Title: Transmitter Jitter Basics: The World of Test
1Transmitter Jitter Basics The World of Test
PC Bus
Steve Sekel Director of Product
Management SyntheSys Research, Inc
2Agenda
- So Whats the Problem?
- Jitter Eye Closure Review
- Side Notes
- The Effect of Timing Reference
- Measurement Depth
- Common Methods
- Eye Diagram Jitter
- Mask Testing
- Bathtub/ BERTScan/ Jitter Peak
- Deep Mask Testing
- Jitter Spectrum
- Summary - Tying it All Together
PC Bus
3Agenda
- So Whats the Problem?
- Jitter Eye Closure Review
- Side Notes
- The Effect of Timing Reference
- Measurement Depth
- Common Methods
- Eye Diagram Jitter
- Mask Testing
- Bathtub/ BERTScan/ Jitter Peak
- Deep Mask Testing
- Jitter Spectrum
- Summary -Tying it All Together
PC Bus
4Overview
- Serial Data Systems share common design
challenges, regardless of the transport standard - PCI-Express
- SATA
- 10GbE
- FibreChannel
- OTN
- .
- Vendor Independent Interoperability is assure
with compliance standards - Much of these specifications focuses in two
areas - Transmitter (Tx)
- Receiver (Rx)
5Whats the Problem?
- Communications Systems
- Its all about the Error Performance
- Systems typically specified to work with 1x10-12
BER or better - Waveform edges must not interfere with the
receiver decision point, or bit errors will occur
Good BER Performance Eye
Receiver decision point
- Problem is how to verify that components
systems can achieve good error performance when
an error every few billion bits is unacceptable.
BER Level
6Aim of Transmitter Testing.
7Agenda
- So Whats the Problem?
- Jitter Eye Closure Review
- Side Notes
- The Effect of Timing Reference
- Measurement Depth
- Common Methods
- Eye Diagram Jitter
- Mask Testing
- Bathtub/ BERTScan/ Jitter Peak
- Deep Mask Testing
- Jitter Spectrum
- Summary - Tying it All Together
PC Bus
8Jitter / Eye Closure Overview
Ideal Edge Placement
Jitter Deviation in time from ideal transition
placement
Jitter Amplitude Displacement in time from
ideal transition placement.(Units generally rms)
Number of edges reaching given amplitude
General Distribution Higher jitter amplitudes
occur less frequently
t
Leading
Lagging
9Simple Jitter Decomposition
Total Jitter (TJ)
Random (RJ)
Deterministic (DJ)
-?
?
-?
?
t
t
Leading
Lagging
Leading
Lagging
- Can be related to data pattern, crosstalk or
other system events - Bounded peak magnitude truncated at some level
- Longer measurement, result isnt any bigger
- Typically noise related
- Unbounded
- Longer measurement, bigger answer
10Jitter Effect on Eye Closure
- Eye diagrams generated from small numbers of
samples can look clean - But usually close when more data is taken
- Factors determining amount of closure
- Magnitude of Random Jitter Component
- Magnitude of Deterministic Jitter Components
- Amount of data which is sampled (measurement time)
11Agenda
- So Whats the Problem?
- Jitter Eye Closure Review
- Side Notes
- The Effect of Timing Reference
- Measurement Depth
- Common Methods
- Eye Diagram Jitter
- Mask Testing
- Bathtub/ BERTScan/ Jitter Peak
- Deep Mask Testing
- Jitter Spectrum
- Summary - Tying it All Together
PC Bus
12What You Measure Depends Upon Your Timing
Reference
2.
1.
3.
Some tracked Some not
Clock Recovery
Clock Recovery
Clock Recovery
Analyzer
Analyzer
Analyzer
a.
- Example Timing Experiment
- 3 Different Measurement Setups (1,2,3)
- Two Different Test Signals (a,b)
b.
(Tracked region)
13What You Measure Depends Upon Your Timing
Reference
2.
1.
3.
Some tracked Some not
Clock Recovery
Clock Recovery
Clock Recovery
Analyzer
Analyzer
Analyzer
1a.
2a.
3a.
1b.
2b.
3b.
See SOME jitter, depends on Jitter frequency
See NO jitter
See ALL jitter
(Tracked region)
14What You Measure Depends Upon Your Timing
Reference
3.
Measured jitter results can vary significantly
depending on the method of triggering. This
applies equally to sampling scopes, BERTs,
real-time scopes etc..
