[Fluxus] puzzle

[Kassen]

2010/1/20 evan.raskob [lists] <lists at lowfrequency.org>

> I like the idea of programmable bands.
>

Me too. With that and perhaps a new method for tracking the most dominant
frequency we could engage more specialised program material (say a didgerido
or a piccolo) in more meaningful ways than the current features can.


> Anything below 120 Hz is pretty much noise - if your lucky, your
> (professional) mic might pick up 80Hz at lowest.
>
>
I'm a bit reluctant to agree here. In many cases you are right but it's not
unusual to have the kick drum of a dance music track at around 80Hz (though
of course a lot of harmonics would be used). For well-produced pieces (where
care was taken to low-cut any channels that have no business mucking up the
low-end) that range will contain meaningful information. I'd anticipate that
cutting below 120Hz and having the centre of the lowest band at -say- 160 to
200Hz would lead to less than satisfactory results when working with styles
like dub. producers like Lee Perry make it quite clear (IMHO) that you can
have musically meaningful material below that range.

Many club systems give up below around 100Hz, but that's not entirely a
matter of lower frequencies being noise but more one of the high cost in
energy needed by the amps to play those back. Because of psychoacoustics the
audience will still experience the fundamental if it's below this so to me
it would make sense to still attempt to track frequencies like that to make
the visuals more closely aligned with how the sound is experienced. This is
-again- assuming well-produced material.

We could of course set the lower end before we set the number of desired
bands. Somebody working with a harpsichord and live mics will have different
needs there from one doing visuals for a dubstep performance, while aside
from that both would likely want to consider the full spectrum in a
evenly divided way.

In my experience anything related to audio will mean compromises and trades
so I'd recommend going with a solution that will set sensible defaults, then
allow the user to make such trades in a effective way.



> I looked into some audio methods awhile back - averaging fft bins, getting
> spectral center frequency, and other interesting music-influenced frequency
> transforms (check out MeapSoft for some open source tools done in Java) but
> lost momentum.  It would be nice to have some audio filtering in fluxus (or
> scheme) so we can do more advanced things with audio... of course, we can
> just use the output of (ga) now and roll our own each time or create a
> scheme module of functions.
>
>
I agree. Small scheme scripts could serve as presets and could be traded
over the list. I experimented a bit with the fft analysis in ChucK and found
that -for me- merely taking a fft is borderline meaningless. One needs to
take it in a way that makes sense in the given context (with regard to
window-size, overlap, etc), then the data needs to be interpreted in some
way. Trains of numbers are just trains of numbers.

That's not to say the current system doesn't work quite well; it's lots of
fun, but if we want to get more advanced and use Fluxus in more contexts we
may need a bit of a plan and a strategy here.

Yours,
Kas.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.pawfal.org/pipermail/fluxus-pawfal.org/attachments/20100120/2d700346/attachment.html>