Author |
Message |
astroid power-up!
Joined: Mar 23, 2004 Posts: 334
G2 patch files: 15
|
Posted: Mon Feb 25, 2008 2:34 am Post subject:
fm/formant resynthesis question for the geniuses |
|
|
rekindling my gas for an fs1r, i stumbled across this:
http://www.xs4all.nl/~niff/fs1r/fseqedit/help/
a formant analysis/resynthesis program. like a good g2 owner i got excited and scurried over here to try to beg some of you geniuses to, uh, basically figure it out for me. lol.
would it be possible to do this kind of thing in real time on the g2?
i mean, i guess it's basically the same thing as a vocoder and all, but any insight would be wonderful. _________________ Astroid Power-Up!: "googleplex" available at:
http://cdbaby.com/cd/googleplex |
|
Back to top
|
|
|
astroid power-up!
Joined: Mar 23, 2004 Posts: 334
G2 patch files: 15
|
Posted: Mon Feb 25, 2008 3:03 am Post subject:
|
|
|
here's a laughable idea, maybe someone can help me make some sense out of it:
the sound goes in, and a pitch tracker follows it.
that tracker plays a sine wave that gets added perfectly out of phase to the original (effectively cancelling out that partial).
that signal is sent to another pitch tracker, which does the same thing with the new signal.
and so on, until you have a list of the formants, in order of their strength.
so why isn't that gonna work? _________________ Astroid Power-Up!: "googleplex" available at:
http://cdbaby.com/cd/googleplex |
|
Back to top
|
|
|
dorremifasol
Joined: Sep 28, 2006 Posts: 814 Location: Barcelona, Spain
Audio files: 7
G2 patch files: 49
|
|
Back to top
|
|
|
Falk
Joined: Mar 07, 2006 Posts: 75 Location: Sweden
G2 patch files: 29
|
Posted: Mon Feb 25, 2008 8:53 am Post subject:
|
|
|
Hi Astroid.
astroid power-up! wrote: | here's a laughable idea, maybe someone can help me make some sense out of it:
the sound goes in, and a pitch tracker follows it.
that tracker plays a sine wave that gets added perfectly out of phase to the original (effectively cancelling out that partial).
that signal is sent to another pitch tracker, which does the same thing with the new signal.
and so on, until you have a list of the formants, in order of their strength.
so why isn't that gonna work? |
There are 2 things wrong with it I am afraid. The pitch tracker (even if it works correctly with complex waveforms) will neither give you the phase nor the amplitude of the partial.....
Sorry.
/Falk |
|
Back to top
|
|
|
astroid power-up!
Joined: Mar 23, 2004 Posts: 334
G2 patch files: 15
|
Posted: Mon Feb 25, 2008 10:27 am Post subject:
|
|
|
what if, instead of setting a sine wave back in to cancel out the partial, it were just notched out with a filter? it'd maybe get the first 8 partials before it crapped out.
hm
i'm gonna make something stupid.
that g2ian patch is great:) _________________ Astroid Power-Up!: "googleplex" available at:
http://cdbaby.com/cd/googleplex |
|
Back to top
|
|
|
Falk
Joined: Mar 07, 2006 Posts: 75 Location: Sweden
G2 patch files: 29
|
|
Back to top
|
|
|
ian-s
Joined: Apr 01, 2004 Posts: 2669 Location: Auckland, New Zealand
Audio files: 42
G2 patch files: 626
|
Posted: Mon Feb 25, 2008 1:37 pm Post subject:
Re: fm/formant resynthesis question for the geniuses |
|
|
astroid power-up! wrote: | http://www.xs4all.nl/~niff/fs1r/fseqedit/help/
a formant analysis/resynthesis program. like a good g2 owner i got excited and scurried over here to try to beg some of you geniuses to, uh, basically figure it out for me. lol.
would it be possible to do this kind of thing in real time on the g2?
|
Looks like a good utility. If someone had access to FS1R sysex, they might be able to do a G2patch + G2ool.py to transfer the formant info to G2 level sequencers.
I tried something similar using praat as the analyser, but the results were not as good as I had hoped.
Alternativly, the whole sample analysis could be done in python to provide a stand alone g2ool utility. Last edited by ian-s on Tue Feb 26, 2008 11:38 am; edited 1 time in total |
|
Back to top
|
|
|
astroid power-up!
Joined: Mar 23, 2004 Posts: 334
G2 patch files: 15
|
Posted: Mon Feb 25, 2008 5:47 pm Post subject:
|
|
|
i guess the thing i'm most curious about is if the g2 could somehow be used as a rudimentary analysis tool. _________________ Astroid Power-Up!: "googleplex" available at:
http://cdbaby.com/cd/googleplex |
|
Back to top
|
|
|
elhardt
Joined: May 14, 2005 Posts: 73 Location: USA
|
Posted: Tue Feb 26, 2008 11:32 am Post subject:
|
|
|
astroid power-up! wrote: | what if, instead of setting a sine wave back in to cancel out the partial, it were just notched out with a filter? it'd maybe get the first 8 partials before it crapped out. |
You don't know how much (as in amplitude) of a sinewave to subtract. If you did, you wouldn't need to analyze the incoming waveform to begin with. And what would be the point of subtracting it out from the original waveform anyway? Plus, waveforms from real acoustic sounds are complex and always changing.
As for using the Nord to analyze sounds, I can't really see how. In Reaktor you can somewhat patch up a crude frequency band analyzer, but it won't do you much good. You'll need a computer to look at the harmonic content and use that info to manually create a patch on the Nord. |
|
Back to top
|
|
|
astroid power-up!
