Add per-laptop speaker tunings, starting with the XPS 14
Laptop speakers ship voiced by the vendor's Windows DSP layer, which Linux does not get. A tuning restores that as a PipeWire filter-chain in front of the internal speaker sink, matched to the machine by DMI string and expected sink. Adding a laptop is a directory under default/audio/tunings with two files and no new code: matching is data. The XPS 14 DA14260 tuning included here was derived by measuring the xps-audio-linux EasyEffects profile (MIT) and fitting a biquad chain to it, so no impulse response or other upstream asset is redistributed. It measures 1.24 dB RMS against that reference, and matches its dynamic range within 0.1 LU -- the reference's multiband compressor turned out to contribute nothing, so a linear chain replaces it. Bass Q is capped deliberately: a closer magnitude fit swung group delay 31 ms across 63-80 Hz, which smears bass transients. The graph runs as its own PipeWire client under its own config name rather than loading into the audio daemon. The daemon only reads its config at startup, so a daemon-loaded tuning could only be switched by restarting PipeWire -- which drops every PulseAudio client's connection, and applications that do not reconnect (Spotify) then have to be restarted by hand. Hosting it separately also contains failure, since a malformed tuning breaks only that service. Three things about the surrounding audio graph needed fixing for this to behave: - Volume must live downstream of the tuning. omarchy-audio-output-sink is now the single definition of which sink an output's volume really uses, shared by the volume keys, the output switcher's OSD and the audio panel, so they cannot disagree. It resolves the current default output, which keeps it correct when headphones are selected while a tuning exists. - The tuning's own output is a movable sink input, so rerouting "all streams" to a newly selected output would drag the processing onto headphones, or into the tuning's own sink, which is a cycle. It is pinned, and stream moves are limited to streams carrying an application.name. - The physical sink a tuning fronts is not independently selectable, since picking it would only bypass the tuning, so it is kept out of the output list. Applying happens at first-run, not finalize-user, because finalize-user also runs in the ISO chroot where there is no audio server and nothing would retry. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
David Heinemeier Hansson
co-authored by
Claude Opus 5
Claude Opus 5
parent
248659de5a
commit
aa9f0c54c5
@@ -0,0 +1,40 @@
|
||||
# Host config for the Omarchy speaker tuning.
|
||||
#
|
||||
# This exists so the tuning gets its own PipeWire client rather than sharing
|
||||
# PipeWire's stock filter-chain.conf. That config merges every fragment in
|
||||
# ~/.config/pipewire/filter-chain.conf.d/, so hosting the tuning there would load
|
||||
# any unrelated filter a user keeps in that directory -- duplicating filters
|
||||
# already hosted elsewhere, and stopping them all when the tuning is switched off.
|
||||
#
|
||||
# Installed as ~/.config/pipewire/omarchy-speaker-tuning.conf with the tuning
|
||||
# graph merged from omarchy-speaker-tuning.conf.d/, and run with
|
||||
# pipewire -c omarchy-speaker-tuning.conf
|
||||
#
|
||||
# The contents are the minimum a filter-hosting client needs, taken from
|
||||
# /usr/share/pipewire/filter-chain.conf.
|
||||
|
||||
context.properties = {
|
||||
log.level = 0
|
||||
}
|
||||
|
||||
context.spa-libs = {
|
||||
audio.convert.* = audioconvert/libspa-audioconvert
|
||||
support.* = support/libspa-support
|
||||
}
|
||||
|
||||
context.modules = [
|
||||
# Boost the audio thread priority.
|
||||
{ name = libpipewire-module-rt
|
||||
args = { }
|
||||
flags = [ ifexists nofail ]
|
||||
}
|
||||
|
||||
# The native communication protocol.
|
||||
{ name = libpipewire-module-protocol-native }
|
||||
|
||||
# Lets this process provide nodes to PipeWire.
|
||||
{ name = libpipewire-module-client-node }
|
||||
|
||||
# Wraps nodes in an adapter with a converter and resampler.
|
||||
{ name = libpipewire-module-adapter }
|
||||
]
