Modeling and Mitigating LED Nonlinearity using Nonlinear ARX model with Wavelet Networks

Abstract

In this paper, a nonlinear autoregressive exogenous (NARX) model with a wavelet network is applied to model and compensate the nonlinearity of the LED in Visible Light Communications (VLC). The NARX model shows the ability to accurately describe the response of the LED. PAM-4 signal with a symbol rate of 5 Msym/s is used to demonstrate the performance of the NARX adaptive compensator. The eyediagrams show that this compensator can substantially improve the distorted signal. The complexity of the NARX adaptive compensator is relatively low, with only 15 units. This also facilitates the adaptive parameters updating process due to the small number of parameters in the NARX adaptive compensator.