Physical-aware systematic multiple defect diagnosis

This study presents a systematic defect diagnosis to identify `culprit physical features´ that are potentially responsible for yield loss. A `single location in-a-cluster´ technique is proposed to diagnose multiple defects that may not be diagnosed by traditional `single location at-a-time´ technique. A statistics technique, `analysis of variance´, is conducted to reduce noise from random defects. Simulations on five ISCAS´89 circuit demonstrate the effectiveness of the authors´ techniques. An experiment on an industrial design manufactured in 55 nm technology discovered a suspected culprit physical feature.