DualSense haptics + speaker work in-game: four fixes between the game and the pad #302

Merged

enricobuehler merged 4 commits from worktree-pad-audio-fidelity into main 2026-08-18 15:02:20 +00:00

Conversation 0 Commits 4 Files Changed 10

+259 -21

enricobuehler commented 2026-08-18 14:55:57 +00:00

Owner

Copy Link

Copy Source

With #291 merged, GE-Proton could finally put the usbip pad into haptics mode — and everything after that was still silent. This PR carries the four fixes that took it from "enable succeeds" to haptics and speaker felt/heard in-game on glass (Spider-Man Remastered, GE-Proton 11-5, .21 host → Deck client → wired DS5).

1. The usbip capture forwarded the pad's hardware quad as the wire's speaker pair (dff2769b)

The ISO endpoint carries the DualSense's own channel map — ch0 = headphone-left, ch1 = the mono speaker, ch2/3 = the voice coils — because the host's UCM split folds everything to hardware channels before our capture point. The 0xD1 wire contract puts the speaker pair on ch0/1. Verbatim forwarding shipped headphone-left silence as wire speaker-left and the real speaker as speaker-right, which the client rendered onto the one Deck split-sink channel that PipeWire never wires to the physical speaker. (The split declares [AUX1,AUX1,AUX2,AUX3] — both channels to the speaker — but empirically only the first claimant reaches it. Measured, not assumed: a test tone showed up on exactly one channel at each hop.)

Fix: normalize_hw_quad — duplicate the hardware speaker channel across the wire pair, pass coils through, drop headphone-left (a remote pad's jack is not a wire surface). The stream-sink path (uhid pads) already emits the logical layout and is untouched.

2. The simulated ISO clock ran ~26 % slow (f7374149)

sleep(interval × packets) per URB, measured from "now", adds every source of slop on top of the nominal period. The virtual pad's ALSA clock measured 35.7 k frames/s against 48 kHz — the PCM backed up into xruns, and on the test box the pad sink became the PipeWire graph driver and dragged the desktop capture to 50 % delivery (severely garbled stream audio).

Fix: a per-endpoint absolute deadline ledger — late completions catch up instead of accumulating; a stall > 20 ms re-anchors instead of fast-forwarding. Measured after: 48,005 frames/s. Two paused-clock tests pin the exact rate and the re-anchor.

3. The speaker lane sounded "insanely compressed" (e356e354)

Both 0xD1 lanes used Application::LowDelay @ 64 kbps CBR. Right for voice-coil rumble; audibly wrong for programme audio. The speaker lane now runs the full Application::Audio coder at 96 kbps (~120 B per 10 ms frame, still far under one MTU). Haptics unchanged — they need the latency, not the fidelity.

4. A packaged WirePlumber policy holds the pad's device for GE-Proton (0a6a49a9)

GE's DS5 haptic router opens the sink's backing hw: device raw whenever it's free — then its own path re-probe EBUSYs against its own handle, invalidates the stream, and spins a 100 Hz "device generation" refresh loop (and, in one game, a buffer race in the same machinery crashed the title). On SteamOS — where that code was developed — PipeWire always holds the device, so GE immediately lands on its well-tested Pulse fallback and none of this fires.

scripts/60-punktfunk-dualsense.conf ships that environment: node.always-process + no-suspend holds the device from card creation; priority.driver = 1 keeps the pad's USB clock from ever driving the graph. Installed by rpm/deb/arch/nix into /usr/share/wireplumber/wireplumber.conf.d/. Covers physically plugged pads on a headless host identically (same failure, same fix).

Gates

• cargo clippy --release --all-targets -p usbip-sim -p pf-inject -p punktfunk-host -- -D warnings: clean (Linux, .21)
• cargo fmt --check (both crates): clean
• Tests: usbip-sim 13/13 (incl. 2 new paused-clock pacing tests), punktfunk-host pad tests 39/39 (incl. 2 new normalization tests)
• On glass: full chain verified end-to-end; in-game haptics + speaker confirmed by the operator

Known follow-ups (documented, deliberately not in this PR)

• UCM: Fedora 44 / Arch alsa-ucm 1.2.16 dropped the SpeakerHaptic device the Deck's 1.2.14 has; the test box runs the Deck's conf files hand-installed. Ship-vs-upstream is an open decision (a sysext can overlay /usr; rpm/deb cannot overwrite another package's files).
• Default volume: WirePlumber defaults a new pad card's sink to 40 % (−24 dB) on both host and client — stacked, that made working haptics imperceptible. Follow-up: pin 0 dB at client pick / host adoption (device.routes.default-sink-volume is global-only, so it can't be scoped in the shipped conf).
• DS5 touchpad click isn't forwarded over the wire (game map-button dead) — separate input-plane gap.

