Linearization of Active Array Transmitters Under Crosstalk via Over-the-Air Observations

Finding computationally feasible solutions to linearize active antenna array transmitters is a timely challenge in modern cellular networks, particularly in 5G and the emerging 6G systems. In this paper, we propose a new modeling and parameter estimation approach to characterize the individual power amplifier (PA) units of an active antenna array subject to crosstalk while relying only on over-the-air observations. Additionally, we describe a beam-sweeping based forward model learning procedure and the corresponding closed-loop digital predistortion (DPD) learning algorithm, to efficiently linearize millimeter-wave phased-array transmitters under crosstalk-induced load modulation. The provided numerical results demonstrate excellent parameter estimation and linearization performance, reaching adjacent channel power ratios (ACPRs) as low as −50 to −60 dB with realistic evaluation assumptions.