From 162ee2ecaf8a5b1e7cfe18937100f9b2933e2f7a Mon Sep 17 00:00:00 2001 From: Keith Burzinski Date: Wed, 22 Apr 2026 14:40:18 -0500 Subject: [PATCH] [i2s_audio] Split speaker into base class and standard subclass (#15404) --- .../components/i2s_audio/speaker/__init__.py | 13 +- .../i2s_audio/speaker/i2s_audio_speaker.cpp | 456 ++++-------------- .../i2s_audio/speaker/i2s_audio_speaker.h | 89 +++- .../speaker/i2s_audio_speaker_standard.cpp | 307 ++++++++++++ .../speaker/i2s_audio_speaker_standard.h | 32 ++ 5 files changed, 515 insertions(+), 382 deletions(-) create mode 100644 esphome/components/i2s_audio/speaker/i2s_audio_speaker_standard.cpp create mode 100644 esphome/components/i2s_audio/speaker/i2s_audio_speaker_standard.h diff --git a/esphome/components/i2s_audio/speaker/__init__.py b/esphome/components/i2s_audio/speaker/__init__.py index d1d1bc3ee..99aa712c6 100644 --- a/esphome/components/i2s_audio/speaker/__init__.py +++ b/esphome/components/i2s_audio/speaker/__init__.py @@ -33,13 +33,16 @@ AUTO_LOAD = ["audio"] CODEOWNERS = ["@jesserockz", "@kahrendt"] DEPENDENCIES = ["i2s_audio"] -I2SAudioSpeaker = i2s_audio_ns.class_( - "I2SAudioSpeaker", cg.Component, speaker.Speaker, I2SAudioOut +I2SAudioSpeakerBase = i2s_audio_ns.class_( + "I2SAudioSpeakerBase", cg.Component, speaker.Speaker, I2SAudioOut ) +I2SAudioSpeaker = i2s_audio_ns.class_("I2SAudioSpeaker", I2SAudioSpeakerBase) CONF_DAC_TYPE = "dac_type" CONF_I2S_COMM_FMT = "i2s_comm_fmt" +I2SCommFmt = i2s_audio_ns.enum("I2SCommFmt", is_class=True) + i2s_dac_mode_t = cg.global_ns.enum("i2s_dac_mode_t") INTERNAL_DAC_OPTIONS = { CONF_LEFT: i2s_dac_mode_t.I2S_DAC_CHANNEL_LEFT_EN, @@ -183,11 +186,11 @@ async def to_code(config): await speaker.register_speaker(var, config) cg.add(var.set_dout_pin(config[CONF_I2S_DOUT_PIN])) - fmt = "std" # equals stand_i2s, stand_pcm_long, i2s_msb, pcm_long + fmt = I2SCommFmt.STANDARD # equals stand_i2s, stand_pcm_long, i2s_msb, pcm_long if config[CONF_I2S_COMM_FMT] in ["stand_msb", "i2s_lsb"]: - fmt = "msb" + fmt = I2SCommFmt.MSB elif config[CONF_I2S_COMM_FMT] in ["stand_pcm_short", "pcm_short", "pcm"]: - fmt = "pcm" + fmt = I2SCommFmt.PCM cg.add(var.set_i2s_comm_fmt(fmt)) if config[CONF_TIMEOUT] != CONF_NEVER: cg.add(var.set_timeout(config[CONF_TIMEOUT])) diff --git a/esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp b/esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp index dde1f70bc..836221e38 100644 --- a/esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp +++ b/esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp @@ -13,36 +13,10 @@ #include "esp_timer.h" -namespace esphome { -namespace i2s_audio { - -static const uint32_t DMA_BUFFER_DURATION_MS = 15; -static const size_t DMA_BUFFERS_COUNT = 4; - -static const size_t TASK_STACK_SIZE = 4096; -static const ssize_t TASK_PRIORITY = 19; - -static const size_t I2S_EVENT_QUEUE_COUNT = DMA_BUFFERS_COUNT + 1; +namespace esphome::i2s_audio { static const char *const TAG = "i2s_audio.speaker"; -enum SpeakerEventGroupBits : uint32_t { - COMMAND_START = (1 << 0), // indicates loop should start speaker task - COMMAND_STOP = (1 << 1), // stops the speaker task - COMMAND_STOP_GRACEFULLY = (1 << 2), // Stops the speaker task once all data has been written - - TASK_STARTING = (1 << 10), - TASK_RUNNING = (1 << 11), - TASK_STOPPING = (1 << 12), - TASK_STOPPED = (1 << 13), - - ERR_ESP_NO_MEM = (1 << 19), - - WARN_DROPPED_EVENT = (1 << 20), - - ALL_BITS = 0x00FFFFFF, // All valid FreeRTOS event group bits -}; - // Lists the Q15 fixed point scaling factor for volume reduction. // Has 100 values representing silence and a reduction [49, 48.5, ... 0.5, 0] dB. // dB to PCM scaling factor formula: floating_point_scale_factor = 2^(-db/6.014) @@ -56,17 +30,21 @@ static const std::vector Q15_VOLUME_SCALING_FACTORS = { 8218, 8706, 9222, 9770, 10349, 10963, 11613, 12302, 13032, 13805, 14624, 15491, 16410, 17384, 18415, 19508, 20665, 21891, 23189, 24565, 26022, 27566, 29201, 30933, 32767}; -void I2SAudioSpeaker::setup() { +void I2SAudioSpeakerBase::setup() { this->event_group_ = xEventGroupCreate(); if (this->event_group_ == nullptr) { - ESP_LOGE(TAG, "Failed to create event group"); + ESP_LOGE(TAG, "Event group creation failed"); this->mark_failed(); return; } + + // Initialize volume control. When audio_dac is configured, this sets the DAC volume. + // When no audio_dac is configured, this initializes software volume control. + this->set_volume(this->volume_); } -void I2SAudioSpeaker::dump_config() { +void I2SAudioSpeakerBase::dump_config() { ESP_LOGCONFIG(TAG, "Speaker:\n" " Pin: %d\n" @@ -75,10 +53,9 @@ void I2SAudioSpeaker::dump_config() { if (this->timeout_.has_value()) { ESP_LOGCONFIG(TAG, " Timeout: %" PRIu32 " ms", this->timeout_.value()); } - ESP_LOGCONFIG(TAG, " Communication format: %s", this->i2s_comm_fmt_.c_str()); } -void I2SAudioSpeaker::loop() { +void I2SAudioSpeakerBase::loop() { uint32_t event_group_bits = xEventGroupGetBits(this->event_group_); if ((event_group_bits & SpeakerEventGroupBits::COMMAND_START) && (this->state_ == speaker::STATE_STOPPED)) { @@ -92,12 +69,12 @@ void I2SAudioSpeaker::loop() { xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::TASK_STARTING); } if (event_group_bits & SpeakerEventGroupBits::TASK_RUNNING) { - ESP_LOGD(TAG, "Started"); + ESP_LOGV(TAG, "Started"); xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::TASK_RUNNING); this->state_ = speaker::STATE_RUNNING; } if (event_group_bits & SpeakerEventGroupBits::TASK_STOPPING) { - ESP_LOGD(TAG, "Stopping"); + ESP_LOGV(TAG, "Stopping"); xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::TASK_STOPPING); this->state_ = speaker::STATE_STOPPING; } @@ -111,10 +88,12 @@ void