Boost Converter for Extreme Cold?

Hi all,

I’m hoping to ask for some advice regarding a DC/DC boost converter suitable for operating under extreme cold conditions.

I’m currently working an Arduino-based logger to be deployed in the Canadian High Arctic, where temperatures can drop to -40°C and below. We’ll be using non-rechargeable lithium thionyl chloride (LiSOCl2) batteries, which have a wide operating range of -55°C to 85°C. While these batteries have a nominal voltage of 3.6V, this can drop to 2.5V or lower under cold temperatures and high current draw (see graph below). The thought is to run a number batteries in parallel to achieve the desired amp-hour (Ah) capacity for operation over an entire year, rather than also running some batteries in series to ensure we always have enough voltage.

What I’m now searching for is the best way to provide a stable voltage to my Adafruit Pro Trinket 3.3 V, while also keeping quiescent current draw very low.

I happened to have on hand an Adafruit PowerBoost Basic 500, which can provide 5V and up to 500 mA from 1.8V, but I feel like this may not be the most efficient route. It also has a higher than desired quiescent draw of 0.15 mA.

I’d appreciate if anyone could offer some feedback, especially on whether or not a DC/DC converter is in fact what I should be after. I just want to make sure I’m on the right track.


A suitable dc/dc converter isn't your only problem.

The arduino itself is probably , even at best, industrial grade.

You need mil-spec stuff for this, including your logging media. You must look at every individual component and it's tolerance at these extremes.

Packaging and sealing also needs care, in particular as the device warms up , when condensation can be a big problem


I've run Arduinos below -40C. They do suffer a little but mostly it's external components and not the Arduino board itself. I've also broken most of the other rules in the datasheet, like using them as flight controls at this temperature.

You have a pretty specialised usage there. Average DC-DC converters won't do very well at that. One problem is the converters are optimised around some specific voltage difference and using them at basically zero voltage difference (the warm battery state) is less efficient. For example, something like this from Pololu will switch to a linear mode if the input exceeds the output and throw away energy.

Looking at the Pololu catalog, the step-up converters with 3.3V output have much worse efficiency than the higher voltages. I think that is because the unavoidable losses in the converter will be fixed voltages (like a diode drop) which are a bigger percentage of 3.3V than higher voltages.

I would build the battery bank to always require a step-down. Make it 2 cells in series. Then you're keeping the step-down regulator in its most efficient range.

Have you evaluated the approach of running a low clock speed ( 4Mhz) and running the Arduino directly from the battery? This might work for you, unless you require the USB Function.

On another note, you might consider a conformal coating on the board. I would consider one of the silicone spray conformal coatings available at electronics suppliers.

Another thought: Not knowing the high temperature the system must contend with, but for the lower temperatures It might be worth considering an insulated housing whose goal is to keep the internally generated heat in the package in order to warm the battery and circuit.

Before deployment you should attempt to thermal cycle the device as much as you can and as many times as you can. This will precipitate any marginal connections.

Thanks everyone for the great feedback!

@allanhurst Thanks for your comment! I spend a lot of time these days inspecting datasheets for the Operating Temperature Range and adding to a master spreadsheet of every possible component I might utilize. Industrial (-40°C to +85°C) is an acceptable temperature range for us, and -55°C is nice when we can get it. We'll be looking and insulating the components in order to maintain a suitable ambient temperature.

@MorganS Running the batteries in series is definitely something we've considered. The thought was if we could run only in parallel, we might be able to save some money on batteries. A step-down converter however sounds like a good option to achieve good efficiency. I am looking forward to investigating this further!

@JohnRob Since purchasing and working with the Pro Trinket 3V, I've discovered I could likely do with a simpler board that is designed more for ultra-low power operation. I've had my eye on the RocketScream Mini Ultra 8 MHz , which decreases the clock speed from 12 to 8 Mhz, and has a quiescent draw of just 1.7 uA (I've gotten the Pro Trinket down to 19.7 uA). Conformal coating is definitely something we've considered, as well. Given the low-cost nature of the datalogger, we've even thought about just epoxying the whole darn thing! This winter we plan to put the components through some solid stress tests to see how they handle things.

Cheers everyone!

