Fixing Audio on the ASUS Zenbook S 14 (Lunar Lake) on Linux

February 8, 2026

Fixing Audio on the ASUS Zenbook S 14 (Lunar Lake) on Linux

The ASUS Zenbook S 14 (UX5406SA) has a quad-speaker Harman/Kardon setup driven by a Cirrus Logic CS42L43 codec and four CS35L56 amplifiers over SoundWire. On Windows, Dolby Atmos does the heavy lifting to make these speakers sound great. On Linux, you’re on your own — and out of the box, the experience ranges from “no sound” to “flat mono from one pair of speakers.”

Here’s everything I had to do to get proper audio working on Arch Linux with PipeWire.

The Hardware Problem

The Lunar Lake audio stack uses Intel SOF firmware talking to the CS42L43 codec via SoundWire. Two things go wrong at the ALSA level:

Speaker switches default to OFF. The ALSA UCM EnableSequence doesn’t trigger the CS42L43’s speaker playback and digital switches during init. Result: flat, mono-sounding audio from only some of the speakers.
Headphone jack detection is broken. The CS42L43’s impedance detection fails because of missing voltage regulators (it falls back to a dummy regulator). The kernel sees a physical insertion but can’t identify the device type, so audio keeps playing through speakers.

Both issues also resurface after suspend/resume.

Fix 1: Speaker Switches

The fix is simple — force the ALSA controls on:

amixer -c "$CARD" cset name='cs42l43 Speaker Playback Switch' on,on
amixer -c "$CARD" cset name='cs42l43 Speaker Digital Switch' on,on

Where $CARD is the CS42L43 card number, found dynamically:

CARD=$(grep -l cs42l43 /proc/asound/card*/id | head -1 | grep -oP 'card\K[0-9]+')

This runs via a systemd oneshot service at boot (after sound.target and alsa-restore) and again on resume from suspend.

Fix 2: Headphone Jack Override

Same idea — force the jack type:

amixer -c "$CARD" cset name='cs42l43 Jack Override' 'Headphone'

I also reset the headphone digital volume to prevent corruption that can happen after kernel updates:

amixer -c "$CARD" cset name='cs42l43 Headphone Digital Volume' 159,159

This also runs at boot (with a 2-second delay for hardware init and retry logic) and on resume. A third trigger runs it after Hyprland monitor profile switches, since audio routing can get confused during those.

Fix 3: Making the Speakers Sound Good

With the switches enabled, audio works — but it sounds noticeably worse than on Windows. That’s because Windows has Dolby Atmos processing the output in real-time, tuned for this specific speaker arrangement.

On Linux, we can approximate this using EasyEffects and its Convolver plugin with Dolby impulse response files. Credit to blog.helgesson.dev for documenting this approach.

Getting the IR Files

The Dolby impulse response files come in a zip with several profiles:

music-balanced.irs — general purpose, good default
movie.irs — emphasis on spatial audio
game.irs — tuned for positional cues
voice.irs — speech clarity
none.irs — reference/bypass

I put these in ~/.local/share/easyeffects/irs/.

EasyEffects Preset

Create an output preset that loads the Convolver with your chosen IR. Mine uses music-balanced:

{
  "output": {
    "blocklist": [],
    "plugins_order": ["convolver"],
    "convolver": {
      "bypass": false,
      "input-gain": 0.0,
      "output-gain": 0.0,
      "kernel-name": "music-balanced",
      "ir-width": 100,
      "autogain": false
    }
  }
}

Save as ~/.local/share/easyeffects/output/Dolby.json.

I also created a Passthrough.json with an empty plugins_order for headphones — no point processing audio that’s already going to proper drivers.

Auto-switching Between Speaker and Headphone Presets

EasyEffects supports autoloading presets based on the active output device. Create files in ~/.local/share/easyeffects/autoload/output/ named after the PipeWire sink and profile:

For speakers (alsa_output.pci-0000_00_1f.3-platform-sof_sdw.HiFi__Speaker__sink:[Out] Speaker.json):

{
  "device": "alsa_output.pci-0000_00_1f.3-platform-sof_sdw.HiFi__Speaker__sink",
  "device-description": "Lunar Lake-M HD Audio Controller Speaker",
  "device-profile": "[Out] Speaker",
  "preset-name": "Dolby"
}

For headphones (alsa_output.pci-0000_00_1f.3-platform-sof_sdw.HiFi__Headphones__sink:[Out] Headphones.json):

{
  "device": "alsa_output.pci-0000_00_1f.3-platform-sof_sdw.HiFi__Headphones__sink",
  "device-description": "Lunar Lake-M HD Audio Controller Headphones",
  "device-profile": "[Out] Headphones",
  "preset-name": "Passthrough"
}

Now when you plug in headphones, EasyEffects automatically drops the convolver processing. Unplug, and the Dolby profile kicks back in.

The Full Stack

Here’s what the complete audio fix looks like across the system:

Layer	What	Why
Kernel/ALSA	Speaker switch override (systemd)	UCM doesn’t enable switches
Kernel/ALSA	Jack detection override (systemd)	Impedance detection fails
PipeWire	SOF firmware + UCM topology	Routes audio to the right amps
EasyEffects	Convolver with Dolby IRs	Software replacement for Dolby Atmos
EasyEffects	Autoload presets	Speaker gets Dolby, headphones get passthrough

Everything is managed through dotfiles with GNU Stow, so a fresh install is just make stow plus enabling two systemd services.

Gotchas

Kernel regressions are real. Kernel 6.17.8 broke audio entirely on this hardware — I had to pin 6.17.7 until 6.17.9 landed. Keep an older kernel around.
Speaker noise above 80% volume was a known issue fixed in kernel 6.15-rc5+. If you’re on an older kernel, don’t crank it.
asound.state corruption can happen. If audio suddenly sounds wrong after an update, delete /var/lib/alsa/asound.state and reboot to let it regenerate.
EasyEffects 8.x stores data in ~/.local/share/easyeffects/, not ~/.config/easyeffects/. Older guides may point you to the wrong path.

Was It Worth It?

Honestly, yes. The difference with the convolver active is immediately noticeable — the speakers go from sounding like a cheap laptop to actually having some depth and separation. It’s not Windows Dolby Atmos, but it’s close enough that I stopped noticing the difference after a few days.

The hardware fixes (speaker switches + jack detection) are the real non-negotiable pieces though. Without those, you’re dealing with broken audio, not just mediocre audio.