In this work we present a knowledge-based qualification approach for solder joint reliability of Ball Grid Array (BGA) component used in server systems. Servers experience very few on-off power cycles during their lifetime. The primary source of BGA thermo-mechanical damage in these systems is, therefore, the temperature fluctuations during use. We employ measured server CPU die temperature during field usage and a novel physics-of-failure metric to define the requirements for qualification in temperature cycling accelerated tests. The result of this computational/empirical approach is compared against the result of the industry-standard based empirical acceleration model. The key conclusion of this work is that standard empirical acceleration models have difficulty accounting for small temperature fluctuations and lead to significant overestimation of the risks. On the other hand, the proposed computational/empirical approach provides a physically meaningful estimation of risk.

This content is only available via PDF.
You do not currently have access to this content.