How do I get accurate ambient temperature?
November 13, 2016 9:43 PM   Subscribe

I've discovered that the variety of indoor and outdoor themometers I have disagree with each other wildly. Is there some device I can buy that will give a temperature that is known to be accurate so I can calibrate the others?

I'm going to admit up front that this is a bit of an obsession for me right now. Logically I know that a couple of degrees here or there doesn't make that much difference. But it's driving me nuts. Consider it a hobby of mine. So if you're willing to go there with me, I'd love any advice you have.

This all started when I noticed that a digital wall thermometer I have disagreed with my HVAC thermostat (right next to it in the same room) by about four degrees. I decided to get a "second opinion" by moving another digital thermometer I had to the same location. It disagreed with both of them, giving me a total temperature span of about six degrees. I've since tried with five different thermometers (counting the thermostat), and none agree with each other.

I read that the "red liquid" thermometers are much more accurate than digital ones (it sounds like mercury might be even better, but apparently those are no longer available), so I bought one. But being obsessed (and not entirely sure what I wanted), I actually bought two, of two different brands. You guessed it. They disagree with each other also, by 3-4 degrees. ARGH!

Is there ANY kind of thermometer I can buy that I can be certain will be accurate to within a degree in normal indoor/outdoor temps (say 32F to 100F)? Or is there some way to calibrate the ones I have? I've read about calibrating thermometers with an ice bath, but that doesn't really work for room/weather thermometers. Even the red liquid ones I have aren't really going to be easy to submerge in an ice bath.
posted by primethyme to Science & Nature (13 answers total) 1 user marked this as a favorite
You can still buy mercury thermometers. $40 or $50 will get you a lab quality one.
posted by fshgrl at 10:11 PM on November 13, 2016

I'd start here:
posted by j.edwards at 10:26 PM on November 13, 2016

Sorry, quick clarification. I'm in California. My understanding is that it's illegal to sell mercury thermometers in California (at least, the merchants I've looked at have said they aren't allowed to ship to CA).
posted by primethyme at 10:30 PM on November 13, 2016

Here's one: +/- 1 degree C accurate, NIST traceable. If it's too cheap, I'm sure you can spend more.

I once wrangled 864 direct reading digital thermometers, accurate to +/- 1/2 degree C, in an array outside on a roof. It took about 15 seconds to read them, and I'm somewhat embarrassed to admit how much fun I had watching time-lapse graphical displays of a day's data.
posted by the Real Dan at 10:31 PM on November 13, 2016 [5 favorites]

To follow up on Real Dan, NIST traceable means, technically, that the chain of calibration standards can be traced back to a NIST-verified source, so it's still possible that a NIST-traceable thermometer is not precise. However, as a matter of practice they are.

But I see that you're looking for +/- 1 degree F, no? The Thomas model is 1 degree C. Believe it or not, thermometers of the conventional sort just aren't that accurate. I think you need a thermocouple-based device. I have been using a ThermoWorks thermocouple for many years now for cooking with excellent results. It may be more than you want to spend, but I don't know a less expensive way to do it. Accuracy gets expensive quickly; a +/- .1 F probe, just the probe, is well over $100. You could also borrow a thermocouple and use it to calibrate your other thermometers.
posted by wnissen at 11:18 PM on November 13, 2016 [1 favorite]

Also, if you do want better data from an outdoor thermometer keep in mind that where you place the sensor can lead to huge errors in the readings, even if your sensor is top notch and highly accurate. Is it too close to a source of heat, like the roof or an asphalt driveway? Is it exposed to the direct rays of the sun without any radiation shield? Is it on the north side of a structure in a cool shady area? Measuring outdoor temperature is non-trivial. The ideal set-up is to have the sensor 5 ft above the ground, in an open area where the nature of the ground is representative of the area as a whole. Also, being in an open area means the instrument will be exposed to sun, but the sensor itself will be in a solar shield that is fan aspirated so the outside air is constantly drawn into the sensor, preventing heat from building up inside the shield as well. This setup is more or less how the official stations operated by the National Weather Service are set up.
posted by Seymour Zamboni at 3:59 AM on November 14, 2016 [2 favorites]

