Comparison of Broadband Single-Sweep and Conventional Banded System On-Wafer S-Parameter Measurements up to 220 GHz

© 2025. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to use any copyrighted component of this work in other works must be obtained from IEEE. Link to publisher version with DOI: 10.1109/ARFTG65332.2025.11168146 Comparison of Broadband Single-Sweep and Conventional Banded System On-Wafer S-Parameter Measurements up to 220 GHz Liam Ausden1, Nick Ridler1, Andrej Rumiantsev2, Jon Martens3, Xiaobang Shang1 1National Physical Laboratory, London, UK, 2MPI Corporation, Chupei Hsinchu, Taiwan, 3Anritsu, Morgan Hill, CA, USA Abstract — This paper presents a comparison of on-wafer S-parameter measurements of coplanar waveguide (CPW) devices using broadband single-sweep and conventional banded systems, up to 220 GHz. Three attenuators and a pair of loads on a commercial calibration substrate were measured using these two different types of systems in the same laboratory, and the results are reported and discussed. Overall, good agreement was observed between the two different approaches. This study benchmarks the performance of the single-sweep system against conventional banded systems using on-wafer measurements and provides insights into the equivalence of these methods for other users. Index Terms — Measurement comparison, millimeter-wave measurement, on-wafer measurement, S-parameters, vector network analyzer. I. INTRODUCTION In recent years, broadband vector network analyzer (VNA) systems capable of measuring S-parameters from low frequencies (below 100 kHz) up to 220 GHz in a continuous sweep have become commercially available. Such systems meet the growing demand for broadband on-wafer characterization, such as for regulatory testing of out-of-band emissions or specific applications like those in [1]-[2]. Compared to the conventional method, which involves using two or more banded systems and stitching together measurement results, broadband systems mitigate several challenges. These include band-to-band discontinuities, the extensive time required for frequent system reconfigurations, and wear on probes and the contact pads of the device under test (DUT). Previous work [3]-[4] has evaluated some key measurement characteristics of a broadband 70 kHz – 220 GHz single-sweep system. In [5] an inter-laboratory comparison of broadband onwafer S-parameter measurements was conducted and the reproducibility of measurements determined. However, despite the potential value of benchmarking a new system, no study has focused primarily on comparing these broadband single-sweep solutions with the conventional approach using banded systems. In this work, on-wafer S-parameter measurements of four passive co-planar waveguide (CPW) devices from a commercially available calibration substrate (MPI TCS-050-100-W), were measured up to 220 GHz using a broadband single-sweep system and conventional banded systems for comparison. Banded measurements were performed, using all compatible systems available at the National Physical Laboratory (NPL), UK, at the time the comparison was undertaken. These were 2.4 mm coaxial (up to 50 GHz), WR10/W-band (75 GHz – 110 GHz), WR6.5/D-band (110 GHz – 170 GHz) and WR5/G-band (140 GHz – 220 GHz). The large frequency coverage allowed detailed comparison over nearly the full operational range of the single-sweep system. Additionally, the probe contact repeatability of each system was evaluated, providing further insight into the comparison. In [6] and [7], on-wafer S-parameter measurements from broadband single-sweep systems were compared with those from banded systems. However, both studies focused on largescale interlaboratory comparisons, with results obtained from multiple operators working in different laboratories and using different probing systems. Additionally, the DUTs were not completely identical, despite being manufactured from the same wafer in the same batch. In [8], a D-band silicon carbide substrate integrated waveguide (SIW) was characterized onwafer using both a 220 GHz single-sweep system and three banded systems. Our work complements these studies by presenting results from measurements on a commercial calibration substrate, which could help other users assess the accuracy of their own measurement results using the same type of calibration substrate. TABLE I DETAILS OF MEASUREMENT SYSTEMS Frequency Band VNA System Probe/Pitch (µm) Broadband single-sweep Anritsu VectorStarTM ME7838G MPI TITANTM/100 