CQI is not varying in my simulation

130 views
Skip to first unread message

Maria Eduarda Veras Martins

unread,
Sep 16, 2024, 2:46:05 PM9/16/24
to 5G-LENA-users
I added mobility to my simulation intending to vary the CQI of my UEs, but when I added a callback to NrGnbPhy, I noticed that the CQI always has the same value of 15. I created the callback and use it with the following address:

Config::Connect("/NodeList//DeviceList//$ns3::NrGnbNetDevice/BandwidthPartMap/*/NrGnbPhy/ReportCqiValues", MakeCallback(&NotifyCqiReport));

how I added mobility:
Screenshot from 2024-09-16 15-40-53.png

 and where I add the callback
Screenshot from 2024-09-16 15-42-05.png

Can anyone tell me how I can vary the CQI or has anyone experienced a similar issue before?

Thank you,
Maria Eduarda.
logs.out

Kent Huns

unread,
Sep 16, 2024, 11:20:37 PM9/16/24
to Maria Eduarda Veras Martins, 5G-LENA-users

15 is a maximum value of CQI. Check if Rxpsd or Txpower is too strong.

Best regards,
Kent


2024年9月17日(火) 3:46 Maria Eduarda Veras Martins <me...@cin.ufpe.br>:
--
You received this message because you are subscribed to the Google Groups "5G-LENA-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 5g-lena-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/5g-lena-users/82810d60-87e7-4b86-a937-f8fe77273e75n%40googlegroups.com.

Katerina Koutlia

unread,
Sep 17, 2024, 5:14:34 AM9/17/24
to 5G-LENA-users
Hi,

You should also add channel variations, because if you have always LoS and the distance from the gNB is not varied significantly, most probably you will have good channel conditions.
Try something like:
CcBwpCreator::SimpleOperationBandConf bandConf(centralFrequencyBand,
bandwidthBand,
numCcPerBand,
BandwidthPartInfo::UMa);

Config::SetDefault("ns3::ThreeGppChannelModel::UpdatePeriod",
TimeValue(MilliSeconds(100))); // 100ms
nrHelper->SetChannelConditionModelAttribute("UpdatePeriod", TimeValue(MilliSeconds(100)));
and you can even nclude shadowing:

nrHelper->SetPathlossAttribute("ShadowingEnabled", BooleanValue(true));

BR,
Kat

Maria Eduarda Veras Martins

unread,
Sep 17, 2024, 2:59:44 PM9/17/24
to 5G-LENA-users
Thank you for the responses, everyone. 
I followed your suggestion and also placed the UEs further away from the gNB, which indeed caused the CQI to change. However, when I move the UEs farther away, my flows are left with only one packet. 
How can I solve this issue?

Here is the output of the FlowMonitor:
Screenshot from 2024-09-17 15-57-57.png

Best regards,
Maria Eduarda.
code.cc

Katerina Koutlia

unread,
Sep 18, 2024, 5:48:55 AM9/18/24
to 5G-LENA-users
Check if your simulation time is too short, if Ues are moved very far away from the gNB, if the packets you send are too few...

BR,
Kat

Kent Huns

unread,
Sep 18, 2024, 1:31:02 PM9/18/24
to 5G-LENA-users
Maybe the packets are too few. Need to set up periodic packet generation like example/cttc-3gpp-indoor-calibration.cc.

Best regards,
Kent
notitle.jpg

Maria Eduarda Veras Martins

unread,
Sep 18, 2024, 1:57:56 PM9/18/24
to 5G-LENA-users
Hi Kent!
Since the bulkSendApplication does not have this attribute (interval) , I will try to implement it and see if it works.
Thank you!

Best Regards,
Maria.

Maria Eduarda Veras Martins

unread,
Sep 20, 2024, 9:49:31 AM9/20/24
to 5G-LENA-users
Hi guys...
I changed the application to OnOff and added SetConstantRate for 100Mbps. My simulation time is 20s, but I am still receiving few packets. I am not sure where else to modify to solve this issue. When I place the UEs near the gNB, everything works normally, but I need the UEs to be far away to vary the CQI.
Can anyone provide some guidance on how I can resolve this?

I have attached my code with the OnOff application and the pcap file from one UE.

Best regards,
Maria
debugUe-n1-i1.pcap
code.cc

Maria Eduarda Veras Martins

unread,
Sep 20, 2024, 11:10:22 AM9/20/24
to 5G-LENA-users
I was able to increase the number of packets by changing the txPower!
I am not sure about the CQI; is it normal for it to vary so much with the UEs stationary? I found the behavior to be strange...

working.out

Kent Huns

unread,
Sep 20, 2024, 11:52:54 AM9/20/24
to 5G-LENA-users
It sounds like you're doing well.
Now you are implicitly setting 3GPP channel model in your code, 
and it uses random number to decide channel environment (=the angles of multipath propagation and the delay spread).

The channel is changed every 100 ms, so the CQI changes drastically.
Config::SetDefault("ns3::ThreeGppChannelModel::UpdatePeriod",
TimeValue(MilliSeconds(100))); // 100ms
nrHelper->SetChannelConditionModelAttribute("UpdatePeriod", TimeValue(MilliSeconds(100)));
I recommend setting UpdatePeriod = 0 (not update) and ShadowingEnabled = true, if make the CQI fluctuations even smaller
nrHelper->SetPathlossAttribute("ShadowingEnabled", BooleanValue(true));

Maria Eduarda Veras Martins

unread,
Sep 20, 2024, 12:23:28 PM9/20/24
to 5G-LENA-users
When commenting out those lines, the CQI indeed remained constant throughout the simulation (which I expected since the UEs are stationary). It also increased the number of packets a little bit, haha. One thing I noticed is that the flow using TCP Cubic sends a lot more packets than the DCTCP flow. Is there any reason for this?
flow 1 is cubic and flow 2 is dctcp:
Screenshot from 2024-09-20 13-22-58.png

Kent Huns

unread,
Sep 20, 2024, 2:10:21 PM9/20/24
to 5G-LENA-users
If there still be a CQI gap between UE1&UE2, maybe they are in different LOS/NLOS condition.
It's a rough way but you can fix it everyone-LOS (pLos=1/pNlos=0) or everyone-NLOS(pLos=0/pNlos=1) by hard coding. 
src/propagation/model/channel-condition-model.cc
Ptr<ChannelCondition>
ThreeGppChannelConditionModel::ComputeChannelCondition(Ptr<const MobilityModel> a,
                                                       Ptr<const MobilityModel> b) const
{
    // compute the LOS probability
    double pLos = ComputePlos(a, b);
    double pNlos = ComputePnlos(a, b);

If CQIs are nearly equal, maybe it's the problem of fairness that MAC scheduler should ensure.
Please refer to my past comments about PF algorithm. It may be better to stop resetting the average throughput.
(And I'd be happy if you check and report the behavior of TdmaRR. I guess UE2's throughput goes to 0.)

Mubanga Nsofu

unread,
Sep 20, 2024, 2:24:06 PM9/20/24
to Maria Eduarda Veras Martins, 5G-LENA-users
I had similar issues in LENA a few years ago and the issue was never resolved 

Maria Eduarda Veras Martins

unread,
Sep 20, 2024, 5:13:39 PM9/20/24
to Kent Huns, 5G-LENA-users
The cqi of UE1 is the same as that of UE2 but they are the same distance from the GNB so I thought it was right... I didn't quite understand, this has to do with the smaller amount of packets with tcp dctcp?

I started my research recently, I'm still learning some concepts…

Reply all
Reply to author
Forward
0 new messages