HOW TO CONVERT SGTL5000 DATA INTO WAV FILE?

[Dipak]

Hello Team,

First Thanks to all Nordic support:)

My problem :

I have interfaced SGTL5000 with Nrf52832 (used SDK 16) I'm getting the audio data on aux connector and also getting data in form of integer value ( negative and positive values ) 

Now I have take a log of this integer values using jlink RTT viewer but I don't know how to convert this values into wav file?

{ I played one song and based on that I take log on RTT viewer  }

2) I also want to transmit this data on ble Nordic uart service and from android I want to convert this data into wav file ( live streaming - some real time delay is acceptable) but I don't how to convert data into wav file?

Thanks! 

[Jørgen Holmefjord]

Hi,

This repository may provide some useful tips and links.

I'm not aware of any specific libraries for converting to WAV format, but as far as I can see, the input from SGTL5000 is PCM format, so any PCM to WAV converters should do the job.

Best regards,
Jørgen

[Dipak]

Thanks for the reply!

Actually, I already used this example in my customize code but I don't know how to convert integer data into a .wav file

I have spent lots of time behind this but still, I have not got success! 

I have attached the recorded data file please let me know if is it the right file for .wav conversion or not?

Please help me out with this  

Thanks!

[Dipak]

If this file format is not right then please let me know how to generate the right file format for .wav file

[Jørgen Holmefjord]

Hi,

I spoke with our developers that made the Python script to convert the WAV file to C in the GitHub repository. They recommend using the built in package "wave" in Python to convert from PCM array to WAV, for instance like this:

import wave

# Raw uint8 values
pcm = [0x01, 0x02, etc.]

# Assume 1 channel, 16-bits (2 bytes), 16 kHz 
# w = open("my_file.wav", "wb")
w.setparams((1, 2, 16000, len(pcm)/2, 'NONE', 'not compressed'))
w.writeframes(bytes(pcm))
w.close()

This should likely work with the endianness used by nRF52 without any explicit converting, but if not, you just needs to re-arrange the bytes in the pcm-array: 

pcm = [0x01, 0x02, 0x03, 0x04]

to:

pcm = [0x02, 0x01, 0x04, 0x03]

Best regards,
Jørgen

[Dipak]

HellO,

Thanks for the reply!

Here is the my I2S setting .

//@@@@@@   I2S setting  @@@@@@@@@@@@@@@@@@@

#define AUDIO_FRAME_WORDS 320
#define I2S_BUFFER_SIZE_WORDS AUDIO_FRAME_WORDS * 2 // Double buffered - I2S lib will switch between using first and second half
static uint32_t m_i2s_tx_buffer[I2S_BUFFER_SIZE_WORDS];
static uint32_t m_i2s_rx_buffer[I2S_BUFFER_SIZE_WORDS];

/* Include car sample to demonstrate that sample can be played from application as well */
#define SAMPLE_LEN 67200
extern const uint8_t car_sample[SAMPLE_LEN];
static uint8_t * p_sample = (uint8_t *)car_sample;
static uint32_t sample_idx = 0;

int16_t data_audio[(AUDIO_FRAME_WORDS/2)];

//@@@@@@@@@@@@@@@@@@@@@@@@@@

//@@@@@@@@@@@@@@ i2c event handler @@@@@@@@@@@@@@@@@

static bool i2s_sgtl5000_driver_evt_handler(drv_sgtl5000_evt_t * p_evt)
{
bool ret = false;
//NRF_LOG_INFO("i2s_sgtl5000_driver_evt_handler %d", p_evt->evt);

//NRF_LOG_INFO("%d",p_evt->param.rx_buf_received.p_data_received);
switch (p_evt->evt)
{
case DRV_SGTL5000_EVT_I2S_RX_BUF_RECEIVED:
{

uint16_t * p_buffer = (uint16_t *) p_evt->param.rx_buf_received.p_data_received;
i2s_buffer_size_words = p_evt->param.rx_buf_received.number_of_words;
int16_t pcm_stream[i2s_buffer_size_words]; // int16_t - i2s_buffer_size_words size; means we only cover half of data_to_send_buffer, which is fine since we are only using LEFT channel

memcpy(data_audio, p_buffer, i2s_buffer_size_words);

ret = true;
}
break;
case DRV_SGTL5000_EVT_I2S_TX_BUF_REQ:
{
//NRF_LOG_INFO("i2s_sgtl5000_driver_evt_handler TX BUF REQ");

/* Play sample! 16kHz sample played on 32kHz frequency! If frequency is changed, this approach needs to change. */
/* Playback of this 16kHz sample depends on I2S MCK, RATIO, Alignment, format, and channels! Needs to be DIV8, RATIO 128X, alignment LEFT, format I2S, channels LEFT. */

uint16_t * p_buffer = (uint16_t *) p_evt->param.tx_buf_req.p_data_to_send;
uint32_t i2s_buffer_size_words = p_evt->param.rx_buf_received.number_of_words;
int16_t pcm_stream[i2s_buffer_size_words]; // int16_t - i2s_buffer_size_words size; means we only cover half of data_to_send_buffer, which is fine since we are only using LEFT channel

/* Clear pcm buffer */
memset(pcm_stream, 0, sizeof(pcm_stream));

/* Check if playing the next part of the sample will exceed the sample size, if not, copy over part of sample to be played */
if (sample_idx < SAMPLE_LEN)
{
/* Copy sample bytes into pcm_stream (or remaining part of sample). This should fill up half the actual I2S transmit buffer. */
/* We only want half becuase the sample is a 16kHz sample, and we are running the SGTL500 at 32kHz; see DRV_SGTL5000_FS_31250HZ */
uint32_t bytes_to_copy = ((sample_idx + sizeof(pcm_stream)) < SAMPLE_LEN) ? sizeof(pcm_stream) : SAMPLE_LEN - sample_idx - 1;
memcpy(pcm_stream, &p_sample[sample_idx], bytes_to_copy);
sample_idx += bytes_to_copy;
ret = true;
}
else
{
/* End of buffer reached. */
sample_idx = 0;
ret = false;
}

/* Upsample the decompressed audio */
/* i < i2s_buffer_size_words * 2 because we have a uint16_t buffer pointer */
for (int i = 0, pcm_stream_idx = 0; i < i2s_buffer_size_words * 2; i += 2)
{
for (int j = i; j < (i + 2); ++j)
{
p_buffer[j] = pcm_stream[pcm_stream_idx];

}
++pcm_stream_idx;
}

}
break;
}

return ret;
}

//@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

//@@@@@@@@@@@@@@ Here I am logging the my audio data @@@@@@@@@@@

// Enter main loop.
for (;;)
{
idle_state_handle();

for(int i = 0; i < (AUDIO_FRAME_WORDS / 2) ; i++)
{
NRF_LOG_RAW_INFO("%d\n,",data_audio[i]);


NRF_LOG_FLUSH();
}

}//FOR LOOP CLOSED BRACKET
}//MAIN LOOP CLOSED BRACKET

Please let me know the right settings for getting audio data. 

[Jørgen Holmefjord]

Hard to say for me by just looking at your code if this is correct or not. Did you test it together with the device? Did you encounter any issues?