27 Nov
2003
27 Nov
'03
6:33 p.m.
Has anyone else seen dropping of notes with EMU10K1 ALSA drivers mentioned by Joerg during
wavetable synthesis?
I have seen dropping of notes with both EMU10K1 and ICE1712 EXTERNAL MPU-401, but I
don't believe this is because of the drivers. Neither Muse nor Rosegarden function
well on my machine, but pmidi has absolutely no problems. I have been assuming that these
dropped notes are due to the larger load of GUI MIDI sequencers versus command-line
sequencers, not the drivers. Rosegarden is worst, Muse next. BTW, Rosegarden is much,
much worse for me than any Windows sequencers I have. I run 2.4.20 with the usual kernel
patches for low-latency.
Just curious about what others have seen. Thanks!
27 Nov
27 Nov
6:42 p.m.
New subject: [linux-audio-user] Recommendatio wanted
On Thursday 27 November 2003 22:31, davidrclark(a)earthlink.net wrote:
Rosegarden is much, much worse [at dropping notes] for
me
than any Windows sequencers I have.
If you're using Rosegarden-4 0.9.1 or earlier, then be aware that the
0.9.5 release -- which will probably come out tomorrow -- has a vastly
improved sequencer-GUI comms layer over the nasty kludge that was in
0.9.1, and hence sequences a lot more smoothly. There are still some
file prefetch issues to be worked out on the audio side, but MIDI is
generally very solid (for me at least) these days.
If you're already using the CVS version and hence already have
everything that'll be in 0.9.5, then... erm... say so, I guess.
Chris
6:54 p.m.
New subject: [linux-audio-user] Recommendatio wanted
At 09:31 AM 28/11/2003, davidrclark(a)earthlink.net wrote:
I run 2.4.20 with the usual kernel patches for
low-latency.
You are aware that 2.4.20 had ext3 issues? Have you tried upgrading your
kernel?
Luke
--
Luke Yelavich
AudioSlack Founder and head package maintainer
Audio software packaged for the Slackware Linux Distribution
http://www.audioslack.com
luke(a)audioslack.com
28 Nov
28 Nov
4:42 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Thu, 27 Nov 2003 davidrclark(a)earthlink.net wrote:
Has anyone else seen dropping of notes with EMU10K1
ALSA drivers mentioned by Joerg during wavetable synthesis?
The bug is well know by the driver developers. Please
have a look at
http://www.alsa-project.org/alsa-doc/doc-php/template.php?company=Creative+…
I have seen dropping of notes with both EMU10K1
and ICE1712 EXTERNAL MPU-401, but I don't believe
this is because of the drivers.
It has nothing to do with the sequencer. Please have
a look at:
http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html
There is an example which always causes droped notes.
Independent of which sequencer you are using.
--
J.Anders, Chemnitz, GERMANY (ja(a)informatik.tu-chemnitz.de)
6:20 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Fri, 28 Nov 2003 06:41 pm, Joerg Anders wrote:
http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html
There is an example which always causes droped notes.
Independent of which sequencer you are using.
Can you describe the fault(s) ? I just played "gap.not" back
thru Timidity in alsaseq mode from Noteedit on an old 1.2ghtz
AMD CPU and it seemed to me to be quite steady... I could not
aurally hear any missed beats or notice obvious glitches.
Also, do the glitches still exist in a rendered wav file ?
(where the sequencer engine does output via ALSA)
--markc
7:05 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Fri, 28 Nov 2003, Mark Constable wrote:
Can you describe the fault(s) ? I just played "gap.not" back
thru Timidity in alsaseq mode from Noteedit on an old 1.2ghtz
This ^^^^^^^ is no
problem. The error occurs if you use
the wavetable synthesizer of the Audigy2 card.
Als already stated: I'd use TiMidity instead. But
TiMidity loads the instruments not until they are really needed.
That is: If the clarinet keeps silent for - say - 20 measures
the clarinet waves are not loaded. In measure 20 TiMidity
starts loading the clarinet. This requires processor time
and leads to a short gap. (Also with my 2.4 GHz Intel)
Perhaps anybody knows how to change this behaviour?
