i2s INMP441 MEMS microphone streamed via ESP32 PICO-D4 to MAX98357A earphone output

[jgoldwein]

I would appreciate help with the below code. At best, I get low volume voice (albeit full volume chord) streaming to the MAX98375A earphones from the INMP441 microphone. I have tried virtually all combinations of channels, bits/channel, sampleRate... Assume the wiring is correct since the chord plays beautifully. Any suggestions, comments... welcomed and much appreciated in advance.

/*
# Simple program to stream audio captured 
#  from:
#   left channel INMP441 MEMS microphone 
#  via:
#         ESP32 PICO-D4
#   to:
#   left Channel MAX98357A earphone output
*/

#include <AudioTools.h>

AudioInfo info(22050, 1, 16);

I2SStream i2s_in;   // in from I2S INMP441 Mems Microphone Board
I2SStream i2s_out;  // out to I2S MAX98357A Board
VolumeStream volume(i2s_in);

ConverterFillLeftAndRight<int32_t> filler(RightIsEmpty);  // fill both channels - or change to RightIsEmpt

StreamCopy copier(i2s_out, volume);  // copies the converted in channel (conv) to i2s_out

// const float sampleRate = 22050;  //44100;  // Sample rate in Hz
int16_t amplitude = 10000;
bool hasRun = false;

void beep(int duration_ms, float freq) {
  unsigned long startTime = millis();
  const float frequency = freq;
  const float sampleRate = 22050;  // Sample rate in Hz
  float phase = 0.0;

  while (millis() - startTime < duration_ms) {
    int16_t sample = (int16_t)(amplitude * sin(phase));
    phase += 2 * PI * frequency / sampleRate;
    if (phase >= 2 * PI) {
      phase -= 2 * PI;
    }
    i2s_out.write((uint8_t *)&sample, sizeof(sample));
  }
}

void setup() {

  pinMode(23, OUTPUT);     // Used for SD pin on MAX98357A Board output.  High = Left, Low=Off
  digitalWrite(23, HIGH);  // Set the pin HIGH

  // Configure i2s_in with the current combination
  auto cfg_in = i2s_in.defaultConfig(RX_MODE);
  cfg_in.copyFrom(info);
  cfg_in.i2s_format = I2S_STD_FORMAT;
  cfg_in.is_master = true;
  cfg_in.port_no = 0;
  //
  cfg_in.pin_bck = 26;
  cfg_in.pin_ws = 25;
  cfg_in.pin_data = 27;
  //
  //  cfg_in.channel_format = audio_tools::I2SChannelSelect::Left; //Left; // I2S_CHANNEL_FMT_ALL_RIGHT;
  i2s_in.begin(cfg_in);

  volume.begin(cfg_in);  // we need to provide the bits_per_sample and channels same as input mems microphone
  volume.setVolume(0.95);


  // Configure i2s_out with the current combination
  auto cfg_out = i2s_out.defaultConfig(TX_MODE);
  cfg_out.copyFrom(info);
  cfg_out.i2s_format = I2S_STD_FORMAT;
  cfg_out.is_master = true;
  cfg_out.port_no = 1;
  //  cfg_out.channel_format = audio_tools::I2SChannelSelect::Left;
  cfg_out.pin_bck = 10;
  cfg_out.pin_ws = 5;
  cfg_out.pin_data = 9;
  i2s_out.begin(cfg_out);
}


void playChord() {
  // Tome (beep) sequence confirms I2S output operating properly
  beep(50, 261.63);  // C note - Root
  delay(5);
  beep(50, 329.63);  // E note - Major third
  delay(5);
  beep(90, 493.88);  // B note - Major 7th
}

void loop() {
  if (!hasRun) {
    playChord();
    hasRun = true;
  }
  copier.copy(filler);
}

[pschatzmann]

Just read the Wiki about Volume Control: you might want to amplify the signal!

[jgoldwein]

First, I want to thank you for this superb set of tools and your amazing support. It is truly appreciated!

Second, thanks for the hint re Volume Control. I was using it improperly and have both corrected the code and added "allow_boost" so I can further amplify the signal. I have posted the now working code below though it seems tenuous since it sometimes fails on reboots for some reason. Considering the INMP441 microphone does indeed use 24 bits within the 32 bit data block, should I use a conversion tool to address this issue? Or, have I addressed that by using I2S_PHILIPS_FORMAT?

