Advanced Jitter and Eye-Diagram Analysis
Project Manager / Keysight Taiwan AEO 余宥浚 Jacky Yu Constructing the Real-Time Eye
E1Data Stream
Eye Crossing Points Unit Interval Left Edge Right Edge Nominal Overlaid Sampling Point transitions
Ideal Sampling Point E0
x = 0 x = 1/2 T x = T 2 2
Non-ideal Real-Time Eye Noise
What happened to our eye opening?
3 3 Non-ideal Real-Time Eye
Jitter
What happened to our eye opening?
4 4 Pause for definitions
A dictionary definition of the verb “jitter”: To make small, quick, jumpy movements. In the digital design world, jitter is defined as: The deviation of the significant instances of a signal from their ideal locations in time.
The significant instances for data signals are the transitions (edges)
The ideal locations for the transitions are determined by the time reference (clock)
5 5 Single Transition Jitter We can see from the eye diagram, jitter effects the transitions of the data stream.
Let’s take a closer look at a single transition.
Ideal Location in Time (Reference)
Transition Early Instant Threshold Late
tEarly 0 1 tLate
Peak-to-peak jitter = JPP = tEarly Pk + tLate Pk
6 6 Jitter and the Real-Time Eye
Single Transition (0 to 1) Ideal Transition Time
Unit Interval
Many Overlaid Transitions (0 to 1 and 1 to 0)
JPP Ideal Sampling Point
Probability Density Function Histogram (PDF)
7 7 It’s all about Bit Error Ratio
➢ Noise and Jitter cause data transmission errors in digital systems
➢ These errors are characterized by the Bit Error Ratio (BER) of a serial data link
➢ BER is the primary measure of the fidelity of a link
BER = # Transmission Errors / Total # Transitions
When data rates were low, designers were mainly concerned with functionality (1s and 0s)
With rates > 1 Gbps, the analog nature of signals becomes significant
Noise and jitter affect system BER and the quality of a data link
8 8 Measuring Jitter: Bit Error Ratio (BER) Testing
❖ The only way to directly measure Total Jitter is with a Bit Error Ratio (BER) test.
❖ Sample at various points along unit interval, 0.5 directly measure BER at 10-3 Gaussian each point. Plot “bathtub” Tails curve.
10-6 BER
10-9 Eye Opening at BER=10-12 -12 10 W
0 0.5TB TB /UI TJ(BER) = UI – W Requires specialized equipment and very slow
9 9 JITTER AND TIME INTERVAL ERROR (TIE)
On an oscilloscope we monitor the waveform transitions and note the jitter at each transition point. This is called the Time Interval Error (TIE) record.
Measures total jitter of the acquisition. The more transitions you measure, the greater TIE will become.
Waveform transitions deviate from expected transition time
10 Generate Time Interval Error (TIE) by measuringKeysight Territory Turbo transitions Program versus reference clock ADVANCED JITTER ANALYSIS WITH REAL - TIME OSCILLOSCOPES
Agenda • Review of Jitter Decomposition • Assumptions and Limitations • Spectral versus Fail Fit Method • Advanced Jitter Analysis with Crosstalk Removal Tool • Scope Random Jitter Removal from Jitter Analysis • Other Tools to consider for Jitter Analysis • Summary
Keysight Territory Turbo Program 11 Total Jitter (TJ)
