High-pitch chipmunk voice on recording from ADC

[EvilBeaver]

Hi. I'm using example streams-i2s-webserver_wav but adopted it for analog microphone and AnalogAudioStream on input.

Sketch is working, but recorded sound is distorted to high-pitch and high-speed, like chipmunk. I've tried to change sample rate but nothing cnahges. What options are useful for tuning voice pitch (or speed) when sampling an analog audio?

    AudioInfo _mainInfo(44100, 1, 16); // TRIED TO CHANGE THIS TO 8k, 22k, 32bit
    AnalogAudioStream _inputStream; 
    auto inputCfg = _inputStream.defaultConfig(RX_MODE);
    inputCfg.copyFrom(_mainInfo);

    _inputStream.begin(inputCfg);
    server.begin(_inputStream, _mainInfo);

[pschatzmann]

Check in your log if you get some mixup between stereo and mono somewhere

[EvilBeaver]

What words to look for in log?

[pschatzmann]

channels!

[pschatzmann]

I committed a correction which updates the audio information (incl. the channels) to the encoded audio stream in the audio server

[EvilBeaver]

You were right it's channels. Here's the sketch and log:

    Serial.begin(115200);
    AudioLogger::instance().begin(Serial,AudioLogger::Info);

    auto config = inputStream.defaultConfig(RX_MODE);
    config.bits_per_sample = 16;
    config.sample_rate = 44100;
    config.channels = 1;
    inputStream.begin(config);
    server.wavEncoder().setAudioInfo(config);

    server.begin(inputStream, config);

[I] AudioTypes.h : 104 - sample_rate: 44100
[I] AudioTypes.h : 105 - channels: 1
[I] AudioTypes.h : 106 - bits_per_sample: 16
[I] AnalogAudioESP32.h : 51 - auto_center
[I] AnalogAudioESP32.h : 94 - RX_MODE
[I] CodecWAV.h : 539 - sample_rate: 44100
[I] CodecWAV.h : 540 - channels: 1
[I] CodecWAV.h : 549 - is_streamed! because length is 2147418112
[I] CodecWAV.h : 539 - sample_rate: 44100
[I] CodecWAV.h : 540 - channels: 1
[I] CodecWAV.h : 549 - is_streamed! because length is 4294967295
[I] CodecWAV.h : 539 - sample_rate: 44100
[I] CodecWAV.h : 540 - channels: 2
[I] CodecWAV.h : 549 - is_streamed! because length is 4294967295

It seems server's channels are not get updated in begin

I committed a correction which updates the audio information (incl. the channels)

It was code error? Should I upgrade my dependency to latest commit?

[pschatzmann]

Yes, if you want to use mono you need to upgrade. Otherwise just change your sketch to use 2 channels.

[EvilBeaver]

Thanks a lot!

[leboff]

I'm experiencing a similar issue with the audio being pitched up with an SI4703 connected to the ADC. Not sure there's any problem with the number of channels however. Sketch below any thoughts?

#include <SparkFunSi4703.h>
#include <Wire.h>
#include <SPI.h>
#include "AudioTools.h"

// SI4703 Connections
#define ESP32_RESET_PIN 23
#define ESP32_I2C_SDA 21
#define ESP32_I2C_SCL 22

// WiFi Credentials
const char *ssid = "SSID";
const char *password = "PASS";

// FM Radio
Si4703_Breakout radio(ESP32_RESET_PIN, ESP32_I2C_SDA, ESP32_I2C_SCL);
int channel = 941; // Example initial channel

AudioInfo info(44100, 1, 16); // 44100 Hz sample rate, mono, 16-bit
AnalogAudioStream analogStream;  // ADC Stream
AudioEncoderServer server(new WAVEncoder(), ssid, password);

void setup() {
    Serial.begin(115200);
    AudioLogger::instance().begin(Serial,AudioLogger::Info);
    
    // Initialize FM Radio
    radio.powerOn();
    radio.setVolume(15);
    radio.setChannel(channel);

    // Configure ADC
    auto cfg = analogStream.defaultConfig(RX_MODE);
    cfg.sample_rate = info.sample_rate;
    cfg.channels = info.channels;
    cfg.bits_per_sample = info.bits_per_sample;
    cfg.is_auto_center_read = true;
    cfg.adc_attenuation = ADC_ATTEN_DB_12;
    cfg.adc_calibration_active = true;
    cfg.adc_channels[0] = ADC_CHANNEL_0; 
    analogStream.begin(cfg);

    // Start Web Server (using WAV encoder)
    server.begin(analogStream, cfg);
    // Serial.println("Server started");
}

void loop() {
    // Handle Web Server
    server.doLoop();
}

[pschatzmann]

The sketch looks right to me:

• Is there anything in the log that is not expected ?
• Check the output fomrat with a desktop audio tool to find out if this is matching with what you have defined.
• I also suggest to add a MeasuringStream into the reading chain, so that you can double check the effective sample rate

ps. In Switzerland the national FM radios have been switched off on the 1.st of January!

