I understand the basics of thermocouples. If you use the wrong extension cable, you create a thermocouple at the end that connects with the thermocouple. However I have been searching high and low to find some information of how to approximate this error (I understand it is probably not linear as well).
So if you have a Type K thermocouple connected to a Type T compensating cable you have three junctions.
1) Ni Cr vs Ni Al (The "main" junction)
2) Ni Cr vs Cu (the +ve side junction)
3) Ni Al vs Cu Ni (the -ve side junction)
What I haven't been able to find is any information about the seebeck (moderator's note: setback?) voltage produced over junctions 2 and 3. My understanding is that if I could get that information I could provide a rough estimate of the error created by having the wrong type of compensating cable.
Why not just fix the problem? Either use the right thermocouple or change the cable. Thermocouples are problematic enough without throwing this kind of unpredictable non-linear error at them. Would you really rather spend hundreds of hours listening to complaints about how your temperature measurement never seems to work properly - or go home and get some sleep?
I'm going to suggest the original poster is misreading the information printed on the cable jacket. Every T/C extension cable I have ever worked on used the 'Tx' identifier on the jacket--regardless of whether it was Type K or Type T or Type J T/C extension cable.
But, I agree with Rob. Sometimes the fix, while it may seem more costly at the time, is easier than living with the problem and trying to come up with all kinds of work-arounds over a longer period of time, which will likely result in the proper fix after an interminable amount of grumbling. While it may be possible to come up with some kind of software "fix", what about the person that comes in behind you and doesn't know the history and matches the T/C with the type of T/C extension cable, and then has problem?
If the cable isn't the proper kind of cable, as Rob suggested how about changing the T/C to match the cable? That would likely be the least expensive solution (not understanding the application). That might be a less-costly solution than replacing the cable (presuming the system/device the T/C is being connected to can work with the type of T/C and T/C extension cable. There's just a lot we don't know about the whole circuit and application.
But, do have a look at the T/C extension cable again to be sure it isn't the right type of cable. That 'Tx' designation causes a lot of confusion. One lead of a Type K T/C is non-magnetic; I don't recall if that true of Type T or not.
With a mismatch you will never get much more than indicator accuracy.
There are simply too many variables unless you know and account for the temperature at each junction.
You won't have any luck with a single offset or span calculation. You
may calibrate for one set of conditions. but it's almost sure to change with seasons or any change in ambient temperature. The worst part is that you will never know if you should believe it Thermocouples are economical and accurate only when done right. And it's really easy to mess them up. even when you know what you are doing.
Exactly. As another poster suggested, either change the extension wire or the thermocouple so they match each other.
If it's not feasible to change either of those, and if the extension wire is shielded, you might be able to place a 4-20ma thermocouple transmitter at the junction point and change your DCS/PLC/instrument input to 4-20ma, and just run the 4-20 signal in the extension wire. I'd imagine it's conductive enough to carry 20ma in whatever gauge they've installed without excessive voltage drop.
> Why not just fix the problem? Either use the right thermocouple or change the cable. Thermocouples are problematic
> enough without throwing this kind of unpredictable non-linear error at them. Would you really rather spend hundreds
> of hours listening to complaints about how your temperature measurement never seems to work properly - or go home and
> get some sleep?
I understand and agree but it would be easier to make a business case if I had some idea of the magnitude of error involved.
The best way to make an estimate of the error is to set up a test system - stick a Type K in a bath at the appropriate temperature, run Type K compensating cable for a bit, connect it to Type T compensating cable, and then connect to your receiving equipment of choice. Now heat the intermediate junction to something like the ambient temperature of the actual thermocouple head in the field.
I think you'll be surprised at the results.
I cannot believe we are discussing the relative merits of doing something in various wrong ways as opposed to just fixing it correctly.
> I cannot believe we are discussing the relative merits of doing something in
> various wrong ways as opposed to just fixing it correctly.
That's right, but it is still a good reminder or case study for similar situations since this concept is not fully understood even by many experienced technicians. I've had to explain it to my guys many times in various similar circumstances, for example inside a control cabinet where the T/C extension cable from the field was connected to a terminal strip at the bottom of the cabinet and routed to the PLC modules at the top of the cabinet with non-compensating wire, the temperature gradient (within the same cabinet) between the PLC and the terminal strip created a very sizeable error, more than 10degC.
