ESP32/main/audio/audio_output.c

#include "audio_output.h"
#include "rtsp_server.h"

#include "audio_resample.h"
#include "dac.h"
#include "led.h"
#include "driver/i2s_std.h"
#include "driver/gpio.h"
#include "esp_check.h"
#include "freertos/FreeRTOS.h"
#include "freertos/task.h"
#include "audio_receiver.h"
#include <inttypes.h>
#include <stdlib.h>

// SIDE NOTE; providing power from GPIO pins is capped ~20mA.
#if CONFIG_I2S_GND_IO >= 0
#define I2S_GND_PIN CONFIG_I2S_GND_IO
#endif
#if CONFIG_I2S_VCC_IO >= 0
#define I2S_VCC_PIN CONFIG_I2S_VCC_IO
#endif

// DACXSMT引脚必须拉高电平才能在外部DAC上启用输出
#define DAC_XSMT_PIN 19

#define TAG           "audio_output"
#define I2S_SCK_PIN   CONFIG_I2S_SCK_IO
#define I2S_BCK_PIN   CONFIG_I2S_BCK_IO
#define I2S_LRCK_PIN  CONFIG_I2S_WS_IO
#define I2S_DOUT_PIN  CONFIG_I2S_DO_IO
#define OUTPUT_RATE   CONFIG_OUTPUT_SAMPLE_RATE_HZ
#define FRAME_SAMPLES 352

// DMA ring-buffer configuration.  Total DMA latency (in samples) is
//   I2S_DMA_DESC_NUM × I2S_DMA_FRAME_NUM
// which at OUTPUT_RATE gives the hardware pipeline delay in µs.
// Keep these in sync with the i2s_chan_config_t initialisation below.
#define I2S_DMA_DESC_NUM  8
#define I2S_DMA_FRAME_NUM 256

/* Max output frames after resampling one input frame */
#define MAX_RESAMPLE_FRAMES \
  ((size_t)((FRAME_SAMPLES + 2) * ((double)OUTPUT_RATE / 44100) + 16))

#if CONFIG_FREERTOS_UNICORE
#define PLAYBACK_CORE 0
#else
#define PLAYBACK_CORE 1
#endif

static i2s_chan_handle_t tx_handle = NULL;
static TaskHandle_t playback_task_handle = NULL;
static volatile bool playback_running = false;
static volatile bool flush_requested = false;
static SemaphoreHandle_t s_audio_mutex = NULL;
static volatile int source_rate = 44100;
static volatile bool resample_reinit_needed = false;
static volatile audio_channel_mode_t channel_mode = AUDIO_CHANNEL_STEREO;

static void apply_volume(int16_t *buf, size_t n) {
#ifndef CONFIG_DAC_CONTROLS_VOLUME
  int32_t vol = airplay_get_volume_q15();
  for (size_t i = 0; i < n; i++) {
    buf[i] = (int16_t)(((int32_t)buf[i] * vol) >> 15);
  }
#endif
}

// Apply the selected channel mode to an interleaved stereo buffer (L,R,...).
// LEFT/RIGHT route the chosen source channel to BOTH outputs so the selected
// track is heard from both speakers; STEREO leaves the buffer untouched.
static void apply_channel_mode(int16_t *buf, size_t frames) {
  audio_channel_mode_t mode = channel_mode;
  if (mode == AUDIO_CHANNEL_STEREO) {
    return;
  }
  size_t src = (mode == AUDIO_CHANNEL_RIGHT) ? 1 : 0;
  for (size_t i = 0; i < frames; i++) {
    int16_t s = buf[i * 2 + src];
    buf[i * 2] = s;
    buf[i * 2 + 1] = s;
  }
}

static void playback_task(void *arg) {
  int16_t *pcm = malloc((size_t)(FRAME_SAMPLES + 1) * 2 * sizeof(int16_t));
  int16_t *silence = calloc((size_t)FRAME_SAMPLES * 2, sizeof(int16_t));
  int16_t *resample_buf = malloc(MAX_RESAMPLE_FRAMES * 2 * sizeof(int16_t));
  if (!pcm || !silence || !resample_buf) {
    ESP_LOGE(TAG, "Failed to allocate buffers");
    free(pcm);
    free(silence);
    playback_task_handle = NULL;
    free(resample_buf);
    vTaskDelete(NULL);
    return;
  }

