Telephony & Wireless

5G NR & 5G Core

How 5G really works end to end: the New Radio (NR) air interface, the cloud-native 5G Core (5GC), and the procedures that tie them together, from registration to voice calls. It is written for interviews, so every section ends with how interviewers probe it and the scenario questions that expose real understanding.

~110 min read 0 interview questions
In 30 seconds
  • NSA (Option 3x) adds an NR carrier to an LTE anchor on the old EPC core via EN-DC; SA (Option 2) connects NR directly to the new 5GC, which is what unlocks VoNR, slicing and RRC_INACTIVE.
  • The 5GC is a service-based architecture: AMF (mobility), SMF (sessions), UPF (user plane), AUSF/UDM (auth and subscription), PCF (policy), NRF (discovery), NSSF (slices), NEF (exposure), talking HTTP/2 over the SBI.
  • 5G QoS uses QoS flows (QFI, 5QI) inside a PDU session; the new SDAP layer maps flows onto radio bearers.
  • NR is flexible: numerology µ sets subcarrier spacing 15 × 2µ kHz, bandwidth parts, beam-based access with SSBs, FR1 sub-6 GHz and FR2 mmWave.
  • Voice on SA is either native VoNR (5QI 1 media, 5QI 5 SIP) or EPS fallback to VoLTE over N26; neither is SRVCC. IMS roaming is usually home-routed.

Why 5G: generations and use cases

Each mobile generation solved the biggest limitation of the one before it. 5G is the first generation designed from day one for three very different kinds of traffic instead of "faster phones only".

GenerationHeadline technologyWhat it addedVoice
1G (1980s)Analog FDMA (AMPS)Mobile voice at allAnalog circuit
2G (1990s)GSM / TDMA, CDMA IS-95Digital voice, SMS, encryption, SIM; GPRS/EDGE packet dataCircuit-switched (CS)
3G (2000s)UMTS/WCDMA, HSPA, CDMA2000Real mobile internet, packet coreCS voice, PS data
4G (2010s)LTE / LTE-A, OFDMA, EPCAll-IP flat core, MIMO, carrier aggregationVoLTE over IMS (CSFB as fallback)
5G (2019+)NR, 5GC service-based coreFlexible numerology, mmWave, massive MIMO, slicing, edge, URLLCVoNR over IMS, or EPS fallback to VoLTE

The three 5G service families

eMBB

Enhanced Mobile Broadband: very high throughput and capacity. Targets of 20 Gbps peak downlink, 10 Gbps peak uplink, around 100 Mbps user-experienced rate. Video, AR/VR, fixed wireless access.

URLLC

Ultra-Reliable Low-Latency Communication: around 1 ms user-plane latency with 99.999% reliability. Factory automation, remote control, V2X safety.

mMTC

Massive Machine-Type Communication: up to 1 million devices per km², long battery life, low cost. Sensors and meters (served in practice by NB-IoT/LTE-M and later NR RedCap).

These targets come from the ITU IMT-2020 requirements. No single deployment hits all of them at once; they define the design envelope. Other targets include mobility up to 500 km/h and much better energy efficiency per bit.

What is genuinely new versus LTE

  • Flexible numerology instead of LTE's fixed 15 kHz subcarrier spacing and 1 ms TTI.
  • Lean carrier: no always-on cell-specific reference signal (CRS); reference signals are sent only when needed, which saves energy and reduces interference.
  • Beam-based design from the start, including initial access, which makes mmWave usable.
  • New channel coding: LDPC for data channels, Polar codes for control channels (LTE used Turbo and convolutional codes).
  • Wide carriers: up to 100 MHz in FR1 and 400 MHz in FR2, with bandwidth parts so a UE does not need to use all of it.
  • Cloud-native core with control/user plane separation, a service-based interface, slicing and edge UPFs.
  • New RRC state RRC_INACTIVE and a new user-plane layer SDAP.
Analogy

Think of 4G as a very good highway built for cars. 5G is a transport authority that runs one road network but can open a wide express lane for trucks carrying bulk cargo (eMBB), a siren-cleared emergency lane with guaranteed arrival times (URLLC), and a vast grid of cheap bicycle paths for millions of slow riders (mMTC). The express lane is eMBB throughput, the emergency lane is URLLC latency and reliability, and the bicycle grid is mMTC device density, all built on shared radio and core infrastructure.

Interview angle A common opener is "What is 5G and why do we need it?" Weak answers say "faster". A strong answer names the three service families with rough numbers, then lists two or three architectural changes that enable them (numerology, beamforming, service-based core, slicing) and notes that most early launches were NSA and only delivered eMBB.

NSA vs SA deployment options

3GPP defined several "options" for combining LTE, NR, the EPC and the 5GC. Operators mostly used two: Option 3 family (NSA) first, then Option 2 (SA).

OptionCoreMaster (anchor) RANSecondary RANName
1EPCLTEnoneLegacy LTE
25GCNRnoneSA
3 / 3a / 3xEPCLTE (eNB)NR (en-gNB)NSA, EN-DC
4 / 4a5GCNReLTE (ng-eNB)NE-DC
55GCeLTE (ng-eNB)noneLTE on 5GC
7 / 7a / 7x5GCeLTE (ng-eNB)NRNGEN-DC

NSA (Option 3 family) and EN-DC

In NSA the UE is attached to the EPC through LTE. LTE carries all control signalling (RRC and NAS). NR is added as extra capacity through EN-DC (E-UTRA NR Dual Connectivity).

  • MN / MCG: the LTE eNB is the Master Node and serves the Master Cell Group (PCell plus any LTE SCells).
  • SN / SCG: the NR en-gNB is the Secondary Node and serves the Secondary Cell Group (PSCell plus NR SCells).
  • X2 connects eNB and en-gNB (X2-C for control, X2-U for user data). The en-gNB has no NG interface to a 5GC.
  • The EPC needs upgrades: the MME must understand the UE's "DCNR" capability, enforce the subscription restriction "NR as secondary RAT not allowed", and support higher bit rates (extended APN-AMBR).

Bearer types and where the split happens

VariantS1-U user plane from S-GW toSplit pointNotes
Option 3eNB onlyeNB PDCP (MCG split bearer)Needs a strong eNB and big X2 backhaul; early deployments.
Option 3aeNB and gNB separatelyIn the core (per bearer)No RAN split; each bearer goes either to LTE or NR.
Option 3xgNB (and eNB for some bearers)gNB PDCP (SCG split bearer)Most common; NR carries the load, LTE helps via X2-U.

With a split bearer, one PDCP entity feeds two RLC legs (LTE and NR). In uplink, the network configures a primary path and a threshold (ul-DataSplitThreshold): below it the UE sends only on the primary path; above it the UE may use both legs.

How NR gets added (SgNB Addition)

  1. LTE attach The UE attaches on LTE, advertising EN-DC capability and band combinations. The eNB broadcasts upperLayerIndication-r15 in SIB2 so the UE knows NR is available in the area.
  2. B1 measurement The eNB configures an inter-RAT NR measurement. When an NR cell exceeds the B1 threshold, the UE sends a measurement report.
  3. SgNB Addition Request The eNB asks the en-gNB over X2. The gNB admits the UE and returns an NR RRC configuration (SCG config) in the Addition Request Acknowledge.
  4. RRC reconfiguration The eNB sends an LTE RRCConnectionReconfiguration with the embedded NR configuration. The UE applies it and replies Complete; the eNB sends SgNB Reconfiguration Complete to the gNB.
  5. Random access on PSCell The UE performs RACH on the NR PSCell and syncs to the SSB beam.
  6. Path switch For Option 3x, the eNB sends an E-RAB Modification Indication to the MME so the S-GW moves the S1-U tunnel to the gNB.
UE            eNB (MN)          en-gNB (SN)         MME / S-GW
 |  LTE attach   |                   |                     |
 |<------------>|<------------------------------------>  |
 |  MeasReport B1|                   |                     |
 |------------->| SgNB Addition Req |                     |
 |               |----------------->|                     |
 |               | SgNB Add Req Ack  |                     |
 |               |<-----------------| (NR SCG config)     |
 | RRCConnReconf (nr-Config)         |                     |
 |<-------------|                   |                     |
 | RRCConnReconfComplete             |                     |
 |------------->| SgNB Reconf Compl |                     |
 |               |----------------->|                     |
 |   RACH on NR PSCell               |                     |
 |--------------------------------->|                     |
 |               |  E-RAB Modification Indication (3x)     |
 |               |--------------------------------------->|
 |               |           S1-U now terminates at gNB    |

SCG failure

If NR fails (radio link failure on the SCG, T310 expiry, reconfiguration failure, sync failure) the UE does not drop the connection. It keeps LTE, suspends the SCG and sends SCGFailureInformationNR to the eNB, which can release or re-add the SN. Repeated SCG failures cause ping-pong adds and releases, a very common field issue.

SA (Option 2)

In SA the gNB connects to the 5GC over NG (N2 for control to the AMF, N3 for user plane to the UPF). NR carries RRC and NAS itself. SA is required for native 5G features: network slicing, RRC_INACTIVE, VoNR, URLLC, edge UPFs and SSC modes.

NSA (Option 3x)

  • Core: EPC (upgraded)
  • Control plane: LTE anchor
  • Voice: VoLTE on LTE
  • Fast to launch, reuses LTE coverage
  • Uplink often limited by LTE anchor
  • Higher UE power (two radios active)
  • No slicing, no RRC_INACTIVE

SA (Option 2)

  • Core: 5GC (service-based)
  • Control plane: NR itself
  • Voice: VoNR or EPS fallback
  • Needs new core and good NR coverage
  • Lower latency, single radio
  • Slicing, RRC_INACTIVE, SSC modes, edge
  • The target end state
Analogy

NSA is like adding a fast new express train line that only runs if you first check in at the old main station: the old station (LTE and EPC) still issues your ticket and keeps your record, and the new line (NR) just carries you faster. SA is a brand-new station with its own ticket office. In the real system the "ticket office" is RRC and NAS signalling: on NSA it stays on LTE and the EPC, on SA it moves to NR and the 5GC.

Common pitfall Saying "NSA uses the 5G core". It does not; Option 3 uses the EPC. Also do not confuse Option 3 (split in the eNB) with Option 3x (split in the gNB).
Interview angle Expect "Where does NR get added in NSA?" Walk through LTE attach, B1 report, SgNB Addition over X2, RRC reconfiguration with NR config, RACH on PSCell and the E-RAB modification. Bonus points for mentioning MCG vs SCG, split bearer, uplink primary path and SCG failure handling.

5G Core service-based architecture

The 5GC replaces the EPC's big boxes with smaller network functions (NFs). Control-plane NFs expose services as REST-style APIs (HTTP/2 + JSON, defined in OpenAPI, usually over TLS) on a common Service-Based Interface (SBI). Any NF can be a consumer or producer, and they find each other through the NRF. NFs are designed to run as cloud-native microservices on containers.

The main network functions

NFRoleClosest EPC equivalent
AMF Access and Mobility ManagementTerminates N1 NAS and N2; registration, reachability, paging, mobility, NAS security; relays session messages to the SMFMME (mobility part)
SMF Session ManagementPDU session create/modify/release, IP address allocation, UPF selection and control over N4, QoS rule enforcementMME session part, SGW-C, PGW-C
UPF User Plane FunctionPacket routing and forwarding, QoS marking, usage reporting, anchor point for mobility; can be placed at the edgeSGW-U, PGW-U
AUSF Authentication ServerRuns 5G-AKA / EAP-AKA' authentication for the home networkPart of HSS/MME
UDM / UDRUDM: subscriber data management, authentication vectors, SUCI deconcealment (SIDF). UDR: the database behind itHSS
PCF Policy ControlAccess/mobility policy, session policy (QoS, charging rules), UE policy such as URSPPCRF
NRF NF RepositoryRegistry and discovery: NFs register their profile, consumers query it to find a producer instanceDNS-based node selection
NSSF Network Slice SelectionDecides the Allowed NSSAI and the AMF set that can serve the requested slicesNew
NEF Network ExposureSecurely exposes network capabilities and events to external application functionsSCEF
AF Application FunctionAn application (for example the IMS P-CSCF) that asks for policy or QoSAF over Rx

Other NFs you may hear about: SCP (Service Communication Proxy, routes SBI traffic, like a service mesh), SEPP (Security Edge Protection Proxy, secures roaming signalling on N32), CHF (charging), NWDAF (network data analytics), BSF (binding support, finds which PCF serves a session), N3IWF (untrusted non-3GPP access such as Wi-Fi, the 5G version of the ePDG), UDSF (unstructured data storage for stateless NFs).

Reference points (interfaces)

InterfaceBetweenProtocol / purpose
N1UE and AMFNAS (5GMM and 5GSM), carried transparently through the gNB
N2gNB and AMFNGAP over SCTP: UE context, PDU session resources, handover, paging
N3gNB and UPFGTP-U user-plane tunnel
N4SMF and UPFPFCP: installs packet detection, forwarding, QoS and usage rules
N5AF (for example P-CSCF) and PCFPolicy authorisation (5G version of Rx)
N6UPF and Data NetworkPlain IP to the internet, IMS or enterprise network
N7SMF and PCFSession management policy
N8AMF and UDMSubscription and registration data
N9UPF and UPFGTP-U between intermediate UPF and anchor UPF
N10SMF and UDMSession subscription data
N11AMF and SMFSession management message relay
N12 / N13AMF and AUSF / AUSF and UDMAuthentication
N15AMF and PCFAccess and mobility policy, UE policy delivery
N16V-SMF and H-SMFHome-routed roaming session control
N22AMF and NSSFSlice selection
N26AMF and MMEInterworking with EPC for seamless 5GS/EPS mobility
XngNB and gNBXnAP handover and dual connectivity; Xn-U data forwarding

In the SBA view, N8, N10, N11 and the others are simply service calls such as Namf_Communication, Nsmf_PDUSession, Nudm_SDM, Npcf_SMPolicyControl. N1, N2, N3, N4, N6, N9 and Xn remain classic point-to-point interfaces.

             NSSF   NEF   NRF   PCF   UDM   AUSF   AF
               |     |     |     |     |     |     |
  ============ Service-Based Interface (HTTP/2, JSON) ============
                     |                    |
                    AMF ------ N11 ------ SMF
                   /  |                    |
               N1 /   | N2                 | N4 (PFCP)
                 /    |                    |
   UE ~~~ radio ~~~ gNB ------ N3 ------- UPF ------ N6 ------ Data Network
                                (GTP-U)          (internet, IMS)

Design principles

  • CUPS (Control and User Plane Separation): SMF controls, UPF forwards. UPFs scale independently and can sit at the network edge for MEC with low latency.
  • AMF / SMF split: mobility and session logic are separate, so one UE can have several PDU sessions handled by different SMFs, possibly in different slices.
  • Stateless NFs where possible, keeping state in UDSF/UDR so instances can fail over.
  • Access-agnostic core: the same AMF serves 3GPP and non-3GPP (Wi-Fi via N3IWF) access.
  • Discovery over configuration: an AMF asks the NRF for an SMF that supports a given DNN and S-NSSAI instead of using static tables.

Roaming: home-routed versus local breakout

A roaming PDU session is either home-routed (HR) or local breakout (LBO). IMS voice almost always uses HR, the 5G counterpart of S8HR (see IMS roaming).

  • Home-routed: the visited network has a V-AMF, V-SMF and usually a V-UPF; the home network keeps the H-SMF and H-UPF (the IP anchor). V-SMF and H-SMF talk over N16; V-UPF and H-UPF over N9. The IMS P-CSCF stays at home. Media trombones; emergency, visited LI and a local ATCF have the same S8HR drawbacks.
  • LBO: visited SMF and UPF only, N6 in the VPLMN. Faster media, local emergency and LI, but every partner needs a roaming interconnect. Rare for the IMS DNN.
  • Control-plane SBI between operators is protected by SEPP on N32, not by Diameter like LTE roaming.

