Signal Integrity Analysis of Gigabit Interconnects Olie Kreidler Tektronix, Inc. 1 Tektronix Confidential Signal Integrity (SI): Digital Becomes Analog “At high frequencies … crosstalk and signal reflections can be perceived as logic triggers, and can be responsible for erroneous signal patterns” – EE Times, April 17, 1998, Special Section on Interconnects 2 Tektronix Confidential Industry Trends and Issues On-going trends and issues – Trends: faster rise times, clock frequencies, increasing interconnect complexity – Requirement: increasing need for signal integrity analysis and SPICE / IBIS interconnect modeling New trends and requirements – Trends: S-parameters and eye diagrams are becoming part of compliance testing for passive PHY All standards currently are differential and serial – Requirements Eye diagram and S-parameter compliance testing must be performed in differential mode Frequency dependent losses need to be modeled 3 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues TDR/T and VNA Measurement Basics App Note: “TDR and VNA Measurement Primer” – Impedance Measurements and IConnect® True Impedance Profile – Time Domain S-parameter Measurements – Eye Diagram Measurements – TDR Probing and Fixturing 4 Tektronix Confidential High-Speed Serial Data Link Analysis The Measurement Challenge - A Closed Eye An “Open Eye” at the Transmitter a “Closed Eye” at the Receiver How to measure this eye? + + - - path + + - - + + - - path + + - - EQUALIZER Tx Rcv Rcv path The risetime of the channel is relatively slow compared to the very fast 1’st channel. i.e. modern channel’s risetime is quite longer than the UI, thus the eye blures and ‘ISI’-s itself. 5 2007/06/10 Confidential V1.1 High-Speed Serial Data Link Analysis The Serial Data’s Solution to a Closed Eye An “Open Eye” at the Transmitter a “Closed Eye” at the Receiver Equalize it! + + path + + + + path + + EQUALIZER Tx Rx + Rcv Rcv EQ. - - - - - - - -- - path Equalization is the answer for the digital receiver. The eye opens at the input of the receiver, the receiver can decode the signal. 6 2007/06/10 Confidential V1.1 High-Speed Serial Data Link Analysis What should the Measurement do? - simple: do what your Tx/Rx does: implement equalization! An “Open Eye” at the Transmitter a “Closed Eye” at the Receiver Equalize it! + + path + + + + path + + EQUALIZER Tx Rx + Rcv Rcv EQ. - - - - - - - -- - path Equalization is the answer to the eSerial receiver, so SW-implemented Equalization on the scope is also the answer to T&M. -opens the eye for display (the scope-‘receiver’) -Lets the user view ‘the inside’ of the Receiver 7 2007/06/10 Confidential This is now in your scope V1.1 Tek Solution DSA8200 with the TDR Module 80E04 reflected rise time: – 35 ps 80E10 reflected rise time – 12ps / 15ps 8 acquisition channels – 8-port TDR – 4-port True Differential TDR Continuously stabilized rho and impedance waveforms All standard measurements available on rho and impedance waveforms 8 Tektronix Confidential TDR Overview - Typical System Reflection Incident Sampler Reflections Incident Step Incident Step Probe 50 W 50 W t = 0 Impedance reference Test device Step Generator TDR Waveform Characteristics TDR systems observe the superposition of incident and reflected signals at source Time separation t1-t0 assures ability to discern difference Vreflected Vincident t0 t1 Time TDR Rho Units Definition Amplitude Vreflected Vincident and 0 at Z Z 0 +1 Vreflected 0 Characteristic (Z = Z0) Vincident -1 t0 t1 Time KCL Applied at Discontinuity Transmission lines support propagation with specific characteristic impedance Z Reflected and forward propagating signals will be such that Si = 0 is satisfied at discontinuity Can easily solve for Z knowing , Z0, and KCL for lumped circuits Forward Incident Z 0 Step Source Reflected (Z = Z0) Discontinuit y Solution for Z Units Where 1 Z Z 0 1 – Z0 is the known reference impedance – the sampling oscilloscope directly measures – Z is the calculated test device impedance Note textbooks usually show reversed expression: Z Z0 Z Z0 TDR Waveforms - Simple Cases Waveforms with Open, Short and 50W terminations Amplitude +1 Open (Z =) reflected =+1 0 (Z = 50W) incident=+1 reflected =-1 -1 Short (Z = 0) t0 t1 Time TDS Measurement Basics TDR Oscilloscope Front Rsource Panel TDR Block Diagram V Vreflected Vincident Cable: Z 0, td DUT: Z DUT Rsource = 50 W Z0 = 50 W then V incident = ½V Ztermination Open circuit V V •Z load / (Z load + Z 0) Matched load ½V Short circuit 0 15 Tektronix Confidential Z load > Z 0 ½V 0 Z load < Z 0 TDR Measurements Basics Inductance and Capacitance Analysis Shunt C discontinuity ½V Z0 Z0 0 Series L discontinuity ½V Z0 Z0 0 L-C discontinuity ½V Z0 Z0 0 C-L-C discontinuity ½V Z0 0 16 Tektronix Confidential Z0 Measurement Tools: Z-line Z-line-Based Measurements t 1 2 L Z ( t )dt 2 t1 t2 t 1 2 1 C dt 2 t1 Z ( t ) t1 t1 t2 17 Tektronix Confidential TDR Measurements Basics TDR Rise Time and Resolution Accepted rule of thumb for resolving two discontinuities tseparate a1 a2 To resolve a1 and a2 as separate discontinuities: tseparate > tTDR_risetime /2 80E04 TDR rise time: 30-40ps at the end of the cable, probe, fixture – Base 1/2trise resolution: 