Investigation of Probe Pitch Influence on On-Wafer Multiline TRL Calibrations up to 110 GHz

Investigation of Probe Pitch Influence on On-Wafer Multiline TRL Calibrations up to 110 GHz Gia Ngoc Phung and Uwe Arz Physikalisch-Technische Bundesanstalt (PTB), Bundesallee 100, 38116 Braunschweig, Germany Abstract—Recently, much progress has been made in the traceability of on-wafer measurements of planar devices on coplanar calibration substrates. However, reliable uncertainties for onwafer S-parameters can only be given for a specific combination of substrate material, planar transmission line and probes, and only if single-mode propagation is ensured. This condition limits the use of uncertainties for probes with different dimensions. Therefore, this paper presents a systematic investigation of the influence of probe pitches on selected devices under test. The effects of probe pitch in conjunction with the influence of neighbourhood effects are investigated via simulations up to 110 GHz and compared with measurement results of example DUTs with expanded uncertainties at a coverage probability of 95 % (k=2). Index Terms—coplanar waveguides, on-wafer measurements, uncertainties. I. INTRODUCTION On-wafer measurements are essential for many end-user applications in communications and electronics technologies such as sixth generation (6G) mobile networks, the Internet of Things (IoT), automotive industry and more. Any unpackaged RF circuit can be measured directly on the wafer for quality assurance or during product development as feedback to the design process. However, on-wafer measurements are known to be challenging, containing many parasitics ([1] and [2]) that make it difficult to achieve traceability for on-wafer S-parameter measurements. So far, traceability has been established in the European PlanarCal project [3] for on-wafer measurements of coplanar waveguides in membrane technology [4] and fused silica [5], and of coplanar waveguides embedded in commercially available alumina substrates [6]. More recently, an uncertainty study of a shunt capacitance in the presence of probe positioning uncertainty, calibration kit process variation, and vector network analyzer electrical repeatability up to 1 THz has been presented in [7]. In [5] it was found that reliable on-wafer S-parameter uncertainties can only be stated for a specific combination of substrate material, planar transmission lines and microwave probes, and only if singlemode propagation is ensured. This requirement restricts the usability of the estimated uncertainties. One of the main influences on on-wafer measurements is probe effects, due to the large number of probes on the market with different probe geometries from different manufacturers, all of which affect the measurement differently [8]. The question is how much difference there is up to 110 GHz between probes from the same manufacturer and the same product line, but with different probe pitches. With this open 0201 open 0202 open 0203 open 0206 open 0207 open 0208 open 0209 open 0606 open 0607 open 0608 open 0609 Fig. 1. Mockup of the calibration substrate TCS-050-100W in CST Studio Suite. CST model pitch = 50 µm measurement pitch = 50 µm CST model pitch = 75 µm CST model pitch = 100 µm Fig. 2. Titan probes model in CST Studio Suite motivation, this paper investigates the influence of probe pitch in conjunction with neighborhood effects. II. INFLUENCE OF PROBE PITCH VARIATION Figure 1 shows the mockup of the coplanar calibration substrate TCS-050-100W in CST Studio Suite [9] used in this study. The design and arrangement of the calibration standards on this calibration substrate [10] are optimized to minimize parasitic effects according to the design rules[11]. Neighborhood and probe effects have been reduced. © 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/ARFTG63706.2025.10989799 0 20 40 60 80 100 Frequency (GHz) 0 0.1 0.2 0.3 0.4 (dB/mm) Titan probes pitch = 50 µm Titan probes pitch = 75 µm Titan probes pitch = 100 µm 0 20 40 60 80 100 120 Frequency (GHz) 2.2 2.3 2.4 2.5 / 0 Fig. 3. Calibrated simulated propagation constants with varied Titan probe pitches: attenuation constants and normalized phase constant. A thru, several line standards of different length and a short as reflect standard of the MPI calibration substrate of Figure 1 are used to perform a