Debug connects with manufacturing test to cut product recall costs

UltraSoC has kicked off a collaboration with PDF Solutions to build a system better able to use runtime information to identify devices that are likely to fail in the field and so reduce the impact of product recalls.

The combined system will take information captured from running systems using UltraSoC’s hardware-based behavioral monitors and feed that to PDF’s machine learning and analytics software that analyzes process and test information captured during manufacturing. UltraSoC claims the “fab to field” analytics framework is could reduce the impact of product recalls in sectors where device reliability is a primary concern, such as automotive industry, where recalls cost a total of $22bn.

More than a hundred semiconductor companies worldwide use PDF’s Exensio software, collecting yield, control, test, and assembly data from more than 21,000 machines worldwide. Models based on this data let the software monitor, diagnose, and identify manufacturing issues. To this, UltraSoC expects to be able to add information on the behavior devices while running in a target system and identify trends. The companies see the analytics framework being used used to generate alerts, actions, and system reports.

“Connecting to UltraSoC’s in-life monitors and data will enable us to extend our analytics and ML offerings to support a total preventive maintenance solution for semiconductor devices,” said Dennis Ciplickas, vice president of advanced solutions at PDF, said.

Rupert Baines, CEO of UltraSoC, added: “The value of quality – or conversely, the cost of poor quality – is too high to ignore. We have seen that, with increasing design and manufacturing complexity, plus system sophistication, product failures and recalls also increase. UltraSoC is already applying its intelligent hardware-based monitoring and analytics to a variety of in-life applications, including cybersecurity, functional safety, and performance optimization. Working with PDF Solutions allows us to tap into comprehensive manufacturing data and advanced ML technology. The resulting fab-to-field analytics framework will have enormous potential to help manufacturers understand the evolving picture of how their products are behaving in real life, and to predict field failures before they actually happen.”