RAN split: CU and DU

The gNB itself can be split into a Central Unit (CU: RRC, SDAP, PDCP) and Distributed Units (DU: RLC, MAC, high PHY) connected by F1. The CU can be further split into CU-CP and CU-UP over E1. Open RAN (O-RAN) standardises these splits so different vendors can be mixed.

Analogy

The EPC is like a large department store where each counter does many jobs. The 5GC is a food court of specialised stalls with a central directory board: the receptionist (AMF) greets you and tracks where you are, the order desk (SMF) arranges your meal, the kitchen conveyor (UPF) delivers the food, the ID checker (AUSF and UDM) verifies who you are, the manager (PCF) sets house rules, the directory board (NRF) tells stalls where to find each other, and the section host (NSSF) seats you in the right area. Every stall orders from others through the same standard form, which is the HTTP/2 service-based interface.

Tip A fast mapping to remember: MME splits into AMF plus part of SMF; SGW-C and PGW-C become SMF; SGW-U and PGW-U become UPF; HSS becomes UDM/UDR (+ AUSF); PCRF becomes PCF; SCEF becomes NEF; NRF and NSSF are new.
Interview angle Interviewers ask you to "draw the 5GC" and then "why is it service-based?". Draw AMF, SMF, UPF with N1/N2/N3/N4/N6 first, then the SBI bus with the other NFs. Explain the benefits (independent scaling, discovery, reuse of cloud tooling, easy introduction of new NFs) and one downside (more inter-NF signalling and operational complexity).

Registration and PDU session call flows

On SA, getting a phone online takes two NAS procedures. Registration (5GMM, handled by the AMF) replaces the LTE Attach and makes the UE known and reachable. PDU Session Establishment (5GSM, handled by the SMF) gives it an IP connection to a Data Network. Unlike LTE, registration does not automatically create a data connection; the UE asks for PDU sessions separately (it can include them in the same signalling burst).

State machines

LayerStatesMeaning
RM (Registration Management)RM-DEREGISTERED, RM-REGISTEREDIs the UE known to the core?
CM (Connection Management)CM-IDLE, CM-CONNECTEDIs there a NAS signalling connection (N1 plus N2) right now?
RRCRRC_IDLE, RRC_INACTIVE, RRC_CONNECTEDRadio connection state between UE and gNB

RRC_INACTIVE is special: the radio is idle-like but the core still sees the UE as CM-CONNECTED.

Registration types

  • Initial registration at power on or when entering 5GS.
  • Mobility registration update when entering a tracking area outside the registration area, or when capabilities or requested slices change.
  • Periodic registration update when timer T3512 expires, to show the UE is still alive.
  • Emergency registration for emergency services, possible with limited service (even without a valid subscription).

Initial registration step by step

  1. RRC setup The UE sends RRCSetupRequest, receives RRCSetup and returns RRCSetupComplete carrying the NAS Registration Request (SUCI or 5G-GUTI, requested NSSAI, UE capabilities, registration type).
  2. AMF selection The gNB selects an AMF (using the GUAMI in the 5G-S-TMSI if known, or the requested NSSAI) and forwards the NAS message in NGAP Initial UE Message.
  3. Identity and authentication The AMF selects an AUSF, which asks the UDM for an authentication vector. The UDM's SIDF deconceals the SUCI into the SUPI. 5G-AKA or EAP-AKA' runs: Authentication Request (RAND, AUTN) and Response (RES*). The AUSF verifies RES* on behalf of the home network.
  4. NAS security Security Mode Command / Complete activates NAS integrity and ciphering with keys derived from KAMF.
  5. UDM registration The AMF registers itself as serving AMF (Nudm_UECM_Registration) and fetches subscription data (Nudm_SDM_Get), including subscribed S-NSSAIs.
  6. Slice and policy If needed, the AMF queries the NSSF for the Allowed NSSAI (or reroutes to another AMF set). It creates an access and mobility policy association with the PCF, which may push URSP to the UE.
  7. Initial Context Setup The AMF sends NGAP Initial Context Setup Request with the security key for the gNB. The gNB runs AS SecurityModeCommand and RRCReconfiguration.
  8. Registration Accept carries the 5G-GUTI, the TAI list (registration area), Allowed and Rejected NSSAI, periodic timer, and 5GS network feature support (for example "IMS voice over PS supported", which the UE needs for VoNR). The UE replies Registration Complete. A voice-centric UE that does not see IMS voice (and cannot fall back) will disable N1 mode and try EPS; a data-centric UE stays on 5GS. That usage setting is a UE policy, not this network bit.
UE           gNB           AMF          AUSF         UDM        PCF / NSSF
 | RRCSetupReq  |             |             |            |            |
 |----------->|             |             |            |            |
 | RRCSetup     |             |             |            |            |
 |<-----------|             |             |            |            |
 | RRCSetupComplete + Registration Request (SUCI, NSSAI) |            |
 |----------->| InitialUEMessage          |            |            |
 |              |----------->| Nausf_UEAuth |            |            |
 |              |             |----------->| Nudm_UEAU (SUCI -> SUPI) |
 |              |             |             |---------->|            |
 | Authentication Request / Response (5G-AKA, RES*)      |            |
 |<=========================>|<---------->|            |            |
 | NAS Security Mode Command / Complete    |             |            |
 |<=========================>|             |            |            |
 |              |             | Nudm_UECM_Registration, Nudm_SDM_Get   |
 |              |             |------------------------->|            |
 |              |             | Nnssf / Npcf AM policy (if needed)     |
 |              |             |-------------------------------------->|
 |              | InitialContextSetupRequest |           |            |
 |              |<-----------|             |            |            |
 | AS SecurityModeCommand, RRCReconfiguration|            |            |
 |<-----------|             |             |            |            |
 | Registration Accept (5G-GUTI, TAI list, Allowed NSSAI, IMS VoPS)    |
 |<=========================|             |            |            |
 | Registration Complete     |             |            |            |
 |=========================>|             |            |            |

PDU Session Establishment step by step

  1. UE request The UE sends UL NAS Transport containing PDU Session Establishment Request: PDU session ID, DNN (the 5G name for APN, for example internet or ims), S-NSSAI, PDU session type (IPv4, IPv6, IPv4v6, Ethernet, Unstructured) and SSC mode.
  2. SMF selection The AMF discovers a suitable SMF through the NRF (by DNN and S-NSSAI) and calls Nsmf_PDUSession_CreateSMContext.
  3. Subscription and policy The SMF checks session subscription with the UDM and creates an SM policy association with the PCF (Npcf_SMPolicyControl_Create), which returns QoS and charging rules.
  4. UPF selection and N4 The SMF selects a UPF (possibly an edge UPF) and sends a PFCP Session Establishment Request with packet detection, forwarding, QoS enforcement and usage reporting rules. It allocates the UE IP address (or IPv6 prefix).
  5. N1N2 transfer The SMF calls Namf_Communication_N1N2MessageTransfer: the N2 part (QoS profiles, UPF uplink tunnel endpoint) goes to the gNB and the N1 part (PDU Session Establishment Accept with QoS rules, session-AMBR, IP address, DNS and P-CSCF addresses in ePCO) goes to the UE.
  6. Radio resources The AMF sends NGAP PDU Session Resource Setup Request. The gNB configures DRBs and SDAP mapping with RRCReconfiguration and delivers the NAS Accept.
  7. Downlink path The gNB returns its downlink tunnel endpoint in the Setup Response. The AMF forwards it with Nsmf_PDUSession_UpdateSMContext, and the SMF updates the UPF over N4. Traffic now flows both ways.
UE          gNB          AMF          SMF          PCF          UPF
 | PDU Session Est. Request (DNN, S-NSSAI, type)          |            |
 |=========================>| CreateSMContext           |            |
 |             |             |----------->| SMPolicyCreate            |
 |             |             |             |----------->|             |
 |             |             |             | PFCP Session Establishment |
 |             |             |             |------------------------>|
 |             |             | N1N2MessageTransfer (N2 QoS + N1 Accept) |
 |             |             |<-----------|             |             |
 |             | PDUSessionResourceSetupReq |             |             |
 |             |<-----------|             |             |             |
 | RRCReconfiguration (DRB, SDAP) + NAS PDU Session Est. Accept         |
 |<-----------|             |             |             |             |
 |             | PDUSessionResourceSetupRsp (DL TEID)     |             |
 |             |----------->| UpdateSMContext           |             |
 |             |             |----------->| PFCP Session Modification |
 |             |             |             |------------------------>|
 |<=================== user data over DRB / N3 GTP-U ================>|

SSC modes

ModeBehaviourUse
SSC 1The anchor UPF never changes; the IP address is kept for the whole sessionDefault, IMS, most traffic
SSC 2Break before make: the network releases the session and the UE re-establishes with a new anchor and IPStateless traffic that benefits from a closer UPF
SSC 3Make before break: a new session with a new anchor is set up before the old one is releasedEdge apps that want continuity plus relocation

Service Request and paging

When a CM-IDLE UE has uplink data, it sends a Service Request to move to CM-CONNECTED and reactivate the user plane of its PDU sessions. For downlink data, the UPF buffers or notifies the SMF, which asks the AMF to page the UE in its registration area; the UE answers with a Service Request.

Common 5GMM and 5GSM reject causes

CauseMeaningFirst thing to check
5GMM #3 / #6Illegal UE / Illegal MEAuthentication failure, blocked IMEI
5GMM #75GS services not allowedSubscription lacks 5G
5GMM #11 / #13PLMN not allowed / roaming not allowed in this TARoaming agreements, SIM PLMN lists
5GMM #15No suitable cells in tracking areaTA restrictions, forbidden TAs
5GMM #22Congestion (with back-off timer T3346)Core overload
5GMM #27N1 mode not allowedUE should disable N1 (SA) and use LTE
5GMM #62No network slices availableRequested NSSAI vs subscribed and allowed slices
5GSM #26Insufficient resourcesSMF/UPF capacity, back-off timer
5GSM #27Missing or unknown DNNDNN name in the data profile vs subscription
5GSM #29User authentication or authorisation failedDN-level authentication, subscription
5GSM #33Requested service option not subscribedDNN or slice not in subscription
5GSM #69Insufficient resources for specific sliceSlice quota or admission control
Analogy

Registration is checking into a hotel: you show ID at the front desk, they verify your booking with head office, give you a room key and tell you which floors you may use. A PDU session is then ordering a specific service, such as room internet or a conference line, each set up separately. The front desk is the AMF, head office is the AUSF and UDM, the room key is the 5G-GUTI and security context, the floors are the Allowed NSSAI and TAI list, and each service order is a PDU session handled by the SMF and UPF.

Common pitfall Saying registration "creates the default bearer". That was LTE Attach. In 5G, registration and PDU session establishment are separate procedures, even if they happen back to back.
Interview angle "Walk me through what happens when a 5G phone powers on" is a classic. Go RRC setup, Registration Request with SUCI, authentication via AUSF/UDM, NAS security, UDM registration and subscription fetch, Initial Context Setup and AS security, Registration Accept, then PDU session for internet and ims. Mention who does what (AMF vs SMF) and the reject causes you would look for.

5G security: SUPI, SUCI and authentication

5G fixes a long-standing privacy weakness: in 2G to 4G the permanent IMSI could be sent in clear text over the air, so "IMSI catchers" could track users. 5G never sends the permanent identity unencrypted.

Identifiers

IdentifierWhat it is
SUPISubscription Permanent Identifier. Usually an IMSI (MCC + MNC + MSIN) or a network access identifier. Never sent in clear over the radio (except the null scheme).
SUCISubscription Concealed Identifier. MCC/MNC and routing indicator stay in clear so the network can route it; the MSIN is encrypted with the home network public key using ECIES (profile A uses Curve25519, profile B uses secp256r1). A fresh ephemeral key makes every SUCI different.
5G-GUTITemporary identity assigned by the AMF (GUAMI + 5G-TMSI), used after registration instead of SUCI.
5G-S-TMSIShortened GUTI used in paging and RRC setup.
PEIPermanent Equipment Identifier, typically the IMEI or IMEISV; requested only after NAS security is active.
GPSIGeneric Public Subscription Identifier, such as the MSISDN, used outside the 3GPP system.

SUCI can be calculated by the USIM or by the modem (ME), as configured on the SIM. If the SIM has no home network public key provisioned, the null scheme is used and the SUCI effectively reveals the MSIN, which is allowed but loses the privacy benefit. The SIDF (Subscription Identifier De-concealing Function) inside the UDM holds the private key and recovers the SUPI.

Authentication

  • 5G-AKA: an evolution of EPS-AKA. The UE verifies the network with AUTN and returns RES*. The AUSF in the home network verifies the result, so the home operator gets proof the UE was present, not only the visited network (increased home control).
  • EAP-AKA': the same AKA math inside the EAP framework; useful for non-3GPP access and private networks. Other EAP methods (for example EAP-TLS) are allowed for private networks.
  • Key hierarchy: K (in USIM and UDM/ARPF) to CK/IK to KAUSF to KSEAF to KAMF, which yields NAS keys (KNASint, KNASenc) and KgNB, which yields RRC and user-plane keys.
  • Algorithms: NEA1/NIA1 (SNOW 3G), NEA2/NIA2 (AES), NEA3/NIA3 (ZUC). NEA0/NIA0 are null algorithms (NIA0 allowed only for emergency calls without authentication).
  • User-plane integrity protection is new in 5G (optional, negotiated per PDU session). LTE only ciphered the user plane.
  • Roaming: SEPP proxies protect inter-operator SBI signalling on N32, replacing the weakly protected SS7/Diameter interconnect.
Analogy

Sending the IMSI in clear is like shouting your passport number across a train station. SUCI is like putting the passport number in an envelope that only your home embassy can open, with the country name written on the outside so the post office knows where to send it. The envelope is ECIES encryption with the home network public key, the country name is the MCC/MNC and routing indicator, and the embassy is the SIDF in the UDM.

Interview angle Expect "What is SUCI and why was it introduced?" and "What changed in 5G-AKA vs EPS-AKA?". Strong answers mention IMSI catchers, ECIES with the home public key, the null scheme as a fallback, the AUSF's home-network confirmation of RES*, and optional user-plane integrity.

QoS flows and 5QI

LTE gave each QoS class its own EPS bearer with its own GTP tunnel. 5G uses a flatter model: a PDU session has one N3 tunnel, and inside it the traffic is divided into QoS flows, each tagged with a 6-bit QFI (QoS Flow Identifier) in the GTP-U header.

The mapping chain

Application packets (SDFs, matched by 5-tuple / app filters)
        |  UPF (downlink) uses PDRs;  UE (uplink) uses QoS rules
        v
   QoS flow  (QFI, with 5QI, ARP, GFBR/MFBR ...)      <-- set by SMF from PCF policy
        |  SDAP layer in gNB and UE
        v
   Data Radio Bearer (DRB)   -- one DRB may carry several QoS flows
        |
        v
   PDCP / RLC / MAC / PHY over the air
  • Downlink: the UPF classifies packets with Packet Detection Rules and marks the QFI. The gNB maps QFI to a DRB using its SDAP configuration.
  • Uplink: the UE uses QoS rules (received from the SMF) to put packets into QoS flows, then SDAP maps them to DRBs.
  • Reflective QoS: the UE can derive uplink rules by mirroring downlink packets marked with the RQI bit, reducing signalling.
  • Every PDU session has a default QoS rule and default QoS flow (usually non-GBR, 5QI 9 for internet, 5QI 5 for IMS).

