1. x / 1 Measurement Accuracy Considerations for Characterizing Low Impedance Stripline and Microstrip PC Board Traces John Rettig Principal Engineer Tektronix, Inc. Beaverton, OR 97077 (503)-627-3232 john.b.rettig@tek.com

2. x / 2 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

3. x / 3 Interconnect Issues  At high-speeds, interconnect limits system performance.  It is desirable to characterize and model interconnect to predict the performance early in the design phase. Impedance Change Transmitted Energy Incident Energy Reflected Energy

4. x / 4 Interconnect Issues  Whenever energy transmitted through any medium encounters a change in impedance, some of the energy is reflected back toward the source  The amount of energy reflected is a function of the transmitted energy and the magnitude of the impedance change  The time between the transmitted energy and the reflection return is a function of the distance and velocity of propagation

5. x / 5  Time Domain Reflectometry - a measure of reflection in an unknown device, relative to the reflection in a standard impedance.  Compares reflected energy to incident energy on a single-line transmission system –Known stimulus applied to the standard impedance is propagated toward the unknown device –Reflections from the unknown device are returned toward the source –Known standard impedance may or may not be present simultaneously with the device or system under test TDR Overview

6. x / 6 TDR Overview - Typical System Incident Step 50 Step Generator Reflections Sampler t = 0

7. x / 7 TDR Overview Elements –High speed stimulus, usually a step generator –High speed sampling oscilloscope –Back (internal) termination –Interconnect system –Device launch –Known standard impedance

8. x / 8 TDR Overview TDR Waveforms - Open, Short and 50  terminations Amplitude Open (Z =  ) (Z = 50  ) Short (Z = 0) Time Reflected + 1  0  - 1  t0t0 t1t1 Incident

9. x / 9 Reflection Coefficient and Impedance Where Z represents the test impedance Z 0 is the reference impedance  is measured by the oscilloscope

10. x / 10 Measuring Impedance    200ps/div div Cursor 1 at 50  Reference Line Cursor 2 at Reflection from Load Resistor

11. x / 11 Measuring Impedance

12. x / 12 Nonlinear Impedance /  Mapping  Everything else equal, lower impedance line measurements can tolerate more  error for a given impedance tolerance  Assumed conditions –250 mV step –50  Reference Line  1 mV or 4 m  error equates to: –0.40  for a 50  test line –0.24  for a 28  test line –0.79  for a 90  test line –1.23  for a 125  test line

13. x / 13 Measuring Impedance

14. x / 14 TDR Overview - Lumped Discontinuities

15. x / 15 TDR Overview - Microstrip Discontinuities Connector Capacitive Discontinuity Inductive Discontinuity Open Circuit Incident Step Volts or  Round Trip Time

16. x / 16 TDR Accuracy Issues - First Order, no Explicit Z 0 Standard  Controlled-impedance environment –Back termination impedance –Interconnect system  Vertical –Step amplitude accuracy –Sampler vertical accuracy  Horizontal –Step position accuracy –Sampler position accuracy

17. x / 17 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

18. x / 18 Employing a Standard Impedance  Inline with launch  Exchange with device under test TDR UNIT 1 2 DUT INTERCONNECT EXCHANGE INLINE

19. x / 19 Standard Impedances  Absolute Standards –Usually air lines built with extreme care employed for control of physical dimensions –Easily disassembled for checking dimensions –Standards Labs can characterize and periodically check –Based on physical equations like coaxial expression

20. x / 20 Standard Impedances  Golden Standards –Absolute accuracy less important –Can be anything stable, repeatable, portable –Often need several sets depending upon supply chain –Characterization needs “round robin” or similar redundant checking

21. x / 21 Microstrip Standards  Microstrip standards depend on many parameters with dispersed controls –Physical dimension (width w, thickness t, height h) –Dielectric media (glass and epoxy characteristics, glass/epoxy ratio, homogeneity) –PCB processes

22. x / 22 Inline Versus Exchange Standard Always in place - remains primary standard in system One connector change Two connector aberrations Measurement zone after standard location Loss in standard can affect accuracy Additional multiple reflection compounding Recommending exchanged standard in this presentation Substituted for DUT - interconnect becomes secondary standard One connector change One connector aberration Measurement zone at standard location No affect from loss in standard No additional multiple reflection compounding InlineExchange

23. x / 23 Standard Impedances  For reasons that will become apparent, want to use standard impedance close to device under test –Error on device measurement is primarily additive  50  is the most commonly available and least expensive line –Can Parallel Two Lines for 25  Using Tee(s) Z1Z1 Z2Z2

