ESP32/main/audio/audio_prompt.c

#include "audio_prompt.h"
#include "audio_prompts_data.h"
#include "audio_output.h"
#include "esp_log.h"
#include "freertos/FreeRTOS.h"
#include "freertos/task.h"
#include "dac_njw1195.h"

static const char *TAG = "AudioPrompt";

void audio_prompt_init(void) {
    ESP_LOGI(TAG, "Audio prompt module initialized");
}

void audio_prompt_play(prompt_id_t id) {
    const int16_t *prompt_data = NULL;
    size_t prompt_len = 0;

    switch (id) {
        case PROMPT_BT_CONNECTED:
            prompt_data = prompt_bt_connected;
            prompt_len = prompt_bt_connected_len;
            break;
        case PROMPT_BT_DISCONNECTED:
            prompt_data = prompt_bt_disconnected;
            prompt_len = prompt_bt_disconnected_len;
            break;
        case PROMPT_LOW_BATTERY:
            prompt_data = prompt_low_battery;
            prompt_len = prompt_low_battery_len;
            break;
        default:
            return;
    }

    if (prompt_data == NULL || prompt_len == 0) {
        return;
    }

    ESP_LOGI(TAG, "Playing prompt tone ID: %d", id);

    // Lock audio output so we don't interleave with BT/AirPlay
    audio_output_lock();

    // Switch DAC to prompt volume
    dac_njw1195_set_prompt_mode(true);

    // Prompts are generated at 44100Hz 16-bit Mono.
    // audio_output_write expects 16-bit Stereo. We duplicate channels.
    // We send in chunks to avoid large memory allocations.
    const size_t CHUNK_SAMPLES = 256;
    int16_t stereo_buf[CHUNK_SAMPLES * 2];

    for (size_t i = 0; i < prompt_len; i += CHUNK_SAMPLES) {
        size_t samples_to_process = prompt_len - i;
        if (samples_to_process > CHUNK_SAMPLES) {
            samples_to_process = CHUNK_SAMPLES;
        }

        for (size_t j = 0; j < samples_to_process; j++) {
            int16_t sample = prompt_data[i + j];
            stereo_buf[j * 2] = sample;     // Left
            stereo_buf[j * 2 + 1] = sample; // Right
        }

        // Write stereo PCM data. audio_output_write internally calls i2s_channel_write.
        audio_output_write(stereo_buf, samples_to_process * 4, portMAX_DELAY);
    }

    // Write silence to flush the DMA buffers.
    // I2S_DMA_DESC_NUM(8) * I2S_DMA_FRAME_NUM(352) = 2816 frames.
    // Writing roughly 3000 frames of silence ensures the prompt is pushed out.
    memset(stereo_buf, 0, sizeof(stereo_buf));
    for (int i = 0; i < 12; i++) { // 12 * 256 = 3072 frames
        audio_output_write(stereo_buf, CHUNK_SAMPLES * 4, portMAX_DELAY);
    }

    // Restore original volume
    dac_njw1195_set_prompt_mode(false);

    audio_output_unlock();
}