-to-CAS delay tCCD MIN 8 tCK(avg)
Internal READ to
PRECHARGE command tRTP MIN max(7.5ns, 8nCK) ns
delay
RAS-to-CAS delay tRCD MIN max(18ns, 4nCK) ns
Row precharge time
tRPpb MIN max(18ns, 3nCK) ns
(single bank)
Row precharge time
tRPab MIN max(21ns, 3nCK) ns
(all banks)
MIN max(42ns, 3nCK) ns
Row active time tRAS Min(9 * tREFI * Refresh Rate, 70.2) us
MAX -
( Refresh Rate is specified by MR4, OP[2:0] )
WRITE recovery time tWR MIN max(18ns, 4nCK) ns
WRITE-to-READ delay tWTR MIN max(10ns, 8nCK) ns
Active bank-A to active
tRRD MIN max(10ns, 4nCK) ns
bank-B
Precharge to Precharge
tPPD MIN 4 tCK
Delay1
Four-bank ACTIVATE
tFAW MIN 40 ns
window
NOTE 1 Precharge to precharge timing restriction does not apply to Auto-Precharge commands.
Page 179
10.3 Temperature Derating for AC timing
Table 90 — Temperature Derating AC Timing

Min/ Data Rate
Max
533 1066 1600 2133 2667 3200 3733 4267
Temperature Derating1
DQS output access time
tDQSCK MAX 3600 ps
from CK_t/CK_c (derated)
RAS-to-CAS delay (derated) tRCD MIN tRCD + 1.875 ns
ACTIVATE-to- ACTIVATE
tRC MIN tRC + 3.75 ns
command period (derated)
Row active time (derated) tRAS MIN tRAS + 1.875 ns
Row precharge time
tRP MIN tRP + 1.875 ns
(derated)
Active bank A to
tRRD MIN tRRD + 1.875 ns
active bank B (derated)
NOTE 1 Timing derating applies for operation at 85 °C to 105 °C.
Page 180
10.4 CA Rx voltage and timing
The command and address(CA) including CS input receiver compliance mask for voltage and timing is
shown in Figure 107. All CA, CS signals apply the same compliance mask and operate in single data rate
mode.
The CA input receiver mask for voltage and timing is shown in the figure below is applied across all CA
pins. The receiver mask (Rx Mask) defines the area that the input signal must not encroach in order for the
DRAM input receiver to be expected to be able to successfully capture a valid input signal; it is not the valid
data-eye.
TcIVW_total
Rx Mask
Vcent_CA(pin mid)
VcIVW
Figure 107 — CA Receiver(Rx) mask
Vcent_CA(pin mid) is defined as the midpoint between the largest Vcent_CA voltage level and the smallest
Vcent_CA voltage level across all CA and CS pins for a given DRAM component. Each CA Vcent level is
defined by the center, i.e., widest opening, of the cumulative data input eye as depicted in Figure 108. This
clarifies that any DRAM component level variation must be accounted for within the DRAM CA Rx mask.
The component level VREF will be set by the system to account for Ron and ODT settings.
CAx CAy CAz
Vcent_CAz
Vcent_CAx
Vcent_CAy
Vref variation
(Component)
Figure 108 — Across pin VREF CA voltage variation

Page 181
10.4 CA Rx voltage and timing (cont’d)
Ck, CK Data-in at DRAM Pin

Minimum CA Eye center aligned
CK_t
CK_c
VcIVW
Rx Mask
CA
DRAM Pin
TcIVW
All CA signals center aligned to the CK
at the DRAM pin
Figure 109 — CA Timings at the DRAM Pins
All of the timing terms in Figure 109 are measured from the CK_t/CK_c to the center(midpoint) of the
TcIVW window taken at the VcIVW_total voltage levels centered around Vcent_CA(pin mid).
tr tf
Rx Mask
Vcent_CA(pin mid)
TcIPW
Note
1. SRIN_cIVW=VcIVW_Total/(tr or tf), signal must be monotonic within tr and tf range.
Figure 110 — CA TcIPW and SRIN_cIVW definition (for each input pulse)
Page 182
10.4 CA Rx voltage and timing (cont’d)
VIHL_AC(min)/2
Vcent_CA
Rx Mask Rx Mask Rx Mask VcIVW
VIHL_AC(min)/2
Figure 111 — CA VIHL_AC definition (for each input pulse)
Table 91 — DRAM CMD/ADR, CS

