Re: antialiased oscillators spec

julian.rohrhuber-QYZGCWsIODmAF8UT6DzBU6xOck334EZe@public.gmane.org Thu, 24 May 2018 10:03:15 +0200

Newsgroups	gmane.comp.audio.supercollider.devel
Message-ID	<[email protected]>

Hi Nathan,

this is indeed very promising!

> width: Pulse width from 0 to 1. Internally, the width is clamped to around 0.01 to 0.99 because the antialiasing methods do not work well for true impulses.

For no aliasing, the clamping would be unnecessary. And wouldn’t the clamping boundary have to be a frequency relative value?

It would be beautiful if the oscillator could become a band limited impulse for short widths. Maybe the width clamping can be internally adjusted carefully?

> 
> Would we want a way to output a hard sync trigger along with the waveform itself? If so, how?

This would be better implemented (I guess) as a separate UGen that takes a sync *input* in form of a trigger.

Unless of course, the time resolution issue that you mention is also relevant in that case (I’m not sure if it really matters): then you’d need a MultiOutUGen, like Pitch, that returns an array.


> I don't plan on supporting negative frequencies either. In general, my attitude is that I want these oscillators to sound pristine when used "properly." If users discover ways to abuse them, that's fine, but I won't put in additional effort to let them be abused.

As long as one doesn’t need to think about the specific implementation of the ugen, I agree.

In my experience, whatever can be done easily enough and with little loss of efficiency should be done, especially if the use is “logical”. By logical I mean that a negative frequency, for example, is consistent with an inversion of the waveform, and a phase that touches or exceeds the [0 … 1] range has a clear mathematical counterpart, because the wave has no bounds.

I mention this only because your intention is to think further ahead, and you have made a great advance in the "combinatorial completeness" of the ugens in terms of rates and quality.

signature.asc (application/pgp-signature, 833 B)