Interconnect II – Class 22

Prerequisite Reading - Chapter 4

12/4/2002
 2

Effects of Frequency Domain

Phenomena on Time Domain Digital
Signals
Key Topics:

 Frequency Content of Digital Waveforms

 Frequency Envelope
 Incorporating frequency domain effects into
time domain signals

Interconnect II 12/4/2002
 3

Decomposing a Digital Signal into Frequency Components

• Digital signals are composed of an infinite number of
sinusoidal functions – the Fourier series

1+2+3
The Fourier
1 +series
2 is shown in its progression to approximate a square wave:
1

0
-  2 3

1+2+3+4+5
1+2+3+4

Square wave: Y = 0 for - < x < 0 and Y=1 for 0 < x < 

Y = 1/2 + 2/pi( sinx + sin3x/3 + sin5x/5 + sin7x/7 … + sin(2m+1)x/(2m+1) + …)

1 2 3 4 5
May do with sum of cosines too.
Interconnect II 12/4/2002
 4

Frequency Content of Digital Signals

• The amplitude of the the sinusoid components are
used to construct the “frequency envelope” – Output
of FT
Tr
0.35
20dB/decade
Tr

Pw
T

40dB/decade

1
T
Harmonic Number
1 3 5 7 9 …...
Interconnect II 12/4/2002
 5

Estimating the Frequency Content

0.35
• Where does that famous equation F  come from?
Tr

• It can be derived from the response of a step function into a

filter with time constant tau t /
V  Vinput (1  e )
• Setting V=0.1Vinput and V=0.9Vinput allows the calculation of the
10-90% risetime in terms of the time constant
t1090%  t90%  t10%  2.3  0.105  2.195
• The frequency response of a 1 pole network is
1 1
F3dB   
2 2F3dB
• Substituting into the step response yields
1.09 0.35
t10 90%  
F3dB F3dB
Interconnect II 12/4/2002
 6

Estimating the Frequency Content

0.35 Edge time
t1090%  factor
F3dB
This equation says:
 The frequency response of the network with time
constant tau will degrade a step function to a
risetime of t10-90%
 The frequency response of the network determines
the resulting rise time ( or transition time)
 The majority of the spectral energy will be
contained below F3dB
• This is a good “back of the envelope” way to estimate
the frequency response of a digital signal.
• Simple time constant estimate can take the form L/R,
L/Z0, R*C or Z0*C.
Interconnect II 12/4/2002
Examining Frequency Content of Digital Signals 7

• The frequency dependent effects described earlier in

this class can be applied to each sinusoidal function in
the series
 Digital signal decomposed into its sinusoidal
components
 Frequency domain transfer functions applied to
each sinusoidal component
 Modified sinusoidal functions are then re-combined
to construct the altered time digital signal
• There are several ways to determine this response
 Fourier series (just described)
 Fast Fourier transform (FFT)
Widely available in tools such as excel,
Mathematica, MathCad…
Interconnect II 12/4/2002
 8

3 Method of Generating a Square Wave

 Ramp pulses
Use Heavy Side function
Used for first pass simulations
 Power Exponential Pulses
Realistic edge that can match silicon performance
Used for behavioral simulation that match silicon
performance.
 Sum of Cosines
Text book identity.
Used to get a quick feel for impact of frequency
dependant phenomena on a wave.

Interconnect II 12/4/2002
 9

Ramp Square Wave

Ramp Pulse Train

RUe( tt )  tt   ( tt )   tt 
1  1
     tt  
   

onepulse( tt  k)  ( RUe( tt  period k)  RUe( tt  pw  k period) )

number_of_pulses
Rpt( t )   onepulse( t  k)
k 0

Rpt  t i

0 50 100 150
ti
ns

Interconnect II 12/4/2002
 10

Power Exponential Square Wave

Power Exponetial Pulse Train

  (   tt ) 
2.5
edge( tt )  Va  1  e  Power Exponential Edge

Ppt ( tt )  edge( tt )   ( tt )

onepulse( tt  k)  ( Ppt ( tt  period k)  Ppt ( tt  pw  k period) )

number_of_pulses
Ppt ( t )   onepulse( t  k)
k 0

Ppt  t i

0 50 100 150
ti
ns
Interconnect II 12/4/2002
 11

Sum Cosine Square Wave

Sum Cosine Pulse Wave
Establish Fourier Spectrum
period C
C   Tedge a 
2 period n   2
freq ( n )  n  1  3  10
period

