Practical De-embedding for Gigabit fixture Ben Chia Senior Signal Integrity Consultant 5/17/2011
Topics Why De-Embedding/Embedding? De-embedding in Time Domain De-embedding in Frequency Domain De-embedding example Compliance test examples Accuracy consideration 2
Terminalolgy De-Embedding Remove fixture effect from the measurement Example: remove probes, SMA+traces Embedding Add a cable to see what happen to the signal Example: test receiver sensitivity by adding a 4-inch cable between the driver and DUT 3
De-embedding Example 4
Measurement Reference Plane 5
6-meter cable de-embedding Does your scope show the real waveform? ` 6
Example: De-embedding Probes 7
More Example: DDR2 Device Fixture DDR2 How to remove this fixture effect? 8
DDR2 Fixture De-Embedding Scope performs de-embedding 9
De-Embedding Calibration Both VNA/TDR have firmware to simplify de-embedding process Probe de-embedding Fixture de-embedding 10
TDR De-embedding Example Calibrate with SMA standard Short Load Connect Port 1 to SMA to measure S11 PCB trace SMA How to capture S11 at the package ball? 11
TDR De-embedding Example Agilent 54754A Option A Step 1: Capture fixture S4P ( Pad+trace +SMA ) Calibrate using SMA standard and Probe standard Measure DUT Port1: Pad Probing Port2: Probing SMA Capture S4P of SMA+Trace in Time Domain 12
Probing Example 13
TDR De-embedding Option A Step 1: Capture fixture S4P for trace with SMA at one end and PCB pads at the other end? Step 2: De-embedding fixture trace using ADS 14
TDR De-embedding Option B De-embedding including SMA using on-board calibration short and load standard short Measurement 50 ohm load DUT 15
Return Loss De-embedding 16
De-Embedding using VNA Equipment : Agilent ENA Calibration : SOLT DUT: Port 1, 2 - SMA Port 3, 4 - USB connectors Require TRL-like PCB calibration standard Fixture delay estimation use port extension to measure fixture delay Port1,2 SMA USB Port3,4 17
USB3.0 De-embedding Example A C How to measure S4P from A to C? 18
USB3.0 De-embedding Example A C How to measure S4P from A to C? 19
De-embedding fixture and Calibration board C A De-embedding fixture Calibration Board 20
Calibration Fixture Calibration Standard Open/Short/Load Port1, 3 use coaxial standard Port 2,4 use calibration board Delay/through Port 1-2 use coaxial standard Port 3-4 use 2x fixture trace Prot 1-3, 2-4, 1-4, 2-3 use 1x fixture trace 1x fixture trace Open/Short/load 1x thru trace 2x thru trace 21
Example : Return Loss Measurement Introduction Return loss measurement When return loss is not passing Return loss fixture removal How to fix the return loss problem Summary 22
Sata Interconnect Return Loss Specification limits the reflected energy 23
Return Loss Definition Return Loss = 20log 10 Z Z o o Z Z L L or 20log 10 V V i r where Zo is the Tx/Rx differential output/input impedance Z L is the link differential impedance Vi is the differential voltage incident upon Rx/Tx Vr is the differential reflected voltage from Tx/Rx Slid
Common Mode vs. Differential Mode Slid
Return Loss Measurement Many design can not meet return loss requirement Return loss accuracy and repeatability become crutial Both TDR ( time domain ) and VNA (frequency domain) can be used for return loss measurement Recommend VNA for marginal design Accuracy noise floor and calibration Repeatability - calibration standard 26
Return Loss Element Termination and IO circuit Ask circuit designer to reduce mismatch in both differential mode and common mode Differential driver circuit state Design special circuit to set IO to static 1 or 0 state Capable to set static de-emphasis state Pattern Blocking Capacitor 27
Blocking Caps The 100nF off chip coupling capacitor leads to a coupling time constant of 10 us. Such a long time constant is needed to limit baseline wander with scrambled data Make sure DC blocking capacitor is in placed for active transmitter return loss measurement - TDR/VNA port may gradually get damaged 28
How to Pass Return Loss Small IO capacitance less reflection Uniform impedance less reflection Shorter and uniform impedance fixture May not require fixture de-embedding if margin is large enough Long fixture trace and via will have better return loss number because reflection are attenuated by the long trace You got wrong result 29
Vector Network Analyzer Frequency Domain Equipment Single ended measurement Convert to differential mode and common mode with software 30
How Return Loss Are Measured today VNA Setup 0 dbm (1mw) Tx Setup One driver on AC high state ( < Voh) Another driver on AC low state (> Vol) Rx Setup Activate terminator 31
Return Loss De-embedding Simulation to remove the SMA, PCB trace, via and fingers. The return loss after de-embedding are usually worse than the VNA data measured at the SMA or PCB fingers 32
SMA Anti-pad Optimization Antipad diameter under trace Antipad diameter above trace Signal via radius Trace moved to different layer Ground via distance SMA Optimized Anti-pad Design Optimization critera -- 50 ohm -- low return loss 33
Return Loss Tuning Strategy : Simulation Pkg Stub Pkg Via Pkg Trace Bond Wire Ri Ci Rterm Agilent ADS or any spice tool can make this simple model 34
Transmit Port B Return Loss Including PCB Red: Model Blue: Measured An Example: XAUI
Receive Port B Return Loss Including PCB Red: Model Blue: Measured 36
Transmit Port B Return Loss Not Including Socket
Receive Port B Return Loss Not Including Socket 38
Pink: Measured Rx Black: Model Rx Violations at 3Ghz Slide What was wrong in this design?
