LoRaWAN Goes to Orbit: NTN Satellite Direct-to-Device Explained

1. Start with a coverage map

Every engineer who has shipped an asset-tracking project has seen the same map: excellent LoRaWAN gateway coverage inside the project area, and a blank ring outside it — ocean routes, cross-border highways, ranchland, desert mining sites, polar research stations.

That blank ring is not a deployment failure. It is the physical boundary of terrestrial networks. No number of towers covers a sea lane 400 km from the nearest coastline. The traditional answer was to switch protocols: adopt a proprietary satellite service (Iridium, Astrocast, Myriota and the like), paying the price in proprietary silicon, proprietary stacks, proprietary billing — and complete separation from the terrestrial network you already run.

Since 2020, there is another option: let LoRaWAN itself go to orbit. The device keeps its LoRa chip and its LoRaWAN stack; the "gateway" simply moves from a rooftop to an orbit 500 km overhead. This is NTN (Non-Terrestrial Network) satellite direct-to-device connectivity, or D2D.

This article explains what makes it work technically (LR-FHSS), who is operating it commercially (three operator architectures), what changes in device design, and when you should not use it.

2. Why LoRaWAN in orbit, instead of switching to a satellite protocol

Satellite IoT is not new. Iridium has run short-burst messaging for decades; Sigfox was born global. What changed is three milestones:

• November 2020: Semtech introduced LR-FHSS (Long Range – Frequency Hopping Spread Spectrum), a new physical layer subsequently incorporated into the LoRaWAN RP002-1.0.2 regional parameters specification.
• 2025: European regulators formally approved satellite-to-low-power-device communications, removing the last regulatory obstacle.
• 2025–2026: Operators including Lacuna Space and Plan-S launched satellites at scale, moving satellite LoRa from pilots to commercial deployment.

The LoRa camp's logic in orbit mirrors its logic on the ground: use an open standard and low-cost silicon to undercut proprietary offerings. Third-party analysis sketches the typical economics: cellular roaming in remote regions quotes around $14 per device per year, while a LoRa tracker node costs about $3. Cheap terminals plus a shared protocol stack mean that a team already running terrestrial LoRaWAN migrates to satellite backfill with minimal learning curve and minimal supply-chain change.

A LoRa Alliance statement from October 2023 listed a long roster of satellite-active members: Actility, EchoStar Mobile, Eutelsat, Lacuna Space, Plan-S, Senet, Wyld Networks and more — spanning satellite operators, network platforms, and device vendors.

3. The technical core: LR-FHSS, a physical layer designed for satellite passes

Conventional LoRa modulation (CSS, chirp spread spectrum) has one fatal weakness for satellites: Doppler shift. A LEO satellite sweeps past at roughly 7.5 km/s, and the resulting frequency offset is enough to break lock on a narrowband chirp signal in a 125 kHz channel.

LR-FHSS is the answer built for exactly this. Its mechanism reduces to three words: fragment, hop, duplicate.

1. Fragment: split an uplink packet into ~50 ms chunks instead of one wide chirp occupying a full channel.
2. Hop: transmit each chunk on pseudo-randomly selected 488 Hz narrowband subchannels within a total occupied bandwidth of 137 kHz, 336 kHz or 1.523 MHz depending on region. The extremely narrow subchannels are inherently tolerant of Doppler shift; hopping spreads collision and interference risk.
3. Duplicate: the packet header (carrying the hopping-sequence seed, data rate, payload length and coding rate) is repeated 2–3 times; the payload is convolutionally encoded at rate 1/3 or 2/3 and split into fragments — recover the original payload once enough fragments arrive (1/3 of fragments for CR 1/3, 2/3 for CR 2/3).

Parameter	LR-FHSS DR8	LR-FHSS DR9
Coding rate	1/3	1/2
Header copies	3	2
Data rate	162 bit/s	325 bit/s
Sensitivity	-137 dBm	-137 dBm

Key qualifiers:

• Uplink only. LR-FHSS applies to the uplink; the downlink remains conventional LoRa-CSS. ACKs, receive windows and Class A timing behave exactly as on a terrestrial network.
• Link budget comparable to SF12 (155 dB class), but with far better interference robustness and capacity thanks to hopping and coded redundancy. Vendor material claims two orders of magnitude more capacity than conventional LoRa and gateway capacity jumping from thousands of nodes to millions — take those as marketing numbers until benchmarked.
• Packets stay small. LR-FHSS maxes out around 50 bytes per packet. It expands "how many devices can send small messages simultaneously," not "how large one message can be." It complements SCHC compression (issue 31): SCHC shrinks headers, LR-FHSS spreads capacity.
• Chip capability ≠ regional data rate. SX1262 gains LR-FHSS support via software upgrade, and LR1121 (Sub-GHz + 2.4 GHz + S-band tri-mode) is the mainstream radio for new satellite-capable devices — but whether a modulation is usable inside LoRaWAN depends on the DR tables of the regional parameters revision you target. This is where vendor slides mislead.

4. The commercial ecosystem: three architectures, three trade-offs

Operators running LoRa-based satellite services commercially fall into three architecture classes.

Dimension	Lacuna Space	Plan-S Connecta	EchoStar Mobile
Orbit	LEO, ~550 km	LEO, ~500 km	GEO, ~36,000 km
Constellation	~8 satellites (early-2026 public figures)	~12–16 satellites	One: EchoStar XXI
Spectrum	Unlicensed ISM	Unlicensed ISM	Licensed S-band
Protocol	LoRaWAN-compatible	LoRaWAN-compatible	LoRa-based proprietary stack
Coverage cadence	Intermittent, ~4–8 passes/day	Intermittent, ~6–10 passes/day	Continuous (regional GEO)
Latency	Minutes to hours	Minutes to hours	Near real-time
Coverage focus	Global (polar orbits)	EMEA/APAC	Europe and adjacent regions

Lacuna Space (UK) is the standards champion. Spun out of the UK Satellite Applications Catapult in 2017, it flew Semtech silicon in 2019 and builds its own in-orbit LoRa receiver payload, LoneWhisper — functionally a LoRaWAN gateway in orbit. Its ground segment hooks directly into The Things Network, ChirpStack and Actility, which is the key to lowest-friction satellite backfill: your application server cannot tell whether a packet arrived from a rooftop gateway or an overhead satellite. Its "Call of the Wild" launches of June 2025 moved it into scale-up mode; a flagship case study is water-quality monitoring along 120 km of the River Teifi in Wales, where terrestrial coverage was too sparse even for basic council decision-making.

Plan-S Connecta (Türkiye) is the vertically integrated fast follower — it builds, launches and operates its own satellites. Its launch cadence is aggressive; constellation density over Türkiye, the Gulf, India and Southeast Asia has already overtaken Lacuna, and a partnership with Türk Telekom pushes satellite IoT through telecom channels. Public case studies concentrate on electricity meter telemetry, fuel tank monitoring and renewable asset health.

EchoStar Mobile is the other extreme: a single geostationary satellite (EchoStar XXI), licensed S-band, and a proprietary stack on top of LoRa rather than native LoRaWAN. GEO removes the Doppler problem and delivers continuous coverage, at the cost of a much worse link budget (36,000 km vs 500 km) and higher device power draw. If you need to send a command and get an ack within seconds, it is the only LoRa-flavored option that delivers — but you give up native-LoRaWAN openness.

Choosing among them is an architecture decision, comparable to choosing Wi-Fi versus Bluetooth: LEO on unlicensed spectrum = lowest terminal cost plus intermittent connectivity; GEO on licensed spectrum = higher power plus continuous connectivity plus interference protection.

