Dmix

From the ALSA wiki

Jump to: navigation, search

PCM plugins extend the functionality of PCM devices allowing low-level sample conversions and copying between channels, files and soundcard devices. The dmix plugin provides for direct mixing of multiple streams.

NOTE: For ALSA 1.0.9rc2 and higher you don't need to setup dmix for analogue output. Dmix is enabled by default for soundcards which don't support hardware mixing. You still need to set it up for digital outputs.

Contents

The Dmix Howto

Written in large parts by Frank Barknecht (feel free to use it any way you like). This howto describes how to set up dmix for OSS applications.

1. Install aoss, alsaplayer, mpg321

Note: If you use Debian, aoss comes with the alsa-oss package. Just type:

apt-get install alsa-oss alsaplayer mpg321 alsaplayer-alsa alsa-base

The aplay command used below comes in the alsa-utils package for Debian, but this conflicts with the udev package, and thus you'll probably not want to install it.

2. Test basic dmix on alsa with:

alsaplayer -o alsa -d plug:dmix some.mp3 &
aplay -D plug:dmix some.wav &
alsaplayer -o alsa -d plug:dmix some.mp3 &

This should work everywhere with ALSA 0.9.7. (It works also with ALSA 0.9.8 on a snd-intel8x0, eg. on nForce2 audio). If the second and/or third audiostream stutters, it might be the buffers-problem mentioned below 7). Don't give up yet.

3. configure .asoundrc for aoss-testing:

# cat ~/.asoundrc:
pcm.dsp0 {
    type plug
    slave.pcm "hw:0"
}
# or:
# pcm.dsp0 pcm.default
# if "default" hasn't been redefined
ctl.mixer0 {
    type hw
    card 0
}

4. Test aoss with mpg123 in oss mode:

aoss mpg123 some.mp3

This should play the file, but not yet "dmix"ed. Now comes the tricky and rather difficult part: redefining dsp0 to behave like dmix.

Note that mpg321 (which is a DFSG-free clone of mpg123) uses libao and has alsa support built-in.

5. The simple approach:

# cat ~/.asoundrc
pcm.dsp0 {
    type plug
    slave.pcm "dmix"
    # A hint is required for listing the device in some GUIs, e.g. Phonon configuration.
    hint {
         show on
         description "My dmix dsp0"
    }
}
# mixer0 can stay unchanged, because
# it isn't used anyway, I guess ;)
ctl.mixer0 {
    type hw
    card 0
}

Test this setup with mpg123 like under 4). This should enable dmix'd OSS playback, but it will not work on cards like the M-Audio Audiophile or other Delta cards, that need a certain audio data format for playback on hw:0 (which dmix uses by default). A fix is below.

6. The complex approach (defining dmix parameters)

If a card needs a certain format conversion we need to create a custom dmix device. Let's call it ossmix.

pcm.ossmix {
    type dmix
    ipc_key 1024 # must be unique!
#   ipc_key_add_uid false   # let multiple users share
#   ipc_perm 0666           # IPC permissions for multi-user sharing (octal, default 0600)
    slave {
        pcm "hw:0,0"      # you cannot use a "plug" device here, darn.
        period_time 0
        period_size 1024 # must be power of 2
        buffer_size 8192  # dito. It
       #format "S32_LE"
       #periods 128 # dito.
       #rate 8000 # with rate 8000 you *will* hear,
       # if ossmix is used :)
    }
    # bindings are cool. This says, that only the first
    # two channels are to be used by dmix, which is
    # enough for (most) oss apps and also lets 
    # multichannel chios work much faster:
    bindings {
        0 0 # from 0 => to 0
        1 1 # from 1 => to 1
    }
}
pcm.dsp0 {
    type plug
    slave.pcm "ossmix" # use our new PCM here
}
# mixer0 like above
ctl.mixer0 {
    type hw
    card 0
}

# You may want to make your new ossmix the default for alsa.
# If your alsa programs are complaining that they can't open
# your sound devices, try uncommenting this next line
#pcm.default pcm.dsp0