Define pulse train of fourier coef i.e. pulse = sum of cosines

Fpt ( tt )    cos  n   a  cos ( freq ( n )  tt )   Va

1 4 1
2

2
  2 
   ( 1  2 a) n n 
1.5

Fpt  t i 0.5

0.5
0 50 100 150
ti
ns
Interconnect II 12/4/2002
Applying Frequency Dependent Effects 12

to Digital Functions
Input signal into lossy t-line Spectral content of waveform
0.35
F
Trise
FT
Volts

FFT
Time Frequency
ly Time domain waveform with
u ltip
Loss characteristics if t-line M frequency dependent losses
With AC losses
Attenuation (V2/V1)

No AC losses

Inverse

Volts
FFT

Time
Frequency

AC losses will degrade BOTH the amplitude and the edge rate
Interconnect II 12/4/2002
 13

Assignment
 Use MathCad to create a pulse wave with
Sum of sine waves
Sum of ramps
Sum of realistic edge waveforms
Exponential powers
 Use MathCad to determine edge time
factor for exponential and Gaussian wave,
10% - 90%
20% - 80%

Interconnect II 12/4/2002
 14

Edge Rate Degradation due to filtering

Remember this equation from a few slides ago?
t1090%  t90%  t10%  2.3  0.105  2.195
This equation says:
• If a step response is driven into a filter with tine constant
tau, the output edge rate is t10-90%
• However, realistic edge rates are not step functions
• RSS the input edge rate with the filter response
Input edge
tr = 300ps
Example:
tout=Output edge
Zo=50 C=5pF

t1090%  2.195  2.195( Zo  C )  2.195(50  5 pF )  548 ps

2 2
tout  Tr  Tout  (300 ps ) 2  (548 ps ) 2  625 ps
Interconnect II 12/4/2002
 15

Additional Effects
Key Topics:

 Serpentine traces
 Bends
 ISI
 Topology

Interconnect II 12/4/2002
 16

Effects of a Serpentine Trace

• Serpentine traces will exhibit 2 modes of propagation
• Typical “straight line” mode
• Coupled mode via the parallel sections
• Causes the signal to “speed up” because a portion of
the signal will propagate perpendicular to the
serpentine
• ”Speed up” is dependent on the spacing and the length
0.65

0.55

0.45

0.35
Volts

0.25
Lp S=5
0.15 S=15

S 0.05
0.0E+00

5.0E-10

1.0E-09

1.5E-09

2.0E-09

2.5E-09

3.0E-09

3.5E-09

4.0E-09
-0.05

-0.15
Time, s
Interconnect II 12/4/2002
 17

Modeling Serpentines
 Assignment – Find a the uncoupled trace length that
matches the delay of the serpentine route below
Use Maxwell Spice/2D modeling of serpentine vs. equal
length wave.
Trace 1 oz copper
5 mil space
route on 5mil width
PWB 5 mil distance to ground plane
Symmetric stripline
Use 50 ohm V source w/ 1ns
rise time (do for ramp and Gaussia
1” 2 port Tline
model 5 mil 2 port
Tline Model
10 port
Transmission Line Spice
Model
Couple length=2 inches
Interconnect II 12/4/2002
 18

Rules of Thumb for Serpentine Trace

• The following suggestions will help minimize the effect of
serpentine traces
• Make the minimum spacing between parallel section (s)
at least 3-4H, this will minimize the coupling between
parallel sections

