I am working on a thermocouple measurement amplifier, and came across INA126 which seems to work well for this type of measurement.

My thermocouples are the grounded type, so the measuring tip is coonected to the braided steel sheath. In the thermocouple article on page 3 they suggest to ground the thermocouple on one side with a 1Meg reistor for the grounded type thermocouple. Why not use a resistor on both sides to get more symmetrical input impedance?

Also, where would you connect the shield to? Ground it directly,? The thermocouple will be inside an electric kiln in close proximity to the heating elements, so I´d like to tie the whole setup to PE.
So should I tie all grounds together and then to PE?

 

I saw this conversation recently. The answer was: If you don't use symmetry in the thermocouple circuit you change the amplifier from differential to single ended and lose the CMRR benefits. I think you have been misled in the recommendation to ground one side of the thermocouple.

 

The advice from the article is not to ground the thermocouple, but to add a high value resistor to ground so that there is a path for the input bias current.
Nontheless my simulation says that with two 1meg resistors the output offset is much less and CMRR much better than with just one.

What do you think about the grounding of the shield? I think the best way would be to connect it to the analog star ground, and this point then to the digital ground and subsequently PE at the power supply.

 

Actually I think that if I connect the shield and thus the tip midpoint of the thermopcouple to ground, I don´t need those 1meg resistors since the input bias current will be taken care of by the shield. Am I correct on this one, or will there be any disadvantages in groudning the shield?

 

My opinion is your post #3. The (2) resistor method. 1 meg high to one side of the thermocouple and 1 meg low to the other side of the thermocouple. Then treat the shield as you said. The braided shield has no business being connected to an input pin. It's just a shield against radiated electric fields. Ground it, but don't ground it to an input pin.

 

The thing is the the shield is somehow connected to the thermocouple wires inside the potted end of the probe and I can´t do anything about that, so I have to treat the shield as a part of the input circuitry.

 

I didn't understand that. Must have been assuming too much. The shield can't be connected to both sides of the thermocouple, so I guess you'll just have to treat it as one of the inputs. Still, the 2 resistor method will center the signal and give you the best CMRR.

Notice that I changed one connection to a 1 meg resistor to get your millivolt input centered in the voltage range. If this isn't the right thing to do because you have a dual supply, ignore it.

 

I use the AD595chip ready made for thermocouples it gives 10mV per deg C, if you use the diagram on page 3, with a 12volt supply,

you can alter the gain by adding extra resistance between pins 8 and 9, to give more mV per C.