Analysis of accelerated LED degradation by statistical methods

In the fast developing LED industry, new generations of LED packages are reaching the market in small time spans. In contrast to that some luminaires, especially in street lighting, should be long lasting and uses the same package over many years. That is one reason why standards are developed to be able to provide lifetime values for the customers. The standard method of calculating the lifetime of an LED package uses the mean values of the luminous flux of a certain number of samples. In a time interval (for example 1000 hours) these values are plotted and fitted by a function, in most cases the exponential function of TM-21 [1] is applied. Our results of LED aging show that often there is no such simple function that is able to describe all the degradation curve. Especially at different conditions (temperature, forward current) the course differ and are not always smooth. Another method to investigate the LED aging is based on statistics. This has been successfully applied in the semiconductor technology (e. g. transistors). In that case every LED package itself gives information, because distributions are the main component of the analysis. The Weibull statistic is tested to the data at different conditions. In order to get a failure rate, the value of a certain percentage of the radiant flux has been taken to define this parameter. The acceleration in the aging test is driven by temperature and forward currents. Both parameters have other effects to the different parts of the LED package. Finally, the acceleration is investigated by the Arrhenius method.