ALSA configuration
Modifying the default configuration
ALSA devices are created and managed by the session manager with the alsa.lua
monitor script. In the default configuration, this script is loaded by
main.lua.d/30-alsa-monitor.lua
, which also specifies an alsa_monitor
global table that can be filled in with properties and rules in subsequent
config files. By default, these are filled in main.lua.d/50-alsa-config.lua
.
The alsa_monitor
global table has 2 sub-tables:
alsa_monitor.properties
This is a simple Lua table that has key value pairs used as properties.
Example:
alsa_monitor.properties = { ["alsa.jack-device"] = false, ["alsa.reserve"] = true, }
The above example will configure the ALSA monitor to not enable the JACK device, and do ALSA device reservation using the mentioned DBus interface.
A list of valid properties are:
["alsa.jack-device"] = false
Creates a JACK device if set to
true
. This is not enabled by default because it requires that the PipeWire JACK replacement libraries are not used by the session manager, in order to be able to connect to the real JACK server.["alsa.reserve"] = true
Reserve ALSA devices via org.freedesktop.ReserveDevice1 on D-Bus.
["alsa.reserve.priority"] = -20
The used ALSA device reservation priority.
["alsa.reserve.application-name"] = "WirePlumber"
The used ALSA device reservation application name.
alsa_monitor.rules
This is a Lua array that can contain objects with rules for a device or node. Those objects have 2 properties. The first one is
matches
, which allow users to define rules to match a device or node. The second property isapply_properties
, and it is used to apply properties on the matched object.Example:
alsa_monitor.rules = { matches = { { { "device.name", "matches", "alsa_card.*" }, }, }, apply_properties = { ["api.alsa.use-acp"] = true, } }
This sets the API ALSA use ACP property to all devices with a name that matches the
alsa_card.*
pattern.The
matches
section is an array of arrays. On the first level, the rules are ORed together, so any rule match is going to apply the properties. On the second level, the rules are merged with AND, so they must all match.Example:
matches = { { { "node.name", "matches", "alsa_input.*" }, { "alsa.driver_name", "equals", "snd_hda_intel" }, }, { { "node.name", "matches", "alsa_output.*" }, }, },
This is equivalent to the following logic, in pseudocode:
if ("node.name" MATCHES "alsa_input.*" AND "alsa.driver_name" EQUALS "snd_hda_intel" ) OR ("node.name" MATCHES "alsa_output.*") then ... apply the properties ... end
As you can notice, the individual rules are themselves also lua arrays. The first element is a property name (ex “node.name”), the second element is a verb and the third element is an expected value, which depends on the verb. Internally, this uses the
Constraint
API, which is documented in the Object Interet API section. All the verbs that you can use onConstraint
are also allowed here.Note
When using the “matches” verb, the values are not complete regular expressions. They are wildcard patterns, which means that ‘*’ matches an arbitrary, possibly empty, string and ‘?’ matches an arbitrary character.
All the possible properties that you can apply to devices and nodes of the ALSA monitor are described in the sections below.
Device properties
PipeWire devices correspond to the ALSA cards. The following properties can be configured on devices created by the monitor:
["api.alsa.use-acp"] = true
Use the ACP (alsa card profile) code to manage the device. This will probe the device and configure the available profiles, ports and mixer settings. The code to do this is taken directly from PulseAudio and provides devices that look and feel exactly like the PulseAudio devices.
["api.alsa.use-ucm"] = true
By default, the UCM configuration is used when it is available for your device. With this option you can disable this and use the ACP profiles instead.
["api.alsa.soft-mixer"] = false
Setting this option to true will disable the hardware mixer for volume control and mute. All volume handling will then use software volume and mute, leaving the hardware mixer untouched. The hardware mixer will still be used to mute unused audio paths in the device.
["api.alsa.ignore-dB"] = false
Setting this option to true will ignore the decibel setting configured by the driver. Use this when the driver reports wrong settings.
