SyntropicOS Embedded Audio Subsystem (src/syntropic/audio/)

The SyntropicOS Audio Subsystem provides lightweight, zero-heap, integer-only audio codecs and PCM streaming engine for microcontrollers.

Architecture

+-----------------------+     +-----------------------+
|  IMA-ADPCM Codec      |     |  Bluetooth SBC Codec  |
|  (syn_adpcm.h/.c)     |     |  (syn_sbc.h/.c)       |
+-----------+-----------+     +-----------+-----------+
            |                             |
            +------------+  +-------------+
                         |  |
                         v  v
            +---------------------------+
            |  PCM Streaming Engine     |
            |  (syn_audio.h/.c)         |
            +-------------+-------------+
                          |
                          v
            +---------------------------+
            |  MCU I2S / SAI / DAC DMA  |
            +---------------------------+

Features

IMA-ADPCM Codec (syn_adpcm)

• Compression: 4:1 lossy compression (16-bit linear PCM <-> 4-bit ADPCM nibbles).
• Footprint: ~200 bytes code, ~8 bytes state (SYN_ADPCM_State).
• Compatibility: Standard WAV IMA-ADPCM format (ffmpeg, Audacity compatible).

Bluetooth SBC Decoder (syn_sbc)

• Specification: Bluetooth SIG A2DP Sub-Band Codec spec compliant.
• Subbands: 4 and 8 subbands, 4/8/12/16 blocks.
• Modes: Mono, Dual Channel, Stereo, Joint Stereo.

PCM Streaming Playback Engine (syn_audio)

• Double-Buffering: Ping-pong PCM buffer preventing audio underrun.
• ISR Hooks: Decoupled syn_audio_isr_half() and syn_audio_isr_complete() handlers.

Voice Activity Detector (syn_vad)

• Algorithms: Multi-feature acoustic onset/voice detector:
• Short-Time Energy (STE)
• Zero-Crossing Rate (ZCR)
• High-Frequency Differential Energy (HFDE)
• Noise Tracking: Adaptive background noise floor tracking via Exponential Moving Average (EMA).
• Hysteresis: Configurable attack frame threshold (attack_frames) and hangover hangover release timing (hangover_frames).
• Presets: SYN_VAD_SENSITIVITY_SENSITIVE, SYN_VAD_SENSITIVITY_NORMAL, SYN_VAD_SENSITIVITY_AGGRESSIVE.

#include <syntropic/audio/syn_vad.h>

static SYN_VAD vad;

void vad_setup(void) {
    SYN_VAD_Config cfg = {
        .sample_rate_hz = 16000U,
        .frame_size_samples = 160U, /* 10 ms frame */
        .sensitivity = SYN_VAD_SENSITIVITY_NORMAL,
        .attack_frames = 2U,
        .hangover_frames = 10U,
    };
    syn_vad_init(&vad, &cfg);
}

void process_audio_frame(const int16_t *pcm_samples, size_t count) {
    bool is_speech = syn_vad_process_frame(&vad, pcm_samples, count);
    if (is_speech) {
        /* Wake wake-word detector or stream to ASR / cloud */
    }
}