The DAC has an INL of about 20 or so LSBs. DNL is +-2LSB. Since it is a 14 bit DAC, is it safe to say that the DAC is monotonic for 13 bit use? (Since DAC is monotonic if DNL is not below -1LSB)

And secondly, what INL is considered good enough for DAC? Of course it should be as close to 0.5LSB, however, is there any other information provided by INL that needs to be considered?


What you have to realize is that the DNL measure is a noise-like measure. Meaning you can use it in noise calculations, simply put it is a RMS measure. That means that it's entirely possible to have low DNL and still not be monotonic. Monotonicity can only be guaranteed architecturally.

Your argument for a reduced resolution implying that it then must be mono-tonic is valid in some cases. As an example if you take the MSB of any DAC and toggle it, every DAC is monotonic. Moving from 14 bit to 13 bit isn't necessarily clear. It depends upon architecture, a Sigma-delta DAC probably would be.

Some DACs are made of cascades of different techniques so it's hard to say without the datasheet.

INL being "good enough" depends upon you application.

  • \$\begingroup\$ Thanks for the answer. I agree with you but I have some queries: 1. DNL is the measure of step increase in output subtracted by expected increase in output. Hence, DNL= a(n+1)-a(n)-1 LSBs. For the DAC to be non-monotonic, DNL should be below -1LSBs. Hence, does it not mean that low DNL is monotonic? 2. The next question is that if a DAC which is 14 bit, has a DNL of +-2LSBs. Is it safe to say that if we use the same DAC as a 13 bit instead, LSB will of course increase 2 times. Hence, the new DNL is +-1LSBs. Is it now monotonic? \$\endgroup\$ Jan 11 '13 at 3:03
  • \$\begingroup\$ Monotonic should have low DNL, Low DNL does not necessarily mean Monotonic because it is an average measure. Imagine a ideal residual error chart that has a step of +2, then a step of -3 and then step of +4 to bring it back to normal. That is clearly non-monotonic but the DNL will only shift a little because of all the other samples in over the span. \$\endgroup\$ Jan 11 '13 at 3:30
  • \$\begingroup\$ Yes, completely agreed, thanks! So, usually you would specify the DNL as a single no. for the entire DAC string. However, in reality it can fluctuate a lot from code to code, some of which can be non-monotonic. \$\endgroup\$ Jan 15 '13 at 10:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.