A Four-port Microwave Measurement System to Speed On-wafer Calibration and Test
Description of a new multiport calibration technique utilizing a four-port wafer-probe system
A four-port measurement system comprising an inexpensive coaxial switch matrix and a comprehensive free software package extends the capabilities of a user's two-port automatic network analyzer and probe station to three- and four-port measurements. The software includes routines for calibration, measurement and correction, and generates normal and differential scattering and impedance parameters in a variety of standard data formats. Any in-line, two-port calibration, including the multiline thru-reflect-line method, may be used to correct the measurement data.
A multiport calibration technique is described.1 This system may be calibrated with any conventional two-port method, including the accurate multiline thru-reflect-line (TRL) calibration.2
Historically, collecting accurate four-port data with a two-port automatic network analyzer (ANA) has been tedious. The multiport data are based on a series of two-port measurements of the device with the unconnected ports terminated with a precision matched load. The ensemble of two-port data is used to fill the three- or four-port scattering parameter (S-parameter) matrix. The analytical method of Tippet and Speciale3 can be used to correct for the imperfect loads used to terminate the unused ports during the two-port measurement. Besides the time and difficulty involved in making multiple connections to complete a single device measurement, the accuracy is limited by the repeatability of the connections and the imperfect termination impedances connected to the unused ports.
These problems are circumvented with automated four-port measurement systems.4 However, the most common method of calibrating these automated systems involves performing six two-port short-open-load-thru (SOLT) calibrations -- one calibration between each pair of ports. In an on-wafer testing environment, the port orientations are orthogonal and it is not possible to connect a thru standard between orthogonal probes. Furthermore, on-wafer SOLT calibrations can reduce accuracy at higher frequencies.5
The new multiport calibration procedure is based on a previously published method1 and utilizes the four-port wafer-probe system architecture shown in Figure 1. This system consists of a conventional two-port ANA and six computer-controlled coaxial switches to route signals from port 1 of the analyzer to the north (N), south (S) and west (W) probes, and signals from port 2 of the analyzer to the north (N), south (S) and east (E) probes. The system is arranged so that a unique termination impedance is applied to each probe when it is not being used to connect the device under test to the analyzer. Figure 2 shows the on-wafer four-port probe system.
Calibrating the system requires three conventional in-line two-port calibrations. The calibrations are performed with port 1 connected to W and port 2 to E, port 1 connected to N and port 2 to S, and port 1 connected to S and port 2 to N. This procedure reduces to an absolute minimum the number of connections made to the standards. Only two connections to the calibration standards are required. The third in-line measurement is made using the switches to swap the N and S ports.
Since it is not required to connect standards between orthogonal probes, this procedure allows on-wafer multiport measurements to be corrected by any standard two-port calibration, including SOLT, line-reflect-match (LRM) and line-reflect-reflect-match (LRRM), plus the most accurate on-wafer calibration, multiline TRL.6 Following calibration, any device with one to four ports can be tested.
This three calibration algorithm works by first correcting all measurements with a W-E one-tier calibration. To translate a measurement made at a N or S port, it is necessary to cascade the scattering-parameter matrix describing the signal path from that port back to the W or E port as required onto the first-tier W-E-corrected scattering-parameter matrix. Therefore, four unique matrices are needed to describe the translations N to W, N to E, S to W and S to E. These matrices are generated by performing two second-tier calibrations, one in the N-S orientation and the other in the S-N orientation, on top of the first-tier W-E calibration. However, these two second-tier calibrations require only one physical connection because the switches swap the N and S ports automatically for the operator.
The system performs four-port measurements from 50 MHz to 26.5 Ghz. Figure 3 shows the automated 3.5 mm coaxial switches, which are off-the-shelf components. Four of the six switches incorporate internal 50 Ω loads that uniquely terminate the test port not in use. These termination impedances are not necessarily 50 Ω; therefore, the actual impedances are measured during the calibration procedure and apply a correction when de-embedding.