I'd use FluidSynth instead. But is creates envelops for
strings as this were a xylophon (or so ...)
Also, do the glitches still exist in a rendered wav
file ?
(where the sequencer engine does output via ALSA)
On http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html there
is now an MP3 file (594K)
The gaps appear after:
12, 39, 43 and 47 seconds.
I used pmidi. But this does not matter. Yiu can use kmid or
tse3play or whatever you want: The gaps appear after
randomly choosen time :-((
--
J.Anders, Chemnitz, GERMANY (ja(a)informatik.tu-chemnitz.de)
7:19 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Friday 28 Nov 2003 11:03 am, Joerg Anders wrote:
I'd use FluidSynth instead. But is creates
envelops for
strings as this were a xylophon (or so ...)
I still don't quite understand what you mean by this. What is the
actual problem, in terms of what you hear? How does a faulty string
envelope sound?
I routinely use Fluidsynth with strings and haven't noticed any
particular problems, except that most freely available soundfonts
have rather too obvious short loops for their string samples. Is it
something particular to the soundfont you're using?
Chris
7:51 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Fri, 28 Nov 2003, Chris Cannam wrote:
I still don't quite understand what you mean by
this. What is the
actual problem, in terms of what you hear? How does a faulty string
envelope sound?
TiMidiy and all other synthezizers produce a long string tone
from 8mbgmsfx.sf2 soundfont if I tie 4 or 5 whole notes.
/---------------------------------------------.......
/
|
|
|
|
|
|
\
\---------------------------------------------- .......
FluidSynth produces such an envelope:
/---\
/ \
| \
| \ ---------------------------------------- ...
|
| /---------------------------------------- ...
| /
| /
\ /
\---/
I routinely use Fluidsynth with strings and haven't noticed any
particular problems, except that most freely available soundfonts
have rather too obvious short loops for their string samples. Is it
something particular to the soundfont you're using?
I use the 8mbgmsfx.sf2 soundfont which was shipped with some
soundcards.
As already mentioned: I know how to "fix" this:
diff -au fluidsynth-1.0.3/src/fluid_voice.c.ori fluidsynth-1.0.3/src/fluid_voice.c.new
--- fluidsynth-1.0.3/src/fluid_voice.c.ori 2003-11-26 12:39:45.000000000 +0100
+++ fluidsynth-1.0.3/src/fluid_voice.c.new 2003-11-26 12:30:25.000000000 +0100
@@ -1359,6 +1359,7 @@
case GEN_VOLENVHOLD: /* SF2.01 section 8.1.3 # 35 */
case GEN_KEYTOVOLENVHOLD: /* SF2.01 section 8.1.3 # 39 */
count = calculate_hold_decay_buffers(voice, GEN_VOLENVHOLD, GEN_KEYTOVOLENVHOLD, 0);
/* 0 means: hold */
+ if (count < 98) count = 19464;
voice->volenv_data[FLUID_VOICE_ENVHOLD].count = count;
voice->volenv_data[FLUID_VOICE_ENVHOLD].coeff = 1.0f;
voice->volenv_data[FLUID_VOICE_ENVHOLD].incr = 0.0f;
@@ -1370,6 +1371,7 @@
case GEN_VOLENVSUSTAIN: /* SF2.01 section 8.1.3 # 37 */
case GEN_KEYTOVOLENVDECAY: /* SF2.01 section 8.1.3 # 40 */
count = calculate_hold_decay_buffers(voice, GEN_VOLENVDECAY, GEN_KEYTOVOLENVDECAY,
1); /* 1 for decay */
+ if (count < 2000) count = 19464;
voice->volenv_data[FLUID_VOICE_ENVDECAY].count = count;
voice->volenv_data[FLUID_VOICE_ENVDECAY].coeff = 1.0f;
voice->volenv_data[FLUID_VOICE_ENVDECAY].incr = count ? -1.0f / count : 0.0f;
But - as further mentioned - this is barbarism. Somebody with SF2 knowledge should clarify
what's wrong!
