5G NTN Overview — NR over Satellite
What Non-Terrestrial Networks are, why 3GPP added them in Rel-17, the use cases, and how NR-NTN differs from a terrestrial cell.
A Non-Terrestrial Network (NTN) is ordinary 5G NR — the same air interface your phone already speaks — delivered through a satellite or a high-altitude platform instead of a rooftop tower. From Release 17, 3GPP made NR work over links that are hundreds or thousands of kilometres long, move at kilometres per second, and cover whole oceans. This page is the map to the whole NTN section.
Introduction
Non-Terrestrial Networks are the 3GPP framework for carrying cellular access — 5G NR, and separately NB-IoT/eMTC — over spaceborne and airborne platforms. Rather than inventing a new radio, NTN reuses the NR protocol stack and adds a set of targeted enhancements ("deltas") so the same waveform survives a link that is very long, rapidly time-varying, and often moving over the ground. The study phase is captured in TR 38.821; the normative NR-NTN behaviour is spread across TS 38.300 and the 38.2xx/38.3xx series.
NTN comes into play wherever terrestrial coverage stops: over oceans and deserts, across remote land, and when the ground network fails in a disaster. From Release 17 the headline target is direct-to-handset service — a normal-looking smartphone getting a bar of signal from a cell that happens to be in orbit — deployed standalone with a 5G core.
Why it matters: terrestrial towers reach only a small fraction of the Earth's surface, and NTN is how the same 5G device and core extend to the rest. This page is the orientation map: what NTN is, the two 3GPP tracks, the platform families and their physics, the payload and spectrum choices, and the three properties that make an NTN cell behave unlike any tower — each of which links to its own deep-dive.
On this page
What NTN actually is
In plain words: think of a normal cell tower put on an extremely tall pole — so tall the pole reaches into space and the "tower" is now a satellite drifting overhead. Your phone still speaks the same language (5G NR); it just has to shout across a much longer distance, at something that may be racing across the sky. Everything special about NTN is the handful of tricks that let the same phone hold that conversation.
Strip away the drama and NTN is a coverage story. Terrestrial cellular reaches only a small fraction of the Earth's surface — the fraction where it is economic to build towers and backhaul. Oceans, deserts, mountains, disaster zones and most of the planet's land area have no signal at all. A satellite sitting hundreds or thousands of kilometres up can illuminate an area a single tower never could, so 3GPP set out to let a normal 5G UE camp on, and get service from, a cell that happens to be in space.
NTN = 5G NR access carried by a spaceborne (satellite) or airborne (HAPS/UAS — High Altitude Platform Station) platform, integrated into 3GPP from Release 17 under the work item "Solutions for NR to support Non-Terrestrial Networks."
Ubiquitous coverage and service continuity: reach the ~90% of the globe towers miss, keep a connection alive over oceans and remote terrain, and add disaster resilience when the ground network is down.
Reuse the NR protocol stack with targeted "deltas" for the huge delay, large Doppler and moving cells. The study phase is captured in TR 38.821; the normative NR-NTN behaviour lands in TS 38.300 and the 38.2xx / 38.3xx series.
One-line placement: NTN is not a new radio — it is NR with a set of enhancements so the same waveform survives a link that is very long, time-varying and possibly moving over the ground.
Physically, an NTN connection has two hops, not one. The service link is the radio link between the UE and the satellite; the feeder link is the link between the satellite and an NTN gateway on the ground, which connects onward to the gNB and the 5G core. In the Rel-17 transparent baseline the satellite simply relays between the two, so the gNB stays on the ground and the whole two-hop path looks to the protocol stack like one long radio link.
Two 3GPP tracks: NR-NTN and IoT-NTN
3GPP split satellite access into two parallel tracks, because a smartphone-class broadband link and a battery-powered sensor have very different needs. This section — and most of this website's NTN pages — is about NR-NTN, the broadband-style track built on 5G NR.
| Track | Radio base | Target device | Spec home |
|---|---|---|---|
| NR-NTN | 5G NR | Handsets, VSAT, broadband-class UEs | TR 38.821, TS 38.300 |
| IoT-NTN | NB-IoT / eMTC (LTE) | Low-power, low-rate sensors / mMTC | TR 36.763 |
The two share the same fundamental physics — the same orbits, the same delay and Doppler problems — but solve them with different tools. If you care about massive machine-type connectivity over satellite, follow ntn-iot; everything else here is NR-NTN.
