Thermometers: What's the Problem with Accuracy?

[(PeteCresswell)]

I've been trying to buy four refrigerator/freezer thermometers and it's frustrating how inaccurate/inconsistent they are.

For Instance:

- 4 thermometers of same make/model, hanging on the same rack.

- Alcohol bulb/glass tube - you'd think basic physics...

- Each reads a different temp from 68 to 74 degrees.

Only thing I can think of is that the card behind the glass thingie is misaligned, but looking at these things, there does not seem to be enough physical room for those kinds of errors.

Also, at home I have a couple of digital indoor/outdoor thermometers.

Bring the outdoor sensor inside and put it right next to the master part, and they're a couple degrees different.

Is there any hope at all of finding four thermometers which:

- All read the same temp when placed side-by-side

- Read the correct temp.

What am I missing?

[hubops]

[dpb]

Looks nice but no spec's on absolute accuracy so no way to know if are any better than the run-of-the-mill cheap consumer thermometers Pete's already tried.

Lee Valley probably can supply data and there's at least a decent chance they may meet the need, but wouldn't bet too much on it without checking first before laying out $50.

Of course, highly accurate thermometers/sensors are available from suppliers for lab/scientific measurements, but one pays for the precision/accuracy.

I don't know otomh what one could find from Omega these days in the lower-end price range, but there's one place to look.

[Ralph Mowery]

The best way to get a thermometer that should be correct is to look at all of them in the store and see how many are showing the same temperature.

If possiable you should take one and place it in a solution of ice and water. It should show 32 deg F or 0 deg C. While this will only give a one point calibration, it should hold close enough for normal ranges of food or human comfort.

[gfretwell]

Spending enough money to get one that has a little quality control in the manufacturing phase? My Bacharach sling always shows to be the same as the lab grade thermometer we use to calibrate the instruments we use for state water samples but they cost $80+ Digital stuff itself is what I call the great lie. They give you a very precise reading, that is usually wrong. I have an $8000 (MSRP) digital water tester from the state that we have to calibrate before and after every use and about 10-15% of the time, they do not pass the verification calibration after the test on at least one parameter.

[micky]

That only works if the bulb is exactly the right size and the amount of alcohol or mercury inside is exactly correct. I don't want to spend too much just to get accuracy that won't really help me.

Last time I bought a thermometer, about 40 years ago (not counting a digital one that claims to measure humidity), a glass one, there were about 8 on the rack and I looked at each temp and bought one whose reading was part of the cluster of readings very close to each other.

What more could I do.

I don't think the card is the problem. Glass thermometers sometimes or always have a scratch, a file mark, at 32 degrees, on one side of the tube, I've read. I've never looked closely and I've never seen the mark, but it makes sense (and it contradicts my first paragraph above, come to think of it). They'd put all the thermometers in a mixture of melting ice and water, maybe even in a refrigerated room. Pull them out one at a time and file the scratch wherever the level happens to be.

Cynicism. Hopelessness.

[rbowman]

There are a lot of slips betwixt that cup and lip. I worked for a company that made pH and ion concentration meters. First you start with the sensor, which was a Ross electrode in this case. They aren't identical out of the gate and their characteristics change over time. The output is fed to an AD converter, which introduces sampling errors depending on how good the converter is and the rate at which you read the output. There is noise that has to be cleaned up and finally a number is obtained. That is fed through the algorithm is question to get a ion concentration or ph. Finally you wind up with the magic 5.376, which is precisely displayed on the readout.

Thermistors, thermocouples, glass electrodes, LVDTs and so forth are a pain in the butt to calibrate and keep in calibration.

Philosophically speaking the entire project to convert the real world to mathematical equations has been a product of human hubris. It's been handy at times but it isn't real.

[gfretwell]

We use what we are used to. OTOH Fahrenheit gives you about twice the precision without resorting to decimals. I am comfortable with both since my science friends are all C

[dpb]

...

...

Precisely...since US doesn't have the culture that guv'mint can enforce stuff like this universally on the general public, popular usage remains (similar as to having kept almost exclusively other US/Imperial measures as well).

I'm a nuclear engineer by BS, MS Physics (Nuclear Science) and spent over 30 year in the traces and still, while I'm capable of thinking through roughly what C corresponds to F and vice versa, it's still only just native F that really resonates; it's just the system that is ingrained.

While in the reactor design world, all the nuclear calculations were in conventional metric units (scaled for the purpose, for example, nuclear absorption cross sections are in "barns", a barn being 10E-24 cm^2) but all the thermo guys were in English units so we consistently used core inlet/outlet temperatures of 555/605F and had (and still remember many) values of the steam tables such as that Tsat at 2200 psia is 649.45 F (although those are like 1977 Table values; they've been updated so probably a hundredth or few off currently-published numbers).

