Hello all,
I would like to hear how some FFT-based convolution sounds with VASP, which claims to leave fewer artifacts than the hann-windowed realtime FFT, but the included example "convolution" does not produce results for me - I load the sound files and send them to the tables via wedit but the final table (I think it's called buf_res) does not receive anything - in other words, no results.
I'm looking to do something like the "timbre stamp" FFT example in Pd's documentation - I've gotten so many wild and beautiful sounds from running tabread4~ based samples against each other through this, but the quality is not always very crisp. If anyone has any patches or info, let me know.
Thanks! Kevin
hi kevin, this could also be a wedit problem, I did not try it with the newest versions of pd. but it might also be a vasp problem, which got a little bit out of my sight during the last months. but you are right, the quality might be better, just because you use only two static soundfiles and don't have to worry about windowing and joining fftslices. could you just send your patch? maye offlist. marius.
Kevin McCoy wrote:
Hello all,
I would like to hear how some FFT-based convolution sounds with VASP, which claims to leave fewer artifacts than the hann-windowed realtime FFT, but the included example "convolution" does not produce results for me - I load the sound files and send them to the tables via wedit but the final table (I think it's called buf_res) does not receive anything
- in other words, no results.
I'm looking to do something like the "timbre stamp" FFT example in Pd's documentation - I've gotten so many wild and beautiful sounds from running tabread4~ based samples against each other through this, but the quality is not always very crisp. If anyone has any patches or info, let me know.
Thanks! Kevin
--
++++ http://pocketkm.blogspot.com
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Hi Kevin,
sorry for not anwering earlier - i'm abroad an not very netty at the
moment.
VASP offers non-granular Fourier transformation, which enables non-
windowed convolution.
Convolution translates to multiplication in the spectral domain,
therefore just fourier-transform your two buffers (zero-padded to
the same size) into the spectral domain, multiply the results,
transform back, there you go.
greetings, Thomas
Am 02.02.2007 um 19:42 schrieb Kevin McCoy:
Hello all,
I would like to hear how some FFT-based convolution sounds with
VASP, which claims to leave fewer artifacts than the hann-windowed
realtime FFT, but the included example "convolution" does not
produce results for me - I load the sound files and send them to
the tables via wedit but the final table (I think it's called
buf_res) does not receive anything - in other words, no results.I'm looking to do something like the "timbre stamp" FFT example in
Pd's documentation - I've gotten so many wild and beautiful sounds
from running tabread4~ based samples against each other through
this, but the quality is not always very crisp. If anyone has any
patches or info, let me know.Thanks! Kevin
--
++++ http://pocketkm.blogspot.com _______________________________________________ PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/ listinfo/pd-list
Wow, didn't know it was that simple. I'll give it a shot later and report back. Thanks!
K
On 2/3/07, Thomas Grill gr@grrrr.org wrote:
Hi Kevin,sorry for not anwering earlier - i'm abroad an not very netty at the moment. VASP offers non-granular Fourier transformation, which enables non-windowed convolution. Convolution translates to multiplication in the spectral domain, therefore just fourier-transform your two buffers (zero-padded to the same size) into the spectral domain, multiply the results, transform back, there you go.
greetings, Thomas
Am 02.02.2007 um 19:42 schrieb Kevin McCoy:
Hello all,
I would like to hear how some FFT-based convolution sounds with VASP, which claims to leave fewer artifacts than the hann-windowed realtime FFT, but the included example "convolution" does not produce results for me - I load the sound files and send them to the tables via wedit but the final table (I think it's called buf_res) does not receive anything - in other words, no results.
I'm looking to do something like the "timbre stamp" FFT example in Pd's documentation - I've gotten so many wild and beautiful sounds from running tabread4~ based samples against each other through this, but the quality is not always very crisp. If anyone has any patches or info, let me know.
Thanks! Kevin
--
++++ http://pocketkm.blogspot.com _______________________________________________ PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list