SAS-2 Zero-Length Test Load Characterization (07-013r7) Barry Olawsky Hewlett Packard (8/2/2007)

Transcription

1 SAS-2 Zero-Length Test Load Characterization (07-013r7) Barry Olawsky Hewlett Packard (8/2/2007) r7 SAS-2 Zero-Length Test Load Characterization 1

2 Zero-Length Test Load Provides ideal connection between compliance point and instrumentation reference plane. May consist of printed circuit board traces, connectors and cables Trace structure variations of width, length and dielectric material are typical. Test board designs need to meet impedance targets but exact effect of insertion loss is considered negligible but usually unknown Connector footprint variations are a common source of varied test board performance. Instrumentation cables vary in length, insertion loss, termination quality and type (rigid, semi-rigid, hand-formable, etc) These effects can be de-embedded when using some instrumentation but can not with most r7 SAS-2 Zero-Length Test Load Characterization 2

3 Zero-Length Test Load Misconceptions All good quality SMA cables have similar electrical characteristics If you use connector footprint provided in the standard you will meet the impedance specifications of the standard The common mode impedance of trace structures on the test fixture has no effect. Only the differential impedance is important r7 SAS-2 Zero-Length Test Load Characterization 3

4 Challenges of Characterization Some compliance points utilize connectors while others don t. How do we account for these differences? Of those that use connectors not all are the same Connector and IC component footprints dictate PCB routing methodology Must resolve the discrepancies or de-embed the PCB effects Cables Printed Circuit Board Connectors r7 SAS-2 Zero-Length Test Load Characterization 4

5 Challenges of Characterization To limit the insertion loss variations an S 21 mask is one possible approach. This approach is problematic if other parameters are less than ideal Can we make that assumption? Just how good do the other parameters have to be? Simulations can help answer this question FAIL PASS r7 SAS-2 Zero-Length Test Load Characterization 5

6 Measurement Objectives Network analyzer measurements are suited to deembedding. Frequency limitations exist with calibration structures but we may be below those limits Time domain instrumentation (oscilloscope) does not yet have the same level of de-embedding features Need to evaluate each usage case for zero-length test load and determine what specifications are necessary and appropriate r7 SAS-2 Zero-Length Test Load Characterization 6

7 Test Fixture Budgeting (Early Targets) Printed Circuit Board Cable Connectors & Footprints Insertion Loss < 3GHz < 3GHz < 3GHz Return Loss < -20dB thru 3GHz < -30dB thru 3GHz < -15dB thru 3GHz Conversion < -30dB thru 3GHz < -30dB thru 3GHz < -30dB thru 3GHz Common Mode? Not Applicable? Since most of these parameters are sensitive to interaction with other components, individual budgeting is of minimal value beyond a first guess estimate r7 SAS-2 Zero-Length Test Load Characterization 7

8 Next Actions The biggest issue is probably how to specify parameters that will yield consistent zero-length fixtures. Simulations should help. Zero-length specification should be influenced by the types of measurements to be performed in SAS-2. We must consider each type r7 SAS-2 Zero-Length Test Load Characterization 8

9 Simulation Approach Three different fixtures will be considered Case #1: Lowest insertion loss (short trace, SMA launch and bulky 24 inch instrumentation cable) Case #2: Lowest return loss (slightly longer trace, better SMA launch and thin 1 meter instrumentation cable) Case #3: 2 meter minisas (includes fixturing losses) Case#4: Mated minisas interface (slightly longer trace, same SMA launch and cable as #2) Characterize each and simulate eye diagrams to compare effects of fixtures at 3 and 6Gbps. Use 1000mV K28.5 transmitter model with fast and slow edge rates r7 SAS-2 Zero-Length Test Load Characterization 9

