Channel-Aware Multilevel Coded Modulation for Coherent Fiber-Optic Communications
The past decades have shown an ever-increasing demand for high-rate Internet services, motivating a great effort to increase the spectral efficiency of optical networks. In general, fiber-optic links are non-Gaussian, and in contrast to additive white Gaussian noise (AWGN) channels, there is no standard framework for quantifying fundamental limits or designing capacity-approaching coding schemes for such channels. In this thesis, some steps are taken toward this challenging goal by first developing a channel model for fiber-optic links and, second, using an information-theoretic design framework to investigate joint design of forward error correction and multilevel modulation, so-called coded modulation (CM), techniques for these channels.
We extend the signal statistics of highly nonlinear single-polarization fiber-optic links with negligible dispersion to the polarization-multiplexed case. Taking chromatic dispersion into account, we derive an analytical discrete-time model for single- wavelength, polarization-multiplexed, non-dispersion-managed (non-DM) links. According to this model, for high enough symbol rates, a fiber-optic link can be described as a linear dispersive channel with AWGN and a complex constant scaling.
We exploit the proposed channel model for highly nonlinear fiber-optic links to devise a new channel-aware multilevel CM scheme based on the minimization of the total block error rate. We introduce a CM system with an N-dimensional constellation constructed from the Cartesian product of N identical one-dimensional constellations. The multidimensional scheme shows better trade-off between complexity and performance than a one-dimensional multilevel CM scheme.
By invoking the introduced channel model for the dispersive non-DM links, we present a four-dimensional CM scheme, which shows a better trade-off between digital signal processing complexity and transparent reach than existing methods. This CM scheme together with a probabilistic signal shaping method is used to devise a rate-adaptive scheme with a single low-density parity-check code. The performance evaluation of the proposed CM scheme for a single-channel transmission fiber-optic system justifies the improvement of the system spectral efficiency for a wide range of transparent reaches, observing more than 1 dB performance gain compared to existing methods.
multidimensional set partitioning
low-complexity de- tector
nonlinear phase noise
Room EE, 6th floor, Hörsalsvägen 11, Department of Signals and Systems, Chalmers University of Technology,
Opponent: Professor Frank R. Kschischang