On May 22, 9:10 pm, Tim Wescott <t...@seemywebsite.com> wrote:
> somdeb wrote:
> > I am trying to understand how bit-width requirements at different
> > integrators can be estimated... couldn't find much in literature.
...
>
> As a guess, I'd say it has to be big enough so that, at a minimum, the
> quantization noise from your integrator is insignificant compared to the
> noise you expect to be getting from the modulator.

my guess is that quantization in the integrator is not the issue, but
saturation.

the first integrator is an up-down counter.  if you assume your input
is not pinned against the rails, how high (or low) does that up-down
counter get?  the base-2 log of that minus 1 is the bit width you
need.  the second integrator sums the output of the first and we ask
the same question regarding it.

that's my amateurish spin on it.

r b-j