The degradation of optocouplers in accelerated tests: A comparison of experimental results with two published models

Abstract

Commercially made optocouplers have been subjected to accelerated endurance tests at temperatures up to 100°C. The patterns of reduction in light output of the light emitting diodes in these devices varied widely, and were generally incompatible with two popular models. However, if the models were used to project long term degradation from early data (up to 168 hours), the resulting distributions of projected lifetimes were similar to the distributions of observed lifetimes. The models could be used as predictors of statistical parameters of groups of components, but not as predictors for individual components. The projected and observed distributions were in closest agreement for the earliest failures - these are the important failures in practical situations. The distributions of lifetimes varied markedly in nature, depending upon the precise definition of failure for determining the end of life. One definition of failure resulted in a rising failure rate, while an alternative definition resulted in a falling failure rate. This result draws attention to the fundamental importance of a clear definition of failure in endurance tests.