5G NTN: When Satellites Speak 5G — Architecture, Projects & What It Means

Picture a geologist collecting samples in the Saharan Hoggar massif, 400 km from the nearest cell tower. Her smartphone registers on a 5G cell. There is no base station in sight, no backhaul cabinet on the ridge. The gNB serving her device is orbiting 550 km overhead at 27,000 km/h, and the entire session is standards-compliant 3GPP NR. This is not science fiction. This is what 5G Non-Terrestrial Networks deliver, and 3GPP has been writing the specifications since Release 15.

The convergence of satellites and mobile standards is one of the most architecturally significant developments in wireless since the introduction of LTE. It rewrites assumptions about coverage, mobility management, and what a “cell” even means. For RF and network engineers, understanding NTN is no longer optional. It is becoming a core competency.

Why 3GPP Chose to Standardize Satellite Access

Before Release 15, satellite communication and cellular were parallel universes. Satellite phones ran proprietary protocols. Cellular standards assumed ground-based infrastructure. The two worlds coexisted but never truly merged.

3GPP changed that calculus by recognizing a simple reality: 80% of the Earth’s surface has no terrestrial mobile coverage. Oceans, deserts, mountains, polar regions, and vast swaths of rural territory remain disconnected. Building towers everywhere is economically impossible. But LEO constellations can blanket the planet with RF coverage for a fraction of the per-square-kilometer cost.

The standardization effort began with TR 38.811 (Study on NR to support NTN) in Release 15, evolved through the channel models and architecture studies in TR 38.821 (Solutions for NR to support NTN), and reached its first implementable milestone in Release 17 with TS 23.501 amendments integrating NTN into the 5G System architecture.

The key design constraints 3GPP had to solve:

Challenge	Terrestrial assumption	NTN reality
Propagation delay	< 1 ms (macro cell)	26 ms one-way for LEO at 600 km
Round-trip time	2-10 ms	~52 ms (LEO), 540+ ms (GEO)
Doppler shift	Negligible at UE speeds	Up to 24 ppm at 27,000 km/h orbital velocity
Cell geometry	Fixed, well-defined	Moving footprint or steered beam
Handover triggers	RSRP/RSRQ thresholds	Ephemeris-based predictive handover
Timing advance	UE-measured	GNSS-assisted pre-compensation

Release 17 addressed each of these by introducing GNSS-assisted timing advance, Doppler pre-compensation at the UE, extended HARQ timers, and a new concept of geographically-fixed Tracking Area Identity (TAI). These are not cosmetic changes. They represent a fundamental rethinking of how the NR air interface handles delay and mobility.

Two Architectures: Transparent vs Regenerative Payload

The most critical architectural decision in NTN design is where the gNB sits. This choice splits the NTN world into two fundamentally different paradigms.

Transparent Payload (Bent-Pipe)

In a Transparent Payload architecture, the satellite is an RF relay. It receives the uplink signal from the UE on the service link, frequency-converts it, amplifies it, and retransmits it on the feeder link toward a ground gateway. All baseband processing, scheduling, and protocol handling happen at the ground-based gNB.

The signal chain looks like this:

UE —> Service Link (NR Uu) —> Satellite (frequency conversion + amplification only) —> Feeder Link —> Ground Gateway —> gNB (on ground)

This is the architecture specified for Release 17 initial deployments. Its advantages are significant:

• Lower satellite complexity. The payload is essentially a transponder. No onboard baseband processing means lower power consumption, lower mass, and lower cost per satellite.
• Easier upgrades. Software updates to the gNB happen on the ground. No need to update satellite firmware in orbit.
• Proven technology. Bent-pipe transponders have decades of heritage in satellite communications.

The tradeoff is latency. Every packet traverses the space segment twice: UE to satellite to ground (service link + feeder link), then ground to satellite to UE on the return path. For LEO at 600 km altitude, this yields approximately 26 ms one-way delay and 52 ms RTT before any terrestrial network delay is added.

Regenerative Payload (Onboard gNB)

In a Regenerative Payload architecture, the satellite hosts the gNB itself. The NR Uu interface terminates onboard. The satellite decodes, schedules, and re-encodes traffic. The feeder link carries backhaul (N2/N3 interfaces to the 5G Core), not raw RF.

