Machine-learning single-photodetector receiver for the simultaneous detection of hybrid RF and baseband fiber-wireless signals

As mobile communication networks advance through each generation, the number and frequency of utilized frequency bands increase. If each supported frequency range is detected using a separate photodetector in a centralized radio access network, the number of photodiodes required grows with the number of supported frequency bands. Although analog radio-over-fiber (RoF) systems remain attractive for their ability to generate high-frequency signals with fewer high-speed oscillators as carrier frequencies rise, the limited bandwidth in wireless communication networks often results in the underutilization of high-speed photodiodes. This unused bandwidth could be repurposed to detect other access network signals, such as passive optical network signals and digital RoF signals. This paper investigates a machine-learning-based single-photodetector framework designed for the simultaneous detection of multiple access network signals, including hybrid bandpass radio frequency signals and baseband fiber-wireless signals. The primary objective is to assess the performance gains achievable by incorporating machine learning to enhance the detection performance of the digital baseband link, which is impaired by the squared amplitude of the co-detected analog RoF signal. Initial results suggest that integrating a machine learning algorithm could improve the detection performance of the single-photodiode framework for simultaneous detection of multiple access network signals. The framework achieves performance comparable to or better than dedicated photodiode detection approaches when the RF link power significantly exceeds that of the baseband link and outperforms the previously demonstrated subtraction method across all tested scenarios.