Low-Complexity Feedback Data Compression for Closed-Loop Digital Predistortion

Abstract

This paper proposes sample combining as a low-complex and effective feedback data compression technique that allows to significantly reduce the computational effort and buffering needs for parameter adaptation in a closed-loop digital predistortion (DPD) system. Compression is achieved by applying an integrate & dump operation to an undersampled feedback signal. The proposed method is experimentally validated for RF measurement based behavioral modeling as well as closed-loop DPD of a 3.5 GHz GaN Doherty PA, taking also quantization effects of the feedback path into account. Our results demonstrate that the proposed technique is as capable as state-of-the-art histogram-based sample selection, however, at a much lower complexity.