Practical Guide to Making Advanced Jitter
Measurements
Get results you can live with!
Pascal GRISON
Digital Application Engineer
pascal_grison@agilent.com
Validating Design Performances
through accurate measurements
PCIe 1.1, 2.5 GT/s
PCIe 2.0, 5.0 GT/s
16” Channel
16” Channel
PCIe 3.0, 8.0 GT/s
16” Channel
2
High Speed Serial Link
Design for Success
There are Three faces to the problem
• How much jitter should the transmit side be allowed to generate
• How much jitter can the receiver side tolerate
• How much degradation is acceptable from transmission line
in the case of local Chip to Chip interconnect (PCI-Express)
in the case of Rack Backplane (ATCA,PCI-Express, AXI-e, VPX…)
in the case of an external cable (SATA,HDMI,DISPLAYPORT,USB…)
A well designed Serial Link mustspecifies properly these 3 points
to guarantee system level performance (bit-error-ratio)
3
Fundamental Signal Integrity Analysis:
The Eye Diagram
The easiest way to get an overall idea of the quality of the serial signal
Using Oscilloscope Software Clock Recovery with PLL Emulation to recover Signal Clock
Eye Diagram is the superposition in the middle of the screen of 3 consecutive bits
Multiple case combined form the Eye (000,001,010,011,100,101,110,111)
Evaluate overall impact of Channel, Crosstalk and RJ/PJ
101 Sequence
011 Sequence
Overlay of all combinations
4
Using OFFLINE Oscilloscope GUI to Analyse ChannelSim
DIA2 1Gb/s Differential signal
ChannelSim ADS
Simulation –> Front Panel
ChannelSim N8900A
Simulation –> Infiniiview
Measurement vs Simulated Eye Diagram Analysis
ChannelSim –> Infiniiview
DUT DSAX93304A Scope Meas
Note
User error on DIA2 Amplitude register
setting during Scope Meas
800mV instead of 1200mV
Using ChannelSim to evaluate Xtalk impact
from SIGA79 Single-Ended signal
DIA2_Xtlk_DIA1_SIGA79_ChannelSimTB_AMI
200 Mbps
Simulation – Infiniiview
DSAX93304A Measurements
Note
User error on DIA2 Amplitude register
setting during Scope Meas
800mV instead of 1200mV
DIA2_Xtlk_DIA1_SIGA79_ChannelSimTB_AMI
400 Mbps
Simulation – Infiniiview
Oscilloscope Measurements
What represents “good enough”?
The eye-mask is the common industry approach to measure the eye opening
Failures usually occur at mask corners
Violating USB FS 12Mb/s Eye Diagram
Good 2.5Gb /sDisplayport Eye Diagram
But How is Defined the Mask Template?
10
Measure DUT Receiver Minimum Eye at BER 10E-12
BERT up to 28Gb/s PRBS Generation
with Calibrated Jitter insertion
and integrated adjustable ISI channel
DUT SerDes in
LoopBack Mode
RX Data
Receiver
Rx latch
AGILENT SI Seminar 2012
by Pascal GRISON
ISI
Channel
DLL
Rx
PLL
Transmitter
Tx latch
JBERT Realtime Error Detector allow
thorough BER Analysis and BER Eye
Opening
Tx
DLL
TX Data
Semiconductor Vendors are Using bert to Caracterize SERDES BER susceptibility
to ISI, Random Jitter and Frequency dependant Periodic Jitter Eye Closure
11
Analysing a serial Link
TX
Clean Source Signal
Channel
Channel
Frequency Response
RX
Closed Eye
Received Signal
We are going to analyse a 12Gb/s Link
Channel will be 9 Inch FR4 PCB
12
Scope Eye & Jitter BreakDown Analysis on TX output
Transmiter 12Gb/s
Intrinsic Jitter Analysis
33GHz 80GSa/s Scope
AGILENT SI Seminar 2012
by Pascal GRISON
RJ: 500fs (RMS)
PJ: 740fs
DCD: 660fs
ISI: 10.52ps
13
Eye Diagram on TX output and Channel Output
Depending on Link
Target Datarate &
Transmission Channel
Losses
AGILENT SI Seminar 2012
by Pascal GRISON
Even with Perfect TX
Eye Opening…
AGILENT SI Seminar 2012
by Pascal GRISON
You may end up with a
completely closed
at Receiver Side
Why is the RX Eye
Closed? ISI Jitter!
