We present the first direct comparison of the impairments due to "inband cross talk" arising from distributed Rayleigh scattering and a single discrete interference path measured on the same system. The understanding of these cases is important because they represent the extremes of the cross talk statistics that are likely to occur in optical networks due to multipath interference (MPI),~1 component cross talk,~2 and four-wave mixing.~4 We find that the commonly used Gaussian model~1 is pessimistic for the single discrete path case. Furthermore, we show that when other intensity dependent noses are accounted for, the worst case two-beam interference model~(1,3) can be corrected. Our analysis provides an empirical correction with well-defined intensity and cross talk scaling, which automatically accounts for system features such as threshold optimization.~2
展开▼