DSP-Based Nonlinear SNR Estimation via Longitudinal Power Monitoring in Commercial Coherent Receivers

The signal-to-noise ratio (SNR) measured on the received symbol constellation, accounting for both amplified spontaneous emission (ASE) noise and Kerr-induced nonlinear interference (NLI), is a key metric for system design and optimization. Consequently, separating these noise contributions enables the optimization of optical networks for more efficient and lower-margin transmissions. However, recent methods for nonlinear SNR estimation, such as artificial neural networks and statistical noise manipulation, require extensive knowledge of transmission parameters or large training datasets. In this work, an alternative approach, which requires minimal information about the system configuration, is described and validated through simulations and experiments. It exploits a linear least squares (LLS)-based longitudinal power monitoring (LPM) algorithm that can be implemented in standard coherent receivers, combined with analytical expressions based on closed-form NLI models (e.g., the GN model). In addition, the practical implementation challenges of using hard-decided symbols to generate reference signals for LLS-based LPM in NLI estimation are explored.