[chuck-users] analog to digital conversion simulation

[Kassen]

eduard;

 a Onepole filter will not
> suffice to get rid off all frequencies above the cutoff. Is there a
> way of having a kind of brickwall filter in chuck?


Oh. dear. I fear we are on the edge of a situation where some massive LiSa
abuse to use her as a FIR filter would become almost sensible (loop at
-sample rate- over a set of LiSa voices that all hold your impulse responce,
writing my_signal.last() to their gain, then starting the sample in
non-looped mode. You need at least as many voices as there are samples in
your impulse-response. For a high cutoff (and thus short IR) that might
actually even work).

Fortunately for LiSa there is very little use, IMHO, in emulating
high-quality DAC's as those sound like a straight (if filtered) signal by
definition. The digital good-stuff (Akai s-612, Roland s-550, Amiga 500,
etc) on the other hand sounds like it does due to a rather non-perfect
implementation which I suspect involves a plain analogue filter with a huge
ripple (which adds character, in retrospect and at second hand prices in a
market hardly anyone cares for....) in order to get a steeper roll off
that's placed as high (in Hz) as they dared to go, trading fidelity for
brightness. I didn't analyse the cerquit but I think it's a good guess that
for the S-612 Akai simply bolted a envelope to the same filter and presented
that as the modulated filter.

So; I'd say add rounding to 8 or 12 bits, try to make the AD and DA stages
slightly non-linear in ways that don't perfectly match and tune the filters
by ear.... But then again, you may have a very different opinion on what the
digital "Good stuff" is, I like the S-612 :¬)

Oh, and for bonus trickery; the S-550 has a digital filter that's not quite
stable at high cutoffs with modulation and particularly a high Q. I suspect
part of the blame lies in it being a 12 bit system and I suspect rounding
errors in filter coefficients act up there. This can be charming, at times.
Perhaps some of our real filter experts will be able to comment on how
likely my theory is, I know it goes semi chaotic in that range, the rounding
is my own theory. I believe rounding everything to the sample frequency
won't improve matters at such bit-rates either, and I always take it down to
15KHz sampling frequency.....

Not sure if that might help or just make it all harder, I'm a bit of a
enthusiast for low-fidelity yet nice sounding stuff.

Yours,
Kas.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.princeton.edu/pipermail/chuck-users/attachments/20071012/bc090bcb/attachment.htm