/*
# Simple program to stream audio captured 
#  from:
#   left channel INMP441 MEMS microphone 
#  via:
#         ESP32 PICO-D4
#   to:
#   left Channel MAX98357A earphone output
*/

#include <AudioTools.h>

I2SStream i2s_in;   // in from I2S INMP441 Mems Microphone Board
I2SStream i2s_out;  // out to I2S MAX98357A Board
VolumeStream volume(i2s_in);

ConverterFillLeftAndRight<int32_t> filler(RightIsEmpty);  // fill both channels - or change to RightIsEmpt

StreamCopy copier(i2s_out, volume, 32);  //  minimize buffer size to minimize latency

const float sampleRate = 96000;  // Use max/optimal sample rate to decrease latency. 44100 and below results in noticable latency
int16_t amplitude = 2000;
bool hasRun = false;

AudioInfo info(sampleRate, 1, 32);

void beep(int duration_ms, float freq) {
  unsigned long startTime = millis();
  const float frequency = freq;
  float phase = 0.0;

  while (millis() - startTime < duration_ms) {
    int16_t sample = (int16_t)(amplitude * sin(phase));
    phase += 2 * PI * frequency / sampleRate;
    if (phase >= 2 * PI) {
      phase -= 2 * PI;
    }
    i2s_out.write((uint8_t *)&sample, sizeof(sample));
  }
}

void setup() {

  pinMode(23, OUTPUT);     // Used for SD pin on MAX98357A Board output.  High = Left, Low=Off
  digitalWrite(23, HIGH);  // Set the pin HIGH

  // Configure i2s_in with the current combination
  auto cfg_in = i2s_in.defaultConfig(RX_MODE);
  cfg_in.copyFrom(info);
  cfg_in.i2s_format = I2S_PHILIPS_FORMAT;
  cfg_in.is_master = true;
  cfg_in.port_no = 0;
  //
  cfg_in.pin_bck = 26;
  cfg_in.pin_ws = 25;
  cfg_in.pin_data = 27;
  //
  // cfg_in.channel_format = audio_tools::I2SChannelSelect::Left;
  i2s_in.begin(cfg_in);

  // configure and set volume.  Boosting required for ability to hear
  auto vol_in_cfg = volume.defaultConfig();
  vol_in_cfg.copyFrom(info);
  vol_in_cfg.allow_boost = true;  // boost amplification
  volume.begin(vol_in_cfg);       // we need to provide the bits_per_sample and channels
  volume.setVolume(100, 0);       // left channel x 100, right channel x 0

  // Configure i2s_out with the current combination
  auto cfg_out = i2s_out.defaultConfig(TX_MODE);
  cfg_out.copyFrom(info);
  cfg_out.i2s_format = I2S_STD_FORMAT;
  cfg_out.is_master = true;
  cfg_out.port_no = 1;
  //  cfg_out.channel_format = audio_tools::I2SChannelSelect::Left;
  cfg_out.pin_bck = 10;
  cfg_out.pin_ws = 5;
  cfg_out.pin_data = 9;
  i2s_out.begin(cfg_out);
}


void playChord() {
  // Tome (beep) sequence confirms I2S output operating properly
  beep(50, 261.63);  // C note - Root
  delay(5);
  beep(50, 329.63);  // E note - Major third
  delay(5);
  beep(90, 493.88);  // B note - Major 7th
}

void loop() {
  if (!hasRun) {
    playChord();
    hasRun = true;
  }
  copier.copy(filler);
}

[joba-1]

according to its datasheet the mic always delivers 32bit frames but with just 24bit used for samples. If you use 16bit samples I think with „normal“ volume control you end up with full amplitudes but only 12 bits used.
I guess the quality is better and the volume automatically correct if you read 32bit and convert data yourself, using all 16bit. Pretty sure this lib has a readymade solution for it somewhere…

[jgoldwein]

Thanks joba-1. I did realize that and even had viewed I's and O's on my scope to make sure I wired the system and implemented the code properly. I now have gotten this to work though (see above response to pschatzmann's suggestion) and think it may be because I use the I2S_PHILIPS_FORMAT. If not, I may have to do a conversion to make sure I have this to a point where I can add filters...

JG

[pschatzmann]

My gut feeling is that it would not make a big difference: but I suggest some small test sketches. Measure the volume with 16 bits vs Measure the volume of 24 or 32 bits converted to 16 bits (using a NumberFormatConverterStream

Also there is a section for the INMP441 in the documentation