Introduction

The capacity of a communication system in a given channel with noise is defined by Claude Shannon as the maximum number of bits per unit bandwidth that can be successfully transmitted (assuming an infinitesimally small error) over a communication link [165]. In a practical communication setup, the electrical signal-to-noise ratio (SNR) requirement for a target bit-error ratio (BER) performance in additive white Gaussian noise (AWGN), the corresponding spectral efficiency, and the information rate are figures of merit for a given modulation scheme. In this book, the spectral efficiency of the modulation scheme is referred to as the uncoded bit rate per unit bandwidth, while the information rate is based on the mutual information of transmitted and received data symbols. Therefore, the information rate represents the maximum achievable rate of the considered modulation scheme which in the given noisy channel can be practically obtained when symbol shaping and forward error correction (FEC) coding are applied [166]. In this context, the Shannon capacity of the communication system is the maximum of the mutual information between the transmitted and received signals in the given noisy channel, where the maximization is performed over all possible input signal distributions.

In a line-of-sight (LOS) communication scenario, where the optical wireless channel is dominated by the LOS signal component, the root-mean-squared (RMS) delay spread of the channel is very small, while the coherence bandwidth is very large, which enables the transmission of signals with equally broad information bandwidth, and very high data rates. Such a scenario can be generally described as a frequency non-selective slow fading channel or a flat fading channel. However, in a practical indoor setup, non-line-of-sight (NLOS) signal components can be reflected by the objects in a room and detected at the receiver. In this case, due to the multipath component, the RMS delay spread of the channel is increased, while the coherence bandwidth is reduced, and it is typically comparable or smaller than the bandwidth of the signal.