That Thomas Traceable thermometer linked by the Real Dan has a working range that easily includes 0°C and 100°C, so if you got one of those you could calibrate it at those temperatures, then use it to make compensation charts for your other thermometers by cross-checking multiple temperatures within their working ranges.
posted by flabdablet at 6:20 AM on November 14, 2016 [2 favorites]

The downside of liquid thermometers is typically the gauge with the temperature lines. What it appears to read can vary with the angle you approach it, the delineations between temperature markings is often quite small, and it is often only loosely connected to the vial of liquid. So even if the liquid is precise, rising to the same point repeatedly for one temperature, if the scale moves, all of a sudden you have the same temperature showing 3-4 degF apart. For calibrating, though, the movement of the scale could be used to your advantage. Rest the base of the liquid thermometer in an ice bath that has been sitting out for a while, and align the scale so its current reading is 32 degF. Then, the key is finding a way to secure the scale, so it doesn't adjust later. If it's secure, it should at least give you a repeatable reading.

As has been said above, digital temperature sensors can get expensive quickly for accuracy, and even faster for calibration. The sensors we use for commercial HVAC testing cost $200-$600 just for the recalibration every few years.

That all being said, the sensor wnissen recommended looks solid. With a +-0.5 degF accuracy, that's about as good as you'll get without spending a few hundred more (and even then, it'll be a marginal improvement, likely to +-0.3 degF). If you get this, and your liquid thermometer is still inaccurate, you could also create your own scale for the thermometer between an ice bath, a warm temperature, and room temperature (and hope that they're in a linear relationship).
posted by BevosAngryGhost at 10:44 AM on November 14, 2016 [1 favorite]

Thanks all.

Any opinions on this? It says +-0.7 degrees F and "Includes NIST-Traceable calibration certificate." The thermocouple wnissen recommended looks like it's meant to be inserted into food, so I wasn't sure if it would work well for air temperature. This thing specifically says it's for air temp. It's slightly less accurate, but still within the 1 degree I was looking for.
posted by primethyme at 11:50 AM on November 14, 2016

That'll work well, too. The main difference you'll see is that the air-specific sensor will be quicker to adjust to temperature changes. The food-oriented model has the same type of sensor (the K-type thermocouple that's mentioned in both), but its ensconces in the probe tip. So any temperature changes will have to affect the thermal mass of the probe before the sensor registers the change.

So if you're only concerned about air temperature, go with the second one. It shouldn't be submerged, though, so if you think you'd like to measure water temperatures in the future, I'd edge toward the first.

Oh, also, it looks like the first sensor has a standard plug connecting the probe to the body. So you could buy different probes (usually ~$20-$60) geared toward air sampling or clamping to a pipe, or replace this one if damaged. But I can't confirm that they didn't do something weird to attach it permanently to the base. Just that the connection that the probe shows to the sensor looks like the common K-type connection.
posted by BevosAngryGhost at 12:53 PM on November 14, 2016 [1 favorite]

The cheapest calibration is to stick the probe in ice water. The phase change means you are holding 0°C to easily within 1 degree.
posted by nickggully at 7:42 PM on November 14, 2016

That's what we did to "calibrate" our lab's bimetallic-type stick thermometers to "close enough for our purposes". Adjustable scale face. Ice water. Make it read zero. Then boil it and check (without moving the scale) 100C. +/- a degree or so was acceptable, outside of that we threw them away.

Even still, I'd see a degree or so difference between two thermometers sometimes, from not-exactly-linear response over the range.
posted by ctmf at 9:28 PM on November 14, 2016

For even more unnecessary precision in the two-point cal, correct boiling temp for ambient pressure with a barometer. (But is that calibrated?)
posted by ctmf at 9:31 PM on November 14, 2016

« Older Comfort when your hero "dies."   |   Rebuild or replace my laptop Newer »
This thread is closed to new comments.