Development of a Neural Approach for Bias-Dependent Scalable Small-Signal Equivalent Circuit Modeling of GaAs HEMTs

This paper presents an approach for small-signal modeling of microwave FETs. The model is based on an equivalent circuit whose elements are extracted by an analytical approach. In order to make model bias-dependent and scalable, artificial neural networks are exploited for modeling of the dependence of the equivalent circuit elements on the bias voltages and the transistor gate width. The proposed approach is exemplified on modeling of three scaled on-wafer GaAs HEMT devices.