QoS parameters

ParameterMeaning
5QIIndex into standardised characteristics: resource type, priority, packet delay budget (PDB), packet error rate (PER)
ARPAllocation and Retention Priority: admission and pre-emption, not per-packet treatment
GFBR / MFBRGuaranteed and maximum flow bit rates for GBR flows
Session-AMBRAggregate cap for all non-GBR flows in a PDU session
UE-AMBRAggregate cap for all non-GBR flows of a UE, enforced by the gNB
Notification controlThe gNB tells the SMF when a GBR flow can no longer be guaranteed (useful for voice codec adaptation)

Important 5QI values

5QIResource typePDBTypical use
1GBR100 msConversational voice (VoNR media)
2GBR150 msConversational video
3GBR50 msReal-time gaming, V2X messages
4GBR300 msBuffered streaming video
65 / 66GBR75 ms / 100 msMission-critical and non-mission-critical push-to-talk voice
5Non-GBR100 msIMS signalling (SIP)
6, 8, 9Non-GBR300 msTCP-based traffic; 9 is the usual default internet flow
7Non-GBR100 msLive streaming, interactive gaming
69 / 70Non-GBR60 ms / 200 msMission-critical signalling / data
82 to 86Delay-critical GBR5 to 30 msURLLC: discrete automation, intelligent transport, power grid, V2X

The QoS characteristics of 5QI 1 to 9 intentionally match LTE QCI 1 to 9, which makes interworking with EPS (EPS fallback, N26 handover) straightforward: the SMF+PGW-C maps QoS flows to EPS bearers.

Do not invert the IMS 5QIs. 5QI 1 is conversational voice (GBR). 5QI 5 is IMS signalling on the default QoS flow of the IMS PDU session. 5G uses a different priority-level scale than LTE (5QI 5 is 10, 5QI 1 is 20; lower is higher priority), but the relationship is the same: SIP is scheduled ahead of other default/non-GBR traffic such as 5QI 9, while voice quality is protected by the GBR reservation. See IMS & VoLTE for the QCI table and the same trap.

Policy chain for a voice call

P-CSCF (AF) --N5 / Rx--> PCF --N7--> SMF --N4--> UPF (marks QFI)
                                        |
                                        +--N11/N2 via AMF--> gNB  (admits GBR flow,
                                                                  maps to voice DRB)
Analogy

In LTE every class of parcel needed its own lorry (one bearer, one tunnel). In 5G one lorry per customer (the PDU session and its N3 tunnel) carries many parcels with coloured priority stickers (QFI), and at the loading dock a sorter (SDAP) puts each sticker colour into the right compartment of the delivery van (the DRB). The sticker rules come from the head office (PCF) through the dispatcher (SMF).

Common pitfall Swapping 5QI 1 and 5QI 5, or quoting LTE priority numbers (1 and 2) as if they were 5QI priority levels. 5QI 1 is voice. 5QI 5 is SIP.
Interview angle "How does 5G QoS differ from LTE bearers?" and "Which 5QI for VoNR?" are near-certain. Say QoS flows inside one PDU session instead of per-bearer tunnels, QFI in GTP-U, SDAP maps flows to DRBs, 5QI 1 GBR for voice media and 5QI 5 for SIP, and ARP decides pre-emption, not packet treatment.

NR air interface: numerology, frames and spectrum

NR is OFDM-based like LTE (CP-OFDM in both directions, with optional DFT-s-OFDM in uplink for better power efficiency at cell edge). The big difference is flexibility: one radio design has to work from 600 MHz to 71 GHz and from IoT to URLLC.

Numerology

Numerology µ sets the subcarrier spacing: SCS = 15 × 2µ kHz. A wider spacing gives shorter OFDM symbols, so a 14-symbol slot is shorter and latency drops. It also tolerates the higher phase noise and Doppler of mmWave. The cost is a shorter cyclic prefix, which handles less delay spread, so low bands use narrow SCS and high bands use wide SCS.

µSCSSlots per 1 ms subframeSlot lengthTypical use
015 kHz11 msFR1 low band, LTE-like, DSS
130 kHz20.5 msFR1 mid band (3.5 GHz TDD) — the workhorse
260 kHz40.25 msFR1 or FR2; only µ with extended CP option
3120 kHz8125 µsFR2 mmWave data
4240 kHz1662.5 µsFR2 SSB only (not data) in Release 15
5 / 6480 / 960 kHz32 / 6431.25 / 15.6 µsFR2-2 (52.6 to 71 GHz), Release 17

Frame structure

  • Frame = 10 ms, split into 10 subframes of 1 ms each (fixed for all numerologies, which keeps LTE/NR timing aligned).
  • Slot = 14 OFDM symbols with normal CP (12 with extended CP). The number of slots per subframe is 2µ.
  • Mini-slots (non-slot-based scheduling, 2, 4 or 7 symbols) let URLLC data start mid-slot without waiting for a slot boundary.
  • Resource block = 12 subcarriers. A carrier has up to 275 RBs; for example 100 MHz at 30 kHz is 273 RBs.
  • Slot formats mark each symbol as downlink, uplink or flexible, configured semi-statically by RRC or dynamically by DCI format 2_0.
Frame 10 ms
|--SF0--|--SF1--|--SF2--| ... |--SF9--|        each subframe = 1 ms
µ=0 : [slot 0                     ]              1 slot / SF
µ=1 : [slot 0      ][slot 1       ]              2 slots / SF
µ=3 : [s0][s1][s2][s3][s4][s5][s6][s7]            8 slots / SF
slot = 14 symbols: | 0 | 1 | 2 | ... | 13 |
                     PDCCH usually in first 1-3 symbols (CORESET)

Frequency ranges

RangeFrequenciesMax carrier bandwidthCharacteristics
FR1 (sub-6)410 MHz to 7.125 GHz100 MHzGood coverage and building penetration. Low band (n71, n28) for coverage, mid band (n77/n78 around 3.5 GHz, n41 at 2.5 GHz) for capacity.
FR2-1 (mmWave)24.25 to 52.6 GHz400 MHzHuge bandwidth and multi-Gbps speeds, but short range, blocked by walls, foliage, hands and bodies. Needs beamforming. Bands n257, n258, n260, n261.
FR2-252.6 to 71 GHzup to 2 GHzRelease 17 extension, higher numerologies.

Band numbers start with "n" for NR (n78) to distinguish them from LTE bands (B3). The ARFCN for NR is the NR-ARFCN; SSBs sit on a coarser GSCN raster so the UE can scan faster.

Bandwidth parts (BWP)

A UE does not have to process a full 100 MHz or 400 MHz carrier. The network configures up to 4 downlink and 4 uplink BWPs per serving cell, each with its own bandwidth, location and numerology; one of each is active at a time.

  • Initial BWP: used during access, derived from CORESET#0 and SIB1.
  • First active BWP: activated after RRC reconfiguration.
  • Default BWP: a narrow BWP the UE falls back to when bwp-InactivityTimer expires, saving power.
  • Switching is by DCI (fast), RRC, timer expiry or random access. Benefits: power saving, support for UEs with smaller bandwidth, and mixing numerologies on one carrier.

Duplexing and TDD patterns

Most mid-band and all mmWave NR is TDD: the same frequency alternates between downlink and uplink over time. The pattern is set in tdd-UL-DL-ConfigurationCommon.

  • DDDSU at 30 kHz (2.5 ms period): 3 downlink slots, 1 special slot, 1 uplink slot. Good latency.
  • DDDDDDDSUU (5 ms period): heavier downlink capacity; used by many operators.
  • The special slot contains downlink symbols, a guard period for the switch and a few uplink symbols (for example 10:2:2).
  • Operators in the same band must use synchronised, compatible patterns, otherwise one operator's downlink interferes with another's uplink.
  • TDD enables channel reciprocity: the gNB estimates the downlink channel from uplink SRS, which helps massive MIMO.

Carrier aggregation, DSS and supplementary links

  • Carrier aggregation: NR supports up to 16 component carriers (PCell + SCells). Combining low band for coverage with mid band for capacity is common. In NSA, LTE CA and NR CA can run at the same time, which is why EN-DC band combinations in UE capability matter so much.
  • DSS (Dynamic Spectrum Sharing): LTE and NR share the same carrier, with the scheduler splitting resources dynamically (every 1 ms). NR must avoid LTE's CRS (rate matching) and often uses MBSFN subframes. It gives fast wide NR coverage on existing LTE spectrum, but capacity is lower than dedicated NR because of overhead.
  • SUL (Supplementary Uplink): an extra low-band uplink carrier paired with a high-band TDD cell to fix cell-edge uplink.
  • SDL (Supplementary Downlink): a downlink-only carrier for extra capacity.

Physical channels and signals

NameDirectionPurpose
PBCHDLCarries the MIB inside the SSB
PDCCHDLDownlink control information (grants), sent in a CORESET and found through search spaces
PDSCHDLDownlink data, SIBs, paging
PUCCHULHARQ ACK/NACK, scheduling requests, CSI reports
PUSCHULUplink data, and UCI when multiplexed
PRACHULRandom access preamble
DMRSDL/ULDemodulation reference signal, sent only with the data it helps decode
CSI-RSDLChannel state and beam measurement; TRS for time/frequency tracking
SRSULUplink sounding, channel reciprocity
PTRSDL/ULPhase tracking, needed at mmWave because of oscillator phase noise

Channel coding: LDPC for PDSCH/PUSCH (high throughput, parallel decoding) and Polar codes for PBCH, DCI and UCI (strong for short blocks).

Analogy

Numerology is like choosing the size of time slots at a busy loading bay. For large slow trucks on long country roads (low band, big delay spread) you book one-hour slots; for fast couriers in the city (mmWave, low latency) you book five-minute slots. A bandwidth part is opening only the lanes of the bay you need so you do not pay to light the whole building. The slot length is the NR slot duration set by µ, and the lanes are the resource blocks of the active BWP.

Tip Keep one calculation ready: at 30 kHz SCS a slot is 0.5 ms, so a DDDSU pattern repeats every 2.5 ms; at 120 kHz a slot is 125 µs. Interviewers like to see you do this arithmetic quickly.
Interview angle Typical probes: "Explain numerology and why multiple SCS exist", "FR1 vs FR2", "Why BWP?", "What is DSS and its trade-off?". Tie each answer back to a requirement: latency, coverage, bandwidth or UE power.

Initial access, beam management and massive MIMO

NR is beam-based even for initial access. At mmWave, path loss is so high that the gNB must focus energy in narrow beams, so even sync signals are swept across beams.

The SSB

  • An SS/PBCH block takes 4 OFDM symbols and 240 subcarriers (20 RBs): PSS, SSS and PBCH (with the MIB and its DMRS).
  • PSS (3 values) and SSS (336 values) together give the Physical Cell ID, 1008 possible PCIs (LTE had 504).
  • SSBs are sent in bursts within a 5 ms half-frame, each SSB index on a different beam. The maximum per burst is 4 or 8 in FR1 and 64 in FR2. The UE assumes a 20 ms burst periodicity during initial search.
  • The MIB tells the UE where to find CORESET#0 and the search space for SIB1; SIB1 carries access parameters and the RACH configuration.

Initial access

  1. Cell search Scan the GSCN raster, detect PSS/SSS, get timing, PCI and the SSB index (which beam).
  2. MIB and SIB1 Decode PBCH, then PDCCH in CORESET#0 and SIB1 on PDSCH.
  3. Random access Send a preamble on the RACH occasion mapped to the best SSB. Because RACH occasions map to SSB indices, the gNB learns the UE's best beam from when it hears the preamble.
  4. RAR, Msg3, Msg4 Random access response with timing advance and uplink grant, Msg3 carries RRCSetupRequest, Msg4 resolves contention and carries RRCSetup. Release 16 adds 2-step RACH (MsgA / MsgB) to cut latency.

Beam management

  • P1 coarse beam selection with the SSB sweep, P2 gNB refines its transmit beam with CSI-RS, P3 the UE refines its receive beam.
  • The UE reports L1-RSRP per beam. The network tells the UE which beam to assume for reception through TCI states (Transmission Configuration Indicator) that reference a quasi-co-located SSB or CSI-RS.
  • Beam failure detection: the UE monitors beam failure detection reference signals; if the counter of beam failure instances reaches beamFailureInstanceMaxCount before the timer expires, beam failure is declared.
  • Beam failure recovery: the UE finds a candidate beam above a threshold and sends a (usually contention-free) RACH on it; the gNB responds on the new beam. This is much faster than a full radio link failure and re-establishment.

Massive MIMO and beamforming

  • Mid-band gNBs commonly use 32T32R or 64T64R antenna panels. Many elements let the gNB steer narrow beams (beamforming) and serve several users on the same resources (MU-MIMO).
  • SU-MIMO sends several layers to one UE: up to 8 layers downlink and 4 uplink in the specification, typically 4 downlink layers on phones.
  • Beamforming types: analog (phase shifters, one beam per RF chain, cheap, used at mmWave), digital (per-element baseband weights, most flexible but power-hungry), hybrid (digital precoding over a few analog-beamformed panels, the practical choice for mmWave).
  • CSI feedback: the UE reports CQI, PMI (Type I or high-resolution Type II codebooks), RI (rank) and CRI. In TDD, SRS-based reciprocity can replace much of this.
  • Phones have several mmWave antenna modules on different edges because a hand can block one; the modem switches modules as the grip changes.
Analogy

An LTE cell is like a ceiling light that brightens the whole room. A mmWave gNB is like a lighthouse with many narrow spotlights: it sweeps them around (SSB burst), you shout back which spotlight you saw best (RACH on that SSB's occasion), and then it keeps a spotlight on you as you walk, switching to another one if someone steps in front of you (beam failure recovery). The spotlights are beams formed by the massive MIMO array.

Interview angle "What is the SSB and why is it beam-swept?" and "What happens when a mmWave beam is blocked?" are common. Explain SSB contents, SSB-to-RACH mapping, P1/P2/P3, TCI states, and beam failure recovery before radio link failure. For device roles, mention antenna module switching and hand blockage.

Protocol stack and RRC states

The NR stack looks like LTE with one new layer and several changes aimed at low latency.

     Control plane                  User plane
  +---------------+
  |      NAS      |  UE <-> AMF (5GMM / 5GSM)
  +---------------+              +---------------+
  |      RRC      |  UE <-> gNB  |     SDAP      |  new in NR: QoS flow -> DRB
  +---------------+              +---------------+
  |     PDCP      |              |     PDCP      |  ciphering, integrity, ROHC,
  +---------------+              +---------------+  reordering, duplication, DC split
  |      RLC      |              |      RLC      |  TM / UM / AM, ARQ, segmentation
  +---------------+              +---------------+
  |      MAC      |              |      MAC      |  scheduling, multiplexing, HARQ
  +---------------+              +---------------+
  |      PHY      |              |      PHY      |  OFDM, LDPC / Polar, MIMO
  +---------------+              +---------------+
LayerRoleWhat changed vs LTE
NASRegistration, authentication, security, PDU sessions (5GMM + 5GSM), end to end with the AMFReplaces EMM/ESM; separate mobility and session protocols; access-independent
RRCSystem information, connection setup/resume/release, measurement config, handover, DRB setupAdds RRC_INACTIVE, on-demand SI, beam and BWP config
SDAPMaps QoS flows to DRBs, marks QFI in uplinkNew layer; exists only when connected to 5GC (not in EN-DC)
PDCPCiphering, integrity (now for user plane too), ROHC, reordering, duplicate discard, split bearer routingPacket duplication over two legs for URLLC reliability; reordering moved here from RLC
RLCSegmentation, ARQ retransmission in AM modeNo concatenation and no in-order delivery, so PDUs can be pre-built before the grant arrives
MACScheduling, logical to transport channel multiplexing, HARQ, BWP and beam proceduresMAC subheader placed before each SDU for pipelined processing; asynchronous HARQ in both directions with flexible timing (K1, K2); up to 16 HARQ processes
PHYModulation up to 256QAM (1024QAM later), coding, MIMO, beamformingFlexible numerology, no CRS, LDPC/Polar

Signalling radio bearers: SRB0 (CCCH, before setup), SRB1 (RRC and piggybacked NAS), SRB2 (NAS, lower priority), and SRB3 (direct NR RRC between UE and the secondary gNB in EN-DC, so SCG reconfigurations need not go through the LTE master).

