Mode Pilot-Tone (MPT)-based High-precision Crosstalk Monitoring for Mode-Division Multiplexing Systems

Efficient and precise monitoring on mode-coupling-induced crosstalk (XT) is crucial for the operation of mode-division multiplexing (MDM) systems, particularly for optimizing digital signal processing (DSP) parameter configurations and maintaining stable transmission performance. In this paper, we propose a mode pilot-tone (MPT) monitoring scheme and investigate its performance for XT (MPT-XT) monitoring in a three-mode MDM transmission system. By deploying a photodiode (PD) with a 1 GHz bandwidth for each mode to extract the MPT responses, the complete XT information from 21 GBaud 16-QAM signals is obtained by the proposed scheme. In the experiment, the root-mean-square-error (RMSE) for XT-prediction achieves 0.11 dB, with a coefficient of determination R2 of 0.9992 across the OSNR range from 15dB to 25dB. By applying the transfer learning-based crosstalk neural network (TL-XT-NN) scheme, which transfers the NN model trained exclusively with simulation data to experimental XT monitoring scenarios, the training overhead required by the NN is reduced more than 48.3%. Moreover, we further evaluate the monitoring performance of the proposed scheme on the multiplexing density, the chromatic dispersion (CD), the differential mode group delay (DMGD), and the mode differential loss (MDL). Simulation results confirm its exceptional and consistent performance across all conditions. Finally, we investigate the impact of MPT modulation on the compensation behavior of multi-input multi-output (MIMO) equalizer and propose an inverse mapping label erasure (IMLE) scheme to mitigate the resulting signal degradation. Through the MPT erasure procedure, the error vector magnitude (EVM) penalty of transmitted signals caused by MPT operation is reduced from 2.92% to 0.19%.