[ previous ] [ next ] [ threads ]
 
 From:  Jim Thompson <jim at netgate dot com>
 To:  "David W.Hess" <dwhess at banishedsouls dot org>
 Cc:  m0n0wall at lists dot m0n0 dot ch
 Subject:  Re: [m0n0wall] Hardened Platform -40C to +85C
 Date:  Sun, 30 Jul 2006 12:14:54 -1000
David W.Hess wrote:
> On Sun, 30 Jul 2006 08:30:40 -1000, you wrote:
>
>   
>> You'll also want to think about the very real possibility of a "cold 
>> start".  That is, if the power fails, and the entire board gets to 
>> sub-zero temps (esp -40C (which is the same as -40F,)) then the crystal 
>> oscillators(s) on the board can freeze.  Without the crystals, you get 
>> no clocks.  Without the clocks, you get nothing from the CPU and other 
>> components that require at least one stable clock.
>>     
>
> When I have done designs for low temperatures in the past other then thermal
> expansion the only problem has been electrolytic capacitors freezing.
>
> Do you happen to know what the actual failure mode was with the crystals?  I
> have seen oscillators fail to start but it was because of low gain and not
> mechanical issues with the crystals.
>   
The problem was somewhere in the POR circuit.  It may have been due to 
lowered resistance (or increased capacitance) due to frozen water vapor 
on the crystal, or elsewhere in the circuit.  (I've seen a test probe 
cause a CXO to not start a POR circuit, I can believe that water vapor 
would induce the same effect.)

Its unlikely that the TCXO on the WRAP/Soekris boards is rated for 
-40C.  (The most common parts are only rated to 0C.)  You can get TCXOs 
that will run -40C to 85C, but they cost more.  I'm in Texas (not 
Hawaii) this week, so I can't read the part #s on the WRAP & Soekris 
boards here, or I'd go pull the spec.

(I have "over 40 year old eyes, and the magnifying glass I use to read 
the little #s on the parts) is back on Oahu.)
>> At the high(er) temps, you'll have difficulty with the decreasing "theta 
>> T" or  difference between the case (and junction) temperature of the 
>> components on the board and the temperature inside the enclosure, as 
>> well as the decreasing theta T between the temperature inside the 
>> enclosure and that of the ambient outside of it.
>>     
>
> Do you mean the allowable temperature rise from junction to ambient temperature?
>   
yes, but this is a non-technical list, and your original query didn't 
read such that I'd assume that you understood the depth of the problem.  
My mistake, and I offer apology.
> If the power dissipation and thermal resistance does not change then the
> difference should not change.  You still need to derate everything to keep sane
> junction temperatures or course which gets tricky at a Ta of 85C.  My rule of
> thumb was generally a maximum Tj of 85C but that would lead to zero power at a
> Ta of 85C.  You end up with odd things like TO-220 regulators using 20 watt heat
> sinks to operate at 1/10th of their maximum load.
>   
Sign seen on the wall (behind the counter) of  a "speed shop" popular 
where I grew up:

"Speed costs money, son.  How fast do you want to go?"

Mounted over this was a piston/rod combo that had obviously "failed in 
service".
>> Be careful, you'll find many companies who claim "-40 - 85C" who can't 
>> really do it.  Test it yourself before you accept more than a small 
>> quantity of any offered solution.
>>     
>
> I suspect it comes down to relying on a company making products for the
> industrial market instead of the consumer or business market.
>   
I ran into these issues a lot back at <my previous employer>.  The CEO 
and sales people (OK, I admit it, I was a nominal "marketing person" 
there, responsible for the womb-to-tomb aspects of the product) would 
say, "we need -40 - 85C!!". 

I'd say "Why?  Why do you need -40C - 85C?"  "-40C I can understand, but 
85C doesn't naturally occur anywhere on earth!"

They would show competitors spec sheets.   I'd claim 'horse-pucky', and 
once actually got one of Crisco's AP products (my employer was a 
supplier of "phased array" WiFi gear) and put it in the environmental 
chamber, just to show that it wouldn't even reliably run at 50C, (122F) 
never mind the 65C (149F) for which it was "rated".

Admittedly, it was the radio which fell apart at the higher temps, but 
this was (ahem) entirely anticipated.  (And it was a "industrial" 
(service provider) box.)

We've gotten a long way from the primary subject of this list (monowall).

Jim
(the rest of the story is that we got a "replacement CEO" about 6 months 
later, and he decreed that the product was "certified for -40C - 85C", 
even though it wasn't, and couldn't be without a refrigeration unit 
tacked onto the side, piping cold air into the box.