Enterprise-wide Yield Management

In this age of outsourcing, partnerships and global cooperation, it is no surprise that the process of yield management is moving toward a more global solution that can be accessed by people in many different places at different times. Simultaneously, the increased risk associated with yield loss, missed time-to-market targets and carrying excess inventory point to a serious need to better manage the supply chain.

It was these issues that motivated the founders of Syntricity Inc. (San Diego) when they began developing enterprise-level software that could be used by fabless firms to track yields of their devices after they were manufactured by the foundry partner. Now the software is being used by fabless companies, foundries and IDMs to track device yields and provide a web-based yield improvement methodology. The software essentially combines the supply chain logistical needs of global corporations with the engineering toolset needs of yield engineers.

"Yield management used to be treated as an engineering, workstation-based application," said Jeff Teza, Syntricity's CEO. "Today, using web-based, browser-based software, we can create an environment where the engineer still has the tools to determine the cause of yield variations, while other parties such as purchasing, management, customers and supply chain partners can easily access that yield information via the Internet."

The latest version of the software, dataConductorEP, may be customized to deliver reports to various parties in the format that they choose. For instance, a fab manager can look at equipment yield impact and utilization, while a design engineer might be looking to see if the first lots of a design yielded as expected.

The Internet application hosting service, dataConductor.com, is a web-based data management, analysis and report-generating solution. The network is monitored 24/7 for maximum security and uptime. (Source: Syntricity)

"Today, each element along the supply chain has its own tools and techniques for managing these problems, and one of the big issues is always communication between them, about what exactly they are measuring, what they are looking at and whether they are in agreement," Teza explained. "We try to resolve these issues by building into dataConductorEP the reports and analytics, but also a way to collaborate around all those shared reports and other documentation." One of the ways communication is facilitated is by access-controlled reports and annotations — sort of an electronic version of Post-it Notes that allow comments to be made on shared documents.

The yield management system automatically monitors for such events as exceeding a minimal yield threshold or other triggers. It can then send an e-mail message to the engineer that has a link to the report showing the live data and results of a first-pass root-cause analysis. To speed recovery from a yield-impacting event, the reports can be made as complex as necessary, to drill down from raw data to lot data to wafer maps, showing yield as a function of critical electrical parameters, etc., and do any number of analyses. It gives yield engineers the methods they need to drill down as far into the data to resolve the yield issue, thereby resolving problems as early as possible, while alerting manufacturing personnel to make accommodations while the problem is being investigated.

When it comes to new product introductions and yield ramping, the ability to forecast yield is critical. "It's not unusual to have someone who is building a cell-phone chip to be ramping the yield from 20 to 80%, so planners all along the supply chain need to prepare and be ready to deliver the appropriate amount of material at the right time," Teza said. "We can collect measurements on wafers in manufacturing from in-line through electrical test, probe and final test, and look at all the results. And we can correlate between an early measurement in the supply chain and what we expect the final yield to be." This type of forecasting, without these capabilities, is not only risky, but it changes day to day, requiring continual updates for all parties along the supply chain.

To date, dataConductor EP data management platform has been deployed at more than 20 customer sites worldwide, and is used by such companies as Chartered Semiconductor, Cirrus Logic, eSilicon, Marvell, Motorola, Silicon Labs, Sun Microsystems and Qualcomm. Because of its web-native roots, the software has the advantage of appearing as an enterprise system to a corporate IT department. It is compatible with other enterprise software systems through Java J2EE standards.

For additional information on yield management, go to www.semiconductor.net/yield