Beamforming

Beamforming is the technique of shaping and steering the radiation pattern of an antenna array toward a specific user, concentrating signal energy where it's needed rather than radiating uniformly. This traces the evolution from LTE through 5G NR to the 6G frontier.

Listen to the Companion Podcast Episode

#1The Evolution of Coverage

Cellular coverage has progressed from broadcasting everywhere to focusing energy precisely:

2G / 3G
Omnidirectional
"Bare Bulb"
All directions — massive waste
LTE
Sectorized
"Floodlight"
Three 120° sectors — moderate
5G NR
Beamformed
"Spotlight"
Narrow beam steered to each UE
6G
Intelligent
RIS"Laser Tracker"
Focused beam + environment shaping

Why "Spray and Pray" Fails at Higher Frequencies

FrequencyWavelengthAntenna SizePath Loss
900 MHzλ = 33 cmLarge antennasLow — easy coverage
3.5 GHzλ = 8.6 cmCompact arraysModerate
28 GHzλ = 1.07 cmTiny elementsHIGH — needs beamforming!
140 GHzλ = 2.1 mmMicro elementsEXTREME — ultra-massive arrays needed

Friis Transmission Equation

Pr = Pt × Gt × Gr × λ² / (4πd)²

As frequency ↑ (λ ↓), received power Pr ↓ dramatically. Beamforming gain (Gt, Gr) compensates.


#2Physics Foundation — How Beams Are Formed

2.1 Single Antenna vs Antenna Array

Single Antenna — Wide, Low Gain

Energy in all directions

Antenna Array — Narrow, High Gain

Main Lobe Side Lobe Side Lobe Energy concentrated → HIGH GAIN

2.2 Constructive & Destructive Interference

2.3 Beamsteering — Visual Explanation

Electronically redirecting the beam by adjusting phase delay across elements — no physical movement:

Beam RIGHT (θ = +30°) 0π/4π/23π/4 BROADSIDE (θ = 0°) 0000 All elements in-phase Beam LEFT (θ = -30°) 3π/4π/2π/40
Drag the angle slider to steer the beam; the per-element phase gradient updates in real time.

2.4 Key Radiation Pattern Terms

Main Lobe Side Lobe Side Lobe Null Null 3 dB Beamwidth
Polar gain pattern: more elements = narrower main lobe + more side lobes. Pink line marks the steering direction.

Beamwidth Relationships


#3Beamforming Architectures

ANALOG 1 RF Chain (DAC) Power Splitter φ₁ φ₂ φ₃ φ₄ ✅ 1 RF chain — low cost/power ✅ Ideal for mmWave ❌ ONE beam at a time ❌ Sequential sweep needed 📍 5G FR2 initial access DIGITAL RF Chain 1 RF Chain 2 RF Chain 3 RF Chain 4 ← Full chain per element ✅ ALL beams simultaneously ✅ Native MU-MIMO ✅ Max flexibility & precision ❌ N RF chains — expensive at mmWave ❌ High power consumption 📍 5G FR1 (sub-6 GHz) HYBRID RF Chain 1 φ φ Sub-array RF Chain 2 φ φ Digital (few) Analog (many) ✅ Multiple beams ✅ Fewer RF chains ✅ Good mmWave gain ❌ Less flexible than digital 📍 5G FR2 dominant 🔮 6G: AI-driven dynamic hybrid

Architecture Comparison

AspectAnalogDigitalHybrid
RF Chains1N (= # elements)K (≪ N)
Simultaneous Beams1NK
MU-MIMONot possibleNativeLimited by K
CostLowHighMedium
FlexibilityLowHighMedium-High
Best FormmWave initial accessSub-6 GHzmmWave data

#4Massive MIMO → Ultra-Massive MIMO → XL-MIMO

AspectLTE (4×4)5G Massive MIMO6G Ultra-Massive6G XL-MIMO
Elements4–832–2561024–2048Continuous aperture
Beamwidth~10°~5–10°< 3°Beam FOCUSING
Spatial ModelFar-field onlyFar-field dominantNear-field emergesNear-field DOMINANT
CapabilityCapacity boostCapacity + coverageTbps ratesTheoretical limits
MIMO TypeSU-MIMOSU + MU-MIMOUltra MU-MIMOCell-free

Near-Field vs Far-Field (XL-MIMO)

FAR-FIELD (5G) Planar wavefronts (parallel) Beam = direction only STEERING only NEAR-FIELD (6G / XL-MIMO) Spherical wavefronts (curved) Beam = direction + DISTANCE

Rayleigh Distance: dR = 2D² / λ


#5Precoding Techniques

5.1 Codebook-Based (LTE & 5G NR)

  Codebook (Predefined Beams)
  ┌────────────────────────────────┐
  │  Beam 0: [1,  1,   1,   1]    │ ← Broadside
  │  Beam 1: [1,  j,  -1,  -j]   │ ← Steered right
  │  Beam 2: [1, -j,  -1,   j]   │ ← Steered left
  │  Beam 3: [1, -1,   1,  -1]   │ ← Steered further
  └────────────────────────────────┘
  UE measures → reports PMI = 2 → gNB applies Beam 2

