BigSound

System-wide audio enhancement for BigCommunity — a five-stage Rust DSP chain routed through PipeWire's filter-chain, with auto-profile per output device and a libadwaita frontend for live tuning.

---

Plays nicely with whatever's already running: Firefox, Spotify, MPV, Discord, games. Pick "BigSound (DSP)" as your audio output and every app on the system goes through the chain.

Screenshot

The window above shows the libadwaita frontend with the BigSound profile applied (the balanced default), running on GNOME with locale pt_BR. Move any slider and the audio responds within milliseconds — no service restarts, no config files to edit.

What it does

Module	Role	Type
BigBass	Psychoacoustic bass enhancement (missing-fundamental synth)	mono ×2
BigClarity	Treble exciter — sparkle on hi-hats, vocals, strings	mono ×2
BigSpace	Stereo widening (Mid/Side) with bass-keep-mono safety	stereo
BigCross	Bauer crossfeed — out-of-head soundstage on headphones	stereo
BigLoud	Stereo-linked compressor + limiter, calibrated make-up gain	stereo

The chain runs at native sample rate inside PipeWire's realtime audio path. No FFT, no oversampling, no allocations during process — just biquads and waveshapers, sample-accurate. Each module is a separate LADSPA .so so PipeWire's module-filter-chain can hot-route them without our daemon being in the audio loop.

DSP details

• BigBass — half-wave-rectified non-linear device (Aarts NLD) on the band-passed sidechain, smoothed with 0.5·(x + √(x²+ε)) to eliminate the corner-aliasing of a hard rectifier. A peak-envelope follower (12 ms attack / 200 ms release) gates the synthesised harmonics so the effect only fires when there's bass content, preventing buzz on vocals.

• BigClarity — high-passed sidechain through tanh saturator (symmetric → only odd harmonics), then a 16 kHz anti-alias low-pass before mixing back. The classical exciter topology used by BBE / Aphex.

• BigSpace — Mid/Side decompose, scale Side by width, recompose. A 200 Hz high-pass on the Side preserves bass mono compatibility: laptop speakers that physically sum to mono don't lose low-end energy.

• BigCross — Bauer crossfeed: 700 Hz low-passed, 280 µs delayed copy of each channel mixed at −9 dB into the opposite ear. Ring-buffer delay (zero allocation in the audio loop). The closest free equivalent to Atmos / Apple Spatial Audio without HRIR convolution.

• BigLoud — feed-forward stereo-linked compressor with soft knee. The make-up gain is calibrated to the limiter ceiling: for each amount value, make_up = ceiling + max_GR so a 0 dBFS input lands exactly on the ceiling after compression. That's what makes the perceived loudness rise instead of dropping — the FxSound trick.

Auto-profile

Plug headphones, switch to Bluetooth, change to HDMI — BigSound detects the active output (sink + port via PipeWire) and applies a matching profile automatically. No clicking, no fiddling. See the Profiles section below for the full catalogue.

Install

Arch / Manjaro / BigCommunity (recommended)

git clone https://github.com/xathay/bigsound.git
cd bigsound/packaging
makepkg -si
systemctl --user enable --now filter-chain.service bigsound-daemon.service

Then open Settings → Sound → Output and pick BigSound (DSP).

From source

git clone https://github.com/xathay/bigsound.git
cd bigsound
./scripts/install.sh

The script builds in release mode, installs LADSPA plugins to /usr/lib/ladspa/ (one sudo prompt — PipeWire 1.6+ refuses to load LADSPA plugins from $HOME), drops the PipeWire filter-chain config in ~/.config/pipewire/filter-chain.conf.d/, and starts the daemon. Everything else is user-local. Requires pipewire >= 1.0, gtk4, libadwaita, gettext, and a Rust toolchain.

