FTnet-based Digital Demodulator for Radio over Fiber Transmission

Abstract

A digital demodulator based on Fourier layer Transformer network (FTnet) is proposed for radio over fiber (RoF) transmission with intensity-modulation and direct-detection (IM-DD). The FTnet is a new deep learning network, which replaces the self-attention mechanism of the Transformer encoder with a Fourier layer. The FTnet-based digital demodulator can directly recover corresponding bitstreams from the impaired receiving signal, which replaces a series of digital signal processing in the RoF traditional digital demodulator, including down-conversion, matched filter, downsampling, phase offset compensation, equalization, decoding, etc. A 10 GHz 16/64QAM 2 Gsym/s 25 km RoF transmission system is established to verify the proposed method. The experimental results show that the receiving sensitivity for 16QAM and 64QAM systems with the proposed demodulator can be improved by 1 dB and 5 dB respectively compared to traditional digital demodulator with LMS equalizer under a bit error rate limit of 3.8×10−3.