VL-155 Practical Differential Pair Design
Slide -1
Practical Differential Pair Design
Dr. Eric Bogatin,
Signal Integrity Evangelist,
Bogatin Enterprises
www.beTheSignal.com
Presented at the Huntsville EMC Symposium, April 2010
 Bogatin Enterprises 2010
VL-155 Practical Differential Pair Design
April 2010
www.beTheSignal.com
Slide -2
Copyright © 2010 by
Bogatin Enterprises, LLC
All rights reserved. No material contained in
this presentation may be distributed or
reproduced in whole or in part without the
written permission of Bogatin Enterprises.
Please respect the great deal of effort that
has gone into the preparation of these
lectures and use these materials for your
personal use only.
Bogatin Enterprises, LLC
26235 W. 110th Terr.
Olathe, KS 66061
v: 913-393-1305
f: 913-393-0929
e: info@BeTheSignal.com
www.BeTheSignal.com
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -3
For More Information
www.BeTheSignal.com
Signal Integrity Certification Programs
Continuing Education Curriculums
Signal integrity public classes
No Myths Allowed webinar series
Streaming recorded lectures
Hands on labs
Feature articles and columns
SI-Insights quarterly publication
Monthly Pop Quiz
Published by Prentice Hall, 2009
My Blog: What I learned this month
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -4
Outline
• Download a copy of the presentation from
beTheSignal.com: under SI content, select
“slides presentation”, PPT-VL-155
• Design Methodology
• Problems to avoid
• Decision factors for coupling
• Exploring Design Space
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -5
Pop Quiz
• Which is better:
Tightly coupled diff pairs?
Loosely coupled diff pairs?
 Bogatin Enterprises 2010
VL-155 Practical Differential Pair Design
www.beTheSignal.com
Slide -6
What is the most common answer to
all SI questions?
“>it depends”
We answer it depends questions by
“putting in the numbers” with analysis:
Rules of thumb
Approximations
Numerical simulation tools
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -7
A Secret to Minimize Confusion
About Differential Impedance
Differential mode
Think:
Differential signals
Common signals
Odd mode
Even mode
Common mode
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -8
Differential and Common Signals
2.0
1.8
V1
Typical LVDS levels
1.6
V2, V
V1, V
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0
2
4
6
8
1.5
V2
Vdiff = V1 – V2
Vcomm = ½ (V1 + V2)
12
14
16
18
20
common
1.0
Vdiff
Vcomm
• Definitions:
10
time, nsec
0.5
differential
0.0
-0.5
0
2
4
6
8
10
12
14
16
18
20
time, nsec
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -9
Every Pair of Signals Has a
Differential and Common Component
2.0
V2, V
V1, V
1.8
1.6
•
Differential and common signals
propagate independently and DO
NOT Interact on the board
•
They each see a different
electrical environment:
Added Skew = RT to one line
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0
2
4
6
8
10
12
14
16
18
20
time, nsec
Diff and comm impedance
Diff and comm prop velocity
Diff and comm attenuation
Diff and comm cross talk
1.5
common
Vdiff
Vcomm
1.0
0.5
•
But>.
Any line to line asymmetry will convert
diff into comm signal and vis versa
differential
0.0
-0.5
0
2
4
6
8
10
12
14
time, nsec
16
18
20
Definitions:
Vdiff = V1 – V2
Vcomm = ½ (V1 + V2)
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -10
A New Design Methodology to Eliminate SI Problems
Before the Design is Released
• An efficient methodology:
Find the root cause
Establish design guidelines to minimize
them- balancing tradeoffs
}
“correct by design”: use analysis tools
to develop pre-layout design rules
specific to your design
Understand the essential
principles
pitch
0.20
NEXT, fraction
Identify the SI problems
0.15
0.10
0.05
0.00
Use post layout analysis tools to verify
the final design
 Bogatin Enterprises 2010
-20
-15
-10
-5
0
5
10
Center to Center Pitch, mils
www.beTheSignal.com
15
20
VL-155 Practical Differential Pair Design
Slide -11
Practical Design Considerations
Performance
(meet specs)
Cost factors:
•
Always do what is free
•
Explore design space with simple estimates, then more accurate
analysis
•
Explore cost- performance trade offs with “virtual prototypes”
•
The most difficult tradeoffs: higher component cost for lower
system cost
•
Consider product lifetime performance
•
Manage risk: buy “insurance” by adding design margin
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -12
Establish a Design Guideline by
Applying the “Youngman Principle”
Read www.bethesignal.com/blog, Nov 9, 2008
“If your arm hurts when you raise it, don't raise your arm.”
