Experimental sensitivity analysis of multi-standard power amplifiers nonlinear characterization under modulated signals

Abstract

This paper proposes an experimental analysis focusing on the sensitivity of three behavioral models, Memory Polynomial (MP), Augmented Hammerstein (AH) and the two hidden layers artificial neural networks (2HLANN) to the characteristics of the input signal driving the power amplifier (PA) to be linearized. The analysis is carried out by changing separately each signal characteristic, respectively the peak to average power ratio (PAPR), the Probability density function (PDF), and the modulation bandwidth and assess the sensitivity of the DPD to that change. When used to linearise a 250 Watt peak-envelop-power Doherty PA, the considered models showed relatively small sensitivity to the variation of these signal characteristics. Yet, the 2HLANN was found to be the most robust model with excellent linearization capabilities.