Adaptive predistorter using NLMS algorithm for nonlinear compensation in visible‐light communication system

Abstract

A predistortion method using the adaptive normalised least mean square (NLMS) algorithm is proposed for the estimation of light-emitting diode distortion. The light emission characteristic could change nonlinearly depending on conditions of the physical environment, such as temperature variation, which cannot be ignored, especially in the case of high-power emissions. Instead of using the fixed values of the conventional memory look-up-table (LUT) for nonlinearity compensation, the proposed adaptive NLMS-based predistorter immediately estimates the distortion change according to environmental changes. The simulation results show that the proposed method has better performance than the conventional predistorter employing an LUT.Introduction: The visual light-emitting diode (LED) has a wavelength range of ∼380-780 nm at a very high-frequency band of 385-789 THz. If this band is used for communication, we can avoid electromagnetic interference from man-made signals, including conventional radio communication, which cannot reach this frequency range. An LED is a digitally controllable element with on/off switching control that is represented as a digital pulse. Human eyes cannot recognise LED blinking; rather, they sense that the LED is continuously turned on if it blinks at a sufficiently high speed. Because of this characteristic, the LED can be utilised simultaneously as lighting and as a communication link. Recent research for visible-light communications (VLCs) has focused on low-speed communications based on the on/off switching scheme [1].In order to apply VLC for wideband data transmission, we need to consider the linear variation of the luminance for high-order modulated signals with M-ary modulation, since the LED switching scheme has limitations in enhancing the transmit data rate. However, M-ary modulations have a critical disadvantage in a VLC system, especially for nonconstant envelope-modulated signals, because of the nonlinear response of the LED. Therefore, this drawback not only reduces the power efficiency of the system, but also degrades the bit error rate (BER) performance [1]. If we distort the bias signal intentionally to compensate for the nonlinearity of the LED luminance, the system performance could be improved [2]. The amount of distortion corresponding to the bias signal is calculated after measuring the characteristic of the LED luminance and storing it in a look-up-table (LUT). Using this LUT, we can constitute a predistorter. However, the performance is degraded in the conventional LUT-based predistorter when the physical characteristic is changed due to abnormal temperatures, LED ageing or similar physical changes.In this Letter, we propose a predistorter that can estimate the LED characteristics due to environmental changes using the adaptive normalised least mean square (NLMS) algorithm.

Precoded Chebyshev-NLMS-Based Pre-Distorter for Nonlinear LED Compensation in NOMA-VLC

“…In this section, we review some recently proposed adaptive pre-distortion techniques which mitigates the nonlinear characteristics of an LED in VLC. The discussion in this section would Adaptive pre-distortion [11] is a popular technique for compensating the varying nonlinear characteristics of an LED in order to invert and track the varying nonlinear LED characteristics.…”

“…Among adaptive pre-distortion techniques, the work in [11] assumes a scaling weight r k as a predistorter, which is multiplied with the input symbol sequence x k at the transmitter. Consequently, it passes through the LED nonlinearity A(.)…”

“…and measurement noise n k is added so as to form an estimated signall k . Consequently, the r k is updated as the following NLMS-based algorithm as proposed in [11]:…”

“…The reason for the choice of Chebyshev expansion in [12] is twofold: a) Chebyshev expansion is a nonlinear function of the input and is therefore more suited to invert nonlinear characteristics of LED, and b) the Chebyshev expansion optimizes the min-max approximation error optimality criterion which is one of the most important reasons for preferring Chebyshev expansion based pre-distortion over other orthogonal polynomialexpansion based pre-distortion techniques. This makes the Chebyshev pre-distorter a better and more suitable pre-distortion technique (as compared to lookup table and adaptive NLMS based pre-distortion as suggested in [11]).…”

“…Among pre-distortion techniques, the simplest technique would be to maintain a lookup table of the estimated nonlinearity. However, such static lookup tables lose their practicability due to varying LED characteristics [11] (due to device aging); hence there is a need for adaptive pre-distorters. Such adaptive pre-distorters have been suggested in the literature which are learnt using normalized least mean squares (NLMS) algorithm [11], and Chebyshev regression using NLMS approach [12].…”