You are on page 1of 36

Crosstalk Measurements for Signal Integrity

Applications
Outline

ı What is crosstalk?
 A brief history of crosstalk
 Definition of crosstalk
 Why is crosstalk important?
 Types of crosstalk
 Impact of crosstalk on signal integrity

ı Measurement Methods for crosstalk


 Time domain measurements
 Frequency domain measurements

28.01.2014 Footer: >Insert >Header & Footer 2


What is Crosstalk?

28.01.2014 Footer: >Insert >Header & Footer 3


A Brief History of Crosstalk

ı Crosstalk terminology form telephone lines


 Communication crosses line from the intended user to
a victim
 Crosstalk if frequency dependent
 Significant contribution to crosstalk identified to be
telephone circuit unbalances
 “Crosstalk set” measures near-end crosstalk of
telephone line at audio frequencies
 Source: L.P. Ferris, R. G. McCurdy: “Telephone Circuit
Unbalances. Determination of Magnitude and
Location”, Pacific Coast Convention of the
A.I.E.E., 1924

28.01.2014 Footer: >Insert >Header & Footer 4


Signal Integrity Problems

Transmission line effects


ı Delay
ı Rise time degradation
ı Attenuation
ı Skin effect
ı Overshoots, undershoots,
ı Ringing
ı Reflections
ı Crosstalk

Other effects
ı Skin losses, via stubs, connectors
ı Proximity effects
28.01.2014 Footer: >Insert >Header & Footer 5
Signal Integrity Problems

ı Effects that are hard (impossible?) to model


 Inherent process variations
 Metal roughness
 Non-ideal skin effects
 Component dielectrics
 How connectors are soldered to the board
 Broadside coupling of signals

28.01.2014 Footer: >Insert >Header & Footer 6


A Systems View
BER = 0.5

Slope of bathtub
curve
BER

Dj Tj Dj

0 1
RjBER RjBER
TjBER = Dj + alphaBER * Rj
Where alpha is related to the slope of the bathtub curve
LeCroy 2008 7
Eye Diagram

Eye Opening
Jitter Some authors believe they can identify
crosstalk by analyzing the eye diagram
Source: Jung CICC 2012

Source: Centric Technologies’ Wireless Cable

28.01.2014 Footer: >Insert >Header & Footer 8


Definition of Crosstalk

ı Crosstalk is the interference between signals that are propagating on various lines in the
system.
ı Crosstalk results from the interaction of electromagnetic fields generated by neighboring
data signals as they propagate through transmission lines and connectors.

28.01.2014 Footer: >Insert >Header & Footer 9


Why is Crosstalk Becoming Important?

ı A thought experiment

Ideal Transmission
Serial Bus Line
Two serial busses in parallel

Add Capacitive Coupling

28.01.2014 Footer: >Insert >Header & Footer 10


Why Crosstalk is Becoming Important

ı Two views of the world:


 Time Domain: Send step into transmission line, see what comes out at the other end

28.01.2014 Footer: >Insert >Header & Footer 11


Why Crosstalk is Becoming Important

ı Two views of the world:


 Frequency Domain: Send step into transmission line, see what comes out at the other
end

28.01.2014 Footer: >Insert >Header & Footer 12


Types of Crosstalk

28.01.2014 Footer: >Insert >Header & Footer 13


Types of Crosstalk

ı Near-end Crosstalk (NEXT)


 The noise induced in the receiving pair due to the signal on the transmitting pair on the
same port. (Source: IEEE1394) Source: Jung CICC 2012

ı Far-end Crosstalk (FEXT)


 The noise induced in the receiving pair due to the signal on the transmitting pair on the
same port. Source: Jung CICC 2012

 Slow transitions  less FEXT

ı Crosstalk Induced Jitter (CIJ)


 Odd mode and even mode have different propagation velocity
 Independent of rise/fall times and signal amplitude
Source: Buckwalter SSC 2006

28.01.2014 Footer: >Insert >Header & Footer 14


Near End Crosstalk (NEXT)

ı Single-Ended Coupled Microstrip


ı NEXT coefficient Kb
ı VNEXTK V (t ) V (t 2t t : propagation time through the trace
b in
f in f

CM LM
ı
1
Kb ( )
4
C Total L Total CM, LM: Mutual Capacitance, Inductance per unit length
CTotal, LTotal: Total Capacitance, Inductance per unit length