10 Zero-Length Fixture Insertion Loss db case1 case2 case3 case GHz r7 SAS-2 Zero-Length Test Load Characterization 10

11 Case #1, 3Gbps, Tr: 66ps Eye Height: 926mV r7 SAS-2 Zero-Length Test Load Characterization 11

12 Case #1, 3Gbps, Tr: 100ps Eye Height: 923mV r7 SAS-2 Zero-Length Test Load Characterization 12

13 Case #2, 3Gbps, Tr: 66ps Eye Height: 794mV r7 SAS-2 Zero-Length Test Load Characterization 13

14 Case #2, 3Gbps, Tr: 100ps Eye Height: 788mV r7 SAS-2 Zero-Length Test Load Characterization 14

15 Case #3, 3Gbps, Tr: 66ps Eye Height: 591mV r7 SAS-2 Zero-Length Test Load Characterization 15

16 Case #3, 3Gbps, Tr: 100ps Eye Height: 583mV r7 SAS-2 Zero-Length Test Load Characterization 16

17 Case #4, 3Gbps, Tr: 66ps Eye Height: 746mV r7 SAS-2 Zero-Length Test Load Characterization 17

18 Case #4, 3Gbps, Tr: 100ps Eye Height: 744mV r7 SAS-2 Zero-Length Test Load Characterization 18

19 Case #1, 6Gbps, Tr: 33ps Eye Height: 893mV r7 SAS-2 Zero-Length Test Load Characterization 19

20 Case #1, 6Gbps, Tr: 50ps Eye Height: 880mV r7 SAS-2 Zero-Length Test Load Characterization 20

21 Case #2, 6Gbps, Tr: 33ps Eye Height: 692mV r7 SAS-2 Zero-Length Test Load Characterization 21

22 Case #2, 6Gbps, Tr: 50ps Eye Height: 678mV r7 SAS-2 Zero-Length Test Load Characterization 22

23 Case #3, 6Gbps, Tr: 33ps Eye Height: 456mV r7 SAS-2 Zero-Length Test Load Characterization 23

24 Case #3, 6Gbps, Tr: 50ps Eye Height: 444mV r7 SAS-2 Zero-Length Test Load Characterization 24

25 Case #4, 6Gbps, Tr: 33ps Eye Height: 626mV r7 SAS-2 Zero-Length Test Load Characterization 25

26 Case #4, 6Gbps, Tr: 50ps Eye Height: 636mV r7 SAS-2 Zero-Length Test Load Characterization 26

27 Simulation Summary 3Gbps (Fast) 3Gbps (Slow) 6Gbps (Fast) 6Gbps (Slow) Case 1 926mV 923mV 893mV 880mV Case 2 794mV 788mV 692mV 678mV Case 3 591mV 583mV 456mV 444mV Case 4 746mV 744mV 636mV 626mV Gbps (Fast) 3Gbps (Slow ) 6Gbps (Fast) 6Gbps (Slow ) Case 1 Case 2 Case 3 Case 4 No pre-emphasis used r7 SAS-2 Zero-Length Test Load Characterization 27

28 Measurements with Pre-emphasis Focusing on cases 1 and 2, how do the eye height and peak-to-peak values compare for the pre-emphasis settings of 0dB, 3dB and 6dB r7 SAS-2 Zero-Length Test Load Characterization 28

