Mapping between layers and antenna ports for Downlink

[pulkit]

Hi Experts,
3GPP specifies one to one mapping between layers and antenna ports for Downlink, that means Precoding is not done for Downlink, is it correct?
Does it also depends on vendor to do precoding in DL?
Because at Rx the precoding matrix can be transparent?

[ankgitm]

Transmission mode 4, 5, 6 use codebook based precoding.
It means that the UE select preferred precoding matrix from the set of matrices available for downlink transmission at the eNB.

[pulkit]

Can you please give the spec number or any other source to read more about it?

[ankgitm]

Transmission mode 8, 9 uses non codebook based precoding.
It means that the UE selects a preferred precoding matrix from a set of matrices defined by 3gpp but the eNB is not constrained to use the same set of matrices.
eNB can use vendor specific weights using SRS reporting.
We called it long term and short term weights.
As of now in NR we applied fixed weights using GoB method but in later release it changes to SRS based.
Note in NR we don’t have any transmission mode.

[Manoj_deka]

@pulkit

3GPP specifies one to one mapping between layers and antenna ports for Downlink, that means Precoding is not done for Downlink, is it correct?
Does it also depends on vendor to do precoding in DL?
Because at Rx the precoding matrix can be transparent?

Not correct.
Precoding is a technique by which a layer is mapped into antenna port.
Y=WX, Y is the AP, W is the Precoding matrix, X is the layer.
For e.g. if there is two layers (RI=2), CW0 is mapped into Layer 1and CW1 into Layer 2.
For 2 APs system, Layer 1 is mapped into both the APs by multiplying with a PM called W (say W0 for mapping into AP0 or W1 for AP1).
Similarly, Layer 2 is also mapped into both APs.
This whole process is also called SFBC (Space Frequency Block Coding).
Space denotes the APs, which are separated by some space.
Freq denotes the Subcarriers in which REs are mapped.
For e.g., RE0 of layer0, is mapped to SC0 of AP0 (multiplying with say W0) and SC1 of AP1 (multiplying with say W1).

[pulkit]

@ankgitm

Transmission mode 8, 9 uses non codebook based precoding.
It means that the UE selects a preferred precoding matrix from a set of matrices defined by 3gpp but the eNB is not constrained to use the same set of matrices.
eNB can use vendor specific weights using SRS reporting.
We called it long term and short term weights.
As of now in NR we applied fixed weights using GoB method but in later release it changes to SRS based.
Note in NR we don’t have any transmission mode.

This confused me, then what are the transmission modes that you mentioned above?

@Manoj_deka

Precoding is a technique by which a layer is mapped into antenna port.
Y=WX, Y is the AP, W is the Precoding matrix, X is the layer.
For e.g. if there is two layers (RI=2), CW0 is mapped into Layer 1and CW1 into Layer 2.
For 2 APs system, Layer 1 is mapped into both the APs by multiplying with a PM called W (say W0 for mapping into AP0 or W1 for AP1).
Similarly, Layer 2 is also mapped into both APs.
This whole process is also called SFBC (Space Frequency Block Coding).
Space denotes the APs, which are separated by some space.
Freq denotes the Subcarriers in which REs are mapped.
For e.g., RE0 of layer0, is mapped to SC0 of AP0 (multiplying with say W0) and SC1 of AP1 (multiplying with say W1).

I am aware of this. My question is related to preciding specifically for DL in NR.
38.211 mentions one to one mapping between layers and antenna ports.
I have limited knowledge on NR parameters as of now therefore doubts are there.

[ankgitm]

@pulkit

This confused me, then what are the transmission modes that you mentioned above?

Those transmission modes applicable only for LTE.

[pulkit]

Ok, thanks for clarification.
Then does DL in NR has precoding?
If yes, is it both codebook and non-codebook or only one of them?

[ankgitm]

Non codebook.
As weight applied at gNB are vendor specific.

2 methods here for NR:

1. GoB based on the basis of csi reporting where fixed weights applied.
2. SRS based where long term and short term weights applied using uplink measurement from SRS.