SPI Master is broken in SDK v1.4.x

[Danny Zadok]

Hello,

 After a long investigation I got to the conclusion that SPI master (spi3) has been broken in the SDK version 1.4.1.

It is working fine in V1.3.1

I have an application that communicates with Analog Devices MEMS, ADXL335, via SPI.

The application was working fine, until I upgraded to the new SDK v1.4.1 due to some bugs in the cloud APIs.

But then I noticed that the SPI stopped working.

Going back to v1.3.1, the application is back working (the Could API is not

Here is my overlay (nrf9160dk_nrf9160ns.overlay)

&spi3 {
compatible = "nordic,nrf-spim";
status = "okay";
sck-pin = <19>;
mosi-pin = <18>;
miso-pin = <17>;
cs-gpios = <&gpio0 10 GPIO_ACTIVE_LOW>;
};

As you see, I am using GPIO 10 as the CS. I noticed that a new member has been added to the spi_cs_control structure (gpio_dt_flags).

struct spi_cs_control           g_spi_cs_ctrl;

g_spi_cs_ctrl.gpio_dev = g_p_gpio_dev;
g_spi_cs_ctrl.gpio_pin = 10;
g_spi_cs_ctrl.delay = 0;
//g_spi_cs_ctrl.gpio_dt_flags = GPIO_OUTPUT_LOW; 

const static struct spi_config spi_cfg =
{
.operation = SPI_OP_MODE_MASTER | SPI_WORD_SET(8) | SPI_LINES_SINGLE | SPI_TRANSFER_MSB, 
.frequency = 500000, /* The frequency in Hz to use when testing in slow mode. Slow: 500000, Fast=16000000 */
.slave = 0, /* Slave number from 0 to host controller slave limit */
.cs = &g_spi_cs_ctrl, /* Chip-Select Control GPIO */
};

In version v.1.3.1 the SPI communication with sensor works fine and I get valid values. In v1.4.1 all the returned values from the sensor are ff or .

I need this issue would be fixed. Because I need both cloud and sensor in my program.

Danny

[Simon]

I just tested an SPI loopback example in NCS v1.4.1 using SPI3 and nRF91 DK. It worked without problems, can you test it out and see if it works for you?

spi_loopback.zip

Best regards,

Simon

[Danny Zadok]

Hi Simon

 Thanks for your reply.

The loopback application is working. I already tested it in the past. This simple application uses MISO and MOSI lines only.

It does not use GPIO as CS line. The problem is with the CS line as I wrote in my original ticket.

Danny

[Simon]

Aah, okay. I was too quick. I'll look into it.

[Simon]

I don't have any sensors with SPI at hand right now that I can test with, but have you looked at the sensor samples in ncs\v1.4.1\zephyr\samples\sensor, and use them as a reference how to go about it. E.g. look at samples/sensor/adxl362, which comes with an nrf52832dk overlay file as well.

Best regards,

Simon

[Danny Zadok]

Hi Simon,

 How don't know how this sample helps me. This sample based on a different driver not SPI.

My claim is that SPI CS is not working in version 1.41 but it is working in version 1.3.1.

Danny

[Simon]

The reason I linked to the sample  samples/sensor/adxl362, is because it seems to use SPI CS and the SPI driver. It may seem like the sample doesn't use the SPI driver, is because it is not the top layer, however if you look at the image under NCS Tutorial part 3-->Use the sensor drivers-->How it works-->The call stack, you can see that it does.

---

However, a colleague just made an SPIM sample using SPI CS, he made an SPIS sample as well that you could test it with. I'll upload both.

spis_thread_test.zip7585.spi_test.zip

Best regards,

Simon

[Danny Zadok]

Thanks Simon

I do not see the master side, but only the salve in this example.

Danny

[Simon]

Sorry, I forgot to press enter between the uploads, and the naming wasn't too good either, but spi_test.zip should contain the SPI master:

Best regards,

Simon