I2SAudioSpeaker::loop() { xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::ALL_BITS); this->status_clear_error(); + this->on_task_stopped(); + this->state_ = speaker::STATE_STOPPED; } - // Log any errors encounted by the task + // Log any errors encountered by the task if (event_group_bits & SpeakerEventGroupBits::ERR_ESP_NO_MEM) { ESP_LOGE(TAG, "Not enough memory"); xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::ERR_ESP_NO_MEM); @@ -133,14 +112,14 @@ void I2SAudioSpeaker::loop() { break; } - if (this->start_i2s_driver_(this->audio_stream_info_) != ESP_OK) { + if (this->start_i2s_driver(this->audio_stream_info_) != ESP_OK) { ESP_LOGE(TAG, "Driver failed to start; retrying in 1 second"); - this->status_momentary_error("driver-faiure", 1000); + this->status_momentary_error("driver-failure", 1000); break; } if (this->speaker_task_handle_ == nullptr) { - xTaskCreate(I2SAudioSpeaker::speaker_task, "speaker_task", TASK_STACK_SIZE, (void *) this, TASK_PRIORITY, + xTaskCreate(I2SAudioSpeakerBase::speaker_task, "speaker_task", TASK_STACK_SIZE, (void *) this, TASK_PRIORITY, &this->speaker_task_handle_); if (this->speaker_task_handle_ == nullptr) { @@ -157,7 +136,7 @@ void I2SAudioSpeaker::loop() { } } -void I2SAudioSpeaker::set_volume(float volume) { +void I2SAudioSpeakerBase::set_volume(float volume) { this->volume_ = volume; #ifdef USE_AUDIO_DAC if (this->audio_dac_ != nullptr) { @@ -166,15 +145,21 @@ void I2SAudioSpeaker::set_volume(float volume) { } this->audio_dac_->set_volume(volume); } else -#endif +#endif // USE_AUDIO_DAC { - // Fallback to software volume control by using a Q15 fixed point scaling factor - ssize_t decibel_index = remap(volume, 0.0f, 1.0f, 0, Q15_VOLUME_SCALING_FACTORS.size() - 1); - this->q15_volume_factor_ = Q15_VOLUME_SCALING_FACTORS[decibel_index]; + // Fallback to software volume control by using a Q15 fixed point scaling factor. + // At maximum volume (1.0), set to INT16_MAX to completely bypass volume processing + // and avoid any floating-point precision issues that could cause slight volume reduction. + if (volume >= 1.0f) { + this->q15_volume_factor_ = INT16_MAX; + } else { + ssize_t decibel_index = remap(volume, 0.0f, 1.0f, 0, Q15_VOLUME_SCALING_FACTORS.size() - 1); + this->q15_volume_factor_ = Q15_VOLUME_SCALING_FACTORS[decibel_index]; + } } } -void I2SAudioSpeaker::set_mute_state(bool mute_state) { +void I2SAudioSpeakerBase::set_mute_state(bool mute_state) { this->mute_state_ = mute_state; #ifdef USE_AUDIO_DAC if (this->audio_dac_) { @@ -184,7 +169,7 @@ void I2SAudioSpeaker::set_mute_state(bool mute_state) { this->audio_dac_->set_mute_off(); } } else -#endif +#endif // USE_AUDIO_DAC { if (mute_state) { // Fallback to software volume control and scale by 0 @@ -196,11 +181,12 @@ void I2SAudioSpeaker::set_mute_state(bool mute_state) { } } -size_t I2SAudioSpeaker::play(const uint8_t *data, size_t length, TickType_t ticks_to_wait) { +size_t I2SAudioSpeakerBase::play(const uint8_t *data, size_t length, TickType_t ticks_to_wait) { if (this->is_failed()) { ESP_LOGE(TAG, "Setup failed; cannot play audio"); return 0; } + if (this->state_ != speaker::STATE_RUNNING && this->state_ != speaker::STATE_STARTING) { this->start(); } @@ -214,8 +200,8 @@ size_t I2SAudioSpeaker::play(const uint8_t *data, size_t length, TickType_t tick size_t bytes_written = 0; if (this->state_ == speaker::STATE_RUNNING) { std::shared_ptr temp_ring_buffer = this->audio_ring_buffer_.lock(); - if (temp_ring_buffer.use_count() == 2) { - // Only the speaker task and this temp_ring_buffer own the ring buffer, so its safe to write to + if (temp_ring_buffer != nullptr) { + // The weak_ptr locks successfully only while the speaker task owns the ring buffer, so it is safe to write bytes_written = temp_ring_buffer->write_without_replacement((void *) data, length, ticks_to_wait); } } @@ -223,7 +209,7 @@ size_t I2SAudioSpeaker::play(const uint8_t *data, size_t length, TickType_t tick return bytes_written; } -bool I2SAudioSpeaker::has_buffered_data() const { +bool I2SAudioSpeakerBase::has_buffered_data() const { if (this->audio_ring_buffer_.use_count() > 0) { std::shared_ptr temp_ring_buffer = this->audio_ring_buffer_.lock(); return temp_ring_buffer->available() > 0; @@ -231,216 +217,27 @@ bool I2SAudioSpeaker::has_buffered_data() const { return false; } -void I2SAudioSpeaker::speaker_task(void *params) { - I2SAudioSpeaker *this_speaker = (I2SAudioSpeaker *) params; - - xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::TASK_STARTING); - - const uint32_t dma_buffers_duration_ms = DMA_BUFFER_DURATION_MS * DMA_BUFFERS_COUNT; - // Ensure ring buffer duration is at least the duration of all DMA buffers - const uint32_t ring_buffer_duration = std::max(dma_buffers_duration_ms, this_speaker->buffer_duration_ms_); - - // The DMA buffers may have more bits per sample, so calculate buffer sizes based in the input audio stream info - const size_t ring_buffer_size = this_speaker->current_stream_info_.ms_to_bytes(ring_buffer_duration); - - const uint32_t frames_to_fill_single_dma_buffer = - this_speaker->current_stream_info_.ms_to_frames(DMA_BUFFER_DURATION_MS); - const size_t bytes_to_fill_single_dma_buffer = - this_speaker->current_stream_info_.frames_to_bytes(frames_to_fill_single_dma_buffer); - - bool successful_setup = false; - std::unique_ptr transfer_buffer = - audio::AudioSourceTransferBuffer::create(bytes_to_fill_single_dma_buffer); - - if (transfer_buffer != nullptr) { - std::shared_ptr temp_ring_buffer = RingBuffer::create(ring_buffer_size); - if (temp_ring_buffer.use_count() == 1) { - transfer_buffer->set_source(temp_ring_buffer); - this_speaker->audio_ring_buffer_ = temp_ring_buffer; - successful_setup = true; - } - } - - if (!successful_setup) { - xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::ERR_ESP_NO_MEM); - } else { - bool stop_gracefully = false; - bool tx_dma_underflow = true; - - uint32_t frames_written = 0; - uint32_t last_data_received_time = millis(); - - xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::TASK_RUNNING); - - while (this_speaker->pause_state_ || !this_speaker->timeout_.has_value() || - (millis() - last_data_received_time) <= this_speaker->timeout_.value()) { - uint32_t event_group_bits = xEventGroupGetBits(this_speaker->event_group_); - - if (event_group_bits & SpeakerEventGroupBits::COMMAND_STOP) { - xEventGroupClearBits(this_speaker->event_group_, SpeakerEventGroupBits::COMMAND_STOP); - break; - } - if (event_group_bits & SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY) { - xEventGroupClearBits(this_speaker->event_group_, SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY); - stop_gracefully = true; - } - - if (this_speaker->audio_stream_info_ != this_speaker->current_stream_info_) { - // Audio stream info changed, stop the speaker task so it will restart with the proper settings. - break; - } - int64_t write_timestamp; - while (xQueueReceive(this_speaker->i2s_event_queue_, &write_timestamp, 0)) { - // Receives timing events from the I2S on_sent callback. If actual audio data was sent in this event, it passes - // on the timing info via the audio_output_callback. - uint32_t frames_sent = frames_to_fill_single_dma_buffer; - if (frames_to_fill_single_dma_buffer > frames_written) { - tx_dma_underflow = true; - frames_sent = frames_written; - const uint32_t frames_zeroed = frames_to_fill_single_dma_buffer - frames_written; - write_timestamp -= this_speaker->current_stream_info_.frames_to_microseconds(frames_zeroed); - } else { - tx_dma_underflow = false; - } - frames_written -= frames_sent; - if (frames_sent > 0) { - this_speaker->audio_output_callback_(frames_sent, write_timestamp); - } - } - - if (this_speaker->pause_state_) { - // Pause state is accessed atomically, so thread safe - // Delay so the task yields, then skip transferring audio data - vTaskDelay(pdMS_TO_TICKS(DMA_BUFFER_DURATION_MS)); - continue; - } - - // Wait half the duration of the data already written to the DMA buffers for new audio data - // The millisecond helper modifies the frames_written variable, so use the microsecond helper and divide by 1000 - const uint32_t read_delay = - (this_speaker->current_stream_info_.frames_to_microseconds(frames_written) / 1000) / 2; - - size_t bytes_read = transfer_buffer->transfer_data_from_source(pdMS_TO_TICKS(read_delay)); - uint8_t *new_data = transfer_buffer->get_buffer_end() - bytes_read; - - if (bytes_read > 0) { - if (this_speaker->q15_volume_factor_ < INT16_MAX) { - // Apply the software volume adjustment by unpacking the sample into a Q31 fixed-point number, shifting it, - // multiplying by the volume factor, and packing the sample back into the original bytes per sample. - - const size_t bytes_per_sample = this_speaker->current_stream_info_.samples_to_bytes(1); - const uint32_t len = bytes_read / bytes_per_sample; - - // Use Q16 for samples with 1 or 2 bytes: shifted_sample * gain_factor is Q16 * Q15 -> Q31 - int32_t shift = 15; // Q31 -> Q16 - int32_t gain_factor = this_speaker->q15_volume_factor_; // Q15 - - if (bytes_per_sample >= 3) { - // Use Q23 for samples with 3 or 4 bytes: shifted_sample * gain_factor is Q23 * Q8 -> Q31 - - shift = 8; // Q31 -> Q23 - gain_factor >>= 7; // Q15 -> Q8 - } - - for (uint32_t i = 0; i < len; ++i) { - int32_t sample = - audio::unpack_audio_sample_to_q31(&new_data[i * bytes_per_sample], bytes_per_sample); // Q31 - sample >>= shift; - sample *= gain_factor; // Q31 - audio::pack_q31_as_audio_sample(sample, &new_data[i * bytes_per_sample], bytes_per_sample); - } - } - -#ifdef USE_ESP32_VARIANT_ESP32 - // For ESP32 16-bit mono mode, adjacent samples need to be swapped. - if (this_speaker->current_stream_info_.get_channels() == 1 && - this_speaker->current_stream_info_.get_bits_per_sample() == 16) { - int16_t *samples = reinterpret_cast(new_data); - size_t sample_count = bytes_read / sizeof(int16_t); - for (size_t i = 0; i + 1 < sample_count; i += 2) { - int16_t tmp = samples[i]; - samples[i] = samples[i + 1]; - samples[i + 1] = tmp; - } - } -#endif - } - - if (transfer_buffer->available() == 0) { - if (stop_gracefully && tx_dma_underflow) { - break; - } - vTaskDelay(pdMS_TO_TICKS(DMA_BUFFER_DURATION_MS / 2)); - } else { - size_t bytes_written = 0; - if (tx_dma_underflow) { - // Temporarily disable channel and callback to reset the I2S driver's internal DMA buffer queue so timing - // callbacks are accurate. Preload the data. - i2s_channel_disable(this_speaker->tx_handle_); - const i2s_event_callbacks_t callbacks = { - .on_sent = nullptr, - }; - - i2s_channel_register_event_callback(this_speaker->tx_handle_, &callbacks, this_speaker); - i2s_channel_preload_data(this_speaker->tx_handle_, transfer_buffer->get_buffer_start(), - transfer_buffer->available(), &bytes_written); - } else { - // Audio is already playing, use regular I2S write to add to the DMA buffers - i2s_channel_write(this_speaker->tx_handle_, transfer_buffer->get_buffer_start(), transfer_buffer->available(), - &bytes_written, DMA_BUFFER_DURATION_MS); - } - if (bytes_written > 0) { - last_data_received_time = millis(); - frames_written += this_speaker->current_stream_info_.bytes_to_frames(bytes_written); - transfer_buffer->decrease_buffer_length(bytes_written); - if (tx_dma_underflow) { - tx_dma_underflow = false; - // Reset the event queue timestamps - // Enable the on_sent callback to accurately track the timestamps of played audio - // Enable the I2S channel to start sending the preloaded audio - - xQueueReset(this_speaker->i2s_event_queue_); - - const i2s_event_callbacks_t callbacks = { - .on_sent = i2s_on_sent_cb, - }; - i2s_channel_register_event_callback(this_speaker->tx_handle_, &callbacks, this_speaker); - - i2s_channel_enable(this_speaker->tx_handle_); - } - } - } - } - } - - xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::TASK_STOPPING); - - if (transfer_buffer != nullptr) { - transfer_buffer.reset(); - } - - xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::TASK_STOPPED); - - while (true) { - // Continuously delay until the loop method deletes the task - vTaskDelay(pdMS_TO_TICKS(10)); - } +void I2SAudioSpeakerBase::speaker_task(void *params) { + I2SAudioSpeakerBase *this_speaker = (I2SAudioSpeakerBase *) params; + this_speaker->run_speaker_task(); } -void I2SAudioSpeaker::start() { +void I2SAudioSpeakerBase::start() { if (!this->is_ready() || this->is_failed() || this->status_has_error()) return; if ((this->state_ == speaker::STATE_STARTING) || (this->state_ == speaker::STATE_RUNNING)) return; + // Mark STARTING immediately to avoid transient STOPPED observations before loop() processes COMMAND_START. + this->state_ = speaker::STATE_STARTING; xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::COMMAND_START); } -void I2SAudioSpeaker::stop() { this->stop_(false); } +void I2SAudioSpeakerBase::stop() { this->stop_(false); } -void I2SAudioSpeaker::finish() { this->stop_(true); } +void I2SAudioSpeakerBase::finish() { this->stop_(true); } -void I2SAudioSpeaker::stop_(bool wait_on_empty) { +void I2SAudioSpeakerBase::stop_(bool wait_on_empty) { if (this->is_failed()) return; if (this->state_ == speaker::STATE_STOPPED) @@ -453,105 +250,16 @@ void I2SAudioSpeaker::stop_(bool wait_on_empty) { } } -esp_err_t I2SAudioSpeaker::start_i2s_driver_(audio::AudioStreamInfo &audio_stream_info) { - this->current_stream_info_ = audio_stream_info; // store the stream info settings the driver will use - - if ((this->i2s_role_ & I2S_ROLE_SLAVE) && (this->sample_rate_ != audio_stream_info.get_sample_rate())) { // NOLINT - // Can't reconfigure I2S bus, so the sample rate must match the configured value - ESP_LOGE(TAG, "Audio stream settings are not compatible with this I2S configuration"); - return ESP_ERR_NOT_SUPPORTED; - } - - if (this->slot_bit_width_ != I2S_SLOT_BIT_WIDTH_AUTO && - (i2s_slot_bit_width_t) audio_stream_info.get_bits_per_sample() > this->slot_bit_width_) { - // Currently can't handle the case when the incoming audio has more bits per sample than the configured value - ESP_LOGE(TAG, "Audio streams with more bits per sample than the I2S speaker's configuration is not supported"); - return ESP_ERR_NOT_SUPPORTED; - } - - if (!this->parent_->try_lock()) { - ESP_LOGE(TAG, "Parent I2S bus not free"); - return ESP_ERR_INVALID_STATE; - } - - uint32_t dma_buffer_length = audio_stream_info.ms_to_frames(DMA_BUFFER_DURATION_MS); - - i2s_chan_config_t chan_cfg = { - .id = this->parent_->get_port(), - .role = this->i2s_role_, - .dma_desc_num = DMA_BUFFERS_COUNT, - .dma_frame_num = dma_buffer_length, - .auto_clear = true, - .intr_priority = 3, - }; - /* Allocate a new TX channel and get the handle of this channel */ +esp_err_t I2SAudioSpeakerBase::init_i2s_channel_(const i2s_chan_config_t &chan_cfg, const i2s_std_config_t &std_cfg, + size_t event_queue_size) { esp_err_t err = i2s_new_channel(&chan_cfg, &this->tx_handle_, NULL); if (err != ESP_OK) { - ESP_LOGE(TAG, "Failed to allocate new I2S channel"); + ESP_LOGE(TAG, "I2S channel allocation failed: %s", esp_err_to_name(err)); this->parent_->unlock(); return err; } - i2s_clock_src_t clk_src = I2S_CLK_SRC_DEFAULT; -#ifdef I2S_CLK_SRC_APLL - if (this->use_apll_) { - clk_src = I2S_CLK_SRC_APLL; - } -#endif - i2s_std_gpio_config_t pin_config = this->parent_->get_pin_config(); - - i2s_std_clk_config_t clk_cfg = { - .sample_rate_hz = audio_stream_info.get_sample_rate(), - .clk_src = clk_src, - .mclk_multiple = this->mclk_multiple_, - }; - - i2s_slot_mode_t slot_mode = this->slot_mode_; - i2s_std_slot_mask_t slot_mask = this->std_slot_mask_; - if (audio_stream_info.get_channels() == 1) { - slot_mode = I2S_SLOT_MODE_MONO; - } else if (audio_stream_info.get_channels() == 2) { - slot_mode = I2S_SLOT_MODE_STEREO; - slot_mask = I2S_STD_SLOT_BOTH; - } - - i2s_std_slot_config_t std_slot_cfg; - if (this->i2s_comm_fmt_ == "std") { - std_slot_cfg = - I2S_STD_PHILIPS_SLOT_DEFAULT_CONFIG((i2s_data_bit_width_t) audio_stream_info.get_bits_per_sample(), slot_mode); - } else if (this->i2s_comm_fmt_ == "pcm") { - std_slot_cfg = - I2S_STD_PCM_SLOT_DEFAULT_CONFIG((i2s_data_bit_width_t) audio_stream_info.get_bits_per_sample(), slot_mode); - } else { - std_slot_cfg = - I2S_STD_MSB_SLOT_DEFAULT_CONFIG((i2s_data_bit_width_t) audio_stream_info.get_bits_per_sample(), slot_mode); - } -#ifdef USE_ESP32_VARIANT_ESP32 - // There seems to be a bug on the ESP32 (non-variant) platform where setting the slot bit width higher then the bits - // per sample causes the audio to play too fast. Setting the ws_width to the configured slot bit width seems to - // make it play at the correct speed while sending more bits per slot. - if (this->slot_bit_width_ != I2S_SLOT_BIT_WIDTH_AUTO) { - uint32_t configured_bit_width = static_cast(this->slot_bit_width_); - std_slot_cfg.ws_width = configured_bit_width; - if (configured_bit_width > 16) { - std_slot_cfg.msb_right = false; - } - } -#else - std_slot_cfg.slot_bit_width = this->slot_bit_width_; -#endif - std_slot_cfg.slot_mask = slot_mask; - - pin_config.dout = this->dout_pin_; - - i2s_std_config_t std_cfg = { - .clk_cfg = clk_cfg, - .slot_cfg = std_slot_cfg, - .gpio_cfg = pin_config, - }; - /* Initialize the channel */ err = i2s_channel_init_std_mode(this->tx_handle_, &std_cfg); - if (err != ESP_OK) { ESP_LOGE(TAG, "Failed to initialize channel"); i2s_del_channel(this->tx_handle_); @@ -559,23 +267,34 @@ esp_err_t I2SAudioSpeaker::start_i2s_driver_(audio::AudioStreamInfo &audio_strea this->parent_->unlock(); return err; } + if (this->i2s_event_queue_ == nullptr) { - this->i2s_event_queue_ = xQueueCreate(I2S_EVENT_QUEUE_COUNT, sizeof(int64_t)); + this->i2s_event_queue_ = xQueueCreate(event_queue_size, sizeof(int64_t)); + } else { + // Reset queue to clear any stale events from previous task + xQueueReset(this->i2s_event_queue_); } - i2s_channel_enable(this->tx_handle_); - - return err; + return ESP_OK; } -bool IRAM_ATTR I2SAudioSpeaker::i2s_on_sent_cb(i2s_chan_handle_t handle, i2s_event_data_t *event, void *user_ctx) { +void I2SAudioSpeakerBase::stop_i2s_driver_() { + if (this->tx_handle_ != nullptr) { + i2s_channel_disable(this->tx_handle_); + i2s_del_channel(this->tx_handle_); + this->tx_handle_ = nullptr; + } + this->parent_->unlock(); +} + +bool IRAM_ATTR I2SAudioSpeakerBase::i2s_on_sent_cb(i2s_chan_handle_t handle, i2s_event_data_t *event, void *user_ctx) { int64_t now = esp_timer_get_time(); BaseType_t need_yield1 = pdFALSE; BaseType_t need_yield2 = pdFALSE; BaseType_t need_yield3 = pdFALSE; - I2SAudioSpeaker *this_speaker = (I2SAudioSpeaker *) user_ctx; + I2SAudioSpeakerBase *this_speaker = (I2SAudioSpeakerBase *) user_ctx; if (xQueueIsQueueFullFromISR(this_speaker->i2s_event_queue_)) { // Queue is full, so discard the oldest event and set the warning flag to inform the user @@ -589,14 +308,47 @@ bool IRAM_ATTR I2SAudioSpeaker::i2s_on_sent_cb(i2s_chan_handle_t handle, i2s_eve return need_yield1 | need_yield2 | need_yield3; } -void I2SAudioSpeaker::stop_i2s_driver_() { - i2s_channel_disable(this->tx_handle_); - i2s_del_channel(this->tx_handle_); - this->tx_handle_ = nullptr; - this->parent_->unlock(); +void I2SAudioSpeakerBase::apply_software_volume_(uint8_t *data, size_t bytes_read) { + if (this->q15_volume_factor_ >= INT16_MAX) { + return; // Max volume, no processing needed + } + + const size_t bytes_per_sample = this->current_stream_info_.samples_to_bytes(1); + const uint32_t len = bytes_read / bytes_per_sample; + + // Use Q16 for samples with 1 or 2 bytes: shifted_sample * gain_factor is Q16 * Q15 -> Q31 + int32_t shift = 15; // Q31 -> Q16 + int32_t gain_factor = this->q15_volume_factor_; // Q15 + + if (bytes_per_sample >= 3) { + // Use Q23 for samples with 3 or 4 bytes: shifted_sample * gain_factor is Q23 * Q8 -> Q31 + shift = 8; // Q31 -> Q23 + gain_factor >>= 7; // Q15 -> Q8 + } + + for (uint32_t i = 0; i < len; ++i) { + int32_t sample = audio::unpack_audio_sample_to_q31(&data[i * bytes_per_sample], bytes_per_sample); // Q31 + sample >>= shift; + sample *= gain_factor; // Q31 + audio::pack_q31_as_audio_sample(sample, &data[i * bytes_per_sample], bytes_per_sample); + } } -} // namespace i2s_audio -} // namespace esphome +void I2SAudioSpeakerBase::swap_esp32_mono_samples_(uint8_t *data, size_t bytes_read) { +#ifdef USE_ESP32_VARIANT_ESP32 + // For ESP32 16-bit mono mode, adjacent samples need to be swapped. + if (this->current_stream_info_.get_channels() == 1 && this->current_stream_info_.get_bits_per_sample() == 16) { + int16_t *samples = reinterpret_cast(data); + size_t sample_count = bytes_read / sizeof(int16_t); + for (size_t i = 0; i + 1 < sample_count; i += 2) { + int16_t tmp = samples[i]; + samples[i] = samples[i + 1]; + samples[i + 1] = tmp; + } + } +#endif // USE_ESP32_VARIANT_ESP32 +} + +} // namespace esphome::i2s_audio #endif // USE_ESP32 diff --git a/esphome/components/i2s_audio/speaker/i2s_audio_speaker.h b/esphome/components/i2s_audio/speaker/i2s_audio_speaker.h index 76b669220..b2644efd0 100644 --- a/esphome/components/i2s_audio/speaker/i2s_audio_speaker.h +++ b/esphome/components/i2s_audio/speaker/i2s_audio_speaker.h @@ -16,10 +16,34 @@ #include "esphome/core/helpers.h" #include "esphome/core/ring_buffer.h" -namespace esphome { -namespace i2s_audio { +namespace esphome::i2s_audio { -class I2SAudioSpeaker : public I2SAudioOut, public speaker::Speaker, public Component { +// Shared constants for I2S audio speaker implementations +static constexpr uint32_t DMA_BUFFER_DURATION_MS = 15; +static constexpr size_t TASK_STACK_SIZE = 4096; +static constexpr ssize_t TASK_PRIORITY = 19; + +enum SpeakerEventGroupBits : uint32_t { + COMMAND_START = (1 << 0), // indicates loop should start speaker task + COMMAND_STOP = (1 << 1), // stops the speaker task + COMMAND_STOP_GRACEFULLY = (1 << 2), // Stops the speaker task once all data has been written + + TASK_STARTING = (1 << 10), + TASK_RUNNING = (1 << 11), + TASK_STOPPING = (1 << 12), + TASK_STOPPED = (1 << 13), + + ERR_ESP_NO_MEM = (1 << 19), + + WARN_DROPPED_EVENT = (1 << 20), + + ALL_BITS = 0x00FFFFFF, // All valid FreeRTOS event group bits +}; + +/// @brief Abstract base class for I2S audio speaker implementations. +/// Provides shared infrastructure (event groups, ring buffer, volume control, task lifecycle) +/// for derived I2S speaker classes. +class I2SAudioSpeakerBase : public I2SAudioOut, public speaker::Speaker, public Component { public: float get_setup_priority() const override { return esphome::setup_priority::PROCESSOR; } @@ -30,7 +54,9 @@ class I2SAudioSpeaker : public I2SAudioOut, public speaker::Speaker, public Comp void set_buffer_duration(uint32_t buffer_duration_ms) { this->buffer_duration_ms_ = buffer_duration_ms; } void set_timeout(uint32_t ms) { this->timeout_ = ms; } void set_dout_pin(uint8_t pin) { this->dout_pin_ = (gpio_num_t) pin; } - void set_i2s_comm_fmt(std::string mode) { this->i2s_comm_fmt_ = std::move(mode); } + + /// @brief Get the I2S TX channel handle + i2s_chan_handle_t get_tx_handle() const { return this->tx_handle_; } void start() override; void stop() override; @@ -63,40 +89,55 @@ class I2SAudioSpeaker : public I2SAudioOut, public speaker::Speaker, public Comp void set_mute_state(bool mute_state) override; protected: - /// @brief Function for the FreeRTOS task handling audio output. - /// Allocates space for the buffers, reads audio from the ring buffer and writes audio to the I2S port. Stops - /// immmiately after receiving the COMMAND_STOP signal and stops only after the ring buffer is empty after receiving - /// the COMMAND_STOP_GRACEFULLY signal. Stops if the ring buffer hasn't read data for more than timeout_ milliseconds. - /// When stopping, it deallocates the buffers. It communicates its state and any errors via ``event_group_``. - /// @param params I2SAudioSpeaker component + /// @brief FreeRTOS task entry point. Casts params to I2SAudioSpeakerBase and calls run_speaker_task_(). + /// @param params I2SAudioSpeakerBase component pointer static void speaker_task(void *params); + /// @brief The main speaker task loop. Implemented by derived classes for mode-specific behavior. + virtual void run_speaker_task() = 0; + /// @brief Sends a stop command to the speaker task via ``event_group_``. /// @param wait_on_empty If false, sends the COMMAND_STOP signal. If true, sends the COMMAND_STOP_GRACEFULLY signal. void stop_(bool wait_on_empty); - /// @brief Callback function used to send playback timestamps the to the speaker task. + /// @brief Callback function used to send playback timestamps to the speaker task. /// @param handle (i2s_chan_handle_t) /// @param event (i2s_event_data_t) /// @param user_ctx (void*) User context pointer that the callback accesses /// @return True if a higher priority task was interrupted static bool i2s_on_sent_cb(i2s_chan_handle_t handle, i2s_event_data_t *event, void *user_ctx); - /// @brief Starts the ESP32 I2S driver. - /// Attempts to lock the I2S port, starts the I2S driver using the passed in stream information, and sets the data out - /// pin. If it fails, it will unlock the I2S port and uninstalls the driver, if necessary. + /// @brief Starts the ESP32 I2S driver. Implemented by derived classes for mode-specific configuration. /// @param audio_stream_info Stream information for the I2S driver. - /// @return ESP_ERR_NOT_ALLOWED if the I2S port can't play the incoming audio stream. - /// ESP_ERR_INVALID_STATE if the I2S port is already locked. - /// ESP_ERR_INVALID_ARG if installing the driver or setting the data outpin fails due to a parameter error. - /// ESP_ERR_NO_MEM if the driver fails to install due to a memory allocation error. - /// ESP_FAIL if setting the data out pin fails due to an IO error - /// ESP_OK if successful - esp_err_t start_i2s_driver_(audio::AudioStreamInfo &audio_stream_info); + /// @return ESP_OK if successful, or an error code + virtual esp_err_t start_i2s_driver(audio::AudioStreamInfo &audio_stream_info) = 0; + + /// @brief Shared I2S channel allocation, initialization, and event queue setup. + /// Called by derived start_i2s_driver_() implementations after building mode-specific configs. + /// @param chan_cfg I2S channel configuration + /// @param std_cfg I2S standard mode configuration (clock, slot, GPIO) + /// @param event_queue_size Size of the event queue + /// @return ESP_OK if successful, or an error code. On failure, cleans up channel and unlocks parent. + esp_err_t init_i2s_channel_(const i2s_chan_config_t &chan_cfg, const i2s_std_config_t &std_cfg, + size_t event_queue_size); /// @brief Stops the I2S driver and unlocks the I2S port void stop_i2s_driver_(); + /// @brief Called in loop() when the task has stopped. Override for mode-specific cleanup. + virtual void on_task_stopped() {} + + /// @brief Apply software volume control using Q15 fixed-point scaling. + /// @param data Pointer to audio sample data (modified in place) + /// @param bytes_read Number of bytes of audio data + void apply_software_volume_(uint8_t *data, size_t bytes_read); + + /// @brief Swap adjacent 16-bit mono samples for ESP32 (non-variant) hardware quirk. + /// Only applies when running on original ESP32 with 16-bit mono audio. + /// @param data Pointer to audio sample data (modified in place) + /// @param bytes_read Number of bytes of audio data + void swap_esp32_mono_samples_(uint8_t *data, size_t bytes_read); + TaskHandle_t speaker_task_handle_{nullptr}; EventGroupHandle_t event_group_{nullptr}; @@ -115,11 +156,9 @@ class I2SAudioSpeaker : public I2SAudioOut, public speaker::Speaker, public Comp audio::AudioStreamInfo current_stream_info_; // The currently loaded driver's stream info gpio_num_t dout_pin_; - std::string i2s_comm_fmt_; - i2s_chan_handle_t tx_handle_; + i2s_chan_handle_t tx_handle_{nullptr}; }; -} // namespace i2s_audio -} // namespace esphome +} // namespace esphome::i2s_audio #endif // USE_ESP32 diff --git a/esphome/components/i2s_audio/speaker/i2s_audio_speaker_standard.cpp b/esphome/components/i2s_audio/speaker/i2s_audio_speaker_standard.cpp new file mode 100644 index 000000000..020346403 --- /dev/null +++ b/esphome/components/i2s_audio/speaker/i2s_audio_speaker_standard.cpp @@ -0,0 +1,307 @@ +#include "i2s_audio_speaker_standard.h" + +#ifdef USE_ESP32 + +#include + +#include "esphome/components/audio/audio.h" +#include "esphome/components/audio/audio_transfer_buffer.h" + +#include "esphome/core/hal.h" +#include "esphome/core/log.h" + +#include "esp_timer.h" + +namespace esphome::i2s_audio { + +static const char *const TAG = "i2s_audio.speaker.std"; + +static constexpr size_t DMA_BUFFERS_COUNT = 4; +static