Nonlinear Frequency-Division Multiplexing

Two signal multiplexing strategies for multi-user optical fiber communication are considered: Wavelength-division multiplexing (WDM) and nonlinear frequency-division multiplexing (NFDM), based on the nonlinear Fourier transform (NFT). The inverse NFT is implemented using an approach that is dual of the forward NFT and does not require solving integral equations. In contrast to prior work where NFDM is applied to single-user channels or combined with WDM, in this paper users’ signals are multiplexed in the nonlinear Fourier domain. Furthermore, each user now sends a sequence of symbols. NFDM orthogonalizes the nonlinear Schrödinger equation in a generalized frequency and a generalized time. Thus all degreesof-freedom are (deterministically) independent. As an example, a data rate of 10.5 bits per complex degree-of-freedom is achieved in a 15-user NFDM with 60 GHz overall bandwidth at average input power P “ ́0.33 dBm. Data rate in a comparable 15user WDM with the same bandwidth and power is 5.26 bits per complex degree-of-freedom. The ratio of the spectral efficiency of NFDM and WDM is 2.23, taking into account bandwidth and time duration of signals at channel input and output. Data rates are plotted as a function of the average input power, indicating that the NFDM rate increases monotonically with transmit power, in contrast to the WDM rate which characteristically vanishes as transmit power is increased more than an optimal value. Thus, as a result of the nonlinear multiplexing of users’ signals, NFDM spectral efficiencies higher than the corresponding WDM spectral efficiencies are demonstrated at high input powers.