What is the purpose of the opamps connected back to back in a thermocouple S.C?

[electronx]

I am trying to design a thermocuple signal conditioner. I came across this structure on the Analog devices LTC 2057 typical application page. Why was such a connection made instead of using a single opamp as a differential amplifier?What is the purpose of the opamps connected back to back in a thermocouple signal conditioner?

(LTC 2057 High Voltage, Low Noise Zero-Drift Operational Amplifier)

[bostonman]

Probably being used as a pre amplifier.

[MarkT]

What is the purpose of the opamps connected back to back in a thermocouple signal conditioner?

To sell Linear Technology chips - its normal for examples in datasheets to only use many devices made by the manufacturer of the datasheet, rather than be a good way to do things.  Here the classic solution would be an instrumentation amp, not this cascaded approach, but that wouldn't involve the chip this datasheet came from...

These days just use a thermocouple chip with digital output for an easy life.

[Nominal Animal]

Both opamps are in the inverting configuration, so that the feedback path of the gain amplifier (LTC2057) does not involve ground; I do believe this helps with common mode rejection ratio, which is exactly why this circuit claims such a high CMRR (122dB).  The LT1991A only buffers the differential signal (and effectively inverts it); it adds no gain.

Edit: That said, MarkT is right: do you need that kind of performance?  You can affect the reading of a small thermocouple by moving around in the same room (by causing airflow changes!), so unless you're doing some kind of scientific experiment involving careful temperature readings (noting that the measurement itself will inject a small bit of heat into the system), it's unlikely you'll need that.  For something else generating slow-changing tiny potential differences, this two-opamp circuit might come handy.  Assuming the figures claimed are achievable in practice.

[David Hess]

The differential connection to the thermocouple helps reject noise, but requires an instrumentation or difference amplifier.  Instrumentation and difference amplifiers typically lack the precision to work directly with a thermocouple, so in this case the LT2057 is used as a precision preamplifier.

Linear Technology published another differential thermocouple amplifier circuit which used the LTC1043 switched capacitor building block to do the differential conversion before precision amplification.

[coppercone2]

because thermocouples can go near very high power devices. Like if you had to build it into the cabinet for a industrial kiln. Lets say its running at 100kW.

Interesting how complicated and sensitive certain ceramics are for manufacturing.

[Terry Bites]

 You need a lot of  gain, a lot of DC precsion and good noise performance to amplify the thermocouple potental. The LTC2057 has excellent 1/f noise and offset characteristcs at low freqencies. The LT1991 isn't doing much lifting here, its used as a unity gain diff amp. It references the measurement to a single point (the red terminal) instead of ground- this lets the themocouple float over the CMV range shown. Every place disimilar metals touch in connectors pcbs and every solder joint is a thermocouple in it own right. Ground becomes a fuzzy temeprature dependent concept in these circuits. Its important to reduce the impact of these potentials much as possble. Referencing the thermocouple measurement to ground will lead to poor performance. 

This circuit isn't necessarily the best soultion. More modern ASICs will problably outperform it.  Beware the musings of the app lab.

[MarkT]

Instrumentation and difference amplifiers typically lack the precision to work directly with a thermocouple

Huh?  That's exactly the sort of application they are designed for, small signals, low bandwidth, good CMRR required...  INA128 at gain=100 has typical gain error of 0.05%, offset voltage 50µV...
The AD8230 is an auto-zeroing instrumentation amp has typ 0.01% gain error and max 10µV offset...

[wasedadoc]

BTW, the configuration in the schematic is not "back to back" in my book.

[electronx]

What is the name given to this type of connection in your book?

[electronx]

Instrumentation and difference amplifiers typically lack the precision to work directly with a thermocouple, so in this case the LT2057 is used as a precision preamplifier.

Yes for the discrete difference amplifier  - partially for the differential opamp containing gain resistors inside waffer -
Is it a bad idea to use high cmrr value (High CMRR: 111 dB minimum, G = 1 V/V ) and Zero drift instrumentation amplifiers (about 40uV output Vos)?
(ADA4255: Zero Drift, High Voltage, Programmable Gain Instrumentation Amplifier with Charge Pump Data Sheet (Rev.0))

Moreover, the differential input seems very logical for my sensitive ADC converter. Also, does it make sense for me to use an instrumentation amplifier with a lower Vos output and use an auto zero fully differential amplifier on its output? (my output should be differential) At the same time, what is the maximum offset so that I can make a sensitive analog signal conditioner at laboratory level?  Of course, cold junction compensation will be made. But how much offset is acceptable at the analog level? Cost is not important. My goal is to achieve as much perfection as possible.

[wasedadoc]

What is the name given to this type of connection in your book?

"Cascaded" or "in tandem" or "in series".  If they were logic gates (output of first to input of second) or two transistors (collector of first feeding base of second) you would not say "back to back".

[David Hess]

Instrumentation and difference amplifiers typically lack the precision to work directly with a thermocouple

Huh?  That's exactly the sort of application they are designed for, small signals, low bandwidth, good CMRR required...  INA128 at gain=100 has typical gain error of 0.05%, offset voltage 50µV...
The AD8230 is an auto-zeroing instrumentation amp has typ 0.01% gain error and max 10µV offset...

Parts like the INA128 and AD8230 are exceptions and not typical, and they still do not meet the performance of the example circuit.  At a gain of 100, the LTC2057 has a typical error of 0.00032%, so more than an order of magnitude better.

The Fluke 80TK thermocouple amplifier design gets by with a single TLC271, but has to operate with a very low supply current.  I have an ongoing project to replace the TLC271 with an OPA187 low power chopper stabilized part to see if it makes any improvement.

Moreover, the differential input seems very logical for my sensitive ADC converter. Also, does it make sense for me to use an instrumentation amplifier with a lower Vos output and use an auto zero fully differential amplifier on its output? (my output should be differential) At the same time, what is the maximum offset so that I can make a sensitive analog signal conditioner at laboratory level?  Of course, cold junction compensation will be made. But how much offset is acceptable at the analog level? Cost is not important. My goal is to achieve as much perfection as possible.

The noise and signal levels for thermocouple measurement do not require anything special on the ADC side.  The differential connection to the thermocouple may be preferred in high noise environments, like if the thermocouple was very remote, however most thermocouple applications use a single ended configuration.  Noise at the output of the amplifier is going to be great enough that nothing special is required on the ADC side.

Low input offset drift is required for the thermocouple amplifier depending on the operating temperature range of the cold junction compensation, which is typically rather limited in an indoor application.  For example with a common type-K thermocouple, the voltage coefficient is 40 uV/C, so a change in input offset voltage of 40 microvolts would be required for a 1C error, and with something like the not so good TLC271 (1uV/C) in the Fluke 80TK thermocouple amplifier, this would mean a 40C change in temperature of the amplifier and cold junction compensation which would be exceptional.  Note that 1 uV/C is actually pretty good for a CMOS operational amplifier.  Bipolar precision parts can be 0.1uV/C, and chopper stabilized parts may be 0.05uV/C or better, but at this point other considerations become more important.

The input offset voltage is calibrated out with a single point calibration of the thermocouple, so is not a concern, although if the offset was low enough and the cold junction compensation was accurate enough, no calibration would be required.

Modern delta-sigma ADCs are good enough to directly digitize a thermocouple output.  No amplifier is required and they make temperature measurement ASICs for exactly this.