Mix_PlayChannel error code empty

Johannes_Prix · May 19, 2002, 9:22am

Hi!

I use Mix_PlayChannel to play sounds in my game and it works GREAT! Yet
sometimes it occurs that collision-sounds occur in large numbers and
very rapid succession and then my Mix_PlayChannel(-1,
Loaded_WAV_Files[Tune] , 0) - Command returns -1. Of course this is
something I have to catch and that won’t be too difficult. I guess the
error might be due to no sound channels left or something. However in
this case Mix_GetError() returns an empty string? Why is this? Is this
normal? Thanks again for any help or suggestions, Johannes.

icculus · May 19, 2002, 9:48am

I use Mix_PlayChannel to play sounds in my game and it works GREAT! Yet
sometimes it occurs that collision-sounds occur in large numbers and
very rapid succession and then my Mix_PlayChannel(-1,
Loaded_WAV_Files[Tune] , 0) - Command returns -1. Of course this is
something I have to catch and that won’t be too difficult. I guess the
error might be due to no sound channels left or something. However in
this case Mix_GetError() returns an empty string? Why is this? Is this
normal? Thanks again for any help or suggestions, Johannes.

The empty error string is probably a bug.

You should allocate more mixing channels with Mix_AllocateChannels() and
see if you still have trouble.

–ryan.

Jonathan_Atkins · May 19, 2002, 12:33pm

Ryan C. Gordon wrote:

I use Mix_PlayChannel to play sounds in my game and it works GREAT! Yet
sometimes it occurs that collision-sounds occur in large numbers and
very rapid succession and then my Mix_PlayChannel(-1,
Loaded_WAV_Files[Tune] , 0) - Command returns -1. Of course this is
something I have to catch and that won’t be too difficult. I guess the
error might be due to no sound channels left or something. However in
this case Mix_GetError() returns an empty string? Why is this? Is this
normal? Thanks again for any help or suggestions, Johannes.

The empty error string is probably a bug.

You should allocate more mixing channels with Mix_AllocateChannels() and
see if you still have trouble.

–ryan.

I saw many places in SDL_mixer, where the error string may not be set, but an error code is returned…–
-==-
Jon Atkins
http://jcatki.2y.net/

slouken · May 19, 2002, 1:52pm

Hi!

I use Mix_PlayChannel to play sounds in my game and it works GREAT! Yet
sometimes it occurs that collision-sounds occur in large numbers and
very rapid succession and then my Mix_PlayChannel(-1,
Loaded_WAV_Files[Tune] , 0) - Command returns -1. Of course this is
something I have to catch and that won’t be too difficult. I guess the
error might be due to no sound channels left or something. However in
this case Mix_GetError() returns an empty string? Why is this? Is this
normal? Thanks again for any help or suggestions, Johannes.

The empty string was a bug, and this is fixed in CVS.

Thanks!
-Sam Lantinga, Software Engineer, Blizzard Entertainment

icculus · May 19, 2002, 7:01pm

I saw many places in SDL_mixer, where the error string may not be set,
but an error code is returned…

Yeah, mostly these just need to be shaken out when someone has time to
stare at it. Beyond bugfixes like the mutex stuff, I have very little
motivation to improve SDL_mixer until it gets ripped apart in the
rewrite…which will hopefully happen soon.

–ryan.

icculus · May 19, 2002, 7:02pm

You got it:

Added a function to query the music format: Mix_GetMusicType()

The empty string was a bug, and this is fixed in CVS.

(Sam, on the other hand, is holding up the fort for both of us. )

–ryan.

David_Olofson · May 21, 2002, 4:35pm

Well, running out of voices - even though it’s a condition that can
result in very audible artifacts - is not a serious error.

There are basically three ways of handling this situation:

1. Allocate more resources, to increase the number of
   simultaneous voices.

	+ You'll never run out of voices.

	- You risk running out of CPU power!
	- Very hard to do reliably on a normal OS,
	  since memory allocation is higly non-
	  deterministic. (ie may cause drop-outs)

2. Drop new sounds until playing voices are finished.
   (Seems to be the default action with SDL_mixer.)

	+ Easy to implement.
	+ No clicking or popping.
	+ No CPU time issues. (Fixed voice pool.)
	+ No system dependant real time issues.

	- New and important sounds may be blocked by
	  barely audible tails of old sounds.

3. Steal voices. (What nearly all synths do when they
   run out of voices.)

	+ New and important sounds will be played.
	+ No CPU time issues. (Fixed voice pool.)
	+ No system dependant real time issues.

	- Tricky to implement without introducing
	  clicks or pops, still without delaying
	  new sounds audibly.
	- Figuring out which channel/voice to steal
	  is rarely as trivial as it may sound.

Unless it’s already supported, I think it would be possible to implement
3) on top of SDL_Mixer, but a good implementation would probably have
to be a part of SDL_Mixer. (It’s better and easier to this kind of stuff
in the real time context of the engine, than “from the outside”.)

Either way, the best solution is usually not to run out of channels at
all.

