BLE_STACK_BUSY error more frequent over time

[Rambo919]

I have a BLE application running Mbed 9.7.2 on an nRF52832 (softdevice 132). I have a connection interval (to the central) of 11.25 ms and specify a slave latency of 3. I have a custom service that houses 3 custom characteristics (notify). Each characteristic is 20 bytes in size and consists of 8 2-byte values and a single 4-byte value. The application will fill a single characteristic with new data every 10 ms; when all 3 characteristics are updated I call BLE::Instance().gattServer().write() to write the new data over BLE during the next connection interval.

My issue is as follows. When I attempt BLE::Instance().gattServer().write() I log the result. When the application first starts the result of this operation is BLE_ERROR_NONE - signifying that all is well. As time progresses the result of BLE::Instance().gattServer().write() more often becomes BLE_STACK_BUSY.

Any insights as to this are very welcome.

[Jørgen Holmefjord]

Hi,

It sounds like you are filling the buffers faster than the stack is able to transfer the data. If you send 20 bytes every 10 ms, and your connection interval is 11.25 ms, you will eventually fill the buffers if no data length extension (DLE) is enabled. Have you tried updating the characteristics at a slower interval, or reduce the connection interval further?

This is a question you should direct to the mbed forums, as this is not code being developed/maintained by Nordic.

Best regards,
Jørgen

[Rambo919]

Hi Jørgen, 

Thanks for your reply. I have also posted on the mbed forums but I think that support is much better here so I thought I'd try in both places. 

Am I mistaken in thinking that it's possible to send more than 20 bytes per connection interval? I can't remember exactly but I remember calculating a theoretical maximum throughput of ~ 5KB/s which allowed about 60 bytes per CI. 

Thanks for your thoughts, this has given much to think about and it sounds like you're pushing me in the right direction. 

Thanks,

Adam