Some tracked Some not
A.
Clock Recovery
Analyzer
Most standards require case 3, a golden PLL
(defined response clock recovery trigger) for
transmitter testing the intention is to
emulate the receiver, and to only have jitter
appear in the measurement that the receiver
cannot track out.
3a.
B.
3b.
C.
See SOME jitter, depends on Jitter frequency
15Test Equipment The Limitations of Shallow
Sampling Views
1.
2.
3.
Clean
Marginal
Closed
?
?
Very different openings at depth
Look very similar on a shallow view
4.
5.
16Eye Diagrams Sample Depth
10 seconds Traditional sampling
scope
1.
It is hard to see what is going on at 10-12 from
a shallow eye diagram. Even higher sampling
rate eye diagrams do not reach down to 10-12
levels.
4.
5.
10 seconds Higher sample
rate scope
2.
Increasing Visibility
lt1 minute BER Contour
(measured depth 10-8)
3.
Increasing Depth
17Agenda
- So Whats the Problem?
- Jitter Eye Closure Review
- Side Notes
- The Effect of Timing Reference
- Measurement Depth
- Common Methods
- Eye Diagram Jitter
- Mask Testing
- Bathtub/ BERTScan/ Jitter Peak
- Deep Mask Testing
- Jitter Spectrum
- Summary - Tying it All Together
PC Bus
18Eye Diagram Jitter
2.
4.
Samples used to form histogram of crossing point.
Edges fall in a variety of places
t
Samples
3.
Eye Diagram Jitter
t
Samples
5.
Samples are taken of edge placements at the
crossing point.
V
Measured Example
6.
1.
Examining the eye diagram crossing point.
19Eye Diagram Jitter Issues
Increasing measurement time
Apparent Opening
1.
2.
3.
4.
Beyond traditional sampling scopes in a
reasonable time. About limit of higher sampling
rate eye diagrams
Common for eye diagrams on sampling scopes,
measured in a few seconds or so.
The eye that standards would like to have
measured
5.
As more samples are taken, more of the actual
eye performance is uncovered.
6.
But which is the right jitter value?
20Eye Diagram Jitter Summary
Weaknesses
Strengths
- Answer changes over time
- Shallow - Doesnt have measurement depth to catch
rare events
- Easy measurement on common equipment.
- Gives quick, intuitive feel for signal quality.
Measurement Speed
- Quick for an answer that only shows gross
problems 10 seconds
Example Measurement Equipment
Example Standards
212. Traditional Mask Testing Basics
Example 10 GbE mask Ref. 802.3ae
Measurements
4.
Overshoot
2.
80
Rise Time
3.
Eye Jitter
1.
130
20
100
80
Generic Mask
Normalized Amplitude ()
50
Keep Out
2a.
Ensures no excessive overshoot
20
0
Keep Out
Ensures jitter isnt excessive
-20
1a.
0
.22
.375
.625
.78
1
Keep Out
Ensures rise fall times are fast enough
Normalized Time (Unit Interval)
3a.
222. Traditional Mask Testing
1.
232. Traditional Mask Testing Mask Margin
1.
Issue 1 How to ensure eye opening at depth is
good, when a mask test is shallow? Traditional
Answer Either the authors of a given standard
expand the mask size when it is written, or users
do to try to ensure only good transmitter pass.
3.
2.
(top bottom regions omitted)
242. Traditional Mask Testing Mask Margin
4.
1.
Issue 2 Good transmitters fail poor yield,
high costs. Example ISI dominated closure may
fail mask but not hurt deep eye opening much
Issue 3 Bad transmitters pass unhappy
customers. Example VCSELs can have large amount
of noise unbounded eye closure
2.
5.
X
7.
Unbounded Effect
Bounded Effect
(Could measure for longer, but this hardly dents
the issue)
6.
3.
252. Mask Testing Summary
Weaknesses
Strengths
- Answer changes over time
- Doesnt have measurement depth to catch rare
events - Standards dont specify points per waveform or
how many waveforms left to the engineer, no
standardization. - Need to use margin to try to guess whats going
on at depth - gamble
- Tests all around the eye diagram, including
- Eye Diagram Jitter
- Rise/Fall Times
- Overshoot
- Very common in standards
Measurement Speed
- Quick for an answer that only shows gross
problems 10 seconds plus
Example Measurement Equipment
Example Standards
- SONET/SDH
- Fibre Channel
- 1 10 GbE
263. Bathtub, BERTScan, Jitter Peak
Bathtub BERTScan (MJSQ)
d.