DQ-1333A DQ-1600/1867 DQ-3200 DQ-4266
Symbol Parameter Unit NOTE
min max min max min max min max
Rx Mask
VcIVW - 175 - 175 - 155 - 145 mV 1,2,4
voltage - p-p
TcIVW Rx timing window - 0.3 - 0.3 - 0.3 - 0.3 UI* 1,2,3,4
CA AC input pulse
VIHL_AC 210 - 210 - 190 - 180 - mV 5,8
amplitude pk-pk
TcIPW CA input pulse width 0.55 0.55 0.6 0.6 UI* 6
Input Slew Rate
SRIN_cIVW 1 7 1 7 1 7 1 7 V/ns 7
over VcIVW
* UI = tck(avg)min
A
The following Rx voltage and absolute timing requirements apply for DQ operating frequencies at or below 1333 for all speed bins.
For example the TcIVW(ps) = 450ps at or below 1333 operating frequencies.
NOTE 1 CA Rx mask voltage and timing parameters at the pin including voltage and temperature drift.
NOTE 2 Rx mask voltage VcIVW total(max) must be centered around Vcent_CA(pin mid).
NOTE 3 Rx differential CA to CK jitter total timing window at the VcIVW voltage levels.
NOTE 4 Defined over the CA internal VREF range. The Rx mask at the pin must be within the internal VREF CA range
irrespective of the input signal common mode.
NOTE 5 CA only input pulse signal amplitude into the receiver must meet or exceed VIHL AC at any point over the
total UI. No timing requirement above level. VIHL AC is the peak to peak voltage centered around Vcent_CA(pin mid)
such that VIHL_AC/2 min must be met both above and below Vcent_CA.
NOTE 6 CA only minimum input pulse width defined at the Vcent_CA(pin mid).
NOTE 7 Input slew rate over VcIVW Mask centered at Vcent_CA(pin mid).
NOTE 8 VIHL_AC does not have to be met when no transitions are occurring.
Page 183
10.5 DRAM Data Timing

t QSH(DQS_t) t QSL(DQS_t)
DQS_c
DQS_t
t QH
t DQSQ
Associated
DQ Pins
Figure 112 — Read data timing definitions tQH and tDQSQ across on DQ signals per DQS group
DQS_c
DQS_t
t QW
DQx
t QW
DQx
t QW
DQz
Figure 113 — Read data timing tQW valid window defined per DQ signal
10.5 DRAM Data Timing (cont’d)
Table 92 — Read output timings
LPDDR4-1600/1867 LPDDR4-2133/2400 LPDDR4-3200 LPDDR4-4266
Parameter Symbol Units* Notes
Data Timing
DQS_t,DQS_c to DQ
Skew total, per group, per tDQSQ - 0.18 - 0.18 - 0.18 - 0.18 UI 1
access (DBIDisabled)
DQ output hold time total

min(tQSH, min(tQSH, min(tQSH, min(tQSH,
from DQS_t, DQS_c tQH - - - - UI 1
tQSL) tQSL) tQSL) tQSL)
(DBI-Disabled)
DQ output window time

total, per pin tQW_total 0.75 - 0.73 - 0.7 - 0.7 - UI 1,4
(DBI-Disabled)
deterministic, per pin tQW_dj TBD - TBD - - TBD - TBD UI 1,4,3
(DBIDisabled)
DQS_t,DQS_c to DQ
Skew total,per group, per
tDQSQ_DBI - TBD - TBD - TBD - TBD UI 1
access
(DBI-Enabled)
DQ output hold time total

min(tQSH_DBI, min(tQSH_DBI, min(tQSH_DBI, min(tQSH_DBI,
from DQS_t, DQS_c tQH_DBI - - - TBD UI 1
tQSL_DBI) tQSL_DBI) tQSL_DBI) tQSL_DBI)
(DBI-Enabled)