Bounded Unbounded Acronyms Deterministic Random Jitter Jitter (DJ) (RJ) • DDJ : Data Dependent Jitter Correlated with Data Uncorrelated with (DDJ) Data (BUJ) • BUJ : Bounded Uncorrelated Jitter InterSymbol Non DutyCycle Periodic Gaussians (s, Interference Periodic Distortion (DCD) Jitter (PJ) RJ ) (ISI) (ABUJ) RMS ABUJ : Aperiodic Bounded Uncorrected Jitter Tr, Tf Settling Time Crosstalk Clocks Thermal Shot Non Linear Reflections Crosstalk Clock Recovery 1/f Non Flat Freq One-Time Response Event Burst
Keysight Territory Turbo Program 12 Dual Dirac Model
Total Jitter (TJ)
Random Random probability Jitter (RJ) Jitter (RJ)
model model approximates approximates measurement measurement
- time error Deterministic + time error Jitter (DJ) Key Assumptions
▪ Total Jitter at a BER can be predicted by a simple model using ‘Deterministic’ and ‘Random’ components. ▪ Gaussian distribution of random noise ▪ Stationarity of jitter distribution
13 Dual Dirac Model – Total Jitter
BER n Total Jitter -8 RJpp / 2 J DJ 1x10 5.73 n = DJPP s 1x10-10 6.47 7s -12 ( s = RJ ) 1x10 7.13 rms = 1x10-14 7.74 s
L R L R x 2 (x − ) 2 (x − ) 2 [ (x − ) + (x − )] exp − = exp − L + exp − R L R 2 TJ can be measured2 directly 2 TJ(BER) = UI –2sW 2s 2s using a BER test
OR
TJ can be calculated using RJ TJ(BER) = 2n*RJrms + DJ and DJ in the Dual Dirac model
14 EZJIT Jitter Analysis Suite For Infiniium Real-Time Oscilloscopes
Jitter Trend, Histogram, Advanced Jitter Vertical Noise and Timing Decomposition: Analysis and Measurements RJ/PJ/DDJ/DCD/ISI/ABUJ Decomposition
EZJIT
EZJIT Plus
EZJIT Complete
15 15 EZJIT Plus – Advanced Jitter Decomposition
➢ Easy to use wizard guides you through jitter measurement setup
➢ Fully compatible with Infiniium Software such as Equalization, PrecisionProbe, and InfiniiSim
➢ Customizable jitter views
16 16 • Review of Jitter Decomposition • Assumptions and Limitations • Spectral vs. Tail Fit Method • Advanced Jitter Analysis with Crosstalk Removal Tool • Scope Random Jitter Removal from Jitter Analysis • Other Tools to Consider for Jitter Analysis • Summary
17 Random Jitter (RJ) Extraction Methods
RJ Extraction Rationale
Methods likely to contain PJ time error
Spectral • Speed/Consistency to Past Measurements PJ threshold
0 • Accuracy in low Crosstalk or Aperiodic 0 freq Bounded Uncorrelated Jitter (ABUJ) conditions
Histogram Object Tail Fit • General Purpose 1.2 1
0.8
• Accuracy in high Crosstalk or ABUJ conditions 0.6
0.4
0.2
0
-0.2 0 5 10 15
18 Measurement Details
Periodic Jitter (PJ) likely to contain PJ threshold is chosen by
experimentation. time error
PJ threshold
0 0 freq
Integrate PSD to derive RJRMS Sum the PJ components for PJRMS
19 Handling of Different RJ, PJ Spectral Content
Separation occurs as described…
Is it RJ What do you do in or PJ? this case?
20 Non-linear Period Jitter (PJ) threshold can help
Wide RJ Bandwidth Analysis Narrow RJ Bandwidth Analysis
Non-Linear PJ Threshold
Linear and Flat PJ Threshold
RJ = 1.06ps RMS RJRMS = 9.66ps PJ = 93.17ps DD PJDD = 27.18ps
Which PJ Threshold or RJ bandwidth analysis do you choose?
21 Wide RJ Bandwidth Narrow RJ Bandwidth
Wide RJ under Narrow RJ more reports TJ accurately reports TJ
Smoothness of slope continuity between measured and extrapolated result on
the bathtub plot indicates the better PJ threshold (RJ bandwidth) method. 22 (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 does not deal with Crosstalk or ABUJ well. The RJ is overestimated severely.
23 23 Amplitude interference uncorrelated with data and not periodic in nature.
