Advanced Digital Systems

Design
(EE6471)

02/10/2002 EE6471 (KR) 1

Intro/Advanced Digital Systems Design

… where digital meets analogue

02/10/2002 EE6471 (KR) 2

Intro/Recommended Literature
Book/Author ISBN Rating EE6471 Relevance EE6471 Relevant Chapter(s)

Johnson/Graham 0-13-395724-1 80% 80% 1-8, 11, 12

”High-Speed Digital Design”

W.Dally/J.Poulton 0-521 59292-5 80% 70% 1-3

”Digital System Engineering”

Tocci/Widmer 0-13-085634-7 80% 70% 8

”Digital Systems”

D.A. Neamen 0256261156 100% 30% Background information

”Electronic Circuit Analysis and Design”

B. Razavi 0-07-118815-0 100% 20% Background information

”Design of Analog CMOS Integrated
Circuits”
Y. Tsividis 0-07-065523-5 100% 10% Fundamentals
”The MOS Transistor”

Boylestad 0-13-927187-2 80% 10% Fundamentals

”Introductory Circuit Analysis”

N.Weste, K.Eshraghian 0-201-53376-6 70% 10% Fundamentals

”Principles of CMOS VLSI Design”

Hayes 0-201-15461-7 60% 0% Fundamentals

”Digital Logic Design”

02/10/2002 EE6471 (KR) 3

Intro/Preface
• This module will highlight and explain analog
circuit principles relevant to high-speed digital
designs
• High speed design isn’t black magic. It’s a
matter of knowing which principles apply
• For their designs to function, high-speed
designers must know and use analog design
principles

02/10/2002 EE6471 (KR) 4

Intro/Many Engineers of Digital Systems…

– Don’t believe currents flow in loops

– Don’t believe in the H-field
– Don’t believe gates are differential amplifiers
– Don’t believe in electromagnetic waves
– Don’t believe that an understanding of EMC will
advance their careers

02/10/2002 EE6471 (KR) 5

Review/Session Objectives…
– Review the fundamentals which are relevant to
advanced digital systems design
• Equivalent circuits (Thevenin, Norton)
• Complex impedances
– time-domain characteristics
– frequency characteristics
• Impedances of passive networks
• Transfer functions
– bode plots
– transient response
– Spectra of square wave signals
– Definition of rise and fall times
– Definition of knee frequency

02/10/2002 EE6471 (KR) 6

Review/Equivalent Circuits…
• Thevenins Theorem
Vcc Vcc
– any two-port circuit of linear elements can be
replaced by an equivalent circuit (TEC) consisting
R1
a single voltage source Vth, and a single series
resistance Rth R2
• 3 steps to determine TEC parameters
– disconnect load
– determine the open-circuit voltage across the
output terminals of the two-port (Vth)
Vcc
– determine the total resistance appearing across the
output terminals of the two-port (Rth)
» consider internal voltage sources shorts
» consider internal currents sources open Rth

Vth

02/10/2002 EE6471 (KR) 7

Review/Inductance

il • arises whereever there is electric current

• currents create magnetic fields
• inductance = reluctance of current to
L vl change
• in high-speed digital design: inductance
mainly due to circuit parasitics
• Unit: H (Henry)

{L} = H = Vs
A

02/10/2002 EE6471 (KR) 8

Review/Inductance/t-domain Equations

il • Time-domain equations:

dil
L vl
vl = L ⋅
dt
1
il = ∫ vl ⋅ dt + il initial
L

• steady state operation: ∫ vl ⋅ dt = 0

02/10/2002 EE6471 (KR) 9
Review/Inductance/t-domain Equations
vr
• Step response of an LR circuit
R di
Vcc L vl V = vl + vr = L + Ri
dt

V −
t

i = 1 − e τ 

60
49.663

40
R 
il ( t)
m ⋅A L
20
Time constant τ =
0 0
R
0 100 200 300 400 500 600
0 t 500
pico ⋅s
Example Parameters:
Vcc=5V, L=10nH, R=100Ω

02/10/2002 EE6471 (KR) 10

Review/Inductance/Ground Bounce
Vcc • Example: Digital gate ground bounce
10
8

0
il ( t)
m ⋅A

vl( t)
L vl m ⋅V
10

− 20 20
Example parameters: 0 5 10 15 20 25
0 t 24.96
•Ground trace inductance:L=5nH n ⋅s
•Current step ∆I=8mA
•Gate output voltage fall time Tf=6ns dil ∆il
•Gate output voltage rise time Tr=2ns vl = L ⋅   → vl = L ⋅
in this example

dt ∆t

02/10/2002 EE6471 (KR) 11

Review/Inductance/f-domain Equations

il • Frequency-domain equations:

Complex impedance
π
L vl j
Zl = j ⋅ Xl = Xl ⋅ e 2
with Xl = ω ⋅ L
Reactance Xl = impedance magnitude

Laplace domain
Zl = pL with p = σ + j ⋅ ω

02/10/2002 EE6471 (KR) 12

Review/Inductance/f-domain Equations

il • Impedance magnitude vs frequency

1 .10
3
628.319
100
L vl 10

Zl( p( fi) ) 1
Ω 0.1

0.01

1 .10
3

0.0011 .10 4
1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10
3 4 5 6 7 8 9

1000 f ( fi) 9
1×10
Hz

Example Parameter:
L=100nH

02/10/2002 EE6471 (KR) 13

Review/Capacitance

• arises whereever there are two

ic conducting bodies charged to different
electric potentials
C vc • two bodies at different electric potential
have an electric field between them
• capacitance = reluctance of voltage to
change
• Unit: F (Farad)

{C} = F = As
V

02/10/2002 EE6471 (KR) 14

Review/Capacitance/t-domain Equations

• Time-domain equations:
ic
dvc
C vc ic = C ⋅
dt
1
vc = ∫ ic ⋅ dt + vcinitial
C

• steady state operation: ∫ ic ⋅ dt = 0

02/10/2002 EE6471 (KR) 15
Review/Capacitance/t-domain Equations
Vcc
• Step response of an RC circuit
vr
1
R vc 0 = vc + vr = ∫ ic ⋅ dt + R ⋅ ic
C C
t
−
5
6 vc = V ⋅ e τ

4
vc( t)
V
2
Time constant τ = RC
0.034 0
0 1000 2000 3000 4000 5000
0 t 5000
pico ⋅s Example Parameters:
Vcc=5V, C=10pF, R=100Ω

02/10/2002 EE6471 (KR) 16

Review/Capacitance/f-domain Equations

• Frequency-domain equations:
ic
Complex impedance
π
C vc Zc =
1
= Xc ⋅ e
−j
2
j ⋅ω ⋅ C
Reactance Xc = impedance magnitude

Laplace domain
1
Zc = with p = σ + j ⋅ ω
pC

02/10/2002 EE6471 (KR) 17

Review/Capacitance/f-domain Equations

• Impedance magnitude vs frequency

ic 1 .10
7
6
1.592 ×10
1 .10
6

C vc 1 .10
5

. 4
Zc( p( fi) ) 1 10
Ω 1 .10
3

100

10
1.592 1
1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10
3 4 5 6 7 8 9

1000 f ( fi) 9
1×10
Hz

Example Parameter:
C=100pF

02/10/2002 EE6471 (KR) 18

Review/Passive Networks

• Simple model for many practical inductors

L

Z = jω ⋅ L + R
?
R
f _ 3dB =
2π ⋅ L
Example Parameter:
L=100nH, R=2Ω

02/10/2002 EE6471 (KR) 19

Review/Passive Networks

• Simple model for many practical inductors

L 1 .10
3
628.322

100
R Zl( p( fi) )
Ω
10

Z = jω ⋅ L + R
2 1
1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10
3 4 5 6 7 8 9
9
R 1000 f ( fi) 1×10
f _ 3dB = Hz
2π ⋅ L
Example Parameter:
L=100nH, R=2Ω

02/10/2002 EE6471 (KR) 20

Review/Passive Networks

• Improved model for practical inductors

L (and resistors!)
C
R

Z = (R + j ω ⋅ L )
1
jω ⋅ C
?
Example Parameter:
L=100nH, C=10pF, R=2Ω

02/10/2002 EE6471 (KR) 21

Review/Passive Networks

• Improved model for practical inductors

L (and resistors!)
C
1 .10
4
4612.318
R
1 .10
3

Zl( p( fi) )
100
Ω

Z = (R + j ω ⋅ L )
1 10

jω ⋅ C 2 1
1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10
3 4 5 6 7 8 9

1000 f ( fi) 9
1×10
Hz

Example Parameter:
L=100nH, C=10pF, R=2Ω

02/10/2002 EE6471 (KR) 22

Review/Passive Networks

• Simple model for many practical capacitors

C

Z=
R

1
jω ⋅ C
+R
?
R commonly refered to as ESR (Equivalent Series Resistance)

