Quick Read

One definition and the points that matter, for each core 5G topic. When one needs more than a minute, the full article is a click away.

5G overview

The system level: what 5G is, how it is deployed, and the capabilities built on top.

  • What is 5G?

    5G is the IMT-2020 generation: a new radio, NR, and a new service-based core, 5GC, specified by 3GPP from Release 15 to meet ITU-R's IMT-2020 requirements.

    • Three usage scenarios — eMBB, URLLC and mMTC — with conflicting design targets.
    • Headline targets: 20 Gbit/s downlink peak, 1 ms user-plane latency, a million devices per km².
    • NSA delivers NR capacity on the LTE core; the full feature set needs SA.
    • mMTC was met largely by LTE-M and NB-IoT, not by NR.

    Full article · 3 min read

  • NSA vs SA

    Two ways to deploy 5G. NSA (Option 3x) adds an NR gNB as secondary node to an LTE eNB on the EPC; SA (Option 2) connects the gNB directly to the 5G Core.

    • NSA runs on EN-DC: LTE carries the control plane, NR adds user-plane capacity.
    • Slicing, QoS flows, VoNR, RedCap and edge UPF need SA, because they are 5G Core features.
    • Throughput is comparable; the SA business case is the core, not the radio.

    Full article · 7 min read

  • 5G Core

    The 5G Core is a service-based architecture: network functions such as the AMF, SMF and UPF expose services over HTTP/2 and discover each other through the NRF.

    • AMF handles access and mobility, SMF handles sessions, UPF forwards packets.
    • Control and user planes separate over N4 (PFCP), so UPFs scale and move to the edge independently.
    • Radio-facing and packet-path interfaces (N1–N4, N6) stay reference points.
    • Slicing falls out of dynamic NF discovery rather than being bolted on.

    Full article · 9 min read

  • Network slicing

    Network slicing runs several logical networks over shared infrastructure, each identified by an S-NSSAI: an 8-bit slice/service type plus an optional 24-bit differentiator.

    • The UE requests slices at registration; the network returns the Allowed NSSAI per registration area.
    • Each PDU session belongs to exactly one slice; several slices mean several sessions.
    • Selecting a slice reserves nothing — isolation comes from RAN, transport and core resource management.

    Full article · 9 min read

  • MEC

    MEC places application compute and a local UPF near the RAN, so traffic breaks out to an edge data network instead of travelling to a central data centre.

    • It shortens transport latency and backhaul load; air-interface latency is unchanged.
    • 3GPP gets traffic there with UPF selection, local data networks and AF traffic influence.
    • SSC modes decide what happens to the session anchor when the UE moves.
    • ETSI MEC defines the application platform; 3GPP defines how traffic reaches it.

    Full article · 3 min read

  • NTN

    NTN adapts NR, from Release 17, to satellites and high-altitude platforms, where long delay, large Doppler and moving beams break terrestrial assumptions.

    • LEO trades coverage for latency; GEO covers continents with a round trip near 540 ms.
    • UEs pre-compensate timing and Doppler using GNSS and broadcast ephemeris.
    • Timing advance and HARQ timers were extended to survive the delay.
    • Transparent payloads keep the gNB on the ground; regenerative payloads put it in orbit.

    Full article · 9 min read

  • 5G-Advanced

    5G-Advanced is 3GPP's name for Releases 18, 19 and 20: the same NR radio and 5G Core, extended with features for efficiency, intelligence and new device classes.

    • Not a new generation: most features arrive as software on existing networks, plus new UEs.
    • Release 18 opened AI/ML for the air interface, network energy saving, L1/L2-triggered mobility and eRedCap.
    • Later releases make the study items normative and trial concepts that 6G will build on.

    Full article · 3 min read

  • Massive MIMO

    Massive MIMO uses arrays with tens of transceiver chains — 32T32R, 64T64R — to serve several users on the same time-frequency resources, separated in space.

    • The capacity gain is MU-MIMO; beamforming alone only improves one link.
    • It is a cell-capacity feature: gains grow with load, not peak rate.
    • TDD reciprocity via SRS gives full channel knowledge with overhead independent of array size.
    • FDD relies on codebook feedback, which limits practical port counts.

    Full article · 8 min read

  • Beamforming

    Beamforming applies per-element phase and amplitude weights so transmissions add constructively towards a receiver, giving array gain of about 10·log₁₀(N) dB.

    • RF-chain count sets the architecture: analog, digital or hybrid.
    • FR1 radios go digital; FR2 goes hybrid, because full RF chains at mmWave are impractical.
    • SSB beam sweeping bootstraps initial access, with up to 64 beams in FR2.
    • Beam management and failure recovery run continuously, not as exceptions.

    Full article · 9 min read

5G NR

The radio, in the order that makes each topic easier than the last.

  • What is 5G NR

    NR is the 5G radio access technology: an OFDM air interface whose numerology scales subcarrier spacing and slot length together, across FR1 and FR2.

    • Numerology µ sets 15 × 2^µ kHz subcarrier spacing; slots shorten as spacing grows.
    • FR1 carries coverage and capacity; FR2 buys bandwidth with a much harder link budget.
    • Symbol-level TDD flexibility and mini-slots are where NR's latency gains come from.
    • Bandwidth parts decouple device capability from carrier width.

    Full article · 9 min read

  • NR architecture

    NR architecture has three domains — UE, NG-RAN and 5G Core — joined by the air interface and the NG interface, with a gNB that can be split into CU and DU.

