Rapid Assessment of BGA Fatigue Life Under Vibration Loading

Thumbnail Image


umi-umd-3802.pdf (1.92 MB)
No. of downloads: 3496

Publication or External Link






Ball Grid Array (BGA) packages are a relatively new package type and have rapidly become the package style of choice. Much high density, high I/O count semiconductor devices are now only offered in this package style. Designers are naturally concerned about the robustness of BGA packages in a vibration environment when their experience base is with products using more traditional compliant gull or J leaded surface mount packages. Because designers simply do not have the experience, tools are needed to assess the vibration fatigue life of BGA packages during early design stages and not have to wait for product qualification testing, or field returns, to determine if a problem exists. This dissertation emphasizes a rapid assessment methodology to determine fatigue life of BGA components. If time and money were not an issue, clearly one would use a general-purpose finite element program to determine the dynamic response of the printed wiring board in the vibration environment. Once the response of the board was determined, one would determine the location and value of the critical stress in the component of interest. Knowing the critical stress, one would estimate the fatigue life from a damage model. The time required building the FEA model, conducting the analysis, and post-process the results would take at least a few days to weeks. This is too time-consuming, except in the most critical applications. It is not a process that can be used in everyday design and what-if simulations. The rapid assessment approach proposed in this research focuses on a physics of failure type approach to damage analysis and involves global and local modeling to determine the critical stress in the component of interest. A fatigue damage model then estimates the life. Once implemented in software, i.e. the new version of CALCE_PWA, the entire fatigue life assessment is anticipated to be executed by an average engineer in real time and take only minutes to generate accurate results.