9 Dec
2002
9 Dec
'02
7:06 a.m.
Some thoughts on that SILENT event for reverb tails and stuff...
(Currently impemented as a "fake" spontaneous state change in
Audiality FX plugins, BTW.)
I would assume that since there is no implicit relation between
Channels on different Bays (remember the input->output mapping
discussion?), this event is best sent from some kind of Master Event
Output Channel. (That is, now we have one Master Event Input Channel,
and one Master Event Output Channel. Each will be in it's own Bay,
and there can be only one and exactly one Channel on each of those.)
So, the SILENT event would need Bay and Channel (but not Slot)
fields, in order to tell the host (or who ever gets the event) which
audio output just went silent.
And it would probably be a rather good idea to have "NOT_SILENT"
event as well, BTW!
Anyway, what I was thinking was: How about allowing plugins to
*receive* SILENT and NOT_SILENT events, if they like?
That way, you could use the plugin API for things like
audio-to-disk-thread "gateways" for recording and that kind of stuff,
without forcing the host to be involved in the details.
Not that recording half a buffer extra of silence would be a
disaster, but I bet someone can or eventually will think of a reason
why their plugin should know the whole thruth about the audio inputs.
Now, there's just one problem: Put a plugin with tail, but without
sample accurate "tail management" support in between a plugin that
sends (NOT_)SILENT events and one that can receive them - and the
information is useless! All you can do is have the host fake the
(NOT_)SILENT events sent to the latter plugin, since the plugin in
the middle thinks only in whole buffers WRT inputs and/or outputs...
And there's another problem: If you would get a (NOT_)SILENT event
*directly* from another plugin, how on earth would you know which one
of *your* audio inputs that other plugin is talking about, when the
event arguments are about *that* plugin's audio outputs?
Only the host knows where audio ports are connected, so the host
would have to translate the events before passing them on.
//David Olofson - Programmer, Composer, Open Source Advocate
.- The Return of Audiality! --------------------------------.
| Free/Open Source Audio Engine for use in Games or Studio. |
| RT and off-line synth. Scripting. Sample accurate timing. |
`---------------------------> http://olofson.net/audiality -'
.- M A I A -------------------------------------------------.
| The Multimedia Application Integration Architecture |
`----------------------------> http://www.linuxdj.com/maia -'
--- http://olofson.net --- http://www.reologica.se ---
9 Dec
9 Dec
9 a.m.
On Mon, Dec 09, 2002 at 01:01:02PM +0100, David Olofson wrote:
So, the SILENT event would need Bay and Channel (but
not Slot)
fields, in order to tell the host (or who ever gets the event) which
audio output just went silent.
I was assuming the SILENT event would be aginst a voice ID.
That way, you could use the plugin API for things like
audio-to-disk-thread "gateways" for recording and that kind of stuff,
without forcing the host to be involved in the details.
That will only work if all your instruments procduce SILENT events and you
recevie all the relevent VOICE_ON events though. Is this even possible? I
bet I can think of a processing graph where it isn't resolvable.
Paticularly if you allow LADSPA plugins in the graph.
For this and the other reasons you mentioned I dont think this is not
possible/practical.
- Steve
9:16 a.m.
On Monday 09 December 2002 14.55, Steve Harris wrote:
It's not about Voices - and Voices become irrelevant anyway, as soon
as the synth has an internal FX with tail... This is all about audio
ports.
On Mon, Dec 09, 2002 at 01:01:02PM +0100, David
Olofson wrote:
Uhm, a Voice is an abstract object that has no known relation to
whatever audio outputs the synth may have... The SILENT event I'm
talking about is what a reverb FX would send when the tail goes below
the threshold, or that a synth would send when *all* Voices that are
routed (dry) to a certain audio output are dead.
So, the SILENT event would need Bay and Channel
(but not Slot)
fields, in order to tell the host (or who ever gets the event)
which audio output just went silent.
I was assuming the SILENT event would be aginst a voice ID. That way, you
could use the plugin API for things like
audio-to-disk-thread "gateways" for recording and that kind of
stuff, without forcing the host to be involved in the details.
