Piecewise Volterra Series Approximation for Improved Non-Linear Channel Modelization and Detection

In satellite communications, the non-linear distortions introduced by the amplifier in the payload have to be overcome. When advanced mitigation techniques are considered at the receiver side, the current channel model is often based on Volterra series derived from an approximation of the non linear transfer function of the on-board amplifier. This nonlinear model is conditioning the performance at the receiver side. In this paper, a new non-linear model is proposed, leading to improved receiver performances. The polynomial approximation is improved considering both the usual model truncation to the $3^{rd}$ order and the signal fluctuation at the input of the amplifier. First, the impact of the polynomial order of the AM/AM and AM/PM curve approximation is studied. Then, a non-linear model is derived based on a piecewise polynomial approximation of the amplifier response. Based on this refined nonlinear model, significant detection performance improvements are shown for both Nyquist and Faster-than-Nyquist rates.