Howdy, trying to figure the relationship between constant-Q and FFT, and if you can make a constant-Q patch in vanilla based on [fft~]. This would be for an android app, so it'd be good if I didn't need an external.
I was checking the [constantq~] external, which is found here https://github.com/grrrr/constantq and in 'deken'. it's based on a paper https://www.google.com.br/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwi7lI6svP_UAhUFW5AKHZ_nCO4QFggoMAA&url=http%3A%2F%2Facademics.wellesley.edu%2FPhysics%2Fbrown%2Fpubs%2FeffalgV92P2698-P2701.pdf&usg=AFQjCNHM8mdNm91QW_NOdtCnkap5wOgY5A by Puckette and Brown. I need help cause I can't read math or figure out code reasonably well, so I don't really know how it works. It seems it could be done just by rescaling the output of an FFT with logarithmic values, but I could be terribly wrong.
Thing is that I already have a pd patch with a logarithmic output here https://github.com/porres/LiveElectronicsTutorial/blob/master/Parte%205%20-%20FFT%20e%20Sinais%20Complexos/25.FFT-IFFT_Analise-Ressintese/2.FFT/6.Log_Freq.pd
this is not vanilla, but made on extended, but easily adaptable for vanilla (I'm also attaching it).
Also, how about just having a bank of filters with center frequencies around the tempered scale, could you have in practice the same thing as from [constantq~]?
thanks
Hi, apart from true CQ Gabor transformations there is a method for estimating the power spectra of arbitrary frequency scales. It's probably what you are looking for. It is based on summing up power spectrum bins by use of triangular windows. I made an implementation which is fairly general. It employs some dynamic patching to generate the individual filter bands. spect_scaled.pd is the main abstraction, it comes with a demo file spect_scaled-help.pd
have fun, Thomas
-- Thomas Grill http://grrrr.org
Am 10.07.2017 um 22:01 schrieb Alexandre Torres Porres porres@gmail.com:
Howdy, trying to figure the relationship between constant-Q and FFT, and if you can make a constant-Q patch in vanilla based on [fft~]. This would be for an android app, so it'd be good if I didn't need an external.
I was checking the [constantq~] external, which is found here and in 'deken'. it's based on a paper by Puckette and Brown. I need help cause I can't read math or figure out code reasonably well, so I don't really know how it works. It seems it could be done just by rescaling the output of an FFT with logarithmic values, but I could be terribly wrong.
Thing is that I already have a pd patch with a logarithmic output here - this is not vanilla, but made on extended, but easily adaptable for vanilla (I'm also attaching it).
Also, how about just having a bank of filters with center frequencies around the tempered scale, could you have in practice the same thing as from [constantq~]?
thanks <6.Log_Freq.pd>_______________________________________________ Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
Btw., just committed the solution to github, additionally implementing the mel frequency scale. https://github.com/grrrr/spect_scaled
Looking forward to contributions. best, Thomas
-- Thomas Grill http://grrrr.org
Am 13.07.2017 um 02:00 schrieb Thomas Grill gr@grrrr.org:
Hi, apart from true CQ Gabor transformations there is a method for estimating the power spectra of arbitrary frequency scales. It's probably what you are looking for. It is based on summing up power spectrum bins by use of triangular windows. I made an implementation which is fairly general. It employs some dynamic patching to generate the individual filter bands. spect_scaled.pd is the main abstraction, it comes with a demo file spect_scaled-help.pd
have fun, Thomas
-- Thomas Grill http://grrrr.org
<hz-log.pd><list-map.pd><log-hz.pd><scl2bin.pd><spect_part.pd><spect_parts.pd><spect_scaled-help.pd><spect_scaled.pd>
Am 10.07.2017 um 22:01 schrieb Alexandre Torres Porres porres@gmail.com:
Howdy, trying to figure the relationship between constant-Q and FFT, and if you can make a constant-Q patch in vanilla based on [fft~]. This would be for an android app, so it'd be good if I didn't need an external.
I was checking the [constantq~] external, which is found here and in 'deken'. it's based on a paper by Puckette and Brown. I need help cause I can't read math or figure out code reasonably well, so I don't really know how it works. It seems it could be done just by rescaling the output of an FFT with logarithmic values, but I could be terribly wrong.
Thing is that I already have a pd patch with a logarithmic output here - this is not vanilla, but made on extended, but easily adaptable for vanilla (I'm also attaching it).
Also, how about just having a bank of filters with center frequencies around the tempered scale, could you have in practice the same thing as from [constantq~]?
thanks <6.Log_Freq.pd>_______________________________________________ Pd-list@lists.iem.at mailing list UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
-
Alexandre Torres Porres -
Thomas Grill