Comparison of Signal Distributions and Capacity Bounds in VLC with Signal-Dependent Noise

This paper investigates the capacity lower bound of single-input single-output (SISO) visible light communications (VLC) in next-generation system (6G). The overall noise is assumed to be additive white Gaussian, where its variance is partially dependent on the current input signal strength. The input signal is constrained by the peak and average power. Firstly, the entropy power inequality (EPI) lower bound and the input entropy lower bound are derived based on the maximum entropy distribution. Using this as a benchmark, a comparison is made between different distributions and bounds. Simulation results show that the maximum entropy distribution and EPI bound are not only suitable for high optical signal-to-noise ratio (OSNR) situation, but also valid at low OSNR case. Finally, the impact of signal-dependent noise is analyzed under the normalized noise power case. Simulations show that the signal-dependent noise will have an obvious impact on capacity.