“If problem A happens when your design has feature B,
then eliminate feature B from your design”
Identify the root cause of a problem and fix the root cause
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -13
Four Chief Problems to Manage in
High Speed Serial Links
0.8
Eye_uniform.Density
• Losses
Conductor loss
Dielectric loss
• Ripple
Impedance mismatches from:
TX, channel, vias, connectors,
RX
What you think you have
0.6
0.4
0.2
0.0
-0.2
0
50 100 150 200 250 300 350 400
time, psec
0.8
• Noise: cross talk
Diff to diff and comm to diff
coupling
• Mode conversion
What you actually have
Line to line asymmetry
Mantra: “losses, ripple, noise, mode conversion”
 Bogatin Enterprises 2010
VL-155 Practical Differential Pair Design
www.beTheSignal.com
Slide -14
Design Solution Options to Achieve
Performance Goals
•
•
•
•
•
Lowest attenuation
Low dissipation factor laminate
Lowest Dk laminate
Wide lines
Smooth copper
Lower impedance
1
Lowest ripple noise
Controlled impedance to a target impedance
Lower target impedance to match lower via impedance
2
Lowest cross talk
Avoid microstrip
Large spacing between channels
Tight coupling when return path is screwed up
Lowest mode conversion
Matched length, or length compensation
Matched cross section lines
Mitigate glass weave skew
Lowest cost features
FR4
Highest interconnect density
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -15
A “Hidden Variable” to Real World
Performance
• Identical boards from different suppliers
• Very different insertion loss: 2x difference- why?
example courtesy of Cisco
 Bogatin Enterprises 2010
VL-155 Practical Differential Pair Design
www.beTheSignal.com
Slide -16
Solutions are Available for
Smoother Copper
Cost will be driven by the market
If you do not ask for it, there will be
no market need
The higher the volume the lower the
cost
(dual flat foil)
Push your fab vendors for:
1. rms roughness characterization data
2. Smoother copper foil
Courtesy of John Andresakis
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -17
Tight or Loose Coupling?
Performance
(meet specs)
Cost factors:
• Performance drivers:
• Cost drivers:
Target impedance
Fewest layers
Widest line width
Lowest cost laminate
Channel to channel cross talk
Highest interconnect density
Glass weave pitch
Narrowest line that is free
Tightest pitch that is high yield
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -18
What is NOT influenced by Coupling
Symmetric,
uncoupled lines
make a perfectly
good differential
pair
uncoupled
tightly coupled
HyperLynx 8.0
•
The degree of coupling has NO impact on reflections or
mismatch
•
The differential signal only sees the differential impedance Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -19
Microstrip:
Differential Impedance and Coupling
Increasing coupling decreases
differential impedance
 Bogatin Enterprises 2010
VL-155 Practical Differential Pair Design
www.beTheSignal.com
Slide -20
Compensate Line Width for Separation to
Keep Differential Impedance Constant
h = 3.5 mils
h = 2.7 mils,
Dk = 4
This line defines design
space for 100 Ohm
differential impedance
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -21
Attenuation and Line Width
R 
S21Len = −4.34 x  Len  dB/inch
L in Ohms/in
 Z0  R
Z in Ohms
For ALL transmission lines:
0
w
b
•
Consequence of wider line width, w:
Lower Resistance
If impedance is constant, lower attenuation
Need thicker b to keep impedance constant
•
What if dielectric thickness, b is fixed? What is impact of wider w?
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -22
S21Len
b
Characteristic Impedance, Ohms
Attenuation and Line Width
R 
= −4.34 x  Len 
 Z0 
w
80
@ 1 GHz
70
Optimum impedance for lowest
conductor loss ~ 35 Ohms
60
50
40
30
20
10
0
0
Stripline, ½ oz copper
Fixed total thickness b = 12.2 mils
5
10
15
20
25
30
35
40
45
50
Line Width, w, mils
Which has lower S21, narrow or wide w?
Insertion Loss, dB/inch
0.00
- Increasing line width, decreases R,
decreasing S21
- Increasing line width decreases Z0,
increasing S21.
-0.05
Conductor loss, rms = 0
-0.10
-0.15
-0.20
0
5
Lower Z0 than 50 Ohms is lower loss
 Bogatin Enterprises 2010
10
15
20
25
30
35
40
45
50
Line Width, w, mils
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -23
How Else to Enable Wide Lines,
And Tight Pitch?
•
Compromise: Loose coupling
s = 2 x w, w = 5 mils
Dk = 4
•
Lower target impedance
•
Thicker H1 = H2 = 13 mils
85 Ohms is target in PCIeII
s = w = 5 mils
Dk = 4
•
Lower Dk = 3.1
s = w = 5 mils
H1 = H2 = 6.5 mils
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -24
Worse Case: Far End Cross Talk in
Microstrip: Single-ended to Diff
Differential Noise, fraction
0.05
(non-interleaved)
0.00
-0.05
-0.10
-0.15
-0.20
-0.25
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
100 Ohm diff
5 mil line, spacing
RT = 100 psec
Len = 10 inches FEXT
time, nsec
-
0
aggressor
~ 5 dB reduction in cross talk from tightest coupling
Differential cross talk from common sources can be -20 dB!
