[i2s_audio] Add more options to speakers and microphones

Now I2S speakers also have the following options that already
exist for microphones:

 * `i2s_mode` (primary/secondary; default: primary)
 * `use_apll` (true/false, default: false)
 * `channel` (left/right/mono/stereo, default: mono)
 * `sample_rate` (default: 16000)
 * `bits_per_sample` (default: 16bit)

Both speakers and microphones also have:

 * `bits_per_channel` (default: equal to `bits_per_sample`)

`channel` has new values:

 * stereo: input/output data is interpreted as a sequence of
   (right, left) pairs (this option works for both speakers
   and microphones);
 * mono: for speakers, this is the old behavior - each sample
   from the input is sent to both left and right channels.

Also, speakers have an extra option:

 * `timeout` (default: 100ms) - how long the background tasks
   should wait before releasing the bus if the buffer is empty.

esphome/components/i2s_audio/__init__.py

@@ -3,7 +3,7 @@
 import esphome.codegen as cg
 
 from esphome import pins
-from esphome.const import CONF_ID
+from esphome.const import CONF_CHANNEL, CONF_ID, CONF_SAMPLE_RATE
 from esphome.components.esp32 import get_esp32_variant
 from esphome.components.esp32.const import (
     VARIANT_ESP32,
@@ -29,6 +29,14 @@
 CONF_PRIMARY = "primary"
 CONF_SECONDARY = "secondary"
 
+CONF_USE_APLL = "use_apll"
+CONF_BITS_PER_SAMPLE = "bits_per_sample"
+CONF_BITS_PER_CHANNEL = "bits_per_channel"
+CONF_MONO = "mono"
+CONF_LEFT = "left"
+CONF_RIGHT = "right"
+CONF_STEREO = "stereo"
+
 i2s_audio_ns = cg.esphome_ns.namespace("i2s_audio")
 I2SAudioComponent = i2s_audio_ns.class_("I2SAudioComponent", cg.Component)
 I2SAudioIn = i2s_audio_ns.class_("I2SAudioIn", cg.Parented.template(I2SAudioComponent))
@@ -42,6 +50,30 @@
     CONF_SECONDARY: i2s_mode_t.I2S_MODE_SLAVE,  # NOLINT
 }
 
+i2s_channel_fmt_t = cg.global_ns.enum("i2s_channel_fmt_t")
+I2S_CHANNELS = {
+    CONF_MONO: i2s_channel_fmt_t.I2S_CHANNEL_FMT_ALL_LEFT,
+    CONF_LEFT: i2s_channel_fmt_t.I2S_CHANNEL_FMT_ONLY_LEFT,
+    CONF_RIGHT: i2s_channel_fmt_t.I2S_CHANNEL_FMT_ONLY_RIGHT,
+    CONF_STEREO: i2s_channel_fmt_t.I2S_CHANNEL_FMT_RIGHT_LEFT,
+}
+
+i2s_bits_per_sample_t = cg.global_ns.enum("i2s_bits_per_sample_t")
+I2S_BITS_PER_SAMPLE = {
+    8: i2s_bits_per_sample_t.I2S_BITS_PER_SAMPLE_8BIT,
+    16: i2s_bits_per_sample_t.I2S_BITS_PER_SAMPLE_16BIT,
+    24: i2s_bits_per_sample_t.I2S_BITS_PER_SAMPLE_24BIT,
+    32: i2s_bits_per_sample_t.I2S_BITS_PER_SAMPLE_32BIT,
+}
+
+i2s_bits_per_chan_t = cg.global_ns.enum("i2s_bits_per_chan_t")
+I2S_BITS_PER_CHANNEL = {
+    8: i2s_bits_per_chan_t.I2S_BITS_PER_CHAN_8BIT,
+    16: i2s_bits_per_chan_t.I2S_BITS_PER_CHAN_16BIT,
+    24: i2s_bits_per_chan_t.I2S_BITS_PER_CHAN_24BIT,
+    32: i2s_bits_per_chan_t.I2S_BITS_PER_CHAN_32BIT,
+}
+
 # https://github.com/espressif/esp-idf/blob/master/components/soc/{variant}/include/soc/soc_caps.h
 I2S_PORTS = {
     VARIANT_ESP32: 2,
@@ -60,6 +92,27 @@
 )
 