total, per pin tQW_total_DBI TBD - TBD - - TBD - TBD UI 1,4
(DBI-Enabled)
Page 184
10.5 DRAM Data Timing (cont’d)
Page 185
Table 92 — Read output timings (cont'd)
LPDDR4-1600/1867 LPDDR4-2133/2400 LPDDR4-3200 LPDDR4-4266
Parameter Symbol Units* Notes
Data Strobe Timing
DQS, DQS# differential
tCL(abs) tCL(abs) tCL(abs) tCL(abs)
output low time tQSL - - - - tCK(avg) 4,5
-0.05 -0.05 -0.05 -0.05
(DBI-Disabled)
tCH(abs) tCH(abs) tCH(abs) tCH(abs)
output high time tQSH - - - - tCK(avg) 4,6
-0.05 -0.05 -0.05 -0.05
(DBI-Disabled)
tCL(abs) tCL(abs) tCL(abs) tCL(abs)
output low time tQSL_DBI - - - - tCK(avg) 5,7
-0.045 -0.045 -0.045 -0.045
(DBI-Enabled)
tCH(abs) tCH(abs) tCH(abs) tCH(abs)
output high time tQSH_DBI - - - - tCK(avg) 6,7
-0.045 -0.045 -0.045 -0.045
(DBI-Enabled)
* Unit UI = tCK(avg)min/2
NOTE 1 DQ to DQS differential jitter where the total includes the sum of deterministic and random timing terms for a specified BER. BER specification and
measurement method are TBD1.
NOTE 2 The deterministic component of the total timing. Measurement method TBD1.
NOTE 3 This parameter will be characterized and guaranteed by design.
NOTE 4 This parameter is function of input clock jitter. These values assume the min tCH(abs) and tCL(abs). When the input clock jitter min tCH(abs) and tCL(abs) is
0.44 or greater of tck(avg) the min value of tQSL will be tCL(abs)-0.04 and tQSH will be tCH(abs) -0.04.
NOTE 5 tQSL describes the instantaneous differential output low pulse width on DQS_t - DQS_c, as measured from on falling edge to the next consecutive rising
edge.
NOTE 6 tQSH describes the instantaneous differential output high pulse width on DQS_t - DQS_c, as measured from on falling edge to the next consecutive rising
edge.
NOTE 7 This parameter is function of input clock jitter. These values assume the min tCH(abs) and tCL(abs). When the input clock jitter min tCH(abs) and tCL(abs) is
0.44 or greater of tck(avg) the min value of tQSL will be tCL(abs)-0.04 and tQSH will be tCH(abs) -0.04.

Page 186
10.6 DQ Rx Voltage and Timing
The DQ input receiver mask for voltage and timing is shown Figure 114 is applied per pin. The "total" mask
(VdIVW_total, TdiVW_total) defines the area the input signal must not encroach in order for the DQ input
receiver to successfully capture an input signal with a BER of lower than TBD1. The mask is a receiver
property and it is not the valid data-eye.
TdIVW_total
Rx Mask
VdIVW Vcent_DQ(pin mid)

Total
Figure 114 — DQ Receiver(Rx) mask
Vcent_DQ(pin_mid) is defined as the midpoint between the largest Vcent_DQ voltage level and the
smallest Vcent_DQ voltage level across all DQ pins for a given DRAM component. Each DQ Vcent is
defined by the center, i.e., widest opening, of the cumulative data input eye as depicted in Figure 115. This
clarifies that any DRAM component level variation must be accounted for within the DRAM Rx mask. The
component level VREF will be set by the system to account for Ron and ODT settings.
DQx DQy DQz
Vcent_DQz
Vcent_DQx
Vcent_DQy
Vref variation
(Component)
Figure 115 — Across pin VREF DQ voltage variation

Page 187
10.6 DQ Rx Voltage and Timing (cont’d)
DQ, DQS Data-in at DRAM Latch DQ, DQS Data-in at DRAM Pin
Internal Componsite Data- Eye Non Minimum Data-Eye/ Maximum Rx Mask
Center aligned to DQS
DQS_c DQS_c
DQS_t DQS_t
tDQS2DQx*
VdIVW_total
DQx,y,z Rx Mask
DRAM Pin
tDQS2DQy*
VdIVW_total
All DQ signals center aligned to the Rx Mask
strobe at the DRAM internal latch DRAM Pin
VdIVW_total tDQS2DQz*
Rx Mask
DRAM Pin
tDQ2DQ
NOTE:
1. tDQS2DQ is measured at the center(midpoint) of the TdiVW window.
2. DQz represents the max tDQS2DQ in this example
3. DQy represents the min tDQS2DQ in this example
Figure 116 — DQ to DQS tDQS2DQ and tDQDQ Timings at the DRAM pins
referenced from the internal latch
All of the timing terms in DQ to DQS t are measured from the DQS_t/DQS_c to the center(midpoint) of the
TdIVW window taken at the VdIVW_total voltage levels centered around Vcent_DQ(pin_mid). In Figure
116 the timings at the pins are referenced with respect to all DQ signals center aligned to the DRAM
internal latch. The data to data offset is defined as the difference between the min and max tDQS2DQ for a
given component.
Page 188
tr tf
Rx Mask
Vcent_DQ(pin mid) VdIVW