Especially when it comes to sound effects, the major problem with
"runaway explosions" and the like seems to be the total output level. Too
many explosions playing at a time will drive the mixer into clipping, or
as in Kobo Deluxe, make the output compressor/limiter “pump”. Neither
sounds particularly good.

Early versions of Kobo Deluxe (as well as many, many other games) “solve"
this by locking each sound effect to a specific voice. That is, there can
be only one instance of each sound effect playing at a time. If a sound
effect is triggered while playing, it just aborts and restarts. No
"buildup” effect - but obviously, the result isn’t all that "massive"
either…

Using a random factor for the retrig/don’t retrig helps reducing that
BzzzzzzzzOOM! effect when large (usually tiled) monsters or buildings
explode, but it’s still far from perfect. Besides, it can also cause
audible clicking, unless the mixer engine deals with that.

In the current version of Kobo Deluxe I’m using a polyphonic soft synth.
Explosions always allocate new voices, and they’re played at slightly
varying pitch. Pan and volume is used for “2D positional” sound. Phatt
sound and no clicking - but the game logic causes other problems.

The biggest one is keeping exploding bases from flooding the audio
engine. Exploding bases are in fact handled by invisible "enemies"
crawling around the base, multiplying when they hit branches. The net
result is that there’s no easy way of controlling the number of enemies
simultaneously and repeatedly firing off explosion sounds. Likewise when
the player shoots nodes at a high rate, having pipes exploding
independently all over the place. Seems like these "exploding pipe"
enemies will need to cooperate better when it comes to the sound effects.

//David Olofson — Programmer, Reologica Instruments AB

.- M A I A -------------------------------------------------.
| Multimedia Application Integration Architecture |
| A Free/Open Source Plugin API for Professional Multimedia |
----------------------------> http://www.linuxdj.com/maia -' .- David Olofson -------------------------------------------. | Audio Hacker - Open Source Advocate - Singer - Songwriter |-------------------------------------> http://olofson.net -'On Sunday 19 May 2002 18:22, Johannes Prix wrote:

Hi!

I use Mix_PlayChannel to play sounds in my game and it works GREAT!
Yet sometimes it occurs that collision-sounds occur in large numbers
and very rapid succession and then my Mix_PlayChannel(-1,
Loaded_WAV_Files[Tune] , 0) - Command returns -1. Of course this is
something I have to catch and that won’t be too difficult. I guess the
error might be due to no sound channels left or something. However in
this case Mix_GetError() returns an empty string? Why is this? Is
this normal? Thanks again for any help or suggestions, Johannes.

Gaetan_de_Menten · May 21, 2002, 4:57pm

Early versions of Kobo Deluxe (as well as many, many other games) “solve"
this by locking each sound effect to a specific voice. That is, there can
be only one instance of each sound effect playing at a time. If a sound
effect is triggered while playing, it just aborts and restarts. No
"buildup” effect - but obviously, the result isn’t all that "massive"
either…
The solution I came up with was to create a simple C++ class (could be done
in C, of course) which allow me to allocate a finite number of channels for
each type of sound. I can also have non-allocated sounds. These go to the
free channels. For example, I allocated 4 channels for exlosions and 2
channels for fire, the rest of the sounds (bonus/etc…) share the other
channels. This gives pretty good results… If anybody wants the code, just
mail me… It’s pretty short.

-Ga?tan.

David_Olofson · May 21, 2002, 6:00pm

Yeah, I’ve been thinking along those lines as well. Actually, I have
something slightly similar already; priority control for voice
allocation. By setting a priority for each channel, I can tell the voice
allocator which channels are the most important. (Currently, I’m only using
that to keep sound effects from stealing voices from the music.)

BTW, I know of at least one studio synth module that does exactly what
you describe; the Roland JV-1080. You can configure the "voice reserve"
for each part in a performance patch if you like. Parts with 0 will fight
over what remains of the 64 voices.

Anyway, in the case of Kobo Deluxe, I’d prefer a “cleaner” solution,
where some kind of “granular explosion synthesizer” keeps a sensible
number of explosion sounds playing, until you tell it to stop.

//David Olofson — Programmer, Reologica Instruments AB

.- M A I A -------------------------------------------------.
| Multimedia Application Integration Architecture |
| A Free/Open Source Plugin API for Professional Multimedia |
----------------------------> http://www.linuxdj.com/maia -' .- David Olofson -------------------------------------------. | Audio Hacker - Open Source Advocate - Singer - Songwriter |-------------------------------------> http://olofson.net -'On Wednesday 22 May 2002 02:01, Ga?tan de Menten wrote:

Early versions of Kobo Deluxe (as well as many, many other games)
“solve” this by locking each sound effect to a specific voice. That
is, there can be only one instance of each sound effect playing at a
time. If a sound effect is triggered while playing, it just aborts
and restarts. No “buildup” effect - but obviously, the result isn’t
all that “massive” either…

The solution I came up with was to create a simple C++ class (could be
done in C, of course) which allow me to allocate a finite number of
channels for each type of sound. I can also have non-allocated sounds.
These go to the free channels. For example, I allocated 4 channels for
exlosions and 2 channels for fire, the rest of the sounds
(bonus/etc…) share the other channels. This gives pretty good
results…