Digital Predistortion of 5G Multiuser MIMO Transmitters Using Low-Dimensional Feature-Based Model Generation

In this article, we present a novel digital predistortion (DPD) system which can be updated quickly and efficiently in response to the dynamic reconfiguration of multiuser multiple-input multiple-output (MIMO) transmitters. By identifying shared properties different power amplifiers (PAs) with two feature extraction stages, nonlinear behaviors PAs are encoded into low-dimensional features. Using extracted features as input, DPD generator is employed synthesize model coefficients directly. way, when levels or beam angles change, fast accurately without capturing output data applying linear identification algorithms. To capture slow-varying dynamics PAs, also calibrated background. Computational complexity operational latency thus reduced significantly. Simulation experimental results demonstrate that proposed approach achieves excellent linearization performance low complexity, making itself promising solution for 5G massive MIMO