PDA

View Full Version : Calibrating high temperature instruments



Bill736
11-28-2012, 02:43 PM
I've been measuring the flue gas temperature of my oil burner furnace lately, and I find that no two thermometers or instruments I have read the same at around 290 degrees C ( 554 degrees F). I have a glass / mercury thermometer , and two type K thermocouple instruments. I've become a bit reluctant to use the mercury filled thermometer any more , since if it breaks off the mercury would fall into my heat exchanger. I can calibrate all of my instruments at 212 degrees F , of course, with boiling water. But how do I calibrate at higher temperatures? I'm not looking for absolute accuracy better than plus or minus 20 degrees F at 500 F , but I do want consistency. A way to calibrate at , say 350 to 400 F would be better than nothing. I've looked at the wood stove thermometers sold, mostly intended for double walled pipe, and they're rather crude and heavily influenced by surface temperature.

Evan
11-28-2012, 02:53 PM
The melting temperature of 63/37 lead/tin solder is exactly 361.4 degrees Fahrenheit. 60/40 solder has a small crystallizing range centered at 370F. The range is about 10 degrees either way. Coat the thermocouple with soot and stick it in a pool of solder as it freezes. Then heat the solder to remove it. The soot will prevent the solder from sticking.

uncle pete
11-28-2012, 03:14 PM
Bill,
I don't have any links handy, but they make a wide range of temperature sticks that are somewhat like a crayon that are also supposed to be quite accurate and well within your accuracy levels. They melt at the range their rated for. Templaque? might be one brand name. Someone here will know the correct name. Brownell's gunsmithing supply at one time sold them, and may still do that.

Pete

The Artful Bodger
11-28-2012, 03:28 PM
Just thinking about this, can the colour of something be "measured" from a digital photograph and thereby find the temperature?

Evan
11-28-2012, 03:51 PM
Not really. There are too many sources of error, especially in the camera. Nearly all camera do automatic white balance which immediately destroys the colour fidelity. Even switched off there are many steps in the chain before it is presented to your eyes where the balance will be altered. It is impossible to present an image as seen by the eye at the same fidelity and colour gamut via any other imaging device. Not difficult but impossible.

It is possible to exactly characterise every step in the chain so that the sources of error are known and quantified but that does not solve the problem of limited colour gamut.

Forestgnome
11-28-2012, 04:08 PM
You could use the melting point of pure lead.

Evan
11-28-2012, 04:16 PM
Finding some pure lead could be very difficult.

achtanelion
11-28-2012, 04:44 PM
Pure (for reasonable values of purity) lead's fairly easy to get. I can buy it at my local metal store (metals supermarkets). It's sometimes available at plumbing supply houses. For the OP, since he's in the states, he can just order it at http://www.rotometals.com/product-p/leadingotpure.htm.

J

Evan
11-28-2012, 04:54 PM
Lead has a rather higher melting point than the OP asked for at 621.5F. He also probably has some solder at hand.

JoeFin
11-28-2012, 05:05 PM
I'll go out on a limb and say the 2 Type K thermocouples are accurate and it is the means you use to measure the resultant Mv from the Thermocouple you are calibrating.

It always is ...

I calibrate 100s if not 1000s of Type J, Type K, Type E thermocouples in the course of commissioning a Combined Cycle Power Plant or a Delayed Coker Petrolieum Refining plant and it is always the same - we measure the Thermocouple's result and calibrate the controler's measurement.

More often then not the Thermocouple works, or it doesn't. Long lengths of interconnecting thermocouple cable, or special purpose thermocouple terminal spaces or the controller/measurement means might have problems with accuracy - but the thermocouple doesn't. It works or it doesn't.

Any Type K Industrial Grade probe you can find off Eboner that passes an "Ice Point" calibration and your boiling point calibration will be plenty accurate for your needs.

I would be much more concerned with what I was measuring it with

philbur
11-28-2012, 05:21 PM
If the mercury thermometer is industrial quality wont it be more than accurate enough to calibrate your thermocouples?

Slide them all inside a steel tube without touching the wall and heat the outside of the tube with a torch to maintain a stable temperature at the level you wish to calibrate.

Phil:)

Evan
11-28-2012, 05:33 PM
This is a calibration I did of a type K compared to a very accurate platinum/palladium calibrated thermocouple. Vertical divisions are 50C to 900 degrees C. The offset is based on the same reading at 50C. If the type K is offset up by about 25C the average error will be approximately split. The calibration was done in very controlled conditions in my heat treat oven with both probes within a few millimetres of each other. Data was recorded from both instruments simultaneously by photography.

http://ixian.ca/pics10/calibrationk.jpg

Evan
11-28-2012, 05:40 PM
Slide them all inside a steel tube without touching the wall and heat the outside of the tube with a torch to maintain a stable temperature at the level you wish to calibrate.

