Use a Peltier cooler to 'extend' uC operating temp range?

[max_torque]

I have an application that i'd love to use an existing uC for, due to being able to re-use all my circuit layout and code as is. (Very tight time scale).  However, the maximum operating temperature of the system "could" exceed the 85 degC limit of the uC.  I say "could" because if this occurred it would be intermittent and very infrequent.

So, i wondered about using a small Peltier cooler, stuck directly to the top of the uC, or coupled to the copper ground plane the uC sits over, that way, i could measure internal temp, and if it got critical, switch on the Peltier and 'pump' heat directly out of the uC, resulting in a higher temp gradient than for a passive heatsink etc?

Anyone done this?  (cost is not important, but Time is)

[mikerj]

How much is the temperature going to exceed the 85C max by?  Could you use a device rated for a higher temperature?  Microchip has 150C rated PICs (12,16,18,24 and 33 series).

[tom66]

Depending on the production quantity, you can simply requalify the parts necessary.
I've seen a manufacturer use 125C rated parts in a 200C environment.
They did this by baking the assembled boards in an oven at 220C.
Any boards that failed were rejected (about 3%.)

[suicidaleggroll]

It might not be the simplest or cheapest solution (you'd want a part specified to a higher temp range for that), but I don't see any reason why it wouldn't work.

I'm doing something similar for battery temperature control on an outdoor autonomous system using LiFePO4 batteries.  The Peltier is on an H-bridge to pump heat into or out of the battery box as necessary to keep it in the 0-45 C range.

[MK]

Most peltiers are getting near their max temp if the cold face is at 85 C, it is antimony and bismuth alloys with a special low temp solder to mount them to the alumina plates, so look carefully at the size you would need and the likely efficiencies at those temps.

[mikerj]

Most peltiers are getting near their max temp if the cold face is at 85 C, it is antimony and bismuth alloys with a special low temp solder to mount them to the alumina plates, so look carefully at the size you would need and the likely efficiencies at those temps.

You can get high temperature TECs good to 200C, but it seems a complex way to do thing unless the OP's temperature requirements are particularly extreme.

[David Hess]

There is no reason this will not work assuming that the temperature limits of the thermoelectric cooler are not exceeded but I agree with tom66.  The easiest way to handle this is to qualify and derate the parts even if you have to do that yourself.

The one thing I would worry about however which would encourage me to use a thermoelectric cooler (or parts with a higher temperature rating) is that floating gate storage (EEPROM and Flash) retention times and reliability are compromised by operation at high temperatures.

[Niklas]

Have you checked the datasheet for the possibility to use the microcontroller at a higher temperature but with limited clock speed? The 85'C rated STM32F103**T6 can be used up to 105'C if the power dissipation is limited.

Being close to 85'C ambient also rings another warning bell: X5R rated ceramic caps. Check their voltage derating or change to X7R. There is a derating chart on page 2 in Murata's MLCC catalog. Failure mode: short circuit

[max_torque]

Thanks all.  Unfortunately, there is no HT rated micro in the same series available "off the shelf" and although we are happy to post production test all our devices, i need to offer the client at least a basic "get out of jail free" route, should it all go pair shaped!!

(later, when we have more time, we will move to a different and temp rated micro controller, and port all the exisiting code over to that new device, but for the first protoype batch, we need to get something that works immediately for demonstration)


I will include a power source and mech foot print on the pcbs for a Peltier and if we need it, it's there ;-)