Capacity Estimation and Near-Capacity Achieving Techniques for Digitally Modulated Communication Systems

This thesis studies potential improvements that can be made to the current data rates of digital communication systems. The physical layer of the system will be investigated in band-limited scenarios, where high spectral efficiency is necessary in order to meet the ever-growing data rate demand. Several issues are tackled, both with theoretical and more practical aspects.

The theoretical part is mainly concerned with estimating the constellation constrained capacity (CCC) of channels with discrete input, which is an inherent property of digital communication systems. The channels under investigation will include linear interference channels of high dimensionality (such as multiple-input multiple-output), and the non-linear optical fiber channel, which has been gathering more and more attention from the information theory community in recent years.

In both cases novel CCC estimates and lower bounds are provided in this thesis. Intuition about the optimal signaling distribution is also provided, which is generally not the standard uniform for high spectral and energy efficiency communications. The practical part deals with tools to approach the CCC with real-life transceivers.

The constellation shaping concept is one such tool. More specifically, the probabilistic shaping concept is of interest to this thesis. A rate-adaptive solution is proposed for designing the mapping function of a probabilistic shaped coded modulation system, which allows for approaching the above mentioned optimal distribution in practice.

This results in increased energy and/or spectral efficiency for both linear and non-linear channels, but also increased maximum reach of the optical link at fixed spectral efficiency. The specific problem of phase noise in digital systems is also studied in this thesis. Phase noise, and particularly non-linear phase noise is especially detrimental to high-speed, high spectral efficiency optical communications.

As part of this work, a low-complexity solution is proposed for tracking, which is able to combat the combined effect of linear and non-linear phase noise in optical fibers, achieving close to the CCC estimate. The main contribution of the PhD project is providing engineers with limits on the data rates that current digital communication systems can achieve, and also with methods and insights for approaching those rates, thus interconnecting theory and practice.