David VanHorn wrote:

"I agree with your approach, up to a point.
It sounds to me like your crystal is crashing into the sides of the can.
Check the specs, but I doubt it's rated for that sort of abuse.

And therein is the fault I see in this method.
You're getting into areas where you're exciting threshold effects, not
wearout effects."

well, yes but no.

No, because what you are doing with this sort of test is, in effect, time
compression. So we expect that crystal to fail. We're just getting it to
fail sooner. There's nothing wrong with taking this test way beyond the
individual component's specs. We are, after all, trying to break it. We're
trying to make the product as a whole more robust.

Yes, because you are right. Taking it beyond what it can possibly survive
will cause it fail, and  under normal conditions that failure mode might not
ever be seen. You do need to apply some common sense. There are physical
limitations, and I do realize that. But, if the crystal is the first thing
that fails in your HALT test, it's a pretty good bet that field failures
will also show a failed crystal. Whether or not that came about due to
vibration. Of course, if vibration is the only thing that causes the crystal
to fail, then yeah, the excess stress in the test is not really valid.

But we don't make delicate lab instruments. We make plumbing tools.
Construction crews drop tools all the time (from ladders, of course). The
tools get wet. They get extremely muddy and dusty. They are left in the back
of a truck in the middle of winter, then taken inside a heated building. Or
they are left in the truck in the middle of summer, and get taken inside an
air conditioned building. All of these stresses add up over time and shorten
the life. Again, we're after time compression. Get them to fail early so you
understand how they fail, and fix it before it ever gets to a customer.

Perhaps you will see a way to cushion the crystal, or even the whole
assembly. Perhaps you choose a different crystal, or switch to an RC if
timing is not critical. And by taking the crystal beyond its specs, you will
likely encounter something else that breaks. Of course, it does not always
make sense to ruggedize a product beyond a point. If the improvement is
relatively expensive, and only yields a small improvement, then perhaps it
is not worth it.

Then there's the company 'attitude' to consider. For the computer industry,
where they want us to buy a new PC every 6 months, this sort of thing
probably does not. But, for someone like my company, whose (newly adopted)
motto is "Tools engineered for maximum uptime" and who offers a life-time
warranty on tools, it makes a lot of sense. Of course, our price reflects
that :) We have always been known for robust products. There's a lot of our
machines made in the late sixties and early seventies that we still service.
Now that we're integrating electronics, we needed a way to get there. Still
learning, but have come a long way.

Ok, now we have drifted sufficiently off the topic of the original question
:) If we continue this discussion, we should probably mark it [OT].

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics