nRF9160 low average power problem

RE: nRF9160 low average power problem

Stian Røed Hafskjold — Thu, 13 Feb 2020 14:08:24 GMT

Hi, thanks for the feedback. I will bring this further to the modem team.

RE: nRF9160 low average power problem

MattiasE — Thu, 13 Feb 2020 13:32:42 GMT

Thank you!

I understand that the issues are related to the Telia network. It is strange that they have these parameters. Maybe you should make it more obvious that current consumption is greatly affected by network parameters. Perhaps you could implement some command to read out important parameters from the modem.

Anyway, we got the Telenor SIM-cards to work yesterday and we got much better values from them. These are idle currents, no data.
Cat-M no eDRX: 0.97mA
Cat-M eDRX 0001 (10.24s): 0.40mA
Cat-M eDRX 0010 (20.48s): 0.23mA

I have attached some images of the current profile, for other users information.

Cat-M no eDRX:

Cat-M eDRX 20.48s:

Cat-M eDRX 20.48 zoomed in on first "bump":

Cat-M eDRX 20.48 zoomed in on second "bump" (same as third):

So I think this concludes this issue. Thank you very much for your help!

Best regards

RE: nRF9160 low average power problem

Stian Røed Hafskjold — Wed, 12 Feb 2020 10:57:27 GMT

I got this answer from the protocol expert, regarding the three RX windows:

Telia network has DL gaps present in eNb configuration that causes the paging reception to be split into three "blocks" in current consumption measurements. In broadcast SIB2-NB:

pcch-Config-r13
defaultPagingCycle-r13: rf256 (1)
nB-r13: oneT (2)
npdcch-NumRepetitionPaging-r13: r256 (8)

Blue field means that one paging is repeated over 256 valid subframes. Timewise this is more than 256ms as there are non-valid subframes for NPDCCH in anchor carrier like synchronization signals and broadcast channels, actual reception is more than 350ms.

Another parameter set in SIB2-NB:

dl-Gap-r13
dl-GapThreshold-r13: n256 (3)
dl-GapPeriodicity-r13: sf256 (2)
dl-GapDurationCoeff-r13: oneHalf (3)

This shows that downlink gaps are used if reception is 256ms or more. In the Telia configuration DL gaps are in effect and the actual gap is half of the 256ms periodicity. This will make reception gaps to the paging monitoring, dividing the reception into three blocks in Telia configuration.

In the customers plot, the first two similar sized blocks are 128ms of reception separated by the configured DL gap. The third block which is a bit smaller is the rest of the valid subframes for paging monitoring also separated with the DL gap.

RE: nRF9160 low average power problem

Stian Røed Hafskjold — Thu, 06 Feb 2020 11:35:25 GMT

Sorry, forgot to mention, yes, the blue is NB and yellow is M1.

I agree that this is probably the reason for the higher current consumption you see. Not really sure why the paging window is so long. I will have to forward the question internally.

RE: nRF9160 low average power problem

MattiasE — Thu, 06 Feb 2020 09:12:16 GMT

Do I understand correctly that the yellow graph is CAT-M and the blue NB-IOT?
Your numbers look good. However, I'm not able to reproduce them.
Connected to Telia with NB-IOT eRDX 20.48s (+CEDRXRDP: 5,"0010","0010","0011") I get 2.47mA average current (idle current is ~60uA, so I get everything to sleep correctly). It is far from your numbers. See attached images.

When comparing my zoomed in paging event to yours, we can see that I get 3 high current periods (after the first bumpy) where you only get one. Do you know why that is so? I feel like this is probably one of the reasons I get so high current consumption. Is that something you/me can control or is that due to some network related setting?

We have received SIM-cards from Telenor now, but unfortunately we did not get them to work for some reason (we have tried with two different modem vendors).. So I cannot compare to other networks yet..
Best regards

RE: nRF9160 low average power problem

Stian Røed Hafskjold — Mon, 03 Feb 2020 14:52:18 GMT

Hi, sorry for the late response. I've been able to do some measurements on NB-IoT now. Below plots show 20.48 second eDRX cycle and PTW="0011", average current and zoomed in on one paging event.

[quote user="MattiasE"]It seams strange that NB-IOT should consume more power when no data is transferred? It can understand that this is the case when transferring a lot of data, but I don't think this should be that case during idle periods?[/quote]

As you can see the total charge for one DRX paging is about three times as much for NB-IoT compared to Cat-M1. This is because the radio has to be listening for a longer time for the paging event. So it's not using three times as much during "idle periods" but since the total average current is dominated by the radio current the average current will be significantly higher if the radio has to be in RX mode for a longer time.

[quote user="MattiasE"]What would be the reason to have 4xPTW if 1xPTW consumes less power? Is it because the modem gets several chances to receive data if signal is bad? In that case shouldn't it be adjusted dynamically (according to the signal) and not set to a static value? It seams a bit strange to me...[/quote]

The PTW and eDRX cycle values tells the base station when the UE is available. So the UE can't decide to not listen to a eDRX paging event on it's own, because it might miss a message from the base station. So therefore, The PTW is negotiated between the UE and the base station during connection. If you want to change PTW dynamically based on signal you will have to reconnect to the base station and renegotiate the PTW value. So that the base station knows when the UE is available.