["device.profile-set"] = "profileset-name"
This option can be used to select a custom profile set name for the device. Usually this is configured in Udev rules but it can also be specified here.
["device.profile"] = "default profile name"
The default active profile name.
["api.acp.auto-profile"] = false
Automatically select the best profile for the device. Normally this option is disabled because the session manager will manage the profile of the device. The session manager can save and load previously selected profiles. Enable this if your session manager does not handle this feature.
["api.acp.auto-port"] = false
Automatically select the highest priority port that is available. This is by default disabled because the session manager handles the task of selecting and restoring ports. It can, for example, restore previously saved volumes. Enable this here when the session manager does not handle port restore.
Some of the other properties that might be configured on devices:
["device.nick"] = "My Device",
["device.description"] = "My Device"
device.description
will show up in most apps when a device name is shown.
Node Properties
Nodes are sinks or sources in the PipeWire graph. They correspond to the ALSA devices. In addition to the generic stream node configuration options, there are some alsa specific options as well:
["priority.driver"] = 2000
This configures the node driver priority. Nodes with higher priority will be used as a driver in the graph. Other nodes with lower priority will have to resample to the driver node when they are joined in the same graph. The default value is set based on some heuristics.
["priority.session"] = 1200
This configures the priority of the node when selecting a default node. Higher priority nodes will be more likely candidates as a default node.
Note
By default, sources have a priority.session
value around 1600-2000 and
sinks have a value around 600-1000. If you are increasing the priority of a
sink, it is not advised to use a value higher than 1500, as it may cause
a sink’s monitor to be selected as a default source.
["node.pause-on-idle"] = false
Pause-on-idle will stop the node when nothing is linked to it anymore. This is by default false because some devices cause a pop when they are opened/closed. The node will, normally, pause and suspend after a timeout (see suspend-node.lua).
["session.suspend-timeout-seconds"] = 5 -- 0 disables suspend
This option configures a different suspend timeout on the node. By default this is 5 seconds. For some devices (HiFi amplifiers, for example) it might make sense to set a higher timeout because they might require some time to restart after being idle.
A value of 0 disables suspend for a node and will leave the ALSA device busy.
The device can then manually be suspended with pactl suspend-sink|source
.
The following properties can be used to configure the format used by the ALSA device:
["audio.format"] = "S16LE"
By default, PipeWire will use a 32 bits sample format but a different format can be set here.
The Audio rate of a device can be set here:
["audio.rate"] = 44100
By default, the ALSA device will be configured with the same samplerate as the global graph. If this is not supported, or a custom values is set here, resampling will be used to match the graph rate.
["audio.channels"] = 2
["audio.position"] = "FL,FR"
By default the channels and their position are determined by the selected Device profile. You can override this setting here and optionally swap or reconfigure the channel positions.
["api.alsa.use-chmap"] = false
Use the channel map as reported by the driver. This is disabled by default because it is often wrong and the ACP code handles this better.
["api.alsa.disable-mmap"] = true
PipeWire will by default access the memory of the device using mmap. This can be disabled and force the usage of the slower read and write access modes in case the mmap support of the device is not working properly.
["channelmix.normalize"] = true
Makes sure that during such mixing & resampling original 0 dB level is preserved, so nothing sounds wildly quieter/louder.
["channelmix.mix-lfe"] = true
Creates “center” channel for X.0 recordings from front stereo on X.1 setups and pushes some low-frequency/bass from “center” from X.1 recordings into front stereo on X.0 setups.
["monitor.channel-volumes"] = false
By default, the volume of the sink/source does not influence the volume on the monitor ports. Set this option to true to change this. PulseAudio has inconsistent behaviour regarding this option, it applies channel-volumes only when the sink/source is using software volumes.
ALSA buffer properties
PipeWire uses a timer to consume and produce samples to/from ALSA devices. After every timeout, it queries the device hardware pointers of the device and uses this information to set a new timeout. See also this example program.