I'm sure in the now almost 40 years since last actually worked inside a reactor vendor (spent last 20+ yr consulting for various utilities and US DOE rather than with vendor) the design protocol will have all moved to metric just for the consistency. We began the process even while I was still at the vendor when teaming with German and French partners; their licensing authorities at the time required it although it never changed our practices in the US.

The argument has raged forever and while there are valid reasons for metric in business and industry and inside those fields things have gone almost universally that way, there's just no overpowering reason that familiar units must change so the status quo remains and new generations continue to be reared with no practical day-to-day acquaintance with them so they also are only innately comfortable with F for temp and the cycle continues...

[Cindy Hamilton]

I've never understood the "precision" argument. Does it matter whether it's 70, 71, or 72 F ? I suppose 31, 32, and 33 F are more significant.

Cindy Hamilton

[(PeteCresswell)]

Per [e-mail]:

But the question is: if you put five of those things side-by-side, how many of them will read the same temperature - and will it be the correct temperature.

[dpb]

Ask Lee Valley for spec's and see if they are anything but fancied up Chinese imports or no...I'd guess probably not, but ya' never know 'til you ask. :)

[TimR]

Exactly.

Unlike the metric system for mass, length, volume, etc., there is no calculation advantage to using C over F.

They are both 100 point scales invented by Europeans. The C scale is 100 degrees between freezing and boiling, the F scale is 100 degrees between how cold and how hot it normally got outside in Europe.

[hubops]

They are accurate. Buy a more expensive one if you want ...

• / - 1 degree

John T.

[hubops]

• / - 2 degrees for $ 10.

John T.

[(PeteCresswell)]

Per Cindy Hamilton:

My take is that when you're out in it a couple degrees, or even 1 degree F can be significant and it seems easier to say "Fifty-One: than "Ten-point-six".

Or do people who use Celsius just drop the decimal and say "Ten-six"?

[dpb]

...

Isotech TTI-22 True Temperature Indicator Thermometer with 0.001°C accuracy and 0.0001°C resolution

And, it's a bargain at _only_ $6,073.00!!!

But, you get:

Includes FREE Ground Shipping FREE Lifetime Tech Support

:)

Omega has a line of low-cost glass lab thermometers but don't specify accuracy on them, either; they have others of lab quality that are $10+ or so each instead of $5-6.

They used to have an ambient-air monitor that was pretty good that we used to use to monitor air temp around the boilers but it doesn't seem to be available any longer and the one other I looked at from another vendor was also listed as discontinued. That's all the time I had...

The problem with inexpensive bulb thermometers is that the capillary dimension has to be exceedingly precise in order to control both the linearity and scaling and doing that drives the price up -- actually, like matching v-belts; to get precise ones they would just select from the run those that matched given calibration but as somebody else noted, the ordinary Walmart/drugstore/etc., ones will not have gone through much if any screening at all; what you get is a measure of the manufacturing process variability by doing the comparisons you've done.

But, the time/effort required to do screening adds up so even moderately accurate ones get to be fairly pricey pretty quickly.

If a bulb thermometer serves the purpose, probably one of the secondary-school lab suppliers or the like will be the best bet.

[dpb]

On 2/4/2018 3:29 PM, Cindy Hamilton wrote: ...

Well, there's the same question around 212, too, from that standpoint... :)

I ran into a minor instance here just recently -- fixed up a monitoring system in the well house to alert me if the temperature drops that indicates I need to go relight the gas heater (occasionally the high KS wind will manage to blow out the pilot). I used a cute little 41F and we had some bitter nights shortly thereafter and it triggered. So, not wanting to try to cut it too close, 1F is ~0.5F so lowered it to 4.5F and added

3/4" of foam packing between the solid wood board mount and the unit.

So far, it's not triggered again but it's not been below zero again since, either; we'll just "hide and watch" and see how it goes.

But, long-winded that got me wondering; are European thermostats/thermometers registering tenths or are they using 1C setpoint tolerances? I can tell difference in our house between a 69F where we normally keep for winter and 70F setpoint; there's a significant run-time difference as well so if it were 20C or 21C that were the choices, the 20C --> 68F is notably chilly and so 70F would be next best granularity could get and that would, over time, be observable in heating bills.

Not huge, but "yes, I think it can make a difference" in ordinary things...

[Boris]

Cindy Hamilton wrote in news:2edfd7c7-d9f7- [e-mail]:

2,40733,40734

For me, the difference is important. The human body can tell a 2 degre difference. 69F is cold for me. 71F is just right. I can tell the difference.

[Vic Smith]

Yep. Pretty easy to tell if the milk is cold or the ice cream is melting. Out of curiosity I checked the 3 thermometers in my house. They are an "alcohol" wall thermo, digital thermostat, and digital meat thermo. All of them read 70F. Maybe those measuring cold are inaccurate.