With #291 merged, GE-Proton could finally put the usbip pad into haptics mode — and everything after that was still silent. This PR carries the four fixes that took it from "enable succeeds" to **haptics and speaker felt/heard in-game on glass** (Spider-Man Remastered, GE-Proton 11-5, .21 host → Deck client → wired DS5). ## 1. The usbip capture forwarded the pad's *hardware* quad as the wire's *speaker pair* (`dff2769b`) The ISO endpoint carries the DualSense's own channel map — ch0 = headphone-left, **ch1 = the mono speaker**, ch2/3 = the voice coils — because the host's UCM split folds everything to hardware channels before our capture point. The 0xD1 wire contract puts the *speaker pair* on ch0/1. Verbatim forwarding shipped headphone-left silence as wire speaker-left and the real speaker as speaker-right, which the client rendered onto the one Deck split-sink channel that PipeWire never wires to the physical speaker. (The split *declares* `[AUX1,AUX1,AUX2,AUX3]` — both channels to the speaker — but empirically only the first claimant reaches it. Measured, not assumed: a test tone showed up on exactly one channel at each hop.) Fix: `normalize_hw_quad` — duplicate the hardware speaker channel across the wire pair, pass coils through, drop headphone-left (a remote pad's jack is not a wire surface). The stream-sink path (uhid pads) already emits the logical layout and is untouched. ## 2. The simulated ISO clock ran ~26 % slow (`f7374149`) `sleep(interval × packets)` per URB, measured from "now", adds every source of slop *on top of* the nominal period. The virtual pad's ALSA clock measured **35.7 k frames/s against 48 kHz** — the PCM backed up into xruns, and on the test box the pad sink became the PipeWire *graph driver* and dragged the desktop capture to 50 % delivery (severely garbled stream audio). Fix: a per-endpoint absolute deadline ledger — late completions catch up instead of accumulating; a stall > 20 ms re-anchors instead of fast-forwarding. Measured after: **48,005 frames/s**. Two paused-clock tests pin the exact rate and the re-anchor. ## 3. The speaker lane sounded "insanely compressed" (`e356e354`) Both 0xD1 lanes used `Application::LowDelay` @ 64 kbps CBR. Right for voice-coil rumble; audibly wrong for programme audio. The speaker lane now runs the full `Application::Audio` coder at 96 kbps (~120 B per 10 ms frame, still far under one MTU). Haptics unchanged — they need the latency, not the fidelity. ## 4. A packaged WirePlumber policy holds the pad's device for GE-Proton (`0a6a49a9`) GE's DS5 haptic router opens the sink's backing `hw:` device **raw** whenever it's free — then its own path re-probe EBUSYs against its *own handle*, invalidates the stream, and spins a 100 Hz "device generation" refresh loop (and, in one game, a buffer race in the same machinery crashed the title). On SteamOS — where that code was developed — PipeWire always holds the device, so GE immediately lands on its well-tested Pulse fallback and none of this fires. `scripts/60-punktfunk-dualsense.conf` ships that environment: `node.always-process` + no-suspend holds the device from card creation; `priority.driver = 1` keeps the pad's USB clock from ever driving the graph. Installed by rpm/deb/arch/nix into `/usr/share/wireplumber/wireplumber.conf.d/`. Covers physically plugged pads on a headless host identically (same failure, same fix). ## Gates - `cargo clippy --release --all-targets -p usbip-sim -p pf-inject -p punktfunk-host -- -D warnings`: clean (Linux, .21) - `cargo fmt --check` (both crates): clean - Tests: usbip-sim 13/13 (incl. 2 new paused-clock pacing tests), punktfunk-host pad tests 39/39 (incl. 2 new normalization tests) - On glass: full chain verified end-to-end; in-game haptics + speaker confirmed by the operator ## Known follow-ups (documented, deliberately not in this PR) - **UCM**: Fedora 44 / Arch alsa-ucm 1.2.16 dropped the `SpeakerHaptic` device the Deck's 1.2.14 has; the test box runs the Deck's conf files hand-installed. Ship-vs-upstream is an open decision (a sysext can overlay `/usr`; rpm/deb cannot overwrite another package's files). - **Default volume**: WirePlumber defaults a *new* pad card's sink to 40 % (−24 dB) on both host and client — stacked, that made working haptics imperceptible. Follow-up: pin 0 dB at client pick / host adoption (`device.routes.default-sink-volume` is global-only, so it can't be scoped in the shipped conf). - **DS5 touchpad click** isn't forwarded over the wire (game map-button dead) — separate input-plane gap.

