Mentor Graphics mines design and test data to improve IC yield and failure analysis

DAC 2011, WILSONVILLE, USA: Mentor Graphics Corp. described a new use model for net-based CAA and scan test diagnosis that minimizes physical failure analysis (PFA) cycle time and maximizes yield across multiple products.

The approach, developed in cooperation with Samsung, employs the combined capabilities of the Calibre YieldAnalyzer product and the Tessent test and failure diagnosis tools to predict yield limiting nets. The method uses silicon defect distributions derived from a combination of certified foundry defect kits and scan test data.

“Our collaborative efforts with Mentor will allow a significant improvement in our ability to predict and understand yield by analyzing it at the net level,” said Kyu-Myung Choi, vice president, Infrastructure Design Center, System LSI Division, Samsung Electronics. “By incorporating critical area analysis (CAA) and volume scan test diagnosis in our yield analysis process, the physical failure analysis cycle time is significantly reduced and the impact of each PFA is significantly higher, allowing us to take faster corrective action and maximize customer yield.”

“We see many opportunities to improve time to market and time to volume by integrating information that has traditionally been isolated and incompatible across the larger flow,” said Joseph Sawicki, vice president and general manager of the Design-to-Silicon division at Mentor Graphics. “Our work with Samsung on net-based CAA is a great example of the potential of collaborative efforts with the foundry to improve designers’ understanding of design and manufacturing interactions and to increase their overall productivity.”

The net-based CAA solution first leverages certified foundry defect kits and net-based CAA data to help ensure the highest accuracy and granularity of net-level yield predictions prior to manufacturing an IC. Once in production, the Tessent Diagnosis product leverages layout information and production test results to refine the defect distribution data. This learned defect data is continuously fed back to the net-based CAA process to maintain high accuracy of net level yield predictions even in a highly variable manufacturing environment. This minimizes PFA cycle time and can be leveraged across the entire product line.

Samsung and Mentor will be jointly demonstrating net-based CAA capability in the Samsung booth, #1404 at the 48th Design Automation Conference, June 6-7 at the San Diego Convention Center.