Packaging optimization to mitigate stress induced parameter shift in precision devices

Stress induced parametric shifts caused by packaging are a major source of yield loss in precision devices. Package stress induced parameter shifts can be decomposed into two parts, a mean shifting component and a distribution broadening component. The mean shifting component is systematic and can be compensated in the circuit design phase or by laser-trimming thin film resistors if package stresses are well characterized. However, the distribution broadening component is random in nature due to the random distribution of filler particles against the die surface. This component is hard to handle by design compensation and laser trimming. This work explores the root cause of the random component by a stochastic finite element method (FEM). The epoxy mold compound is modeled as a two-phase composite material consisting of randomly distributed filler particles in an epoxy resin. A predictive methodology that accounts for filler effects and material property mismatch of the fillers and the resin is studied for the first time. The effect of a compliant coating on top of die was simulated by the FEM. The residual stress effect of the compliant coating layer was also accounted for in the simulation. The present work thus incorporates both the systematic and random components of stress in silicon, and suggests yield improvements through package and assembly process optimization. A methodology of generating a conformal FEM mesh of a composite material with a large number of close-packing spheres in a matrix is also presented. This methodology overcomes the difficulty of Boolean operation failures in commercial FEM software packages and provides a feasible way to account for the random distribution of fillers in the mold compound.