Nonlinear Fourier Transformation Based Detection

A mathematical tool for calculating the nonlinear Fourier transform (NFT) of a time-domain signal is the Ablowitz-Ladik iteration. After having revisited this algorithm, it is applied to coherently detected 16-GBd BPSK bursts transmitted over a few spans of standard SMF fiber in lab. For decision on a transmitted burst we compare its NFT spectrum with a set of calculated reference spectra. Decision on the continuous part of the nonlinear spectrum did not lead to an improvement in the nonlinear transmission regime could not be confirmed in the experiments. The main reason for this degradation appears to be peaks in the amplitude spectrum which exhibited a large variance for a noise load and require a high precision of receiver's channel estimation. On the other hand, decision based on the discrete part of the nonlinear spectrum worked successfully. Discrete complex frequency points (eigenvalues) exhibit a low statistical spreading for a noisy signal and allow low BER detection at high burst power levels up to 12-dBm burst power over up to 3 fiber spans. However, the results also indicate that with increasing power and link length the detection is limited by a growing discrepancy between NFT spectrum obtained from signal propagation on realistic multi-span lossy fiber link and spectrum calculated from lossless fiber of the NFT channel model.