[srslte-users] [Commercial UE] Impossible to see the created LTE network

Fri Jun 29 08:34:05 UTC 2018

Hi Vincent,

you provide a neat checklist for debugging srsLTE setups.

With recent Ubuntu 18.04 we observe much more late/overflow/underflow etc. 
compared to 17.10 or 16.04, which is just what you experience with Arch. I 
cannot tell if this is "a lot" or not.

I just want to point out that you'd test some other smartphone than the S6. I 
couldn't get a (branded) S6 to see the network at all. Same for iPhone 8 and 
X. Rooting shouldn't make much difference. Maybe some smartphones reject the 
Test Network PLMN (MCC 01 MNC 001), but not to my experience.

Merlin

Am Mittwoch, 27. Juni 2018, 16:05:13 CEST schrieb Falconieri  Vincent:
> Dear srsLTE community,
> 
> Since about one week, I'm trying to have a working setup (ENC + ENB +
> Commercial UE) : I want the UE to be able to connect to the ENB.
> Unfortunately, I can't get it working, even with reading most of the old
> SRSLTE_list e-mails, documentation provided and information found online.
> Sorry for the big e-mail, I tried to solved it by myself in vain, and tried
> to compile all available information once, here.
> 
> ## Setups
> I tried 2 setups, both on the same machine (Laptop ASUS FX550V, i5-6300HQ
> 3.2 Ghz + 8Gb RAM) 1. Arch-linux (performance mode for CPU, packets up to
> date) :
>     + srsENC (compiled manually)
>     + srsENB (compiled manually)
>     + UHD (compiled manually and also from AUR repository. Right now : 
> UHD_3.11.1.0-0-unknown) + USRP B210 (default firmware downloaded via the
> python script provided, connected on USB 3.0) + 2 VERT900 antennas
> (connected on A:RX2 and A:RX/TX, 824 to 960 MHz, 1710 to 1990 MHz from
> documentation) + Samsung Galaxy S6 Edge+ (SM-G928F, not rooted)
> 2. Ubuntu 17.04 (performance mode for CPU, packets up to date, installed on
> a USB Key 3.0) : + srsENC (compiled manually)
>     + srsENB (compiled manually)
>     + UHD (from repo (v3.10 I guess) and also compiled manually with version
> UHD 3.12.0.HEAD-0-gec786351) + USRP B210 (default firmware downloaded via
> uhd_images_downloader, connected on USB 3.0) + 2 VERT900 antennas
> (connected on A:RX2 and A:RX/TX, 824 to 960 MHz, 1710 to 1990 MHz from
> documentation) + Samsung Galaxy S6 Edge+ (SM-G928F, not rooted)
> 
> ## Problems
> I can't see the created LTE network by the setup from the smartphone
> (Parameters > Mobile networks > Network operators > Networks lookup) even
> if the setup seems to work. I also tried to get more precise information
> with the help of apps like "Network Signal Guru", "G-NetTrack Lite", "Cell
> Signal Monitor" but without a rooted phone, it displays only information on
> the connected cell. (As I can't even detect mine, I can't get any
> information from these apps). The smartphone can see commercial network
> without problem. (At least 3 commercial networks listed outside Faraday
> cage)
> 
> ## Work
> srsENC : runs flawlessly. No errors in logs.
> 
> srsENB : connects correctly to srsENC.
> * On setup 1. (=Arch) : A lot of UOSDL (even in performance mode),
> The last errors I get, which I never resoled is :
>     [ERROR] [STREAMER] recv packet demuxer unexpected sid 0x190038
>     [ERROR] [STREAMER] recv packet demuxer unexpected sid 0xfffe000e
> 
> * On setup 2. (=Ubuntu) : nearly no UOSDL. Still some entry in enb.log :
>     12:57:55.434535 [RF  ] [W] Overflow
> No other errors displayed. No tentative of connection from UE in the logs.
> 
> * On both setups : Tried with out-of-the-box conf files, tried to increase
> gain (Tx 86, Rx 60), tried other earfcn (1800, 1949, 1600, ...), recompile
> srsENB with and without srsGUI, run with different version/source of uhd
> ...  without improvement. Enable pregenerate_signals allowed to get less
