Page 43 - Building A Succesful Board-Test Strategy

What Is a Test Strategy? 29

reasonable to estimate the scanner-stage failure rate of the copier/fax at 0.5 percent,
subject to the same hedging as before.
On the other hand, a copier company would have had less experience with
the modem stage of a fax/copier. Estimating its failure rate might simply mean
assuming that it will fail at least as often as any other part of the circuit and design-
ing the test strategy accordingly.
When a company introduces a new product that clearly departs from
previous product lines, any failure-mechanism estimates will necessarily be less
precise. A personal-computer company that branches out to make printers must
contend with a wide range of electromechanical problems in addition to any data-
processing failure modes that resemble computer failures. A test strategy could base
testing of the computer-like functions on previous experience, then concentrate on
tactics required to test the product's more unique characteristics. This approach
helps avoid unwelcome surprises, where the unfamiliar portion of the circuit creates
the most headaches, while allowing the manufacturer to gather information and
adjust the production process so that estimating failure modes for the next printer
product will be much easier.
In addition to deciding where failures will likely occur, a test engineer must
try to determine what types of failures they are likely to be. If a paste printer
has been known to create large numbers of shorts or off-pad solder joints on
production boards, it is reasonable to expect that new boards will exhibit these
same problems.
The need for familiarity with the manufacturing and test characteristics
of a large number of different board types helps to explain the move to contract
manufacturing. The diversity of their client base generally means that contractors
have experience with problem-solving techniques that individual client companies
likely have never seen.
To illustrate how a fault spectrum influences strategic decisions, consider
the hypothetical spectrum in Figure 1-8. Assume an overall first-pass yield of
50 percent.
More than half of the failures, and therefore more than 25 percent of the
boards through the process, contain shorts. Removing those shorts will improve
board quality more than removing any other fault type. Removing them can
involve finding them at test and correcting them through board repair or
determining the cause of this large number of shorts and altering the process to
prevent their creation.
A lot of shorts on a surface-mount board may suggest a problem with the
paste-printer calibration. On a through-hole board, the culprit may be (and often
is) the wave-solder machine. Missing, wrong, and backwards components gener-
ally result from operator errors. Bent leads could be a problem in manual assem-
bly or with automatic-insertion equipment. Unless the production process changes,
analog specification, digital-logic, and performance problems are the only faults
that will likely change dramatically from one board design to another.
In fact, in this example, shorts, opens, missing components, backwards
components, wrong components, and bent leads—the easiest faults to identify