Ripple Method for Evaluating Residual Errors

Transcription

1 Ripple Method for Evaluating Residual Errors H. Heuermann Univ. of Applied Sciences Aachen, Institute of High Frequency Tech., Eupener Str. 7, D-5266 Aachen, Germany Heuermann HF-Technik GmbH Am Zirkus 4a; D Stolberg, Germany

2 Outline Motivation Brief Repetition Error Sources Classical Ripple Test Modified Ripple Test and Renormalization Technique Automatic Error-Bar Calculation with ErrorBar-S Numerical Investigations Wafer-Level Implementation Conclusion 2

3 Motivation This presentation addresses users of standard Vector Network Analyzers (VNA) Many users believe their very flat S-parameter measurements, but errors are still there Reflection measurements have larger errors Only worst case estimations of S-parameter measurement errors 3

4 Motivation ErrorBar-S: S-parameter results with error bars for any user 4

5 Outline Motivation Error Sources Classical Ripple Test Improvements of the Ripple Test Numerical Investigations Wafer-Level Implementation Conclusion 5

6 Error Sources Reflection or transmission measurement: r=s ii, t=s ij, i,j=,2,3, 6

7 Error Sources Non-ideal calibration standards Problems with matched load and open Calibration method high Cable phase instability Repeatability of connections Temperature Noise low 7

8 Outline Motivation Error Sources Classical Ripple Test Improvements of the Ripple Test Numerical Investigations Wafer-Level Implementation Conclusion 8

9 Classical Ripple Test For small measurement errors All errors are included in the final Source Match with the error vector r q VNA error model for ripple test 9

10 Classical Ripple Test Calculation of the Source Match: long matched transmission line + short Setup for the r q - calculation for ripple test

11 Classical Ripple Test Perfect VNA measurement: Z = Z q Results of the reflection coefficient measurement (matched line loaded with the short)

12 Classical Ripple Test Real VNA measurement: Z Z q Results of the reflection coefficient measurement (matched line loaded with the short) 2

13 Classical Ripple Test Source of ripple: real reflection (r w ) measured reflection (r m ) Model: emergence of ripples in the r-plane 3

14 Classical Ripple Test How to calculate error? see ARFTG 27, EuMW 27 and IMS 26 What else? Measurement errors over the smith chart Theory of modified ripple test Theory of renormalization technique Results of coaxial measurements Results of on-wafer measurements 4

15 Outline Motivation Error Sources Classical Ripple Test Improvements of the Ripple Test Numerical Investigations Wafer-Level Implementation Conclusion 5

16 Improvements of the Ripple Test Modified Ripple Test Using two-port measurement of reference line: match of the reference line represents the Source Match Calculation of the Source Match can be automated (software) 6

17 Improvements of the Ripple Test Renormalization Technique Impedance of reference line: Skin depth change impedance over frequency Planar reference lines are dispersive Solution: Line characterization (e.g. NIST methods) Renormalization from/to the line Z 7

18 Re(Zo), Ohm Improvements of the Ripple Test Renormalization Technique Line Impedance Z Determination of the reference line impedance Frequency (GHz) Impedance of a coplanar waveguide line (CSR-8, SUSS MicroTec) with the respect to the NIST method 8

19 Magnitude (db) Improvements of the Ripple Test Renormalization Technique Ohm Reflection S Renormalized Measurement results of a coplanar line Frequency (GHz) Return loss of a matched transmission CPW line (CSR-8, SUSS MicroTec): in the 5 system renormalized 9

20 Improvements of the Ripple Test Automatic Error-Bar Calculation with ErrorBar-S New product includes Modified and classical ripple test Renormalization S-parameter illustration and export Source match illustration and export 2

21 Outline Motivation Error Sources Classical Ripple Test Improvements of the Ripple Test Numerical Investigations Wafer-Level Implementation Conclusion 2

22 Numerical Investigations Test modified ripple test with measurement values from Matlab: Experiment with perfect conditions Shows accuracy of modified ripple test Remember: Ripple test is only an error estimation! 22

23 Magnitude (db) Numerical Investigations Accuracy estimation: simulation results Reflection S Decreasing error Return loss of a mismatched 25 transmission line after perfect SOLT and 6dB reproducibility error Frequency (GHz) 23

24 Magnitude Error (db) Numerical Investigations Errors over the Smith-Chart: Magnitude Reflection error of load: -2 db Short and open: perfect; Using SOLcalibration Working at one frequency point Imaginary Part - - Real Part 24

25 Phase Error (Degree) Numerical Investigations Errors over the Smith-Chart: Phase 3 2 Reflection error of load: -2 db Short and open: perfect; Using SOLcalibration Imaginary Part - - Real Part 25

26 Magnitude Error (db) Numerical Investigations Source Match over Smith-Chart -2 db Imaginary Part Real Part.5 Source match is the match of the load: -2dB Reference measurement of a perfect load for all S- parameters 26

