Characterizing Device Nonlinearity in Optical Communication Systems Using Frequency-Resolved Nonlinear Noise-to-Signal Ratio Estimation

In modern high-speed optical communications, imperfections in transceivers become the primary source of signal degradation as baud rates and modulation orders increase. It is essential to accurately identify, characterize, and specify both linear and nonlinear imperfections. While techniques for addressing linear imperfections are well-established, tackling nonlinear imperfections remains challenging due to their dependence on both the system and the input signal. The noise power ratio (NPR) method, developed decades ago, is effective for Gaussian input signals. This method involves intentionally notching a specific frequency component of the input signal and measuring the resulting re-growth component at that frequency in the output. The NPR is defined as the ratio of this re-growth component's power to the total output power. Despite its simplicity, the NPR method does not accurately estimate nonlinear impairments for systems with non-Gaussian input because the signal behaves differently with and without the notch. To address this issue, we propose a new method for measuring the nonlinear noise-to-signal ratio (NSR) without altering the input signal. Our approach involves dividing the test input signal into numerous frequency subbands and using the correlation between the received and transmitted data within each subband to calculate the NSR. We conduct a performance analysis of this estimator and derive its theoretical mean and variance using Taylor expansion. The effectiveness of our method is evaluated through numerical simulations of a finite Volterra series up to the third order, with comparisons made against the orthogonal method that separates linear and nonlinear components. Results demonstrate that our proposed method provides an accurate estimation of nonlinear impairments compared to the benchmark.