Page 40 - Building A Succesful Board-Test Strategy

26 BUILDING A SUCCESSFUL BOARD-TEST STRATEGY

have to recreate those vectors, another example of that popular game "reinventing
the wheel."
Limiting board-level device tests to looking only for gross failures applies to
other complex devices as well. Sophisticated microprocessors, for example, should
be treated in this way. The only alternative would be to test them thoroughly on
the boards, an unpromising prospect, In-circuit testers, which make contact with
individual devices through a bed-of-nails fixture, find achieving necessary test
speeds difficult or impossible. Capacitance contributed by the fixture nails coin-
promises signal integrity.
At functional test, which examines the board as a unit from the edge con-
nector or other accessible point, intervening circuitry between the test's input and
device pins, and between device outputs and the board's access point, enormously
complicate the task of creating comprehensive test programs.
Even design-for-testability techniques do not reduce such efforts by much.
Boundary scan, which allows access to internal logic segments from the board's
edge, provides only serial signals. For long tests, this approach is painfully slow. A
good device self-test helps, but if the self-test exists, why not execute it before
expending the labor to install the device on the board and risking the added expense
of removing a bad device later? The most cost-effective solution remains ensuring,
through vendor testing or other means, that board assembly proceeds with good
components only.

1.6.3 Board Characteristics
What do the boards look like? How large are they? A 5" x 8" board presents
different problems from one that is 15" x 18". Similarly, does the board have 600
nodes or 2000? Some test equipment cannot handle the larger boards or higher
pin counts. With a low-pincount board and a high-pincount tester, it may be con-
venient to test several small boards of one or more board types on a single bed-
of-nails fixture.
Are there pads available for probing by a bed-of-nails during in-circuit testing
or by a technician for fault isolation during functional test? How many layers does
the board contain? Have bare boards been tested? If not, can inner board layers
be repaired if a fault emerges? Are there embedded components (resistors and
capacitors) that require testing within the board itself?
What are the average lot sizes? Lots of 100 instrument boards suggest a very
different strategy from disk-drive-controller board lots of 25,000 or locomotive-
engine board lots of one.
What is the product mix? How often does the test operation change board
types? One prominent disk-drive manufacturer began production of one product
in a U.S.-based facility, devoting one line to each board type and therefore requir-
ing no changeover. When he transferred the product to Malaysia, he discovered
an entirely different philosophy. There, the entire facility turned out a single board