Ferkans — Interactive Telecom Tutor

Massive MIMO Meets Standards — The Beam Management Challenge

Chapter 18 showed that massive MIMO can deliver extraordinary spectral efficiency through spatial multiplexing of many users. Translating this into a deployable standard requires solving a practical challenge: with 64--256 antennas forming narrow beams, how does the network and UE find each other? At sub-6 GHz, the base station can transmit cell-wide reference signals and rely on digital beamforming. At mmWave, the link budget demands beamformed reference signals, but the UE does not initially know which beam to listen on. NR's beam management framework — the P1/P2/P3 procedures — provides a systematic solution using SSB beam sweeping, CSI-RS beam refinement, and beam failure recovery. This is arguably the most complex new physical layer procedure in NR compared to LTE.

Definition:
NR CSI Framework — Type I and Type II Feedback

NR defines two CSI feedback types for closed-loop MIMO:

Type I (single-panel, codebook-based):

The UE selects one beam from a DFT-based codebook and reports a PMI (Precoding Matrix Indicator).
Wideband $\mathbf{W}_{1}$ selects a beam group; subband $\mathbf{W}_{2}$ selects the best beam and co-phase within the group.
Overhead: $\sim$ 10--20 bits per reporting instance.
Suitable for FDD with moderate antenna counts (up to 32 ports).

Type II (multi-beam, high-resolution):

The UE reports a linear combination of $L$ beams from the oversampled DFT codebook: $\mathbf{w} = \sum_{l=1}^{L} c_l\, \mathbf{b}_l$ where $\mathbf{b}_l$ are orthogonal beams and $c_l$ are complex combining coefficients.
Wideband: beam indices $\{l\}$ and wideband amplitudes.
Subband: phase and amplitude of each beam per subband.
Overhead: $\sim$ 50--200 bits per instance.
Near-optimal for MU-MIMO with large antenna arrays.

Type II CSI trades higher feedback overhead for significantly better MU-MIMO performance. The key insight is that a linear combination of $L = 4$ beams can represent most practical channels with high fidelity, whereas Type I's single-beam selection misses the spatial structure needed for multi-user interference management.

,

SSB Burst Set Structure — SSB burst set within a 5 ms half-frame for FR2 ( $L_{\max} = 64$ ). Each SSB occupies 4 OFDM symbols $\times$ 20 RBs and is transmitted on a different spatial beam direction, enabling the UE to identify the strongest beam during initial access by measuring L1-RSRP on each SSB index.

Definition:
NR Beam Management Procedures (P1/P2/P3)

NR beam management is a hierarchical framework for beam acquisition, refinement, and tracking:

P1 — Beam Sweeping (Initial Acquisition): The gNB transmits SSBs (SS/PBCH Blocks) in different beam directions. Up to $L_{\text{SSB}}$ beams are swept per SSB burst set (4 in FR1, 64 in FR2). The UE measures RSRP on each SSB and reports the best beam index.

P2 — gNB Beam Refinement: After P1 identifies a coarse beam, the gNB transmits CSI-RS on a finer grid of beams around the P1 direction. The UE reports the best refined beam.

P3 — UE Beam Refinement: The UE sweeps its own receive beams (analog Rx beamforming) while the gNB holds its Tx beam fixed. This aligns the UE's receive beam to the gNB's transmit beam.

Beam Failure Recovery: If the serving beam quality drops below a threshold ( $Q_{\text{out}}$ ), the UE initiates beam failure recovery by identifying a new candidate beam (from SSB or CSI-RS measurements) and sending a BFRQ on PRACH or PUCCH.

The P1/P2/P3 hierarchy reflects a practical constraint: exhaustive beam search over all gNB and UE beam combinations would take $L_{\text{gNB}} \times L_{\text{UE}}$ measurements, which is prohibitive for large arrays. The hierarchical approach reduces this to $L_{\text{gNB}} + L_{\text{refined}} + L_{\text{UE}}$ measurements.

