OmniX-Space/xiaozhi-esp32
1
0
Fork 0
Code Issues Pull Requests Actions Releases Activity

Update to 2.0.1

Browse Source
This commit is contained in:
苏雨洛
2025-09-15 22:04:01 +08:00
parent 5c43129024
commit 7d7f5eae3d
74 changed files with 5253 additions and 439 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
main/audio/audio_codec.cc
View File

@@ -35,17 +35,28 @@ void AudioCodec::Start() {
}
// 保存原始输出采样率
if (original_output_sample_rate_ == 0){
if (original_output_sample_rate_ == 0) {
original_output_sample_rate_ = output_sample_rate_;
ESP_LOGI(TAG, "Saved original output sample rate: %d Hz", original_output_sample_rate_);
}
if (tx_handle_ != nullptr) {
ESP_ERROR_CHECK(i2s_channel_enable(tx_handle_));
esp_err_t err = i2s_channel_enable(tx_handle_);
if (err == ESP_ERR_INVALID_STATE) {
// 已经启用，忽略
ESP_LOGW(TAG, "TX channel already enabled");
} else {
ESP_ERROR_CHECK(err);
}
}
if (rx_handle_ != nullptr) {
ESP_ERROR_CHECK(i2s_channel_enable(rx_handle_));
esp_err_t err = i2s_channel_enable(rx_handle_);
if (err == ESP_ERR_INVALID_STATE) {
ESP_LOGW(TAG, "RX channel already enabled");
} else {
ESP_ERROR_CHECK(err);
}
}
EnableInput(true);

2
main/audio/audio_codec.h
View File

@@ -59,4 +59,4 @@ protected:
virtual int Write(const int16_t* data, int samples) = 0;
};
#endif // _AUDIO_CODEC_H
#endif // _AUDIO_CODEC_H

1
main/audio/audio_service.cc
View File

@@ -672,6 +672,7 @@ void AudioService::SetModelsList(srmodel_list_t* models_list) {
models_list_ = models_list;
}
void AudioService::UpdateOutputTimestamp() {
last_output_time_ = std::chrono::steady_clock::now();
}

2
main/audio/audio_service.h
View File

@@ -108,9 +108,7 @@ public:
bool ReadAudioData(std::vector<int16_t>& data, int sample_rate, int samples);
void ResetDecoder();
void SetModelsList(srmodel_list_t* models_list);
void UpdateOutputTimestamp();
private:
AudioCodec* codec_ = nullptr;
AudioServiceCallbacks callbacks_;

