Linear and Nonlinear SNR Estimation using Spectral and Temporal Correlations
Vinod Bajaj, Pétros Ramantanis, Fabien Boitier, Patricia Layec · 2024
We propose a method to estimate linear and nonlinear signal-to-noise ratio (SNR) by using normalized correlations of symbol rate spaced frequencies [1] and amplitude correlation functions [2], which are derived from received signal in optical coherent transmission. These correlations are utilized as input features to a neural network (NN) for estimation. We evaluate our proposed method on 64 GBaud dual-polarized (DP) 16QAM (0.1 roll-off) transmission data generated using split-step Fourier method (SSFM) simulations with standard single-mode fiber parameters. Our method achieves estimation errors within ±0.5 dB for linear SNR and ±2 dB for nonlinear SNR (for values ≤ 25 dB).