Dynamic neural network based pre-distortion in satellite communication

With OFDM, WCDMA and other high peak-to-average power ratio (PAPR) modulation technique applied in satellite communications, and the ever-incensing transmission speed of satellite communication, traditional power amplifier pre-distortion technology has revealed inefficiency to meet the requirement of satellite communication. Aiming at this problem, we present a dynamic neural network predistortion method. The amplitude and phase of the input signal are separated ahead, while the proposed dynamic neural network based predistortion is processed on this basis respectively. The computational complexity reduction and the performance pre-distortion improvement are proved through mathematical derivation. Simulation results and their analysis verify preliminarily that the proposed method can be more effective in suppressing the power amplifier's nonlinear characteristics and the memory effect; compared with third-order Volterra, it is more suitable for practical applications.