1
f _ 3dB = Example Parameter:
2π ⋅ RC C=100pF, ESR=100Ω

02/10/2002 EE6471 (KR) 23

Review/Passive Networks

• Simple model for many practical capacitors

C
1 .10
7
6
1.592 ×10
1 .10
6

1 .10
R
5
Zc( p( fi) )
Ω
1 .10
4

1 .10
3

1
Z= +R 100.013 100
jω ⋅ C 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10
3 4 5 6 7 8 9

1000 f ( fi) 9
1×10
Hz
R commonly refered to as ESR (Equivalent Series Resistance)

1
f _ 3dB = Example Parameter:
2π ⋅ RC C=100pF, ESR=100Ω

02/10/2002 EE6471 (KR) 24

Review/Passive Networks

• Popular model for practical capacitors

C

ESR

ESL
?
1
Z = ESR + jω ⋅ ESL +
jω ⋅ C Example Parameter:
C=100pF, ESL=10nH, ESR=1Ω

02/10/2002 EE6471 (KR) 25

Review/Passive Networks

• Popular model for practical capacitors

C
1 .10
7
6
1.592 ×10
1 .10
6

ESR 1 .10
5

. 4
Zc( p( fi) ) 1 10
Ω 1 .10
3

ESL 100

10
1.004 1
1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10
3 4 5 6 7 8 9

1000 f ( fi) 9
1×10
Hz

1
Z = ESR + jω ⋅ ESL +
jω ⋅ C Example Parameter:
C=100pF, ESL=10nH, ESR=1Ω

02/10/2002 EE6471 (KR) 26

Review/Passive Networks
Fairly accurate model for
L_lead SMT chip inductor

ESR
C_stray L := 1⋅ µ ⋅ H ESR := 1⋅ Ω L_lead := 10⋅ n ⋅ H C_stray := 30⋅ pico⋅ F

L
Z( p ) := 2p ⋅ L_lead + parallel p ⋅ L + ESR , 
1
 p ⋅ C_stray 
L_lead
1 .10
5

1 .10
4

1 .10
3

Z( p( fi) )
100
Ω
10

Note the simplicity 1

of the Mathcad 0.1.

1 10
3
0.01 0.1 1 10 100 1 .10
3

source code… f ( fi)

Meg ⋅Hz

02/10/2002 EE6471 (KR) 27

Review/Transfer Functions
• Voltage Transfer Function
Z1
vout ( j ⋅ ω )
vin Z2 vout T ( j ⋅ω ) =
vin( j ⋅ ω )

Z 2( j ⋅ ω )
T ( j ⋅ω ) =
Z1( j ⋅ ω ) + Z 2( j ⋅ ω )

Visualisation of transfer functions through Bode Plots

Gain( f ) = 20 ⋅ log( T ( j ⋅ 2π ⋅ f ) ) " dB" Phase( f ) = arg(T ( j ⋅ 2π ⋅ f ))

02/10/2002 EE6471 (KR) 28

Review/Transfer Functions

Generalised Transfer Function: Normal Form

constant gain Q real zeroes

 
Q

∏  1 + j ⋅ ω 
q =1 
2π ⋅ fz 
T ( j ⋅ω ) = K
q

R  
2
M
    2d   j ⋅ω  
( j ⋅ω ) N
⋅ ∏ 1 + j ⋅ ω
π ⋅  ⋅ ∏ 1 +   j ⋅ ω +  

m =1 
2 fp m  r =1 2π ⋅ fn r  2π ⋅ fn r
    

N poles M real poles R complex conjugate poles

at origin

02/10/2002 EE6471 (KR) 29

Review/Transfer Functions
– Linear Approximations for sketching bode plots…
• each pole introduces a change in the slope of –20dB/decade
• each zero introduces a change in the slope of +20dB/decade
• linear approximations ok for transfer functions with real poles and real zeroes.