Does that mean that this
link will never Work?
Well it Depends….
Black GUI Offline Analysis Application: Infiniiview
14
What are Inter-Symbol Interferences?
AGILENT SI Seminar 2012
by Pascal GRISON
ISI Jitter is coming from Signal Distorsions in Transmission Channel
15
Impact of TX De-Emphasis on RX Signal
To reduce ISI at RX Side, Most TX implement De-Emphasis
AGILENT SI Seminar 2012
by Pascal GRISON
Press ESC during Video to Skip Video
16
-12dB TX De-Emphasis -> RX Eye Opening
From Zero RX Eye
Opening with no TX
De-Emphasis
RX Eye Opening of
25mV X 27.5ps
Was achieved with
-12dB De-Emphasis
AGILENT SI Seminar 2012
by Pascal GRISON
Note: Measure is
done on D+ only
So Differential Eye
Opening is 2X SE
Opening
=50mV X 27.ps
Much better!
But is it enough?
Infiniiview Offline Eye Diagram Analysis of Waveform captured on scope
17
Scope can Emulate Receiver EQUALIZATION
Modern SerDes are embbeding
RX EQUALIZATION
AGILENT SI Seminar 2012
by Pascal GRISON
Using Oscilloscope Equalization
we can emulate most DUT RX EQ
configurations:
FeedForward EQ
Continuous Time EQ
Decision Feedabck EQ
Let’s Emulate a Typical configuration:
Upper Eye:
FFE 2Taps -> CDR
DFE 5 Taps ->Data
Lower Eye
FFE 2Taps -> CDR (no EQ on DATA)
DSO91304A#014 or N5465A
18
Emulate Receiver EQUALIZATION on Oscilloscope
From almost Zero
RX Eye Opening
with no TX DeEmphasis and
No RX EQ
RX Eye Opening of
132mV X 65ps
Was achieved with
EQUALIZATION
AGILENT SI Seminar 2012
by Pascal GRISON
Note: Measure is
done on D+ only
So Differential Eye
Opening is 2X SE
Opening
=264mV X 65ps
Very Good Eye
opening !!
You MUST Emulate your RX Equalization in Oscilloscope to Analyze True RXEye Diagram
Press ESC during Video to Skip Video
19
Jitter Components
Total Jitter
(TJ)
Bounded
UnBounded
Deterministic
Jitter (DJ)
Correlated with
Data (DDJ)
DutyCycle
Distortion
(DCD)
Tr, Tf
D
InterSymbol
Interference
(ISI)
Settling Time
Reflections
Non flat Freq
Response
Random Jitter
(RJ)
Uncorrelated
with Data (BUJ)
Non
Periodic
(ABUJ)
Periodic
(PJ)
Gaussians
Xtalk
Clocks
Thermal
Non Linear
CR
Xtalk
Shot
Events
(s, RJRMS)
1/f
Burst
20
Where Does Jitter Come From?
Aggressor Lane A
Aggressor Lane B
Transmitter
Aggressor Lane C
Receiver
Lane under Study
•Lossy Channel interconnect (ISI)
•Impedance mismatches (ISI)
•Crosstalk with ABC Lanes (BUJ)
•Thermal Noise (RJ)
•Local Oscillator (RJ/PJ)
•Bias shift (DCD)
•Power Supply Noise (RJ, PJ)
•On chip coupling (PJ, ISI)
•Termination Errors (ISI)
21
High Probability Determinisic Jitter
is reported as Peak-Peak
Ideal Location in Time (Reference)
Transition
Instant
Early
Late
0
22
DtEarly
DtLate
Threshold
1
JPP=DtEarly Pk + Dtlate Pk
Random Jitter is Measured as RMS
•
JPPRJ is unbounded
•
For pure random jitter the BER defines the JPPRJ:
•
Total Jitter (TJ), JTJ, for a given BER:
J TJ  n  s
DJ
 J PP
RJ
DJ
 n  J rms
 J PP
23
BER = 10-12  = JPPRJ = 14.1 JrmsRJ
Pure random or periodic jitter:
Relation between RMS and PP Jitter
For 6 Sigma Statistics (BER=3.4*10-6) and pure random jitter:
Jitter pp ~ 9 * Jitter RMS.