(Source: Sohn, Advanced Packaging V24(4), 2001)

28.01.2014 Footer: >Insert >Header & Footer 15


Far End Crosstalk (FEXT)

ı Single-Ended Coupled Microstrip


ı FEXT coefficient Kf
ı V Kt
FEXT
d
V (t t ) tf: propagation time through the trace
f f in f
dt
CM LM
ı
1
Kf ( )
2
C Total L Total CM, LM: Mutual Capacitance, Inductance per unit length
CTotal, LTotal: Total Capacitance, Inductance per unit length

(Source: Sohn, Advanced Packaging V24(4), 2001)

28.01.2014 Footer: >Insert >Header & Footer 16


Types of Crosstalk

ı Alien Crosstalk (AXT)


 Crosstalk within a group or bundle of cables
 Alien Near-End Crosstalk (ANEXT) (IEEE 802.3 terminology)
 Alien Far-End Crosstalk (AFEXT) (IEEE 802.3 terminology)

ı Power sum near end crosstalk (PSNEXT)


 power sum of NEXT of all other wire pairs on crosstalk in one pair (in UTP cables)

ı Equal Level Far end crosstalk (ELFEXT)


 FEXT minus attenuation of cable

ı Power Sum Equal Level Far end Crosstalk (PSELFEXT)


 power sum of ELFEXT of all other wire pairs on crosstalk in one pair (in UTP cables)

28.01.2014 Footer: >Insert >Header & Footer 17


Types of Crosstalk(con’ed)

ı ICR: Insertion crosstalk ratio


 ICR = |IL – PSXT|
 Similar to PSELFEXT but includes NEXT

ı ICN: Integrated crosstalk noise


 Takes into account spectrum of excitation signal (Source: Sercu, DesignCon 2010)

28.01.2014 Footer: >Insert >Header & Footer 18


Sources of Crosstalk

Source: Mukherjee, ECTC, 2013


Source: Wu, EMC V55(4), 2013

ı Crosstalk happens even in ideal transmission lines


ı Crosstalk on TSVs (through silicon vias)
ı Crosstalk in packages Source: Lim, EMC V55(4), 2013

ı Launch pattern for BGAs


ı Crosstalk on vias in PCB,
ı Crosstalk through difference in propagations velocity of different modes in coupled
stripline/microstrip

Source: Hsu, ECTC, 2012

28.01.2014 Footer: >Insert >Header & Footer 19


Sources of Crosstalk

ı Increased number of features in computers systems


ı Data rates increase

ı Board size decreases (or stays the same)

ı Need for “right the first time”

28.01.2014 Footer: >Insert >Header & Footer 20


Crosstalk Measurements

28.01.2014 Footer: >Insert >Header & Footer 21


Typical VNA measurements for SI Engineers

ı Insertion loss
 Common mode and differential
ı Return loss
 Common mode and differential
ı Crosstalk

ı Within Channel measurements


 insertion loss, return loss, pair-to-pair near-end crosstalk loss (NEXT), power sum
NEXT loss, pair-to-pair attenuation to crosstalk ratio, far-end (ACRF), power sum
ACRF, return loss, and delay
ı Between Channel measurements
 alien crosstalk parameters, power sum alien attenuation to crosstalk ratio, far-end
(PSAACRF) and power sum alien NEXT

28.01.2014 Footer: >Insert >Header & Footer 22


Measurement Techniques

ı Bit error rate test-set


 Not suitable to evaluate the amount of Crosstalk
ı Real-time Oscillscopes
 Some information on crosstalk via eye diagram
 Statistical correlation between different source are conceptually possible
ı Time domain Reflectometry (TDR)
ı Frequency domain measurements
 Using Vector Network Analyzer

28.01.2014 Footer: >Insert >Header & Footer 23


Time Domain 1/2

ı Eye diagram
 Eye height
 Eye width
ı Advantages:
 Related to system performance
ı Disadvantages
 Not easy to figure out what part of eye closure is due to crosstalk
 In presence of ISI, crosstalk can not be identified
 Requites large data sets (can’t do PRBS31)
 No information on how to fix the problem

28.01.2014 Footer: >Insert >Header & Footer 24


Time Domain 2/2

ı TDR
 Measure reflected/transmitted energy and frequency content
 Based on equivalent time oscillosopes
ı Advantages:
 Intuitive Measure
 Information on impedance as function of electrical length
ı Disadvantages
 Not very accurate at high frequency
 Calibration is questionable
 Low dynamic range
 Repeatability

