back to list

spectral mapping techniques

🔗sethares@...

10/31/2001 5:10:21 PM

>if you have a MIDI file at hand, I'd be glad to play with it

I'll happily send a copy... can I email it to you as an attachment?

>I'd much rather pick your brain regarding preparing altered timbres for
>one or more complete GM voice sets.

This is a huge undertaking, though I understand why it would be a good
thing. Several years ago I made up a large number of sounds (not quite
a whole GM set, but about 50 instruments) with the spectra mapped to
10-tet. I actually offered it for sale for a while through "Chicken
Systems" software, and 2 people bought it, plus I gave copies to about
6 or 7 others. I'd be happy to distribute it (give it away) again.
But (there's always a "but") its formatted for the ASR-10 sampler... so you'd
need the sampler in order to make use of it. Since Ensoniq
(the maker of the ASR-10) has been bought out, they no longer make
samplers, so its kind of a dead machine. Im toying with the
idea of eventually moving the samples over to another format,
but this is itself a major job...

>It seems as if a fair amount of
>tinkering with the overtones is almost inaudible to the ear

Well, yes and no. Certainly small deviations are no problem,
but when the deviations get large there are (at least) three things
that can happen...

(1) the pitch can change
(2) the note can split, essentially sounding like two notes
(3) you can hear artifacts of the manipulation process

Assuming none of these occur, then you can really get
"new" consonances at unusual intervals.

>How hard would it be for a guy like me to acquire and use the tools that
>would be required for such a project?

I dont think theres any "easy" way. Right now I see three possiblities...

(1) use Csound
(2) use Metasynth
(3) roll your own (this was my approach several years ago because (1) and (2)
were not available.) Actually, Csound existed, but the transform modules
did not function.

Id be happy to share my code, but as I warned before, its in Matlab
and is absolutely not user friendly.

Hope this helps,

Bill Sethares

🔗John A. deLaubenfels <jdl@...>

11/2/2001 4:56:54 AM

[I wrote:]
>>if you have a MIDI file at hand, I'd be glad to play with it

[Bill S wrote:]
>I'll happily send a copy... can I email it to you as an attachment?

Kyool! Sure, send to jdl"at"adaptune.com .

[JdL:]
>>I'd much rather pick your brain regarding preparing altered timbres
>>for one or more complete GM voice sets.

[Bill:]
>This is a huge undertaking, though I understand why it would be a good
>thing.

Dang, I was afraid you were going to say that! Maybe just the piano for
starters would be a more manageable, and very interesting, goal.

>Several years ago I made up a large number of sounds (not quite
>a whole GM set, but about 50 instruments) with the spectra mapped to
>10-tet.

Which has much wilder partial alignment than, say, 12-tET, yes?

>I actually offered it for sale for a while through "Chicken
>Systems" software, and 2 people bought it, plus I gave copies to about
>6 or 7 others. I'd be happy to distribute it (give it away) again.
>But (there's always a "but") its formatted for the ASR-10 sampler... so
>you'd need the sampler in order to make use of it. Since Ensoniq
>(the maker of the ASR-10) has been bought out, they no longer make
>samplers, so its kind of a dead machine. Im toying with the
>idea of eventually moving the samples over to another format,
>but this is itself a major job...

And, in truth, 10-tET is not my big interest, but thanks for your
generous offer.

[JdL:]
>>It seems as if a fair amount of tinkering with the overtones is almost
>>inaudible to the ear

[Bill:]
>Well, yes and no. Certainly small deviations are no problem,
>but when the deviations get large there are (at least) three things
>that can happen...

>(1) the pitch can change
>(2) the note can split, essentially sounding like two notes
>(3) you can hear artifacts of the manipulation process

Understood...

>Assuming none of these occur, then you can really get
>"new" consonances at unusual intervals.

[JdL:]
>>How hard would it be for a guy like me to acquire and use the tools
>>that would be required for such a project?

[Bill:]
>I dont think theres any "easy" way. Right now I see three possiblities...

>(1) use Csound
>(2) use Metasynth
>(3) roll your own (this was my approach several years ago because (1)
>and (2) were not available.) Actually, Csound existed, but the
>transform modules did not function.

I've got to reveal my extreme ignorance and ask a really basic question:
which of these builds sound from the ground up, and which, if any, allow
the manipulation of actual samples from an acoustic instrument?

The former, building sounds from the ground up, is, I believe, easier,
since one can specify all the frequencies to be used. Margo's FM
synthesis uses such a method, yes? But if I've understood you
correctly, you have done the latter; that is, started with a sampled
sound, analyzed and subtracted out the highly pitched portions, leaving
the characteristic unpitched (or poorly pitched) noises that make the
instrument sound "real", shifted the pitched portions and added them
back in? The guitar in the 19-tET piece, for example; did that not
start life as an acoustic sample?

OK, yes, I just went back and read your post,

/makemicromusic/topicId_1152.html#1152

Reading this makes manipulating the partials of sampled waves sound
almost easy (using Matlab). Moving things from the time domain into
the frequency domain and back. And I gather that transients are somehow
preserved (since FFT's require integrals across time, I'm a little vague
on how this works).

>Id be happy to share my code, but as I warned before, its in Matlab
>and is absolutely not user friendly.

Sorry, what does this code do again? Remap the partials of a wave with
the push of a button?

My thanks for your help, Bill.

JdL