Variation in the average SINR Values Across Different MCS Indexes

[Sangwon Seo]

Hi, 

I am conducting an experiment based on "cttc-nr-demo.cc," where I have fixed the MCS index for various simulation scenarios while keeping other parameters, such as transmission power, constant.

However, I have observed different average SINR values across the MCS indexes. Specifically, the CDF of SINR values (attached image, got from RxPacketTrace.txt) shows that for MCS index 25, the average SINR is 12.8, whereas for MCS index 20, it is 11.1.

I would like to understand this phenomenon in more depth. I suspect that the variation might be related to the different optimized beta values used in the EESM. Could you confirm if this is the case, or provide some insights into the underlying reasons?

Thank you for your time and assistance.

Best regards,
Sangwon

[Tom Henderson]

I don't know for sure, but I can speculate one possible cause to
investigate, which is that as you lower the MCS, you decrease the coding
efficiency, which leads to needing to use more bandwidth (PRBs) to send
the same transport block size, and hence increases the noise power. So
I would recommend to check that the bandwidth used for MCS 20 is the
same as that for MCS 25 (I don't know offhand how to check that; please
look at the documentation).

- Tom

On 10/22/24 06:23, Sangwon Seo wrote:
> Hi,
>
> I am conducting an experiment based on "cttc-nr-demo.cc," where I have
> fixed the MCS index for various simulation scenarios while keeping other
> parameters, such as transmission power, constant.
>
> However, I have observed different average SINR values across the MCS
> indexes. Specifically, the CDF of SINR values (attached image, got from
> RxPacketTrace.txt) shows that for MCS index 25, the average SINR is
> 12.8, whereas for MCS index 20, it is 11.1.
>

[Kent Huns]

I thought the randomness of propagation loss might be one of the causes.
To be precise, shadowing & multipath fading of 3GPP channel model.

Could you change the seed value by RngSeedManager::SetSeed()? And would the results still be similar trend?
https://www.nsnam.org/docs/release/3.28/doxygen/classns3_1_1_rng_seed_manager.html

[Tom Henderson]

I agree that this is another possible cause.

We recommend changing the run number using the '--RngRun=2' argument
(using '2' as an example value here), or else via environment variable:

NS_GLOBAL_VALUE="RngRun=2" ./ns3 run ...

You can change either run or seed number to get different random
numbers, but the author of the RNG generator recommends to change the
run number and leave the seed unchanged, to keep the random variables
uncorrelated with one another.

- Tom

[Sangwon Seo]

Thank you very much for your assistance. I have completed the scenarios for 'RngRun=2' and 'RngRun=11'. The attached plots show the results, with the left figure representing RngRun=2 and the right one representing RngRun=11.

It appears that there are differences in the average SINR values across the random streams. However, the overall trend remains consistent.
Best regards,

Sangwon

2024년 10월 22일 화요일 오후 7시 24분 19초 UTC+2에 Tom Henderson님이 작성:

[Kent Huns]

Thank you for sharing the result. It feels strange but interesting.

Then, I guess we would better to check Tom's first comment.

Please check the correlation between "SINR" and "tbSize" listed in Rxpackettrace.txt.

You may get something if this prediction is correct.

* If you've not modified cttc-nr-demo.cc so much, the setting of MAC scheduler is now TdmaRR.

   In TDMA, all PDSCH resources in a slot(all PRBs and all symbols) are allocated for 1UE.

   So allocated PRBs are the same in each simulation.  And allocated symbols are "nSymbol" in Rxpackettrace.txt.

[Katerina Koutlia]

Hi all,

Could you check if your scenario is always LoS or not and if you have channel updates?
Also, maybe the fact that there is no interference from other gnbs also plays some role. What if you try with 2 gnbs working in the same freq?

BR,
Kat

--
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/929a9609-f778-4b96-8665-323dccdc1a65n%40googlegroups.com.

[Sangwon Seo]

Hi all, 

Thank you! The original packet size is 100 Bytes.  TB size for MCS 25 and 20 are 168 and 173 in the simulation, respectively. Here, I had only used UL and DL (not flexible slot because it makes different TB sizes).  and a single gNB and 5 UEs were activated. 

To remove all interference, I conducted a simulation with 1 gNB and 1 UE case (Please refer to the plot below). 

It shows a 1.8 dB difference between MCS 25 and MCS 20...

BR,

Sangwon

2024년 10월 23일 수요일 오후 7시 33분 37초 UTC+2에 Katerina Koutlia님이 작성:

[Kent Huns]

Sorry for the silly question, but did you remove NACKed data("corrupt"=1) when you made these plots?

If you are checking link curve now, they are perfectly reasonable results. High MCS needs high SINR. (Please add the results of MCS=15,10,5)

https://www.researchgate.net/figure/BLER-curves-under-different-MCS-values-for-AWGN-channel_fig2_331343403

[Sangwon Seo]

Thank you for the mail but it is included all the SINR values regardless corrupted or not…

BR,

Sangwon

You received this message because you are subscribed to a topic in the Google Groups "5G-LENA-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/5g-lena-users/0rZB1wMXMA4/unsubscribe.
To unsubscribe from this group and all its topics, send an email to 5g-lena-user...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/5g-lena-users/cf7702fe-d06a-4c1e-86ab-619239a1d87en%40googlegroups.com.

[Sangwon Seo]

Thank you very much for sharing the link to the BLER curves. I appreciate your assistance. Attached, you will find a plot showing the SINR distribution across MCS indexes 25, 20, 15, 10, and 5, based on measured SINR values that encompass all transmission trials.

While the paper you provided offers insights on BLER curves in the AWGN channel, it does not seem to address the variability in SINR values (with the same transmission power) observed across different MCS indexes. 

BR,

Sangwon

2024년 10월 24일 (목) 오후 5:46, Sangwon Seo <wwtkd...@gmail.com>님이 작성:

[Kent Huns]

Thank you for the additional work. Yes, the image I attached was just an example of link budget curve.

On the other hand, what we are watching is received SINR including NACK as you said so.

I guess it's a great hint that the plots are equal when their MCS is low. (now I'm ignoring the part of the CDF below 10^-2)

Received power density may change when the packet size is too small for the allocated PDSCH region. This is what Tom said at the beginning.

This story is still just a hypothesis. However, this impact will be small when the packet size is large enough or PDSCH are shared with many UEs by OFDMA.