• Minimize the length of the parallel sections (Lp) as

much as possible

• Embedded microstrips and striplines exhibits less

serpentine effects than normal m9ictrostirpsd

Interconnect II 12/4/2002
 19

Effects of bends
• Virtually every PCB design will exhibit bends
• The excess area caused by a 90o bend will increase
the self capacitance seen at the bend
• Empirically inspired model of a 90o bend is simply 1
square of excess capacitance
C90o _ bend  C11W Capacitance of
1 extra square
• Measurements have shown increased
delays due to the current components
“hugging” the corner increasing the
mean length
• 2 rights do not necessarily equal a
left and a right, especially for
wide traces
• 45o bends, round and chamfered
bends exhibit reduced effects
Interconnect II 12/4/2002
 20

Inter Symbol Interference

• Inter symbol interference (ISI) is reflection noise that
effects both amplitude and timing
 The nature of this interference is cause by a signal not settling
to a steady stated value before the next transition occurs.
 Can have an effect similar to crosstalk but has completely
different physics
Volts

Ideal waveform beginning transition

from low to high with no reflections or losses

Timing
difference Waveform beginning transition from low to high
with unsettled noise cased by reflections.

Receiver switching threshold

Time

Different starting point due to ISI

Interconnect II 12/4/2002
 21

Inter Symbol Interference

• ISI can dramatically affect the signal quality
 Depending on the switching rate/pattern, significant differences in waveform
shape can be realized – one or two patterns won’t produce worst case
 If the designer does not account for this effect, switching patterns that are
unaccounted for result in latent product defects.
400 MHz switching
4 200 MHz switching

2
Volts

-1
Ideal 400 MHz waveform
-2
0.E+00

1.E-09

2.E-09

3.E-09

4.E-09

5.E-09

6.E-09

7.E-09

8.E-09

9.E-09

1.E-08
Time, s
Interconnect II 12/4/2002
 22

Topology – the Key to a sound design

• What about the case where there is more than one
receiver, or more than one driver (e.g., a Multi-
processor FSB)
Zo2 Receiver 1
Rs=Zo L1
L2 (L1=L2)
Vs 0-2V Zo1 L3
Zo3
Receiver 2

• There will be an impedance discontinuity at the

junction
 The equivalent input impedance looking into the junction will
be the parallel combination of Zo2 and Zo1
Z || Z  Z o1
  o 2 o3
Z o 2 || Z o 3  Z o1
• This model can be simplified and solved with lattice diagrams
 Valid when L1=L2
Rs=Zo L1 L2
Vs 0-2V Zo1 Z=Zo2||Zo3

Interconnect II 12/4/2002
 23

Topology – the Key to a sound design

• Now, consider the case where L2 and L3 are NOT Equal
Zo2 Receiver 1
Rs=Zo
L2 L
Vs
Zo1 3
0-2V Zo3
Receiver 2
2.5

2.0
Voltage at receiver

1.5  The reflections from the receiver discontinuities will

not arrive at the same time; the 2 segment
simplification is not applicable
1.0

 This topology will ring with a frequency dependant in

0.5 L2 and L3

 This topology can be solved with a multi-segment

0.0 lattice diagram
0.0 2.0 4.0 6.0 8.0 10.0
Time, ns
Interconnect II 12/4/2002
 24

Topology – the Key to a sound design

Zo R1
Rs=Zo In J
Zo
Vs Zo Vinitial  2 1
0-2V T3 T2 Zo Zo  Rs
R2
1  0  2 3 Zo
 Zo
2 1
 2  3  
Zo 3
 Zo
2
2
T2  T3  1   
A 3
2 2 4
A’ A  
3 3 3
B 2 2 2 2 8
B    
B’ 3 3 9 9 9
C 2 2 2 2 2 2 2 2 16
C        
3 3 9 9 3 3 9 9 9
2 2 4
A'   
3 3 3
2 2 2 2 2 2 20
B'       
3 3 3 3 9 9 9
In J R1 R2
Interconnect II 12/4/2002