nRF52832 (SPI master function) only detects accelerometer KX224 (SPI slave) when JTAG debugger is connected; but SPI bus oscillogram is perfect

[matthk]

Hello together,

I have a strange problem and don't know how to proceed...

At my customer-specific electronic a Kionix KX224 is placed next to the BLE module which contains the nRF52832. When I connect my jTracePro Debugger everything works fine. But when using that electronic without the debugger, all other sensors using GPIOs or TWI work well and the results are transmitted by BLE as planned. Only the KX224 is not detected by the nRF52832, although the oscilloscope shows perfect answer from the KX224; the SPI decoding module from the oscilloscope shows the expected data.

I tried already to switch on internal Pull-Ups - no change.

Does anybody have an idea?

BR

matthk

[Kenneth]

Hi,

I suggest you add UART to output debug information to try to find what is stopping the Kionix KX224 to work as intended.

You may also want to check out these two Errata which may have an impact here:
http://infocenter.nordicsemi.com/topic/com.nordic.infocenter.nrf52832.Rev1.errata/anomaly_832_108.html?cp=2_1_1_1_1_33 
http://infocenter.nordicsemi.com/topic/com.nordic.infocenter.nrf52832.Rev1.errata/anomaly_832_109.html?cp=2_1_1_1_1_34 

Best regards,
Kenneth

[matthk]

Hi Kenneth,

thanks - I will check that. But please note that it's not the KX224 - this one works perfect and the oscillogram shows a perfect response - it's the nRF52832 which doesn't want to read the data transmitted by SPI (at MISO) when the debugger is connected. But the /CS, CLK and MOSI are generated perfectly from the nRF...

I will also double-check with another electronics of the same type to be sure that it's not just a partly damaged nRF...

Best regards,

Matthias

[matthk]

Hi Kenneth,

as I have not already activated DMA and furthermore I use SDK13.0.0 I suppose that this is not the problem; should I anyway add:

CFLAGS += -DNRF52_PAN_109 and

ASMFLAGS += -DNRF52_PAN_109

to my makefile?

I have tested now 3 electronics; all show the same behavior.

I have additionally tested the behavior if I start with connected JTAG debugger; then let run and disconnect  while running - this works! So it seems that a starting condition is different when a JTAG debugger is connected.

Do you have another idea why the nRF52832 doesn't want to read-in the data at MISO pin?

Best regards,

Matthias

[matthk]

Ups - I forgot the EasyDMA which is enabled for TWI1 and SPI0 in my project...

[matthk]

No more ideas?

[Kenneth]

Not really, what happens if you physically disconnect the Kionix and wrap MOSI->MISO instead. That should basically loop-back all data.

Having a debugger connected may affect timing and prevent low power modes, you may achieve the same by removing sleep and add some delays after power up for most part. I can't see any obvious problems here, other than try and fail some until you find the problematic behavior.