Band-Limited Digital Predistortion Based on Iterative Learning Control With Input Signal Reconstruction for Wideband RFPAs

Abstract

To address the analog-to-digital converter (ADC) sampling rate bottleneck in wideband, a novel band-limited (BL) digital predistortion (DPD) method based on iterative learning control (ILC) with input signal reconstruction (ISR) is proposed. Unlike conventional approaches, this method updates the BL input signal in the BL domain and subsequently reconstructs the full band input using a forward model. Notably, the regression matrix of the forward model is consistently constructed from the original baseband signal, thereby ensuring numerical stability throughout the iterative process. Experimental results demonstrate that the proposed ISR-ILC strategy outperforms existing algorithms. Moreover, under a relaxed oversampling rate, it achieves superior linearization performance while significantly reducing computational complexity.