15-20ps – 0.1”-0.12” in FR4 80E10 TDR rise time: 12-16ps at the end of the cable, probe, fixture – Base 1/2trise resolution: 6-8ps – 0.04”-0.048” in FR4 18 Tektronix Confidential TDR Measurements Basics TDR Rise Time and Resolution More real case: resolving a single discontinuity tsingle a1 a1 is not resolved if tsingle << tTDR_risetime Going beyond the TDR resolution and risetime: relative techniques – Signal integrity modeling – JEDEC standard – Failure analysis – golden device comparisons 19 Tektronix Confidential TDR Measurements Basics TDR Rise Time and Resolution If 30-40 ps (or 12 ps) fast TDR How in the world a 80 ps rise time does not resolve it …. signal rise time will???????????? Conclusion: for SI analysis, use the actual DUT rise time! (filter down the rise time, if necessary) 20 Tektronix Confidential TDR Measurements Basics Differential TDR Differential serial link analysis Virtual ground plane Even and odd mode measurements TDR Oscilloscope Front Panel Rsource V V Rsource 21 Tektronix Confidential Cable: Z 0, td Virtual ground Vincident Vreflected Cable: Z 0, td DUT: Z DUT TDR Measurements Basics Good Measurement Practices Perform calibration routines regularly Minimum warm-up time 20 minutes Maintain constant temperature in the lab and check the instrument t° Zoom in on the DUT – but include all the DUT signature transitions (more to follow) Use torque wrenches when mating SMA or other RF connectors 22 Tektronix Confidential Time and Frequency Domains VNA Block Diagram Rsource Vincident1 Vref lected1 DUT Port 2 V Calibration procedures: - SOLT - TRL - LRRM Port 1 VNA Front Panel Cable: Z 0, td Vtransmitted2 VNA: Vector Network Analyzer Similar diagram can be drawn for reverse measurements (port 2 to port 1) Differential VNA: 4-port measurements 23 Tektronix Confidential Time and Frequency Domains Equations for TDR vs. VNA TDR Vreflected Vincident S11 VNA Vreflected1 Vincident1 Z load Z 0 Z load Z 0 Z input ( DUT ) Z 0 Z input ( DUT ) Z 0 Z DUT 1 Z0 1 Z input ( DUT ) 1 S11 Z0 1 S11 S11 ( f ) Duration Limited FFT ( (t )) S21 ( f ) Duration Limited FFT ( (t )) 24 Tektronix Confidential Time and Frequency Domains TDR vs. VNA TDNA (Time Domain Network Analysis) – – – – – – Based on TDR/T measurements: Transient Broadband More intuitive for a digital designer Dynamic range: about 50-60dB Less expensive FDNA (Frequency Domain Network Analysis) – – – – – – Based on VNA measurements: Steady-state measurements Narrow-band More intuitive for microwave/RF designer More expensive Higher dynamic range (up to 110 dB) 25 Tektronix Confidential Time and Frequency Domains Time or Frequency Domain? SI measurements do not require high dynamic range VHIGH VLOW -40dB equals 1% in time domain 1% (-40dB) Xtalk Compliance testing does not require high DR – About –10 dB for insertion loss – -25 to –35 dB for return loss – Higher for frequency domain crosstalk 26 Tektronix Confidential Impedance Accuracy TDR Basic Equations Vreflected Vincident Vreflected Z DUT Z load Z 0 Z load Z 0 2 td Vincident Z0 Z Vincident × DUT Z DUT Z 0 V incident V reflected 1 Z0 Z0 1 V incident V reflected 27 Tektronix Confidential Vmeasured = Vincident +Vreflected 0 Z0 V measured 2 V incident V measured Impedance Accuracy TDR Multiple Reflection Effects Issue: impedance accuracy suffers due to signal re-reflection inside the DUT Z0 Z1 Z2 Vtransmitted1 Vreflected1 Vreflected2 t0 Time 28 Direction of propagation Tektronix Confidential Z3 Z4 Impedance Accuracy IConnect Computation of the True Impedance Profile V reflected 1 1V incident 1 V reflected 2 t 12 2V incident 1 1V incident 2 V reflected 3 ( t 12 t 22 32 t 12 22 1 )V incident 1 t 12 2V incident 2 1V incident 3 Vreflected1 k1 V reflected 2 k 2 Vreflected 3 k3 Vreflectedn k n 29 Tektronix Confidential 0 0 k1 0 k2 k1 k n1 kn 2 0 Vincident1 V 0 0 incident 2 0 Vincident 3 0 k1 Vincidentn Impedance Accuracy Board Trace IConnect® Z-line Multiple reflections in TDR waveform Scope reads here about 44 Ohm instead of 50 Ohm 30 Tektronix Confidential Impedance Accuracy Board Trace IConnect® Z-line Accurate impedance profile in IConnect® 31 Tektronix Confidential Impedance Accuracy Package Trace IConnect® Z-line Raw TDR: confusing multiple reflections Impedance profile in IConnect®: Exact failure location, improved resolution 32 Tektronix Confidential Impedance Accuracy IConnect® Software Z-line Algorithm TDR measurements suffer from multiple reflections – No limited to TDR, also in TD in VNA IConnect removes multiple reflections – Ensures accurate impedance measurements in multi-impedance DUT – Direct and accurate readout of Z, td, L, C – Different and more accurate than Z readout in the scope Attention!: – Data noise may interfere with accuracy Use scope averaging Use software noise filtering – Line loss is extracted separately 33 Tektronix Confidential Frequency Dependent S-parameters Why S-parameters Compliance testing – Insertion loss (around 6-10 dB) – Return loss (around 20-30 dB) – Frequency domain crosstalk Link performance evaluation and simulation – Simulate S-parameters directly 34 Tektronix Confidential Frequency Dependent S-parameters Equations for TDR vs. VNA TDR Vreflected Vincident S11 VNA