1Adata said:
yes, but this does not mean that electronic components do not depend on
temperature; an overheating makes very strong impact on electronic chips.
---------
Sure, I agree with that.
One metric I would suggest to you, is whether the equipment function
stopped due to the temperature. The processor inside the box should
"crash", once it gets past a certain temperature. Going an additional
number of degrees past that point, could lead to physical damage.
If the equipment operated throughout the overheated interval,
I would argue that is a crude metric that critical temperatures
were not exceeded.
Some temperatures
1) Silicon damage (info from our fab many many years ago) = 135C
The full 135C can be survived if the chip is housed in
a ceramic package. They no longer use ceramic for most
applications. The chip packaging material is now "organic",
and may be limited to 90C to 100C, as the packaging material
degrades at higher temperatures. So the chip itself can
stand 135C, while the packaging is more susceptible.
This was the value from our fab. YMMV
2) Simulation temperature (operation not guaranteed correct) = 105C
A chip designer tries to prove correct operation at
this temperature. Whatever the value is (105C-110C), this
would be considered an upper limit during design.
3) Processor may crash (only rough figure) = 70C to 90C perhaps
Anecdotal values seen on various processor families.
I suppose, it would be possible for a hardware design, to
continue to run error free, at the same time as the organic
packaging is being degraded. But it is more likely, you'd
see some functional degradation before that happens. If
the box continued to run, without observable performance
degradation, I would argue it is just fine. If it crashed,
then it may have overheated significantly.
When a device is convection cooled, and you seal the top vent,
it would be easy for temperatures to shoot way up. The value
of cooling effectiveness of convection is not that great.
Having to fall back to conduction cooling, because the
vent is blocked, is not going to work very well if the
product has plastic packaging. It might behave marginally
better if the casing was metal.
There have been cases of small devices like this in the past,
where the internal temperatures were really too high for
long life. Some people had first generation gigabit routers
die, because the designers didn't do enough to cool them.
So there are products, where the silicon inside is tortured.
Some people added ventilation to their own products (voiding
the warranty), and their products survived. So not all
hardware designers are kind to the silicon.
Some chip types *love* to run hot. Back when ECL logic was
popular, you could get skin burns from touching the tops
of some of the chips. And the chips worked best, when they
were that hot. CMOS has different behavior, and gets slower
at high temperatures.
Paul