  size_t written;
  while (playback_running) {
    if (resample_reinit_needed) {
      resample_reinit_needed = false;
      audio_resample_init((uint32_t)source_rate, OUTPUT_RATE, 2);
    }
    if (flush_requested) {
      flush_requested = false;
      audio_resample_reset();
      i2s_channel_disable(tx_handle);
      i2s_channel_enable(tx_handle);
    }
    size_t samples = audio_receiver_read(pcm, FRAME_SAMPLES + 1);
    if (samples > 0) {
      if (s_audio_mutex) xSemaphoreTakeRecursive(s_audio_mutex, portMAX_DELAY);
      ESP_LOGD(TAG, "Read %u samples from receiver", (unsigned int)samples);
      int16_t *play_buf = pcm;
      size_t play_samples = samples;
      if (audio_resample_is_active()) {
        play_samples = audio_resample_process(pcm, samples, resample_buf,
                                              MAX_RESAMPLE_FRAMES);
        play_buf = resample_buf;
      }
      ESP_LOGD(TAG, "Resampled to %u samples", (unsigned int)play_samples);
      apply_volume(play_buf, play_samples * 2);
      apply_channel_mode(play_buf, play_samples);
      led_audio_feed(play_buf, play_samples);
      i2s_channel_write(tx_handle, play_buf, play_samples * 4, &written,
                        portMAX_DELAY);
      if (s_audio_mutex) xSemaphoreGiveRecursive(s_audio_mutex);
      ESP_LOGD(TAG, "I2S write: %u bytes written", (unsigned int)written);
      taskYIELD();
    } else {
      // ESP_LOGW(TAG, "Receiver underflow - playing silence");
      led_audio_feed(silence, FRAME_SAMPLES);
      i2s_channel_write(tx_handle, silence, (size_t)FRAME_SAMPLES * 4, &written,
                        pdMS_TO_TICKS(10));
      vTaskDelay(1);
    }
  }

  free(pcm);
  free(silence);
  playback_task_handle = NULL;
  vTaskDelete(NULL);
}

esp_err_t audio_output_init(void) {
  if (s_audio_mutex == NULL) {
    s_audio_mutex = xSemaphoreCreateRecursiveMutex();
  }

  i2s_chan_config_t chan_cfg =
      I2S_CHANNEL_DEFAULT_CONFIG(I2S_NUM_0, I2S_ROLE_MASTER);
  chan_cfg.dma_desc_num = I2S_DMA_DESC_NUM;
  chan_cfg.dma_frame_num = I2S_DMA_FRAME_NUM;

  ESP_RETURN_ON_ERROR(i2s_new_channel(&chan_cfg, &tx_handle, NULL), TAG,
                      "channel create failed");

  i2s_std_config_t std_cfg = {
      .clk_cfg = I2S_STD_CLK_DEFAULT_CONFIG(OUTPUT_RATE),
      .slot_cfg = I2S_STD_PHILIPS_SLOT_DEFAULT_CONFIG(I2S_DATA_BIT_WIDTH_16BIT,
                                                      I2S_SLOT_MODE_STEREO),
      .gpio_cfg =
          {
              .mclk = I2S_SCK_PIN,
              .bclk = I2S_BCK_PIN,
              .ws = I2S_LRCK_PIN,
              .dout = I2S_DOUT_PIN,
              .din = I2S_GPIO_UNUSED,
          },
  };
#ifdef I2S_GND_PIN
  gpio_reset_pin(I2S_GND_PIN);
  gpio_set_direction(I2S_GND_PIN, GPIO_MODE_OUTPUT);
  gpio_set_level(I2S_GND_PIN, 0);
#endif
#ifdef I2S_VCC_PIN
  gpio_reset_pin(I2S_VCC_PIN);
  gpio_set_direction(I2S_VCC_PIN, GPIO_MODE_OUTPUT);
  gpio_set_level(I2S_VCC_PIN, 1);
#endif

  ESP_RETURN_ON_ERROR(i2s_channel_init_std_mode(tx_handle, &std_cfg), TAG,
                      "std mode init failed");
  ESP_RETURN_ON_ERROR(i2s_channel_enable(tx_handle), TAG,
                      "channel enable failed");
  ESP_LOGI(TAG, "I2S initialized: Rate=%u, DMA_Desc=%d, DMA_Frame=%d",
           (unsigned int)OUTPUT_RATE, I2S_DMA_DESC_NUM, I2S_DMA_FRAME_NUM);