For pure periodic Time Intervall Error (Jitter):
Jitter pp ~ 2*sqrt(2) Jitter RMS ~ 2.828 * Jitter RMS
For BER = 10-12 and pure random Jitter
Jitter pp = 14.1 * Jitter RMS
Page #
Tx
f Noise
Pre-emphasis
Delay
Ground Bounce
ISI
Skew
Frequency Response
Crosstalk
Reflections
Skew
Topics
Review of Jitter Measurement
Jitter Decomposition
Four Critical Areas
• Your control of the jitter measurement
• Examples and tips for Good Measurements
Evaluating ‘BUJ’ from Crosstalk
Other Considerations
Noise
Match
Equalization modeling
Clock Recovery/PLL Performance
Review of Jitter Measurement
On an oscilloscope we monitor the waveform transitions and note the jitter at
each transition point. This is called the Time Interval Error record
The Problem with Jitter…
Jitter pk-pk vs # Transitions (fixed record length)
Jitterpk-pk
(ps)
58
56
54
52
50
48
46
44
42
0.25
Jitter will statistically grow over:
Max
Average
Minimum
0.5
1
2
4
8
• increasing number of Acquired
Waveforms
• Increasing observation time
16
Transitions (M)
[Acquisition Length constant at 8MPt]
Jitter for 1 Million Transitions
100
Jitterpk-pk
90
Max
80
Average
70
Minimum
60
50
40
64
32
16
8
4
Acquisition Length(MPt)
2
Phase Noise Plot
Character of Jitter
Many contributors to Jitter
• Most of these are Bounded… they have limited distributions of jitter.
• Others are grouped in the UnBounded classification…
Unbounded Jitterpkpk will grow over time of measure
The distributions of these contributors convolve together to
compose the Total Jitter Histogram.
Jitter Components
Total Jitter
(TJ)
Bounded
UnBounded
Deterministic
Jitter (DJ)
Correlated with
Data (DDJ)
DutyCycle
Distortion
(DCD)
Tr, Tf D
InterSymbol
Interference
(ISI)
Settling Time
Reflections
Non flat Freq
Response
Random Jitter
(RJ)
Uncorrelated
with Data (BUJ)
Non
Periodic
(ABUJ)
Periodic
(PJ)
Gaussians
Xtalk
Clocks
Thermal
Non Linear
CR
Xtalk
Shot
Events
(s, RJRMS)
1/f
Burst
Approach to Resolve ‘random nature’:
the Dual Dirac Assumption
Fit the tails of the jitter PDF to two Gaussian curves
DJDD
Jitterpp(BER) =DJDD + n s
N = f(target BER)
For instance for BER = 10-12 n ~ 14
sL
The jitter that composes DJDD
comes from the deterministic
components…
7s for 10-12 BER.
sR
L
R
Jitter Decomposition Overview
Waveform
Acquisition
Clock
Reference
Evaluate TIE
DDJ Analysis
RJ Extraction
Dual Dirac
Analysis
Complete T.I.E Record
DDJ: T.I.E per Bit
RJ/PJ T.I.E Record
Reported Values of TJ, RJ, DJDD
Four Critical Areas
1
Waveform
Acquisition
Clock
Reference
2
Evaluate TIE
1
Signal Fidelity in Connection
Oscilloscope Settings
2
Clock Recovery Type &Setting
PLL Parameters
3
Pattern Type/Length Expected
4
Gaussian Jitter Estimation
Method
DDJ Analysis
3
RJ Extraction
4
Dual Dirac
Analysis
Inattention to these areas will compromise your result.
Measurement Signal Fidelity
Connection path to signal isn’t perfect
1
Waveform
Acquisition
Scope Probe/Connection Flatness
Test Point Access Fixture BW/Flatness
Skew
Match
You want
not
Your Device
Tx
Test
Fixture
Degradation in performance of any of these will cause DDJ increase
in your result, and affect RJ as well.
Frequency Response of Infiniium DSO91304A
Agilent DSO91304A 13GHz FLAT Response
Agilent DSOX93204A 32GHz ULTRA-FLAT Response
Magnitude Flatness +/-0.25dB up to
Key Observations
32GHz
•
Agilent meets specified bandwidth will all of it settings.
•
Agilent provides the flattest frequency response by using the DSP magnitude and phase compensation technology.
