The Evolution of Non-Terrestrial Networks: From 5G NTN to the 6G Horizon

Non-Terrestrial Networks (NTN) have moved from the margins of 3GPP to a substantial body of normative specification. What began as a study into whether 5G New Radio could operate over satellites and high-altitude platforms now covers satellite access, mobility, IoT connectivity, regenerative payloads and, increasingly, autonomous satellite operation.

The evolution is not a replacement of terrestrial networks by satellites, and 3GPP is not converging every satellite system onto a single design. What the standards provide are common interfaces and procedures that let cellular technology operate across very different propagation environments and satellite architectures.

Releases 15 and 16 established the technical foundations. Release 17 created the first normative NR-NTN framework. Release 18 enhanced mobility and operational capability. Release 19, frozen in December 2025, added regenerative payloads, Store-and-Forward operation and further IoT capability. Release 20 now carries the 5G-Advanced NTN work forward while opening the 6G study phase, with normative 6G work reserved for Release 21.

1. The foundation: Releases 15 and 16

The journey began with study work rather than normative specification. 3GPP examined how terrestrial-oriented NR could be adapted to environments where propagation delay, Doppler, satellite motion and coverage geometry differ fundamentally from conventional cellular deployment.

Two technical reports formed the basis:

• 3GPP TR 38.811 — Study on New Radio (NR) to support Non-Terrestrial Networks
• 3GPP TR 38.821 — Solutions for NR to support Non-Terrestrial Networks

For IoT, the parallel study is TR 36.763 (NB-IoT/eMTC over NTN), which reuses much of the TR 38.821 reference framework.

1.1 Deployment scenarios

NTN is not a single orbital configuration. Different platforms produce very different radio conditions, and TR 38.821 defines reference scenarios for each.

Low Earth Orbit (LEO). Typically 500–2,000 km. A satellite at 600 km moves at roughly 7.56 km/s relative to the ground, which is where most of the difficulty comes from: large and rapidly varying Doppler, and beam footprints of 50–1,000 km that sweep across the Earth. Round-trip propagation delay is modest by satellite standards — about 25.8 ms at 600 km with a transparent payload, or 12.9 ms with a regenerative one.

Medium Earth Orbit (MEO). Roughly 7,000–25,000 km. MEO was not among the TR 38.821 reference scenarios, which concentrated on LEO, GEO and HAPS; MEO systems fall between the two extremes on both delay and coverage.

Geostationary Earth Orbit (GEO). 35,786 km above the equator, effectively stationary relative to a point on the ground. Doppler is almost negligible (0.93 ppm) and beam footprints reach 3,500 km, but round-trip delay is 541 ms in the worst case for a transparent payload, or 271 ms over the service link alone with the payload regenerative.

High-Altitude Platform Stations (HAPS). Stratospheric rather than orbital, typically up to 20 km. One-way delay is about 1.5 ms and Doppler is dominated by platform motion — a radio environment much closer to terrestrial systems than either GEO or LEO.

The framework is deliberately general: 3GPP NTN addresses multiple non-terrestrial deployment models rather than being designed around one constellation.

2. The engineering problem: making NR work in space

Terrestrial cellular assumes short propagation distances and stable cell geometry. NTN breaks both assumptions. Three consequences drove most of the design work.

Propagation delay

A terrestrial macro cell of 10 km radius has a maximum one-way delay of about 33 µs. A GEO link has a worst-case round trip of 541 ms — four orders of magnitude larger. Differential delay within a single cell reaches 10.3 ms for GEO and about 3.1 ms for a 600 km LEO cell, meaning UEs in the same beam experience materially different timing.

This affects timing advance, random access, HARQ, scheduling, retransmission and mobility procedures.