It is important, that this "ossmix" PCM works with your card in ALSA mode. That is, the following should produce sound:

alsaplayer -o alsa -d ossmix some.mp3

7. Dmixing the ICE1712 chip

The config above does still not work (well) with an Audiophile card (ICE1712 chip) or a VIA VT82xx (snd-via82xx). The error is:

$ alsaplayer -d ossmix
error on set_format SND_PCM_FORMAT_S16_LE
Unavailable hw params:
ACCESS: RW_INTERLEAVED
FORMAT: S32_LE
SUBFORMAT: ALL
SAMPLE_BITS: ALL
FRAME_BITS: ALL
CHANNELS: 2
RATE: 48000
PERIOD_TIME: (21333 21334)
PERIOD_SIZE: 1024
PERIOD_BYTES: ALL
PERIODS: (6 7)
BUFFER_TIME: (136520 136521)
BUFFER_SIZE: 6553 <<<<====take a look here!!
BUFFER_BYTES: ALL
TICK_TIME: ALL

Jaroslav Kysela did know the reason: The max buffer size of the Delta card is "6553" as marked above. So we need to specify a buffer size in our dmix definition that is: a) a power of two and b) less than 6553. Here's the fixed OSS-dmix .asoundrc for a ICE1712 and VT82xx card:

# cat ~/.asoundrc
# our ICE1712 dmix:
pcm.ossmix {
    type dmix
    ipc_key 1024
    slave {
    pcm "hw:0,0"
        period_time 0
        period_size 1024
        buffer_size 4096 # buffer size < 6653, but pow(x, 2)
        rate 44100 # we want to play CDs only
        format S32_LE # needed in alsa 1.0.10 for some reason
    }
    bindings {
        0 0
        1 1
    }
}
# Everything shall be dmixed, so redefine "default":
pcm.!default {
    type plug
    slave.pcm "ossmix"
}
# OSS via aoss should d(mix)stroyed:
pcm.dsp0 {
    type plug
    slave.pcm "ossmix"
}
ctl.mixer0 {
    type hw
    card 0
}

Comment by Jan. I have an Audiophile (Ice1712) and the above .asoundrc does not change anything. I still get the same error message "error on set_format SND_PCM_FORMAT_S16_LE". As it seems i am the only one... If somone has the same problem please write (here).

Comment by Nils: Jan, as it was some time ago last time this page was updated I suppose that you probably have your sound working right now. If not, I think that the solution is to prefix your dmix device (ossmix) by "plug:" when using it. I have an M-Audio Delta 66 and that works for me except for the KDE4 multimedia settings configuration (using Phonon in Fedora 11) as Phonon doesn't seem to add the "plug:" prefix to the dmix device. I worked-around that problem by creating a .asoundrc file like this:

pcm.dmixplug {
  type plug
  slave {
    pcm "dmix"
  }
  hint {
    show on
    description "DMix"
  }
}

Like that you use this device without the "plug:" prefix:

aplay -D dmixplug yoursound.wav

Comment by Thomas De Schampheleire: I have a VIA VT82xxx chipset, but I found that a smaller buffer size in .asoundrc caused XMMS to hang on high load or IO-transfer (presumably because the buffer has run empty). Making the buffer size bigger resolved this problem, even above 6653. I currently have 16384 which works well.

buffer_size 16384

In XMMS I use ALSA output, with Audio device: "pcm.ossmix", and Mixer device: PCM.

8. Dmixing the Audiophile 192 (ice1724)

You can enjoy dmixed sound at a cool 96khz with this card. Here is my \~/.asoundrc file

pcm.!default {
   type plug
   slave.pcm "dmixer"
}

pcm.dmixer  {
   type dmix
   ipc_key 1024
   slave {
      pcm "hw:0,0"
      format S32_LE
      period_time 0
      period_size 1024
      buffer_size 8192

      rate 96000
   }
   bindings {
      0 0
      1 1
   }
}

ctl.dmixer {
   type hw
   card 0
   device 0
}
pcm.dsp {
    type plug
    slave.pcm "dmixer"     # use our new PCM here
}
ctl.mixer {
    type hw
    card 0
}

Dmixing with Amarok

For some strange reason, in Amarok 1.4.4 you will not be able to use ALSA as output stream with the Envy24HT chip. To overcome this and yet still enjoy dmixing. Use the aoss application available from HG

rsync -avz --delete rsync://alsa.alsa-project.org/hg/alsa-oss .

