Has anyone seen up to now UL MIMO in 5G?

Hi All.

Question about 5G: has anyone seen up to now UL MIMO in 5G?

I understand that there is in 5G SA UL MIMO 2x2 for very few UEs, but not in 5G NSA.

Admin note: this post was updated with image below.

I’ve seen only in test devices so far, either prototype or from OEMs but non commercial.

1 Like

Here is UL MIMO with Snapdragon Insider Phone in Elisa Finland 5G SA network.

There is some commercial UE’s support UL MIMO like Mi 11 Ultra: Xiaomi Mi 11 Ultra (Global) (M2102K1G) | 4G/5G Bands & Combos

1 Like

Hi All.

Anyone have seen more than 8 UEs per TTI in 5G Uplink?

Per slot you mean?

At every PUSCH 8 users?

Yes. Have you seen it?

But you should mention: mu-mimo or su-mimo?

Just going to enable it for SU-MIMO.

Checking perf right now.

Later for MU-MIMO.

So you are gonna enable it for Throughput UL/DL testing?

In SU-MIMO increasing users in PDSCH / PUSCH will not bring much benefit in DL/UL throughput.

Oppositely even can decrease UE throughput, little improvement of cell edge throughput and improve the latency due to more UEs to be scheduled in slot.

Real impact of increasing number of users per slot visible at MU-MIMO.

Of course when you set parameters given here does not mean gNB will able to schedule those each time, it is depend on mostly on Aggregation level, (AL PDCCH) usage, because depends on al allocated gNB can assign REs for multiples UEs at the same time.

Actually you started a nice topic but I will extend it to MU-MIMO. :wink:

Did you mean CCE not enough (high used) so mu-user affected?

Yes.

Number of users can be paired is totally depend on if you have enough CCE allocated.

If anyone knows what is the correlation factor in MU-MIMO please ping.

I’d like to understand how gNB calculates this via CS1 & CS2 lists.

Correlation factor means the correlation between phases according to each UE location.

For example for 32 csi-rs ports for UE1 is needed those phases a1, a2, a3 … a32 and for UE2 is needed those phases b1, b2, b3 … b32.

Correlation factor is functions between those phases of UE1 and UE2.

If correlation is low UE can be paired, if it’s is high cannot be paired.

More about correlation here:

If you understand this picture you understand MU-MIMO:

This is from this link: Sharetechnote - 5G_CSI_RS_Codebook

But the thresholds to pair an UE are parameters usually.

For example default value of correlation between 2 UEs to be paired is 0,3.

Which means that if correlation is 0,4 between 2 UEs they won’t be paired

But if I change the threshold from 0.3 to 0.4 will be paired with important interference one to another.

This is why is important not to change correlation threshold for pairing from default one

There was a question of how is the throughput between NR and LTE for same bandwidth.

This picture provides the answer:

Theoretical peak of NR FDD performance (SA)

In Nokia, it is 0.33.

But also there is cqi threshold, generally set to very low.

All higher values for sake of NR because of F-OFDMA used only in NR.

That CQI is to make a distinction between PMI and SRS weights.

Cool, now I can continue to understand 2d dft taperin, there is also threshold for that tapering in horizontal and vertical in term of DB.

parameter names : dlmumimotaperingV, h in DB for anyone would like to check.

My understanding, after pairing to meet some SINR requirement gNB needs to perform taperin on beams.