nRF9160 NB-IoT Cell Selection

[Macro Yau]

Hi Nordic Team,

I have deployed dozens of nRF9160-based devices in urban environments, where some of the devices are within coverage of multiple NB-IoT cells. My devices have the nRF9160, running the Serial LTE Modem (SLM) firmware, connected to a host MCU and communicated using AT commands. The nRF9160 is powered up once an hour to transmit some data over HTTPS. It also logs the RSRP and RSRQ measurements (via AT+CESQ), as well as the ID of the connected cell (via AT+CEREG). After completing (or aborting) the data transmission and telemetry logging, the nRF9160 is completely powered off (power-gated using a MOSFET switch).

When I analyzed such telemetry data, I noticed that quite a number of devices (all fixed in the same location over time) tend to jump between different cells in each time the nRF9160 is powered on. I am interested in learning more about how the nRF9160 selects a cell to camp on, for further optimizing my device's energy usage and reducing transmission failure due to poor connectivity:

1. I understand 3GPP TS 36.304 (Section 5.2) stipulates that initial cell selection depends on the network operator's pre-defined RSRP and RSRQ thresholds. As your previous field-testing blog post mentioned that "RSRP is used when doing cell selection/reselection and handover," I am curious whether RSRQ is taken into account in the cell selection process?
2. In case the operator-defined thresholds are sufficiently low (say, -140 dBm / -23 dB as I intercepted from the SIB1-NB message via modem trace capture), such that all cells in vicinity satisfy the 3GPP-stated criteria, would the nRF9160 just pick the first detected cell, or is there any mechanism to pick the "best" cell to camp on?
3. I am aware of some previous forum posts mentioned that commanding AT+CFUN=0 would have the nRF9160 saved the current cell information to the NVM for faster connection in the future. I tried turning off the modem with and without this command, but in either case, the modem could switch to different cells between power cycles instead of staying on the previously-connected cell, with the connection time more or less in the same range. I am wondering if this is the expected behavior? (Note: I define "connection time" as the time spent between powering on the nRF9160 and obtaining an IP address after completing registration; typically it takes some 10 to 60 seconds in my test and deployment environments.)

My nRF9160 hardware and software configurations:

• SiP: nRF9160-SIBA (with a mix of B0 and B1A revisions)
• Modem firmware version: mfw_nrf9160_1.2.3
• SLM application version: 1.4
• PSM and eDRX disabled
• Band-locked to LTE Band 8

Thank you very much!

[helsing]

Hi Marco,

Thank you for your questions. I will check with the team an get back to you later this week.

Best regards,

Håkon

[Chraebs]

helsing did you get some information about this. It is an interesting topic.

[Charlie]

Hi Macro and Matthias,

Sorry for the late reply. My colleague has a sick leave and I will provide you with a reply soon.

Best regards,

Charlie

[Charlie]

Hi,

The nRF9160 is powered up once an hour to transmit some data over HTTPS. It also logs the RSRP and RSRQ measurements (via AT+CESQ), as well as the ID of the connected cell (via AT+CEREG). After completing (or aborting) the data transmission and telemetry logging, the nRF9160 is completely powered off (power-gated using a MOSFET switch).

I doubt if this is a good design. The power-off does save power, but keep in mind the process of power-on will also consume more power than the normal running period, especially when you enable PSM for your application. Searching and connecting with cellular stations are power-hungry. A one-hour interval is quite short for a cellular IoT application, I do not think the power-off period can compensate for the power-on process power consumption. I suggest you do some experiments to compare their difference. I expect only when the interval is serval days or longer then it is worth trying your power-off method, otherwise you should keep the power on and use PSM for your application.

I am curious whether RSRQ is taken into account in the cell selection process?

would the nRF9160 just pick the first detected cell, or is there any mechanism to pick the "best" cell to camp on?

According to our cellular modem expert, cell selection is based on the S criteria following 3GPP TS 36.304 specification, so RSRQ should be used as 

When I analyzed such telemetry data, I noticed that quite a number of devices (all fixed in the same location over time) tend to jump between different cells in each time the nRF9160 is powered on.

I am wondering if this is the expected behavior?

I am not sure why this happens, maybe you are on the edge of several cells. It would be better to get some information about the neighbor cells for further analysis, you can refer to the discussion on this ticket: nrf9160 AT%NCELLMEAS reports only 1 cell and send back one modem trace with AT+NCELLMEAS experiments. By the way,10 to 60 seconds of searching time sounds like a normal result. 

Best regards,

Charlie