Quantifying distortion of RF power amplifiers for estimation of predistorter performance

This paper demonstrates a method to quantify the accuracy of memory models and the effectiveness of digital predistortion of power amplifiers with memory. By using assumptions of periodic stationarity, coherent ensemble averaging and a stable measurement system, we are able to decompose the waveform distortion into memoryless, deterministic memory, and random memory contributions. We demonstrate how this can be used to evaluate the performance of a power amplifier and project its optimal performance with predistortion. We also show experimentally that the dynamic deviation reduction-based Volterra series digital predistortion technique has convergent behavior, moving towards these quantified targets when applied to a class AB power amplifier implemented with GaN FETs.