HARQ vs ARQ: HARQ in MAC is fast (sub-millisecond) and soft-combines failed attempts; ARQ in RLC is a slower safety net that retransmits whole segments reliably.

RRC states

StateUE context in RANCore viewMobilityReached by
RRC_IDLENoneCM-IDLEUE-controlled cell reselection; core paging over the registration areaRRC Release
RRC_INACTIVEStored in UE and the anchor gNBCM-CONNECTED (N2 and N3 kept)Cell reselection within the RAN Notification Area (RNA); RAN-initiated pagingRRC Release with suspendConfig
RRC_CONNECTEDActiveCM-CONNECTEDNetwork-controlled handover based on measurement reportsRRC Setup or RRC Resume

How RRC_INACTIVE works

  1. Suspend The gNB sends RRCRelease with suspendConfig: an I-RNTI (identity of the stored context), the RNA and a periodic RNA update timer.
  2. Move The UE reselects cells freely inside the RNA without signalling; leaving the RNA triggers an RNA update.
  3. Resume On uplink data or RAN paging, the UE sends RRCResumeRequest with the I-RNTI. The new gNB fetches the context from the anchor gNB over Xn (Retrieve UE Context) if needed.
  4. Connected again RRCResume and RRCResumeComplete: security and bearers are restored with a few messages and no NAS signalling to the core. If the context cannot be found, the network falls back to RRCSetup.

Release 17 adds Small Data Transmission (SDT), which lets a UE send small packets in RRC_INACTIVE without fully resuming.

Power saving

  • C-DRX in connected mode, idle DRX and eDRX between paging occasions.
  • BWP switching to a narrow default BWP during low traffic.
  • Wake-up signal (DCI format 2_6) tells the UE whether to wake for the next DRX on-duration.
  • SCell dormancy and cross-slot scheduling reduce unnecessary PDCCH monitoring.
  • RedCap (Reduced Capability, Release 17) NR devices for wearables and sensors: narrower bandwidth (20 MHz in FR1), fewer antennas, lower cost and power.
Analogy

RRC_CONNECTED is sitting at a restaurant table with a waiter. RRC_IDLE is leaving; next time you queue and get seated from scratch. RRC_INACTIVE is stepping outside for a phone call while the waiter keeps your table and order: when you return you just say your table number and carry on. The table number is the I-RNTI, the stored order is the UE context in the anchor gNB, and "the neighbourhood you can wander in" is the RAN Notification Area.

Common pitfall Saying the core sees an inactive UE as idle. It does not; RRC_INACTIVE is CM-CONNECTED, and paging for it is done by the RAN, not the AMF.
Interview angle "What is SDAP and why was it needed?" and "Explain RRC_INACTIVE" are favourites. Also be ready for "What does PDCP do?" and "HARQ vs ARQ". Good answers link each change to a goal: SDAP to the QoS flow model, RLC without concatenation and new MAC layout to latency, PDCP duplication to URLLC reliability.

Voice on 5G: VoNR and EPS fallback

There is no circuit-switched domain in 5G. All voice is IMS (SIP signalling plus RTP media), exactly as in VoLTE; only the access and core underneath change. For the SIP ladder itself see IMS & VoLTE and call flows.

Voice options by deployment

DeploymentWhere the call runsUser sees
NSA (Option 3x)VoLTE on the LTE anchor (IMS via EPC, QCI 1 bearer); the NR SCG may be kept or released during the call depending on operator config5G icon may stay
SA with EPS fallbackUE camps on NR; at call setup it is moved to LTE and the call runs as VoLTE5G changes to LTE during the call
SA with VoNRNative IMS voice on NR with a 5QI 1 QoS flow5G stays
RAT fallbackMoved to LTE connected to 5GC (ng-eNB, Option 5); rareLTE

VoNR requirements

  • IMS PDU session on DNN ims with a default 5QI 5 QoS flow for SIP; P-CSCF addresses delivered in ePCO.
  • Registration Accept must indicate "IMS voice over PS session supported over 3GPP access" in 5GS network feature support. Without it the UE will not attempt IMS voice on NR.
  • Policy: the P-CSCF talks to the PCF over N5 (or Rx through interworking) so the SMF can create a 5QI 1 GBR QoS flow (and 5QI 2 for video).
  • gNB voice features: admission of GBR flows, ROHC header compression, C-DRX tuned for 20 ms voice frames, configured grant (the NR cousin of LTE SPS), uplink coverage aids (PUSCH repetition / slot aggregation, the cousin of TTI bundling), RLC UM for the voice DRB, and inter-system handover to LTE when NR coverage ends.
  • UE: IMS stack and modem support voice over NR; codecs AMR-WB and EVS; carrier configuration enabling VoNR; a voice-centric usage setting that will leave 5GS if IMS voice is not actually available.
  • N26 for seamless VoNR to VoLTE handover when leaving NR coverage. 5G-SRVCC to 3G CS exists in Release 16 but is rarely deployed.

VoLTE versus VoNR

AspectVoLTEVoNR
When it runsLTE, including the NSA LTE anchor5G SA only
QoSQCI 1 dedicated bearer5QI 1 QoS flow (do not call 5QI 5 the voice flow)
PolicyRx / GxN5 / N7 / N4
SIP ladderSame IMS coreSame IMS core
Setup if voice is not supported on this RATCSFB or stay on LTE without VoLTEEPS fallback to VoLTE, or RAT fallback to ng-eNB
In-call coverage exitSRVCC to CSN26 handover to VoLTE; 5G-SRVCC is rare
Mouth-to-earG.114 ~150 ms target; jitter buffer and codec dominateSame target; shorter NR slots help a little, setup time vs EPS fallback helps a lot

Full SIP, S8HR, MOS/E-model, CMR/ANBR and the LTE radio set (SPS, TTI bundling) are on IMS & VoLTE.

Voice-centric versus data-centric (5GS usage setting)

TS 24.501 gives the UE a 5GS usage setting, independent of the IMS VoPS bit in Registration Accept.

  • Voice-centric: if 5GS cannot provide IMS voice (VoPS not indicated and emergency/EPS fallback not a usable substitute for normal calls), the UE disables N1 mode for that PLMN and moves to EPS, where it applies the same rule again. This is a common reason a "5G SA is live" device still camps only on LTE or NSA.
  • Data-centric: the UE stays on 5GS for data even when voice is unavailable; a dial then triggers EPS fallback or fails.

EPS fallback flow

  1. IMS call starts on NR The UE is registered to IMS over the NR IMS PDU session. An MO INVITE is sent (or an MT INVITE arrives after paging).
  2. Voice QoS flow requested The P-CSCF asks the PCF for voice media resources over N5; the SMF asks the AMF to set up a 5QI 1 QoS flow, and the AMF sends NGAP PDU Session Resource Modify Request to the gNB.
  3. gNB triggers fallback The gNB is configured for EPS fallback (or the UE or radio conditions do not support VoNR). It rejects the QoS flow with cause "IMS voice EPS fallback or RAT fallback triggered" and may first configure a B1 measurement of LTE.
  4. Move to LTE Either an N2 inter-system handover (Handover Required to AMF, Forward Relocation to MME over N26, UE handed to LTE) or an RRC Release with redirection to E-UTRA, after which the UE performs a Tracking Area Update on LTE.
  5. Voice bearer on LTE The combined SMF+PGW-C maps the IMS PDU session to an EPS PDN connection (same IP address, so the SIP session continues) and creates the QCI 1 dedicated bearer.
  6. Call completes as VoLTE Preconditions are met, the callee is alerted, and media flows over LTE.
  7. Return to 5G After the call, the eNB can release the UE with redirection or priorities toward NR ("fast return"); otherwise the UE returns only by idle-mode reselection, which can take longer.
UE        gNB        AMF       SMF+PGW-C    PCF      P-CSCF       MME       eNB
 | INVITE (over NR IMS PDU session) ----------------------->|          |         |
 |          |          |          |         |<-- N5 -----|           |         |
 |          |          |<-- request 5QI 1 --|           |           |         |
 |          | PDUSessResourceModifyReq      |           |           |         |
 |          |<--------|          |         |           |           |         |
 |          | reject: "EPS fallback triggered"          |           |         |
 |          |-------->|          |         |           |           |         |
 |   (B1 LTE meas.)   |          |         |           |           |         |
 |          | HO Required -> AMF --- N26 Forward Relocation -------->|         |
 |<== HO command / or RRCRelease with redirect to E-UTRA ==|         |         |
 |========================== attach to LTE / TAU =========================>|
 |          |          |          | create QCI 1 dedicated bearer ---->|------>|
 | 183 / PRACK / UPDATE / 180 / 200 OK continue as VoLTE                         |

Why EPS fallback is slower

  • It adds an inter-RAT move to call setup: roughly a few hundred ms with handover, and often 1 to 3 seconds with redirection plus TAU.
  • Without N26, the UE must attach or do PDN connectivity with a handover indication on LTE, which is slower again and risks failed calls.
  • Long delays can race SIP timers, causing 408/480 or cancelled MT calls, and a poor "5G to LTE and back" user experience.

Emergency calls

If the network does not support emergency calls over NR, the Registration Accept indicates Emergency Services Fallback support and the UE sends a Service Request with an emergency fallback indication, which moves it to LTE for the emergency call. The emergency call uses a dedicated emergency PDU session (or PDN on LTE).

EPS fallback

  • At call setup (before or during INVITE)
  • 5G NR to 4G LTE
  • Stays PS/IMS: call becomes VoLTE
  • Deployment bridge until VoNR

SRVCC

  • During an active call
  • LTE (or NR) PS to 2G/3G CS
  • IMS call anchored and transferred to CS
  • Coverage edge continuity for legacy networks
Analogy

Imagine a new airport terminal (5G SA) where check-in and shops work, but the gates for one airline (voice) are not finished. When you book that airline, staff walk you through a connector to the old terminal (LTE) where the gate works, and you fly from there. VoNR is when the new terminal's gates finally open. The connector is N26 and the inter-system handover; the flight is the same booking throughout, which is the same IMS SIP session and IP address.

Interview trap Mixing up EPS fallback and SRVCC. EPS fallback moves at call setup and stays on IMS (becomes VoLTE). SRVCC moves an already-active call from PS to CS. Also, a user complaint "5G drops to LTE every time I call" is usually EPS fallback working as designed, not a device bug.
Interview angle Expect to draw the EPS fallback flow and explain who triggers it (the gNB, after the SMF/AMF request a 5QI 1 flow), handover vs redirection, the role of N26, how the IP address is preserved (SMF+PGW-C anchor), the latency impact, and how VoNR removes it. Senior loops ask for a migration plan: NSA with VoLTE, then SA with EPS fallback, then VoNR cell by cell with EPS fallback as backstop.

Network slicing

A network slice is a logical end-to-end network (RAN resources, core NFs and transport) running on shared physical infrastructure, tuned for a service type or customer. It is possible because the 5GC is software NFs that can be instantiated and selected per slice.

Identifiers

  • S-NSSAI (Single Network Slice Selection Assistance Information) = SST (Slice/Service Type, 8 bits) + optional SD (Slice Differentiator, 24 bits) to separate slices of the same type (for example two enterprise eMBB slices).
  • Standard SST values: 1 eMBB, 2 URLLC, 3 MIoT (massive IoT), 4 V2X, and later 5 HMTC (high-performance machine type) and 6 HDLLC. Operators can also define their own values.
  • NSSAI = a list of up to 8 S-NSSAIs. Types: Configured NSSAI (provisioned in the UE per PLMN), Requested NSSAI (sent in Registration Request), Allowed NSSAI (returned in Registration Accept, valid in the registration area), Rejected NSSAI (with a cause), and Subscribed S-NSSAIs in the UDM, some marked default.

How a slice is selected

  1. Registration The UE sends Requested NSSAI; the gNB may use it to pick an AMF that supports those slices.
  2. NSSF The AMF checks the subscription and, if needed, asks the NSSF, which returns the Allowed NSSAI and possibly a different AMF set (the registration is then rerouted).
  3. URSP The PCF sends UE Route Selection Policy rules to the UE. Each rule has a traffic descriptor (application ID, DNN, IP descriptors, connection capabilities) and one or more route selection descriptors (S-NSSAI, DNN, SSC mode, PDU session type, access type).
  4. PDU session per slice When an app's traffic matches a URSP rule, the UE uses (or creates) a PDU session with that S-NSSAI. The AMF selects an SMF for that slice through the NRF, and the SMF picks slice-specific UPFs.
  5. RAN enforcement The gNB applies slice-aware scheduling and admission (resource quotas or priorities per slice).

Release 16 adds Network Slice-Specific Authentication and Authorisation (NSSAA), where a third party (for example an enterprise AAA server) must approve slice access. Release 17 adds slice quotas (maximum UEs or sessions per slice) and slice-aware cell reselection.

Examples

Consumer eMBB

SST 1, best-effort 5QI 9, large central UPF, shared RAN.

Voice or enterprise voice

A slice with guaranteed 5QI 1 admission and protected capacity, isolated from data congestion.

URLLC factory

SST 2, edge UPF on site, delay-critical GBR 5QI, mini-slots, PDCP duplication.

Private enterprise

SST 1 with an SD value per customer, NSSAA to the enterprise AAA, dedicated DNN.

Analogy

Slicing is like one office building rented to several companies. Each company gets its own floor, access badge rules and reserved lifts, but they share the foundations, power and plumbing. The building is the physical RAN and core; each floor is a slice identified by S-NSSAI; the badge list is the Allowed NSSAI; and the reception desk that tells visitors which floor to use for which purpose is URSP on the phone and the NSSF in the network.

Common pitfall Thinking slicing is just QoS. QoS differentiates flows inside a session; a slice can have separate NFs, UPFs, policies, capacity quotas and even authentication. Also, slicing needs SA; NSA has no slicing.
Interview angle Interviewers ask "What is a slice and how is it selected?", "What is S-NSSAI?", and on Android "How does an app end up on a slice?" (URSP, traffic descriptor, route selection descriptor, a separate PDU session). Mention the NSSF, Allowed vs Configured NSSAI, and reject cause 5GMM #62.

Mobility: handovers and idle mode

Measurement events

EventConditionTypical use
A1 / A2Serving becomes better / worse than a thresholdStop / start measurements, trigger fallback
A3Neighbour better than serving by an offsetIntra-frequency handover (most common)
A4Neighbour better than a thresholdLoad balancing, adding cells
A5Serving below threshold 1 and neighbour above threshold 2Coverage-based handover
A6Neighbour better than an SCell by an offsetSCell change
B1Inter-RAT neighbour better than a thresholdNR addition in EN-DC, EPS fallback target
B2Serving below threshold 1 and inter-RAT neighbour above threshold 2NR to LTE handover at coverage edge

Measurement quantities on NR are SS-RSRP, SS-RSRQ, SS-SINR (from the SSB) and CSI-RSRP/RSRQ/SINR. Hysteresis and time-to-trigger avoid ping-pong.