Vreflected1 Vincident1 Z load Z 0 Z load Z 0 Z input ( DUT ) Z 0 Z input ( DUT ) Z 0 Z DUT 1 Z0 1 Z input ( DUT ) 1 S11 Z0 1 S11 S11 ( f ) Duration Limited FFT ( (t )) S21 ( f ) Duration Limited FFT ( (t )) 35 Tektronix Confidential Frequency Dependent S-parameters Single-ended TDR TDR stimulus on channel 1, TDR stimulus on channel 2, response on channel 1 response on channel 1 S11 TDR11 S TDT 21 21 S12 TDT12 S 22 TDR22 TDR stimulus on channel 1, response on channel 2 36 Tektronix Confidential TDR stimulus on channel 2, response on channel 2 Frequency Dependent S-parameters Correct Data Acquisition DUT waveform to settle to steady DC level 37 Tektronix Confidential Frequency Dependent S-parameters Return Loss = TDR Measure TDR, compute S11 (return loss) in IConnect What is wrong with this RL picture? 38 Tektronix Confidential Frequency Dependent S-parameters Insertion Loss = TDT Measure TDT, compute S21 (insertion loss) in IConnect Test case for loss extraction 39 Tektronix Confidential Frequency Dependent S-parameters Differential and Mixed S-parameters Differential TDR stimulus, differential response (most important) Sdd 11 TDRdd 11 S dd 21 TDTdd 21 S cd 11 TDRcd 11 S cd 21 TDTcd 21 S dd 12 TDTdd 12 S dc 11 TDRdc11 S dd 22 TDRdd 22 S dc 21 TDRdc 21 S cd 12 TDTcd 12 S cc11 TDRcc11 S cd 22 TDTcd 22 S cc 21 TDTcc 21 Differential TDR stimulus, common mode response (useful in time domain for EMI troubleshooting) 40 Tektronix Confidential Common mode TDR stimulus, differential response (useful in time domain for EMI troubleshooting) S dc12 TDTdc 12 S dc 22 TDRdc 22 S cc12 TDTcc12 S cc 22 TDRcc 22 Common mode TDR stimulus, common mode response (less important) Frequency Dependent S-parameters Differential TDR = S11diff Several InfiniBand traces of different length. Differential return loss. 41 Tektronix Confidential Demo of TDR and S parameter measurements 42 Tektronix Confidential Additional Material 43 Tektronix Confidential 44 Tektronix Confidential Frequency Dependent S-parameters Power Plane Resonance Observe plane impedance profile (Z-line) Resonances between planes 45 Tektronix Confidential Frequency Dependent S-parameters Electrical Compliance Testing Need fixture to interface to interconnects for compliance testing Reference thru Traces DUT Connection Traces Infiniband Connector 46 Tektronix Confidential Frequency Dependent S-parameters VNA Fixture De-embedding For VNA requires additional standard to de-embed properly Insertion loss: fixture insertion loss must be subtracted from DUT insertion loss Return loss: no way to de-embed without additional standards! – Fixture return loss ends up being lumped with the DUT return loss – Can be a problem even with quality fixtures 47 Tektronix Confidential Frequency Dependent S-parameters VNA Real Fixture Limitation Example Spec: -10 dB at 1.25 GHz With fixture, the cable assembly is failing the spec! Fixture is failing the assembly with VNA measurements 48 Tektronix Confidential Frequency Dependent S-parameters TD-VNA Fixture De-embedding Simplicity of calibration allows simple fixture de-embedding Spec: -10 dB at 1.25 GHz Fixture de-embedded with TDNA, the assembly is passing 49 Tektronix Confidential Frequency Dependent S-parameters Correlation with Network Analyzer 50 Tektronix Confidential Frequency Dependent S-parameters Correlation with Network Analyzer -5 -10 VNA SDD11, dB -15 IConnect® S11.wfm(dBMag). No TDR calibration -20 -25 51 Tektronix Confidential 9.5 9.0 8.5 8.0 7.5 7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 FREQ, Ghz 0 Frequency Dependent S-parameters Correlation with Network Analyzer Insertion Loss VNA TDA Comparison Data courtesy Kieran Kelly, Samtec,Inc. 0 -2 IL (dB) -4 VNA TDA -6 -8 -10 -12 000.0E+ 1.0E+9 0 2.0E+9 3.0E+9 4.0E+9 5.0E+9 6.0E+9 Frequency (Hz) 52 Tektronix Confidential 7.0E+9 8.0E+9 9.0E+9 10.0E+9 Frequency Dependent S-parameters Correlation with Network Analyzer: 65 GHz The TD-VNA bandwidth is These data were measured using the PSPL Model 4022 and a 70 GHz sampler -5 S21 magnatude (dB) directly related to TDR/T rise time 0 -10 -15 -20 -25 -30 -35 VNA S21 -40 4022 S21 -45 -50 0 S-parameters correlate to 10 20 30 40 50 60 frequency (GHz) 65 GHz 0 Picosecond Pulse Labs S11 magnatude (dB) Courtesy: Kipp Schoen, -10 -20 -30 VNA -40 4022 -50 -60 0 10 20 30 40 frequency (GHz) 53 Tektronix Confidential 50 60 Frequency Dependent S-parameters Calibrated Results: SOLT Dashed line – Tektronix 11801 with SOLT cal Solid line – Agilent 8510 VNA Excellent correlation between TDNA and FDNA data 54 Tektronix Confidential Frequency Dependent S-parameters Noise Floor and Dynamic Range N avg N DR( N , N avg ) DR( N 0 , N avg0 ) N0 N avg 0 To reduce noise floor: – Increase number of averages Navg – Increase number of points Npoints – Decrease acquisition window length (increase effective incident power) 55 Tektronix Confidential Frequency Dependent S-parameters TD-VNA Incident Effective Power The TD-VNA 4022 10ps 4022 10 ps 54754A25 25psps TDR -20 magnatude (dB) bandwidth directly related to the risetime of the TDR and TDT signals 0 -40 -60 -80 -100 -120 -140 0 20 40 60 frequency (GHz) 56 Tektronix Confidential 80 100 Frequency Dependent S-parameters TD-VNA