b93327dc26
The daemon now requests the Bluetooth headset profile itself (pactl set-card-profile) the moment a take starts, instead of waiting for WirePlumber to notice the stream and run its switch timer. The microphone link comes up 0.15-0.18 s after the key instead of 0.21-0.27 s; WirePlumber still restores A2DP when the stream goes away. First real audio is now at 0.82-0.86 s, which is the headset's own floor. When the plugin is updated while the daemon runs, the daemon sees its source change, exits with code 4 once idle, and the shell service relaunches it without counting that as a crash. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Xoc9DJCViR7dg9eKzzAcfQ
2.9 KiB
2.9 KiB
How we cut Bluetooth microphone startup time
Measured on a Sony WH-1000XM5 over an Intel AX210, PipeWire 1.6.8, WirePlumber 0.5.17, kernel 7.2.3. Times are after the dictation key press.
| before | after | |
|---|---|---|
| mic link (eSCO) up | 0.7–1.0 s, sometimes never | 0.15–0.18 s |
| first real audio at the daemon | 1.5–1.7 s, or never | 0.82–0.86 s |
| "Listening" in the bar | 2.1 s on a timer, even with no audio | 0.86–0.94 s, only on real audio |
"After" is with the daemon recording straight from bluez_input.<addr>; via
camera-effects-mic add ~0.25 s (its 250 ms poll and processing).
What changed
| Change | Where it lives | Gain |
|---|---|---|
| Daemon indicator: green on real audio only (RMS > 0.0005 or three consecutive non-silent chunks), no timer | daemon/sttd.py |
the bar cannot claim to listen to a dead mic |
Daemon asks for the headset profile itself at capture start (pactl set-card-profile), before WirePlumber notices the stream |
daemon/sttd.py |
link up ~0.1 s sooner (0.21–0.27 s → 0.15–0.18 s); WirePlumber still restores A2DP |
Record from bluez_input.<addr> instead of camera-effects-mic |
~/.config/speech-to-text/config.json |
~0.25 s sooner, no noise suppression on dictation |
| WirePlumber profile-switch timeout 500 ms → 50 ms | ~/.local/share/wireplumber/scripts/device/autoswitch-bluetooth-profile.lua |
link up ~0.45 s sooner |
| A2DP auto-connect rule narrowed to the XM5 | ~/.config/wireplumber/wireplumber.conf.d/bluetooth-a2dp-autoconnect.conf |
no 5 s pages of absent devices at login |
btusb driver patch 0002 |
/usr/lib/modules/<kernel>/updates/btusb.ko |
mic link no longer comes up silent when a page overlaps it |
PipeWire bluez5 patches 0001 + 0003 |
~/.local/lib/spa-0.2/bluez5/ via SPA_PLUGIN_DIR drop-ins |
no stale-error dead takes; back-to-back takes work |
Only the first two rows are part of this repo. The rest are machine-level and are described, with patches and measurements, in HANDOFF.md.
Why ~0.85 s is the floor
key press
0.00 s ─┬─ daemon asks for the headset profile, starts pw-record
│
0.07 s ─┼─ eSCO connect() reaches the kernel
│ headset leaves sniff mode, radio sets up the link
0.16 s ─┼─ eSCO link up
│
│ headset sends encoded digital silence while its own
│ mic path starts and mSBC syncs: ~0.6 s
│ (measured at the decoder; Linux cannot shorten it)
│
0.80 s ─┼─ first non-zero samples leave the headset
│ decode + 50 ms chunking
0.85 s ─┴─ first real audio reaches the daemon
Nothing measurable is left on the Linux side.
The headset's silence after link-up and the eSCO setup are hardware. Words
spoken in the first ~0.9 s are lost unless the microphone is already in
hands-free mode, which drops playback to mono, so it is off by default
(warmMic).