[sdiy] Generating a large number of CV outputs

[David Kantowitz]

The plot with 4 segments is what happens when the single, constant input
value is processed for 1024 time steps (ie. I did not copy & paste
to graphically put together a longer image).  To me, it looks like 2nd
order converter takes 512 time steps to get through its entire cycle.

Frequency domain analysis seems necessary to understand 2nd order
and beyond.  Especially when  you consider 2nd order isn't good enough
to make commercial sigma-delta DACs.

The time domain plots do make a few things clearer:
  a) 1st order is better (for audio) than PWM and reverse-PWM and is as
easy to implement.
  b) 2nd order is much better than 1st, but requires(?) software or fpga.
  c) Things get really crappy at the highest and lowest input values.

For a practical converter, as Tom wrote, you really need to not use the
top/bottom ~10% and rescale.  Also, I read that a 2nd order sigma-delta
has some pathological signal input cases where the feedback goes
unstable at the high and low values ... which is also solved by not using
some % of the highest and lowest values.

Which left me not sure how exactly to create a useful spectral view.

Someone else asked about errors from edge shape & jitter.  I'm guessing
jitter doesn't matter much at the expected oversampling rates of 64 to 256.
Especially if that jitter is random.  Edge asymmetry, while barely a problem
for PWM, is a significant source of error for the noise shaping approaches.
Schlappi alluded to this when he described the analog section for his fpga
based converters in his 3 body module.



On Tue, Dec 12, 2023 at 2:20 AM <rburnett at richieburnett.co.uk> wrote:

> Fascinating!  The reversed PWM one still looks very regular, which
> should not come as a surprise.  It would be interesting to see the
> resulting spectra for these different techniques too.  The Delta-Sigma
> is obviously better at "shaping" the switching noise up to high
> frequencies than the PWM based methods.
>
> Does the 2nd order delta-sigma pattern still repeat after 256 bits?  Or
> did you just show the first 256 bits of a longer sequence?
>
> -Richie,
>
>
> On 2023-12-12 05:32, David Kantowitz via Synth-diy wrote:
> > ... and 2nd-order sigma delta.  I might have the algorithm
> ________________________________________________________
> This is the Synth-diy mailing list
> Submit email to: Synth-diy at synth-diy.org
> View archive at: https://synth-diy.org/pipermail/synth-diy/
> Check your settings at: https://synth-diy.org/mailman/listinfo/synth-diy
> Selling or trading? Use marketplace at synth-diy.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20231212/df003303/attachment.htm>