28
main/audio/codecs/es8388_audio_codec.cc
View File

@@ -6,13 +6,14 @@
Es8388AudioCodec::Es8388AudioCodec(void* i2c_master_handle, i2c_port_t i2c_port, int input_sample_rate, int output_sample_rate,
gpio_num_t mclk, gpio_num_t bclk, gpio_num_t ws, gpio_num_t dout, gpio_num_t din,
gpio_num_t pa_pin, uint8_t es8388_addr) {
gpio_num_t pa_pin, uint8_t es8388_addr, bool input_reference) {
duplex_ = true; // 是否双工
input_reference_ = false; // 是否使用参考输入，实现回声消除
input_channels_ = 1; // 输入通道数
input_reference_ = input_reference; // 是否使用参考输入，实现回声消除
input_channels_ = input_reference_ ? 2 : 1; // 输入通道数
input_sample_rate_ = input_sample_rate;
output_sample_rate_ = output_sample_rate;
pa_pin_ = pa_pin; CreateDuplexChannels(mclk, bclk, ws, dout, din);
pa_pin_ = pa_pin;
CreateDuplexChannels(mclk, bclk, ws, dout, din);
// Do initialize of related interface: data_if, ctrl_if and gpio_if
audio_codec_i2s_cfg_t i2s_cfg = {
@@ -144,13 +145,21 @@ void Es8388AudioCodec::EnableInput(bool enable) {
if (enable) {
esp_codec_dev_sample_info_t fs = {
.bits_per_sample = 16,
.channel = 1,
.channel_mask = 0,
.channel = (uint8_t) input_channels_,
.channel_mask = ESP_CODEC_DEV_MAKE_CHANNEL_MASK(0),
.sample_rate = (uint32_t)input_sample_rate_,
.mclk_multiple = 0,
};
if (input_reference_) {
fs.channel_mask |= ESP_CODEC_DEV_MAKE_CHANNEL_MASK(1);
}
ESP_ERROR_CHECK(esp_codec_dev_open(input_dev_, &fs));
ESP_ERROR_CHECK(esp_codec_dev_set_in_gain(input_dev_, 24.0));
if (input_reference_) {
uint8_t gain = (11 << 4) + 0;
ctrl_if_->write_reg(ctrl_if_, 0x09, 1, &gain, 1);
}else{
ESP_ERROR_CHECK(esp_codec_dev_set_in_gain(input_dev_, 24.0));
}
} else {
ESP_ERROR_CHECK(esp_codec_dev_close(input_dev_));
}
@@ -175,6 +184,9 @@ void Es8388AudioCodec::EnableOutput(bool enable) {
// Set analog output volume to 0dB, default is -45dB
uint8_t reg_val = 30; // 0dB
if(input_reference_){
reg_val = 27;
}
uint8_t regs[] = { 46, 47, 48, 49 }; // HP_LVOL, HP_RVOL, SPK_LVOL, SPK_RVOL
for (uint8_t reg : regs) {
ctrl_if_->write_reg(ctrl_if_, reg, 1, &reg_val, 1);
@@ -200,7 +212,7 @@ int Es8388AudioCodec::Read(int16_t* dest, int samples) {
}
int Es8388AudioCodec::Write(const int16_t* data, int samples) {
if (output_enabled_) {
if (output_enabled_ && output_dev_ && data != nullptr) {
ESP_ERROR_CHECK_WITHOUT_ABORT(esp_codec_dev_write(output_dev_, (void*)data, samples * sizeof(int16_t)));
}
return samples;

2
main/audio/codecs/es8388_audio_codec.h
View File

@@ -29,7 +29,7 @@ private:
public:
Es8388AudioCodec(void* i2c_master_handle, i2c_port_t i2c_port, int input_sample_rate, int output_sample_rate,
gpio_num_t mclk, gpio_num_t bclk, gpio_num_t ws, gpio_num_t dout, gpio_num_t din,
gpio_num_t pa_pin, uint8_t es8388_addr);
gpio_num_t pa_pin, uint8_t es8388_addr, bool input_reference = false);
virtual ~Es8388AudioCodec();
virtual void SetOutputVolume(int volume) override;

