From 9431848bea57f31d068008c984bf34f19d5529d8 Mon Sep 17 00:00:00 2001 From: John Boiles Date: Sat, 11 Jan 2025 03:02:38 -0800 Subject: [PATCH] SPDIF Speaker support --- CODEOWNERS | 2 + esphome/components/spdif_audio/__init__.py | 7 + .../spdif_audio/speaker/__init__.py | 61 ++ .../components/spdif_audio/speaker/spdif.cpp | 90 +++ .../components/spdif_audio/speaker/spdif.h | 52 ++ .../spdif_audio/speaker/spdif_speaker.cpp | 540 ++++++++++++++++++ .../spdif_audio/speaker/spdif_speaker.h | 132 +++++ 7 files changed, 884 insertions(+) create mode 100644 esphome/components/spdif_audio/__init__.py create mode 100644 esphome/components/spdif_audio/speaker/__init__.py create mode 100644 esphome/components/spdif_audio/speaker/spdif.cpp create mode 100644 esphome/components/spdif_audio/speaker/spdif.h create mode 100644 esphome/components/spdif_audio/speaker/spdif_speaker.cpp create mode 100644 esphome/components/spdif_audio/speaker/spdif_speaker.h diff --git a/CODEOWNERS b/CODEOWNERS index 404ad35efc..609f3ddd03 100644 --- a/CODEOWNERS +++ b/CODEOWNERS @@ -382,6 +382,8 @@ esphome/components/smt100/* @piechade esphome/components/sn74hc165/* @jesserockz esphome/components/socket/* @esphome/core esphome/components/sonoff_d1/* @anatoly-savchenkov +esphome/components/spdif_audio/* @johnboiles +esphome/components/spdif_audio/speaker/* @johnboiles esphome/components/speaker/* @jesserockz @kahrendt esphome/components/spi/* @clydebarrow @esphome/core esphome/components/spi_device/* @clydebarrow diff --git a/esphome/components/spdif_audio/__init__.py b/esphome/components/spdif_audio/__init__.py new file mode 100644 index 0000000000..e03a07528e --- /dev/null +++ b/esphome/components/spdif_audio/__init__.py @@ -0,0 +1,7 @@ +import esphome.codegen as cg + +CODEOWNERS = ["@johnboiles"] +DEPENDENCIES = ["esp32"] +MULTI_CONF = True + +spdif_audio_ns = cg.esphome_ns.namespace("spdif_audio") diff --git a/esphome/components/spdif_audio/speaker/__init__.py b/esphome/components/spdif_audio/speaker/__init__.py new file mode 100644 index 0000000000..397911ab85 --- /dev/null +++ b/esphome/components/spdif_audio/speaker/__init__.py @@ -0,0 +1,61 @@ +from esphome import pins +import esphome.codegen as cg +from esphome.components import speaker +from esphome.components.i2s_audio import I2SAudioComponent +import esphome.config_validation as cv +from esphome.const import ( + CONF_DATA_PIN, + CONF_DEBUG, + CONF_ID, + CONF_SAMPLE_RATE, + CONF_TIMEOUT, +) + +from .. import spdif_audio_ns + +DEPENDENCIES = ["i2s_audio"] +CODEOWNERS = ["@johnboiles"] + +SPDIFSpeaker = spdif_audio_ns.class_( + "SPDIFSpeaker", cg.Component, speaker.Speaker, I2SAudioComponent +) + +CONF_BUFFER_DURATION = "buffer_duration" +CONF_NEVER = "never" + +CONF_I2S_AUDIO_ID = "i2s_audio_id" + +CONF_FILL_SILENCE = "fill_silence" + +CONFIG_SCHEMA = speaker.SPEAKER_SCHEMA.extend( + { + cv.GenerateID(): cv.declare_id(SPDIFSpeaker), + cv.GenerateID(CONF_I2S_AUDIO_ID): cv.use_id(I2SAudioComponent), + cv.Required(CONF_DATA_PIN): pins.internal_gpio_output_pin_number, + cv.Optional(CONF_SAMPLE_RATE, default=48000): cv.positive_int, + cv.Optional( + CONF_BUFFER_DURATION, default="500ms" + ): cv.positive_time_period_milliseconds, + cv.Optional(CONF_TIMEOUT, default="500ms"): cv.Any( + cv.positive_time_period_milliseconds, + cv.one_of(CONF_NEVER, lower=True), + ), + cv.Optional(CONF_FILL_SILENCE, default=True): cv.boolean, + cv.Optional(CONF_DEBUG, default=False): cv.boolean, + } +).extend(cv.COMPONENT_SCHEMA) + + +async def to_code(config): + var = cg.new_Pvariable(config[CONF_ID]) + await cg.register_component(var, config) + await cg.register_parented(var, config[CONF_I2S_AUDIO_ID]) + await speaker.register_speaker(var, config) + + cg.add(var.set_data_pin(config[CONF_DATA_PIN])) + cg.add(var.set_sample_rate(config[CONF_SAMPLE_RATE])) + cg.add(var.set_buffer_duration(config[CONF_BUFFER_DURATION])) + if config[CONF_TIMEOUT] != CONF_NEVER: + cg.add(var.set_timeout(config[CONF_TIMEOUT])) + cg.add_define("SPDIF_FILL_SILENCE", config[CONF_FILL_SILENCE]) + cg.add_define("SPDIF_DEBUG", config[CONF_DEBUG]) diff --git a/esphome/components/spdif_audio/speaker/spdif.cpp b/esphome/components/spdif_audio/speaker/spdif.cpp new