Zebra's oscillators noise

[hastoy]

Hi,

I'm trying to transfer single cycle waveforms from samples to Zebra using wav2zebra. Everything works good but I have a sound quality issue. In my sound editor I compress the wav length to 128 samples which is what Zebra uses I've read. Then I compare the sound of Zebra with the sound looped in my sound editor. At mid pitch it sounds very similar but at low pitch Zebra has artifacts, as if the detuning of the osc was of poorer quality than in my sound editor. I get the kind of noise you get when you loop a sound outside of zero crossings, could this be a bug ?? Are there synths with cleaner oscillators than others ?

Cheers

Guillaume

[xh3rv]

Zebra's oscillators are pretty clean in that they should be harmonically well-structured, and very nice about aliasing. The amplitude discontinuities of non-zero-crossings seems unlikely to me - you'd be able to see it, plus I believe less nasty than in a sampler.

How low do you have to go before things go weird? I wonder if there are some effects of the fundamental dropping out of range, or high-range stuff coming into more prominent range. Dunno why this would be worse than naively re-pitching in a sampler, though. Damping high range with osc fx or otherwise makes sense sometimes, depending on what's being sampled and cooked up for Zebra.

[hastoy]

It's obvious with a sine wave, FMO sounds much cleaner than OSC in the low bass notes, it gets noisy with OSC. I had posted that earlier but now I realize it gets even worse with complex waveforms. You can test this easily in your editor by simply copy/paste many instances of a waveform to get a second or 2 of sound, detune several octaves and listen how cleaner it is than in Zebra, it's really weird :-/ As you pointed out I could use osc fx like Filter or Choplift but there are only 2 fx available so I don't want to waste those I don't know if I can post an mp3 here to illustrate my message ?

[3ee]

Depends on what mode you model the osc's wave.

The morph modes offer "full" resolution and the blend modes offers "256"? partial spectral... so it's limited...

[Urs]

Ah… there was a bug indeed where a clean sine wave would be off by a tiniest amount. If you go to our latest builds thread, build 2157 should fix this. If not, then the next build we're going to publish in a few days will (which we post exactly to test this kind of stuff)

[hastoy]

3ee, since I import a wavefrom it must be GeoBlend

Urs, thanks for your answer, I'm looking forward to test the next build