Doppler

At 600 km, maximum Doppler shift for an earth-fixed UE is 24 ppm — roughly ±48 kHz at 2 GHz, rising to ±480 kHz at 20 GHz and ±720 kHz at 30 GHz. The shift also changes quickly: TR 38.811 records a maximum variation rate of −544 Hz/s at 2 GHz and −8.16 kHz/s at 30 GHz. Terrestrial NR was never designed for frequency offsets of this magnitude changing at this rate.

Moving coverage

In terrestrial networks a cell maps to a fixed geographic area. With LEO NTN, the satellite and its beams move across the Earth, and with earth-moving beams a fixed point on the ground may be served by a given spotbeam for only a few seconds. The network must handle changing beam visibility, satellite motion and very frequent mobility events.

3. Release 17: from study to specification

Release 17 turned feasibility work into normative specification, and did so on two parallel tracks that are worth keeping distinct:

• NR-NTN — 5G NR over satellite, for smartphones and broadband terminals.
• IoT-NTN — NB-IoT and eMTC (LTE-M) over satellite, for low-power devices.

Both were built on transparent payloads, in which the satellite provides the radio link while RAN processing stays on the ground. This was a pragmatic choice: it extended NR to satellite links without requiring a complete base station on board every spacecraft, and it allowed the ground segment to be upgraded conventionally.

3.1 Timing compensation and the GNSS assumption

Adapting NR timing to satellite distances was the central problem. The Release 17 solution assumes a GNSS-capable UE: the device uses its own position together with broadcast satellite ephemeris to pre-compensate timing advance and frequency offset before transmitting.

This is worth stating plainly, because the assumption is now being revisited. Release 19 system work introduced GNSS-independent operation and enhanced positioning, and the Release 20 NR-NTN project is assessing the impact of not relying on GNSS information for initial access and connected-mode procedures, with a decision on how to proceed after a one-year study. The trajectory is therefore from GNSS as a baseline requirement toward GNSS as an optimisation.

3.2 Doppler compensation

The principle mirrors the timing case: compensate the predictable, geometry-driven component of the frequency shift at both ends, and leave only the residual associated with the UE's specific position and motion. Network-side pre-compensation on the feeder link is combined with UE-side pre-compensation on the service link. This differs from terrestrial NR, where Doppler is smaller and changes far more slowly.

3.3 Spectrum

Release 17 defined FR1-NTN operation in L- and S-band (bands n255 and n256), which is what makes direct-to-device service to unmodified handsets possible. Release 18 extended NR-NTN to FR2-NTN — Ku and Ka-band — for VSAT terminals on aircraft and vessels, a very different service class with directive antennas and much higher throughput.

4. NTN versus terrestrial NR

HARQ illustrates the consequence. Terrestrial HARQ relies on short feedback loops; at GEO round-trip times, stop-and-wait operation stalls the process pipeline. NTN therefore allows HARQ feedback to be disabled per process, shifting reliability to RLC and higher layers where the round trip makes retransmission uneconomic.

The objective throughout is not to discard terrestrial procedures but to modify them where the physics of the link makes the terrestrial assumption inefficient.

5. Release 18: operational efficiency

Release 18, the first 5G-Advanced release, moved NTN beyond baseline connectivity toward mobility and operational efficiency.

The clearest example is RACH-less handover for NR-NTN, which allows the UE to make its initial uplink transmission during handover without the conventional random-access procedure. With satellite beams sweeping past a UE and handovers occurring far more often than in terrestrial deployment, removing a round trip from each one materially reduces cumulative interruption time.

Conditional Handover (CHO) is often discussed alongside this, but CHO is an established NR mobility mechanism rather than an NTN invention. What Release 18 contributed was its application to moving satellite coverage, alongside improved TN–NTN handover management, FR2 support for VSAT terminals, and power-saving enhancements for IoT devices operating in discontinuous coverage.

6. Network-verified UE location

3GPP created a dedicated Release 18 study on network-verified UE location for NTN, documented in TR 38.882.

