7 Commits
Author SHA1 Message Date
Dejvino 25ad8aa17c M314-style radar: dual-mic blips, sweep reveal, fade lifecycle 2026-11-12 20:04:40 +01:00
Dejvino 315cdbd177 Learn the populated microphone slots and DMA packing at runtime 2026-11-12 14:54:47 +01:00
Dejvino 129c670ff4 Drop mic gain out of saturation and soften the noise gate
The hardware log showed the live mic slots sitting at ~0.6 RMS in
ambient noise. At 37.5 dB the ES7210 was clipping, so the noise floor
slowly climbed to the saturated level and the gate never fired again:
the blip appeared once and then vanished forever.

Use 18 dB analog gain so ambient levels have headroom, lower the gate
to 1.5x the floor so real sounds stand out, and widen the loudness
window to -50..-6 dBFS. The debug line now also prints the
floor-relative signal levels.
2026-11-12 14:33:01 +01:00
Dejvino 72f80bfdd8 Make microphone slot discovery self-healing
The ES7210 power-up transient is loud enough to look like the first
sound. It placed one centred blip right after init and then pinned the
mic pair to whichever slots were loudest at that instant. If that was
the wrong pair, later claps were ignored forever.

Ignore the first 800 ms after starting the codec, and re-evaluate the
mic pair on every clear sound instead of locking once. The pair only
changes after repeating for a few blocks, or immediately when the
current pair is silent while another one is loud. Also enable the
per-second rms/pair debug line so the slot mapping can be verified on
hardware.
2026-11-12 14:10:17 +01:00
Dejvino b8de153e5a Fix microphone stereo detection on the T-Deck Plus
I2S_CHANNEL_FMT_ALL_LEFT copies a single channel into every slot, so
both microphones read the same signal and the balance stayed at zero.
Use I2S_CHANNEL_FMT_MULTIPLE (the real TDM mode) instead.

The board only fits two of the four ES7210 mic footprints, but the
schematic carries all four. Rather than hardcode MIC1/MIC3, the detector
now measures the TDM slots and locks onto the two that carry signal,
treating the lower slot as the left microphone. Equal maximum gain is
applied to all inputs so the pair amplifies consistently.

Also widen the balance range (claps produce only a few dB of difference
between the closely spaced mics) and recalibrate loudness to the dBFS
range seen with 37.5 dB analog gain.
2026-11-12 14:04:31 +01:00
Dejvino 2ba92a65d1 Drive radar blips from the T-Deck Plus dual microphones
Add a SoundDetector that reads the ES7210 as a 4-slot TDM stream
(slot 0 = MIC1/left, slot 2 = MIC3/right), tracks a per-channel RMS
envelope with an adaptive noise floor, and publishes a stereo balance
plus a dB-mapped loudness.

The radar sketch now slides a single reused blip left/right from the
balance and moves it nearer as the loudness rises, instead of firing a
hardcoded demo blip. The device's own ping/click mutes the detector for
200 ms so it cannot hear itself.

The trackball stays detached: GPIO0 is unavailable while the ES7210
microphones are running.
2026-11-12 08:20:38 +01:00
Dejvino 7dfdc06e6f Add NoiseSensor radar example and fix short sound playback
The sound task used to zero the I2S TX DMA ring as soon as the event
queue drained. i2s_write() only queues samples, so anything shorter than
the ~64 ms ring was erased before it reached the speaker. Enable
tx_desc_auto_clear instead, which clears each descriptor after the DMA
has actually transmitted it.

Also add the NoiseSensor radar example and point the default PlatformIO
environment at it.
2026-11-12 08:14:19 +01:00