> I guess I ought to know this but I never really found out: do companies
> making regular commercial grade components usually test each component in
> some way??! I never exepected that, I always thought that components for
> non-critical applications (i.e. consumer electronics) were only spot
> checked. I could understand maybe testing each one of the industrial or
> aerospace/military versions for saftey's sake (although, you would also
> expect the final equipment manufacturer to test each final device under
> these circumstances).

In many cases, testing individual devices merely to see if they meet cust-
omer specifications is not terribly useful; a crate of devices that happen
to "barely" meet specs may turn into half a crate of devices that fail in
use.  Instead, it's desirable to determine what types of defects or modes
of failure exist, and then to work at identifying/preventing those.

My impression of normal silicon manufacturing processes is that devices are
at minimum somehow checked to ensure that no stray dust particle has prod-
uced a crater or mountain somewhere.  While chips are always made in clean
room conditions, even a room with one particle for every thousand liters of
air will have a few dust particles that may land somewhere and wreak havoc
on one chip on one wafer without affecting anything else; even if 99.99% of
parts don't have such defects, most customers would like another nine at
the end of that.

For (E)PROM devices, my impression is that many such devices, even OTPROM's,
are indeed programmed and tested before packaging; until the wafer has been
sealed in the expoxy, there's nothing to prevent its erasure.  If the dev-
ices program and operate correctly at various extremes of temperature and
radiation exposure, they are very unlikely to fail in the field; my imp-
ression is that all shipped Microchip EPROM-related products are tested for
data retention at moderately high temperatures, and that some "sacrificial"
units from each batch are tested at excessive temperatures.

Once devices have been tested on the wafer, they are placed into either plas-
tic of ceramic packages.  My impression is that when Microchip says that the
JW parts are characterized but not tested at extreme temperatures, what they
mean is:

(1) All their parts are tested at temperatures somewhat outside the operating
    range specification; the sacrificial parts are tested outside the "absol-
    ute" specification.  The wafers themselves do not fail under such abuse.

(2) The plastic-cased parts will suffer package-related failure at temperature
    extremes which the ceramic cases can survive.  Microchip does not, how-
    ever, test individual parts to ensure full-range temperature reliability
    after they are packaged.  While Microchip does not expect /JW packages to
    fail due to temperature extremes, it is possible that, e.g., thermal exp-
    ansion and contraction could weaken one of the die bonds.

(3) Exposing individual parts to temperature extremes significantly outside
    rated specifications may degrade their long-term reliability.  It would
    not make sense for Microchip to take years off the lives of their parts
    by exposing them all to extreme temperatures even if they'd found that
    most parts treated that way would suffer little if any apparent damage.

Note that these are only my impressions; since I don't work at Microchip
anything above may or may not be right, and should be taken with 4 grains
of salt and 2 grains of oregano.