Module 4: D2D / Direct-to-Cell Explained

How direct-to-device NTN works, key players, architectures, and current performance reality.

What Direct-to-Device Means

Your phone communicates directly with a satellite — no ground antenna relay needed.

How a Regular Phone Talks to Space

Direct-to-Device (D2D) NTN lets a standard smartphone — the same one in your pocket right now — communicate directly with a satellite orbiting roughly 550 km overhead. The critical insight is asymmetric link budgets: your phone has a tiny antenna and transmits at about 200 mW, but the satellite compensates with a massive phased-array antenna (some spanning over 60 square meters) and extremely sensitive receivers. This combination closes the link budget across hundreds of kilometers of free-space path loss.

The physics are demanding. A signal traveling 550 km to LEO experiences roughly 150 dB of path loss in sub-2 GHz bands. The satellite's large antenna provides enough gain (30-40 dBi) to make up the difference, but the margins are thin — which is why D2D works best outdoors with a clear view of the sky.

How D2D Differs from Traditional Satellite Phones

Traditional satellite phones from Iridium, Thuraya, or Globalstar use specialized hardware with purpose-built antennas and proprietary protocols. They cost $500-$1,500 per handset and $1-$2 per minute of voice. D2D flips this model entirely: the phone is standard, the intelligence and antenna gain live on the satellite. Your existing phone number, your existing plan, your existing device — just with coverage that extends to anywhere you can see the sky.

D2D NTN turns every compatible smartphone into a satellite-capable device — no special hardware, no external antenna, no satellite phone subscription required. The addressable market jumps from millions of satellite phone users to billions of smartphone owners.

MNO Partnerships and Spectrum

D2D services operate in licensed mobile spectrum — the same frequencies your phone already uses for LTE or 5G. This requires partnerships between satellite operators and mobile network operators (MNOs) who hold the spectrum licenses. T-Mobile partnered with SpaceX, AT&T and Verizon with AST SpaceMobile. The MNO contributes spectrum rights and subscriber relationships; the satellite operator contributes space infrastructure.

Coverage and Limitations

Coverage extends to anywhere with an unobstructed view of the sky — open fields, oceans, deserts, rural roads. However, D2D does not work indoors. Building walls introduce 15-25 dB of additional signal loss that the already-thin link budget cannot absorb. Urban canyons with tall buildings also degrade performance. Typical throughput today sits below 1 Mbps for most services, sufficient for messaging and basic data but not video streaming.

Think of D2D as the ultimate coverage gap filler: it will not replace your local cell tower, but it ensures you are never truly without connectivity when you step outside.

Why it matters: D2D is the most disruptive NTN model — it turns every phone into a satellite-capable device.

Transparent vs Regenerative Architectures

Bent-pipe satellites relay signals to ground; regenerative satellites process them in orbit.

Transparent Architecture (Bent-Pipe)

In a transparent D2D architecture, the satellite is essentially a radio mirror in orbit. The signal path goes: phone transmits to satellite, satellite amplifies and frequency-shifts the signal, then relays it down to a ground-based gNB (base station) connected to the mobile core network. All baseband processing — scheduling, decoding, error correction — happens on the ground. The satellite itself has no knowledge of the protocol it is carrying.

This approach keeps the satellite simple and cheap to manufacture. A bent-pipe transponder is well-understood technology with decades of heritage in GEO communications. However, it requires a ground station within the satellite's footprint at all times. Over oceans or remote regions far from any gateway, service may be unavailable. Transparent architecture is what 3GPP Release 17 standardized for NTN.

In transparent mode, the satellite is protocol-agnostic — it does not know whether it is relaying LTE, NR, or any other signal. All intelligence lives on the ground.

Regenerative Architecture (On-Board Processing)

In a regenerative architecture, the satellite carries a full gNB (or at least the gNB-DU distributed unit) on board. The phone's signal is demodulated, decoded, and processed in orbit. The satellite then communicates with the core network via a feeder link to a gateway, or routes traffic to other satellites through inter-satellite links (ISLs). This eliminates the extra ground hop, saving 50-100 ms of round-trip latency compared to transparent mode.

