Re: [yoshimi-user] Different sound in Yoshimi and ZynAddSubFX.

  • From: Andreas Pierrou <pierrou.andreas@xxxxxxxxx>
  • To: yoshimi-user@xxxxxxxxxxxxxxxxxxxxx
  • Date: Thu, 17 Oct 2013 13:13:25 +0200

On 10/17/2013 04:41 AM, David Adler wrote:

On 10/16/13, Andreas Pierrou wrote:
On 10/15/2013 11:45 PM, David Adler wrote:
On 10/15/13, Andreas Pierrou wrote:
On 10/15/2013 08:33 PM, Andreas Pierrou wrote:
...
When I compare the sound generated by ZynAddSubFX and Yoshimi for the
xiz-file attached to this mail (after opening the attached xmz-file),
it
sounds noticably different even though Yoshimi is said to generate the
same sound as ZynAddSubFX. Even when I compare instruments in the
default instrument bank, the sound differs. It is possible for any of
you to reproduce this? :'( Are the synthesizing algorithms slightly
different in Yoshimi and ZynAddSubFX?
...
Btw, I forgot: I'm using ZynAddSubFX v2.4.3 and Yoshimi v1.1.0 with a
linux-rt kernel.

https://image.bayimg.com/344efbe1a0ee4d330895f185b7ca3152243807e5.jpg

These are 30 cycles of a C5 note, displayed in Jaaa with peak hold.
Yoshimi on the left, Zyn on the right. Master volume is turned to
the lowest possible setting in both synths.
Same versions as used by Andreas.

So yes, they look and sound different. Yoshimi sounds a bit brighter.
Judging from sound and spectrograms (and peak values), the
generating algorithms aren't identical; though I don't recall any
such changes being intended or announced here on the list.

Ah, hmm, that's interesting. Why were the algorithms changed? (To
provide a lower latency than ZynAddSubFX?)

As Will now confirmed, they should not have changed. Latency of
the sound generation is then determined by Jack settings, so
the objective wasn't to provide lower latency on the synth side
but to potentially allow for lower xrun-free latency settings in Jack.

Hmm, OK. But are the synth algorithms the only cause of xruns? After I
installed the linux-rt kernel I almost haven't had any problems with
xruns at all (while having a lot more xruns before my linux-rt
installation). Maybe the complaints about large amounts of xruns were
because of a bad system configuration instead of the synth algorithms?
Are the synth algorithms in ZASFX really that much resource consuming?

Or c) we manage to track down the cause of the differences and
fix that in Yoshimi (my favourite solution).
As I'm using Yoshimi exclusively and as the instruments from the
bank still sound very usable, I am personally not affected that
much -- but Yoshi really ought to be a drop-in replacement for Zyn.

That's the solution I prefer too, if the Yoshimi developers agree on
that. But who changed the algorithms? (I don't know the names of the
developers, except for Cal and Jeremy, and I don't know if Cal is active
on this mailing list.) Maybe that person had good reasons to change them
after all. Maybe we should listen to him before deciding whether or not
to change back to the ZASFX synth algorithms so that the project won't
be messed up.

Actually, I have in fact had a few problems with the sound in ZASFX. In
many positions of the knobs I hear "scratches". What's interesting is
that even though this phenomenon gets worse when the volume is very high
(suggesting it's a symptom of clipping), it still occurs when the volume
is moderate, but then in seemingly randomly knob positions (suggesting
that its not only a result of clipping or that clipping occurs even for
moderate volumes). I don't know if this is something theoretically
unavoidable or just a bug in the software. Does anyone recognize these
"scratchy sounds" in ZASFX?

Best regards
Andreas


Other related posts:

  1. Home
  2. yoshimi
  3. Archive
  4. 10-2013