# Ultra low power heat switch

I have a battery operated device which needs to be temperature enabled. Once the temperature goes above 60C, the device should start operation. Ideally it'll be like a fuse, and will continue even if the temperature drops below 60C after initial activation.

The device has a microcontroller, so I currently plan to put it into deep sleep and wake it to check the temperature every minute or so, however this will consume almost a microamp of current continuously (to operate the timer needed to wake).

I'd like to reduce that even further. I'm not aware of any circuit elements that could accomplish this, though, and any circuit designs appear to consume more power. Further complicating matters is that I have less than 32mm^3 volume available for any additional components to implement this.

I'm imagining a piece of wax separating two sprung contacts, but I don't want to develop something myself, I'm hoping such a device already exists at this tiny size.

Are there other options I can consider?

• ..this will consume almost a microamp of current continuously Sounds excellent to me. You assume that 1 uA is too much but you fail to show why that is so. In a quick calculation I estimate the battery life with a CR2032 coin cell to be a couple of years (only for that 1 uA though). Jan 10, 2017 at 14:26
• They make things called thermal cutouts... maybe they make one that is normally open and closes at the trip point. I don't know how accurate the temperature trip point is. Jan 10, 2017 at 14:29
• Search for "thermal switch" ... apparently there are "normally open" ones. Too big? ebay.co.uk/itm/… The "normally open" ones close on reaching temperature.
– user16324
Jan 10, 2017 at 14:29
• @FakeMoustache I assert that 1uA is too much. It's actually closer to 100nA, but for the purposes of this question anything below 1uA is worth investigating. I'm not going to cloud the issue with the reasons or calculations behind this assertion - either there's a solution given the requirements or there isn't. Jan 10, 2017 at 14:30
• @BrianDrummond Yes, too big. Also these do the reverse as FakeMoustache points out. Jan 10, 2017 at 14:32

A bimetallic switch seems a likely candidate, that consumes no power:

Another fuse-like option is one made of Wood's metal. Standard alloy melts at 70 C, perhaps a different alloy melts at 60 C. Would operate once-only.

Maybe these are too big, but search for Thermal switch, or thermostats- mechanical on digikey... we use the Cantherm ones.

The ones in a TO-220 pack look nice!

A bunch of years ago when the wakeup timer in a PIC took too much current, I used a external circuit to make a low power wakeup. It used a couple of transistors and a few resistors, and a cap for the timing component. It took well under 1 µA. I sortof remember around 100 nA, but it's been so many years that the details have gotten fuzzy.

I think my timer used two pins, one output to reset the timer and a input that would get pulled low when the timer tripped.

I do remember that shortly after that, Microchip came out with a "ultra low power wakeup timer", that would have been good enough for my purposes. It think it used a little more current than my discrete solution, but not much.

Search for PICs with "ultra low power wakeup timer", or something similar to that. I may have the marketing name for this a little wrong, but I know this is a official feature that some PICs have. I'd look around in the 16F1xxx series for starters.

One possible option is to use a microprocessor that supports extreme deep sleep as your temperature sensor. The Atmel ATMega328P-PU (the Picopower version) can wake-up on pin change from Power-Down mode @ 100nA. Once awake it only consumes 200 uA if you keep most of the peripherals off.
The processor has an onboard temperature sensor, which when you wake-up must obviously be at ambient, so for many seconds at least it tells you your environment temperature.
To get a regular wake-up signal you could use the TI Nano-power Watchdog timer. It consumes about 30 nA and has a timer value of 64 Seconds. You could of course use the '328P Timer 2 to count multiple transitions from the TPL5000 to get any reasonable time value you need for wake-up.

That would appear to get you down to about 130 nA current consumption in sleep state and probably well less than 500 nA average even with wake-ups every 64 seconds.

The last question is of course...why aren't you using a processor like the ATMega328P as your main compute element??

One last thought....if the microprocessor you have has acceptable sleep state current you could wake up and use a Maxim 18B20 1-Wire temperature sensor. These have a very high supply current of 3 uA, but you only use them for readings so can keep them powered down until needed.