Long Short-Term Memory Networks for Behavioral Modeling of A GaN Sequential Power Amplifier

Abstract

this paper investigates wideband behavioral modeling of Gallium Nitride (GaN) power amplifiers (PAs) using long short-term memory (LSTM) networks. Due to the memory mechanisms used in LSTM networks, they have the capability of accurately capturing both the short term and long term memory effects presenting in GaN PAs. The LSTM network-based model is verified experimentally on a GaN sequential power amplifier (SPA) under wideband multi-channel modulated signals, with showing good alignment between the modeled and measured data.