nrf52840 SPIM0 consecutive TX bug

[gat]

Hi,

I am writing my own SPIM driver, when I try to send a 2 bytes message and then a 26 bytes message the second message "crashes" after 2 bytes.

But if I send the 26 bytes message and then the 2 bytes message, the two messages are correctly sent.

Any known issue with TXD.MAXCNT ?

My code :

#define FLAG_CLEAR() 	NRF_SPIM0->EVENTS_ENDRX = 0;\
						NRF_SPIM0->EVENTS_ENDTX = 0;\
						NRF_SPIM0->EVENTS_STARTED = 0;\
						NRF_SPIM0->EVENTS_STOPPED = 0;\
						NRF_SPIM0->EVENTS_END = 0;

void SPI_setup()
{
	NRF_SPIM0->ENABLE = 0;
	
	NRF_SPIM0->PSEL.SCK = (SPI0_CLK_Port<<5UL) + SPI0_CLK_Pin;
	NRF_SPIM0->PSEL.MISO = (SPI0_MISO_Port<<5UL) + SPI0_MISO_Pin;
	NRF_SPIM0->PSEL.MOSI = (SPI0_MOSI_Port<<5UL) + SPI0_MOSI_Pin;
	NRF_SPIM0->FREQUENCY = SPIM_FREQUENCY_FREQUENCY_M2;
	NRF_SPIM0->CONFIG = SPIM_CONFIG_CPHA_Msk;
	
	NRF_SPIM0->ENABLE = SPIM_ENABLE_ENABLE_Enabled;                           
}

void SPI_write(void* data, uint32_t length)
{
	NRF_SPIM0->RXD.MAXCNT = 0;
	NRF_SPIM0->TXD.PTR = (uint32_t)data;
	NRF_SPIM0->TXD.MAXCNT = length;
	NRF_SPIM0->TASKS_START = 1;
	while(NRF_SPIM0->EVENTS_ENDTX == 0);
	NRF_SPIM0->TASKS_STOP = 1;
	FLAG_CLEAR();
}

When I send 2b and then 26b, you can see the weird crash with additional clock edges...

[gat]

Our investigation concluded that SPIM0 crashes when trying to read RAM at a specific location (0x20006E50)

[JONATHAN LL]

So the workaround/fix with sending CS  between messages fixes the original issue but leads to crash ? or did the crash occur regardless? could you possible share more details around the circumstance?

Regards,
Jonathan

[gat]

Sorry, I got confused during tests but it is definitely reading from another location in RAM that solved the problem. I added a local array variable in the function calling the SPI_write() containing the frame to send (values hardcoded). Also adding another global array containing the exact same bytes worked too (it was a global array at first).

[JONATHAN LL]

Hi,

Great to see that there was a fix for the issue. If this is the solution the feel free to mark the comment as suggested answer. 

Regards,
Jonathan