Xn handover (gNB to gNB directly)

  1. Measurement report The UE reports an A3 event.
  2. Handover Request The source gNB sends Handover Request over Xn with the UE context.
  3. Admission The target reserves resources and returns an RRCReconfiguration with reconfigurationWithSync (new cell, C-RNTI, RACH resources) inside the Acknowledge.
  4. Handover command The source forwards it to the UE and starts forwarding buffered data over Xn-U (SN Status Transfer keeps PDCP sequence numbers).
  5. Access target The UE performs random access on the target and sends RRCReconfigurationComplete.
  6. Path switch The target sends NGAP Path Switch Request to the AMF; the SMF tells the UPF to move the N3 downlink tunnel. The UPF sends end-marker packets on the old path.
  7. Release The target sends UE Context Release to the source over Xn.
UE         Source gNB        Target gNB        AMF        SMF / UPF
 | MeasReport A3 |                 |               |              |
 |----------->| HO Request (Xn)  |               |              |
 |              |---------------->|               |              |
 |              | HO Request Ack (RRCReconfiguration inside)       |
 |              |<----------------|               |              |
 | RRCReconfiguration (reconfigurationWithSync)   |              |
 |<-----------| SN Status Transfer + data forwarding (Xn-U)     |
 |              |---------------->|               |              |
 |  RACH + RRCReconfigurationComplete             |              |
 |------------------------------>| Path Switch Req |              |
 |              |                 |-------------->| N4: switch N3 DL |
 |              |                 |               |------------->|
 |              |                 | Path Switch Ack |              |
 |              | UE Context Release |<------------|              |
 |              |<----------------|               |              |

N2 handover (through the AMF)

Used when there is no Xn between the gNBs, or the AMF or UPF must change. The source sends Handover Required to the AMF, the (possibly new) AMF sends Handover Request to the target, the target acknowledges, the AMF sends Handover Command to the source, the UE moves, and the target sends Handover Notify. More signalling and slightly higher interruption than Xn.

Advanced handover types

  • Conditional Handover (CHO), Release 16: the network prepares one or more targets in advance with conditions; the UE executes when a condition is met, which avoids failures when the radio drops too fast to send a report.
  • DAPS (Dual Active Protocol Stack) handover, Release 16: the UE keeps receiving from the source until the target is up, for near-zero interruption.
  • RLF and re-establishment: after radio link failure (T310 expiry, max RLC retransmissions, RACH failure) the UE tries RRCReestablishment on a suitable cell; if the context is unavailable it falls back to a new setup.

Inter-system mobility (5GS and EPS)

  • With N26: AMF and MME exchange UE context, so connected-mode handover and idle-mode moves keep sessions and the IP address with little signalling. Needed for good EPS fallback and VoNR to VoLTE continuity.
  • Without N26: the UE re-registers or attaches on the other system and uses a "handover" indication in session requests to keep the IP address via the combined SMF+PGW-C anchor. Slower, and in-flight sessions may break.
  • A combined SMF+PGW-C and UPF+PGW-U is the common anchor that makes PDU sessions and PDN connections map to each other.

NSA mobility

In EN-DC, mobility is driven by the LTE master. Moving between NR cells under the same eNB is a PSCell change (SN modification or SN change); an LTE handover may keep, change or release the SN. SCG failures are reported to the master without dropping the connection.

Idle mode mobility

  • Cell reselection is UE-controlled using priorities and thresholds broadcast in SIBs (SIB2 serving, SIB4 inter-frequency NR, SIB5 LTE). Higher-priority layers are measured always; equal or lower priority only when serving quality drops.
  • Registration area: the AMF gives the UE a TAI list (a set of tracking areas). The UE moves freely inside it and does a mobility registration update only when entering a TA outside the list. A larger list means fewer updates but a larger paging area.
  • Periodic registration on T3512 expiry; if the UE is unreachable for longer, the AMF implicitly deregisters it.
  • MICO mode (Mobile Initiated Connection Only): an IoT UE that is not paged at all, saving power.
  • RRC_INACTIVE mobility is tracked per RAN Notification Area by the RAN, not per TA by the AMF.
Analogy

An Xn handover is a relay race where runners pass the baton directly; an N2 handover is when the runners cannot see each other and the coach (AMF) passes the baton. Idle mobility is like telling your family which city you are in rather than which street: you only phone home when you move to a city outside the agreed list. The baton is the UE context, the agreed list of cities is the TAI list, and phoning home is a mobility registration update.

Interview angle Classic questions: "Xn vs N2 handover", "What is a path switch?", "What does N26 give you?", "Why do we have registration areas?". Show the balance between signalling and paging load, and mention CHO/DAPS and RLF re-establishment for extra depth. For NSA, mention PSCell change and SCG failure.

Android 5G: icons, network type and data

On Android the framework has to translate modem state into what the user sees and what apps can use. This is where a lot of "no 5G icon" bugs actually live. For RIL and modem internals see Telephony, RIL & modem.

Network type vs display type

  • On SA, the data registration reports NETWORK_TYPE_NR.
  • On NSA, the registered network is still LTE (NETWORK_TYPE_LTE), because NAS and RRC run on LTE. The 5G icon comes from a separate override.
  • TelephonyDisplayInfo (Android 11+) carries both the network type and an override network type: OVERRIDE_NETWORK_TYPE_NONE, LTE_CA, LTE_ADVANCED_PRO, NR_NSA, NR_ADVANCED (Android 12+, for mmWave or high-bandwidth NR; replaces the deprecated NR_NSA_MMWAVE). Apps listen with TelephonyCallback.DisplayInfoListener.

How the NSA icon is decided

The framework's NetworkTypeController computes the override from the NR state, physical channel configuration and carrier config.

NR stateMeaningSource
NONENo EN-DC capability in this cellSIB2 has no upperLayerIndication
RESTRICTEDCell supports EN-DC but the core says this UE may not use NR (DCNR restricted)NAS attach/TAU accept
NOT_RESTRICTEDEN-DC available and allowed, but no SCG yetSIB2 + NAS
CONNECTEDSCG (NR PSCell) is activePhysical channel config shows an NR cell

The carrier config key 5g_icon_configuration_string maps states to icons, for example:

connected_mmwave:5G_Plus,connected:5G,not_restricted_rrc_idle:5G,not_restricted_rrc_con:None

This is why carriers differ: one shows 5G whenever the cell is 5G-capable (even in LTE idle), another only when the SCG is actually connected. Grace timers (5g_icon_display_grace_period_string and the secondary grace period) keep the icon from flickering when the SCG is released briefly. NR_ADVANCED can be driven by mmWave frequency range, aggregated bandwidth thresholds (nr_advanced_threshold_bandwidth_khz_int) or configured bands.

Enabling NSA and SA

  • KEY_CARRIER_NR_AVAILABILITIES_INT_ARRAY in carrier config declares whether the carrier supports NSA, SA or both on the device.
  • Allowed network types are set with setAllowedNetworkTypesForReason() using bitmasks that include NETWORK_TYPE_BITMASK_NR (user preference, carrier, power reasons combine).
  • The radio HAL exposes setNrDualConnectivityState(), so the framework can disable EN-DC, for example for power or thermal reasons.
  • VoNR is controlled by carrier config (KEY_VONR_ENABLED_BOOL, KEY_VONR_SETTING_VISIBILITY_BOOL) and the user toggle through TelephonyManager.setVoNrEnabled().
  • NR cell info comes as CellIdentityNr (NCI, PCI, TAC, NR-ARFCN, bands) and CellSignalStrengthNr (SS-RSRP, SS-RSRQ, SS-SINR and CSI variants).

DataNetworkController basics

Since Android 13 the data stack is built around DataNetworkController (it replaced the older DcTracker).

  1. Requests ConnectivityService sends network requests (INTERNET, IMS, MMS, ENTERPRISE, PRIORITIZE_LATENCY and so on) that become telephony network requests.
  2. Evaluation The controller checks data enabled, roaming, default data subscription, radio state and policy, then selects a DataProfile (the APN/DNN plus, on 5G, a traffic descriptor).
  3. Setup A DataNetwork state machine calls setupDataCall through the RIL with access network NGRAN or EUTRAN, the data profile and optional slice info. The response carries the interface, addresses, DNS, P-CSCF and QoS.
  4. Lifecycle DataRetryManager handles back-off and retry after failures (using the 5GSM or ESM cause), and AccessNetworksManager handles transport handover between cellular and IWLAN.

Slicing on Android: URSP rules from the network map app traffic to a TrafficDescriptor (DNN, OS app ID) and NetworkSliceInfo. Apps request capabilities such as NET_CAPABILITY_ENTERPRISE, NET_CAPABILITY_PRIORITIZE_LATENCY or NET_CAPABILITY_PRIORITIZE_BANDWIDTH, and the telephony stack sets up a PDU session on the matching slice.

Handy debug commands

adb shell dumpsys telephony.registry      # ServiceState, nrState, TelephonyDisplayInfo
adb shell dumpsys carrier_config           # 5g_icon_configuration_string, NR availabilities
adb shell dumpsys phone                    # data networks, NetworkTypeController state
adb logcat -b radio                        # RIL requests, PhysicalChannelConfig, data setup
Analogy

The 5G icon is like a restaurant's "open" sign that the manager can wire in different ways: one manager lights it when the kitchen is capable of cooking (NR available), another only when food is actually being served (SCG connected), and a delay switch keeps it from blinking when the chef steps away for a moment. The wiring rules are 5g_icon_configuration_string, the kitchen status is the NR state, and the delay switch is the grace timer.

Interview angle Device-side interviews ask "Why does the phone show 5G but the speed is LTE-like?" or "Why is there no 5G icon?". Explain that NSA registers on LTE, the icon is a display override from NR state and carrier config, and walk the chain from SIB2 and NAS to modem indications, the RIL, NetworkTypeController and SystemUI.

Quick revision

  • 5G targets three families: eMBB (throughput), URLLC (about 1 ms, 99.999%), mMTC (1 million devices per km²).
  • NSA Option 3/3a/3x uses the EPC with an LTE anchor and NR as secondary via EN-DC; SA Option 2 uses the 5GC with NR only.
  • In EN-DC the eNB is Master Node (MCG) and the en-gNB is Secondary Node (SCG), linked by X2.
  • Option 3 splits at the eNB PDCP, 3a splits in the core, 3x splits at the gNB PDCP (most common).
  • NR addition in NSA: LTE attach, B1 report, SgNB Addition over X2, RRC reconfiguration with NR config, RACH on PSCell, E-RAB modification.
  • An SCG failure keeps LTE up; the UE sends SCGFailureInformationNR.
  • 5GC NFs: AMF mobility and NAS, SMF sessions and UPF control, UPF user plane, AUSF auth, UDM subscription, PCF policy, NRF discovery, NSSF slice selection, NEF exposure.
  • Key interfaces: N1 NAS, N2 NGAP, N3 GTP-U, N4 PFCP, N6 to data network, N9 UPF to UPF, N11 AMF to SMF, N16 V-SMF to H-SMF, N26 AMF to MME, Xn gNB to gNB.
  • SBI is HTTP/2 + JSON with NRF-based discovery; CUPS lets UPFs sit at the edge.
  • Registration (AMF) replaces Attach and does not create a data session; PDU Session Establishment (SMF) does.
  • Registration types: initial, mobility update, periodic (T3512), emergency.
  • Registration Accept carries 5G-GUTI, TAI list, Allowed NSSAI and the IMS voice over PS indicator.
  • PDU session: DNN, S-NSSAI, type (IPv4/IPv6/IPv4v6/Ethernet/Unstructured), SSC mode 1/2/3.
  • SUCI hides the SUPI with ECIES and the home network public key; the SIDF in the UDM recovers it.
  • 5G-AKA lets the home AUSF verify RES*; user-plane integrity protection is new in 5G.
  • QoS flows (QFI) live inside one PDU session tunnel; SDAP maps flows to DRBs.
  • 5QI 1 GBR conversational voice, 5QI 2 video, 5QI 5 IMS signalling (not voice), 5QI 9 default internet, 82 to 86 delay-critical GBR. Do not invert 1 and 5.
  • ARP is admission and pre-emption priority, not packet treatment.
  • SCS = 15 × 2µ kHz; slots per subframe = 2µ; 14 symbols per slot.
  • FR1 is 410 MHz to 7.125 GHz (100 MHz carriers); FR2-1 is 24.25 to 52.6 GHz (400 MHz carriers, beamforming); FR2-2 is 52.6 to 71 GHz in Release 17.
  • Up to 4 BWPs per direction per cell, one active; default BWP saves power.
  • SSB = PSS + SSS + PBCH over 4 symbols and 20 RBs; 1008 PCIs; up to 64 beams in FR2.
  • SSB-to-RACH mapping tells the gNB the UE's best beam; beam failure recovery is faster than RLF.
  • LDPC for data, Polar for control; no always-on CRS in NR.
  • DSS shares one LTE carrier with NR dynamically at the cost of overhead.
  • SDAP is new; PDCP adds duplication; RLC drops concatenation; HARQ is asynchronous both ways.
  • RRC_INACTIVE keeps context in UE and anchor gNB, core sees CM-CONNECTED, resume with I-RNTI, RAN paging in the RNA.
  • VoNR needs SA, IMS PDU session, IMS VoPS indicator, 5QI 1 flows, gNB voice support (ROHC, C-DRX, configured grant, PUSCH repetition, RLC UM).
  • Voice-centric UE leaves 5GS when IMS voice is not available; data-centric stays. Usage setting is not the VoPS bit.
  • IMS roaming on 5G is usually home-routed (V-SMF/H-SMF on N16, N9 between UPFs), the S8HR analogue; LBO is uncommon for the IMS DNN.
  • EPS fallback: gNB rejects the 5QI 1 flow, moves the UE to LTE by handover (N26) or redirection, call completes as VoLTE.
  • EPS fallback is at call setup and stays IMS; SRVCC moves an active call to CS.
  • S-NSSAI = SST (8 bits) + SD (24 bits); up to 8 in an NSSAI; URSP maps apps to slices.
  • Xn handover ends with a Path Switch; N2 handover goes through the AMF.
  • The TAI list is the registration area; leaving it triggers a mobility registration update.
  • On Android NSA the network type is LTE and the 5G icon is an override from NetworkTypeController and 5g_icon_configuration_string.