That will not be accurate. There will be a hot side and a cold side with significant differences in the internal infrared emission. To make it work the sensors will need to be in an insulating jacket such as a ceramic tube within the steel tube.

philbur
11-28-2012, 06:00 PM
Depends on how you use the torch. Covering with sand is also a possibility

The point was you can use the mercury thermometer to do the calibration.

Phil:)


That will not be accurate. There will be a hot side and a cold side with significant differences in the internal infrared emission. To make it work the sensors will need to be in an insulating jacket such as a ceramic tube within the steel tube.

Bill736
11-28-2012, 10:06 PM
Thanks for the ideas . I agree that good quality type K thermocouples ( mine are industrial made in the USA Thermoelectric ) are probably uniform and accurate enough, but it's my instruments that have questionable calibrations. Going through my " stuff", I've found an additional thermocouple pyrometer, and a new old stock dial type thermometer. I think I'll try them all at the same temperature in the flue when it stabilizes, throw out the low and high reading devices, and use the one that is closest to the average of the remaining three. It will probably be accurate enough, and as long as it's consistent will do well for my uses. I'll keep a record of the readings of the other devices as backups. Oddly enough, it's the glass/mercury laboratory thermometer that reads about 40 degrees F low, and it's supposed to be a high quality made in USA device ( from about 25 years ago).

philbur
11-29-2012, 04:40 PM
I would be careful about ignoring the mercury thermometer. I think there is nothing related to its function that could go out of wack. So unless it was low by 40 degree F when new there no reason why it should be in error today. Is it in error in boiling whater at 212 degrees F?

Phil:)


Oddly enough, it's the glass/mercury laboratory thermometer that reads about 40 degrees F low, and it's supposed to be a high quality made in USA device ( from about 25 years ago).

ckelloug
11-29-2012, 05:12 PM
Thermocouples are usually calibrated with stuff like this which checks to see that the voltage returned is correct.

No affiliation but I've been meaning to buy a couple of cooking obsessionists a couple of their handheld thermapen thermometers.

http://www.thermoworks.com/products/calibration/

Bill736
11-29-2012, 05:22 PM
philbur- You make a good point, and that's one reason I'm confused that three type K thermocouple pyrometers read so much higher than the glass thermometer. I just did check the glass thermometer in boiling water, and it's right on the money at 100 degrees C. ( I'm pretty close to sea level , so I don't make any corrections.) The thermometer is a 16 inch long Kessler brand, of Westbury, NY, and was bought at a surplus equipment sale years ago , at the lab where I worked. I do pay attention to the 3 inch insertion line. The maximum temperature on the thermometer scale is 400 C . The mystery continues, but I ordered another type K pyrometer last night to add to the fray.

Evan
11-29-2012, 06:21 PM
I don't know how it might apply to other type K thermocouples but the one I calibrated has a large non-linearity between 50C and 200C. Something to watch for.

h12721
11-29-2012, 09:01 PM
How do you calibrate a thermocouple? Isn't the TC made from two dissimilar metals for a particular TC and that is it.
So how do you calibrate them? I would say you cal the read out instrument.
Hilmar

beanbag
11-29-2012, 09:34 PM
A type K thermocouple is not quite linear, although not as much variation as Evan showed. (must be due to unknown QED effects, lol) My guess is that the two readers are using slightly different compensation algorithms.

One thing you can do is drill a small hole in a block of alu and stick both probes into it. Drill another hole for the solder to melt. Then heat it with a torch and let it cool. See if the two meters agree.

Bill736
11-29-2012, 11:05 PM
After refreshing my memory on thermocouples and thermocouple readout instruments , there are a couple of problems that may be affecting my readings. I've long noticed that one of my instruments gives slightly different temperatures if it has been stored in my cold garage, vs. my warm house. That probably means that the internal ice point compensator circuit is not fully functional. I also read that some thermocouple based instruments are sensitive to electrical noise, and there's plenty of that next to an oil burner with a constant arc between the electrodes. I may take a hotplate into the basement, and measure the boiling point of water with the thermocouple instruments next to the running furnace. I'm still not satisfied with the spread of 55 degrees F between my glass thermometer and my industrial quality probe/dial thermometer , and another 30 degrees higher for one thermocouple instrument, and another 20 degrees higher for the second thermocouple instrument. I've tried switching the type K thermocouples among readout instruments, and the thermocouples themselves don't vary at all.

