nRF9160 - CFUN=0 - order of operations?

[Achim Kraus]

I searched the forum and found quite a lot of hits for CFUN, where CFUN=0 is faced delays.

My understanding of the other tickets is, that the EPS-detach sometimes takes longer, in my cases up to more than 60s.

CFUN=0 also saves the configuration to the NVM.

With the large times, what is executed first? The EPS-detach or the save to NVM? Or is the order independent?

(Just to mention: in my experience, some network provider seems to not answer to the EPS-detach while others do. That makes  EPS-detach with a that long timeout somehow a wast of energy.) 

[Hakon]

Hello,

I have asked the modem team. They should know this.

[Achim Kraus]

Any update?

Let me add:

I know the ommon explanation for the longer times to at+cfun=0 , but I feel strange about that.

If my modem is in sleeping mode and wakes up to send data, that works very frequently fast (from LTE-M 0.5 - NB-IoT up to 5s), but switching off takes in difference very often a lot longer.

From other "close/shutdown" definitions, one of the most common failure in the implementation is to try to implement "weak close" in a reliable way. e.g. if the implementation must anyway handle cases, where the "close" didn't work, the specification commonly uses a "weak close". That works usually by a simple close request and the other respond and closes. The respond may get lost, but a retry of the close request will not work, because the other has already closed. Not sure, if that is the root-cause. At least spending that long time in switching off seems to indicate, that there are issues.

[Sigurd]

Hi ,

Håkon is out of office, so I'm answering instead.

what is executed first? The EPS-detach or the save to NVM?

EPS-detach is performed before save to NVM.

[Achim Kraus]

Thanks!

Then the question, why that takes that long gets more and more important.

Hope I can spend some time next week in providing more captures, but for me it seems to be a significant difference between moving back to RRC connected state in order to send the next message or to deregister.