0
0
Example Parameter:
20 Single (real) pole
Pole Frequency fp=1MegHz
gain( T ( p( fi) ) )
40
Tp_app( f ( fi) )

60

− 60 80 3
1 .10 1 .10
3
0.01 0.1 1 10 100
0.001 f ( fi) 1000
Meg ⋅Hz

02/10/2002 EE6471 (KR) 30

Review/Transfer Functions/1st Order
• Voltage Transfer Function
R 1
vin vout Zc ( j ⋅ ω ) j ⋅ω ⋅ C
T ( j ⋅ω ) =
C 1
= =
R + Zc ( j ⋅ ω ) R + 1 1 + j ⋅ ω ⋅ RC
j ⋅ω ⋅ C
Pole
Gain Plot Phase Plot
0 0
− 4.343×10
−6 − 0.057

20

gain( T ( p( fi) ) ) phase( T ( p( fi) ) )

40 50
−3 − 45

60

− 60 80 3 − 89.943 100 3
1 .10 1 .10
3
1 .10 1 .10
3 0.01 0.1 1 10 100
0.01 0.1 1 10 100
0.001 f ( fi) 1000 0.001 f ( fi) 1000
Meg ⋅Hz Meg ⋅Hz

Example Parameter:
R=1000Ω, C=159pF
02/10/2002 EE6471 (KR) 31
Review/Transfer Functions/2nd Order
T ( j ⋅ω ) =
1
2nd Order Transfer Function 2
 2d   j ⋅ω 
– damping ratio d determines 1 +   j ⋅ ω +  
type of pole  2π ⋅ fn r   2π ⋅ fn r 
• d<1 underdamped (complex
poles) Gain Plot
• d>1 overdamped (real poles) 20
13.979

• d=1 critically damped

0
gain( T_2nd ( p( fi) , 0.1 ) )

gain( T_2nd ( p( fi) , 0.5 ) )

20
gain( T_2nd ( p( fi) , 1) )

gain( T_2nd ( p( fi) , 2) )

40

− 40.569 60
0.1 1 10
0.1 f ( fi) 10
Meg ⋅Hz

02/10/2002 EE6471 (KR) 32

Review/Transfer Functions/2nd Order
T ( j ⋅ω ) =
1
L R 2
 2d   j ⋅ω 
vin C vout 1 +   j ⋅ ω +  
 2π ⋅ fn   2π ⋅ fn 

T ( j ⋅ω ) =
• Z0: characteristic impedance 1
of the resonant circuit 1 + j ⋅ ω ⋅ RC + ( j ⋅ ω ) LC
2

• d: damping ratio
• Q: Quality factor 1 L R
fn = Z0 = d=
• δ: decay constant 2π ⋅ LC C 2⋅ Z0
(oscillations decay with exp(-δt))
• fn: natural resonant critical damping : d = 1 ⇔ R = 2 ⋅ Z 0
frequency of the undamped
system fr = fn 1 − d 2 for d < 1
• fr: resonant frequency of the 1 R
damped system Q= δ = d ⋅ ωn =
d 2L

02/10/2002 EE6471 (KR) 33

Review/Transfer Functions/2nd Order 2
2

Importance of damping 1
bitstream_sampledj
for digital systems… bitstream_RLC j
0

• Excessive ringing can

d=0.1 −1
– cause digital gates to 1
0 200 400 600 800 1000 1200
0 j 1023
malfunction
2
2
– can destroy digital
gates bitstream_sampledj
1

– cause EMI problems bitstream_RLC j

0

d=0.5 −1 1
0 200 400 600 800 1000 1200
0 j 1023

R L 2
2

C
1
bitstream_sampledj

bitstream_RLC j
Example Parameter: 0

C=10pF, L=0.7nH
Random bitstream 100MBit/s d=1.0 −1 1
0 200 400 600 800 1000 1200
0 j 1023

02/10/2002 EE6471 (KR) 34

Basics/Common Rise and Fall Time Definitions
– 10%-90% (sometimes 20%-80%) – most common vsignal
• Advantages 0.9 Vmag
– simple to measure (2-point measurement) Vmag
– many scopes can automatically make this measurement 0.1 Vmag
t
• Disadvantages
– sensitivity to noise, ringing, and asymptotic behaviour
Tr10% −90%
– Center slope rise time
• Advantages
– less sensitivity
Tr20% −80%
• Disadvantages
– awkward to measure (few scopes directly support it) Vmag
– Maximum slope rise time Trcs =
dvsignal
• Advantages dt 50%
– less sensitivity
– relevant for cross-coupling effects Vmag
• Disadvantages Trms =
– really awkward to measure (fewer scopes directly support it) max  dvsignal 
 dt 

