Page 52 - Building A Succesful Board-Test Strategy

38 BUILDING A SUCCESSFUL BOARD-TEST STRATEGY

A "Hot-mockup" test takes place in a model of the real system. Hot mockup
of a disk-drive controller board, for example, requires a system that is complete
except for the board under test. Other than testing an assembled system at the end
of the production line, hot mockup is the only way to ensure that a particular board
actually performs its intended function.
Strategic choices directly affect how engineers generate test programs. Some
techniques, such as emulation, permit only manual generation and require that the
person responsible intimately understand the board's function and technology,
Shorts-and-opens testers and some forms of inspection can learn their programs
from one or more known-good boards, with little human intervention. Program-
ming in-circuit and functional testers ranges from completely manual to completely
automatic. Vendor-supplied or manufacturer-developed tools can help with
manual steps.
How well a product incorporates design-for-testability and concurrent-
engineering principles may expand or limit strategic choices. Inclusion of bound-
ary-scan or self-test circuitry on devices or board segments significantly simplifies
test generation and enhances final product quality. Burn-in or ESS in a manufac-
turing strategy will affect the test steps around it.
Much also depends on the required depth of failure diagnostics. If bad prod-
ucts end up in the garbage, engineers identify failing components but not specific
faults except as necessary for feedback into the process. In military-system and
other modular applications, testing identifies the failing module for replacement
and goes no further. On the other hand, if test diagnostics must find the specific
component that failed, as is done for process monitoring, test steps and program
generation must take that fact into account.
The likelihood that a board device will fail affects the diligence with which a
test should verify that it works exactly as expected. A device that rarely fails falls
into the same category as an ASIC that used to work. Board-level test need only
confirm that it is there, inserted correctly, and alive. Conversely, if devices are
suspect, board-level tests must more completely exercise them.
When planning a test strategy, test-engineering managers must know who
their engineers are and with what machines, strategies, and technologies they are
already familiar. If the people have no experience with functional testing, for
example, any strategy that includes that option will require a longer learning curve
than one that does not. This constraint does not preclude using a strategy element,
but including that element brings with it additional baggage.
Similarly, experience with a particular vendor or machine may help the
decision-making process. Programmers familiar with one tester type will have an
easier time developing tests for that machine than for something else. By the same
token, if the experience is negative, the range of strategic choices can be narrowed
by eliminating that vendor or machine from consideration.
Some manufacturers would like to use the same or similar equipment in
product development, production, and field service. That approach limits the
number of possibilities in the factory. Bed-of-nails test, for example, is rarely