Improving short-term stability of fiber-optic radio frequency transfer using long short-term memory prediction

A phase compensation method (PCM) based on the long short-term memory (LSTM) network is proposed for improving the short-term stability of fiber-optic radio frequency (RF) transfer systems. The proposed LSTM-PCM can predict the phase fluctuation within the next propagation delay and realize the precompensation of the system. For verifying the feasibility of the proposed method, we transmitted the 2.4 GHz RF signal over the optical fiber links from 200 to 1000 km. The phase fluctuations obtained from the experiment were used to construct an LSTM network. We simulated the frequency stability of the LSTM-PCM, and the findings demonstrate that the short-term stability of the system is improved more significantly with the increase of the transmission distance. With a transmission distance of 1000 km, the classical PCM achieves an overlapping Allan deviation (OADEV) of 1.02 x 10(-12) at 1 s, whereas the proposed method can reduce the OADEV to 6.60 x 10(-13) at 1 s. The LSTM-PCM can effectively suppress the residual delay phase fluctuations and has the potential for application in long-haul fiber-optic RF transfer systems.