The $250 million mistake.

Executives like John VanNewkirk usually find themselves courted by Wall Street, not running low-cost sellers of test systems. Yet that is exactly what the president and chief executive of CheckSum (checksum.com) is doing. Armed with degrees from Brown and Harvard, VanNewkirk is applying lessons learned as a management consultant with Bain & Co., where he developed and implemented growth strategies for Fortune 500 companies, to the Arlington, WA-based company, which he and an investor group acquired in 2003. CheckSum is proving less can be more. With a product line that includes ICT, combinational and functional testers starting at $10,000, the company claims an installed base of over 2,500 systems. VanNewkirk spoke with CIRCUITS ASSEMBLY in April. Excerpts (see full text at circuits assembly.com/cms/content/view/1514):

[ILLUSTRATION OMITTED]

CA: What is the biggest (read: most expensive) mistake test engineers are making day-in and day-out?

JVN: Electronics manufacturers are wasting money by running every test job across traditional in-circuit testers. In the majority of cases, the capabilities of "big iron" in-circuit testers are not needed. Ironically, often the more complex the board, the less the capabilities of traditional ICT are actually applicable.... [P]rogramming and fixturing costs are the largest part of the total cost of ownership of ICT. There is an alternative: use low-cost ICT for as many boards as possible and use the expensive ICT only when you really need its capabilities. Test is waste. We're not adding value to products, merely detecting defects in our processes.

CA: According to your estimates, $250 million is spent needlessly each year on test. Do you think test engineers feel the same way?

JVN: Probably not. First, the waste is invisible to most because it's sprinkled throughout standard methods that have worked for years. Test engineers get beat up when yields drop or when defects escape, but don't get much credit for dealing with new technologies or reduced test budgets. Most have big iron testers, so they're going to use them. They don't think about the operating costs. Most contract manufacturers care even less, because they just pass the costs on to their OEM customers in the form of non-recoverable engineering (NRE). Fixture, program and support costs seem small when viewed on a project-by-project basis, but add up to a more significant cost than the tester itself.

Second, test engineers haven't fully recognized the implications of the fault spectrum shift. Most engineers agree that digital devices don't fail like they used to, but haven't recognized that digital backdrive doesn't help them get the job done on most boards, even though it is expensive and hard to implement. It is very difficult for a test engineer to reduce test, even when the test isn't doing them any good.

The most successful manufacturers have analyzed their fault spectrum and focus their efforts on testing for defects that actually occur, addressing the root causes of these defects. Simple assembly and soldering errors make up the majority of faults.... By our calculations, the entire AOI, AXI and boundary-scan industries can be funded by eliminating the waste of traditional big-iron ICT.

CA: With studies forecasting increased numbers of BGAs, gate arrays and FPGAs, how will higher complexity coupled with fewer nets affect your thesis?

JVN: Actually, increased complexity is one reason manufacturers should spend less on ICT. Increased board densities, higher operating speeds and device complexity invariably lead to loss of electrical access. Increasingly complex digital and mixed-signal chips such as SOC make developing the digital vector tests used by traditional ICT extremely expensive and time-consuming--assuming they can be developed at all. Today's devices are increasingly reliable. Why write test vectors for ICT when it doesn't add real fault coverage? We estimate about 80% of today's boards could be effectively tested on low-cost ICT [without] a change in board yields.

CA: Are board speeds reaching the point where test pads are no longer viable and should not be designed in?

JVN: Electrical access is disappearing as operating speeds increase and circuits grow increasingly dense. The loss of test pads and access is another reason why digital vector-based testing is less useful. If you can't access all the nets to test a device, test coverage is going to suffer. Yet we believe ICT will remain the most effective and economical means to inspect for manufacturing faults. AOI and AXI, while promising, remain augmenting--not replacement--technologies.

CA: What recommendations do you have for storing fixtures so as to avoid bent or broken pins, etc.?

JVN: [M]any customers don't know how to take care of the probes, which really are the critical element in ICT tester reliability. Sub-0.075" test probes are dramatically more fragile; simply brushing them with the same stiff brush used on the more robust 0.100" probes can bend or break the smaller probes. [Manufacturers] need to treat fixtures for what they are: precision tools with complex mechanical and electronic components that need periodic care and repair.