Fiber-optic Communication using Discrete Spectral Modulation

Abstract

This thesis presents signal-processing and encoding methods for fiber-optic communication using the discrete spectrum of the nonlinear Fourier transform (NFT) of signals. Under the action of the NFT, the channel model for an optical fiber, a noisy, nonlinear and dispersive medium, is transformed into parallel scalar channels. Consequently, the NFT spectra (comprised of the continuous and discrete spectrum) have recently been suggested as more suitable signal degrees-of-freedom for information transmission in optical fiber. This thesis explores the potential of modulating the discrete spectrum alone. We propose a multi-eigenvalue communication scheme in which information encoded using only the locations of the discrete eigenvalues. Two encoding methods, multi-eigenvalue position encoding and a lower-complexity trellis encoding, are presented and shown to generate multisoliton signal sets with spectral efficiencies greater than 2.5 bit/s/Hz. These multisoliton signals do not do not undergo any pulse broadening, but we find that they are significantly limited by bandwidth expansion if the link length is not much smaller than a characteristic parameter dispersion length parameter. This limitation implies that encoding information in the eigenvalues alone is only meaningful when dispersion is very small and dominated by nonlinearity, e.g., close to the zero-dispersion wavelength at 1300 nm. We experimentally demonstrate, for the first time, the successful modulation and error-free detection of three-eigenvalue nonlinear frequency division multiplexed (NFDM) signals over an 1800 km link with Raman amplification and digital coherent receivers. The three eigenvalues are modulated by independent on-off keying signals, thus forming 3-bit NFDM symbols for nonlinear fiber transmissions. A novel bi-directional algorithm for computing the discrete spectral amplitudes is proposed, which addresses the significant problem of rounding errors inherent in previously-known techniques. We use the proposed method to obtain (for the first time) accurate spectral-domain noise statistics for 2-soliton signals using numerical simulation.