Embed Size (px)
Citation preview
This document is owned by Agilent Technologies, but is no longer kept current and may contain obsolete or
inaccurate references. We regret any inconvenience this may cause. For the latest information on Agilent’s
line of EEsof electronic design automation (EDA) products and services, please go to:
www.agilent.com/fi nd/eesof
Agilent EEsof EDA
HUAWEI TECHNOLOGIES CO., LTD.Page 1
Serial link Simulation using
Advanced Design System
Serial link Simulation using
Advanced Design System
Page 2
Serial Link Challenge • Analog Components in a Serial Link
– Interconnects • Modeling impedance mismatch, crosstalk, dispersion, dielectric losses
– Frequency domain models derived from simulation or measurements
– I/O Models• IBIS and non-linear transistor level models
• Digital Components in a Serial Link
– SERDES Models• De-emphasis• Decision Feedback Equalizer• Feed Forward Equalizer• Clock and Data Recovery• Gain Controls
Best represented in frequency domain
Best represented in time domain
Best represented in numeric domain
Accurate representation of RF, analog, and digital components using ADS
Page 3
Agenda
• High-Speed Serial Link: Design Flow
• Transmitter: Jitter Modulation, De-emphasis Modeling
• Analog Passive Channel: Interconnect Modeling
• Receiver: Jitter Modulation , Equalization Modeling
• Post-Processing: Statistical Eye, BER/Bathtub
• Validation of Results and Examples
Paper contributed by: Huang Chunxing ([email protected]),Cory Edelman ([email protected]), & Sanjeev Gupta ([email protected])
Transmitter Analog Passive Channel Receiver
Digital + Analog Analog Analog + Digital
Page 4
High-Speed Serial Link Design
Data Signal FEC & Coding Pre-emphasisRJ, DJ,DDJ,PJ Tx
Transmitter
Rx Feed Forward Equalizer
Decision Feedback Eq.
CDR
Receiver
1. Component designer – Ensures that
components meet jitter budget specifications
2. Link Architect – Link jitter analysis, BER, De-
emphasis, FFE, DFE
Statistical post processing
Analog Passive Channel
Page 5
High Speed Serial Link - Design Flow
MATLAB Co-simulation in ADS
Transmitter
Analog Channel
Receiver
Page 6
Why Huawei used ADS?A time-domain simulator that accurately finds the response of frequency-domain measurements and models – how channels are characterized
ADS 2006 Update 1 has a new, exclusive, patent-pending simulator that accurately
predicts time domain response directly from multi-port S-parameter models
A link-level tool that can model equalizers and pre-emphasis, analog channel, and verify performance by analyzing jitter
ADS Ptolemy supports RF and DSP co-simulation. ( mixed signal simulation)
Deterministic jitter is bounded and may be simulated within an acceptable time
Co-simulation with external tools
ADS supports MATLAB co-simulation
Able to use fast behavioral driver, buffer and receiver models
ADS 2006 Update 1 supports IBIS models
Page 7
DCD: amplitude of nominal eye crossing is other than the 50% level
Data CorrelatedTJ(t)=RJ(t) + PJ(t) + DDJ(t)+…
Transmitter - Jitter Component
Data Un-Correlated
Transmitter Digital + Analog
ISI: largest of the difference between the earliest falling/rising edge and latest falling/rising edges.
DDJ: difference in the position of the earliest edge (rising or falling) and the latest (rising or falling) edge.
Page 8
Transmitter Modeling - Jitter ModulationUsing Voltage Controlled Delay to add jitterTransmitted jitter includes Periodic Jitter or PJ, and Duty Cycle Distortion or DCDRandom Jitter or RJ will be considered in statistical post-processing
∑=
+=N
iiiitotal twAtPJ
0
)cos()( θSinusoidal Source
DCD
Pattern Generator Voltage ControlDelay
)]2/()2/([*5.0 WxWxDCD ++−= δδ
Signal with jitter
Page 9
Transmitter Modeling – De-emphasisPre-distortion of a digital signal to compensate for high frequency losses in a channel
Both pre-emphasis and de-emphasis could be expressed as Finite Impulse Response or FIR filter
0 0.1 0.2 0.3 0.4 0.50
0.5
1
1.5
2
2.5
3
3.5
4Frequency Response
Normalized
dB
Page 10
High-Speed Serial Channel Components
Transmitter Analog Passive Channel Receiver
• Connectors• Line-cards• Backplane
Page 11
Serial Link- Channel Representation
S-Parameter
• Measurement-based models – Network Analyzer + • Physical models – EM solution of Maxwell’s equations +• Analytical models – closed form representation
+
+
Page 12
Channel ModelsChannel models, including forward response and crosstalk, are usually best expressed using S-parameters
ADS provide accurate time domain simulations directly from S-parameter/Touchstone files obtained from Network Analyzer measurements or EM solvers
Channel Impulse response:Derived from direct S-parameter simulation using ADS High FrequencySPICE
Comparison of S-Parameters derived from the impulse response of ADS SPICE simulator with original data
Analog Passive Channel Analog
Page 13
ADS Ptolemy for Serial Link AnalysisFebruary, 2007Page 13
Co-Simulation in ADS ( Analog/Digital Mixed Signal Simulation) (Molex 10Gb Backplane)
Page 14
Receiver Modeling- EqualizationThe Impulse Response h(t) embodies all of the
information contained in a circuit element.