•
Notice the amplitude gain/attenuation variations are controlled to the minimum amount throughout the bandwidth,
Oscilloscope Settings
Scale Setting
Threshold Settings
Hysteresis
Acquisition Length
Your Device
Test
Fixture
1
Waveform
Acquisition
Oscilloscope settings: Input Scaling
1
Waveform
Acquisition
Scaling = Volts/division selection
A poor selection will Amplify scope noise floor to affect your
measurement….
Tip for Good Measurement
Choose a scale that gets the ‘raw’ signal close to full screen.
Push Knob to access Vernier DO NOT OVERDRIVE the SCOPE
Oscilloscope Settings: Scale and Jitter 1
5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
80
70
60
50
40
30
Jitter Pk-Pk
RJrms
Jitter vs Full Scale
20
10
0
RJ
TJ
Full
Half
Qtr
Eighth
Scale
Dependent on slope of
signal, noise on signal and
noise of scope
Waveform
Acquisition
Why is your Oscilloscope Vertical Noise
Floor Impacting your Jitter Results?
Let’s consider a theoretical signals with Zero jitter, fixed voltage noise
presenting three different edge speed and crossing a Threshold at
50%
1.
2.
3.
4.
Voltage noise translate directly in Jitter
Higher Vertical Noise Floor translate in Higher Jitter
Slow Edges will dramatically transform vertical Noise into Jitter
At constant Edge Speed, best Measurement Noisefloor translate into
Lowest RJ and TJ Jitter, Best Eye Diagram Opening and more repeatable
results!
Oscilloscope Noise Impacts Measured Jitter
Measure AC rms measurement at proper Volts/Div scale for DUT signal
Agilent 86100D/86108B Series: ~ 640 uV
(at 35 GHz BW Setting) & 140mV/div setting
Agilent 90K X-Series: ~ 6.1 mV
(at 137 mV/div and 32 GHz BW Setting)
Note - single-ended noise measurements since we’re performing a comparison using singleended signals (analyzing P and N from the same DUT)
Manually Determine Induced Jitter due to Scope
Noise and Signal’s Slew Rate
RN = Random Noise(rms)
Slew Rate = rate of change of signal in V / ns
= Delta V/ Delta T
Induced Jitter due to scope noise:
1. 86100D / 86108B DCA-X
Noise = 640 uV
Slew Rate = 173.3 mV / 8.34 ps
= 20.8 V/ns
Induced Jitter = RN / SlewRate
= 640uV / 20.8V/ns
Induced Jitter = 31 fs
Delta V
Delta T
2. 90K X-Series Oscilloscope
Noise = 6.1 mV
Slew Rate = 26 V/ns
Induced Jitter = RN /SlewRate
= 6.1mV / 26 V/ns
Induced Jitter = 234 fs
The faster the edge, the smaller the problem! And vice-versa!
Estimate Jitter due to Intrinsic Scope Jitter/Noise
and Signal’s Slew Rate (AM-to-PM Conversion)
Example: 86100D / 86108B
1. DUT Random Jitter = 200 fs
2. Scope Random Jitter = 50 fs
 Random Timing Jitter = 206 fs
= SQRT [(200^2)+(50^2)]
3. Noise Induced Jitter from scope
= 31 fs (see previous page)
Example: 90K X-Series
1. DUT Random Jitter = 200 fs
2. Scope Random Jitter = 75 fs
 Random Timing Jitter = 213 fs
= SQRT [(200^2)+(75^2)]
3. Noise Induced Jitter from scope
= 234 fs (see previous page)
Measured Jitter = SQRT [(Timing Jitter)^2 + (AM-to-PM Jitter)^2)]
Measured Jitter
= SQRT [(206)^2 + (31)^2)]
= 208 fs
Measured Jitter
= SQRT [(213)^2 + (234)^2)]
= 317 fs
Scope jitter results include noise induced jitter (AM-to-PM conversion).
Results change due to signal slew rate and random noise.
Summary - BaNoise / Slew Rate
As random noise (RN) increases, random jitter
increases. Especially problematic with slower
edge speeds!
Minimize oscilloscope noise. Use only enough
BW to capture signal.
Case Study: Observing the 4.8Gbps (FB-DIMM like)
Signal with Various Edge Rates (at 55ps)
4.8Gbps: Fundamental Freq = 2.4GHz, 3rd Harmonics = 7.2GHz, 5th Harmonics = 12GHz
6GHz Scope
6GHz scope only captures
fundamental frequency.