29 Case #1, 3Gbps, Tr: 66ps, 3dB Eye Height / Max Eye: 690mV / 946mV r7 SAS-2 Zero-Length Test Load Characterization 29

30 Case #1, 3Gbps, Tr: 100ps, 3dB Eye Height / Max Eye: 690mV / 925mV r7 SAS-2 Zero-Length Test Load Characterization 30

31 Case #2, 3Gbps, Tr: 66ps, 3dB Eye Height / Max Eye: 649mV / 848mV r7 SAS-2 Zero-Length Test Load Characterization 31

32 Case #2, 3Gbps, Tr: 100ps, 3dB Eye Height / Max Eye: 646mV / 840mV r7 SAS-2 Zero-Length Test Load Characterization 32

33 Case #1, 6Gbps, Tr: 33ps, 3dB Eye Height / Max Eye: 686mV / 923mV r7 SAS-2 Zero-Length Test Load Characterization 33

34 Case #1, 6Gbps, Tr: 50ps, 3dB Eye Height / Max Eye: 686mV / 907mV r7 SAS-2 Zero-Length Test Load Characterization 34

35 Case #2, 6Gbps, Tr: 33ps, 3dB Eye Height / Max Eye: 622mV / 772mV r7 SAS-2 Zero-Length Test Load Characterization 35

36 Case #2, 6Gbps, Tr: 50ps, 3dB Eye Height / Max Eye: 623mV / 751mV r7 SAS-2 Zero-Length Test Load Characterization 36

37 Case #1, 3Gbps, Tr: 66ps, 6dB Eye Height / Max Eye: 493mV / 943mV r7 SAS-2 Zero-Length Test Load Characterization 37

38 Case #1, 3Gbps, Tr: 100ps, 6dB Eye Height / Max Eye: 493mV / 944mV r7 SAS-2 Zero-Length Test Load Characterization 38

39 Case #2, 3Gbps, Tr: 66ps, 6dB Eye Height / Max Eye: 473mV / 846mV r7 SAS-2 Zero-Length Test Load Characterization 39

40 Case #2, 3Gbps, Tr: 100ps, 6dB Eye Height / Max Eye: 473mV / 850mV r7 SAS-2 Zero-Length Test Load Characterization 40

41 Case #1, 6Gbps, Tr: 33ps, 6dB Eye Height / Max Eye: 481mV / 910mV r7 SAS-2 Zero-Length Test Load Characterization 41

42 Case #1, 6Gbps, Tr: 50ps, 6dB Eye Height / Max Eye: 486mV / 910mV r7 SAS-2 Zero-Length Test Load Characterization 42

43 Case #2, 6Gbps, Tr: 33ps, 6dB Eye Height / Max Eye: 463mV / 795mV r7 SAS-2 Zero-Length Test Load Characterization 43

44 Case #2, 6Gbps, Tr: 50ps, 6dB Eye Height / Max Eye: 462mV / 791mV r7 SAS-2 Zero-Length Test Load Characterization 44

45 Simulation Summary Case1 3Gbps Case1 6Gbps Case2 3Gbps Case2 6Gbps DE: 0dB (Slow) 923mV 880mV 788mV 678mV DE: 0dB (Fast) 926mV 893mV 794mV 692mV DE: 3dB (Slow) 925mV 907mV 840mV 751mV DE: 3dB (Fast) 946mV 923mV 848mV 772mV DE: 6dB (Slow) 944mV 910mV 850mV 791mV DE: 6dB (Fast) 943mV 910mV 846mV 795mV Three and six db measurements are pk-pk at eye midpoint of the emphasized bit. Zero db measurements are of the eye opening r7 SAS-2 Zero-Length Test Load Characterization 45

46 Simulation Summary Case1 3Gbps Case1 6Gbps Case2 3Gbps Case2 6Gbps DE: 0dB (Slow) DE: 0dB (Fast) DE: 3dB (Slow) DE: 3dB (Fast) DE: 6dB (Slow) DE: 6dB (Fast) Zero db results are slightly misleading r7 SAS-2 Zero-Length Test Load Characterization 46

47 Eye Height vs. Insertion Loss To better understand how insertion loss effects eye height simulations were repeated with PCB trace models. The trace structure was selected to obtain 100 ohms differential and noticeable skin effect losses. The trace length was computed to obtain 0.5 to 6dB of insertion loss at 3GHz in increments of 0.5dB Eye height measurements were made by taking approximating the average value obtained after transition bits r7 SAS-2 Zero-Length Test Load Characterization 47

48 Measurement Correlation All previous simulations utilized S-Parameter files of the interconnect. To validate the simulation results, measurements were made with using the actual hardware of case #1 and #2. Due to hardware limitations, only the 0dB cases are measured r7 SAS-2 Zero-Length Test Load Characterization 48

49 Case #1, 3Gbps, 0dB Eye Height Measured / Simulated: 890mV / 926mV r7 SAS-2 Zero-Length Test Load Characterization 49

50 Case #2, 3Gbps, 0dB Eye Height Measured / Simulated: 759mV / 794mV r7 SAS-2 Zero-Length Test Load Characterization 50

51 Case #1, 6Gbps, 0dB Eye Height Measured / Simulated: 844mV / 893mV r7 SAS-2 Zero-Length Test Load Characterization 51

52 Case #2, 6Gbps, 0dB Eye Height Measured / Simulated: 679mV / 692mV r7 SAS-2 Zero-Length Test Load Characterization 52

53 Trace Model Insertion Loss db case1 0.5dB 1dB 3dB GHz r7 SAS-2 Zero-Length Test Load Characterization 53

54 Eye Height vs. Channel Loss 100% Reduced Eye Height 90% 80% 70% 3Gbps 6Gbps 60% Channel Loss (db) (Insertion Loss Measured at 3GHz) r7 SAS-2 Zero-Length Test Load Characterization 54

55 Comparison to Nyquist for 6Gbps 100% Reduced Eye Height 90% 80% 70% 3Gbps 6Gbps 60% Channel Loss (db) (Insertion Loss Measured at 3GHz) r7 SAS-2 Zero-Length Test Load Characterization 55

56 Fixture De-embedding De-embedding appears to be a reasonable method for removing errors in near-end measurements induced by test fixtures and cables However, some level of expertise in modeling tools and the availability instrumentation that can characterize fixturing and cables is required Can we obtain a crude estimate of bit amplitude by measuring insertion loss at the Nyquist frequency and by applying fractional power x? Estimate = S 21 (Nyquist) x For this case, x = 0.72 is selected r7 SAS-2 Zero-Length Test Load Characterization 56