Total
TdIPW
Note
1. SRIN_dIVW=VdIVW_Total/(tr or tf), signal must be monotonic within tr and tf range.
Figure 117 — DQ TdIPW and SRIN_dIVW definition (for each input pulse)
VIHL_AC(min)/2
Vcent_DQ
Rx Mask Rx Mask Rx Mask VdIVW_total
VIHL_AC(min)/2
Figure 118 — DQ VIHL_AC definition (for each input pulse)

Page 189
Table 94 — DRAM DQs In Receive Mode
1600/1867A 2133/2400 3200 4266
Symbol Parameter Unit NOTE
min max min max min max min max
VdIVW_total Rx Mask voltage - p-p total - 140 - 140 - 140 - 120 mV 1,2,3,5
Rx timing window total
TdIVW_total - 0.22 - 0.22 - 0.25 - 0.25 UI* 1,2,4,5
(At VdIVW voltage levels)
Rx timing window 1 bit toggle 1,2,4,
TdIVW_1bit - TBD - TBD - TBD - TBD UI*
(At VdIVW voltage levels) 5,14
DQ AC input pulse amplitude
VIHL_AC 180 - 180 - 180 - 170 - mV 7,15
pk-pk
Input pulse width
TdIPW DQ 0.45 0.45 0.45 0.45 UI* 8
(At Vcent_DQ)
tDQS2DQ DQ to DQS offset 200 800 200 800 200 800 200 800 ps 9
tDQDQ DQ to DQ offset - 30 - 30 - 30 - 30 ps 10
DQ to DQS offset temperature
tDQS2DQ_temp - 0.6 - 0.6 - 0.6 - 0.6 ps/°C 11
variation
DQ to DQS offset voltage
tDQS2DQ_volt - 33 - 33 - 33 - 33 ps/50mV 12
variation
Input Slew Rate over
SRIN_dIVW 1 7 1 7 1 7 1 7 V/ns 13
VdIVW_total
* UI = tCK(avg)min/2
A
The following Rx voltage and absolute timing requirements apply for DQ operating frequencies at or below 1333 for all speed bins. For example the TcIVW(ps) = 450ps at or below 1333
operating frequencies.
Page 190
Notes to Table 94.
NOTE 1 Data Rx mask voltage and timing parameters are applied per pin and includes the DRAM DQ to DQS voltage
AC noise impact for frequencies >20MHz and max voltage of 45mv pk-pk from DC-20MHz at a fixed temperature on
the package. The voltage supply noise must comply to the component Min-Max DC operating conditions.
NOTE 2 The design specification is a BER (TBD1). The BER will be characterized and extrapolated if necessary
using a dual dirac method.
NOTE 3 Rx mask voltage VdIVW total(max) must be centered around Vcent_DQ(pin_mid).
NOTE 4 Rx differential DQ to DQS jitter total timing window at the VdIVW voltage levels.
NOTE 5 Defined over the DQ internal VREF range. The Rx mask at the pin must be within the internal VREF DQ range
irrespective of the input signal common mode.
NOTE 6 Deterministic component of the total Rx mask voltage or timing. Parameter will be characterized and
guaranteed by design. Measurement method TBD1.
NOTE 7 DQ only input pulse amplitude into the receiver must meet or exceed VIHL AC at any point over the total UI.
No timing requirement above level. VIHL AC is the peak to peak voltage centered around Vcent_DQ(pin_mid) such
that VIHL_AC/2 min must be met both above and below Vcent_DQ.
NOTE 8 DQ only minimum input pulse width defined at the Vcent_DQ(pin_mid).
NOTE 9 DQ to DQS offset is within byte from DRAM pin to DRAM internal latch. Includes all DRAM process, voltage
and temperature variation.
NOTE 10 DQ to DQ offset defined within byte from DRAM pin to DRAM internal latch for a given component.
NOTE 11 TDQS2DQ max delay variation as a function of temperature.
NOTE 12 TDQS2DQ max delay variation as a function of the DC voltage variation for VDDQ and VDD2. It includes the
VDDQ and VDD2 AC noise impact for frequencies > 20MHz and max voltage of 45mv pk-pk from DC-20MHz at a fixed
temperature on the package. For tester measurement VDDQ=VDD2 is assumed.
NOTE 13 Input slew rate over VdIVW Mask centered at Vcent_DQ(pin_mid).
NOTE 14 Rx mask defined for a one pin toggling with other DQ signals in a steady state.
NOTE 15 VIHL_AC does not have to be met when no transitions are occurring.