8GHz Scope
8GHz scope captures both
fundamental and 3rd harmonics, but
not 5th. The eye pattern changes
dramatically.
12GHz Scope
Although 12GHz scope captures 3rd and 5th harmonics, at 55ps
rise time, there is no difference between eye patterns of 8 and
12GHz scope even the signal rate stays at 4.8Gbps.
This is because the signal has no 5th harmonics freq content.
It is the “edge rate” that determines required BW, not 3rd and 5th harmonics.
43
Rise Time vs. Bandwidth and Required Sampling Rate
Scope BW and Measurement Accuracy
fmax
Scope Digital Filter Type
Measurement Error of Tr
20%
10%
3%
Sampling Speed
(With sin (x)/x interpolation feature)
0.5 / Rise Time (10%-90%)
0.4 / Rise Time (20%-80%)
Gaussian
Brickwall
Scope BW
1.0 fmax
1.3 fmax
1.9 fmax
4 x BW
1.0 fmax
1.2 fmax
1.4 fmax
2.5 x BW
For more info, see application note 5988-8008EN
• A simple calculation matrix to determine the required scope bandwidth and the
sampling rate to characterize a given signal accurately.
• Notice, due to the different amount of “out of bandwidth” signal frequency contents
that each filter response captures (i.e. becomes the source of aliasing), in order to
characterize the signal with desired accuracy, a scope with a “Gaussian” filter
response requires more bandwidth and more sampling rate than a scope with a
“Brickwall” filter response.
45
Oscilloscope Settings:
Threshold Settings
1
Waveform
Acquisition
Choose:
• Fixed Threshold ONLY
• Threshold Value
0.0 mV threshold
10.0 mV threshold
Tip for Good Measurement
Use halfway point in the signal swing. Most differential buses will
stipulate 0 volts as the threshold. Examine rise and fall time differences.
Oscilloscope Settings:
Hysteresis Settings
1
Waveform
Acquisition
Setting Hysteresis:
You are setting how you discern an edge.
If the setting is too low:
the scope will interpret multiple edges.
If the setting is too high:
the scope will miss edges altogether.
Hysteresis settings
Hysteresis settings
1 region
0 region
1
threshold
0 region
1 region
threshold
0
Tip for Good Measurement
Use Halfway point between threshold value and the smallest 0-1-0 or 1-0-1
swing
Oscilloscope Settings:
Memory Depth
1
Waveform
Acquisition
It’s a balance:
If the setting is too low:
- can’t do PLL clock recovery
- won’t see enough of the signal edges
If the setting is too high:
- the scope responsiveness suffers
- may start including more 1/f noise than you want
Tip for Good Measurement
Use the Setup Wizard. Experiment for repeatable and consistent
results.
Check things out…
You can quickly analyze the T.I.E. Trend...
?
!!
Tip for Good Measurement
Before performing Jitter separation, check the T.I.E trend, spectrum
for ‘reasonableness’.
Checking for ‘Reasonableness’
More on T.I.E. Trend….
Unsmoothed
Smoothed and Expanded
Smoothed
Peak to Peak trend measurements will let you know if you are in the ballpark…
If you get something like 10nSeconds on a 2Gbs signal, you likely have issues
you need to resolve before doing jitter decomposition
Checking for ‘Reasonableness’
Analyze the T.I.E. Spectrum….
Short time record
Longer (32x) record
T.I.E. Spectrum measurement will let you see frequency components. Higher
resolution may demonstrate frequency spacings of clock harmonics, DDJ
spacings, or multiple jitter sources.
Clock Reference
2
Jitter measurement demands a reference. It may be:
From previous edges in the signal
Externally Available
Recovered from a Hardware clock recovery unit
Constant Clock estimation
Software PLL
Clock
Reference
SW PLL and Constant Clock
Constant Clock
0.5 MHz Sine injected.
1/f noise content seen
2
Clock
Reference
2nd Order SW PLL
0.5 MHz Sine wave is reduced 18 dB
and there is no other low freq content
Quick Review - Clock Recovery (CR) Basics
o Provides a recovered clock for receiver
o Manages jitter in the system
o Standards specify CR Phase Locked Loop (PLL) order, bandwidth, peaking, or damping factor
Sampler
(Receiver)
Data Input
Phase Detector
Phase
Error
Amplifier
Voltage Controlled
Input
Oscillator (VCO)
Signal
Data relative to a
“clean” clock
(narrow loop BW)
Recovered Clock
Narrow
CR Loop
BW
Basic CR Block Diagram
PLL “Jitter Transfer Function” (JTF)
JTF  Closed loop gain

fout
A( s)

 G( s)  G( s) e jf ( s )
fin 1  A( s)
BEWARE of Clock Recovery
(PLL) Definitions!
