[LAU] Heads Up! ZynAddSubFX (Initial Feedback)

[Frank Neumann]

Hi list,
on Fri, 17 Jul 2009 15:24:04 +0200 <hollunder at gmx.at> wrote:

> True. I just tried and pushed and in normal use cases there's nothing
> to complain about. Changing patches doesn't cause it to disconnect from
> jack anymore. I often hear a slight crackle but that doesn't register
> as xrun.

I made some tests in the last days as well, but I am not quite as happy -
my system is an AMD64 (Ubuntu 9.04), running on an M-Audio Audiophile 2496
with 256 frames, 48kHz, 2periods.

Yes, there are no xruns in QJackCtl, but I only need to use a rather light
patch like "DX Rhodes 1" from the "Rhodes" group, and by pressing some
"staccato" chords, I can easily create crackles. The CPU load does not go
beyond some 13-15%, and these are no "overdrive" crackles from stacking up
too many voices.

If I record such a "session" in, say, timemachine and analyze the file
afterwards in a sample editor like Sweep, I can clearly detect the
"0-buffers" (I even counted their size manually: exactly 256 frames).
So, while it's certainly great that ZynAddSubFX is not zombified anymore
(and that the fonts finally look ok :-), the fact that I both *hear* crackles
AND *get* them in the recorded file still makes Zyn in a JACK environment
rather useless for me. Now, if THAT could be fixed.. :-).

If I recompile it for standard OSS output, there are no such crackles. So
it's clearly related to JACK mode usage.

As far as I understood from the Zyn mailing list archives, people are trying
to get this fixed, but it's obviously a non-trivial task.

Greetings,
Frank