 
+def i2s_component_schema(base_schema, klass, default_channel, default_bits_per_sample):
+    return base_schema.extend(
+        {
+            cv.GenerateID(): cv.declare_id(klass),
+            cv.GenerateID(CONF_I2S_AUDIO_ID): cv.use_id(I2SAudioComponent),
+            cv.Optional(CONF_I2S_MODE, default=CONF_PRIMARY): cv.enum(
+                I2S_MODE_OPTIONS, lower=True
+            ),
+            cv.Optional(CONF_USE_APLL, default=False): cv.boolean,
+            cv.Optional(CONF_CHANNEL, default=default_channel): cv.enum(I2S_CHANNELS),
+            cv.Optional(CONF_SAMPLE_RATE, default=16000): cv.int_range(min=1),
+            cv.Optional(CONF_BITS_PER_SAMPLE, default=default_bits_per_sample): cv.All(
+                cv.float_with_unit("bits", "bit"), cv.enum(I2S_BITS_PER_SAMPLE)
+            ),
+            cv.Optional(CONF_BITS_PER_CHANNEL): cv.All(
+                cv.float_with_unit("bits", "bit"), cv.enum(I2S_BITS_PER_CHANNEL)
+            ),
+        }
+    ).extend(cv.COMPONENT_SCHEMA)
+
+
 def _final_validate(_):
     i2s_audio_configs = fv.full_config.get()[CONF_I2S_AUDIO]
     variant = get_esp32_variant()
@@ -83,3 +136,19 @@ async def to_code(config):
         cg.add(var.set_bclk_pin(config[CONF_I2S_BCLK_PIN]))
     if CONF_I2S_MCLK_PIN in config:
         cg.add(var.set_mclk_pin(config[CONF_I2S_MCLK_PIN]))
+
+
+async def register_i2s_component(config):
+    var = cg.new_Pvariable(config[CONF_ID])
+    await cg.register_component(var, config)
+    await cg.register_parented(var, config[CONF_I2S_AUDIO_ID])
+    cg.add(var.set_i2s_mode(config[CONF_I2S_MODE]))
+    cg.add(var.set_channel(config[CONF_CHANNEL]))
+    cg.add(var.set_sample_rate(config[CONF_SAMPLE_RATE]))
+    cg.add(var.set_bits_per_sample(config[CONF_BITS_PER_SAMPLE]))
+    bits_per_channel = config.get(
+        CONF_BITS_PER_CHANNEL, i2s_bits_per_chan_t.I2S_BITS_PER_CHAN_DEFAULT
+    )
+    cg.add(var.set_bits_per_channel(bits_per_channel))
+    cg.add(var.set_use_apll(config[CONF_USE_APLL]))
+    return var

esphome/components/i2s_audio/i2s_audio.h

@@ -11,9 +11,27 @@ namespace i2s_audio {
 
 class I2SAudioComponent;
 
-class I2SAudioIn : public Parented<I2SAudioComponent> {};
+class I2SAudioInOut : public Component, public Parented<I2SAudioComponent> {
+ public:
+  void set_i2s_mode(i2s_mode_t mode) { this->i2s_mode_ = mode; }
+  void set_channel(i2s_channel_fmt_t channel) { this->channel_ = channel; }
+  void set_sample_rate(uint32_t sample_rate) { this->sample_rate_ = sample_rate; }
+  void set_bits_per_sample(i2s_bits_per_sample_t bits_per_sample) { this->bits_per_sample_ = bits_per_sample; }
+  void set_bits_per_channel(i2s_bits_per_chan_t bits_per_channel) { this->bits_per_channel_ = bits_per_channel; }
+  void set_use_apll(uint32_t use_apll) { this->use_apll_ = use_apll; }
+
+ protected:
+  i2s_mode_t i2s_mode_{};
+  i2s_channel_fmt_t channel_;
+  uint32_t sample_rate_;
+  i2s_bits_per_sample_t bits_per_sample_;
+  i2s_bits_per_chan_t bits_per_channel_;
+  bool use_apll_;
+};
+
+class I2SAudioIn : public I2SAudioInOut {};
 
