A Simple, Yet Powerful Method to Characterize Differential Interconnects

Transcription

1 A Simple, Yet Powerful Method to Characterize Differential Interconnects

2 Overview Measurements in perspective The automatic fixture removal (AFR) technique for symmetric fixtures Automatic Fixture Removal (AFR) Process for a Mirror-Image Symmetrical Fixture Examples and accuracy estimates Revealing the secrets to practical applications 2

3 The Roles of Interconnect Measurements (S-Parameters) Emulate system performance with a behavioral model Characterize a component Verify a component to a spec Validate simulation/model to hardware Extract materials properties Debug a problem 3

4 The Problem: Directly Measuring a Behavioral Model of a Specific Structure What I want to measure is embedded in the middle of a bunch of interconnect I don t care about I just want the via structure, or the connector, or the cable, or the interface, or the uniform trace, 4

5 The Solution: De-embedding Traditional Calibration SOLT TRL (thru, reflect, line) LRM (line, reflect, match) De-embedding using Thru Line1 Line2 Measured text fixtures Calculated test fixtures by 3D full wave field solver Calculated text fixtures by approximation (port extension) The new way: really simple, automatic fixture removal (AFR) Must have a symmetric, 2x thru of the fixture only 5

6 De-Embedding Automagically What we want: performance S D What we measure: composite measurement of and fixtures S A S D S B With the separate fixtures S-Parameters, we can deembed the alone from the composite measurements S A S D S B The challenge: getting the S A and S B de-embed 6

7 The Automatic Fixture Removal (AFR) Process for a Mirror-Image Symmetrical Fixture TX/RX Port 1 Port 2 PACKAGE TYPE A USB3.0 RECEPTACLE USB3.0 CABLE TYPE B USB3.0 RECEPTACLE PACKAG E TX/ RX PCB INTERONNECT we measure S, we want SA, SB PCB INTERONNECT a1 S A S21A S B S21B b2 Port 1 S11A S22A S11B S22B Port 2 b1 S12A S12B Apply: Symmetry in the fixture: S11A = S22B, S22A = S11B, S21A = S12A = S21B = S12B Network Theory Calculate S-parameter file for each mirror image half of the fixture. a2 7

8 A Key Step in the Complete Process: Use Time Domain Gating to Get S11A, S22B Use time domain gating to get T11A from T11 Turn T11A into S11A 8

9 Final Step: Calculating Just the Half-Fixture S-Parameter Files S A S B S A S B The result: The half fixture S-parameters Use these to de-embed the, using PLTS or ADS 9

10 The Complete Process of Extracting The from the + Fixture + fixture thru ref fixture to get de-embed Step 3: De-embed only from + fixture Step 4: Analyze the de-embedded 10

11 An Example: 4-port Measurement of a Samtec Board to Board Connector Side view 2x thru ref fixture Top view 11

12 Samtec Connector + fixture Step 3: De-embed only from + fixture 12

13 Samtec Connector Always Look at Your Data You can observe a lot by looking -- Yogi Berra + fixture T11 + fixture Step 3: De-embed only from + fixture 1 2 Fixture connector T11 + fixture T22 + fixture NEXT T31, T42 Real world issue #1: not all the fixtures are identical or symmetric 13

14 Samtec Connector + fixture + fixture 2x thru ref fixture Real world issue #2: 2x thru ref fixture fixture with Step 3: De-embed only from + fixture + fixture: T11, T22, T33, T44 14

15 Samtec Connector + fixture T11A half fixture T11 fixture Step 3: De-embed only from + fixture T22B half fixture T22 fixture T22B half fixture, not perfect match to T22 fixture Real world issue #3: 2x thru ref fixture not perfectly symmetrical 15

16 Samtec Connector De-embed file A + fixture Step 3: De-embed only from + fixture De-embed file B 16

17 Samtec Connector + fixture Step 3: De-embed only from + fixture Insertion and return loss + fixture, only NEXT, FEXT + fixture, only 17

18 Samtec Connector T11 fixture + + fixture Step 3: De-embed only from + fixture T11 only At this rise time ~ 50 psec, connector impedance is lower than as measured in the fixture Most of the attenuation is due to the fixture Resonance dips probably due to coupling to other open lines 18

19 How Well Does The AFR Method Really Work? Plan Model a real physical system with an equivalent circuit model to create a known response Use a single via with transmission line feeds and typical launches Create S-parameters for a simulated and a simulated fixture Change features in fixture and compare de-embedded with actual Explore the sensitivity of the fixture on the de-embedded A real Via Courtesy of Wild River Technologies Measured TDR response What circuit elements should we include in the model for this + fixture? 19

20 Building Calibration Examples Using Synthesized S-Parameter Data Using Agilent ADS to synthesize precision S-parameter to test the accuracy of the AFR Method Courtesy of Wild River Technologies A real Via (this way we know what the answer is supposed to be) A Simple Model of a Thru Via in ADS A simulated Via 0.25 inch long, 50 Ohm line feed on top and bottom 0.25 pf capture pad on the top, bottom surface 64 mil long uniform 60 Ohm transmission line 20

21 Synthesize Precision S-Parameters for a Simulated Via Inside a Fixture The simulated model: fixture + via Courtesy of Wild River Technologies Fixture on either side is: 3 inch uniform, lossy transmission line, with launches same as the launch 2x thru ref fixture only Apply AFR to three cases: Case 1: fixture is uniform lossy transmission line Case 2: fixture is non uniform, lossy line Case 3: fixture is not perfectly mirror image symmetrical, non uniform, lossy line Case 4: fixture is perfectly mirror image symmetrical, but 2x ref thru is not The simulated model: 2x thru ref fixture 21

