Improved Accuracy of On-Wafer Measurements Using the MMAVERIC Calibration Technique

Abstract

Besides calculating the usual coefficients of a 3-term (one-port) or 12-term (two-port) error model, the MMAVERIC technique provides 1) a corruption factor, 2) the phase and loss of the transmission line in which the calibration is performed, and 3) the reflection coefficient of one of the terminations used in the procedure. Each of these outputs is a valuable tool for analyzing and minimizing residual calibration errors. The corruption factor is a sensitive measure of the aggregate contributions of all departures from the ideal situation. For example, if one of the calibration standards is a short, we can determine how closely it approaches the ideal magnitude, |Γ|= 1, and phase, argΓ = 180/°. Our experimental evidence has shown that, at 40 GHz, |Γ| for a coplanar short directly contacted by the probe is about 0.7 dB smaller than that for a short seen through a length of coplanar line on the substrate. Thus, lower corruption and residual calibration errors are obtained if the reference plane is defined on the substrate at some distance away from the probe tips. We have corroborated the work of others(4) by modeling the short as a small reactance and shown how further improvements are achieved by separating the substrate from the wafer chuck by a low-dielectric spacer.