Behavioral modeling of RF power amplifiers with memory effects using orthonormal hermite polynomial basis neural network

Abstract

Behavioral modeling technique provides an efficient and convenient way to analyze and predict the performance of the RF power amplifiers (PAs) in system-level, and thus helps to construct a suitable predistorter to linearize the PA system. To accurately describe the nonlinear dynamic characteristics of PAs, an orthonormal Hermite polynomial basis neural network (OHPBNN) is utilized to represent the PA behavioral model, which outperforms, mainly in respect of modeling accuracy, the classic feedforward neural network using sigmoid activation functions. In addition, we apply an adaptive algorithm to determine the appropriate memory depth of PA behavioral model. Simulation results show that the proposed model provides more accurate prediction of the PAs output signal compared with classic neural network models.