Then simply start Amarok with "aoss amarok" and choose either ALSA or Autodetect for the sound output. If you use the \~/.asoundrc file I provided above, you should be in blue skies for years to come. :)

Additional Info

Bob Rossi asked and Frank Barknecht responded with some useful hints and tips from the alsa-user mailing-list.

I am *really* interested to see if dmix is right for me. I have been setting up my sound system in Linux for 2 weeks now. I've got it to the point where I can play sounds and use the MIC. I really want to get all of the applications to work without using a sound daemon. I am frustrated with the documentation regarding dmix and so far, have only heard that people have not got dmix to do what I want it to. I am hoping that dmix is the solution to my sound daemon problems. My questions are...

Well, /dev/dsp is part of the old sound architecture on Unix, that has been implemented on Linux in the OSS kernel sound driver. ALSA is the new Linux sound system and it still provides backwards compatible /dev/dsp devices (as well as the other (OSS devices)? required for OSSEmulation). ALSA also provides its own devices or PCMs, which can be extended in various ways that in general are much more flexible than the OSS layer was. One of these devices is the dmix plugin. It enables sound mixing even on hardware that doesn't support this directly. BUT: to use ALSA dmix you have to use software that is ALSA enabled. You cannot use old OSS software with the ALSA devices, unless you convince this software to use ALSA instead of OSS. There are mainly two way to achieve this: 1) rewrite the software or 2) swindle and cheat the software to let it *think* that it uses OSS devices when it is in fact using an ALSA device. 2 is, what the aoss-library does.

So: No, dmix technically does not get between /dev/dsp and the sound driver, but aoss gets between the application and the real /dev/dsp and routes the audio data over to ALSA's dmix.

You should be able to set it up like I described.

Yes, that's the goal. It already works with a lot of software that directly uses ALSA. Depending on your needs you might want to take a look at JACK, too, if you're into low latency inter-application sound sharing.

With ALSA you shouldn't worry so much about the /dev devices, but more with the PCM definitions. You don't really need to think in terms of /dev/snd anymore. Better think in terms of plug devices like ```hw:0```, default or ```plughw:surround51.```

See below. It works.

Thanks for the answers, Frank.


Elsewhere on the list Frank suggested and Jaroslav Kysela added...

Normally (without hardware mixing) you cannot use /dev/dsp multiple times directly. Only ALSA's ```plug:dmix``` device supports stream mixing outside of sound servers. It should be possible to run OSS applications with the aoss wrapper on dmix, but somehow I couldn't get this to work yet. It's becoming a FAQ: You must set period size and buffer size to power of two for the dmix plugin, because OSS API does not allow other values.

Getting mplayer to work

Gstreamer and gst-plugins 0.8.2 aren't playing nice with the alsasink output element. However dmix mixing though esd works for now (with help from the quick and dirty HOWTO). Using gconf-editor, change osssink to esdsink in /system/gstreamer/0.8/default/audiosink.

pcm.!default {
    type plug
    slave.pcm "dmixer"
}
pcm.dsp0 {
    type plug
    slave.pcm "dmixer"
}
pcm.dmixer {
    type dmix
    ipc_key 1024
    slave {
        pcm "hw:0,0"
        period_time 0
        period_size 1024
        buffer_size 8192
        #periods 128
        rate 44100
     }
     bindings {
        0 0
        1 1
     }
}
ctl.mixer0 {
    type hw
    card 0
}

KDE

The easiest way to make KDE notifications work as well, together with other sound, go to Control Center -> Sounds & Multimedia -> System Notifications -> Player Settings -> Use external player. The player you specify should be able to play different kinds of sounds, e.g. mp3, wav, ogg, ... I currently use mplayer for this.

