Page 75 - Building A Succesful Board-Test Strategy

Test Methods 61

generation usually means “merely” learning correct responses from a good board.
Therefore, board manufacturers can position these testers to prescreen more expen-
sive and more difficult-to-program in-circuit and functional machines.
Unfortunately, the convenience of self-learn programming depends on the
availability of that “good” board early enough in the design/production cycle to
permit fixture construction. Therefore, manufacturers often create both fixture drill
tapes and test programs simultaneously from computer-aided engineering (CAE)
information.
Shorts-and-opens testers, as the name implies, directly detect only shorts and
opens. Other approaches, such as manufacturing-defects analyzers, can find many
more kinds of failures at only marginally higher cost and greater programming
effort. Also, the surface-mount technology on today’s boards makes opens much
more difficult to identify than shorts. As a result, shorts-and-opens testers have
fallen into disfavor for loaded boards (bare-board manufacturers still use them),
having been replaced by inspection and more sophisticated test alternatives.
In addition, as with all bed-of-nails testers, the fixture itself represents a
disadvantage. Beds-of-nails are expensive and difficult to maintain and require
mechanically mature board designs. (Mechanically mature means that component
sizes and node locations are fairly stable.) They diagnose faults only from nail to
nail, rather than from test node to test node. Sections 2.3.5 and 2.3.6 explore fixture
issues in more detail.

2.3.3 Munufucturing-Defects Analyzers
Like shorts-and-opens testers, manufacturing-defects analyzers (MDAs)
can perform gross resistance measurements on bare and loaded boards using
the op-amp arrangement shown in Figures 2-4 and 2-5. MDAs actually calculate
resistance and impedance values, and can therefore identify many problems that
shorts-and-opens testers cannot find. Actual measurement results, however, may
not conform to designer specifications, because of surrounding-circuitry effects.
Consider, for example, the resistor triangle in Figure 2-6. Classical calcula-
tions for the equivalent resistance in a parallel network,
1
1
-- --+- 1
RM Rl &+R3
produce a measured resistance of 6.67 kQ. Like shorts-and-opens testers, MDAs
can learn test programs from a known-good board, so 6.67kR would be the
expected-value nominal for this test, despite the fact that R1 is actually a 10-kR
device.
An MDA tester might not notice when a resistor is slightly out of tolerance,
but a wrong-valued part, such as a 1-kQ resistor or a 1-MQ resistor, will fail. Creat-
ing an MDA test program from CAE information can actually be more difficult than
for some more complex tester types, because program-generation software must con-
sider surrounding circuitry when calculating expected measurement results.