WDM monitoring through blind signal separation

WDM monitoring in optical networks can be carried out after the separation in the electronic domain of the individual baseband channels, from which suitable performance parameters can then be measured. We have proposed to apply blind signal separation based on higher-order statistics. The specific BSS method considered provides an approximate optimal solution (in the maximum-likelihood sense) for the case of two channels, and has a computational cost of O(N⁵2/ L) when processing L-sample blocks of an N-channel WDM signal. For the signal distributions typically occurring in WDM monitoring, the method presents no undesired solutions. In addition, the case of spectrally white channels can also be handled, thus allowing beneficial reductions in the rates at which the photocurrents are sampled. Although the suggested procedure operates on signal blocks (batch processing), fast adaptive implementations can easily be designed as well.