- 學位論文
電子封裝體受熱循環應力之疲勞壽命分析與可靠度研究
Thermal Fatigue Life and Quantitative Reliability Analysis of Wafer-Level Chip Scale Packages
摘要
近年來，關於電子封裝體受溫度循環測試所產生應力、應變之有限元素模擬，通常將幾何尺寸、材料性質等參數設為定值，因此，經疲勞壽命預估公式代入軟體模擬出之應力、應變，所計算出封裝體疲勞壽命為ㄧ定値。然而，經由實際測試得知，封裝體之疲勞壽命往往具有相當的離散性，對此，本研究依實際加工之精度範圍，以適當隨機取樣的封裝體錫球尺寸代入有限元素模擬，來探討製造過程中不準確性對於疲勞壽命之影響，並以可靠度觀點加以描述。此外，本研究亦針對疲勞壽命預估公式中之參數變異情形，觀察其對疲勞壽命分佈及其可靠度。本研究結果顯示，錫球之加工誤差，使得封裝體疲勞壽命範圍介於1130至2403 cycles，因此對封裝體疲勞壽命的離散性產生一定程度之影響；而經由疲勞壽命預估公式變異，即可描述真實情況中，封裝體各參數之不確定性所造成疲勞壽命分佈之離散情形，且藉由疲勞壽命分佈情形，進而可評估封裝體使用之可靠度。
並列摘要
In study the reliability of electronic packages from mechanics point of view, the results of stress and strain obtained from finite element analysis are deterministic values. The fatigue life of the package predicted based on a modified Coffin-Manson equation or its equivalent is also a deterministic value. However, the real test outcomes reflect the package life is, in fact, a random variable following a certain probability distribution. Consequently, the tested lives of a certain type of package are frequently plotted in a Weibull probability paper but not a unique and deterministic value. In order to find out possible causes of the contradiction, some parameters involved in the finite element and life prediction analyses are considered random variables in the present paper. A certain type of wafer-level chip-scale package is used as an example. With regard to the finite element analysis, it is found, through parametric sensitivity study, the size of the solder bump affects the fatigue life of the package the most. Therefore, in the first stage of the nondeterministic analysis, the solder-bump size of the package is modeled a random variable in consideration of real manufacturing tolerance. Its effect on the random fatigue life prediction and quantitative reliability assessment of the package is investigated and discussed through statistical analysis of the simulation result. In the second stage of the nondeterministic analysis, parameters appearing in the modified Coffin-Manson equation are modeled as random variables in consideration of data scatter of the fatigue test result. Their effects on random fatigue life prediction of the package are investigated and discussed through either analytical derivation or simulations. The degrees of influence of all studied parameters on the fatigue life and reliability prediction of the package are discussed in detail in the present paper.