Investigation of Fatigue Life of Wafer-Level Chip-Scale Packages under Thermal Cycling Conditions by Acceleration Models
| Autor: | Po-Lun Chou, 邱柏倫 |
|---|---|
| Rok vydání: | 2012 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 100 Wafer-Level Chip-Scale Packages (WLCSP) have become more and more popular due to their light-weights and small sizes. To reduce time-to-market of a product as well as its development cost, the electronics industry often employs accelerated tests to find the life of the product. It appears that test outcome of the product’s life is not determin-istic but a random variable following a certain probability distribution. Therefore, in the present study, a finite element analysis taking uncertainty into consideration of a WLCSP subjected to various JEDEC prescribed thermal cyclic loading conditions is performed. A life prediction model of the WLCSP is constructed based on the analytical result. The study is divided into three parts. In the first part, a few parameters involved in the package size, material property are assumed random to account for their uncer-tainties. In particular, the influence of these uncertainties on fatigue life of the package is investigated. In the second part of the study, nonlinear regression analyses are carried out to find parametric values of several acceleration models. The Norris-Landzberg model is selected after comparing its performance with those of others. In the last part, in addition to parameters of the prediction model, thermal mechanical fatigue properties are taken into account and sensitivity analysis is performed for improvement of the pre-diction accuracy. The final result of the present study indicates the maximum value of the cycling temperature has great impact on life prediction of the WLCSP. In particular, for the studied WLCSP, the maximum temperature of 110℃ is suggested to be selected for dividing all JEDEC prescribed thermal cycling conditions into two groups when predicting lives of the package subjected to different conditions. The prediction errors are found to be within 1.59% by doing so. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|