Satellite links and telemetry: what Starlink actually changes

What RF telemetry cannot do

A conventional telemetry link is superb at exactly one thing: delivering a few megabits per second from an aircraft to a ground station with about a millisecond of latency, deterministically, for as long as the aircraft stays in sight of the antenna. That last clause is the whole limitation.

The radio horizon for an aircraft at thirty thousand feet is roughly three hundred and fifty kilometres. At five thousand feet it is about a hundred and forty. Beyond that line the link stops, and with it the real-time visibility: the data is still recorded, but nobody on the ground knows what is happening until the aircraft lands.

For a campaign flown in a test corridor over the range, that is a non-problem. For an oceanic leg, a long-range cruise validation, a polar route, an endurance flight or an unmanned platform working two hundred nautical miles offshore, it is the defining constraint of the whole campaign. Historically the answers were a chase aircraft, a relay, a mobile ground station, or simply accepting that the flight is blind.

What a low-orbit constellation brings

A satellite path replaces the geometry entirely: the aircraft looks up rather than sideways, and the coverage question becomes "is there a satellite overhead" rather than "am I within line of sight of the antenna". For a constellation of thousands of satellites in low orbit, the answer is almost always yes.

That table already contains the honest conclusion. A geostationary link at half a second of latency cannot support a real-time test point. A narrowband L-band service — the sort used today for over-the-horizon tracking — carries a few hundred parameters at a low rate and nothing more, which is genuinely useful and not telemetry. A low-orbit broadband constellation is the first option that is fast enough and wide enough to be interesting, and it comes with its own set of problems.

The asymmetry nobody mentions

Satellite broadband was designed to deliver video to homes and passengers. Its capacity is therefore heavily biased towards the downlink — from the satellite to the terminal. The return direction, from the terminal upwards, gets a fraction of it.

Telemetry needs the opposite. The whole point is to move data from the aircraft to the ground, which is the constrained direction on a system whose brochure figure describes the other one. An aviation-grade terminal announcing hundreds of megabits per second is announcing the direction you do not need; the number that governs your campaign is the uplink, and it is typically an order of magnitude smaller.

This is not a fatal objection, it is a specification discipline: when you evaluate a service for telemetry, the only figure that matters is sustained uplink throughput under real flight conditions, and it is rarely the one printed first.

Latency, and why determinism beats bandwidth

Tens of milliseconds sounds excellent next to half a second, and it is. But a flight-test engineer clearing a point does not care about the average latency; they care about the worst one, and about whether the link can stall.

A conventional telemetry link either works or does not, with a fixed delay you can write in a document. A satellite IP path is a packet network: it retransmits, it queues, it hands over between satellites, and occasionally it stops for a second or two. A one-second stall is invisible on a video call and unacceptable while a flutter point is being flown.

The real revolution is the return path

Everything above treats a satellite link as a bigger, slower, wider version of what already exists. The genuinely new capability is different: for the first time, the link goes both ways.

Classical PCM telemetry is a broadcast. The aircraft transmits a frame it was configured with before take-off, and nothing on the ground can influence it. Everybody who has run a campaign knows the consequence: the parameter you wish you had put in the telemetered frame is the one you discover you need at point four of sortie two, and the answer is to land, reconfigure and fly again.

A bidirectional path removes that. Change the telemetered subset in flight. Raise the rate on three channels for the next point. Ask for the ten seconds around the event you just saw, at full recorded rate, instead of the reduced stream. Pull a video clip. Push a new frame format. Check that the recorder is healthy and has the space it claimed. None of that is exotic — it is ordinary network behaviour, and instrumentation has simply never had it.

The operating concept changes accordingly, from "record everything and hope the right thing is in the downlink" to "record everything and fetch what matters". That is a bigger shift than any bandwidth figure.

What it changes concretely

• Beyond-line-of-sight campaigns get real-time visibility for the first time: oceanic legs, ferry flights, endurance sorties, offshore work, polar routes.
• A three-tier architecture becomes natural — full rate recorded onboard, reduced rate over the deterministic range link inside the footprint, and a satellite path for awareness and retrieval everywhere else.
• Turnaround collapses. A three-hour sortie is a hundred and sixty gigabytes and will not go up a satellite link — that is fourteen hours at twenty-five megabits per second. But the two hundred megabytes the analyst actually asked for takes about a minute, and can be on their desk before the aircraft has landed.
• A distributed campaign becomes possible: the aircraft in one country, the specialists in another, watching the same data without anybody boarding a plane.
• And the largest market is not flight test at all — it is operational fleets sending health, usage and mission data off the aircraft continuously instead of on a memory card once a week.

