STATISTICAL-ANALYSIS OF GAAS-MESFET S-PARAMETER EQUIVALENT-CIRCUIT MODELS

We compare the ability of three different equivalent-circuit extraction methods to give ensembles of model parameters that accurately predict not only average S-parameters but the S-parameter statistics, i.e., the standard deviations and intercorrelations between the real and imaginary parts. Measurements were made for 400 GaAs MESFETs fabricated on a single wafer with an MBE-grown active layer. Data is compared for different biases. We find that bimodal distributions give correlations that the equivalent-circuit models fail to model. The possibility of using uncorrelated equivalent-circuit values is also discussed.