constexpr size_t I2S_EVENT_QUEUE_COUNT = DMA_BUFFERS_COUNT + 1; + +void I2SAudioSpeaker::dump_config() { + I2SAudioSpeakerBase::dump_config(); + const char *fmt_str; + switch (this->i2s_comm_fmt_) { + case I2SCommFmt::PCM: + fmt_str = "pcm"; + break; + case I2SCommFmt::MSB: + fmt_str = "msb"; + break; + default: + fmt_str = "std"; + break; + } + ESP_LOGCONFIG(TAG, " Communication format: %s", fmt_str); +} + +void I2SAudioSpeaker::run_speaker_task() { + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::TASK_STARTING); + + const uint32_t dma_buffers_duration_ms = DMA_BUFFER_DURATION_MS * DMA_BUFFERS_COUNT; + // Ensure ring buffer duration is at least the duration of all DMA buffers + const uint32_t ring_buffer_duration = std::max(dma_buffers_duration_ms, this->buffer_duration_ms_); + + // The DMA buffers may have more bits per sample, so calculate buffer sizes based on the input audio stream info + const size_t ring_buffer_size = this->current_stream_info_.ms_to_bytes(ring_buffer_duration); + const uint32_t frames_to_fill_single_dma_buffer = this->current_stream_info_.ms_to_frames(DMA_BUFFER_DURATION_MS); + const size_t bytes_to_fill_single_dma_buffer = + this->current_stream_info_.frames_to_bytes(frames_to_fill_single_dma_buffer); + + bool successful_setup = false; + std::unique_ptr transfer_buffer = + audio::AudioSourceTransferBuffer::create(bytes_to_fill_single_dma_buffer); + + if (transfer_buffer != nullptr) { + std::shared_ptr temp_ring_buffer = RingBuffer::create(ring_buffer_size); + if (temp_ring_buffer.use_count() == 1) { + transfer_buffer->set_source(temp_ring_buffer); + this->audio_ring_buffer_ = temp_ring_buffer; + successful_setup = true; + } + } + + if (!successful_setup) { + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_ESP_NO_MEM); + } else { + bool stop_gracefully = false; + bool tx_dma_underflow = true; + + uint32_t frames_written = 0; + uint32_t last_data_received_time = millis(); + + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::TASK_RUNNING); + + // Main speaker task loop. Continues while: + // - Paused, OR + // - No timeout configured, OR + // - Timeout hasn't elapsed since last data + while (this->pause_state_ || !this->timeout_.has_value() || + (millis() - last_data_received_time) <= this->timeout_.value()) { + uint32_t event_group_bits = xEventGroupGetBits(this->event_group_); + + if (event_group_bits & SpeakerEventGroupBits::COMMAND_STOP) { + xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::COMMAND_STOP); + ESP_LOGV(TAG, "Exiting: COMMAND_STOP received"); + break; + } + if (event_group_bits & SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY) { + xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY); + stop_gracefully = true; + } + + if (this->audio_stream_info_ != this->current_stream_info_) { + // Audio stream info changed, stop the speaker task so it will restart with the proper settings. + ESP_LOGV(TAG, "Exiting: stream info changed"); + break; + } + + int64_t write_timestamp; + while (xQueueReceive(this->i2s_event_queue_, &write_timestamp, 0)) { + // Receives timing events from the I2S on_sent callback. If actual audio data was sent in this event, it passes + // on the timing info via the audio_output_callback. + uint32_t frames_sent = frames_to_fill_single_dma_buffer; + if (frames_to_fill_single_dma_buffer > frames_written) { + tx_dma_underflow = true; + frames_sent = frames_written; + const uint32_t frames_zeroed = frames_to_fill_single_dma_buffer - frames_written; + write_timestamp -= this->current_stream_info_.frames_to_microseconds(frames_zeroed); + } else { + tx_dma_underflow = false; + } + frames_written -= frames_sent; + + // Standard I2S mode: fire callback immediately for each event + if (frames_sent > 0) { + this->audio_output_callback_(frames_sent, write_timestamp); + } + } + + if (this->pause_state_) { + // Pause state is accessed atomically, so thread safe + // Delay so the task yields, then skip transferring audio data + vTaskDelay(pdMS_TO_TICKS(DMA_BUFFER_DURATION_MS)); + continue; + } + + // Wait half the duration of the data already written to the DMA buffers for new audio data + // The millisecond helper modifies the frames_written variable, so use the microsecond helper and divide by 1000 + uint32_t read_delay = (this->current_stream_info_.frames_to_microseconds(frames_written) / 1000) / 2; + + size_t bytes_read = transfer_buffer->transfer_data_from_source(pdMS_TO_TICKS(read_delay)); + uint8_t *new_data = transfer_buffer->get_buffer_end() - bytes_read; + + if (bytes_read > 0) { + this->apply_software_volume_(new_data, bytes_read); + this->swap_esp32_mono_samples_(new_data, bytes_read); + } + + if (transfer_buffer->available() == 0) { + if (stop_gracefully && tx_dma_underflow) { + break; + } + vTaskDelay(pdMS_TO_TICKS(DMA_BUFFER_DURATION_MS / 2)); + } else { + size_t bytes_written = 0; + + if (tx_dma_underflow) { + // Temporarily disable channel and callback to reset the I2S driver's internal DMA buffer queue + i2s_channel_disable(this->tx_handle_); + const i2s_event_callbacks_t null_callbacks = {.on_sent = nullptr}; + i2s_channel_register_event_callback(this->tx_handle_, &null_callbacks, this); + i2s_channel_preload_data(this->tx_handle_, transfer_buffer->get_buffer_start(), transfer_buffer->available(), + &bytes_written); + } else { + // Audio is already playing, use regular write to add to the DMA buffers + i2s_channel_write(this->tx_handle_, transfer_buffer->get_buffer_start(), transfer_buffer->available(), + &bytes_written, DMA_BUFFER_DURATION_MS); + } + + if (bytes_written > 0) { + last_data_received_time = millis(); + frames_written += this->current_stream_info_.bytes_to_frames(bytes_written); + transfer_buffer->decrease_buffer_length(bytes_written); + + if (tx_dma_underflow) { + tx_dma_underflow = false; + // Enable the on_sent callback and channel after preload + xQueueReset(this->i2s_event_queue_); + const i2s_event_callbacks_t callbacks = {.on_sent = i2s_on_sent_cb}; + i2s_channel_register_event_callback(this->tx_handle_, &callbacks, this); + i2s_channel_enable(this->tx_handle_); + } + } + } + } + } + + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::TASK_STOPPING); + + if (transfer_buffer != nullptr) { + transfer_buffer.reset(); + } + + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::TASK_STOPPED); + + while (true) { + // Continuously delay until the loop method deletes the task + vTaskDelay(pdMS_TO_TICKS(10)); + } +} + +esp_err_t I2SAudioSpeaker::start_i2s_driver(audio::AudioStreamInfo &audio_stream_info) { + this->current_stream_info_ = audio_stream_info; + + if ((this->i2s_role_ & I2S_ROLE_SLAVE) && (this->sample_rate_ != audio_stream_info.get_sample_rate())) { // NOLINT + // Can't reconfigure I2S bus, so the sample rate must match the configured value + ESP_LOGE(TAG, "Incompatible stream settings"); + return ESP_ERR_NOT_SUPPORTED; + } + + if (this->slot_bit_width_ != I2S_SLOT_BIT_WIDTH_AUTO && + (i2s_slot_bit_width_t) audio_stream_info.get_bits_per_sample() > this->slot_bit_width_) { + // Currently can't handle the case when the incoming audio has more bits per sample than the configured value + ESP_LOGE(TAG, "Stream bits per sample must be less than or equal to the speaker's configuration"); + return ESP_ERR_NOT_SUPPORTED; + } + + if (!this->parent_->try_lock()) { + ESP_LOGE(TAG, "Parent bus is busy"); + return ESP_ERR_INVALID_STATE; + } + + uint32_t dma_buffer_length = audio_stream_info.ms_to_frames(DMA_BUFFER_DURATION_MS); + + i2s_role_t i2s_role = this->i2s_role_; + i2s_clock_src_t clk_src = I2S_CLK_SRC_DEFAULT; + +#if SOC_CLK_APLL_SUPPORTED + if (this->use_apll_) { + clk_src = i2s_clock_src_t::I2S_CLK_SRC_APLL; + } +#endif // SOC_CLK_APLL_SUPPORTED + + // Log DMA configuration for debugging + ESP_LOGV(TAG, "I2S DMA config: %zu buffers x %lu frames", (size_t) DMA_BUFFERS_COUNT, + (unsigned long) dma_buffer_length); + + i2s_chan_config_t chan_cfg = { + .id = this->parent_->get_port(), + .role = i2s_role, + .dma_desc_num = DMA_BUFFERS_COUNT, + .dma_frame_num = dma_buffer_length, + .auto_clear = true, + .intr_priority = 3, + }; + + // Build standard I2S clock/slot/gpio configuration + i2s_std_clk_config_t clk_cfg = { + .sample_rate_hz = audio_stream_info.get_sample_rate(), + .clk_src = clk_src, + .mclk_multiple = this->mclk_multiple_, + }; + + i2s_slot_mode_t slot_mode = this->slot_mode_; + i2s_std_slot_mask_t slot_mask = this->std_slot_mask_; + if (audio_stream_info.get_channels() == 1) { + slot_mode = I2S_SLOT_MODE_MONO; + } else if (audio_stream_info.get_channels() == 2) { + slot_mode = I2S_SLOT_MODE_STEREO; + slot_mask = I2S_STD_SLOT_BOTH; + } + + i2s_std_slot_config_t slot_cfg; + switch (this->i2s_comm_fmt_) { + case I2SCommFmt::PCM: + slot_cfg = + I2S_STD_PCM_SLOT_DEFAULT_CONFIG((i2s_data_bit_width_t) audio_stream_info.get_bits_per_sample(), slot_mode); + break; + case I2SCommFmt::MSB: + slot_cfg = + I2S_STD_MSB_SLOT_DEFAULT_CONFIG((i2s_data_bit_width_t) audio_stream_info.get_bits_per_sample(), slot_mode); + break; + default: + slot_cfg = I2S_STD_PHILIPS_SLOT_DEFAULT_CONFIG((i2s_data_bit_width_t) audio_stream_info.get_bits_per_sample(), + slot_mode); + break; + } + +#ifdef USE_ESP32_VARIANT_ESP32 + // There seems to be a bug on the ESP32 (non-variant) platform where setting the slot bit width higher than the + // bits per sample causes the audio to play too fast. Setting the ws_width to the configured slot bit width seems + // to make it play at the correct speed while sending more bits per slot. + if (this->slot_bit_width_ != I2S_SLOT_BIT_WIDTH_AUTO) { + uint32_t configured_bit_width = static_cast(this->slot_bit_width_); + slot_cfg.ws_width = configured_bit_width; + if (configured_bit_width > 16) { + slot_cfg.msb_right = false; + } + } +#else + slot_cfg.slot_bit_width = this->slot_bit_width_; +#endif // USE_ESP32_VARIANT_ESP32 + slot_cfg.slot_mask = slot_mask; + + i2s_std_gpio_config_t gpio_cfg = this->parent_->get_pin_config(); + gpio_cfg.dout = this->dout_pin_; + + i2s_std_config_t std_cfg = { + .clk_cfg = clk_cfg, + .slot_cfg = slot_cfg, + .gpio_cfg = gpio_cfg, + }; + + esp_err_t err = this->init_i2s_channel_(chan_cfg, std_cfg, I2S_EVENT_QUEUE_COUNT); + if (err != ESP_OK) { + return err; + } + + i2s_channel_enable(this->tx_handle_); + + return ESP_OK; +} + +} // namespace esphome::i2s_audio + +#endif // USE_ESP32 diff --git a/esphome/components/i2s_audio/speaker/i2s_audio_speaker_standard.h b/esphome/components/i2s_audio/speaker/i2s_audio_speaker_standard.h new file mode 100644 index 000000000..7b7f8b647 --- /dev/null +++ b/esphome/components/i2s_audio/speaker/i2s_audio_speaker_standard.h @@ -0,0 +1,32 @@ +#pragma once + +#ifdef USE_ESP32 + +#include "i2s_audio_speaker.h" + +namespace esphome::i2s_audio { + +enum class I2SCommFmt : uint8_t { + STANDARD, // Philips / I2S standard + PCM, // PCM short + MSB, // MSB / left-justified +}; + +/// @brief Standard I2S speaker implementation. +/// Outputs PCM audio data directly to an I2S DAC using the standard I2S protocol. +class I2SAudioSpeaker : public I2SAudioSpeakerBase { + public: + void dump_config() override; + + void set_i2s_comm_fmt(I2SCommFmt fmt) { this->i2s_comm_fmt_ = fmt; } + + protected: + void run_speaker_task() override; + esp_err_t start_i2s_driver(audio::AudioStreamInfo &audio_stream_info) override; + + I2SCommFmt i2s_comm_fmt_{I2SCommFmt::STANDARD}; +}; + +} // namespace esphome::i2s_audio + +#endif // USE_ESP32