RE: nRF9160 low average power problem

MattiasE — Fri, 24 Jan 2020 10:55:34 GMT

Thanks you for your reply.

It seams strange that NB-IOT should consume more power when no data is transferred? It can understand that this is the case when transferring a lot of data, but I don't think this should be that case during idle periods?

Your power numbers looks good and this is what we would like to achieve.

What would be the reason to have 4xPTW if 1xPTW consumes less power? Is it because the modem gets several chances to receive data if signal is bad? In that case shouldn't it be adjusted dynamically (according to the signal) and not set to a static value? It seams a bit strange to me...

The problem is likely the Telia network, as you say. We will try to get hold of other SIM-cards and play around with different settings and providers.

Kind regards.

RE: nRF9160 low average power problem

Stian Røed Hafskjold — Thu, 23 Jan 2020 14:47:31 GMT

NB-IoT consumes much more current than Cat-M. About 3x as much, because the bit rate is much slower so the radio has to be on for a longer time.

From the response you get from the network, +CEDRXRDP: 5,"0010","0010","0011", I can see that the paging window (PTW) is set to 10.24 seconds (0011), or in other words 4x one DRX cycle. Meaning that in each eDRX event you get 4x paging cycles, which you can see in the last plot you posted.

Try to set PTW to 0, in order to only get one paging cycle each eDRX event: AT%XPTW=5,"0000"

This should cut the current significantly. I don't have the possibility to measure on NB-IoT at the moment, but I'm attaching some plots for CAT-M1, to show you the expected numbers for eDRX cycle: "0010", and PTW: "0011" or "0000". Both with and without SIM card current:

20,48 eDRX, 4x PTW cycles, including SIM:

20,48 eDRX, 4x PTW cycles, excluding SIM:

20,48 eDRX, 1x PTW cycles, including SIM:

20,48 eDRX, 1x PTW cycles, excluding SIM:

[quote user="MattiasE"][/quote]

We have tried AT%XRAI=4 but no difference. (According to the spec you only support RAI for NB-IOT for now. Do you know when you will support RAI for Cat-M?)

According to the datasheet you are compliant with 3GPP LTE release 14 Cat-NB2. Do you know when you will support LTE release 14 of Cat-M?

I will have to come back to you on this.

RE: nRF9160 low average power problem

MattiasE — Wed, 22 Jan 2020 12:51:32 GMT

Hi, thanks for your reply

You are correct, we don't get eDRX on Cat-M. Okay, maybe Telia Cat-M network is not fully functional/configured yet? We are working on getting Telenor SIM-cards.

However, I was a bit unclear. We do get eDRX on Telia NB-IOT (using Telia SIM). Have you heard anything about Telia NB-IOT net in Sweden?

We get this response when using NB-IOT:

+CEDRXRDP: 5,"0010","0010","0011"

Even though we get eDRX on NB-IOT we get very high average current. Below is a image of the current consumption when connected with NB-IOT with eDRX "0010" and no data transfer. As noted before, average current is about 1.2mA, far too high. It should not be this high, right?

I have attached a trace when connecting to the network and being idle for some time.devzone.nordicsemi.com/.../trace_2D00_2020_2D00_01_2D00_22T12_2D00_43_2D00_19.427Z.bin

We have tried AT%XRAI=4 but no difference. (According to the spec you only support RAI for NB-IOT for now. Do you know when you will support RAI for Cat-M?)

According to the datasheet you are compliant with 3GPP LTE release 14 Cat-NB2. Do you know when you will support LTE release 14 of Cat-M?

Kind regards

RE: nRF9160 low average power problem

Stian Røed Hafskjold — Wed, 22 Jan 2020 09:07:13 GMT

Hi, seems like you are not getting into eDRX idle mode. If you enable eDRX through the AT command +CEDRXS in mode 2, you will get notifications from the network saying which parameters you got. You can also read the parameters directly using +CEDRXRDP. What is the network reporting back?

We've gotten reports from other customers saying that the Telia network in Sweden is not really supporting eDRX. You should contact Telia and ask. You can also do a modem trace and post the results here. Then we can analyze and see what is going on. Here's a guide: https://devzone.nordicsemi.com/nordic/nordic-blog/b/blog/posts/how-to-get-modem-trace-using-trace-collector-in-nrf-connect

In Sweden the iBasis SIM card is only working with Telia, and it's in roaming mode. Telia might have chosen to support less parameters in roaming mode. Like PSM mode is probably not supported.

I think the best thing for you now is to get a native Telenor SIM card to do the testing.

RE: nRF9160 low average power problem

MattiasE — Mon, 20 Jan 2020 16:29:44 GMT

We have managed to capture the current profile of the device when it is connected to Cat-M network, see image below. Maybe this is useful. This pattern is repeated every 640ms. The upper plot is the current consumption and the lower plot is the COEX2 signal.

We really need to get lower current consumption than this.

Kind regards

Mattias