Still Hearing Stuttering?

I got pretty serious stuttering on mp3 playback in both amaroK via its builtin xine support, and JuK? via artsd, but increasing the buffer sizes in the "slave" block fixed it just fine:

slave {
    pcm "hw:0,0"
    period_time 0
    buffer_time 0
    period_size 2048 # jm: much smoother than 1024/8192!
    buffer_size 32768
    rate 44100
}

If this isn't harmful, I'd even go as far as suggesting that it should be part of the default "complex" example...

Dmix KDE(arts), ESD and SDL quick and dirty HOWTO

moved to separate page Dmix Kde - arts, ESD and SDL quick and dirty HOWTO. Always try to use the 'simple approach' of above. The 'complex approach' can lead to unexpected behaviour in some applications. The examples of that approach use card specific information that can be different. So try the simple approach, and then start playing with the complex.

Gentoo ALSA dmix setup

http://gentoo-wiki.com/HOWTO_ALSA_sound_mixer_aka_dmix

Comments

NEEDED: an example situation of how the dmix plugin solution can solve a problem that would otherwise be difficult or impossible any other way ?

(Csan: For example: using quake3 and teamspeak2 at the same time is still not possible with snd-intel8x0, ALSA 0.9.8 and .asoundrc set up as instructed below. Starting up teamspeak with aoss ./teamspeak is fine but running quake3 using aoss quake3 afterwards still hangs at "sound initialization" until teamspeak exits.

This is because aoss does not support libc's fopen() call which is utilized by teamspeak. See aoss page for possible fix. The latest version of aoss does support fopen. Playing back more music files simultaneously works, though.) I heard this can be solved by altering some /proc settings for the OSS playback, check this link:

http://www.linux-gamers.net/modules/wfsection/article.php?articleid=34

Also, TakashiIwai comes to the rescue with a new plugin that lets the user combine dmix/dsnoop into a single PCM device which can then be used via the aoss script.. More info when it's finished and in cvs.. It is in 1.0.2 and it's called asym. See the page.

Question: I have the setup as shown in #6 Complex example above, and it works great so far. Just tested with xmms and alsaplayer and it works great. How can I get any arbitrary application to work with it? --Dave

Multi -> Dmix support(?)

From what I can see, one can not send output from the multi-plugin to the dmix-plugin. This would be convenient, for certain setups. Am I wrong?

Can alsa input be dmixed? If can't do so, for example in glame:

I use ```plughw:0,0``` for input, while default (dmixed) for output. When I record a music, the input and output didn't co-operate well, there's from 100ms to 2s delay after I speak something. :-(

In some config examples redefined ctl.!default instead ctl.mixer0. What's better?

This may be a stupid question, but _why_ does the plugin not work transparently? Since ALSA "gets between" the software and the sound hardware, should it not be possible to arrange so that everyone who wants output has it routed through software mixing first? This fiddling with output devices seems kind of pointless to achieve what ought to be a very simple goal.

For ALSA 1.0.9rc2 and higher you don't need to setup dmix. Dmix is enabled as default for soundcard which doesn't support hw mixing.

Does that mean one can drop the "-d plug:dmix" part? Is this supposed to work?

alsaplayer -o alsa some.mp3

dmix & surround sound

I wanted to play a DVD with xine through the dmix plugin, with all 6 channels, and having other ALSA-aware applications mixed in (regardless of channel count).

I wrote this in my /etc/asound.conf:

pcm.dmixs51 {
    type dmix
    ipc_key 1024
    ipc_key_add_uid false # let multiple users share
    ipc_perm 0660 # IPC permissions (octal, default 0600)
    slave {
        pcm "hw:0,1" # see below
        rate 48000
        channels 6
        period_time 0
        period_size 1024
        buffer_time 0
        buffer_size 4096
    }
}

pcm.asym51 {
           type asym
           playback.pcm "dmixs51" 
           capture.pcm "hw:0,0" # this might be "dsnoop:0"
}

pcm.dsp0 {
         type plug
         slave.pcm "asym51"
}

The "hw:0,1" in the dmixs51 slave section is because I have a C-Media PCI CMI8738-MC6, that uses the second device for surround output.

