Is functional test necessary?

Functional printed-circuit-board (PCB) test has had a seesaw relationship with other board-test techniques. When surface-mount parts and ball-grid arrays first appeared, limited circuit-node access hindered the application of traditional in-circuit test (ICT) approaches, and it seemed that functional test would retain a vital role. Now, boundary-scan techniques are augmenting ICT, making it easy for manufacturers to locate structural defects before the final-test stage despite the lack of full bed-of-nails access. Earlier in the manufacturing cycle, optical and x-ray inspection systems are finding defects before the first application of electrical power.

Consequently, some structural-test proponents have hinted that some combination of the following approaches should be sufficient: full test of the individual ICs that will populate a board, board inspection, in-circuit test, and boundary-scan test. Only the paranoid, these proponents hint, need functional board test.

Those in the trenches are less sanguine. Bill Ecklow, who, as manager of embedded test technology, tests boards at Cisco Systems, told participants in a board-test panel October 10 at the International Test Conference 2002 (Ref. 1) that signal-integrity issues are limiting the effectiveness of in-circuit tests of many PCB components, including the high-speed serial data-communications circuits common on the boards that Cisco manufactures. For such circuits, an in-circuit tester's bed of nails acts as a series of antennas that conduct interference into a board under test or as antennas that act as inappropriate loads on PCB nodes.

Access, too, remains an issue, despite boundary scan. In an e-mail exchange after the ITC panel, Bob Stasonis, product marketing manager at Teradyne's Assembly Test Division, told me, "The use of functional test at the board or subassembly level for the commercial world is driven by test access, or really the lack of test access."

A classic example, he said, is "cell phones that, because of the size and type of electrical components, are only testable at the in-process stage by noncontact optical or x-ray inspection and by functional test. Similar examples can be cited for other technologies where the fault opportunities or escapes are significant for a specific type of process test. In the end, a comprehensive functional test is the only means to verify product functionality."

Gordon Robinson, senior member of the technical staff at 3MTS, a manufacturer of chip testers, suggested at the ITC panel that chip test plus structural board test could never ensure that a product would work satisfactorily. What's needed at conferences like the ITC, he said, is not more information on complicated ways of detecting mechanical assembly faults, but more information on detecting subtle interactions that keep products from working properly even when they are made perfectly from perfect parts.

In an interview after the conference, panel organizer and test-engineering consultant Ben Bennetts elaborated on this point: "In the chip world, there has been a definite move away from functional test once the manufacture of the chip moves into volume production. In prototype chip debug on first silicon, test objectives are based on both functional and structural test, but once there is sign-off on the functional behavior of prototype silicon, the emphasis turns to structural test only."

At this point, he explained, the chip engineer's viewpoint is, "Why do I need to constantly and continuously re-prove that I got the design right? All I need to do now is check that each manufactured instance of the chip is without manufacturing defect."

In the board-test environment, this strategy hasn't fared so well. Bennetts told me that although board development and production follows a similar prototype-to-production pattern, board manufacturers aren't following the chip world's structural-plus-functional to structural-only test path: "Board OEM companies are not so willing to commit to such a black-and-white solution. The cost if we get it wrong is much higher. There is still the desire and, some would argue, the overwhelming need to continue to apply functional checks on even mass-produced boards."

Kamran Firooz, general manager at Agilent Technologies' manufacturing test division (Loveland, CO), told me his views on why that's the case. "After in-circuit test, you know that the right parts are there. What you don't know is whether they will work together." Certain combinations of fast and slow gates or variations in component load characteristics can prevent components that meet their own specified propagation-delay and impedance specs from working together properly in a completed product. What prevents them from working together, he said, are design problems that can only be found at the functional-test stage.

If the conclusion is that functional test is required, how should it be applied? Combinational test—a single test system that performs both in-circuit and functional test—used to be appealing. Such an approach offers the benefit of minimizing test insertions.

