A predistorter design for a memory-less nonlinearity preceded by a dynamic linear system

Abstract

We propose a data predistorter design scheme for a memory-less nonlinearity which is preceded by a linear system with memory. This system configuration is often found in telecommunications. The predistorter technique is useful since the compensation of the nonlinearity of high-power amplifiers allows the efficient use of the power resource and bandwidth, while maintaining the prescribed signal spectral distribution. We use either a modified indirect learning architecture or a stochastic gradient method for training the predistorters. As a predistorter structure, we use a Volterra series model or a time-delayed neural network. We apply our approach to the compensation of various nonlinear systems including TWT-type nonlinearities. The results show that our approach is very effective in compensating the memory-less nonlinearity preceded by a linear system with memory. We show the results for nonlinear systems with a TWT-type nonlinearity.