DAC Accuracy

Discussion in 'General Electronics Chat' started by laguna92651, Nov 16, 2008.

  1. laguna92651

    Thread Starter Active Member

    Mar 29, 2008
    101
    0
    I've run an experiment on an active filter. I using the NI Elvis prototyping board. The frequency response is taken with the NI Bode analyzer and I saved the output data in a text file. The Bode analyzer amplitude spec is give as 12 or 16 bits, how do I calculate the amplitude error? The measurements are in dB.
     
  2. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    You really can't until you pin the resolution down. 12 bits is 1 part in 4096, 16 bits is 1 in 32,768. Is that resolution selectable somehow?
     
  3. laguna92651

    Thread Starter Active Member

    Mar 29, 2008
    101
    0
    The spec sheet says it is "12 or 16 bits, DAC device dependent". I would just pick one, say 12 bit. If I had a output reading of 6.54dB, how would I calculate the error for the reading.
     
  4. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    That depends on the stated output accuracy. If it is +/- one bit (kinda sloppy), that is one part in 4096. You need to establish the level for 0 dB, translate that to a 6.54 dB voltage gain, and take that over 4096 (or whatever the conversion error is stated to be).

    Or you can take the 0 dB level over 4096 and calculate that in dB. That should be a constant error figure.
     
  5. laguna92651

    Thread Starter Active Member

    Mar 29, 2008
    101
    0
    I'm not sure I follow. The input is 1 v ac and when I measured the center frequency (the circuit is a passband active filter) I get 9.71 dB at 5150Hz. I calculated the gain to be 3.06. I just need to calculate the measurement error on the 9.71 dB reading, and the other data readings.
     
  6. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
  7. laguna92651

    Thread Starter Active Member

    Mar 29, 2008
    101
    0
    I can calcultate dB, it is incorporating the error into the reading that I'm not sure about. If the 12 bit error is 1 part in 4096, would the error in measuring 9.71 be 9.71/4096 = .0024dB?
     
  8. Distort10n

    Active Member

    Dec 25, 2006
    429
    1
    This seems correct, but this assumes a total error of +/-1 bit which, in my mind, includes offset, DNL, INL, noise, gain error, temp drift, etc.

    So you want to know how much of the result is due to the error, correct?
     
  9. laguna92651

    Thread Starter Active Member

    Mar 29, 2008
    101
    0
    Yes, that's correct.
     
  10. Distort10n

    Active Member

    Dec 25, 2006
    429
    1
    Then it should be easy to calculate the % error. You can also convert % error into parts per million (ppm) or in dB.

    Calculate the ideal gain of the filter. Assume the components are their ideal value so they have no tolerances, parasitics, tempco's, etc.
    Also assume the active filter has no error either; i.e., no offset, bias currents, PSRR or CMRR errors.
    Finally assume that the DAC is ideal so it has no offset, DNL, INL, or gain errors.

    You have a known input amplitude and known 'ideal' gain. Compare this calculation to the resulting measurement from your circuit. The difference is the error.

    Finding out all the error sources and compensating for them is another problem, but that is not your question. ;)

    I hope this helps.
     
  11. laguna92651

    Thread Starter Active Member

    Mar 29, 2008
    101
    0
    thanks for the help.
     
Loading...