Digital signal processing in nonlinear frequency division multiplexing systems

Abstract

The field of nonlinear Fourier transform (NFT) and nonlinear frequency division multiplexing (NFDM) has gained attention in the community. The NFT framework decomposes the nonlinear Schrodinger equation (NLSE) into parallel channels and decomposes the signal into nonlinear spectral components, which can be obtained by solving the Zakharov-Shabat problem from the Lax pair operators corresponding to the NLSE. In this case, the NFT/inverse-NFT (INFT) operators transform the signal between time-domain and the nonlinear frequency domain. Hasegawa and Nyu first proposed to use the NFT concept for nonlinear fiber transmissions in which information can be transmitted on the eigenvalues of the nonlinear channels without interfering with each other [1]. The concept is further combined with advances in digital coherent technology and generalized into NFDM. NFDM is considered a generalized theoretical framework for nonlinear fiber transmissions in the sense that continuous spectrum resembles conventional orthogonal frequency division multiplexing (OFDM) while discrete spectrum (eigenvalues) resemble solitons. In practical transmission, various impairments like fiber loss, optical amplification, and amplified spontaneous emission (ASE) noise significantly degrade and complicate the transmission of NFDM signals. Characterizing the impairments and exploit relative DSPs is important in the development of the NFDM system. This thesis focuses on discrete-spectrum modulation, analyzing interaction distortion among neighboring solitons during transmission. Various DSP schemes are proposed which includes designing a specific soliton-train on the transmitter side and incorporate with receiver multi-symbol DSPs to compensate inter-symbol interference after transmission.