That approach has fallen out of favor, said Firooz, primarily because in a combinational tester expensive in-circuit instrumentation sits idle during the long periods in which the tester applies functional tests. Stasonis quantified that point: "In general, in-circuit test programs are an order of magnitude faster than functional test programs. Consequently, adding any significant amount of functional test to a predominantly in-circuit program can compromise the production-line beat rate."

As an alternative, Firooz said you could use an in-circuit tester to cost-effectively provide functional tests of certain circuit blocks within a board. Ken Hallman, marketing manager at CheckSum, elaborated on this point in an e-mail interview: Many electronic assemblies such as terminators, filters, and interfaces can be effectively functionally tested with the use of in-circuit testers. For passive components, he said, an in-circuit test can provide the equivalent test coverage of functional test, especially if the test programmer takes care to use a subassembly's I/O pins to apply the tests.

Eric Starkloff, platform manager for PXI and instrument control at National Instruments, told me he still sees applications for combinational tests. "For many applications, a combined approach is very economical—the platform costs are shared and the system is compact so it takes up less rack or floor space." For other applications, he said, you should separate process test from functional test for several reasons: For example, having process test upstream from functional test can catch faults early, before you tie up expensive time in functional test; and in-circuit test can catch errors like short circuits that can be potentially damaging to a functional test system.

Monica Lobetti-Bodoni, design-for-test specialist at Siemens Mobile Communications (Milan, Italy), recounted for the ITC panel an odyssey among several conferences from NEPCON West in California to Productronica in Germany, hearing at each one that "No, this isn't the conference for board-test information." Further, she added, board test isn't a topic that excites much academic interest. Board test today, she said, invites a cocktail of test techniques, "but there are no formalized methods for choosing the cocktail ingredients" and determining the fault coverage that a particular recipe might provide.

Teradyne's Stasonis told me he sees some progress in this area: "With the development of process-analysis software tools over the past few years, one is at least able to now quantify the escapes from in-circuit or noncontact test techniques. These escapes can then be used to decide what to test at the functional test stage. These tools facilitate the linkage between the functional and in-circuit test engineering groups, which historically have had very little interaction and have not easily collaborated on overall test strategies."

One problem is that board test seems to get no respect. Kenneth Parker, board-test engineer and scientist at Agilent, told the ITC panel that people "marvel at the technology that goes into testing an IC with 50,000 transistors, but yawn at the prospect of testing a board containing hundreds of these ICs along with hundreds of analog components sprinkled in."

CheckSum's Hallman told me that "in the engineering food chain, test engineers are often thought to be inferior to design engineers." He doesn't buy that: "It can take more skill to figure out some other person's design than to create the design in the first place. If you don't believe that, ask an engineer to explain how a board he designed several years ago works; then ask a test engineer to explain the same circuit" after a similar time lapse. Nevertheless, the perception persists.

To achieve design-and-test harmony, Hallman suggests that companies offer financial incentives to teams that cooperate: If a design leads to high test costs, low quality, reduced sales, and curtailed financial rewards, "then even design engineers will take note of the burden they have inadvertently placed on the test-engineering department." He cautioned that there is a danger to involving design engineers too closely to the test process: "Design engineers have an inherent inclination to promote parametric test—an expensive and unnecessary step that leads to the time-consuming process of tracking down false failures."

Are designers amenable to adapting test strategies? According to consultant Bennetts, yes. He told me, "I lecture extensively on 1149.1 boundary scan: sometimes to board people and sometimes to chip people. I insist on telling chip people about boundary scan in terms of what it is and what problem it solves—a board-test problem. As a result, chip designers especially leave my courses knowing both the 'what' and 'why' of boundary scan. I have never had chip designers object to inserting boundary scan once they have understood the reason for it."

Concluded Starkloff in an interview, "We believe that there is a real opportunity to bring the worlds of design, design verification, and functional test closer together. This will enable test programs to be designed much more quickly as new designs are moved into manufacturing. Functional test will ultimately be viewed as an extension of a test strategy that starts in the earlier stages of a product's design."