Platforms and altitudes
"Satellite" is not one thing. The altitude of the platform sets almost everything that matters — how far the signal travels, how fast the platform moves across the sky, how big a patch of ground it lights up, and how long a given cell stays over your head. The four platform families 3GPP considers span from the edge of the atmosphere to geostationary orbit.
| Platform | Altitude | Character |
|---|---|---|
| GEO (geostationary) | ~35,786 km | Appears fixed over a point; huge footprint; largest delay; negligible Doppler |
| MEO | ~7,000–25,000 km | Between the two extremes |
| LEO (low Earth orbit) | ~300–1,500 km | Small delay but large Doppler; footprint sweeps rapidly across the ground |
| HAPS / UAS | ~8–50 km | Airborne (stratospheric); short link, quasi-stationary |
The delay and Doppler numbers for each family are worked out in detail on ntn-orbits. The headline: GEO gives one enormous, near-constant delay; LEO gives a small but rapidly changing delay plus severe Doppler.
Payloads and spectrum
Two more choices define an NTN deployment: what the satellite does to the signal, and which band it uses.
Payload: transparent vs regenerative
A transparent (or "bent-pipe") payload just filters, frequency-shifts and amplifies — the gNB stays on the ground and the satellite is invisible to the protocol stack. This is the Rel-17 baseline. A regenerative payload puts the gNB (or a gNB-DU) on board the satellite, so it demodulates and decodes; this is a Rel-18 addition. The full comparison lives on ntn-architecture.
Spectrum: L/S-band vs Ka-band
For direct-to-handset service, NTN uses low frequencies — FR1 L-band and S-band — where the link budget is kind to a tiny phone antenna. For fixed VSAT terminals with real dishes, it uses FR2 Ka-band for bandwidth. The band plan is detailed on ntn-spectrum.
Spec anchor: transparent-payload NR-NTN is the Rel-17 baseline; regenerative payloads and store-and-forward come later. See TR 38.821 for the architecture options studied.
The three things that make NTN "not a normal cell"
If you remember nothing else, remember these three. Every NTN-specific mechanism exists to tame one of them, and each links to its own deep-dive.
Round-trip time is tens to hundreds of milliseconds — versus tens of microseconds terrestrially — and it drifts as the satellite moves. This breaks timing advance and every timer built around it. See ntn-timing-advance.
A LEO satellite races across the sky at ~7.5 km/s, shifting the carrier by tens of kHz and drifting continuously. See ntn-doppler.
The patch of ground a cell covers may slide beneath you (earth-moving beams), forcing frequent, time-driven mobility. See ntn-mobility.
Because of the first two, the NTN UE is required to have a GNSS receiver: knowing its own position and the satellite's ephemeris lets the UE pre-compensate most of the delay and Doppler before it ever transmits. That single assumption — the UE knows where it is — is what makes NR-NTN tractable, and it flows from the satellite's broadcast assistance in ntn-sib19.
Use cases and deployment
NTN is deployed standalone (SA) with a 5G core, using a UE that advertises NTN capability and carries GNSS. The headline Rel-17 goal is direct-to-handset service in FR1 — an unmodified-looking smartphone getting a bar of signal from space. Beyond that, the use cases fan out:
| Use case | What it delivers |
|---|---|
| Coverage / ubiquity | Service where no tower exists — oceans, deserts, remote land |
| Service continuity | Seamless roaming onto satellite when terrestrial coverage drops |
| Backhaul / trunking | Satellite backhaul for isolated terrestrial cells |
| IoT / mMTC | Global low-rate sensor connectivity (via ntn-iot) |
| Public safety | Disaster-resilient links when infrastructure fails |
| Broadcast / multicast | Wide-area content delivery to many terminals at once |
TN ↔ NTN: a terrestrial NR (TN) cell and an NTN cell run the same NR waveform, numerology and protocol stack — the differences are all consequences of distance and motion. TN round-trip delay is tens of microseconds and timing advance is a small, near-static correction; NTN delay is tens to hundreds of milliseconds and drifts, so the UE must pre-compensate a large TA from GNSS and ephemeris rather than rely on the network measuring it. TN Doppler is modest; NTN Doppler (LEO) is tens of kHz and continuous. TN cells are fixed; NTN beams may sweep the ground, turning mobility into a time-driven, geography-driven event. And TN needs no positioning receiver, whereas an NR-NTN UE is required to have GNSS. Same radio, harder link.