file mode 100644 index 0000000000..dfde370085 --- /dev/null +++ b/esphome/components/spdif_audio/speaker/spdif.cpp @@ -0,0 +1,90 @@ +#include "spdif.h" + +#include "esphome/core/defines.h" +#include "esphome/core/log.h" + +#include +namespace esphome { +namespace spdif_audio { + +static const char *const TAG = "spdif"; + +/* + * 8bit PCM to 16bit BMC conversion table, LSb first, 1 end + */ +static const uint16_t BMC_TABLE[256] = { + 0x3333, 0xb333, 0xd333, 0x5333, 0xcb33, 0x4b33, 0x2b33, 0xab33, 0xcd33, 0x4d33, 0x2d33, 0xad33, 0x3533, 0xb533, + 0xd533, 0x5533, 0xccb3, 0x4cb3, 0x2cb3, 0xacb3, 0x34b3, 0xb4b3, 0xd4b3, 0x54b3, 0x32b3, 0xb2b3, 0xd2b3, 0x52b3, + 0xcab3, 0x4ab3, 0x2ab3, 0xaab3, 0xccd3, 0x4cd3, 0x2cd3, 0xacd3, 0x34d3, 0xb4d3, 0xd4d3, 0x54d3, 0x32d3, 0xb2d3, + 0xd2d3, 0x52d3, 0xcad3, 0x4ad3, 0x2ad3, 0xaad3, 0x3353, 0xb353, 0xd353, 0x5353, 0xcb53, 0x4b53, 0x2b53, 0xab53, + 0xcd53, 0x4d53, 0x2d53, 0xad53, 0x3553, 0xb553, 0xd553, 0x5553, 0xcccb, 0x4ccb, 0x2ccb, 0xaccb, 0x34cb, 0xb4cb, + 0xd4cb, 0x54cb, 0x32cb, 0xb2cb, 0xd2cb, 0x52cb, 0xcacb, 0x4acb, 0x2acb, 0xaacb, 0x334b, 0xb34b, 0xd34b, 0x534b, + 0xcb4b, 0x4b4b, 0x2b4b, 0xab4b, 0xcd4b, 0x4d4b, 0x2d4b, 0xad4b, 0x354b, 0xb54b, 0xd54b, 0x554b, 0x332b, 0xb32b, + 0xd32b, 0x532b, 0xcb2b, 0x4b2b, 0x2b2b, 0xab2b, 0xcd2b, 0x4d2b, 0x2d2b, 0xad2b, 0x352b, 0xb52b, 0xd52b, 0x552b, + 0xccab, 0x4cab, 0x2cab, 0xacab, 0x34ab, 0xb4ab, 0xd4ab, 0x54ab, 0x32ab, 0xb2ab, 0xd2ab, 0x52ab, 0xcaab, 0x4aab, + 0x2aab, 0xaaab, 0xcccd, 0x4ccd, 0x2ccd, 0xaccd, 0x34cd, 0xb4cd, 0xd4cd, 0x54cd, 0x32cd, 0xb2cd, 0xd2cd, 0x52cd, + 0xcacd, 0x4acd, 0x2acd, 0xaacd, 0x334d, 0xb34d, 0xd34d, 0x534d, 0xcb4d, 0x4b4d, 0x2b4d, 0xab4d, 0xcd4d, 0x4d4d, + 0x2d4d, 0xad4d, 0x354d, 0xb54d, 0xd54d, 0x554d, 0x332d, 0xb32d, 0xd32d, 0x532d, 0xcb2d, 0x4b2d, 0x2b2d, 0xab2d, + 0xcd2d, 0x4d2d, 0x2d2d, 0xad2d, 0x352d, 0xb52d, 0xd52d, 0x552d, 0xccad, 0x4cad, 0x2cad, 0xacad, 0x34ad, 0xb4ad, + 0xd4ad, 0x54ad, 0x32ad, 0xb2ad, 0xd2ad, 0x52ad, 0xcaad, 0x4aad, 0x2aad, 0xaaad, 0x3335, 0xb335, 0xd335, 0x5335, + 0xcb35, 0x4b35, 0x2b35, 0xab35, 0xcd35, 0x4d35, 0x2d35, 0xad35, 0x3535, 0xb535, 0xd535, 0x5535, 0xccb5, 0x4cb5, + 0x2cb5, 0xacb5, 0x34b5, 0xb4b5, 0xd4b5, 0x54b5, 0x32b5, 0xb2b5, 0xd2b5, 0x52b5, 0xcab5, 0x4ab5, 0x2ab5, 0xaab5, + 0xccd5, 0x4cd5, 0x2cd5, 0xacd5, 0x34d5, 0xb4d5, 0xd4d5, 0x54d5, 0x32d5, 0xb2d5, 0xd2d5, 0x52d5, 0xcad5, 0x4ad5, + 0x2ad5, 0xaad5, 0x3355, 0xb355, 0xd355, 0x5355, 0xcb55, 0x4b55, 0x2b55, 0xab55, 0xcd55, 0x4d55, 0x2d55, 0xad55, + 0x3555, 0xb555, 0xd555, 0x5555, +}; + +// BMC preamble +static const uint32_t BMC_B = 0x33173333; // block start +static const uint32_t BMC_M = 0x331d3333; // left ch +static const uint32_t BMC_W = 0x331b3333; // right ch +static const uint32_t BMC_MW_DIF = (BMC_M ^ BMC_W); +static const uint8_t SYNC_OFFSET = 2; // byte offset of SYNC +static const uint32_t SYNC_FLIP = ((BMC_B ^ BMC_M) >> (SYNC_OFFSET * 8)); + +// initialize S/PDIF buffer +void SPDIF::setup() { + ESP_LOGCONFIG(TAG, "Setting up SPDIF..."); + + int i; + uint32_t bmc_mw = BMC_W; + + for (i = 0; i < SPDIF_BLOCK_SIZE_U32; i += 2) { + spdif_block_buf_[i] = bmc_mw ^= BMC_MW_DIF; + } + ESP_LOGD(TAG, "SPDIF buffer initialized to %zu bytes", sizeof(spdif_block_buf_)); + + spdif_block_ptr_ = spdif_block_buf_; +} + +esp_err_t SPDIF::write(const uint8_t *src, size_t size, TickType_t ticks_to_wait) { + const uint8_t *p = reinterpret_cast(src); + + while (p < (uint8_t *) src + size) { + // convert PCM 16bit data to BMC 32bit pulse pattern (which is 64 i2s bits to emulate BMC) + // We cast to int16_t to avoid sign extension issues when XOR-ing + *(spdif_block_ptr_ + 1) = + (uint32_t) (((static_cast(BMC_TABLE[*p]) << 16) ^ static_cast(BMC_TABLE[*(p + 1)])) << 1) >> + 1; + + p += 2; + spdif_block_ptr_ += 2; // advance to next audio data + + if (spdif_block_ptr_ >= &spdif_block_buf_[SPDIF_BLOCK_SIZE_U32]) { + // set block start preamble + ((uint8_t *) spdif_block_buf_)[SYNC_OFFSET] ^= SYNC_FLIP; + + esp_err_t err = block_complete_callback_(spdif_block_buf_, sizeof(spdif_block_buf_), ticks_to_wait); + if (err != ESP_OK) { + return err; + } + + spdif_block_ptr_ = spdif_block_buf_; + } + } + + return ESP_OK; +} + +} // namespace spdif_audio +} // namespace