You are in the realms of measuring very small temperature differences. Do you have access to any micro-scale welding apparatus and know someone who can use it! I would suggest that you make your own thermocouples using very fine wire, adding radiation shielding if necessary. For industrial applications the standard bead is several mm in diameter and enclosed in a sheath (for robustness). You need something much, much finer than that.

I don't know where you are located, but if stuck, talk to thermocouple suppliers in Europe or North America (USA or Canada). Many will give free knowledge/advice if you are doing fundamental research.

You will need a measuring device - it will need high input impedance. In addition, if your thermocouple output is in tiny fractions of a milli-volt, then I would suggest that you use a true second thermocouple in an ice bath rather than an automatic (assume internally generated) second junction.

Trust that the above helps.

Regards Stephen.

Harun,

I would look at using a differential thermocouple, rather than two separate thermocouples.

https://cdn4.explainthatstuff.com/measuring-temperature-with-thermocouple.png

Take another K type, and wire them so that you have:

J1 J2

A----------/\_______/\--------------B

The wire from A to J1 is one of the K type metals, the wire between J1 and J2 is the other type. And again, J2 to B is the first metal type.

The temperature difference between J1 and J2 will directly lead to a voltage.

Or, if you prefer, build a potential divider out of platinum resistance thermometers.

(larger voltages, fewer problems with noise and amplification)

Getting much better than +-0.5K with a TC is hard - unless you've calibrated it, and have shielded twisted pairs all the way, and a really good reference bath.

I suggest you check wire and welding arrangement for classic DTA (differential thermal analysis) with two type K thermocouples at Figs. 5.1 and 5.2 (pp. 200 and 201) at my Ph.D. thesis ― Please note that DTA is a typical case of measuring a very low voltage signal (temperature difference) from two thermocouples heads, often within the μV to mV range, while at least one temperature signal (mV range) is also measured:

Thesis Cristalização de Biomateriais Vitrocerâmicos e Mineralização...

James,

Do you mean getting better accuracy than +-0.5K is hard even for differantial thermo couple configuration? If it is so, I think no need to effort for prepearing differential thermocouple because 0.5K is also high.

Do you think two distinct DIN EN 60751 class AA Pt-100 is appropirate for this process?

Harun,

No.

WIth a *single* thermocouple (a hot junction, and then a cold reference junction some distance away, and a very nice amplifier) it is challenging to do better than +-0.5K.

https://www.thermocoupleinfo.com/type-t-thermocouple.htm

However, if you have a differential arrangement, as I and Carlos described, where the two junctions are very close physically, then one generally can do far better - much better than having two separate thermocouples.

I add a drawing.

You can either measure V1 and V2 (upper sketches) and take the difference, and do some arithmetic. (the cold junction references are not shown in the upper sketch, for clarity)

Or, measure V3.

The choice of thermocouple over Pt elements depends on how rapidly your temperatures might be changing - and the environment that they are in.

Pt elements are generally more massive, slower to respond, but because of their flat nature, might be easier to stick to whatever you are measuring.

TCs are smaller, faster to respond, and a little hard to attach reliably.

You are in the realms of measuring very small temperature differences. Do you have access to any micro-scale welding apparatus and know someone who can use it! I would suggest that you make your own thermocouples using very fine wire, adding radiation shielding if necessary. For industrial applications the standard bead is several mm in diameter and enclosed in a sheath (for robustness). You need something much, much finer than that.

I don't know where you are located, but if stuck, talk to thermocouple suppliers in Europe or North America (USA or Canada). Many will give free knowledge/advice if you are doing fundamental research.

You will need a measuring device - it will need high input impedance. In addition, if your thermocouple output is in tiny fractions of a milli-volt, then I would suggest that you use a true second thermocouple in an ice bath rather than an automatic (assume internally generated) second junction.

Trust that the above helps.

Regards Stephen.

Thank you James, Carlos and Stephen. I could not respond quick since I was not able to work on it. I will share my feedback here for people who may have similar problem in future as soon as I test it .

Stephen,

I also have Omega Class 1 K-type thermocouple wire, It's diameter is 1 mm (including insulation). I was not aware of existance of micro-scale welding appartus, is it crucial? I have standart thermocouple welding appartus which was used for welding thicker wires before. As far as I know, thinner thermocouples are better for fast response time, do thickness also efffect accuracy?

Best Regards.

Thick wires affect accuracy in several ways, e.g. - (i) conduction along the wire, (ii) Radiation losses away from bulb and radiation to it if it is near any hot source (you may want to measure gas temperature near a hot wall), and (iii) thermal mass in a transient system (response time, etc.). There is quite a lot of info. on thermocouples in the literature, especially that published in the last century - it's not all out of date. If I can find my old lecture notes I'll try and up-load a few pages. I was trying to clear space the other day, saw their age and the rest is history.

By micro-scale welding apparatus I'm just thinking of the ability to make a very fine, hot flame and experience in using it as it will completely melt and blow away a very fine wire unless carefully controlled. If you have any means of testing the thermocouple, you could simply lay one wire on top of the other and place the pair under a fine spot welder. You don't want contamination from the welder tip or loss of the liquid weld - just enough energy for fusion. You simply then trim away the surplus wire.

