A New Frontier for Power Amplifier enabled by Machine Learning


This article focuses on the recent studies of introducing machine learning techniques for radio frequency power amplifiers online operational conditions optimization, primarily at sub6GHz frequency of 5G. We report two demonstrators of advanced power amplifier architecture designed with cutting edge 0.15um-gallium nitride (GaN) HEMT technology, namely: a digital Doherty power amplifier (DDPA) and a novel digitally assisted ultra-wideband mixed mode dual-input power amplifier based on frequency-periodic load modulation (FPLM). For both applications, compact data-driven ML techniques have been applied to boost power amplifiers performance significantly. It is attempted to bring another approach for RF engineers for dealing with sophisticated power amplifiers design and operating tasks.