mixer_speaker.h

Go to the documentation of this file.

#pragma once
 
#ifdef USE_ESP32
 
#include "esphome/components/audio/audio.h"
#include "esphome/components/audio/audio_transfer_buffer.h"
#include "esphome/components/speaker/speaker.h"
 
#include "esphome/core/component.h"
#include "esphome/core/helpers.h"
 
#include <freertos/FreeRTOS.h>
#include <freertos/event_groups.h>
 
#include <atomic>
 
namespace esphome {
namespace mixer_speaker {
 
/* Classes for mixing several source speaker audio streams and writing it to another speaker component.
 *  - Volume controls are passed through to the output speaker
 *  - Source speaker commands are signaled via event group bits and processed in its loop function to ensure thread
 * safety
 *  - Directly handles pausing at the SourceSpeaker level; pause state is not passed through to the output speaker.
 *  - Audio sent to the SourceSpeaker must have 16 bits per sample.
 *  - Audio sent to the SourceSpeaker can have any number of channels. They are duplicated or ignored as needed to match
 *    the number of channels required for the output speaker.
 *  - In queue mode, the audio sent to the SourceSpeakers can have different sample rates.
 *  - In non-queue mode, the audio sent to the SourceSpeakers must have the same sample rates.
 *  - SourceSpeaker has an internal ring buffer. It also allocates a shared_ptr for an AudioTranserBuffer object.
 *  - Audio Data Flow:
 *      - Audio data played on a SourceSpeaker first writes to its internal ring buffer.
 *      - MixerSpeaker task temporarily takes shared ownership of each SourceSpeaker's AudioTransferBuffer.
 *      - MixerSpeaker calls SourceSpeaker's `process_data_from_source`, which transfers audio from the SourceSpeaker's
 *        ring buffer to its AudioTransferBuffer. Audio ducking is applied at this step.
 *      - In queue mode, MixerSpeaker prioritizes the earliest configured SourceSpeaker with audio data. Audio data is
 *        sent to the output speaker.
 *      - In non-queue mode, MixerSpeaker adds all the audio data in each SourceSpeaker into one stream that is written
 *        to the output speaker.
 */
 
class MixerSpeaker;
 
class SourceSpeaker : public speaker::Speaker, public Component {
 public:
  void dump_config() override;
  void setup() override;
  void loop() override;
 
  size_t play(const uint8_t *data, size_t length, TickType_t ticks_to_wait) override;
  size_t play(const uint8_t *data, size_t length) override { return this->play(data, length, 0); }
 
  void start() override;
  void stop() override;
  void finish() override;
 
  bool has_buffered_data() const override;
 
  void set_mute_state(bool mute_state) override;
  bool get_mute_state() override;
 
  void set_volume(float volume) override;
  float get_volume() override;
 
  void set_pause_state(bool pause_state) override { this->pause_state_ = pause_state; }
  bool get_pause_state() const override { return this->pause_state_; }
 
  size_t process_data_from_source(std::shared_ptr<audio::AudioSourceTransferBuffer> &transfer_buffer,
                                  TickType_t ticks_to_wait);
 
  void apply_ducking(uint8_t decibel_reduction, uint32_t duration);
 
  void set_buffer_duration(uint32_t buffer_duration_ms) { this->buffer_duration_ms_ = buffer_duration_ms; }
  void set_parent(MixerSpeaker *parent) { this->parent_ = parent; }
  void set_timeout(uint32_t ms) { this->timeout_ms_ = ms; }
 
  std::weak_ptr<audio::AudioSourceTransferBuffer> get_transfer_buffer() { return this->transfer_buffer_; }
 
 protected:
  friend class MixerSpeaker;
  esp_err_t start_();
  void enter_stopping_state_();
  void send_command_(uint32_t command_bit, bool wake_loop = false);
 
  static void duck_samples(int16_t *input_buffer, uint32_t input_samples_to_duck, int8_t *current_ducking_db_reduction,
                           uint32_t *ducking_transition_samples_remaining, uint32_t samples_per_ducking_step,
                           int8_t db_change_per_ducking_step);
 
  MixerSpeaker *parent_;
 
  std::shared_ptr<audio::AudioSourceTransferBuffer> transfer_buffer_;
  std::weak_ptr<RingBuffer> ring_buffer_;
 
  uint32_t buffer_duration_ms_;
  uint32_t last_seen_data_ms_{0};
  optional<uint32_t> timeout_ms_;
  bool stop_gracefully_{false};
 
  bool pause_state_{false};
 
  int8_t target_ducking_db_reduction_{0};
  int8_t current_ducking_db_reduction_{0};
  int8_t db_change_per_ducking_step_{1};
  uint32_t ducking_transition_samples_remaining_{0};
  uint32_t samples_per_ducking_step_{0};
 
  std::atomic<uint32_t> pending_playback_frames_{0};
  std::atomic<uint32_t> playback_delay_frames_{0};  // Frames in output pipeline when this source started contributing
  std::atomic<bool> has_contributed_{false};        // Tracks if source has contributed during this session
 
  EventGroupHandle_t event_group_{nullptr};
  uint32_t stopping_start_ms_{0};
};
 
class MixerSpeaker : public Component {
 public:
  void dump_config() override;
  void setup() override;
  void loop() override;
 
  void init_source_speakers(size_t count) { this->source_speakers_.init(count); }
  void add_source_speaker(SourceSpeaker *source_speaker) { this->source_speakers_.push_back(source_speaker); }
 
  esp_err_t start(audio::AudioStreamInfo &stream_info);
 
  void set_output_channels(uint8_t output_channels) { this->output_channels_ = output_channels; }
  void set_output_speaker(speaker::Speaker *speaker) { this->output_speaker_ = speaker; }
  void set_queue_mode(bool queue_mode) { this->queue_mode_ = queue_mode; }
  void set_task_stack_in_psram(bool task_stack_in_psram) { this->task_stack_in_psram_ = task_stack_in_psram; }
 
  speaker::Speaker *get_output_speaker() const { return this->output_speaker_; }
 
  uint32_t get_frames_in_pipeline() const { return this->frames_in_pipeline_.load(std::memory_order_acquire); }
 
 protected:
  static void copy_frames(const int16_t *input_buffer, audio::AudioStreamInfo input_stream_info, int16_t *output_buffer,
                          audio::AudioStreamInfo output_stream_info, uint32_t frames_to_transfer);
 
  static void mix_audio_samples(const int16_t *primary_buffer, audio::AudioStreamInfo primary_stream_info,
                                const int16_t *secondary_buffer, audio::AudioStreamInfo secondary_stream_info,
                                int16_t *output_buffer, audio::AudioStreamInfo output_stream_info,
                                uint32_t frames_to_mix);
 
  static void audio_mixer_task(void *params);
 
  esp_err_t start_task_();
 
  esp_err_t delete_task_();
 
  EventGroupHandle_t event_group_{nullptr};
 
  FixedVector<SourceSpeaker *> source_speakers_;
  speaker::Speaker *output_speaker_{nullptr};
 
  uint8_t output_channels_;
  bool queue_mode_;
  bool task_stack_in_psram_{false};
 
  TaskHandle_t task_handle_{nullptr};
  StaticTask_t task_stack_;
  StackType_t *task_stack_buffer_{nullptr};
 
  optional<audio::AudioStreamInfo> audio_stream_info_;
 
  std::atomic<uint32_t> frames_in_pipeline_{0};  // Frames written to output but not yet played
};
 
}  // namespace mixer_speaker
}  // namespace esphome
 
#endif