Dynamic Range (with PSPL Module) 57 Tektronix Confidential Frequency Dependent S-parameters IConnect Produces S-parameters Differential, mixed mode and single ended Insertion, return loss and frequency domain crosstalk Performance with base DSA8200: 50-60 dB dynamic range (vs. 100 dB for VNA), 12 GHz bandwidth Performance with 80E10 up to 50GHz bandwidth Cost ½ of a comparable VNA solution Intuitive, easy to use and more than adequate dynamic range for digital designers 58 Tektronix Confidential TDT and IConnect Eye Diagram Efficient S-parameters Testing in IConnect Easy, quick, efficient S-parameter measurements and electrical compliance testing – Insertion, return loss, frequency dependent crosstalk – Excellent correlation with traditional VNA techniques – Cost-effective and quick Minimal calibration required – Only reference at the end of the fixture – Easy fixture de-embedding 59 Tektronix Confidential TDT and IConnect Eye Diagram Why Eye Diagram in IConnect? Eye Diagram for Interconnects – Specification mask testing – Not just communication standards, also for new serial link standards IConnect benefit: no pattern generator required for interconnect eye diagram analysis – De-embed deterministic / interconnect jitter – No active component jitter 60 Tektronix Confidential TDT and IConnect Eye Diagram Eye Diagram Degradation in Interconnects Interconnect losses Pattern-dependent, crosstalk induced jitter Method to improve the eye – Equalization, pre-emphasis and de-emphasis – Other signal conditioning techniques Only deterministic jitter exists in interconnects, no random component! 61 Tektronix Confidential TDT and IConnect Eye Diagram Eye Diagram Options TDT easily gives the eye diagram degradation – Deterministic jitter only 62 Tektronix Confidential TDT and IConnect Eye Diagram New Eye Mask and Jitter Measurements 63 Tektronix Confidential TDT and IConnect Eye Diagram Why is TDT Based Eye Better? Easy to de-embed fixture – The same improvement as for S-parameter measurements! No jitter from the pattern generator 64 Tektronix Confidential TDT and IConnect Eye Diagram Predicted and Measured Eye Diagrams Pattern Generator Based IConnect K28.5 65 Tektronix Confidential IConnect PRBS 210-1 TDT and IConnect Eye Diagram Predicted and Measured Eye Diagrams 2^10-1 pattern generator measurement Data Courtesy FCI Electronics 66 Tektronix Confidential 2^10-1 IConnect eye from TDT measurement TDT and IConnect Eye Diagram Predicted and Measured Eye Diagrams 1.5Gb/s (Gen 1) Simulated Measured 3.0Gb/s (Gen 2) Simulated Measured Serial ATA data courtesy Molex, Inc. 6.0Gb/s (Gen 3) Simulated 67 Tektronix Confidential Measured TDT and IConnect Eye Diagram Efficient Eye Diagram Testing in IConnect Easy, quick, efficient eye diagram measurements and compliance testing – Excellent correlation with traditional pattern generator techniques – Cost-effective and quick Minimal calibration required – Only reference at the end of the fixture – Easy fixture de-embedding 68 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues TDR/T Probing and Fixturing App Note: “TDR Measurement Primer” App Note: “TDR Techniques for Characterization and Modeling of Electronic Packaging” Quick Guide:“Interconnect Probing Quick Guide” Interconnect SPICE / IBIS Modeling and Model Validation 69 Tektronix Confidential Probing and Fixturing TDR Measurement Setup TDR probe with signal and ground connection TDR oscilloscope 70 Tektronix Confidential Probing and Fixturing Probing and Fixturing Issues Probing is the weakest link! Start with a probe – – – – – – 50 Ohm for TDR measurements must be rugged and inexpensive ensure stable repeatable contact large pitch* means small bandwidth variable pitch means poor repeatability ensure sufficient compliance * Pitch: center-to-center signal to ground pad spacing 71 Tektronix Confidential Probing and Fixturing Package and Connector Probing Use a high-quality probe (and positioner) Need an interface adapter or fixture to probe Fixturing requirements – Reproduce the real application environment – Ensure easy fixture de-embedding (reference short and open structures may be needed) Reference short and open Vias to package leads 72 SignalGround Probe PCB provides ground connections Fixture ground plane Package Tektronix Confidential SignalGround Probe HighSpeed Connector PCB provides ground connections PCB trace to connector lead Fixture ground plane Via to ground for reference measurements Probing and Fixturing Board Probing Ensure good contact to a via – Difficult for a microwave probe – Use TDA’s QuickTDR™ probe – Probes also available from TDR manufacturers Ensure ground contacts near your signals Variable pitch: a sad necessity – Available from from TDR manufacturers and probe manufacturers (Cascade Microtech, ICM) – Measurements suffer from poor repeatability and decrease the instrument usable bandwidth 73 Tektronix Confidential Probing and Fixturing Probing vs. Fixturing Probing advantages: Fixturing advantages: Maximum flexibility for multiple Evaluate the DUT in its intended device measurements No fixture de-embedding required But: environment of use (example: package on a board) Great flexibility for specific DUT But: Requires DUT to have easily accessible contact areas Positioning system may be expensive Difficult to change after the fixture has been designed Must de-embed fixturing from measurements These approaches are complementary! Fixturing is thinking ahead about how you will probe! 