Jitter Peak
b.
- BERT decision point samples center of the eye,
measures BER. (a.) - Decision point is moved in time, BER measured
again. - Towards crossing point, measured BER increases.
(b.) - Jitter is measured as width of crossing point in
BER. - Resulting graph is bathtub curve (c. d.)
c.
BERT Decision Point
a.
273. Bathtub, BERTScan, Jitter Peak Summary
Weaknesses
Strengths
- Only jitter dimension measured
- Older equipment can give slow and inconsistent
results
- If measured using BER, based on deep sample
depth, little extrapolation needed to low BERs - Can give direct measurement of TJ
- One method of obtaining RJ/DJ separation
Measurement Speed
Equipment dependent, can be slower than shallow
eye diagram jitter measurements
Example Measurement Equipment
Example Standards
- BERTs
- BERTScope
- Estimates from some sampling scopes etc on
shorter patterns.
- MJSQ (and therefore Fibre Channel)
284. BER Contour A Path to Deep Mask Testing
4.
1.
3.
Can extend the bathtub idea by taking multiple
radial slices through the eye with the decision
point to cover all of the inside of the eye.
2.
294. Deep Mask Testing
Compare deep compliance mask with measured BER
Contour
2.
Traditional Mask
1.
Deep Mask
3.
4.
Measured XFI Device Example
304. Deep Mask Testing Summary
Weaknesses
Strengths
- Mandated devices must pass at depth, but not
specified how to test to get there - Arent many instruments that can measure this
directly
- Specifies eye opening down where it is needed, at
10-12 BER - Measures all around the inside of the eye like
conventional mask testing
Measurement Speed
- Not a fast manufacturing test 1 minute plus
Example Measurement Equipment
Example Standards
- OIF CEI
- XFI
- SFP electrical signal to module input
315. Jitter Spectrum
Lo F
1.
Hi F
Sampling for 1 minute for transmission
standards. Imagine edge positions recorded, then
FFTd Frequency spectrum gt80 MHz measured for
10Gb/s Many analyzers dont show spectrum, just
show answer.
2.
5.
4.
6.
Jitter (edge movement) frequencies
3.
325. Jitter Spectrum
Weaknesses
Strengths
- Jitter Test Only
- Traditionally, repeatability issues from
instruments with higher than ideal intrinsic
jitter
- Gives idea ofwhat will and will not be tracked
by Rx PLL - Can be a good diagnostic test
Measurement Speed
- Not a fast manufacturing test 1 minute
measurement
Example Measurement Equipment
Example Standards
- SONET Functional Testers
- BERTScope DCRj, CRj
- SONET/ SDH, e.g. 9.95 Gb/s
- OTN, e.g. 10.709, 11.095 Gb/s
33Agenda
- So Whats the Problem?
- Jitter Eye Closure Review
- Side Notes
- The Effect of Timing Reference
- Measurement Depth
- Common Methods
- Eye Diagram Jitter
- Mask Testing
- Bathtub/ BERTScan/ Jitter Peak
- Deep Mask Testing
- Jitter Spectrum
- Summary - Tying it All Together
PC Bus
34Why Do I See Many Different Jitter Results?
(Other Factors Also at Play, Not Covered Here)
Jitter Frequencies Included in Measurement
Banded to 80 MHz
LBW ? Bit Rate
Broadband, 0 ? Bit Rate
With Clock Recovery
Shallow
80M
20k
10G
LBW
80M
20k
10G
4M
- OmniBER
- DCRj Jitter Spectrum
Deep
Depth
80M
20k
10G
4M
80M
20k
10G
LBW
80M
20k
10G
4M
35Summary
PC Bus
- Transmitter test is designed to ensure error free
system operation. - Test methods vary between standards, partly
because of the available equipment at the time
they were written. - Older standards specified mask tests but not test
depth. - Telecom standards also worry about the frequency
spectrum of jitter. - In order to assure correct operation at BER
levels of system operation, deep measurements are
preferable. - Many newer standards require testing at deep BER
levels, either with deep masks or Bathtub Jitter.
Note All the jitter methods discussed here,
including for 11.1Gb/s 10GbE LAN PHY, are
available from the BERTScope family.