Calibrating, measuring, viewing and formatting multiport data is ac complished with a self-contained Windows-compatible LabVIEW executable library.7 The front panel, shown in Figure 4, is used to set up, save and retrieve a calibration menu, and call up the measurement and calibration routines.
Figure 5 shows the measurement menu. The top toggle key, which is labeled Measure West/East Cal. Stds., selects the switch configuration and standards to be measured. In the W-E configuration, port 1 of the analyzer is connected to the W probe and port 2 is connected to the E probe, and the analyzer performs a single two-port measurement. Toggling the top key changes the system to Measure North/South Cal. Stds. mode. In this configuration, the software automatically uses the switch matrix to measure each standard twice, first with port 1 connected to N and port 2 connected to S, and then with port 1 connected to S and port 2 connected to N.
Independent file paths for storing the calibration standard measurements and for device under test (DUT) measurements, shown towards the bottom of the measurement window, ease the organization of complex measurement protocols.
Figure 6 shows the TRL calibration menu. Parameters corresponding to particular properties of the calibration standards, estimated dielectric constant or waveguide cutoff frequency, and output reference impedance may be edited and saved to a calibration menu. Following execution, de-embedded measurements can be displayed to facilitate troubleshooting and help the user verify that the calibration is satisfactory. The measurement software also accommodates port extensions or added adapters, as shown in Figure 7. This feature may be used to account for additional test fixtures or probe-head transitions between the measurement reference planes and the multiport device.
The final menu in this sequence is used to de-embed and save corrected multiport device data. It allows the user to assign port numbers to each probe, as well as select the number of ports used in the data output. Corrected device data may then be selected and plotted. For example, Figure 8 shows the magnitude plots for all sixteen S-parameters of a four-port device, divided into reflection terms and transmission terms, respectively. Various popular file formats for standard or differential S or Z (impedance)-parameters can be specified to facilitate analysis, design or modeling using commercial software packages.
The system can be used to make fully calibrated measurements on any device with up to four ports. Many potential applications exist in the coaxial and on-wafer environments, including packaged components, electronic package characterization8 and multiconductor transmission-line analysis.9 For example, Figure 9 presents measurement results for an asymmetric coupled transmission line on a silicon substrate.10 This graph shows the measured resistance per unit length of line (solid curve) for the two asymmetric transmission lines, labeled Rc11 and Rc22 , and the coupled line parameter, Rc12 . Also plotted in dashed lines are the lower and upper 95 percent confidence intervals and the line parameters calculated from the quasi-analytical method of Grotelüshen, et al.11
| || |
| || |
NIST MULTIPORT ROUND-ROBIN
Accurate measurement of multiport devices has been hindered by the lack of a verifiable calibration and test methodology. This exciting multiport round robin allows you to compare your measurements to NIST's 7-mm four-port measurements. The five devices provided by NIST include a directional coupler, a 3 dB power divider, a 90 hybrid coupler and a differential transmission line shown in Figure 10. To learn more or to participate, visit the NIST Web site at http://www.boulder.nist.gov/micro, which features project descriptions, a comprehensive list of on-line references and a convenient downloadable version of the multiport measurement software. *
1. D.F. Williams and D.K. Walker, "In-line Multiport Calibration," 51st ARFTG Conference Digest, June 12, 1998, pp. 8890.
2. R.B. Marks, "A Multiline Method of Network Analyzer Calibration," IEEE Transactions on Microwave Theory and Techniques, Vol. 39, No. 7, July 1991, pp. 12051215.
3. J.C. Tippet and R.A. Speciale, "A Rigorous Technique for Measuring the Scattering Matrix of a Multiport Device with a 2-port Network Analyzer," IEEE Transactions on Microwave Theory and Techniques, Vol. 30, No. 5, May 1982, pp. 661666.
4. SPTS-4 Four-port S-parameter VNA Test System, ATN Microwave Corp., North Billerica, MA.
5. D.F. Williams, R.B. Marks and A. Davidson, "Comparison of On-wafer Calibrations," 38th ARFTG Conference Digest, December 1991, pp. 6881.
