This paper describes an experiment in which it was determined that properly designed burn-in can not only eliminate a weak subpopulation, but also stabilize the healthy one. Accelerated life tests have been carried out on 30W sign lamps to show that the time-to-failure pattern is bimodal. The accelerated life data have been used to design burn-in process parameters, and to determine the deterioration caused in the healthy subpopulation by the chosen burn-in. The results showed no decrease in mean life of the healthy lamps. Rather, increased stability (lower standard deviation in life) has been observed. The burn-in procedure with parameters 240V and duration in the range 15-20s was successfully applied to the 30W sign lamps in order to eliminate the weak subpopulation. Due to fast turnaround of the procedure it can be implemented at the end of the production line. The high temperature thermal treatment of filaments during burn-in seems to relax structural stresses caused by fabrication process. The fact that the power coefficient obtained here differs from the one quoted by GE Technical Information TP-110 as well as by IES Lighting Handbook (Reference Volume 1984) should alert users of accelerated reliability tests to be wary of uncritical application of experimental factors taken from the literature, including this paper.
Read full abstract