Joined: Mar 23, 2004 Posts: 334
G2 patch files: 15
|
Posted: Tue Feb 26, 2008 1:39 pm Post subject:
|
|
|
elhardt wrote: | astroid power-up! wrote: | what if, instead of setting a sine wave back in to cancel out the partial, it were just notched out with a filter? it'd maybe get the first 8 partials before it crapped out. |
You don't know how much (as in amplitude) of a sinewave to subtract. If you did, you wouldn't need to analyze the incoming waveform to begin with. And what would be the point of subtracting it out from the original waveform anyway? Plus, waveforms from real acoustic sounds are complex and always changing.
As for using the Nord to analyze sounds, I can't really see how. In Reaktor you can somewhat patch up a crude frequency band analyzer, but it won't do you much good. You'll need a computer to look at the harmonic content and use that info to manually create a patch on the Nord. |
it's clear to me now that my first idea wouldn't work. i tried my second idea, of notching out the bands, and that didn't work too well either. the point of subtracting the partials/bands out is so that the analysis could be done in stages (get rid of the strongest one first, etc.). it's not some ideal solution, just something i wanted to see if i could hack together for the hell of it.
where i'm coming from is more of an abstract sound design standpoint. i'm not trying to make a realistic guitar sound per se, but think it would be lovely to have the strongest 8 partials of a guitar or any realtime signal splayed out in additive form for mangling. autechre/richard devine over wendy carlos in this regard.
i'd take 'realtime' over 'perfect copy' any day.
i'll still keep hacking away until 1 of 3 things happen.
1. i get a kyma or hartmann
2. vst makers make a resynthesis plug with more of the philosophy of the synths in 1.
3. they update the g2 to have spectral tools _________________ Astroid Power-Up!: "googleplex" available at:
http://cdbaby.com/cd/googleplex |
|
Back to top
|
|
|
jksuperstar
Joined: Aug 20, 2004 Posts: 2503 Location: Denver
Audio files: 1
G2 patch files: 18
|
Posted: Tue Feb 26, 2008 2:59 pm Post subject:
|
|
|
astroid power-up! wrote: | i guess the thing i'm most curious about is if the g2 could somehow be used as a rudimentary analysis tool. |
Unfortunately, this is (IMHO) where the G2 is weakest, regarding alternative synthesis techniques (ie- not subtractive). The analysis options are limited at best. |
|
Back to top
|
|
|
elhardt
Joined: May 14, 2005 Posts: 73 Location: USA
|
Posted: Tue Feb 26, 2008 3:25 pm Post subject:
|
|
|
astroid power-up! wrote: |
i'll still keep hacking away until 1 of 3 things happen.
1. i get a kyma or hartmann
2. vst makers make a resynthesis plug with more of the philosophy of the synths in 1.
3. they update the g2 to have spectral tools |
Since you mention you'd be satisfied with only the first 8 harmonics or so, that gave me an idea. If you want to analyze a typical pitched sound that has the natural harmonic series and you have control over the pitch of the sound to be analyzed, then the following might work. If for example decide you want to always analyze sounds at a pitch of 1000 Hz, then you know that the harmonics will be 2000, 3000, 4000 Hz and so on, spaced 1000 Hz apart. You could set up what would be like the front part or analysis part of a vocoder with the bandpass filters set to freqs of 1000, 2000, 3000 up to 8000 Hz. Then in place of the synthesis part of a vocoder, which is usually another set of bandpass filters and VCAs, you instead use 8 VCAs, each which controls the level of a single sinewave osc, then you might be able to pull it off somewhat. Keep in mind you would either need the audio input signal to keep remaining on the input all the time to keep the VCAs open at their specified amounts, or you would have to come up with a way to freeze the VCAs at their levels kind of the way some vocoders have a freeze button that holds the vowel sound in the VCAs for the bandpass filters. You'd probably have to use 8 Sample & Hold modules for that. The sinewave oscs of course could be controlled from the keyboard so you can play your creation musically if that's what you intend.
Okay, so get to it. |
|
Back to top
|
|
|
astroid power-up!
Joined: Mar 23, 2004 Posts: 334
G2 patch files: 15
|
Posted: Tue Feb 26, 2008 3:50 pm Post subject:
|
|
|
elhardt wrote: | Keep in mind you would either need the audio input signal to keep remaining on the input all the time to keep the VCAs open at their specified amounts, or you would have to come up with a way to freeze the VCAs at their levels kind of the way some vocoders have a freeze button that holds the vowel sound in the VCAs for the bandpass filters. |
right, or some kind of noise gate i guess. maybe track and hold...
that's interesting. maybe i could combine this vocoder idea with the pitch tracking (if i can work out how not to have it so glitchy), so the fundamental (or at least something) is being followed by the filterbank. i'm gonna see how many bands i could get out of this. probably more like 32 if it were splayed out through the buses.
thanks for the brainstorming, that vca linking idea is a good one. _________________ Astroid Power-Up!: "googleplex" available at:
http://cdbaby.com/cd/googleplex |
|
Back to top
|
|
|
astroid power-up!
Joined: Mar 23, 2004 Posts: 334
G2 patch files: 15
|
Posted: Thu Feb 28, 2008 3:18 pm Post subject:
|
|
|
i tried the vocoder/follower hybrid idea out. it sounds pretty much like you'd expect it to sound-a glitching vocoder. lol.
the vocoder/osc part alone, without pitch following, sounds kinda like you'd guess too, a chord that just kinda feebly tries to adjust tone color along with the audio.
it's actually kind of a fun/controllable tone coloring device if you attenuate the poop out of it, but it's not anywhere near "useful" yet. i may take the vca idea that ken added and head back to my original thing to see if that gets any closer. _________________ Astroid Power-Up!: "googleplex" available at:
http://cdbaby.com/cd/googleplex |
|
Back to top
|
|
|
|