How to maintain a temperature of -5°C in 1 square inch in an office?
February 25, 2016 9:22 AM   Subscribe

Hi, We need to calibrate some sensors at a temperature of -5 degrees C, also -4, -3... +3, +4, +5. The sensor is similar in size of your headphone jack, so we only need 1 square inch of environment at these temperatures. There's also a cord, which will prevent a completely sealed environment.

We bought a Peltier cooling kit which should get down to -3°C, but usually runs at slightly above freezing in my office environment and it rises to 40°C at the slightest perturbation, such as swapping in the next sensor, and it takes 20 or minutes to get back down to freezing.

We've tried automotive anti-freeze (glycol), and dry ice, and salted ice, but those methods don't allow precise control, and have limited cooling capacity and therefore timeframe before they need to be replenished. They're also wet and messy.

The Peltier also has the disadvantage that, even if it where the desired temp. on the plate, it's room temp a mere centimeter away, so small positioning differences make a huge different in temp.

Should I buy a second Peltier? Get a dorm refrig./freezer? What's the least expensive route to precise temp. control in the -5 to +5 C range?
posted by cyclicker to Science & Nature (14 answers total)
Why don't you just use a fridge?
posted by If only I had a penguin... at 9:23 AM on February 25, 2016

put the peltier unit in a fridge or other cold environment.
if you buy a fridge or freezer, avoid self-defrosting units as their temperatures vary quite a bit during the defrosting times.
posted by sciencegeek at 9:26 AM on February 25, 2016

If you can get your hands on a recirculating water bath and fill it with antifreeze, that's a pretty straightforward setup. If the sensors can't get wet you could put it up against an empty metal container placed in the bath. They tend to be kind of slow to change temperature, but very stable, so you could measure all the sensors at -5C and then change temp.
posted by tchemgrrl at 9:38 AM on February 25, 2016

See if the manufacturer will calibrate them. It'll be more accurate and will save you a ton of headache in the long run.

If you really need a volume of air at that accurate of a temperature, you're looking at some sophisticated environmental test equipment. You're not going to be able to hold -5C at less than a degree of tolerance with a dorm fridge. If this is something you really need to do yourself (and actually be that accurate), look into commercial environmental testing sites like NTS. You'll need to write up a test procedure for them and pay them for use of the equipment and an operator. If you can do the whole thing in a day, you're looking at a cost on the order of thousands of dollars.

A good chunk of my job is environmental testing, ask me anything!
posted by backseatpilot at 9:39 AM on February 25, 2016 [1 favorite]

presumably you have two sensors (perhaps not the same): one you trust and one you are calibrating. if so, then you don't need to control the temperature precisely, you just need to have both sensors be in close thermal contact, cool everything down t below -5, and then let the heat rise. record the values from the sensor being calibrated as the trusted sensor goes through the values -5, -4, ...

to keep the two sensors in thermal contact and to make the rise in temperature slow enough, you might use a fairly large block of metal, with two holes drilled in, close to each other, where the sensors are inserted (perhaps with thermal paste to make a good connection to the metal).

then cool the entire thing down (in whatever way is cheapest - you don't need any fine control) and let it warm up.
posted by andrewcooke at 9:51 AM on February 25, 2016

You need a 'Dry Block Calibrator'

The Fluke 9009 looks perfect, but I'll bet it's excitingly expensive.
posted by Confess, Fletch at 10:02 AM on February 25, 2016 [2 favorites]

In a lab environment, it shouldn't be too hard. You want an outside container (large beaker) to hold a water bath. To the water, add healthy amount of salt. Chill it down with ice. Not sure of the quantities, but you could go as low as 0F (sort of by definition). You will probably need to stir. This is ice cream maker theory. Probably, you will start too cold, and do you calibrations as it warms.

In the bath, you suspend a test chamber (small beaker with cover), held down (but not necessarily covered) in the water somehow. It needs to have a thermometer probe and your test sensors.