3GPP Release 19 standardizes the regenerative NTN payload, defining how the gNB-DU/CU split works in orbit. This allows the satellite to make real-time scheduling decisions, adapt to changing demand, and operate autonomously without continuous ground station visibility.

Latency and Trade-Offs Compared

Transparent mode adds an extra ground hop: phone to satellite (25 ms), satellite to ground gNB (25 ms), then back — totaling roughly 50-100 ms additional latency versus a regenerative satellite that processes locally. Regenerative payloads save this but come with trade-offs: the satellite is heavier, more expensive, consumes more power, and is harder to upgrade once launched. Transparent satellites can be upgraded by swapping ground equipment; regenerative payloads are fixed at launch.

The industry is moving from transparent (Rel-17) toward regenerative (Rel-19), but most D2D satellites flying today still use the simpler bent-pipe approach. The transition will accelerate as on-board processing hardware becomes lighter and more power-efficient.

The Three gNB Placement Variants

3GPP defines three specific deployment options for where the base station (gNB) sits in an NTN architecture:

Variant 1 — Full gNB on satellite (regenerative): The satellite carries the complete gNB (radio and processing). The phone connects via NR-Uu air interface, the satellite processes everything on board, and communicates with the core network via a feeder link. Lowest latency but heaviest, most complex satellite.

Variant 2 — gNB on ground, satellite as relay (transparent): The gNB sits at the ground station behind the gateway. The satellite uses a proprietary satellite radio interface (SRI) to relay signals between the phone and the ground gNB. The satellite does no baseband processing — it just amplifies and frequency-converts. Simplest satellite but requires constant ground station visibility.

Variant 3 — Split gNB (DU in orbit, CU on ground): The satellite carries only the gNB Distributed Unit (DU) — the real-time radio processing — while the Centralized Unit (CU) runs on the ground. This is a pragmatic middle ground that brings some processing to orbit without the full complexity of Variant 1. This split allows inter-satellite handover via the Xn interface.

Earth-Fixed Beams and Satellite Handover

LEO satellites move at roughly 7.5 km/s, crossing the sky in about 5-10 minutes. Without compensation, the coverage area (beam footprint) would sweep across the ground with the satellite, forcing constant handovers. The solution is earth-fixed beams: the satellite uses active beam steering (phased array antennas) to keep its beams locked to fixed areas on the ground as it flies overhead. From the phone's perspective, the beam stays in place — only the satellite behind it changes.

When a satellite passes out of range, an inter-satellite handover occurs. The 3GPP mechanism works like this: the ground gNB broadcasts SIB19 (System Information Block 19) containing satellite ephemeris data — the precise orbital position and trajectory of both the current and incoming satellite. The phone uses this data plus GPS to predict when handover should occur. The UE is pre-configured with the incoming satellite's radio parameters, and when conditions are met (elevation angle drops below threshold, or a timer fires), the UE initiates a conditional handover — synchronizing with the new satellite before releasing the old one (make-before-break). The phone must continuously compensate for Doppler frequency shift caused by the satellite's velocity, and adjust its timing advance to account for the changing distance.

Conditional handover for NTN requires 3GPP Release 17+ devices and modern chipsets like Qualcomm Snapdragon X80 or later. The phone does significant work — tracking satellite positions, compensating for Doppler, and managing timing — that would be unnecessary in a terrestrial cell.

Why it matters: Architecture choice determines latency, capacity, and operational complexity of the D2D service.

Key Players

Starlink D2C (T-Mobile), AST SpaceMobile (AT&T/Verizon), and Lynk Global lead the D2D race.

Starlink Direct-to-Cell

SpaceX's Starlink D2C is the largest D2D constellation by far. Out of 10,000+ active Starlink satellites, over 650 carry dedicated D2C antennas. The service launched commercially in July 2025 through a partnership with T-Mobile, initially offering text messaging and expanding to app-based data services. By early 2026, Starlink D2C reports over 6 million registered users. SpaceX acquired S-band spectrum from EchoStar to supplement T-Mobile's PCS spectrum, giving the service more bandwidth headroom for future capacity growth.

Starlink's strategy relies on sheer constellation density: with thousands of satellites, there is almost always one overhead, enabling near-continuous service rather than intermittent passes.

AST SpaceMobile

AST SpaceMobile takes a fundamentally different approach: fewer but much larger satellites. Their BlueBird 6 satellite, launched in December 2025, carries the largest commercial phased-array antenna ever deployed — approximately 2,400 square feet (223 square meters). This massive aperture provides enough gain to support not just messaging but voice calls and video at usable quality. In controlled testing with BlueBird 1-5, AST demonstrated 14 Mbps downlink to an unmodified smartphone.

AST has MNO partnerships with AT&T, Verizon, Vodafone, and Rakuten, giving it multi-market reach. The company is targeting a constellation of 45-60 satellites by end of 2026 for broad commercial service.

AST SpaceMobile's BlueBird arrays are so large they can establish voice-grade links with unmodified smartphones — a first in commercial satellite history. A single BlueBird satellite can support thousands of simultaneous connections across its coverage footprint.

Lynk Global and Other Players

Lynk Global pioneered the D2D concept with early demonstration flights and currently operates 3 licensed satellites. The company holds 36 commercial contracts across 50 countries, focusing on SMS-first connectivity for underserved markets. Lynk is merging with Omnispace and has announced a multi-orbit partnership with SES to extend its reach.

Skylo operates as an NTN service layer rather than launching its own satellites. It provides NB-IoT connectivity through existing satellite infrastructure, targeting IoT and asset-tracking use cases rather than consumer smartphone connectivity.

The D2D market is splitting into two strategic camps: SpaceX bets on thousands of small-antenna satellites for ubiquitous coverage, while AST SpaceMobile bets on fewer satellites with massive antennas for higher per-user throughput. Both approaches have technical merit — the market will determine which wins.

Why it matters: Competition between approaches (large arrays vs many small sats) is driving rapid innovation.

Phone Compatibility

Starlink D2C works with existing LTE phones. Some services may need NTN-capable chipsets.

Why Existing LTE Phones Already Work

Starlink's Direct-to-Cell service broadcasts standard LTE signals from orbit — the satellite essentially acts as an eNodeB (LTE base station) in the sky. When your phone scans for cell towers, it sees the satellite as just another tower. It connects using the same LTE protocols it would use on the ground. No firmware update, no app download, no hardware modification required. AST SpaceMobile works the same way, targeting unmodified handsets with standard cellular protocols.

This backward compatibility is what makes D2D so commercially powerful. There are over 5 billion smartphones in use globally, and a significant fraction already support the LTE bands used by D2D services.

Technical Challenges Satellites Must Solve

Making a phone believe it is connecting to a normal tower from 550 km away requires the satellite to handle several technical challenges transparently. Extended timing advance is the biggest: terrestrial LTE supports a maximum timing advance of about 0.67 ms (corresponding to ~100 km cell radius), but LEO satellites need up to 20 ms. The satellite pre-compensates so the phone never exceeds its normal timing range. Doppler shift is another challenge — a LEO satellite moves at 7.5 km/s, causing frequency shifts of up to plus or minus 24 kHz. The satellite pre-corrects its transmission frequency so the phone sees a stable signal. HARQ (Hybrid Automatic Repeat Request) timing is also adjusted, since the standard 8 ms HARQ round-trip assumed by LTE is impossible with satellite propagation delays.

The satellite does all the heavy lifting: it pre-compensates for Doppler shift, extends timing advance ranges, and adjusts HARQ timing — so your phone never knows it is talking to space instead of a tower down the road.

NTN-Specific Chipsets and Software

While existing phones work for basic D2D, next-generation chipsets add dedicated NTN capabilities for better performance. Qualcomm's Snapdragon X75 and X80 modems include an NTN modem alongside the terrestrial cellular modem. MediaTek's Dimensity platform adds satellite messaging support. Samsung has built NTN support into its Exynos chipsets. These NTN-aware modems reduce connection setup time, improve power efficiency during satellite searches, and enable smoother handoffs between terrestrial and satellite cells.

Google's Android 15 introduced the Android Satellite API, giving app developers a standardized interface for satellite connectivity features. This means messaging, emergency, and location apps can detect and use satellite connections without carrier-specific workarounds.

Future NTN-optimized handshakes will dramatically improve battery life during satellite connectivity — current phones waste significant power scanning for and maintaining links that the baseband was never designed to handle at satellite distances.

