EDN's 20th Annual Innovation Awards
Category: Design, debug, and production test, yield analysis
Finalist: Tessent YieldInsight yield-analysis tool, Mentor Graphics
The Tessent YieldInsight software product is specialized for understanding and identifying semiconductor yield loss from production scan test data. Augmenting traditional methods, this solutions reduces the time to root cause by 75 to 95%, according to Mentor Graphics. To investigate the cause of yield loss, IC manufacturers collect large amounts of data from manufacturing and test equipment, and they use yield-management software to ask questions about the data. Based on this initial analysis, failure-analysis engineers use PFA (physical failure analysis) techniques such as emission microscopes and nanoprobing to identify the defects in specific failing devices. Corrective action such as a process, design, or test-program change is then taken to improve yield. The challenge is to know which questions to ask about the data and select the right devices for PFA that represent the systematic yield loss mechanisms. At advanced process nodes such as 65 nm, it is increasingly difficult to separate random and systematic defects, often resulting in wasted PFA and slower yield ramp.
Tessent YieldInsight instead mines the data with respect to designs and provides the right questions to ask. While the software visualizes results consistent with yield analysis practices, it also automatically identifies systematic defects and includes drill-drown capability to effectively pick die for PFA. The solutions leverages layout-aware diagnosis that provides a very rich and detail data set for each failing device. The statistical analysis is designed to especially work on this data set. Because the solution is based on diagnosis of production test fail data, it does not require manufacturing equipment data, which might not be available for fabless semiconductor manufacturers. Tessent Yield Insight customers have been able to accelerate their cycle time to root cause from months to days, reduce the amount of wasted PFA, and uncover previously hidden yield limiters representing the last 1 to 2% in high-volume manufacturing.