How to amplify thermocouple readings

[LesHammer]

> With the 500kHz source powered down the TCs would read similar temperatures,

> but when the 500kHz source was powered up (to be used as a heating element)

> one TC would increase in temperature and the other would decrease. 

How do the connectors and wires look?  Any crud on them?  My first guess would be a bad connection - sure it's making some contact so it measures the voltage on the thermocouple, but when you turn on the 500kHz "radio" the thermocouples act like an antenna, the signal gets rectified allong the way and you see it as an additional dc component.

--Les

[UCLA-engineer]

Everything is clean and if I plug an Omega handheld TC meter into the line instead of the NI system I get realistic results that do not fluctuate when the 500kHz line is energized.

[LesHammer]

> if I plug an Omega handheld TC meter into the line instead of the NI system I get realistic results

Hmm, a handheld would have excellent common mode rejection (battery operation does that for you.  🙂

However, a wall powered NI system wouldn't ... Is your circuit grounded any way?  Are you generating common mode voltages with that 500kHz?

--Les

[UCLA-engineer]

After sampling the output from the 500kHz generator with an oscilliscope I realized the frequency fluctuates slightly at around 450kHz.  I would like to just oversample at 2x the frequency and average the results but sadly no device samples at 1MS/s.  I tried increasing the sample rate to 333kS/s but the "DAQ Assistant" maxed me out at 125kS/s for reading one TC and even lower for reading multiple TCs.  I am not sure why the 2Hz lowpass filter isn't fixing this problem for me.  Would putting a capacitor from ground to the copper line and another from ground to the constantan line help (Type T TC).  Can anyone think of another method of solving this problem?

[tbob]

You have to get rid of the noise being induced by the 500KHz generator.  No amount of oversampling or undersampling or grounding will fix your problem.  In fact, extra grounding will probably make it worse.  There are several ways to fix the problem:

Induction can only occure where there is a loop.  If a 500KHz (or any frequency) square wave is run through a single wire, it will radiate an electrical field consisting of many harmonic frequencies well above 500KHz (the harmonics depend on the rise time of the square wave).  To stop the radiation, you should either run the 500KHz through a coax or a twisted pair.  If neither of these are happening, you have a large loop that can be described as starting from the 500KHz source and going to some ground plane that is physically separated from the actual signal physical path (wire or PCB trace).  The larger the distance between the signal path and the signal return path (often called ground), the more radiated energy will be emitted by the loop.  Coax or twisted pair will greatly reduce the problem.

If you can't fix the source of the radiation, then you can try to fix the susceptibility of the thermocouple circuit to induced voltages.  Again, twisted pair wire should be used to connect the thermocouple to the amplifier.  The wires should be twisted all the way to, or as close as possible to the amplifier.  If the wire is twisted up to a connector on a PCB, and traces run the signal to the amplifier input, the traces on the PCB should be as close together as possible from the twisted pair connection to the amplifier differential input.  (I hope you are using a differential amplifier).  Any separation of wires or PCB traces will cause a loop that can pick up noise from any radiating source.

If you can't fix the close proximity of the PCB traces, you can try to put a balanced filter inline (see picture below).  Put equal resistors (you will have to play with the values) inline with BOTH leads of the thermocouple.  Put a capacitor (play with value) from the end of one resistor that goes to the amp to the end of the other resistor that goes to the amp.  This method will not cut out as much of the noise as the first two methods above, but it may work.

Message Edited by tbob on 09-29-2005 10:55 AM

[jacbenimble]

I also had the same problem.

Please see the

"Thermocouple readings NG on scxi chassis but ok using FLUKE calibrator. "

Thread