Glossary

5G-AKA
5G Authentication and Key Agreement; challenge-response authentication where the home AUSF verifies the UE's RES*.
5G-GUTI
Temporary UE identity assigned by the AMF, used instead of the permanent identity after registration.
5GC
5G Core network, the service-based core used by SA.
5GMM / 5GSM
5G NAS mobility management and session management protocols.
5QI
5G QoS Identifier; an index to standard QoS characteristics such as delay budget and error rate.
AMF
Access and Mobility Management Function; terminates NAS, handles registration, paging and mobility.
ARP
Allocation and Retention Priority; decides admission and pre-emption under congestion.
AUSF
Authentication Server Function in the home network.
Beam failure recovery
Fast procedure where the UE switches to a new beam through RACH instead of declaring radio link failure.
BWP
Bandwidth Part; a sub-band of the carrier with its own numerology on which the UE operates.
CORESET
Control Resource Set; the time-frequency region where PDCCH is sent.
CUPS
Control and User Plane Separation; splitting SMF-type control from UPF forwarding.
DAPS
Dual Active Protocol Stack handover; UE keeps the source link until the target is ready.
DNN
Data Network Name; the 5G equivalent of an APN.
DRB
Data Radio Bearer carrying user data over the air.
DSS
Dynamic Spectrum Sharing; LTE and NR on the same carrier, split dynamically.
EN-DC
E-UTRA NR Dual Connectivity; the NSA mode with LTE master and NR secondary.
EPS fallback
Moving a 5G SA UE to LTE at voice call setup so the call runs as VoLTE.
FR1 / FR2
Frequency Range 1 (sub-7 GHz) and Frequency Range 2 (mmWave).
gNB
The NR base station; en-gNB when it serves as secondary node in EN-DC.
HARQ
Hybrid ARQ; fast MAC-layer retransmission with soft combining.
H-SMF / V-SMF
Home and visited Session Management Functions in a home-routed roaming PDU session, connected over N16.
I-RNTI
Identifier of a suspended UE context, used to resume from RRC_INACTIVE.
Massive MIMO
Antenna arrays with many elements for beamforming and multi-user spatial multiplexing.
MCG / SCG
Master Cell Group and Secondary Cell Group in dual connectivity.
N16
Interface between V-SMF and H-SMF for home-routed roaming session control.
N26
Interface between AMF and MME for seamless 5GS and EPS interworking.
NEF
Network Exposure Function; exposes network capabilities to external applications.
NGAP
NG Application Protocol used on N2 between gNB and AMF.
NRF
NF Repository Function; registry and discovery service for network functions.
NSA
Non-Standalone; NR added to an LTE anchor on the EPC.
NSSAI
List of S-NSSAIs; configured, requested, allowed or rejected.
NSSF
Network Slice Selection Function.
Numerology
The parameter µ that sets subcarrier spacing and slot length.
PCF
Policy Control Function; QoS, charging and UE policies.
PDU session
A connection between the UE and a data network, the 5G version of a PDN connection.
PFCP
Packet Forwarding Control Protocol on N4 between SMF and UPF.
PSCell
Primary cell of the secondary cell group.
QFI
QoS Flow Identifier carried in the GTP-U header.
RNA
RAN Notification Area; area in which an RRC_INACTIVE UE moves without notifying the network.
RRC_INACTIVE
NR RRC state with stored context allowing fast resume.
S-NSSAI
Single slice identifier made of SST and optional SD.
SA
Standalone; NR connected directly to the 5GC.
SCS
Subcarrier spacing, 15 × 2µ kHz.
SDAP
Service Data Adaptation Protocol; maps QoS flows to DRBs.
SIDF
Subscription Identifier De-concealing Function inside the UDM.
SMF
Session Management Function; manages PDU sessions and controls UPFs.
SSB
Synchronization Signal / PBCH block used for cell search and beam selection.
SSC mode
Session and Service Continuity mode (1, 2, 3) controlling anchor UPF changes.
SUCI
Subscription Concealed Identifier; encrypted form of the SUPI.
SUPI
Subscription Permanent Identifier, usually the IMSI.
TAI list
List of tracking areas forming a UE's registration area.
TCI state
Transmission Configuration Indicator; tells the UE which reference beam to assume.
TDD
Time Division Duplex; uplink and downlink share a frequency at different times.
UDM / UDR
Unified Data Management and its repository; the 5G equivalent of the HSS.
UPF
User Plane Function; forwards user packets and enforces QoS.
URLLC
Ultra-Reliable Low-Latency Communication.
URSP
UE Route Selection Policy; rules mapping app traffic to slices, DNNs and sessions.
Usage setting
UE policy, voice-centric or data-centric, that decides whether the UE may stay on 5GS without IMS voice. Not the network IMS VoPS indication.
VoNR
Voice over NR; IMS voice carried natively on 5G SA.
Xn
Interface between gNBs for handover and dual connectivity.

Interview questions

Fundamentals

What is 5G and why was it needed?

5G is the fifth mobile generation: the NR radio plus the service-based 5G Core. It was designed for three service families that 4G could not all serve well: eMBB (multi-Gbps throughput), URLLC (about 1 ms latency with 99.999% reliability) and mMTC (up to 1 million devices per km²). Key enablers are flexible numerology, mmWave spectrum, massive MIMO and beamforming, a cloud-native core with CUPS and edge UPFs, and network slicing.

NSA vs SA: what is the difference?

NSA (Option 3/3a/3x) keeps the 4G EPC and an LTE anchor for all control signalling, adding NR as a secondary carrier through EN-DC for throughput. Voice stays VoLTE. SA (Option 2) connects NR directly to the 5GC, with RRC and NAS on NR. SA enables slicing, RRC_INACTIVE, VoNR, URLLC and edge features. NSA was faster to launch; SA is the end state.

What is EN-DC?

E-UTRA NR Dual Connectivity: the UE is connected at the same time to an LTE eNB (Master Node, control anchor) and an NR en-gNB (Secondary Node, extra capacity), with the EPC as core. It is the basis of NSA Option 3 deployments. The eNB and gNB coordinate over X2.

What are MCG and SCG?

In dual connectivity, the Master Cell Group is the set of cells served by the master node (PCell plus SCells) and the Secondary Cell Group is the set served by the secondary node (PSCell plus SCells). In EN-DC the MCG is LTE and the SCG is NR. Losing the SCG does not drop the connection; losing the MCG does.

Name the main 5G Core network functions.
  • AMF: registration, NAS, mobility, paging.
  • SMF: PDU sessions, IP allocation, UPF control.
  • UPF: user-plane forwarding and QoS enforcement.
  • AUSF: authentication.
  • UDM/UDR: subscriber data.
  • PCF: policy.
  • NRF: service registry and discovery.
  • NSSF: slice selection.
  • NEF: capability exposure to external apps.
AMF vs SMF vs UPF in one line each?

AMF is the mobility brain (who and where the UE is), SMF is the session brain (which data connections exist and with what QoS), UPF is the muscle that forwards packets according to rules the SMF installs over N4.

Map EPC nodes to 5GC functions.

MME splits into AMF (mobility) and part of SMF (session); SGW-C and PGW-C become SMF; SGW-U and PGW-U become UPF; HSS becomes UDM/UDR with AUSF; PCRF becomes PCF; SCEF becomes NEF. NRF and NSSF are new.

What is a PDU session and a DNN?

A PDU session is the connection between the UE and a data network through a UPF, the 5G equivalent of an LTE PDN connection. It has an ID, type (IPv4, IPv6, IPv4v6, Ethernet, Unstructured), S-NSSAI, SSC mode and QoS flows. The DNN (Data Network Name) identifies the data network, the 5G name for APN, for example internet or ims.

How is 5G registration different from LTE attach?

LTE attach both registers the UE and creates a default bearer. 5G registration (5GMM, with the AMF) only authenticates and registers the UE and gives it a registration area and allowed slices. Data connectivity is a separate PDU Session Establishment (5GSM, with the SMF). Registration also has explicit types: initial, mobility update, periodic and emergency.

What is NR numerology?

Numerology µ sets the subcarrier spacing: 15 × 2µ kHz, giving 15, 30, 60, 120, 240 kHz (and 480/960 kHz in Release 17). A slot is always 14 symbols, so higher SCS means shorter slots (1 ms at 15 kHz, 0.5 ms at 30 kHz, 125 µs at 120 kHz) and lower latency, plus better robustness to mmWave phase noise. Lower SCS has a longer cyclic prefix, better for large cells with long delay spread.

FR1 vs FR2?

FR1 is 410 MHz to 7.125 GHz with up to 100 MHz carriers: good coverage and penetration. FR2 (mmWave) is 24.25 to 52.6 GHz (FR2-1) with up to 400 MHz carriers: multi-Gbps capacity but short range, easily blocked, and dependent on beamforming. Release 17 adds FR2-2 up to 71 GHz.

What is a bandwidth part (BWP)?

A contiguous subset of the carrier with its own numerology on which a UE operates. Up to 4 downlink and 4 uplink BWPs can be configured per cell, one active at a time. It saves power (narrow BWP when traffic is low), supports UEs that cannot handle the full carrier, and allows mixing numerologies. Switching is by DCI, RRC or an inactivity timer back to the default BWP.

What is the SSB?

The SS/PBCH block: PSS, SSS and PBCH over 4 OFDM symbols and 20 RBs. It gives timing, the Physical Cell ID (1008 values) and the MIB, which points to CORESET#0 and SIB1. SSBs are beam-swept in bursts (up to 64 beams in FR2), so the UE also uses them to find its best beam and for RSRP measurements.

What is a QoS flow and a QFI?

A QoS flow is the finest level of QoS differentiation in a PDU session. All packets in a flow get the same treatment. Each flow has a QFI (6-bit QoS Flow Identifier) carried in the GTP-U header on N3 and a QoS profile (5QI, ARP, bit rates). Many flows share one PDU session tunnel.

Which 5QI is used for VoNR media and signalling?

Do not invert them. 5QI 1 (GBR, 100 ms delay budget) is conversational voice media. 5QI 5 (non-GBR) is IMS SIP signalling on the default flow of the IMS PDU session. Video adds 5QI 2. 5G priority levels use a different scale (5QI 5 is 10, 5QI 1 is 20) but the relationship matches LTE: SIP is high priority relative to other default flows such as 5QI 9; voice is protected by being GBR.

What is SDAP and why is it new?

Service Data Adaptation Protocol is the top user-plane layer in NR when connected to the 5GC. It maps QoS flows to DRBs and marks the QFI on uplink packets. LTE did not need it because each EPS bearer mapped one-to-one to a radio bearer; 5G's QoS-flow model needs an explicit flow-to-DRB mapping. It is not used in EN-DC (EPC).

What are the RRC states in NR?

RRC_IDLE (no context in RAN, core paging), RRC_CONNECTED (active, network-controlled mobility) and the new RRC_INACTIVE, where the UE and anchor gNB keep the context and the core still sees the UE as CM-CONNECTED, allowing fast resume with little signalling.

What is VoNR?

Voice over NR: native IMS voice on 5G SA. SIP signalling runs on the IMS PDU session (5QI 5) and media on a 5QI 1 QoS flow over NR. It needs SA core, IMS integration with the PCF, gNB voice support (ROHC, C-DRX, configured grant, PUSCH repetition, RLC UM), UE support and the IMS voice over PS indicator in Registration Accept. The SIP flow is the same as VoLTE. NSA does not use VoNR; voice stays on the LTE anchor.

What is EPS fallback?

An interim voice solution for SA networks without VoNR. When a voice call starts, the gNB rejects the 5QI 1 flow and moves the UE to LTE by inter-system handover (over N26) or RRC release with redirection. The call completes as VoLTE on the same IMS session. The user sees 5G switch to LTE during calls.

What is network slicing and what is an S-NSSAI?

Slicing creates several isolated logical end-to-end networks on shared infrastructure, each tuned for a service (broadband, low latency, IoT, enterprise). A slice is identified by an S-NSSAI = SST (8-bit slice/service type, for example 1 eMBB, 2 URLLC, 3 MIoT, 4 V2X) plus an optional 24-bit SD (slice differentiator).

SUPI vs SUCI?

SUPI is the permanent subscription identity (usually the IMSI). SUCI is its concealed form: the MSIN is encrypted with the home network's public key using ECIES, while MCC/MNC and routing indicator stay visible for routing. It defeats IMSI catchers because the permanent ID is never sent in clear. The UDM's SIDF decrypts it.

Xn vs N2 handover?

Xn handover is prepared directly between gNBs over Xn; the core is only involved at the end through a Path Switch Request. N2 handover goes through the AMF (Handover Required, Request, Command, Notify) when no Xn exists or the AMF/UPF must change. Xn is faster with less core signalling.

What is massive MIMO?

Base station antenna arrays with many elements (32 or 64 transmit/receive chains, many more physical elements) that form narrow beams toward users and serve several users on the same time-frequency resources (MU-MIMO). It increases capacity and coverage, especially in mid-band TDD, and is essential for mmWave.

Why is mmWave hard to use?

High path loss and poor penetration: walls, foliage, rain, hands and bodies block the signal. Range is typically a few hundred metres with line of sight. It needs beamforming with fast beam tracking, several antenna modules on the phone, dense small cells, and uses more UE power and heat.

Going deeper

Explain Option 3 vs 3a vs 3x.

All use the EPC with LTE as master. Option 3: the S-GW sends user data only to the eNB, and the eNB PDCP splits it to NR over X2 (MCG split bearer); heavy load on the eNB and X2. Option 3a: the core sends each bearer directly to either eNB or gNB; no RAN split. Option 3x: the S-GW sends data to the gNB, whose PDCP splits some to LTE (SCG split bearer); most common because NR carries the bulk.

Walk through how NR is added in an NSA call.
  1. UE attaches on LTE, reporting EN-DC capability and band combinations; SIB2 upperLayerIndication shows NR is available.
  2. eNB configures a B1 measurement for NR; UE reports a good NR cell.
  3. eNB sends SgNB Addition Request over X2; gNB replies with the NR SCG configuration.
  4. eNB sends RRCConnectionReconfiguration with the NR config; UE replies Complete; eNB sends SgNB Reconfiguration Complete.
  5. UE performs RACH on the PSCell.
  6. For 3x, eNB sends E-RAB Modification Indication so the MME/S-GW move S1-U to the gNB.
What is a split bearer and how does uplink work on it?

A bearer whose single PDCP entity sends through two RLC legs, one on LTE and one on NR. In downlink the anchor node's PDCP decides which leg to use based on flow control. In uplink the network sets a primary path (usually NR or LTE) and ul-DataSplitThreshold: while the uplink buffer is below the threshold the UE transmits only on the primary path; above it the UE may use both legs.

What happens on an SCG failure?

The UE detects a problem on NR (radio link failure on the PSCell, T310 expiry, RACH failure, reconfiguration or sync failure). It suspends SCG transmission but keeps the LTE MCG connection, and sends SCGFailureInformationNR with the cause and measurements. The eNB can release the SN or re-add NR on a better cell. Data continues on LTE.

Describe the service-based architecture and the NRF's role.

Control-plane NFs expose services (for example Nsmf_PDUSession) over HTTP/2 with JSON on the SBI. Consumers call producers directly or via an SCP. Each NF registers its profile (type, supported DNNs and slices, capacity) with the NRF; consumers query the NRF to discover a suitable instance. This allows scaling out, failover and adding new NFs without reconfiguring every peer.

Name the N1 to N11 interfaces.

N1 UE-AMF (NAS); N2 gNB-AMF (NGAP); N3 gNB-UPF (GTP-U); N4 SMF-UPF (PFCP); N5 AF-PCF; N6 UPF-data network; N7 SMF-PCF; N8 AMF-UDM; N9 UPF-UPF; N10 SMF-UDM; N11 AMF-SMF. Also useful: N12 AMF-AUSF, N13 AUSF-UDM, N15 AMF-PCF, N16 V-SMF to H-SMF (home-routed roaming), N22 AMF-NSSF, N26 AMF-MME, N32 SEPP to SEPP.

What is CUPS and why does it matter for 5G?

Control and User Plane Separation: the SMF makes decisions and the UPF only forwards. UPFs can be scaled separately and placed close to users (edge or on-premises) for low latency and local breakout (MEC), while control stays central. Several UPFs can be chained on N9 (intermediate UPF plus anchor UPF).

Walk me through 5G initial registration.
  1. RRC setup; RRCSetupComplete carries Registration Request (SUCI or GUTI, requested NSSAI, capabilities).
  2. gNB selects an AMF and sends Initial UE Message.
  3. AMF selects AUSF; UDM deconceals SUCI and generates vectors; 5G-AKA runs and the AUSF verifies RES*.
  4. NAS Security Mode Command/Complete.
  5. AMF registers with UDM and fetches subscription; may query NSSF and set up PCF policy.
  6. Initial Context Setup to the gNB; AS security and RRC reconfiguration.
  7. Registration Accept (5G-GUTI, TAI list, Allowed NSSAI, network feature support); Registration Complete.
Walk me through PDU session establishment.
  1. UE sends PDU Session Establishment Request (session ID, DNN, S-NSSAI, type, SSC mode) in UL NAS Transport.
  2. AMF discovers an SMF via NRF and calls CreateSMContext.
  3. SMF checks UDM subscription and gets policy from PCF.
  4. SMF selects a UPF, allocates the IP and sends PFCP Session Establishment.
  5. SMF sends N1N2MessageTransfer: N2 QoS info to the gNB and N1 Accept to the UE.
  6. gNB sets up DRBs with RRCReconfiguration and returns its downlink tunnel ID.
  7. SMF updates the UPF with the downlink tunnel; data flows.
What registration types exist and when is each used?

