Uploaded November 2012 | Updated September 2026, 1 day ago
Skip to about 1:45 if you're impatient; it picks up at 2:00.
Unlike my last few algorithmic videos, this is 12EDO, with more constraints applied to the scale/chord networks to make the result more musical. The sounds and mix also respond to the scales that are being used. You can hear it change at about 4:40.
I'm testing a few synthesis techniques here that seem to be novel. The first is a way avoid or reduce aliasing by limiting frequencies to integer divisions of the sample rate. This introduces some tuning error, which can be corrected by either slowly switching the frequency between the nearest two points (vibrato), or rapidly modulating it in a random fashion (noise shaping). I find that the artifacts that result are much less objectionable than typical aliasing. I intended this to be used with high sample rates (384 or 768 kHz), but here I'm doing it at 44.1 kHz, and deliberately using a lot of high pitches and harmonics as a sort of "worst case" test. The artifacts are definitely audible, but I don't find them unpleasant.
The second technique is a sort of phase distortion that can smoothly interpolate between an arbitrary waveform and a sine. It's pretty simple, based on the identity f(x) = sin(arcsin(f(x))), but I don't know of any previous distortion synthesis algorithm that works this way. It makes a nice alternative to crossfading for wavetable or vector synthesis, and permits interesting "overmodulation" effects. Basically, it "morphs" between waveforms by distorting the shape.
I'm using both of these techniques for all sounds (except drums). I'll try to give more details later in a more appropriate place (i.e. blog post), but this is all I have for now.
For anyone interested, I posted the PD abstractions in this thread: forum.pdpatchrepo.info/topic/6759/new-anti-aliasing-and-phase-distortion-abstractions
Skip to about 1:45 if you're impatient; it picks up at 2:00.
Unlike my last few algorithmic videos, this is 12EDO, with more constraints applied to the scale/chord networks to make the result more musical. The sounds and mix also respond to the scales that are being used. You can hear it change at about 4:40.
I'm testing a few synthesis techniques here that seem to be novel. The first is a way avoid or reduce aliasing by limiting frequencies to integer divisions of the sample rate. This introduces some tuning error, which can be corrected by either slowly switching the frequency between the nearest two points (vibrato), or rapidly modulating it in a random fashion (noise shaping). I find that the artifacts that result are much less objectionable than typical aliasing. I intended this to be used with high sample rates (384 or 768 kHz), but here I'm doing it at 44.1 kHz, and deliberately using a lot of high pitches and harmonics as a sort of "worst case" test. The artifacts are definitely audible, but I don't find them unpleasant.
The second technique is a sort of phase distortion that can smoothly interpolate between an arbitrary waveform and a sine. It's pretty simple, based on the identity f(x) = sin(arcsin(f(x))), but I don't know of any previous distortion synthesis algorithm that works this way. It makes a nice alternative to crossfading for wavetable or vector synthesis, and permits interesting "overmodulation" effects. Basically, it "morphs" between waveforms by distorting the shape.
I'm using both of these techniques for all sounds (except drums). I'll try to give more details later in a more appropriate place (i.e. blog post), but this is all I have for now.
For anyone interested, I posted the PD abstractions in this thread: forum.pdpatchrepo.info/topic/6759/new-anti-aliasing-and-phase-distortion-abstractions