 Standards (and scopes)
describe PLL requirements
differently.
1.2
Clock Recovery PLL Response
Jitter Transfer Function (JTF)
and
Observed Jitter Transfer Function (OJTF)
1
Jitter Multiplier
• indicates how much of the jitter on the
input signal is “transferred” to the
recovered clock (output)
• low-pass filter function (LPF)
OR
0.8
Data relative to
recovered clock
(wide loop BW)
OR
Wide CR
Loop BW
“Observed Jitter Transfer Function” (OJTF)
• indicates the jitter that is “observed” by the
receiver (scope)
• high frequency jitter on the data stream is
“transferred” to the receiver (HPF)
0.6
0.4
OJTF  1  JTF
0.2
 1 - G( s)  1  G( s) e jf ( s )
0
1.0E+3
10.0E+3 100.0E+3 1.0E+6
Frequency (Hz)
10.0E+6 100.0E+6
Agilent 86100C/D Sampling Scope

CR loop BW setting configures JTF

JTF Example: Ethernet, SONET/SDH
Agilent 90K Series Real-time Scope

CR loop BW setting configures OJTF

OJTF : SATA/SAS
Jitter Spectrum
Magnitude
To understand how the CR PLL response impacts low frequency jitter, it is
useful to observe jitter in the frequency domain
Frequency
Offset frequency
Jitter Spectrum
Shows distribution of low frequency jitter and impact of clock recovery
Narrow CR loop bandwidth
Wide CR loop bandwidth
Spectral lines indicate deterministic jitter
(including SSC and its odd harmonics)
Jitter floor (without tones)
is random jitter
 Observe all incoming jitter
 Track out low frequency jitter
Clock Recovery response greatly impacts amount of jitter
seen by receiver, and/or measured by an oscilloscope!
Clock Recovery Models
1st Order PLL:
JTF BW = OJTF BW
Peaking/DF = none
Roll-Off: 20 dB/decade
- Less ability to track out low
frequency jitter and stay locked
- Real hardware CR does not
behave this way
3rd Order PLL:
JTF BW > OJTF BW
- Specify zero, gain, pole
frequencies.
- Roll-Off: 60 dB/decade
below zero frequency
- Use “PLL Response Tutorial”
workbook to model.
2nd Order, Type 2 PLL:
Bandwidth: JTF BW > OJTF BW
Peaking/Damping Factor: need to specify
Roll-Off: 40 dB/decade
(tracks out low jitter more than 1st order PLL)
Desired SW CR Loop Response
HW CR Loop Response
HW CR response may
have higher peaking in
OJTF than “desired”.
e.g. match a standard exactly
Less Peaking
This will amplify jitter in
this region.
Jitter Spectrum
Note – significance
depends on DUT jitter
spectrum.
Jitter Spectrum
Jitter Spectrum Analysis and SW Clock Recovery
Emulation using Agilent 86100D/86108B-JSA
86108A/B Module
Device
Under
Test
Data or Clock
Signal
(“Jitter Filter”)
•
•
•
Integrated
Hardware
Clock
Recovery
Ideal Software
Clock Recovery
Emulation
(“Jitter Fitler”)
Filtered
Signal
•
•
•
“Real” CR PLL response
Adjustable Loop Bandwidth
Adjustable Peaking (discrete)
“Ideal”, flexible CR PLL response
Adjustable Loop Bandwidth
Adjustable Peaking (continuous)
Desired SW CR Loop Response
e.g. match a standard exactly
Less Peaking
HW CR response may
have higher peaking in
OJTF than “desired”.
Jitter Spectrum
Apply “ideal” PLL
using Software
Clock Recovery
Emulation
Jitter Spectrum
Jitter amplification will
occur in region where
unwanted peaking
exists.
Note – “how much” of
an increase depends
on DUT jitter spectrum.