Initial at power on or first entry to 5GS; mobility registration update when entering a TA outside the TAI list or when capabilities or slices change; periodic registration update when T3512 expires; emergency registration for emergency services, possible in limited service.

Explain SSC modes.

Session and Service Continuity modes decide whether the anchor UPF (and IP address) can change. SSC 1: anchor never changes (IMS, default). SSC 2: break before make; the session is released and re-established with a new anchor. SSC 3: make before break; a new session is created before releasing the old one. Modes 2 and 3 let edge applications move closer to the user.

How does 5G QoS differ from LTE bearers?

LTE had one GTP tunnel per EPS bearer and a one-to-one bearer to radio bearer mapping, with TFTs to classify traffic. 5G uses one N3 tunnel per PDU session, with QoS flows marked by QFI inside it. The UPF uses packet detection rules (downlink) and the UE uses QoS rules (uplink). SDAP maps flows to DRBs, and several flows can share one DRB. 5QI replaces QCI. It is flatter and needs less signalling to add a flow.

What is reflective QoS?

A mechanism where the UE derives uplink QoS rules from downlink traffic. The UPF sets the RQI (Reflective QoS Indicator) on downlink packets; the UE creates a matching uplink rule (with swapped addresses and ports) with the same QFI. It avoids explicit signalling for many short-lived flows. Both the UE and network must support it.

What is ARP and how is it different from 5QI?

ARP (Allocation and Retention Priority) is used at admission and under congestion: whether a flow may be admitted, whether it can pre-empt others, and whether it can be pre-empted. 5QI defines packet treatment: resource type, priority level, delay budget and error rate. An emergency call has high ARP so it can pre-empt other GBR flows.

At 30 kHz SCS, how long is a slot, and what is the period of a DDDSU pattern?

µ = 1, so there are 2 slots per 1 ms subframe: a slot is 0.5 ms, and each of its 14 symbols is about 35.7 µs including CP. DDDSU is 5 slots, so it repeats every 2.5 ms. DDDDDDDSUU would be 10 slots, a 5 ms period.

How do TDD patterns affect performance?

More downlink slots give more downlink capacity but fewer uplink opportunities, which hurts uplink throughput and adds latency for HARQ feedback and uplink grants. Short periods (DDDSU) reduce latency; long periods reduce switching overhead. Neighbouring operators in the same band must align patterns to avoid downlink-to-uplink interference. The special slot's guard period sets the maximum cell range.

What is DSS and what are its trade-offs?

Dynamic Spectrum Sharing runs LTE and NR on the same carrier, sharing resources dynamically per scheduling interval. Benefit: fast wide NR coverage (often low band) without refarming spectrum. Costs: overhead from avoiding LTE CRS and control regions (rate matching, MBSFN subframes), so NR capacity on DSS is lower than on a dedicated carrier; NR must use 15 kHz SCS to align with LTE.

How does carrier aggregation work in NR, and why do band combinations matter?

NR aggregates up to 16 component carriers (PCell plus SCells), possibly mixing FDD low band and TDD mid band. In NSA, LTE CA and NR carriers are combined as an EN-DC band combination. The UE must declare each supported combination in its capability; if the network's deployed combination is missing from the UE capability, NR or CA simply will not be configured.

What are SUL and SDL?

SUL (Supplementary Uplink) pairs a high-band TDD cell with an extra low-frequency uplink-only carrier to extend uplink coverage at the cell edge, where UE power is the limit. SDL (Supplementary Downlink) is a downlink-only carrier added for capacity.

Explain RRC_INACTIVE in detail.

The gNB sends RRCRelease with suspendConfig (I-RNTI, RAN Notification Area, periodic RNA update timer). The UE and anchor gNB keep the AS context; N2 and N3 stay up so the core sees CM-CONNECTED. The UE reselects freely within the RNA. Downlink data triggers RAN paging in the RNA. To resume, the UE sends RRCResumeRequest with the I-RNTI; a new gNB fetches the context over Xn, then RRCResume and Complete restore security and bearers without NAS signalling. Benefits: lower latency and battery, less core signalling.

HARQ vs ARQ, and how does NR use HARQ?

HARQ (MAC) is fast retransmission with soft combining of failed attempts; ARQ (RLC AM) is a slower, reliable retransmission of segments that HARQ missed. NR HARQ is asynchronous in both directions, supports up to 16 processes, and uses flexible timing: K1 (PDSCH to HARQ-ACK) and K2 (grant to PUSCH) are signalled per transmission. It can also retransmit parts of a block (code block group based HARQ).

What is PDCP duplication?

PDCP sends the same packet over two RLC legs (two carriers in CA or two nodes in DC); the receiver discards duplicates. It improves reliability and cuts latency without waiting for retransmission, and is a key tool for URLLC. It can be activated and deactivated dynamically by MAC CE.

EPS fallback vs SRVCC?

EPS fallback happens at call setup, moves the UE from NR to LTE and keeps the call in IMS (it becomes VoLTE). SRVCC happens during an active call, moving it from PS (VoLTE, or NR in 5G-SRVCC) to 2G/3G CS, with IMS access transfer. Different trigger, timing and target domain.

Voice-centric versus data-centric on 5GS?

The 5GS usage setting (TS 24.501) is a UE policy, not the IMS VoPS bit in Registration Accept. A voice-centric UE that cannot get IMS voice on 5GS disables N1 mode for that PLMN and tries EPS. A data-centric UE stays on 5GS for data and only leaves when a call actually needs fallback. This pair of flags is a common reason a device never camps on SA even though SA cells are on air.

How does 5G roaming carry an IMS PDU session?

Usually home-routed: visited V-SMF and V-UPF, home H-SMF and H-UPF (the IP and P-CSCF stay at home), N16 between SMFs and N9 between UPFs. That is the S8HR idea on 5GC interfaces. LBO (visited SMF/UPF and local N6) is specified but uncommon for IMS. Inter-operator SBI is protected by SEPP on N32. Same S8HR costs apply: emergency, visited LI, no local ATCF, trombone delay.

What is N26 and why does it matter?

The interface between AMF and MME. It lets them exchange UE mobility and session context, so moves between 5GS and EPS (EPS fallback, VoNR to VoLTE handover, idle mobility) keep the IP address and sessions with minimal signalling. Without N26, the UE has to re-register or attach on the other system, which is slower and less reliable.

Explain NSSAI types and how the NSSF is involved.

Configured NSSAI is provisioned in the UE per PLMN; the UE sends a Requested NSSAI at registration; the network returns Allowed NSSAI (valid in the registration area) and possibly Rejected NSSAI with causes. Subscribed S-NSSAIs are in the UDM. The AMF may ask the NSSF, which decides the Allowed NSSAI and whether another AMF set should serve the UE (causing rerouting).

What is URSP?

UE Route Selection Policy, delivered by the PCF through the AMF. Each rule has a precedence, a traffic descriptor (application ID, DNN, IP or FQDN descriptors, connection capabilities) and route selection descriptors (S-NSSAI, DNN, SSC mode, session type, access type). The UE OS uses it to decide which PDU session, and therefore which slice, carries an app's traffic.

How does idle-mode mobility work and what is a registration area?

In idle, the UE picks cells itself using priorities and thresholds from SIBs. The AMF assigns a TAI list (registration area); the UE moves freely within it and sends a mobility registration update only when entering a TA outside the list, plus periodic updates on T3512. Paging for downlink data is sent across the registration area. A bigger list means fewer updates but more paging load.

Which measurement events trigger handovers?

A3 (neighbour better than serving by an offset) for most intra-frequency handovers; A5 (serving below one threshold and neighbour above another) for coverage-based handover; A2/A1 to start/stop inter-frequency measurements; B1 (inter-RAT neighbour above threshold) for NR addition in EN-DC or LTE target in EPS fallback; B2 for NR to LTE at coverage edge. Hysteresis and time-to-trigger prevent ping-pong.

How is 5G-AKA different from EPS-AKA?

The math (RAND, AUTN, MILENAGE or TUAK in the SIM) is similar, but in 5G-AKA the UE returns RES* and the home network's AUSF confirms it, so the home operator has proof the UE was authenticated even when roaming. It works with SUCI concealment, has a new key hierarchy (KAUSF, KSEAF, KAMF), and EAP-AKA' is an equal alternative.

How do Service Request and paging work in 5G?

A CM-IDLE UE with uplink data sends a Service Request to move to CM-CONNECTED and reactivate user-plane resources for its PDU sessions. For downlink data, the UPF notifies the SMF, which asks the AMF (Namf_Communication_N1N2MessageTransfer) to page the UE across the registration area; the UE responds with a Service Request. For RRC_INACTIVE UEs, the RAN pages instead.

Why does NR use LDPC and Polar codes?

LDPC is used for data (PDSCH/PUSCH) because it decodes in parallel with high throughput and low latency at large block sizes, and supports incremental redundancy for HARQ. Polar codes are used for control (DCI, UCI, PBCH) because they perform well for small blocks. LTE's Turbo codes did not scale as well to multi-Gbps rates.

Advanced

Why did NR remove the always-on CRS, and what replaced it?

LTE's CRS is sent in every subframe across the whole band even with no users, which wastes energy, adds interference and fixes the antenna port structure. NR is a "lean carrier": the SSB (sparse, periodic) handles sync and idle measurements, DMRS is sent only with data for demodulation, CSI-RS is configured per UE for channel and beam measurement, TRS for tracking and PTRS for phase noise. This saves network energy and makes beamforming of all signals possible.

How does SSB-to-RACH mapping help beam-based initial access? What is 2-step RACH?

Each SSB index is associated with specific RACH occasions and preambles. The UE sends its preamble on the occasion mapped to its best SSB, so the gNB learns the best downlink beam from when and where it receives the preamble and can answer on that beam. 2-step RACH (Release 16) combines Msg1 and Msg3 into MsgA (preamble plus PUSCH payload) and Msg2 and Msg4 into MsgB, reducing access latency, useful in small cells and unlicensed spectrum.

Explain beam management P1, P2, P3, TCI states and beam failure recovery.

P1: gNB sweeps SSB beams and the UE picks the best (coarse). P2: gNB refines its transmit beam with CSI-RS on narrower beams. P3: the UE sweeps its receive beam on a fixed gNB beam. The gNB tells the UE which beam to assume via TCI states (activated by MAC CE, indicated in DCI) that reference a quasi-co-located SSB or CSI-RS. For BFR, the UE counts beam failure instances on its detection reference signals; when the count reaches the maximum it selects a candidate beam above threshold and sends a contention-free RACH on it; the gNB answers on that beam. BFR recovers in tens of ms rather than a full RLF and re-establishment.

Analog vs digital vs hybrid beamforming?

Analog: phase shifters in RF steer one beam per RF chain; cheap and low power but only one direction at a time per chain. Digital: each element has its own baseband chain, allowing many simultaneous beams and precise MU-MIMO, but costly and power-hungry at wide bandwidth. Hybrid: a few digital streams each drive an analog-beamformed sub-array; the practical choice for mmWave. Mid-band massive MIMO is mostly digital.

Explain the gNB CU/DU split.

The gNB can be split into a Central Unit (RRC, SDAP, PDCP; often virtualised in a data centre) and Distributed Units (RLC, MAC, high PHY; near the radio) connected by F1. The CU can be split into CU-CP and CU-UP over E1. This pools higher-layer processing, simplifies dual connectivity and mobility within a CU, and lets the UP be placed near edge UPFs. O-RAN further splits the DU and radio unit (split 7.2) with open fronthaul.

How does 5G achieve URLLC targets?

Radio: higher numerology and mini-slots for short transmissions, configured-grant (grant-free) uplink to skip scheduling requests, pre-emption of eMBB traffic (downlink pre-emption indication), low-rate robust MCS tables, fast HARQ with short K1, PDCP duplication over two carriers, and multi-TRP diversity. Core: edge UPF to cut transport delay, delay-critical GBR 5QIs, redundant N3 tunnels or dual PDU sessions. Reliability of 99.999% comes from diversity and conservative link adaptation.

Why might VoNR have lower latency than VoLTE?

Shorter slots with higher numerology and faster HARQ timing reduce air-interface delay; RRC_INACTIVE allows quick resume for MT calls; there is no inter-RAT move at setup (unlike EPS fallback); and the flatter 5GC with possibly local UPFs can shorten the media path. In practice mouth-to-ear latency is dominated by codec and jitter buffer, so gains are modest, while call setup time gains over EPS fallback are large.

EPS fallback: handover-based vs redirection-based. Which is better and why?

Handover-based (N2 inter-system handover through N26): the target LTE cell is prepared in advance, the UE moves in one step and bearers are ready; typically a few hundred ms added. Needs B1 measurements and N26. Redirection-based: the gNB releases the UE with an LTE carrier; the UE must find the cell, do RRC setup and a TAU, then the dedicated bearer is created; often 1 to 3 s added and sensitive to LTE coverage. Handover is faster and more reliable; redirection is simpler and a fallback when measurements are not configured.

How is IP continuity preserved when moving between 5GS and EPS, with and without N26?

Both rely on a combined SMF+PGW-C and UPF+PGW-U anchor, so the same node serves the PDU session and the PDN connection. With N26, the AMF passes context to the MME, and the SMF+PGW-C maps QoS flows to EPS bearers, so the session moves transparently. Without N26, the UE attaches or registers on the target and requests the session with a "handover" indication; the combined node recognises it and reuses the IP. The UE must support the relevant interworking mode, and there is a longer interruption.

How do emergency calls work on 5G SA without emergency support over NR?

The Registration Accept indicates support for Emergency Services Fallback. When the user dials an emergency number, the UE sends a Service Request with an emergency services fallback indication; the AMF triggers the gNB to hand over or redirect the UE to LTE (or to E-UTRA on 5GC), where the emergency PDN and emergency IMS call are set up. If emergency over NR is supported, the UE creates an emergency PDU session and places the call natively.

What are Conditional Handover and DAPS handover?

CHO: the source prepares one or more candidate targets in advance and sends the UE their configurations with execution conditions (for example A3-like). The UE executes when a condition is met, which avoids failures when the radio degrades too fast to send a report and receive a command. DAPS: during handover the UE keeps receiving (and sending) on the source until the target link is established, giving near-zero interruption; it needs dual-stack UE support.

Walk through an Xn handover including the path switch and end marker.

A3 report, Handover Request over Xn with UE context, target admits and returns RRCReconfiguration with reconfigurationWithSync. The source sends it to the UE, sends SN Status Transfer and forwards buffered and in-flight downlink data over Xn-U. The UE does RACH on the target and sends Complete. The target sends NGAP Path Switch Request; AMF tells SMF, which updates the UPF over N4 to send downlink to the target. The UPF sends end marker packets on the old path so the source knows forwarding is finished and the target can deliver in order. The target then releases the UE context at the source.

How is SUCI computed, and what is the null scheme?

The USIM holds the home network public key and its identifier. The USIM or ME generates an ephemeral ECC key pair, derives a shared secret with the home public key (ECDH), and encrypts the MSIN with a key derived from it, adding a MAC tag. SUCI contains: SUPI type, MCC/MNC, routing indicator, protection scheme ID (null, profile A X25519, profile B P-256), home key ID, and the scheme output (ephemeral public key, ciphertext, MAC). With the null scheme the MSIN is sent unencrypted, used when no key is provisioned, and for some emergency cases.

Describe the 5G key hierarchy.

