Digital Predistortion Function Synthesis using Undersampled Feedback Signal

Abstract

This letter presents a new approach to synthesize the digital predistortion (DPD) function using an undersampled feedback signal. First, an expression for the DPD update algorithm that accommodates undersampling of the feedback signal is derived. This includes a direct learning algorithm that iteratively identifies the DPD function coefficients. Then, a delay estimation and alignment algorithm that employs a fractional delay filter is presented for estimating and compensating the non-integer delay between the sampled input and undersampled output signals of the power amplifier (PA). The new proposed approach is found to have comparable linearization capability compared to a conventional full-rate based indirect-learning DPD, even with a significantly undersampled feedback signal. For instance, it was successfully applied to linearize a 20 W GaN Doherty PA driven by a wideband modulated signal of up to 80 MHz bandwidth, and yield an ACLR of -49 dBc after linearization using a complex feedback signal sampled at 80 complex MSPs as opposed to 400 complex MSPs that would be required for conventional sampling.