Multi-basis weighted memory polynomial for RF power amplifiers behavioral modeling

Abstract

In this paper, two multi-basis weighted memory polynomial models are proposed for radio frequency power amplifiers' behavioral modelling. In these models, the conventional memory polynomial function of the generalized and hybrid memory polynomial models is replaced by a weighted version of it. Experimental validation was performed on a power amplifier prototype exhibiting strong memory effects, and driven by a 20 MHz LTE signal with 1001 configuration. Proposed weighted generalized memory polynomial and hybrid memory polynomial models show superior performance when compared to their memory polynomial based conventional counterparts. Indeed, an NMSE improvement of 2 dB to 3 dB is obtained for the same complexity, and a reduction of almost 50% in the number of coefficients are achieved for the same performance.