LEO, MEO and GEO

Choosing an orbital regime is the first architectural decision a satellite system makes, and it is effectively irreversible. Everything downstream follows: how many satellites are needed, what the user terminal looks like, how much the constellation costs to build and replace, what radiation the electronics must survive, and what happens at end of life.

This page works through each of those consequences quantitatively. The Orbit types and regimes page covers the orbital mechanics; this one covers the system engineering that follows from it.

The three regimes

The bottom two rows are where the architectures diverge irreconcilably. A GEO system can be one satellite. A LEO system is never one satellite, because any individual satellite is over a given user for only a few minutes at a time. That single fact cascades into completely different businesses: GEO is a capital project with a fifteen-year payback, LEO is a manufacturing operation with continuous replacement.

Why a satellite stays put at one altitude

The defining property of geostationary orbit is worth stating precisely, because it is often described loosely. A satellite is geostationary when its orbital period exactly matches Earth’s sidereal rotation period of 86 164.09 s, its orbit is circular, and its inclination is zero.

That radius is 35 786 km of altitude. There is exactly one such orbit, it is a ring rather than a shell, and its circumference is 265 000 km. Satellites in it must be separated in longitude to avoid radio interference, typically by 2°, which puts a hard bound of a few hundred usable slots per frequency band. This is why GEO slots are coordinated internationally and treated as an economic asset in a way that no other orbit is.

The sidereal day matters here. Using the solar day of 86 400 s gives an altitude about 90 km too high, and a satellite placed there would drift westward roughly one degree per day, which is not geostationary in any useful sense.

Coverage geometry

How much of Earth one satellite can see is set by geometry alone. From Ground tracks and coverage, the Earth-central angle to the edge of the visible region is

and the fraction of Earth’s surface within that cap is

Evaluating at a 25° elevation mask, which is representative of a small flat-panel user terminal:

The last column is a lower bound and not a design. It assumes the footprints tile the sphere perfectly with no overlap and no gaps, which no real constellation achieves. Real systems need two to four times these numbers, because circles do not tile a sphere, because coverage must be continuous as satellites move, and because inclined orbits do not cover the poles at all.

Starlink’s first shell uses 1584 satellites at 550 km where the bound says 184. The factor of roughly nine covers the tiling inefficiency, continuous coverage with satellites in motion, and capacity rather than mere coverage. That last point is important: above a certain scale, constellations are sized by how many users need service simultaneously, not by whether anyone can see a satellite at all.

Latency

Propagation delay is fixed by the speed of light and cannot be engineered away.

These are best-case figures at zenith, and they exclude the terrestrial network on either end. Real systems add ground segment routing, so an observed LEO round trip of 20 to 40 ms against a theoretical 7 ms is normal and mostly reflects the terrestrial path, not the space one.

The GEO figure is the one that shapes applications. A round trip approaching half a second is imperceptible for broadcast video, tolerable for web browsing, awkward for voice, and disqualifying for interactive applications, competitive gaming, and any protocol whose throughput depends on round-trip time. TCP’s congestion window recovery is proportional to round-trip time, so a GEO link delivers far less than its raw capacity to an unmodified TCP session, which is why GEO operators deploy performance-enhancing proxies that terminate and spoof TCP at both ends.

This latency difference, more than bandwidth, is what created the market for LEO broadband. GEO capacity was never the limiting factor for most consumers; responsiveness was.

Link budget

Path loss scales with the square of distance, so range enters the link budget directly. Using the free space path loss expression from the communications subsystem:

The LEO advantage is nearly 31 dB against GEO, a factor of more than 1200 in received power. That is what makes a small flat-panel user terminal possible at all: the same terminal that closes a LEO link would need roughly a thousand times the antenna area, or a thousand times the satellite transmit power, to close a GEO link at the same data rate.

There is a subtlety worth drawing out. Note that the LEO slant range at 25° elevation is 1123 km, not 550 km. A satellite near the horizon is more than twice as far as one overhead, costing an extra 6.2 dB. Link budgets must close at the worst-case elevation, which means the geometric advantage of LEO is smaller than the altitude alone suggests, and it means low-elevation passes are the design case rather than an edge case.

