5G NR Overview

Introduction

5G NR (New Radio) is the 3GPP radio-access technology standardised from Release 15 onward and specified across the TS 38-series (physical layer in TS 38.2xx, protocol layers in TS 38.3xx, overall description in TS 38.300). It is the air-interface half of the 5G System; the 5GC (5G Core, TS 23.501) is the other half. Where LTE was one radio tuned for broadband, NR is deliberately a configurable platform — the same framework re-shapes itself for a smartphone, a factory robot, or a battery-powered water meter.

NR is where the UE lifecycle begins on the radio: the device finds a cell via the SSB, reads system information, runs random access, and enters RRC_CONNECTED — all over the NR Uu interface to a gNB. Every later topic on this site (numerology, the resource grid, MIMO, scheduling, mobility) is a detail of one of the design choices introduced here.

This overview matters because almost every 5G design decision is downstream of a single premise: the next decade of demand is not one application but three families of them, so the radio itself must be a knob-set rather than a fixed pipe. Hold that idea and the rest of the system — flexible numerology, bandwidth parts, slicing, the disaggregated RAN — reads as consequences rather than a list to memorise.

Why a new radio was needed

Each earlier generation optimised for one dominant application — voice for 2G, packet data for 3G, mobile broadband for 4G — and then stretched to fit everything else. The ITU-R chartered 5G under IMT-2020 (Recommendation M.2083) to serve three radically different traffic classes natively, with performance targets studied by 3GPP in TR 38.913. No single rigid waveform can be simultaneously the widest pipe, the lowest-latency link, and the most energy-frugal connection, so the concrete purpose of NR is to expose the parameters — subcarrier spacing, slot length, bandwidth, antenna behaviour, logical slice — that let an operator dial in whichever corner a service needs.

What 5G NR is, and why it had to exist

5G NR (NR = New Radio) is the 3GPP radio-access technology for 5G, first standardised in Release 15 (2018). "New" is literal: unlike the step from 3G to LTE, NR is not an incremental upgrade of an existing air interface — it is a fresh physical layer and a fresh set of protocols, designed so that a single technology can be tuned for wildly different services and can run anywhere from 600 MHz to millimetre-wave.

The intuition worth carrying: the earlier generations each optimised for a single killer application (voice for 2G, packet data for 3G, broadband for 4G). 5G's premise is that the next decade of demand is not one application but three families of them, so the radio itself must be a configurable platform rather than a fixed pipe. Everything else on this page is a consequence of that one decision.

The configurability is concrete and numeric. NR's time/frequency grid is parameterised by a numerology index μ, defined in TS 38.211, from which the subcarrier spacing and slot length follow directly:

with μ ∈ {0,1,2,3,4} giving 15/30/60/120/240 kHz and slots of 1 ms down to 62.5 μs. A resource element is one subcarrier × one OFDM symbol; a resource block is a fixed 12 subcarriers in frequency at any numerology. That single scaling law is what lets the same radio be robust and wide-reaching at 15 kHz or fast and low-latency at 120 kHz.

IMT-2020: the three usage scenarios and their KPIs

The ITU-R defined 5G's ambitions in the IMT-2020 vision (Recommendation M.2083), and everything NR does traces back to three usage scenarios. You should be able to name all three and the KPI each one stresses.

The ITU expresses these ambitions as eight capability KPIs, and the clever part of the IMT-2020 picture is that no single scenario needs all eight at maximum — each scenario pushes a different corner. That is exactly why one rigid radio could not serve them and why NR is a configurable platform.

These KPIs also line up with the standardised slice types the core uses to serve each scenario. An S-NSSAI (TS 23.501) is an 8-bit SST (Slice/Service Type) plus an optional 24-bit SD (Slice Differentiator); the standardised SST values map one-to-one onto the triangle: SST 1 = eMBB, SST 2 = URLLC, SST 3 = mMTC (and SST 4 = V2X). So "which corner of the triangle" and "which slice value" are the same question asked at the radio and at the core.

SA vs NSA — the two ways to deploy 5G

There are two deployment options, and the difference is which core and which control anchor the device uses. In Non-Standalone (NSA), 5G is bolted onto an existing LTE network: the device anchors on an LTE eNB for control signalling and adds the NR carrier purely as a data booster, with everything still terminating in the LTE core (EPC). This is the EN-DC (E-UTRA–NR Dual Connectivity) architecture, option 3. In Standalone (SA), the gNB connects directly to a 5G Core — NR radio and 5G core end to end, no LTE dependency. This is option 2, and it is where the 5G features that need the new core actually live.

