How Well Do You Really Know Thermocouples

[Archie]

Brian123 and g.mccormick answered it in a simpler and more concise way than I could have.

When asked exactly how and why thermocouples work this is how I explain the theory....

Thermocouples work on a principle known as the Seebeck effect. A physicist named Seebeck first discovered when a loop was made from wires of dissimilar metals and one junction heated, an electric current would flow (technically he said it was a magnetic field). The key point to note is that only one junction was heated, therefore a temperature differential existed between the two junctions.

In my question I specified a J type thermocouple, but there is nothing special about that type other than it being a common type, so I will continue to use that type to explain. Using a piece of J type thermocouple wire (let's say 500cm long), strip both ends then twist the two wires together at each end essentially making 2 thermocouple junctions. If one end was kept at room temperature of 72F and the other end heated to 200F, a current flow would be induced in the loop, therefore a possible measureable voltage. Now if both ends were heated to 200F, the current flow would stop, therefore 0 volts would be generated. Now heat both ends to 300F and once again there will be no current flow nor voltage. No matter how hot you make the junctions, if both junctions are equal, there will be no current flow and no voltage.

This may bring to question, if the thermocouple voltage can be 0 volts at 200F as well as 0 volts at 300F, then how is it possible to measure temperature with a thermocouple? As several have mentioned, this is where a cold junction compensation element comes into play. If you were to look up the specifications for J type thermocouple, you will find it generates -8.095mV at -346F and 69.553mV at 2193F. The mV values found in standard thermocouple tables are typically based on one junction being at 32F(0C). If the "cold" end were to warm up while maintaining the same temperature at the measuring end, the voltage will decrease and if it were to cool down, the voltage would increase, the generated voltage is based on the temperature difference between the 2 junctions. The purpose of the CJC element is measure the temperature at the terminals connecting the thermocouple to the device and use that value to compensate for the temperature not being at 32F (0C), which is what the standard tables use for their voltages specified.

With that being said, if the device measures 0 volts on the input, it must assume the measuring junction of the thermocouple is at the same temperature as the end connected to the device, which would be the temperature measured by the CJC element. I did state in the question the device is setup for thermocouple which was intended to mean the device would have a CJC for accurate measurement.

If a piece of copper wire were used to jumper the input terminals, it would force the input to 0 volts, therefore a reading of the CJC temperature would be the expected display temperature. What if the terminals are not copper, wouldn't that create 2 thermocouple junctions that could generate a voltage? The answer is yes. However keeping in mind there must be a differential temperature to generate a voltage and the input terminals are typically so close together that the temperature difference is negligible.

What use does this tidbit of information really have? Consider a scenario in which you have a PLC with a TC card and it is giving suspected bad readings. A very quick & dirty (Q&D) test is to jumper the input with a piece of common wire and see if it is giving a temperature reading close to the temperature of the TC card. This can give an indication if the card is working properly.

An interesting piece of trivia.... The same effect that makes thermocouples work can also be used to generate electricity. Several space probes use what is essentially a bank of thermocouples and a radioactive material (for heat) to generate electrical power for the probe's electronics. The mars rover, Curiosity, uses thermocouples to generate 110 watts of electrical power.

 

[wes.s]

Before I read the thread I'm thinking 32°F since a J T/C puts out 0v at 32F.

 

[danw]

Thermocouple is ALWAYS two measurements

The lesson is that a thermocouple circuit has to make TWO measurements to come up with a final indicated temperature.

One measurement is the thermocouple's mV output which represents a non-CJ compensated temperature.

The other measurement is the cold junction measurement, which should be the temperature of the connection terminals.

Those two measurements are added together to get the indicated (computed) temperature value.

A quick check of any direct connect thermocouple is to short the input with a piece of copper wire (exactly like Archie proposed) and the reading will be the temperature of the terminal connection block. That could be close to ambient temperature, or it might be several degrees off, depending on how much heat from the electronics is conducted through to the terminal connection block. I've seen terminal connection blocks 10-15 degrees warmer than the ambient air.

Why is the themocouple temperature reading consistently low?

If the cold junction circuit is dysfunctional or not enabled, then an error which is the difference between the freezing point of water (32°F, 0°C) and the terminal connnection block will be evident. Whenever there are posts reporting a consistent low temperature error of about 40° to 70°F or 20° to 40°C, the immediate suspicion is that cold juction is not working or is disabled.

Question: Should the CJ box be checked?
Checking the CJ box enables CJ. So, yes, check the CJ box.

There are very, very few instances where the CJ box should not be checked. In fact, other than one laboratory application where accuracy was so critical that they used an actual ice bath, I can't think of one off-hand.

