5G NR Explained: Numerology, BWP, HARQ, and Frame Structure Interview Prep

5G NR Numerology (subcarrier spacing, slot duration, channel bandwidth)

Sample interview questions

1. Why does an NR macro cell with ~2 µs maximum excess delay prefer 30 kHz numerology over 60 kHz for downlink data?
   • A. At 60 kHz the slot becomes 250 µs, which falls below the FR1 TTI floor that 5G NR Release 15 requires for downlink scheduling.
   • B. At 60 kHz the cyclic prefix shrinks to about 1.2 µs, below the ~2 µs maximum excess delay, so multipath leaks across symbols and breaks subcarrier orthogonality, while the 30 kHz CP of ~2.3 µs still covers it. ✓
   • C. At 60 kHz the HARQ-ACK K1 indicator is undefined, so the UE cannot acknowledge any PDSCH on this numerology in FR1 deployments.
   • D. At 60 kHz the gNB is forced into mini-slot scheduling, which prevents URLLC traffic from sharing the same carrier with eMBB users.

   Option B is correct because CP duration scales inversely with subcarrier spacing: 15 kHz gives roughly 4.7 µs of CP, 30 kHz gives 2.3 µs, and 60 kHz gives 1.2 µs. When the channel's maximum excess delay (~2 µs here, i.e. the delay of the last significant multipath tap) exceeds the 60 kHz CP but still fits inside the 30 kHz CP, the 60 kHz numerology lets multipath copies of one OFDM symbol leak into the next, breaking the per-bin equalization assumption and creating irreducible ISI, while 30 kHz stays clean. Option A is incorrect because 5G NR has no 1 ms TTI floor — sub-millisecond slot durations are deliberately supported at higher numerology and are part of the URLLC design intent. Option C is incorrect because the K1 PDCCH-to-PUCCH timing indicator is defined for every FR1 numerology in Release 15 (TS 38.213); it is not restricted to 15 and 30 kHz. Option D is incorrect because mini-slot scheduling is decoupled from SCS: 2/4/7-symbol mini-slots run on any numerology, and 30 kHz with mini-slots routinely carries URLLC alongside eMBB on the same carrier.

2. Why does 5G NR HARQ timing use a dynamic K1 indicator in DCI instead of LTE FDD's fixed n+4 HARQ-ACK timing?
   • A. K1 supports a fixed n+4 timing inherited from LTE so legacy HARQ scheduling continues to work transparently on FR1 NR carriers with no spec-level changes.
   • B. K1 forces HARQ-ACK to be sent on the next subframe boundary, mirroring LTE timing assumptions to preserve backward compatibility with eNB-class scheduling.
   • C. K1 lets the gNB choose the HARQ-ACK slot per transmission, decoupling ACK timing from numerology so self-contained slots can carry both PDSCH and its ACK at higher SCS. ✓
   • D. K1 is a CSI report timer used by beam management procedures and the HARQ-ACK slot stays fixed at n+4 across both FR1 and FR2 numerology configurations.

   Option C is correct because NR slot duration scales with numerology (1 ms at μ=0 down to 125 µs at μ=3), so a fixed n+4 rule would force the ACK eight times further away in wall-clock time at low SCS than at high SCS. The K1 indicator (TS 38.213 §9.2.3) lets the gNB place HARQ-ACK in the same slot as the PDSCH when the DL/UL configuration permits, enabling self-contained slot operation and tight URLLC closed-loop latency. Option A is incorrect because the misconception that K1 retains a fixed n+4 timing is wrong — Release 15 defines K1 as a per-DCI dynamic offset across all numerologies, not a legacy n+4 reuse. Option B is incorrect because HARQ-ACK timing in NR is decoupled from subframe boundaries and from the n+4 LTE rule; the misconception that NR preserves LTE-style timing for "backward compatibility" misreads the design intent. Option D is incorrect because K1 is a PDSCH-to-HARQ-ACK timing offset, not a CSI report timer, and HARQ-ACK is not fixed at n+4 in NR — the dynamic K1 field is the timing source.