⚠ Common pitfalls / gotchas
- Treating NTN as a new radio. It is NR with deltas — the waveform, numerologies and stack are the same; assuming a bespoke air interface leads you to look for changes that are not there.
- Forgetting the GNSS requirement. An NR-NTN UE without a working GNSS fix cannot pre-compensate delay/Doppler and effectively cannot access the cell; "no signal" here is often "no position."
- Confusing service and feeder links. The UE↔satellite hop (service link) and the satellite↔gateway hop (feeder link) have different budgets and Doppler; a delay/frequency problem on one is not the other.
- Assuming GEO Doppler is a problem. GEO's headache is delay (~541 ms RTT), not Doppler; LEO is the reverse. Applying LEO intuition to GEO (or vice versa) misdiagnoses the link.
- Expecting terrestrial timers to just work. RACH windows, TA timers and HARQ/RLC timers all assume terrestrial round trips; NTN extends or reworks them, so a "timer expired" fault may be a mis-set NTN offset, not a coverage loss.
Q. In which 3GPP release did NR-NTN first become normative, and where is the study captured?
A. Release 17, under the work item "Solutions for NR to support Non-Terrestrial Networks." The study phase is in TR 38.821; the normative behaviour is spread across TS 38.300 and the 38.2xx/38.3xx series with NTN deltas.
Q. Why must an NR-NTN UE have GNSS?
A. The propagation delay and Doppler are enormous and time-varying. Knowing its own position plus the satellite ephemeris (broadcast in SIB19) lets the UE pre-compensate timing advance and frequency offset before transmitting, which the network cannot do for it blindly.
Q. What are the two 3GPP satellite-access tracks?
A. NR-NTN (broadband, built on 5G NR) and IoT-NTN (NB-IoT/eMTC, in TR 36.763). They share the orbital physics but target different devices.
Q. What is the difference between the service link and the feeder link?
A. The service link is the radio link between the UE and the satellite; the feeder link is the link between the satellite and the ground-based NTN gateway, which connects onward to the gNB and 5G core. In a transparent payload the satellite just relays between the two.
Summary
NTN extends 5G NR — the very same air interface — to satellites and high-altitude platforms so a normal device can get service where no tower reaches. It is not a new radio but NR plus targeted deltas, standardised from Release 17 (study in TR 38.821, normative behaviour across TS 38.300 and the 38.2xx/38.3xx series), and it splits into two tracks: NR-NTN for broadband/handsets and IoT-NTN for low-power sensors. The platform's altitude — GEO, MEO, LEO or HAPS — sets the delay, the Doppler, the footprint size and how fast a cell moves overhead, while the payload (transparent vs regenerative) and band (L/S-band for handsets, Ka-band for VSAT) fix the rest of the deployment.
Three properties make an NTN cell unlike any tower: enormous varying delay, large Doppler, and a moving footprint. Every NTN-specific mechanism exists to tame one of them, and all three lean on the UE having GNSS plus the ephemeris broadcast in SIB19, so it can pre-compensate delay and frequency before it ever transmits. Hold onto the link picture — UE over the service link to the satellite, satellite over the feeder link to the gateway and the ground gNB/5GC — and the deep-dive pages on orbits, timing advance, Doppler and mobility each slot into place.
Where to go next
Now that you have the shape of NTN, dig into the pieces that make it work: how the network is wired, the orbits that set the physics, and the delay problem that dominates everything.