On the numerical stability of RF power amplifier´s digital predistortion

Digital predistortion (DPD) is a highly cost-effective technique to linearize high power amplifiers (HPAs) in wireless applications. The polynomial-type behavior models are widely used both for HPAs and predistorters, but they often suffer from the numerical stability problem when the higher order nonlinearities are included. When wideband signals are transmitted, the envelope impedances of HPAs can not always keep constant, therefore the envelope frequency components caused by the even-order nonlinearities become the main source of the memory effect of HPAs. In this paper, we propose a new polynomial predistorter model considering the influence of the envelope frequency components, which shows better performance in out-of-band spectral suppression; then, we propose a dual-orthogonal-basis polynomial model for complex Gaussian inputs, which alleviate the numerical instability of digital predistortion.