There is test I use to evaluate one important aspect of all synthesis systems. It tests oscillator accuracy.
The patch is by Jean Claude Risset and is an additive concept he called frequency domain grating, and is analogous to diffraction grating used in spectroscopy.
Here is Hartmanns paper about it http://www.pa.msu.edu/acoustics/fdg.pdf
You take as many sines as the system will handle, typically a thousand or so, and sum them. All must start on exactly the same phase. Now, if we had a series of _all_ frequencies it would give us an impulse, but instead set the difference between each oscillator to be 1 cycle + delta, where delta is very small, maybe 1Hz or less.
What you will hear in Csound on a 64 bit system is a ** beautiful ** effect as a "rainbow" of all frequencies apparently plays in sequence.
Now, if the oscillators are very good the spacing will be equal and the effect will sound crystal clear with individual frequencies popping in and out in a regular stream.
If the oscillators are innacurate (because of distortion, bit depth, asymmetry etc) then you will hear a muddled effect. Any deviation from a perfect sine must introduce other hamonics and these come out in the sequence, so it's an empirical/practical way of testing the quality of a digital system without any special test equipment or measuring other than your ears.
Hardly very scientific, but roughly from the few chances I've had to try it on different systems...
Csound - the King, all bow before Csound Nyquist/CLM - a close second on some machines Puredata - Not bad, but not good, you can tell something is broken Max - pretty damn awful Reaktor - forget it
So I believe there is a difference between Max and Pd and Max is the loser. But of course all software is in development. And it must be said;
- this only test oscillator accuracy but may also test fp accuracy (I haven't
thought it through in detail)
- It depends how you do it. For example, using table oscillators that precompute
is different than taking the cosine (maybe Taylor or poly approximated) of a phasor.
If I had all systems running and time I'd try it again, but I don't. If we could define the exact parameters for a patch to eliminate variables I think Rissets diffraction is a very good test that reveals the quality of digital synthesis software.
On Thu, 06 Mar 2008 21:02:00 -0500 marius schebella marius.schebella@gmail.com wrote:
what I experience sometimes when I do very basic stuff like using phasors, is that I hear weird comb filtering of my environment after I put down my headphones. similar as if you look into bright light and then close the eyes, and you still see a review-image. regarding the difference between pd and max: are you talking about the music that people produce or are you talking about the digital signal process? m.
Damian Stewart wrote:
hey,
i was talking to a Portuguese musician tonight (Miguel Cardoso is his name) and he was saying that he thought that Pd sounded much better than Max - a fuller sound with the oscillators, he said.
i hadn't really thought about this before, but i do know that to my ears my Pd patches sound a lot richer than most Max/MSP stuff that I've heard - not sure whether that's my source material or patches or whether it's at a deeper architectural level than that.
anyone have any evidence, anecdotal or otherwise, to confirm this? reasons why this might be the case?
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list