Page 49 - Building A Succesful Board-Test Strategy

What Is a Test Strategy? 35

or functional testing, or may spot-check the contractor's process by testing
lot samples, PC system manufacturers and other makers of "commodity'" pro–
ducts often purchase tested boards directly from third parties, performing only
burn-in or other environmental stress screening (ESS) and system-level testing
before shipment.
Sometimes, contract manufacturers do everything from board assembly to
boxing systems and shipping them to distributors and customers. In those cases,
the system "manufacturer" (the company whose name goes on the product)
must trust that test, repair, and other quality control steps produce high-quality,
reliable systems.
At what level does test occur? Are bare boards tested before assembly? Do
loaded-board tests examine individual components, or do bed-of-nails access lim–
itations require resorting to logic clusters? Does the board contain independent
functional units for functional testing, or must a test exercise the entire board at
once? Are parts of the board testable only by one technique or the other? Do some
boards or parts of boards defy conventional testing altogether, requiring inspec-
tion or a comprehensive self-test?
Bare-board testing is not expensive, but it does require investing in equip-
ment, floor space, work-in-process inventory, and people. Most companies buy
their bare boards from suppliers who test the boards and ensure their quality.
Therefore, board vendors should pay severe penalties if any of their supposedly
good boards fails.
Sometimes leaving bare-board test with the vendor is inappropriate. Testing
very complex boards that demand specialized test techniques, such as the 50-layer
monster described earlier, usually requires work by the board designer—in this case
the system manufacturer.
At times a manufacturing-process decision demands an unusual test strategy.
In an attempt to minimize environmental damage, one company abandoned freon-
based board cleaning, adopting an aqueous technique instead. Unfortunately,
deionized water could not dissolve fluxes from the soldering step, so the engineers
resorted to organic-acid fluxes, which represent a serious reliability problem if left
on the board. Therefore, the company had to create a test strategy that involved
testing the boards again after cleaning.
Is it possible to test boards automatically during assembly? This type of test
would catch through-hole leads crumpled during automated insertion, for example.
Although many manufacturers have considered this approach for a number of
years, few, if any, actually do it.
Can the test time meet throughput requirements? High-volume production
often demands that a board come off the line every, say, 30 seconds. That means
that no holdup in the line can last more than that 30 seconds. The manufacturer
must "spend" that time very carefully. Meeting this specification might require
resorting to sampling techniques or testing only "trouble spots" if overall yields
are high enough to permit that approach.
Is testing necessary at the board level at all? As process yields from board
production increase, more and more manufacturers are asking this important