Microelectronic Yield, Reliability, and Advanced Packaging
Reliability is one of the major keys in product development. While reliability test are conducted in almost every manufacturing plant, the analysis of reliability test data is hardly rigorous, and engineers mainly rely on the softwares that come with the reliability test equipment to perform the test data analysis. Although this is usually sufficient, the underlying assumptions of the analysis are seldom known, and misleading conclusions might be resulted. Also, the sample size for a reliability test is generally determined from an engineering specification, and variation cannot be made when special circumstances arise. Furthermore, confidence interval estimation of MTTF and T50, outlier points identification from test data are usually not given. This could make the test data are usually not given. This could make the test analysis meaningful since point estimate can lead to erroneous decision, and so are the outlier points. In addition, a specified distribution, in particular, the exponential distribution is usually assumed in the data analysis. However, in practical problem, reliability test may be affected by other failure mechanisms. Thus, test data could be from mixture of distributions, and different models need to be identified and analyzed separately. Therefore, to ensure that the reliability test data can be analyzed accurately, the analysis must include sample size determination, parameter and confidence interval estimations, outlier point identification, and failure mode identification. Sample size determination is required so that desirable confidence level can be obtai4ned from test data with acceptable confidence interval. Outlier point identification is required so that undesirable data points can be eliminated and correct analysis for the remaining desirable test dat can be done. Failure mode identification is required so that each failure mode can be analyzed separately as they tend to have different life distributions. In this paper, reliability data analysis software developed by us will be presented that take into account of the above-mentioned, and hence an accurate and complete reliability test data analysis can be performed.
https://spie.org/Publications/Proceedings/Volume/4229?SSO=1
