0% found this document useful (0 votes)
159 views52 pages

Signal Integrity in The Real World

The document outlines an agenda for a one-day seminar on signal integrity topics. It includes the following sessions: - Registration and breakfast from 8:00-8:45am - Three morning sessions on signal integrity applications, measurements, and test-as-a-service from 8:45-11:30am - Lunch from 11:30am-12:20pm - Two afternoon sessions on oscilloscope fundamentals and arbitrary waveform generation from 12:20-3:40pm - A wrap-up session from 3:40-3:45pm

Uploaded by

patarinw
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
159 views52 pages

Signal Integrity in The Real World

The document outlines an agenda for a one-day seminar on signal integrity topics. It includes the following sessions: - Registration and breakfast from 8:00-8:45am - Three morning sessions on signal integrity applications, measurements, and test-as-a-service from 8:45-11:30am - Lunch from 11:30am-12:20pm - Two afternoon sessions on oscilloscope fundamentals and arbitrary waveform generation from 12:20-3:40pm - A wrap-up session from 3:40-3:45pm

Uploaded by

patarinw
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 52

Duration Time

45m 8:00 am- 8:45 am Registration and Continental Breakfast


15m 8:45 am – 9:15 am Break: Product Fair and Ask the Experts
1hr 9:15 am – 10:15 am Signal Integrity in the real-world
1hr 10:20 am – 11:20 am Signal Integrity Measurements and Network Analysis
10m 11:20 am – 11:30 am Keysight Test-as-a-Service
50m 11:30am -12:20 pm Break: Giveaways and Lunch
1hr 12:20 pm – 1:20 pm High-Speed Oscilloscope Fundamentals
1hr 1:20 pm - 2:20 pm Fundamentals of Arbitrary Waveform Generation
20m 2:20 pm – 2:40 pm Break: Product Fair and Ask the Experts
1hr 2:40pm – 3:40 pm Fundamentals of Bit-Error Ratio Testing
3:45 pm Wrap Up and Giveaways

2
Tim Wang Lee 2018.04.19
SI/PI Application Engineer
Transmitter Receiver

Channel

4
CPU Graphic card Cable On board video processor LED display

Interconnects

Signal integrity is about the problems interconnects


introduce and how to avoid them.
– Dr. Eric Bogatin
5
The case of the failing virtual channel

1 Evaluate the channel

2 Investigate the channel

3 Understand the root cause

4 Solve the problem


6
Data rate: 32 Gbps
Nyquist: 16 GHz

Channel

7
PRBS: Pseudo-Random Bit Sequence
By sending PRBS, we are testing how the channel affects all the possible transmitted data pattern.
PRBSX: The 2x-1 pseudo-random binary sequence combines every permutation of x bits.

1 0 1 1 1 1 0 0 1 1 0 1 1 1 0 1 1 1 0 0 1 0 1 0 1 0 0 1 0 1

8
2UI = 0.2 nsec 2UI = 0.2 nsec 2UI = 0.2 nsec

9
10
PRBS at the transmitter Received PRBS at Receiver

11
Transmitter Receiver

Data rate: 32 Gbps


Nyquist: 16 GHz

12
The eye is closed.
The channel corrupts the PRBS pattern
significantly so that the receiver cannot
correctly detect 1 or 0.

13
3-inch microstrip line cross-section

14 mil
3-inch microstrip line
W
H

Rule of thumb: W/H = 2


Via structure FR4 = ~50 Ohm

3-inch stripline
3-inch stripline cross-section

7.5 mil

14
Date Rate: 32 Gbps
Nyquist Frequency: 16 GHz
Estimated Loss (FR4): ~ 0.1 dB/in/GHz
Estimated Delay (FR4): 6 in/nsec

3-inch 3-inch
Structure Via
microstrip stripline
Round Trip Delay
(nsec)
Estimated Loss
(dB at Nyquist)
Impedance
(Ohm)
15
Date Rate: 32 Gbps
Nyquist Frequency: 16 GHz
Estimated Loss (FR4): ~ 0.1 dB/in/GHz
Estimated Delay (FR4): 6 in/nsec

3-inch 3-inch
Structure Via
microstrip stripline
Round Trip Delay
1
(nsec)
Estimated Loss
5
(dB at Nyquist)
Impedance
<50
(Ohm)
16
Date Rate: 32 Gbps
Nyquist Frequency: 16 GHz
Estimated Loss (FR4): ~ 0.1 dB/in/GHz
Estimated Delay (FR4): 6 in/nsec

3-inch 3-inch
Structure Via
microstrip stripline
Round Trip Delay
1 1
(nsec)
Estimated Loss
5 5
(dB at Nyquist)
Impedance
<50 ~50
(Ohm)
17
Date Rate: 32 Gbps
Nyquist Frequency: 16 GHz
Estimated Loss (FR4): ~ 0.1 dB/in/GHz
Estimated Delay (FR4): 6 in/nsec

3-inch 3-inch
Structure Via
microstrip stripline
Round Trip Delay
1 Small 1
(nsec)
Estimated Loss
5 Small 5
(dB at Nyquist)
Impedance
(Ohm)
<50 ? ~50
18
Open

19
Open
50 Ohm <50 Ohm ~50 Ohm

50 Ohm ?

0.5 nsec 1.5 nsec 1.6 nsec 2.6 nsec

3-inch 3-inch
Structure Via
microstrip stripline
Round Trip Delay
(nsec)
1 Small 1
Impedance
(Ohm)
<50 ? ~50
20
3-inch 3-inch
Structure
microstrip stripline
Round Trip Delay
1 1
(nsec)
Impedance
<50 ~50
(Ohm)

Structure Via

Round Trip Delay


< 0.1 nsec
(nsec)
Impedance
~35
(Ohm)

21
• Rise Time
• Data rate (Unit Interval) 0

 : Reflection Coefficient 0


Z 2  Z1

Z1  Z 2
22
There are two impedance
discontinuities in the via structure.

