I am not that experienced using thermocouples (E and K type) and may be overlooking a glaringly obvious fact.

When attaching thermocouples to a National Instruments signal conditioner, I am having difficulty achieving stable temperature readings using a LabView virtual instrument written by a former employee. The leads appear to be securely attached, but when I gently touch the thermocouple lead wire, the temperature readings fluctuate wildly and even go to a different nominal value 20 - 30 F apart. The hot junction is not in contact with anything but the same air.

In my acquaintance with thermocouples, this is the first time I have experienced this kind of problem.

What am I not considering?

You might have some defective thermocouple wires with internal shorts, or poor (oxidized) junctions/connections.

Might also be that you are acting as a shield/ground when you touch the TC wires. How is the board isolated relative to ground? Are you using shielded TC wire or running it thru EMT, is the shield/EMT grounded to the same reference as the A/D board sees, and what environment does the TC wire travel through in terms of electrical/magnetic fields?

Finally, some longer-term drift errors occur due to local heating/cooling of the board's temperature sensor it uses for cold junction compensation. Are there any rapid temperature swings that the board might be experiencing, i.e. a sudden cold draft when the door to its room is opened, or the AC kicking on/off?

Make sure your NI box has a good ground. I had to run a ground wire directly from the box to a known ground to get the readings to stabilize. In my situation, we had some heavy EMF running around due to heavy duty power inverters and I suspect those were causing some weird stuff to go on. With a direct ground, all was fine.

rp

Definitely sounds like a grounding problem. Non-isolated (bare) thermocouples with line powered instruments are very susceptible to grounding issues. Grounding the instrument may make the problem worse. Jacketed thermocouples solve many of these problems but they are more expensive and slower in thermal response. I would never use bare thermocouples in an industrial environment without instrumentation specifically designed to provide the ground isolation. This makes the instrument more expensive.

Problems like this are common. We always use ungrounded junction thermocouples, because ground loops are a common problem when the junctions themselves are grounded.

Quite possibly it is a result of your body acting as an antenna picking up power-frequency (50 Hz or 60 Hz) noise which is then entering the very sensitive T/C preamplifier when you touch the thermocouple wires. Look for the earthing (grounding) requirements for the NI amplifier. There is almost certainly a specific way that NI intended you to earth the instrument, and there countless ways which they didn't intend you to earth it.

Ditto on the use of ungrounded TCs. I have also used shielded wires in electrically noisy environments, with the shield only grounded to the amp.

ISZ

I don't mean to hack this thread, but I have a question related to just this. You guys are talking about the grounding the thermo couple.
Say in my testing I am measuring temperatures at 3 TC points of an electrical equipment. I generally run the thermo couple from a point where I want to measure to the temperature reader.

Which grounding are you guys taling about? Do you mean that the equipment must gave a proper grounding wire connection?

Talk to an electrical engineer and they'll teach you about ground loops. I can tell you only the little I know.

If you have a thermocouple which has its junction connected electrically to its metallic sheath, that thermocouple junction is considered to be "grounded". If the junction is embedded in MgO insulation such that it is in thermal contact but not in electrical contact with the sheath, or the junction is exposed beyond the end of the sheath, that junction is considered "ungrounded". Whenever we have used grounded junction thermocouples in the interest of having the fastest responding and most accurate junction possible, we have very often experienced problems with getting good data out of them due to ground loops or other electrical interference.

If you use a "grounded" junction T/C, what you have to worry about is a difference in potential between the ground that the sheath is connected to electrically (i.e. the equipment) and the ground that the thermocouple amplifier/reading device is connected to (i.e. the instrumentation ground). If there's an isolating transmitter in there, the problem is reduced of course, but ONLY if it has electrical isolation as part of its design. MANY thermocouple input cards for I/O systems do not have proper isolation as part of their design, and hence they are subject to ground loop problems. Of course when the tech comes around and plugs the T/C into his battery-powered (i.e. isolated) thermocouple reader/calibrator, everything looks fine.

EMI/RFI is a different matter. Unshielded wires, particularly unshielded wires connected to a very high input impedance amplifier (such as, say, unshielded thermocouple extension wires connected to a thermocouple reading device) are particularly good antennas to pick up this kind of noise, which can be generated in great quantities by devices such as motor speed controllers etc. Twisting the wires helps to reduce the noise, but a shield helps more still. That shield has to be connected to ground somewhere, but not such that it becomes a ground loop path itself. But it is still necessary to keep the T/C junction isolated electrically from the sheath ground.

Does this make things clearer?

To expand on MM a little, one has to realize that TC's have very small outputs. If you look at a K-type, there is only ~4.1mV difference between 0°C and 100°C. Which is roughtly 0.0041mV or 4.1 micro-volts per degree C. (FYI - unlike other types, this one is fairly linear over short ranges)

Now if we look at a ground loop it isn't hard to see that any voltage drop in a ground path, with a grounded TC, will result in an erroneous output. (And realize that this voltage drop will change with temp, humidity, stray voltage from nearby electrical equipment, dust, etc.) Next comes a person with some amount of static charge, and viola, output changes again.

The easy answer is to use ungrounded whenever possible. Generall the response time penanty isn't an issue, but if you have to go ungrounded you must take special precautions to prevent this.

ISZ