5.2 Non-Codebook (Eigen-Based)

  H = U × Σ × Vᴴ
              ↑  Precoding matrix (V) applied at gNB

  ✅ More accurate   ❌ High feedback overhead
  📍 5G NR non-codebook UL, TDD reciprocity

5.3 Grid of Beams (GoB)

Beam 0 Beam 1 Beam 2 Beam 3 gNB

5.4 6G: AI/ML-Optimized Precoding

Traditional (5G)

Measure → Select from codebook → Apply PMI

Reactive, fixed beam set

AI-Driven (6G)

Sense → Predict channel → Optimize beam weights

Proactive, infinite beam space, sensor + environment data


#6SSB Beam Sweeping — Initial Discovery

The gNB sweeps its SSB across multiple beams sequentially. UE measures SSB-RSRP per beam index and reports the best:

SS Burst Set — One Sweep Cycle Beam 0Slot 0 Beam 1Slot 1 Beam 2Slot 2 Beam 3Slot 3 ··· Beam 63Slot 63 Max SSB Beams: FR1 < 3 GHz → 4 FR1 3-6 GHz → 8 FR2 mmWave → 64 6G THz → 256+ Periodicity: 5, 10, 20, 40, 80, or 160 ms LTE: CRS (wide beam, no per-beam sweep) → 5G: SSB per-beam sweeping
SSB sweep: each slot activates in sequence; the UE reports the beam with the highest RSRP.

#7CSI-RS Beam Refinement — Coarse to Fine

SSB Beams (Coarse)

4–64 wide beams, broadcast periodically

CSI-RS Beams (Fine)

Hundreds of narrow beams, per-UE dynamic

MetricWhat It MeasuresAvailabilityLimitation
SSB-RSRPCoarse beam coverageAlways (broadcast)Doesn't reflect data beam quality
CSI-RSRPFine data beam qualityWhen configuredPer-UE, requires RRC setup

⚠️ Field Trap

In mmWave, a cell can show excellent SSB-RSRP (-83 dBm) but terrible CSI-RSRP due to obstacles blocking the narrow service path. Never judge mmWave on SSB-RSRP alone.


#8Beam Management — P1, P2, P3

P1 — Initial Beam Pair gNB Tx UE Rx gNB sweeps Tx (SSB) • UE sweeps Rx Best (Tx, Rx) pair P2 — gNB Tx Refinement CSI-RS Fixed gNB sweeps fine CSI-RS • UE Rx fixed Best gNB Tx beam P3 — UE Rx Refinement Fixed UE Rx gNB Tx fixed • UE sweeps Rx Optimal UE Rx beam

TCI States — Beam Indication

  TCI State = { Reference Signal ID, Beam Type }
    TCI 1 → { SSB-Index 3, QCL-TypeD }      → Beam aligned with SSB #3
    TCI 2 → { CSI-RS-Resource 7, QCL-TypeD } → Beam aligned with CSI-RS #7
  Signaled via: MAC-CE (semi-static) or DCI (dynamic, per-slot)

LTE had none of this. Beam management was implicit.


#9Beam Failure Recovery (BFR)

Stage 1: DETECTION BLER > threshold → BFI counter → beamFailureInstanceMaxCount Stage 2: NEW BEAM IDENTIFICATION Scan SSBs/CSI-RS → find candidate above threshold Stage 3: RECOVERY REQUEST Dedicated PRACH on preconfigured recovery resources Stage 4: RECOVERY RESPONSE gNB acknowledges → Service resumes on NEW beam Timeline: tens of ms (vs hundreds for handover) KPI Target: BFR Success Rate > 95% (mmWave indoor)
Beam Failure Recovery: Detection → New Beam ID → Recovery Request → Recovery Response.

LTE: Radio Link Failure → full RRC reconnect → seconds. 6G: ML models predict failure before it happens.


#10Reconfigurable Intelligent Surfaces (RIS)

RIS shifts beamforming from the transmitter to the environment itself:

Traditional — Blocked gNB WALL UE ╳╳ RIS-Assisted — Virtual LOS gNB RIS░░░░ UE RIS applies per-element phase shifts → reflected waves combine constructively at UE location Passive • Low power • Walls, ceilings, windows, UAVs
RIS-assisted path: toggle the intelligent surface and move the UE to see the reflected virtual LOS path.
Coverage Extension
Virtual LOS around obstacles
Interference Suppression
Steer nulls at cell edges
Indoor Enhancement
Transparent RIS on windows
Physical Layer Security
Focus to legitimate UE, null eavesdropper
UAV-Mounted RIS
Mobile, repositionable coverage
SWIPT
Simultaneous info + power transfer

#11Holographic Beamforming & CAPA

Discrete Array (5G)

N beams from N discrete elements

Continuous Aperture — CAPA (6G)

▓▓▓▓▓▓▓▓▓▓▓▓▓

Infinite beam shapes, continuous control

Three 6G Concepts


#12Cell-Free Massive MIMO — No More Cell Edges

Cell-Centric (LTE / 5G)

gNB₁ gNB₂ handover zone Cell-edge degradation

Cell-Free (6G)

AP₁ AP₂ AP₃ AP₄ AP₅ UE ALL APs serve this UE No edges • No handovers
Cell-free beamforming: all active APs cooperatively beamform toward the UE. Drag the UE to see dynamic steering.