Higher
Accuracy
Hardware only clock recovery
“Ideal” SW Clock Recovery Model
Clock Recovery Comparison
Always use similar clock recovery models – “Apples-to-Apples” setup
9000090K
X-Series
Agilent
X-Series
Agilent 86100D with 86108B
10 Gb/s Jitter Measurement – Matching CDR
“Perform a jitter measurement using 2nd Order CR response with
10MHz OJTF and 0.707 DF.”
10G Pattern
Generator
D+
JTF: 2nd Order, 20 MHz Loop BW, 2dB Peaking
OJTF: 2nd Order 10 MHz Loop BW, 0.707 DF
D-
OJTF: 2nd Order 10 MHz Loop BW, 0.707 DF
We are using the same CR setup now, but are there other things we should look at?
Pattern Type/Length
Your Device
Tx
ISI Channel
Periodic Pattern
is generally preferred
algorithm is robust/efficient
Arbitrary Pattern
requires dynamic estimation
of ISI channel so is less
efficient.
3
Data
Dependence
Analysis
Data Out
Data Out Pattern
Extraction
Non Repeating
Arbitrary
Repeating
Short: 27, 29, 211, 215
Long: 223, 231
Periodic/Arbitrary
Arbitrary
Tip for Good Measurement
If Arbitrary mode is a must, analyze step response if possible.
Pattern Type/Length
Your Device
Tx
ISI Channel
Data
Dependence
Analysis
3
Data Out
Arbitrary Data: Analyze Step
Measure step length in terms
of Unit Intervals, N. Use value
as starting point in determining
optimal setting for Arbitrary
Mode ISI Filter.
N
PRBS 231 3Gbs
4
6
Minimize RJ
Trade off with time
8
10
RJ
RJ Extraction
4
Waveform
1 Acquisition
Clock
Reference
RJ
Extraction
2
Evaluate TIE
DDJ
Analysis
Characterize the tails
of the distribution
4
3
We will now deal with your
algorithmic options in the
evaluation of the RJ component
RJ
Extraction
We are here
RJ Extraction
4
RJ
Extraction
Jitter Measurement
Algorithm on
Oscilloscope
Extraction Method
Rationale
Narrow Bandwidth
Speed/Consistency to Past
Accuracy in low BUJ cases
Presence of Low Freq RJ
Wide Bandwidth
Known Bounded Noise
Spectral
Gaussian Tail Fit
General Purpose/
ABUJ (Xtalk) Conditions
Spectral Extraction Method
4
time error
Measurement Detail
likely to contain PJ
RJ
Extraction
PJ threshold is
chosen by
experimentation.
PJ threshold
0
freq
time error
0
likely to contain PJ
Integrate PSD to derive d,
or, RJRMS. Sum the PJ
components for PJRMS
PJ threshold
0
0
freq
Spectral Extraction
4
RJ
Extraction
Separation occurs
as described…
What do you do in
this case?
Is it RJ
or PJ?
Spectral Extraction
4
RJ
Extraction
Non-linear Threshold in limited acquisition sizes can help this…
Wide RJ BW analysis
RJ=.88ps
PJDD=10.4ps
Narrow RJ BW analysis
RJ=1.68ps
PJDD=4.0ps
Tip for Good Measurement
Choosing longer sampling time and/or selecting Narrow Mode
will spread the spectrum around (greatly alias) and will have the
effect of the flattening the noise.
RJ Spectral Extraction: Wide vs Narrow
4
RJ
Extraction
Tip for Good Measurement
Analyze the bathtub plot for slope continuity between measured data
and extrapolated result
ABUJ: Crosstalk or Ground bounce
Amplitude interference uncorrelated with data
and not periodic in nature.
Victim
No crosstalk
Dv
Aggressor
Victim Out
With crosstalk
Dt
Dt = Dv/Slopevictim
Crosstalk Interference Model
ABUJ Observations and Measurement
(ABUJ= Aperiodic Bounded Uncorrelated Jitter)
Something is wrong here..
Using the slope continuity concept we expect
the extrapolated curve to look like this.
The RJ/PJ spectral extraction doesn’t deal with ABUJ well.
The RJ is overestimated severely.
1. View ABUJ in time domain
2. Techniques to Evaluate
RJ Extraction with Crosstalk (ABUJ)
Spectral vs Gaussian RJ Extraction.