Gaussian Victim
Slope continuity
No crosstalk Bathtub and RJ,PJ Histogram Aggressor v
Non-Gaussian Victim Out Slope discontinuity
t = v/Slopevictim With crosstalk t Bathtub and RJ,PJ Histogram
24 MEASUREMENT DETAIL Histogram and Gaussian fit to right tail 1.4 1. Fit a Gaussian characteristic to
1.2 the right and left extremes of the RJ/PJ histogram distribution. 1
0.8
0.6 3. Find low probability Histogram Fits. True RJrms = 2, PJmax = 5 0.4 event (crosstalk) at the 0.9
0.2 0.8 end of tail that does not
0.7 fit Gaussian 0 -30 -20 -10 0 10 20 30 0.6 characteristics. jitter, ps 0.5
0.4
2. Actual data is 0.3 never smooth 0.2 0.1
0
2 4 6 8 10 12 14 25 Measurement Detail
Histogram Object 1.2
1 High Precision Low accuracy 0.8
Fit Window 0.6
0.4 Low Precision 0.2 High accuracy
0 DJ end Noisy data -0.2 0 5 10 15
error 0.8
0.6 Curve fit error
0.4 Hard to detect Crosstalk events out in 0.2 0 the tail. Might take longer time for Tail Fit -0.2 -0.4 results to converge. -0.6 -0.8 6 7 8 9 10 11 12 13 14 15
26 SPECTRAL VS. TAIL FIT EXTRACTION
No Crosstalk With Crosstalk
Spectral Extraction Spectral Extraction
Slope discontinuity. Over reports RJ.
Tail Fit Extraction Tail Fit Extraction
Analyze the bathtub plot with both RJ extraction modes to explore the presence of crosstalk or ground bounce. 27 • Review of Jitter Decomposition • Assumptions and Limitations • Spectral vs. Tail Fit Method • Advanced Jitter Analysis with Crosstalk Removal Tool • Scope Random Jitter Removal from Jitter Analysis • Other Tools to Consider for Jitter Analysis • Summary
Advanced Jitter Analysis 28 Crosstalk Identification • Which signals are coupling onto your victim? Crosstalk Quantification • How much error and jitter do each aggressor add to your victim? Crosstalk Removal for Jitter Analysis • What would your signal look without crosstalk present on victim? • How much jitter margin can be recovered without crosstalk? • If the signal was failing the jitter spec, can it pass without crosstalk?
Assist in making important design decisions: • Is it worth reducing crosstalk impact in design? • Where to improve?
29 Power supply aggressor signal
Original serial data victim signal
Serial data victim signal with crosstalk removed
Eye diagram with crosstalk
Eye diagram without crosstalk
30 1. Analyze up to four signals (victim or aggressor) at once. 2. Remove Near-End Crosstalk (NEXT), Far-End Crosstalk (FEXT) and Power Supply Crosstalk from Victim signal. 3. Plot waveform without crosstalk on the scope which can be: • Used for eye diagram, jitter decomposition, de- embedding, equalization and mask test • Saved as a waveform file
31 1. Probe up to 4 signals (Aggressors or victims). No simulation models or inputs are required.
2. Setup the victim signal.
4. The app reports the amount of crosstalk from each aggressors and return a waveform without crosstalk for analysis.
32 32 Power Supply Crosstalk on Victim
No Power Supply Aggressor With Power Supply Aggressor on the Transmitter PLL
Noise and Data TIE correlates Clean Power Noise Power Supply Supply
TIE Trend TIE Trend
33 Removing Power Supply Crosstalk from Victim
With Power Supply Crosstalk on the Power Supply Crosstalk Removed with Transmitter PLL Improvement on Data TIE Trend
TIE Trend with glitch removed
Noise Power Noise Power Supply Supply
TIE Trend TIE Trend
Advanced Jitter Analysis 34 34 Removing Power Supply Crosstalk from Victim Measured Victim Without Crosstalk
Measured Victim with Power Supply Crosstalk
Eye diagram correlates
Victim after Power Supply 35 Crosstalk removed Compare jitter results before and after crosstalk removal.
