Keep 'em flyin'

High-reliability, low-volume products, such as avionics and vectronics (space and other vehicle electronics), present different challenges from more familiar consumer products. They demand zero failures—either perfect manufacturing and test processes or sufficient redundancy to compensate for any system faults that emerge during operation. No one wants an airplane's guidance system to fail over the Atlantic or a communications satellite to stop serving its customers.

One way to increase reliability—reducing parts count—relies on a few complex parts rather than many simpler ones and introduces its own test challenges. Kamill Hilberth, CTO at Superior Electronics, a contract manufacturer and independent test-program developer for military and other high-reliability applications, explained, "Testing the parts on a complex board can produce diminishing returns, discouraging comprehensive functional test as well as bed-of-nails test. In addition, component cost and complexity represent a powerful incentive for introducing noninvasive verification to avoid damaging them [the parts] during test." So, manufacturers of high-value boards—like others in the industry—rely increasingly on inspection. "Because machine vision deals with structure rather than function," said Hilberth, "it can find faults hidden even to customized solutions like built-in test."

Yet, inspection of such PCBs also presents challenges. Double-sided PCBs containing multichip modules and other complex components thwart attempts at manual or automated optical inspection. Manufacturers of these systems rely heavily on thermal and x-ray techniques. Avionics, because of low manufacturing volumes and frequent design and engineering changes, can permit much wider variations of "acceptable" than more conventional products do.

"Under these circumstances," Hilberth said, "matching observed images with reference files resembles human signature recognition. No two signatures will be identical. Nevertheless, we need to develop accept/reject criteria through extensive image libraries and special filtering algorithms."

Designers must find ways to balance decision criteria to avoid failing good boards or passing faulty ones. To improve fault coverage, inspection systems must generally produce higher-resolution images.

Because increasing image resolution requires acquiring more information—a narrower field-of-view and more "snapshots," for example—inspection often involves trading off between test time and comprehensiveness. In high-volume applications where throughput is critical, manufacturers might risk missing a fault during inspection and finding it later. Avionics and other high-reliability products encourage the opposite approach. A "failure is not an option" strategy means accepting higher costs and longer test times to ensure that no failures survive to the final product.

Hilberth also advocates incorporating corresponding noninvasive inspection steps into field and depot testing and into so-called "health management systems"—monitoring techniques that report on the mission readiness of electronics during normal operation or as part of field maintenance. To employ inspection in this way requires a common fault database between the two venues to allow the manufacturer to correlate results and thereby more easily detect patterns or trends. Inspection can also reduce the number of hidden faults during manufacturing, making health-management decision engines more effective.

Manufacturers of avionics and vectronics must address the same issues as their more traditional counterparts. Although boundary conditions, quality criteria, and economic benefits differ dramatically, inspection increasingly provides viable solutions.