Digital predistortion based on sample selection with memory effect

Abstract

In this article, a novel digital predistortion (DPD) method for radio frequency (RF) power amplifiers (PAs) based on adaptive sample selection is proposed. In the traditional approach, the samples are selected randomly based only on the signal magnitude. This kind of selection implies the potential instability problem. To deal with this issue, we propose a new sampling strategy that takes the influence of memory terms into account, achieving better accuracy of normalized mean square error (NMSE) and more stable convergence than state‐of‐the‐art approaches. To further validate our method, a 100‐MHz OFDM signal is used for experiments. Results show that the proposed method trained with just 0.6% of complete data can achieve comparable linearization performance compared with that trained with complete data, and stable convergence is also validated.