Thermocouple errors 1

[YesSaidJake]

Hi,

I've set up a temperature measurement rig for a project measuring the temperature of various indoor and outdoor temperatures, for a building physics investigation, and I'm getting some strange errors. The system works fine in the lab, and seemingly fine at one site, but at another site it's giving errors.

Most days, between about 12:30 and 13:30, the readings go silly for half an hour to an hour. The temperatures are around 15 to 20 C, but for the stated time, one thermocouple is reading around 30, while another is reading around -5 to -10 C. There's simply no way that the building in question is really that temperature. Could there be some sort of interference taking place? At the moment, I don't have a spare data logger to test, but it does work fine in other locations. It's a Campbell Scientific CR1000, with K type thermocouples. Other thermocouples connected to the same data logger work (seemingly) fine, even at that site, and the whole system seems to work at other sites - both before and after the strange measurements.

Suggestions very welcome!

 

[YesSaidJake]

On further investigation, it seems to be some sort of interference between the different thermocouple channels, but I'm not sure why, yet.

 

[VEBill]

There's some tidbits listed here:

http://www.omega.com/temperature/Z/PracticalGuidelinesforTemperatureMeasurement.html

But it sounds like something is 'broken'. Maybe the recorder is getting too hot from the noonday sun?

 

[analogkid2digitalman]

Is direct sunlight falling on the CR1000 unit during that timeframe?
On the thermocouple that swings negative, could the polarity of the leads be reversed?
Any specific equipment in operation just during the time in question?
Any potential ground loops between the T/C"s? Are they attached to anything metal or just sensing ambient air?

-AK2DM

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"It's the questions that drive us"
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 

[danw2]

My list mirrors Akid2Dman's list.

I'd be checking that negative reading T/C for being wired backwards; red is negative (L), yellow is positive (H).

Whether you're sensing air or a solar exposed surface makes a huge difference. I once stapled a thermocouple to the building's southern exposure wood siding on a cold spring day, air temp 1-2 Deg C. The surface temperature of the wood siding got up to 19 Deg C at about noon. Air temp stayed at 1-2 Deg C.

Somewhere in the setup there's a setting for line frequency, so the sampling will synch and not alias-sample noise. But that kind of noise is jittter or periodic, not way off, like yours.

The spec sheet is vague about operating limits. -25 to 50 C (?). If the datalogger itself bakes in the sun, it could overheat and go screwy.

I assume that the T/Cs are in sheath. Hard to imagine picking up common mode with what you're doing with sheathed T/Cs.

 

[itsmoked]

Don't forget every T/C reader HAS to have an ice-point correction. It is normally done by measuring the temp of the actual terminals where the field wires connect and then running an inverse transform look-up to come up with the local connection temp to subtract from the field measurement. That isothermal structure around the field terminals is often the sketchiest aspect of the entire system. If the sensor required to measure that is not happy or direct heating from sunlight or a draft or several other things happen then the resulting measured field temp will be off by many 10's of degrees. Especially onerous is if the two terminals are not exactly the same temp.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[YesSaidJake]

Thanks for all the suggestions - I've narrowed it down, sort of. I'm still no wiser as to the actual problem, but now I've managed to replicate the problem in the lab as well.

I have four thermocouples, going into channels 2, 3, 5 and 6 on the datalogger. The thermocouples on channels 2 and 5 seem to be giving correct readings but channels 3 and 6 are not. Somehow, channel 3 is also responding to the signal from channel 2 (as well as its own signal), and channel 6 is responding to the signal on channel 5. If I increase the temperature of T2, this shows up as a temperature drop on T3 (even if T3 is in reality held at a constant temperature). The signal from T3 is still being used, and if T3 alone is heated, it responds (and T2 stays constant). Heating both causes T2 to rise correctly, but T3 to only rise by the amount it's heated above T2. T5 and and T6 behave in the same way.

This leads me to think it could be something to do with the reference temperature as Keith suggests, except I'm using an internal temperature sensor in the datalogger as the reference, and that's the same for all the measurements.

 

[itsmoked]

However, the single temp sensor may be doing a bad job of indicating the actual temps of all the various field terminals if they aren't isothermal. The heavy metal plate may have fallen off.

With this more info you've provided there are other possibilities.

Your logger uses analog switches to swap the various T/C inputs into the the single instrumentation amp that's actually doing the voltage measurements. Those analog switches are single chips that can fail and cause what you're seeing. They also have notoriously low common mode voltage abilities. Violating the common mode spec causes all sorts of strange behavior like tying the signals to other channels or ground or VCC.

If with 3 inch T/Cs plugged into all the channels you see this bad behavior then the analog switches are probably screwed up. If you don't see a problem until the T/Cs are really long or near high alternating currents, or in radio frequency noise, or the T/Cs are metallicaly touching something that adds a high common mode voltage violating the common mode spec.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[YesSaidJake]

Given that the problem is consistently happening on the two channels and not appearing at all on the other two, I'd guess it's probably more likely to be the switches rather than the plate? If it's due to an error with the reference temp would you expect the errors to change if the logger is at different temps/in sunlight? Whereas if the switches are playing up I can imagine that might cause a more consistent error like this. I'll have a play around with different types of thermocouples and see what I get.

 

[IRstuff]

Do you not have another meter?

Have you tried swapping TCs around?

TTFN
I can do absolutely anything. I'm an expert!
faq731-376 forum1529

 

[danw2]

> ice-point correction. It is normally done by measuring the temp of the actual terminals where the field wires connect and then running an inverse transform look-up to @ come up with the local connection temp to subtract from the field measurement.

The cold junction (CJ) measurement adds to (not subtracts from) the thermocouple measurement.

Case in point, when both ends of a T/C are at ambient room temperature, there is no thermal gradient, therefore there is no T/C EMF - zero mV output. Without CJ, the reported temperature would be freezing point of water. The CJ terminal block measurement is added to the T/C value, in this case, the 0mV freezing point of water temperature, in order to calculate the device's 'reported' temperature.

I suspect that Campbell Scientific does a reasonable job of cold junction measurement because they claim leadership in that field, according to this blurb from the manual:

While the CJ might be fairly decent, the common mode limit is 16V. Not terribly robust.

 

[YesSaidJake]

Solved At least, I think so, subject to more extensive testing.

I'm feeling a little bit sheepish, though I don't think it was entirely unreasonable of me to expect that the 4 things labelled as thermocouple outputs would be, well, thermocouple outputs (moan); but it turns out that channels 3 and 6 report the signal from two thermocouples in each case. Here's the wiring diagram of how the inputs are wired:
[pre]
' 2H ------ green (+) ----------------------------------------
' | |
' Thermocouple 1 |
' | |
' 2L ------ white (-) ------------------------ |
' | |
' AG ------ TC 11 shield Thermocouple 2 |
' | |
' 3H ------ green (+) ------------------------ |
' |
' |
' 3L ---------------------------------------------------------
'
' AG ------ TC 12 shield
[/pre]

So it's possible to recover the values for all the thermocouples from the 4 outputs, but the reported outputs from the datalogger don't actually correspond to the signal on each channel as I expected. (Am I alone in thinking this is a strange way to wire it? No doubt there's a good reason for it, but it would be nice to be told about this sort of thing in the manual!)

Thanks everyone for your help - I know much more about the hardware I'm using than I did a couple of days ago, even if it turned out to be a (sort of) software issue!