Behavioral modeling of power amplifier with long term memory effects using recurrent neural networks

Abstract

This paper describes recurrent neural network (RNN) technique for behavioral modeling of power amplifier (PA) with short and long term memory effects. RNN can be trained directly using the input-output data without the internal details of the circuit and the trained models can reflect the behavior of nonlinear circuit. Additional signals representing slow memory effects are extracted from the PA input and output signals and are used as extra inputs to RNN model in order to effectively represent long term memory. Examples of RNN modeling of power amplifier with short and long term memory effects are presented.