Measuring Temperature Down to -100C

[SparkFly]

Hi everyone,

I have been working on a project to measure (among other things) the temperature on a balloon flight up to around 40km altitude. I've read some papers indicating that temperatures could be around -60C, but I would like some wiggle room; measuring to around -80C to -100C (if possible), with an accuracy of around 1C.

I began by searching through the various types of temperature sensors readily available through Digikey, Farnell, etc, and it seems to me that the use of a thermocouple would allow me to achieve these sorts of measurements. However, I come across the problem of selecting an appropriate reference IC, as these seem to be available down to only (!) -20C http://www.digikey.co.uk/product-search/en/integrated-circuits-ics/interface-sensor-and-detector-interfaces/2556696.

I was wondering if anyone might be able to offer suggestions as to how I may be able to measure temperatures down to these ranges, bearing in mind the entire system will also be exposed to such temperatures. I am open to being told to scrap the thermocouple approach if needed.

I'm not too concerned about staying within manufacture's temperature tolerances, unless they actually affect the output significantly (eg; I have flown microcontrollers rated at -40C to temps below -55C before). I have a budget of around £30 (~$45 USD) for the temperature sensing circuit, but can go a little above if needed. Output-wise, just about anything is okay.

I hope I've written everything needed, and everything clear enough to be understood. If not, let me know and I will try to explain again.

Many thanks for your time in advance.

SF

[Nauris]

I would use thin-film PT1000 sensor. Not much point using thermocouples if the entire system is at same temperature.

[SparkFly]

Many thanks for the reply, Nauris.

I have tried searching Digikey and Farnell for thin film sensors, however the lowest temperatures that could be measured (that I could find) was only around -70C. Unless there is something which I am completely missing, in which case I am very sorry.

SF

[nali]

Platinum resistance sensor - have a look @ Farnell p/n 541102

[Cloud]

What about using that type of measuring as they use on weather probes. With little resistance wire and measuring how much energy you need to put in to stay at the same temperature.

[Kleinstein]

Pt100 / Pt1000 sensors are an option. However the accuracy of the thin film versions is not that good at low temperatures.
An alternative might be a diode type sensor - they are available for even lower temperature, but I am not sure about the price, tested calibrated sensors might be to expensive. Normal diodes are reasonable good after a single point (e.g. 0 C) adjustment - though not specified and tested. 
Thermocouples are tricky as the reference temperature might also change quite a lot. Also accuracy is not that good for the simple versions.

[Marco]

What about using that type of measuring as they use on weather probes. With little resistance wire and measuring how much energy you need to put in to stay at the same temperature.

That's interesting ... but you'd need a really stable thermal resistance to the environment. Do they put it in a metal tube with some heatsink fins on it or something?

[danadak]

Here is an approach that might work -


http://www.eetimes.com/document.asp?doc_id=1279718

http://www.ti.com/lit/an/sboa019/sboa019.pdf


Its ratiometric in current, not absolute accuracy.

Of course the other issue you have is semiconductor operation down to very cold temps, and reliability.
If you don't care about the latter, eg. having failed flights, then you are good to go. Look at clock generation
in the processor you are going to use, if it is PLL based may be a big issue, buts if its a fully static part, if it
will clock down to DC you should be OK.

Also look at the storage temp on parts, that can be an indication of package integrity at T extremes.



Regards, Dana.

[doktor pyta]

It might be a challenge to make the whole electronics work correctly at -60 ...-100'C, sending telemetry etc.

measuring to around -80C to -100C (if possible), with an accuracy of around 1C.

Do You really really need that ? (or this 1'C was once written in the proposal and just stayed there without big reason...)

[John Heath]

A DS18S20 is good for -55C to +125C . I suspect you could push your luck to  -70C. 3 pin package serial communication power leached from the RS 232 source that is reading it. It looks like a 2n2222 transistor but has a full microprocessor in it. The interface is TTL type RS232 so 0 to 5 volts is fine. The good news is it is cheap so you may prototype to test the waters at  -70C. Digi sells them. I can say from experience that these puppies do not care for thermal shock.  A blast of freeze mist could be the last time it answers the phone. Little squirts of freeze mist to bring it down gently to -70C. It will transmit temperature in Kelvin , C or F if requested ,, if that helps.

[Ian.M]

The DS18S20 does *NOT* use RS232 framing on its serial bus.  It uses the Dallas 1-Wire protocol.  One of the easier ways of interfacing 1-Wire in hardware abuses a UART at a non-standard baid rate, with bit pattens calculated to generate the correct pulses, one 1-Wire bit per UART character, but many libraries simply bit-bang it.

