Fatigue life and reliability prediction of electronic packages under thermal cycling conditions through FEM analysis and acceleration models

Abstract

Finite element analysis (FEA) is employed frequently in studying the reliability of electronic packing. In real practice of electronics industry, accelerated life testing (ALT) has also been employed extensively on finding the life and reliability of an electronic product or component. Under the common concern for reliability, an analytical process combining FEA with ALT in studying the electronic packaging is proposed in the present paper. A Wafer-Level Chip-Scale Package (WLCSP) subjected to various JEDEC prescribed thermal cycling conditions is illustrated as a numerical example. First, parameters concerning package size and material property in FEA are assumed to be random to account for their uncertainties. Fatigue life distributions and quantitative reliabilities of the package under various loading conditions are found. The influence of parametric uncertainties on fatigue life distribution and reliability is investigated. Secondly, regression analysis is conducted with results from the finite element analysis in order to find the parametric values of several acceleration models. The most appropriate acceleration model is then selected. The fatigue life and reliability of the package under various loading conditions are predicted based on the acceleration model together with the result of FEA. Finally, in addition to parameters of acceleration and life prediction models, thermal-mechanical properties are taken into account, and sensitivity analysis is performed to improve the life prediction accuracy. The result indicates that, in FEA, although parametric uncertainty influences the fatigue life distribution of the package, it affects little about the package's mean life. It is found that, compared with other acceleration models, the Norris-Landzberg model is more appropriate for ALT analysis of the studied WLCSP. It is also found that the maximum value of the cyclic temperature has significant impact on fatigue life prediction of the package. Based on a modified acceleration model, the prediction errors are found to be within 1.59%.