Fix mic distortion for Inmp441Max98357aFastLed

PlatformIO/platformio.ini

@@ -33,7 +33,7 @@ lib_deps =
    https://github.com/marvinroger/async-mqtt-client.git
    https://github.com/me-no-dev/AsyncTCP.git
    https://github.com/knolleary/pubsubclient.git
-   https://github.com/bblanchon/ArduinoJson.git @ ^6.19.4
+   https://github.com/bblanchon/ArduinoJson.git
    ESP Async WebServer
    m5stack/M5Atom
    fastled/FastLED

PlatformIO/src/devices/Inmp441Max98357aFastLed.hpp

@@ -12,34 +12,24 @@
 #define NUM_LEDS 5
 CRGB leds[NUM_LEDS];
 
-// I2S pins on ESp32cam MIC
-// GPIO2 <--> WS
-// GPIO14 <--> SCK
-// GPIO15 <--> SD
-#define I2S_SCK 25
-#define I2S_WS 32
-#define I2S_SD 33
-#define I2S_PORT I2S_NUM_0
-
-// I2S pins on ESp32Cam Speakers
-// GPIO16 <--> LRC
-// GPIO13 <--> BCLK
-// GPIO12 <--> DIN
-#define I2S_LRC 12
-#define I2S_BCLK 14
-#define I2S_DIN 27
-
-#define I2S_PORT_TX I2S_NUM_1
-
-#define I2S_OUTPUT_SAMPLE_RATE (22050)
-#define I2S_SAMPLE_RATE (16000)
-#define I2S_SAMPLE_BITS (16)
-#define I2S_READ_LEN 512
-// LEDs
-#define LED_FLASH 4
-
-#define KEY1_GPIO GPIO_NUM_34
-#define KEY_LISTEN KEY1_GPIO
+// I2S pins & parameters for MIC
+#define MIC_I2S_BCLK 25
+#define MIC_I2S_FS 32
+#define MIC_I2S_DIN 33
+#define MIC_I2S_PORT I2S_NUM_0
+#define MIC_SAMPLE_RATE (16000)
+#define MIC_I2S_SAMPLE_BITS 32
+#define MIC_I2S_SAMPLE_BYTES (MIC_I2S_SAMPLE_BITS / 8)
+#define I2S_READ_LEN (256 * MIC_I2S_SAMPLE_BYTES)
+
+// I2S pins & parameters for speaker
+#define SPK_I2S_FS 12
+#define SPK_I2S_BCLK 14
+#define SPK_I2S_DOUT 27
+#define SPK_I2S_PORT I2S_NUM_1
+#define SPK_SAMPLE_RATE (22050)
+#define SPK_I2S_SAMPLE_BITS 16
+#define SPK_I2S_SAMPLE_BYTES (SPK_I2S_SAMPLE_BITS / 8)
 
 class Inmp441Max98357aFastLED : public Device
 {
@@ -51,6 +41,7 @@ class Inmp441Max98357aFastLED : public Device
 
   void setWriteMode(int sampleRate, int bitDepth, int numChannels);
   void writeAudio(uint8_t *data, size_t size, size_t *bytes_written);
+  void setGain(uint16_t gain);
 
   int numAmpOutConfigurations()
   {
@@ -79,6 +70,7 @@ class Inmp441Max98357aFastLED : public Device
 
 private:
   char *i2s_read_buff = (char *)calloc(I2S_READ_LEN, sizeof(char));
+  uint16_t m_gain;
   long currentMillis, startMillis;
 };
 