>since a J T/C puts out 0v at 32F
ANY thermocouple puts out (generates/creates) 0.0mV when it is isothermal, that is, same temperature at both ends. The condition of an isothermal T/C, same temperature at both ends, can occur at any temperature.

Archie described a thermocouple that was 200° or 300° at both ends that does not generate any mV output, because there is no temperature gradient to create of voltage.

That is can actually happen for room temperature measurements. If the T/C terminal block is not heated by conduction from the electronics, it's temperature will be room temperature. Both ends of the thermocouple are at room temperature. The thermocouple is isothermal (same temperature at both ends, no temperature gradient) so it generates 0.0mV, which is represents "no temperature value" to the electronics. That no temperature value is added to the CJ value, which is room temperature. The indicated, reported temperature is room temperature.

If the terminal connection block is warmer (or colder) than ambient then its measured temperature is the value used and added to the temperature from the thermocouple.

The reason the thermocouple tables show 0.0mV at 0/0°C is because the measurment assumes cold junction compensation, such as the temperature of a 2nd thermocouple in an ice bath is connected in series to the working thermocouple. The tables assume ice point compensation or cold junction compensation.

That's why a millivolt measurement of the cold end of a thermocouple will be in error (the measured value will low for working temps above freezing) from the correct value in the T?C table by the difference between the temperature of the cold end and the freezing point of water. The ice point or cold junction compensation temp measurement is missing.

>since a J T/C puts out 0v at 32F
In real life, where the terminal connection block is probably room temperature or above, a thermocouple measuring the freezing point of water is generating a negative voltage which is interpreted as a negative temperature, which when added to the CJ's positive temperature of the terminal connection block results in a value of the freezing point of water, 32°F or 0°C

 

[g.mccormick]

archie and danw nice write ups.

The importance of the cjc is why I mentioned being careful of the ambient elements around the input card. The input card needs to see consistent (or very slowly changing) ambient temperatures so that there is not a difference between the connector and the cjc inside the card. Even mounting orientation can effect this. We have equipement that had the rack mounted horizontally have a less accurate temperature measurement due to the temperature gradient of the card itself.

 

[danw]

Archie

When I teach thermocouples, I have the guys bring their Fluke voltmeter to class.

Everyone gets a 2 foot length of K wire that is stripped on the cold end, stripped and twisted on the hot end. They can't play with it until we do the 'exercise'. It sits on the desk while we do other things so that it becomes isothermal at room temperature.

I have one of the big circular room thermometers on the wall to show room temperature. I bring up the word ambient and what it means because most of them think I'm mispronouncing a drug name.

I have them turn on their meters to warm up for 5 minutes.

Then I have them look up the mV output for a K thermocouple from a printed table (which row?, which column?)

Exercise: Then they're to be careful to not touch the thermocouple, and to hook up the voltmeter to the thermocouple, which is where 'thermocouple red is negative in the US' comes up.

When they get readings of 0mV, I ask, was Mr. Seebeck wrong? The table says x.xx mV for this temperature. How could that not be the case?

It demonstrates
- an isothermal thermal couple generates no voltage because there is no temperature gradient (0mV)
- one of the two required measurements is missing because the mV reading does not match the table (nor will it ever, even if there were a temperature gradient and the thermocouple was producing an EMF)
- the indicated value of 0.0mV as 32°F or 0°C shows a low error - typical of a disabled CJ

 

[Archie]

danw

I have done a similar demonstration with a volt meter and piece of thermocouple wire to show the generated voltage. To further try to emphasize the concept of the temperature differential I used a cup of hot water to show when the twisted end was put in the water, the mV signal would increase. Then put only the meter probe ends (the cold junction) in the hot water and the mV signal would go negative. And finally put all junctions in the water to show the signal would drop back to 0mV when all junctions are at an equal temperature no matter what that temperature was.

 

[Tharon]

So now I have a question about the way we used to do some Thermocouple checks. (Different now, using TC Calibration devices, but used to do it another way).

We had a high limit controller, and a controller. Dual TC with one TC going back to each controller. We would measure the voltage at the controller terminals with a Fluke multimeter and make sure the temperature readout on the device matched the mV lookup charts for the TC Type. Adjust bias as needed.

I'm thinking now this would work only as long as the TC remained connected to the controller which has its CJC. But if I disconnected the TC and had it running straight to the Fluke DMM, I would no longer get mV readings corresponding to the actual temp at the hot junction. Instead a differential reading from Hot side to my DMM?

 

[isiah_s]

FYI, the positive leg of a type T and type U thermocouple is pure copper.