esphome diff --git a/esphome/components/spdif_audio/speaker/spdif.h b/esphome/components/spdif_audio/speaker/spdif.h new file mode 100644 index 0000000000..38b1786cbe --- /dev/null +++ b/esphome/components/spdif_audio/speaker/spdif.h @@ -0,0 +1,52 @@ +#pragma once + +#include +#include +#include + +// Number of samples in a SPDIF block +static const uint16_t SPDIF_BLOCK_SAMPLES = 192; +// A SPDIF sample is 64-bit +static const uint8_t SPDIF_BITS_PER_SAMPLE = 64; +// To emulate bi-phase mark code (BMC) (aka differential Manchester encoding) we are send +// twice as many bits per sample so that we can generate the transitions this encoding requires. +static const uint8_t EMULATED_BMC_BITS_PER_SAMPLE = SPDIF_BITS_PER_SAMPLE * 2; +#define SPDIF_BLOCK_SIZE_BYTES (SPDIF_BLOCK_SAMPLES * (EMULATED_BMC_BITS_PER_SAMPLE / 8)) +#define SPDIF_BLOCK_SIZE_U32 (SPDIF_BLOCK_SIZE_BYTES / sizeof(uint32_t)) // One block, 1536 bytes + +namespace esphome { +namespace spdif_audio { + +class SPDIF { + public: + /// @brief Initialize the BMC lookup table and working buffer + void setup(); + + /// @brief Function to call when a block of data is complete (called from write) + void set_block_complete_callback( + std::function callback) { + block_complete_callback_ = std::move(callback); + } + + /// @brief Convert PCM audio data to SPDIF BMC encoded data + /// @param src Source PCM audio data + /// @param size Size of source data in bytes + /// @return esp_err_t as returned from block_complete_callback_ + esp_err_t write(const uint8_t *src, size_t size, TickType_t ticks_to_wait); + + /// @brief Reset the SPDIF block buffer + void reset() { spdif_block_ptr_ = spdif_block_buf_; } + + protected: + // BMC lookup table for converting 8-bits to 16-bit emulated BMC waveform + static const uint16_t BMC_LOOKUP_TABLE[256]; + + std::function block_complete_callback_; + + // Working buffer that holds an entire SPDIF block ready for I2S output + uint32_t spdif_block_buf_[SPDIF_BLOCK_SIZE_U32]; + uint32_t *spdif_block_ptr_{nullptr}; +}; + +} // namespace spdif_audio +} // namespace esphome diff --git a/esphome/components/spdif_audio/speaker/spdif_speaker.cpp b/esphome/components/spdif_audio/speaker/spdif_speaker.cpp new file mode 100644 index 0000000000..5a31005436 --- /dev/null +++ b/esphome/components/spdif_audio/speaker/spdif_speaker.cpp @@ -0,0 +1,540 @@ +#include "spdif_speaker.h" + +#ifdef USE_ESP32 + +#include +#include + +#include "esphome/components/audio/audio.h" + +#include "esphome/core/application.h" +#include "esphome/core/hal.h" +#include "esphome/core/log.h" + +namespace esphome { +namespace spdif_audio { + +static const size_t DMA_BUFFERS_COUNT = 4; + +static const size_t TASK_STACK_SIZE = 4096; +static const ssize_t TASK_PRIORITY = 23; + +static const size_t I2S_EVENT_QUEUE_COUNT = DMA_BUFFERS_COUNT + 1; + +static const char *const TAG = "spdif_audio.speaker"; + +#if SPDIF_FILL_SILENCE +// A full BMC block's worth of 16-big stereo samples +int16_t silence[SPDIF_BLOCK_SAMPLES * 2]; +#endif + +enum SpeakerEventGroupBits : uint32_t { + COMMAND_START = (1 << 0), // starts the 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 + STATE_STARTING = (1 << 10), + STATE_RUNNING = (1 << 11), + STATE_STOPPING = (1 << 12), + STATE_STOPPED = (1 << 13), + ERR_TASK_FAILED_TO_START = (1 << 14), + ERR_ESP_INVALID_STATE = (1 << 15), + ERR_ESP_NOT_SUPPORTED = (1 << 16), + ERR_ESP_INVALID_ARG = (1 << 17), + ERR_ESP_INVALID_SIZE = (1 << 18), + ERR_ESP_NO_MEM = (1 << 19), + ERR_ESP_FAIL = (1 << 20), + ALL_ERR_ESP_BITS = ERR_ESP_INVALID_STATE | ERR_ESP_NOT_SUPPORTED | ERR_ESP_INVALID_ARG | ERR_ESP_INVALID_SIZE | + ERR_ESP_NO_MEM | ERR_ESP_FAIL, + ALL_BITS = 0x00FFFFFF, // All valid FreeRTOS event group bits +}; + +// Translates a SpeakerEventGroupBits ERR_ESP bit to the coressponding esp_err_t +static esp_err_t err_bit_to_esp_err(uint32_t bit) { + switch (bit) { + case SpeakerEventGroupBits::ERR_ESP_INVALID_STATE: + return ESP_ERR_INVALID_STATE; + case SpeakerEventGroupBits::ERR_ESP_INVALID_ARG: + return ESP_ERR_INVALID_ARG; + case SpeakerEventGroupBits::ERR_ESP_INVALID_SIZE: + return ESP_ERR_INVALID_SIZE; + case SpeakerEventGroupBits::ERR_ESP_NO_MEM: + return ESP_ERR_NO_MEM; + case