Model for Inverse Determination of Process and Material Parameters for Control of Package-on-Package Warpage

Abstract

Package-on-package (PoP) assemblies may experience warpage during package fabrication and later during surface mount assembly. Excessive warpage may result in loss of coplanarity, open connections, misshaped joints, and reduction in package board-level reliability under environmental stresses of thermal cycling, shock, and vibration. Previous researchers have shown that warpage may be influenced by a number of design and process factors including underfill properties, mold properties, package geometry, package architecture, board configuration, underfill and mold dispense and cure parameters, and package location in the molding panel. A comprehensive inverse model incorporating a full set of design and process parameters and their effect on PoP and PoP assembly warpage is presently beyond the state of the art. In this paper, data have been gathered on multiple PoP assemblies under a variety of assembly parameters. The packages have been speckle coated. The warpage of the PoP assemblies have been measured using a glass-top reflow oven using multiple cameras. Warpage measurements have been taken at various temperatures of the reflow profile between room temperature and the peak reflow temperature. Finite-element models have been created, and the PoP warpage predictions have been correlated with the experimental data. The experimental data set has been augmented with the simulation data to evaluate configurations and parameter variations, which were not available in the experimental data set. Statistical models have been developed to capture the effect of single and multiple parameter variations using principal component regression and ridge regression. The best subset variables obtained from stepwise methods have been used for model development. The developed models have been validated with the experimental data using a single factor design of the experimental study and are found to accurately capture material and geometry effects on part warpage. The results show that the proposed approach has the potential of predicting both single and coupled factor effects on warpage.