Post Go back to editing

LTC2499 datasheet recommended buffer seems like an odd choice

Category: Datasheet/Specs says "The LTC6078 is an excellent amplifier for this function." in regards to an external buffer.

My understanding is the autocallibration will remove Vos and drift.

Looking at the LTC6078 datasheet I saw this graph of Vos vs Vcm:


For a single-ended sensor am I reading this correctly in concluding that I should expect up to +/-  40uV (~7 bits?) of nonlinearity from poor CMRR and that this will not be removed by the autocallibration?

If that is the case, is either an OPA141 or OPA196 a reasonable choice? Any other recommendations for a particularly suitable buffer for this ADC? (This is to avoid impedance mismatch in reading a  single-ended sensor with a variable impedance from ~200k to 200R).


Top Replies

  • Hi, 

    I would recommend moving or posting this question on (+) Q&A - Amplifiers - EngineerZone ( someone here may be able to assist you. 

    We will also contact the product owner if you have any specific question/concern about LTC2499. 



  • Before I do that, I'd like to get clarity on selection criteria for a buffer for the LTC2499 to extract as many accurate noise free bits out of a variable impedance sensor as the 2499 is capable of. The datasheet proposes the LTC6078 but doesn't specify why. In particular exactly what the self-calibration will remove isn't clear as are any speed requirements related to the input switching due to auto-calibration.

    My current guesses:

    Vos and drift do not matter, they are removed in calibration

    Low voltage and current noise is essential and anything much worse than 1uV peak to peak <10Hz will degrade effective resolution.

    The different Vos near the positive and negative rails of most rail to rail output amps will degrade accuracy

    High CMRR is essential

    Low input bias current is obviously needed for high impedance sources

    Recommendations around using the buffer as an active low pass filter (to limit noise bandwidth) would also be helpful.


  • Hi  ,

    Primarily, the reason for using the LTC6078 is due to its high input impedance and low offset. It also has rail-to-rail input and output so that it can operate on a single 5V supply and swing from 0-5V for both input and output.



  • I thought the LTC2499 automatically calibrated out the offset. Is there some advantage to using a low-offset amp in this application?

    Is my interpretation of the LTC6078 Vos vs Vcm graph correct in concluding that the 6078 will introduce ~7.7 bits of INL (ln(60/1000000)/(5/(2^24))/ln(2)) with particularly high nonlinearity around 4V? (reading the graph as roughly typical 60uv offset v CM) vs the 2ppm of vref 

    I had sort of concluded that the driver switchover offset inherent in rail-to-rail output op amps was going to be a problem for anything that was trying to wring maximum performance out of this ADC with VREF=5V and for that reason and had resigned myself to at least a higher voltage rail for the buffer.

  • Hi,

    The external opamp buffers are included in the calibration loop, so their DC errors are removed. But I think one should be very careful: it only works if the signal range is within both the input and output range of the opamp. I can’t see in the datasheet how it is implemented, but the auto-zero step is typically done by disconnecting the positive input from the signal, and tying it to the negative input. You can find this scheme in the AN-1464 application note:

    You’ve mentioned single-ended application, when the COM pin can be grounded. In such a case during an auto-zero cycle the positive input can get grounded too, and the output of the opamp buffer can be saturated (the output can’t reach the supplies perfectly). In this case you must disable the calibration.

    I think the LTC6078 is an excellent choice as the datasheet says, even without calibration (much better than the competitor’s opamps you’ve mentioned):

    • In differential mode (and if both input signals are in the linear range of the opamp) the offset and its nonlinearity is removed in the auto-zero cycle.
    • If you can’t use calibration, offset and its nonlinearity should be considered. What you’ve calculated is relative to the perfect 24-bit linearity. You got 7 bits, that means that the linearity level is 24-7 = 17 bits, i.e. about 8 ppm. So 40uV of offset means about 1 LSB at 17-bit resolution (38uV at 5V full-scale). I think it is excellent, compare this to the accuracy of your sensor.

    One more important sentence from the datasheet: 

    "Highest linearity is achieved with fully differential drive and a constant common-mode voltage (Figure 38b). Other drive schemes may incur an INL error of approximately 50ppm."

    So, for example in single-ended mode the INL can be higher than the LTC6078 opamp’s contribution.