WAV 8-bits mono

[hash6iron]

Hi @pschatzmann .

I'm trying to use NumberFormatConverterStreamT<int16_t, uint8_t> nfc(kitStream), to transform line-in signal to 8-bits. Later I'd like to copy this signal to wav file but nothing is recorded. If I remove nfc all is ok but 16-bits.

How Can I record on WAV 8-bits MONO signal from Line-in?
Thanks in advance

[pschatzmann]

I don't see any reason why this should not work!
Did you eventually forgot to call begin on the converter ?

You could try to replace the output with a CsVOutput or a HexDumpOutput to see if you get any data...

[hash6iron]

Maybe I need to configure something to setup WAV at 8-bits or is not necessary?

[hash6iron]

Hi @pschatzmann .

I try all possible but the result always is the same, WAV is empty. This is my sketch. What is wrong?

kitStream is configured for AudioBoard audiokit v2.2 ESP32 A1S and works fine with WAV, MP3 playing and WAV 16 bits recording, but not 8-bit recording. kitStrem uses Line-in for recording and Line-out for sound monitoring.

  auto new_sr = kitStream.defaultConfig(RXTX_MODE);
  new_sr.sample_rate = 44100;

  kitStream.setAudioInfo(new_sr);

  char  file_name[20] = "/test.wav";
  wavfile = SD_MMC.open(file_name, FILE_WRITE);

  EncodedAudioStream encoder(&wavfile, new WAVEncoder()); 

  NumberFormatConverterStream nfc(kitStream);
  nfc.begin(kitStream.audioInfo().bits_per_sample, 8 );

  MultiOutput multi;

  //out for WAV
  multi.add(encoder);
  //out for Line_out
  multi.add(kitStream);
  //from Line_in of kitStream to multiOutput  
  StreamCopy copier(multi,kitStream);
  // sinco
  copier.setSynchAudioInfo(true);

  //audio settings
  auto ecfg = encoder.defaultConfig();
  ecfg.sample_rate = DEFAULT_WAV_SAMPLING_RATE_REC;
  ecfg.bits_per_sample = 16;
  ecfg.channels = 1; // Mono
  copier.begin(multi, nfc);
  encoder.begin(ecfg)
  
  while (!STOP)
  {
      copier.copy();
  }
  
  copier.end();
  encoder.end();
  wavfile.flush();
  wavfile.close();

[pschatzmann]

What did you get when you replaced the File with a HexDumpOutput ?

If you don't get any output, create the most imple test sketch that just tests this functionality w/o any additional complications: then share this test sketch and the log output with log level info!

I can't compile and run your code above, and it does not match with what you initially described...

[pschatzmann]

I was testing with the following sketch and was getting the correct result.

AudioInfo info(8000, 1, 16);
SineWaveGenerator<int16_t> sineWave( 32000);  
GeneratedSoundStream<int16_t> sound( sineWave); 
//CSVOutput<uint8_t> out;
HexDumpOutput out(Serial);
WAVEncoder wav;
EncodedAudioStream encoder(&out, &wav); 
NumberFormatConverterStreamT<int16_t, uint8_t> converter(encoder);
StreamCopy copier(converter, sound);

void setup() {
  Serial.begin(115200);
  AudioToolsLogger.begin(Serial, AudioToolsLogLevel::Info);
  delay(5000);

  // Setup sine wave
  auto cfgs = sineWave.defaultConfig();
  cfgs.copyFrom(info);
  sineWave.begin(info, N_B4);

  converter.begin(info);
  encoder.begin(converter.audioInfoOut());
  out.begin();