For once I am in complete agreement with Bob, the one thing that you can say for sure is that it will be wrong under nearly all circumstances. That doesn't inspire confidence. And the workarounds are more time and trouble than fixing it. It wouldn't even be dependable as an indicator because it will be inconsistently wrong depending on season, activity, etc. anything that changes the temperature differential from end to end. It's not something you'd do to someone you like. Much worse than simple inaccuracy.
I can quite believe it - especially if the guy responsible for the original design is now responsible for writing the cheques to get things fixed. I was once in a situation where I was responsible for a large furnace management PLC with some issues relating to performance under partial shut-down - I was expected to make it work properly but any modifications had to be approved by the guy who had designed the original program and there was no way he'd admit there was an issue.
>I cannot believe we are discussing the relative merits of
>doing something in various wrong ways as opposed to just
>fixing it correctly.
I have a situation where the wrong TC was installed in a heater bundle. The cabling is of a different type as the TC. We have a contractor that is arriving to assist with a startup of a very critical station and there is no time to 1. Replace the TC in the bundle and 2. Replace the cable back to the house. I can appreciate that someone would want to know the exact symptoms of using a mismatched setup.
In a perfect world everything is available when we want it with as much time as necessary to do it the "Right" way. I would really like to live there...can someone give me an address?
> I understand and agree but it would be easier to make a business case if I had some idea of the magnitude of error involved.
Hi Ben, Yes I thought this might be the case. (I've been there like you :) ) Its false economy. You're trying to save a few hundred bucks and you've already wasted much more than that in your salary time just trying to figure out what to do. Until you fix this, you're going to keep wasting more time and more money because there is only one solution. Here's the justification you need for your company accountant (who incidentally cares nothing about technical details) ....
It doesn't work. It needs replaced.
the error is minimum,
not one you want to make. but for most applications is is just the confusion it creates, rather than the error. Of course the confusion will burden the maintenance department for years, until the wiring or the sensor is replaced.
Here's my rationale on the topic:
I don't think the error is a linearity question related to thermocouple composition. The error has to do with temperature gradient.
If there's no temperature gradient across the copper wire, and the connection points from T/C wire to copper extension wire are isothermal (both legs/sides at the same temperature) then there's no error.
Copper lead wire would be the equivalent of really long, extended analog input terminal screws. Terminal screws are not made of thermocouple alloys, they are the same fasteners used for other signal connections, nickel plated brass or whatever. This is the Law of intermediate metals/intermediate junctions. If it works for terminal screws on transmitters, AI cards for PLCs, RTU's DCS's, recorders, indicators and controllers, it works for longer than normal 'screws' made from copper wire, too.
The AI's Cold Junction compensation measurement done at the AI terminal connections assumes isothermal conditions; whether that is reality or not, that's what's done. And it also assumes that the measurement is in close proximity to the sensed point.
An error from copper extension wire would arise from whatever the temperature gradient is across the copper wire because the EMF generated in the circuit is the gradient across the thermocouple and thermocouple wire, which terminates short of the terminals where CJ is measured. The gradient across the copper wire is not 'visible' to the AI, so it becomes the major error factor.
There could be some error due to voltage drop across the copper, but copper's resistivity is much lower than that of thermocouple wire. The main error component is the temp gradient across the copper, which could be positive or negative, depending on the temperature of the T/C-copper junction point relative to the cold end.
Take an example:
Under correctly wired conditions using all thermocouple extension wire:
if the hot end is 200°C and the AI terminal block is 30°C, the EMF 'seen' at the AI for whatever type T/C = 170°C (200°-30°=170°). 30°C cold junction compensation is added to the 'measured' EMF temperature of 170°C to provide an indicated temperature of 200°C.
With copper extension wire:
if the hot end is 200°C, the T/C-copper junction is at 25°C, and the AI terminal block is 30°C, the EMF 'seen' at the AI is 200° minus 25° or 175°. With the 30° cold junction added to the 175° EMF equivalent, the indicated temperature shows up as 205°C. The 5°C copper wire temp gradient shows up as a 5°C error.