24. x / 24 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

25. x / 25 TDR Resolution  Insufficient TDR resolution –Results from closely spaced discontinuities being smoothed together –Can miss details of device under test –May lead to inaccurate impedance readings

26. x / 26 TDR Resolution SMA through F-F barrel (8.6 mm dielectric) Each end individually loosened 2.5 turns (1.8 mm) Both ends loosened 2.5 turns; risetime limits –Top -- Full (~28 ps) –Mid -- 50 ps –Bottom -- 75 ps

27. x / 27 TDR Resolution  TDR system risetime is related to resolution –Reflections last as long as the incident step and display as long as the system risetime –First discontinuity reflection is witnessed –Twice the propagation delay between discontinuities elapses –Second discontinuity reflection is witnessed –Ideally, leading corner of reflection from second discontinuity arrives back at first discontinuity no earlier than lagging corner of reflection from first discontinuity, thus

28. x / 28 TDR Resolution  Previous rule assumes 0-100% ramp model; real world specifies 10-90% quadratic-type responses  System rise time is characterized by fall time of reflected edge from ideal short at test point  Other second order factors enter picture  System rise time approximated by:

29. x / 29 TDR Resolution  Resolution Factors –Rise Time –Settling Time –Foot and Preshoot Settling Time Foot Rise Time Preshoot

30. x / 30 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

31. x / 31 Stimulus Amplitude Error  Percentage Z error roughly doubles percentage  error near Z 0  Amplitude shift with DC termination –6-8 m  typical 0 - 50  –6-8 m  typical 50  - open  Can be entirely compensated by directly measuring incident amplitude Amplitude impacts here

32. x / 32 Stimulus Amplitude Compensation  Measure incident amplitude during calibration cycle  Keep same DC loading on TDR system during device test  Calibration cycle can be same as standard impedance check  Note that use of standard impedance close to device minimizes any residual errors

33. x / 33 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

34. x / 34 Sampler Nonlinearity Error  Samplers have limited dynamic range relative to general purpose oscilloscopes  Extremes of dynamic range may have small nonlinear behavior  Use of limited range stimulus minimizes this  Lower than Z 0 impedance measurement environment is less vulnerable because of inverted reflection  Usually very small effect

35. x / 35 Sampler Nonlinearity Error  Lower than Z 0 impedance measurement environment is less vulnerable because of inverted reflection Open (Z =  ) (Z = 50  ) Short (Z = 0) Time Reflected + 1  0  - 1  t0t0 t1t1 Incident

36. x / 36 Sampler Nonlinearity - Recommendations  Preserve identical gain and possibly offset settings for all portions of a given standard impedance calibration followed by DUT measurements

37. x / 37 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

38. x / 38 Rho Zero Error  Start of incident edge may not register as zero   Ohms calibration usually assumes -1  = 0 

39. x / 39 Rho Zero Error  Physical hardware may not use zero volt baseline - baseline instead restored in software  Typical 2 m  hysteresis total (  1 m  ) in software algorithm

40. x / 40 Rho Zero Error  Can be entirely compensated - use amplitude measurements for incident and reflected amplitudes  Disable software baseline restoration if present  In absence of other accuracy improvement techniques, use of a set zero rho feature along with standard impedance close to device, provides remarkable improvement

41. x / 41 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

42. x / 42 Stimulus / Sampler Aberrations  Aberrations - foot, preshoot, ringing, longer term effects –Can amount to several percent –Short or long term –All reflections are replicas of the original step, filtered by reflecting device –Foot distorts response before the reflection, affecting establishment of baseline  With linear devices, it doesn’t matter how aberrations are apportioned between stimulus and sampler

43. x / 43 Aberration Replication Settling replica Preshoot replica Incident Reflected

44. x / 44 Stimulus / Sampler Aberrations  Typical specifications –10 ns to 20 ps before step: ±3% or less –<300 ps after step: +10%, -5% or less –300 ps to 5 ns after step: ±3% or less –Elsewhere: ±1% or less

45. x / 45 Stimulus / Sampler Aberrations 100 ps/div 1 ns/div 10 ns/div 100 ns/div

46. x / 46 Stimulus / Sampler Aberrations  Measurement over a zone to establish  levels and amplitude includes: –Device reflections –Stimulus / Sampler aberrations incident –Stimulus / Sampler aberrations reflected  Recommended measurement technique will involve: –Use of exchanged impedance standard –Measurement over identical zones for both standard impedance and device