No Crosstalk
4
RJ
Extraction
w/Crosstalk
X
Examine slope continuity
Spectral Extraction
Spectral Extraction
Gaussian Tailfit
Extraction
Compare actual Data with RJ estimates of both methods
Gaussian Tailfit
Extraction
Tip for Good Measurement
Analyze the bathtub plot with both extraction modes to explore
presence of crosstalk or ground bounce.
What makes Tail fitting hard
4
RJ
Extraction
Histogram Object
1.2
Measurement Detail
1
High Statistical Precision
Low accuracy Extrapolation
0.8
Fit Window
0.6
Lower Precision
Statistics
Higher accuracy
Extrapolation
0.4
0.2
0
-0.2
Noisy data
DJ end
0
5
10
More data may be required to get reliable consistent results
15
Aperiodic Bounded Uncorrelated Jitter
Time Domain Views
Victim
Aggressor
Aggressor at transition
Aggressor
Victim
Aggressor
Crosstalk, time aligned for illustration.
Aggressor in middle of eye
Adjusting the Crosstalk in phase
Two Ways to Analyze ABUJ
1. Two Pass Spectral Extraction Approach
 Assumes you have control of the Xtalk interferer(s)
 Assumes conveyed jitter of interferer is all ABUJ
2. If Interferers cannot be desactivated
 Use Gaussian Tailfit Extraction
ABUJ/Crosstalk Analysis
1. Two Pass approach
a) Turn off crosstalk element(s).
b) Measure Jitter components
1.47 ps
c) Turn on crosstalk element(s)
d) Enter RJrms value for RJ (‘specify’)
e) Crosstalk (ABUJ) will go into
bounded portion of jitter which will prevent
overestimation of RJ and Total Jitter.
ABUJ/Crosstalk Analysis
Two Pass Approach
With interferer
No interferer
Victim
Aggressor
Aggressor at transition
ABUJ/Crosstalk Analysis
2. Gaussian Single Pass RJ Tailfit Extraction
No interferer
With interferer
Victim
Aggressor
Aggressor at transition
Total Jitter TailFit estimation in this case is within 2% of Two Pass Analysis!
ABUJ is a bit tricky. Use every tool you have available.
Other Jitter Measurement Considerations
Gain Margin by removal of Scope contribution to RJ
DUT Tx
DUT Tx
ISI
Channel
With no Scope RJ removal
With Scope RJ removal
Other Jitter Measurement Considerations
Simulate Crosstalk to Evaluate Effect of Aggressor on Victim
Ch B
Tx
.snp
Tx
Ch A
Tx
Tx
+
Scope Front End HW
Tx
Tx
.s2p or
.s4p
.s2p or
.s4p
Great correlation
+
Actual Measurement
Simulation
Other Jitter Measurement Considerations
Analyze the Amplitude components of your signal
Analyze anywhere in the Unit Interval
Summary
Histogram Fits. True RJrms = 2, PJmax = 5
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
2
Dual Dirac Model
4
6
8
10
Your device and Environment
Tx
f Noise
14
ABUJ (Crosstalk) Analysis
Four Critical Areas
Pre-emphasis
Delay
Ground Bounce
ISI
Skew
12
Frequency Response
Crosstalk
Reflections
Skew
Use tools available
Agilent’s Oscilloscope
Agilent’s Oscilloscope
Portfolio Portfolio
Real-time
Bandwidths from
20 MHzBandwidth
to 90 GHz from 50 MHz to 63 GHz
Entry
Handheld
U1600B
USB
U2700
InfiniiVision Series
Infiniium Series
2000X
3000X
6000A
7000B
9000A
90000A
90000X
90000Q
DSO/MSO
DSO/MSO
DSO/MSO
DSO/MSO
DSO/MSO
DSO/DSA
DSO/DSA
DSO/DSA
the Infiniium 90000 Q-Series
InfiniiumIntroducing
90000 Q-Series
Achieve
Your Real Edge
Achieve Your
Real Edge
4 Channels
•
•
20 GHz, 25 GHz, & 33 GHz analog bandwidth
Up to 80 GS/s sample rate
2 Channels
•
•
50 GHz & 63 GHz analog bandwidth
Up to 160 GS/s sample rate
Industry
Leading
Signal
Integrity
Up to
2 Gpts
acquisition
memory
Industry’s
most
complete
software
Industry’s
only 30 GHz
probing
system