RF power amplifier behavioral modeling and predistortion using polynomial functions

Abstract

This thesis considers behavioral modelling and digital predistortion (DPD) linearization techniques of radio frequency (RF) power amplifiers (PAs) used in long term evolution (LTE) base stations. Due to the use higher transmission data rates and complex modulation schemes in recent communication standards, memoryless and quasi-memoryless digital predistorters are no longer efficient and the need of more advanced models has become a must. In this work, two different Doherty PAs suitable for the amplification of 4th generation LTE signals were used for model validation in forward and reverse contexts. Moreover, Complex baseband input and output samples of two LTE signals with different average power, peak-to-average power ratios (PAPR) and carrier configuration were used to identify different Volterra reduced dynamic models. A Novel weighted memory polynomial model is proposed and compared with the state-of-the-art memory polynomial based models. The new model shows a comparable performance with the generalized memory polynomial model with a significant reduction in the number of coefficients that is almost 50%. Moreover, the model was experimentally validated against the memory polynomial model in the reverse context and also shows superiority in the reduction of coefficients' number. For the reverse modeling, Indirect learning architecture DPD was deployed. The main contribution in this work is a new behavioral model suitable for direct and reverse modelling of high efficient PAs driven by wideband LTE signals.