The shift register approach is nice if you don't want to optimize the power down to the last microamp. However, if you have e.g. a button input that may be pressed once a day or once an hour, you don't want to keep sampling it every 10ms. Better let the MCU go to deep sleep mode (EM2) and wake up based on interrupt whenever that may occur.

I think you should be able to do some simple debounce in the ISR by using the RTCC counter value as timestamp. The BLE stack uses RTCC for timing reference so it's running in the background anyway.

For each GPIO edge, you can sample RTCC and calculate the delay compared to the previous edge. This way it should be easy to filter out glitches.

If I begin development with the Thunderboards, will I face a learning-curve and many software changes ? 

The Bluetooth stack API is basically the same, regardless of which kit or module you work with. So in that sense it does not really matter which kit you start with. The main difference is in the peripherals (Thunderboards have more sensors).

AFAIK the BRD4001A may is more powerful (and with more feature) than the one on BG22 thunderboard.

I don't think the larger main board (4001a) has that much more features. It does have the LCD but I find it often quite useless. Both the large and small devkits have on-board J-Link that can be used for debugging and also virtual COM port which can be really helpful.

There are some benefits of using the 4001a kit: 

1) it can be used as programmer / debugger for external devices, too (for example, when you build your first prototype you can use the kit to program it)

2) it can be used for some accurate sleep current measurements (AEM = Advanced Energy Monitoring). Can be quite handy when optimizing your own design

 

As conclusion, you can get started with BG22 thunderboard and purchasing an extra 6020b kit may be worth checking, too. In terms of learning the Bluetooth part, both are equally good. I have both type of kits available and it's easy to switch back and forth between these.

 

Did you press the generate button for the bootloader project? It needs to be run once after creating the project.

. How do you flash to board ?

In Simplicity Studio projects, there's a subfolder "binaries" where you can find the hex file. Right-click that file and select "flash to device" from the drop-down menu.

 

Then of course repeat the same for the application project (Thunderboard demo). 

. How do you flash to board ?

In Simplicity Studio projects, there's a subfolder "binaries" where you can find the hex file. Right-click that file and select "flash to device" from the drop-down menu.

Based on the API ref, those two commands are functionally equivalent. And same command applies to both client and server.

"2.6.1.17 cmd_gatt_set_max_mtu

Set the maximum size of ATT Message Transfer Units (MTU). Functionality is the same as gatt_server_set_max_mtu and this setting applies to both GATT client and server. "

 

As far as I know, this setting must be configured BEFORE the connection is established. I doubt that calling gatt_server_set_max_mtu after the connection is already established has any effect. (It's easy to test of course)

Sounds really odd. I would have guessed that deep sleep is enabled but you already checked that.

Is the expansion board (the one with joystick) connected? Try unplugging it and see if it helps.

Sure it can. The Silabs BLE stack (and all EFR32BG devices that run this stack) can support multiple master and slave connections at the same time.

See for example this link for more info on the topic:

https://docs.silabs.com/bluetooth/latest/general/connections/dualtopology

Generally, you can configure the number of simultanous connections (this affects the RAM footprint of the stack) and then you can open as many connections as master or slave, as long as the maximum number of connections is not exceeded.