SI Design & Measurement Principles and Best Practices

Transcription

1 I ment Principles and Best Practices 13 May, 2015 Heidi Barnes enior Application Engineer High peed Digital Design Keysight EEof EDA Division In collaboration with: Ben Chia enior ignal Integrity Consultant Granite River Labs Mike Resso ignal Integrity Application cientist Keysight Technologies Richard Zai enior ignal Integrity Consultant PacketMicro

2 ignal Integrity Talker Listener The Tx to Rx connection: Digital standards refer to it as the Physical Layer or the PHY for short. imulation tools call it the channel eminar Page 2

3 Typical 100 Gbps Telecom ystem eminar Page 3

4 Backplanes are a Critical Link eminar Page 4

5 Backplane Connectors Are Advanced High Precision Molded Components Differential ignal Traces urface Mount Terminals Double sided shield eminar Page 5

6 Why Do I Need ignal Integrity? Talker Backplane Connector and Via Field Listener 40Gb/s What I imulate What I Measure 1) I should find a new job. 2) What is the fixture? 3) As-fabricated materials? 4) How can I tell what is wrong? 5) What should I fix first? A Transparent Connector Connector + Fixture eminar Page 6

7 Eric Bogatin s Rule #9 Rule # 9: Never do a measurement or simulation without first anticipating what you expect to see. imulations and measurements minimize risk Knowing what to expect saves time and money If you get it right you feel good If you get it wrong you will learn something ignal and Power Integrity expert Eric Bogatin is the author of: ignal Integrity and Power Integrity implified, Prentice Hall, 2 nd edition 2010 eminar Page 7

8 ignal Integrity Basics Material Loss : Conductor Dielectric Impedance Loss and -Parameters Reflections Crosstalk eminar Page 8

9 Frequency Dependent Channel Loss from Materials Dielectric Loss Current Density Conductor Loss Dielectric Loss 2.3 f example : 2.3(1GHz) r tan db/inch 4 (0.02) db/inch 1oz Copper Loss 36 w Z 36 example : (10mils)(50 ) 0 f tan db/inch 1GHz 0.07 db/in Chapter 9 in Eric Bogatin s ignal Integrity and Power Integrity, 2010 Losses from materials: Dielectric loss caused by dipole movement with rapidly changing fields. Proportional to frequency. Conductor loss with skin effect pushing currents to lowest inductance path. Proportional to square root of frequency. eminar Page 9

10 Example: Channel Loss with PCB Materials My Application Length = 40mm LossTan=.006 Width=0.3 mm Cu = 5.6e7 dk = 3.6 Worst Case vs Length 100mm to 25mm top 25 mm tart 100 mm Worst Case vs Loss Tangent 0.02 to top tart 0.02 Worst Case vs Width 0.1mm to 0.5mm Worst Case vs Conductivity 1e7 /m to 1e8 /m Worst Case vs dk 3.0 to 4.4 top 0.5mm top 1e8 /m top mm tart 1e7 /m tart 4.4 Worst Case: 100mm Length, 0.02 LossTan, 0.1mm Width, 1e7 /M, 4.4 dk eminar Page 10

11 PCB Material Loss in the Time Domain tep Response ingle Bit Response (BR) 10 Gb/s 40 Gb/s 40 Gb/s PRB9 Waveform Worst case PCB Materials Eye Diagram ocket Application Best Case PCB Material Worst Case PCB Material eminar Page 11

12 Real World PCB - Effective Loss Tangent Leveraging old PCB technology Copper Foil Profile surface roughness New I requirements for the PCB fabrication document: -Glass Weave -Copper Profile -Test tructures for asfabricated PCB losses. Glass Weave dk variation Pictures from Lee Ritchey of peeding Edge, 13-TU2 Breaking the 32 Gb/s Barrier: PCB Materials, imulations, Measurements, DesignCon2015 eminar Page 12

13 Via tubs Create Capacitive Loads How is a Via tub Created? ignal current splits in two directions and sees two 50 ohm lines in parallel (25 ohms) Excess capacitance is created by a 25 ohm segment of equivalent circuit Reflections and poor signal integrity results Good (no stub) i Bad (stub) i 1 i i Design i& 2 Measure eminar Page 13

14 Backplane Data Rates are Increasing eminar Page 14

15 Impedance Reflections Reflections are the reality when time traveling is not allowed Vin Time Delay Vout Vin does not show up instantaneously at Vout, and therefore Vin requires a round trip delay to adjust to the Vout termination. Initial Vin only sees the characteristic impedance Zo. Thought Experiment for Impedance Reflections OPEN - Positive Voltage Reflection Current must flow back to the source open HORT - Negative Voltage Reflection Voltage must go to zero short V V time time +2V 0V eminar Page 15