SpeakerEventGroupBits::ERR_ESP_NOT_SUPPORTED: + return ESP_ERR_NOT_SUPPORTED; + default: + return ESP_FAIL; + } +} + +/// @brief Multiplies the input array of Q15 numbers by a Q15 constant factor +/// +/// Based on `dsps_mulc_s16_ansi` from the esp-dsp library: +/// https://github.com/espressif/esp-dsp/blob/master/modules/math/mulc/fixed/dsps_mulc_s16_ansi.c +/// (accessed on 2024-09-30). +/// @param input Array of Q15 numbers +/// @param output Array of Q15 numbers +/// @param len Length of array +/// @param c Q15 constant factor +static void q15_multiplication(const int16_t *input, int16_t *output, size_t len, int16_t c) { + for (int i = 0; i < len; i++) { + int32_t acc = (int32_t) input[i] * (int32_t) c; + output[i] = (int16_t) (acc >> 15); + } +} + +// 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) +// float to Q15 fixed point formula: q15_scale_factor = floating_point_scale_factor * 2^(15) +static const std::vector Q15_VOLUME_SCALING_FACTORS = { + 0, 116, 122, 130, 137, 146, 154, 163, 173, 183, 194, 206, 218, 231, 244, + 259, 274, 291, 308, 326, 345, 366, 388, 411, 435, 461, 488, 517, 548, 580, + 615, 651, 690, 731, 774, 820, 868, 920, 974, 1032, 1094, 1158, 1227, 1300, 1377, + 1459, 1545, 1637, 1734, 1837, 1946, 2061, 2184, 2313, 2450, 2596, 2750, 2913, 3085, 3269, + 3462, 3668, 3885, 4116, 4360, 4619, 4893, 5183, 5490, 5816, 6161, 6527, 6914, 7324, 7758, + 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 SPDIFSpeaker::setup() { + ESP_LOGCONFIG(TAG, "Setting up SPDIF Audio Speaker..."); + + this->event_group_ = xEventGroupCreate(); + + if (this->event_group_ == nullptr) { + ESP_LOGE(TAG, "Failed to create event group"); + this->mark_failed(); + return; + } + + this->spdif_->setup(); + this->spdif_->set_block_complete_callback([this](uint32_t *data, size_t size, TickType_t ticks_to_wait) -> esp_err_t { + size_t i2s_write_len; + return i2s_write(this->parent_->get_port(), data, size, &i2s_write_len, ticks_to_wait); + }); + +#if SPDIF_FILL_SILENCE + memset(silence, 0, sizeof(silence)); +#endif +} + +void SPDIFSpeaker::loop() { + uint32_t event_group_bits = xEventGroupGetBits(this->event_group_); + + if (event_group_bits & SpeakerEventGroupBits::STATE_STARTING) { + ESP_LOGD(TAG, "Starting Speaker"); + this->state_ = speaker::STATE_STARTING; + xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::STATE_STARTING); + } + if (event_group_bits & SpeakerEventGroupBits::STATE_RUNNING) { + ESP_LOGD(TAG, "Started Speaker"); + this->state_ = speaker::STATE_RUNNING; + xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::STATE_RUNNING); + this->status_clear_warning(); + this->status_clear_error(); + } + if (event_group_bits & SpeakerEventGroupBits::STATE_STOPPING) { + ESP_LOGD(TAG, "Stopping Speaker"); + this->state_ = speaker::STATE_STOPPING; + xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::STATE_STOPPING); + } + if (event_group_bits & SpeakerEventGroupBits::STATE_STOPPED) { + if (!this->task_created_) { + ESP_LOGD(TAG, "Stopped Speaker"); + this->state_ = speaker::STATE_STOPPED; + xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::ALL_BITS); + this->speaker_task_handle_ = nullptr; + } + } + + if (event_group_bits & SpeakerEventGroupBits::ERR_TASK_FAILED_TO_START) { + this->status_set_error("Failed to start speaker task"); + xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::ERR_TASK_FAILED_TO_START); + } + + if (event_group_bits & SpeakerEventGroupBits::ALL_ERR_ESP_BITS) { + uint32_t error_bits = event_group_bits & SpeakerEventGroupBits::ALL_ERR_ESP_BITS; + ESP_LOGW(TAG, "Error writing to I2S: %s", esp_err_to_name(err_bit_to_esp_err(error_bits))); + this->status_set_warning(); + } + + if (event_group_bits & SpeakerEventGroupBits::ERR_ESP_NOT_SUPPORTED) { + this->status_set_error("Failed to adjust I2S bus to match the incoming audio"); + ESP_LOGE(TAG, + "Incompatible audio format: sample rate = %" PRIu32 ", channels = %" PRIu8 ", bits per sample = %" PRIu8, + this->audio_stream_info_.sample_rate, this->audio_stream_info_.channels, + this->audio_stream_info_.bits_per_sample); + } +} + +void SPDIFSpeaker::set_volume(float volume) { + this->volume_ = volume; + // 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]; +} + +void