Bridging DFM Analysis and Volume Diagnostics for Yield Learning - A Case Study

Abstract

Volume diagnostics introduces important means for yield learning as conventional techniques become more expensive and insufficient in identifying systematic yield limiters. Integrating DFM practices within the design flows requires faster identification and ranking of systematic yield limiters in the design. This paper presents a paradigm for identifying outliers in the fail signatures obtained from volume fail data using fail rate prediction from chip-level CAA analysis. Results from case study shows that a comparative analysis between predicted and observed fail rates can highlight potential yield limiters.