A NN Predistortion Method for Nonlinear HPAs in OFDM System
基于BPNN的OFDM系统的HPA预失真

To circumvent the transmission performance degradation of Orthogonal Frequency Division Multiplexing (OFDM) systems due to the nonlinear HPA, a new predistorter is presented to predistort HPA which consists of two similar single-input and single-output Neural Networks (NNs) in series. The former NN is the inverse model of HPA’s AM-AM characteristics to compensate its amplitude and the latter the model of AM-PM characteristics to compensate its phase. The phase predistortion is implemented just by its model rather its inverse model, which ensures it of much higher precision. Simulation results show that even though input back-off is as low as 2.93dB, the spectral regrowth can be reduced about 10dB. The proposed scheme can adaptively predistort HPA effectively and simply, and improve transmission performance of OFDM system greatly.