K (in USIM and ARPF/UDM) produces CK and IK during AKA. From these, KAUSF is derived (in AUSF and UE), then KSEAF (anchor key for the serving network), then KAMF. KAMF yields KNASint and KNASenc for NAS, and KgNB, which yields KRRCint, KRRCenc, KUPint and KUPenc. On handover, KgNB is refreshed through NH/NCC chaining so a compromised gNB cannot derive keys of earlier or later hops.

Why was user-plane integrity protection added in 5G, and why is it optional?

LTE only ciphered user data, and research showed attackers could flip bits in encrypted packets (for example DNS redirection attacks). 5G adds optional UP integrity (NIA algorithms) negotiated per PDU session by the SMF policy. It is optional because it costs processing and a 4-byte MAC-I per packet, which is expensive at multi-Gbps rates; it is typically enabled for sensitive or low-rate traffic.

What are NSSAA and slice quotas?

NSSAA (Network Slice-Specific Authentication and Authorisation, Release 16) adds a second, per-slice authentication with an external AAA server (for example an enterprise) using EAP after primary authentication; the slice is only added to the Allowed NSSAI after success. Slice quotas (Release 17) limit the number of UEs or PDU sessions per slice, enforced by a network slice admission control function; excess requests are rejected with slice-specific causes and back-off.

How does VoNR differ from VoLTE?

Same IMS SIP ladder and codecs. VoNR needs SA, a 5QI 1 QoS flow (not 5QI 5), N5/N7 policy, and NR radio features (configured grant, PUSCH repetition, C-DRX, ROHC, RLC UM). VoLTE uses a QCI 1 bearer, Rx/Gx, SPS and TTI bundling on LTE. NSA voice is still VoLTE. If VoNR is missing, the call uses EPS fallback at setup; if NR coverage ends mid-call, an N26 handover to VoLTE should keep the session. Mouth-to-ear still tracks G.114 ~150 ms; the big win over EPS fallback is setup time, not MOS.

An operator is migrating from NSA to SA. What is the voice strategy?

Phase 1 (NSA): VoLTE on the LTE anchor, SRVCC or CSFB for legacy edges. Phase 2 (early SA): SA for data, EPS fallback for voice, with N26 and handover-based fallback to keep setup time low, plus fast return to NR after calls. Phase 3: enable VoNR cluster by cluster once NR coverage and uplink are good, keep EPS fallback and VoNR-to-VoLTE handover as the backstop. KPIs: call setup time, setup success, fallback rate, drop rate, time to return to NR. Device side: carrier config for VoNR, IMS feature tags, IOT testing.

How does Android decide the 5G icon on NSA?

The modem reports registration info including EN-DC availability (from SIB2 upperLayerIndication), DCNR restriction (from NAS) and physical channel configs. The framework derives an NR state: NONE, RESTRICTED, NOT_RESTRICTED or CONNECTED. NetworkTypeController combines that with RRC state, frequency range and bandwidth, and carrier config (5g_icon_configuration_string, grace timers, NR advanced thresholds) to produce the TelephonyDisplayInfo override (NR_NSA or NR_ADVANCED), which SystemUI renders. The underlying network type stays LTE.

What does DataNetworkController do, and how does 5G slicing reach an app on Android?

DataNetworkController (Android 13+) receives telephony network requests from ConnectivityService, evaluates whether data is allowed (data switch, roaming, DDS, policy), picks a DataProfile, and creates DataNetwork state machines that call setupDataCall through the RIL (with access network NGRAN, slice info and traffic descriptor). Retries are handled by DataRetryManager. For slicing, URSP rules from the network are matched against the app; an app requesting NET_CAPABILITY_ENTERPRISE or PRIORITIZE_LATENCY gets a network backed by a PDU session on the matching S-NSSAI with the right TrafficDescriptor.

Which RLC and MAC changes in NR reduce latency?

RLC no longer concatenates SDUs and no longer delivers in order (reordering moved to PDCP), so RLC PDUs can be built before the uplink grant size is known. MAC places a subheader immediately before each MAC SDU, allowing the PDU to be processed on the fly instead of reading a header block first. HARQ is asynchronous with signalled timing, and processing times are shorter (UE capability 2 for fast processing).

What is RedCap and why was it introduced?

Reduced Capability NR (Release 17) targets wearables, industrial sensors and video surveillance that need more than LTE-M/NB-IoT but much less than a smartphone. RedCap UEs support narrower bandwidth (20 MHz FR1, 100 MHz FR2), 1 or 2 receive antennas, half-duplex FDD option and lower modulation, plus extended DRX. It lowers cost and power while staying on the 5G core and supporting slicing.

How is roaming signalling secured in 5G?

Each operator deploys a SEPP (Security Edge Protection Proxy) at its border. SBI messages between visited and home networks go SEPP to SEPP over N32, with TLS and application-layer protection (PRINS) that can encrypt or integrity-protect selected JSON fields while letting intermediate IPX providers modify only allowed parts. This replaces the weakly authenticated SS7/Diameter interconnect. Home control over authentication (AUSF checks RES*) also limits fraud by visited networks.

Scenario & debugging

On NSA, LTE data works but NR is never added. How do you debug?
  1. Check the cell: does SIB2 carry upperLayerIndication? Is the UE in a DCNR-restricted state from NAS (subscription "NR as secondary RAT not allowed")?
  2. Check UE capability: is EN-DC enabled, and does it list the band combination the network uses (LTE anchor plus NR band)? Check Android allowed network types and whether EN-DC was disabled via setNrDualConnectivityState (power or thermal).
  3. Check RRC: was a B1 measurement configured, did the UE report NR, was the threshold met?
  4. Check X2: SgNB Addition Request rejected, or PSCell misconfigured?
  5. Check RACH on PSCell and SCG failure reports.

Use modem RRC OTA logs and compare against a reference device on the same cell.

NSA users see frequent SCG failures and the NR leg keeps being added and released. What do you look at?

Read the SCGFailureInformationNR cause: T310 expiry or RLF (NR coverage or interference), RACH failure on the PSCell (uplink power, PRACH config), sync or reconfiguration failure (config compatibility). Check B1 thresholds too low (adding NR at the edge), missing A2-based SN release, TDD pattern misalignment with neighbours, and uplink path (NR uplink at edge). The fix is often raising B1 thresholds, adding hysteresis, making LTE the uplink primary path at edge, or fixing the PSCell config. Also check UE-side blacklisting timers that stop re-adds.

The phone is in a 5G area but shows no 5G icon. How do you root-cause?
  1. Is the device NSA or SA? On SA check data network type is NR; on NSA check the NR state in dumpsys telephony.registry.
  2. If NR state is NONE: the cell does not advertise EN-DC or the modem does not report it.
  3. If RESTRICTED: the core restricts NR for this subscription (plan or provisioning issue).
  4. If NOT_RESTRICTED but no icon: check 5g_icon_configuration_string; some carriers only show 5G when CONNECTED.
  5. If CONNECTED but no icon: framework issue in NetworkTypeController or SystemUI, or physical channel config not reported by the modem.
  6. Also check carrier config NR availabilities, allowed network types (user may have chosen LTE), and SIM or APN provisioning.
The phone shows 5G but throughput is LTE-like. Why?

Common reasons: the icon is configured to show 5G when NR is only available (NOT_RESTRICTED in idle), not actually connected; the SCG is added but carrying little traffic (Option 3 with split at eNB, or flow control favouring LTE); the NR carrier is DSS or low band with limited bandwidth; poor NR radio (low SINR, low rank); uplink limitation on the LTE anchor for TCP; or core or backhaul bottleneck. Check the physical channel config, SCG bandwidth, MCS and rank, and the bearer split in modem logs.

On SA, voice calls take 3 to 5 seconds longer to connect than on LTE. What is happening?

Almost certainly EPS fallback, probably redirection-based or without N26. Timeline: INVITE, 5QI 1 request rejected by gNB, release with redirect, LTE cell search, RRC setup, TAU, dedicated bearer, then SIP continues. Verify in logs: NGAP modify reject with EPS fallback cause, RRC release with redirected carrier, TAU timing. Improvements: handover-based fallback with B1 measurements, deploy N26, tune LTE target carrier, start fallback early (on INVITE detection), and ultimately enable VoNR.

Users on SA miss incoming calls; the caller hears ringing then the call fails. Where do you look?

MT EPS fallback adds paging on NR, RRC setup, INVITE delivery, then the move to LTE. If this exceeds network SIP timers (for example the terminating side's no-answer or 180 timers) the call is cancelled. Check paging success on NR, whether the UE was in RRC_INACTIVE and resume succeeded, the EPS fallback duration, TAU success on LTE, and whether the INVITE retransmitted or a CANCEL arrived. Also check that the IMS registration was still valid over NR and that the P-CSCF could reach the UE after the move.

A user says "5G drops to LTE every time I make a call". Is it a bug?

Usually not. On SA without VoNR it is EPS fallback, working as designed. On NSA the icon may drop because the operator releases the SCG during VoLTE calls. Confirm by checking whether VoNR is supported by the network and enabled in carrier config, and whether the network sends the fallback trigger. It is a bug only if VoNR is supposed to be active (network and device configured) and the UE still falls back, or if it fails to return to 5G after the call.

After a call ends, the phone stays on LTE for minutes. Why, and how is it fixed?

After EPS fallback the UE is connected on LTE; if the eNB simply releases it without NR priority or redirection, the UE returns only through idle reselection, which depends on NR priority in SIB24 (NR frequencies for LTE reselection) and thresholds, or the UE stays connected due to background data. Fixes: "fast return" (release with redirection to NR or B1-triggered handover to NR), correct reselection priorities, and checking the UE is not blocking NR (thermal, battery saver, user setting).

The device never camps on SA, only NSA, even though SA is live. What do you check?

Carrier config KEY_CARRIER_NR_AVAILABILITIES_INT_ARRAY must include SA; the modem's SA enablement for that carrier; the SIM or subscription must allow 5GS (otherwise 5GMM #7 or #27 N1 mode not allowed, after which the UE disables N1 mode); allowed network types; UE support for the SA band; network cell reselection priorities pointing to the SA layer; and whether the UE previously got a reject and is in a back-off or disabled state. Also check the usage setting: a voice-centric UE that does not see IMS VoPS on SA (and has no usable EPS fallback path) will disable N1 and stay on LTE/NSA. Check NAS logs for a Registration Request on NR.

Registration is rejected with 5GMM cause #62. What does it mean?

"No network slices available": none of the requested S-NSSAIs can be allowed (not subscribed, not supported in this TA, or rejected by NSSF), and no default slice is available. Check the Requested NSSAI the UE sends (from configured NSSAI or URSP), the UDM subscribed S-NSSAIs and default flags, TA slice support, and Rejected NSSAI causes in the reject. Often it is a provisioning mismatch between the UE's configured NSSAI and the subscription.

PDU session setup fails on SA with 5GSM #27 but works on LTE. Why?

#27 is "missing or unknown DNN". The DNN in the device's data profile may differ from what the 5GC subscription expects (for example the LTE APN name works through PGW mapping but the SA subscription lists a different DNN or requires a slice). Check the requested DNN and S-NSSAI in the NAS message, UDM session subscription for that DNN in that slice, and the device's APN/DNN config for NGRAN. Also check PDU session type (IPv4v6 vs IPv6 only) and SSC mode.

IMS registration succeeds on LTE but fails on SA. How do you debug?
  1. Did the IMS PDU session come up (DNN ims, correct S-NSSAI)? Look for 5GSM rejects.
  2. Did ePCO include P-CSCF addresses?
  3. Did Registration Accept indicate IMS voice over PS over 3GPP? Without it the UE may not register IMS for voice on NR.
  4. Is the IMS stack using the right access type and feature tags for NR, and is VoNR enabled in carrier config?
  5. SIP trace: 403 or timeouts at the P-CSCF can mean the IMS core does not accept NR access or the IP pool differs.
A VoNR call drops when the user leaves NR coverage. What should happen and what failed?

The gNB should configure B2 (NR below threshold, LTE above) and perform an N2 inter-system handover to LTE via N26, with the 5QI 1 flow mapped to a QCI 1 bearer, keeping the IMS call. Failures: no B2 or LTE neighbours configured, no N26, handover preparation rejected by the MME (QoS mapping or capacity), UE handover failure, or the move happening too late (RLF before the command). Check measurement config, NGAP Handover Required/Command, and RRC logs. If no LTE is available, 5G-SRVCC to 3G would be needed but is rarely deployed.

mmWave throughput is poor and unstable. What do you suspect?

Blockage (hand, body, glass), non-line-of-sight, frequent beam switches or beam failure recovery, wrong antenna module chosen, BWP stuck on a narrow default, low rank or MCS, thermal throttling or SAR back-off reducing uplink, and uplink carried on LTE or FR1 in NSA. Check SSB/CSI-RS RSRP per beam, BFR counts, active BWP, rank and MCS, antenna module switching logs and thermal state.

Battery drain is much higher on 5G NSA than LTE. Why and what can be done?

NSA keeps two radios (LTE and NR) active, NR wideband receive and mmWave modules consume a lot, and the SCG may stay configured during low-traffic periods. Mitigations: network-side C-DRX tuning, BWP switching to a narrow default, SCG release on inactivity, UE-side dynamic EN-DC disable for low-throughput apps or screen off (Android can disable EN-DC through the radio HAL), and moving to SA with RRC_INACTIVE and wake-up signals.

Uplink throughput in NSA at the cell edge is poor even though NR downlink is fine. Why?

At the edge, NR mid band (TDD, higher frequency) uplink is power-limited, and the UE shares total power between LTE and NR (dynamic power sharing, or single-uplink operation for some band combinations). The primary uplink path may be NR when LTE would be better. Fixes: set LTE as uplink primary path at edge, tune ul-DataSplitThreshold, use SUL or low-band NR, and check that the UE's TDM pattern for single uplink is configured correctly.

An enterprise app should use a dedicated slice but its traffic goes over the default internet slice. How do you debug?

Check the chain: does the network deliver URSP rules to the UE (PCF, UE policy container in NAS)? Does the rule's traffic descriptor match the app (OS ID and app ID, DNN or FQDN)? Does the app request the right capability (for example NET_CAPABILITY_ENTERPRISE) and does the device profile support it? Is the target S-NSSAI in the Allowed NSSAI in the current area? Did the PDU session on that slice get rejected (5GSM #69 or subscription)? Look at URSP evaluation logs and the setupDataCall request's slice info.

RRC resume from RRC_INACTIVE often fails and falls back to a full RRC setup. What might be wrong?

The new gNB cannot retrieve the context: no Xn to the anchor gNB, anchor gNB released the context (timer too short or overload), or I-RNTI or resume MAC-I verification failure (security keys out of sync). Also periodic RNA update failures, or the UE moved outside the RNA without updating. Check Retrieve UE Context failures on Xn, RNA configuration and timers, and gNB context retention settings. Impact: extra latency and NAS signalling.

After moving from 5G SA to LTE, data stalls for several seconds. What do you check?

Whether N26 is deployed; without it the UE must attach or re-establish the PDN with a handover indication, and apps see a gap or even an IP change. Check the TAU or attach sequence, whether the IP address was preserved (SMF+PGW-C anchor), whether the Android data stack tore down and rebuilt the network (new interface, sockets reset), and DNS or MTU changes. Also check that QoS flows mapped to EPS bearers correctly and the default bearer came up without ESM rejects.

On a DSS carrier, NR throughput is much lower than expected. Is this normal?

Partly. DSS shares the carrier with LTE, so NR loses resources to LTE CRS, LTE control region and LTE traffic, often 10 to 30% overhead; if LTE load is high, NR gets less. Low band DSS carriers are also narrow (10 to 20 MHz). Check the LTE and NR load split, rate-matching configuration, MBSFN usage and scheduler sharing policy. For capacity, a dedicated NR mid-band carrier is required; DSS is mainly for coverage and the 5G icon.