OSNR Monitoring Using Support Vector Ordinal Regression for Digital Coherent Receivers

An OSNR monitoring method assisted by support vector ordinal regression (SVOR) is proposed for digital coherent receivers. According to Godard's error of equalized signals obtained after polarization demultiplexing, the optical signal-to-noise ratio (OSNR) value could be effectively estimated by using SVOR. The effectiveness of the proposed OSNR monitoring method is verified via numerical simulations with 28-Gbaud PDM-QPSK, PDM-8PSK, PDM-8QAM, PDM-16QAM and PDM-32QAM signals. The simulation results show that the 100% accuracies for the five modulation formats can be obtained over a wide OSNR range. The feasibility is further demonstrated by proof-of-concept experiments in 28-Gbaud PDM-QPSK, PDM-8PSK, PDM-8QAM and PDM-16QAM systems. The errors of averaged values of estimated OSNR for all the signals are less than or equal to 0.2 dB.