Observations on the sensitivity of on-wafer cascode cell S-parameter measurements due to probing uncertainties

On-wafer probing accuracy limitations affect measurement repeatability for potentially unstable circuits. Here we present a 2×25 μm pHEMT cascode cell on 2 mil GaAs substrate measured from 0.045-110 GHz. The S-parameter measurements carried out for the same device using two different on-wafer measurement systems, one PNA based (Sys-1) and the other a 8510XF VNA (Sys-2), differed by up to 2 dB in gain above 50 GHz. Measurements were performed using a Cascade Microtech probe station after applying on-wafer LRRM calibration on WINCAL XE using the same set of standards. Sys-1 used Cascade infinity probes; Sys-2 Cascade ACP probes. A maximum difference in measured gain of 1.1 dB at 85 GHz was observed with variation in probe position. As |S22| for the cascode configuration is > 1, measured gain is highly sensitive to the output termination. The extra pad metallization due to differing probe position was modeled as a negative shunt capacitance (-5 fF), effectively de-embedding the extra pad capacitance.