[leboff]

Nothing obvious in the logs (snippet below). Checked the output with MediaInfo and that looks right as well. Will work on adding the MeasuringStream and report back...

Complete name                            : download (1).wav
Format                                   : Wave
Format settings                          : PcmWaveformat
File size                                : 3.41 MiB
Duration                                 : 40 s 540 ms
Overall bit rate mode                    : Constant
Overall bit rate                         : 706 kb/s

Audio
Format                                   : PCM
Format settings                          : Little / Signed
Codec ID                                 : 1
Duration                                 : 40 s 540 ms
Bit rate mode                            : Constant
Bit rate                                 : 705.6 kb/s
Channel(s)                               : 1 channel
Sampling rate                            : 44.1 kHz
Bit depth                                : 16 bits
Stream size                              : 3.41 MiB (100%)

[I] AnalogConfigESP32V1.h : 139 - I2S_MODE_ADC_BUILT_IN
[I] AnalogAudioStream.h : 66 - bool audio_tools::AnalogAudioStream::begin(audio_tools::AnalogConfig)
[I] AnalogDriverESP32V1.h : 41 - virtual bool audio_tools::AnalogDriverESP32V1::begin(audio_tools::AnalogConfigESP32V1)
[I] AnalogDriverESP32V1.h : 700 - channels: 1, max: 6
[I] AnalogDriverESP32V1.h : 710 - ADC channel 0 is on pin 36
[I] AnalogDriverESP32V1.h : 725 - sample rate eff: 44100, range: 20000 to 2000000
[I] AnalogDriverESP32V1.h : 685 - adc bit width: 12, range: 9 to 12
[I] AnalogDriverESP32V1.h : 747 - bits per sample: 16
[I] AnalogDriverESP32V1.h : 514 - buffer_size: 512 samples, conv_frame_size: 1024 bytes
[I] AnalogDriverESP32V1.h : 529 - adc_continuous_new_handle successful
[I] AnalogDriverESP32V1.h : 552 - dig_cfg.sample_freq_hz: 44100
[I] AnalogDriverESP32V1.h : 553 - dig_cfg.conv_mode: 1 (1: unit 1, 2: unit 2, 3: both)
[I] AnalogDriverESP32V1.h : 554 - dig_cfg.format: 0 (0 is type1: [12bit data, 4bit channel])
[I] AnalogDriverESP32V1.h : 556 - dig_cfg.adc_pattern[0].atten: 3
[I] AnalogDriverESP32V1.h : 557 - dig_cfg.adc_pattern[0].channel: 0
[I] AnalogDriverESP32V1.h : 558 - dig_cfg.adc_pattern[0].unit: 0
[I] AnalogDriverESP32V1.h : 559 - dig_cfg.adc_pattern[0].bit_width: 12
[I] AnalogDriverESP32V1.h : 568 - adc_continuous_config successful
[I] AnalogDriverESP32V1.h : 781 - enabled calibration for ADC0 with atten 3 and bitwidth 12
[  1714][V][esp32-hal-periman.c:160] perimanSetPinBus(): Pin 36 successfully set to type ADC_CONT (8) with bus 0x1
[I] AnalogDriverESP32V1.h : 604 - 1 FIFO buffers allocated of size 520
[I] AnalogDriverESP32V1.h : 606 - Setup ADC successful
[I] CodecWAV.h : 524 - sample_rate: 44100
[I] CodecWAV.h : 525 - channels: 1
[I] CodecWAV.h : 533 - is_streamed! because length is 2147418112
[I] CodecWAV.h : 524 - sample_rate: 44100
[I] CodecWAV.h : 525 - channels: 1
[I] CodecWAV.h : 533 - is_streamed! because length is 4294967295

[pschatzmann]

I confirm: nothing unexpected!

[leboff]

Here's what I get after adding a measuringstream. Should I be seeing the 44.1khz sample rate here?

==> Samples per second: 28000
==> Samples per second: 36000
==> Samples per second: 36000
==> Samples per second: 36000

[pschatzmann]

Yes or something close to that.
What ESP32 core version are you using ?
As a work around you can pass a corrected cfg with 36000 to the server.

[leboff]

Works well enough with the web server set to 36000 so going to go forward with that. I'm using 3.X (had to upgrade for some reason I can no longer remember). Really appreciate the quick responses, thank you for that and this project! and kind of a bummer about FM in Switzerland!

[pschatzmann]

https://esp32.com/viewtopic.php?t=40200