Improved reliability prediction through reduced-stress temperature cycling

Abstract

Standards for conditions and criteria of reliability stresses, in the absence of proven reliability models, have long been based on the capabilities and historical performance of processes and materials. A widely accepted model for temperature cycling has evolved in this decade. The author uses this model and case studies of specific failure mechanisms to show that the use of the commonly accepted condition C in temperature cycling can create unreasonably high acceleration of thermal stresses, possibly masking failure mechanisms more relevant to field applications. This often results in the pursuit of corrective action for mechanisms unlikely to occur in application environments, and may actually prevent detection of mechanisms more likely to occur in the field. For temperature cycling of present technology encapsulated packages, it is concluded that 1000 cycles of condition B is a superior criterion to 500 cycles of condition C. As an extension of this discussion, it is proposed to design ongoing and future reliability evaluations around evolving understanding of the physics of failure and the real needs of applications.