Nonlinear Interference Mitigation: Methods and Potential Gain

We explore the potential benefits of digital nonlinearity compensation (NLC) techniques in fully loaded coherent wavelength-division multiplexed (WDM) transmission systems. After providing an overview of the various classes of nonlinear interference noise (NLIN) and digital signal processing approaches for their mitigation, we consider the two practically most relevant digital NLC methods known today: back-propagation (BP) and equalization of nonlinear phase and polarization rotation noise (PPRN). We consider a wide range of system configurations including a variety of modulation formats from quadrature phase-shift keying (QPSK) to 256-ary quadrature amplitude modulation (QAM), single-carrier and digital subcarrier multiplexed (SCM) optical superchannels, as well as both point-to-point line systems and optically routed networks (ORNs). Using theoretical predictions from the time-domain model for NLIN, we show that the gain in peak signal-to-noise ratio in fully loaded WDM systems using single-channel and three-channel joint BP is typically limited to 0.5 and 1 dB. The additional gain provided by increasing the number of jointly back-propagated channels beyond three is limited to 0.1 dB per additional back-propagated channel. The remarkably slow growth of the BP gain with the bandwidth of the back-propagated signal is shown to apply also for systems in which the receiver employs PPRN removal. We additionally explore the potential benefits of SCM across a wide range of system configurations, including the impact of BP and PPRN removal on SCM systems. We show that while the BP gain is similar to single-carrier systems, PPRN removal can have a much stronger effect, particularly for higher order QAM systems, implying that SCM can be advantageous not only for constant modulus formats such as QPSK, but also for high spectral efficiency systems. In the context of ORNs, we find that SCM not only improves system tolerance to nonlinearities, but also induces significantly smaller performance variations in various ORN scenarios.