hurolura,
problem is that you can convert integer values to float values. but its difficult when it comes to converting integer arrays to float (1....0...-1) arrays.
@shroomz, I guess I understand how the waldorf oscillator works generally.
the first 64 samples are used for the lower octaves through the fact that downshifting doesn`t introduce aliasing
lets say the first 64 samples represent c3, then c2,c1 and c0 is done by downpitching the first 64 samples.
c4 is represented by the next 32 samples as pitching one octave higher should half the the wavecycle....
and so on
c5 16 samples
c6 8 samples
c7 4 samples
c8 2 samples
thats just how i image the generell workflow, I did not verify this.
the higher octaves representations may be antialiased to deliver better pitching quality.
thats why the look of the waves may change in the higher region.
so basically the first 64 samples represent the wave in its pure form, the higher regions the pitched versions.
just a wuick try to explain whats going on, but its not unlikely that I am wrong on this.
that would also explain the inferior quality of the waldorf osc in the down region, where it has no "pressure" and "depth" through a weaker downpitching.
it would be awesome if SC would build an oscillator with 2048 samples in lowest pitch (c0) and the other pitches are represented with fractions of this base picth : 1024 (c1) and so on, So you could use the high quality pitching of a wave editor to generate the octave pitch steps.....
and between the octaves another perfromance algo blend s through the picthings in a satisfieying way.
some sort of flexible and quality wave based oscillator.
just some sort of brainstorming from my side.
and similar stuff 
