# Hi, # Recently I heard the name of Gabor, a phyisict who had applied the idea of quanta in acoustics. I didn't read any of his papers but as I understand what I read about him, he tried to set the mathematical basis of producing any sound from "tiny sound particles". And today this idea is called "granular synthesis". There is a book about it "Microsound" by Curtis Roads. I will look at the book and investigate the idea further. # Here, I'm attaching some patches about the subject which I studied this morning and want to share with you :-). there is a unit generator which produces sinusoidal oscillations enveloped with gaussian distribution curve. (kleineWelle.pd) The output is a small wave packet. (you can also look at kleineWelle-help.pd) metroWelle, produces periodic wave packets. [metroWelle length_of_wave_packet freq_of_osc period] granulator.pd is the main patch and the area for free study. You can add more metroWelle's and listen to their sum. # This is a very primitive kind of granular synthesis. As I now realized granular synthesis needs enormous control parameters. Preparing every wave packet by hand is very hard. # Have a nice day! -ugur guney-
Hallo, U??ur GÃŒney hat gesagt: // U??ur GÃŒney wrote:
# Recently I heard the name of Gabor, a phyisict who had applied the idea of quanta in acoustics. I didn't read any of his papers but as I understand what I read about him, he tried to set the mathematical basis of producing any sound from "tiny sound particles". And today this idea is called "granular synthesis". There is a book about it "Microsound" by Curtis Roads. I will look at the book and investigate the idea further. # Here, I'm attaching some patches about the subject which I studied this morning and want to share with you :-). there is a unit generator which produces sinusoidal oscillations enveloped with gaussian distribution curve. (kleineWelle.pd) The output is a small wave packet. (you can also look at kleineWelle-help.pd) metroWelle, produces periodic wave packets. [metroWelle length_of_wave_packet freq_of_osc period] granulator.pd is the main patch and the area for free study. You can add more metroWelle's and listen to their sum.
Some hints: For granular synthesis you definitely want to avoid line~ and use vline~ instead. Alternative approaches could be made with phasor~ and samphold~. The Vosim patch I posted recently is also a granular technique. In Vosim, the waveform is a squared sin, the envelope is a exponentiall falling step function for each sin^2 peak, but you could change that of course to produce different waveforms and envelopes, just replace the cos~ stuff with tabread4~s and scale the phasors accordingly. (tabread4~ would be faster in your patch as a replacmenent for expr~, too)
And yes, you need many copies of each, so something like [polypoly~] or [nqpoly4] will help.
Ciao
Frank Barknecht _ ______footils.org__
# Thanks for the hints. After I changed expr~ with tabread~ the cpu consumption decreased dramatically. # As a I understand the Vosim patch, sinusoidal harmonics model of sound is not useful for speech. It is very interesting that repeating N successive decaying sin^2 peaks sounds like a vowel. # In my PD distro (0.39.2-extended-test7) [polypoly~], [nqpoly4] does not exist. So I can't use them. # The attachment includes improved version of the study patch. Main patch is granuler03.pd. You can open it and listen to the sound of cicadas in the middle of night :-) -ugur-
On Dec 15, 2007 1:08 PM, Frank Barknecht fbar@footils.org wrote:
Some hints: For granular synthesis you definitely want to avoid line~ and use vline~ instead. Alternative approaches could be made with phasor~ and samphold~. The Vosim patch I posted recently is also a granular technique. In Vosim, the waveform is a squared sin, the envelope is a exponentiall falling step function for each sin^2 peak, but you could change that of course to produce different waveforms and envelopes, just replace the cos~ stuff with tabread4~s and scale the phasors accordingly. (tabread4~ would be faster in your patch as a replacmenent for expr~, too)
And yes, you need many copies of each, so something like [polypoly~] or [nqpoly4] will help.
Ciao
Frank Barknecht _ ______footils.org__
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Hallo, U??ur GÃŒney hat gesagt: // U??ur GÃŒney wrote:
# Thanks for the hints. After I changed expr~ with tabread~ the cpu consumption decreased dramatically. # As a I understand the Vosim patch, sinusoidal harmonics model of sound is not useful for speech. It is very interesting that repeating N successive decaying sin^2 peaks sounds like a vowel.
Well, Vosim was developed for speech, so I guess at least it was useful for that, but personally I didn't really play that much with it. A single vosim generator as in my patch will exhibit a signal with one strong formant. Combine three of them, and you can do some almost nice vowel sounds by putting the formants at the typical frequencies for a, e, i ...
Although vosim is a granular technique, it's purpose is mainly in the realm of what Miller calls "designer spectra" in his book. There he also describes some more flexible techniques.
# In my PD distro (0.39.2-extended-test7) [polypoly~], [nqpoly4] does not exist. So I can't use them.
They are just abstractions, you could get them from CVS and then use them. Both really help with creating and managing many copies of an abstraction.
Ciao
Frank Barknecht _ ______footils.org__
is it just about getting rand. varieble amplitude sinusoid ?
i have made a nice object called 0sc~ - which is like an unstable oscillator (amplitude wise, and frequency wise if Q is small) it is just a cascade of vcf~ with noise~ (and pink~) put trough that
mine isn't making sound similar to yours ..i'm not sure does it fall in the granular concept ?
On Dec 16, 2007 1:02 AM, errordeveloper@gmail.com wrote:
is it just about getting rand. varieble amplitude sinusoid ?
i have made a nice object called 0sc~ - which is like an unstable oscillator (amplitude wise, and frequency wise if Q is small) it is just a cascade of vcf~ with noise~ (and pink~) put trough that
# Granularity comes from the fact that the overall sound is the sum of short sound segments (which are 5 to 30 msec long). To avoid glitches due to discontinuity these segments are calculated by enveloping a sinusoidal with a gaussian. # In my patch I prepared some sound segments with definite duration and frequency and triggered them with regular time intervals.
mine isn't making sound similar to yours ..i'm not sure does it fall in
the granular concept ?
# In this view, I think your object works in continuous signal domain. -ugur-
-
errordeveloper＠gmail.com -
Frank Barknecht -
Uğur Güney