28.01.2014 Footer: >Insert >Header & Footer 25


Frequency Domain Measurements

ı S-parameters
 Frequency/phase response of a channel
ı Advantages:
 Highly accurate
 Large dynamic range
 Well-known calibration procedures/embedding/de-embedding
 Spatial information via IFT
 Up to 500+ GHz
ı Disadvantages
 Has reputation of being complicated

28.01.2014 Footer: >Insert >Header & Footer 26


How does a VNA work?

ı VNA consists of Generator, directional


element and receiver
ı Generator sends out pure sine-wave
ı Incident wave is measured with reference
receiver
ı Reflected wave is measured with one
measurement receiver
ı Transmitted wave is measured with another
measurement receiver

28.01.2014 Footer: >Insert >Header & Footer 27


Requirements Reflectometer 2

Meas. Receiver

Ref. Receiver PORT 2

ı In principle, 2n port device can me measured with a 2-


Bias Tee

Reflectometer 4

DUT
Meas. Receiver

port VNA Ref. Receiver PORT 4

Bias Tee

 True differential measurements require at least 4- Reflectometer 1

Meas. Receiver

ports with 2 coherent sources Ref. Receiver


PORT 1

Bias Tee

 Modern VNAs provide up to 48 ports via switch matrix Reflectometer 3

ı Accuracy of models depends on accuracy of S-


Meas. Receiver

Ref. Receiver PORT 3

parameter measurements Bias Tee

 Stability of setup is crucial


 Connecting/reconnecting cables is error prone

28.01.2014 Footer: >Insert >Header & Footer 28


Are TDR and VNA measurements equivalent?

ı TDR instruments are a lot easier to set up, why bother with a VNA?
ı Arguments for TDR
 Easier to set up/use
 Cheaper
ı Arguments for VNA
 At high speed much lower uncertainty (TDR @50 GHz: 12 dB uncertainty)
 TDR dynamic range: 35 dB, VNA: 100+ dB
 Sources of VNAs are much cleaner than for TDRs
 Can not adjust step amplitude of TDRs
 No bias-T option for TDRs
 TrueDifferential
 Sophisticated calibration procedures
 Easy to de-embed probes, cables, fixtures

28.01.2014 Footer: >Insert >Header & Footer 29


How to handle crosstalk

28.01.2014 Footer: >Insert >Header & Footer 30


How to Reduce Crosstalk

ı Design:
 Increase spacing between traces
 Guard traces, serpentine microstrip lines, spiral layout
 Segmented transmission lines using Genetic Algorithms (Seki, EDAPS 2012)
 High quality connectors
 Backdrilled VIAS
ı Compensation
 Active X-talk cancellation
 Amplitude (Pelard, JSSC, 2004)
 Timing
 TX side
 RX side

28.01.2014 Footer: >Insert >Header & Footer 31


Equalization Techniques

ı Various active/passive techniques proposed


ı Receiver side equalization
 Noise enhancement
ı TX side pre-emphasis
 Coupling of energy into adjacent channels

28.01.2014 Footer: >Insert >Header & Footer 32


Summary and Outlook

28.01.2014 Footer: >Insert >Header & Footer 33


System Level Approach

ı Much of today’s design flow is driven by systems specs


 PCB/component/package/device specs are not always well
defined
 Specs can be traded off against each other as long as
system requirements are met
ı Design margins are eroding
 Trend to higher speed and higher integration

ı Crosstalk is next frontier in conquering high-speed designs


 Nearly impossible to spec crosstalk on a systems level
 Successful designs will require integrated
modeling/characterization cycles that integrate crosstalk
Source: Mukherjee, ECTC, 2013
mitigation on device/package and PCB level

28.01.2014 Footer: >Insert >Header & Footer 34


For More Information
ı Download complete slide presentation via Slideshare

slideshare.net/rohdeschwarzNA

ı Access app notes, white papers and other supporting material via our
Twitter feed @RohdeSchwarzNA
• Download complete slide presentation via Slideshare

slideshare.net/rohdeschwarzNA

• Access app notes, white papers and other supporting material via our Twitter
feed
@RohdeSchwarzNA

You might also like