Moving beyond boundary scan and inspection
Steve Scheiber, Contributing Technical Editor -- Test & Measurement World, 8/1/2005
The most evident twin changes in the electronics industry in the past 20 years have been the increase of circuit functionality and the ever-declining access to board nodes to ensure that a circuit works.
Test engineers point to design-for-testability as one solution, but design engineers have never met that concept with enthusiasm. The same can be said of built-in self-test, which rarely covers more than about a third of possible failures. At one point, we thought we had the ultimate solution in boundary scan, but that, too, has been called into question—manufacturers are facing the limitations of boundary scan to verify a board's function as well as the likelihood that boundary scan makes boards more susceptible to security breaches.
So what do we do? How do we ensure proper circuit operation? Inspection has addressed some of the issues, like voids in ball-grid arrays, but it can find only a subset of possible faults. Hot mockup is time-consuming and extremely labor-intensive, and its results tend to be more qualitative than quantitative; also, mockup systems degrade as they age (called "hot mockup system wearout") until you can't tell if a failure comes from the system or the board under test. How do you schedule calibration/maintenance/replacement for hundreds of systems? On system failure? On a rotating basis? All at once?
Both OEMs and contract manufacturers have to get away from the notion that a single test strategy can address every situation. Contract manufacturers who truly want to be ready to accept any task that walks in the door must remain flexible enough to devise unique strategies specifically optimized for each situation rather than rely on a menu of predetermined choices. Those who specialize in certain kinds of boards will find the path somewhat less complicated—and less crowded.
Some new(ish) ideas are emerging. Some board-test manufacturers are turning to a variation of in-circuit emulation to address some of the problems. Today's device designers have to incorporate test routines in their silicon because it is the only way to determine if the device works. Ireland-based International Test Technologies has adapted emulation techniques to microprocessor-based boards by taking advantage of those device-level test routines, knitting them together the way in-circuit test programs were assembled years ago. Other companies are adopting similar approaches. Although not appropriate for every board category, on the subset of boards for which they work, these techniques are quite successful.
Mostly we need to take the action that experts recommend for all innovation—we must think outside the box. Because as the box grows ever smaller, much less of what we face every day will fit inside it.
|