#13AI/ML-Driven Beam Management

5G NR — Reactive Beam Fails Detect (BFI) Find New Beam Recovery → Interruption 6G — Predictive / Proactive Digital Twin+ Sensor Fusion AI/ML Modelpredicts degradation Pre-switch Beambefore failure ZERO interruption

ISAC — Integrated Sensing and Communication

6G gNB uses the same waveform for both data communication and radar-like sensing. Sensing data feeds beam management: "Person approaching UE path → pre-configure alternative beam."


#14THz Beamforming — 6G Frontier Band

100 GHz – 10 THz → enormous bandwidth → Tbps data rates for holographic video, digital twins, immersive XR.

ChallengeImpactSolution
Molecular absorptionFrequency-specific loss spikes (O₂, H₂O)Frequency-aware beam selection
Extreme beam narrownessSlightest movement breaks linkUltra-massive MIMO (2048+)
Beam squintDirection shifts across wide-band subcarriersTrue-time-delay (not just phase)
Blockage sensitivityHuman hand/leaf/rain = full blockMulti-connectivity + RIS
Weak THz hardwarePAs ~0 dBm, lossy phase shiftersDynamic hybrid sub-arrays

#15Beam Correspondence — FR2 / THz Pitfall

UE must use the same beam direction for Tx as it measured for Rx. If calibration is off:

✅ Correspondence OK

gNB UE Reaches gNB ✅

❌ Correspondence FAILS

gNB UE Signal lost ❌

More severe at FR2/THz — narrower beams mean even small calibration errors cause failure.


#16FR1 vs FR2 — Opposing Beam Dynamics

AspectFR1 (Sub-6 GHz)FR2 (mmWave 24–52 GHz)
BeamwidthWide (30–60°)Very narrow (5–15°)
SSB beams/cell4–832–64
Beam recoveryRareFrequent
Primary roleCapacity boostLink survival
ArchitectureFully digital viableHybrid mandatory
BlockageLow (diffraction)Very high
Measurement trapSSB ≈ data qualitySSB ≠ data quality

#17Complete Evolution — LTE → 5G → 6G

AspectLTE5G NR6G (~2030)
Antenna elements2–832–2561024–2048+ (XL-MIMO)
Beamforming roleCapacity enhancementMandatory at mmWaveBackbone of entire radio
Beam managementImplicit (eNB)P1/P2/P3 frameworkAI-predictive, sub-ms
Reference signalsCRS (cell-wide)SSB + CSI-RS (per-beam)AI-generated, reduced overhead
Beam recoveryRLF (seconds)BFR (tens of ms)Predictive (pre-failure)
PrecodingCodebookCodebook + non-codebookAI/ML + near-field
Spatial modelFar-field onlyFar-field dominantNear-field dominant
EnvironmentPassivePassiveActive (RIS, holographic)
TopologyCell-centricCell-centricCell-free, user-centric
FrequencySub-3 GHzSub-6 + mmWaveSub-6 + mmWave + THz
AI/MLNoneRel-18+Foundational
SensingNoneNoneISAC (radar + comms)

#18Practical Field Considerations

Common Pitfalls

PitfallImpactBest Practice
SSB-RSRP alone in mmWaveFalse confidenceAlways correlate with CSI-RSRP
Ignoring beam correspondenceUL failures, RACH misfireCorrelate SSB-Index with PUCCH-Resource
Too-long SSB periodicityDelayed beam selectionBalance overhead vs speed
No tracking in high-mobilityFrequent beam failuresDoppler-aware beam management
Treating FR1 and FR2 sameWrong expectationsDifferent strategies per band

Blockage Sources at 28 GHz

SourceAttenuationRecovery
Human body20–30 dBBFR (fast beam switch)
Vehicle (metal)30–40 dBHandover or RIS bypass
Foliage10–20 dBPower ramping, beam switch
Rain (heavy)5–10 dB/kmPower margin, beam switch
Tinted glass20–40 dBRIS or outdoor-only
Concrete wall40–60 dBCannot penetrate — RIS/relay needed

#19Further Reading

3GPP Specifications

Research Papers

Podcast

Beamforming Deep Dive — Telecom Leaders Podcast

6G AI-Native MAC Topics

☆ Cell-Free MAC → ☆ ISAC Scheduling → ☆ NTN MAC →