However I would question the sanity of using an active semiconductor sensor in a plastic package directly exposed to such low temperatures for an extended endurance mission with no possibility of intervention if it fails.   A passive sensor is likely to be considerably more robust.

Keeping the electronics working at such low temperatures is likely to be the most challenging aspect.  If the electronics payload is carefully thermally insulated, it may be possible to keep it warm enough from its ordinary power dissipation that a DS18S20 could monitor the payload temperature, then a pair of thermocouples could be used to measure the difference between that and the exterior air temperature.

Dry ice will get you down to -78.5 deg C so DIY  environment chamber testing of your electronics payload is quite practical.

[TurboTom]

For my DIY air liquefier I used plain and simple LL4148 Minimelf Si PN diodes to measure temperature. Forward threshold voltage at constant current excitation is a pretty negative linear slope down to about 100K. Actually, calibration at the triple point of water should be sufficient for an accuracy of +- 1K over your intended measurement range (if this is sufficient for you). If you have access to dry ice or liquid nitrogen, you may use their temperatures for additional calibration (though at the 77K of LN2 the slope tends to flatten out a tiny bit). Depending on the thermal coupling to the environment, a pulsed measurement may be benefical to reduce the influence of diode heating due to the test current.

Cheers,
Thomas

[Ian.M]

@TurboTom: Your DIY air liquifier sounds interesting.  Care to start a topic and tell us about it?

[sarepairman2]

you want to use a speical low temperature RTD or thermistor for this job. they exist  down to much lower temperatures.
http://www.thermometricscorp.com/cryogenic-temp-sensor.html

What kind of thermal mass are you measuring? I.e. something that can be disturbed or something you cannot afford to introduce heat to?


I would use a RTD with a rational polynomial best fit equation.



you might also want to use AC excitation due to potential thermocouple errors. I would use the AD7730.
http://electronicdesign.com/embedded/eliminate-rtd-self-heating-errors

http://www.mosaic-industries.com/embedded-systems/microcontroller-projects/temperature-measurement/platinum-rtd-sensors/resistance-calibration-table#using-a-rational-polynomial-function

[Bryan]

Can you not keep the electronics for the temperature sensor in some sort of heated controlled environment, such as a hacked OCXO chassis. This will keep the electronics in a controlled, stable environment.

[Ian.M]

The application (High altitude ballooning under UK CAA restrictions) is seriously constrained by the payload mass, determined by the maximum theoretical lift. Assuming hydrogen, >100,000 ft target altitude and payload INSIDE a spherical balloon of maximum permitted dimensions (2m), gives a max total mass of about 1lb.  Practical payloads will be considerably less.

As such, I doubt that ovenising a significant proportion of the electronics is an option, due to weight of the oven and extra battery capacity to run it.   However insulating the payload to run the electronics at  about 50 deg C above ambient, starting with the payload at ambient can keep the electronics in a reasonable operating temperature range, assuming launch on a cool day to avoid payload overheating pre-launch.  As a foam casing is required for safety anyway, this is probably the best option.

[Pjotr]

RTD's up to cryo range are not hard to find, i.e.: http://www2.mouser.com/ProductDetail/Heraeus/32207399/?qs=%2fha2pyFaduhue3bUnFFfqkcopKTfDGEtbd2QEHSp1EE%3d

More information: http://heraeus-sensor-technology-us.com/en/home/landingpage.aspx

Be aware that Pt RTD's are not perfectly linear. Manufacturers usually supply resistance tables. Measuring temperature with a MCU can be done quite accurately without calibration because you can measure ratio-metric (relative to your V_ref) and compensate with a look-up table derived from the manufacturers resistance table.

[SparkFly]

Thank you all for your excellent advice.

I think I will begin by examining the use of the platinum resistance sensors in greater detail, although the use of PN junction devices also looks really cool and should be worth an investigation; if only for curiosity.

WRT keeping electronics in a heat controlled environment; this was never really a problem. I have flown about a dozen flights up to ~35km at temperatures that go below -55C and (with simplistic extrapolation) are likely to be hitting -65/-70C. The real killer for me though, is the mass of such a system. I would have to slightly disagree with Ian; the CAA would (has!) allow for balloons much greater than 2m, just need to get 'special' permissions. This is really besides the point as it's more fun trying to see what you can take out/reduce in a payload and still get it working. I've recently made a tracker with a GS module, RF downlink, and coin cell charger that weighs <7g. An oven-type arrangement would be well heavy.

Thanks again to everyone.

[Pjotr]

Some good reading about the basics: http://www.minco.com/~/media/WWW/Resource%20Library/White%20Papers/Resistance-Thermometry.ashx

[SparkFly]

Thank Pjotr, the comparisons were especially useful.