Le 2011-10-12 à 01:32:00, hardoff goes bananas a écrit :
no idea what causes the transpose/chord shift in this patch, but if you leave it for a while, it inevitably does happen.
You are greatly amplifying error accumulation coming from [phasor~]. With a samplerate of exactly 44100 and a frequency of exactly 2, an ideal [phasor~] would increase by an amount of 2/44100 at each sample, leading to a new 0 at every 22050. The truth is a bit different.
If you do it with regular pd floats (float32), the amount 2/44100 gets computed as
12466118/pow(2,15+23) instead, because that's the closest fraction that this format can express.
For reasons of error accumulation, [phasor~] uses a «double» (float64) for storing the phase, but then it stores 1/44100 as a float, an computes 2*(1/44100) as a float too, therefore a [phasor~ 2] doesn't have the exact frequency of 2.
BTW, have you tried [phasor~ 2.01] and other things like that ? At least on my computer, it's more interesting than [phasor~ 2].
______________________________
______________________________
__________
| Mathieu BOUCHARD ----- téléphone :
+1.514.383.3801 ----- Montréal, QC