FarEnd
Far
EndNoise,
Noisefraction
in dB
victim
Len
RT
spacing
Coupling
+
~
Coupling cases:
Uncoupled: s = 3 x w
Loose: s = 2 x w
Tight s = w
-10
-20
-30
-40
-50
FEXT often limits max trace length in PCIe to < ~16 inches
5
10
15
20
25
30
35
40
Spacing, mils
 Bogatin Enterprises 2010
www.beTheSignal.com
45
50
VL-155 Practical Differential Pair Design
Slide -25
Worst Case Near End Cross Talk in
Stripline
Coupling
+
Which is more important
influencing NEXT: coupling or
spacing?
spacing
aggressor
~ 1 dB reduction in cross
talk from tightest
coupling
Reduce cross talk by
increasing spacing to
aggressor!
Near End Cross Talk, in dB
victim
0
-10
For less than -50 dB xtk,
keep spacing > 3 x w
-20
-30
-40
-50
-60
Three coupling cases:
Tight s = w
Loose: s = 2 x w
Uncoupled: s = 3 x w
-70
-80
-90
-100
0
1
2
3
4
Spacing/line width
 Bogatin Enterprises 2010
VL-155 Practical Differential Pair Design
www.beTheSignal.com
Slide -26
Coupling and Differential Cross Talk
•
When the return path is a wide, uniform plane, tighter
coupling has little impact on differential cross talk ( ie, in
controlled impedance board traces)
•
When the return path is not a uniform plane, tighter
coupling can dramatically decrease differential cross
talk
•
Always use tight coupling between lines in a differential
pair when the return path is not a wide uniform plane:
Gaps
Vias
Connectors
Leaded, 2 layer packages
Sockets/interposers
Flex/ribbon cable
 Bogatin Enterprises 2010
www.beTheSignal.com
5
VL-155 Practical Differential Pair Design
Slide -27
Local Dk Variation Causes
“Weave Induced Skew”
Worst case if pitch = (1/2 + n) x glass pitch
Typical glass weave pitch ~ 15-25 mils
1080, 2116 are 17 mils pitch
Best case is if routing pitch matches glass weave
pitch: ~16-20 mils
8 mil wide line, 17 mil pitch
Brist et al. PCD&F Nov 2004
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -28
Time Delay Measurements of
Different Traces
8 inch long Stripline in
2116 glass
straight
8 psec in 8
inches ~ 1
psec/inch
zig-zag
Courtesy of Altera
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -29
Measured Skew in 4 inch Test Lines
Photo courtesy of Jeff Loyer, Intel
-
+
Glass Dk ~ 6
Resin Dk ~ 3
Higher local Dk
Slower speed
Longer delay
40,000 TDR measurements
Lower local Dk
higher speed
shorter delay
Typical glass weave skew ~ 2.5
psec/inch
Worst case glass weave skew maybe
~15 psec/inch
Courtesy of Jeff Loyer, Intel Corp.
Typical case: 20 inches x 2.5 psec/inch = 50 psec skew.
Possible problem for > 2 Gbps
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -30
Which is Better:
Tight or Loose Coupling?
If loss is important,
consider using
loose coupling
Lower Conductor Loss
Higher Interconnect Density
If interconnect density is
most important, always
use tight coupling
tight
Sweet spot s ~ 2w
 Bogatin Enterprises 2010
loose
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -31
Which is Better,
Tight or Loose Coupling?
It depends:
•
Why loose coupling:
Lower loss
Risk reduction for glass weave skew mitigation
•
Why tight coupling
Higher interconnect density
Lower cost
•
What is not critical
Differential impedance control
Channel to channel cross talk
•
What else:
Lower Df
Lower Dk
Smoother copper
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -32
Practical Guidelines
• If bit rate is < ~ 1 Gbps
Loss not an important driver
Always consider tight coupling
• If bit rate is > ~ 5 Gbps
Loss very important
Consider looser coupling
Route on a pitch equal to the glass weave pitch
• Regardless of bit rate, always do your own analysis
 Bogatin Enterprises 2010
www.beTheSignal.com
VL-155 Practical Differential Pair Design
Slide -33
The End!
www.BeTheSignal.com
Signal Integrity Certification Programs
Continuing Education Curriculums
Signal integrity public classes
No Myths Allowed webinar series
Streaming recorded lectures
Hands on labs
Feature articles and columns
SI-Insights quarterly publication
Monthly Pop Quiz
Published by Prentice Hall, 2009
My Blog: What I learned this month
 Bogatin Enterprises 2010
www.beTheSignal.com