End-to-end learning strategy based on a frequency domain feature decoupling network emulator with joint probabilistic shaping and equalization for a 300-Gbit/s OAM mode division multiplexing transmission

Abstract

Mode coupling and device nonlinear impairment appear to be a long-standing challenge in the orbital angular momentum (OAM) mode division multiplexing (MDM) of intensity modulation direct detection (IM/DD) transmission systems. In this paper, we propose an end-to-end (E2E) learning strategy based on a frequency domain feature decoupling network (FDFDnet) emulator with joint probabilistic shaping (PS) and equalization for an OAM-MDM IM/DD transmission with three modes. Our FDFDnet emulator can accurately build a complex nonlinear model of an OAM-MDM system by separating the signal into features from different frequency domains. Furthermore, a FDFDnet-based E2E strategy for joint PS and equalization is presented with the aim of compensating the signal impairment for the OAM-MDM IM/DD system. An experiment is carried out on a 300 Gbit/s carrierless amplitude phase-32 (CAP-32) signal with three OAM modes over a 10 km ring-core fiber transmission, and the results show that the proposed FDFDnet emulator outperforms the traditional CGAN emulator, with improvements in the modelling accuracy of 30.8%, 26.3% and 31% for the three OAM modes. Moreover, the receiver sensitivity of the proposed E2E learning strategy is higher than for the CGAN emulator by 3, 2.5, 2.2 dBm and the real channel by 5.5, 5.1, and 5.3 dBm for the three OAM modes, respectively. Our experimental results demonstrate that the proposed FDFDnet emulator-based E2E learning strategy is a promising contender for achieving ultra-high-capacity interconnectivity between data centers.