Why it matters: No special hardware means instant addressable market of billions of existing phones.

Current Performance Reality

Sub-1 Mbps today, with LTE-like speeds expected outdoors as constellations grow.

Real Measured Performance Today

Starlink D2C outdoor tests show throughput of 1-5 Mbps depending on satellite pass geometry, constellation congestion, and environmental conditions. Most users experience the lower end of this range during early service rollout as constellation density is still ramping. AST SpaceMobile claimed 14 Mbps downlink in a controlled demonstration with its BlueBird 1-5 test satellites, though real-world multi-user performance will be lower. Voice calls have been demonstrated at acceptable quality over both platforms — a milestone that seemed unlikely just a few years ago.

Latency for LEO D2D services sits at 30-50 ms typical round-trip, which is comparable to terrestrial cellular in rural areas and dramatically better than GEO satellite services (600+ ms). For messaging and web browsing, this latency is essentially imperceptible.

The Indoor Challenge

Indoor connectivity is the fundamental limitation of D2D satellite services and is unlikely to be fully solved. The link budget from a phone to a LEO satellite is already operating with thin margins outdoors. Building walls, roofs, and windows introduce 15-25 dB of additional signal attenuation — equivalent to reducing signal strength by a factor of 30 to 300. Low-E glass windows (common in modern energy-efficient buildings) are particularly problematic, attenuating signals by 20-30 dB. The physics here are unforgiving: D2D is an outdoor and near-window technology.

Current D2D services prioritize coverage over speed: the goal is connectivity where none existed before, not replacing your local cell tower. Think of it as going from zero bars to some bars — that is the transformation.

Constellation Density and Service Continuity

The number of satellites in a constellation directly determines how often a user has one overhead and for how long. With today's Starlink D2C fleet of 650+ equipped satellites, coverage passes are frequent but not continuous in all areas. As more D2C-equipped satellites launch, service continuity improves — eventually reaching near-100% outdoor availability. AST SpaceMobile's approach with fewer, larger satellites means coverage passes may be less frequent initially, but each pass delivers higher throughput.

Performance Roadmap

The industry roadmap progresses through clear phases. Messaging services are available now (2025-2026). Data applications like maps, email, and weather follow as constellations grow. Voice calling becomes reliable at scale in 2026-2027 as AST SpaceMobile and Starlink both expand capacity. Broadband-like experiences approaching 10+ Mbps outdoors are targeted for 2027 and beyond, contingent on next-generation satellite hardware and denser constellations.

Do not expect D2D to replace your home broadband or even match a good urban cell connection. But for the 90% of Earth's surface that has zero cellular coverage today, going from nothing to messaging and basic data is genuinely transformative.

Why D2D is Technically Hard: The 5G Core Challenges

D2D satellite connectivity introduces several challenges that terrestrial 5G networks never face:

Tracking Area Code (TAC) Dynamics: In terrestrial networks, cells are fixed and have static TACs. In NTN, the tracking areas are fixed to the ground, but the satellite serving them changes as it orbits — the TAC assignment must be updated as different satellites sweep overhead. Cell sizes are typically 100-500 km (compared to a few km for terrestrial), and can reach 1,000 km for satellites at 1,200 km altitude.

HARQ and Error Handling: Hybrid Automatic Repeat Request (HARQ) retransmission mechanisms designed for terrestrial millisecond-scale latency break down when the round trip is 50+ ms. NTN modifies HARQ to use more robust modulation and coding schemes that tolerate the increased delay, trading peak throughput for reliability.

Power Control: A phone needs 10-20 dB more transmit power to reach a satellite than a terrestrial tower — that is 10 to 100 times more power. This drains batteries rapidly. Adaptive power control and efficient beam-forming between the UE and satellite are essential to manage energy consumption while maintaining link quality.

Doppler Shift Compensation: LEO satellites move at 7.5 km/s relative to the ground, causing frequency shifts of up to ±24 parts per million. Both the satellite and the phone must continuously compensate for this shift in real time — something terrestrial phones never need to do.

Why it matters: Setting realistic expectations prevents overpromising — D2D is transformative for coverage, not a tower replacement.