Modeling and Optimization of Frequency‐Doubled Radio‐Over‐Fiber Link Based on Deep Learning Technique

Deep learning (DL) technology has shown extensive potential for modeling radio‐over‐fiber (RoF) systems. Meanwhile, frequency multiplication technology serves as an effective approach to reduce the bandwidth requirements of core components in RoF links. However, the introduction of harmonic interference, nonlinear distortion, and phase noise in frequency‐doubled RoF systems poses additional challenges to DL‐based modeling and optimization. A DL‐based frequency doubled RoF link end‐to‐end (E2E) modeling and optimization method is proposed. The DL model consist of three models: a signal preprocessing model for symbol mapping and signal generation, a fiber channel model for transmission link modeling in addition with a receiver model for demodulation. Numerical analysis is carried out based on 16QAM RoF signal with a bit rate of 2 Gbit/s, which indicates that the DL model is capable of modeling the frequency doubled RoF link which incorporates harmonic and nonlinear distortion effects. Moreover, geometry shaping and probabilistic shaping are performed based on E2E framework to further enhance the performance of the system. The results show that the optimized link has better bit error rate performances, indicating that our approach holds potential applications in nonlinear RoF link modeling and could valuable insights for advanced communication systems.