STM32F407 I2S High speed trouble

[alexandr]

Posted on March 09, 2013 at 17:28

Hello peolpes,

maybe somebody can help me. For a testing I use two I2S - one(I2S3) for transmitting, second (I2S2) - for receiving, phisicaly connected at the PCB. At both I use Double Buffer DMA. I fill the sending buffer with the constant, and looking for a receiving buffer to get the same values. I use external clock at PC9 49MHz (by the way - what is the maximum alloud frequency?). At all speeds up to 96K it works fvery goood, but at 192K it totaly not work! Everytime I get a failed values in the receiving buffer, also frame error at I2S2 (receiver). I'm doing everithing according to the datasheet and errata sheet, but I thin something wrong in initialisation (maybe I forget to clear some flags?)

// this is receiver r init:
void I2S2_RCV_Init(void)
{
I2S_InitTypeDef I2S2_InitStructure;
//------------------------------------------------
 I2SGPIO_Init(); 
//
I2S_StructInit(&I2S2_InitStructure);
I2S2_InitStructure.I2S_AudioFreq =I2S_AudioFreq_Default;
I2S2_InitStructure.I2S_CPOL = I2S_CPOL_Low;
I2S2_InitStructure.I2S_Mode = I2S_Mode_SlaveRx;
I2S2_InitStructure.I2S_MCLKOutput = I2S_MCLKOutput_Disable;
I2S2_InitStructure.I2S_Standard = I2S_Standard_MSB;
 I2S2_InitStructure.I2S_DataFormat = I2S_DataFormat_32b;
// wait for WS=1 the for WS=0, i.e. \_ front
WaitForWCLK(1));
WaitForWCLK(0));
I2S_Init(SPI2, &I2S2_InitStructure);
I2S2_DMA_Init();
DMA_Cmd( Rx_DMAStream, ENABLE );
I2S_Cmd(SPI2, ENABLE);
}
// WS waiting function:
static u8 WaitForWCLK(u8 edge)
{
TOTmr1ms = WCLK_TIMEOUT; // TOTmr1ms is decremented in SysTick interrupt
if(edge>0){ // wait for WCLK=1
while (0==(GPIOB->IDR & GPIO_Pin_12)){
if(TOTmr1ms==0) 
return 0;
} 
return 1;
}else{ // wait for WCLK=0
while (0!=(GPIOB->IDR & GPIO_Pin_12)){
if(TOTmr1ms==0) 
return 0;
} 
return 1;
}
}

DMA Function:

static __INLINE void I2S2_DMA_Init( void ) // ******* D M A RECEIVER ****************
{
DMA_InitTypeDef DMA_InitStruct;
NVIC_InitTypeDef NVIC_InitStructure;
DMA_Cmd( Rx_DMAStream, DISABLE );
while( DMA_GetCmdStatus( Rx_DMAStream ) == ENABLE ); // wait for the last DMA end
RCC_AHB1PeriphClockCmd(RCC_AHB1Periph_DMA1, ENABLE);
// From:
DMA_InitStruct.DMA_PeripheralBaseAddr = (uint32_t)&(SPI2->DR); // I2S Data
// To:
DMA_InitStruct.DMA_Memory0BaseAddr = (u32)&DMARCVBUF[0][0]; // memory start
DMA_InitStruct.DMA_BufferSize = DMARCVBUF_SIZE; // length
// With parameters:
DMA_InitStruct.DMA_Channel = DMA_Channel_0; 
DMA_InitStruct.DMA_DIR = DMA_DIR_PeripheralToMemory; // direction
DMA_InitStruct.DMA_PeripheralInc = DMA_PeripheralInc_Disable; // 
DMA_InitStruct.DMA_MemoryInc = DMA_MemoryInc_Enable; 
DMA_InitStruct.DMA_PeripheralDataSize = DMA_PeripheralDataSize_HalfWord; // 16 bit per
DMA_InitStruct.DMA_MemoryDataSize = DMA_MemoryDataSize_HalfWord; // 16 bit memory
DMA_InitStruct.DMA_Mode = DMA_Mode_Circular; // circ
DMA_InitStruct.DMA_Priority = DMA_Priority_High; // prority
DMA_InitStruct.DMA_FIFOMode = DMA_FIFOMode_Disable;
DMA_InitStruct.DMA_MemoryBurst = DMA_MemoryBurst_Single; 
DMA_InitStruct.DMA_PeripheralBurst = DMA_PeripheralBurst_Single;
DMA_Init( Rx_DMAStream, &DMA_InitStruct ); // Init New
DMA_ITConfig( DMA1_Stream3, DMA_IT_TC, ENABLE ); 
DMA_DoubleBufferModeConfig(Rx_DMAStream, (uint32_t)RCVBuffer_2, DMA_Memory_0 );
DMA_DoubleBufferModeCmd( Rx_DMAStream, ENABLE );
SPI_I2S_DMACmd( SPI2, SPI_I2S_DMAReq_Rx, ENABLE);
}

P.S. CPU - STM32F407VG #stm32f4xx-i2s-dma

[waclawek.jan]

Posted on March 15, 2013 at 12:19

It took me some time to get to setting this up.

