SPI MISO seems to be pulled down in NCS

[Mandrii]

Hello. I migrated my code from nRF5 SDK for Thread and Zigbee to nRF Connect SDK (NCS). After that SPI code stopped working :( After debugging with the oscilloscope, I found that SS and MOSI signals are correct, the ADCxx1Sxx1 chip is sending the correct values to MISO line, but MISO pin on the nRF52840 side is reading all zeroes (ADC chip is separated from nRF by a digital isolator ISO772x). When I connect VCC to MISO with a wire however, the software can read all ones instead of zeroes (so SPI is working). 

I guess, this is an issue of internal pull-down resistor on SPI? In the devicetree documentation of Zephyr 2.5.0 I found the "miso-pull-up" and "miso-pull-down" options. The latter doesn't change anything for me, while the former changes the values read from SPI to all ones, unless the MISO pin is shorted to GND.

Is there a fix for this? Can I disable pulldown? Is there a "no pull" option for MISO in NCS?

For MISO, I am using pin P0.09. (CONFIG_NFCT_PINS_AS_GPIOS=y is set in prj.conf.)

P.S. Versions:

• NCS Toolchain: 1.5.1
• Zephyr: 2.4.99
• west: 0.9.0
• dtc: 1.4.7
  

[Håkon Alseth]

Hi,

 

Is there a fix for this? Can I disable pulldown? Is there a "no pull" option for MISO in NCS?

For no pull, remove the "miso-pull-*" in your DT overlay, and recompile.

 

For MISO, I am using pin P0.09. (CONFIG_NFCT_PINS_AS_GPIOS=y is set in prj.conf.)

its important that you verify this by reading the register:

https://infocenter.nordicsemi.com/topic/ps_nrf52840/uicr.html?cp=4_0_0_3_4_0_5#register.NFCPINS

 

via nrfjprog:

nrfjprog --memrd 0x1000120c -f nrf52

 

 

nRF52840 side is reading all zeroes (ADC chip is separated from nRF by a digital isolator ISO772x). When I connect VCC to MISO with a wire however, the software can read all ones instead of zeroes (so SPI is working). 

 This is quite strange. Have you tried reducing the frequency of the SPIM to see if this has an impact on the read data?

Does this issue happen on more than one board?

 

PS: are you using the nRF52840-DK for testing? If yes, then the P0.09/P0.10 GPIOs are not routed by-default to the pin headers. You then need to connect some solder bridges: 

https://infocenter.nordicsemi.com/topic/ug_nrf52840_dk/UG/dk/hw_nfc_if.html?cp=4_0_4_7_10 

 

Kind regards,

Håkon

[Mandrii]

Håkon Alseth said:

For no pull, remove the "miso-pull-*" in your DT overlay, and recompile.

Tried that. Originally, there were no "miso-pull" options in my DTS file.

Håkon Alseth said:

its important that you verify this by reading the register:

I am getting:

$ nrfjprog --memrd 0x1000120c -f nrf52
0x1000120C: FFFFFFFE

– looks correct. 

Håkon Alseth said:

Have you tried reducing the frequency of the SPIM to see if this has an impact on the read data?

Yes, SPI clock speed is at the minimum (125 kHz).

Håkon Alseth said:

are you using the nRF52840-DK for testing?

No, we are using our custom board with MS88SF2 v1.0 module.

Håkon Alseth said:

Does this issue happen on more than one board?

Unfortunately, we have only one custom dev board of this kind for now. Because it works fine with different firmware (the one compiled with nRF5 SDK for Thread and Zigbee) we didn't think that the board could have an issue. But we will look into that too.

[Mandrii]

We fixed the problem. 

It was completely on our software side: second SPI CS pin was active in the new firmware, while I tested another pin.

Thanks for your help anyway