Then, I configured xine to use plug:dmixs51 for every output, with these directives (excerpt from ~/.xine/config):

audio.device.alsa_default_device:plug:dmixs51
audio.device.alsa_front_device:plug:dmixs51
audio.device.alsa_surround40_device:plug:dmixs51
audio.device.alsa_surround51_device:plug:dmixs51

I also told mpd to use that device (excerpt from /etc/mpd.conf):

ao_driver      "alsa09"
ao_driver_options   "dev=plug:dmixs51"

The only thing I could not get to work in this setup is Skype, but I think it is Skype's fault. They say an ALSA-aware version is in the works, let's just hope.

dakkar

I have mixed the concepts and constructed a .asoundrc which has a default device upmixing Stereo to 5.1 and then putting it into a 5.1 dmix and then to the snd_card. Mine works on nForce2. Check it out, hope it helps. Works perfectly for me with MPlayer, XMMS, xine, aoss mpg123, etc.:

----------- Stereo to 5.1 Upmixing, dmixing .asoundrc ---------------
pcm.snd_card {
     type hw
     card 0 # change to your cards number or name
}

# 6 channel dmix:
pcm.dmix6 {
     type dmix
        ipc_key 1024
        ipc_key_add_uid false # let multiple users share
        ipc_perm 0660 # IPC permissions (octal, default 0600)
        slave {
                pcm snd_card # see below
                rate 48000
                channels 6
                period_time 0
                period_size 1024 # try 2048 against skipping
                buffer_time 0
                buffer_size 5120 # in case of problems reduce this
                                 # in case of skipping, try increasing
        }
     }

# upmixing: 
pcm.ch51dup {
        type route
        slave.pcm dmix6
        slave.channels 6
        ttable.0.0 1
        ttable.1.1 1
        ttable.0.2 1
        ttable.1.3 1
        ttable.0.4 0.5
        ttable.1.4 0.5
        ttable.0.5 0.5
        ttable.1.5 0.5
   }

pcm.duplex {
     type asym
     playback.pcm "ch51dup" # upmix first
#     playback.pcm "dmix6"  # just pass to 6 channel dmix
#     capture.pcm "dsnoop:0" # doesn't work for me
     capture.pcm "snd_card"
}

# change default device:
pcm.!default {
     type plug
     slave.pcm "duplex"
}

# for aoss
pcm.dsp "duplex"

pcm.dsp1 "duplex"

DJtm

-- Hi, nice try. But I'm getting only "ALSA lib pcm_params.c:2152:(snd_pcm_hw_refine_slave) Slave PCM not usable". It's just working, when I use "slave.pcm "surround51"" and do not pass everything to dmix at the ch51dup section. But this isn't desirable, isn't it? Since the blocking soundcard problem got not solved, it's only a workaround for a simplier, but also blocking configuration I had before. Any suggestions would be welcome; I'm using ALSA 1.0.10-ubuntu4 with a ca0106 chip.

Dmix Surround7 and more

I needed my soundcard to have different dmixed channels so i could direct my music player and everything on the desktop to my speakers, and skype to my headset, and myth-frontend to my TV. Now i have dmixed and stereo on the channels i need, setup as diffent "sound cards" for each application to use.

pcm.hda_intel {
    type hw
    card 0
}

pcm.dmixer {
   type dmix
   ipc_key 1099
   ipc_perm 0660
   slave {
       pcm hda_intel
       rate 48000
       channels 8
       periods 128
       period_time 0
       period_size 1024
       buffer_size 4096
   }
}

pcm.dsnooper {
   type dsnoop
   ipc_key 1098
   ipc_perm 0660
   slave {
       pcm hda_intel
       rate 44100
       periods 128
       period_time 0
       period_size 1024
       buffer_size 4096
   }
   bindings {
       0 0
       1 1
   }
   slowptr true
}