02/10/2002 EE6471 (KR) 35

Basics/Composite Rise and Fall Time
n
Trcomposite = Tr12 + Tr 2 2 + LTrn 2 = ∑ Tri
2

i =1

– Relates the rise (fall) time of a complete system to the rise time of each of its components
– Equation is valid for cascaded linear systems
• Example: Pulse generator, scopes, scope probes
– Equation is not applicable to nonlinear systems
• e.g. saturating amplifiers, logic gates etc. where a regeneration of a fresh rising (falling) edge
takes place
– Equation holds (approximately) true for the different definitions of rise (fall) times
• just use the same definition for all rise (fall) times involved
• robustness of the equation stems from a property of the convolution operator
(comment: when cascading systems together, their impulse response convolve)

02/10/2002 EE6471 (KR) 36

Basics/Spectrum of Periodic Signals
Fourier series for arbitrary periodic functions (general form)

f (t ) = a 0 + ∑n =1 [a n cos(2π ⋅ f 0 ⋅ n ⋅ t ) + bn sin (2π ⋅ f 0 ⋅ n ⋅ t )]

∞

f (t ) = ∑n = −∞ c n ⋅ e j 2π ⋅ f 0⋅n⋅t
∞

Any given periodic function (satisfying certain Note:

conditions) can be expressed in terms of a fourier •periodic functions ⇒ discrete spectrum
series. •fundamental frequency f0
Comment: The idea can be extended to non-periodic •harmonics at n*f0
functions (fourier integrals). •real fourier coefficients an and bn
•complex fourier coefficient cn

02/10/2002 EE6471 (KR) 37

Basics/Spectrum of Square Wave Signals
1
f0=
Tr Tf Tp

A Tw Tw
Duty cycle d =
t Tp
Tp

Fourier series for periodic square wave signals

⋅ [sin (2π ⋅ f 0 ⋅ n ⋅ t )]
Example Parameter:
f (t ) =
A 2A 1
+ ∑ n=1,3,5K
∞
f0=100kHz. A=1
2 π n
150
116.076
Equation valid for:
•duty cycle d=50%
100
•Tr=Tf=0
(
dBµV spectrum_sq m ⋅2 )
Note: 50

•Only sine terms (ac content is an odd function)

•Only odd harmonics 0 0
1 .10
3
100
•Amplitude of harmonics decrease 20dB/decade 100 1 1000
m⋅
Tp
k ⋅Hz

02/10/2002 EE6471 (KR) 38

Basics/Spectrum of Square Wave Signals
1
f0=
Tr Tf Tp

A Tw Tw
Duty cycle d =
t Tp
Tp
Example Parameter:
f0=100kHz. A=1, Tr=Tf=100ns (0%-100%)
Above a certain frequency harmonics
roll off with >20dB/decade 116.076
150

•the frequency where this happens is

sometimes called “knee frequency”
100
•the knee frequency is not a function of the ( )
dBµV spectrum_sqm ⋅2

fundamental frequency
Ampl ( m)
•the knee frequency is a function of rise 50

and fall times Tr and Tf

0 0
1 .10 1 .10 1 .10 1 .10
3 4 5 6
100
100 1 819200
m⋅
Tp
k ⋅Hz
02/10/2002 EE6471 (KR) 39
Basics/Definition of Knee Frequency
K
f 3dB = Pulse Type K
Tr10% −90% Exponential 0.350
(1st order)
Note:
•f3dB is often referred to as “knee frequency” 2nd order 0.344
of a signal (critically
•Pulse shape hardly matters damped)
•K=0.35 typically used
•sometimes K=0.5 is used for convervative Gaussian 0.339
approximations

Significance of the knee frequency:

•simple definition only based on signal rise (fall) time
•Use the knee frequency to make a quick judgement of frequency effects
•but bear its limitations in mind: knee frequency is an imprecise measure of spectral content

02/10/2002 EE6471 (KR) 40

Basics/Spectrum of Square Wave Signals
Linear Pulse Gaussian Pulse
150 150
116.076 116.077

100 100
(
dBµV spectrum_sq m ⋅2 ) (
dBµV spectrum_sq q ⋅2 )
Ampl ( m) Ampl ( q)
50 50

0 0 0 0
1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10
3 4 5 6 3 4 5 6
100 100
100 1 819200 100 1 819200
m⋅ q⋅
Tp Tp
k ⋅Hz k ⋅Hz

Harmonics above the knee frequency depend on pulse type.

Common types:
•Linear transition
•Exponential transition
•2nd order (critically damped) Example Parameter:
•Gaussian transition f0=100kHz. A=1, Tr=Tf=100ns (0%-100%)

02/10/2002 EE6471 (KR) 41