Jitter with Crosstalk
Jitter without Crosstalk TJ = 158ps PJdd = 58ps DJdd = 68ps An Improvement of 20% to Total TJ = 124ps Jitter without Crosstalk. PJdd = 27ps DJdd = 33ps
36 • The crosstalk application can remove the ideal, ISI and return “Unknown Crosstalk + Noise” (residual) content. • Perform further analysis on this residual waveform with measurements such as FFT, markers, etc. to root cause the source of aggressor.
Remove: Ideal + ISI of Victim Show: Only “Unknown Crosstalk + Noise”
Measured waveform = Ideal + ISI + Unknown XT + Noise
150 MHz clock coupled into the serial data signal inside the Simulated waveform = Unknown XT + Noise package
FFT on “Unknown Crosstalk + Noise”
37 37 • Review of Jitter Decomposition • Assumptions and Limitations • Spectral vs. Tail Fit Method • Advanced Jitter Analysis with Crosstalk Removal Tool • Scope Random Jitter Removal from Jitter Analysis • Other Tools to Consider for Jitter Analysis • Summary
38 Random jitter will vary with slew rates.
1. Every scope has intrinsic vertical noise floor. This vertical noise can translate into horizontal jitter. 2. As signal slew rate decreases, vertical noise increases the random jitter. 3. Measured random jitter is a function of signal slew rate, scope noise and scope sample clock jitter.
39 Calibrate and remove scope random jitter contribution
• Scope RJ calibration is available to remove the contribution of scope noise to measured RJ.
• User is asked to disconnect the signal from Channel to measure the ACVrms noise for the current Vertical setting.
40 Gain Margin by removal of Scope contribution to RJ
Signal with Fast Rise Time
DUT TX
With no Scope RJ removal
Lossy DUT TX Channel After Scope RJ removal
Gain margin through scope RJ removal.
41 • Review of Jitter Decomposition • Assumptions and Limitations • Spectral vs. Tail Fit Method • Jitter Analysis with Crosstalk Removal Tool • Scope Random Jitter Removal from Jitter Analysis • Other Tools to Consider for Jitter Analysis • Summary
42 Estimate Jitter and Eye Opening to various BER level
– Specify the BER eye contours you want the scope to plot.
– Specify which BER contour to highlight in red.
BER Eye Contours
Eye Contour at BER 10-12
43
Jitter Analysis with De-embedding and Equalization Connector Txp Connector Rxp + + Tx Channel Rx EQ - Txn Rxn -
After Scope De-embedding to the TX point Measurement Node After Scope Equalization
44 • Review of Jitter Decomposition • Assumptions and Limitations • Spectral vs. Tail Fit Method • Jitter Analysis with Crosstalk Removal Tool • Scope Random Jitter Removal from Jitter Analysis • Other Tools to Consider for Jitter Analysis • Summary
Advanced Jitter Analysis 45 45 N5400A EZJIT Plus for N8823A EZJIT Complete E2688A High-Speed N8827A PAM-4 Clock Jitter Analysis and RJ for Vertical Noise SDA for Reference Recovery Scope Removal Calibration Analysis Clock Recovery and Eye Analysis
N8833A Crosstalk Analysis N5461A Serial Data N5465A InfiniiSim BER Eye Contour and Removal Application Equalization Software De-embedding Comes standard with Software E2688A and N8823A
46 Histogram Object 1.2
likely to contain PJ 1
0.8 time error 0.6
0.4
PJ threshold 0.2
0 0 0 freq -0.2 0 5 10 15 Dual Dirac Model for Jitter Spectral vs. Tail Fit for ABUJ Decomposition (Crosstalk) Jitter Analysis
Slope discontinuity
Use Smoothness of Slop Continuity on the Use Crosstalk Removal Tool to Recover Bathtub Curve Jitter Margins
Scope Random Jitter Removal BER Eye Contour, De-embedding and Equalization for Jitter Analysis
47