Fast connecting to a bonded device

[db_lw]

Hello all,

I have a BLE central application based on nRF Connect SDK V1.6.1 and a BLE peripheral application based on nRF5 SDK 17.1.0. The peripheral wakes up on interrupt and starts advertising. The central scans all the time for the peripheral and connects to it as soon as it's possible. The two devices use Coded PHY and are bonded with Just Works and LESC.

It is very important that the peripheral can notify the central as soon as possible with its data. I would like to minimize the connection establishment time so I already set the connection intervals to the lowest value of 7.5ms but the time the security exchange needs is still higher than I would like.

Is it normal that the security exchange process between two bonded devices takes around ~50ms?
Can this time be reduced further?

Best regards,
David

[Einar Thorsrud]

Hi David,

Just connecting and encrypting an already bonded link should be fast. Can you make a sniffer trace showing the packet exchange on air? Do you call bt_conn_set_security() immediately after the connection was established?

[db_lw]

I attached my sniffer trace. Maybe you can see something that's causing a delay.

Edit: Sniffer trace with key for decryption.

ble_lesc_conn_already_bonded_with_key.pcapng

[Einar Thorsrud]

Hi David,

The sniffer trace shows well what happens on air, and there are a few connection events between LL_ENC_REQ/RSP and LL_START_ENC_REQ/RSP. Unfortunately I was not able to conclude if it is possible to reduce the time more or not today, so I will have to get back to you early next week.

Edit: I discussed a bit with the BT stack team and they did not see much room for improvement, but may be that it could be slightly faster with 1 Mbps instead of coded phy. Could you test that just for comparison?

[db_lw]

Hi Einar,

thanks for your response. Ok, I guess then I have to reconsider using bonding for this application.

Einar Thorsrud said:

Could you test that just for comparison?

I probably won't do that because in this case the use of coded phy is mandatory.

Best regards,
David

[Einar Thorsrud]

Hi David,

I tested a similar setup on my end but with 1 Mbps and see the same. I will check more with the stack team, but it looks like it is difficult to speed this up.

[db_lw]

Ok, thank you for your help. I will then not use bonding at the moment.

[Einar Thorsrud]

I understand. That seems to be the only option for now.

In case other find this thread, it turns out that the Zephyr Bluetooth controller implementation now takes 4 connection events to secure the link, due to control procedure handling moved to lower priority Upper Link Layer (previously it was 3 connection events).