Abstract
The paper attracts the attention that the accelerated lifetime of commercial concentrator lattice match triple junction GaInP/GaInAs/Ge cells is better fitted by lognormal than Weibull distribution that has been adopted by most of the researchers in the field. A fair number of statistical tests are used to analyze a real-time dataset from accelerated life testing (ALT) that significantly favors the lognormal distribution. For comparison purposes, the Arrhenius-Weibull and lognormal stress relationships are used to predict the lifetime model under nominal conditions. They provide comparable estimates to the nominal meantime to failure (MTTF) and activation energy of the cells; yet, the two models possess different behaviors, especially at their tails and peaks. Moreover, an intensive Monte Carlo simulation is conducted to examine the distribution robustness towards censoring. The results again affirm that the censored samples of Lognormal are more efficient than those of Weibull in estimating the distribution parameters.