STM32U5: Spi delay after start

[FKaes.1]

Hello,

after setup SPI1 and setting CSTART=1, I expected that SPI is send immidiatly on bus.

(I toogled PE00 with CSTART=1)

I couldn't find some that explains that behaviour. It behave  the same, if I use DMA or writing directly to TXDR.

Do you have an idea ?

My testcode.

  SPI1->IFCR = 0xFFFF; // Clear all flags
  MODIFY_REG(SPI1->CR2, SPI_CR2_TSIZE, BufferSize);
  SPI1->CR1 |= (SPI_CR1_SPE);
  SPI1->CR1 |= SPI_CR1_CSTART;

  GPIOE->ODR &= ~(1 << 0);
  for (u32 i = 0; i < BufferSize; i++)
  {
    LL_SPI_TransmitData8(SPI1, pBuffer[i]);
  }

CPU is set to 160Mhz and SPI Prescaler is 256

[Saket_Om]

Hello @FKaes.1 

This bit is set to start SPI communication. However, communication cannot start if there is no data available in the TX FIFO.

[FKaes.1]

Hi, I changed it for testing to "front load" the FIFO with some bytes. Bu still same behavior.

  SPI1->IFCR = 0xFFFF; // Clear all flags
  MODIFY_REG(SPI1->CR2, SPI_CR2_TSIZE, BufferSize);
  SPI1->CR1 |= (SPI_CR1_SPE);

  for (u32 i = 0; i < 8; i++) // front load before start
  {
    LL_SPI_TransmitData8(SPI1, pBuffer[i]);
  }

  SPI1->CR1 |= SPI_CR1_CSTART; // start

  GPIOE->ODR &= ~(1 << 0);
  for (u32 i = 8; i < BufferSize; i++) // send the remaining bytes
  {
    LL_SPI_TransmitData8(SPI1, pBuffer[i]);
  }

[FKaes.1]

For testing, I set the SPI Prescaler from 256 to 32. Then the delay was reduced by factor 8, too. So, I assume it is not caused by SW.

At the end, this isn't a problem for me. I was just interessed, if  this was caused by a wrong setup or usage by my side.

[LLECH.1]

Hello FKaes.1,

Such delay can appear if MSSI bits of CFG2 register are not set to their reset values.

Could you verify this register please ? 

Also I suppose your SPI is configured in Master mode, otherwise it is the received master clock which triggers the start of transfer and not CSTART bit.

Best regards,

[FKaes.1]

This are the settings+config rigth before setting CSTART = 1.

MSSI = 0

Master = 1

HW SS is not indented to be used. Chip select is controlled by SW.

For testing, I set MSSI to the max. value and the delay increased. So, the direction seems to be right.

[LLECH.1 · Accepted solution]

Hello FKaes.1,

In fact, hardware needs 5 clock cycles to have first clock edge from CSTART assertion. It corresponds to 5*256/150 = 8µs. Therefore, the SPI behavior that you can observe is the expected one.

Best regards,

[FKaes.1]

ok, thank you. That explains it. I was just not able to find that in the datasheet / ref manual.