> UOSDL. Add some devices_args ("recv_frame_size=1024", ...) allowed to get
> new errors message, but not a working setup.
> 
> UHD Driver : "make test" only provides 1 error. "./tests_messages" outputs
> some errors : (In faraday cage with sudo)
>     Test Burst ACK    ->     9 successes,   2 failures
>     Test Underflow    ->     7 successes,   5 failures
>     Test Time Error   ->     4 successes,   7 failures
>     Test Late Command   ->    16 successes,   0 failures
> "./benchmark_rate --tx_rate 6.25e6 --rx_rate 6.25e6 --duration 30" gives no
> error on USB 3.0. Correctly detects the B210.
> Can receive data from the USRP (./rx_ascii_art_dft --rate 1000000 --freq
> 1700 --rx-gain 60 gives a "moving graph")
> 
> USRP B210 Card : LEDs work correctly (blink of the LED in case of UOSDL),
> used by someone else, it worked for another project but not on these
> frequencies and not for this usage.
> 
> Laptop : working for diverse other usages. Never used it for "real time
> hardware" applications (so USB 3.0 ports never tested in equivalent
> settings) Smartphone : working for daily usage. Roaming is working. 4G LTE
> verified. More details on frequencies this UE can handle :
> https://www.frequencycheck.com/models/vDm7R/samsung-sm-g928f-galaxy-s6-edge
> -lte-a-64gb-samsung-zen
> 
> ## Aside notes
> * All tests are conducted in a Faraday cage.
> * I may have access to another smartphone (that I can root) and might get
> another computer by the end of the week for further testing. * I have no
> GPSDO on the B210. It seems that some people (even in this list) managed to
> make it work without this. Am I making a mistake ? Is it mandatory ? *
> Monitoring the CPU gives 30% charge when EPC/ENB running, 10% without.
> Numerous L(ate),D(rop),O(verflow),U(ndeflow) in powersaving mode, nearly
> none in performance mode (but still some. 0 to 2 per minute in the best run
> I ever had). * USB 2.0 on the same machine provides a lot of UOSDL without
> improvement. * I have a configurable sim card (sysmocom) that I keep for
> later steps. Not used now. * Antennas are placed orthogonally from each
> other.
> * With the notes I took, and what I tested, I think I have enough to start a
> wiki .. we could talk about it when my problem will be solved, if you are
> interested. :)
> 
> 
> ## Questions
> * Some gave the advice to reduce the bandwidth to 5 Mhz to prevent timing
> issues. Do we agree we can't control this parameter directly, but only
> through the number of ressource block ? (tried n_prb with almost all values
> displayed in the conf.example : 6,15,50, ...) * Do you know if
> mcc/mnc/tac/... should have a particular value to work with the commercial
> UE ? As far as I understand, it should be fine with default values, isn't
> it ? It should behave like roaming. * Do you have an idea of what can get
> wrong ?
> 
> 
> ## Useful informations :
> Some captures and complementary informations are joined to this e-mail. Feel
> free to ask anything I can try that could be relevant. - Error with cache :
> with "pregenerate_signals" there is less UDLO, but still. -
> Error_No_network_visible (2) : One run (as all the other, but this one has
> nothing else "special") where the smartphone can't see the network -
> Error_Recv_packet_demuxer (2) : Runs where a major error occured. It seems
> that unpluging and repluging the card (so the firmware is copied again)
> solve the issue. Seems it happens when there is 2 runs in a row, with
> different enb.conf. I unfortunately did not manage to get a "nice and
> smooth" run today. I don't know why. If I managed to get one, I'll give you
> the logs.
> 
> Thanks for your time,
> Hope we'll be able to find a solution or at least ideas to explore
> Best regards,
> F. Vincent