Coaxial (2.4 mm) Keysight PNA-X N5247B MPI TITANTM/75 WR10/W-band Keysight PNA-X N5247B with VDI extenders GGB Picoprobe®/100 WR6.5/D-band GGB Picoprobe®/100 WR5/G-band GGB Picoprobe®/75 II. EXPERIMENTAL SETUP All measurements were performed using the MPI TS150THZ manual probe station. Key details of the VNA systems and on-wafer probes, including pitch sizes, are given in Table 1. A. Broadband Single-Sweep System Broadband measurements were performed using an Anritsu VectorStarTM ME7838G broadband VNA equipped with 100 µm pitch MPI TITANTM T220A GSG0100 probes. The Anritsu system comprises a base microwave VNA integrated with millimeter-wave modules. Multiplexing of the sources and receivers provides continuous coverage from 70 kHz to 220 GHz. Up to 54 GHz, source radio frequency (RF) is provided directly from the base VNA. To generate the higher frequencies this base signal is multiplied up using an array of source multipliers in the modules. Module mixers are used for down-conversion above 30 GHz with the base VNA handling the lower frequencies. A more detailed description of this system can be found in [9]. The MPI probes are interfaced with the millimeter-wave modules via non-threaded 0.6 mm outer conductor diameter coaxial connectors. B. Banded Systems The coaxial (2.4 mm) system comprised of a Keysight PNA-X N5247B connected to 75 µm pitch MPI TITANTM T50A GSG75 probes using coaxial RF cables. For the three waveguide banded systems, the same VNA was used but with VDI Inc. frequency extenders to access frequency bands beyond the VNA’s standard range. GGB Industries Inc. Picoprobes were connected directly to the frequency extenders. For the WR10 and WR6.5 bands, 100 µm pitch probes (models 120-GSG-100-BT-M and 170-GSG-100-BT-M) were used. For the WR5 band, 75 µm pitch (model 220-GSG-75-BT-M) were used. The WR15 (50 GHz – 75 GHz) and WR12 (60 GHz – 90 GHz) systems at NPL were not included in the comparison, because, at the time this study was conducted, only 150 µm pitch probes were available, which are not compatible with the selected calibration substrate. C. Calibration Substrate The MPI TCS-050-100-W calibration substrate was used for calibration and test measurements. This substrate is 254 μm thick alumina and contains several groups of lumped standard elements, as well as five CPW transmission lines. In addition, it also contains 3 dB, 6 dB and 10 dB symmetric attenuators, and these were chosen as devices-under-test (DUTs) for comparison of the broadband and banded systems over a range of different transmission values. A pair of 50 Ω loads was selected as a fourth DUT to investigate measurements of a lowFig. 2. Measured transmission coefficient magnitude (S21) of nominally 3 dB, 6 dB and 10 dB attenuators using the broadband single-sweep system (orange) and banded systems: coaxial 2.4 mm (blue); WR10 (red); WR6.5 (purple); WR5 (green). Fig. 3. Difference in measured linear transmission coefficient (S21) magnitude using the banded systems (coaxial 2.4 mm in blue, WR10 in red, WR6.5 in purple and WR5 in green) with respect to broadband single-sweep system of nominally 3 dB, 6 dB and 10 dB attenuators. The “b” and “ss” stand for banded and single-sweep systems. (a) (b) (c) (d) Fig. 1. Microscope images of devices on MPI TCS-050-100-W calibration substrate selected as DUTs in comparison. (a) 3 dB attenuator; (b) 6 dB attenuator; (c) 10 dB attenuator; (d) 50 Ω loads. reflect device. Microscope images of the four DUTs are shown in Fig. 1. D. Calibration Method and Test Settings All acquired data was calibrated with the multiline Thru-Reflect-Line (mTRL) algorithm [10], which is widely regarded as one of the most accurate calibration techniques for on-wafer measurements. Thru and line standards used had effective lengths of 150 µm, 360 µm, 610 µm, 1175 µm, 2425 µm, and 5200 µm. The calibration reference planes were shifted to the probe tips. The reference impedance was left set to the characteristic impedance of the line standards, i.e., no renormalization to 50 Ω was performed. The exact same calibration standards as well as DUTs were measured by all systems and by the same operator to ensure fair comparison. The calibration substrate was placed on a ceramic chuck during measurements, which were performed with a 0.25 GHz frequency step, 100 Hz intermediate frequency bandwidth (IFBW), and no averaging. The power levels for the broadband single-sweep system were -12 dBm for frequencies up to 54 GHz and -18 dBm for the higher frequencies. The coaxial