Behavioral modeling and digital predistortion linearization for wideband RF power amplifiers with neural networks

Abstract

In this talk, we will focus on discussing the capability of a few neural networks for modeling and linearizing different RF power amplifiers. The AM/AM and AM/PM characteristics and spectrum comparison will be utilized to evaluate the performance of different neural networks in modeling and preditortion linearization. At first, a few BP based feedforward neural networks will be used to simulate the dynamic nonlinearity of RF power amplifiers. Then they will be utilized to linearize the power amplifiers. After that, a few RBF neural networks will be applied to construct the behavioral model and digital predistortion linearizers for the wideband RF power amplifiers. And furthermore, a hybrid-structure neural network is presented to mimic the dynamic nonlinear properties of a broadband RF power amplifiers. Finally, the future trend for modeling and linearization of RF power amplifiers with neural networks will be discussed.