|
||||
@@ -0,0 +1,137 @@
|
||||
# Dell XPS 14 DA14260 speaker tuning.
|
||||
#
|
||||
# Biquad chain fitted to the measured response of the xps-audio-linux EasyEffects
|
||||
# profile under a dense pink-weighted multitone of 104 bin-aligned tones,
|
||||
# followed by a lookahead limiter. Measures 1.24 dB RMS against that reference
|
||||
# (0.97 dB weighted over the fit's own error metric).
|
||||
#
|
||||
# Q below 200 Hz is capped at 1.8 on purpose. A closer magnitude fit is possible
|
||||
# with high-Q sections, but the reference produces its narrow bass features by
|
||||
# convolution, and reproducing them with high-Q biquads swung group delay 31 ms
|
||||
# across 63-80 Hz, which smears bass transients. The cap costs 0.33 dB and
|
||||
# halves the swing.
|
||||
#
|
||||
# This is a plain filter-chain sink rather than a WirePlumber smart filter. A
|
||||
# smart filter is the better shape -- it would leave the real device as the
|
||||
# default output instead of adding a second one -- but on PipeWire 1.6.8 /
|
||||
# WirePlumber 0.5.15 this graph loads and links correctly as a smart filter and
|
||||
# then passes audio through unprocessed: its controls are present and
|
||||
# mpv -> filter -> sink links are made, yet the filter's input monitor and the
|
||||
# speaker sink's monitor measure identically. Revisit when that is understood.
|
||||
#
|
||||
# Channels are wired explicitly because the limiter is a stereo plugin; a mono
|
||||
# graph is duplicated per channel and would limit each side independently,
|
||||
# shifting the stereo image on bass transients.
|
||||
|
||||
context.modules = [
|
||||
{ name = libpipewire-module-filter-chain
|
||||
args = {
|
||||
node.description = "Laptop Speakers"
|
||||
media.name = "Laptop Speakers"
|
||||
|
||||
filter.graph = {
|
||||
nodes = [
|
||||
{ type = builtin name = s0_l label = bq_highpass control = { "Freq" = 60.9 "Q" = 1.0 } }
|
||||
{ type = builtin name = s1_l label = bq_highpass control = { "Freq" = 60.9 "Q" = 1.0 } }
|
||||
{ type = builtin name = s2_l label = bq_peaking control = { "Freq" = 83.4 "Q" = 1.8 "Gain" = -8.0 } }
|
||||
{ type = builtin name = s3_l label = bq_peaking control = { "Freq" = 100.4 "Q" = 1.59 "Gain" = 7.47 } }
|
||||
{ type = builtin name = s4_l label = bq_peaking control = { "Freq" = 250.5 "Q" = 2.966 "Gain" = -4.7 } }
|
||||
{ type = builtin name = s5_l label = bq_peaking control = { "Freq" = 419.8 "Q" = 3.0 "Gain" = -5.83 } }
|
||||
{ type = builtin name = s6_l label = bq_peaking control = { "Freq" = 631.3 "Q" = 2.515 "Gain" = -10.33 } }
|
||||
{ type = builtin name = s7_l label = bq_peaking control = { "Freq" = 894.4 "Q" = 4.0 "Gain" = -2.42 } }
|
||||
{ type = builtin name = s8_l label = bq_peaking control = { "Freq" = 1355.7 "Q" = 2.884 "Gain" = 6.92 } }
|
||||
{ type = builtin name = s9_l label = bq_peaking control = { "Freq" = 1707.2 "Q" = 1.311 "Gain" = -6.54 } }
|
||||
{ type = builtin name = s10_l label = bq_peaking control = { "Freq" = 3100.0 "Q" = 0.5 "Gain" = -10.09 } }
|
||||
{ type = builtin name = s11_l label = bq_peaking control = { "Freq" = 3200.0 "Q" = 1.048 "Gain" = 3.09 } }
|
||||
{ type = builtin name = s12_l label = bq_highshelf control = { "Freq" = 6015.2 "Q" = 1.5 "Gain" = -1.34 } }
|
||||
|
||||
{ type = builtin name = s0_r label = bq_highpass control = { "Freq" = 60.9 "Q" = 1.0 } }
|
||||
{ type = builtin name = s1_r label = bq_highpass control = { "Freq" = 60.9 "Q" = 1.0 } }
|
||||
