• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 46 subscribers
  • Views 1693 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

to characterize 24-bit ADC using very high resolution DAC

DHAVAL PATEL
Prodigy 30 points
Other Parts Discussed in Thread: ADS1675, ADS1672

hello sir,

my project requirement is to characterize 24 bit resolution ADC using higher resolution DAC output signal. i want to use 28 bit resolution DAC but practically it is not possible. can i use parallel Current output DAC combination of 16-bit and 12-bit resolution DAC and merge them to get 28 bit resolution???

suggestions are invited..

thanks,

dhaval patel

  • 0
    TI__Guru* 98145 points
    Dhaval,


    I never tried to combine a 16-bit DAC and a 12-bit DAC to make 28-bits. I don't want to say you can't do it, but I'm not convinced it would ever work.

    First let's start with what you need to characterize. What exactly do you need to measure?

    Often times, we use a DAC as an input signal (a low noise version), and then use a precision multimeter (like an Agilent 3458A to measure the input and the reference voltage. With a precision measurement of the input and the reference, we can get an ideal output code and compare that with what the ADC actually gives us.

    Please think about what it is that you want to characterize and we'll see what we can come up with.


    Joseph Wu
  • 0 in reply to Joseph Wu
    Prodigy 30 points
    sir,
    I want to characterize 24-bit resolution ADC "AD1672" and want to get atleast 20 bit performance.
    for that characterization, one of the method is to generate and apply signal with DAC which resolution is higher than ADC(i.e. 28 bit resolution DAC) but maximum resolution DAC available in market is 20 bit.
    what would you suggest to characterize 24 bit ADC???
  • 0 in reply to DHAVAL PATEL
    TI__Guru* 98145 points
    Dhaval,


    Generally with DC characteristics, we make measurements as I described in the previous post. We take a low noise DAC and measure back with a precision multimeter. Since the multimeter is known to be very linear as well as accurate, we can compare the input and the reference, determine the expected code and compare with what the ADC gives.

    With good characterization and the topology of delta-sigma ADCs, no missing codes and monotonicity is reliable.

    We don't try to input a DAC code and get the same ADC code back when we test these devices. That's not how we generally characterize devices. We will large numbers of codes in the our own initial characterization, but it's not done on each device. Imagine trying to test even 2^20 codes in a DC application. You don't try to set the DAC a million times and resolve a reading for each one.

    Now, what characteristic is it that you want to measure? INL? Gain Error? Offset? I know more about DC measurements because I generally look after the low speed precision data converters. The ADS1675 has a much higher bandwidth and involves a lot more AC measurements. This requires some precision equipment and I think that we've used some from Rohde and Schwarz or Bruel and Kjaer.


    Joseph Wu
  • 0 in reply to Joseph Wu
    Prodigy 30 points

    hi,

    suggest all possible methods to characterize low speed AD1672 for DC as well as AC characteristics, I want at least 20 bit performance with all effective parameters like offset, DNL , INL etc

  • 0 in reply to DHAVAL PATEL
    TI__Guru* 98145 points

    Dhaval,

    I'll cover a few of the characteristics that you can measure. There will be some explanation with some example data.

    I'll attach an excel spreadsheet with a 21 point measurement that measures INL, offset, and gain error. This data is for a different ADC, but the ADS1672 would use a similar method for DC measurements. In delta sigma data converters (with 1-bit feedback), the ADC is monotonic when you discount the effect of noise. That's why normally, these ADCs never specify a DNL.

    As I mentioned before, you need a DAC that drive the input but it does not need to be high resolution, it needs to be low noise. It's not my intention to try to measure every single code. For a 24-bit ADC that's 16 million codes. I use two DACs so that I can adjust the input independently. For this ADC, the input range is +5V to -5V (or 2xVREF based on a 2.5V reference).

    I also need to measure the reference. This way I can get an accurate measurement of the gain error. Note that I'm not measuring the gain error of the system. I'm only measuring the gain error of the ADC. I only care about what ratio of input to the reference to get what I would call an ideal output.

    Inputs are set so that there are 21 inputs distributed through the entire input range (dividing -4.95V to +4.95V into 20 equal parts). I don't push all the way out to the edge of the measurement for fear of over ranging the input. Also, inputs are set so that the input has a common mode at mid-scale for all measurements. We take an odd number of input voltage points so that we include an input of 0V as an input. This way I can record the offset.

    For each measurement, I take many points and average them. Here, it's 100 points for each measurement. This gives me my noise vs input signal. I use the average of each of measurements, but I can also record the standard deviation.

    Referencing the attached excel file, Column B is the ideal input with what I want to generate, while column C and D are the inputs for AIN+ and AIN- from the DAC that are used generate the differential input with a common mode value of 2.5 V.

    Column E is the measured input of from the Agilent 3458A, while column F is the output code of the ADC that has been averaged for noise.  Column G is the voltage read by the ADC, scaling the code read from the DUT with the reference voltage measured by a multimeter (the reference is measured to be 2.5001661V). Note this is my "gold standard" measurement. I know that the 3458A has the best accuracy and linearity that I can get to make measurement.

    With column H, I calculate the ideal output based on the measured input. I take the slope of the line created from the endpoints of the output code from the DUT, and the measured value of the input. Assuming that the measured value of the input is golden, I calculate the expected ideal value of the output code. Then I take the measured value, and subtract the expected ideal value. Column I shows the result of this which is also the calculation of the INL. As a bonus, column J shows the INL error in parts per million (ppm) in relation to 2^24 codes.

    Plotting the INL of column I with the input voltage of column E, gives the plot in the file.

    We do much more, but this method gives you a way of determining four important parameters. There are some notes in some of the excel spreadsheet to explain what measurements are being made and how they are calculated.

    I'll see what we do with AC parameters, but most of that requires specialized equipment. For audio and lower data rate ADCs, we've used an Audio Precision System 2 to make some measurements. Regardless, it requires much more specialized equipment to do this type of measurement.

    Joseph Wu

    INL example.xlsx