Statistical methods for stress screen development

Abstract

Stress screening during design, development, and production of electronic hardware is a quality improvement technique which can be employed to reduce defects in a product. However, due to the variety of electronic hardware types which may be screened and the number of stresses which may be applied for screening, there are no commercial standards which describe how to develop an effective stress screen. This paper describes a non-product-specific screen development technique which utilizes statistical analysis methods to achieve an effective and efficient stress screen. Statistical applications for various aspects of stress screen development are suggested, including Pareto analysis, Exploratory Data Analysis (EDA), Weibull analysis of time-to-failure data, comparison of means, analysis of variance (ANOVA), use of statistical process control charts (CUSUM, X-bar R), Duane plots of reliability growth, and use of the Poisson distribution for determining sample screen sizes. The techniques outlined involve test and analytical activities applied throughout product development; from first prototypes through to volume production. The use of statistical methods allows for development of an effective screen to remove defects and for an effective risk assessment of the effect of defects through numerical quantification of defect probabilities.