-class I2SAudioOut : public Parented<I2SAudioComponent> {};
+class I2SAudioOut : public I2SAudioInOut {};
 
 class I2SAudioComponent : public Component {
  public:

esphome/components/i2s_audio/media_player/__init__.py

@@ -12,6 +12,10 @@
     I2SAudioOut,
     CONF_I2S_AUDIO_ID,
     CONF_I2S_DOUT_PIN,
+    CONF_LEFT,
+    CONF_RIGHT,
+    CONF_MONO,
+    CONF_STEREO,
 )
 
 CODEOWNERS = ["@jesserockz"]
@@ -30,12 +34,12 @@
 CONF_I2S_COMM_FMT = "i2s_comm_fmt"
 
 INTERNAL_DAC_OPTIONS = {
-    "left": i2s_dac_mode_t.I2S_DAC_CHANNEL_LEFT_EN,
-    "right": i2s_dac_mode_t.I2S_DAC_CHANNEL_RIGHT_EN,
-    "stereo": i2s_dac_mode_t.I2S_DAC_CHANNEL_BOTH_EN,
+    CONF_LEFT: i2s_dac_mode_t.I2S_DAC_CHANNEL_LEFT_EN,
+    CONF_RIGHT: i2s_dac_mode_t.I2S_DAC_CHANNEL_RIGHT_EN,
+    CONF_STEREO: i2s_dac_mode_t.I2S_DAC_CHANNEL_BOTH_EN,
 }
 
-EXTERNAL_DAC_OPTIONS = ["mono", "stereo"]
+EXTERNAL_DAC_OPTIONS = [CONF_MONO, CONF_STEREO]
 
 NO_INTERNAL_DAC_VARIANTS = [esp32.const.VARIANT_ESP32S2]
 

esphome/components/i2s_audio/media_player/i2s_audio_media_player.h

@@ -23,7 +23,7 @@ enum I2SState : uint8_t {
   I2S_STATE_STOPPING,
 };
 
-class I2SAudioMediaPlayer : public Component, public media_player::MediaPlayer, public I2SAudioOut {
+class I2SAudioMediaPlayer : public Component, public Parented<I2SAudioComponent>, public media_player::MediaPlayer {
  public:
   void setup() override;
   float get_setup_priority() const override { return esphome::setup_priority::LATE; }

esphome/components/i2s_audio/microphone/__init__.py

@@ -2,19 +2,17 @@
 import esphome.codegen as cg
 
 from esphome import pins
-from esphome.const import CONF_CHANNEL, CONF_ID, CONF_NUMBER, CONF_SAMPLE_RATE
+from esphome.const import CONF_NUMBER
 from esphome.components import microphone, esp32
 from esphome.components.adc import ESP32_VARIANT_ADC1_PIN_TO_CHANNEL, validate_adc_pin
 
 from .. import (
-    CONF_I2S_MODE,
-    CONF_PRIMARY,
-    I2S_MODE_OPTIONS,
-    i2s_audio_ns,
-    I2SAudioComponent,
-    I2SAudioIn,
-    CONF_I2S_AUDIO_ID,
     CONF_I2S_DIN_PIN,
+    CONF_RIGHT,
+    I2SAudioIn,
+    i2s_audio_ns,
+    i2s_component_schema,
+    register_i2s_component,
 )
 
 CODEOWNERS = ["@jesserockz"]
@@ -23,29 +21,14 @@
 CONF_ADC_PIN = "adc_pin"
 CONF_ADC_TYPE = "adc_type"
 CONF_PDM = "pdm"
-CONF_BITS_PER_SAMPLE = "bits_per_sample"
-CONF_USE_APLL = "use_apll"
 
 I2SAudioMicrophone = i2s_audio_ns.class_(
     "I2SAudioMicrophone", I2SAudioIn, microphone.Microphone, cg.Component
 )
 