multiline Thru Reflect Line (mTRL) reference calibration according to [12]. The investigated calibration substrate consists of an alumina substrate with a thickness of 254 µm. In the measurements, the calibration substrate is placed on a ceramic chuck with dielectric constant of 10.2 to avoid the propagation of the surface waves. To reproduce a realistic measurement scenario, the entire wafer placed on a ceramic chuck of 10.2 and the Titan probes are modeled in CST Studio Suite according to [13] and [10]. For the excitation a waveguide port is placed at the coaxial line and the wave in the coaxial waveguide is converted into a CPW mode between the needles, thus feeding the DUT. For the systematic study, the pitches of the Titan probes are varied between 50, 75 and 100 µm as shown in Figure 2. After simulating all the calibration elements, the multiline TRL calibration process was applied to the simulated data for the 50, 75 and 100 µm probe pitches. The first results of the mTRL calibration are the propagation constants shown in Figure 3. Comparing the attenuation constants of the 50, 75 and 100 µm probe pitches, the results are almost identical. Comparing the normalized phase constant however, there are slight differences in the simulated results for 75 µm. On the other hand, the results for 50 and 100 µm lie on top of each other. The question is how these admittedly relatively small differences between the three cases affect the calibration of a DUT. Two DUTs, an open structure and a 3 dB attenuator (Figure 4), located multiple times on the calibration substrate, are selected for investigation using the corresponding multiline TRL calibration. Figure 5 shows the results of the open 0201 located in the first row and second column of the calibration substrate (Figure 1) excited with Titan probes of different pitches. Comparing the scattering parameters, there are only small differences of less than 0.2 dB for the reflection coefficients |S11|and |S22|. There are no significant differences Open Att 3 dB Fig. 4. Selected DUTs for comparison: open and 3 dB attenuator 0 20 40 60 80 100 Frequency (GHz) -0.8 -0.6 -0.4 -0.2 0 |S11| (dB) 0 20 40 60 80 100 Frequency (GHz) -80 -60 -40 -20 |S21| (dB) 0 20 40 60 80 100 Frequency (GHz) -80 -60 -40 -20 |S12| (dB) 0 20 40 60 80 100 Frequency (GHz) -0.8 -0.6 -0.4 -0.2 0 |S22| (dB) Titan probes pitch = 50 µm Titan probes pitch = 75 µm Titan probes pitch = 100 µm Fig. 5. EM simulation of an open 0201 with varied Titan probes pitches: Magnitude of reflection coefficient |S11|and |S22|and of transmission coefficient |S21|,|S12|. in the transmission coefficients |S12|and |S21|. As expected, there is little difference in the reflection and transmission behavior due to the different probe pitches. The level of probe crosstalk remains the same. This proves that the influence of the pitch of the probe is negligibly small. The next question is what happens if the same DUT is placed in different positions on the calibration substrate and whether the probes would radiate differently due to the different positions. For this reason all the open structures denoted as open 0201, 0202, 0203 ... 0609 are examined. They are located at different positions as shown in Figure 1. Figures 6 – 8 show the corresponding results of the open structures for the 50, 75 and 100 µm probe pitches. Comparing the results, there are only small differences for the reflection coefficients |S11|and |S22|in the order of ∆|Sii|= +/− 0.004 and the transmission coefficients |S12|and |S21|in the order of ∆|Sij|= +/−0.005 . This indicates that the effect of neighbouring effects is negligibly small for all three Titan probe pitches when varying the position of the DUT. 0 20 40 60 80 100 Frequency (GHz) -5 0 5 10 |S11|-|S11,med| 10-3 Open 0201 Open 0202 Open 0203 Open 0206 Open 0207 Open 0208 Open 0209 Open 0606 Open 0607 Open 0608 Open 0609 median 0 20 40 60 80 100 Frequency (GHz) -5 0 5 |S21|-|S21,med| 10-4 0 20 40 60 80 100 Frequency (GHz) -4 -2 0 2 |S12|-|S12,med| 10-4 0 20 40 60 80 100 Frequency (GHz) -4 -2 0 2 4 |S22|-|S22,med| 10-3 Fig. 6. EM simulation of differently located opens with Titan probes pitch of 50 µm: Difference plot to the median value of reflection coefficient |S11 | and |S22|and of transmission coefficient |S21|,|S12|. 