system used 0 dBm, while the VDI extenders’ default power levels were used for the waveguide banded systems. III. RESULTS AND DISCUSSIONS A. Measurement Results Figs. 2-4 compare the measured S-parameters of the 3 dB, 6 dB and 10 dB attenuators between the broadband singlesweep and banded systems. Fig. 2 shows the S21 magnitude in dB (S12 is similar to S21 and therefore, for clarity, not displayed), whilst Fig. 3 shows the differences in measured linear magnitude of S21 using the banded systems with respect to the broadband single-sweep system. Fig. 4 shows the magnitude of the reflection coefficients (in dB) with S11 plotted using a solid line and darker shade compared to the dashed line and lighter shade used to display S22. Fig. 5 compares the S11 and S22 responses of the pair of 50 Ω loads between the broadband single-sweep system and banded systems. To assess the probe contact repeatability of the measurement systems, 10 repeat landings per probe were made on a thru standard. Repeat landings involved re-alignment of the probe to the contact pads of the thru. During repeat landings of the probe connected to the VNA’s port 1, the probe connected to the VNA’s port 2 remained landed on the thru and vice versa. The standard deviations of the calibrated S11 and S22 from these repeat measurements are shown in Fig. 6. In Figs. 2-6, responses are shown starting from 2.5 GHz, despite the coaxial (2.4 mm) system and broadband single-sweep system being capable of covering frequencies down to 10 MHz and 70 kHz respectively. B. Discussions Overall, as shown in Figs. 2-5, all DUTs have demonstrated good agreement between the five measurement systems listed in Table 1. Particularly good agreement is seen between the broadband single-sweep and coaxial systems, while the waveguide banded systems are in good agreement with each other, but generally show larger differences when compared with the broadband system. The measured attenuations of the three attenuators (Fig. 2) show a relatively flat frequency response up to around 100 GHz, with attenuations larger than the 3 dB, 6 dB and 10 dB design values. At the higher frequencies, the attenuations Fig. 5. Measured reflection coefficient magnitude of 50 Ω loads using the broadband single-sweep system (orange) and banded systems: coaxial 2.4 mm (blue); WR10 (red); WR6.5 (purple); WR5 (green). S11 is denoted using a solid line and S22 with a dashed line and paler color. Fig. 4. Measured reflection coefficient magnitude of nominally 3 dB, 6 dB and 10 dB attenuators using the broadband single-sweep system (orange) and banded systems: coaxial 2.4 mm (blue); WR10 (red); WR6.5 (purple); WR5 (green). S11 is denoted using a solid line and S22 with a dashed line and paler color. generally decrease with frequency. Figs. 2-3 show good agreement between the waveguide banded systems at 110 GHz (i.e. the WR10 and WR6.5 band overlap) and between 140 GHz and 170 GHz (i.e. the WR6.5 and WR5 band overlap). More notable differences are seen when comparing results of these three banded systems to those of the broadband single-sweep system. These differences are most likely related to different probe geometries (GGB Picoprobes were used in the banded systems compared to MPI TITANTM probes for the singlesweep system). Other factors expected to have a small effect include probe contact repeatability (discussed later), probe coupling, differences in probe pitch size, wear to the contact pads on calibration standards/DUTs, differences in hardware (i.e. VNA and frequency extenders), and system drift. A check on system drift was performed by measuring the S-parameters of a thru both before and after the comparison measurements. For all systems, the maximum differences in S21/S12 between “before” and “after” measurements were within 0.06 dB. The reflection responses of the three attenuators (Fig. 4) and the pair of 50 Ω loads (Fig. 5) show the broadband single-sweep system in close agreement with the banded systems up to 50 GHz (i.e., with coaxial 2.4 mm) and above 140 GHz (i.e., with WR6.5 and the higher frequencies of WR5). Slightly poorer agreement is observed with the WR10 measurements. The relatively higher values of S11 and S22 seen at the lowest frequencies in Fig. 5 is likely due to the reference impedance deviating from 50 Ω (no re-normalization to 50 Ω was performed). Probe contact repeatability (see Fig. 6) was demonstrated to be reasonably good, with all systems showing standard deviations of the linear magnitude of S11 and S22 