SPIS of nrf52 could not read the data accuracy

[hzhuxin]

I am working with SPIS example ( sdk version 15.3). SPI master send data regulaly. For example, it sent 5ms cs low, but there are only two bytes transferred, the first byte at begin of 5ms, other one at the end. SPIS of nrf52 not read the second one, what can I do with it?

[Edvin]

Hello,

From your main.c file, you have:

APP_ERROR_CHECK(nrf_drv_spis_buffers_set(&spis, m_tx_buf, 0, m_rx_buf, 1));

Try to increase the length of the receive buffer, which you currently have set to 1.

Can you try:

APP_ERROR_CHECK(nrf_drv_spis_buffer_set(&spis, m_tx_buf, 2, m_rx_buf, 2));

[hzhuxin]

I had tried from 2 to 20, the same result.

[hzhuxin]

Yes, i create the graph.
but no any info lost.

CLK only had two 8 bits cycle, one at the cs got low, another at before cs to high.

CLK & CS relationship, please check the attachment.

The data in gray box was lost. All other data are no problem.

[Edvin]

Hello,

I assume the yellow line is your CLK and the blue is the MOSI? To me it looks like there is nothing in the MOSI in the yellow box. It looks like it belongs to the 3rd 8 but cycle on the CLK. 

However, where do you set up your SPIS to read? Can you send some snippets?

[hzhuxin]

blue line is CS

I read the SPIS in spis_event_handler, the code were post 3 days ago!

[hzhuxin]

I post another photo to show the MOSI and CLK, so you can clear to read the detail infomation, yellow line is the CLK and blue line is MOSI.

[Edvin]

Hello,

When you set the buffer to 20 bytes (the length), what are the values in the buffer when the event is triggered? 

BR,

Edvin

[hzhuxin]

Only 1 byte was received, that's the reason i launch the item.

[hzhuxin]

Hi, Edvin,

I found something wrong using the logic analyzer.

The problem was cause by the second cs signal, it's stay on high just take 3.5us which was too small than the others(34us).

Is there any thing i can changed to solve the problem? 

BR,

Xin

[Edvin]

Hello,

According to the datasheet the minimum high of the CS/SS is 300ns, so 3.5us should really be fine. 

Are you using the following function to check whether the data is received?

NRF_LOG_INFO(" Transfer completed. Received: %s",(uint32_t)m_rx_buf);

can you try the following:

        NRF_LOG_RAW_INFO("received string len %d, data: 0x", sizeof(m_rx_buf));
        for (uint8_t i=0; i<sizeof(m_rx_buf); i++)
        {
            NRF_LOG_RAW_INFO("%02x:", m_rx_buf[i]);
        }
        NRF_LOG_RAW_INFO("\n");

I am sorry I can't give a concrete answer, but I don't have any HW capable of simulating your device. Please test this with the unmodified spis example, except for the printing of the string. I don't understand exactly how you handle the SPI events in the main file that you sent.

BR,

Edvin

[hzhuxin]

I had tried unmodified spis again, only add the instructions you give to me, no any more changed, the second bytes was lost, and many bytes not printed in log info, but only the first was printed with server 0x00. i think maybe print log info take so much time, so some bytes was lost.

I tried to store all data to a big buffer like the amount reached to 200 without printing any log info, i tried setup the break point when the 200 data had stored to the buffer, then to find the second bytes, result tell me i was fail again.

So i can sure, your minimum high of the CS 300ns is wrong, because i can get the second byte with logic analyzer but the 52832 could not. Could you give me any advice?

[Edvin]

Is it possible to save the file that this screenshot is from?

I would like to measure some timings. It may be that the nRF doesn't have time to start the clocks. Can you try to enable the HCLK and keep it running? Does that change the behavior?

BR,

Edvin