Use

After installing, BigSound appears in the GNOME app launcher. Open it for live sliders + profile dropdown. From the terminal:

bigsound show                       # snapshot of current parameters
bigsound profile list               # all profiles
bigsound profile apply Audiophile   # switch profile manually
bigsound set bigloud:amount 0.7     # tune one parameter live
bigsound profile save MyMix         # save current state as a new profile

Architecture

apps  ──► BigSound (PipeWire sink)
              │
              └── filter-chain
                    ├── bigbass_l/r       (mono)
                    ├── bigclarity_l/r    (mono)
                    ├── bigspace          (stereo)
                    ├── bigcross          (stereo)
                    └── bigloud           (stereo)
              │
              ▼
       default real output (analog / HDMI / BT / USB)

D-Bus ◄──── bigsound-daemon ────► pw-cli set-param (live tuning)
GUI/CLI                              (no service restart)

Each DSP module is a separate Rust crate compiled to a LADSPA .so; the PipeWire module-filter-chain strings them together. The Rust daemon (com.bigcommunity.BigSound1) caches parameters in memory, pushes to PipeWire, and watches for output-device changes to auto-apply the matching profile.

Profiles

BigSound ships with 15 profiles. Five auto-apply by output device; ten are manual presets you pick from the dropdown for a specific intent.

Auto-applied (by output device)

Profile	Detects	Character
Laptop Speaker	alsa_output.pci-*::analog-output-speaker	Aggressive bass synthesis + heavy compression — tiny speakers can't reproduce bass, so we synthesise the harmonic series and let your brain reconstruct the missing fundamental
Headphones	*::analog-output-headphones, headsets	Gentle tuning with crossfeed on — escapes the "music inside your skull" feeling that pure stereo causes on headphones
Bluetooth	bluez_output.*	Codec-aware: extra treble exciter to claw back the SBC/AAC roll-off above 12 kHz, and a bass enhancer tuned for what BT codecs actually preserve
HDMI / TV	*.hdmi-*, HDMI sinks	Conservative neutral — TV speakers vary too much to be aggressive
BigSound	fallback when no rule matches	Balanced, hi-fi-friendly — the default starting point

Manual presets (by intent)

Pick from the dropdown when you want a specific character:

• Studio Reference — almost passthrough. Use for mixing, A/B reference, or critical listening of mastered material. Everything zeroed except the peak limiter as a safety net.

• Audiophile — invisible processing, but with 3D. Like Studio Reference but with strong crossfeed (0.55) so headphone listening escapes the in-head image. Best for FLAC/WAV on dedicated headphones.

• Atmos / Cinema — the "free Dolby Atmos" preset. Wide widening (1.55)

  • crossfeed (0.5) + bass push (+4 dB). Works on both headphones AND speakers. For films, trailers, big-room mixes.

• Headphones - Cinema — cinema preset, max for headphones only. Pushes Bauer crossfeed (0.85) and widening (1.45) further for stereo cinema content on headphones. DSP-only; full HRTF binaural will arrive in a later release.

• Punchy — the FxSound feel. Bass +6 dB, full compression, max squash. The "passthrough is half the volume" effect. Great for laptop / car speakers / any "I want obvious impact" scenario. Fatiguing on good headphones.

• Gaming — positional + impact. Maximum widening (1.6) for enemy localization, light compression (0.2) so footsteps survive, crossfeed for headset out-of-head soundstage.

• Voice / Podcast — talk-radio level. Narrow stereo (0.85, below unity — speech is centred and widening worsens sibilance), heavy compression for consistent dialogue level, vocal-band exciter at 2.8 kHz where intelligibility lives.

• Bass Heavy — EDM, hip-hop, dubstep. Maximum sub-bass synthesis (drive 0.85, mix 0.65, +8 dB), restrained treble (already-hot mixes don't want more), hard compression for club-loud impact.

• Late Night — quiet listening that doesn't disturb. Bass rolled off (low frequencies travel through walls), heavy compression so sudden loud bits don't startle, vocal clarity boosted to compensate for the Fletcher-Munson dip at low volume.

