Spi transfert uint16_t

[Sani300]

Hi, 

I trying to convert a stm32 function code to nRF52. The goal is to use a spi periperal on my custom board. 

In this application we need to have 2 bytes transfert. That why I need to have an uint16_t function. 

In spi exemple and driver nrf_drv_spi, we have this function nrf_drv_spi_transfer. ANd this function take only uint8_t parameters. 

This is stm32 function : 

uint16_t SpiInOut( Spi_t *obj, uint16_t outData )
{
    uint8_t rxData = 0;

    if( ( obj == NULL ) || ( SpiHandle[obj->SpiId].Instance ) == NULL )
    {
        assert_param( FAIL );
    }

    __HAL_SPI_ENABLE( &SpiHandle[obj->SpiId] );

    CRITICAL_SECTION_BEGIN( );

    while( __HAL_SPI_GET_FLAG( &SpiHandle[obj->SpiId], SPI_FLAG_TXE ) == RESET );
    SpiHandle[obj->SpiId].Instance->DR = ( uint16_t ) ( outData & 0xFF );

    while( __HAL_SPI_GET_FLAG( &SpiHandle[obj->SpiId], SPI_FLAG_RXNE ) == RESET );
    rxData = ( uint16_t ) SpiHandle[obj->SpiId].Instance->DR;

    CRITICAL_SECTION_END( );

    return( rxData );
}

Another advantage of this function, that we can send and receve SPI information !

Do you have some advise/help to help me for this translation? 

Sani300

[Turbo J]

That specific STM32 function will only ever work properly with the low 8 bits of data:

 SpiHandle[obj->SpiId].Instance->DR = ( uint16_t ) ( outData & 0xFF );

Note the AND 0xFF part in this line.

The actual amount of tranferred bits (per DR write) is in the SPI configuration somewhere - anything but 8 bits is highly unlikely.

You can safely port this function to the 8 bits used in NRF hardware.

[Sani300]

Thanks for your reply, so if I want exacly the same I do this :

    uint8_t rxData = 0;
  
    if (obj->initialized != true)
        return;
    
    tx1 = outData
    tx2 = 0xFF

    nrf_drv_spi_transfer(&obj->Instance, tx1, 1, NULL, 0);
    nrf_drv_spi_transfer(&obj->Instance, tx2, 1, NULL, 0);

    nrf_drv_spi_transfer(&obj->Instance, NULL, 0, rx1,1);
    nrf_drv_spi_transfer(&obj->Instance, NULL, 0, rx2,1);

    rxData = rx1 & rx2;

    return( rxData );

Is that right? 

[Turbo J]

That seems very different from the STM32 code above.

[Sani300]

If i have understand STM 32 code : 

while( __HAL_SPI_GET_FLAG( &SpiHandle[obj->SpiId], SPI_FLAG_TXE ) == RESET );
SpiHandle[obj->SpiId].Instance->DR = ( uint16_t ) ( outData & 0xFF );

Send by SPI : OutData & 0xFF

while( __HAL_SPI_GET_FLAG( &SpiHandle[obj->SpiId], SPI_FLAG_RXNE ) == RESET );
rxData = ( uint16_t ) SpiHandle[obj->SpiId].Instance->DR;

Receive by SPI : reply from slave

New code : 

Build tx frame : 

      uint8_t rxData = 0;
      
      uint8_t tx[1];
      uint8_t rx[1];

      tx[0] = outData & 0xFF;

Send and receive frame by SPI 

      nrf_drv_spi_transfer(&obj->Instance, tx, 1, rx, 1);

      rxData = *rx;

      return( rxData );