MDG and SNR Estimation in SDM Transmission Based on Artificial Neural Networks

Abstract

The increase in capacity provided by coupled space division multiplexing (SDM) systems is fundamentally limited by mode-dependent gain (MDG) and amplified spontaneous emission (ASE) noise. Therefore, monitoring MDG and optical signal-to-noise ratio (SNR) is essential for accurate performance evaluation and troubleshooting. Recent works show that the conventional MDG estimation method based on the transfer matrix of multiple-input multiple-output (MIMO) equalizers optimizing the minimum mean square error (MMSE) underestimates the actual value at low SNRs. Besides, estimating the optical SNR itself is not a trivial task in SDM systems, as MDG strongly influences the electrical SNR after the equalizer. In a recent work we propose an MDG and SNR estimation method using artificial neural networks (ANNs). The proposed ANN-based method processes features extracted at the receiver after digital signal processing (DSP). In this paper, we discuss the ANN-based method in detail, and validate it in an experimental 73-km 3-mode transmission link with controlled MDG and SNR. After validation, we apply the method in a case study consisting of an experimental long-haul 6-mode link. The results show that the ANN estimates both MDG and SNR with high accuracy, outperforming conventional methods.