SPDIFSpeaker::set_mute_state(bool mute_state) { + this->mute_state_ = mute_state; + if (mute_state) { + // Software volume control and scale by 0 + this->q15_volume_factor_ = 0; + } else { + // Revert to previous volume when unmuting + this->set_volume(this->volume_); + } +} + +size_t SPDIFSpeaker::play(const uint8_t *data, size_t length, TickType_t ticks_to_wait) { + if (this->is_failed()) { + ESP_LOGE(TAG, "Cannot play audio, speaker failed to setup"); + return 0; + } + if (this->state_ != speaker::STATE_RUNNING && this->state_ != speaker::STATE_STARTING) { + this->start(); + } + + size_t bytes_written = 0; + if ((this->state_ == speaker::STATE_RUNNING) && (this->audio_ring_buffer_.use_count() == 1)) { + // Only one owner of the ring buffer (the speaker task), so the ring buffer is allocated and no other components are + // attempting to write to it. + + // Temporarily share ownership of the ring buffer so it won't be deallocated while writing + std::shared_ptr temp_ring_buffer = this->audio_ring_buffer_; + bytes_written = temp_ring_buffer->write_without_replacement((void *) data, length, ticks_to_wait); + } + + return bytes_written; +} + +bool SPDIFSpeaker::has_buffered_data() const { + if (this->audio_ring_buffer_ != nullptr) { + return this->audio_ring_buffer_->available() > 0; + } + return false; +} + +void SPDIFSpeaker::speaker_task(void *params) { + SPDIFSpeaker *this_speaker = (SPDIFSpeaker *) params; + uint32_t event_group_bits = + xEventGroupWaitBits(this_speaker->event_group_, + SpeakerEventGroupBits::COMMAND_START | SpeakerEventGroupBits::COMMAND_STOP | + SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY, // Bit message to read + pdTRUE, // Clear the bits on exit + pdFALSE, // Don't wait for all the bits, + portMAX_DELAY); // Block indefinitely until a bit is set + + if (event_group_bits & (SpeakerEventGroupBits::COMMAND_STOP | SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY)) { + // Received a stop signal before the task was requested to start + this_speaker->delete_task_(0); + } + + xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::STATE_STARTING); + + audio::AudioStreamInfo audio_stream_info = this_speaker->audio_stream_info_; + + const uint32_t bytes_per_ms = + audio_stream_info.channels * audio_stream_info.get_bytes_per_sample() * audio_stream_info.sample_rate / 1000; + + const size_t dma_buffers_size = + DMA_BUFFERS_COUNT * SPDIF_BLOCK_SAMPLES * audio_stream_info.channels * audio_stream_info.get_bytes_per_sample(); + + int task_delay_ms = bytes_per_ms * DMA_BUFFERS_COUNT / 2; + + // Ensure ring buffer is at least as large as the total size of the DMA buffers + const size_t ring_buffer_size = + std::max((uint32_t) dma_buffers_size, this_speaker->buffer_duration_ms_ * bytes_per_ms); + + if (this_speaker->send_esp_err_to_event_group_(this_speaker->allocate_buffers_(dma_buffers_size, ring_buffer_size))) { + // Failed to allocate buffers + xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::ERR_ESP_NO_MEM); + this_speaker->delete_task_(dma_buffers_size); + } + + if (!this_speaker->send_esp_err_to_event_group_(this_speaker->start_i2s_driver_(audio_stream_info))) { + xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::STATE_RUNNING); + + bool stop_gracefully = false; + uint32_t last_data_received_time = millis(); + bool tx_dma_underflow = false; + + while (!this_speaker->timeout_.has_value() || + (millis() - last_data_received_time) <= this_speaker->timeout_.value()) { + event_group_bits = xEventGroupGetBits(this_speaker->event_group_); + + if (event_group_bits & SpeakerEventGroupBits::COMMAND_STOP) { + break; + } + if (event_group_bits & SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY) { + stop_gracefully = true; + } + + if (this_speaker->audio_stream_info_ != audio_stream_info) { + // Audio stream info has changed, stop the speaker task so it will restart with the proper settings. + + break; + } + + i2s_event_t i2s_event; + while (xQueueReceive(this_speaker->i2s_event_queue_, &i2s_event, 0)) { + if (i2s_event.type == I2S_EVENT_TX_Q_OVF) { +#if SPDIF_DEBUG + int64_t last_overflow_log_time = 0; + const int64_t min_log_interval_us = 1000000; + int64_t current_time = esp_timer_get_time(); + if (current_time - last_overflow_log_time >= min_log_interval_us) { + ESP_LOGE(TAG, "I2S_EVENT_TX_Q_OVF"); + last_overflow_log_time = current_time; + } +#endif +#if SPDIF_FILL_SILENCE + // Queue DMA a couple buffers full of silence when we don't have anything else to play + this_speaker->spdif_->reset(); + this_speaker->spdif_->write(reinterpret_cast(silence), sizeof(silence), 0); + this_speaker->spdif_->write(reinterpret_cast(silence), sizeof(silence), 0); + // this_speaker->spdif_->write(reinterpret_cast(silence), sizeof(silence), 0); + // this_speaker->spdif_->write(reinterpret_cast(silence), sizeof(silence), 0); +#endif + tx_dma_underflow = true; + } + } + + size_t bytes_to_read = dma_buffers_size; + + size_t bytes_read = this_speaker->audio_ring_buffer_->read((void *) this_speaker->data_buffer_, bytes_to_read, + pdMS_TO_TICKS(task_delay_ms)); + if (bytes_read > 0) { + if ((audio_stream_info.bits_per_sample == 16) && (this_speaker->q15_volume_factor_ < INT16_MAX)) { + // Scale samples by the volume factor in place + q15_multiplication((int16_t *) this_speaker->data_buffer_, (int16_t *) this_speaker->data_buffer_, + bytes_read / sizeof(int16_t), this_speaker->q15_volume_factor_); + } + + this_speaker->spdif_->write(this_speaker->data_buffer_, bytes_read, portMAX_DELAY); + +#if SPDIF_DEBUG + static uint64_t total_bytes = 0; + static uint64_t last_log_time = 0; + static uint64_t last_log_bytes = 0; + + total_bytes += bytes_read; + int64_t current_time = esp_timer_get_time(); + + if (last_log_time == 0) { + last_log_time = current_time; + last_log_bytes = total_bytes; + } + + // Check if it's time to log sample statistics (every minute) + if (current_time - last_log_time >= 5000000) { + uint64_t elapsed_time = current_time - last_log_time; + uint64_t bytes_since_last_log = total_bytes - last_log_bytes; + // 4 bytes per 16-bit stereo sample + uint64_t samples = bytes_since_last_log / 4; + float seconds = elapsed_time / 1000000.0f; + float hz = samples / seconds; + + ESP_LOGD(TAG, "%llu samples in %.2fs (%.2fHz)", samples, seconds, hz); + + // Reset for next log + last_log_time = current_time; + last_log_bytes = total_bytes; + } +#endif + + tx_dma_underflow = false; + last_data_received_time = millis(); + } else { + // No data received + if (stop_gracefully && tx_dma_underflow) { + break; + } + } + } + + xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::STATE_STOPPING); + + i2s_driver_uninstall(this_speaker->parent_->get_port()); + + this_speaker->parent_->unlock(); + } + + this_speaker->delete_task_(dma_buffers_size); +} + +void SPDIFSpeaker::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; + + if (this->speaker_task_handle_ == nullptr) { + xTaskCreate(SPDIFSpeaker::speaker_task, "speaker_task", TASK_STACK_SIZE, (void *) this, TASK_PRIORITY, + &this->speaker_task_handle_); + } + + if (this->speaker_task_handle_ != nullptr) { + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::COMMAND_START); + this->task_created_ = true; + } else { + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_TASK_FAILED_TO_START); + } +} + +void SPDIFSpeaker::stop() { this->stop_(false); } + +void SPDIFSpeaker::finish() { this->stop_(true); } + +void SPDIFSpeaker::stop_(bool wait_on_empty) { + if (this->is_failed()) + return; + if (this->state_ == speaker::STATE_STOPPED) + return; + + if (wait_on_empty) { + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY); + } else { + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::COMMAND_STOP); + } +} + +bool SPDIFSpeaker::send_esp_err_to_event_group_(esp_err_t err) { + switch (err) { + case ESP_OK: + return false; + case ESP_ERR_INVALID_STATE: + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_ESP_INVALID_STATE); + return true; + case ESP_ERR_INVALID_ARG: + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_ESP_INVALID_ARG); + return true; + case ESP_ERR_INVALID_SIZE: + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_ESP_INVALID_SIZE); + return true; + case ESP_ERR_NO_MEM: + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_ESP_NO_MEM); + return true; + case ESP_ERR_NOT_SUPPORTED: + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_ESP_NOT_SUPPORTED); + return