57 Results of Eye Height Estimation and Error Eye Height Estimated from S21 at Nyquest Estimation Error 100% 3% Reduced Eye Height 90% 80% 70% 6Gbps 3GHz Error 2% 1% 0% -1% -2% % Channel Loss (db) -3% Channel Loss (db) (Insertion Loss Measured at 3GHz) r7 SAS-2 Zero-Length Test Load Characterization 57

58 Fixture De-embedding by Estimation from Nyquist This simple sample indicates that rigorous modeling of test fixture effects may not be required in order to obtain accurate estimates of transmitter waveforms This assumption may fail if return loss effects are no longer negligible r7 SAS-2 Zero-Length Test Load Characterization 58

59 Fixture Return Loss Clearly the insertion loss is a major factor in the waveform degradation. The return loss may also be a factor. Three fixtures were measured to determine typical return loss values. Note that only the SMA launch and fixture trace are included r7 SAS-2 Zero-Length Test Load Characterization 59

60 Estimated Test Fixture Return Loss db Fixture 1 Fixture 2 Fixture 3 GHz r7 SAS-2 Zero-Length Test Load Characterization 60

61 Instrumentation Return Loss It was noted that previous return loss data neglects instrumentation return loss. To include these additional imperfections, a collection of sampling scopes, real-time sampling scopes and other SMA-connected instrumentation was characterized with a network analyzer. The scope data is divided into two groups with one having a specified bandwidth below 10GHz and the other one above r7 SAS-2 Zero-Length Test Load Characterization 61

62 Instrumentation BW < 10GHz db GHz RT Sampling 1 RT Sampling 2 RT Sampling 3 RT Sampling 4 RT Sampling 5 RT Sampling 6 RT Sampling 7 RT Sampling r7 SAS-2 Zero-Length Test Load Characterization 62

63 Instrumentation BW > 10GHz db Sampling 1 Sampling 2 Sampling 3 Sampling 4 RT Sampling 1 RT Sampling 2 RT Sampling 3 RT Sampling 4 RT Sampling 5 RT Sampling 6 RT Sampling 7 RT Sampling 8 GHz r7 SAS-2 Zero-Length Test Load Characterization 63

64 Instrumentation Return Loss From the instrumentation data, assume a flat return loss curve estimate up to 6GHz This level will then need to be added to the previous fixture estimate. Insertion loss and return loss of the fixture itself further reduce the observed return loss of combined fixture/instrumentation configuration. This will also be accounted for r7 SAS-2 Zero-Length Test Load Characterization 64

65 Estimated Combined Fixture/Instrumentation Return Loss db Fixture 1 Fixture 2-21 GHz r7 SAS-2 Zero-Length Test Load Characterization 65

66 Measured Combined Fixture/Instrumentation Return Loss db GHz Fixture1+Sampling Fixture2+Sampling r7 SAS-2 Zero-Length Test Load Characterization 66

67 Measurement Results The resonances of shorter electrical length appear to have reduced the estimated value r7 SAS-2 Zero-Length Test Load Characterization 67

68 Return Loss Magnitudes S11 (db) Reflected Power (db) Reflected Power (mag) Transferred Power (mag) Transferred Amplitude (mag) Amplitude Reduction (db) % 99.0% 99.5% % 98.2% 99.1% % 96.8% 98.4% % 94.4% 97.1% % 90.0% 94.9% r7 SAS-2 Zero-Length Test Load Characterization 68

69 Uncompensated Oscilloscope Insertion Loss The possibility that measurement errors exist due to non-ideal compensation of insertion loss at the oscilloscope front end was discussed at the last face to face. Three classes of oscilloscopes were tested RT with BW < 10GHz, RT with BW < 10GHz and an ET Scope with BW > 10GHz Amplitude error due to interconnect loss and accuracy of the sinusoidal source was deembedded by measuring connecting the signal reference point to a spectrum analyzer r7 SAS-2 Zero-Length Test Load Characterization 69

70 Measured Oscilloscope Amplitude (223mVrms Sinusoid / 0dBm) 223mVrms Sinusoidal Amplitude at Oscilloscope (Cable De-embedded) Vrms >10GHz RT <10GHz RT >10GHz ET GHz r7 SAS-2 Zero-Length Test Load Characterization 70

71 Measurement Results The traces demonstrate that some instruments do have uncompensated insertion loss that increases with frequency and will generate ISI Both RT scope traces show ~10% reduction in amplitude as the sinusoidal input is varied from 100MHz to 3GHz This approximates the entire loss profile of case #1! The measured waveform is clearly reshaped and uncompensated insertion loss should be modeled and de-embedded from the measurement r7 SAS-2 Zero-Length Test Load Characterization 71

72 07-013r7 SAS-2 Zero-Length Test Load Characterization 72