74 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues – Impedance, S-parameters and eye diagram measurements and compliance testing Interconnect SPICE / IBIS Modeling and Model Validation – Z-line, lossy line, and automatic behavioral modeling 75 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues Interconnect SPICE / IBIS Modeling and Model Validation Measurement Based Interconnect Analysis – Behavioral Modeling: MeasureXtractor™ – Topological Modeling: TDT and Lossy Line Modeling Impedance Profile (Z-line) Transmission Line Modeling L and C JEDEC computation – Examples: Power Plane Analysis Backplanes and cable assemblies 76 Tektronix Confidential Measurement-Based Link Design How to Analyze System Signal Integrity? Tx Board Connector Connector (simplified model) Cable Assembly Connector Daughtercard Connector Backplane Rx Is this similar to your application? App Note: “Signal Integrity Modeling of Gigabit Backplanes, Cables and Connectors Using TDR” 77 Tektronix Confidential Measurement-Based Link Design Interconnect Measurement Based Design Board Connector Connector (simplified model) Cable Assembly Connector Daughtercard Linearize the link input and termination for initial analysis 78 Tektronix Confidential Connector Backplane Receiver simplified Measurement-Based Link Design Measurement Based Design Details Impedance measurement => reflections TDR/T or S-parameters => losses, jitter, eye diagram degradation – System losses is a result of losses in components – Eye diagram is a result of losses and crosstalk in components 79 Tektronix Confidential IConnect Modeling Methodology Goals and Model Validity Goal: accurately predict interconnect performance via simulations – Need an accurate SPICE /IBIS models Model required range of validity is defined by the fast corner rise time of the driver – Equivalent bandwidth estimated as: fbw=0.35 / trise or harmonics of clock It may be desired to extend the required range of model validity beyond trise and fbw – Have a confidence guard band 80 Tektronix Confidential IConnect Modeling Methodology Modeling Technique: Behavioral Behavioral Modeling: MeasureXtractor™ – A universal, fully automatic, exact modeling technique – Can use time or frequency domain data – Matches exactly both time and frequency response – Perfect for… Connector, package or socket modeling Model for a characterization fixture for a connector, a package or a cable Model for a daughtercard board When behavioral model is acceptable – Can create large model for a large interconnect such as a backplane or cable assembly 81 Tektronix Confidential IConnect Modeling Methodology Modeling Technique: Topological Lossy line and coupled lossy line modeling – When need to predict losses and crosstalk Large lossy backplanes and motherboards Cables and cable assemblies Impedance profile (Z-line) models – When losses are small – Need to predict impedance reflections, crosstalk only Small daughtercards, boards Electrically long connectors, packages 82 Tektronix Confidential IConnect Modeling Methodology Modeling Technique: Topological JEDEC technique for L and C computation – Industry standard technique for electrically short interconnects – Electrically short: trise >> tprop delay – Packages, connectors, sockets tprop delay trise 83 Tektronix Confidential IConnect Modeling Methodology Behavioral or Topological? Behavioral Measurement Requires two-port or fourRequirements port measurements Topology Automatic, no user selection intervention Extraction Automatic, no user intervention Type of “Black-box,” no internal models changes allowed Limitation for Large model for long long interconnects (backplanes, interconnects cable assemblies) 84 Tektronix Confidential Topological Just TDR (reflection) is sufficient User-controlled (easy and intuitive from TDR measurements) User-driven; more labor intensive than behavioral Intuitive, easy “what-if” scenario analysis Efficient model extraction processes exist for large interconnects IConnect Modeling Methodology Methodology: Gbit Ethernet Example Launch, high-speed connector: Z-line modeling Any piece can be modeled in MeasureXtractor™ 85 Tektronix Confidential Cable and Test cards: lossy line modeling Lumped pieces can be modeled with JEDEC technique IConnect Modeling Methodology Measurement Based Approach TDR/T or VNA Measurements Extracted interconnect, instrument source models SPICE Direct link to simulators Automatic comparison of simulation and measurement in IConnect waveform viewer 86 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues Interconnect SPICE / IBIS Modeling and Model Validation Behavioral Modeling: MeasureXtractor™ – Topological Modeling: TDT and Lossy Line Modeling Impedance Profile (Z-line) Transmission Line Modeling L and C JEDEC computation – Examples: Power Plane Analysis Backplanes and cable assemblies 87 Tektronix Confidential Behavioral Modeling MeasureXtractor™ Modeling