A Scalable Bandwidth and Frequency-Dependent DPD Linearizer for User Equipment Power Amplifiers With Nonflat Frequency Response

This article presents a scalable digital predistortion (DPD) linearizer for handset power amplifiers (PAs) with nonflat frequency response and assuming a 5G scenario with dynamic resource blocks (RBs) reallocation. The scalable DPD model is designed to be both bandwidth and frequency-dependent, including some design constraints targeting a resource-efficient hardware implementation. In order to extract the most relevant basis functions that better trade-off the linearization performance and the required resources of a hardware implementation of the DPD, a feature selection technique based on a constrained version of the doubly orthogonal matching pursuit (DOMP) is proposed. The linearization performance of the proposed DPD method is evaluated on a PA system-on-chip (SoC), considering 5G new radio (NR) signals at different frequency locations and with different bandwidths within a 100 MHz channel.