,

Theorem: SSB Beam Sweep Latency and Coverage

An SSB burst set with $L$ beams and periodicity $T_{\text{SSB}}$ has the following properties:

Angular coverage per beam: For a ULA with $N_{\text{ant}}$ antennas, each beam covers approximately:

$\Delta\theta \approx \frac{2}{N_{\text{ant}}} \;\text{radians} = \frac{114.6^{\circ}}{N_{\text{ant}}}$

To cover a sector of angular width $\Theta_{\text{sector}}$ (typically $120^{\circ}$ ), the minimum number of beams is:

$L_{\min} = \left\lceil\frac{\Theta_{\text{sector}}}{\Delta\theta}\right\rceil = \left\lceil\frac{N_{\text{ant}}}{2} \cdot \frac{\Theta_{\text{sector}}}{57.3^{\circ}}\right\rceil$

Initial access latency (worst case):

$T_{\text{access}} = T_{\text{SSB}} + T_{\text{RACH}} + T_{\text{RAR}}$

where $T_{\text{SSB}}$ is the SSB periodicity (5--160 ms, default 20 ms) and $T_{\text{RACH}}$ accounts for PRACH timing alignment.

More antennas create narrower beams (higher gain) but require more beams to cover the sector, increasing sweep time. This is the fundamental antenna gain vs. access latency trade-off in mmWave systems. With $N = 64$ antennas, $\sim$ 32 beams are needed for $120^{\circ}$ coverage, each with $\sim 3.6^{\circ}$ beamwidth.

Proof

Beamwidth of a ULA

The half-power beamwidth of a ULA with $N$ elements and half-wavelength spacing is approximately:

$\text{HPBW} \approx \frac{0.886 \times 2}{N} \approx \frac{2}{N} \;\text{radians}$

(The exact 3 dB beamwidth depends on the taper/window, but $2/N$ is a standard approximation for uniform weighting.)

Number of beams for sector coverage

A $120^{\circ}$ sector spans $2\pi/3$ radians. With beam spacing equal to the HPBW to ensure coverage with at most 3 dB loss at beam edges:

$L_{\min} = \left\lceil\frac{2\pi/3}{2/N}\right\rceil = \left\lceil\frac{\pi N}{3}\right\rceil \approx N$

For $N = 64$ : $L_{\min} \approx 67$ . FR2 supports up to 64 SSBs, which is nearly sufficient.

Access latency

The UE must wait for the SSB carrying the best beam. In the worst case (beam at the end of the sweep), this is one full SSB periodicity $T_{\text{SSB}}$ . After detection, RACH and RAR add 3--5 ms.

Total worst-case: $T_{\text{SSB}} + 5$ ms. With $T_{\text{SSB}} = 20$ ms: $\sim$ 25 ms. $\blacksquare$

,

SSB Beam Sweeping for Initial Access

Watch the gNB sweep SSB beams sequentially across a 120° sector. Each beam illuminates a narrow angular region. When the beam pointing toward the UE is transmitted, the UE detects the highest RSRP and reports the best beam index — completing the P1 procedure.

Eight SSB beams sweep a 120° sector. The UE (red) detects beam 5 (green) as the strongest.

SSB Beam Sweep Pattern

Visualise the SSB beam sweep pattern for different configurations. The plot shows the beam directions in the angular domain, the beam gain pattern, and the time-domain SSB burst structure. Adjust the number of SSB beams to see how coverage and sweep time trade off, and vary the periodicity to observe its impact on initial access latency.

Parameters

Number of SSB beams8

SSB periodicity (ms)20

NR CSI Feedback — Type I vs Type II

Compare the spectral efficiency achievable with Type I (single-beam) and Type II (multi-beam) CSI feedback as a function of SNR. Adjust the number of antenna ports, the number of beams $L$ in Type II, and the transmission rank. Observe how Type II with $L = 4$ beams approaches ideal (genie-aided) CSI performance for MU-MIMO, while Type I saturates due to quantisation error.