3. Why does 5G NR use Polar codes for PDCCH and LDPC codes for PDSCH rather than one code family for both channels?
   • A. Polar codes outperform LDPC across every block length and rate, but Polar decoders cannot run at the throughput PDSCH requires, so PDSCH falls back to LDPC for hardware reasons alone.
   • B. Polar coding was carried forward from LTE control channels for backward compatibility; LDPC was introduced for PDSCH because turbo codes had been deprecated by IEEE before the NR study item.
   • C. Polar and LDPC are mathematically equivalent under SCL decoding; the split between PDCCH and PDSCH is a 3GPP marketing choice rather than a coding-theoretic decision.
   • D. Polar codes excel at short, low-rate control blocks with maximum-likelihood-class performance under SCL decoding; LDPC codes scale efficiently to large transport blocks with parallel iterative decoding for high-throughput PDSCH. ✓

   Option D is correct because the two code families occupy different operating points: Polar codes (with successive-cancellation list decoding plus CRC-aided list selection) achieve near-ML performance on short, low-rate blocks of the kind that PDCCH carries (tens of bits), while LDPC codes have a quasi-cyclic structure designed for long transport blocks and parallel iterative decoding at multi-Gbps throughput — the natural fit for PDSCH (TS 38.212 §5.3.1/5.3.2). Option A is incorrect because Polar codes do not beat LDPC at every block length and rate; LDPC performs better as block length grows, which is one of the reasons it was selected for PDSCH. Option B is incorrect because LTE used convolutional codes for PDCCH, not Polar, and turbo codes (used in LTE PDSCH) were dropped in NR because LDPC offered higher throughput and better high-rate performance, not because of any IEEE deprecation. Option C is incorrect because Polar and LDPC are very different code families with distinct decoder structures; the PDCCH-versus-PDSCH split is a coding-theoretic optimization for block-length regime, not a marketing label.

4. In an FR2 cell sweeping up to 64 SSBs in a 5 ms SS burst set (default 20 ms periodicity), how does the UE bind its RACH preamble to the best SSB so the gNB picks the right MSG2 beam?
   • A. The UE encodes the SSB index in the upper 4 bits of the PRACH preamble sequence index; the gNB decodes the sequence and extracts the index from those bits.
   • B. rach-ConfigCommon maps SSBs to RACH occasions and/or preamble subsets; the UE transmits on the resource mapped to its strongest SSB, and the gNB infers the preferred SSB beam from that mapped PRACH resource. ✓
   • C. The UE sends a dedicated SR on PUCCH before PRACH, with the SR resource ID encoding the chosen SSB index for the gNB receiver.
   • D. The gNB measures a UL SRS sounding burst that precedes PRACH and selects the MSG2 beam from the strongest SRS direction, so no SSB-to-preamble binding is required.

   Option B is correct because rach-ConfigCommon (signaled in SIB1) defines an SSB-to-RACH-occasion / SSB-to-preamble mapping (TS 38.213 §8.1). The UE picks a preamble from the subset associated with its strongest SSB; when the gNB detects the preamble on a specific RACH occasion, it learns which SSB beam the UE measured as best and points MSG2 (RAR on PDSCH) in that beam direction. Option A is incorrect because the PRACH preamble is a Zadoff-Chu sequence indexed by root and cyclic shift; there are no "upper 4 bits" carrying SSB index — the binding flows entirely through the occasion/preamble subset mapping. Option C is incorrect because SR-on-PUCCH is a CONNECTED-mode mechanism; an idle UE has no PUCCH resource and cannot send SR before PRACH. Option D is incorrect because the UE has no UL grant before MSG2, so it cannot send SRS at this point of the initial-access procedure.

5. How does a URLLC mini-slot preempt an ongoing eMBB PDSCH on the same carrier, and how is the eMBB UE told to drop the affected REs?
   • A. The gNB tears down the eMBB transport block via RRC Release, schedules URLLC on the freed slots, and the eMBB UE re-establishes the bearer from scratch on a new BWP.
   • B. The URLLC mini-slot rides orthogonal subcarriers reserved by SIB1, so it does not overlap eMBB resources and a preemption indication is not transmitted at all.
   • C. The gNB schedules the URLLC mini-slot inside the eMBB PDSCH and sends DCI format 2_1 on a group-common PDCCH so the eMBB UE marks the preempted REs as erased before LDPC decoding. ✓
   • D. The eMBB UE detects preemption autonomously from CRC failure patterns and drops the entire transport block without any explicit signaling from the gNB.

   Option C is correct because 5G NR defines DCI format 2_1 (TS 38.213 §11.2) as a group-common PDCCH message scrambled with INT-RNTI. The eMBB UE monitors INT-RNTI on its serving cell; when the gNB has stolen part of the UE's PDSCH for a URLLC mini-slot, DCI 2_1 carries a bitmap of preempted time-frequency regions. The eMBB UE then marks those REs as erasures (LLR=0) before LDPC decoding, which dramatically improves recovery rates over treating them as random noise. Option A is incorrect because RRC Release is a connection-teardown procedure with seconds-class latency, while URLLC preemption is a sub-millisecond mechanism inside an active PDSCH. Option B is incorrect because there is no SIB1-reserved orthogonal URLLC band; mini-slot scheduling overlaps eMBB resources by design, and that overlap is the whole point of preemption. Option D is incorrect because without explicit preemption indication the eMBB UE has no way to distinguish stolen REs from low-SNR noise, and CRC-only feedback wastes the soft information already gathered on the unaffected REs.

