Event-based use conditions method for thermo-mechanical reliability risk assessment

Abstract

The traditional approach of thermo-mechanical (T-M) reliability modeling is based on power cycle events. This approach is not useful for products which are rarely powered down, because power cycles alone do not capture all the reliability stress from temperature variation over these products' use life.This paper describes a methodology to determine the temperature cycle requirements for products like smartphones and tablets which accounts for the temperature variation associated with usage events, which we call "mini-cycles".The T-M model is based on the distribution of individual users' histories as a series of events over time, which is then translated into a temperature vs. time trace for each user. These temperature traces are then used as the main inputs to T-M models, for example using the Norris-Landzberg (N-L) acceleration model to evaluate solder damage for each user. Results are summarized in a distribution of T-M damage across all users. This new methodology improves the understanding of thermo-mechanical reliability requirements due to the impact of "mini-cycles".KEY WORDS: ther mo -mechanical, solder joint reliability, event-based use conditions, knowledge-based qualification. INTRODUCTIONPresently, the majority of industry uses either JEDEC or power cycles [1, 2] to determine the package temperature cycle requirements. Power cycles include both on-to-off and on-to-suspend event cycles. The temperature cycle (TC) requirement for a product is determined by assigning temperatures to each distinct, time-weighted state (on, suspend, off), and accounting for the number of events over the product lifetime as the input to the T-M reliability model, which may be based on Norris-Landzberg (N-L) model. However, for products that are not regularly power cycled, like smartphones and tablets, power cycles alone do not capture all the reliability stress from temperature variation over time. There is significant temperature (T) variation associated with application (app) usage which should be accounted for towards determining the TC requirement.Usage events, such as phone calls, represent a temperature vs. time behavior which we call a mini-cycle, to distinguish it from power cycles. A distribution of actual user app traces can be converted by modeling to a sequence of mini-cycles. It has been shown [3, 4] that mini-cycles may have a significant impact on T-M failure. The main subject of this paper is to convert user histories of events to T vs. time traces, which can be used as inputs for T-M modeling.New data was acquired for this work because previous user behavior studies [5,6] did not provide the detailed usage vs.