And a really really well insulated container so that any heat generated by components (there's always some) is used beneficially in raising the internal temperature above ambient.

Epoxy has a large thermal expansivity probably not the thing to use for low temperature potting -
silicone rubber (beware the acid curing sort though) is probably available that won't apply any
stresses across the whole temperature range, being compliant.

Mil spec components are usually ceramic to specifically avoid the epoxy thermal expansion problem,
which can damage gold bondwires for instance in chip packaging. Most ceramics have a low expansion
coefficient similar to silicon (alas gold itself is rather higher).

I'm not sure insulation is useful, keeping the temperature constant (ie close to ambient) is going to
reduce cycling and give the highest reliability I think (assuming the device is intermittently powered).
Diurnal variations are presumably not an issue in the high artic?

The Atmel 328P - xx has an absolute max operation specification of -55 to +125 °C and a storage temperature of -60 to +150°C (i'm looking at this file: Atmel-8271-8-bit-AVR-Microcontroller-ATmega48A-48PA-88A-88PA-168A-168PA-328-328P_datasheet.pdf)

Atmel appears to not make a military version of this device (maybe not any device).

It also states that with a 4 Mhz clock the chip will operate down to 1.8V

I personally have no issue with pushing the limits on a device for a one off project and I've seldom had any issues. However one also must consider the consequences of a failure. For me the worse the consequences the more conservative I become. I'm sure you would be bummed if your data logger failed after one or two days.

Below are some thoughts that might help.

  • The most likely failures in cold are mechanical (bad connection) and a failure to start coming out of sleep.
  • When purchasing a board the thing you want the most is good workmanship (soldering). I've seen many boards with simply horrible solder connections. Fine for a hobby but won't stand up to mechanical stress.
  • Thermal cycling is cumulative if your devices passes one cycle it does not guarantee it will pass 10 or 20 cycles. Also, the speed of the temperature change is important, the faster the temperature change the greater the stress.
  • If you can eliminate the DC-DC you would be much better off.
  • For potting you would be much better off with an electronics spray silicone conformal coating. If that is not an option folks have had good luck with GE II. Make sure it is absolutely positively cured if you wish to then pot it in epoxy. As others have mentioned you don't want to pot the board directly with a hard potting.

@MarkT I quickly plotted some historical air temperature data for the month of January at our field site in the graph below. The site is at 82.4° latitude, and it's interesting to see how much variability there can be during an Arctic winter in 24-hour darkness. While temperatures tend to stay quite cold, they can vary as much as 15°C over the span of a couple days.

It's also good to know about the concerns with epoxy under such temperatures. Conformal coating definitely sounds like the way to go.

@JohnRob Thanks very much for your input! It has really helped bring a number of things to my attention, which is exactly what I need. We will most likely be running the Arduino-based logger alongside our existing instruments in order to gauge: a) how well they operate over the course of a year and b) how reliable the resulting measurements are compared to existing commercial solutions. It's very much a proof of concept project.

The DC-DC converter idea has been put aside, and running the batteries in series/parallel now sounds like the better solution moving forward. I've been searching for an efficient step-down converter, and I'm wondering if it would be best to just bypass the Arduino's onboard regulator and supply a stable 3.3V directly. I'm also aware that the internal 1.1V bandgap can greatly vary with temperature, and a stable external reference voltage may also be handy to have. I measured a change of 1.077 to 1.073V in the bandgap of my Pro Trinket just by placing it in the freezer (+20°C to -5°C). I can only imagine how -40°C would affect it.

Thanks again everyone for your help!

Well, a step-down converter is a DC-DC converter. If you can tolerate the variable input voltage then it's best to run it directly off the battery.

Applying regulated 3.3V directly to the Arduino 3.3V pin (for 3,3V Arduinos) is the right way to go. If I was really concerned about power consumption, I'd make a custom board without the original regulator. You don't know if it's going to back-feed a trickle.

I've purchased the good conformal coatings before. They offer features like "solder-through" which means you can just apply a soldering iron directly and the coating burns off harmlessly. Great for repairs and adjustments on your prototype. They smell/feel/look exactly like clear ladies' nail polish. Now I buy all my conformal coating in the makeup aisle. They perform exactly the same as the expensive stuff.