Yes, and you can even measure the temperature and calculate the
individual errors. My reasoning is that if there were no temperature gradient and it wasn't variable, you probably wouldn't be interested in putting a thermocouple there. If you terminate both ends in a temperature controlled space, you're home free, but you usually don't.
Oops, I stand corrected. In my earlier post, I was thinking ordinary copper wire was the extension wire, not Type T extension wire.
Type T wires are copper and constantan, not plain copper. The Law of Intermediate metals would not apply to a junction to dissimilar metals.
I agree that Rob's statement about unknown non-linearity applies to a case of the mixed thermocouple/extension wire.
an alternative (might be of help).
why bother about individual junctions, you have a thermocouple system with 3 junctions. treat the system as a single unit, vary the temperature for your working range and note down the voltage your system gives. (its not clear which thermocouple type indicator/controller/input card is being used so lets keep it up to voltage measurement only). you already know the voltage for a k type thermocouple. the two tables can be compared to determine error (for this system).
more or less the same thing that i said in my last post, has already been said by Bruce Durdle in this thread.
sorry, hadn't read.
As far as I know, it is not feasible to treat this as a single system as you suggest, since you would have to consider TWO combinations of temperatures, one of them the actual process temperature, and another the temperature of the terminal block where the K T/C connects to the T extension. Unless you neglect the K-to-T junction, which is probably not a good idea since normally the temperature of the T/C head or terminal block is not the same as the actual process junction or the instrument input terminals.
Yep - there are two extremes.
If the terminals in the thermocouple head are at more or less the same temperature as those at the tip - for instance if the head is covered in insulation - then you will effectively have a Type T thermocouple rather than a Type K. If they are very close to ambient temperature, then there will be little error and the Type T extension wires will have little effect.
Between these two, there is a lot that can happen. Like I said, the easiest way to get an estimate of the error is to set up a test and try it. But eg a thermocouple in boiling water with copper wires from the head (also exposed to steam from the water bath) may read as low as 45 degC.
Cost of a replacement Type T T/C - about $300? Cost of your time at $xx /hour? - probably about $5xx by now!
Exactly. There's no way that's going to work satisfactorily. You'd have to characterize both of your two oddball junctions vs temperature, AND know what the temperature is at the junction point (which is going to require a new measurement point) so you can back out the effects. PLUS, even if you did all that, you'd still be tripling your error because in effect you're making three measurements and subtracting out two.
It's like trying to do differential pressure measurement with two single-ended transmitters and subtracting. You increase your error to the point the measurement may no longer even be useful, and that's with just TWO measurements rather than three AND not having to invent the sensor technology (new T/C types) as in this case.
Easier to just do it right.
ok forum, one more chance :)
let's make this whole thing a k type thermocouple.
the actual thermocouple is k type, attach some k type wires and run them a little so they get at atmospheric temperature. now attach the t type wires and run them as long as needed. now at the end, again attach some k type wires at the appropriate ends(that too, at atmospheric temperature). and that goes to the indicator.
so the 4 outer junctions are all at atmospheric temperature and the sensing junction senses the temperature to be measured. this is just a k type thermocouple. theoretically, there is no error with this system (practically, there might, because of the variations of the atmospheric temperature at the 4 outer junctions).
Aye, there's the rub. The two ends are unlikely to be at the same temperature in the real world. Quite often one end is in the control room and the other end is in the head of a TC well, perhaps stuck into a furnace, or a freezer. From the TC type, probably someplace hot.
Thanks CWW for the reply,
>The two ends are unlikely to be at the same temperature in the real world.
we can always take a few meters of k type wire so that it goes up to a point, it comes to open air and we make junction there. similarly for the k type wire at the indicator side.
if a strictly errorless condition is required, these 2 points can be made artificially isothermal.
regarding the point of making the 3 junctions at the same temperature ( we discussed it @ 2 February 2013 - 6:59 pm, 2 February 2013 - 12:27 pm, just to avoid confusion), if we make it a t type thermocouple then we can expect it to work satisfactorily as a t type. of course, a t type indicator will be needed. so that might also be a possible approach. but here we are not dependent on the atmospheric temperature to be same at two points.
