Page 37 - Building A Succesful Board-Test Strategy

What Is a Test Strategy? 23

If the process mean does not coincide with the nominal value of the test, the
number of board-level and system-level failures will increase. To allow for that
situation, Davis [1994] suggests applying a variance of ± 1.50 to the 60 nominal,
As a result, the process will produce 3.4ppm failures. With 5000 parts and process
steps, approximately 2 systems out of every 100 will fail, too high to avoid testing
in most cases.
It is impossible to overemphasize the necessity for tight process control,
Achieving a six-sigma process and maintaining it are not the same. In an
automated-assembly operation, if someone loads a reel of incorrect components
onto a pick-and-place machine and no one catches it, a large number of consecu-
tive boards will fail, and the process will no longer qualify for six-sigma status.
In addition, no one can achieve a six-sigma process overnight. A manager
cannot legislate .002 ppm failures per process step. People will react by throwing
up their hands and saying, "We can't get there." The factory floor is no place for
lofty theoretical goals. Rather, opt for incremental improvements.
Process analysis will indicate steps and components most likely to fail. Acting
on those elements first reduces overall board and system failure rates. Over a long
time, the total process will begin to approach six-sigma levels.

1.6.2 What Do You Test?
As stated earlier, before creating a test strategy, you must clearly examine
what you will be testing, both initially and during the life of the product, the
product line, and the facility. Will the boards be heavily analog, as with electro-
mechanical systems such as automobile engine controllers, for example? Will
boards be primarily digital, destined for PCs and related products? Will there be a
lot of mixed-signal technology, as with communications boards? Will product
requirements confine the board to a maximum size, as with boards for cellular
phones and other hand-held products?
How complex are the electronic designs, and what portion of the overall
product manufacturing cost do the electronics represent? How high are total
production volumes?
An elevator, for example, is a fairly expensive piece of low-volume hardware
that must be extremely reliable. Its electronic technology is both primitive and inex-
pensive compared to, say, a high-end PC. The system motor responds to the press-
ing of a button either in the elevator car or on the building floor. It needs to know
whether to go up or down, where to stop, when to open the door, and how long
to leave it open. A sensor on the door's rubber bumper has to know to open the
door if it encounters an object (such as a person) while it is trying to close.
The main controller may rely on a many-generations-old microprocessor that
is fairly easy to test either in-circuit or functionally (even with relatively inexpen-
sive and unsophisticated equipment) and has a very low failure rate. Circuit logic
is fairly shallow, facilitating test-program development.
In addition, although an elevator system may cost hundreds of thousands
of dollars, costs for the electronics and for electronics test constitute a very small