Neural network predistortion based on error-filtering algorithm in OFDM system

Orthogonal frequency division multiplexing (OFDM) system is equipped with the solid state power amplifier (SSPA) which causes severe nonlinear distortions when it works near saturation. In this paper, a neural network predistorter is designed based on the adaptive inverse control technique. The error-filtering algorithm is applied to improve the classical adaptive structure, hence the convergent process turns to be rapid by using the LMS algorithm. Simulation results demonstrate high efficiency and low complexity of the method, and the compensated system is characterized by the approximately ideal amplifier within the range of saturated input value.