srsRAN gNB with srsUE over usrp b210 frontends

[thaxir]

Issue Description

I'm following the recent tutorial 'srsRAN gNB with srsUE' from here https://docs.srsran.com/projects/project/en/latest/tutorials/source/srsUE/source/index.html. However, the UE (srsue) never obtains an IP from srsgnb staying at this :

RF status: O=3, U=0, L=0
Scheduling request failed: releasing RRC connection...
Random Access Transmission: prach_occasion=0, preamble_index=0, ra-rnti=0x39, tti=1934

Setup Details

e.g. Network configuration, Operation System, Hardware, RF front-end, library and driver versions

• 2 USRP b210 as RF frontend
• 2 NUCs
• Open5gs
• srsRAN_Project for gnb
• srsRAN_4G for the UE
• Internal clock/no GPSDO
• UHD_3.15. in both nucs

Expected Behavior

Cannot use srsue since it is not properly connected to the gnb

Actual Behaviour

[What happens instead e.g. error message]
Seems like UE attempted a scheduling request but didn't receive a response from the gNB. Not sure though

Steps to reproduce the problem

Additional Information

[Any additional information, configuration or data that might be necessary to reproduce the issue]
gnb.log.txt
ue.log.txt

ue.conf.txt

gnb-config.yml.txt

[pgawlowicz]

UL signal is too strong at the gnb. Please try to tune down srsUE tx_gain and gnb rx_gain.

[thaxir]

Hi @pgawlowicz thank you for the reply.
I have experimented with a lot of changes in both srsUE and gnb as you said but could have any significant result. The only thing i managed to see traffic in the UL was with the tx gain 75 rx gain 35 (see screeshot below) .Other that these i had received one prb with txgain/rxgain at 48/38 but that it only one. The two usrps are very close to each other like 10 cm .

srsue :

Attached are also the latest logs of gnb and srsue and the confs.
ue.log.2.txt

gnb.log.2.txt

gnb.conf.txt
srsue.conf.txt

[thaxir]

any suggestions on this topic ??

[pgawlowicz]

could you try to apply the patch from this comment to srsUE, recompile and try again?
#826 (comment)

[thaxir]

@pgawlowicz Piotr we have done the changes you suggested .
After restarting serveral times srsue and gnb the srsue logs are changed as you see in the snapshot below

Gnb reports this:

and amf in open5gs which is in the same machine with gnb reports this :

which seems that amf rejects the srsue . Do you know why? (

And another question is why do i need to restart several times both srsue and gnb in order to have this connection open and not the Scheduling request failed : releasing RRC connection?
I attach also the logs for ue and gnb.
ue.log.3.txt
gnb.log.3.txt

gnb.log.3.txt

[thaxir]

Any suggestions on this?

[pgawlowicz]

it seems the problem is on the core network side, did you add the UE to the subscriber database?

[thaxir]

Hi @pgawlowicz again . So far i have tried to experiment with the position of the usrps if the radio link and the line of site was the error but i am getting on the srsue

On the gnb

I was able to get some movement over the amf

i started cross checking the database i see the same values in the srsue and open5gs (deleting and re-adding the subscriber)

Searching for others with this issue i came across this #451 which seems to have a lot of similarities with my case. However, on the newer version of srsran gnb i don't know where to put the inactivity_timer: 7200

Does this inactivity timer can be added somewhere in the gnb ?

[pgawlowicz]

could you share gnb logs in debug mode?

[thaxir]

@pgawlowicz yes here
i don t know if i have captured all because there are a lot of msgs.
gnb-debug.txt

[pgawlowicz]

it seems those are srsUE logs.

[thaxir]

Apologies,
here
gnb-debug.txt

[pgawlowicz]

Could you tell me which code version (commit id) do you use? it is missing in the logs.

[thaxir]

How do i see this?

[pgawlowicz]

git log -n1

[thaxir]

ec29b0c1ff79cebcbe66caa6d6b90778261c42b8

[pgawlowicz]

ok, this is very old version, could you update and try again?

[thaxir]

Should i use binaries from Package Manager on ubuntu or build it from the source?

[pgawlowicz]

both ways work

[thaxir]

@pgawlowicz i just installed srsran-project (apt-get install) from the package manager.
nuc6@nuc6:/usr/share/srsran$ gnb --version
srsRAN 5G gNB version 23.10.1 ()

Should i proceed to testing ?

[pgawlowicz]

hmm, it that case please build from the source code.

[thaxir]

Ok done

No i get this

It seems like a config problem
Here is the config of gnb

[pgawlowicz]

You have to install uhd driver if you are building from source. Please follow our installation guide https://docs.srsran.com/projects/project/en/latest/user_manuals/source/installation.html

[thaxir]

I have already installed the uhd from the previous run

Do i need to redo this installation?

[pgawlowicz]

could you check if cmake detects the UHD correctly?

[thaxir]

No it is not detected so i build 3.15 again . I will send you the update tomorrow. Thank you very much this far for the help

[thaxir]

Hi @pgawlowicz again, so the update : I have installed uhd 3.15 .
Gnb :

Here the logs
gnb-debug(2).txt
Gnb config:
gnb.yml.txt

The srsue

[pgawlowicz]

and the srsUE config and logs?

[thaxir]

Ue logs:
ue.txt.zip
Ue conf:
ue-conf.txt

[pgawlowicz]

PUSCH SNR is very low, probably caused by a very high CFO. Do you have an external clock to connect to b210s?

[thaxir]

No i don t have ext clock. I could order one but it will take time . I don t know if i could connect the two usrps through coaxial cable for using on usrp as master timer and one as slave .
Is there any configuration i can do now without a gpsdo?

[pgawlowicz]

you might try to tune TX an RX gains for both gnb and srsUE, also time_adv_nsamples in srsUE config. Some users are able to run this setup without an external clocks. Did you check if it works correctly over zmq?

[thaxir]

Yes in the past i did use zmq before shifting to ueransim. For the time being i need to have radio link between ue and gnb. Do you have any reference values for start playing with the gnb /ue /time_adv ?
If i replace srsue with a cots ue would this still be a problem?

[pgawlowicz]

Usually, tx_gain: 60 and rx_gain:30-40 and 'time_adv = auto or 300'. Did you apply the srsUE patch that I shared in one of the previous posts?
I think b210 should work correctly with COTS. You might give it a try.

[thaxir]

Did you apply the srsUE patch that I shared in one of the previous posts?
Yes the ue is patched.