Background: I have two test/measurement setups that are measuring the same phenomona. I am using the same thermocouple harness in both setups. This harness comprises of 8 K-type thermocouples paralleled for averaging. By shooting a signal into the connector for the harness, we calibrated the system. The harness was verified (for function) earlier. It checks out to w/in tenths of a degree F. The thermocouples typically read between 800-1500F. The connector heats up to about 400F (give or take about a 100F).
Problem: I am loosing about 8 degrees F from one setup to another. I tend to believe the 8 degrees difference.
Any ideas?
Also, any recomendations for a good book or something (in addition to Omega's stuff) for learning about thermocouples. I understand the basics about Seebeck, but it is still voodoo chemistry to me. How do you model such circuits. How would you model a short to a known medal kinda stuff. That 8 TC harness in parallel would be nice to understand.
Background: I have two test/measurement setups that are measuring the same phenomona. I am using the same thermocouple harness in both setups. This harness comprises of 8 K-type thermocouples paralleled for averaging. By shooting a signal into the connector for the harness, we calibrated the system. The harness was verified (for function) earlier. It checks out to w/in tenths of a degree F. The thermocouples typically read between 800-1500F. The connector heats up to about 400F (give or take about a 100F).
Problem: I am loosing about 8 degrees F from one setup to another. I tend to believe the 8 degrees difference.
Any ideas?
Also, any recomendations for a good book or something (in addition to Omega's stuff) for learning about thermocouples. I understand the basics about Seebeck, but it is still voodoo chemistry to me. How do you model such circuits. How would you model a short to a known medal kinda stuff. That 8 TC harness in parallel would be nice to understand.
"Joshua Guthrie" wrote in message news: snipped-for-privacy@l41g2000cwc.googlegroups.com...
I am troubled by the idea of "averaging" the thermocouples by putting them in parallel. Thermocoules are voltage sources and you "never" average voltage sources by placing them in parallel. For them to give an assemblance of a true average, the junction resistance, lead length, termination resistance, etc. would have to be exactly equal on all devices for a proper average to occur, otherwise they rob from each other in unintended ways. I suspect your proplem is related to this un-equal resistance issue. Thermocouples should be averaged by placing them in series then dividing the resulting millivoltage by the number of couples, but this requires more advance signal conditioning. Usually multiple thermocouples are arranged with each on it own separate channel then any required averaging done by a computer that's also doing data logging. I am also concerned about the connector being at 400 deg. Is this system properly copper compensated? The cold junction has to be known and either controlled in temperature or compensated for electronically to get accurate readings. If the connector is a true chromel-alumel block and wired properly this may not be a problem, but if it's copper or brass, it could be. I think I would look at the averaging scheme and answer why the connector is so hot. Needless to say that "any" thermocouple placed in the same environment should give nearly the same result and not be off by eight degrees. Good Luck. Bob
I really don't suspect the harness. Someone a whole lot sharper than me came up with it and it was tested prior to our runs. In fact, I don't honestly know much about the harness. All I could dig up on it was an elementry schematic where it appeard to be 8 TC's in ||. It looked like they were doing the averaging through choosen reistances on each leg. But the schematic was only an overview and really told me nothing. However, the design has been in use for about 20 years. And the same harness was used in both tests.
I would like to learn more about the harness and plan to do so Monday (esp in light of this discussion).
It looks like they are using K type connectors. I saw an Al stamp on one side (the other one was crimped in an unfortunate place). They are not copper or gold. I did notice that they look to be using a steel nut on top of the connector. I assume it would make a junction, but would it affect the overall circuit? I would think the steel side would just be on open circuit and just floa --that it would just add noise, if anything.
Bob is right on the mark here; your paralleled thermocouples force current through each other with any temperature differences, eliminating any possibility of accuracy. The hot connector may also be a problem unless it is at a uniform temperature and has identical thermocouple grade wire (not extension grade) on both sides.
