Experimental analysis of optical spectrum based power distribution analysis for intermediate node monitoring in optical networks using shallow multi-task artificial neural network

In this work, we propose an intelligent solution to cost-effective intermediate node monitoring in optical networks using optical spectrum-based power distribution analysis (OSPDA) and shallow multi-task artificial neural network (SMT-ANN). The proposed technique is used to realize simultaneous identification of modulation format (MF) and multi-parameter optical performance monitoring (OPM) such as identification of launch power (LP), chromatic dispersion (CD), differential group delay (DGD), and optical signal-to-noise ratio (OSNR) estimation. OSPDA is based on comparing optical spectrums with and without impairment to determine the power level deviations and correlation for simultaneous OPM. It involves features derived from OSPDA as input to the proposed SMT-ANN for executing both identification and estimation of OPM. The experimental validation has been carried out for 10 different MFs such as 4 Quadrature Amplitude Modulation (QAM), 8 QAM, 16 QAM, 32 QAM, 64 QAM, Binary Phase Shift Keying (BPSK), Quadrature Phase Shift Keying (QPSK), Offset QPSK (OQPSK), 8 Phase Shift Keying (PSK), and 16 PSK at three LP for both back-to-back (B2B) and 50 km optical fiber transmission link. The various levels of CD and DGD were introduced using different lengths of optical fiber. The best results achieved from the analysis include 99.87 %, 99.81 %, 98.72 %, and 98.36 % identification accuracy for MF, LP, CD, and DGD respectively. The minimum average mean absolute error (MAE) obtained for OSNR estimation is 0.218 dB. Thus, the proposed method is practically feasible for simultaneous OPM at intermediate nodes for real-time robust and reconfigurable optical networks.