The requirement is largely regulatory. A satellite beam does not respect national borders, so an operator licensed in one country may need to demonstrate that a served UE is actually within its authorised territory. Related drivers include emergency services, location-based service authorisation, prevention of spoofing and satellite service authorisation more generally.

The significant point is architectural: NTN is incorporating mechanisms that let the network verify UE location rather than simply trusting a UE-reported position. Accuracy targets are use-case and regulator dependent, not a single fixed figure applied to every device.

7. Release 19: regenerative payloads

Release 19 added support for regenerative architectures alongside — not instead of — transparent ones.

The interesting part is how the decision was reached. The question was whether a satellite should carry a full gNB or only a gNB-DU, leaving the central unit on the ground and running F1 over the feeder link. The argument ran for roughly a year across four TSG RAN meetings before settling on the full gNB on board.

For IoT-NTN, the corresponding Release 19 work item (IoT_NTN_Ph3) specifies a full eNB as a regenerative payload — LTE, not NR. The distinction matters when reading the specifications: gNB for NR-NTN, eNB for IoT-NTN.

A full gNB on board changes the architecture. The satellite becomes an active network node rather than a radio relay, which brings:

• native support for inter-satellite links and the Xn interface between satellites,
• inter-satellite mobility and handover with minimal core-network involvement,
• reduced round-trip time for all gNB–UE procedures, including random access and HARQ,
• operation through feeder-link outages rather than dependence on a continuous ground link,
• Store-and-Forward as a buildable capability.

The trade-off is payload complexity and power. Onboard processing adds requirements for compute, memory, power budget, thermal management, radiation tolerance, hardware redundancy and — often underestimated — software lifecycle management for a base station that cannot be physically serviced.

8. Store-and-Forward

Release 19 also introduced Store-and-Forward (S&F) operation, in which a satellite receives data while a user is in view, buffers it on board, and forwards it when a feeder link or gateway becomes available.

One scoping point is important: in Release 19, S&F is specified only for IoT UEs, built on the regenerative eNB payload and including support for feeder-link switchover. It is a delay-tolerant service, not a general NR capability.

That still opens real applications — delay-tolerant IoT, remote sensing, maritime, asset tracking, and any deployment where feeder-link availability is intermittent. The architectural significance is larger than buffering: it is the first case in 3GPP of a satellite performing network functions without a continuous connection to the ground.

9. IoT-NTN in Release 19

The IoT_NTN_Ph3 work item covers:

• Store-and-Forward with a full eNB regenerative payload, including feeder-link switchover,
• uplink capacity enhancement, including multiplexing of multiple UEs via Orthogonal Cover Codes (OCC) and reduced signalling to complete an EDT transaction,
• NB-IoT public warning system support,
• a TDD mode for NB-IoT NTN, enabling half-duplex operation and extending IoT-NTN beyond legacy FDD.

Peak data rate is not the objective. A satellite beam covers a very large area with limited spectrum and power, so the constraint is how many low-power devices can share the resource. Better multiplexing translates directly into device density across agriculture, logistics, environmental monitoring, industrial sensing, maritime tracking and infrastructure monitoring.

Release 19 also specified UE-Satellite-UE communication and RedCap device support in FR1-NTN, alongside improvements for air-to-ground and UAS scenarios.

10. How much compute belongs in orbit?

Regenerative NTN poses a question 3GPP does not answer: how much processing should move into space?

A transparent satellite leaves the network on Earth and can be upgraded by replacing ground equipment. A regenerative satellite carries a base station that must survive years of radiation exposure and thermal cycling, run on a constrained power budget, and be updated over the air.

This is where developments outside 3GPP matter. Advanced semiconductor packaging and 3D heterogeneous integration (3DHI) are being pursued across industry and government research programmes, and could eventually improve compute density and thermal performance in constrained environments. These are enabling technology trends that make regenerative payloads more practical over time — not requirements written into any release.