• Live / Concert — Tiny Desk, bootlegs, sessions. Wide image to suggest the venue, exciter on cymbals/brushes (live recordings often sound dull), moderate compression so audience-noise spikes don't startle.

Picking what to use

If you want…	Pick
Hear it loud, with obvious impact	Punchy or Bass Heavy
Hear it accurately	Audiophile or Studio Reference
Hear cinema / films	Atmos / Cinema (any output) or Headphones - Cinema (headphones only)
Listen quietly without disturbing	Late Night
Speech / podcasts / voice content	Voice / Podcast
Game audio	Gaming
EDM / hip-hop / heavy electronic	Bass Heavy
Live recordings / concerts	Live / Concert

Profiles live in ~/.config/bigsound/profiles/ (user-saved as 99-user-<name>.json) and /usr/share/bigsound/profiles/ (bundled). The daemon merges both, with user files winning on conflict — your customisations survive package upgrades.

Internationalisation

Strings are wrapped with gettext. Source language is English (the .po template), Brazilian Portuguese is shipped (crates/gtk-app/po/pt_BR.po). To add another language, copy pt_BR.po, translate the msgstr entries, name it <locale>.po and reinstall.

Troubleshooting

No audio at all after selecting BigSound (DSP)   Check systemctl --user status filter-chain.service — the service is disabled by default on PipeWire ≥ 1.0. Enable with systemctl --user enable --now filter-chain.service.

App says "bigsound-daemon doesn't seem to be running"   systemctl --user start bigsound-daemon.service. Check journalctl --user -u bigsound-daemon for errors.

Sliders don't move the audio   PipeWire suspends idle sinks and zeros their parameter struct on resume. The daemon's periodic re-push (every 3 s) fixes this automatically the moment audio starts flowing — just play something for ~3 s and the cache will reach the audio path.

Plugged headphones, profile didn't switch   Some codecs report headphone insertion as an active port change on the same sink (rather than a sink rename). The daemon detects both, but only when the new port is available. If the jack-detect on your laptop is unreliable, the active port may revert to speaker. Workaround: switch in Settings → Sound → Output manually, or save your headphone settings as a manual user profile.

Want JamesDSP back temporarily   systemctl --user stop filter-chain.service bigsound-daemon.service disables BigSound without uninstalling. Run JamesDSP normally; reverse with systemctl --user start ... to bring BigSound back.

Develop

The workspace is a Cargo multi-crate project:

BigSound/
├── crates/
│   ├── big-bass/             ← DSP library
│   ├── big-bass-ladspa/      ← LADSPA cdylib wrapper
│   ├── big-clarity/  + -ladspa/
│   ├── big-space/    + -ladspa/
│   ├── big-cross/    + -ladspa/
│   ├── big-loud/     + -ladspa/
│   ├── daemon/               ← bigsound-daemon (D-Bus server)
│   ├── cli/                  ← bigsound (CLI client)
│   └── gtk-app/              ← bigsound-app (libadwaita frontend)
├── configs/pipewire/         ← filter-chain config template
├── packaging/                ← PKGBUILD + bigsound.install
└── scripts/                  ← install.sh, ab-listen.sh

Build everything:

cargo build --release

Run a single LADSPA module's offline-WAV test harness (legacy validation tool for BigBass, before the realtime chain existed):

cd crates/big-bass
cargo run --release --bin big-bass-cli -- --help

A/B helper for offline listening tests:

./scripts/ab-listen.sh /tmp/bigsound-ab/   # plays each WAV with a banner

License

GPL-3.0-or-later. See LICENSE.

Credits

Built by the BigCommunity Team. Maintainer: Leonardo Athayde.

DSP techniques credited to: Aarts/Larsen (psychoacoustic bass), Aphex / BBE (exciter topology), Benjamin Bauer (1961, stereo crossfeed), Robert Bristow-Johnson (RBJ biquad cookbook), Linkwitz-Riley (crossover phase coherence).

Built on PipeWire, GTK 4, libadwaita, zbus, Rust.