To obtain the optimum tap coefficients, we need to
invert the process:
δ (t) h(t)
h(t) e(t)Equalizer
Delay . . .Delay
+
x x x x
Transmission path
Page 15
Receiver Modeling - Linear Feedforward Equalizer (LFE)
Delay . . .Delayr(n) r(n − 1) r(n − 2) r(n − (N-1))
+
x xTap0 Tap1 xTap2 xTapN-1
e(n)
“Taps” are the corrections the equalizer uses to open the eye• A weight is applied to N-1 preceding bits to correct
the current bit…
∑−
=
−=1
0
)()()(N
k
knrkfne
f(k) are the LFE taps, an LFE “filter”
∑−
=
−=1
0
)()(N
kk knrTapnegives or
Page 16
Decision Feedback Equalizer
M tap Feedbackb(n-τ)
N tap LFEf(n)
SymbolDetector
r(n) + e(n)+
−
Delay, τ
LFE taps are tuned to 1. remove ISI over N pre-cursors and 2. to minimize noise gain, f∗fThe M feedback taps, b(n), cancel the residual post-cursor ISI
after the LFE.
Page 17
Decision Feedback Equalizer
LFET . . .Tr(n) r(n − 1) r(n − 2) r(n − (N-1))
+x xf(0) f(1) xf(2) xf(N-1)
++
−Feedback filter
T T . . .s(n) s(n−2)s(n−1) s(n−(Μ−1))
x x x xb(0) b(1) b(2) b(M−1)
+
e(n)
Delay, τ
SymbolDetector
^ ^ ^ ^
ADS
Page 18
Equalizer- Algorithms
Algorithm calculates the optimal sampling point and equalizer coefficients
LMS adaptive algorithm
The LFE taps are set to maximize the eye-opening.Optimization techniques seek to determine the maximum of a function…The problem:
Find values for fn to minimize the function
Zero Forcing algorithm
Solve the following N-equations
( )21
0
2 )()()()()( ∑ ∑∑ ⎥⎦
⎤⎢⎣
⎡−−=−
−
=n
N
knknrkfnsnens
0)()()(
0)0()()0(
1
0
1
0
=−−
=−−
∑
∑
−
=
−
=
N
k
N
k
kNrkfNs
krkfs
M
See William Press, Numerical Recipes, Chapter 10
But zero forcing • does not give the best SNR • is not unique
Page 19
Equalization Type
∑−
=
−=1
0)()()(
N
kknrkfne
• Non-adaptive EqualizersTaps coefficient are fixed
• Adaptive EqualizersAdaptive equalizers adjust their taps on the fly
Argument for including a known training pattern in the protocol
Training pattern the ideal decision assumption is true There are many proprietary adaptive equalizers without trainingsequences: “blind equalizer.”
Page 20
Decision Feedback Equalization (Non Adaptive)
DFE
Page 21
Decision Feedback Equalization - Adaptive
DFE
Analog Channel
Training Sequence
Page 22
DFE – 1 Sample per Bit
DFE is decision feed-back equalizer
Page 23
Continuous-Time Equalizer
CTE stands for continuous time equalizer. A CTE model is an ideal circuit that
can adjust zeros and pole positions to get the desired frequency response.
As with FFE, CTE has a noise gain problem. CTE with only one pole and
one zero can be expressed by the following equation:
This is typically implemented in the analog domain.