-i2s_channel_fmt_t = cg.global_ns.enum("i2s_channel_fmt_t")
-CHANNELS = {
-    "left": i2s_channel_fmt_t.I2S_CHANNEL_FMT_ONLY_LEFT,
-    "right": i2s_channel_fmt_t.I2S_CHANNEL_FMT_ONLY_RIGHT,
-}
-i2s_bits_per_sample_t = cg.global_ns.enum("i2s_bits_per_sample_t")
-BITS_PER_SAMPLE = {
-    16: i2s_bits_per_sample_t.I2S_BITS_PER_SAMPLE_16BIT,
-    32: i2s_bits_per_sample_t.I2S_BITS_PER_SAMPLE_32BIT,
-}
-
 INTERNAL_ADC_VARIANTS = [esp32.const.VARIANT_ESP32]
 PDM_VARIANTS = [esp32.const.VARIANT_ESP32, esp32.const.VARIANT_ESP32S3]
 
-_validate_bits = cv.float_with_unit("bits", "bit")
-
 
 def validate_esp32_variant(config):
     variant = esp32.get_esp32_variant()
@@ -61,21 +44,12 @@ def validate_esp32_variant(config):
     raise NotImplementedError
 
 
-BASE_SCHEMA = microphone.MICROPHONE_SCHEMA.extend(
-    {
-        cv.GenerateID(): cv.declare_id(I2SAudioMicrophone),
-        cv.GenerateID(CONF_I2S_AUDIO_ID): cv.use_id(I2SAudioComponent),
-        cv.Optional(CONF_CHANNEL, default="right"): cv.enum(CHANNELS),
-        cv.Optional(CONF_SAMPLE_RATE, default=16000): cv.int_range(min=1),
-        cv.Optional(CONF_BITS_PER_SAMPLE, default="32bit"): cv.All(
-            _validate_bits, cv.enum(BITS_PER_SAMPLE)
-        ),
-        cv.Optional(CONF_USE_APLL, default=False): cv.boolean,
-        cv.Optional(CONF_I2S_MODE, default=CONF_PRIMARY): cv.enum(
-            I2S_MODE_OPTIONS, lower=True
-        ),
-    }
-).extend(cv.COMPONENT_SCHEMA)
+BASE_SCHEMA = i2s_component_schema(
+    microphone.MICROPHONE_SCHEMA,
+    I2SAudioMicrophone,
+    default_channel=CONF_RIGHT,
+    default_bits_per_sample="32bit",
+)
 
 CONFIG_SCHEMA = cv.All(
     cv.typed_schema(
@@ -88,7 +62,7 @@ def validate_esp32_variant(config):
             "external": BASE_SCHEMA.extend(
                 {
                     cv.Required(CONF_I2S_DIN_PIN): pins.internal_gpio_input_pin_number,
-                    cv.Required(CONF_PDM): cv.boolean,
+                    cv.Optional(CONF_PDM, default=False): cv.boolean,
                 }
             ),
         },
@@ -99,10 +73,8 @@ def validate_esp32_variant(config):
 
 
 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])
+    var = await register_i2s_component(config)
+    await microphone.register_microphone(var, config)
 
     if config[CONF_ADC_TYPE] == "internal":
         variant = esp32.get_esp32_variant()
@@ -112,11 +84,3 @@ async def to_code(config):
     else:
         cg.add(var.set_din_pin(config[CONF_I2S_DIN_PIN]))
         cg.add(var.set_pdm(config[CONF_PDM]))
-
-    cg.add(var.set_i2s_mode(config[CONF_I2S_MODE]))
-    cg.add(var.set_channel(config[CONF_CHANNEL]))
-    cg.add(var.set_sample_rate(config[CONF_SAMPLE_RATE]))
-    cg.add(var.set_bits_per_sample(config[CONF_BITS_PER_SAMPLE]))
-    cg.add(var.set_use_apll(config[CONF_USE_APLL]))
-
-    await microphone.register_microphone(var, config)

esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp

@@ -58,7 +58,7 @@ void I2SAudioMicrophone::start_() {
       .tx_desc_auto_clear = false,
       .fixed_mclk = 0,
       .mclk_multiple = I2S_MCLK_MULTIPLE_256,
-      .bits_per_chan = I2S_BITS_PER_CHAN_DEFAULT,
+      .bits_per_chan = this->bits_per_channel_,
   };
 
   esp_err_t err;
@@ -167,21 +167,24 @@ size_t I2SAudioMicrophone::read(int16_t *buf, size_t len) {
     return 0;
   }
   this->status_clear_warning();
-  if (this->bits_per_sample_ == I2S_BITS_PER_SAMPLE_16BIT) {
-    return bytes_read;
-  } else if (this->bits_per_sample_ == I2S_BITS_PER_SAMPLE_32BIT) {
-    std::vector<int16_t> samples;
-    size_t samples_read = bytes_read / sizeof(int32_t);
-    samples.resize(samples_read);
-    for (size_t i = 0; i < samples_read; i++) {
-      int32_t temp = reinterpret_cast<int32_t *>(buf)[i] >> 14;
-      samples[i] = clamp<int16_t>(temp, INT16_MIN, INT16_MAX);
+  // ESP-IDF I2S implementation right-extends 8-bit data to 16 bits,
+  // and 24-bit data to 32 bits.
+  switch (this->bits_per_sample_) {
+    case I2S_BITS_PER_SAMPLE_8BIT:
+    case I2S_BITS_PER_SAMPLE_16BIT:
+      return bytes_read;
+    case I2S_BITS_PER_SAMPLE_24BIT:
+    case I2S_BITS_PER_SAMPLE_32BIT: {
+      size_t samples_read = bytes_read / sizeof(int32_t);
+      for (size_t i = 0; i < samples_read; i++) {
+        int32_t temp = reinterpret_cast<int32_t *>(buf)[i] >> 14;
+        buf[i] = clamp<int16_t>(temp, INT16_MIN, INT16_MAX);
+      }
+      return samples_read * sizeof(int16_t);
     }
-    memcpy(buf, samples.data(), samples_read * sizeof(int16_t));
-    return samples_read * sizeof(int16_t);
-  } else {
-    ESP_LOGE(TAG, "Unsupported bits per sample: %d", this->bits_per_sample_);
-    return 0;
+    default:
+      ESP_LOGE(TAG, "Unsupported bits per sample: %d", this->bits_per_sample_);
+      return 0;
   }
 }
 

esphome/components/i2s_audio/microphone/i2s_audio_microphone.h

@@ -10,7 +10,7 @@
 namespace esphome {
 namespace i2s_audio {
 
-class I2SAudioMicrophone : public I2SAudioIn, public microphone::Microphone, public Component {
+class I2SAudioMicrophone : public microphone::Microphone, public I2SAudioIn {
  public:
   void setup() override;
   void start() override;
@@ -30,13 +30,6 @@ class I2SAudioMicrophone : public I2SAudioIn, public microphone::Microphone, pub
   }
 #endif
 
-  void set_i2s_mode(i2s_mode_t mode) { this->i2s_mode_ = mode; }
-
-  void set_channel(i2s_channel_fmt_t channel) { this->channel_ = channel; }
-  void set_sample_rate(uint32_t sample_rate) { this->sample_rate_ = sample_rate; }
-  void set_bits_per_sample(i2s_bits_per_sample_t bits_per_sample) { this->bits_per_sample_ = bits_per_sample; }
-  void set_use_apll(uint32_t use_apll) { this->use_apll_ = use_apll; }
-
  protected:
   void start_();
   void stop_();
@@ -48,11 +41,6 @@ class I2SAudioMicrophone : public I2SAudioIn, public microphone::Microphone, pub
   bool adc_{false};
 #endif
   bool pdm_{false};
-  i2s_mode_t i2s_mode_{};
-  i2s_channel_fmt_t channel_;
-  uint32_t sample_rate_;
-  i2s_bits_per_sample_t bits_per_sample_;
-  bool use_apll_;
 
   HighFrequencyLoopRequester high_freq_;
 };