Standard Improvement Form JEDEC
The purpose of this form is to provide the Technical Committees of JEDEC with input from the industry
regarding usage of the subject standard. Individuals or companies are invited to submit comments to
JEDEC. All comments will be collected and dispersed to the appropriate committee(s).
If you can provide input, please complete this form and return to:
JEDEC Fax: 703.907.7583
Attn: Publications Department
3103 North 10th Street
Suite 240 South
Arlington, VA 22201-2107
1. I recommend changes to the following:

Requirement, clause number
Test method number Clause number
The referenced clause number has proven to be:

Unclear Too Rigid In Error
Other
2. Recommendations for correction:
3. Other suggestions for document improvement:
Submitted by
Name: Phone:
Company: E-mail:
Address:
City/State/Zip: Date:
Rev. 8/13

Jesd209 4

Uploaded by

Copyright:

Available Formats

Jesd209 4

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

Jesd209 4

Uploaded by

Copyright:

Available Formats

JEDEC

Low Power Double Data Rate 4

JEDEC SOLID STATE TECHNOLOGY ASSOCIATION

This document is copyrighted by the JEDEC and may not be

For information, contact:

JEDEC Solid State Technology Association

or refer to www.jedec.org under Standards-Documents/Copyright Information

LOW POWER DOUBLE DATA RATE 4 (LPDDR4)

2 Package Ballout & Pin Definition

2.1 Pad Order

19 VSSQ 59 VSSQ 119 VSSQ 159 VSSQ

2.2 Package Ballout

Using Variation VFFCDB for MO-273

NOTE 1 15mm x 15mm, 0.4mm ball pitch

2.2.2 200-ball x32 Discrete Package, 0.80mm x 0.65mm using MO-311

NOTE 1 0.8mm pitch (X-axis), 0.65mm pitch (Y-axis), 22 rows.

DQS0_t_ DQS1_t_ DQS0_t_ DQS1_t_

NOTE 1 0.5 mm ball pitch. MO-TBD

DQS0_c_ DQS1_c_ DQS0_c_ DQS1_c_

DQS0_t_ DQS1_t_ DQS0_t_ DQS1_t_

JEDEC Standard No.209-4

2.3 Pad Definition and Description

Table 1 — Pad Definition and Description

Data Mask Inversion: DMI is a bi-directional signal which is driven HIGH

LPDDR4-SDRAM is a high-speed synchronous DRAM device internally configured as 2-channel and 8-

These devices contain the following number of bits:

4Gb has 4,294,967,296 bits

1. As of publication of this document, under discussion by the formulating committee.

3.1 LPDDR4 SDRAM Addressing

3.2 Simplified LPDDR4 State Diagram

MPC SRE REF All MPC MPC

Figure 1 — LPDDR4: Simplified Bus Interface State Diagram-1

A) FIFO Based Write / Read Timing

MPC MPC MRW

Figure 2 — LPDDR4: Simplified Bus Interface State Diagram-2

3.3 Power-up, Initialization and Power-off Procedure

Table 3 — MRS defaults settings

3.3.1 Voltage Ramp and Device Initialization

Table 4 — Voltage Ramp Conditions

tINIT5=2us(min) tZQCAL = 1us(min) tZQLAT = Max(30ns, 8tCK)(Min)

CA[5:0] MRW ZQ Cal ZQ Cal CA BUS Write DQ

DQs Valid Valid Valid

Figure 3 — Power Ramp and Initialization Sequence

3.3.1 Voltage Ramp and Device Initialization

Table 5 — Initialization Timing Parameters

tINIT2 10 - ns Minimum CKE low time before RESET_n high

tINIT3 2 - ms Minimum CKE low time after RESET_n high

3.3.2 Reset Initialization with Stable Power

2. Repeat steps 4 to 10 in 3.3.1, Voltage Ramp and Device Initialization.

Table 6 — Reset Timing Parameter

3.3.3 Power-off Sequence

Table 7 — Power Supply Conditions

3.3.4 Uncontrolled Power-Off Sequence

Table 8 — Timing Parameters Power Off

3.4 Mode Register Definition

3.4.1 MR0 Register Information (MA [7:0] = 00H)

OP[7] OP[6] OP[5] OP[4] OP[3] OP[2] OP[1] OP[0]

00B: RZQ Self-Test Not Supported

CATR 0B: CA for this rank is not terminated