Efficient FET model parameter extraction using multi-plane data-fitting and bidirectional search technique

A novel FET small-signal model parameter extraction technique is presented. It uses a novel optimization approach, in which data fitting is carried out on two reference planes instead of only one, and the objective function, is minimized by a bidirectional search. As a result, all parameters of a commonly used 15-element small-signal FET equivalent circuit are clearly identified from only one set of measured S-parameters. Starting values are self-consistently and simply generated from measured S-parameters in the passive pinch-off operating mode. Moreover, by applying multibias data fitting, which is performed without increasing the number of ordinary optimization variables, both the overall bias-independent parasitic and bias-dependent intrinsic elements are robustly determined. Demonstration results are presented for a 0.5- mu m MESFET.<
>