Circuit for Return Loss Calculation CL: Cell Capacitance RS: Cell Series Resistance RL: Termination resistance Z0: Reference Resistance for S11 Calculation Z0 S11 RL CL RS 40
Rx Return loss model vs measured SMA Pcb via Package Pad/ball/via Ci 83ps Slide
Rx Return Loss Debugging Identify the elements from S11 plot 3GHz = 333ps Reflection at 180 degree 165ps Round trip delay 83ps Delay from PCB pad to Pakage pad ~83ps Fix: Increase the impedance to 50 ohm at the PCB via and package via in the model Slide
Pink: measured Rx Black: modified Rx model Replace PCB via and Package pads to 50 ohm Slide
TX : Increase PCB and Pads to 50 ohm Slide
Return Loss Summary Measurement based return loss provides accurate results after de-embedding SMA and lead-in traces on the text fixture SMA Impedance discontinuity can affect return loss significantly Optimized SMA anti-pad design can improve return loss accuracy Return loss measurement can be verified by careful modeling the package and PCB interconnect transistion 45
Hands-on Lab USB Compliance test Configuration Different fixture design comparison Insertion loss Eye diagram Conclusions 46
USB 3.0 Tx Compliance Channels Compliance Channels are defined to test Transmitter for worst case conditions. Device Tx TP1 Host Tx TP0 Worst Case Channel for Hosts 5 Device PCB Trace 3M Cable Host Tx TP1 3 Meter USB 3.0 Cable Worst Case Channel for Devices 11 Device PCB Trace 3M Cable Device Tx TP0 47
GRL Case Study: Three USB 3.0 Probing Setups for Device Testing TP0 USB-IF Fixture with 4 USB3.0 Cable INTEL_TIER2.S4P Agilent Fixture with 4 USB3.0 Cable TP0 AGILENT_TIER2.S4P Scope Measurement Plane TP0 LITEK_TIER2.S4P LiTek Fixture with 0 Cable TP0 (Device Connector) FixtureX.S4P Measurement Plane 48
Embed of USB 3.0 Cable + Host Channel Compliance Test Point 100mV Eye Mask CTLE In Scope SW TP1 Transmit Channel TP0 cascade_cable_back.s4p 49
Differential Insertion Loss Comparison 2.5GHz Nyquist Frequency 50
Results LiTek + 0 Cable Agilent + 4 Cable Intel + 4 Cable @ Scope Inputs @ TP0 DUT Connector @ TP1-Eq Compliance Point 250mV 234mV 228mV 51
Conclusions LiTek with 0 cable provides 6-9% Higher Voltage Margin than 4 Cable based fixtures. Intel Fixture, which is recommended by the USB-IF provides the least amount of margin and represents the worst case of the three fixtures tested. 52
Thank You 53
Backup - VNA SOLT Reference Plane 54
Transmit-Reflect-Line Calibration TRL Calibration SOLT Calibration Calibration standard designed on the same PCB as the device under test 55
How to Improve TRL Calibration Reduce connector reflection Increase trace width to reduce excessive solder variation Ensure soldering has minimum variation PCB trace Use wider trace to reduce impedance variation Use better material to reduce variation in impedance and delay Layout trace 45 degree with woven glass Specify denser woven glass for FR4 material 56
New: Two Tier Adaptor Removal Calibration MicroProbe Tips 57
Probing Example of TRL kit 58