A fully automatic algorithm for conversion of VNA S-parameter or TDR/T data into SPICE or IBIS model – – – – – Passivity of the model guaranteed Compact and efficient Fully automated Not an optimization Behavioral models 88 Tektronix Confidential Behavioral Modeling Lack of Passivity Produces Oscillations Slide courtesy: Instability is a very bad thing! TERASPEED CONSULTING GROUP 89 Tektronix Confidential Behavioral Modeling Sources of Passivity Issues Insufficient attention to measurements or calibration – Interconnects do not amplify signals! – Even if individual measurements are passive, combined system measurements can have amplification properties Simulator extrapolation and interpolation based on model, not original measurement Finite measurement acquisition window (in the limit, the data is infinite!) 90 Tektronix Confidential Behavioral Modeling Example: Correlation of Measurement and Model Exact correlation in time and frequency domains 91 Tektronix Confidential Behavioral Modeling Example: Model Listing … .subckt DUT port1 gnd_ r14 port1 15 -1898.27 r1 port1 2 -24977.9 r15 port1 16 -971048 r2 port1 3 23982.9 r16 port1 17 2103.52 c34 35 gnd_ -7.10021e-016 r3 port1 4 -12111 r17 port1 18 15884.8 r91 36 37 176812 r4 port1 5 3124.2 r18 port1 19 326913 r5 port1 6 -2348.66 r19 port1 20 6566.94 r6 port1 7 14099.3 r20 port1 21 1508.55 r7 port1 8 -4392.43 r21 port1 22 -885.906 r8 port1 9 1444.76 r22 port1 23 4789.5 r9 port1 10 1.1046e+007 r23 port1 24 2788.21 r10 port1 11 -301858 r24 port1 25 -14075.7 r11 port1 12 3622.64 r25 port1 26 -18495.7 r12 port1 13 -1221.14 r26 port1 27 -1786.85 r13 port1 14 6965.14 r27 port1 28 -27328.6 … … 92 Tektronix Confidential … … c33 34 gnd_ 7.10021e-016 r90 35 gnd_ 3622.1 r92 36 gnd_ 57641.7 c35 36 gnd_ 7.15572e-016 r93 37 gnd_ 3932.13 c36 37 gnd_ -7.15572e-016 r94 38 gnd_ -2488.86 c37 38 gnd_ -3.07009e-015 … r95 39 gnd_ 5982.67 c38 39 gnd_ -2.49077e-014 .ends Behavioral Modeling MeasureXtractor™ Summary Converts S-parameters or TDR/T data into an exact-match model Passivity is guaranteed If you can measure it, and want model it with little effort, use MeasureXtractor™! 93 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues Interconnect SPICE / IBIS Modeling and Model Validation Topological Modeling: TDT and Lossy Line Modeling App Note: “Practical Characterization of Lossy Transmission Lines Using TDR” Impedance Profile (Z-line) Transmission Line Modeling L and C JEDEC computation – Examples: Power Plane Analysis Backplanes and cable assemblies 94 Tektronix Confidential TDT and IConnect Lossy Lines TDT and Lossy Line Modeling Is For: Long lossy transmission lines in backplanes and motherboards Long lossy cables 95 Tektronix Confidential TDT and IConnect Lossy Lines Loss Example: Time and Frequency Domain 96 Tektronix Confidential TDT and IConnect Lossy Lines Skin Effect vs. Dielectric Loss Typical FR-4 50 Ohm Trace 97 Tektronix Confidential TDT and IConnect Lossy Lines Different Loss Modeling Approaches Lumped (behavioral) – Defined for all frequencies – Slow for long lines Distributed – Based on parameters (Rskin, Gdielectric) Defined for all frequencies Not as general – Based on RLGC data General for quasi-TEM Not defined for all frequencies 98 Tektronix Confidential TDT and IConnect Lossy Lines Causality in TEM Models From basic physics, the real and imaginary parts of the dielectric constant are tightly related to ensure causality. The same is true of the permeability constant, m In TEM modeling, this means that R and L are related, and G and C are related Models based on RLGC data (or S-parameters) should address this issue 99 !!! Tektronix Confidential TDT and IConnect Lossy Lines Our Model Extraction Approach Assume standard simulator equations: Z Rdc Rac f jL Y Gdc Gac f jC Two extraction methods: – Open circuit reflection (TDR, one port) – Matched circuit transmission (TDR, TDT, two-port) Extract loss parameters: Rdc, Rac, Gdc, Gac, L, C Write resulting model in various formats – Lumped – Distributed with parameters – Distributed with RLGC data 100 Tektronix Confidential TDT and IConnect Lossy Lines Example: Extraction Results (Transmission) Extracted skin effect and dielectric loss parameters Simulated and measured transmission 101 Tektronix Confidential TDT and IConnect Lossy Lines Symmetrical Lossy Coupled Line Model TDR source 1 Board lines TDR source 2 Assumptions: – The lines are symmetrical – TDR steps are symmetrical – TDR steps arrive at the lines at the same time at the beginning of both lines 102 Tektronix Confidential IConnect Differential TDR Techniques Even/Odd vs. Common/Differential Z odd t odd l Lself Lm C tot C m L self Lm C tot C m Z differential 2 Z odd t differential t odd 103 Tektronix Confidential Z even t even l Z common Lself Lm C tot C m L self Z even t common t even 2 Lm C tot C m IConnect Differential TDR Techniques Even and Odd Impedance Profile Example Zeven>Zself>Zodd teven>tself>todd Note: - Odd mode = differential measurement (two TDR sources of opposite polarity) - Even mode = common mode measurement (two TDR sources of the same polarity) 104 Tektronix Confidential TDT and IConnect Lossy Lines Example: Extraction Results (Reflection, Coupled) Both