enricobuehler added 4 commits 2026-08-18 14:55:59 +00:00

fix(pad): the usbip capture forwarded the pad's hardware quad as the wire's speaker pair ... dff2769ba9

The isochronous endpoint carries the DualSense's own channel map — ch0 = headphone LEFT,
ch1 = headphone RIGHT and the built-in mono speaker, ch2/3 = the voice coils — because
everything a game writes has already been folded to hardware channels by the host's UCM
split (or written there directly) before it reaches the endpoint. The 0xD1 wire contract
instead puts the *speaker pair* on ch0/1. Forwarding the quad verbatim shipped headphone-
left (silence) as wire speaker-left and the actual speaker as wire speaker-right, which
the client renders onto the one split-sink channel that current PipeWire never wires to
the physical speaker. Field-diagnosed on glass end-to-end: haptics felt, speaker dead,
a test tone measured on exactly one channel at each hop.

Duplicate the hardware speaker channel across the wire's speaker pair; pass the coils
through; drop headphone-left deliberately (a remote pad's jack is not a wire surface).
The stream-sink capture path (uhid pads) already emits the logical layout and is
unchanged.

fix(usbip): pace ISO completions against an absolute deadline, not relative sleeps ... f737414949

The simulator slept `service_interval × packets` per URB, measured from "now" at handling
time — so tokio timer slop, socket I/O and handler lock waits all added ON TOP of the
nominal period, every URB. The virtual pad's audio clock ran measurably slow under load
(~26 %: hw_ptr advanced ~35.7k frames/s against a 48 kHz stream), the PCM backed up into
xruns, and — because snd-usb-audio's clock IS URB completion — anything clocked off the
device dragged with it. On the test box the pad sink became the PipeWire graph driver and
pulled the desktop capture down to 50 % delivery ("insanely distorted" stream audio).

Pace against a per-endpoint absolute deadline ledger instead: each URB advances the
endpoint's deadline by exactly its nominal duration and sleeps until that instant, so
overhead eats into the next sleep rather than accumulating. A stall beyond 20 ms
re-anchors to now instead of fast-forwarding a burst. Measured after: 48005 frames/s.
Two paused-clock tests pin the exact rate and the re-anchor.

fix(pad-audio): the speaker lane gets the music coder, the haptics lane keeps LowDelay ... e356e354f2

Both 0xD1 lanes encoded with Application::LowDelay at 64 kbps CBR — right for voice-coil
rumble (felt latency, band-limited content), audibly wrong for the pad speaker, which
carries real programme audio: on glass it "sounded insanely compressed". The speaker lane
now uses the full Application::Audio coder at 96 kbps (~120 bytes per 10 ms frame, still
far under one MTU); its few ms of extra algorithmic delay are inaudible on a speaker.
Haptics unchanged.

feat(packaging): a WirePlumber policy holds a DualSense's sound card for GE-Proton ... 0a6a49a9aa

GE-Proton's DS5 haptic router opens the pad sink's backing hw: device RAW whenever it is
free — then its own path re-probe EBUSYs against its own handle, invalidates the stream,
and spins a 100 Hz "device generation" refresh loop: haptics dead, speaker dead, and in
one game a buffer race in the same machinery crashed the title outright. On SteamOS,
where that code was developed, PipeWire always holds the device, so GE lands on its
well-tested Pulse-routing fallback immediately and none of this fires.

Ship the SteamOS-shaped environment: node.always-process + no suspend keeps PipeWire
holding the device from the moment the card appears, and priority.driver = 1 keeps the
pad — whose USB audio clock (virtual or physical) is nobody's idea of a house clock —
from ever driving the graph. Installed by rpm/deb/arch/nix into
/usr/share/wireplumber/wireplumber.conf.d/. Matches both DS5 product-string spellings;
covers physically plugged pads on a headless host identically.

enricobuehler merged commit 412991f6a3 into main 2026-08-18 15:02:20 +00:00

enricobuehler deleted branch worktree-pad-audio-fidelity 2026-08-18 15:02:29 +00:00

enricobuehler referenced this pull request 2026-08-18 17:14:22 +00:00

The RPM %install tried to execute a %files path, so main builds no RPMs #306

enricobuehler referenced this issue from a commit 2026-08-18 22:18:30 +00:00

release: 0.31.0 — version bump, notes, CHANGELOG, Play notes

Sign in to join this conversation.

Reviewers

No Reviewers

Labels

No items

No labels

Milestone

No items

No Milestone

Projects

Clear projects

No projects

Assignees

Clear assignees

FernandoMX HitFrostbite enricobuehler (Enrico Bühler) kszlim luxus nannku neoney sassycrown

No Assignees

1 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: unom/punktfunk#302