AVR Freaks

I am intrigued.   The SAM3XE has only got SPI0 but it has USART0, USART1, USART2 with available SCK pins.

On the Mega2560 they carefully removed access to all of the USART XCK pins.   Preventing you using any of the 4 USARTs in MSPI mode.

The Due can access PA17, PA16 but not PB24.   So you can use both USART0 and USART1 in SPI mode.

I don't see the problem.   The SPI0 peripheral can use 8-16 bits for SPI transfers.   The USART_SPI can only use 8 bit transfers.

If you want native 16-bit just use SPI0.

If you need to use the USART hardware pins,  just do two 8-bit transfers.

Both SPI0 and USARTn peripherals have TX and RX buffers.   Both can use DMA.

This is all from my reading of the datasheets.  I don't have intimate experience of SAM3XE.

It seems fairly unimportant for a handful of 16-bit transfers.    But would give impressive performance when or if you want to transfer millions of 16-bit data e.g. to graphics controller chips.

David.

SPI always writes TX and receives on RX.   i.e. the hardware always does a transfer.

So if you want to receive 16-bit you write a 8-bit dummies.   And read the two 8-bit data that was received.

The buffering makes it important that you do the correct number of writes and reads.   A simple way to implement this is to write a single 8-bit function:

uint8_t usart_spi_transfer(uint8_t c);

Then you always keep in step.

Most SPI devices require a command before you receive the appropriate response e.g.

     usart_spi_transfer(command);

     uint8_t hi = usart_spi_transfer(0);  //dummy

     uint8_t lo  = usart_spi_transfer(0); //dummy

     return (hi << 8) | lo;

Of course you can write a single uint16_t usart_spi_transfer16(uint16_t c);

using the previous example.

David.