Polynomial-Assisted Neural Network Behavioral Model of Wideband Radio Frequency Power Amplifiers for 5G New Radio

In this paper, a novel polynomial-assisted neural network is proposed for behavioral modeling of wideband radio frequency power amplifiers for 5G new radio. It is achieved by embedding the polynomial basis function of conventional models into the neural network structure to reduce the computation complexity and improve the convergence of the neural network. Experimental results employing a wideband GaN PA using 5G new radio signal shows that the proposed model that integrates memory polynomial (MP) with vector decomposition time delay neural network (VDTDNN) can achieve either better performance or lower complexity compared with the conventional models.