--
J.Anders, Chemnitz, GERMANY (ja(a)informatik.tu-chemnitz.de)
8:13 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Friday 28 Nov 2003 11:50 am, Joerg Anders wrote:
TiMidiy and all other synthezizers produce a long
string tone
from 8mbgmsfx.sf2 soundfont if I tie 4 or 5 whole notes.
[...] FluidSynth produces such an envelope:
So a strong attack, then fading away rapidly to almost nothing?
I haven't got that soundfont, so I can't compare here. Certainly with
the soundfonts I do have, Fluidsynth produces a long level string
tone when I play a long note. The fonts I'm using probably just
contain fairly straightforward sample loops.
You could always just try another soundfont, although one supplied
with a card is probably more likely to be consistent and well
balanced. Still, the one I use most (Musica Theory 2) has pretty
adequate solo strings. I think I got it from Hammersound.
Chris
1 Dec
1 Dec
6:46 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Fri, 28 Nov 2003, Chris Cannam wrote:
On Friday 28 Nov 2003 11:50 am, Joerg Anders wrote:
Meanwhile I found the same happens with FluidR3 soundfont. And
this soundfont is freely available.
Please have at look at the end of
http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html
There you find the sound of 2 tied whole
notes created by the hardware synthesis, TiMidity,
and FluidSynth on basis of FluidR3 soundfont. Again:
FluidSynth creates "attacking" (and very silent) strings.
I also added the MIDI file. Do you really get different results
with FluidR3 soundfont and FluidSynth ?
--
J.Anders, Chemnitz, GERMANY (ja(a)informatik.tu-chemnitz.de)
TiMidiy and all other synthezizers produce a long
string tone
from 8mbgmsfx.sf2 soundfont if I tie 4 or 5 whole notes.
[...] FluidSynth produces such an envelope:
So a strong attack, then fading away rapidly to almost nothing?
I haven't got that soundfont, so I can't compare here. Certainly with
the soundfonts I do have, Fluidsynth produces a long level string
tone when I play a long note. The fonts I'm using probably just
contain fairly straightforward sample loops.
7:09 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Monday 01 Dec 2003 10:45 am, Joerg Anders wrote:
Do you really get different results
with FluidR3 soundfont and FluidSynth ?
Well, one problem is that I don't think you've ever actually said
which patch you're using, except that it's "strings". The result I
get varies wildly depending on which patch I use. With the Fluid3
soundfont, I get:
Violin (41), viola (42), cello (43): level tones, entered smoothly,
building up rather than attacking. A bit unexpressive but nice
enough.
Tremolo Strings (45): a very glitchy sound with a smooth envelope.
Sounds nasty, surely something wrong with the font or synth here but
it's not the same effect as you're describing.
String Ensemble 1 (49): pretty much the same as you. I'm guessing
this is what you're using.
String Ensemble 2 (50): a nice smooth string wash, no attack.
Synth Strings 1 (51): a smooth sound with some pretty bizarre tuning.
Synth Strings 2 (52): similar but more so (reminds me of that
out-of-tune synth in Joy Division's The Eternal).
So you see, I've generally assumed that most of these fonts were just
designed to have the string ensemble 1 sound a bit more vigorous for
faster passages and the string ensemble 2 be the friendly wash. That
is actually quite a useful distinction. It hadn't occurred to me
that other synths would play it differently.
Given that this isn't exhibited with all string patches, it really
might be worth taking a bit more time to check that the hardware
synth is actually the correct one. Just because you like it better
doesn't necessarily mean it is (although it seems probable). I
certainly agree with you that the difference is rather strange.
Chris
7:14 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Monday 01 Dec 2003 11:08 am, Chris Cannam wrote:
So you see, I've generally assumed that most of
these fonts were
just designed to have the string ensemble 1 sound a bit more
vigorous for faster passages and the string ensemble 2 be the
friendly wash. That is actually quite a useful distinction.
