LTC2984 Thermocouple Wiring

I'm designing a PCB using the LTC2984 IC. I'm planning on reading as many K-type thermocouples as possible with a diode as the cold junction.

I've attached the thermocouple wiring suggestion in the LTC2984 datasheet above. I have 2 questions:

  1. What is the advantage of using a differential pair to read the thermocouple if one is shorted to ground as in the 2nd example? In my PCB design should I just have the 2nd ADC shorted directly to the ground plane?
  2. What is the purpose of the resistors in line with the thermocouple? In the Maxim 31855 ICs I've used before, we just directly connect the thermocouples to the ADC without resistors. If they are necessary, what are the recommended values for them?

I'm still studying in college so a simpler explanation would be very appreciated.

Thanks in advance!!

Top Replies

    •  Analog Employees 
    May 6, 2020 +1 verified


    First question - diff vs. single-ended: 

        The advantage of the differential input is that any interference or offset in the common mode voltage (common mode means common to both parts of…

  • +1
    •  Analog Employees 
    on May 6, 2020 9:45 PM 2 months ago


    First question - diff vs. single-ended: 

        The advantage of the differential input is that any interference or offset in the common mode voltage (common mode means common to both parts of the differential pair) does not affect the measured signal of interest, the differential voltage. We show the wiring suggestion with one leg tied to GND just as an easy way to ensure the thermocouple's common mode is held at an acceptable value (the inputs can go below GND on the LTC2984, ensuring cold temperatures can be measured). Without some sort of bias voltage the high input impedance buffers of the LTC2984 could allow the thermocouple to "float" - the common mode voltage could exceed the input range of the part and cause an error in the reading. You are welcome to tie a pair to a known voltage other than GND. 

       The obvious downside to the differential input is that it requires two input pins on the LTC2984. For your application I would suggest going single-ended, but watch out! With a single ended input, COM is the negative reference, so it needs to be tied to the same potential as the negative leg of the thermocouple. You say you have a ground plane in your design, this is good. You'll want to make sure the negative lead of the thermocouple and the COM pin of the LTC2984 are at the same potential - with a very low resistance ground plane this is almost always the case, but keep in mind that even the ground plane has some resistance - if the two connections to it are far away and there is another component sending current into GND, this resistance could cause a difference in the two potentials. Keep your ground plane whole (i.e. no necking or other restrictions) and you should be OK. 

    Second question - in-line resistances: 

        As this diagram is a wiring suggestion, the resistors are optional, much like the chicken and vegetables in the Top Ramen pictures. 

        Here, along with the capacitors shown, they form an RC low-pass filter. They can also serve as overvoltage protection by limiting the current into the LTC2984. Refer to my LTJournal article for more details:

        They are optional, but you may want to include footprints for them and do the initial build of your PCB with low values, try 100 ohm and 100 pF to start - that will filter out very high frequency interference without impacting settling time or open-circuit detect too much. 

        Note that the application of these series resistors is only possible because the LTC2984 has buffered inputs with a maximum of 1 nA leakage currents. The part you've previously used has a maximum 100 nA on the thermocouple inputs, meaning any errors across series resistors from leakage currents could be up to 100x worse for the same resistance values. That is why those parts do not recommend placing resistors in-line with the thermocouple. 

    Hope it helps,