Effects of signal PDF on the identification of behavioral polynomial models for multicarrier RF power amplifiers

This paper proposes an experimental study of the sensitivities of the power amplifier modelling and their influences on the system's identification. Two memory polynomial models are widely investigated in the behavioural modelling and linearization of RF power amplifiers (PAs). In order to improve the accuracy of the behavioural modeling versus identification, we assess the performances of these models under different signal bandwidths, statistics and signal distributions. For this purpose, normalized mean square error has been used to compare the model output to measured data when the PA is driven under multi-carrier input signals. Measurement results and simulation have been carried out and the results demonstrated the effects of the signal characteristics on the performances of the model. Multi-carrier wideband code-division multiple access and multi-tone signals were used with an experimental Doherty amplifier. The obtained results revealed a degradation of the polynomial model performances when the statistics of the input signals change with similar peak-to-average power ratio.