Aha, and I see they are actually coded as Strings and Slow Strings.
My mistake.
Looking at the font in Swami, the strings (ensemble strings 1 as I was
referring to it before) have distinctly shorter hold and release than
the slow strings... obviously. Hmm.
Chris
7:32 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Mon, 1 Dec 2003, Chris Cannam wrote:
On Monday 01 Dec 2003 10:45 am, Joerg Anders wrote:
... Ok, my
mistake!
Do you really get different results
with FluidR3 soundfont and FluidSynth ?
Well, one problem is that I don't think you've ever actually said
which patch you're using, except that it's "strings".
String Ensemble 1 (49): pretty much the same as you. I'm guessing
this is what you're using.
Yes, that's what I'm talking about!
String Ensemble 2 (50): a nice smooth string wash, no attack.
Yes, but these are slow strings. They always come too late.
Synth Strings 1 (51): a smooth sound with some pretty bizarre tuning.
Synth Strings 2 (52): similar but more so (reminds me of that
out-of-tune synth in Joy Division's The Eternal).
The synth strings are unusable in most cases. Only patch 49
behaves and sounds like real strings.
So you see, I've generally assumed that most of
these fonts were just
designed to have the string ensemble 1 sound a bit more vigorous for
faster passages and the string ensemble 2 be the friendly wash. That
is actually quite a useful distinction. It hadn't occurred to me
that other synths would play it differently.
The string ensemble 2 always comes too late. If you want an accompainmant
you cannot use string ensemble 2.
Given that this isn't exhibited with all string patches, it really
might be worth taking a bit more time to check that the hardware
synth is actually the correct one...
... TiMidity gives also smooth strings. The same happens with AWE32/64.
On M$-Windows are also non-attacking strings. Believe me: The
soundfont creator certainly had no attacking strings in mind!
Just because you like it better
doesn't necessarily mean it is (although it seems probable). I
It is certainly a FluidSynth error.
certainly agree with you that the difference is rather
strange.
My question is: What is the conclusion ? Is there any FluidSynth
programmer on this list ?
If not, that means: There is no soft synth which creates a
good bigband sound. The hardware sound is the only option.
But the Audigy2 driver produces gaps:
http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html
And apparently nobody will fix this. Thus, I'll throw away my
Audigy2 soundcard and I'll follow the recommendation of
On Wed, 26 Nov 2003, Hartmut Z Noack wrote:
If SB-Live does not work sufficiently, why dont trying
to use Terratecs
EWX24/96? The card does not have Hardware SF-Support but you can load
SFs into RAM (if you have enough...) and it works perfectly well for me.
Conclusion: try Delta 66 or some similar Pro-Gear or try it with a
ICE-chipbased card like terratecs muse, whitch are verywell supported by
alsa and can play SF from RAM.
The whole dilemma comes from: The AWE64 is an ISA card.
Otherwise, I'd further use AWE64 ... :-((
--
J.Anders, Chemnitz, GERMANY (ja(a)informatik.tu-chemnitz.de)
7:50 a.m.
New subject: [linux-audio-user] Recommendatio wanted
Hi,
I looked at this patch in Swami. It has lots of layers (8).
Unfortunately I can't listen to it right now so I can't do anything
really constructive...
Judgeing from the sound sample you sent I'd venture a guess that one or
two of the layers are audible, the rest are for some reason not played.
The tuning differance is probably due to this also, the layer is
supposed to be played with the others, to make it thicker.
/Robert
String Ensemble
1 (49): pretty much the same as you. I'm guessing
this is what you're using.
Yes, that's what I'm talking about! 
2:21 p.m.
New subject: [linux-audio-user] Recommendatio wanted
måndagen den 1 december 2003 12.31 skrev Joerg Anders:
Yes, but these are slow strings. They always come too
late.
My question is: What is the conclusion ? Is there any FluidSynth
programmer on this list ?
If not, that means: There is no soft synth which creates a
good bigband sound. The hardware sound is the only option.
