Neural Network based Nonlinear Forward Model Identification for Digital MIMO Arrays Under Load Modulation

In multi-antenna transmitters, antenna coupling may lead to load modulation of the PAs, making their nonlinear responses beam-dependent and compromising traditional digital predistortion solutions. This paper presents a methodology for nonlinear PA array modeling using time-delay neural networks (TDNNs), with specific focus on digital MIMO systems. RF measurements on an emulated <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$4 \times 1$</tex> array at 2.1 GHz demonstrate the effectiveness of the proposed approach in accurately modeling the PA array under load modulation, while showing clear gains over current polynomial-based and TDNN-based state-of-the-art models.