104
main/audio/codecs/no_audio_codec.cc
View File

@@ -105,8 +105,8 @@ NoAudioCodecSimplex::NoAudioCodecSimplex(int input_sample_rate, int output_sampl
.slot_cfg = {
.data_bit_width = I2S_DATA_BIT_WIDTH_32BIT,
.slot_bit_width = I2S_SLOT_BIT_WIDTH_AUTO,
.slot_mode = I2S_SLOT_MODE_MONO,
.slot_mask = I2S_STD_SLOT_LEFT,
.slot_mode = I2S_SLOT_MODE_STEREO,
.slot_mask = I2S_STD_SLOT_BOTH,
.ws_width = I2S_DATA_BIT_WIDTH_32BIT,
.ws_pol = false,
.bit_shift = true,
@@ -136,6 +136,9 @@ NoAudioCodecSimplex::NoAudioCodecSimplex(int input_sample_rate, int output_sampl
chan_cfg.id = (i2s_port_t)1;
ESP_ERROR_CHECK(i2s_new_channel(&chan_cfg, nullptr, &rx_handle_));
std_cfg.clk_cfg.sample_rate_hz = (uint32_t)input_sample_rate_;
// RX 使用单声道 LEFT
std_cfg.slot_cfg.slot_mode = I2S_SLOT_MODE_MONO;
std_cfg.slot_cfg.slot_mask = I2S_STD_SLOT_LEFT;
std_cfg.gpio_cfg.bclk = mic_sck;
std_cfg.gpio_cfg.ws = mic_ws;
std_cfg.gpio_cfg.dout = I2S_GPIO_UNUSED;
@@ -214,6 +217,84 @@ NoAudioCodecSimplex::NoAudioCodecSimplex(int input_sample_rate, int output_sampl
ESP_LOGI(TAG, "Simplex channels created");
}
NoAudioCodecSimplexPdm::NoAudioCodecSimplexPdm(int input_sample_rate, int output_sample_rate, gpio_num_t spk_bclk, gpio_num_t spk_ws, gpio_num_t spk_dout, i2s_std_slot_mask_t spk_slot_mask,gpio_num_t mic_sck, gpio_num_t mic_din) {
duplex_ = false;
input_sample_rate_ = input_sample_rate;
output_sample_rate_ = output_sample_rate;
// Create a new channel for speaker
i2s_chan_config_t tx_chan_cfg = I2S_CHANNEL_DEFAULT_CONFIG((i2s_port_t)1, I2S_ROLE_MASTER);
tx_chan_cfg.dma_desc_num = AUDIO_CODEC_DMA_DESC_NUM;
tx_chan_cfg.dma_frame_num = AUDIO_CODEC_DMA_FRAME_NUM;
tx_chan_cfg.auto_clear_after_cb = true;
tx_chan_cfg.auto_clear_before_cb = false;
tx_chan_cfg.intr_priority = 0;
ESP_ERROR_CHECK(i2s_new_channel(&tx_chan_cfg, &tx_handle_, NULL));
i2s_std_config_t tx_std_cfg = {
.clk_cfg = {
.sample_rate_hz = (uint32_t)output_sample_rate_,
.clk_src = I2S_CLK_SRC_DEFAULT,
.mclk_multiple = I2S_MCLK_MULTIPLE_256,
#ifdef I2S_HW_VERSION_2
.ext_clk_freq_hz = 0,
#endif
},
.slot_cfg = {
.data_bit_width = I2S_DATA_BIT_WIDTH_32BIT,
.slot_bit_width = I2S_SLOT_BIT_WIDTH_AUTO,
.slot_mode = I2S_SLOT_MODE_MONO,
.slot_mask = spk_slot_mask,
.ws_width = I2S_DATA_BIT_WIDTH_32BIT,
.ws_pol = false,
.bit_shift = true,
#ifdef I2S_HW_VERSION_2
.left_align = true,
.big_endian = false,
.bit_order_lsb = false
#endif
},
.gpio_cfg = {
.mclk = I2S_GPIO_UNUSED,
.bclk = spk_bclk,
.ws = spk_ws,
.dout = spk_dout,
.din = I2S_GPIO_UNUSED,
.invert_flags = {
.mclk_inv = false,
.bclk_inv = false,
.ws_inv = false,
},
},
};
ESP_ERROR_CHECK(i2s_channel_init_std_mode(tx_handle_, &tx_std_cfg));
#if SOC_I2S_SUPPORTS_PDM_RX
// Create a new channel for MIC in PDM mode
i2s_chan_config_t rx_chan_cfg = I2S_CHANNEL_DEFAULT_CONFIG((i2s_port_t)0, I2S_ROLE_MASTER);