@@ -90,54 +82,69 @@ void Inmp441Max98357aFastLED::init()
   // Speakers
   i2s_config_t i2sConfig_tx = {
       .mode = (i2s_mode_t)(I2S_MODE_MASTER | I2S_MODE_TX),
-      .sample_rate = I2S_OUTPUT_SAMPLE_RATE,        // I2S_SAMPLE_RATE,
-      .bits_per_sample = I2S_BITS_PER_SAMPLE_16BIT, // I2S_BITS_PER_SAMPLE_16BIT,
+      .sample_rate = SPK_SAMPLE_RATE,                                                   // I2S_SAMPLE_RATE,
+      .bits_per_sample = static_cast<i2s_bits_per_sample_t>(I2S_BITS_PER_SAMPLE_16BIT), // I2S_BITS_PER_SAMPLE_16BIT,
       .channel_format = I2S_CHANNEL_FMT_ONLY_LEFT,
-      .communication_format = (i2s_comm_format_t)(I2S_COMM_FORMAT_STAND_I2S),
+      .communication_format = I2S_COMM_FORMAT_STAND_I2S,
       .intr_alloc_flags = ESP_INTR_FLAG_LEVEL1,
       .dma_buf_count = 2,
       .dma_buf_len = 512,
       .tx_desc_auto_clear = true,
   };
 
   i2s_pin_config_t pin_config_tx = {
-      .bck_io_num = I2S_BCLK, .ws_io_num = I2S_LRC, .data_out_num = I2S_DIN, .data_in_num = -1};
+      .bck_io_num = SPK_I2S_BCLK, .ws_io_num = SPK_I2S_FS, .data_out_num = SPK_I2S_DOUT, .data_in_num = -1};
 
-  err += i2s_driver_install(I2S_PORT_TX, &i2sConfig_tx, 0, NULL);
+  err += i2s_driver_install(SPK_I2S_PORT, &i2sConfig_tx, 0, NULL);
   if (err != ESP_OK) {
     Serial.printf("Failed installing headphone driver: %d\n", err);
     while (true)
       ;
   }
 
-  err += i2s_set_pin(I2S_PORT_TX, &pin_config_tx);
+  err += i2s_set_pin(SPK_I2S_PORT, &pin_config_tx);
   if (err != ESP_OK) {
     Serial.printf("Failed setting headphone pin: %d\n", err);
     while (true)
       ;
   }
   Serial.println("I2S headphone driver installed.\n");
 
+  //----- Microphone
+  /*
+    - the INMP441 MEMS microphone has a specified sensitivity of -26 dBFS @ 1 Pa,
+      or 120 dB SPL @ FS.
+    - Even a loud speech signal won't exceed 80 dB, so we "waste" the top 40 dB
+      of dynamic range.
+    - The microphone sends 24 bit samples (which is not to say it has a true
+      24 bit dynamic range).
+    - If we read 16 bit samples (nominal 96 dB dynamic range), and waste the top
+      40 dB, then we get a ~56 dB dynamic range speech signal.
+    - Might bet better to read 24 bit samples, and amplify by 40 dB ?
+    - in my experiments, a x20 gain (26 dB) will result in a 50% FS signal, i.e.
+      not clipping, with a loud voice at 30cm distance
+  */
   i2s_config_t i2s_config = {.mode = (i2s_mode_t)(I2S_MODE_MASTER | I2S_MODE_RX),
-                             .sample_rate = I2S_SAMPLE_RATE,
-                             .bits_per_sample = i2s_bits_per_sample_t(I2S_SAMPLE_BITS),
+                             .sample_rate = MIC_SAMPLE_RATE,
+                             .bits_per_sample = static_cast<i2s_bits_per_sample_t>(MIC_I2S_SAMPLE_BITS),
                              .channel_format = I2S_CHANNEL_FMT_ONLY_LEFT,
-                             .communication_format = i2s_comm_format_t(I2S_COMM_FORMAT_STAND_I2S),
+                             .communication_format = I2S_COMM_FORMAT_STAND_I2S,
                              .intr_alloc_flags = 0,
                              .dma_buf_count = 2,
-                             .dma_buf_len = 512,
+                             .dma_buf_len = I2S_READ_LEN,
                              .use_apll = 1};
 