> Just how much extension wire that needs to be replaced is involved here?
considering this question as- the amount of the t type wire that has to be replaced by k type wire,
just enough to bring the junction points to open air. the precise amount depends on the precise locations of the thermocouple head and the indicator.
@ Roger, you are correct but i suppose you have misunderstood me. by treating as a single unit and varying the temperature, i meant all the junctions(and that equals the system) always getting equal temperature and that common temperature being varied for the working range.
one possible way to do it practically -take up the bare wire junction (just the junction and small lengths of the two wires) of the k type and attach the wires of the t type(as said in the problem) to the appropriate ends. put this whole thing in the temperature to be measured. and so all 3 junctions are at the same temperature.
well, i haven't done this thing till now, but i don't see any reason why it won't work.
oh, my mistake.
if all the junctions get to the same temperature, it would just reduce to a t type thermocouple.
what can be done is to keep the outer 2 junctions at some fixed temperature and just vary the temperature of the inner junction only, to see the response.
Hi, reading through the conversation over - it seems my current problem is discussed over, but I would appreciate if anyone can comment on if I have been able to pinpoint the cause of my problem.
I have a K type thermocouple measuring the throw temperatures of a gas compressor. The measurement is correct before the startup and at the initial start, but then drops and deviates with temperatures when compared to manual reading done with a infrared gun, see reading under:
MBC Bravo temperature reading HMI Degree C 94 96 114 92 90 115
MBC Bravo Temperature reading with infrared gun Degree C 118 118 117 119 98 96
Thanks in advance!
The junction errors at those temps will only be at +/-2 DegC, worst case.
Copper extension wires are not that uncommon.
Just how precise are you trying to measure. Typically bellow 100 C or so you have to use RTD's, but that is only a judgement call.
That said, you should not be getting what looks like 20-30 C errors peak from the measurement. Those may be real.
my 2 cents,
You should keep in mind that infrared thermometers can not read very accurately (emissivity, type of surface, etc.). Usually, the reference to calibrate an IRed thermometer is a thermocouple. If you use copper extension wire or cable, you will get inaccuracies because of the cold junction compensation, which is inside the thermocouple module. This compensation measures the temperature at the module inputs and subtracts this value from the signal coming in. Therefore, if the junction box temperature (where copper and thermocouple wires are connected to each other) and the temperature at the PLC module location are the same, the error will not be significant. If not, the error caused by this will be that of the temperature difference between the 2 locations. If terminal blocks are used, it is best to use K type terminals as well. I hope this helps.
I'm old and grumpy enough to remember calibrating thermocouple devices before the days of digital displays.
The measuring device usually consisted of a standard cell, Wheatstone bridge and Null balance galvanometer. They always had a thermometer mounted beside the terminals and another potentiometer which you set for the temperature compensation.
What this has to do with the topic I'm not sure.
Just do it right!
And a bucket of ice for the cold junction for pre-startup steam turbine testing
>I'm old and grumpy enough to remember calibrating
>thermocouple devices before the days of digital displays.
>The measuring device usually consisted of a standard cell,
>Wheatstone bridge and Null balance galvanometer. They always
>had a thermometer mounted beside the terminals and another
>potentiometer which you set for the temperature
>What this has to do with the topic I'm not sure.
>Just do it right!
A friend inherited his Dad's process instrument field calibration business and his Dad's notebook.
The notebook lists locations of "Cold Junction" reference points for various plants, building or units, for instance, NW corner of the rolling mill, 6 feet west, 3 feet south, 4 feet deep, 54 Deg F.
Someone had found a stable temperature point, had measured the point and that's where the (separate) cold junction thermocouple junction was landed.
The cold junction thermocouple was run to the instrument, wired in series with the measuring thermocouple and instrument was then calibrated for a 54 Deg F cold junction reference, not a 32 Deg F ice point.
Likewise, not thread relevant, but interesting as to how clever instrument design and implementation was before microprocessors.