The programmes that have no ground segment at all

Everything above quietly assumes you already own a telemetry ground station. A large part of the industry does not, and for those programmes the comparison is not satellite against the range link. It is satellite against nothing at all.

It is worth spelling out what a ground segment actually is. An autotracking antenna of two to four metres on a steerable pedestal, with a radome or a shelter, a foundation and a survey, sited somewhere the aircraft can be seen from. A telemetry receiver, a combiner, a bit synchroniser, a frame synchroniser and a decommutator. A frequency allocation, obtained and coordinated. And somebody to operate all of it during every single flight. That is a capital project with civil works in it, and it has to be within line of sight of wherever you intend to fly.

The swap is capital expenditure for operating expenditure, and that is exactly the right way round for a programme that flies for six weeks, or from a different airfield each time, or once. No asset to depreciate, no shelter to build, nobody to keep trained on equipment used a month a year, and no logistics when the campaign moves country. The ground station becomes a laptop and a connection — which also means the specialists do not have to travel, a line that matters more on a small budget than on a large one.

That is what opens the door for whole categories of programme that have never had real-time telemetry:

• Unmanned platforms of every size, which fly beyond line of sight by definition and frequently already carry a satellite terminal for command or for the payload.
• eVTOL and advanced air mobility programmes — often young companies with no telemetry heritage, flying from small airfields with no range behind them.
• Light aircraft certification, retrofits and modification approvals, where the whole programme budget would not buy a tracking antenna.
• Research institutes, universities, balloon and high-altitude work, and one-off scientific campaigns.
• Established manufacturers running a campaign away from their home range, where the alternative is a mobile station and a fortnight of logistics.
• And ferry or delivery flights, for which nobody has ever built a ground station.

The boundary deserves naming too, because the argument is not universal. A subscription recurs, and a facility that flies every day for ten years may well be cheaper owning its antenna. And there is a third option this comparison should not hide: renting a mobile telemetry station, or buying time on somebody else range. Satellite wins clearly on short, occasional and geographically scattered campaigns — and it wins outright on the ones that would otherwise fly with nobody watching.

The obstacles, honestly

A satellite terminal is airborne equipment, and the list of what that implies is long enough to decide the schedule.

• The antenna. A broadband aero terminal is a substantial flat-panel installation with mass, drag, power consumption and a structural modification behind it. On a transport aircraft that is routine. On a compact aircraft, a helicopter or a small unmanned platform it is often simply impossible.
• Attitude. The terminal looks up. A high-bank turn, a pull-up, an inverted pass — precisely what flight test does — puts the airframe between the antenna and the satellite. The failure geometry is different from a downward-looking telemetry antenna, and on an agile platform it is worse.
• Certification. DO-160, EMC, a new emitter on the aircraft, and an installation approval. And the range authority has to accept an additional transmitter operating during your test.
• Sovereignty and security. Test data is often classified or commercially sensitive, and a commercial constellation routes it through gateways and jurisdictions you do not control. Encryption at the source addresses confidentiality; it does not address metadata, availability or the question of who can turn the service off.
• Service guarantees. A commercial operator offers a subscription, not a campaign commitment. Coverage can be restricted over some areas, and priority is not yours. A test range antenna, whatever its limitations, obeys you.
• Cost model. A subscription and a data volume rather than equipment you own — an advantage for an occasional campaign and a liability for a facility that flies every day, as the section above sets out.

None of these is a reason not to do it. They are the reasons the answer is a second path alongside the existing one rather than a replacement for it, and the reason the first serious users are large aircraft on long routes rather than compact aircraft over a test range.

What the equipment has to be able to do

If a satellite path is going to be useful, the acquisition system has to be built for a link that is intermittent, asymmetric and best-effort. Four capabilities decide whether it works.

1. An IP output alongside the PCM one. The data has to leave the unit as packets, on Ethernet, at a rate that can be set independently of the recorded rate.
2. Store and forward. When the link stalls, the stream queues and resumes; when it returns, the gap is filled from the recording rather than lost. A satellite path without gap-fill is a link that lies about what it delivered.
3. On-demand retrieval. A request from the ground — this channel, this time window, at full rate — served from the onboard recording. This is the feature that justifies the whole architecture.
4. Rate adaptation and encryption at the source. The stream has to shrink when the link does, and it has to be encrypted before it leaves the aircraft, because the path is not yours.

The first of those already exists on any modern recorder that streams over Ethernet — which is the quiet reason this transition is closer than it looks. The remaining three are software on both ends of a link, not new hardware. The revolution, if there is one, is not in orbit. It is in the ground software learning to ask questions of an aircraft that can finally answer.