Limitations to estimating yield based on in-line defect measurements

To estimate yield loss based on data from in-line defect measurements, certain assumptions must be made so the data can be made to fit a given yield model. For the assumptions to be credible, inherent limitations due to detection, review sampling, and classification groupings must be understood and dealt with. Methodologies such as correlations of in-line measurement data to test data can be misinterpreted if issues such as multiple-failed die are not considered. This paper discusses some of the inherent limitations of in-line measurements and how they can affect the intended outcome of yield estimations.