Phase-Normalized Neural Network for Linearization of RF Power Amplifiers
Arne Fischer-Bühner, Lauri Anttila, Manil Dev Gomony, Mikko Valkama
Abstract
This letter proposes a methodology for phase-normalization of the complex-valued I/Q inputs of a real-valued time delay neural network (RVTDNN). The normalization enables modeling of the nonlinear behavior of a radio frequency (RF) power amplifier (PA) in a more efficient way, by complying with the physical characteristics of the distortions at RF. The presented digital predistortion (DPD) linearization experiments with a Doherty GaN PA at 3.5 GHz show a 4-dB improvement in the output linearity compared to state-of-the-art neural network (NN) and polynomial-based DPD models, allowing linearization to below <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$-$</tex-math> </inline-formula> 50 dBc adjacent channel leakage ratio (ACLR) levels with feasible processing complexity.