pcm.duplexasym {
   type asym
   playback.pcm "dmixer"  # playback on all speakers
   capture.pcm "dsnooper" # recording 
}

pcm.duplex { # plug and play on all speakers
   type plug
   slave.pcm "duplexasym"
}

pcm.!default {
   type plug
   slave.pcm "duplexasym"
}

ctl.duplex {
   type hw
   card 0
}

pcm.headset { # playback only on frontpanel headset
 type route
 slave.pcm dmixer
 slave.channels 8
 ttable.0.0 1 # headphones front L
 ttable.1.1 1 # headphones front R
}

pcm.speakers { # playback only on desktop speakers
 type route
 slave.pcm dmixer
 slave.channels 8
 ttable.0.4 1 # speakers L
 ttable.1.5 1 # speakers R
}

pcm.TV { # playback only on my TV (using Side)
 type route
 slave.pcm dmixer
 slave.channels 8
 ttable.0.6 1 # TV L
 ttable.1.7 1 # TV R
}

DocNielsen

Does dmix affect sound quality?

A question more than a comment: how does dmix affect sound quality? I use a Chaintech AV710, which uses the snd-ice1724 driver. After it was set to use dmix by default by an upstream ALSA, I noticed that alsamixer was showing sample rate 48000 while I was playing back a CD (all CDs are 44100). If I forced it to 44100 with rate locking so nothing could automatically change it and played the CD, it was obviously wrong (too slow). This suggests to me that something in the chain is forcing audio played through dmix to be resampled to 48KHz, which is not optimal for sound quality. When I commented out the parts of ICE1724.conf that enable dmix, I found this no longer happens - alsamixer now displays 44100 when I'm playing a CD. So for audio geeks like me who deeply care about audio quality, is dmix not advisable?


Answer:

Dmix by default uses 48kHz sample rate. So, if your source is 44.1kHz, it will be upsampled to 48khz. If you want pure 44.1kHz output, you should set the rate parameter:

pcm.swmixer {
    type dmix
    ipc_key 1234
    slave {
        pcm "hw:0,0" #for ICE1724's analog output
        #pcm "hw:0,1" # for ICE1724's digital output
        #format S32_LE # needed only for ICE1724's digital output.
        period_size 1024
        buffer_size 4096
        rate 44100
    }
}

Also, you should keep in mind, that software mixing is possible only, if all signals are at the same sample rate. So, if you want to play DVD's, which use 48kHz sample rate for audio, with this kind of plug, everything will be downsampled to 44.1kHz. Unfortunatelly it does'nt work wery well (there are some syncronization problems with some players). In this case you have two opportunities, either use a separate plug (and loose ability to playback mp3 and dvd on the same time) or tell your DVD player to sample audio down to 44.1khz, for example:

mplayer -af resample=44100

Saul: Also a question more than a comment:

Can dmix be used within a single program to effectively mix different "voices"? Does its latency permit scheduling with (at least) single-frame accuracy (e.g., I start VoiceA and wish to setup and schedule VoiceB to be "added to the mix" at a certain point)? If someone knows how to do this, some guidance or a nudge in the right direction would be greatly appreciated.


Answer: Latency depends on period and buffer size. AFAIK dmix is not designed as sample-accurate (because it is usually not needed on general purpose audio system and will make the design quite complex). If you need sample-accurate and fully syncronized mixer/router, you should consider using the Jack http://jackaudio.org/. Large number of Linux audio applications, where sample-accurate mixing is needed, support Jack.

On the other hand, overall latency of the audio system will still mostly depend on period size. As small as your period size can be, as low latecy you will get. Of course, low latency needs good task scheduler in kernel. If you start getting xruns on low latency audio setup, you should use some low-latency/realtime kernel patches. I reccomend Ingo Molnar's realtime preempt http://people.redhat.com/mingo/realtime-preempt/ or Con Colivas patch set http://members.optusnet.com.au/ckolivas/kernel/.

Retrieved from "http://alsa.opensrc.org/Dmix"

Categories: ALSA plugins | Configuration

GITHUB | EDIT