Radiation

Radiation dose does not vary monotonically with altitude. It peaks in the middle, which makes MEO the harshest commercial regime.

See Space radiation for the belt structure. Two consequences follow for system design.

MEO demands the most expensive electronics. A GNSS satellite at 20 180 km sits in a worse radiation environment than a GEO satellite, which is one reason navigation satellites are costly and long-lead.

LEO permits commercial parts. Below the belts, dose rates are low enough that carefully screened commercial components can survive for the five-year design lives typical of constellation satellites. That is not a minor detail: it is a precondition for building satellites at the cost and rate that large constellations require, and it connects the choice of orbit directly to the manufacturing economics discussed in Launch and mission economics.

Lifetime and disposal

This table is why constellation altitudes have converged downward. Starlink’s 550 km shell decays naturally within about five years of a failure, so a dead satellite removes itself. A constellation at 1200 km does not have that property, and OneWeb’s higher shell must actively deorbit every satellite, which requires the propulsion system to still work at end of life.

The regulatory position has tightened around exactly this. The FCC’s five-year post-mission disposal rule for US-licensed spacecraft, effective September 2024, supersedes the long-standing 25-year international guideline and effectively rules out passive disposal above roughly 600 km. See Space debris and traffic management.

What each regime is actually for

GEO is chosen when the satellite must appear fixed. That property allows a non-tracking user antenna, which is what made consumer satellite television possible: a dish bolted to a wall and never moved again. It also suits weather observation, where a continuous view of one hemisphere is worth more than resolution, and wide-area broadcast, where one transmission serves a continent.

MEO is a compromise that is rarely optimal and occasionally ideal. It is the natural home of navigation, because GNSS needs many satellites visible simultaneously from anywhere with good geometric diversity, and MEO delivers that from a few dozen satellites with orbital periods long enough that the constellation geometry changes slowly. It also suits regional trunking, where a handful of satellites with moderate latency serve high-capacity fixed links.

LEO is chosen for proximity. Close range means low latency, favourable link budgets, and high angular resolution for imaging. The cost is that no single satellite provides continuous service, so LEO systems are inherently constellations, and constellations are inherently manufacturing problems.

The economics diverge

The regime choice determines the shape of the business more than any other decision.

GEO

One or a few satellites, each costing 150 to 400 million dollars, built over three to five years, operated for fifteen. A capital project. Failure of one satellite is a company-scale event.

MEO

Tens of satellites, moderate individual cost, replaced in batches. Usually government-funded for navigation, since the service has no direct revenue model.

LEO

Hundreds to thousands of satellites at hundreds of thousands to a few million dollars each, produced continuously on a line, with a five-year design life and permanent replacement. A manufacturing operation.

Consequence

The LEO model only closes if satellite unit cost and launch cost per kilogram are both low. It is a direct consequence of reusable launch and commercial-parts avionics, not an independent development.

The replacement obligation is the part most often underestimated. A constellation of 1584 satellites with a five-year design life must launch roughly 320 satellites per year forever simply to stand still. That is a production rate of about six per week, sustained indefinitely, and it means the manufacturing line is the constellation, not a phase that precedes it.

References

• Wertz, J. R., Everett, D. F. and Puschell, J. J. (eds.) Space Mission Engineering: The New SMAD, Microcosm Press, 2011, chapters 9 and 13.
• Maral, G., Bousquet, M. and Sun, Z. Satellite Communications Systems, 6th ed., Wiley, 2020.
• Wertz, J. R. Mission Geometry: Orbit and Constellation Design and Management, Microcosm Press, 2001.
• FCC Report and Order 22-74, Mitigation of Orbital Debris in the New Space Age, adopted 2022, effective for most licensees September 2024.
• ITU Radio Regulations, Article 22 and Appendix 30, on geostationary orbital slots and coordination.