27 Magnitude Error (db) Numerical Investigations Delta Match over Smith-Chart - Source match is the match of the load: -2dB Imaginary Part Real Part.5 Reference measurement of a perfect load for all S- parameters 27

28 Magnitude Error (db) Numerical Investigations Errors over the Smith-Chart From modified ripple test calculated measurement error -4.5 Imaginary Part Real Part.5 28

29 Phase Error (Degree) Numerical Investigations Errors over the Smith-Chart From modified ripple test calculated measurement error 5.5 Imaginary Part Real Part 29

30 Magnitude Error (db) Numerical Investigations Errors over the Smith-Chart Errors of standards: Load: -3 db Short:.5 Open: Imaginary Part Real Part.5 Using real SOL-data 3

31 Phase Error (Degree) Numerical Investigations Errors over the Smith-Chart Errors of standards: Load: -3 db Short:.5 5 Open: 2 Using real SOL-data Imaginary Part - - Real Part 3

32 Numerical Investigations Errors over the Smith-Chart Modified ripple-test: Source match is the match of the load: -3dB Reference measurement of a perfect load for all S-parameters Using real SOL-data 32

33 Magnitude Error (db) Numerical Investigations Errors over the Smith-Chart From modified ripple test calculated measurement error -4 Imaginary Part - - Real Part Using real SOL-data 33

34 Phase Error (Degree) Numerical Investigations Errors over the Smith-Chart From modified ripple test calculated measurement error.5 Imaginary Part Real Part.5 Using real SOL-data 34

35 Magnitude Error (db) Numerical Investigations Errors over the Smith-Chart From modified ripple test calculated measurement error -4.5 Imaginary Part Real Part Using real SOL-data & -6dB-source match 35

36 Phase Error (Degree) Numerical Investigations Errors over the Smith-Chart From modified ripple test calculated measurement error.5 Imaginary Part Real Part Using real SOL-data & -6dB-source match 36

37 Outline Motivation Error Sources Classical Ripple Test Improvements of the Ripple Test Numerical Investigations Wafer-Level Implementation Conclusion 37

38 Magnitude (db) Wafer-Level Implementation Accuracy estimation: measurement results Reflection S Frequency (GHz) Return loss (magnitude) of a 25 load (CSR-8, SUSS MicroTec) and calculated error bars 38

39 Phase (degree) Wafer-Level Implementation Accuracy estimation: measurement results Reflection S Return loss (phase) of a 25 load (CSR-8, SUSS MicroTec) and calculated error bars Frequency (GHz) No ripple of measurement is real 39

40 Magnitude (db) Wafer-Level Implementation Accuracy estimation: measurement results Reflection S Return loss (magnitude) of a load (CSR-8, SUSS MicroTec) and calculated error bars Frequency (GHz) 4

41 Phase (degree) Wafer-Level Implementation Accuracy estimation: measurement results Reflection S Return loss (phase) of a load (CSR-8, SUSS MicroTec) and calculated error bars Frequency (GHz) Ripple of measurement is real 4

42 Magnitude (db) Wafer-Level Implementation Reflection S Frequency (GHz) Accuracy estimation: measurement results Return loss of a 25 stub line (CSR-8, SUSS MicroTec) and calculated error bars Ripple of measurement is real 42

43 Outline Motivation Error Sources Classical Ripple Test Improvements of the Ripple Test Numerical Investigations Wafer-Level Implementation Conclusion 43

44 Conclusion The classical ripple test was presented Improvements with modified ripple test and renormalization technique were a good start to come to automatic error bar calculation Numerical investigations of modified ripple shows the good quality The wafer-level implementation demonstrate good accuracy estimations The first commercial product ErrorBar-S includes all these techniques: combines classical and modified ripple test 44

45 References European co-operation for Accreditation, Expression of the Uncertainty of Measurement in Calibration, EA-4/2, Heuermann, H., Old and New Accuracy Estimation of S-Parameter Measurements with the Ripple- Test, MTT-S International Microwave Symposium Workshop TMB, San Francisco, Juni 26. Eul, H.J., Schiek, B., A Generalized Theory and New Calibration Procedures for Network Analyzer Self-Calibration, IEEE Trans. Microwave Theory Tech., MTT-39, Apr. 99, pp Heuermann, H., Schiek, B., Robust Algorithms for Txx Network Analyzer Procedures, IEEE Trans. Instrum. Meas., Feb. 994, pp Marks, R., Williams, D., Characteristic Impedance Determination Using Propagation Constant Measurements, IEEE Microwave and Guided Wave Lett., vol., pp. 4-43, June 99. A. Rumiantsev, H. Heuermann, S. Schott, A Robust Broadband Calibration Method for Wafer-Level Characterization of Multiport Devices, 69 th ARFTG Conference Digest, June 27 Rumiantsev, A., Doerner, R., Thies, S., Calibration Standards Verification Procedure Using the Calibration Comparison Technique, 36 th European Microwave Conference Digest, September