But the Audigy2 driver produces gaps:
There's another sad thing about using soundfonts via soundblaster live/extigy
etc..: Not possible to use any decent reverb! :-( If I've understood it
correctly there are no such things as good effects for the onboard
soundblaster dsp processors (I bet it's a hassle loading/unloading/routing
too, if there are any good effects).
Am I completely wrong about this? Has anyone else on the list had any success
with these?
/Mathias
28 Nov
28 Nov
8:07 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Fri, 28 Nov 2003 09:03 pm, Joerg Anders wrote:
Als already stated: I'd use TiMidity instead. But
TiMidity loads the instruments not until they are really needed.
And this causes audible glitches in it's playback for you ?
That is: If the clarinet keeps silent for - say - 20
measures
the clarinet waves are not loaded. In measure 20 TiMidity
starts loading the clarinet. This requires processor time
and leads to a short gap. (Also with my 2.4 GHz Intel)
Perhaps anybody knows how to change this behaviour?
*If* you happen to be using kernel 2.4.20 and ext3 with
Timidity then that could've been the source of the problem.
On http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html
there
is now an MP3 file (594K)
Excellent, thanks... any chance of the FluidSynth string
envelope problem and dropouts from Timidity in mp3 format too :)
The gaps appear after: 12, 39, 43 and 47 seconds.
Wow, yes, very obvious. Enough to make composing unbearable.
I used pmidi. But this does not matter. Yiu can use
kmid or
tse3play or whatever you want: The gaps appear after
randomly choosen time :-((
Well if it doesn't happen with Timidity in alsaseq mode and,
as you point out being across many apps, then it must obviously
be something to do with the emu10k1 driver. Perhaps there might
be some buffer fragments and buffer size adjustments but I'm
not sure what values to suggest (if possible at all).
--markc
9:32 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Fri, 28 Nov 2003, Mark Constable wrote:
Excellent, thanks... any chance of the FluidSynth string
envelope problem ...
... yes if anybody explains the jack usage to me:
- I have installed jack-audio-connection-kit-0.90.0
- I have stared jackd -d alsa
- I have recompiled fluidsynth with jack support
- I have started fluidsynth -j <soundfont>
And how can I determine the WAV file to put the FluidSynth
output there ?
--
J.Anders, Chemnitz, GERMANY (ja(a)informatik.tu-chemnitz.de)
10:25 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Fri, 28 Nov 2003, Mark Constable wrote:
Excellent, thanks... any chance of the FluidSynth
string
envelope problem ....
Ok, I got it! At the end of:
http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html
You'll find 3 files:
- strings_hardware.mp3 (67K)
- strings_timidity.mp3 (121K)
- strings_fluid.mp3 (71K)
produced by hardware midi synthesizer, TiMidity++, and
FluidSynth. It is the same midi and the same
soundfont (8mbgmsfx.sf2).
In contrast to TiMidity++ and the hardware midi synthesizer
FluidSynth produces a clear attack.
Inside a big orchestra this leads to inaudible strings.
Or - if you increase the volume - to "attacking" strings.
--
J.Anders, Chemnitz, GERMANY (ja(a)informatik.tu-chemnitz.de)
10:42 a.m.
New subject: [linux-audio-user] Recommendatio wanted
Hi,
Joerg Anders wrote:
On Fri, 28 Nov 2003, Mark Constable wrote:
Excellent!
Excellent, thanks... any chance of the FluidSynth
string
envelope problem ....
Ok, I got it! At the end of:
http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html
You'll find 3 files:
- strings_hardware.mp3 (67K)
- strings_timidity.mp3 (121K)
- strings_fluid.mp3 (71K)
Crappy headphones and lots of background noise, but isn't there a layer
missing from the fluid output?
It definitely sound different...
/R
produced by hardware midi synthesizer, TiMidity++,
and
FluidSynth. It is the same midi and the same
soundfont (8mbgmsfx.sf2).
In contrast to TiMidity++ and the hardware midi synthesizer
FluidSynth produces a clear attack.
Inside a big orchestra this leads to inaudible strings.
Or - if you increase the volume - to "attacking" strings.