self and mutual parameters are extracted 105 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues Interconnect SPICE / IBIS Modeling and Model Validation Topological Modeling: Impedance Profile (Z-line) Transmission Line Modeling App Note: “PCB Interconnect Characterization from TDR Measurements” App Note: “Characterization of Differential Interconnect from TDR Measurements” L and C JEDEC computation – Examples: Power Plane Analysis Backplanes and cable assemblies 106 Tektronix Confidential IConnect Single-ended TDR Techniques Single Transmission Line Modeling Is For: Short (lossless) transmission lines Electrically long packages (longer than Trise) Electrically long connectors (longer than Trise) 107 Tektronix Confidential IConnect Single-ended TDR Techniques Transmission Line Z and td Directly available from impedance profile Eliminate confusion about: – Exact impedance value – Exact electrical length of the lines Z01 Z02 td 108 Tektronix Confidential IConnect Single-ended TDR Techniques Via L and C t 1 2 L Z ( t )dt 2 t1 t2 t 1 2 1 C dt 2 t1 Z ( t ) t1 t1 t2 109 Tektronix Confidential IConnect Single-ended TDR Techniques IConnect® Modeling Process Measure and acquire 110 Process data Tektronix Confidential Extract model Simulate, compare and verify IConnect Single-ended TDR Techniques Modeling in IConnect Software * Name: Automatically Generated .subckt Single port1 port2 gnd_ ****** Partition #1 c1 port1 gnd_ 456f l1 port1 1 1.05n ****** Partition #2 t1 1 gnd_ 2 gnd_ Z0=50.8 TD=125p ………….. ****** Partition #4 t3 3 gnd_ port2 gnd_ Z0=48.2 TD=190p .ends 111 Tektronix Confidential IConnect Single-ended TDR Techniques Prepare to Simulate and Validate 112 Tektronix Confidential IConnect Single-ended TDR Techniques Simulation and Validation Results 113 Tektronix Confidential IConnect Single-ended TDR Techniques Using Rise Time Filtering to Achieve Simple Models 114 Tektronix Confidential IConnect Differential TDR Techniques Coupled Transmission Line Modeling Is For: Differential transmission lines Differential connectors and packages that are electrically long (longer than Trise) Crosstalk prediction (differential and single ended, forward, backward) Crosstalk induced jitter prediction 115 Tektronix Confidential IConnect Differential TDR Techniques Differential Line Modeling Short interconnect – Use lumped-coupled model Long interconnect – Split lines in multiple segments Longer yet interconnect – Symmetric distributed coupled line model – For longer lines, use lossy approach instead Zodd, todd Zodd, todd -Z odd/2, todd Zev en/2, tev en 116 Tektronix Confidential IConnect Differential TDR Techniques IConnect Differential Line Modeling 117 Tektronix Confidential IConnect Differential TDR Techniques Composite Model Generation * Name: Automatically Generated .subckt Symmetric 1 2 3 4 5 ****** Partition #1 t1 1 5 6 5 Z0=49.7 TD=92.3p t2 3 5 7 5 Z0=49.7 TD=92.3p ****** Partition #2 l1 6 8 19n c1 8 5 6.44p l2 7 9 19n c2 9 5 6.44p c3 8 9 716f k1 l1 l2 207m .ends 118 Tektronix Confidential IConnect Differential TDR Techniques Model Validation in IConnect 119 Tektronix Confidential IConnect Differential TDR Techniques Coupled LC Computation in IConnect 120 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues Interconnect SPICE / IBIS Modeling and Model Validation Topological Modeling: L and C JEDEC computation App Note: “TDR Techniques for Characterization and Modeling of Electronic Packaging” – Examples: Power Plane Analysis Backplanes and cable assemblies 121 Tektronix Confidential IConnect Short Interconnect Modeling When is Interconnect Lumped? tprop delay trise Practical rule of “short” or “lumped” (RLC) interconnect trise > tprop delay• (2 or 3) 122 Tektronix Confidential IConnect Short Interconnect Modeling Short Interconnect Modeling is Used For: IC packages Connectors Sockets Vias on the board Use only when short compared to Trise !!! 123 Tektronix Confidential IConnect Short Interconnect Modeling Single Parasitic Inductance TDR Oscilloscope Front Panel Rsource V VTDR Vref short 124 L Tektronix Confidential Vincident L Cable: Z 0, td Vreflected t Z0 1 2 L Z ( t )dt 2 t1 V V TDR t1 Vref short dt IConnect Short Interconnect Modeling Single-Ended TDR: Package Lead Inductance L Measurements Measure reflection Pa ck ag e le ad s TDR into the package lead Package under test Short all the leads to ground on the inside of the package. Short the leads that are not being measured to ground on the outside of the package. Short waveform: – TDR into the “short”; connect the probe signal contact to ground on a conductive (metal) pad 125 Vbackground noise Lself TDR into package lead. Lmutual TDR waveform: Vinduced L Wshort Short lead ends to ground Measure near-end crosstalk in adjacent lead – W TDR Tektronix Confidential Z 0 (WTDR Wshort )dt 2 V 0 Lmutual Z 0 (Winduced Wbackground )dt 2 V 0 Induced waveform: Measure near end crosstalk with far end of the victim shorted Background waveform: Corrects for the noise and scope DC offset IConnect Short Interconnect Modeling TDR Oscilloscope Front Rsource Panel Single Parasitic Capacitance V Vref open Vincident Cable: Z 0, td Open Vreflected t C VTDR 126 Tektronix Confidential 1 