)()()(
0
0
0
0
pszs
zpafHCTE +
+=
Page 24
Determining Link Performance – Statistical Post Processing
Calculate Forward channel PDF:
1. Interpolate through response to sufficiently
small time interval
2. Overlap each bit in one Unit Interval (UI)
range
3. At each UI sampling point, compute PDF
of Overlapping data.
Zero
Amplitude PDF at each point
Why statistical post processing?
• Predict BER performance with minimal bit pattern
Page 25
Statistical Post Processing - Random Jitter Effects on Link Performance
Zero
Amplitude PDF at each point
Signal PDF
Random Jitter distribution
Overall PDF
∫∞
∞−
⋅+= υυυττ dPISIPISIP RJ )(),(),(
2
2
2
21 σ
σπ
t
RJ eP−
=
Page 26
Statistical Post-Processing Using MATLAB®
Co-simulation in ADSStatistical Eye “StatEye”
Statistical eye is a set of probability contours
Horizontal axis is with respect to sampling point
Vertical axis is with respect to signal amplitude
Different colored lines represent BER
MatlabSinkM6
NumberOfFirings=1MatlabWrapUp=""MatlabFunction=""MatlabSetUp=""DeleteOldFigures=YESScriptDirectory=""
MatlabCx_MM1
MatlabWrapUp=""MatlabFunction=""MatlabSetUp=""DeleteOldFigures=YESScriptDirectory=""
Matlab_MM8
MatlabWrapUp=""MatlabFunction=""MatlabSetUp=""DeleteOldFigures=YESScriptDirectory=""
MatlabSinkFM7
NumberOfFirings=1MatlabWrapUp=""MatlabFunction=""MatlabSetUp=""DeleteOldFigures=YESScriptDirectory=""
MatlabFCx_MM2
MatlabWrapUp=""MatlabFunction=""MatlabSetUp=""DeleteOldFigures=YESScriptDirectory=""
MatlabF_MM3
MatlabWrapUp=""MatlabFunction=""MatlabSetUp=""DeleteOldFigures=YESScriptDirectory=""
Page 27
Bathtub
Intercept through statistical eye along horizontal axis to get horizontal Bathtub at definite voltage
Intercept through statistical eye along vertical axis to get vertical Bathtub at definite sampling time
Horizontal
Vertical
Statistical Post-Processing Using MATLAB®
Co-simulation in ADS
Page 28
Simulated vs. Measured
Validation of Simulation Results– Using Agilent N4903 J-BERT signal generator; 86100C Infinium wideband scope
– N4903 internal random jitter (RJ) used in all cases shown
0.010 Unit Interval RJ 0.015 Unit Interval RJ
Page 29
Validation of Simulation Results (2)
0.05 Unit Interval PJ, .011 RJ 0.015 Unit Interval PJ, .018 RJ
Simulated vs. Measured
Page 30
Example - Channel pre-simulation
Assumed conditions:
10Gbps NRZ, PRBS 23, 100000 bits, amplitude 800mvpp, Tr(f) = 24ps
0.15UIpp DJ = 0.05UI DCD+0.1UI PJ, 0.15UIpp RJ @10-12BER
BER = 1e-12 (2*Q=14.069)
3-tap de-emphasis, 5-tap DFE
Slice voltage: 10mV
Forward Channel Insertion Loss: Eight crosstalk channels:
Page 31
Example - Channel pre-simulation
Time Offset (UI)
Am
plitu
de
Statistical Eye=0.028, Width=0.222
0 0.2 0.4 0.6 0.8 1-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
1
2
3
4
5
6
7
8
9
10
11
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-12
-10
-8
-6
-4
-2
0
Time Offset (UI)
BE
R
Bathtub at threshold=0.010
Simulation results:
28mv eye-height
0.222UI eye-width
This indicates that the channel could support 10Gbps data transmission
Page 32
ADS2006 Update 2 will provide root cause analysis of jitter problems
• Provides confidence that the BER simulation results will match the measured performance.
• Provides jitter diagnosis, which can reduce the product development cycle from weeks to hours.
