CSI-RS Beam
How CSI-RS Beams Refine the 5G NR Link
The 3GPP NR beam-management framework defines three procedures, labeled P1, P2, and P3. P1 is the coarse phase: the gNB sweeps a set of Synchronization Signal Block beams during the half-frame burst, up to 64 of them in FR2, and the UE selects the strongest SSB index for initial access. The CSI-RS beam enters at P2 and P3. After the UE has reported its best SSB, the network configures one or more CSI-RS resources that are spatially associated with that SSB through a Quasi Co-Location Type D relationship, meaning the UE can assume the same departure and arrival angles. The gNB then transmits a sweep of narrower CSI-RS beams so the UE can pick a direction far more precisely than SSB alone permits.
In the P2 procedure the gNB sweeps multiple narrow transmit beams while the UE keeps its receive beam fixed; the UE measures L1-RSRP on each CSI-RS resource and reports the winning CSI-RS Resource Indicator, or CRI. In P3 the roles reverse: the gNB holds one transmit beam steady and repeats the CSI-RS so the UE can sweep its own receive beams and lock its analog spatial filter. This pairing of transmit-side and receive-side refinement is what allows a millimeter-wave link to stay aligned as the user moves, and it is why CSI-RS resource sets carry a repetition flag that tells the UE whether the gNB beam is changing or being held constant.
CSI-RS serves a second, distinct role beyond beam selection. Configured with multiple ports and higher density, a non-zero-power CSI-RS becomes the pilot for full channel state information estimation, from which the UE derives rank, precoding matrix, and channel quality reports. Beam-management CSI-RS, by contrast, is usually single-port and low-density because only received power matters. Network planners must budget the resource-element overhead of both modes against the periodicity needed to track fast-changing mmWave channels.
CSI-RS Beam Reporting Equations
RSRPL1 = (1/N) × ∑ |hk · wtx|2 (linear, averaged over N resource elements)
Reported CRI (best resource of the set):
CRI = arg maxi RSRPL1(i), i ∈ {0 … K−1}
Approximate 3 dB beamwidth of an N-element array:
θ3dB ≈ 0.886 × λ / (N × d × cosθ0) rad
Where hk = channel on RE k, wtx = transmit beamforming weight vector, K = resources per CSI-RS set (up to 64), λ = wavelength, N = element count, d = element spacing, θ0 = steering angle. Example: 8-element array, d = λ/2, broadside → θ3dB ≈ 12.7°; 32 elements → ≈ 3.2°.
Reference Signal Roles in 5G NR Beam Management
| Reference Signal | Primary Role | Beamwidth | UE Report | Procedure | Typical Periodicity |
|---|---|---|---|---|---|
| SSB | Initial access, coarse beam sweep | ~10° to 15° | SSBRI + L1-RSRP | P1 | 5 to 20 ms |
| CSI-RS (beam mgmt) | Tx/Rx beam refinement | ~2° to 5° | CRI + L1-RSRP | P2, P3 | 10 to 40 ms or aperiodic |
| CSI-RS (NZP, MIMO) | Channel estimation, CSI feedback | Beamformed or wide | RI / PMI / CQI | n/a | 5 to 80 ms |
| TRS (CSI-RS burst) | Time/frequency tracking | Follows serving beam | None (UE internal) | n/a | 10 to 40 ms |
| SRS | Uplink sounding, reciprocity | UE-formed | gNB measures | Uplink beam mgmt | 1 to 40 ms |
Frequently Asked Questions
What is the difference between SSB-based and CSI-RS-based beam management?
SSB beams are coarse, up to 64 lobes of roughly 10° to 15° swept in the FR2 burst for initial access and the P1 wide-beam sweep. CSI-RS beams are the refinement layer: after the UE reports its best SSB, the gNB configures CSI-RS spatially tied to that SSB by a QCL Type D relationship and sweeps narrower 2° to 5° beams. P2 sweeps the gNB transmit beam with a fixed UE receive beam; P3 fixes the transmit beam while the UE sweeps its receive beams. CSI-RS supplies the angular resolution SSB cannot, which is decisive at 24 to 52 GHz where beams must be narrow to fight path loss.
How does a UE report the best CSI-RS beam to the gNB?
The UE measures L1-RSRP on each CSI-RS resource in the set and reports a CSI-RS Resource Indicator (CRI) for the best resource plus its measured L1-RSRP, with up to four beams per report. L1-RSRP is quantized from about −140 dBm to −44 dBm in 1 dB steps, and weaker beams are reported as 2 dB differentials. Reporting can be periodic on PUCCH, semi-persistent, or aperiodic when triggered by DCI. The gNB uses the CRI to set the serving beam and the Transmission Configuration Indicator state for later PDSCH and PDCCH reception.
What CSI-RS resource configuration is used for beam refinement versus channel estimation?
Beam-management CSI-RS uses the repetition parameter (on or off) and is typically single-port at a density of one resource element per resource block, since only received power matters. Tracking uses a TRS, a special CSI-RS burst for fine time and frequency sync. Full MIMO channel estimation uses non-zero-power CSI-RS with up to 32 ports and higher density so the UE computes rank, PMI, and CQI rather than only L1-RSRP. Zero-power CSI-RS creates rate-matching gaps so a neighbor cell's CSI-RS does not collide with this cell's PDSCH.