Mono audio source in Flash

[Alexmouse]

I want to play an audio clip over i2s from Flash memory. My application is mono and a mono source will fit in flash, but not stereo, it is too big.

The HAL i2s interface does not appear to offer a mechanism to duplicate mono source bytes (16b) onto the stereo stream. Am I missing something? (STM32F411)

[MM..1]

Try header files...

/** @defgroup SAI_Mono_Stereo_Mode SAI Mono Stereo Mode
 * @{
 */
#define SAI_STEREOMODE 0x00000000U
#define SAI_MONOMODE ((uint32_t)SAI_xCR1_MONO)

located ... Users\you\STM32Cube\Repository\STM32Cube_FW_F4_V1.26.2\Drivers\STM32F4xx_HAL_Driver\Inc\...

[Alexmouse]

#if defined(STM32F427xx) || defined(STM32F437xx) || defined(STM32F429xx) || defined(STM32F439xx) ||\

  defined(STM32F446xx) || defined(STM32F469xx) || defined(STM32F479xx) || defined(STM32F413xx) ||\

  defined(STM32F423xx)

Looks like I'm out of luck with my STM32F411 :relieved_face:

[MM..1]

Then you simply in flash store mono and in RAM mini buffer copy stereo. MCU will have work.

[Alexmouse]

OK, I've got this playing my audio clip from Flash, stored in a mono array and converted to "stereo" via a RAM buffer. My amplifier is mono, MAX98357.

All looking good except, replay appears to be correct pitch but playing at double speed and so half the replay duration.

If I play games with the replay sample rate, I can get correct replay rate but pitch is half of that expected.

I've exported a 16KHz sampling rate 16 bit audio source from Audacity, so I'm confused.

As it is simply sampled data, how can pitch be correct but rate doubled?

[MM..1]

You dont show code, or explain method DMA or ? But seems as you send samples and overwrite every second sample or other mistake.

How is count of your mono samples ? Used wide is half word ?

[Alexmouse]

I'm using Audacity to create my WAV source, and HxD to turn this into a header. I'm trying it out with an identical 3.5 second clip, in mono and stereo formats. No DMA just yet, I probably don't need it as the processor won't be very busy during playback.

In stereo:

#define SOURCE_SIZE 225394

const uint8_t source_buff[SOURCE_SIZE] = {

0x52, 0x49, 0x46, 0x46, 0x6A, 0x70, 0x03, 0x00, 0x57, 0x41, 0x56, 0x45,

0x66, 0x6D, 0x74, 0x20, 0x10, 0x00, 0x00, 0x00, 0x01, 0x00, 0x02, 0x00,

0x80, 0x3E, 0x00, 0x00, 0x00, 0xFA, 0x00, 0x00, 0x04, 0x00, 0x10, 0x00,

0x64, 0x61, 0x74, 0x61, 0x48, 0x6F, 0x03, 0x00, 0xFB, 0xFF, 0xED, 0xFF,....

if(HAL_I2S_GetState(&hi2s1) == HAL_I2S_STATE_READY){

status = HAL_I2S_Transmit (&hi2s1, source_buff, SOURCE_SIZE, 100);

}

Audio quality good, but truncated to about 1 second.

In mono:

#define SOURCE_SIZE 126104

const uint8_t source_buff[SOURCE_SIZE] = {

0x52, 0x49, 0x46, 0x46, 0x90, 0xEC, 0x01, 0x00, 0x57, 0x41, 0x56, 0x45,

0x66, 0x6D, 0x74, 0x20, 0x10, 0x00, 0x00, 0x00, 0x01, 0x00, 0x01, 0x00,

0x80, 0x3E, 0x00, 0x00, 0x00, 0x7D, 0x00, 0x00, 0x02, 0x00, 0x10, 0x00,

0x64, 0x61, 0x74, 0x61, 0x6E, 0xEB, 0x01, 0x00, 0x03, 0x00,....

if( playback_active == true ) // time to play audio clip

{

buffer_content = 0;

if(HAL_I2S_GetState(&hi2s1) == HAL_I2S_STATE_READY){

while( (load_ptr < BUFFER_SIZE) && (transfer_ptr < SOURCE_SIZE) ) // load 16 bit buffer with chunk of 8 bit source

{

audio_buffer[load_ptr] = source_buff[transfer_ptr+1] + (source_buff[transfer_ptr] << 8);

audio_buffer[load_ptr+1] = audio_buffer[load_ptr]; // make into "stereo" source

transfer_ptr = transfer_ptr + 2; // two bytes

load_ptr = load_ptr + 2; // for each 16 bit sample

buffer_content = buffer_content + 1; // number of samples in this batch

}

}

if( transfer_ptr >= SOURCE_SIZE )

{

playback_active = false; // reached end of source material

transfer_ptr = 0; // reset pointer from 8 bit source

load_ptr = 0; // reset pointer in 16 bit buffer

}

if(HAL_I2S_GetState(&hi2s1) == HAL_I2S_STATE_READY)

{

load_ptr = 0; // reset pointer in 16 bit buffer

status = HAL_I2S_Transmit (&hi2s1, audio_buffer, buffer_content, 100);

}

}

Plays the full clip, but pitch is low, play rate is too fast. I imagined that pitch and rate would vary together.

[Alexmouse]

Looks like I need to go stereo & DMA. Seems a waste when I will discard 50% of the data.

[Alexmouse]

One sample per call seems to work much better, thanks for the idea!