6. R.B. Marks, "A Multiline Method of Network Analyzer Calibration," IEEE Transactions on Microwave Theory and Techniques, Vol. 39, No. 7, July 1991, pp. 12051215.
7. LabVIEW is a registered trademark of National Instruments Corp., Austin, TX. Windows is a registered trademark of Microsoft Corporation. The use of the these trade names and software does not represent an endorsement by the National Institute of Standards and Technology, and other products may work as well or better.
8. D.C. DeGroot and D. F. Williams, "National Institute of Standards and Technology Programs in Electrical Measurements for Electronic Interconnections," 7th Topical Meeting on Electrical Performance of Electronic Packaging, October 2628, 1998, pp. 4549.
9. D.F. Williams, J.E. Rogers and C.L. Holloway, "Multiconductor Transmission Line Characterization: Representations, Approximations and Accuracy," IEEE Transactions on Microwave Theory and Techniques, Vol. 47, No. 4, April 1999, pp. 403409.
10. U. Arz, D.F. Williams, D.K. Walker, J.E. Rogers, M. Rudack, D. Treytnar and H. Grabinski, "Broadband Measurement of Asymmetric Coupled Lines Built in a 0.25 µm CMOS Process," MTT-S International Microwave Symposium Digest, June 1116, 2000, pp. 609612.
11. E. Grotelüschen, L.S. Dutta and S. Zaage, "Quasi-analytical Analysis of the Broadband Properties of Multiconductor Transmission Lines on Semiconductor Substrates," IEEE Transactions on Comp., Packaging and Manufacturing, Tech.Part B, Vol. 17, August 1994, pp. 376382.
David K. Walker graduated from Hastings College in Hastings, Nebraska, in 1980 with Bachelor's degrees in physics and mathematics, followed by Bachelor's and Master's degrees in EE from Washington University in St. Louis in 1982 and '83, respectively. He spent eight years in industry working on microwave semiconductor device design and fabrication before joining NIST in 1991 as part of the MMIC Project in the Microwave Metrology Group. His work at NIST includes semiconductor fabrication and network analyzer calibration and measurement in the on-wafer environment. He holds five patents related to microwave technology.
Dylan F. Williams received a PhD in electrical engineering from the University of California at Berkeley in 1986. He joined the Electromagnetic Fields Division of the National Institute of Standards and Technology in 1989 where he develops metrology for the characterization of monolithic microwave integrated circuits and electronic interconnects. He has published over 80 technical papers and is the recipient of the Department of Commerce Bronze and Silver Medals, the Electrical Engineering Laboratory's Outstanding Paper Award, two ARFTG Best Paper Awards, the ARFTG Automated Measurements Technology Award and the IEEE Morris E. Leeds Award.
Allen Padilla graduated with a Bachelor's degree in electrical engineering from the University of Colorado at Boulder in 2000. He worked as an intern with the RF Electronics Group at NIST in Boulder in 19992000. He is currently employed by Motorola, Inc., Phoenix, AZ.
Uwe Arz received the Dipl.-Ing. degree from the University of Hannover, Germany, in 1994, and is currently working toward the PhD degree in the area of on-chip interconnect measurement methods. Since 1995, he has been a research and teaching assistant at the Laboratory of Information Technology, University of Hannover. His research interests include broadband characterization methods for high speed digital interconnects and packages. Mr. Arz was the recipient of the first ARFTG Microwave Measurement Fellowship Award.
Hartmut Grabinski received the Dipl.-Ing. degree from Fachhochschule Hannover, Germany, in 1977, and the Dipl.-Ing. and Dr.-Ing. degrees from the University of Hannover in 1982 and 1987, respectively. In 1993 he was granted the Dr.-Ing. habil. degree in `Theoretische Elektrotechnik.' Presently he is serving as a lecturer at the department of electrical engineering in Hannover. Since 1987 he has been the manager of the division Design & Test of the Laboratory for Information Technology, University of Hannover. His research interests include electrodynamics of interconnects and electrical performance of electronic packaging.