Page 105 - Building A Succesful Board-Test Strategy

Test Methods 91

vectors each time and read the outputs again. The person then reports the name
of the failure, and the machine learns that information. Test engineers can gener-
ate input vectors manually or obtain them from a fault simulator or other CAE
equipment. For very complex boards, engineers may prefer entire CAE-created test,
programs.
Teaching instead of programming cuts down on programming time. Users
of this technique have reported program-development times measured in days
or weeks, a vast improvement over the months that developing some full
functional programs requires. In addition, because the program learns from
experience, delaying a product's introduction because its test program is incom-
plete is unnecessary.
Another advantage to teaching the tester about the board is that it permits
margin testing and some analog testing. The engineer can submit a number of
good boards for testing, declaring that waveforms and other parametric variations
are within the normal range. This permits the test to detect timing problems
and other faults that do not fit into the conventional "stuck-at" category. Even
with only one good board, a test engineer can exercise that board at voltages
slightly above and below nominal levels, telling the tester that the responses are
also acceptable. In this way, normal board-to-board variations are less likely to fail
during production.
Anytime a good board does fail, the test engineer so informs the tester. It
then incorporates the information into the database, reducing the likelihood that
a similarly performing board will fail in the future.
Software to make an expert system work is extremely sophisticated. The chief
proponent of the technique packed it in a few years ago, and as yet no other
company has taken up the challenge. I continue to talk about it in the hope
that some enterprising test company will resurrect it, like a phoenix from the ashes.
Thus far, this personal campaign has not succeeded. Even in its (relative) heyday,
most test engineers remained unaware of this option, and therefore could not
evaluate its appropriateness for their applications. The learning curve both before
and after purchase was not insignificant, but users who tried the approach reported
great success.

2.3.72 Two Techniques/ One Box

Combinational testers provide in-circuit and functional test capability in a
single system. This solution offers many of the advantages of each approach and
some of the drawbacks of each.
Board access requires a special bed-of-nails fixture containing two sets of
nails. A few long probes permit functional testing at a limited number of critical
nodes. All remaining fixture nodes contain shorter pins. Vacuum or mechanical
fixture actuation occurs in two stages. The first stage pulls the board down only far
enough for contact with the longer pins. Second-stage actuation pulls the board
into contact with the shorter pins as well for an in-circuit test. Manufacturers can
conduct either test first, depending on the merits of each strategy.