STM32 + CS4344 I2S audio DAC

• STM32 + CS4344 I2S audio DAC
  Posted by gab_22 on 10 Jan, 2024 21:31
• Hi everyone!
  
  I'm trying to understand how to use the I2S feature of STM32 cores, in particular, on the STM32F303RE and a low-cost CS4344-based dac board (this one eBay auction: #174915064977 ).
  
  My first goal is to play a sine wave through I2S fed by the DMA channel.
  
  Here, in the attachment, are the configurations for the I2S module and the DMA
  
  And here, is my code
  
  

  Code: [Select]

    /* USER CODE BEGIN 2 */
  HAL_I2S_Init(&hi2s2);
  HAL_I2S_Transmit_DMA(&hi2s2, (uint16_t*)sine2496, 512);
  /* USER CODE END 2 */
  

  Code: [Select]

  /* USER CODE BEGIN 0 */
void HAL_I2S_TxCpltCallback(I2S_HandleTypeDef *hi2s)
{
HAL_I2S_Transmit_DMA(&hi2s2, (uint16_t*)sine2496, 512);
}
/* USER CODE END 0 */
  

  Code: [Select]

  const uint16_t sine2496[] = {

    16383,
    16383,
    16785,
    16785,
    17187,
    17187,
                    ....
  The samples of the sine table (despite its name it's a 16bits/48kHz sampling rate computed table) are duplicated to have the same signal on both stereo channels.
  
  But the sound at the output is not a sine wave. Fortunately, I have a logical analyzer and I sampled that :
  
  Marker 4: Frame one (see the table above) expected value on both channels
  Marker 6: Unexpected values. I expected 16785 on the left and right channels.
  Marker 7: Correct values but wrong place (expected at marker 6)
  Marker 8: Unexpected values. I expected 17187 on the left and right channels.
  Marker 9: Correct values but wrong place (expected at marker 8 )
  etc
  
  I don't understand why I have these unexpected values instead of the correct values.
  
  The I2S signal frequency seems ok, I just need to check the MCLK with an oscilloscope because of the measured min/max/mean frequencies on the screenshot.
  
  Do you have any ideas?
  
  Thank you
  
  
  
  

  ---

  

  

  

  
• #1 Reply
  Posted by mskeete on 11 Jan, 2024 16:33
• Have you tried setting the data width to "half word" instead of byte?
• #2 Reply
  Posted by gab_22 on 11 Jan, 2024 20:02
• Yes, and also with "word" setting. But the output is still not a sine wave.
• #3 Reply
  Posted by DavidAlfa on 12 Jan, 2024 00:32
• DMA configuration is wrong, width must be half-word, both src and dst.
  Also once the DMA ends, no more data is streamed through i2s, so you'll get "silence" chunks between each reloading, distorting the output.
  Set DMA to circular mode and it'll keep going forever without needing that ISR (Remove HAL_I2S_TxCpltCallback), the DMA will reset itself to the beginning of the buffer when done.
  
  Remember HAL_I2S_Transmit_DMA size is number of dma transfers, not bytes.
  With DMA in half-word, sine2496 must have 512 samples here:
  HAL_I2S_Transmit_DMA(&hi2s2, (uint16_t*)sine2496, 512);
  
  If playing a signal, use half-transfer and transfer complete ISRs to update each half of the buffer, this way you' won't run into buffer underrun problems, if you get clipping noises the buffer might be too small, causing underrun.
  I played with this during the pandemic, check:
  https://github.com/deividAlfa/STM32F411-Black-pill-USB-WAV-MP3-player