47. x / 47 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

48. x / 48 Launch resistance  Indiscernible from device impedance  Effect is additive, always positive, and similar to an equal amount of standard impedance error Incident Step 50 Step Generator Reflections Sampler t = 0 R contact

49. x / 49 Launch resistance  May not necessarily be repeatable  Most often is reasonable bounded (tens of m  ) –Can be measured with 4-wire measurement - replicate several in series if necessary  May also be present to some degree in ground contact  Use of exchanged standard impedance will cancel launch resistance provided it is constant  Otherwise, make estimate of mean value and include for every measurement

50. x / 50 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

51. x / 51 Launch Inductance  Caused by –Non-characteristic launches –Air gaps at launch –contorted launch paths –Poor attention to ground attach  Time constant is longer for smaller Z 0

52. x / 52 Launch Inductance  May affect measurement zone –Lumped time constant decay usually lasts much longer than propagation through –Multiple reflections carried into zone –If C present, may have second order ring (though usually only a larger Z DUT is more vulnerable)  May be partially compensated –Must be absolutely repeatable attach geometry –Standard impedance Z 0 must be very close to Z DUT

53. x / 53 Launch Inductance Launch from 0.141” semi-rigid coax to microstrip through center and ground wires, with gap L-R: gap = 0/1/2/3 mm

54. x / 54 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

55. x / 55 Crosstalk  The coupling of energy from one line to another  Three Elements Contribute to Crosstalk: –Port terminations –Stimulus –Moding on transmission system  Generalities: –Proportional to the line length –Proportional to rise time of driving signal –Can be positive or negative (inductive or capacitive) –Occurs in both forward (near-end) and backward (far- end) directions –Non-characteristic port terminations make it worse

56. x / 56 Crosstalk Measurement Techniques  Set up TDR on aggressor line  Observe victim lines with TDT  Take care to terminate all other lines TDR 1 2 + 50 DUT

57. x / 57 Crosstalk Crosstalk between adjacent 50  runs on FR4 with W=2.5 mm S=2 mm Far end 50  terminations Aggressor: 200 m  /div Victim: 4%/div

58. x / 58 Crosstalk  Adjacent microstrips and striplines will always crosstalk to a degree if fringing fields overlap  Impedance measurements include loading of adjacent lines, whether intended or not  Three key questions: –Is the geometry measured representative of the geometry in the application? –Are the port terminations of adjacent lines representative? –Are the driving conditions of adjacent lines representative?

59. x / 59 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

60. x / 60 Superposition of Reflections Z1Z1 Z2Z2 Z3Z3 Z4Z4 Z0Z0 Z0Z0 Effect of multiple impedance discontinuities

61. x / 61 Problems with Multiple Reflections  Multiple reflections –Can appear anywhere in a waveform –Can look like anomaly or Z-shift –Might be controllable, but always are predictable  TDR trace is a reflection profile, not a true impedance profile  Resolution may be lost due to rise time degradation

62. x / 62 Problems with Multiple Reflections  TDR trace may be difficult to interpret in the presence of multiple reflections  Can be deconvolved with DSP if impedance profile and other information is needed  Can be compensated with techniques using exchanged standard impedance, if only impedance is needed

63. x / 63 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

64. x / 64 Transmission Line Losses  Affects TDR and TDT measurements –Long dribble up time constants –Impedance accuracy affected –High frequency details attenuated –Risetime (and therefore resolution) increased  Loss mechanisms –DC resistive losses –Skin effect losses –Dielectric losses –Coupled losses (crosstalk loss)

65. x / 65 Transmission Line Losses 100 mm 50  microstrip on FR4 Same with 1.5 meters RG58 inserted in front of DUT

66. x / 66 Transmission Line Losses  Generally, skin effect losses dominate at microwave frequencies  Skin effect loss effect on impedance measurements can be compensated within reason  Resolution cannot be regained easily

67. x / 67 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

68. x / 68 Sequential Timebase Architecture Trigger Recognizer Gated Oscillator Counter Delay Vernier Integer (30 bits)Fractional (18 bits) Programmed Delay Register (48 bits) (11801C Illustrated) T delay = 0 - 2.62144 ns T GVCO = 2.62144 ns Strobe drive Sampling Head Delay Vernier Compensation Strobe drive Sampling Head Delay Vernier Compensation

69. x / 69 Horizontal Error Sources  Gated oscillator error  Segment mismatch error  Delay vernier error

70. x / 70 Horizontal Error Sources Time from trigger event (programmed) (actual) Gated oscillator period error Trigger to event error T 0 Segmentation Segment length T segment mismatch error T i Time from trigger event vernier error  (t) Delay Ideal 1:1 slope Gated oscillator period T GVCO

71. x / 71 Overall Timing Error  Term-by-term components of error: –Gated oscillator period error –Net segment mismatch (sum of all segment mismatches T i up to time t) –Delay vernier error  (t), with the argument normalized to range of (0, T segment ).

