I have been requested to design a switching regulator that will operate down to -65C. No manufacturer seems to make a controller chip that is rated to operate at that temperature. Most will only go to -40C (some Mil spec parts will go as low as -55C, but not any of the chips I want to use). So, the question is: what will happen if a part that is rated at -40C is operated at -65C? I suspect that it will work fine, but I’d like to know what the failure mechanisms are possible. I’m planning on using all ceramic caps, so those shouldn’t be a problem.
I don’t know. But two specialty areas that do operate at least some semiconductor electronic parts at temperatures below -55 are infrared detectors and low noise cameras. I could guess some shift in the populations of carriers in elevated energy states, and extreme thermal expansions, could be trouble areas.
Based on some of your other posts I have to guess you already know these things, or even know better than I do and could correct me. Just trying to help on the off chance I can…
On deep space probes, they’ll put in Radioisotope heaters to keep parts warm. Something like that could be a possible approach.
I’ve taken a few boards over the years down well below the min temp spec of their components. Whenever I’ve done it, the boards always continued to work down to the coldest temp that I could get with the setup I had. These were fairly short tests though, and the boards weren’t designed or even expected to run long term at those temperatures. In my experience, for what it’s worth, components tend to survive much better below their cold temp spec than they do above their hot temp spec.
I know that’s not quite the answer you are looking for, but I hope it’s at least somewhat helpful.
Any chance you can stick a power resistor in there that switches on when the temperature gets below a certain threshold?
I was also lowering the temperature slowly, which is going to minimize the mechanical stresses due to thermal contraction. Things are more likely to crack or snap off if the rate of cooling is faster. Are these regulators expected to be able to be dumped from a hot environment directly into that cold environment and just work, or will they be in an environment where the temperature change will be much more gradual?
Thanks for the answers, folks.
It’s an interesting problem, from an engineering perspective. The regulator itself will generate some heat, so it won’t have to operate at the minimum temperature for very long. However, it won’t be put into operation until after it’s deployed (it’s for an outside-of-an-aircraft device), so it may have to start operating at the lowest rated temperature (which then brings up TCE issues).
I’m going to talk to the end user next week - my guess is that they will tell me to ignore the ratings, because it is stored in a warmer environment, and then deployed into the ambient, where it rarely gets that cold anyway.
We’ll see.
Suddenly deploying something warm into a cold environment creates the possibility of condensation problems, too.
Condensation is an issue when deploying a cold object into a warm environment (water vapor in the warm air condenses on the object). In the OP’s case, that might be an issue when the device is brought back into the aircraft, but not when it is placed outside.
Perhaps it could be thermally insulated and mounted in contact with something that has significant thermal mass, so it takes a longish time to reach the very low ambient temperature?
Perhaps (as suggested above) the device could include a resistive heater sufficient to keep it at or above -40?
If failures are tolerable, the best approach may be to simply ignore the issue during design, and see how normal parts fare at abnormal temperatures. Note that it shouldn’t be all that difficult to do some testing at -65C, using a setup cooled by liquid nitrogen.
How many units are you making? One? Or many?
You can get one unit to work by tweaking the circuit. But you will have a problem if many units are to be made. That’s because you won’t know the values/parameters for many of your components. Everything is a function of temperature (resistance values, capacitance values, semiconductor voltages, etc.), and at -65 C the manufactures will not provide tolerance values. As I’m sure you’re aware, the tolerance on the capacitor and/or inductor in a switching regulator can be critical.
First, the condensation issue: the units will be potted in epoxy, so condensation shouldn’t be a problem.
The initial run is for 500 devices. Component value drift is a concern, so I’m going to use low TC metal film precision resistors to set the feedback voltage. I’m going to assume (and test if possible), that their value drift is fairly linear with temperature. I can get 50 ppm/C resistors, which means their value could change as much as .5% over the temperature range. This is acceptable as it is, and since the regulator voltage is set by a divider, if both resistors drift in the same direction, the total inaccuracy will be less.
Yea, this is generally how it’s done. I’m working on a portable data acquisition system right now that contains a heater and a thermistor (for measuring ambient temperature). At very low temperatures the heater kicks on, and the electronics are not allowed to turn on until the temperature is above -20 C.
Since it’s a switching regulator, do you have any capacitors or inductors that have critical values? Not sure about inductors, but many capacitors have large tempcos.
I don’t think any are particularly critical.
There’s the input Cap, which just needs to be above some minimum value, so I can over-rate it to be safe. There’s also a feedback cap, but that will work over a 10:1 value range.
Still, I’m going to try to get the lowest tempco caps I can find.
Have you tried looking outside the U.S. military?
I’ll see how serious the customer is. There are probably NASA-certified parts that do what I want, but the customer is unlikely to want to spend 100x (or 1,000x) over an off-the-shelf part.
And the reason I’m asking… most of the time, the tolerance on a bypass or coupling capacitor is not critical. Heck, ± 20% is often good enough. But some switching regulator circuits (especially high frequency/high efficient designs) are very sensitive to capacitor and inductor values. I remember building one circuit that specified a 0.1 μF cap on the output. I stuck a 1 μF on the output and it acted erratically. When they said to use a 0.1 μF cap, they meant it!
For your application, I would use COG/NPO caps, since they have the lowest tempco. But they’re multilayer, which means they’re prone to cracking due to mechanical/thermal stresses induced by the potting compound (epoxy or whatever). If this is a concern, consider using polycarbonate or polystyrene caps.
Building a reliable switching regulator is not simply a matter of looking at a schematic and laying out a PCB. Component placement and layout are very critical for switching regulator circuits, especially the high frequency ones. Pay close attention to stray capacitance, stray inductance, the grounding scheme, and inductance of through-hole component leads and PCB traces.
Read as much as you can about proper design and layout for switching regulator circuits. As Don Lancaster once said, “An hour in the library is worth a month in the lab.”
Good advice.
I’ve designed a number of switching regulators. I always try to stick close to the recommend PCB layout, since I figure that the IC designers know what they are talking about. I’ve only had problems on one design, and that was a complicated 0-30v output, 6-40v input, buck/boost regulator driving a motor - it had problems due to the back EMF of the motor when the output voltage was suddenly dropped from 30v to 0.
I would be worried about solid-solid phase transitions that would cause characteristics to suddenly change. Certainly the materials being used in IC will have phase diagrams available, unless they are proprietary. It is certainly my understanding that processors actually run better cold. Aren’t there people that overclock their computers by dipping the IC in liquid nitrogen? I’m certain that there are computers designed to work this way. Look at at any computers designed for large scale computations like modeling complex systems like weather and molecules. I’m certain that I’ve heard of computers that run that way.
Also keep in mind that getting it to work on the bench is the easy part. Getting it to pass qual is another mess of problems.
For qual, do you have to do shock & vibe? Temperature cycling? Temperature shock?
If there is a large mismatch between the mechanical tempco of the potting compound, the tempco of solder, and the tempco of the components, prepare yourself for cracked components and/or cracked solder joints. After you temperature cycle, I would x-ray the circuit and look for cracks. You may also want to cross-section it to look for evidence of mechanical creep and other stresses.
And of course, the circuit needs to operate at the temperature extremes. This is where electrical tempcos will create all kinds of havoc.
Perhaps a silly question, but is this going to be deployed as a pod hung from a hardpoint? If so, can you not draw power from the aircraft for a heating circuit?