{ type = builtin name = s2_r label = bq_peaking control = { "Freq" = 83.4 "Q" = 1.8 "Gain" = -8.0 } }
|
||||
{ type = builtin name = s3_r label = bq_peaking control = { "Freq" = 100.4 "Q" = 1.59 "Gain" = 7.47 } }
|
||||
{ type = builtin name = s4_r label = bq_peaking control = { "Freq" = 250.5 "Q" = 2.966 "Gain" = -4.7 } }
|
||||
{ type = builtin name = s5_r label = bq_peaking control = { "Freq" = 419.8 "Q" = 3.0 "Gain" = -5.83 } }
|
||||
{ type = builtin name = s6_r label = bq_peaking control = { "Freq" = 631.3 "Q" = 2.515 "Gain" = -10.33 } }
|
||||
{ type = builtin name = s7_r label = bq_peaking control = { "Freq" = 894.4 "Q" = 4.0 "Gain" = -2.42 } }
|
||||
{ type = builtin name = s8_r label = bq_peaking control = { "Freq" = 1355.7 "Q" = 2.884 "Gain" = 6.92 } }
|
||||
{ type = builtin name = s9_r label = bq_peaking control = { "Freq" = 1707.2 "Q" = 1.311 "Gain" = -6.54 } }
|
||||
{ type = builtin name = s10_r label = bq_peaking control = { "Freq" = 3100.0 "Q" = 0.5 "Gain" = -10.09 } }
|
||||
{ type = builtin name = s11_r label = bq_peaking control = { "Freq" = 3200.0 "Q" = 1.048 "Gain" = 3.09 } }
|
||||
{ type = builtin name = s12_r label = bq_highshelf control = { "Freq" = 6015.2 "Q" = 1.5 "Gain" = -1.34 } }
|
||||
{ type = lv2
|
||||
name = limiter
|
||||
plugin = "http://lsp-plug.in/plugins/lv2/limiter_stereo"
|
||||
control = {
|
||||
# Both default to enabled: "alr" regulates level toward the
|
||||
# threshold and "boost" normalises the threshold up to full
|
||||
# scale. A fixed tuning must switch them off or its tone drifts
|
||||
# with programme level.
|
||||
"alr" = 0
|
||||
"boost" = 0
|
||||
"g_in" = 0.5456
|
||||
"th" = 0.891
|
||||
}
|
||||
}
|
||||
]
|
||||
|
||||
links = [
|
||||
{ output = "s0_l:Out" input = "s1_l:In" }
|
||||
{ output = "s1_l:Out" input = "s2_l:In" }
|
||||
{ output = "s2_l:Out" input = "s3_l:In" }
|
||||
{ output = "s3_l:Out" input = "s4_l:In" }
|
||||
{ output = "s4_l:Out" input = "s5_l:In" }
|
||||
{ output = "s5_l:Out" input = "s6_l:In" }
|
||||
{ output = "s6_l:Out" input = "s7_l:In" }
|
||||
{ output = "s7_l:Out" input = "s8_l:In" }
|
||||
{ output = "s8_l:Out" input = "s9_l:In" }
|
||||
{ output = "s9_l:Out" input = "s10_l:In" }
|
||||
{ output = "s10_l:Out" input = "s11_l:In" }
|
||||
{ output = "s11_l:Out" input = "s12_l:In" }
|
||||
{ output = "s12_l:Out" input = "limiter:in_l" }
|
||||
{ output = "s0_r:Out" input = "s1_r:In" }
|
||||
{ output = "s1_r:Out" input = "s2_r:In" }
|
||||
{ output = "s2_r:Out" input = "s3_r:In" }
|
||||
{ output = "s3_r:Out" input = "s4_r:In" }
|
||||
{ output = "s4_r:Out" input = "s5_r:In" }
|
||||
{ output = "s5_r:Out" input = "s6_r:In" }
|
||||
{ output = "s6_r:Out" input = "s7_r:In" }
|
||||
{ output = "s7_r:Out" input = "s8_r:In" }
|
||||
{ output = "s8_r:Out" input = "s9_r:In" }
|
||||
{ output = "s9_r:Out" input = "s10_r:In" }
|
||||
{ output = "s10_r:Out" input = "s11_r:In" }
|
||||
{ output = "s11_r:Out" input = "s12_r:In" }
|
||||
{ output = "s12_r:Out" input = "limiter:in_r" }
|
||||
]
|
||||
|
||||
inputs = [ "s0_l:In" "s0_r:In" ]
|
||||
outputs = [ "limiter:out_l" "limiter:out_r" ]
|
||||
}
|
||||
|
||||
audio.channels = 2
|
||||
audio.position = [ FL FR ]
|
||||
|
||||
capture.props = {
|
||||
node.name = "omarchy_speaker_tuning"
|
||||
media.class = Audio/Sink
|
||||
}
|
||||
playback.props = {
|
||||
node.name = "omarchy_speaker_tuning_output"
|
||||
node.passive = true
|
||||
target.object = "@SPEAKER_SINK@"
