Go/no-go testing is no-go

E-mail

The time of pass/fail testing is history. Although it's still necessary to weed out bad parts, that's only a small component of the modern test process.

Janusz Rajski, chief scientist and director of DFT engineering at Mentor Graphics, said in a recent interview that adaptive test and advanced diagnostics are becoming critical as processes shrink. Rajski, who serves as program chair for this year's International Test Conference (October 21–26, Santa Clara, CA), said attendees at this year's ITC (www.itctestweek.org) will learn specific techniques they can apply to cope with the “staggering complexity and small feature sizes of 65-nm devices.” (Read the interview.)

The goals of a modern test process are twofold (at least). One goal is to improve test coverage, and that's where adaptive test comes in. Rajski described adaptive test as an approach in which you perform analysis as well as test to determine the most frequent ways in which devices fail; based on your analysis, you try to develop more effective subsequent tests.

Another goal is to improve design and production processes, which is where advanced diagnosis comes into play. Rajski said that technologies are emerging that provide links to the physical domain. Those links enable diagnosis down to the level of specific layout features to help in the calibration of DFM rules and in the development of design guidelines.

The key lesson from emerging test technologies is that test yields value. In an online column, Dr. Fang Xu, senior scientist at Teradyne's Semiconductor Test Division, makes the case that even a simple pass/fail test “is at least as important and valuable as any other stage in production,” adding that screening good parts from bad “is not fundamentally different from extracting silicon from its dioxide in order to make wafers.” (Read the commentary.)

As engineers deal with the staggering device complexity that Rajski cited, new challenges are emerging. According to Tom Farkas, president and founder of Metrikos, the focus today is shifting to test of devices that acquire environmental or other external nondigital information, with analog stimulus in and bits out—often over an RF link. As described in this month's cover story, Farkas is working with the Bennington Microtechnology Center to develop test strategies for such devices.

However these strategies develop, you can be sure that significant benefits will come from the effective analysis and use of test results. Also in this month's cover story, BMC executive director Henry Klim emphasizes, “There is significant value in acquiring physical test data from functioning devices.”