23
Low impedance (Z < 50 Ohm)
Gamma < 0
Reflection is negative

High impedance (Z >> 50 Ohm)


Gamma =1
Reflection is positive 24
Channel length

Expect signal at ~1.25 nsec

25
Date Rate: 32 Gbps
Unit Interval: 31.25 psec

Inter-symbol Interference:
Also known as ISI, it’s when
current symbol (pulse) is
interfering adjacent symbols.

One Unit Interval

26
Physically, channel has a low pass response
and is attenuating higher frequencies more
than lower frequencies.

Fourier Transform

Equalization techniques aims to redistribute


the frequency content reduce ISI.
27
Freq = f0

S21 (Port 2 excited by port 1)


• Transmission coefficient
• Insertion loss
S11 (Port 1 excited by port 1)
• Reflection coefficient
• Return loss

Expectation:
• From impedance profile, S11
should be below -20 dB.
• From single pulse response,
S21 should decrease with
frequency.

Not what we expect!

28
3-inch microstrip line 3-inch 3-inch
Structure Via
microstrip stripline
Estimated Loss
Via structure 5 Small 5
(dB at Nyquist)

3-inch stripline Simulated Loss


4.2 4 5.2
(dB at Nyquist)

29
3-inch microstrip line

3-inch stripline

Conclusion so far:
After removing the via
structure and simulate, we
find that via structure is the
root cause of eye closing.

30
Side view
Data rate: 32 Gbps
Nyquist: 16 GHz
Bandwidth: 5*16 = 80 GHz
Wavelength: 6 in/nsec/80 ~ 75 mil

75 mil

~75 mil

Transmission line:
Voltages and currents can vary
in magnitude and phase over
physical length.
31
Two conductors make a transmission line
Two important (uniform) transmission line parameters
1. Characteristic Impedance (Z0)
2. Time Delay (TD)
Top View

Cross-section View
W
TD

Dk H

1 H W: width of trace
Z0   H: height of substrate Len
C Dk  W Dk: dielectric constant TD  Len: length
Vprop: velocity of propagation
V prop
32
Top View

 : Reflection Coefficient
Z 2  Z1

Z1 = 50
Z1  Z 2

Z2
Z2
Z2 short Z2 < 50 Z2 = 50 Z2 >50 Z2 Open
(Ohm)

Γ -1 Negative 0 Positive 1

33
Top View
V

34

Delay 
 2
Len 
4
V

 1
35
V


Len 
4
V

 1
36
V


Len 
4
Quarter-wave stub resonance
At frequency where the physical length if the stub is a
quarter of a wave length, it seems like nothing is being
transmitted.

 1
37
 v v in
len     4  len f res   v 6
4  4  len nsec
1.5
f res (GHz)  For FR4, expect fres = 20 GHz
len (in)
Len ~ 75 mil

Stub resonance
Via

38
Len = 72.8 mil

Expect fres = 20.6 GHz

39
Tline model

Via The Root Cause:


We know via stub is the root cause
of eye closing after investigating
transmission line stub resonance
and modeling the via structure with
transmission lines.
40
Via stub
removed
Via Stub

41
Channel with via stub

Channel with via stub


removed

42
43
• Use the channel (with via stub) to transmit data at lower data rate
Date Rate: 32 Gbps Date Rate: 18 Gbps
Nyquist Frequency: 16 GHz Nyquist Frequency: 9 GHz

Loss at Nyquist: ~14 dB Loss at Nyquist: ~6 dB

44
45
Channel frequency response: Equalization:
attenuates higher frequencies more than Distributes the frequency-dependent loss
lower ones. evenly throughout frequencies.

• Continuous Time Linear Equalization (CTLE)


• Feed-forward Equalization (FFE)
• Decision Feedback Equalization (DFE)

CTLE FFE DFE


Linear Linear Non-linear
Freq. Domain Time Domain Time Domain
Analog Digital Digital
Post-cursor Pre and post- Post-cursor
cursor

46
Cursor

Pre-cursor Post-cursor
Visualization of EQ taps:
Using the maximum value as
the main cursor, the cursor
diagram tells us how many pre-
and post- cursor taps we need
for equalization.

47
Decision Feedback Equalizer:
At the arrival of received data
(symbols), DFE algorithm detects
and makes a decision. Assuming
the decision is correct, proper tap
values are chosen and feedback to
the originally received data.

48
49
1. Evaluate the channel 2. Investigate the channel

4. Solve the problem


3. Understand the root cause

50
Calibration

Measuring a 32-port channel

Fixture and Fixture Removal

51
https://tinyurl.com/y9wah5ak

S-parameters: Signal Integrity


Analysis in the Blink of an Eye
https://tinyurl.com/ybqobvyv
https://youtu.be/FYelztbksSo

https://tinyurl.com/ycmbvbgx
https://community.keysight.com/people/timwanglee/content

http://www.electronicdesign.com/embedded-revolution/xip-linux-new-spin-embedded-architecture
http://www.hisdigital.com/us/product2-720.shtml
Picture credit: https://sonyglobal.scene7.com/is/image/gwtprod/2fce0810e5f97ed5b68b4d4a4d6be7ad?fmt=pjpeg&bgcolor=FFFFFF&bgc=FFFFFF&wid=2515&hei=1320
https://auvidea.com/product/70901/
52
http://cdn8.bigcommerce.com/s-jte32/images/stencil/1280x1280/products/227694/231045/led-1366-768-15.6-(tg)__47702.1480641666.jpg?c=2&imbypass=on

You might also like