ESP_ERROR_CHECK(i2s_new_channel(&rx_chan_cfg, NULL, &rx_handle_));
i2s_pdm_rx_config_t pdm_rx_cfg = {
.clk_cfg = I2S_PDM_RX_CLK_DEFAULT_CONFIG((uint32_t)input_sample_rate_),
/* The data bit-width of PDM mode is fixed to 16 */
.slot_cfg = I2S_PDM_RX_SLOT_DEFAULT_CONFIG(I2S_DATA_BIT_WIDTH_16BIT, I2S_SLOT_MODE_MONO),
.gpio_cfg = {
.clk = mic_sck,
.din = mic_din,
.invert_flags = {
.clk_inv = false,
},
},
};
ESP_ERROR_CHECK(i2s_channel_init_pdm_rx_mode(rx_handle_, &pdm_rx_cfg));
#else
ESP_LOGE(TAG, "PDM is not supported");
#endif
ESP_LOGI(TAG, "Simplex channels created");
}
NoAudioCodecSimplexPdm::NoAudioCodecSimplexPdm(int input_sample_rate, int output_sample_rate, gpio_num_t spk_bclk, gpio_num_t spk_ws, gpio_num_t spk_dout, gpio_num_t mic_sck, gpio_num_t mic_din) {
duplex_ = false;
input_sample_rate_ = input_sample_rate;
@@ -280,25 +361,30 @@ NoAudioCodecSimplexPdm::NoAudioCodecSimplexPdm(int input_sample_rate, int output
int NoAudioCodec::Write(const int16_t* data, int samples) {
std::lock_guard<std::mutex> lock(data_if_mutex_);
std::vector<int32_t> buffer(samples);
// 立体声交织输出：L,R,L,R ... （每声道32位）
std::vector<int32_t> buffer(samples * 2);
// output_volume_: 0-100
// volume_factor_: 0-65536
int32_t volume_factor = pow(double(output_volume_) / 100.0, 2) * 65536;
for (int i = 0; i < samples; i++) {
int64_t temp = int64_t(data[i]) * volume_factor; // 使用 int64_t 进行乘法运算
int32_t s32;
if (temp > INT32_MAX) {
buffer[i] = INT32_MAX;
s32 = INT32_MAX;
} else if (temp < INT32_MIN) {
buffer[i] = INT32_MIN;
s32 = INT32_MIN;
} else {
buffer[i] = static_cast<int32_t>(temp);
s32 = static_cast<int32_t>(temp);
}
// 交织到左右声道
buffer[2 * i] = s32; // Left
buffer[2 * i + 1] = s32; // Right（复制）
}
size_t bytes_written;
ESP_ERROR_CHECK(i2s_channel_write(tx_handle_, buffer.data(), samples * sizeof(int32_t), &bytes_written, portMAX_DELAY));
return bytes_written / sizeof(int32_t);
ESP_ERROR_CHECK(i2s_channel_write(tx_handle_, buffer.data(), (samples * 2) * sizeof(int32_t), &bytes_written, portMAX_DELAY));
return bytes_written / sizeof(int32_t) / 2; // 返回每声道样本数
}
int NoAudioCodec::Read(int16_t* dest, int samples) {
@@ -329,4 +415,4 @@ int NoAudioCodecSimplexPdm::Read(int16_t* dest, int samples) {
// 计算实际读取的样本数
return bytes_read / sizeof(int16_t);
}
}

3
main/audio/codecs/no_audio_codec.h
View File

@@ -32,7 +32,8 @@ public:
class NoAudioCodecSimplexPdm : public NoAudioCodec {
public:
NoAudioCodecSimplexPdm(int input_sample_rate, int output_sample_rate, gpio_num_t spk_bclk, gpio_num_t spk_ws, gpio_num_t spk_dout, gpio_num_t mic_sck, gpio_num_t mic_din);
NoAudioCodecSimplexPdm(int input_sample_rate, int output_sample_rate, gpio_num_t spk_bclk, gpio_num_t spk_ws, gpio_num_t spk_dout, i2s_std_slot_mask_t spk_slot_mask, gpio_num_t mic_sck, gpio_num_t mic_din);
int Read(int16_t* dest, int samples);
};
#endif // _NO_AUDIO_CODEC_H
#endif // _NO_AUDIO_CODEC_H