Module capture 

Voice-stream capture state manager.

A D-STAR voice transmission on the wire is a sequence of protocol events:

VoiceStart(header) -> VoiceFrame x N -> VoiceEnd

Each transmission carries two parallel data streams inside the voice frames:

• AMBE voice data (9 bytes/frame, 50 frames/sec): the compressed speech audio that will be decoded to PCM and then to MP3.

• Slow data (3 bytes/frame, 50 frames/sec): a low-bandwidth side channel carrying things like text messages, DPRS position reports, and squelch codes. Each logical slow-data block spans several frames and is reassembled by SlowDataAssembler. Frame 0 of every 21-frame superframe is a sync frame (slow-data content 0x555555), not a data fragment — the assembler handles the descrambling internally.

This module provides two types that encapsulate the per-stream and cross-stream bookkeeping:

• StreamCapture — one in-progress capture: header metadata, all accumulated AMBE frames, and the slow-data assembler with its recovered text/DPRS fields.

• CaptureManager — maps stream IDs to active captures, so that voice frames from multiple concurrent transmissions (different modules, different reflectors) can be dispatched to the correct capture without cross-contamination.

Slow-data assembly details

dstar_gateway_core::SlowDataAssembler::push takes one 3-byte fragment at a time and returns Some(SlowDataBlock) when a block completes. For SlowDataBlock::Text, the .text field holds the 20-character status message. For SlowDataBlock::Gps, the payload is a raw NMEA-style DPRS sentence that we feed into dstar_gateway_core::parse_dprs to extract latitude and longitude. All other block kinds (header retx, fast data, squelch, unknown) are ignored for persistence purposes — they aren’t part of the streams table schema.

Structs

CaptureManager 🔒

Manages multiple concurrent voice-stream captures.

StreamCapture 🔒

State for one in-progress voice-stream capture.