Development Issues When Using the nRFx Library in Zephyr

[joey]

HI，

I am currently developing with nRF54L15 and SDK 3.0.0. I have encountered some issues:
1. Problems using the nrfx library in Zephyr
I have defined the configuration information of the SPIM30 node in the devicetree. According to the Zephyr manual, this configuration is initialized before running the application. Therefore, I directly obtained the instance of spim30 in the program and used the nrfx spim library to send data. However, I found that the originally set 8M in devicetree was actually only 1M, and the CS pin output level was supposed to be high after transmission but remained low. But when I directly used the Zephyr spi API, there was no problem. The nrfx is shown in the following figure. Why is this happening?How to deal with this problem?

2. I have used nrfx in Zephyr and set CONFIG_DEVICE_PM_RUNTIME. Does it have no effect?
3. In my program, I am using grtc + dppi + spim, all of which are using nrfx library functions. How should the power management of spim be handled? Since I am using the nrfx library and only using dppi to trigger?

Thanks!

[Vidar Berg]

Hello,

The configurations from your spi30 node in the devictree will not be automatically applied when you are using the nrfx drivers directly like they are when using the Zephyr SPI driver API. Please verify that .frequency value passed to nrfx_spim_init() is configured correctly.

https://github.com/zephyrproject-rtos/zephyr/blob/e9cf74b581a1371f77862eea1e3cf24cadd457c6/samples/boards/nrf/nrfx_prs/src/main.c#L145 

Best regards,

Vidar

[joey]

Thank you for your reply

After initializing SPIM with nrfx_spim_init, I was able to communicate normally by calling nrfx_spim_xfer. Then, I configured the DMA array list and used nrfy_spim_task_trigger(spim30_inst.p_reg, NRF_SPIM_TASK_START) to trigger the task. However, when observing the oscilloscope, there were no waveforms and the program was not running properly. What could be the problem?

[joey]

However, these functions of mine are undefined.,How should it be set to be enable?

[Vidar Berg]

This was a bug which as been addressed in SDK v3.1.0. I recommend upgrading if possible. If not, you can change NRF_SPIM_HAS_ARRAY_LIST to '1' in your SDK.

[joey]

OK,

In the program, I used the nrfx library to operate the spim, especially combining with dppi to operate the spim. Now, regarding the original power management of Zephyr, how should I operate it to ensure that the device can enter sleep mode normally?

[Vidar Berg]

Zephyr will enter sleep (System ON) independently of the SPI activity. CONFIG_PM_DEVICE_RUNTIME only affects Zephyr drivers which support the runtime power management which usually involves disabling the peripheral during idle periods. Was the goal to reduce the system current consumption when you decided to use SPI with DPPI and GRTC, or is it to meet timing requirements in your application? The CPU overhead with triggering the SPI transactions is usually negligible (i.e. no significant impact on the average power consumption) 

[joey]

I was originally developing the SPI using Zephyr's APIs, but found that using a k_timer to trigger SPI data transmission every 1 second resulted in a power consumption of about 29 μA. I want to directly use the GRTC to trigger the SPIM via DPPI for data acquisition to reduce CPU power consumption. However, since the program uses the DPPI+SPIM approach, could Zephyr's power management shut down these operations during sleep? If it could, how can I ensure that the DPPI+SPIM operations remain active while also allowing the peripherals to enter low power mode after each operation?

[Vidar Berg]

joey said:

I was originally developing the SPI using Zephyr's APIs, but found that using a k_timer to trigger SPI data transmission every 1 second resulted in a power consumption of about 29 μA.

I don't think you will see much improvement considering the long interval between your transactions. Did you confirm that the added current was because of the CPU, and not leakage, current draw by the SPI slave, etc.?

joey said:

could Zephyr's power management shut down these operations during sleep?

No.