22 Case 1: Fixture is Uniform, 50 Ohm, Lossy Symmetric Transmission Line + fixture S11 Via + Fixture Step 3: De-embed only from + fixture S11 2x thru ref fixture S21 2x thru ref fixture S21 Via + Fixture 22

23 Case 1: Fixture is Uniform, 50 Ohm, Lossy Symmetric Transmission Line + fixture 2x thru ref fixture only Left half of fixture S21 S11 AFR Step 3: De-embed only from + fixture Right half of fixture 23

24 Case 1: Fixture is Uniform, 50 Ohm, Lossy Symmetric Transmission Line S11 Via + Fixture + fixture Step 3: De-embed only from + fixture Left half of fixture Right half of fixture Via Model Only (de-embedded) 24

25 Case 1: Fixture is Uniform, 50 Ohm, Lossy Symmetric Transmission Line + fixture De-embedded via model Step 3: De-embed only from + fixture Correct via model De-embedded via model Correct via model The de-embedded via model is a pretty good match to the actual via model 25

26 Case 2: Fixture is Non-uniform, 50 Ohm, Lossy Symmetric Transmission Line T11 Via + Fixture + fixture Step 3: De-embed only from + fixture S11 Via + Fixture S11 2x Thru ref S21 2x Thru ref S21 Via + Fixture 26

27 Case 2: Fixture is Non-uniform, 50 Ohm, Lossy Symmetric Transmission Line + fixture Via + Fixture Step 3: De-embed only from + fixture Via + Fixture De-embedded via De-embedded via 27

28 Case 2: Fixture is Non-uniform, 50 Ohm, Lossy Symmetric Transmission Line De-embedded via model + fixture Step 3: De-embed only from + fixture Correct via model Correct via model De-embedded via model 28

29 Case 3: Fixture is Non-uniform, Asymmetric Lossy Transmission Line + fixture Step 3: De-embed only from + fixture 53 Ω 53 Ω T11 Via + Fixture 47 Ω T11 2x ref fixture Fixture 47 Ω 29

30 Case 3: Fixture is Non-uniform, Asymmetric Lossy Transmission Line Via + Fixture + fixture Step 3: De-embed only from + fixture De-embedded via De-embedded via Via + Fixture 30

31 Case 3: Fixture is Non-uniform, Asymmetric Lossy Transmission Line + fixture De-embedded via model Step 3: De-embed only from + fixture Correct via model Correct via model De-embedded via model Even with asymmetric fixture, agreement is pretty good 31

32 Case 3: Fixture is Non-uniform, Asymmetric Lossy Transmission Line T11 Via + Fixture + fixture Step 3: De-embed only from + fixture T11 Correct via model T11 Correct via model T11 De-embedded via model Even with asymmetric fixture, agreement is pretty good 32

33 Case 4: 2x Ref Thru is Asymmetric Lossy Transmission Line, Fixture is Symmetric + fixture T11 Via + Fixture Step 3: De-embed only from + fixture 53 Ω T11 2x ref fixture Fixture 47 Ω Note: 2x thru ref fixture is different from the + fixture 33

34 Case 4: 2x Ref Thru is Asymmetric Lossy Transmission Line, Fixture is Symmetric De-embedded via model + fixture Step 3: De-embed only from + fixture Correct via model De-embedded via model Correct via model 34

35 Case 4: 2x ref thru is asymmetric lossy transmission line, fixture is symmetric + fixture De-embedded via model Step 3: De-embed only from + fixture Correct via model 35

36 Observations About ANY De-embed Process ANY de-embedding technique relies on: the reference structures being identical to the fixture structures The fixture feeds on both ends of the being identical The accuracy of ANY de-embed process is only as good as the quality of the fixture and references The Automatic Fixture Removal (AFR) process is simple to implement Can be applied to single ended interconnects Can be applied to differential interconnects 36

37 Application Methods Always design in a 2x thru ref fixture to be identical to the actual fixture to the Always split the reference plane in a uniform transmission line region Otherwise, the fringe fields are specific to the probe and intrinsic to the Always design the fixture as transparent as possible Short length, low loss Uniform transmission lines matched to 50 Ohms Short via stubs Signal vias surrounded by return vias Optimize barrel diameter, clearance holes to make 50 Ohm via Minimum coupling between the lines What if the fixtures on the two ends are not the same? Build two different 2x thru fixtures- both symmetric Extract S-parameter for each half fixture Use a different de-embed on each end Always compare T11, T22 of assembly,t11, T22 of fixture 37

38 Revealing the Secrets to Practical Measurements Uniform transmission line and loss per length Via or circuit board feature Connector models Cable properties 38

39 Web Resources: For more information on Agilent Technologies products, applications or services, please contact your local Agilent office. The complete list is available at: Contact Agilent at: Americas Canada (877) Brazil (11) Mexico United States (800) Asia Pacific Australia China Hong Kong India Japan 0120 (421) 345 Korea Malaysia Singapore Taiwan Other AP Countries (65) Europe & Middle East Belgium 32 (0) Denmark Finland 358 (0) France * *0.125 /minute Germany 49 (0) Ireland Israel /544 Italy Netherlands 31 (0) Spain 34 (91) Sweden United Kingdom 44 (0) For other unlisted countries: Product specifications and descriptions in this document subject to change without notice. Agilent Technologies, Inc Published in USA, July 17, EN