Posts

5G Beam Management Explained | DL Beamforming, SSB, CSI-RS, Codebook Types

Mohamed ELAdawi

13 Apr 2026 — 12 min read

Introduction

Downlink beam management is one of the most important parts of 5G Massive MIMO. It is the process used by the network to identify, refine, and maintain the best beam direction for each UE, starting from initial access and continuing during connected mode.

Unlike 4G, where beamforming was mainly tied to traffic channels, 5G extends beamforming to common, control, and traffic-related signals. This makes beam management much more important from both coverage and capacity perspectives.

This article focuses mainly on digital beamforming in mid-band 5G NR. It explains the beamforming life cycle, the roles of SSB and CSI-RS, the difference between Eigen-based beamforming and Grid of Beams, and how Type I and Type II codebooks support traffic beam selection.

Content

This article will cover:

What 5G DL beam management is
What are the main functions and stages of beam management
Detailed technical breakdown
Key parameters and where to find them
Practical example walkthrough
Summary key takeaways

(1) What 5G DL beam management is

5G DL beam management is the process of identifying and maintaining the best downlink beam for the UE. In practice, it starts in idle mode with SSB beam sweeping, then moves to beam pairing, followed by CSI-RS-based refinement, and finally traffic beam allocation for channels such as PDSCH.

At a high level, beam management answers three key questions:

Which beam should the UE use first to access the cell?
Which finer beam should be used after the UE enters connected mode?
Which traffic beam should the gNB use to maximize link quality and throughput?

Massive MIMO makes this process more powerful than traditional MIMO because it uses more antenna elements. More elements create sharper and more directive beams, which improves antenna gain, reduces interference, and supports better MU-MIMO performance.

Another key advantage is 2D beamforming. In 5G Massive MIMO, the beam can be controlled in both the azimuth and elevation domains. This is especially useful in dense urban areas and high-rise environments.

To build the beam correctly, the transmitter needs channel-related information. In practice, this can come from:

SRS in uplink, mainly for reciprocity-based beamforming in TDD
CSI-RS feedback, where the UE measures downlink signals and reports CSI back to the network

[Figure 1: Why Massive MIMO creates sharper beams and how 2D beamforming improves gain, SINR, and MU-MIMO support]

(2) What are the main functions and stages of beam management?

The beamforming life cycle in 5G can be summarized as follows:

Initial beam acquisition in idle mode
Beam pairing during access and connection setup
Beam refinement in connected mode
Dynamic traffic beam allocation
Beam mobility management
Beam failure detection and recovery

The UE first finds the best SSB beam, then the network learns this beam through the PRACH association, then the beam is refined using CSI-RS, and later the traffic beam is allocated for actual user data transmission.

The goal is not only to find a beam once. The real objective is to keep the best beam over time while the UE moves, the channel changes, or the traffic conditions vary.

[Figure 2: 5G beamforming life cycle from idle mode to traffic beam allocation]

Which 5G channels and signals can be beamformed?

5G supports beamforming across more channels and signals than 4G.

A simple classification is:

SSB -> common beam, usually with predefined static weight
CSI-RS -> connected-mode measurement beam, usually predefined for measurement purposes
PDCCH -> control beam, static or dynamic depending on implementation
PDSCH/PUSCH -> traffic beams, usually dynamic and based on CSI or SRS

The main reference signals involved in beam management are:

Downlink
- SSB in idle mode
- CSI-RS in connected mode
Uplink
- PRACH during initial access
- SRS in connected mode

[Figure 3: Common beam vs control beam vs traffic beam, and the main signals used in beam management]

(3) Detailed technical breakdown

3.1 SSB-based initial beam acquisition: Procedure 1

The first beam-management step happens in RRC Idle. The gNB transmits an SS Burst using different SSB beam directions, and the UE measures them to find the best one.

At the beginning, the UE does not know which beam is the best. It scans the transmitted SSBs, compares their radio conditions, and selects the beam with the best result, typically based on RSRP.

After decoding the SSB, the UE reads SIB1. From SIB1, it learns the RACH resources and the association between the selected SSB and the corresponding PRACH opportunities.

The UE then sends a PRACH preamble. Since the gNB already knows which PRACH resources are associated with which SSB, it can infer which SSB beam the UE selected. This is how the network identifies the initial beam pair.

In simple terms:

The UE finds the best downlink SSB beam
The UE sends PRACH using the associated resources
The gNB infers the selected beam
Initial DL/UL beam pairing is established

The standard maximum number of SSBs can be summarized as follows:

FR1 up to 3 GHz -> up to 4 SSBs
FR1 from 3 to 6 GHz -> up to 8 SSBs
FR2 -> up to 64 SSBs

💡

From an implementation point of view, the exact sweeping pattern may vary. Some deployments may use a single wide beam for initial acquisition instead of multiple narrow SSB beams.

[Figure 4: SSB beam sweeping, SIB1-based RACH association, and initial DL/UL beam pairing]

After the initial beam is established, the next step is to move to a finer beam that is more suitable for connected-mode unicast transmission.

This is where CSI-RS-based beam refinement is used.

