UDP audio debug

The firmware can tee its microphone capture to a LAN UDP target so a host machine can listen along with aplay / nc. Opt-in via the behavior.audio_debug_udp_target field in STACKCHAN.RON; empty disables the stream entirely (the task parks and consumes no resources).

Enable

In /sd/STACKCHAN.RON:

behavior: (
    audio_debug_udp_target: "192.168.1.42:5005",
)

Substitute the IP of the host you’ll receive on (must be on the same LAN as the avatar) and any free UDP port. Reboot the avatar.

Receive

nc -lu 192.168.1.42 5005 | aplay -r 16000 -f S16_LE -c 1 -t raw

Decoder parameters:

• -r 16000 — sample rate matches firmware SAMPLE_RATE_HZ
• -f S16_LE — 16-bit signed little-endian, matches the on-wire layout
• -c 1 — mono
• -t raw — no header

Wire format

Each datagram is one 20 ms frame from AUDIO_FRAME_PUBSUB: 320 × i16 little-endian = 640 bytes of payload. No header, no sequence number — drops show up as audible gaps. At 50 frames/sec the stream is ~32 KB/s of UDP traffic.

When to use

• Verifying the ES7210 capture path works end-to-end without flashing a wake-word build.
• Inspecting environmental noise at a deployed unit to tune EmotionFromVoice / IntentFromLoud thresholds.
• Recording a sample for off-device model training.

The stream is plaintext UDP on the LAN; not suitable for a shared network. Leave the field empty in any deployment outside a controlled bench.

Related

• Behavior flags — audio_debug_udp_target is one of the live-reloadable behavior: block fields.
• Voice agent onboarding — the wake-word + sidecar loop uses the same audio capture path tapped here.
• Sidecar agent — the agent the wake-word path forwards audio to once audio_debug_udp_target confirms capture works.