Parameters

Antenna ports32

Type II beams L4

Transmission rank2

Example: Beam Management Design for FR2

A 5G NR mmWave gNB at 28 GHz uses a 64-element ULA panel covering a $120^{\circ}$ sector.

(a) How many SSB beams are needed for full sector coverage? (b) With SSB periodicity of 20 ms, what is the worst-case initial access latency? (c) After P1, the gNB refines within a $\pm 5^{\circ}$ window using CSI-RS. How many P2 beams are needed at twice the angular resolution? (d) Estimate the total beam alignment time (P1 + P2 + P3) if the UE has 4 Rx beams.

Solution

SSB beam count

(a) Beamwidth $\approx 2/64$ rad $= 1.79^{\circ}$ . Beams for $120^{\circ}$ : $\lceil 120/1.79 \rceil = 67$ . NR FR2 supports $L_{\max} = 64$ SSBs, so we use 64 beams (slight under-coverage at sector edges, acceptable with overlap from adjacent panels).

Access latency

(b) Worst case: UE must wait for the entire SSB burst. SSB burst duration: 64 SSBs $\times$ 4 symbols $\times$ $(8.33 + 0.59)$ $\mu$ s $\approx 64 \times 35.7 = 2.28$ ms. Plus periodicity wait: up to 20 ms. Total worst-case P1: $\sim 22$ ms.

P2 refinement beams

(c) P2 window: $\pm 5^{\circ} = 10^{\circ}$ . P2 beamwidth: $1.79/2 \approx 0.9^{\circ}$ (doubled resolution). P2 beams: $\lceil 10/0.9 \rceil = 12$ .

Total alignment time

(d) P1: 22 ms (worst case). P2: 12 beams at 1 CSI-RS per slot ( $0.125$ ms each) = 1.5 ms. P3: 4 UE Rx beams $\times$ 0.125 ms = 0.5 ms. Total: $22 + 1.5 + 0.5 = 24$ ms.

In practice, P2 and P3 can partially overlap, reducing this to $\sim 23$ ms. $\blacksquare$

Quick Check

What is the purpose of the P3 procedure in NR beam management?

To select the initial gNB transmit beam during cell search

To refine the gNB transmit beam using CSI-RS

To align the UE receive beam while the gNB holds its transmit beam fixed

To handle beam failure recovery when the serving beam is lost

Correction:

To align the UE receive beam while the gNB holds its transmit beam fixed

P3 is UE Rx beam refinement. The gNB transmits on its selected beam (from P1/P2) while the UE sweeps its analog receive beams to find the best Rx direction. This is particularly important for mmWave UEs with analog Rx beamforming (phased arrays).

🎓CommIT Contribution(2013)

JSDM: Joint Spatial Division and Multiplexing

A. Adhikary, J. Nam, J.-Y. Ahn, G. Caire — IEEE Trans. Information Theory, vol. 59, no. 10

Adhikary, Nam, Ahn, and Caire proposed JSDM, a two-stage precoding framework for FDD massive MIMO that separates users into groups based on their channel covariance eigenspaces and applies a pre-beamforming matrix per group followed by a MU-MIMO precoding matrix within each group. The pre-beamformer exploits the long-term spatial statistics (which change slowly and require low feedback overhead), while the inner precoder handles fast fading.

The NR Type I CSI dual-stage codebook ( $\mathbf{W}_{1} \mathbf{W}_{2}$ structure) is a direct standardisation of this concept: $\mathbf{W}_{1}$ selects a wideband beam group (analogous to the JSDM pre-beamformer) and $\mathbf{W}_{2}$ selects the subband precoder within that group. JSDM demonstrated that FDD massive MIMO is feasible without prohibitive CSI feedback — a result that directly informed the 3GPP NR MIMO design.

