Advanced Simulation of Semiconductor Devices by Artificial Neural Networks

Abstract

Characterization and modelling of transistors for advanced microwave electronics is very useful to understand and improve device performances.

Here, a modelling procedure of HEMT transistors for microwave applications is described. There are many different approaches to this problem, such as direct extraction methods, global or decomposed fitting techniques and inverse modeling issues. In this paper, we use an Artificial Neural Network (ANN) to extract a 17-element small signal circuit model. By this procedure we are able to identify the model from one measured [S] parameters set. Several circuit model values have been extracted by varying the temperature and the bias operating conditions. To our knowledge, these are the first results of a complete equivalent circuit extracted by an ANN technique without the need for any final tuning of the variables.