I'm having a strange issue with getting SPI running through the transaction manager whilst running a S140 softdevice (SDK13 w/ nrf52840 5.0.0-2Alpha - I know there's a newer version but it's not being used just yet. It's there so I don't accidentally code something that's not compatible with the SD when I eventually get around to the BT part of the project)
I am able to get SPI and SPI manager running normally without the SD - they work perfectly into my current test which is CLK and MOSI into a logic analyser (the target will be a screen with no read line - very similar to the screen in the SPI manager example code).
However, when I load up the SD + base code from 'experimental_ble_nfc_pairing_reference' and modify to init. SPI after all of the BT + NFC etc (as you can see, nothing extra on the example apart from the "testing()" call to start the SPI:
int main(void)
{
logs_init();
timer_init();
bool erase_bonds = buttons_init();
ble_stack_init();
/* Set BLE device name. */
ble_set_device_name(DEVICE_NAME);
gap_params_init();
gatt_init();
advertising_init();
conn_params_init();
peer_manager_init(erase_bonds);
nfc_pairing_init();
// Init SPI
testing();
// Start execution.
NRF_LOG_INFO("NFC Connection Handover BLE peripheral device example.\r\n");
// Enter main loop.
while(true)
{
UNUSED_RETURN_VALUE(NRF_LOG_PROCESS());
}
}
This then goes into the following test function:
NRF_SPI_MNGR_DEF(m_nrf_spi_mngr, 1, 0);
static ret_code_t init_spi0_master(void)
{
// SPI0 (with transaction manager) initialization.
nrf_drv_spi_config_t const m_master0_config =
{
.sck_pin = 3,
.mosi_pin = 4,
.miso_pin = NRF_DRV_SPI_PIN_NOT_USED,
.ss_pin = NRF_DRV_SPI_PIN_NOT_USED,
.irq_priority = APP_IRQ_PRIORITY_HIGHEST,
.orc = 0xFF,
.frequency = NRF_DRV_SPI_FREQ_500K,
.mode = NRF_DRV_SPI_MODE_0,
.bit_order = NRF_DRV_SPI_BIT_ORDER_MSB_FIRST
};
return nrf_spi_mngr_init(&m_nrf_spi_mngr, &m_master0_config);
}
void testing(void) {
uint8_t test_str[] = "HELLO";
nrf_spi_mngr_transfer_t transfers =
{
.p_tx_data = (uint8_t const *) &test_str,
.tx_length = (uint8_t) sizeof(&test_str)+1,
.p_rx_data = (uint8_t *) NULL,
.rx_length = (uint8_t) 0,
};
ret_code_t ret = init_spi0_master();
APP_ERROR_CHECK(ret);
ret = nrf_spi_mngr_perform(&m_nrf_spi_mngr, &transfers, 1, NULL);
APP_ERROR_CHECK(ret);
NRF_LOG_FLUSH();
However the issue I'm having is I'm getting terrible synchronisation issues:
And at higher speeds, this becomes even more noticable (the 500K speed in the code is for the sake of making sure it's not something silly like my logic analyser underperforming etc.)
At 8mhz:
Closer look at the data packet:
This early clock firing causes an offset between the data and clock as well which is causing the logic analyser to decode the "HELLO" as garbage.
Interestingly, if I run the transfer in a loop, I have a tiny bit more luck in the fact that the first packet will be the same messed-up clock, but subsequent packets are correctly lined up. This being said, the original clock shift is there causing decoding to fail.
One final thing, enabling / disabling EasyDMA has no impact. As far as the debug logs go? They think everything's alright:
PI:INFO:Transfer tx_len:5, rx_len:0.
SPI:DEBUG:Tx data:
SPI:DEBUG:
48 45 4C 4C 4F 00 00 00 00 00 00 00 E9 69 02 00 HELLO.......éi..
00 00 00 00 ....
SPI:INFO:Function: nrf_drv_spi_xfer, error code: NRF_SUCCESS.
SPI:DEBUG:SPI: Event: NRF_SPI_EVENT_READY.
SPI:DEBUG:SPI: Event: NRF_SPI_EVENT_READY.
SPI:DEBUG:SPI: Event: NRF_SPI_EVENT_READY.
SPI:DEBUG:SPI: Event: NRF_SPI_EVENT_READY.
SPI:DEBUG:SPI: Event: NRF_SPI_EVENT_READY.
SPI:INFO:Transfer rx_len:0.
SPI:DEBUG:Rx data:
Update: Just attempted to recompile with SD 132 - Still having the same issue.
Update 2: Project folder uploaded (project relative path'd to ../sdk/nRF5_SDK_13.0.0_04a0bfd/components etc.)