6. In a 5G NR CU/DU split gNB, which protocol-layer functions land in the CU versus the DU, and how does F1-AP differ from F1-U?
   • A. PHY and MAC run in the CU; RLC, PDCP, and RRC run in the DU. F1-AP carries baseband IQ samples and F1-U carries radio statistics for OAM telemetry.
   • B. All protocol layers run in the CU; the DU is a remote radio head doing DAC, ADC, and PA functions. F1-AP and F1-U both carry IQ samples over eCPRI fronthaul.
   • C. PDCP runs in the DU and RRC runs in the CU; F1-AP carries the RLC ARQ state across the split, and F1-U is reserved for sidelink relay traffic between adjacent DUs.
   • D. RRC, SDAP, and PDCP run in the CU; RLC, MAC, and PHY run in the DU. F1-AP carries control-plane signaling (UE context, mobility, bearer setup) over SCTP, while F1-U carries PDCP user-plane PDUs over GTP-U tunnels. ✓

   Option D is correct because the 3GPP F1 split (TS 38.401) places higher layers (RRC, SDAP, and PDCP) in the CU and lower layers (RLC, MAC, PHY) in the DU. F1-AP (TS 38.473) is the control-plane protocol over SCTP that carries UE context setup, bearer configuration, mobility commands, and CU-DU coordination; F1-U is the user-plane protocol over GTP-U tunnels that carries PDCP PDUs from CU to DU for downlink (and the reverse for uplink). This split centralizes RRC mobility control while keeping latency-sensitive RLC/MAC scheduling at the DU. Option A is incorrect because PHY and MAC cannot move into the CU — the fronthaul latency budget (sub-millisecond HARQ-ACK loop) makes that impractical. Option B is incorrect because that describes a Lower-Layer Split (option 7 or 8) on eCPRI, not the F1 split; the CU/DU split places PDCP at the boundary, not the PHY. Option C is incorrect because the standardized split puts PDCP in the CU rather than the DU, and F1-U carries PDCP PDUs, not sidelink traffic.

Frequently asked questions

What is the difference between 5G NR sub-6 GHz and mmWave?

Sub-6 GHz (FR1, roughly 410 MHz to 7.125 GHz) trades raw bandwidth for propagation: signals diffract around obstacles and penetrate buildings, so coverage cells are large and mobility-friendly, but per-user data rates are bounded by 100-MHz-class channels. mmWave (FR2, 24.25-52.6 GHz and beyond) inverts that tradeoff: hundreds of MHz of contiguous spectrum enable multi-Gbps peaks, but path loss, oxygen absorption near 60 GHz, and brittle line-of-sight links force dense beamforming, narrow cell radii, and aggressive beam-management procedures (SSB sweeping, beam refinement, beam failure recovery). Interviewers expect you to articulate this physical-layer / system-design coupling, not just quote frequency ranges.

How do Bandwidth Parts (BWP) work in 5G NR?

A BWP is a contiguous subset of the carrier bandwidth that a UE actively monitors and transmits on at a given time. The gNB configures up to four downlink and four uplink BWPs per UE; one is active at a time. BWPs decouple UE RF capability from carrier width — a low-power UE can sit on a narrow BWP for long stretches and only widen when traffic warrants. They also enable numerology switching (different subcarrier spacings on different BWPs), reduced UE power consumption (narrow active BWP idles more cleanly), and forward compatibility for use cases that need different waveform parameters. Switching is triggered via DCI or expiry timers, not just RRC reconfiguration.

What is HARQ and why does it matter in 5G NR?

HARQ (Hybrid ARQ) combines forward error correction with retransmissions: when a transport block fails CRC, the receiver stores the soft bits, requests a retransmission, and combines incoming redundancy versions with the buffered copy via Chase combining or incremental redundancy. In 5G NR, HARQ is asynchronous and adaptive on the downlink, runs over up to 16 parallel processes per carrier, and is critical to meeting URLLC latency targets (~1 ms one-way) because a fast NACK plus a finely-tuned redundancy-version schedule recovers errors without invoking RLC ARQ. Interviewers probe whether you understand the soft-buffer sizing, the role of the redundancy version, and how HARQ feedback timing interacts with self-contained slot structure.

How does 5G NR numerology differ from LTE?