0 20 40 60 80 100 Frequency (GHz) -5 0 5 10 |S11|-|S11,med| 10-3 Open 0201 Open 0202 Open 0203 Open 0206 Open 0207 Open 0208 Open 0209 Open 0606 Open 0607 Open 0608 Open 0609 median 0 20 40 60 80 100 Frequency (GHz) -2 -1 0 1 2 |S21|-|S21,med| 10-4 0 20 40 60 80 100 Frequency (GHz) -2 -1 0 1 2 |S12|-|S12,med| 10-4 0 20 40 60 80 100 Frequency (GHz) -2 -1 0 1 2 |S22|-|S22,med| 10-3 Fig. 7. EM simulation of differently located opens with Titan probes pitch of 75 µm: Difference plot to the median value of reflection coefficient |S11 | and |S22|and of transmission coefficient |S21|,|S12|. III. MEASUREMENT RESULTS To complement and verify the results of Fig. 6, we performed measurements with Titan probes with a probe pitch of 50 µm. The calibration standards and the DUTs are measured and the mTRL calibration is applied to the measured data. Figure 9 shows the comparison between the simulated and measured data with Titan probes with a probe pitch of 50 µm for the opens. Overall, the agreement for both quantities, reflection coefficient |S11|and |S22|and transmission coefficient |S21|,|S12|is good. There are only small deviations, but these are due to the lack of knowledge of the exact material properties and the exact probe geometries. Both simulation and measurement results prove that the calibration substrate does indeed exhibit negligible calibration errors as targeted by the optimization. So far, we have compared the 0 20 40 60 80 100 Frequency (GHz) -5 0 5 10 |S11|-|S11,med| 10-3 Open 0201 Open 0202 Open 0203 Open 0206 Open 0207 Open 0208 Open 0209 Open 0606 Open 0607 Open 0608 Open 0609 median 0 20 40 60 80 100 Frequency (GHz) -4 -2 0 2 |S21|-|S21,med| 10-4 0 20 40 60 80 100 Frequency (GHz) -4 -2 0 2 |S12|-|S12,med| 10-4 0 20 40 60 80 100 Frequency (GHz) -2 -1 0 1 2 |S22|-|S22,med| 10-3 Fig. 8. EM simulation of differently located opens with Titan probes pitch of 100 µm: Difference plot to the median value of reflection coefficient |S11 | and |S22|and of transmission coefficient |S21|,|S12|. 0 20 40 60 80 100 Frequency (GHz) -1.5 -1 -0.5 0 |S11| (dB) Open 0201 simulated Open 0202 simulated Open 0203 simulated Open 0206 simulated Open 0207 simulated Open 0208 simulated Open 0209 simulated Open 0201 measured Open 0202 measured Open 0203 measured Open 0206 measured Open 0207 measured Open 0208 measured Open 0209 measured 0 20 40 60 80 100 Frequency (GHz) -80 -60 -40 -20 |S21| (dB) 0 20 40 60 80 100 Frequency (GHz) -80 -60 -40 -20 |S12| (dB) 0 20 40 60 80 100 Frequency (GHz) -0.4 -0.3 -0.2 -0.1 0 0.1 |S22| (dB) Fig. 9. EM simulation and measurement of differently located open with Titan probes pitch of 50 µm: Magnitude of reflection coefficient |S11|and |S22|and of transmission coefficient |S21|,|S12|. measurement results for the 50 µm with the em simulation. To analyze the influence of the other Titan probe pitches, we estimated a comprehensive uncertainty budget for on-wafer S-parameter measurements of the devices with 50 um pitch probes on this calibration substrate, including instrumentation errors, connector repeatability, and calibration standard uncertainties. The uncertainty analysis was performed based on the method in [5]. For comparison, we use the open and the attenuator shown in Figure 4. Figures 10 and 11 show the error-corrected reflection and transmission S-parameter measurements including expanded uncertainties at a 95 % coverage probability (k=2) of the open and 3 dB attenuator in comparison with the EM simulation. The simulated results for both DUTs with different Titan probe pitches are within the uncertainties of the measured results. In summary, for the DUTs investigated on this commercial Fig. 10. Error-corrected reflection and transmission S-parameter measurements including expanded uncertainties at a coverage probability of 95 % (k=2) for the 50 µm pitch Titan probes of the open measurement vs. simulation results of the differently located open with different Titan probe pitches. Fig. 11. Error-corrected reflection and transmission S-parameter measurements including expanded uncertainties at a coverage probability of 95 % (k=2) for the 50 µm pitch Titan probes of the 3 dB attenuator measurement vs. simulation results of the differently located 3 dB attenuator with different Titan probe pitches. calibration substrate up to 110 GHz, the influence of probe pitches seems to be relatively small. For higher frequency measurements excited