System approach for optimizing a low-complex digital predistortion model for RF power amplifier linearization

Overview & Synopsis

This Ph.D. work contributes to the digital predistortion linearization technique of power amplifiers. Power Amplifier is one of the most critical elements of radiocommunication systems, which exhibits static nonlinearities and nonlinear memory effects. Achieving a good trade-off between the linearity of the power amplifier and its efficiency is becoming more crucial.Digital predistortion is a powerful linearization technique that aims to compensate for power amplifier distortions and provides linear amplification with good efficiency. The predistortion principle consists of implementing a nonlinear function, the so-called predistorter upstream of the power amplifier. The predistorter ideally has the inverse characteristics of the power amplifier. This operation allows us to consider the memory effects of the power amplifier, and in particular, the long-term memory. Several behavioral models have been used as predistorter.Most of them are classified into two families: global models derived from the Volterra series and models based on the segmentation approach. The comparative and analysis study of these models is one of the focuses of this dissertation, in which three aspects are used for comparison: linearization performance, complexity, and hardware implementation properties.By focusing on models based on the segmentation approach, this dissertation proposes an approach to design an optimal model according to a trade-off between linearization performance and model complexity.This model is used to linearize a dual-input Doherty power amplifier. A global optimization algorithm combined with a control process is proposed to enhance efficiency while maintaining a good linearity level according to a proposed adaptive cost function