-  i2s_pin_config_t pin_config = {.bck_io_num = I2S_SCK, .ws_io_num = I2S_WS, .data_out_num = -1, .data_in_num = I2S_SD};
+  i2s_pin_config_t pin_config = {
+      .bck_io_num = MIC_I2S_BCLK, .ws_io_num = MIC_I2S_FS, .data_out_num = -1, .data_in_num = MIC_I2S_DIN};
 
-  err += i2s_driver_install(I2S_PORT, &i2s_config, 0, NULL);
+  err += i2s_driver_install(MIC_I2S_PORT, &i2s_config, 0, NULL);
   if (err != ESP_OK) {
     Serial.printf("Failed installing mic driver: %d\n", err);
     while (true)
       ;
   }
 
-  err += i2s_set_pin(I2S_PORT, &pin_config);
+  err += i2s_set_pin(MIC_I2S_PORT, &pin_config);
   if (err != ESP_OK) {
     Serial.printf("Failed setting mic pin: %d\n", err);
     while (true)
@@ -207,31 +214,51 @@ void Inmp441Max98357aFastLED::animateBlinking(StateColors colors)
 void Inmp441Max98357aFastLED::setWriteMode(int sampleRate, int bitDepth, int numChannels)
 {
   if (sampleRate > 0) {
-    i2s_set_clk(I2S_PORT_TX, sampleRate, static_cast<i2s_bits_per_sample_t>(bitDepth),
+    i2s_set_clk(SPK_I2S_PORT, sampleRate, static_cast<i2s_bits_per_sample_t>(bitDepth),
                 static_cast<i2s_channel_t>(numChannels));
   }
 }
 
 void Inmp441Max98357aFastLED::writeAudio(uint8_t *data, size_t size, size_t *bytes_written)
 {
-  i2s_write(I2S_PORT_TX, data, size, bytes_written, portMAX_DELAY);
+  i2s_write(SPK_I2S_PORT, data, size, bytes_written, portMAX_DELAY);
 }
 
 bool Inmp441Max98357aFastLED::readAudio(uint8_t *data, size_t size)
 {
+  size_t samples_requested = size / sizeof(int16_t);
+  size_t bytes_requested = MIC_I2S_SAMPLE_BYTES * samples_requested;
   size_t bytes_read;
-  i2s_read(I2S_PORT, (void *)i2s_read_buff, size, &bytes_read, portMAX_DELAY);
-  uint32_t j = 0;
-  uint32_t dac_value = 0;
-  for (int i = 0; i < size; i += 2) {
-    dac_value = ((((uint16_t)(i2s_read_buff[i + 1] & 0xff) << 8) | ((i2s_read_buff[i + 0]))));
-    data[j++] = 0;
-    data[j++] = dac_value * 256 / 2048;
+  // we skip every other sample, so read 2x desired # of samples
+  i2s_read(MIC_I2S_PORT, (void *)i2s_read_buff, bytes_requested, &bytes_read, portMAX_DELAY);
+
+  int16_t *from = (int16_t *)i2s_read_buff;
+  int16_t *to = (int16_t *)data;
+  size_t nsamples = bytes_read / MIC_I2S_SAMPLE_BYTES;
+  size_t i;
+  int32_t y;
+  /*
+    according to https://www.esp32.com/viewtopic.php?t=11023 , incoming samples
+    are in flipped order, i.e. 1,0,3,2,5,4,... instead of 0,1,2,3,4,5,...
+    As of ESP-IF 4, this is not true anymore. However, reading from the
+    microphone as int16_t appears to repeat every sample twice, so we read
+    as int32_t instead, and ignore the 16 LSBs
+  */
+  for (i = 0; i < nsamples; i++) {
+    y = from[1]; // read 16 MSB of the 32 bit value
+    y *= m_gain;
+    *to++ = y;
+    from += 2;
   }
   return true;
 }
 
 void Inmp441Max98357aFastLED::updateBrightness(int brightness)
 {
   FastLED.setBrightness(brightness);
+}
+
+void Inmp441Max98357aFastLED::setGain(uint16_t gain)
+{
+  m_gain = gain;
 }