11:06 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Fri, 28 Nov 2003, Robert Jonsson wrote:
but isn't there a layer missing from the fluid output?
... that's something I also assume. It must be a relatively
rare SF2 feature. Perhaps layers.(?)
BTW: The same happens with the clarinet (GM-No: 71). But not with
"choir_Aahs" (GM-No: 52) although this has a similar envelope.(??)
And: It never happens with TiMidity++ and/or hardware synthesis.
--
J.Anders, Chemnitz, GERMANY (ja(a)informatik.tu-chemnitz.de)
11:33 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Friday 28 Nov 2003 2:23 pm, Joerg Anders wrote:
- strings_hardware.mp3 (67K)
- strings_timidity.mp3 (121K)
- strings_fluid.mp3 (71K)
Hmm, yes. The difference is pretty obvious there.
Is that the string ensemble patch? I do find that one has quite a bit
of attack; I thought you were referring to the solo strings, which
have some attack as well but not so much as to make one suspect the
code rather than the whim of the soundfont designer.
I made my own comparative mp3s, but then I discovered they were
actually playing at different pitches from yours and not in an
obvious 44.1:48KHz ratio either. So now I'm all confused and have to
go and find my tuning fork.
Chris
11:35 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Sat, 29 Nov 2003 12:23 am, Joerg Anders wrote:
http://rnvs.informatik.tu-chemnitz.de/DROP/gap.html
You'll find 3 files:
- strings_hardware.mp3 (67K)
- strings_timidity.mp3 (121K)
- strings_fluid.mp3 (71K)
produced by hardware midi synthesizer, TiMidity++, and
FluidSynth. It is the same midi and the same
soundfont (8mbgmsfx.sf2).
Totally cool to be able to hear what you are talking about.
Are you sure they are all meant to be the same pitch ? The
fluidsynth notes seem to be almost a semi-tone too high. I've
got all 3 in a loop in xmms and the pitch difference is really
obvious... perhaps there's some simple off-by-one value in
there somewhere and perhaps it's effecting both the envelope
and pitch change.
There is also a stereo-ish pan and nicer delay to the timidity
sample so I suspect you did not use -EFreverb=0 -EFchorus=0 etc.
--markc
11:50 a.m.
New subject: [linux-audio-user] Recommendatio wanted
On Friday 28 Nov 2003 3:33 pm, Mark Constable wrote:
Are you sure they are all meant to be the same pitch ?
The
fluidsynth notes seem to be almost a semi-tone too high.
Oh yes! No wonder I couldn't get my own version to match up.
Here's the same thing using first the violin patch and then the
"tremolo strings" (string ensemble) patch of Musica Theory 2
in Fluidsynth:
http://www.all-day-breakfast.com/rosegarden/tmp/violins.mp3
http://www.all-day-breakfast.com/rosegarden/tmp/strings.mp3
As you hear, they both have some attack but the second is far more
pronounced. The first is the patch that I originally thought the
discussion was about.
Chris
12:40 p.m.
New subject: [linux-audio-user] Recommendatio wanted
On Friday 28 November 2003 05:03 am, Joerg Anders wrote:
Als already stated: I'd use TiMidity instead. But
TiMidity loads the instruments not until they are really needed.
That is: If the clarinet keeps silent for - say - 20 measures
the clarinet waves are not loaded. In measure 20 TiMidity
starts loading the clarinet. This requires processor time
and leads to a short gap. (Also with my 2.4 GHz Intel)
Perhaps anybody knows how to change this behaviour?
This would be an ugly workaround, but could you (using your example) force
TiMidity to load the sound at the beginning by putting in a clarinet note
with velocity 0?
|)
|)enji
7460
days inactive
7464
days old
linux-audio-user@lists.linuxaudio.org
24 comments
8 participants
participants (8)
-
Benjamin Flaming -
Chris Cannam -
davidrclark＠earthlink.net -
Joerg Anders -
Luke Yelavich -
Mark Constable -
Mathias Lundgren -
Robert Jonsson