Experimental approach for robust identification of radiofrequency power amplifier behavioural models using polynomial structures

Abstract

Here, the robustness of polynomial model identification for radiofrequency power amplifiers (PAs) driven by wideband code division multiple access (WCDMA) signals is investigated. The ill-conditioning problem is highlighted, and the sensitivity of the model identification to disturbance is demonstrated. These issues are mainly related to the correlated character of the WCDMA signal. The identification robustness is then enhanced using, instead of the WCDMA input, a synthetic wideband signal with a parametrised probability density function (pdf) close to that of the WCDMA signal, which is of the Rayleigh type. The pdf of the synthetic signal is chosen in order to have the best conditioning values. The signal characteristics, including bandwidth and average power, are controlled to maintain the same amplifier average operating conditions during the characterisation step. Measurement results performed on a 300 W Doherty PA illustrate satisfactory modelling accuracy and significant improvement in the identification robustness achieved by using the proposed approach.