2 1 1 C dt 2 t1 Z ( t ) Z 0V C V ref open t1 VTDR dt IConnect Short Interconnect Modeling Single-Ended TDR: Package Lead Capacitance C Measurements Vinduced Wopen Measure reflection Pa ck ag e le ad s TDR into the package lead TDR into package lead Short the leads that are not being measured to ground on the outside of the package TDR into the “open”;disconnect the probe from the DUT or remove the DUT from the fixture 127 1 (Wopen WTDR )dt 2 Z 0 V 0 Open waveform: – Vbackground noise C self TDR waveform: Cmutual W TDR Keep lead ends open Measure near-end crosstalk in adjacent lead – C Package under test Tektronix Confidential C mutual 1 (Winduced Wbackground )dt 2 Z 0 V 0 Induced waveform: – Measure near end crosstalk with far end of the victim open-ended Background waveform: – Corrects for the noise and scope DC offset IConnect Short Interconnect Modeling Input Die Capacitance Measurement 128 Tektronix Confidential IConnect Short Interconnect Modeling Even-Odd Mode L and C Measurements ag e le Package under test Pa ck Measure odd mode TDR with differential stimulus, and even mode TDR with common mode stimulus ad s Compute C and L in even and odd mode TDR into the two adjacent socket lead with differential and common mode stimulus Cself Ceven Cmutual 129 Ceven Codd Tektronix Confidential 2 Lself Leven Lodd Lmutual 2 Leven Lodd 2 IConnect Short Interconnect Modeling Even-Odd Mode Impedance Profile ag e le Package under test Pa ck Measure odd mode TDR with differential stimulus, and even mode TDR with common mode stimulus ad s Compute C and L from even and odd Z-line TDR into the two adjacent socket lead with differential and common mode stimulus Lself 1 2 Z even teven Z odd todd 1 Lm Z even t even Z odd t odd 2 t 1 t C tot odd even 2 Z odd Z even t 1 t C m odd even 2 Z odd Z even Ctotal = Cself + Cm 130 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues Interconnect SPICE / IBIS Modeling and Model Validation Examples: Power Plane Analysis Backplanes and cable assemblies 131 Tektronix Confidential Power Plane Analysis Power Distribution Network (PDN) Test Vehicle Probe placement at point of power application 3" Reference short connection for inductance measurement Top plane 1 3/8" (35mm) Via connection to bottom plane Pad connection to top plane Point of power delivery 132 Tektronix Confidential Power Plane Analysis PDN Equivalent Models R R L C C R C L R C1 L C2 133 Tektronix Confidential Power Plane Analysis PDN Capacitance Measurements 134 Tektronix Confidential Power Plane Analysis PDN Impedance 135 Tektronix Confidential Power Plane Analysis PDN Resonance: Analysis for Bypass Caps 136 Tektronix Confidential Power Plane Analysis PDN Model Validation 137 Tektronix Confidential Power Plane Analysis PDN Model Accuracy 138 Tektronix Confidential Power Plane Via Power Via Inductance Lself Lself Leven Lodd 2 1 Z even t even Z odd t odd 2 139 Tektronix Confidential Lmutual Lmutual Leven Lodd 2 1 Z even t even Z odd t odd 2 Power Plane Via Example: Via Modeling Correlation between simulation and measurement ***** Partition #1 l1 port1 1 1.9n l2 port3 2 1.9n k1 l1 l2 200m 140 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues Interconnect SPICE / IBIS Modeling and Model Validation Examples: Backplanes and cable assemblies 141 Tektronix Confidential Putting It All Together Complete Topological Modeling Methodology Connectors, packages: – Short structures => use lumped elements (LC) or lossless T-lines – Use the true impedance profile approach Cables – lossy transmission line Backplane traces – lossy transmission line Combine the model and verify the accuracy with simulations Note that MeasureXtractor™ can do any of that! (behaviorally) 142 Tektronix Confidential Single-ended Example Example: IConnect-Extracted Model LC L=700pH C=280fF 143 T-line Z=53 Ohm Td=520ps (short => lossless) Tektronix Confidential CLC L=6nH C=1.5pF W-line L=198nH, C=69.7pF Ro=0.18 Ohm Rs=0.2uOhm Gd=6.7nS (nMho) (long => lossy) Single-ended Example Simulation Results 144 Tektronix Confidential Differential Example Backplane Example Courtesy FCI Electronics 145 Tektronix Confidential Differential Example Backplane Example: PCI-X Eye Diagram 146 Tektronix Confidential Differential Example Backplane Example Differential measurement, full mode analysis Backplane Daughter Card Daughter Card Connector Connector 147 Tektronix Confidential Differential Example Daughter Card and Backplane Models Daughter Card Model (odd mode) Backplane Model (odd mode) 148 Tektronix Confidential Differential Example Full Daughter Card Modeling 149 Tektronix Confidential Differential Example Full Backplane Modeling 150 Tektronix Confidential Differential Example Simulation Results 151 Tektronix Confidential Differential Example Jitter De-Embedding: Daughter Card Only 152 Tektronix Confidential Differential Example Jitter De-Embedding: Backplane Only 153 Tektronix Confidential Outline Interconnect Measurement Accuracy Issues – Impedance, S-parameters and eye diagram measurements Interconnect SPICE / IBIS Modeling and Model Validation – Z-line, lossy line, and automatic behavioral modeling 154 Tektronix Confidential Visit Us at the Tektronix Booth 155 Tektronix Confidential