Problems with reading through SPI with DMA on STM32F407

[bmak]

Hi!

So I configured my SPI to work with DMA and with external CS. Before my main loop I set CS high, and everytime before I initialize communication on SPI with DMA i set CS low. Then, after communication is done I go to interrupt on that event and set CS high again (don't know if it is the best way on doing that, let me know if there is one better :) ).

Okay, so in while(1) loop I transmit-receive data from LIS3DSH. I sent 3 commands and receive 3 sets of data. The trick is that it's working really strange.

So, when I'm sending WHO_AM_I (0x8F) I should receive 0x3F, but whoami value is 0x00.

When I'm sending INFO_1 (0x8D) I should receive 0x21, but info1 value is 0x3F.

And when I'm sending INFO_2 (0x8E) I should receive 0x00, but info2 value is 0x21.

I check tx-rx on oscilloscope, and communication looks perfect, just the values of my variables are weird.

I know that using DMA on SPI causes it go on full speed, maybe I'm just too fast and my variables are changing too slow? I'm not sure.

Here is my code and configuration in CubeMX.

I'm using:

• STM32F407VG
• LIS3DSH (it's not its problem, he works fine)
• Atollic TrueStudio 9.3
• CubeMX 6.2.1

uint8_t whoami, info1, info2;
 
int main(){
   HAL_GPIO_WritePin(SPI_CS_GPIO_Port,SPI_CS_Pin, GPIO_PIN_SET); //CS=1
   while (1)
   {
	  whoami = LIS3DSH_Read(WHO_AM_I);
	  info1 = LIS3DSH_Read(INFO1);
	  info2 = LIS3DSH_Read(INFO2);
     }
}

uint16_t LIS3DSH_Read(uint16_t address){
	uint16_t rx;
 
	HAL_GPIO_WritePin(SPI_CS_GPIO_Port,SPI_CS_Pin, GPIO_PIN_RESET); //CS=0
 
	uint16_t tx = (0x80 | address) << 8; //RW=1
	HAL_SPI_TransmitReceive_DMA(&hspi1,(uint16_t*)&tx, (uint16_t*)&rx, 1); 
	return rx;
}
 
void HAL_SPI_TxRxCpltCallback(SPI_HandleTypeDef *hspi){
	HAL_GPIO_WritePin(SPI_CS_GPIO_Port,SPI_CS_Pin, GPIO_PIN_SET); //CS=1
}

void HAL_SPI_MspInit(SPI_HandleTypeDef* hspi)
{
  GPIO_InitTypeDef GPIO_InitStruct = {0};
  if(hspi->Instance==SPI1)
  {
  /* USER CODE BEGIN SPI1_MspInit 0 */
 
  /* USER CODE END SPI1_MspInit 0 */
    /* Peripheral clock enable */
    __HAL_RCC_SPI1_CLK_ENABLE();
 
    __HAL_RCC_GPIOA_CLK_ENABLE();
    /**SPI1 GPIO Configuration
    PA5     ------> SPI1_SCK
    PA6     ------> SPI1_MISO
    PA7     ------> SPI1_MOSI
    */
    GPIO_InitStruct.Pin = GPIO_PIN_5|GPIO_PIN_6|GPIO_PIN_7;
    GPIO_InitStruct.Mode = GPIO_MODE_AF_PP;
    GPIO_InitStruct.Pull = GPIO_NOPULL;
    GPIO_InitStruct.Speed = GPIO_SPEED_FREQ_VERY_HIGH;
    GPIO_InitStruct.Alternate = GPIO_AF5_SPI1;
    HAL_GPIO_Init(GPIOA, &GPIO_InitStruct);
 
    /* SPI1 DMA Init */
    /* SPI1_RX Init */
    hdma_spi1_rx.Instance = DMA2_Stream0;
    hdma_spi1_rx.Init.Channel = DMA_CHANNEL_3;
    hdma_spi1_rx.Init.Direction = DMA_PERIPH_TO_MEMORY;
    hdma_spi1_rx.Init.PeriphInc = DMA_PINC_DISABLE;
    hdma_spi1_rx.Init.MemInc = DMA_MINC_ENABLE;
    hdma_spi1_rx.Init.PeriphDataAlignment = DMA_PDATAALIGN_HALFWORD;
    hdma_spi1_rx.Init.MemDataAlignment = DMA_MDATAALIGN_HALFWORD;
    hdma_spi1_rx.Init.Mode = DMA_NORMAL;
    hdma_spi1_rx.Init.Priority = DMA_PRIORITY_LOW;
    hdma_spi1_rx.Init.FIFOMode = DMA_FIFOMODE_DISABLE;
    if (HAL_DMA_Init(&hdma_spi1_rx) != HAL_OK)
    {
      Error_Handler();
    }
 
