Genetic Algorithm Optimized Memory Polynomial digital pre-distorter for RF power amplifiers

Abstract

Digital pre-distortion (DPD) is an efficient way of linearizing RF power amplifiers in wireless communications transmitters. Memory Polynomial and Generalized Memory Polynomial methods are two such successful methods capable of reducing spectral regrowth of high power amplifiers with memory effect. However, these methods often need a large number of coefficients, which makes these methods less cost efficient. In this paper we present an effective method based on Genetic Algorithm to simultaneously reduce the number of coefficient and optimize the performance of Memory Polynomial (MP) and Generalized Memory Polynomial (GMP) Radio Frequency (RF) power amplifier pre-distorters. The proposed method is validated using a single carrier WCDMA signal using an indirect learning architecture. In comparison with the MP model, the proposed model shows improved adjacent channel power ratio performance in the DPD application with 42% reduction in the number of coefficients. In comparison with the GMP model, the proposed model achieves higher model accuracy and better DPD performance, but reduces 25% of coefficients.