RE: nrf51822 SPI timing

FUNG — Thu, 25 Jul 2019 07:50:10 GMT

Hi niebert,

Thank you for your reply.

And the code is from spi example in SDK12.2:

int main(void)
{
    APP_ERROR_CHECK(NRF_LOG_INIT(NULL));

    NRF_LOG_INFO("SPI example\r\n");

	nrf_drv_spi_config_t spi_config = NRF_DRV_SPI_DEFAULT_CONFIG;
    spi_config.ss_pin   = SPI_SS_PIN;
    spi_config.miso_pin = SPI_MISO_PIN;
    spi_config.mosi_pin = SPI_MOSI_PIN;
    spi_config.sck_pin  = SPI_SCK_PIN;
		
	spi_config.frequency = NRF_DRV_SPI_FREQ_125K;
	spi_config.mode = NRF_DRV_SPI_MODE_3;
	spi_config.bit_order = NRF_DRV_SPI_BIT_ORDER_LSB_FIRST;
		
//    APP_ERROR_CHECK(nrf_drv_spi_init(&spi, &spi_config, spi_event_handler));
    APP_ERROR_CHECK(nrf_drv_spi_init(&spi, &spi_config, NULL));

	m_tx_buf[0] = 0x5A;
	m_tx_buf[1] = 0x16;
	m_tx_buf[2] = 0x01;
	m_tx_buf[3] = 0x80;
	m_tx_buf[4] = 0x00;

    while (1)
    {
        // Reset rx buffer and transfer done flag
        memset(m_rx_buf, 0, m_length);
//        spi_xfer_done = false;

		APP_ERROR_CHECK(nrf_drv_spi_transfer(&spi, m_tx_buf, 5, m_rx_buf, 5));

//        while (!spi_xfer_done)
//        {
//            __WFE();
//        }

        nrf_delay_ms(1000);
    }

}

without the softdevice, the delay is little bit shorter.

I trace the code nrf_drv_spi.c and can't know where the time is come from.

Thanks.

RE: nrf51822 SPI timing

niebert — Tue, 23 Jul 2019 08:24:50 GMT

(a) I see no good reason to develop for Nrf51822 today. Nrf52810 is cheaper and MUCH more powerful. Check it out.

(b) That said, the delays you observe are in the low hundreds of CPU cycles and can well be due to software latency in the library and in your code. I do not think they are deliberate in the SDK, but maybe in some other code you are using? Without a code snippet, impossible to tell. By optimisation or by writing your own integrated driver, you might be able to reduce this latency, but never get completely rid of it.