LTE locked subcarrier spacing at 15 kHz with a fixed 1-ms subframe and 0.5-ms slot. 5G NR generalizes this: subcarrier spacing scales as 15 × 2^μ for μ in {0,1,2,3,4}, giving 15, 30, 60, 120, and 240 kHz. Higher μ shrinks the OFDM symbol and the slot (a slot has 14 symbols under normal CP, 12 under extended CP), enabling sub-millisecond TTIs at FR2 and tight feedback loops for URLLC. Larger SCS shrinks the absolute cyclic prefix (reducing delay-spread tolerance) and improves resilience against phase noise, while the fractional CP overhead stays roughly constant — that is why mmWave deployments lean on 120 or 240 kHz. The flip side: low μ tolerates large delay spreads (suitable for outdoor sub-6 macro). Picking μ is a deployment-driven physical-layer tradeoff, not a free knob.

What role does beam management play in 5G NR?

Beam management is the procedure stack that lets a gNB and UE acquire, maintain, refine, and recover narrow analog or hybrid beams. It builds on SSB-based initial access (the UE sweeps SSB beams and reports the strongest), CSI-RS-based refinement (finer probing for data-channel beams), beam reporting via L1-RSRP, beam indication via TCI states in DCI, and beam failure recovery (BFR) when the serving beam fails consecutive thresholds. At mmWave it is mandatory; at sub-6 with massive MIMO it is still used to manage SU/MU-MIMO precoders. Interview questions often probe the timeline (P1/P2/P3 procedures, SSB periodicity) and what happens when the chosen beam drops mid-transmission.

What is CORESET and how is it used in 5G NR?

A CORESET (Control Resource Set) is a configured time-frequency region — a few OFDM symbols by a set of contiguous PRBs — where a UE blindly decodes PDCCH candidates to find downlink control information. Unlike the LTE control region, which spanned the first 1-3 symbols across the entire carrier, NR CORESETs are localized and per-UE configurable, which lets the gNB schedule control flexibly inside a slot, place CORESET#0 to bootstrap RRC connection, and overlay multiple CORESETs with different search-space configurations. Knowing CORESET layout is essential to reasoning about scheduling latency, blind decoding load, and how DCI is multiplexed with data.

What is the difference between SA and NSA 5G NR deployments?

Non-Standalone (NSA) 5G NR keeps the EPC (4G core) as the control-plane anchor and adds an NR carrier in dual-connectivity alongside an LTE anchor — Option 3x in 3GPP terminology, also called EN-DC. The UE uses the eNB for RRC and mobility while the gNB (acting as the secondary cell group, SCG) carries the high-throughput user-plane data over a split or duplicate bearer. Standalone (SA) 5G NR replaces the EPC with the 5G Core (5GC) and uses the gNB for both control and user planes through the NG-RAN interfaces (N2 to AMF, N3 to UPF); the UE registers with AMF over N1 and there is no LTE anchor. Operators usually deploy NSA first because it lets them light up 5G data over an existing LTE footprint without rebuilding the core; the migration to SA unlocks URLLC, network slicing (see /topics/network-slicing-explained), and service-based architecture features that EPC cannot provide. Sample question oQ2cA6Fh on the CU/DU split assumes the SA architecture — F1-AP only exists once the gNB owns its own RRC, which is not the case under EN-DC.

How does 5G NR multiplex URLLC and eMBB traffic on the same carrier?

eMBB and URLLC have radically different reliability and latency targets — eMBB tolerates BLER around 10^-1 with retransmission, while URLLC needs BLER around 10^-5 with one-shot delivery in 1-2 ms. NR multiplexes both on the same carrier through three layered mechanisms. First, dedicated MCS Table 3 (qLowSE / low-spectral-efficiency table) provides URLLC-grade BLER for the URLLC PDSCH while eMBB uses MCS Tables 1 or 2. Second, mini-slot scheduling (2/4/7 symbols) lets a URLLC transport block start mid-slot without waiting for the next slot boundary. Third, pre-emption indication on DCI format 2_1 (scrambled with INT-RNTI) lets the gNB steal symbols mid-eMBB-PDSCH for an urgent URLLC mini-slot and tell the eMBB UE which REs to erase before LDPC decoding. Sample question iJ7zX3Lp drills into the pre-emption mechanism in detail.

How does a UE find and access an NR cell from RRC_IDLE?

A UE acquires an NR cell through a fixed sequence: detect the SSB (PSS, SSS, PBCH+MIB), use the MIB to find CORESET-0, decode SIB1 on the PDCCH-scheduled PDSCH, then run the RACH procedure (4-step by default, 2-step when applicable). The RACH MSG2 carries timing-advance and a temporary C-RNTI, MSG3 carries the UE identity, and MSG4 confirms contention resolution. See /topics/nr-initial-access-ssb-rach for the full initial-access flow including contention-free RACH used in handover, and /topics/nr-bwp-and-numerology for the BWP / numerology layer the UE operates on after access.

Related topics