UE —> Service Link (NR Uu) —> Satellite (onboard gNB: full L1/L2/L3 processing) —> Feeder Link (N2/N3 backhaul) —> Ground Gateway —> 5GC

The regenerative approach cuts the air interface latency in half for the NR segment because the HARQ loop closes onboard the satellite rather than round-tripping to the ground. It also enables inter-satellite links (ISL) to route traffic between satellites without touching the ground, which is essential for serving users in areas without a ground gateway in view.

The cost: dramatically higher satellite complexity. Onboard processing requires radiation-hardened baseband ASICs, more power, and more thermal management. Satellite replacement cycles of 5-7 years mean the onboard gNB hardware may become outdated before the satellite deorbits. Release 17 does not yet fully specify the regenerative architecture; it is expected to mature in Release 18 and Release 19.

Criterion	Transparent	Regenerative
gNB location	Ground	Onboard satellite
Air interface RTT (LEO)	~52 ms	~26 ms
Satellite complexity	Low (transponder)	High (full gNB)
Software upgradability	Easy (ground-based)	Difficult (in-orbit)
ISL routing	Not supported	Supported
3GPP maturity	Release 17	Release 18+
Current example	Starlink D2C	Thales Alenia Space (R&D)

Earth-Fixed Cell vs Earth-Moving Cell: The Geometry Problem

A LEO satellite at 550 km altitude crosses the sky in roughly 10 minutes. Its footprint on the ground sweeps across thousands of kilometers during a single pass. This creates a fundamental question: does the cell move with the satellite, or does it stay fixed on the ground?

Earth-Moving Cell (EMC)

In the EMC model, the cell footprint is attached to the satellite. As the satellite moves, the cell sweeps across the ground. Users enter and exit the cell as it passes overhead. This is the traditional approach inherited from legacy satellite systems like Iridium and Globalstar.

EMC is simpler to implement because the antenna beam pattern is fixed relative to the satellite body. But it creates a cascade of mobility management problems. Users experience frequent handovers as cells sweep past them. The Tracking Area changes continuously. Paging becomes complex because the network must predict which satellite cell will be overhead when a page needs to be delivered.

Earth-Fixed Cell (EFC)

In the EFC model, the cell footprint is geographically anchored. The satellite uses electronically steered phased-array antennas to keep the beam locked on a fixed area on the ground, even as the satellite moves overhead. When the satellite moves too far to maintain the beam angle, it hands off the fixed cell to the next satellite in the constellation.

This is the approach championed by AST SpaceMobile with its BlueWalker 3 and BlueBird satellites. Their massive phased-array antennas (64 m2 aperture on BlueBird) enable beam steering that maintains fixed ground cells.

EFC is what Release 17 targets with the geographically-fixed TAI concept. From the 5G Core perspective, a TAI corresponds to a fixed geographic area. Regardless of which satellite is currently serving it, the TAI does not move. This dramatically simplifies paging, location updates, and session continuity.

The engineering penalty is antenna complexity. Steering a beam to maintain a fixed ground footprint while the satellite moves at 7.5 km/s requires sophisticated phased-array control with millisecond-level recalculation. The satellite must also coordinate with the next satellite in the orbital plane for seamless cell handover.

Aspect	Earth-Moving Cell	Earth-Fixed Cell
Cell anchor	Satellite body	Geographic coordinates
User experience	Cell sweeps past user	User stays in cell
Handover frequency	Very high	Low (satellite-to-satellite)
TAI management	Dynamic, complex	Fixed, standard
Antenna requirement	Fixed beam	Steered phased array
Paging complexity	High	Low (fixed TAI)
Example	Iridium, Globalstar	AST SpaceMobile

U-DESERVE 5G: France Bets on Direct Satellite-to-Device

Among the most ambitious NTN validation programs in Europe is U-DESERVE 5G (Universal Direct accEss to SatellitE for Resilient VEhicular 5G), funded under France’s France 2030 investment plan. This project brings together a consortium that reads like a who’s who of European space and telecom:

• Thales Alenia Space — satellite platform and payload integration
• Capgemini — system architecture and software
• Thales — defense and security applications
• Orange — operator expertise and spectrum
• SES — multi-orbit satellite operations (MEO/GEO experience)
• Qualcomm — NTN-capable modem chipsets
• Loft Orbital — agile satellite hosting platform

The project targets validation of direct satellite-to-device 5G NR communication, meaning a standard UE equipped with an NTN-capable Qualcomm modem connecting directly to a satellite without any intermediate ground relay for the user plane. This is distinct from Starlink’s current Direct-to-Cell approach, which uses a proprietary LTE-like air interface. U-DESERVE aims for full 3GPP NR NTN compliance.

