BLE radio goes silent

[PB]

Hello,

Over the years we've occasionally been experiencing a issue with nRF52840 and nRF52832 chips in case of which any BLE activity would stop happening. A chip mounted on a PCB works perfectly fine at first, after some unknown event it still works ok as to UART, SWD and peripherals but no BLE events ever happen. It doesn't advertise nor scan anymore. It's not that signal strength is weaker but that the device can't do any BLE anymore. So far we've accepted this as the scale was very small but yesterday this happened to 2 nRF52840 chips in a row while flashing a few times and we started to wonder what is causing it.

As to technical details and things checked so far:

1. Happens on nRF52832 and nRF52840 chips without any clear reason. Sometimes we just pick up a device flashed & tested months before and it's not BLE capable anymore.
2. It will persist between different firmware. Eg. Zephyr RTOS HCI USB normaly works perfectly fine on the device but if BLE fails it will also make HCI not scan anything anymore with bluetoothctl.
3. We tried using DTM on the failing boards but we could not get it to execute commands - the DTM UI in pcnRFConnect failed on the boards which had BLE issues. This is interesting as we know that UART works well with our firmware so there should be no issues with controlling DTM firmware over the same channel.
4. Since it happened over a longer time period it's unlikely related to any specific batch of chips
5. The PCBA wasn't damaged. Other peripherals work perfectly fine. It's unlikely ESD because the board was mounted in the tester and nobody touched the PCBA between the moment it worked and until it stopped
6. Device is battery-less, powered directly from USB. No issues here
7. We don't know any way to recover when it happens

I've found some information that the high frequency clock can cause similar issues but we are yet to inspect the hardware more in depth. Right now we're trying to understand what can be the reason so we can avoid it in the future.

Best regards

[PB]

Hello Bendik,

This falling devices are not able to start with DTM Firmware so we're not able to check constant carrier. They still start on our Firmware, but without any BLE. Also we checked 32MHz quartz (along with its Cp capacitors) and replacing it didn't fix the issue. 

The crystal used is ECS (ECS-320-18-33-JGN-TR)

Best regards,
PB

[Bendik Heiskel]

Hi,

PB said:

This falling devices are not able to start with DTM Firmware so we're not able to check constant carrier. 

Have you check if the radio test firmware works?

https://github.com/nrfconnect/sdk-nrf/tree/main/samples/peripheral/radio_test

You can use it to start a constant carrier and measure the frequency accuracy of the 32MHz crystal by measuring the frequency offset from the selected channel of the carrier.

The frequency stability of the crystal is high, which means that the worst case frequency deviation from 32MHz could be 50ppm. The BLE specifications have a limit of 40ppm carrier offset, so a few devices with a bad crystal could be out of spec.

PB said:

Also we checked 32MHz quartz (along with its Cp capacitors) and replacing it didn't fix the issue.

Could you try replacing the crystal of a failing device with the crystal from one that's working?  It it still doesn't work the crystal is not the problem.

 

Best regards,

Bendik