Math behind the pairing UEs in MU-MIMO

RAN
,
16

Thanks.

I am asking this because N1 N2 O1 O2 doing things complicated when from logical to physical antenna mapping, which I try to figure out also.

In my understanding, those paired users must be seperated logically first with different DFT beams and signal is mapped to physical antenna ports.

This is the most complicated part.

17

Paired users must be in different O1 and O2.

You canot pair 2 UEs that are in same O1 and same O2 but in different N1 and N2.

18

Please check thatā€¦

Not so sure about that, as FDM features may allow that, as i said this is vendor specific feature.

19

FDM feature is Nokia concept.

20

I think you are correct, since N1 N2 O1 O2 values pair for single direction means a physical UE1 location.

And since we know that the UEs must be separated spatially the value sets of these parameters canā€™t be duplicate.

If they are duplicated this means that the 2 beams for 2 UEs are directed at the same location.

21

Number of bubles here should equal to antenna element in the physical antenna right?

5G_CSI_RS_Codebook
5G_CSI_RS_Codebook884Ɨ429 93.6 KB

From Sharetechnote - 5G_CSI_RS_Codebook

22

No it is the number of ports multiplied by the oversampling factor O.

For example:

N1 = 4
O1 = 4
There will be 4Ɨ4 = 16 beam horizontally

N2 = 2
O2 = 4
There will be 2Ɨ4 = 8 beam vertically

23

Thanks, I think I understood MU-MIMO now :smiley:

Just the question those dft beams are similar to CSI-RS beams (32 ports for instance).

What is the relation between csirs ports and dft beams?

24

When N2 is bigger than 1 the antenna array are Uniform planner array and you can do beamforming (move the beam) in vertical and horizontal direction (2D).

When N2 = 1 you have a uniform linear array and you can move the beam horizontally only (1D).

I think it is explained down in Sharetechnote page. I personally didnā€™t link those concepts together yet in my mindā€¦

This post can help you understand this image in more detail, for the record this method is called grid of beams Precoding ā†’ Grid of Beams GoB precoding in 5G

25

I think what was confusing here and sharetechnote article: gNB actually does not need csi.rs for DL channel estimation and adjust weights, SRS is good at cell center.

@Jaeku_Ryu (ShareTechNote) always implies csi-rs and codebook togetherā€¦ which confused me a lot. :slight_smile:

26

It suggest reading this short article too: Types of beamforming in 5G:

27

You can use either, SRS or CSI-RS from my understanding.

Please check this Samsung Technical White Paper about the topic here now. :point_down:

28

Really good article to tell people the general principles. :+1:

I think I figured out the type 1 and type 2 from 5G NR Bullets, which I read again and again but only now it is clear. :slight_smile:

Thanks guys brainstorming was really good! :clap:

1 Like
29

From Samsung white paper,

I think type 1 uses either SRS or CSI-RS.

Type 2 uses SRS.

30

Nopeā€¦ per my understanding both PMI and SRS possible for type 2.

31

Right, you are correct :+1:t2:

32

What indeed determines the performance of type 2?

Which algorith is used, zero forcing or non zero forcing?

And now I have to understand the differenceā€¦ (Holy, never finish :laughing:)

If we continue like this brainstorming 1 week, we will figure out all about MU-MIMO! :joy:

Those are quite complicated topics.

33

Correct :white_check_mark: :100:

34

This is a good (advanced) video on MU-MIMO calculations:

MU - MIMO Channel Estimation in 4G / 5G | 2 UE Single Layer Case
MU - MIMO Channel Estimation in 4G / 5G | 2 UE Single Layer Case1246Ɨ641 108 KB

MU - MIMO Channel Estimation in 4G / 5G | 2 UE Single Layer Case II

YouTube: :point_right: https://www.youtube.com/watch?v=2EupE_Vb-jo

1 Like
35

I tried to put a little bit more detailed / consolidated note on this topic :
https://www.sharetechnote.com/html/5G/5G_MassiveMIMO_MU_MIMO.html

1 Like