JSDMmassive-MIMOFDDtwo-stage-precodingCSI-feedbackView Paper →

Why This Matters: Massive MIMO Theory Behind NR MIMO

The massive MIMO concepts in Chapter 18 — channel hardening, favourable propagation, TDD reciprocity-based precoding, and pilot contamination — are the theoretical foundations for NR's MIMO design. The NR beam management framework (P1/P2/P3) is the practical answer to the initial access problem that massive MIMO creates: narrow beams increase gain but require beam alignment before data transmission can begin. The MIMO book in this library provides the full theoretical treatment, including DPC-achieving precoding, RZF/ZF/MMSE precoder analysis, and the cell-free massive MIMO paradigm that may define 6G.

Why This Matters: Reconfigurable Intelligent Surfaces as Passive Beamformers

The beam management overhead in NR motivates research into reconfigurable intelligent surfaces (RIS) (Chapter 28), which create additional controllable propagation paths without active RF chains. RIS can assist beam alignment by reflecting the gNB beam toward the UE, potentially reducing the number of SSB beams needed for coverage — especially in NLOS scenarios where direct beam paths are blocked.

Key Takeaway

Beam management is the price of massive MIMO at mmWave. Narrow beams from large arrays provide the gain needed to close the mmWave link budget, but they require a multi-stage alignment procedure (P1/P2/P3) before any data can flow. The worst-case initial access latency of $\sim$ 25 ms at FR2 is an order of magnitude longer than LTE's cell search, creating a fundamental tension between array gain and access speed.

Common Mistake: Beam Failure in Dynamic Environments

Mistake:

Relying solely on the initial P1/P2/P3 alignment without continuous beam tracking, assuming the beam remains stable after initial alignment.

Correction:

In mobile scenarios, the optimal beam direction changes continuously. Without beam tracking, a UE moving at 30 km/h can exit a 3 $^{\circ}$ mmWave beam in $\sim$ 50 ms. NR provides beam failure detection (BFD) via periodic L1-RSRP measurements and beam failure recovery (BFR) via PRACH-based beam re-selection. Additionally, SSB and CSI-RS measurements must be configured at appropriate periodicity (10--40 ms) to track beam drift proactively.

SSB (SS/PBCH Block)

Synchronisation Signal / Physical Broadcast Channel Block: a time-frequency resource carrying PSS, SSS, PBCH, and DM-RS, transmitted in a specific beam direction. NR supports up to 4 (FR1) or 64 (FR2) SSBs per burst set for beam sweeping.

Beam Management

The NR framework for beam acquisition, refinement, and maintenance using hierarchical procedures: P1 (SSB sweep for initial beam), P2 (CSI-RS for gNB beam refinement), P3 (UE Rx beam refinement), and beam failure recovery.

CSI-RS (Channel State Information Reference Signal)

A configurable reference signal in NR used for channel measurement, beam management (P2 refinement), and CSI acquisition for link adaptation and precoding. Unlike LTE's always-on CRS, CSI-RS is transmitted only when configured, reducing overhead.

5G NR MIMO and Beam Management

Massive MIMO Meets Standards — The Beam Management Challenge

Definition: NR CSI Framework — Type I and Type II Feedback

SSB Burst Set Structure

Definition: NR Beam Management Procedures (P1/P2/P3)

Theorem: SSB Beam Sweep Latency and Coverage

Beamwidth of a ULA

Number of beams for sector coverage

Access latency

SSB Beam Sweeping for Initial Access

SSB Beam Sweep Pattern

Parameters

NR CSI Feedback — Type I vs Type II

Parameters

Example: Beam Management Design for FR2

SSB beam count

Access latency

P2 refinement beams

Total alignment time

Quick Check

JSDM: Joint Spatial Division and Multiplexing

Why This Matters: Massive MIMO Theory Behind NR MIMO

Why This Matters: Reconfigurable Intelligent Surfaces as Passive Beamformers

Key Takeaway

Common Mistake: Beam Failure in Dynamic Environments

SSB (SS/PBCH Block)

Beam Management

CSI-RS (Channel State Information Reference Signal)

Definition:
NR CSI Framework — Type I and Type II Feedback

Definition:
NR Beam Management Procedures (P1/P2/P3)