An artificial neural network based nonlinear behavioral model for RF power transistors

In this paper, a frequency domain, nonlinear, behavioral model for RF power transistors, based on an artificial neural network (ANN), is proposed and validated. The model is identified using the back-propagation algorithm from the incident and scattered wave data of the RF transistor. The model has been extracted and validated on Cree GaN HEMT device. Both simulation and measurement examples are presented. Compared with existing nonlinear transistor behavioral models, which are all input-power dependent, the new model is able to effectively predict the behavior of a transistor over the entire Smith chart, at different levels of input power with only a single set of model coefficients.