Linearisation of radio frequency power amplifiers exhibiting memory effects using direct learning‐based adaptive digital predistoriton

Abstract

This study presents an adaptive predistortion algorithm based on direct learning approach to compensate for the non-linearities of a power amplifier (PA) exhibiting memory effects. The proposed algorithm implements the steepest descent technique on an odd-order memory polynomial model to optimise the predistorter coefficients. The performance of the proposed algorithm is validated using a harmonically tuned broadband PA driven by long-term evolution 20 MHz signal. Measurement results confirmed the robustness of the proposed technique by adapting the coefficients of the predistorter to the changes in average input power, drain bias, and gate bias of the PA. The linearisation using the proposed algorithm is compared to the traditional uncompensated case and results are presented. For changes in average input power, gate bias and drain bias levels, the normalised mean square error shows substantial enhancement when the predistortion coefficients are updated using the proposed algorithm.