1. Signal-Dependent Noise for B-Modulation NFT-Based Transmission
- Author
-
Jaroslaw E. Prilepsky, Maryna Pankratova, Anastasiia Vasylchenkova, Sergei K. Turitsyn, and Nikolay B. Chichkov
- Subjects
Physics ,Discrete mathematics ,symbols.namesake ,Fourier transform ,Modulation ,Gaussian noise ,Domain (ring theory) ,Spectrum (functional analysis) ,symbols ,Time domain ,Covariance ,Noise (electronics) - Abstract
The nonlinear Fourier transform (NFT) is a potentially promising way to mitigate nonlinear signal distortions in fiber-optic communication systems [1–4]. Recently, the new approach named the b-modulation was introduced [1] that allows to control the temporal duration of NFT-generated signals. In the "traditional" NFT-based communication, the nonlinear spectrum, r (ξ) is used to modulate data. Then the signal duration depends on the signal energy and is not predetermined and controlled. The utilisation of the NFT scattering coefficient b (ξ) gives a possibility to control the pulse duration in time domain under condition that the Fourier spectrum of b (ξ) is localized. In particular, the OFDM format satisfies that condition. It was already shown that the properties of the effective noise in the NFT domain differ from a circular Gaussian noise. More specifically, such a noise is signal-dependent and non-circular [2, 3]. It was shown in [4] that the covariance B 1 and pseudo-covariance B 2 characteristics related to b-coefficient of NFT spectrum depend on the propagation distance. Here, we extend our previous work and examine noise characteristics for a novel b-modulation transmission introduced in [1]. We demonstrate that the variances behaviour is generic and do not depend on a specifics of NFT modulation. The presence of amplifier spontaneous emission in the optical link translates into the effective noise in the nonlinear spectral domain as an effective additive term N (ξ, b (ξ, 0)): $e^{-4 i \xi^{2L}b}(\xi, L) = b(\xi, 0) + N(\xi, b(\xi, 0), L)$ , where $\xi$ is the “nonlinear frequency” (NFT spectral parameter): $\langle N (\xi)N^*(\xi ')\rangle = 2DL\pi \delta (\xi -\xi')B_{1}(\xi, L), \langle N (\xi)N(\xi')\rangle = 2DL\pi \delta (\xi - \xi ')B_{2}(\xi, L) $
- Published
- 2019
- Full Text
- View/download PDF