from repeated probe landings on a thru as less than 9 mU. These findings suggest that probe contact repeatability is a relatively small contributor to the differences between the systems seen in Figs. 2-5. Differences in probe contact repeatability between setups could be attributed to (1) frequency range (with measurements generally becoming less repeatable with increasing frequency), (2) probe design (including ease of realignment for the operator), and (3) probe condition (e.g., the probe connected to VNA port 2 in the WR10 system was known to be in relatively poor condition). IV. CONCLUSION This paper has presented a comparison of on-wafer S-parameter measurements between broadband single-sweep and conventional banded systems up to 220 GHz. Relatively good agreement between the different systems was achieved. Deviations between the results from different setups was discussed in detail. The information presented could assist endusers understand better the level of equivalence which can be expected between measurements from broadband single-sweep and banded systems. ACKNOWLEDGEMENT This work was supported, in part by the European Partnership on Metrology (EPM) project 23IND10 OnMicro, and in part by the National Measurement System Programme of the UK government’s Department for Science, Innovation and Technology (DSIT). The project (23IND10 OnMicro) has received funding from the European Partnership on Metrology, co-financed from the European Union’s Horizon Europe Research and Innovation Programme and by the Participating States. REFERENCES [1] H. Morales, L. Dunleavy, and S. Skidmore, “Ultra-broadband characterization of Schottky diodes,” 79th ARFTG Microw. Meas. Conf. Dig., pp. 1-4, June 2012, doi: 10.1109/ARFTG79.2012.6291198. [2] T. Li, L. Li, and J. C. M. Hwang, “Validity of room-temperature calibration for on-wafer measurements up to 220 GHz, 125 °C, and 48 h,” 101st ARFTG Microw. Meas. Conf. Dig., pp. 1-4, June 2023, doi: 10.1109/ARFTG57476.2023.10278622. [3] A. Rumiantsev, J. Martens, and S. Reyes, “Calibration, repeatability and related characteristics of on-wafer, broadband 70 kHz–220 GHz single-sweep measurements,” 95th ARFTG Microw. Meas. Conf. Dig., pp. 1-4, June 2020, doi: 10.1109/ARFTG47271.2020.9241377. [4] A. Rumiantsev, J. Martens, and S. Reyes, “A differential broadband single-sweep 70 kHz-220 GHz wafer-level system: first calibration and measurement characteristics,” 100th ARFTG Microw. Meas. Conf. Dig., pp. 1-4, January 2023, doi: 10.1109/ARFTG56062.2023.10148890. [5] A. Rumiantsev, R. Doerner, J. Martens, and S. Reyes, “Interlaboratory comparison of on-wafer broadband 70kHz-220GHz single-sweep measurements,” 51st Eur. Microw. Conf. Dig., pp. 425-428, April 2021, doi: 10.23919/EuMC50147.2022.9784277. [6] X. Shang, N. Ridler, U. Arz, G. N. Phung, I. Roch-Jeune, G. Ducournau, K. Haddadi, T. Flisgen, R. Doerner, D. Allal, D. Jayasankar, J. Stake, R. Schmidt, G. Fisher, and F. Mubarak, “Interlaboratory investigation of on-wafer S-parameter Fig. 6. Standard deviations of linear reflection coefficient magnitudes from 10 repeat landings on thru. Calibration was applied prior to calculation of standard deviation. Results are from the broadband single-sweep system (orange) and banded systems: coaxial 2.4 mm (blue); WR10 (red); WR6.5 (purple); WR5 (green). S11 is denoted using a solid line and S22 with a dashed line and paler color. measurements from 110 GHz to 1.1 THz,” 53rd Eur. Microw. Conf. Dig., pp. 624-627, September 2023 doi: 10.23919/EuMC58039.2023.10290550. [7] F. Mubarak, G. N. Phung, U. Arz, K. Haddadi, I. Roch-Jeune, G. Ducournau, T. Flisgen, R. Doerner, D. Allal, D. Jayasankar, J. Stake, R. Schmidt, G. Fisher, N. Ridler, and X. Shang, “An interlaboratory comparison of on-wafer S-parameter measurements up to 1.1 THz,” IEEE Trans. Terahertz Sci. Technol., doi: 10.1109/TTHZ.2025.3537461. [8] L. Li, S. Reyes, M. J. Asadi, D. Jena, H. Xing, P. Fay, and J. C. M. Hwang, “Single-sweep vs. banded characterizations of a Dband ultra-low-loss SiC substrate-integrated waveguide,” 99th ARFTG Microw. Meas. Conf. Dig., pp. 1-4, June 2022, doi: 10.1109/ARFTG54656.2022.9896408. [9] J. Martens, and T. Roberts, “Broadband 220 GHz network analysis: structures and performance,” 94th ARFTG Microw. Meas. Conf. Dig., pp. 1-5, January 2020, doi: 10.1109/ARFTG47584.2020.9071774. [10] R. B. Marks, “A multiline method of network analyzer calibration,” IEEE Trans. Microw. Theory Techn., vol. 39, no. 7, pp. 1205-1215, July 1991, doi: 10.1109/22.85388.