Efficient Inverse Covariance Matrix Estimation for Low-Complexity Closed-Loop DPD Systems

Abstract

This paper studies closed-loop digital predistortion systems, with special focus on linearization of mmW active antenna arrays. Considering the beam-dependent nonlinear distortion and very high DPD processing rates, a modified self-orthogonalized (SO) learning solution is proposed, which is capable of reducing the computational complexity compared to other similar solutions, while at the same time obtaining a comparable linearization performance. The modified SO consists of a novel method for efficiently calculating the inverse of the input data covariance matrix. Thorough RF measurement results at 28 GHz band featuring a state-of-the-art 64 element active array and channel bandwidths up to 800 MHz, are reported. A complexity analysis is also carried out which, together with the obtained results, allow to asses the performance-complexity trade-offs. Altogether, the results show that the proposed methods can facilitate efficient mmW active antenna array linearization.