16 Telegrapher s Equations Voltages and Currents are changing with Time and Distance (Magnitude and Phase) Create a simple model of a transmission line. Oliver Heaviside For small R and G inusoidal Input Utilize calculus to analyze the model when summing a series of incremental length sections. Resulting Relationships eminar Page 16

17 Characteristic Impedance Z 0 Derivation from Telegrapher Equations: Z 0 Z 0 L C Derivation from transmission line charging: Independent of Length 1 vc L Q x C CL x, I, Q CV t t v CL xv V 1 then I vclv and Z x I vcl v eminar Page 17

18 Time Travel is not Allowed What is high speed? V=IR The Channel has finite length: peed of Tx : ignal Rise-time Type of Data : Data Rate Gb/s peed of Channel: Time Delay Rise Time velocity 20 80% 12 Dk 0.22 Frequency 3dB 1 Unit Interval DataRate NRZ mils ps ps eminar Page 18

19 Magnitude of the Reflection Rise time 25ps Reflection : Z Z x x Z Z 0 0 Time Delay : TD 1inch vl dk ns v 12 mil Example FR4 : ps (100) 17 ps Rise time Bit Rate ~1/10 th Rise Time Feature ize High peed Feature ize 1 MBit 3 m (10 ft) Matched Termination 10 MBit 30 cm (12 in) T-Line Zo 100 MBit 3 cm (1.2 in) Connector Zo 1 GBit 3 mm (120 mils) Passive MT Zo 5 GBit 0.6 mm (24 mils) Via Zo 10 GBit 0.3 mm (12 mils) Die, Package, PCB Co-sim 40 GBit mm (6 mils) Machining Tolerances eminar Page 19

20 Reflections from a tub Resonator w= 11 mils L=.75 in 250 mil w= 11 mils L=.75 in 2.92mm 2.92mm 342 mil 250 mil eminar Page 20

21 tub Resonator (L=λ/4) L= ¼ of the wavelength or electrical length p/4 p 42 t t L g 4 g Given: p v OPEN 1 p= time to travel 1 wavelength p Destructive Interference eminar Page 21

22 Parallel tub Resonator (L=λ/4) Quarter-wavelength stub (L=λ/4) 1. Destructive interference 2. Virtual short 3. Cancellation of waves 4. Minimum transmission (21) eminar Page 22

23 tub Resonator (L=λ/2) L= ½ of the wavelength or electrical length p/2 L g 2 t OPEN 1 p t p= time to travel 1 wavelength g Given: p v p Constructive Interference eminar Page 23

24 Parallel tub Resonator (L=λ/2) Half-wavelength stub (L=λ/2) 1. Constructive interference 2. Virtual open 3. Addition of waves 4. Maximum transmission (21) eminar Page 24

25 Full-Path imulation to Measurement Correlation Parallel tub Resonator Test tructure AD Line-Type Model with Embedded Fixture Measurement vs. imulation in AD CMP-28 tripline Resonator 10 Gb/s, PRB 9 IMULATED MEAURED N4951A Pattern Generator eminar Page 25

26 Circuit Model with Reflections cattering Parameters Behavioral Model Return Loss 11 2-Port 21 Insertion Loss PORT 1 -Parameter Behavioral Model PORT 2 Insertion Loss Return Loss referenced to the return path -Parameters (11, 21, 12, 22).s2p Touchstone file eminar Page 26

27 Transmission Lines are Differential eminar Page 27

28 ingle Ended Parameters Port 1 Port 3 Four-port single-ended device Port 2 Port 4 Return Loss or TDR Insertion Loss or TDT Near End Crosstalk (NEXT) Far End Crosstalk (FEXT) Frequency Domain Parameters Time Domain Parameters FFT or IFFT eminar Page 28

29 Response Common- Mode Response Differential- Mode timulus Common-Mode timulus Port 1 Port 2 Port 1 Port 2 Differential- Mode Response ingle-ended to Differential -Parameters Port 1 Port 3 ingle-ended Port 2 Port 4 Balanced port 1 Balanced Balanced port 2 timulus Naming Convention: mode res., mode stim., port res., port stim Port 1 Port 2 Port 1 Port 2 DD 11 DD 21 CD 11 CD 21 DD 12 DD 22 CD 12 CD 22 DC 11 DC 21 CC 11 CC 21 DC 12 DC 22 CC 12 CC 22 eminar Page 29

30 Differential -parameters eminar Page 30

31 Crosstalk for Microstrip vs tripline Analysis EM simulation for complex radiation and crosstalk. Identical 12mil Traces, 5mil Gap MICROTRIP Microstrip Loss Microstrip FEXT Microstrip 5Gbps Eye TRIPLINE tripline Loss tripline FEXT tripline 5Gbps Eye eminar Page 31

