-55 degree operation and micros

everyone

You could always use an Intel processor. Seriously, I think you're on the right track with your heater. Depending on the size and thermal conductivity of your case, a resistor or two would probably do the job. If you cut them out once the temperature gets above (say) -30C, you can turn them off and start your processor. Again, depending on the insulating properties of your case, you may find that the heat produced by your circuitry is enough to keep the thing warm (less cold). My usual problems are at the other end of the temperature scale.

Cheers, Alf snipped-for-privacy@remove.the.obvious.ieee.org

Depending on the part and the service, sometimes you can use self-heat to keep the processor above -40.

One thing you have to be careful about is electrolytic caps. Some eletrolytes freeze that cold.

Just as a guess, his problem isn't RUNNING temp, it is STARTING temp. Any aircraft engine, for example, has to have a controller that can start the engines when the ambient temperature is as low as -55dC. Heaters are quite normal to get the CPU up to a legal temperature before turning it on. The heaters are quite common peltier devices that can be shut off after the processor is "self-heating" and can even turn the current around if the CPU gets too hot (maybe because the controller is within the engine cowling itself and temporary eddys keep airflow from cooling properly).

One of the FAEs in the comp.arch.fpga group posted about using self heat on an FPGA to start below spec temp. He just kept reconfiguring it while watching the core temp. I belive he used an on chip diode to measure the temp, but I am not sure. Either way, the concept is sound if you have other circuitry that will control this operation.

I've never had to do anything at the low end of the temperature range, but have always been curious about what is there about solid-state electronics that would cause them to not work at extreme low temperatures. The electrons freeze?

Thanks for the info, the heater seems to be the way to go. I guess I should have been more specific, it is the start up at -55 I'm worried about not the operation. the MIL spec says soak at -55 for 3 hours and then turn on. It looks like I'll use a PTC that will shut it self off as the temp comes up. But as per most designs I'm sure the customer will freak out and tell me that it goes over the power budget and I should find another way. I had an engineer work outthe numbers real quick and I needed about 10 watts to bring it up to -35.

everyone

You could also consider special qualification of the parts. -55 is not really an issue for most chips other than possibly start up current (which tends to increase, sometimes dramatically). They just don't test them at that temp. You can do your own testing which will likely be cheaper than getting the vendor to do the testing for you.

Anyone know of any other issues with startup or operation of CPUs at low temps? Of course crystal oscillators are a different issue, but you can easily get those rated for -55C.

BE wrote:

I find a large, friendly woman to cuddle with.

Yes, there's not much of an alternative.

-- Regards, Albert

---------------------------------------------------------------------- AM Research, Inc. The Embedded Systems Experts

----------------------------------------------------------------------

Apart for some mechanical problems like some materials becoming very brittele in low temperatures, one has to remember that the temperature _range_ becomes quite wide. A component rated for only 0..+70 C will need to function in a 70 K wide temperature range, which is about 25 % compared to a slightly elevetad room temperature of 300 K. However, a device rated for -55.. +125 C will have to operate in a 180 K temperature range, which is 60 % of 300 K.

All components have some kinds of temperature coefficients, both mechanical end electric.

Dissimilar expansion coefficients can cause high stresses at the extremes of the temperature ranges when dissimilar materials are joined.

A silicon transistor will have a Vbe temperature coefficient of -2.2 mV/K, so a 70 K temperature range will change the Vbe by 150 mV around the nominal 700 mV, while the 180 K range by 400 mV (from about 500 mV to 900 mV). Unless the circuit is well temperature compensated, you may end up with troubles, in particular with low voltage battery powered systems, in which the battery performance may also suffer.

When you look at specifications for components, which are available in commercial and some extended temperature ranges, the typical parameters (at room temperature) are the same, but the worst case specifications are much worse for the extended temperature device, since the worst case situation usually occurs at the temperature extremes. The performance within the commercial temperature range would be the same.

When using some component outside its rated temperature range, use at least a component, which is also available at a larger temperature range and thus specified for that temperature range, so you can get some educated guesses how the limited temperature range device will operate at extreme temperatures. The manufacturer may of course have selected the better units for the extended temperature range, so the limited temperature range components do not perform that well at extreme temperatures.

Even if the chips are practically identical, the package may be different and thus behave differently at extreme temperatures.

Paul

I am thinking of digital devices -- how does the larger temperature range effect them (electronically)?

I can understand package and battery problems and operational problems for analog circuits.

I find a horse to cuddle with.

"Digital components are made of analog parts" - Don Vonada, Design Leader of Digital Equipment Corporation, in early 1970's.

This still valid.

Tauno Voipio tauno voipio @ iki fi

Quite true, but digital circuits don't have to be calibrated. Unless rise and fall times become excessive, we don't care how digital circuits get from a high to a low or vice versa nor do we care how high a high is unless it gets too low.

Yeah, I do. As transistors get smaller and smaller, colder temperatures allow adequate mobility of the electrons that they can jump across gates intended to keep them from flowing (i.e. short circuit, for example). The smaller the circuit, the higher that temperature is. -55dC becomes a short circuit problem in the current generation of high end devices, IIRC.

For anything made at the 1 micron scale or LARGER, it isn't much of an issue, though. IIRC, I think it starts becoming an issue at .65 micron scale or so. However, it has been 5 years since I had to last deal with those device physics issues, and I have never been a device physicist :-P.