The concept is straightforward:

SSB provides the initial beam
CSI-RS provides a finer beam decision

Each SSB beam can be refined into multiple narrower CSI-RS beams. A common theoretical illustration is 4 CSI-RS beams per SSB, which means that 8 SSB beams could become 32 CSI-RS beams across the cell.

The UE measures these CSI-RS resources and selects the best one based on radio quality, especially RSRP. This leads to a more accurate beam decision, with better link budget and lower interference.

Narrower CSI-RS beams give the UE a better view of the downlink channel. This later helps the gNB form a better traffic beam for PDSCH.

In practice, the exact mapping between SSB beams and CSI-RS beams may vary depending on vendor implementation and deployment strategy.

💡

Additional context: UE-side receive-beam sweeping exists as a concept, but it was not part of the main technical focus here.

[Figure 5: CSI-RS transmit-end refinement inside the initial SSB sector]

3.3 Beamforming types Eigen-based vs Grid of Beams

5G beamforming can be grouped into two high-level categories:

Eigen-Based Beamforming (EBB)
Grid of Beams (GoB)

Eigen-Based Beamforming

In Eigen-based beamforming, the beam follows the UE more dynamically. The gNB uses SRS-based uplink measurements and channel reciprocity to estimate the downlink channel, then updates the beamforming weights accordingly.

This approach is mainly suitable for TDD, because uplink and downlink use the same frequency and reciprocity can be exploited.

Key characteristics:

Dynamic beam update
Good beam tracking
Strong dependence on SRS and reciprocity
Higher baseband processing demand

Grid of Beams

Grid of Beams uses a predefined set of beam directions. Instead of continuously building a new beam pattern, the network switches between already available beams.

This can be implemented in two forms:

Beamformed GoB
- Examples: SSB beams or beamformed CSI-RS beams
- The UE reports the best beam using SSBRI or CRI
Non-beamformed GoB
- The CSI-RS is transmitted without narrow-beam precoding
- The UE reports the best PMI, which helps the gNB choose the best precoding matrix

💡

Grid of Beams is more flexible across FDD and TDD and usually requires less real-time baseband processing than Eigen-based beamforming.

[Figure 6: Eigen-based beamforming vs Grid of Beams, including beamformed and non-beamformed CSI-RS cases ]

3.4 Codebook-based traffic beams Type I vs Type II

Once the UE has measured CSI-RS and the channel is better understood, the network can allocate the traffic beam for PDSCH.

Two main traffic-beam approaches are commonly used:

Reciprocity-based beamforming using SRS
Codebook-based beamforming using CSI feedback

The downlink focus here is the codebook-based path.

What does codebook mean?

A codebook is a predefined set of precoding matrices. The UE measures CSI-RS, identifies the best candidate, and reports feedback so that the gNB can apply the most suitable beamforming weights.

Type I codebook

Type I is the simpler approach and is mainly associated with SU-MIMO.

The process is:

The gNB has a predefined beam set
The UE selects one beam
The UE also reports the co-phase information
The gNB uses this feedback to form the final traffic beam

This is easier to implement, widely used, and effective for single-user transmission.

Type II codebook

Type II provides richer information and is more suitable for MU-MIMO.

Instead of selecting only one beam, the UE can report:

A group of beams
Their relative importance
Amplitude scaling
Co-phasing / linear combination information

This gives the gNB more spatial information and helps it separate users more efficiently in MU-MIMO scheduling.

The trade-off is clear:

Better MU-MIMO support
Better spatial separation
Higher feedback overhead
Stronger UE capability dependency

💡

Additional context: Enhanced Type II and AI-based feedback are part of later evolution paths, but they are beyond the main scope here.

[Figure 7: Type I vs Type II codebook reporting and how the gNB builds the final traffic beam — Type I]

[Figure 7: Type I vs Type II codebook reporting and how the gNB builds the final traffic beam — Type II]

3.5 How the gNB chooses the beam from CSI feedback

The gNB chooses the beam by using the CSI feedback sent by the UE.

The flow is:

The gNB transmits NZP-CSI-RS and CSI-IM
The UE measures the downlink channel
The UE reports CSI, which includes:
- CQI
- RI
- PMI
The gNB calculates the beamforming weight vector
The gNB sends the PDSCH using the chosen traffic beam

A practical interpretation of PMI is:

i11 -> mainly related to the azimuth / horizontal direction
i12 -> mainly related to the vertical / elevation direction

This means different UEs can report different PMI combinations depending on their location. For example:

A UE on street level may need one horizontal beam and a low vertical beam index
A UE in a high-rise building may need a different vertical beam index
Another UE in a different street direction may need a different horizontal beam index

This is one of the clearest ways to understand how CSI feedback is translated into a traffic beam decision.

