Elgato_dark/onju-v2
1
0
Fork 0
mirror of https://github.com/justLV/onju-v2 synced 2026-04-21 15:47:55 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Fix onjuino interaction description (VAD, not double-tap), update m5_echo README terminology

Browse source
This commit is contained in:
justLV 2026-04-08 13:49:09 -07:00
parent 2943b07596
commit 260fbea9eb
1 changed files with 7 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
m5_echo/README.md
View file

@ -9,20 +9,20 @@ Firmware for the [M5Stack ATOM Echo](https://shop.m5stack.com/products/atom-echo
| **Board** | ESP32-S3 (custom PCB) | ESP32-PICO-D4 (M5Stack ATOM Echo) |
| **Mic** | Standard I2S (separate ADC) | PDM (SPM1423, on-chip) |
| **Speaker** | MAX98357A, 6 NeoPixel LEDs | NS4168, 1 SK6812 LED |
| **Interaction** | Capacitive touch: tap to start call, double-tap to end | Physical button: hold to talk, release to listen |
| **Mic control** | VAD-driven timeout (server extends via 0xDD) | Push-to-talk only - mic active while button held |
| **Call lifecycle** | Manual start/end via touch | Auto-starts on boot, persistent until bridge disconnects |
| **Interaction** | Capacitive touch: tap to wake, VAD detects end of speech | Physical button: hold to talk, release to listen |
| **Mic control** | VAD-driven (server detects silence and processes) | Push-to-talk only - mic active while button held |
| **Call lifecycle** | Tap to activate, VAD handles turn-taking | Auto-starts on boot, persistent until server disconnects |
| **PSRAM** | Yes (2MB, large playback buffer) | No (smaller buffers, more DMA buffers to compensate) |
| **Opus** | Full encoder + decoder | Decoder only (mic sends mu-law) |
| **I2S** | Single config for simultaneous TX+RX | Switches between PDM RX (mic) and I2S TX (speaker) |
## PTT behavior
- **Boot**: Connects to WiFi, announces on multicast with `PTT` flag. Bridge auto-starts a call and opens a persistent TCP connection.
- **Idle**: Bridge streams Opus audio over the persistent TCP connection. Device decodes and plays through the speaker.
- **Boot**: Connects to WiFi, announces on multicast with `PTT` flag. Server auto-starts a call and opens a persistent TCP connection.
- **Idle**: Server streams Opus audio over the persistent TCP connection. Device decodes and plays through the speaker.
- **Button press**: Immediately stops speaker playback (I2S switches to PDM mic mode). Opus task keeps reading the TCP stream but discards frames. Mic audio is captured, mu-law encoded, and sent via UDP.
- **Button release**: I2S switches back to speaker mode. Opus task resumes decoding and playing. The TCP connection stays alive throughout - no reconnection needed between turns.
- **Call end**: Bridge closes the TCP connection. Device returns to idle, waiting for a new connection.
- **Call end**: Server closes the TCP connection. Device returns to idle, waiting for a new connection.
## Hardware pin mapping (ATOM Echo)
@ -51,7 +51,7 @@ Uses the same TCP/UDP protocol as onjuino:
- **TCP 3001**: Server commands (0xAA audio, 0xBB LED set, 0xCC LED blink, 0xDD mic timeout)
- **Multicast**: Device announces as `"m5-echo m5echo PTT"` on the configured multicast group
The `PTT` token in the multicast announcement tells the bridge to auto-start a call on discovery and keep the TCP connection open persistently (no silence-based disconnection).
The `PTT` token in the multicast announcement tells the server to auto-start a call on discovery and keep the TCP connection open persistently (no silence-based disconnection).
## Serial commands