• Allows the designer to systematically tweak the link for best performance by understanding the underlying issues
Going Forward…
““ADS is the only commercially available EDA product that
provides jitter diagnosis and results that can be trusted”
Page 33
Jitter Analysis Capability in ADS2006 Update 2
-1.0E-11-8.0E-12-6.0E-12-4.0E-12-2.0E-120.02.0E-124.0E-126.0E-128.0E-12
-1.2E-11
1.0E-11
100
200
300
400
0
500
indep(TJHist)
TJHi
st
-8.0E-12-6.0E-12-4.0E-12-2.0E-120.02.0E-124.0E-126.0E-128.0E-12
-1.0E-11
1.0E-11
20406080
100
0
120
indep(RJPJHist)
RJPJ
Hist
-4E-13
-2E-13
0 2E-13
4E-13
-6E-13
6E-130.2
0.4
0.6
0.8
0.0
1.0
indep(DDJHist)
DDJH
ist
indep(DDJFHist)
DDJF
Hist
indep(DDJRHist)
DDJR
Hist
DDJ Composite Histogram-1.0E-11-8.0E-12-6.0E-12-4.0E-12-2.0E-120.02.0E-124.0E-126.0E-128.0E-12
-1.2E-11
1.0E-11
100
200
300
400
0
500
indep(TJHist)
TJHi
st
indep(RJPJHist)
RJPJ
Hist
indep(DDJHist)
DDJH
ist
Composite Histogram
Page 34
Bathtub and Q Plots ADS2006 Update 2
0.2 0.4 0.6 0.80.0 1.0
1E-13
1E-9
1E-5
1E-1
1E-17
3E-1
UI
BTDa
taBT
Mdl
0.2 0.4 0.6 0.80.0 1.0
-10
-8
-6
-4
-2
-12
0
UI
BTQD
ata
BTQM
dl
TJpp
3.319E-11
RJrms
2.057E-12
DJdd
3.860E-12
PJdd
3.799E-12
PJrms
1.635E-12
ISIpp
3.248E-13
DCD
3.458E-14DDJpp
3.248E-13
Page 35
Overcoming the Serial Link Design Challenge Using ADS
• Analog Components in a Serial Link– Interconnects
• Modeling impedance mismatch , crosstalk, dispersion, dielectric losses– Frequency domain models derived from simulation or measurements
– I/O Models• IBIS and non-linear transistor level non linear models
• Digital Components in a Serial Link
– SERDES Models• De-emphasis• Decision Feedback Equalizer• Feed Forward Equalizer• Clock and Data Recovery• Gain Controls
Best represented in frequency domain
Best represented in time domain
Best represented in numeric domain
Accurate representation of RF, analog, and digital components using ADS
Page 36
Summary:Introduced: Serial Link Design and Verification Design Flow
The basic methodology of statistical eye simulation in ADS has been presented
ADS is shown to be an ideal platform to realize statistical eye simulation:
Jitter Modulation at transmitter
Forward channel, crosstalk channel models
Include pre-emphasis or de-emphasis
Equalizers: Adaptive LFE, DFE
Statistical post-processing: StatEye, Jitter separation, BER bathtub
Presented: A brief overview of new jitter analysis capabilitiesin ADS
www.agilent.com/fi nd/emailupdatesGet the latest information on the products and applications you select.
www.agilent.com/fi nd/agilentdirectQuickly choose and use your test equipment solutions with confi dence.
Agilent Email Updates
Agilent Direct
www.agilent.comFor more information on Agilent Technologies’ products, applications or services, please contact your local Agilent office. The complete list is available at:www.agilent.com/fi nd/contactus
AmericasCanada (877) 894-4414 Latin America 305 269 7500United States (800) 829-4444
Asia Pacifi cAustralia 1 800 629 485China 800 810 0189Hong Kong 800 938 693India 1 800 112 929Japan 0120 (421) 345Korea 080 769 0800Malaysia 1 800 888 848Singapore 1 800 375 8100Taiwan 0800 047 866Thailand 1 800 226 008
Europe & Middle EastAustria 0820 87 44 11Belgium 32 (0) 2 404 93 40 Denmark 45 70 13 15 15Finland 358 (0) 10 855 2100France 0825 010 700* *0.125 €/minuteGermany 01805 24 6333** **0.14 €/minuteIreland 1890 924 204Israel 972-3-9288-504/544Italy 39 02 92 60 8484Netherlands 31 (0) 20 547 2111Spain 34 (91) 631 3300Sweden 0200-88 22 55Switzerland 0800 80 53 53United Kingdom 44 (0) 118 9276201Other European Countries: www.agilent.com/fi nd/contactusRevised: March 27, 2008
Product specifi cations and descriptions in this document subject to change without notice.
© Agilent Technologies, Inc. 2008
For more information about Agilent EEsof EDA, visit:
www.agilent.com/fi nd/eesof