Put out some helper signal onto PB14, see modified source below (btw. d=(uint32_t)SPI3->DR; // to clear RXNE bit I know it was an act of desperation, but SPI3???). Turns out the problem is that you spend too much time between detecting the WCLK edge (see picture - red cursor A) and the I2S is enabled (blue cursor B). Pure luck the gcc I use compiled a similar sequence than your Keil... 🙂 Simply place the ''wait for WCLK edge'' just before the I2S enable - and don't forget to decrease the latencies by disabling interrupts. JW ------

//----------------------------------------------

u8 InitI2S2(void){ // slave receiver GPIO_InitTypeDef GPIO_InitStructure; // NVIC_InitTypeDef NVIC_InitStructure; I2S_InitTypeDef I2S2_InitStructure; volatile u32 d; RxDMADisable(); // RCC_I2SCLKConfig(RCC_I2S2CLKSource_Ext); RCC_APB1PeriphClockCmd(RCC_APB1Periph_SPI2, ENABLE); // GPIO GPIO_PinAFConfig(GPIOB,GPIO_PinSource12, GPIO_AF_SPI2); // WCLK GPIO_PinAFConfig(GPIOB,GPIO_PinSource13, GPIO_AF_SPI2); // BCLK GPIO_PinAFConfig(GPIOB,GPIO_PinSource15, GPIO_AF_SPI2); // SDATA GPIO_StructInit(&GPIO_InitStructure); GPIO_InitStructure.GPIO_Pin = GPIO_Pin_12 | GPIO_Pin_13 | GPIO_Pin_15; // WCLK-BCLK-SDATA GPIO_InitStructure.GPIO_Mode = GPIO_Mode_AF; GPIO_InitStructure.GPIO_PuPd = GPIO_PuPd_NOPULL; GPIO_InitStructure.GPIO_Speed = RCV_Speed; GPIO_Init(GPIOB, &GPIO_InitStructure); GPIO_StructInit(&GPIO_InitStructure); GPIO_InitStructure.GPIO_Pin = GPIO_Pin_14; GPIO_InitStructure.GPIO_Mode = GPIO_Mode_OUT; GPIO_InitStructure.GPIO_PuPd = GPIO_PuPd_NOPULL; GPIO_InitStructure.GPIO_Speed = RCV_Speed; GPIO_Init(GPIOB, &GPIO_InitStructure); GPIOB->BSRRL = 1 << 14; // set high //------------------------------------------------ I2S_StructInit(&I2S2_InitStructure); I2S2_InitStructure.I2S_AudioFreq = I2S_AudioFreq_Default; I2S2_InitStructure.I2S_CPOL = I2S_CPOL_Low; I2S2_InitStructure.I2S_Mode = I2S_Mode_SlaveRx; I2S2_InitStructure.I2S_MCLKOutput = I2S_MCLKOutput_Disable; I2S2_InitStructure.I2S_Standard = I2S_Standard_MSB; I2S2_InitStructure.I2S_DataFormat = I2S_DataFormat_32b; GPIOB->BSRRH = 1 << 14; // set low if(0==WaitForWCLK(1)) return 0; // return if timeout if(0==WaitForWCLK(0)) return 0; // ---//---- GPIOB->BSRRL = 1 << 14; // set high I2S_Init(SPI2, &I2S2_InitStructure); I2S2_DMA_Init(); DMA_Cmd( Rx_DMAStream, ENABLE ); GPIOB->BSRRH = 1 << 14; // set low I2S_Cmd(SPI2, ENABLE); GPIOB->BSRRL = 1 << 14; // set high d=(uint32_t)SPI3->DR; // to clear RXNE bit __ISB(); return 1; } //********************************************************************************************************* ________________ Attachments : t.png : https://st--c.eu10.content.force.com/sfc/dist/version/download/?oid=00Db0000000YtG6&ids=0680X000006HtXz&d=%2Fa%2F0X0000000aRb%2FOX_zfcbhBdrT8jSs6ugYR1DJdxCyUkz.go08_C6xdTY&asPdf=false

[alexandr]

Posted on March 15, 2013 at 13:28

> (btw.   d=(uint32_t)SPI3->DR;   // to clear RXNE bit  

>I know it was an act of desperation, but SPI3???). Probably copy-paste error, when I cut the nesessary parts from all project. btw, I don't know is it must or not to clear this bit here (of course, with actual SPIx :)) >Turns out the problem is that you spend too much time between detecting the WCLK edge (see picture - red cursor A) and the I2S is enabled (blue cursor B). Pure luck the gcc I use compiled a similar sequence than your Keil... 🙂 >Simply place the ''wait for WCLK edge'' just before the I2S enable - and don't forget to decrease the latencies by disabling interrupts. >JW Thank you very much Jan, I will try asap. I can't imagine that:   I2S_Init(SPI2, &I2S2_InitStructure);   I2S2_DMA_Init();   DMA_Cmd( Rx_DMAStream, ENABLE ); are so long, in comparison to WCLK half-period! Id describes many thinkg, like good operation (out of spec) at 384kbps - it was simly next-next WCLK period, and strange dependancy of the optimisation level and ''time'' switch. Again, thank you! Let you know the results!

[emalund]

Posted on March 15, 2013 at 15:34

just curious

never heard of using I²S for interprocessor communications, why did you choose that one

Erik

[alexandr]

Posted on March 15, 2013 at 17:41

Erik, where did I say that I wish to use I2S for interprocessor communication?! SPI or UART are faster and simpler.

It was only for I2S-Slave testing, really I2S will come from external source. Alex.

[emalund]

Posted on March 15, 2013 at 18:48

Erik, where did I say that I wish to use I2S for interprocessor communication?! 

here: For a testing I use two I2S - one(I2S3) for transmitting, second (I2S2) - for receiving, physically connected at the PCB. you did not say ''For a testing something else'' anyhow, you satisfied my curiosity, I was wondering if someone had come up with some  new interprocessor comms method, never too old to learn Erik