Unlocking next-Gen InP HBTs performance: empirical validation of 16-term calibration in bias-dependent RF characterization
Fig. 1. Relative standard deviation of the extracted fmax values using 8-term calibration and 16-term calibration for InP HBTs with emitter widths of 850 nm and 500 nm against measured bias conditions.
As next-generation RF transistors push toward unprecedented operational speeds, conventional characterization methods are reaching their limits. The maximum frequency of oscillation (fmax), obtained by extrapolating data from low-frequency Mason’s gain (U) measurements, remains the key benchmark for transistor performance. However, fmax extracted using conventional 8-term calibrated data is no longer reliable for ultra-scaled, high-speed devices. On-wafer setups introduce severe systematic errors that distort data integrity and compromise development cycles. To address this challenge, we have established a 16-term calibration solution that eliminates these measurement vulnerabilities and enables more reliable high-frequency semiconductor characterization.
We investigated indium phosphide (InP) heterojunction bipolar transistors (HBTs) fabricated using FBH’s TRM process technology. Devices with emitter widths of 850 nm and 500 nm were characterized using both a conventional calibration procedure (i.e., 8‑term mTRL calibration) and an advanced 16‑term calibration. The conventional approach does not adequately compensate for on‑wafer parasitic effects. As a result, extraction of fmax by extrapolating Mason’s gain can lead to substantial under- or over-estimation of the intrinsic transistor performance [1].
In contrast, the improved calibration approach provides a consistent estimate of fmax across a wide range of measurement frequencies, bias conditions, and technology nodes. The values of fmax were extracted by averaging the data for frequencies above 30 GHz. For devices with an emitter width of 850 nm, both calibration techniques yield comparable mean fmax values of 306 GHz and 310 GHz, measured at VCE = 1.5 V and IB = 0.8 mA. For down‑scaled devices with an emitter width of 500 nm, measured under the same bias conditions, the standard method yields a mean fmax value of 410 GHz with a standard deviation of 50 GHz. Application of the advanced 16‑term method further reduces the uncertainty in the fmax extraction, resulting in a mean value of 433 GHz with a standard deviation of 16 GHz. Fig. 1 shows that the relative standard deviation of the extracted fmax remains below 4 % for both devices when the advanced calibration is employed. The relative standard deviation increases to as much as 12 % for the smaller‑node devices when the conventional calibration is used. The statistical analysis underscores the critical importance of 16‑term calibration for accurate high‑frequency characterization of transistors with aggressively scaled emitter widths [2].
Our current 23IND10 OnMicro project aims to extend these investigations by conducting inter-laboratory measurements and further access the validity of the improved calibration at even higher frequencies. This work will contribute to establishing reliable and accurate on-wafer calibration and measurement methodologies for sub-THz device characterization.
This work was supported in part by the European Partnership on Metrology (EPM) project under the grant 23IND10 OnMicro. The 23IND10 OnMicro project is co-financed by European Union’s Horizon European Research and Innovation Program and by Participating States.
Publications
[1] A. Kanitkar, R. Doerner, T. K. Johansen, W. Heinrich, T. Flisgen, “On-Wafer 16-Term Calibration for Characterization of InP HBTs Featuring Sub-THz fmax,” 55th European Microwave Conference (EuMC), Utrecht, Netherlands, 2025.
[2] A. Kanitkar, R. Doerner, T. K. Johansen, W. Heinrich, T. Flisgen, “On-Wafer 16-Term Calibration Approach for Dual-Setup, Bias-Dependent RF Characterization of Scaled InP HBTs,” International Journal of Microwave and Wireless Technologies (accepted for publication).