  // DAC XSMT: pull high to enable external DAC output
#ifdef DAC_XSMT_PIN
  gpio_reset_pin(DAC_XSMT_PIN);
  gpio_set_direction(DAC_XSMT_PIN, GPIO_MODE_OUTPUT);
  gpio_set_level(DAC_XSMT_PIN, 1);
  ESP_LOGI(TAG, "DAC XSMT pin %d set high", DAC_XSMT_PIN);
#endif

  // MCLK/BCLK/LRCK are now running. Some codecs need this edge to finish their
  // clock setup; amplifiers that manage power from board RTSP events can ignore
  // the hook.
  dac_on_i2s_started();

  audio_resample_init(44100, OUTPUT_RATE, 2);

  return ESP_OK;
}

void audio_output_start(void) {
  if (playback_task_handle != NULL) {
    return; // already running
  }
  playback_running = true;
  xTaskCreatePinnedToCore(playback_task, "audio_play", 4096, NULL, 7,
                          &playback_task_handle, PLAYBACK_CORE);
}

void audio_output_stop(void) {
  if (playback_task_handle == NULL) {
    return;
  }
  playback_running = false;
  // Wait for task to exit cleanly
  int timeout = 40;
  while (playback_task_handle != NULL && timeout-- > 0) {
    vTaskDelay(pdMS_TO_TICKS(50));
  }
  if (playback_task_handle != NULL) {
    ESP_LOGW(TAG, "Playback task did not exit within timeout");
  } else {
    ESP_LOGI(TAG, "Playback task stopped");
  }
}

esp_err_t audio_output_write(const void *data, size_t bytes, TickType_t wait) {
  size_t written = 0;
  if (s_audio_mutex) xSemaphoreTakeRecursive(s_audio_mutex, portMAX_DELAY);
  esp_err_t err = i2s_channel_write(tx_handle, data, bytes, &written, wait);
  if (s_audio_mutex) xSemaphoreGiveRecursive(s_audio_mutex);
  return err;
}

void audio_output_set_sample_rate(uint32_t rate) {
  // Only safe to call when no writer task is actively using I2S
  // (AirPlay playback task must be stopped, BT calls this before
  // the I2S writer task starts consuming data)
  ESP_LOGI(TAG, "Setting sample rate to %" PRIu32 " Hz", rate);
  if (tx_handle == NULL) {
    ESP_LOGW(TAG, "I2S 通道未初始化，跳过采样率时钟配置");
    return;
  }
  i2s_channel_disable(tx_handle);
  i2s_std_clk_config_t clk_cfg = I2S_STD_CLK_DEFAULT_CONFIG(rate);
  i2s_channel_reconfig_std_clock(tx_handle, &clk_cfg);
  i2s_channel_enable(tx_handle);
}

void audio_output_flush(void) {
  flush_requested = true;
}

void audio_output_lock(void) {
  if (s_audio_mutex) xSemaphoreTakeRecursive(s_audio_mutex, portMAX_DELAY);
}

void audio_output_unlock(void) {
  if (s_audio_mutex) xSemaphoreGiveRecursive(s_audio_mutex);
}

void audio_output_set_source_rate(int rate) {
  if (rate > 0 && rate != source_rate) {
    source_rate = rate;
    resample_reinit_needed = true;
  }
}

uint32_t audio_output_get_hardware_latency_us(void) {
  return (
      uint32_t)(((uint64_t)I2S_DMA_DESC_NUM * I2S_DMA_FRAME_NUM * 1000000ULL) /
                OUTPUT_RATE);
}

audio_channel_mode_t audio_output_cycle_channel_mode(void) {
  audio_channel_mode_t next;
  switch (channel_mode) {
  case AUDIO_CHANNEL_STEREO:
    next = AUDIO_CHANNEL_LEFT;
    break;
  case AUDIO_CHANNEL_LEFT:
    next = AUDIO_CHANNEL_RIGHT;
    break;
  default:
    next = AUDIO_CHANNEL_STEREO;
    break;
  }
  channel_mode = next;
  ESP_LOGI(TAG, "Channel mode: %s",
           next == AUDIO_CHANNEL_LEFT    ? "LEFT only"
           : next == AUDIO_CHANNEL_RIGHT ? "RIGHT only"
                                         : "STEREO");
  return next;
}

audio_channel_mode_t audio_output_get_channel_mode(void) {
  return channel_mode;
}