Help with opamp selection for thermocouple ADC

[[email protected]]

I have an ADC which takes an input signal in the range of +/- 64mV
relative to ground. I am using this to read a thermocouple for
measuring exhaust gas temperature, and it works fine if I use a probe
with an ungrounded hot junction; I tie one side of the cold junction to
ground at the ADC and measure the other side. In order to use a more
common grounded probe, which is my goal, I need a differential stage
before the ADC.

The +/- 64mV range is just about right for my purposes, so I do not
want any gain, nor do I want to switch ADCs if I can help it. If it
simplifies the analog stage I can deal with a 0-64mV input and output
range ( i.e. I do not need to measure probe temperatures below the
"cold" junction temp, although it wouldn't be undesirable to be able to
do so.) I do need to reliably read values very near 0 volts.

I understand how to use an opamp to convert the difference in voltage
between the two halves of the cold junction to a ground-referenced
voltage. I am having trouble selecting the correct opamp. I want to
minimize errors due to offset voltage or nonlinearity, but if possible
I don't want to spend $8/ea for the parts.

I am using a single +5v supply.

Any advice will be appreciated.

Thanks

 

[[email protected]]

I have an ADC which takes an input signal in the range of +/- 64mV
relative to ground. I am using this to read a thermocouple for
measuring exhaust gas temperature, and it works fine if I use a probe
with an ungrounded hot junction; I tie one side of the cold junction to
ground at the ADC and measure the other side. In order to use a more
common grounded probe, which is my goal, I need a differential stage
before the ADC.

The +/- 64mV range is just about right for my purposes, so I do not
want any gain, nor do I want to switch ADCs if I can help it. If it
simplifies the analog stage I can deal with a 0-64mV input and output
range ( i.e. I do not need to measure probe temperatures below the
"cold" junction temp, although it wouldn't be undesirable to be able to
do so.) I do need to reliably read values very near 0 volts.

I understand how to use an opamp to convert the difference in voltage
between the two halves of the cold junction to a ground-referenced
voltage. I am having trouble selecting the correct opamp. I want to
minimize errors due to offset voltage or nonlinearity, but if possible
I don't want to spend $8/ea for the parts.

I am using a single +5v supply.

Any advice will be appreciated.

Thanks

Unless you can be sure that your thermocouple's common mode voltage is
above 0 volts your circuit won't work. An op amp must have a power
supply greater than the range of input voltages (at least all the ones
I've used). For example you won't be able to buffer a signal that
varies 64 mV above or below -3V with an op amp that has a 0 to +5V
supply.

Typically you use a grounded thermocouple when you're trying to measure
the temperature of some hot metal chunk in a somewhat noisy
environment. Problems arise if there are several different types of
metal between your hot junction and your adc. These different metals
form thermocouple junctions that will affect what you read from your
hot junction.

I'm afraid you only have 3 options.

1...Continue to use an ungrounded thermocouple. They make some nice
expensive ones that are in a stainless steel sheath. These may be too
fat for your application.

2...Verify that the metals between your ground point and the
thermocouple always produce a positive common mode voltage between 0
and 5 volts. If someone changes the type of metal used somewhere in the
system you might get hozed.

3...Bite the bullet and design in a negative supply for your op amp. If
you have something like a Max232 chip in your system you can sometimes
tap -10V off of that. You won't need more than a few milliamps of
current.

Something else to consider is your op amp will have some minimum offset
voltage, typically hundreds of microvolts. Depending on the type of
thermocouple and type of op amp used this can result in an error of
several tens of degrees. If you have a micro in your system, you can
get around this by shorting the thermocouple from time to time and
seeing what your adc reads. Save this value and subtract it from your
thermocouple reading before doing any kind of temperature lookup.

Mark