How good is your calibration? A post-mortem examination and recalibration

Summary form only given. It is pointed out that the standard network analyzer calibration procedure suffers from several drawbacks. The test patterns, for instance, often reside on a wafer different from that being tested. The normal procedure is to measure only a single set of calibration patterns instead of collecting a statistically significant sample spanning the wafer. Wafer thickness uniformity, for example, affects the coplanar to microstrip launch discontinuity, which impacts the quality of the calibration. The parasitic probe-to-probe capacitances also differ between the calibration and test setups. Rather than propose solutions to these and other calibration issues, the authors examine the statistical quality of measurements after calibration and test. Reciprocity and symmetry are examined for several hundred passive components from a special test wafer. The frequency-dependent deviations from perfect reciprocity and symmetry are used to evaluate the quality of the original calibration. It is shown how this information can be used to recenter the calibration and quantify the intrinsic frequency degradation