Post Go back to editing

How to measure negative temperatures with a thermocouple on a battery powered device.

I am struggling to figure out the best/most affordable way to accurately measure a thermocouple with a single positive supply (2 AA batteries).  First off, I am trying to measure temps down to -200C (liquid nitrogen), the board itself will be at room temperature (~25 C).

My current circuit works OK, but doesn't track well as the battery level changes.  The batteries start at ~3.7V and drop down to 2.5V before failing.  I am currently using the ADA4051-2.  Channel 1 is a differential amplifier with gain ~= 196.  Rf=Rg=332K and R1=R2=1690.  Channel 2 is a level shifter and provides an offset equal to Vref*.5.  I chose .5 because it scales well over the range 3.7V-2.5V. 

For example when the battery is at 3.7V, the offset would be equal to 1.85.  Assume we are measuring liquid nitrogen and the temp is -200C.  Using the <a href =>table</a> for a K-Type thermocouple, the voltage into the differential amplifer should be -5.891 mV.  With a gain of 196 we would expect the output of the differential amplifier to be -.005891*196 = -1.1546V.  Because the A/D cannot accept negative voltages we shift this to a level our A/D can read.  I've attached an image to show how the level shifter is wired.  Vref is equal to Vdd * R4/(R4+R3).  Right now I have R3=R4=1K, so Vref = Vdd*.5.  So at Vdd==3.7, Vref = 1.85.  After the level shifter we get -1.1155+1.85 = .6954V.  This is the voltage into the 10-bit A/D.

Here is where it starts getting a little tricky. 

The Reference of the A/D is whatever the battery level is.  Also the offset voltage is dependent on the battery level.  Without having a precise measurement for the battery level all other measurements fail.  This also leaves some intense tables/math that need to be made if you want to be able to get the actual thermocouple temperature.  Ultimately there are three variables, junction temperature (we use a thermistor and a seperate A/D reading, again simple voltage divider, which allows the A/D output to be independent of battery level), battery level and thermocouple voltage difference.

You might ask, why not throw a precision switching regulator at the output of the battery terminals and regulate the board down to 2.x Volts.  My concern with this is battery life.  If the device lasts for 1+ years, will this still be achievable with a regulator?  Also we could use a precision reference IC but this adds some complexity and currently the board is setup to only use the battery level as a reference.

The best way would be to have the circuit battery level independent.  One way I thought of was to have the gain of the diff amp change with as the voltage changes using a FET or VCA, but then theres the issue of figuring out what the gain is...

Lastly, to use the AD8495 to measure temps of -200C what would the reference need to be?  What would you reccomend?

If anyone is interested here is how I determined the actual temp from the A/D output.

Say the A/D output in HEX is 160.  160 in DEC = 352.  We used a 10-bit A/D (2^10=1024) and had Vref = 3.00 V so the output of the diff amp was 352/1024*3.00 = 1.03125 V.  Assume for this example the offset is equal to 1 V.

1.03125 - 1 = .03125 (output after being amplifed, with offset accounted for)

The gain of the amp is 196.

Thermocouple output = .03125/196 = .00015943 (look this value up on the table, .15943mV == ~4 degrees C)

Determine the junction temperature, we have an onboard thermistor wired into a seperate A/D.  For simplicity lets assume the board/junction temp is equal to 25 C.  At 25C the thermcouple compensation is 1mV.  Remember the thermocouple measures the difference in temp between the board and whatever it is your temping, so you must compensate for the board temperature.

Simply add the two (1mV + .15942mv) to get the temperature at the thremocouple probe tip.  So 1.15942mV.  Look this value up and that is the temp you are measuring. (~29C).


An ADC is a math function:  a divider.   Vin/Vref = 0 to 0.999999 converted to binary.  You are trying to measure an absolute voltage,

and compare it to the reference pin, which is proportional to the battery voltage.  We do make an AD510, a 1.0 volt shunt reference,

but it needs 100 uA minimum to meet specs.  The ADR280 is another idea, it's 16 uA.

Two AA cells are 3.7V??  What type of batteries?  

I'll think about this some more......


  • Harry, thank you for the reply.  Yes, I've mesured the output of 2 AA L91's and on occasion have seen voltages up to 3.7V, typicaly they output around 3.5V.  The ADR820 looks like a nice product.  I guess we could use that for the A/D Ref and Voltage offset.

    Would this also be used for the AD8595?


  • Harry, yes my apologies it was meant to read ADR280.  I don't understand how the shunt reference IC's work.  Is it just a voltage supply with a series resistor that regulates the current into the diode?

    Yes I was referring to the AD8495.  If we used this IC, the power would be sourced from the microcontroller.  We can turn the device on, take a measurement and turn it off, so the consumption of 250uA is not unbearable.

    Thank you for helping!

  • jr,

      You can also look at fig 26 on the OP295 datasheet:

    You can take that basic idea with an ADA4051-1 and an ADR280 and come up with a good design.

    You mentioned "ADR820";  I think that's a typo.

    Thanks for the link on the Lithium/Iron DiSulphide batteries.  I was not familiar with them.

    I don't understand:  "Would this also be used for the AD8595?"

    Are  you refering to the AD8495?  It's 250 uA, which I am assuming is too high for you?

    Another group handles the AD8495, so I will send an email to apps engineer that covers the part

    to join the discussion.


  • jr,

      Think of a shunt reference as a Zener diode.  You have a resistor from V+ to the diode to ground.  As the current to the

    load changes, the shunt takes more or less current to keep the voltage constant.

    IC shunt references have been around since at least 1971.  They have much lower dynamic resistance than a Zener.  See:


  • This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

    Thank you,
    EZ Admin