That will only work if all your instruments procduce SILENT events
and you recevie all the relevent VOICE_ON events though. Is this
even possible? I bet I can think of a processing graph where it
isn't resolvable. Paticularly if you allow LADSPA plugins in the
graph.
Yes - but then you'd either have to insert meter plugins to detect
silence, or just tell the user "Sorry, but I can't reliably
autodetect silence in this net."
For this and the other reasons you mentioned I dont
think this is
not possible/practical.
Well, I agree that there probably isn't much point in spending much
time on this. If it fits in nicely somewhere, fine - but otherwise,
just allow nice plugins to send SILENT events whenever an audio
output port changes between silence and non-silence, and/or accept
NULL pointers for input buffers.
The only part of this that hosts *have* to support would be the
SILENT event, since that means the plugin may not even touch
subsequent buffers. They have to be cleared, or if the receiving
plugin supports the "NULL buffer for silence" feature, that may be
used.
//David Olofson - Programmer, Composer, Open Source Advocate
.- The Return of Audiality! --------------------------------.
| Free/Open Source Audio Engine for use in Games or Studio. |
| RT and off-line synth. Scripting. Sample accurate timing. |
`---------------------------> http://olofson.net/audiality -'
.- M A I A -------------------------------------------------.
| The Multimedia Application Integration Architecture |
`----------------------------> http://www.linuxdj.com/maia -'
--- http://olofson.net --- http://www.reologica.se ---
10:23 a.m.
On Mon, Dec 09, 2002 at 03:11:51PM +0100, David Olofson wrote:
OK, that sounds useful, but also really hard. Realisticly a reverb can't
tell when its gone to zero except by inspection and guesswork. If it has
nonlienar internal behaviour it may not even be possible/practical. I
tried to cacluate the RT60 time for Josep's gong model, based on the
feedback levels. It turned out to be /seriously/ hard. Basicly not
possible without actually running it ;)
I though you wanted it so that the host would know when instrument voices
had finsihed (as in saol or csound).
If you want to render tails I think its just better to have a postroll
time like ardour (and presuamably protools).
- Steve
I was assuming
the SILENT event would be aginst a voice ID.
Uhm, a Voice is an abstract object that has no known relation to
whatever audio outputs the synth may have... The SILENT event I'm
talking about is what a reverb FX would send when the tail goes below
the threshold, or that a synth would send when *all* Voices that are
routed (dry) to a certain audio output are dead.
11:13 a.m.
On Monday 09 December 2002 16.18, Steve Harris wrote:
On Mon, Dec 09, 2002 at 03:11:51PM +0100, David
Olofson wrote:
OK, that sounds useful, but also really hard. Realisticly a reverb
can't tell when its gone to zero except by inspection and
guesswork.
Well, just don't do it if it's too hard. Keep in mind though, that by
doing that, you're only moving the problem over to the host and/or
the user.
Besides, since plugins have access to their internal buffers and
variables, I would think that they practically always have a better
chance of doing a good job of this than the host or the user.
I was
assuming the SILENT event would be aginst a voice ID.
Uhm, a Voice is an abstract object that has no known relation to
whatever audio outputs the synth may have... The SILENT event I'm
talking about is what a reverb FX would send when the tail goes
below the threshold, or that a synth would send when *all* Voices
that are routed (dry) to a certain audio output are dead. If it has nonlienar internal behaviour it may not
even
be possible/practical. I tried to cacluate the RT60 time for
Josep's gong model, based on the feedback levels. It turned out to
be /seriously/ hard. Basicly not possible without actually running
it ;)
Well, there's always the output meter hack. :-)
Yes, the host could do that - but it won't work for delays and that
kind of FX, because they can be totally silent for long periods of
time, while there is still output to come.
I though you wanted it so that the host would know
when instrument
voices had finsihed (as in saol or csound).
Well, maybe that could be of some use, but frankly, I can't think of
anything right now.
If you want to render tails I think its just better to
have a
postroll time like ardour (and presuamably protools).
Yeah, that's probably sufficient for practically every user.