(I did something like this in a P-Chem class, but the prof provided the apparatus.)
posted by SemiSalt at 10:07 AM on February 25, 2016

Similar to SemiSalt's answer - prepare a solution of saltwater that will freeze at exactly -5 C. Per Blagden's law, this should be about 0.74 mol (or about 43g) of sodium chloride per kg of water - it will probably be a bit off due to impurities in the salt (such as iodide, d'oh). Freeze a small amount in a regular freezer, maybe in an ice cube tray; put some frozen pieces into a beaker. Right as this solution melts a bit and becomes slushy it should be stable at -5 C for a few minutes. As others point out, you'll probably need to calibrate this with a thermometer you trust, so.

Also something sounds hinky with your Peltier setup; if nothing else those things usually pull a temperature gradient quickly. Are you using an under-powered power supply? Are you literally letting it sit there with the hot side down, as in that product photo, such that the hot air circulates up around the cold platform? You should put a barrier between them or at least find a way to point the cold side down.
posted by Joey Buttafoucault at 10:53 AM on February 25, 2016

See if the manufacturer will calibrate them.

Themo Fisher will do any random thermocouple (or whatever) for me to ISO 9000 standards for $75 (CAD) and to ISO 17025 for around $300. the differences are tolerance and how sure you want to be about measurement uncertainty.

Alternately, Fisher sells traceable, calibrated, digital thermometers for $42 (USD) for that temperature range. So that's your cost floor for a one off.

If you've got a fridge available as If I only had a penguin suggests, and you don't really care about accuracy that much, I'd do it that way.

But what is your traceability? What's your temperature reference? How accurately (and precisely) do you know the temperature of what you're calibrating with? Does that matter?

I might just buy the thermometer and leave it at that.
posted by bonehead at 11:40 AM on February 25, 2016

(The Fluke 9009 looks perfect, but I'll bet it's excitingly expensive: over $5k USD)
posted by bonehead at 11:43 AM on February 25, 2016

Dry ice-ethanol or dry ice-methanol slushies are another common route for preparing a moderate-temperature cold bath. You could make a slushy, and then immerse your system in the slushy and *heat* your apparatus to your target temperature.
posted by janell at 11:55 AM on February 25, 2016

What's the calibration process? If, as mooted above, it's measuring (say) a voltage or a current when the sensor is at a temperature reported by another sensor, and it's not noisy, and your sensor has low thermal inertia, and that it doesn't have to be too accurate (these are the details that really matter in terms of defining a procedure, and you haven't said), then you might be able to get away with

1. Strap sensor-under-test to calibrated sensor
2. Buy tub of ice-cream at local shop, returning quickly to base
3. Put party of the first part into party of the second part
4. Note s-u-t readings at the moment the reported temperatures pass your calibration points
5. Mourn the fact you've been staring at ice-cream all day, and now it's melted

But depending on the details (not) mentioned above, you might be looking at a multi-thousand dollar task rather than a five dollar task. Are these sensors going to kill someone if they're not right?
posted by Devonian at 12:05 PM on February 25, 2016

Going below 0C in ambient conditions can be a real problem because of condensation and freeze up. If you need to do this for any length of time, or through multiple changes/temperature cycles, you will really want a humidity-controlled, low humidity environment, otherwise everything will ice up.

As a rule of thumb as well, for calibration, I generally want something that 10x better than what I'm trying to calibrate. In other words, if you want an accuracy and precision of +/- 1C, your calibration device should be at least accurate to 0.1C and be certified to measure within +/- 0.1C. That gives a fair margin of uncertainty, quite a bit more than 99.5% (or six-sigma) requirements. That means errors in the final calibration are most likely to come from the calibration procedure, not from the calibrating device.
posted by bonehead at 3:06 PM on February 25, 2016

Response by poster: Lots of things to think about. Thanks!
posted by cyclicker at 8:09 AM on March 3, 2016

« Older 6 mile run in DC   |   tired of getting the run-around on this race Newer »
This thread is closed to new comments.