Exploring predistortion training algorithms in a Cartesian feedback-trained digital predistortion system for RF power amplifier linearization

Abstract

A Cartesian feedback-trained digital predistortion system for RF power amplifier linearization offers many advantages with its combination of two different linearization techniques. This thesis describes such a system, focusing on the important issue of predistorter training. It examines and analyzes in great detail the promising loop filter pre-charging optimization and the tradeoffs associated with such training, developing a model that provides many valuable system design insights. In order establish a means to experimentally verify the theory and explore predistortion training algorithms, this thesis presents the design, development, and characterization of a mock-up prototype that models the essential features of the actual Cartesian feedback-trained digital predistortion system. The mock-up serves as a standalone proof-of-concept system that demonstrates the benefits and tradeoffs of loop filter pre-charging in predistorter training. It confirms the theory while also revealing practical issues pertaining to the limits on performance.