nrf52: after disconnect, device doesn't go into sleep again

[bearsh]

Description

BLE_HeartRate example, release profile, advertising interval changed to 250ms: during advertising, the current consumption is about 76uA. Once connected the controller keeps alive (I don't know if that's normal) and draws a little more then 3.3mA. But once disconnected, it still draws 3.3mA. Furthermore the spike while sending the advertisement packet is much higher (around 9mA compared to 6.5mA).

Target: nrf52
Toolchain: arm-none-eabi-gcc v7.3.1
Tools: mbed-cli v1.8.3
Mbed OS: c966348 (HEAD, tag: mbed-os-5.11.1, origin/mbed-os-5.11) Merge pull request #9208 from ARMmbed/release-candidate

1. 0s - Blue line (~0.9s): advertising
2. Blue line (~0.9s) - Red line (~4.6s): connected
3. Red line (~4.6s) - ...: advertising

Issue request type

• Question
• Enhancement
• Bug

[ciarmcom]

Internal Jira reference: https://jira.arm.com/browse/MBOCUSTRIA-394

[bearsh]

The same also happens if I set a advertising timeout. Once the timeout is reached, advertising stop but current consumption increases dramatically.

[TacoGrandeTX]

@bearsh Thanks for pointing this out and we noticed it ourselves while investigating #9093. We have shown that this is related to the CryptoCell 310 (CC310), and @pan- has an idea to resolve it. The higher current floor we observe related to using the CC310 (mentioned in #9093) needs to be re-evaluated.

[bearsh]

thanks, applying the workaround from #9093 (comment) reduces the current consumption to the desired value.

[TacoGrandeTX]

@bearsh Are you content and can we close the issue?

[bearsh]

@TacoGrandeTX I didn't do any testing on the latest release but once #9372 get merged the issue will be closed automatically, so I don't see any reason to close it. furthermore, as far as I got it, there's no proof that #9372 will really solve this issue, even with #9019 merged...