Dynamically configurable pHEMT model using neural networks for CAD

A novel approach to developing CAD microwave device models is presented. Traditional CAD devices are implemented using static empirical equations to describe electrical behavior. Recently, neural networks have been used in place of empirical equations to model device behavior. This paper describes the implementation of a CAD device model that utilizes a dynamically configurable combination of empirical equations and neural networks to increase the flexibility of the model's capabilities. The model was developed for pHEMT devices but can be customized to work with other device structures such as HBTs. The framework for this model is a common-source large-signal equivalent FET circuit. With the exception of the drain current source, all of the nonlinear elements of the circuit are configurable to either empirical or bias-dependent neural network controlled components. The neural network architecture employed is based on the knowledge-based algorithm.