Measure-based modeling and FPGA implementation of RF Power Amplifier using a multi-layer perceptron neural network

This paper is focused on the development of an emulation of a Memorial Polynomial Model (MPM) for Radio Frequency Power Amplifier (PA) with Artificial Neural Network (ANN) using back propagation algorithm (BP) considering the nonlinear and memory effects. This model is based on the accurate capacities of artificial neural networks to fit functions. We demonstrate that it is a practical tool to emulate precisely the behavior of a power amplifier since its measured AM-AM and AM-PM conversion curve. Results are given to show how precisely the characteristics of three amplifiers are modeled thanks this technique. These ANN and MPM are implemented on the DSP-FPGA Development Kit, Cyclone®III Edition-Altera. The reduction of computational complexity together with the fast processing involved in the DSP Board give a proper behavioural modelling of the PA, and leave open the option to introduce different digitally/analogically modulated signals.