Enhancing the Accuracy of Microwave Element Models by Artificial Neural Networks

Abstract

In the recent PSpice programs, five types of the GaAs FET model have been implemented. However, some of them are too sophisticated and therefore very difficult to measure and identify afterwards, especially the realistic model of Parker and Skellern. In the paper, simple enhancements of one of the classical models are proposed first. The resulting modification is usable for the accurate modeling of both GaAs FETs and pHEMTs. Moreover, its updated capacitance function can serve as an accurate representation of microwave varactors, which is also important. The precision of the updated models can be strongly enhanced using the artificial neural networks. In the paper, both using an exclusive neural network without an analytic model and cooperating a corrective neural network with the updated analytic model will be discussed. The accuracy of the analytic models, the models based on the exclusive neural network, and the models created as a combination of the updated analytic model and the corrective neural network will be compared.