How it works

The pieces

flowchart LR
  subgraph App["Desktop app (Tauri 2)"]
    UI["React UI<br/>Basics · Sound · Light"] -- "typed commands" --> Hub["hub.rs<br/>device, settings, PipeWire, recording"]
  end
  CLI["pd200x (CLI)"] --> Core
  Hub --> Core["pd200x-core"]
  Core -- "USB HID<br/>/dev/hidraw*" --> Mic["PD200X / PD200XS"]
  Core -- "wpctl, pw-dump" --> PW["PipeWire<br/>system volume"]
  Core -- "pw-record, pw-play" --> PW
  Core --> S["~/.config/pd200x-control/settings.json"]

• crates/core is the logic: the USB protocol, the live level, the system volume, the calibration, the recording analysis and the settings. Most of it is pure functions with tests; the hardware and the programs it runs are behind small traits, so the tests use a simulated microphone and fake programs.
• crates/cli is the pd200x command, a thin layer over the core.
• app/src-tauri exposes the core to the frontend as typed Tauri commands (tauri-specta writes app/src/lib/bindings.ts), watches for the microphone, and adds the tray.
• app/src is the React frontend: TanStack Router and Query, Tailwind, components in the shadcn/ui style, texts in German and English through Paraglide JS. Outside Tauri it runs against an in-memory mock backend.

The USB protocol

The microphone is a USB audio device plus a HID interface for its settings (352f:0104). The app writes and reads that HID interface through /dev/hidraw* – no driver, no libusb, no libudev.

Every setting is a 16-bit register. A packet of 65 bytes either writes a value (4B C4 0B 00 00 03 <register> <value> <checksum>) or asks for one (… 09 00 00 04 <register>); the microphone answers queries with the same layout. The checksum is the negated sum of the bytes. The microphone also sends its level on its own, register 0x2034, about 18 times a second; that is the live meter.

Registers	Meaning	Encoding
0x2022	mute	0/1
0x2023	gain	0–100
0x2026, 0x2027	compressor on, threshold	0/1, 2000 + dB × 10
0x202C, 0x202D	limiter on, threshold	0/1, 2000 + dB × 10
0x2030	headphone volume	0–100
0x2031, 0x2070	monitoring (written together)	2000/3000, 0 = on
0x2034	level (sent by the microphone)	2000 + dB × 10
0x2036–0x2039	light on, brightness, colour, linkage	0/1, 0–100, 0–8, 0/1
0x203A, 0x203B	noise cancellation on, strength	0/1, (s + 120) × 10
0x204D…	7 EQ bands × (on, type, Hz, gain, Q)	gain 2000 + dB × 10, Q × 100

The packet layout, the registers and the eight EQ presets were worked out by pd200x (advnpzn, MIT) from USB captures of MAONO Link. pd200x-control adds reading back the EQ, dynamics and light (the microphone answers those queries too) and detects which preset the microphone has. The linkage register reads back values that do not match what was written, so the app remembers that one itself.

The system microphone volume

The PipeWire source of the microphone is found with pw-dump (or wpctl status -n where pw-dump is missing) by its name, alsa_input.usb-352f_PD200X…. wpctl set-volume <id> <volume> sets it. PipeWire volumes are cubic: 1.75 means an amplitude of 1.75³ ≈ 5.36, that is +14.6 dB, which the desktops show as 175 %. WirePlumber keeps it across restarts.

Protection from apps

WebRTC clients (Teams, Slack, browsers) change the source volume through the PulseAudio interface of PipeWire. The switch writes a pipewire-pulse drop-in that gives every PulseAudio client the block-source-volume quirk and restarts pipewire-pulse; drop-ins append to pulse.rules. wpctl talks to PipeWire directly and is not affected, so the app can still change the volume.

The level finder

From five seconds of the microphone's level reports the app takes the 98th percentile as the speech peak (a single bump or click does not count). The difference to −9 dB goes to the hardware gain first (0.42 dB per step, up to 100) and the rest to the system volume (up to +18 dB); when it is too loud, the system volume comes down first.

The test recording

pw-record --target <PD200X source> --format s16 --rate 48000 --channels 1 writes a WAV file to the temporary folder, stopped after 30 seconds at the latest with SIGINT, so the WAV header is complete. The core reads it back for the peak, the RMS average and a 120-point waveform. pw-play plays it on the default output.

What is stored where

Where	What
The microphone	everything it can be set to; it keeps it without power
WirePlumber	the system microphone volume
~/.config/pipewire/pipewire-pulse.conf.d/	"Protect from apps", when it is on
~/.config/pd200x-control/settings.json	own EQ profiles, the last EQ, dynamics and light the app set, "light off while muted", the options
The temporary folder	the test recording, until the app quits