A Grouped-ANN Equalizer for the Rotated PS 64QAM OFDM in W-Band ROF System

Abstract

In the letter, a grouped artificial neural network (G-ANN) based equalizer is proposed and experimentally demonstrated for the rotated and probabilistically shaped 64 quadrature amplitude modulation (RPS 64QAM) orthogonal frequency division multiplexing (OFDM) in W-band radio-over-fiber (ROF) system. PS OFDM can achieve capacity improvement and flexible rate adaptability. Then, the rotated QAM scheme is applied after PS to increase the noise tolerance and resist selective frequency fading in ROF system. In addition, a deep learning equalizer using G-ANN is proposed for RPS 64QAM OFDM. The numerical analysis indicates that the proposed G-ANN equalizer has a better performance and lower training overhead compared with the ANN equalizer. The experimental results show that, after 20 km standard single mode fiber (SSMF) and 1 m wireless transmission, compared to RPS without and with ANN equalizer, the sensitivity improvement of RPS OFDM with the proposed G-ANN equalizer can achieve 1.5dB and 0.6dB at BER of $10^{-5}$ . Meanwhile, the receiver sensitivity of RPS can be improved 1.6 dB compared with PS at normalized generalized mutual information (NGMI) of 0.83.