What are the steps of beam failure recovery?

BFR follows a four-step sequence: Step 1 - Candidate Identification: After BFD declares failure, the UE scans all configured candidate beam resources (SSB indices and/or CSI-RS) to find beams above the Qin threshold (L1-RSRP > -128 dBm, L1-SINR > -2 dB). The UE selects the strongest candidate. If no candidate is found, the UE enters contention-based recovery using any available PRACH resource. Step 2 - BFRQ Transmission: The UE transmits a beam failure recovery request (BFRQ) using dedicated PRACH resources (contention-free random access) associated with the selected candidate beam. The preamble index maps to the candidate beam direction. Step 3 - gNB Detection: The gNB monitors all configured BFR-PRACH occasions. Upon detecting the preamble, it identifies the requested beam and prepares a new beam configuration. Step 4 - Response: The gNB transmits a BFR response (BFRS) on PDCCH addressed to the C-RNTI, completing the beam switch.

How long does beam failure recovery take?

BFR latency depends on several factors: Detection: 10-40 ms (BFD timer + BFI count accumulation). Candidate search: 5-20 ms (depends on SSB periodicity, typically 20 ms). PRACH transmission: 1-10 ms (next available PRACH occasion). gNB processing: 3-5 ms (beam reconfiguration). PDCCH response: 1-2 ms. Total: 20-77 ms typical, with 30-50 ms being the most common range. If BFRQ fails (no response within ra-ResponseWindow, typically 10-40 ms), the UE retransmits up to max BFR retransmissions (configured by RRC). Each retry adds 20-40 ms. After all retries exhausted (total ~200 ms), the MAC layer declares beam failure recovery failure and triggers RRC connection re-establishment, which requires full synchronization, security re-authentication, and context restoration, taking 200-500 ms and causing a noticeable service disruption.

What is the difference between BFR and handover?

BFR and handover both change the serving beam but differ fundamentally: BFR is UE-initiated, intra-cell: the UE detects beam failure and autonomously selects a new beam from the same serving cell (same gNB, same cell ID). No core network involvement. MAC layer procedure. Latency: 20-50 ms. Handover is network-initiated, inter-cell: the gNB decides based on measurement reports that the UE should be served by a different cell (potentially different gNB). Involves RRC signaling, core network path switch, and security key derivation. Latency: 30-100 ms. BFR handles transient blockages where the cell still has available beams. Handover handles mobility where the UE moves beyond the cell's coverage. In mmWave deployments, BFR occurs 10-100x more frequently than handover due to the high blockage rate of narrow beams.

5G NR Beam Management

Beam Failure Recovery (BFR)

/beem FAYL-yur ree-KUV-uh-ree/

The 5G NR MAC-layer procedure (TS 38.321) that re-establishes the radio link after beam failure detection. Four steps: 1) candidate beam identification above Q_in (>−128 dBm), 2) BFRQ via dedicated PRACH, 3) gNB detection and reconfiguration, 4) BFRS response on PDCCH. Total latency: 20 to 50 ms. If all retries fail, RRC re-establishment triggers (200 to 500 ms disruption). BFR occurs 10 to 100x more frequently than handover in mmWave deployments.

Latency: 20–50 ms

Q_in: >−128 dBm

Fallback: RRC re-est.

Understanding Beam Failure Recovery

BFR is the second half of the 5G NR beam resilience mechanism, activated after BFD declares that the serving beam is no longer usable. The procedure must find an alternative beam and switch to it fast enough that higher-layer protocols (TCP, RTP) do not trigger retransmissions or timeouts. The 20 to 50 ms recovery window is typically below the TCP retransmission timeout (200 ms) and just within the VoNR jitter buffer (40 to 60 ms).

The candidate beam search phase is critical: the UE must find a beam that satisfies Q_in (approximately 1% BLER, corresponding to L1-RSRP > −128 dBm). This is a significantly higher quality bar than Q_out (10% BLER), ensuring that the new beam provides reliable service rather than immediately re-failing.

BFR Timing Budget

Nominal BFR Timeline:
BFD detection: 10–40 ms
Candidate search: 5–20 ms (SSB period)
PRACH transmission: 1–10 ms
gNB processing: 3–5 ms
PDCCH response: 1–2 ms
Total: 20–77 ms (typical 30–50 ms)

Failure Escalation:
Each BFRQ retry: +20–40 ms
Max retries exhausted: ~200 ms
RRC re-establishment: +200–500 ms
Total worst case: ~700 ms

BFR vs. Handover Comparison

Feature	Beam Failure Recovery	Handover
Trigger	UE-initiated (BFD)	Network-initiated (A3)
Scope	Intra-cell (same gNB)	Inter-cell (new gNB)
Signaling	MAC (PRACH + PDCCH)	RRC + core network
Latency	20–50 ms	30–100 ms
Frequency (FR2)	10–100x per minute	~1x per minute
Core network	Not involved	Path switch required

Common Questions

Frequently Asked Questions

BFR procedure steps?

1) Candidate beam scan (RSRP > −128 dBm). 2) BFRQ via dedicated PRACH (preamble maps to candidate). 3) gNB detects, reconfigures. 4) BFRS on PDCCH. If no candidate found, contention-based RA fallback.

How long does recovery take?

BFD: 10 to 40 ms. Candidate: 5 to 20 ms. PRACH: 1 to 10 ms. Response: 3 to 7 ms. Total: 20 to 77 ms. Retries: +20 to 40 ms each. RRC re-establishment fallback: 200 to 500 ms.

BFR vs. handover?

BFR: UE-initiated, intra-cell, MAC layer, 20 to 50 ms, no core network. Handover: network-initiated, inter-cell, RRC + core, 30 to 100 ms. BFR 10 to 100x more frequent in FR2 (mmWave blockage).

Related Terms

← Beam Failure Detection Beamformer IC →

5G NR Systems

Precision RF Components

RF Essentials provides precision terminations and custom waveguide assemblies for 5G NR beam management test systems and OTA measurement equipment.

Request a Quote