Experimental Characterization of the Time Correlation Properties of Nonlinear Interference Noise

Abstract

We demonstrate a method for experimentally characterizing the time-evolution statistics of nonlinear interference noise (NLIN). For the first time in an experimental setting strong temporal correlations are observed beyond the phase-noise NLIN component. The ability of measuring these correlation is imperative for designing NLIN mitigation schemes. IntroductionModern wavelength division multiplexed (WDM) systems are often limited by nonlinear interference between different WDM channels. From the standpoint of each individual WDM receiver, this interference appears random and is registered as noise. It is often referred to as nonlinear interference noise, or NLIN. In recent years, extensive efforts have been invested into the characterization of NLIN and the exploration of methods for its mitigation. An important finding in this context was that the cross-phase-modulation (XPM) component of NLIN (the predominant component in most practical scenarios) can be modeled very accurately as an inter-symbol interference (ISI) process with random and time varying ISI coefficients 1-3 . Hence, the characterization of NLIN can be translated into the characterization of the corresponding ISI coefficients. Devising reliable experimental methods for doing that is imperative for assessing the prospects of NLIN mitigation and for designing NLIN equalization techniques 4-7 . Almost all existing research to date has focused on the zeroth-order ISI coefficient, whose contribution to NLIN is phase and polarization rotation noise (PPRN) [9][10][11] . Yet, in most cases of interest, it is also important to account for higher order ISI terms, whose accumulated effect may exceed the PPRN contribution. One of the obstacles to doing so is the difficulty in evaluating the higher order ISI terms experimentally. While the zeroth-order ISI term produces clear kidney-shaped spots in the received constellation points, the effect of higher order ISI terms is not geometrically obvious, and can be easily confused with additive white noise.In this paper, we describe a method for measuring general-order ISI coefficients and evaluat-

Machine Learning Based Linear and Nonlinear Noise Estimation

“…Although the ISI matrices are time-varying, their statistical properties can be studied through the autocovariance function (ACF). In [16], a metric capable of extracting their temporal properties from the post-DSP received symbols,â (n) r , was derived: Figure 1 shows the evolution of the real part of the most relevant orders of ACF for SSMF, ELEAF and TWC fibres over 10 spans, accounting for a total transmission of 1000km. The real part is the most significant contribution for the considered cases.…”

Nonlinear Interference Noise in Raman-Amplified WDM Systems

“…In long-haul dispersion-uncompensated systems, the NLPN exhibits long auto-correlation times, that allow for efficient filtering of this noise contribution. This effect has been characterized in recent experiments [8][9][10], and filtering techniques based on auto-regressive and Kalman filtering have been investigated [11].…”