Empirical results on the relationship between die yield and cycle time in semiconductor wafer fabrication

Abstract

A negative correlation between die yield and cycle time is frequently hypothesized for semiconductor wafer fabrication. Methods that aggregate die yield and cycle time statistics over time are shown to exaggerate correlation coefficients. A lot-by-lot analysis of die yield and cycle time data from four volume manufacturing facilities is performed. The results indicate that the correlation coefficient is often statistically insignificant. Where the correlation coefficient is significant, outlying data points are checked for assignable causes and removed pending proper explanation. In addition, statistical models regressing die yield on cycle time are poor, and thus should not be used as the basis of decision-making in production control.