Predistortion With Low Feedback Sampling Rate for Wideband Nonlinear Satellite Downlinks

In the satellite channel, the traveling wave tube amplifier (TWTA) and output multiplexing (OMUX) filter introduce the distortions to degrade the signal quality. Traditional signal predistortion methods often compensate for the distortions separately by sampling TWTA output and OMUX filter output, which leads to high sampling cost and system complexity. In this letter, a low sampling rate predistortion method based on deep neural network (DNN) is proposed for satellite downlink. This method introduces a novel DNN-based signal recovery module in the feedback path to recover the TWTA output through the OMUX filter output, and then the distortions of the TWTA and OMUX filter are compensated respectively. To reduce implementation complexity, the DNN onboard is trained offline on the ground with historical data. The experimental results show that the proposed method can achieve the same compensation effect as traditional method while requiring 28% sampling rate.