Don't forget that you will need compensating cable to go between the thermocouple and the measuring device. For the measuring device, you may want to talk to a US manufacturer called "Fluke". They do good hand held apparatus. If you are going to a data log then get the correct input amplifier, especially for transient analysis. Any energy needed to drive the input section of any device, even the best available, must come from the thermocouple and for precision sensing you need the minimum loss. Modern technology helps here, but I don't know your budget.

Do you know the temperature range that you are interested in? Remember that radiation is a T**4 function and that you work in degrees absolute. I'm not trying to teach you the obvious - I'm sure that you know it; just be aware.

Keep plugging away but know where you are heading. Remember the adage - research is 1% inspiration, and 99% persperation.

If you are going to use two thermocouples, try to:

1) choose exactly the same lengths of wires for both thermocouples.

2) during the calibration, keep the thermocouple beads as close as possible to one another.

3) once installing, place the beads at the same radial distance form the tube center line. If one is close to the tube wall and the other is close to the tube center, it may show a higher temperature difference.

Dear Harun,

Using a differential configuration as proposed by James Garry, I have measured temperature differences of the order of a few hundreds of Kelvin using K-junctions. Commercial thermocouples are very thick and useless. Indeed using ice bath as reference is not useful at all.

A key problem is heat conduction along the wires and radiation losses. I used 30 micron thick wires. Especially in measuring temperature of air or gasses it is very difficult to avoid heat transfer along the wires. The problem is even more dramatic when the gasses are almost stagnant (not flowing hard). I stretched thermocouple wires of length/diameter ratio of the order of 300 in the gas flow to avoid heat conduction along the wire. Also a good thermal contact of the reference wire with the wall is a key factor. In low temperature research very thin wires of Copper-Constantan junctions are used (down to 10 micron). While K thermocouples are difficult to make for very thin wires, the Copper-Constantan wire are easily soldered.

Note that the good news is that if you use one junction to a wall, you can use the same junction as reference of multiple other junctions. I did use the same reference junction in a wall with 5 measuring points in air. Of course measuring temperature differences along your wall is very easy. I was measuring the difference between the reservoir temperature in a wind tunnel and the (adiabatic) stagnation of a wire stretched across a free air jet flow (Mach numbers between 0.1 and 2).

Good luck

Mico

First of all I would like to thank you all for your precious suggestions. After doing search for a while I realized that devices in the laboratory are not suitable for such measurement. Our data logger is like a black box, it just give the measured temperature I couldn't get voltage information. It has built-in cold junction compensation but I am suspicious about accuracy of its referance temperature measurement. Therefore I decided to apply this solution; I will keep the temperature difference as large as possible and I will measure it with Pt100, after I found heat loss, I will try to measure heat conduction through insulation if I catch agreement between these two measurements I will consider heat loss through conduction for rest of the experiments. Since temperature difference between surface of heater and outer surfeace of insulation is much bigger I think this method would be much easier.

Dear Harun,

Please note that if you have a voltmeter able to measure micro-volts you can carry out differential thermocouple measurements without any other equipment. For a K-type thermocouple the signal is 41 micro-volt/ K. The thermocouple wires are standard if you restrict yourself to 0.1 mm wires. I never used compensation wires, but only thermocouple wires. Compensation wires are only useful if you place your voltmeter far from the set-up. Again, the great advantage of measuring directly the temperature difference (between a reference junction at the wall and a second junction) is that you do not need any complex apparature and you do directly measure the relevant temperature difference.

Good luck

Mico

Small diameter RTDs are the most accurate generally. However, by using two small diameter, bare-wire, type K Cromals, one at each end but using the inlet one as the cold junction it is possible to measure very small DTs very accurately. However they need to be calibrated at two points usually (as mV vs Tis linear) as the precision is not good for CrAl.

Dear Dr. Hirschberg,

Would you mind sharing the accuracy of voltmeter that you use for such small temperature difference?

I carried out the measurement back in 1973-75. At that time I used a vaccum tubes voltmeter with a high impedance ( even for that time the voltmeter was old fashion. At least 10 years old). Results were surprisingly good. Reproducibility was a few micro-volts. I guess that modern micro-voltmeters are much better.

You may have thought of this already, but have you considered the instruments that the technicians use when aligning satellite aerials (domestic and otherwise). They are looking to measure microvolts and the instruments have a high input impedance. They are mechanically robust. I don't know what they cost and in some places satellites are a sensitive subject, so be prudent. In the UK, my first approach would be to Radiospares. In the USA I would look at one of the Californian outlets. For the instrument itself, Fluke do a good range as do AVO.

Regards Stephen.

I actually used a voltmeter in its lowest range. I tested my temperature measurement by comparing results in liquids with measurement with a old fashion mercury thermometer with sensitivity of 0.01 degree celcius.

As mentioned by Stephen Lloyd, Fluke makes excellent instruments. I used them later a lot. Typically the Fluke 8845A /Fluke 8846A volmeter has a minimum range of 100 mV but 6.5 digits and an accuracy of about 0.0035 % of measurements + 0.0005% of the range. The measurements uncertainty is therefore about 5 uV. One should also look at digital data acquisition system. Using such instruments with a battery might increase accuracy/stability .