by other types of probes on calibration substrates with smaller element spacings, the impact of the probe pitch may be larger and not negligible. Future work will include final validation and measurements with different probe pitches and from different manufacturers. IV. CONCLUSION In this paper, we have investigated the influence of probe pitch in conjunction with neighborhood effects using a commercial calibration substrate with minimized parasitic effects. In conclusion, the comparison of the expanded uncertainty intervals of 50 um probe pitch measurements with the em simulations of the different probe pitches shows that these differences are small, demonstrating that the variations seen in em simulations of the different probe pitches are small compared to other uncertainty components of the measurement uncertainty budget. ACKNOWLEDGMENT This work was supported by the European Partnership on Metrology (EPM) 23IND10 OnMicro project which 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. The authors are also grateful to Dr. Andrej Rumiantsev for providing us the CST model of the Titan probes. The authors gratefully acknowledge the computing resources granted by PTB Sec. Q.45 High Performance Computing (HPC). REFERENCES [1] N. Mahjabeen, Y. Zhang, A. M. Dave, J. Um, A. Harpel, B. Stadler, R. Franklin, and R. Henderson, “Limitations and Importance of EM Models for On-wafer High Frequency Performance Evaluation,” in 2022 IEEE/MTT-S International Microwave Symposium - IMS 2022, 2022, pp. 475–478. [2] C. Yadav, M. Deng, S. Fregonese, M. Cabbia, M. De Matos, B. Plano, and T. Zimmer, “Importance and Requirement of Frequency Band Specific RF Probes EM Models in Sub-THz and THz Measurements up to 500 GHz,” IEEE Transactions on Terahertz Science and Technology, vol. 10, no. 5, pp. 558–563, 2020. [3] European Metrology Programme for Innovation and Research JRP Number 14IND02, “Microwave Measurements for Planar Circuits and Components.” https://planarcal.ptb.de. [4] U. Arz, S. Zinal, T. Probst, G. Hechtfischer, F.-J. Schm¨ uckle, and W. Heinrich, “Establishing Traceability for On-wafer S-parameter Measurements of Membrane Technology Devices up to 110 GHz,” in 2017 90th ARFTG Microwave Measurement Symposium (ARFTG), 2017, pp. 1–4. [5] U. Arz, K. Kuhlmann, T. Dziomba, G. Hechtfischer, G. N. Phung, F. J. Schm¨ uckle, and W. Heinrich, “Traceable Coplanar Waveguide Calibrations on Fused Silica Substrates up to 110 GHz,” IEEE Transactions on Microwave Theory and Techniques, vol. 67, no. 6, pp. 2423–2432, 2019. [6] U. Arz, G. N. Phung, and A. Rumiantsev, “Traceable Lumped-Element Calibrations up to 110 GHz on Commercial Calibration Substrates,” in 2023 100th ARFTG Microwave Measurement Conference (ARFTG), 2023, pp. 1–4. [7] R. D. Jones, J. Cheron, B. F. Jamroz, W. R. Deal, M. Urteaga, D. F. Williams, A. D. Feldman, P. H. Aaen, C. J. Long, and N. D. Orloff, “OnWafer Capacitor Characterization Including Uncertainty Estimates Up to 1.0 THz,” IEEE Transactions on Terahertz Science and Technology, vol. 14, no. 5, pp. 734–744, 2024. [8] G. N. Phung, F. J. Schm¨ uckle, R. Doerner, B. K¨ ahne, T. Fritzsch, U. Arz, and W. Heinrich, “Influence of Microwave Probes on Calibrated OnWafer Measurements,” IEEE Transactions on Microwave Theory and Techniques, vol. 67, no. 5, pp. 1892–1900, 2019. [9] “CST Studio Suite,” 2023. [Online]. Available: https://www.3ds.com/ products-services/simulia/products/cst-studio-suite/ [10] A. Rumiantsev, R. Doerner, and G. N. Phung, “Calibration Substrate Design for Accurate mm-Wave Probe-Tip Calibration,” in 2020 94th ARFTG Microwave Measurement Symposium (ARFTG), 2020, pp. 1–4. [11] M. Spirito, U. Arz, G. N. Phung, F. J. Schm¨ uckle, W. Heinrich, and R. Lozar, “Guidelines for the Design of Calibration Substrates, including the Suppression of Parasitic Modes for Frequencies up to and including 325 GHz,” EMPIR 14IND02 – PlanarCal, 2018, PhysikalischTechnische Bundesanstalt (PTB), 2018. [12] R. B. Marks, “A Multiline Method of Network Analyzer Calibration,” IEEE Trans. on Microwave Theory and Techniques, vol. 39, no. 7, pp. 1205–1215, 1991. [13] G. N. Phung, U. Arz, and W. Heinrich, “Precise Modeling of Coplanar Device Measurements Under Realistic Conditions up to G-Band,” IEEE Transactions on Components, Packaging and Manufacturing Technology, vol. 13, no. 10, pp. 1584–1590, 2023.