Chip operating temperatures

[tom cloud]

>>From many years experience in the aerospace industry 
>operating temperatures are much more important than 
>storage temps.  Mounting the components within the 
>passenger compartment will keep commercial quality 
>components well within specs regardless of outside temp.
>
>The mil-spec components price will shock your shorts off-

Yep

>so remember the big difference at high temps is the heat
>dissipation capabilities of the chips.  Think not?

the parameters of the chip itself change with delta-T, so the chips
are spec'd within certain limits.  As I said in another post, the
next biggie after what the manufacturer guarantees the chip
to do is the physical limitation of package (epoxy vs. ceramic).
Epoxy gets soft and melts / burns.  It also allows moisture to creep
into the chip.  So, the same chip *might* function at high
temp were it put into a better header -- then again, it might
not, as the mfgr selects the better chips to charge more money
for them!

>I know 
>many PC's that people paid top dollar for are running pushed
>processors (lower speed rated chips operated at higher 
>clock speeds) with a big fan glued to the chip.

Yep,  the heat in a digital circuit is mostly generated during
the rise and fall time -- not while its HI or LO, so slower chips,
pushed to faster PRR (i.e. clock period), will get hotter, as the
transitions are occuring more often per unit of time.  Therefore
a heat sink is necessary.

>Built a few 
>myself (with the customer knowing it). I could go into why
>hot chips get hotter - suffice to say that as temp goes up,
>resistance goes down so current go's up causing temp to go
>up so resistance goes down - ad naseum until either temp

Depends on whether we're talking about bipolar transistors for field
effect transistors.  Bipolars experience what is called *thermal
runaway* due to their Beta increasing with temperature.  In FET's, a
drop in resistance actually produces lower heat dissipation, since
P=EI and lower R means lower E dropped across the switch.  When selecting
an FET switch, one of the major parameters is lowest R(on).  What
we perceive as a high temp actually means a *real* hot spot(s)
inside on the chip.  At temps over something like 175 C or 200 C or
so, migration of carriers begins and the switching speeds begin
to deteriorate (permanently), causing more heating -- and on and on.

>stabilizes or chip fries. Pesky laws of physics again.
>
>Do like PC's do.  Mount cooling fans to flow air over chips. 
>Consider Pelzior effect cooling chips - I use them on Cyrix 686-166
>chips in mini-towers.  Makes cold directly from current.  Available
>from many sources.  But the best cure is to condition the signals 

Can mount metal tabs from the main heat generating sources to the
enclosure (or mount them directly on the enclosure) with fins on
the outside to help remove heat -- should be adequate for any condition
you're willing to live in.

        [ snip sommore ]