    • User plane: SDAP maps QoS flows, PDCP secures, RLC segments, MAC schedules, PHY transmits.
    • RRC terminates in the gNB; NAS terminates in the AMF, carried transparently.
    • The F1 split puts RRC, SDAP and PDCP in the CU; RLC, MAC and PHY in the DU.

    Full article · 6 min read

  • Frame structure

    NR keeps LTE's 10 ms frame and 1 ms subframe, but slot length depends on numerology: 1/2^µ ms, carrying 14 OFDM symbols (12 with extended cyclic prefix at 60 kHz).

    • Slot formats mark each symbol as downlink, uplink or flexible, configured in layers.
    • Mapping type B mini-slots start at any symbol and run 2–7 symbols.
    • Slot aggregation stretches a transmission across several slots for coverage.
    • All timing is an integer multiple of Tc, about 0.5 ns, locked to LTE's.

    Full article · 9 min read

  • Numerology

    Numerology µ scales NR subcarrier spacing as 15 × 2^µ kHz. Symbols per slot stay at 14 and the subframe stays at 1 ms, so a slot lasts 1/2^µ ms.

    • Powers of two keep the grids aligned, so one carrier can host several numerologies.
    • Small µ favours coverage and delay spread; large µ favours latency, Doppler tolerance and high carrier frequencies.
    • 30 kHz in FR1 and 120 kHz in FR2 are the practical defaults.

    Full article · 9 min read

  • Resource grid

    An NR resource block is 12 subcarriers, addressed through four coordinate systems: common resource blocks from Point A, bandwidth parts, and physical and virtual resource blocks.

    • CRBs are absolute, anchored at Point A for every numerology.
    • A UE can have up to four bandwidth parts configured and one active, switched by DCI, timer or RRC.
    • PRBs count from the BWP start; VRBs are what the DCI names.
    • BWPs enable power saving, narrowband devices and mixed numerologies.

    Full article · 14 min read

  • PUSCH

    PUSCH carries uplink data, plus uplink control when the two share a slot. Its design is driven by a power-limited UE transmitter more than by the channel.

    • Transform precoding selects DFT-s-OFDM: lower PAPR and better coverage, single layer only.
    • CP-OFDM supports multiple layers where the link budget allows.
    • Grant-based access costs a scheduling round trip; configured grants remove it by reserving resources.
    • Fractional path-loss compensation deliberately holds back cell-edge UEs to protect neighbouring cells.

    Full article · 9 min read

  • PDSCH

    PDSCH carries downlink user data. The gNB picks an MCS and a resource allocation, and the transport block size formula turns them into a byte-aligned block for LDPC.

    • Processing: CRC, LDPC with redundancy versions, scrambling, then layer and resource-element mapping.
    • Mapping skips DM-RS, SSB, CSI-RS, the scheduling PDCCH and configured reserved resources.
    • A pre-emption indication tells the UE which resources URLLC traffic took.
    • Slot aggregation repeats the block with rotating redundancy versions for coverage.

    Full article · 22 min read

  • PUCCH

    PUCCH carries uplink control — HARQ-ACK, scheduling requests and CSI — when the UE has no PUSCH to multiplex it onto. NR defines five formats.

    • Two axes: short (1–2 symbols) or long (4–14), and up to 2 bits or more.
    • Format 0 carries no DM-RS; the information sits in the cyclic shift.
    • Long formats buy energy for cell-edge UEs, not extra payload.
    • Format 4 is Format 3 with spreading, so UEs can share resources.

    Full article · 8 min read

  • PRACH

    Random access is NR's one unscheduled uplink transmission, breaking the circularity of needing a grant to ask for a grant.

    • Msg1 is a preamble on a PRACH occasion tied to an SSB beam, carrying no identity.
    • Msg2 returns timing advance, a Msg3 grant and a TC-RNTI.
    • Msg4 resolves contention; without it, two UEs could end up sharing one identity.
    • Two-step RACH saves a round trip by transmitting payload before timing advance is known.

    Full article · 10 min read

  • SRS

    SRS is an uplink reference signal the gNB uses to measure a UE's channel — and, in TDD, to build the downlink precoder through reciprocity.

    • One SRS transmission replaces heavy CSI feedback for large antenna arrays.
    • Reciprocity covers the propagation channel, not the transceivers, so calibration is required.
    • Antenna switching sounds receive antennas the UE does not transmit from.
    • Usages: beam management, codebook, non-codebook and antenna switching.

    Full article · 8 min read

  • SSB

    The SSB packs PSS, SSS and PBCH into 4 OFDM symbols and 240 subcarriers. It is the first thing a UE finds, and the reference for beams across NR.

    • A burst set holds up to 4, 8 or 64 SSBs in a 5 ms half-frame, one per beam.
    • Burst sets repeat every 5 to 160 ms.
    • Cases A–E set where blocks start in the half-frame for each SSB subcarrier spacing.
    • The SSB index acts as a beam index for measurement, RACH occasion selection and paging.

    Full article · 11 min read

  • Scheduling

    Every slot, the MAC scheduler decides which UEs transmit, on which resources, at what MCS and on which layers; logical channel prioritisation then fills the resulting transport block.

    • Who, how much, how hard and where are vendor algorithms; LCP is standardised.
    • LCP serves each channel up to its prioritised bit rate, then by strict priority.
    • Uplink scheduling is blind: an SR is one bit, a BSR is grouped and quantised.
    • Configured grants remove the scheduling-request round trip.

    Full article · 10 min read