|
||||
# This stream is the filter's output and is a movable sink input like any
|
||||
# other, so anything that reroutes "all streams" to a newly selected
|
||||
# output would drag the processing along with it -- onto headphones, or
|
||||
# into the tuning's own sink, which is a cycle. Pin it.
|
||||
node.dont-move = true
|
||||
# If the speaker sink is not present yet -- the tuning host can start
|
||||
# before the device is discovered -- WirePlumber would otherwise link this
|
||||
# output to whatever default exists, quietly tuning the wrong device while
|
||||
# the tuning sink still looks healthy. Wait for the named target instead.
|
||||
# Both are needed: without linger, WirePlumber destroys the node rather
|
||||
# than waiting (see its scripts/linking/find-defined-target.lua).
|
||||
node.dont-fallback = true
|
||||
node.linger = true
|
||||
}
|
||||
}
|
||||
}
|
||||
]
|
||||
@@ -0,0 +1,31 @@
|
||||
## Dell XPS 14 DA14260 internal speakers.
|
||||
##
|
||||
## Thirteen biquads and a lookahead limiter, applied as a PipeWire filter-chain
|
||||
## in front of the internal speaker sink. The stock Linux path already loads
|
||||
## Dell's Cirrus smart-amplifier firmware; this adds the perceptual voicing the
|
||||
## Windows Waves layer provides and Linux does not.
|
||||
|
||||
description="Dell XPS 14 (2026) speakers"
|
||||
## Matched on the DMI product name plus the presence of the sink below, which
|
||||
## together are specific enough that no per-model predicate script is needed.
|
||||
## Hardware needing a sharper test can set match_command to any predicate.
|
||||
match_dmi="XPS 14 DA14260"
|
||||
## Unescaped dots: this is passed to awk as a string, where a backslash escape
|
||||
## would be consumed before the regex sees it.
|
||||
sink_pattern='^alsa_output.*sof_sdw.*HiFi__Speaker__sink$'
|
||||
|
||||
## Provenance. Derived by measuring the response of the xps-clone EasyEffects
|
||||
## profile from https://github.com/spencerbull/xps-audio-linux (MIT) and fitting
|
||||
## a biquad chain to it. No upstream asset is redistributed: the convolution
|
||||
## impulse response is not carried, so this tuning has no binary blob and is
|
||||
## sample-rate agnostic.
|
||||
derived_from="xps-audio-linux xps-clone (MIT, spencerbull)"
|
||||
validated_by="dhh"
|
||||
validated_on="2026-07-24"
|
||||
|
||||
## Measured against that reference under a dense pink-weighted multitone of 104
|
||||
## bin-aligned tones. See docs/AUDIO-TUNING.md for how to reproduce these.
|
||||
magnitude_rms_db="1.24"
|
||||
bass_group_delay_swing_ms="13.2"
|
||||
limiter_headroom_db="1.6" ## worst-case peak on a hot master vs -1 dBFS
|
||||
dynamic_range_delta_lu="0.1"
|
||||
@@ -0,0 +1,32 @@
|
||||
[Unit]
|
||||
Description=Omarchy speaker tuning filter-chain
|
||||
Documentation=https://github.com/basecamp/omarchy/blob/master/docs/AUDIO-TUNING.md
|
||||
# WirePlumber does the linking, so starting before it is up risks the output being
|
||||
# linked before the speaker device has been discovered.
|
||||
After=pipewire.service wireplumber.service
|
||||
Requires=pipewire.service
|
||||
Wants=wireplumber.service
|
||||
# Restart with the audio daemon, since the filter-chain loses its connection when
|
||||
# PipeWire goes away.
|
||||
PartOf=pipewire.service
|
||||
|
||||
[Service]
|
||||
Type=simple
|
||||
# Hosts the tuning as a PipeWire *client* rather than loading it into the daemon
|
||||
# from pipewire.conf.d, which is only read at daemon startup. That is what lets
|
||||
# the tuning be switched on and off without restarting pipewire-pulse -- a
|
||||
# restart drops every PulseAudio client's connection, and applications that do
|
||||
# not reconnect (Spotify) have to be restarted by hand.
|
||||
#
|
||||
# It also contains failure: a malformed tuning breaks only this service, where a
|
||||
# bad drop-in in the daemon's own config stops PipeWire from starting at all.
|
||||
#
|
||||
# The config name is deliberately not PipeWire's stock filter-chain.conf, which
|
||||
# merges every fragment in ~/.config/pipewire/filter-chain.conf.d/ and would make
|
||||
# this service host unrelated user filters too.
|
||||
ExecStart=/usr/bin/pipewire -c omarchy-speaker-tuning.conf
|
||||
Restart=on-failure
|
||||
RestartSec=2
|
||||
|
||||
[Install]
|
||||
WantedBy=graphical-session.target
|
||||
Reference in New Issue
Block a user