If the materials of the instrument made a difference, how do you account for instruments that can be software-configured to read various different types of thermocouples with no impact to accuracy?

Because different materials give off different voltages. By software you just pick the different ranges, but all of them give some voltage range.

 

[danw]

danw

I have done a similar demonstration with a volt meter and piece of thermocouple wire to show the generated voltage. To further try to emphasize the concept of the temperature differential I used a cup of hot water to show when the twisted end was put in the water, the mV signal would increase. Then put only the meter probe ends (the cold junction) in the hot water and the mV signal would go negative. And finally put all junctions in the water to show the signal would drop back to 0mV when all junctions are at an equal temperature no matter what that temperature was.

Yup. Part II
And connect the voltmeter backwards (reverse polarity) so that heating the hot end drives drives the mV downwards.

 

[Archie]

We had a high limit controller, and a controller. Dual TC with one TC going back to each controller. We would measure the voltage at the controller terminals with a Fluke multimeter and make sure the temperature readout on the device matched the mV lookup charts for the TC Type. Adjust bias as needed.

I'm thinking now this would work only as long as the TC remained connected to the controller which has its CJC. But if I disconnected the TC and had it running straight to the Fluke DMM, I would no longer get mV readings corresponding to the actual temp at the hot junction. Instead a differential reading from Hot side to my DMM?

Measuring the mV signal with a volt meter will only give you part of the picture. You would not be able to directly look up the expected value in the chart. The signal would have to be adjusted based on the temperature of the terminals in which the thermocouple is connected to. This applies the same whether connected directly to the temperature controller or connected directly to your meter.

For example, if the temperature at the thermocouple measuring end is 100F and you connect the other end to a meter which is also at 100F, the voltage reading would be 0 volts. Looking that up in a chart would tell you the measured temperature is 32F, therefore a seemingly 68 degree error.

So if the thermocouple is left connected to the controller, then the temperature of those terminals are the cold reference that must be accounted for. If you remove the connections and connect directly to the DMM probes, those now become the cold reference. That value can easily be influenced by simply using your fingers to hold the wires tighter on the probes, therefore heating your junctions.

 

[danw]

So now I have a question about the way we used to do some Thermocouple checks. (Different now, using TC Calibration devices, but used to do it another way).

We had a high limit controller, and a controller. Dual TC with one TC going back to each controller. We would measure the voltage at the controller terminals with a Fluke multimeter and make sure the temperature readout on the device matched the mV lookup charts for the TC Type. Adjust bias as needed.

I'm thinking now this would work only as long as the TC remained connected to the controller which has its CJC. But if I disconnected the TC and had it running straight to the Fluke DMM, I would no longer get mV readings corresponding to the actual temp at the hot junction. Instead a differential reading from Hot side to my DMM?

Your measurement/bias technique drove the indicated/reported temperature readings low by whatever the CJ correction value was.

The only way your technique would work correctly is if the cold end terminal block (where you were measuring) were held at the freezing point of water. That's what the thermocouple 'tables' assume, CJ is at the freezing point of water. Since the terminal block was not at the freezing point of water, your mV measurement was low by the difference between the freezing point of water and the terminal block temperature.

Your applied negative bias counteracted what the CJ was attempting to report as a correct temperature value.

 

[Tharon]

Seems I have a few things to think about with these TC calibrations.

Thanks.

 

[proof]

I disagree with your statement that heating both junctions to 200 or three hundred gives no volts. They do give volts, but they cancel each other out, so there is no current flow. You had the right conclusion, but for the wrong reason. Measuring the voltage across a thermocouple with a DVM is also a misleading test. The DVM lead to on wire on the thermocouple creates a junction, and the other DVM lead to the other side also creates a thermocouple. So now you are not measuring the thermocouple voltages, but a new voltage from the three junctions. This voltage would be created from the difference between temperatures in the three junctions. Of course this also assumes the DVM circuitry does not have any effect on the circuit, which is probably not a correct assumption in practice.

Putting a short jumper of copper across a thermocouple creates two junctions, creating voltages with one positive, and one negative, with the sum the same as the voltage that is created by connecting the thermocouple wires together directly.

Thermocouples are very simple and easy to understand. Figuring out how to measure the voltage is the hard part.

 

[g.mccormick]

Yup. Part II
And connect the voltmeter backwards (reverse polarity) so that heating the hot end drives drives the mV downwards.

When everything is at room temperature its hard to find a backwards wired tc. If everything is same temp, the value will be room temp (as we've discussed since no differential between joint and input card).

Its always fun when your temps read "right at room temp" then when you heat up whatever you're measuring the temps go super cold!

 

[nykommling]

How does two TC in parallel act?