The idea is: 16 ssb beams cover whole cell, for eg 128°, initial access UE and gNB know which SSB index, can locate UE within 8° of each SSB.
Then in connected, locate UE within 2° of each csi-rs beams?
Refined beam possible with FR1 only.
What you need to understand for csi-rs ports is this:
A PDSCH layer is sent to each of the csi-rs port (let’s say 16 of them) and each csi-rs port will apply a different weight so all 16 signals will reach a speciifc Ue antenna (no 3 let’s say) so will sum up to reach maximum level.
Followings short notes for each process.
This is very well explained in Sharetechnote:
A -> B : This is the process of the raw data (data with DMRS) to CRS-RS antenna port. This mapping is done by multiplying the data with PMI matrix(codebook). The number of columns of this matrix is same as the number of layers (i.e, the number of DMRS antenna ports) and the number of raws of this matrix is same as the number of CSI-RS antenna ports. This mapping can be illustrated as follows.
FR2 also have refined beams.
Really? In my understanding it’s with digital beamforming only.
I think beam forming is target of FR2.
All mmWave I have seen had Csi beams.
Hope this is what you meant by refined beams.
I think FR1 has digitial beamforming while FR2 has analogue beamforming.
But most vendors use hybrid versions of AAU where both analog and digital beamforming is used.
Yes, I mean refined beams only valid with FR1 only.
Yes, correct. Hybrid beamforming is future concept as of now.
I think we have discussed this with some logs long time back, where I have shared snaps of 4 refined beams in mmwave DT csi-rsrp of the best refined csi beam was chosen.
Maybe was this discussion: Beam Management, Beam Switching and Beam Refinement in 5G NR - #22 by ankgitm
Very well explained, thanks!
1 Like
Anyone can share idea about CSI resource?
3GPP TS 38.215 V16.4.0 (2020-12)
CSI reference signal received power (CSI-RSRP), is defined as the linear average over the power contributions (in [W]) of the resource elements of the antenna port(s) that carry CSI reference signals configured for RSRP measurements within the considered measurement frequency bandwidth in the configured CSI-RS occasions.
L1 RSRP explicitly mentioned used for beam switching it will be use when we set report quantity to cri RSRP.
Let me explain how I understand csi-rs reference signals.
Each csi-rs reference signal is transmitted from one antenna port 3000s.
So if Ue has 3 pdsch layers with 8 csi-rs ports then there will be 8 csi-rs reference signals and UE will use 3 antenna to receive the pdsch layer.
Accordingly the PMI matrix will be exactly what UE measures with 3 antennas, each of the 3 antenna measuring 8 csi-rs referrence signals.
So PMIL will be LXP where L is the pdsch layers and P is the csi-rs ports.
So UE will measure port0 with 4 antenna, port1 with 4 antenna, port 2 with 4 antenna and so on.
So in our case of 3 pdsch layers and 8 csi-rs ports UE will report a PMI that is 8x3=24 elements inside it.
Ideea is that PMI cannot have any value (it woudl be too complicated) so 3gpp limited the values of PMI to a precoded book.
So UE has to chose from a set of predefined PMIs and chose the best one that fits him.
Overall understanding is correct, but there are some mis-concepts.
- CSI-RS port number corresponds to the number of RF chains, the maximum number of allowed layers, and the maximum number of the digital beamforming PMI, here it is 8.
- There should be 8 CSI-RS reference signal for UE channel sounding. Since UE has 4 antenna, it should do channel estimates with size of 4x8, allowing maximum 4 layers PDSCH transmission. However since gNB indicates 3 layers transmission, thus UE only selects PMI size of 8x3 report back to gNB for digital beamforming.
2 Likes
It was a nice discusion dears.
Now I am clear with it.
Each csi-rs port is broadcasted over the whole footprint of the cell.
Hello I was reading all the discussion and for me it’s more or less clear also the mapping between CSI-RS ports with thei own RS.
We know that we can have up to 32 APs for CSI-RS today, in fact, is the approach that is used by some vendors like Ericsson, this enables the possibility for the UE to measure better the CSI for different TX chains but it’s more complex for the UE in terms of calculations and with 32 APs CSI-RS it’s not possible to do beamforming of the CSI-RS.
Now my doubts it’s about when the Beamforming of the CSI-RS is used that I believe is the approach of Huawei and Nokia, anyway, as per my understanding if the beamforming of CSI-RS is used then the CRI is needed, right? is this strictrly necessary? (the CRI, I mean)
My doubt is, in the supposed scenario of having 8-ports CSI-RS with beamforming each port will be mapped on how many beams? and typically those beams will be refined beams inside the SSB beam right?
Yes, as per my understanding the CSI-RS can be beamformed, don’t know if the term “beamforming” can be confusing. I’m not refering like the PDSCH than with the processing of SRS can be “beamformed” towards a specific user but to create beams were the CSI-RS are transmitted. E.g. multiple CSI-RS beams within SSB beams. In all this cases I’m talking about FR1.
In this sense, I’m not sure how many beams can be formed for CSI-RS using 8-ports (according to the literature will depend on N1 and N2) and if CSI-RS beams are formed within the SSB beams all the 8-ports will be transmited INSIDE that SSB? and then for the next SSB then again the 8-ports and so on.
And lastly there is the role of the CRI
1 Like