Modelling reliability of flip chip on board assemblies implementing a correction function approach comparing analytical and finite element techniques

To determine the reliability performance of electronic components, environmental tests, or accelerated life tests, are used to apply stresses to electronic packages that exceed the stress levels experienced in the filed. In theory, these elevated stress levels are used to generate the same failure mechanisms that are found in the field, only at an accelerated rate. Therefore, an acceleration factor is typically used to correlate (extrapolate) the accelerated life testing data to a field failure rate for a specified use condition. Often times this data is time consuming and expensive to obtain , hence a need exists for reducing the time to data for electronic components in reliability testing. A methodology is presented whereby existing reliability data can be leveraged to obtain" correction functions\´ which can be used to modify a mean time to failure, MTTF, estimated analytically or numerically. A suggested analytical model is presented in addition to the statistics based methodology that can be used obtain correction functions. The correction function approach is similar to approaches used for modifying fatigues strengths in engineering alloys. Fatigue strengths or endurance limits are modified to account for physical differences between the actual parts in that were used to obtain the fatigue data. The methodology presented allows for the use of numerous correction functions to adjust estimated life times of component level assemblies based on key correction factors that account for effects difficult or impractical to incorporate in the base prediction models. The methodology is effective in that it can leverage the utility of the life prediction enabled by finite element modelling. The potential correction factors are presented in a fishbone diagram accounting for effects such as substrate metallization, underfill delamination, solder joint voids, underfill voids, intermetallic thickness, etc. Using existing reliability data, the correction functions are determined via multiple linear regression analysis. To illustrate the utility of the life prediction methodology, a case study is presented for flip chip on board assemblies. The uncorrected fatigue life of the solder interconnects is estimated using a trilayer stack analytical model predicting plastic s- train and incorporating correction functions for the glass transition temperature of the underfill, an area ratio for the solder joint interconnect pads, and the substrate bond pad metallization.