[Figure 8: Example of CQI/RI/PMI feedback and how i11/i12 help the gNB steer the beam to different UEs]

(4) Key parameters (and where to find them)

Parameter	What it controls	Where it appears (SIB/RRC/DCI/log)	Practical note
Active SSB beams / SS burst positions	Defines how many SSB beam directions are active	SIB1 / cell configuration / logs	In mid-band, up to 8 SSBs is a common standard view, but deployment may use fewer
SSB-to-PRACH association	Maps the selected SSB to the PRACH opportunities	SIB1 / initial access configuration	This is the key link that lets the gNB infer the selected initial beam
`nzp-CSI-RS-ResourceSetId`	Groups the CSI-RS resources used for channel measurement	RRCReconfiguration -> `csi-MeasConfig`	Start from the report config, then trace the referenced resource set
`nzp-CSI-RS-ResourceId`	Identifies each individual CSI-RS resource inside the set	RRCReconfiguration -> `csi-MeasConfig`	Multiple resource IDs may mean multiple CSI-RS beams/resources
`nrofPorts`	Number of CSI-RS antenna ports transmitted to the UE	CSI-RS resource configuration	More ports give more granularity, but also more overhead and stronger UE capability requirements
Codebook type	Selects the feedback/reporting method	RRCReconfiguration / UE capability	Type I is simpler and common for SU-MIMO; Type II is richer and better for MU-MIMO
`N1`, `N2`	Logical antenna dimensions in horizontal and vertical domains	Codebook configuration in RRC	Useful for understanding horizontal vs vertical beam granularity
`O1`, `O2`	Oversampling factors for the codebook grid	Codebook configuration in RRC	Higher oversampling means more predefined beam choices
`frequencyDomainAllocation`	CSI-RS starting subcarrier pattern	CSI-RS resource mapping	Helps identify where the CSI-RS sits in frequency domain
`firstOFDMSymbolInTimeDomain`	Starting OFDM symbol for CSI-RS	CSI-RS resource mapping	Useful when reading traces or visualizing resource placement
`density` / `cdm-Type`	CSI-RS RE pattern and multiplexing method	CSI-RS resource mapping	These affect mapping, overhead, and how ports share REs
`periodicityAndOffset`	Slot timing of periodic / semi-persistent CSI-RS	CSI-RS resource configuration	Critical when checking if the UE is measuring the expected CSI-RS occasions
`powerControlOffset` / `powerControlOffsetSS`	Relative power between CSI-RS, PDSCH, and SS signals	CSI-RS resource configuration	Can influence CSI quality and access behavior
PMI `i11` / `i12`	Beam direction information reported by the UE	CSI report / logs	`i11` is commonly linked to azimuth and `i12` to elevation in practical interpretation
`supportedCSI-RS-ResourceList` / `maxNumberTxPortsPerResource`	UE capability for CSI-RS reception	UE Capability Information	Always check this before enabling higher CSI-RS port configurations

[Figure 9: Example from RRCReconfiguration showing codebook type, CSI-RS resources, and supported ports ]

[Figure 9.1: Example from UE Capability Information showing maximum CSI-RS supported ports ]

(5) Practical example (walkthrough)

Consider a UE in idle mode in a mid-band 5G cell. The cell transmits several SSB beams. The UE measures all of them and finds that SSB beam 3 has the best radio condition.

The UE then reads SIB1, learns the associated PRACH resources, and sends the PRACH preamble using the occasions linked to that SSB. The gNB now understands which initial beam the UE selected, so the initial DL/UL beam pair is established.

After the UE enters connected mode, the gNB configures CSI-RS. Instead of keeping only the initial wide beam view, the network now uses a finer beam decision for user data.

The UE measures the configured NZP-CSI-RS resources and sends CSI feedback. In a simple Type I example, the UE reports the best beam direction through PMI, along with CQI and RI. The gNB then uses this to build the final PDSCH traffic beam.

Now consider two different deployment options:

32 CSI-RS ports with strong codebook granularity
- Better beam-selection granularity
- More overhead
- Stronger UE capability dependency
8 CSI-RS ports with beamformed CSI-RS resources
- Lower overhead
- Often better CSI-RS coverage
- Less traffic-beam granularity than the higher-port case

If the scenario contains many high-rise buildings, a configuration with more vertical control may be preferred. If the area is mostly flat, more horizontal beam granularity may provide better value.

This shows that beamforming design is not only a standards topic. It is also a deployment trade-off topic.

[Figure 10: End-to-end example from SSB beam selection to CSI feedback and final PDSCH beam allocation ]

Summary (Key takeaways)

5G beam management starts with SSB-based initial acquisition, then moves to beam pairing, CSI-RS refinement, and finally traffic beam allocation.
Massive MIMO improves beamforming because more antenna elements create sharper, higher-gain, more directive beams.
5G supports beamforming across common, control, and traffic-related signals, unlike 4G where the focus was mainly on traffic channels.
Eigen-based beamforming is more dynamic and reciprocity-driven, while Grid of Beams uses predefined beam positions and switching/reporting.
Type I codebook is simpler and common for SU-MIMO, while Type II codebook provides richer feedback and better MU-MIMO support at the cost of higher overhead.
Real performance depends not only on the standard view, but also on vendor implementation, UE capability, and deployment choices.

References

3GPP TR 38.802, Section 6.1.6.1
3GPP TS 38.214
3GPP TS 38.331
5G NR in Bullets
Massive MIMO for New Radio (Samsung white paper)