Neural-Network-Based Digital Predistortion for Active Antenna Arrays Under Load Modulation
Alberto Brihuega, Lauri Anttila, Mikko Valkama
Abstract
In this letter, we propose an efficient solution to linearize mmWave active antenna array transmitters that suffer from beam-dependent load modulation. We consider a dense neural network that is capable of modeling the correlation between the nonlinear distortion characteristics among different beams. This allows providing consistently good linearization regardless of the beamforming direction, thus avoiding the necessity of executing continuous digital predistortion parameter learning. The proposed solution is validated, conforming to 5G new radio transmit signal quality requirements, with extensive over-the-air RF measurements utilizing a state-of-the-art 64-element active antenna array operating at 28-GHz carrier frequency.