Accelerated estimation of spectral degradation of white GaN-based LEDs

Abstract

The intensity and colour (spectral) degradation of light-emitting diodes (LEDs) can significantly affect the reliability of demanding lighting products for which a certain level of intensity and colour homogeneity is required. Intensity degradation is sometimes characterized by manufacturers in a form of a mean half-lifetime, i.e. the time in which the initial intensity of LEDs decreases to 50%. However, additional information on spectral degradation in a batch of LEDs is required for designing and/or maintaining demanding state-of-the-art lighting products. For this purpose, a current-accelerated prediction of spectral degradation of white GaN-based LEDs is presented. The method is based on periodical automated spectral acquisitions of LEDs driven by above nominal currents. The obtained information on current-accelerated spectral degradation is then used for a model-based prediction of spectral degradation of LEDs driven by an arbitrary current. The proposed method was tested on 640 ultra-bright white GaN-based LEDs (from 5 to 20 cd) from six manufacturers. For this purpose, the predicted and measured spectral degradations at the application-specific nominal currents were compared. The method has proved useful for predicting spectral degradation of the eight tested batches by the exponential lifetime prediction model. The achieved up to sixfold acceleration and the mean absolute lifetime prediction error of 15% indicate that the proposed method is feasible for an accelerated estimation of spectral degradation of white GaN-based LEDs. As such, the method may be a valuable tool for designing and maintaining lighting products in numerous existing and emerging applications.