A few references from my old notes, hope they still work:
we use the same idea on grease pumps where the hose needs to be heated, there are 4 K thermo's running along the feed pipe in 4 evenly spaced locations around the pipe. they are all tied together at the controller. the idea is to get a M(mean)/average reading of what the center of the stream would be and it works out well.
Paralleling thermocouples means that the average is weighted by the resistance of the individual thermocouples (including the contact resistance of any connectors). The legs of the thermocouples have TCRs so you may be messing things up fairly seriously if you're running the legs at high temperatures. A few modest precision resistors >> the thermocouple loop resistance (at room temperature, and with no gradients nearby) could work wonders in swamping those effects if you keep them cool and with the junctions at each end at the same temperature. You should NOT run the connectors at anything like 400°F if you expect to see reasonable accuracy. Finally, CRIMPED? That's very bad. It's possible to crimp gas-tight connections with the right kind of connectors and a very expensive crimper, but generally speaking that's anathema to accuracy. Welded connections for high or low temperature and tight screw terminal connections for around room temperature. Oh, and it should go without saying that the junctions must be isolated from each other for this to work, otherwise you'll have an unholy mess indeed. Even to achieve 5°C accuracy can require a bit of care in thermocouple systems.
Best regards, Spehro Pefhany (temporararily in Li4 jiang1)
Thanks... You guys are increasing my education by leaps and bounds.
Out of curiosity, why is it bad to crimp relatively high temp connectors? I wouldn't have thought 400F that hot. Corrosion making excess junctions? To make you feel better, the harness isn't crimped, just the connector to my system. But it is in a hot location.
The responses are correct about averaging is weighted by loop resistance. But the engine exhaust people have beeen using exhaust thermocouple rakes for years. They typically use sensors all of the same length, style, geometry, etc. so that the individual loop resistances are the same.
You can easily check to see if your loops are the same with a DMM. If they are not then that could be your problem.
As already mentioned, another problem source could be your interconnections. Especially if you are making connections in thermal gradients. Try seeing if local heating of a connection alters the output of the sensor by passing a heat gun or torch across the connection to give a sharp gradient across the hardware. You shouldn't see a change in the output but if you do then you may have a problem at that connection.
You can do this same test all along the length of a thermocouple to look for thermoelectric inhomogeneities away from the junction. It's not a precise test but it can highlight trouble spots. Type K can have quite a bit of calibration shift due to short range ordering, grain growth, alloy diffusion, and cold work in bent areas. Have the 2 setups you are comparing been through the same processes or is one new and the other used?
Also note that standard limits of error for type K at 1500°F is 11°F so consider that when you are trying to reconcile the 8°F difference. And saying that 2 averaging thermocouple rakes calibrate to within a few tenths of a degree for type K is stretching it. I would imagine the calibration uncertainty is much larger than a few tenths. Check out NIST's website for calibration uncertainty for type K.
Someone has already recommended Kerlin's book and he had another with Bob Shepherd from NIST called Industrial Temperature Measurement. Professor Kerlin is at U of Tennessee and he holds the copyright now to the book. There's also ASTM's "Manual on the Use of Thermocouples in Temperature Measurement". NIST also maintains a good site and the people there are very helpful. Bentley from Australia CSIRO has a more recent set of 3 books called "Industrial Temperature Measurement" or Thermoemetry or something like that.
Because the crimp joint will probably loosen and will likely not remain gas-tight due to thermal cycling. This will result in a bad connection- in your case one of the thermocouples might stop contributing to the average, in more typical systems the break protection circuitry will cause an error in the reading.
Best regards, Spehro Pefhany (temporararily in Li4 jiang1)
I have another question. Somewhat unrelated. A good portion of the literature I've seen so far relates thermocouple EMF's to the substances Seebeck Coeficents. However, I've not seen anyone allude to where a person can find a good list of these tables. I've found lists for the normal TC types. But when you're using Seebeck Coefs. wouldn't you normally be interested in something weird, like a jucntion between Chromel and Steel? Any ideas. Perhaps I'm still missing something.
