Introduction

In optical wireless communication (OWC), the light intensity of a light emitting diode (LED) is modulated by a message signal. After propagating through the optical wireless channel, the light message is detected by a photodiode (PD). Key characteristics of the optical transmitter (Tx) and receiver (Rx) include their optical spectral response, electrical modulation bandwidth, radiation/detection patterns, optical power output of the LED, the photosensitive area, and the noise figure of the PD. The optical wireless channel has been shown to be a linear, time-invariant, memoryless system with an impulse response of a finite duration [21]. The primary characteristic of the channel is the path loss. An accurate representation of the light distribution in an indoor setup can be obtained by means of a Monte Carlo ray-tracing (MCRT) simulation. Channel modeling confirms that the path loss in dB is linear over logarithmic distance, and it ranges between 27 dB and 80 dB for line-of-sight (LOS) and non-line-of-sight (NLOS) communication scenarios in the considered setup [16, 66]. At high data rates, where the signal bandwidth exceeds the channel coherence bandwidth, the channel can be characterized as a frequency selective channel due to dispersion [67]. There is no fast fading, but the signal undergoes slow fading. Ray-tracing simulations show that slow fading can be modeled as a log-normally distributed random variable [16]. The delay spread of dispersive optical wireless channels can be accurately modeled by a rapidly decaying exponential impulse response function. Root-mean-squared (RMS) delay spreads between 1.3 ns and 12 ns are reported for LOS links, whereas RMS delay spreads between 7 ns and 13 ns are reported for NLOS links [21]. In order to counter the channel effect, single-carrier pulse modulation techniques, e.g. multi-level pulse position modulation (M-PPM) and multi-level pulse amplitude modulation (M-PAM), employ a linear feed-forward equalizer (FFE) or a non-linear decision-feedback equalizer (DFE) with zero forcing (ZF) or minimum mean-squared error (MMSE) criteria at the expense of increased computational complexity.