I believe Oli’s patchers are using oversampling to reduce the aliasing artifacts of digital waveforms. The advantage of oversampling is that it can be used on any waveform generator, but the amount of oversampling required may vary by waveform, and by what is an acceptable noise floor vs. performance tradeoff.
There’s a couple of alternative approaches to band-limited oscillators in the gen~ examples folder, one using Discrete Summation Formula BLITs, and the other using feedback FM. Both have their strengths & weaknesses too. The feedback FM is pretty cheap, but the aliasing is reduced but still present. The DSF BLIT is more involved, but is certainly cleaner.
I started working on a minBLEP based oscillator example, but it’s not working yet.
the patch I posted is 8x oversampled. I think that a higher order DPW could produce better results without such heavy oversampling.
The paper "Alias-suppressed oscillators based on differentiated polynomial waveforms" explains more about this technique as does "Digital Sound Generation" from ICST
- would be interested to see the gen~ minBLEP when it is done.