ROVA: A Practical, Parsimonious Volterra Model for Modulated Microwave Systems

Abstract

This article presents a novel Volterra model particularly suited to describe modulated microwave systems. The seRies expansiOn of the Volterra Approximation (ROVA), (ReeksOntwikkeling van de Volterra Approximatie in Dutch) model uses the Taylor series expansion of the frequency-domain Volterra kernels and their symmetries to avoid a combinatorical explosion in model parameters. ROVA is moreover linear in its model parameters and can hence be solved in linear least squares sense. It is parsimonious compared to other state-of-the-art models used for similar applications. Additionally, the interpretability of ROVA avoids the need for heuristics to tune its hyperparameters. The performance of the ROVA approach is compared to current state-of-the-art memory polynomials. Simulation and experimental validation on power amplifiers indicate that ROVA outperforms the memory polynomials with less model parameters and lower modeling error.