Oh, also, I took apart my terminal. They were silver soldered. They looked like they were just crimped because of the way they put additional insullation over the terminals (chafing has been a past problem).
You can get some nominal information on the thermoelectric power of metal elements in some reference books. Goodfellow's catalog has a good listing for elements and some alloys but this is nominal information. There's a little bit of information on binary alloys if you search around. Try Kinzie's book "Thermocouple Temperature Measurement". You might also look into the Symposium series, "Temperature, Its Measurement and Control in Science and Industry". As well as IMEKO and TEMPMEKO.
Finally the National Metrology Labs like NIST and CSIRO are a good place to start or to send questions.
But you are asking for information that may not be there. For example,
300 series stainless steels weren't developed for thermoelectric properties and so there's little motivation to characterize those properties. Also slight chemical or metallurgical variations could change the properties so it would be difficult to establish a standard reference table or function.
I read in sci.electronics.design that snipped-for-privacy@westmouthbay.com wrote (in ) about 'Thermocouple Problem...', on Thu, 24 Mar 2005:
Google with 'Seebeck coefficients' gave 5600 hits.
To understand thermocouples, you need to look at the metals and the temperatures all round the whole current loop. You also need to understand that the thermal voltages are generated in **the wires that have a different temperature at each end**, not at the junctions between dissimilar metals.
In a simple case, with one chromel-alumel junction feeding a moving-coil meter with a Manganin swamp resistor, the loop looks like (use Courier font):
If the whole Brass-Copper-Managanin-Brass part is all at the same temperature, the only voltages left are those generated in the Chromel and Alumel wires, because they are the only ones that have a temperature gradient between their ends.
You can put a piece of any metal anywhere in the loop and it has no effect IF it is all at the same temperature, so you don't need to know its thermoelectric properties.
--
Regards, John Woodgate, OOO - Own Opinions Only.
There are two sides to every question, except
'What is a Moebius strip?'
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk
Hi, Tony. Google seems to have eaten my reply, so here's a short and late version.
About 10 hours north-west of Kunming by road (Southwest China). Beautiful place with great food.
Just a temporary break, but if I end up going to Yunnan University for a semester in the fall, I'll probably have to take a longer one-- high speed ADSL internet connections there generally are barely usable, IME (though domestic sites seem to load quite fast). At 1500+ messages a week, SED is pretty high traffic.
Best regards, Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Speff, that would be great if you can take class there. You might do better than you think for web access. I do all my surfing with Win 3.1 on
28.8k dialup as this is available in most parts of the world. Many hotels offer dialup connections.
It's plenty fast enough. I have no problems keeping up with news sites, slashdot, and all the newsgroups and mailing lists. Very often the connect time dominates. Most of the time, I turn images off. The eye candy adds little to the information content. The only problem is Acrobat files. Newer formats can be unreadable.
The old software is immune to viruses and trojans. Yes, it crashes regularly, but reloading the offending program ususally fixes it. I never have to reformat and reinstall. All the config files are plain ascii text and easy to modify. I never have to mess with the registry.
A set of rather intensive courses in Mandarin mainly. I know a fellow who taught some courses in China in a very esoteric branch of control theory. He was amazed at the size of the audiences for his lectures- and that they all seemed to be nodding appropriately, indicating at least some level of understanding.
Yes, it is. How did you come in contact with it? You see the prosciutto-like hams all over the place (in high end venues). Like the Parma version, it is not cheap. I didn't try it. Their coffee is pretty decent, they make an okay sherry-like red wine and apparently their tobacco is good. Interesting pictographs among the Naxi minority. BTW, the level of abundance in Chinese supermarkets has increased tremendously. Imagine having 5 different kinds of cooked chicken feet to choose from! Gasoline is roughly at US prices at the pump (about Y4.5 per litre).
Best regards, Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.