The use cases targeted include:

1. Vehicular resilience. Emergency service vehicles maintaining 5G connectivity in areas without terrestrial coverage during disaster response.
2. Maritime. Ships in European waters accessing 5G data services via LEO constellation.
3. Rural broadband supplement. Providing baseline 5G NR coverage to areas where terrestrial deployment is economically unviable.

U-DESERVE represents the European industrial strategy for NTN: sovereign technology, 3GPP compliance, and multi-vendor interoperability. It is a direct response to the de facto dominance of US-based players (SpaceX/Starlink, AST SpaceMobile) in the D2D satellite market.

The Broader NTN Ecosystem in 2026

U-DESERVE does not operate in a vacuum. The NTN field is intensely competitive:

AST SpaceMobile completed voice calls, video calls, and 4G/5G data sessions from unmodified smartphones via its BlueWalker 3 test satellite. Its commercial BlueBird constellation targets the Earth-Fixed Cell model with phased-array antennas capable of creating multiple fixed ground cells per satellite. AT&T, Vodafone, and Rakuten are commercial partners.

Starlink Direct-to-Cell has moved from SMS-only to data services (WhatsApp, Google Maps confirmed on T-Mobile in the US). SpaceX’s approach uses a proprietary LTE-band air interface rather than full 3GPP NTN, which gives it speed-to-market advantage but creates an interoperability question long-term.

Qualcomm’s Snapdragon Modem-RF platforms now include NTN support in the X75 and X80 modem families. This is critical because NTN adoption is ultimately gated by chipset availability. When the modem supports NTN natively, the device OEM does not need to add a separate satellite modem.

MediaTek has also demonstrated NTN capability in its Dimensity platform, ensuring multi-vendor competition at the chipset level.

The European Space Agency (ESA) and GSMA Foundry have committed significant funding to space-mobile convergence, recognizing that NTN is not an alternative to terrestrial 5G but a complement that extends coverage universally.

What This Means for RF and Network Engineers

NTN is not a niche topic for satellite specialists. It is entering the mainstream 5G architecture, and it changes the engineer’s toolkit in concrete ways.

Timing and synchronization. NTN requires GNSS-assisted timing advance and Doppler pre-compensation. If you work on UE conformance testing or protocol stacks, you need to understand how TA is derived from GNSS position rather than measured from the downlink.

Handover planning. With EFC, the handover is satellite-to-satellite for the same fixed cell. The triggers are ephemeris-based (predictable orbital mechanics) rather than measurement-based (RSRP/RSRQ thresholds). This is a paradigm shift in mobility management.

Coverage modeling. Traditional RF propagation models (Okumura-Hata, COST 231) do not apply to satellite links. NTN uses free-space path loss with atmospheric attenuation, scintillation, and rain fade models. If you are doing coverage planning for hybrid terrestrial-NTN networks, your propagation toolchain needs updating.

KPI frameworks. How do you measure “coverage quality” when the serving cell moves at 27,000 km/h? When the RTT is inherently 52 ms before any core network delay? NTN requires new KPI definitions and new acceptance criteria for field validation.

Protocol analysis. NTN introduces new RRC and NAS signaling elements: extended discontinuous reception (eDRX) configurations for satellite, NTN-specific SIB extensions, and GNSS-assisted procedures. If you decode Layer 3 messages, expect new information elements you have not seen before.

The bottom line: 5G NTN is not a separate technology. It is 5G NR with a different propagation environment. The same RRC, the same NAS, the same QoS framework — but with timing, mobility, and geometry constraints that challenge every assumption built on terrestrial experience.

---

Want to go deeper on NTN terminology? Check the HiCellTek glossary for detailed definitions of NTN, gNB, TAI, HARQ, and every other acronym in this article. If you are working on NTN field validation or hybrid terrestrial-satellite testing, contact us — this is exactly the kind of network complexity where specialized expertise matters.