true; + default: + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_ESP_FAIL); + return true; + } +} + +esp_err_t SPDIFSpeaker::allocate_buffers_(size_t data_buffer_size, size_t ring_buffer_size) { + if (this->data_buffer_ == nullptr) { + // Allocate data buffer for temporarily storing audio from the ring buffer before writing to the I2S bus + ExternalRAMAllocator allocator(ExternalRAMAllocator::ALLOW_FAILURE); + this->data_buffer_ = allocator.allocate(data_buffer_size); + } + + if (this->data_buffer_ == nullptr) { + ESP_LOGE(TAG, "Failed to allocate data_buffer_"); + return ESP_ERR_NO_MEM; + } + + if (this->audio_ring_buffer_.use_count() == 0) { + // Allocate ring buffer. Uses a shared_ptr to ensure it isn't improperly deallocated. + this->audio_ring_buffer_ = RingBuffer::create(ring_buffer_size); + } + + if (this->audio_ring_buffer_ == nullptr) { + ESP_LOGE(TAG, "Failed to allocate audio_ring_buffer_"); + return ESP_ERR_NO_MEM; + } + + return ESP_OK; +} + +esp_err_t SPDIFSpeaker::start_i2s_driver_(audio::AudioStreamInfo &audio_stream_info) { + if (this->sample_rate_ != audio_stream_info.sample_rate) { // NOLINT + // Can't reconfigure I2S bus, so the sample rate must match the configured value + ESP_LOGE(TAG, "SPDIF only supports a single sample rate"); + return ESP_ERR_NOT_SUPPORTED; + } + + // Currently only 16-bit samples are supported + if (audio_stream_info.bits_per_sample != 16) { + ESP_LOGE(TAG, "SPDIF only supports 16 bits per sample"); + return ESP_ERR_NOT_SUPPORTED; + } + + // Currently only stereo is supported + if (audio_stream_info.channels != 2) { + ESP_LOGE(TAG, "SPDIF only supports stereo"); + return ESP_ERR_NOT_SUPPORTED; + } + + if (!this->parent_->try_lock()) { + return ESP_ERR_INVALID_STATE; + } + + constexpr uint32_t i2s_bits_per_sample = 32; + uint32_t sample_rate = this->sample_rate_ * 2; + i2s_config_t config = { + .mode = static_cast(I2S_MODE_MASTER | I2S_MODE_TX), + .sample_rate = sample_rate, + .bits_per_sample = static_cast(i2s_bits_per_sample), + .channel_format = I2S_CHANNEL_FMT_RIGHT_LEFT, + .communication_format = I2S_COMM_FORMAT_STAND_I2S, + .intr_alloc_flags = 0, + .dma_buf_count = DMA_BUFFERS_COUNT, + .dma_buf_len = SPDIF_BLOCK_SIZE_U32, + .use_apll = true, +#if SPDIF_FILL_SILENCE + .tx_desc_auto_clear = false, +#else + .tx_desc_auto_clear = true, +#endif + .fixed_mclk = 0, + .mclk_multiple = I2S_MCLK_MULTIPLE_256, + .bits_per_chan = I2S_BITS_PER_CHAN_DEFAULT, + }; + + esp_err_t err = + i2s_driver_install(this->parent_->get_port(), &config, I2S_EVENT_QUEUE_COUNT, &this->i2s_event_queue_); + if (err != ESP_OK) { + // Failed to install the driver, so unlock the I2S port + this->parent_->unlock(); + return err; + } + + i2s_pin_config_t pin_config = { + .mck_io_num = -1, + .bck_io_num = -1, + .ws_io_num = -1, + .data_out_num = this->data_pin_, + .data_in_num = -1, + }; + + err = i2s_set_pin(this->parent_->get_port(), &pin_config); + + if (err != ESP_OK) { + // Failed to set the data out pin, so uninstall the driver and unlock the I2S port + i2s_driver_uninstall(this->parent_->get_port()); + this->parent_->unlock(); + } + + return err; +} + +void SPDIFSpeaker::delete_task_(size_t buffer_size) { + this->audio_ring_buffer_.reset(); // Releases ownership of the shared_ptr + + if (this->data_buffer_ != nullptr) { + ExternalRAMAllocator allocator(ExternalRAMAllocator::ALLOW_FAILURE); + allocator.deallocate(this->data_buffer_, buffer_size); + this->data_buffer_ = nullptr; + } + + xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::STATE_STOPPED); + + this->task_created_ = false; + vTaskDelete(nullptr); +} + +} // namespace spdif_audio +} // namespace esphome + +#endif // USE_ESP32 diff --git a/esphome/components/spdif_audio/speaker/spdif_speaker.h b/esphome/components/spdif_audio/speaker/spdif_speaker.h new file mode 100644 index 0000000000..4cdad84822 --- /dev/null +++ b/esphome/components/spdif_audio/speaker/spdif_speaker.h @@ -0,0 +1,132 @@ +#pragma once + +#ifdef USE_ESP32 + +#include "esphome/components/i2s_audio/i2s_audio.h" + +#include + +#include +#include +#include + +#include "spdif.h" +#include "esphome/components/audio/audio.h" +#include "esphome/components/speaker/speaker.h" + +#include "esphome/core/component.h" +#include "esphome/core/gpio.h" +#include "esphome/core/helpers.h" +#include "esphome/core/ring_buffer.h" + +namespace esphome { +namespace spdif_audio { + +class SPDIFSpeaker : public Parented, public speaker::Speaker, public Component { + public: + SPDIFSpeaker() : spdif_(new SPDIF()) {} + float get_setup_priority() const override { return esphome::setup_priority::PROCESSOR; } + + void setup() override; + void loop() override; + + 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_data_pin(uint8_t pin) { this->data_pin_ = pin; } + void set_sample_rate(uint32_t rate) { this->sample_rate_ = rate; } + + void start() override; + void stop() override; + void finish() override; + + /// @brief Plays the provided audio data. + /// Starts the speaker task, if necessary. Writes the audio data to the ring buffer. + /// @param data Audio data in the format set by the parent speaker classes ``set_audio_stream_info`` method. + /// @param length The length of the audio data in bytes. + /// @param ticks_to_wait The FreeRTOS ticks to wait before writing as much data as possible to the ring buffer. + /// @return The number of bytes that were actually written to the ring buffer. + 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 play(data, length, 0); } + + bool has_buffered_data() const override; + + /// @brief Sets the volume of the speaker. Uses the speaker's configured audio dac component. If unavailble, it is + /// implemented as a software volume control. Overrides the default setter to convert the floating point volume to a + /// Q15 fixed-point factor. + /// @param volume between 0.0 and 1.0 + void set_volume(float volume) override; + + /// @brief Mutes or unmute the speaker. Uses the speaker's configured audio dac component. If unavailble, it is + /// implemented as a software volume control. Overrides the default setter to convert the floating point volume to a + /// Q15 fixed-point factor. + /// @param mute_state true for muting, false for unmuting + void set_mute_state(bool mute_state) override; + + protected: + /// @brief Function for the FreeRTOS task handling audio output. + /// After receiving the COMMAND_START signal, allocates space for the buffers, starts the I2S driver, and 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, + /// stops the I2S driver, unlocks the I2S port, and deletes the task. It communicates the state and any errors via + /// event_group_. + /// @param params I2SAudioSpeaker component + static void speaker_task(void *params); + + /// @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 Sets the corresponding ERR_ESP event group bits. + /// @param err esp_err_t error code. + /// @return True if an ERR_ESP bit is set and false if err == ESP_OK + bool send_esp_err_to_event_group_(esp_err_t err); + + /// @brief Allocates the data buffer and ring buffer + /// @param data_buffer_size Number of bytes to allocate for the data buffer. + /// @param ring_buffer_size Number of bytes to allocate for the ring buffer. + /// @return ESP_ERR_NO_MEM if either buffer fails to allocate + /// ESP_OK if successful + esp_err_t allocate_buffers_(size_t data_buffer_size, size_t ring_buffer_size); + + /// @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 uninstall the driver, if necessary. + /// @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 nstalling 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); + + /// @brief Deletes the speaker's task. + /// Deallocates the data_buffer_ and audio_ring_buffer_, if necessary, and deletes the task. Should only be called by + /// the speaker_task itself. + /// @param buffer_size The allocated size of the data_buffer_. + void delete_task_(size_t buffer_size); + + SPDIF *spdif_{nullptr}; + TaskHandle_t speaker_task_handle_{nullptr}; + EventGroupHandle_t event_group_{nullptr}; + + QueueHandle_t i2s_event_queue_; + + uint8_t *data_buffer_{nullptr}; + std::shared_ptr audio_ring_buffer_; + + uint32_t buffer_duration_ms_; + + optional timeout_; + uint8_t data_pin_; + uint32_t sample_rate_; + + bool task_created_{false}; + + int16_t q15_volume_factor_{INT16_MAX}; +}; + +} // namespace spdif_audio +} // namespace esphome + +#endif // USE_ESP32