Page 60 - Building A Succesful Board-Test Strategy

46 BUILDING A SUCCESSFUL BOARD-TEST STRATEGY

analysis. Visual inspection on a through-hole board, for example, might spot solder
splashes from a poor wave-solder process or components that are obviously missing
or backwards.
Determining the success of any test strategy requires examining the effec-
tiveness of each step, then making adjustments as required. Not long ago, a loco-
motive manufacturer discovered that, after assembling his system from tested and
supposedly good boards and beginning system-level test, the locomotive horn
began to blast, and he could not find a way (other than powering down the system)
to turn it off.
The problem turned out to be a faulty field-effect transistor (FET). Although
an in-circuit test for that part was feasible, the program running at the time
included no such test. The automatic program generator did not know how to
turn the FET on and off, so it did not create the test. The overall strategy was
adequate, but a test engineer had to add the FET test manually to achieve
sufficient comprehensiveness. In that case, the tester could not attain the voltage
levels necessary to turn the part on and off, necessitating extra circuitry in the
fixture. Because the product was very expensive, almost any effort to improve fault
coverage was justifiable.
Again, knowing actual or projected defect levels and likely defects helps in
planning an effective strategy. Also, any strategy must be flexible enough to accom-
modate new information as painlessly as possible.

1.9 Statistical Process Control
Statistical process control (SPC) is a strategy for reducing variability in
products, delivery times, manufacturing methods, materials, people's attitudes,
equipment and its use, and test and maintenance. It consists of answering the
following questions:

* Can we do the job correctly?
* Is the process correct?
* Is the product correct?
* Can we do the job better?

Electronics manufacturers who diligently monitor their processes with those
questions in mind will see fewer and fewer failures as the product matures. Even-
tually, a well-controlled process will produce no failures at all.
At times, however, especially in high-volume processes, even zero defects
is not enough. Manufacturing engineers need to anticipate problems from the
production cycle before they occur to avoid the considerable time, expense, and
schedule disruption of having to fix large numbers of defective products or the
inconvenience of product parameters that "drift" in the field to the point where
the products fail.