Real-time low-distortion digital PWM modulator for switching converters

Abstract

A compensation scheme to reduce the inherent baseband distortion in uniform PWM modulators (UPWM) is presented in this paper. The method is based on real-time mapping of the switching times of UPWM to those of natural PWM that, as it is well-known, exhibits far less in-band distortion. Two alternatives are presented: one based on an exact, analytic algorithm recently reported in the literature, and another one that uses an artificial neural network (ANN). Both methods are designed for arbitrary, band-limited modulating signals and they are not restricted to single-frequency sinusoids, as other techniques presented in the literature. Simulation results and experimental measurements of a FPGA implementation demonstrate a significant reduction of the distortion in real-time applications. The performance of both alternatives is compared for several modulating signals, and the results of industry-standard distortion tests are also reported. Finally, some guidelines for choosing the best alternative for ASIC or FPGA implementations are provided