PMI vs SRS at different SINR

Hello 5G/LTE Experts.

Why pmi beam weight at low sinr will be enough for estimation?
And why at high sinr SRS beam weight will be enough for estimation?
Any diagram performance that show pmi vs srs at different SINR?

I think SRS-based weight calculation measures uplink SRS RSRP and its DL channel quality is based on uplink status that’s why high SINR is required.
Particularly In a TDD system, air interface channels naturally have reciprocity, but RF channels require calibration to ensure the reciprocity between TX and RX channels.
While, PMI-based weight calculation only estimates DL channel quality and no calibration require with UL so low SINR will be sufficient for channel estimation

Any research diagram shows that?
I mean shows SRS vs PMI over SINR that we can conclude from it that at high SINR best to use srs otherwise PMI used?
Do you mean because at low SINR I will not be able to get correctly the SRS from the UE so PMI estimation is used because it depends on DL signal?

Yes as in PMI calculation UE will only measure DL CSI RS RSRP – Further, gNodeB can adaptively select SRS or PMI based weights for the downlink so that the weights can more accurately reflect the quality of traffic channels.
And furthermore, if numbers of TX and RX antennas are less than four, PMI-based weights are always used and SRS-based weights are not supported(this could be vendor specific).

I may didn’t get you. :frowning:

But in both cases UE will need to send srs or csi rs?
So… I mean okay we said at bad sinr means the srs that UE send to gNB will not arrive at bad sinr to gNB so SRS beam weight isnt efficient.
Now if gNB decides in bad sinr to use pmi weight it still need from UE to send to it csi-rs for pmi estimation at gnb?
So also csi rs at bad sinr that UE send to gNB will not be decoded correctly because bad sinr.
So why we say that at bad sinr we use pmi if we still need at bad sinr from UE csi-rs to be sent?

Bad sinr here means UL signal quality received by eNB, not the reported CQI.
If UL quality is poor then it is not properly reflecting actual DL channel because eNB can transmit at high power but UE is power limited.
Therefore, switch to csi-rs that is measuring DL channel.
If UL quality is good, then DL. Channel is either equal or better, hence use SRS.
If both eNB and UE can transmit at same power then the channel is a mirror and SRS will be only thing needed.

1 Like

SRS has lower uplink coverage compared to CSI feedback. So, SRS is used in good coverage while once the UE moves to bad coverage, decisions are supposed to be based on CSI Feedback (PMI).
Now why gNB can decode CSI feedback in bad coverage but not decode SRS - it is because SRS is more of a wideband channel so the UE needs to send SRS over a much larger frequency allocation which means the UE’s power is divided over a much larger number of subcarriers. However, CSI feedback (PMI) is a small message and thus would have a higher power per subcarrier.