Access Official Technical PDF

Share Summary:LinkedInEmail

As optical transmission systems expand beyond the conventional C band, estimating nonlinear interference becomes more complicated. Inter-channel stimulated Raman scattering (ISRS) and frequency-dependent fiber loss can cause different parts of a wavelength-division multiplexed (WDM) signal to follow different power profiles along a link. The ISRS GN model extends the Gaussian noise (GN) model to account for that frequency-dependent evolution when predicting the effect of Kerr nonlinearity.

For network planners and optical-system designers, the practical issue is not simply whether Raman scattering changes channel power. A channel’s nonlinear interference also depends on how the interacting frequency components propagate together during four-wave mixing. A model that assumes identical power evolution across the spectrum can therefore misestimate performance in ultra-wideband systems.

Why the conventional Gaussian noise model needs an extension

The GN model estimates nonlinear interference (NLI) by treating the transmitted signal as Gaussian and applying a first-order perturbation analysis of fiber nonlinearity. Its relatively modest computational complexity makes it useful for system design, performance estimation, and physical-layer network optimization. Under many dispersion-uncompensated, multi-span conditions, its assumptions provide a practical approximation.

A central assumption in conventional GN approaches is that every frequency component experiences the same power evolution along the fiber. That is often adequate over narrower optical bandwidths, but it becomes less reliable when the system spans the C and L bands or wider. ISRS transfers power between spectral components: lower-frequency components gain power at the expense of higher-frequency components. Frequency-dependent attenuation can create additional differences, while distributed Raman amplification can produce nonuniform power evolution across the spectrum.

These effects influence the Kerr interaction that generates NLI. To represent that interaction accurately, a model must track the power profile of each frequency participating in the mixing process, rather than applying the profile of the channel being evaluated to all of them.

How the ISRS GN model represents nonlinear interference

The model calculates NLI using the transmitted signal’s spectral power distribution and the normalized power profiles of the frequency components involved in four-wave mixing. In simplified terms, it accounts for the propagation of the channel under evaluation and of the other frequencies that combine to produce nonlinear energy at that channel.

This distinction matters because a frequency-dependent signal profile affects the nonlinear mixing integral along the fiber. If all components are treated as though they attenuate or gain in the same way, the calculated NLI can be biased. The ISRS GN model instead allows each participating frequency to follow its own power evolution, making the framework suitable for arbitrary frequency-dependent profiles.

For a link with identical spans, the single-span result can be extended to multiple spans using a phased-array term that represents how nonlinear contributions accumulate. If span profiles differ, for example because of non-ideal gain-flattening filters, the link must instead be treated using its full power profile.

Analytical or semi-analytical implementation?

When a closed-form power profile is practical

For lumped amplification and negligible variation in fiber attenuation, the signal power profile can be represented analytically by approximating the Raman gain spectrum as linear over the operating bandwidth. Substituting this profile into the NLI calculation produces the analytical ISRS GN model. The approximation is described as suitable for bandwidths up to approximately 15 THz; beyond that range, the Raman gain spectrum can no longer be treated as linear for this purpose.

The analytical form reduces the complexity of evaluating the model. It can be useful for assessing broad trends, comparing launch-power conditions, and estimating NLI in systems operating across the C+L bands, provided the assumptions are appropriate.

When numerical power-profile calculation is needed

For more general cases, the channel powers are calculated by numerically solving coupled ordinary differential equations. These equations represent Raman power transfer between WDM channels alongside frequency-dependent fiber attenuation. The resulting profiles are then used in the NLI calculation, giving a semi-analytical approach.

This route is needed when attenuation varies materially with frequency or when the system uses hybrid amplification that includes distributed Raman gain. The additional numerical step increases computational complexity, but lets the model represent power evolution that the simpler closed-form profile does not capture.

What validation and wideband results indicate

The model was compared with split-step numerical simulations for Gaussian-modulated signals. In the reported validation, the maximum deviation in NLI power was 0.1 dB. This result applies to the conditions studied; it should not be interpreted as a universal error bound for every link configuration.

In a modeled C+L-band system occupying about 10 THz, ISRS changed NLI power by as much as 2 dB at the optimum launch power. The size and direction of the change depend on spectral location and launch conditions: channels experiencing net Raman gain and those experiencing net loss do not see the same NLI response. The study also found that ISRS affects the coherence factor, which describes how nonlinear contributions accumulate across spans, although that change was relatively small for the system examined.

These results illustrate why a spectrum-wide average or a single power profile may be insufficient for wideband link analysis. Channel-level performance estimates can depend on where a channel sits in the optical spectrum and on how its interacting frequencies evolve along the link.

Practical implications for system design

For engineers evaluating ultra-wideband transmission, the model provides a way to incorporate ISRS into estimates of channel-dependent NLI and signal-to-noise ratio. Its outputs can inform analysis of launch-power distributions, link capacity, and physical-layer constraints used in network planning. The modeling approach can also help compare system configurations without relying exclusively on computationally intensive propagation simulations.

Model selection should follow the link’s physical characteristics. A closed-form profile is appropriate when lumped amplification and nearly uniform attenuation make its assumptions reasonable. For distributed Raman amplification, appreciable frequency-dependent loss, or non-identical span profiles, the numerical power-profile approach is more suitable. In either case, the model’s assumptions should be checked against the bandwidth, fiber, and amplification scheme being evaluated.

Conclusion

The ISRS GN model addresses a key limitation of conventional GN analysis: the assumption that all frequencies experience the same power evolution. By tracking the distinct profiles of the components involved in Kerr-driven mixing, it supports more informed NLI estimates for C+L-band and other wideband optical systems. The practical choice is between a lower-complexity analytical formulation for suitable lumped-amplified links and a semi-analytical formulation for cases requiring numerical power-profile calculation. Matching that choice to the transmission system is essential for useful channel-level performance estimates.

Access Official Technical PDF

About The Author