72. x / 72 Interval Error  Mismatch and GVCO contributions cancel outside of (t 1, t 2 ) interval  For short time intervals, delay vernier error terms dominate. Single measurement (t 1, t 2 ):

73. x / 73 Interpreting Specifications   10 ps interval: 1 ps + 4 ppm of position, typical  100 ps interval: 2.5 ps + 4 ppm of position, typical   1 ns interval: 4 ps + 40 ppm of interval + 4 ppm of position, typical

74. x / 74 Does This Matter?  No, for most TDR testing  Yes, for a few situations: –Absolute accuracy on long run characterizations –Matching long runs (clock distribution)  Accuracy improvement technique “Picosecond Time Interval Measurements”, John B. Rettig and Laszlo Dobos, IEEE Transactions on Instrumentation And Measurement, Vol. 44, No., 2, April 1995, pp. 284-287.

75. x / 75 Agenda  Introduction  Sources of TDR error –Standard impedance error –Inadequate resolution –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Stimulus / Sampler aberrations  Recommended TDR measurement technique –Launch resistance –Launch inductance –Device coupling error –Multiple reflections –Line loss –Strobe placement error

76. x / 76 TDR Accuracy Improvement IPC - TM - 650 Method  Eliminates These Errors –Stimulus amplitude error –Sampler nonlinearity error –Rho zero error –Most Stimulus / Sampler aberrations –Launch resistance and inductance, provided it is duplicated on the standard impedance –Multiple reflections up to exchange point –Line loss

77. x / 77 TDR Accuracy Improvement - Steps:  Preliminary Steps –Preset TDR –Turn off Rho cal –Turn off Baseline correct TDR 1 2 DUT INTERCONNECT STD IMPEDANCE

78. x / 78 TDR Accuracy Improvement  Characterize interconnect as secondary standard –Disconnect standard impedance and measure size of step incident on standard impedance V I that arrives back at sampler. Re-connect standard impedance and measure size of step reflected from standard impedance V R, that arrives back at sampler. Calculate Z INT : Z INT = Z STANDARD (V I - V R ) / (V I + V R ) VIVI intercon Open VRVR intercon standard Open

79. x / 79 TDR Accuracy Improvement  Measure DUT impedance against interconnect –Disconnect DUT and measure size of step incident on DUT V I that arrives back at sampler. –Re-connect DUT and measure size of step reflected from DUT V R that arrives back at sampler. Calculate Z DUT :Z DUT = Z INT (V I + V R ) / (V I - V R ) VIVI intercon Open VRVR intercon DUT Open

80. x / 80 TDR Accuracy Improvement - Advantages  Reduces problem to three critical measurements –Interconnect (secondary standard) impedance Z INT –Incident step at DUT V I –Reflected step from DUT V R  Employs primary traceable standard - exchange method  Eliminates or minimizes most controllable sources of error

81. x / 81 TDR Accuracy Improvement - Caveats  Always measure with last line or DUT open circuited  Always measure V I and V R over identical time zones –Reduces sensitivity to aberrations, dribble-up –Reduces chance of measuring re-reflections  Minimize interconnect / TDR cable length and loss

82. x / 82 TDR Accuracy Improvement - Caveats  Use short, repeatable connector between interconnect and standard impedance or interconnect and DUT  Keep impedances close between impedance standard and DUT –Helps keeping baseline and topline on same screen –V R will be small, making high accuracy possible –Sensitivity to accuracy of V R will be very low

83. x / 83 TDR Accuracy Improvement Example Standard (25  ) VI = 243.04 mV VR = -84.32 mV DUT VI = 243.04 mV VR = -54.56 mV Z INT = 51.56  Z DUT = 32.66 

84. x / 84 Conclusions  Control as best as possible--or know limits of-- those items that can’t be compensated: –Standard impedance error –Inadequate resolution –Connector repeatability uncertainties  Employ stable, repeatable, and durable interconnect and launches  Keep the standard impedance close to the DUT impedance  Employ TDR accuracy improvement to compensate those items that can be dealt with