NSA got 5G to market quickly because operators could light up NR on top of a mature LTE core and coverage layer. SA is the destination: only with a 5GC do you get native network slicing, the ultra-low-latency and high-reliability behaviours URLLC needs, and the service-based signalling of the modern core. Most networks launched NSA and are migrating to SA. (See the dedicated SA vs NSA topic for the full deployment-option map.)

FR1 vs FR2 — the spectrum NR runs in

NR is defined across two frequency ranges, and they behave so differently that many design details (bandwidth, subcarrier spacing, beamforming intensity) follow directly from which one you are in.

The propagation gap between the two ranges shapes almost every downstream design decision. At mmWave, free-space loss is far higher and signals barely penetrate buildings, so a cell's reach shrinks from kilometres to a few hundred metres and every link leans on a tightly focused beam that the network must actively track as the device moves. FR1, by contrast, can cover a whole neighbourhood from one site and can run with much wider beams. In practice operators layer the two: a low/mid-band FR1 blanket for ubiquitous coverage, with FR2 small cells dropped into stadiums, transit hubs and city cores where raw capacity is the constraint.

The key NR enablers

What actually lets one framework span 20 Gbit/s broadband, 1 ms control loops and a million sensors? A handful of design choices, each of which you will meet again as its own topic.

Notice how these enablers line up with the three scenarios. Flexible numerology and mini-slots serve URLLC; massive MIMO and wide bandwidth serve eMBB; slicing and the service-based core let mMTC and everything else coexist on shared metal; and the CU/DU/RU split plus BWP are the plumbing that makes deploying and powering all of this economical.

How NR differs from LTE

NR is easiest to understand as a set of deliberate departures from LTE. LTE fixed most parameters; NR made them configurable. This is the single table to internalise.

The recurring pattern: LTE broadcast a lot of always-on signalling (like CRS) that wasted energy and pinned parameters; NR strips the carrier "lean," transmits reference signals only when needed, and makes the grid, the codes and the antenna behaviour configurable. That is what turns a broadband radio into a service-agnostic platform.

3GPP releases and the RAN vs Core division of work

NR did not arrive all at once. 3GPP delivers 5G as a rolling series of releases, each adding capability. Knowing which release introduced what is standard interview fodder.

Two Rel-17 items deserve a callout. RedCap (reduced-capability NR) is 5G's answer to mid-tier IoT — wearables, industrial sensors, video cameras — devices that need more than mMTC throughput but far less than a smartphone, at lower cost and power. NTN brings satellite and high-altitude platforms into 3GPP, extending 5G coverage to oceans, aircraft and remote land where no tower reaches.

Finally, the split that frames the whole system: the RAN owns the radio problem and the Core owns the service problem. The diagram below shows how the gNB (itself split into CU/DU/RU) faces the device over the air and hands off to the 5GC over the NG interface.

Keep the division sharp: the RAN is responsible for everything about getting bits reliably across the air — synchronisation, scheduling, HARQ, beam management, mobility over the radio. The Core is responsible for everything about you as a subscriber — authentication, registration, session and IP management, policy, charging, and steering your traffic to the internet. The gNB never decides who you are; the 5GC never schedules a subframe.

Summary

5G NR is best understood as one decision and its consequences. The decision: build a service-agnostic, configurable air interface instead of a fixed pipe, because the IMT-2020 vision demands three incompatible profiles at once — eMBB (bigger pipe), URLLC (faster and surer), mMTC (more and cheaper). The consequences are the enablers: scalable numerology (SCS = 15×2^μ kHz, slots of 1 ms/2^μ), bandwidth parts, massive MIMO, lean-carrier reference signals, LDPC/Polar coding, a sliceable service-based 5GC, and a disaggregated CU/DU/RU RAN.

The two orthogonal deployment axes are worth keeping separate in your head: spectrum (FR1 sub-7 GHz for coverage, FR2 mmWave for hotspot capacity) and architecture (NSA anchored on LTE/EPC vs SA on a native 5GC). A cell picks one of each independently.

Finally, the RAN/Core division frames every other topic: the RAN gets bits across the air, the Core decides who you are and where your data flows. Every deeper page on this site — numerology, the resource grid, MIMO, scheduling, the 5GC functions — is a zoom into one box of that picture.