Trouble exporting BLE server from Nucleo to custom board.

[Mickael1]

Hello,

Working on a custom BLE server (peripheral role), I face a trouble exporting the working sources from a Nucleo project (STM32WB55) to a custom board (STM32WB55) project.

Everything is working fine on the Nucleo: advertissing, connection, bidirectional communication etc...

On the custom board, the BLE master sees the advertissing slave, but there is no way to connect to it. It fails connecting!

2 softwares were used on the Master side: ST BLE Toolbox and nRF Connect. Both give same results.

The nRF connect show an additional log error 133(0x85): GATT Error.

The custom board does not implement LSE crystal. HSE is used instead. I follow this URL but cannot find CFG_BLE_LSE_SOURCE in BLE configuration.

(community.st.com/s/article/configuring-stm32wb-for-bluetooth-le-without-an-lse-crystal).

I was expecting a trouble from this point, and found several ticket on the knowledge base, but no one have solution:

- community.st.com/s/question/0D50X0000C3BaK3SQK/migration-from-nucleo-to-my-own-board-doesnt-work

- st.force.com/community/s/question/0D53W000006qrigSAA/stm32wb55-pnucleowb55-advertising-works-but-i-cant-connect-when-building-application-using-code-generated-by-cubemx

No uart debug is possible on the custom board. Is there a way to use custom server application debug trace though STLink/JTAG?

The following items were verified on the custom board:

- Master clock is the expected one, verified on MCO pin.

- The application heartbeat is the expected. The system boot properly as well as the BLE full firmware stack.

- Debugging though STlink/JTAG is working.

- The RTC_WKUP_IRQHandler is verified woked up every 100ms (by toggling a pin).

- The advertising name, device type, tx power level and flags are properly received on the BLE master.

- Connect and Disconnect requests from master trigger the slave to go through Custom_APP_Notification CUSTOM_CONN_HANDLE_EVT and CUSTOM_DISCON_HANDLE_EVT cases on the custom application. This looks like a timeout is triggered on the master, not receiving properly the connection request.

Do you have any clue to help debug the situation?

[Remy ISSALYS · Accepted solution]

Hello,

Since your custom board doesn't implement LSE, you should use HSE/1024 for RF wakeup clock, and for this, you should enable the calibration for the RF system wakeup clock source in STM32CubeWB package v1.14.1, so you should have this configuration:

/**
 * LsSource
 * Some information for Low speed clock mapped in bits field
 * - bit 0:   1: Calibration for the RF system wakeup clock source   0: No calibration for the RF system wakeup clock source
 * - bit 1:   1: STM32W5M Module device                              0: Other devices as STM32WBxx SOC, STM32WB1M module
 */
 
#if defined(STM32WB5Mxx)
  #define CFG_BLE_LSE_SOURCE  (SHCI_C2_BLE_INIT_CFG_BLE_LSE_CALIB | SHCI_C2_BLE_INIT_CFG_BLE_LSE_MOD5MM_DEV)
#else
  #define CFG_BLE_LSE_SOURCE  (SHCI_C2_BLE_INIT_CFG_BLE_LSE_CALIB | SHCI_C2_BLE_INIT_CFG_BLE_LSE_OTHER_DEV)
#endif

Best Regards

[Mickael1]

Dear Remy,

This is it!

Thank you so much.

One thing to take care:

CubeMX generates the SHCI_C2_BLE_INIT_CFG_BLE_LSE_NOCALIB for CFG_BLE_LSE_SOURCE, even when HSE/1024 is seleted.

The correction of the RF system wake up clock source calibration needs to be handled manually after cubeMx code generation, between USER CODE BEGIN/END.

Sincerely

[Remy ISSALYS]

Hi,

Yes, you have right, in the next release (v1.15.0), the define CFG_BLE_LSE_SOURCE will be managed by cubeMX according to RF system wake up clock configuration.

Best Regards