Structural Similarity as a Method to Reduce Qualification Tests

Abstract

To efficiently select qualification and reliability monitoring programs, structural similarity rules for integrated circuit designs, wafer fabrication processes and/or package designs are currently used by the industry. By following the package structural similarity rules, the numbers of reliability qualification tests may be greatly reduced. However, when looking at the present rules it is clear that they are not reliably defined. For instance, geometrical parameters such as die-to-pad ratio are not quantitatively included and it seems that linear relationships are assumed. Besides that, these rules are mainly deducted from experience and industrial trial and error results, not from reliability physics. Driven by the present development trends of microelectronics (miniaturization, integration, cost reduction, etc) it is urgently needed to develop 'structural similarity rules' based on reliability physics (physics of failures), to meet the industrial development trends. In this study, we have used simulation-based optimization techniques to deduct such structural similarity rules. Parametric 3D non-linear FE models are used to explore the responses of the complete BGA family for both the thermo-mechanical and moisture-diffusion responses as function of six parameters among which the die-to-pad ratio and the body size. In this way, structural similarity rules are deduced which can be used to shorten design cycles. Even more, by using the accurate 3D nonlinear reliability prediction models, an Excel-based tool is created for package designers. By using this tool, the number of reliability qualification tests can be reduced. More importantly, possible failure mechanisms can be (better) understood and predicted