32 Differential tandard Through Vias vs. hort Microvias -Parameters Differential Return Loss of standard via Differential Return Loss of microvia Differential Insertion Loss of microvia Differential Insertion Loss of standard via eminar Page 32

33 Differential tandard Through Vias vs. hort Microvias PRB9 10 Gbps PRB9 20 Gbps eminar Page 33

34 Cascading -Parameters Cascading Two 2-Port Networks Total Network T a a Port 1 A1 A B B2 Port 2 Network Network b b A1 B 2 Multiplying -Parameter Matrices Doesn t Work T A B T T Convert to Normalized Incident and Reflected Wave T-Matrix From to T det T T T T T A T B From T to T T det T T T 22 T T eminar Page 34

35 imulation to Measurement Correlation Down to the last ripple Model with AFR Fixture Model with MBM Fixture Measured Fixture + DUT OLT Cal eminar Page 35

36 Do I trust the imulation or the Measurement? IMPLE CHANGE IN PCB TRACE WIDTH Which one is correct? PCB tripline IMULATED MEAURED eminar Page 36

37 Measurement Fixtures tandard coaxial OLT Calibrations only calibrate to the end of the coaxial cable! A PCB Fixture transitions from the coax connector to the planar PCB transmission line. Coaxial Connector Coaxial Interface Planar PCB Footprint Page eminar 37

38 MA hort Length Fixture Can I ignore the fixture losses if I shorten the length? 1 Fixture A eries Resonant Beatty tructure W 3*W W ~2.54 cm (1 in) MA Fixture B INERTION LO A short path does not eliminate impedance reflections from the connector to PCB transition horter fixture can be worse! hort 20mil Long 700mil eminar Page 38

39 What is the Reference Plane YMMETRICAL 2x FIXTURE THROUGH PATH tep 1 plitting of the -Parameters FIXTURE A -PARAMETER FIXTURE B -PARAMETER Reference Plane Agilent PLT AFR Algorithm FIXTURE + Resonant Beatty tructure tep 2 T-Line Change in Z Fabrication Properties FIXTURE DE-EMBED Reference Plane T-Matrix eminar Page 39

40 Fixture Removal Benefits FIXTURE + Resonant Beatty tructure T-Line Change in Z Fabrication Properties FIXTURE DE-EMBED T-Matrix -Parameters before Fixture De-Embed Measured -Parameters after Fixture De-Embed FIXTURE DE-EMBED T-Matrix eminar Page 40

41 Fixture De-Embedding with -Parameters De-Embed imulation using -Parameters eminar Page 41

42 Using TDR to Verify Fixture Port 1 Time Domain Verifies Fixture Data Matches with the DUT Fixture TDR from 11 eminar Page 42

43 Using TDR to Verify Fixture Port 2 Time Domain Verifies Fixture Data Matches with the DUT Fixture TDR from 22 eminar Page 43

44 Before and After Fixture Removal Frequency Domain Time Domain horter Electrical Length after Fixture Removal DUT with Fixture Removed in RED DUT with Fixture in Blue eminar Page 44

45 Fixture Measurement-Based-Model (MBM) Model matches in the frequency and time domains Fixture Left- ide Fixture Right- ide MBM Easy to adjust PCB routing length for flexible fixture length. eminar Page 45

46 Full Path Embedded Fixture + DUT imulation Fixture -Parameter from AFR and from deconstructed MBM Measured Fixture + DUT Model with MBM Fixture Model with AFR Fixture eminar Page 46

47 tellar Measurement to imulation Correlation with Accurate Fixture Calibrations Looking Good! IMULATED 32 Gb/s, PRB9 MEAURED eminar Page 47

48 The Future: The Internet of Things (IOT) eminar Page 48

49 Example: DDR2 Device Fixture DDR2 How to remove this fixture effect? eminar Page 49

50 DDR2 Fixture De-Embedding cope performs de-embedding eminar Page 50

51 ATA De-embedding Example How to measure 4P A from A to C? C eminar Page 51

52 De-embedding fixture and Calibration board C A De-embedding fixture Calibration Board eminar Page 52

53 UB Type-C High peed Cable Fixture Calibration tructures Through Reflect Lines (TRL) eminar Page 53

54 Error Correction Techniques eminar Page 54

55 High Frequency Probing When connectors are not an option.. PCB channel has no connectors: chip to chip. MA - MA Direct measurement of the in-situ path is required. Reduced cost for repetitive production testing. eminar Page 55

56 Tools and Accessories eminar Page 56

57 Example GHz Probes and Probe tations eminar Page 57

58 .in ummary A theory is something nobody believes, except the person who made it. An experiment is something everybody believes, except the person who made it. Albert Einstein A simulation is something nobody believes, except the person who made it. A measurement is something everybody believes, except the person who made it. Paul Huray Always verify simulation and measurements with standards eminar Page 58