Temperature and voltage references

[Maxlor]

I recently played with different sensors, among them an NTC thermistor. It works quite well, but, trying a couple of different parts I notice that while their precision seems to be better than 0.1°C (at room temperatures anyway), the accuracy is a lot worse. But hey, I can correct for that in firmware, right? ... If I only knew what to correct for exactly. I have several thermometers around my home, but they vary by around 1°C.

So this got me thinking... where might I actually get an accurate temperature reference, accurate to 0.1 degrees or better? Is there something like the DMMCheck but for temperatures instead of voltages?

And more generally: I know I can get my equipment calibrated at calibration offices. But whats their reference? Maybe a more expensive calibration office, but ultimately, what's the physical phenomenon that is measured? And what about temperature?

[Fsck]

temperature: triple point cells = triple point of material.
voltage: josephson junction arrays =  josephson effect

[dannyf]

Boiling water + ice water.

[Lightages]

Actually boiling water temperature changes with air pressure so you need to account for this and you need to use basically pure water.

http://www.csgnetwork.com/h2oboilcalc.html

[Maxlor]

Actually, using water might work. I'm thinking that some pseudo-distilled water together with air pressure readings from either my weather station or the local meteo station should give me a boiling temperature accurate to maybe within 0.2°C. According to the Javascript calculator, 5mbar of pressure difference cause a 0.15K difference in boiling temperature, and according to this guy's experiment, one spoon of salt causes a 1.5K difference in boiling temperature. I can probably stay two orders of magnitude below one spoon.

I'm making a few assumptions here, namely that salts are the most common thing affecting the purity of my water, that all salts are in a similar order of magnitude as far as affecting the boiling point is concerned, and that a worn but clean pot won't add anything noticeable to the water. hopefully I haven't missed too much. I'm thinking that I might need to figure out some way to wrap the wires leading to the NTC, in order to keep them dry. Even though distilled water is supposed to be non-conductive, I doubt the stuff you can buy is really all that pure.

Another accurate temperature point I can get from a clinical thermometer. Those seem to be fairly accurate, even if their range is very limited. But I'll test a few first to see what the variation is.

As for the freezing point of water, that seems to be affected a lot more than the boiling point by impurifications. I could try saturating the water with salt, which should give me a fixed well know point.... except that all the salt I can get here contains Iodine, which probably affects the results as well. Oh. And it just occurred that satured salt water freezes at -21.1°C, and my freezer doesn't actually go that low. Doh.

[Maxlor]

temperature: triple point cells = triple point of material.
voltage: josephson junction arrays =  josephson effect

Are you saying that measuring the triple points are the official way to create measurements?

Exploiting the Josephson effect seems to be the way to create the volt standard, in any case. It's clever. And the guy got a nobel price for the idea, nice! Too bad that super conductors are still a bit out of range for a hobbyist

[sleemanj]

http://youtu.be/EkFmrWsSzgA

[cellularmitosis]

Ever since watching this video on "ice bending",

http://youtu.be/sQdLttUh_b0

I've wondered, when the water flashes into ice, does it also absorb heat from its surroundings and become exactly 0C?  Or does it remain ~-2C?

[cellularmitosis]

In this follow up video he measures the water at -4.5C, and he also states the water "warms up" when it flashes into ice, which made me wonder if it  "warms up" to exactly 0C?

[DTJ]

For your 0C reference use crushed ice & water & continuously stir it.

IF unstirred the most dense water will be at the bottom (obviously!)  - its temperature will be around 4C.

[mazurov]

In this follow up video he measures the water at -4.5C, and he also states the water "warms up" when it flashes into ice, which made me wonder if it  "warms up" to exactly 0C?

No. The water warms up by energy released during phase transition from liquid to solid. The final temperature will depend on the temperature to which water was super-cooled. They set this temperature intentionally to have solid slightly below freezing - you need all 3 states to establish a triple point.

[Fsck]

temperature: triple point cells = triple point of material.
voltage: josephson junction arrays =  josephson effect

Are you saying that measuring the triple points are the official way to create measurements?

Exploiting the Josephson effect seems to be the way to create the volt standard, in any case. It's clever. And the guy got a nobel price for the idea, nice! Too bad that super conductors are still a bit out of range for a hobbyist

it's the calibration reference

[ejeffrey]

Are you saying that measuring the triple points are the official way to create measurements?

Yes, the Kelvin scale (and Celsius) is defined in terms of the triple point of water.  However, it is even worse than that.  Unlike all the other fundamental units, temperature is an intensive parameter (as opposed to extensive).  This means that if you 'add together' two objects at 300 K, you don't get an object at 600K.  Compare this to kilograms, where if you add two, you get 2 kg.  The lack of ability to add and subtract temperatures makes it much harder to define a temperature scale at points other than the reference point.  Instead we have the ITS-90 (international temperature scale of 1990) which specifies a set of fixed points that closely approximates the true thermodynamic temperature, along with guidelines for interpolating between them.

[Galaxyrise]

How much does the shape of the thermistor curve vary between parts?  I would guess a linear RTD, or a AD590 would be much easier to get accuracy across the temperature range.

[SeanB]

They will be within a set tolerance at the calibration temperature, either 1%, 2% or 5% to 20% for the cheapies. The curve will be close to constant per batch, but again will vary from batch to batch but all will be within 1% or better in a batch. You pay more for those that are cut and matched to a closer spec, as then the reject rate is a lot higher and testing is more than a single measurement and a trim.

[SArepairman]

http://youtu.be/EkFmrWsSzgA

wow, I thought a triple point cell required a specially built vacuum chamber... but maybe thats for liquid nitrogen.

that method is pretty accessible to common people.

[SeanB]

Triple point cell really needs just a freezer to get it mostly to ice and then an insulated box where it holds the inside stable while the ice melts.

[Maxlor]

How much does the shape of the thermistor curve vary between parts?  I would guess a linear RTD, or a AD590 would be much easier to get accuracy across the temperature range.

I bought two NTC thermistors specced as 0.5% (NTCLE203E3103SB0). Putting them side by side on my desk and measuring them results in a 0.3°C temperature difference.

And yes, thermistors are quite nonlinear, they're characterized by the Steinhart-Hart polynom which is 1/T = a0 + a1 · ln r + a3 · (ln r)^3 (source). Ugh, 3 constants in there so three measurements would be necessary for calibration. So I'm thinking that building a thermometer with 0.1K accuracy using a 1kOhm platinum RTD and a 22bit ADC might actually be doable for me, with my very modest hobbyist equipment. I can calibrate that with a single measurement point (I'll try distilled ice water. Just have to find a working method for mixing electronics and water. A latex/rubber glove will probably work.)

Once I have the RTD based thermometer, I can use that to compare the NTCs against.

The one thing I don't have yet is an instrument that measures resistance with 0.03% accuracy or better. But I only need that once, I'll figure something out.

[Vgkid]

Your best way to acclomplish this without a super accurate dmm is to buld a resistance bridge.
look up Linear AN 43, or look through here.
https://www.eevblog.com/forum/projects/rtd-pt1000-circuits/

[leppie]

Just have to find a working method for mixing electronics and water. A latex/rubber glove will probably work.

Use deionized distilled water (or PC water-cooling liquid). That should not be electrically conductive when unused.