5. Five things that change in device design

A satellite LoRa device is not a terrestrial LoRa device with a different antenna. Five design logics must be rewritten:

1. Pass windows are scarce. A LEO satellite is overhead 5–15 minutes per pass; the next pass may be 90 minutes or 12 hours away depending on constellation density. If a device wakes to send a heartbeat with no satellite in view, that transmission energy is wasted. Mature firmware plans wake-ups from orbital almanac data and transmits only inside known pass windows — Lacuna calls this "almanac provisioning." Managing an orbital calendar inside an MCU with no internet connection is the most underestimated software task in these projects.
2. Each pass offers 30–60 useful seconds per device, shared with every other device under the same footprint. This forces a healthy discipline: compute the absolute minimum payload, encode it tightly, stop.
3. The antenna stops being an afterthought. A satellite at low elevation sits at a worse angle than a gateway on a tower. Buoys and livestock collars cannot guarantee orientation: either the antenna pattern is omnidirectional enough, or firmware retries until a high-elevation pass arrives.
4. Battery-life claims need a sanity check. "10-year battery" is the most common figure in satellite IoT marketing — achievable on paper, but only with very low message rates, large primary cells, and clean sleep between transmissions. All three conditions, every time.
5. The duty-cycle math changes. With intermittent connectivity, "how often do we report" is no longer just a business question; it is a joint optimization of pass frequency, message importance and battery budget.

6. What it is not: three honest boundaries

• Not low latency. Under LEO architectures, data latency runs minutes to hours; the only near-real-time option is GEO (EchoStar), which carries its own power penalty.
• Not bidirectional real-time control. Downlink opportunities on LEO operators are sparse. Intermittent connectivity cannot run valve switching. Satellite LoRa serves "report state upward," not "push commands downward."
• Not a terrestrial replacement. It is backfill. Wherever the ground can cover, terrestrial wins on cost, latency, capacity and bidirectionality. The right question is not "should we use satellite" but "how many of these devices will never fall inside terrestrial coverage."

7. Decision checklist

Good fits for satellite LoRa:

1. Maritime and inland-waterway container and vessel tracking — no terrestrial coverage along the route
2. Livestock and wildlife tracking — collar devices, arbitrary orientation, one position per day is enough
3. Remote environmental and water-quality monitoring — boreholes, rivers, rangeland, where manual inspection is expensive
4. Cross-border assets and cold chain — cost-sensitive terminals, low message volume (a few messages per day)

Poor fits:

1. Buildings, factories, campuses — terrestrial coverage wins everywhere that matters
2. Devices needing second-level downlink control — intermittent connectivity cannot do it (unless GEO, accepting the power cost)
3. High-frequency telemetry (minute-level reporting) — LEO capacity and pass cadence cannot carry it; NB-IoT or proprietary satellite fits better
4. ABP legacy fleets expecting a drop-in network swap — session management on satellite NS still follows the OTAA path; device firmware changes are required

8. The platform view: how ThinkLink fits in

ThinkLink is a terrestrial LoRaWAN network server supporting 1.0.2 / 1.0.3 / 1.0.4. The official knowledge base currently carries no dedicated satellite documentation, so platform-capability statements here stay conservative pending official confirmation. The integration path, however, is clear and ecosystem-proven:

• The standard route for satellite messages into a platform is operator ground-segment bridging. Lacuna's service integrates with TTN, ChirpStack and Actility and emits standard NS-side device messages. For platforms like ThinkLink, the path is to treat the satellite operator (or its bridge) as a special "gateway" or upstream NS — thing-model decoding, alert rules, and MQTT forwarding all keep working unchanged.
• Chip capability ≠ platform capability. LR-FHSS support in SX1262 or LR2021 lives at the radio-chip level; usability inside LoRaWAN depends on regional-parameter DR tables. Radio selection and platform selection are two separate confirmations on a satellite project.
• Dual-NS architecture gains a new use here. The terrestrial NS handles routine telemetry while the satellite link acts as a low-frequency backfill channel — uplink dual-send, downlink single-send remains the rule, with the satellite side designed as a status-reporting channel, not a control channel.

One sentence to take away: LoRaWAN in orbit extends the boundary of a terrestrial network from the top of a gateway pole to low Earth orbit. For teams already running terrestrial LoRaWAN, it is not a new technology stack — it is the same stack stretched across the gatewayless parts of the planet, provided you accept intermittent connectivity as a premise and rewrite device design around pass windows.

---

References: LoRaWAN RP002 regional parameters, Semtech LR-FHSS technical material, LoRa Alliance member announcements (Oct 2023), Lacuna Space press releases (Jun 2025), third-party operator comparisons (2026). Specific specifications and pricing subject to official vendor publications.