  Serial.println("Test started...");
}

void loop() { 
  copier.copy();
}

Log

...
10:26:14.079 -> [I] CodecWAV.h : 589 - is_streamed! because length is 2147418112
10:26:14.111 -> [I] AudioStreamsConverter.h : 373 - begin 16 -> 8 bits
10:26:14.111 -> [I] AudioTypes.h : 126 - out: sample_rate: 8000 / channels: 1 / bits_per_sample: 8
10:26:14.111 -> [I] AudioEncoded.h : 75 - virtual void audio_tools::EncodedAudioOutput::addNotifyAudioChange(audio_tools::AudioInfoSupport&)
10:26:14.111 -> [I] CodecWAV.h : 577 - sample_rate: 8000
10:26:14.143 -> [I] CodecWAV.h : 578 - channels: 1
10:26:14.143 -> [I] CodecWAV.h : 579 - bits_per_sample: 8
10:26:14.143 -> [I] CodecWAV.h : 589 - is_streamed! because length is 4294967295
10:26:14.143 -> Test started...
10:26:14.143 -> [I] CodecWAV.h : 631 - Writing Header
10:26:14.143 -> 52 49 46 46 1C 0 FF 7F  - 57 41 56 45 66 6D 74 20 
10:26:14.143 -> 10 0 0 0 1 0 1 0  - 40 1F 0 0 40 1F 0 0 
10:26:14.143 -> 1 0 8 0 64 61 74 61  - 24 0 FF 7F 7F AE D6 F1 
10:26:14.143 -> FB F3 DA B2 84 54 2B F  - 3 9 21 47 75 A5 CF ED 
10:26:14.174 -> FB F6 E0 BB 8D 5D 33 13  - 4 6 1B 3F 6C 9C C8 E8 
10:26:14.174 -> FA F9 E6 C3 97 67 3A 18  - 5 4 16 37 63 92 C0 E3 
10:26:14.174 -> F8 FA EB CB A0 70 43 1E  - 8 3 11 2F 59 89 B7 DD 
10:26:14.174 -> F5 FB EF D2 A9 7A 4B 24  - B 3 D 28 50 7F AE D6 
10:26:14.174 -> F2 FB F3 D9 B2 83 54 2B  - E 3 9 22 48 76 A6 CF 
...

I committed the following changes:

• log bit size in WAV codec
• accept AudioInfo in NumberFormatConverterStreamT: otherwise you would need to call setAudioInfo(info); begin(); in two steps
• Corrected the NumberFormatConverterStreamT log to report bits instead of bytes

I was double checking the hex dump with Chatgpt

fmt Subchunk ✅ NOW CORRECT FOR 8-BIT!
text
66 6D 74 20 "fmt "
10 00 00 00 Chunk size = 16 bytes
01 00 Audio format = 1 (PCM)
01 00 Channels = 1 (mono)
40 1F 00 00 Sample rate = 8000 Hz ✓
40 1F 00 00 Byte rate = 8000 bytes/sec ✓
01 00 Block align = 1 byte ✓
08 00 Bits per sample = 8 bits ✓

[hash6iron]

But.

I'm using LINE-IN as source. Your sketch works fine for me too, but not if I use kitStream LINE-IN, although if I don't use converter 16 to 8, all is ok from LINE-IN.

[pschatzmann]

The data source does not matter, you can just take anything!
The important thing is that WAV expects unsigned 8 bits and not signed 8 bits, all bigger sizes are signed (e.g. 16, 24 ...)!
Please note that the AudioKit can't directly provide and process 8bit data via I2S: you need to use 16 bits and then convert.

[hash6iron]

Yes.
This is the reason that I'm using
NumberFormatConverterStreamT<int16_t, uint8_t> converter(kitStream);

I'm converting directly audio source (line-in) to 8 bits. At the moment all modifications were unsuccessfull.

Now I'm trying uses HexDumpOutput but nothing is showed on Serial monitor.

0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0  - 0 0 0 0 0 0 0 0
...

I think that converter is not communicating kitStream (line-in audio) to multiOutput, but I don't know the reason.

Now. I'm testing with a similar sketch at this below, but doesn't work.

  auto new_sr = kitStream.defaultConfig(RXTX_MODE);
  new_sr.sample_rate = 44100;
  kitStream.setAudioInfo(new_sr);

  AudioInfo info(44100, 1, 16);
  
  EncodedAudioStream encoder(&wavfile, new WAVEncoder());
  HexDumpOutput out(Serial);
  NumberFormatConverterStreamT<int16_t, uint8_t> converter( kitStream );

  MultiOutput multi;
  multi.add(encoder);
  multi.add(kitStream);
  multi.add(out);

  StreamCopy copier(multi,kitStream);
  copier.setSynchAudioInfo(true);

  // Start converter
  converter.setAudioInfo(info);
  converter.begin();

  // Start encoder
  encoder.begin(converter.audioInfoOut());
  // Start copier
  copier.begin(multi, converter);
  // Start HEXdump
  out.begin();

  while(!STOP)
  {
     copier.copy();
  }

[hash6iron]

Hi again.

I have tested only with WAV enconder and not workig, then the problem is NumberFormatConverterStreamT with AudioBoardStream.