11. Release 20: the bridge to 6G

3GPP concluded that two releases are needed to specify 6G: Release 20 for studies, Release 21 for normative work. Release 20 combines a third wave of 5G-Advanced enhancement with the formal 6G study phase, with Stage-1 frozen in June 2025, Stage-2 through 2026 and Stage-3 targeted for March 2027.

NTN appears on both tracks:

• NR-NTN studies the impact of operating without GNSS for initial access and connected-mode procedures — the most consequential open question in NTN radio design, since removing the GNSS dependency changes both device cost and the addressable device population.
• IoT-NTN specifies radio improvements to enable voice support over GEO systems.
• SA1's satellite access Phase 4 study covers emergency and mission-critical services over satellite, multi-orbit access, unreachability notification, and IMS voice over GEO.

For 6G itself, the ITU target for technology proposals is early 2029, with full specifications submitted by mid-2030. Release 21 is where 6G NTN becomes normative.

12. IMT-2030 and the ITU process

The 6G framework is being set under IMT-2030, defined in Recommendation ITU-R M.2160, which identifies six usage scenarios:

1. Immersive Communication
2. Hyper Reliable and Low-Latency Communication
3. Massive Communication
4. Ubiquitous Connectivity
5. Artificial Intelligence and Communication
6. Integrated Sensing and Communication (ISAC)

At its February 2026 meeting, ITU-R Working Party 5D completed the draft report Minimum requirements related to technical performance for IMT-2030 radio interface(s), containing 20 technical performance requirements. At its June 2026 meeting WP 5D completed the companion evaluation guidelines report. Both were submitted to Study Group 5 for approval in December 2026. WP 5D will accept candidate RIT/SRIT submissions from February 2027.

Two clarifications are worth carrying: WP 5D defines the radio interface requirements only — core network, architecture and security are 3GPP's remit — and completing the requirements is not the same as standardising an air interface. No 6G radio interface, and no single terrestrial–satellite architecture, has been selected.

Ubiquitous Connectivity is the scenario that matters most here. It reframes satellite access as one domain within a coordinated fabric spanning terrestrial cellular, HAPS, satellite, edge compute and sensing — rather than a separate service bolted onto the side. The practical objective is service continuity where terrestrial infrastructure is absent, overloaded or damaged: disaster response, remote communities, maritime, aviation, logistics and environmental monitoring.

13. ISAC and the satellite view

Integrated Sensing and Communication proposes that wireless infrastructure derive information about the physical environment as well as transport information across it — object detection, localisation, mapping, tracking, environmental monitoring and high-precision positioning.

The evaluation framework now reflects this concretely. The IMT-2030 guidelines introduce extended channel models covering near-field propagation, spatial non-stationarity and ISAC-specific behaviour, and define seven test environments including three that did not exist in the IMT-2020 framework: Indoor Factory-HRLLC, Indoor Factory-ISAC and Urban Macro-ISAC.

For NTN the appeal is obvious — satellites already observe very wide areas — but no ISAC test environment is currently satellite-based, and NTN sensing remains a research topic rather than a specified capability.

14. Where the standards meet deployment

Standardisation and commercial reality are not moving in lockstep, and the gap is the most interesting part of the current picture.

On the 3GPP side, Release 17 IoT-NTN is in commercial service, with Skylo operating an NB-IoT NTN network and Release 17 NTN chipsets shipping from Qualcomm and MediaTek. Release 19 froze in December 2025, and its regenerative and RedCap ecosystem will mature through 2026 and 2027.

In parallel, several direct-to-device operators are moving partly outside the 3GPP path. SpaceX's Direct to Cell service established that unmodified handsets can connect to LEO satellites, and its subsequent move toward dedicated spectrum, higher beam density and satellite-tuned protocols points at treating the satellite layer as another cellular access rather than a special case. Regulatory frameworks have followed: the FCC's Supplemental Coverage from Space rules created a US licensing pathway for handset-grade satellite service using mobile operator spectrum.