    __HAL_LINKDMA(hspi,hdmarx,hdma_spi1_rx);
 
    /* SPI1_TX Init */
    hdma_spi1_tx.Instance = DMA2_Stream3;
    hdma_spi1_tx.Init.Channel = DMA_CHANNEL_3;
    hdma_spi1_tx.Init.Direction = DMA_MEMORY_TO_PERIPH;
    hdma_spi1_tx.Init.PeriphInc = DMA_PINC_DISABLE;
    hdma_spi1_tx.Init.MemInc = DMA_MINC_ENABLE;
    hdma_spi1_tx.Init.PeriphDataAlignment = DMA_PDATAALIGN_HALFWORD;
    hdma_spi1_tx.Init.MemDataAlignment = DMA_MDATAALIGN_HALFWORD;
    hdma_spi1_tx.Init.Mode = DMA_NORMAL;
    hdma_spi1_tx.Init.Priority = DMA_PRIORITY_LOW;
    hdma_spi1_tx.Init.FIFOMode = DMA_FIFOMODE_DISABLE;
    if (HAL_DMA_Init(&hdma_spi1_tx) != HAL_OK)
    {
      Error_Handler();
    }
 
    __HAL_LINKDMA(hspi,hdmatx,hdma_spi1_tx);
 
    /* SPI1 interrupt Init */
    HAL_NVIC_SetPriority(SPI1_IRQn, 0, 0);
    HAL_NVIC_EnableIRQ(SPI1_IRQn);
  /* USER CODE BEGIN SPI1_MspInit 1 */
 
  /* USER CODE END SPI1_MspInit 1 */
  }
 
}

[T J]

You need to pick up the Rx after the DMA has completed.

also, you have to send the address before you can read the output,

suggest :

    uint16_t LIS3DSH_Read(uint16_t address){
    	uint16_t rx[2];
     
    	HAL_GPIO_WritePin(SPI_CS_GPIO_Port,SPI_CS_Pin, GPIO_PIN_RESET); //CS=0
     
    	uint16_t tx = (0x80 | address) << 8; //RW=1
    	HAL_SPI_TransmitReceive_DMA(&hspi1,(uint16_t*)&tx, (uint16_t*)&rx, 2); // extra 1 to receive the result
    	return 0;
    }
     
    void HAL_SPI_TxRxCpltCallback(SPI_HandleTypeDef *hspi){
    	HAL_GPIO_WritePin(SPI_CS_GPIO_Port,SPI_CS_Pin, GPIO_PIN_SET); //CS=1
        RxCompleted = true;    // data is available in rx[1];    <--- poll this flag in your foreground loop
    }

[bmak]

Hi, thanks for your answer.

It worked after increasing clock speed, i couldn't delete my post.

I will check your suggestion, thanks! :)

[bmak]

Hi, sorry, but I'm not quite sure what you mean by "pick up the Rx after the DMA has completed".

[Sebastiaan · Accepted solution]

With HAL_SPI_TransmitReceive_DMA, you essentially launch a process (it's not a "cpu thread", but you activate the peripheral to use the configured TX and RX buffers to be used for the spi_transceive functionality). The peripheral will execute that process at its own pace. For example, if you want to transmit 1kbyte, at clock speed of 8MHz, it will require 1 msec in total.

But your function "HAL_SPI_TransmitReceive_DMA" does not wait for that timeout, it only launches the process!

So in fact, only when the callback is completed ("HAL_SPI_TxRxCpltCallback"), the data is present in the rx buffer.

You tried to increase the clock speed such that the DMA process works faster, and such that rx is filled faster, but that will only cause the same problem when you want to write or read longer bytestreams, so this is absolutely not the way to go.

In order to fix it, you could either do a simple approach by raising a flag (e.g. spi_done=1) in the callback, and set it to 0 before launching the DMA. After DMA, you should put a while loop waiting for spi_done to become 1.

Or a much better solution is to use freeRTOS operating system and use thread synchronization mechanisms to inform the calling task that the DMA process was done (so it's not stuck in a waiting loop).

I hope that's more clear.

[bmak]

Thank you so much!

Now i understand it :) It works perfect.