I guess there is a reason why most plugin APIs have this feature -
but there must also be a reason why so many plugins don't implement
it. :-)
//David Olofson - Programmer, Composer, Open Source Advocate
.- The Return of Audiality! --------------------------------.
| Free/Open Source Audio Engine for use in Games or Studio. |
| RT and off-line synth. Scripting. Sample accurate timing. |
`---------------------------> http://olofson.net/audiality -'
.- M A I A -------------------------------------------------.
| The Multimedia Application Integration Architecture |
`----------------------------> http://www.linuxdj.com/maia -'
--- http://olofson.net --- http://www.reologica.se ---
12:54 p.m.
On Mon, Dec 09, 2002 at 05:08:29PM +0100, David Olofson wrote:
Good! If the host really wants to know it was waste cycles doing it. Most
of the time the host wont want to know, so it will just waste the plugin
cycles trying to calculate when its silent (especially if it has to
inspect internal buffers). I guess you could indicate to the plugin wether
you care, but thats getting more complex.
OK, that
sounds useful, but also really hard. Realisticly a reverb
can't tell when its gone to zero except by inspection and
guesswork.
Well, just don't do it if it's too hard. Keep in mind though, that by
doing that, you're only moving the problem over to the host and/or
the user. Well, there's always the output meter hack. :-)
Yep, and that requires no support from the plugin.
I guess there is a reason why most plugin APIs have
this feature -
but there must also be a reason why so many plugins don't implement
it. :-)
One thing we should have is a latency indication. Its blantantly missing
from ladspa.
Though its probably only useful on inline processing plugins, not
instruments.
- Steve
3:51 p.m.
On Monday 09 December 2002 18.49, Steve Harris wrote:
Exactly.
Yes indeed. Missing in most other APIs as well. (VST as well, IIRC.)
On Mon, Dec 09, 2002 at 05:08:29PM +0100, David
Olofson wrote:
Good! If the host really wants to know it was waste cycles doing
it. Most of the time the host wont want to know, so it will just
waste the plugin cycles trying to calculate when its silent
(especially if it has to inspect internal buffers). I guess you
could indicate to the plugin wether you care, but thats getting
more complex.
Yeah... *hehe* Good point.
OK, that
sounds useful, but also really hard. Realisticly a
reverb can't tell when its gone to zero except by inspection
and guesswork.
Well, just don't do it if it's too hard. Keep in mind though,
that by doing that, you're only moving the problem over to the
host and/or the user. Well,
there's always the output meter hack. :-)
Yep, and that requires no support from the plugin. I guess there
is a reason why most plugin APIs have this feature
- but there must also be a reason why so many plugins don't
implement it. :-)
One thing we should have is a latency indication. Its blantantly
missing from ladspa. Though its probably only useful on inline processing
plugins, not
instruments.
On the contrary, I think it's *very* important for instruments as
well; especially with percussive sounds. If you want to use two
synths together, and one or both for some reason have non-zero event
timestamp -> audio output latency, you'd really want the host to be
able to compensate for it, so you can start tweaking with the outputs
in exact sync.
I've mixed percussion sounds from SoundFonts on PCI cards with MIDI
syths. Need I say more? :-)
//David Olofson - Programmer, Composer, Open Source Advocate
.- The Return of Audiality! --------------------------------.
| Free/Open Source Audio Engine for use in Games or Studio. |
| RT and off-line synth. Scripting. Sample accurate timing. |
`---------------------------> http://olofson.net/audiality -'
.- M A I A -------------------------------------------------.
| The Multimedia Application Integration Architecture |
`----------------------------> http://www.linuxdj.com/maia -'
--- http://olofson.net --- http://www.reologica.se ---
5:33 p.m.
On Mon, Dec 09, 2002 at 09:46:09 +0100, David Olofson wrote:
I was pretty sure its in VST, but I could be wrong.
Yup, good point.
- Steve
One thing we
should have is a latency indication. Its blantantly
missing from ladspa.
Yes indeed. Missing in most other APIs as well. (VST as well, IIRC.) Though its
probably only useful on inline processing plugins, not
instruments.
On the contrary, I think it's *very* important for instruments as
well; especially with percussive sounds. If you want to use two
7901
days inactive
7901
days old
linux-audio-dev@lists.linuxaudio.org
7 comments
2 participants
participants (2)
-
David Olofson -
Steve Harris