The connection of NRF52832 individual mobile phone models is not stable

[LuBei]

• Nrf52832 chip is used, and the software package is the “nRF5_SDK_16.0.0_98a08e2” version.

• According to customer feedback, Bluetooth connection is generally unstable at present, which is mainly manifested as follows: on the following specific mobile phone models, if we use our app to connect Bluetooth, it will be disconnected after 1-5 minutes of successful connection, while other models of mobile phones are normal。

  Mobile phones with unstable connection: SDK在iPhone X, iPhone 11, Samsung Galaxy S9

[LuBei]

Latest test progress:

• We use the development board plus test routines in the SDK，Chain breaks can also occur.

• The SD version used for the test is “ nRF5_SDK_16.0.0_98a08e2”,The test routine uses the "ble_app_uart".

• The nrf sniffer log information detected is as follows:DEMO PCB+ble_app_uart.pcapng

[Einar Thorsrud]

Hi,

I do not see anything obvious in the sniffer trace. 

LuBei said:

We use the development board plus test routines in the SDK，Chain breaks can also occur.

So you see this also when testing on the nRF52840 DK? What if you test with an unmodified ble_app_uart example?

One typical problem that could cause this behavior is clock inaccuracies. What is the 32 kHz clock source you are testing with? Can you try to increase the margin by changing NRF_SDH_CLOCK_LF_ACCURACY in sdk_config.h to 1 (500 ppm)? Does that have any effect?

[LuBei]

• I'm really sorry for not seeing your reply until now.

• The external crystal oscillator we use is 32.768KHz.I'm going to test the frequency offset first.

• I'll let you know if there is any progress.

[Einar Thorsrud]

Hi,

Sounds good. I look forward to your feedback. As you are using the 32.768 kHz crystal, it could be good to know which crystal you use and which load caps you use. Potentially also seeing the whole schematics and layout (at least around the clock). If for instance you use wrong load cap values, that would offset the clock frequency.