Digital Predistortion of RF Power Amplifiers With Phase-Gated Recurrent Neural Networks
Tugce Kobal, Yue Li, Xiaoyu Wang, Anding Zhu
Abstract
In this article, we present a novel recurrent neural network (RNN)-based behavioral model to linearize radio frequency (RF) power amplifiers (PAs) under wideband excitations. Based on the lightweight Just Another NETwork (JANET) unit, we propose a new neural network structure that is especially suitable for modeling the complex behavior of the RF PAs. A novel signal preprocessing technique is developed to model the complex interaction between amplitude information and phase information in the digital predistortion (DPD) model. By integrating the preprocessing stage into the RNN model, the complete phase-gated JANET (PG-JANET) model can provide enhanced linearization performance with lower complexity compared to the existing models.