3.1 - The Clustering Effect on Defect Level Modeling

Authors: Jose T. de Sousa
Univ/Comp: Bell Labs.
Presenter: Jose T. de Sousa - sousa@research.bell-labs.com

******** Abstract:
A defect level (DL) model that considers the clustering effect is
proposed. It uses negative binomial statistics, completely unifying
yield (Y) and DL theories. It is well known that the clustering effect
produces significantly more optimistic yield estimates. This paper shows
that clustering also produces significantly more optimistic DL
estimates. This is one of the important factors that explain why in
reality the defect coverage requirement is significantly less stringent
than predicted by some of the previous models.