Paul Alciatore
11-30-2012, 03:00 AM
I seem to recall that thermocouples do not read absolute temperature. They read a difference between the temperature at the junction vs the temperature at the colder end of the two wires. In theory, the colder or ambient end should be in an ice bath to maintain it at 0 deg C. In most instruments, ambient end is at (or perhaps hotter than) room temperature and compensation is made electronically.

Thus, depending on the actual temperature of the ambient end and how the compensation is made, you may have some error due to the actual temperature at the ambient end of the thermocouple's wires. I have a great degree of faith in glass-mercury thermometers and doubt that a good one would fall out of calibration by very much. They either work or they are completely broken. Your 40 degree difference between the thermocouple and the glass-mercury thermometer makes me suspicious that this compensation is either not being made at all or is being applied incorrectly, perhaps with an incorrect ambient temperature being assumed or improperly measured. Are you sure that the temperature at the point where the two thermocouple wires end is actually being used for the compensation? In any case, I would trust the glass-mercury thermometer over the thermocouple until that is proven wrong.

Also, as Evan's curves show, the thermocouple is not exactly linear. But the instrument reading it probably is. Actually all thermometers are non-linear to some degree; even the glass-mercury ones. Thus, marking 100 equal divisions between the freezing and boiling points of water is not completely accurate. And thermometers of different construction will have different inaccuracies. The whole thing is a bit fuzzy. For more on this check this out:

http://en.wikipedia.org/wiki/Scale_of_temperature

Barrington
11-30-2012, 12:57 PM
Assuming the thermocouples are connected using the appropriate cable and connectors, there are a couple of things that can be tested fairly simply.

A quick way of checking the cold junction compensation is to put a wire short across the thermocouple input terminals. The readout should then display the ambient temperature.

If that looks o.k., and you have the facility, then feed in a small voltage via a resistive divider.
(e.g. a 9 volt battery accross 3300 ohms in series with 10 ohms: The voltage to the thermocouple input taken from accross the 10 ohm resistor should then be 9 * 10/(3300+10) = 27.2mV. Best to measure the actual voltage if possible.)

Then, using a K-type reference table, find what temperature this corresponds to. e.g. 27.2mV : 654C, and the display should read 654 + ambient.

A K-type table: http://srdata.nist.gov/its90/download/type_k.tab

Cheers

.

Evan
11-30-2012, 01:09 PM
One thing you can do is drill a small hole in a block of alu and stick both probes into it. Drill another hole for the solder to melt. Then heat it with a torch and let it cool. See if the two meters agree.

Excellent plan, that should work well.

Evan
11-30-2012, 01:32 PM
I just graphed the NIST chart for the type K and it looks extremely linear. Now I must investigate and see what is up with mine. Although, it makes no difference at the temperatures I am interested in for heat treating.

ckalley
11-30-2012, 04:22 PM
The 'Standard Limits of Error' for Type K thermocouples is 2.2*C or .75%, whichever is greatest. That could be some of what you are seeing. For a reasonable comparision between different probes, each one needs to be in the same location (depth) and orientation in the in the pipe and the gas flow going past it must be the same. What you are trying to measure is not a simple proposition. There are a lot of variables. No two probes will ever measure the exact same temperature.

The effects of radiant energy can also cause some significant errors, but I wouldn't expect that to have much effect at your temperature.

Craig

Evan
11-30-2012, 04:26 PM
2.2*c ????????

Barrington
12-01-2012, 09:48 AM
The 'Standard Limits of Error' for Type K thermocouples is 2.2*C or .75%, whichever is greatest.
Or, to put it another way, the limit of error for a 'standard' grade type K is 2.2C up to 293.33C and 0.75% from 293.33C upwards. (to 10.29C at 1372C)

A 'special' grade unit has about half that error.

The error referred to is the deviation from the ideal value listed in the reference table, and applies when the thermocouple is new. Various types of degradation may take it outside these initial limits.

Cheers

.

Bill736
12-01-2012, 12:33 PM
Thanks all for your input, and thanks to Barrington for his tip on shorting the inputs of the thermocouple instrument to check on the accuracy of the ice point compensation. I never knew that, but I went out to my shop last night, and tried it on three thermocouple instruments that had been sitting in the garage for at least a day. At a ( reliable) shop temp. of 47 F, one instrument showed an ambient with shorted input of 36 F. One showed 45 F ( pretty good), and the third indicated 67F . That third one is one that I "calibrated" recently ( with thermocouple plugged in) by adjusting the two little trim pots on the circuit board. ( It's a Taylor type K , model 9800 from about 20 years ago.) One trim pot seemed to be a coarse adjustment, and the other a fine adjustment. However, now I wonder if that fine adjustment is really an adjustment of the ice point compensator circuit. I guess there's one way to find out.