Thermocouple problems


Thread Starter


Hi, I'm having problems with temperature readings. I'm using a Mitsubishi FX2n PLC with 3 thermocouples extension block (fx2n-4ad-tc). Each block has 4 channels, so i can read 12 independent temperatures.

The system controls the temperature of a zone of a pipe line, the thermocouple is welded to that pipe line zone or just moored, so the thermocouple is grounded.

The measurement circuit is as follows:
Thermocouple->male connector->female connector->extension cable (like 20 meters)->male connector->female panel connector->compensated cable (like 1 meter)->thermocouple block channel (fx2n-4ad-tc).

All cables and connectors are K type.

The following problems happen independently for differents sets of thermocouples (differents pipe lines) and for anyone of the three blocks:
1 - When only one block's channel is used, the reading is OK, but when i connect other channel of the same block the reading of the first channel goes down about 30-40 Celcius.
2 - When two or more channels seems to work fine, a temperature drop happen suddenly of about 80 Celcius when they reach the 600 Celcius app.
3 - Slow oscillations readings happen when two or more channels are used.

The answer probably lies in the "miscellaneous" spec on page two of the FX2n-4AD-TC user's guide:

"Isolation: No isolation between analog channels."

Given that you have grounded the T/Cs at the hot end by welding them to the pipe, and all 4 inputs appear to be single ended with a common ground, I would guess that you most likely have ground loops.

Your description of varying amplitudes depending on which T/C is read by the multiplexor is a typical description of a ground loop or common mode problem with multiplexed, non-isolated, single ended analog inputs.


1) Use an ungrounded thermocouple element in a mineral insulated MgO sheath and fasten the sheath to the pipeline. This has a fairly high probability of eliminating the ground loop because the T/C element is not electrically in contact with the pipeline. It will be slightly slower response than a welded TC, but is the least expensive approach.

2) If you must weld the T/C element to the pipe, then you should consider using

a) a thermocouple card with isolated, differential inputs. I have no idea as to whether Mitsubishi has one or not.

b) a 2 wire, loop powered thermocouple transmitter WITH AN ISOLATED OUTPUT for each thermocouple and an appropriate 4-20ma analog input card. The transmitter takes a T/C input and converts it to 4-20mA. The isolation is needed to prevent a ground/common mode problem with an analog 4-20 input card like you have with the thermocouple card. You'll have to re-wire with copper wire, not T/C extension wire.

Dozens of instrumentation manufacturers have TC transmitters. Be sure to get isolated output.

4) You concluded with a mention of some cycling. Cycling could be real temperature variation, or it could be noise, or aliased noise. Mitsubishi recommends shielded T/C extension wire, but shielding can not diminish ground loop or common mode, only induced EMI, and your description doesn't sound like induced EMI. I would recommend twisted T/C extension cable for noise reduction, but don't tackle noise reduction until you get rid of the ground/common mode problem.


Curt Wuollet

I would suspect stray currents coming from someplace. I'd try grounding the pipes at the point where the TCs are attached.


Sadly the T/Cs must be grounded due to a work policy.

Apparently Mitsubishi does not have a T/C block with isolated and/or differential inputs, so the only solution seems to be the T/C transmitter and the analogs blocks.

Thanks for your answer.
I would challenge that policy of grounded t/cs. Ungrounded is much safer and is the solution to your problem. However, singled ended inputs are a nightmare for thermocouples used in your application. Did You try the TC transmitters? Did it work? I would like to know. Older PLCs do a poor job with thermocouple inputs.

Dan Farnan
[email protected]
Hi ,
I am making a data acqusation system with 20 K type thermocouples. my connection are:

thmermocouple> male connector> female connector> shelded wire of 15-20 cm > wire end soldered on pcb> mux dg407> ad595cq> muxDG406> adc ad574> 89C52> pc

I have got the same prolem of fluctions. when only one thermocouple is connected i get small data varation of 2-3 oC but if 2 thermocouples are connected the temp get out of control to 200 to 500. even at room temp. also when i put my thermocouple in hot water.

I have earth the the body of thermocouple and grounded pin-1 of ad595. if i do not ground it my output goes to zero even if thermocouple body is earthed.

plz if any one can suggest a solution. or tell me what and where the probem is.

Re your grounded thermocouples:

It seems obvious that the channels of your PLC are not isolated as several others pointed out. so as I see it you have 3 possible solutions:

1/ buy a transmitter for each and feed the signals into a 4-20 input module.

2/ see if you can purchase an input module with isolated inputs.

3/ use a 2 pole multiplexer so that the PLC is only reading one thermocouple at the time.

Solution 2 would probably be the cheapest since for 1 you would need a new 4-20 module anyway plus a transmitter for each channel.

Solution 3 would require some PLC program changes and a discrete output module to operate as a multiplexer directly (2 channels per thermocouple) or drive some good quality 2 pole relays. The advantage of multiplexing is all thermocouples would be read by the same A/D.

Hope this helps

Curt Wuollet

The thing to do here is take a sensitive multimeter and look at the voltages around the ckt. These voltages are really small and it's not hard to mess them up. The mux may not be suitable for handling signals at this level accurately, you might want to amplify first. And you have a couple of extra thermocouples in the circuit with the solder connections and whatnot. The companies that make instrumentation amp IC's usually have good application notes dealing with these error terms. Accurate TC measurements can not ignore the error sources because some are as large or larger than the signal.