A Robust Augmented Complexity-Reduced Generalized Memory Polynomial for Wideband RF Power Amplifiers

Abstract

This paper proposes a robust model for the accurate behavioral modeling and digital predistortion (DPD) of wideband radio-frequency power amplifiers (PAs). It is constructed using a complexity-reduced generalized memory polynomial (MP) (GMP) (CR-GMP) connected with a nonlinear memory effect (NME) subblock in parallel. The CR-GMP is a complexity-reduced but accuracy-degraded version of the conventional GMP, and its performance is augmented by the extra NME subblock. Hence, the proposed model is termed as augmented CR-GMP (ACR-GMP). The resultant ACR-GMP model can achieve comparable performance as the GMP model, but with much fewer coefficients and lower complexity. Its performance is experimentally assessed both in forward modeling and DPD linearization. Comparisons are conducted between the ACR-GMP model and some state-of-the-art models, such as the MP, the PLUME, and the GMP. Experimental results have been given for a 1.9-GHz 35-W peak-power GaN Class-AB PA driven by two signal scenarios: a 15-MHz bandwidth long-term-evolution signal and a 20-MHz bandwidth wideband-code-division-multiple-access 1001 signal (with the middle two carriers OFF). All the results show that the ACR-GMP model outperforms both the MP and the PLUME models in terms of performances and the GMP model in terms of complexity (at comparable performances).