By default, PipeWire handles ALSA batch devices differently from non-batch devices. Batch devices only get their hardware pointers updated after each hardware interrupt. Non-batch devices get updates independent of the interrupt. This means that for batch devices we need to set the interrupt at a sufficiently high frequency (at the cost of CPU usage) while for non-batch devices we want to set the interrupt frequency as low as possible (to save CPU).
For batch devices we also need to take the extra buffering into account caused by the delayed updates of the hardware pointers.
Most USB devices are batch devices and will be handled as such by PipeWire by default.
There are 2 tunable parameters to control the buffering and timeouts in a device
["api.alsa.period-size"] = 1024
This sets the device interrupt to every period-size samples for non-batch devices and to half of this for batch devices. For batch devices, the other half of the period-size is used as extra buffering to compensate for the delayed update. So, for batch devices, there is an additional period-size/2 delay. It makes sense to lower the period-size for batch devices to reduce this delay.
["api.alsa.headroom"] = 0
This adds extra delay between the hardware pointers and software pointers. In most cases this can be set to 0. For very bad devices or emulated devices (like in a VM) it might be necessary to increase the headroom value. In summary, this is the overview of buffering and timings:
Property
Batch
Non-Batch
IRQ Frequency
api.alsa.period-size/2
api.alsa.period-size
Extra Delay
api.alsa.headroom + api.alsa.period-size/2
api.alsa.headroom
It is possible to disable the batch device tweaks with:
["api.alsa.disable-batch"] = true
It removes the extra delay added of period-size/2 if the device can support this. For batch devices it is also a good idea to lower the period-size (and increase the IRQ frequency) to get smaller batch updates and lower latency.
ALSA extra latency properties
Extra internal delay in the DAC and ADC converters of the device itself can be
set with the latency.internal.*
properties:
["latency.internal.rate"] = 256
["latency.internal.ns"] = 0
You can configure a latency in samples (relative to rate with
latency.internal.rate
) or in nanoseconds (latency.internal.ns
).
This value will be added to the total reported latency by the node of the device.
You can use a tool like jack_iodelay
to get the number of samples of
internal latency of your device.
This property is also adjustable at runtime with the ProcessLatency
param.
You will need to find the id of the Node you want to change. For example:
Query the current internal latency of an ALSA node with id 58:
$ pw-cli e 58 ProcessLatency
Object: size 80, type Spa:Pod:Object:Param:ProcessLatency (262156), id Spa:Enum:ParamId:ProcessLatency (16)
Prop: key Spa:Pod:Object:Param:ProcessLatency:quantum (1), flags 00000000
Float 0.000000
Prop: key Spa:Pod:Object:Param:ProcessLatency:rate (2), flags 00000000
Int 0
Prop: key Spa:Pod:Object:Param:ProcessLatency:ns (3), flags 00000000
Long 0
Set the internal latency to 256 samples:
$ pw-cli s 58 ProcessLatency '{ rate = 256 }'
Object: size 32, type Spa:Pod:Object:Param:ProcessLatency (262156), id Spa:Enum:ParamId:ProcessLatency (16)
Prop: key Spa:Pod:Object:Param:ProcessLatency:rate (2), flags 00000000
Int 256
remote 0 node 58 changed
remote 0 port 70 changed
remote 0 port 72 changed
remote 0 port 74 changed
remote 0 port 76 changed
Startup tweaks
Some devices need some time before they can report accurate hardware pointer positions. In those cases, an extra start delay can be added that is used to compensate for this startup delay:
["api.alsa.start-delay"] = 0
It is unsure when this tunable should be used.
IEC958 (S/PDIF) passthrough
S/PDIF passthrough will only be enabled when the accepted codecs are configured on the ALSA device.
This can be done in 3 different ways:
Use pavucontrol and toggle the codecs in the output advanced section
Modify the
["iec958.codecs"] = "[ PCM DTS AC3 MPEG MPEG2-AAC EAC3 TrueHD DTS-HD ]"
node property to something.Use
pw-cli s <node-id> Props '{ iec958Codecs : [ PCM ] }'
to modify the codecs at runtime.