Prediction of yield for flip-chip solder assemblies

Abstract

A model has been developed to predict the yield of solder flip-chip assemblies. Yield was found in relation to the mean and the standard deviation of the solder volume distribution, the assembly warpage, and the chip size. In this study, solder volume was represented by the diameters of the solder balls. The model identified failed joints using the Surface Evolver model and a regression model. The results showed that the volume variation corresponding to a diameter variance of 25 /spl mu/m for solder balls produced very low yields. Warpage greatly decreased the maximum variance allowed for 100% yields. For example, the diameter variance is /spl plusmn/11 /spl mu/m for 150 /spl mu/m solder balls; this value is reduced to 7.6 /spl mu/m with a warpage of 6.6 /spl mu/m across the diagonal of a 1/spl times/1 cm chip. In addition to volume variation and warpage, the pad radius also had an effect on the assembly yield. However, this effect is nonlinear. In some cases, larger solder balls could result in lower yields than smaller solder ball sizes with the same pad radius. Finally, larger chip sizes would not decrease yields significantly if warpage level were below 8 /spl mu/m. This model is a good guide for use in design of flip-chip assemblies, although it is not yet complete. Further development must include reliability modeling.