The open question for the next few years is whether direct-to-device converges on 3GPP NTN or splits into standardised and proprietary tiers.

15. The evolution in stages

The progression can be read as a sequence of capabilities rather than a single transition. (These are stages of maturity, not protocol layers.)

Stage 1 — Radio adaptation. NR adapted to long delay, large Doppler and moving coverage. Stage 2 — Mobility. Handling rapidly changing satellite and beam geometry. Stage 3 — IoT. Low-power devices reaching satellites for wide-area, intermittent connectivity. Stage 4 — Regenerative processing. Base station functions moving on board. Stage 5 — Store-and-Forward. Operation through intermittent ground connectivity. Stage 6 — AI and sensing. Communications, computing, AI and sensing combined. Stage 7 — Multi-domain integration. Terrestrial, aerial and orbital nodes operating as coordinated components of one architecture.

16. What changes for operators

For mobile operators, NTN is a network-extension technology rather than a replacement for terrestrial RAN. It fills coverage and resilience gaps that terrestrial infrastructure cannot economically address.

The model is terrestrial RAN + NTN + Wi-Fi + HAPS + edge/cloud, not terrestrial RAN versus satellite.

That raises a set of practical questions:

• Which geographies are uneconomic to cover terrestrially, and what is the marginal cost of covering them from orbit?
• Which customer segments actually require ubiquitous connectivity, and will they pay for it?
• Which spectrum arrangements support NTN, and does Supplemental Coverage from Space or an equivalent framework apply?
• How should NTN traffic integrate with the existing core, and how should TN–NTN mobility be managed?
• Which services justify direct-to-device, and at what latency and throughput?
• How much processing should stay on the ground versus move into the satellite — and does that answer change with feeder-link availability?
• Partner with a satellite operator, or build?

The answers vary substantially by market and by regulatory regime.

Conclusion: from satellite connectivity to a multi-domain fabric

NTN represents one of the most significant extensions of cellular technology beyond terrestrial infrastructure, and the questions driving it have changed with each release.

Releases 15 and 16 asked whether 5G NR could operate reliably over satellites. Release 17 answered with the first normative NR-NTN and IoT-NTN specifications. Release 18 asked how to make that operationally efficient, and delivered mobility improvements including RACH-less handover. Release 19 asked how much of the network could move into the satellite itself, and answered with a full gNB on board and Store-and-Forward for IoT. Release 20 is now asking whether the GNSS assumption underpinning the whole design can be removed, while opening the 6G study phase that Release 21 will make normative.

The 6G question is broader: whether communication, sensing, computing, AI and heterogeneous access can operate as one coordinated system. IMT-2030 points in that direction, but the radio interface has not been selected and the evaluation process runs into 2029.

What has already changed is the boundary. The long-term network is likely to consist of many node types — ground-based, aerial and orbital — coordinated through common standards and increasingly intelligent control. The real evolution of NTN is not from Earth to space. It is from isolated networks toward an integrated multi-domain fabric.

References

• 3GPP TR 38.811 — Study on New Radio (NR) to support Non-Terrestrial Networks
• 3GPP TR 38.821 — Solutions for NR to support Non-Terrestrial Networks
• 3GPP TR 36.763 — Study on NB-IoT/eMTC support for Non-Terrestrial Networks
• 3GPP TR 38.882 — Study on requirements and use cases for network verified UE location for NTN
• 3GPP TR 22.865 — Study on satellite access, Phase 3
• 3GPP Release 19 work items: NR_NTN_Ph3, IoT_NTN_Ph3, 5GSAT_Ph3
• Recommendation ITU-R M.2160 — Framework and overall objectives of the future development of IMT for 2030 and beyond
• Draft Report ITU-R M.[IMT-2030.TECH PERF REQ] and M.[IMT-2030.EVAL], ITU-R WP 5D, 2026