LEO vs GEO: Orbital Mechanics, Latency Bounds, and Doppler Shift Explained

The orbit is only the first term

A geostationary satellite circles above the equator at approximately 35,786 km and appears stationary to an observer on Earth. Low Earth orbit extends to about 2,000 km; broadband constellations occupy particular shells within it. Medium Earth orbit sits between LEO and GEO. A lower orbit reduces propagation distance but requires moving beams, handovers and a larger constellation to sustain coverage. VSAT describes a small earth terminal and does not, by itself, specify an orbit.

For an ideal overhead GEO bent-pipe path, a request and its reply traverse four ground-to-space legs. The propagation floor is 4 × 35,786 / 299,792, or about 0.477 seconds. Real slant ranges, terrestrial routing and queues make it longer. For two idealised 550 km ground-space legs each way, the corresponding floor is about 7.3 ms. That calculation is not a promise of single-digit Internet latency.

Doppler and handover

Relative radial velocity shifts the apparent carrier frequency. The first-order relationship is Δf ≈ vᵣf/c. At 12 GHz and an illustrative 7 km/s radial velocity, the magnitude would be approximately 280 kHz. Actual radial velocity changes throughout a pass and is not simply the satellite’s orbital speed. Modems need timing and frequency compensation; users should leave this to supported terminal firmware.

Beam or satellite handovers may change delay and packet delivery even when average signal strength is healthy. Evaluate the distribution of round-trip time and burst loss, rather than using one speed test as evidence of suitability for remote desktops or industrial telemetry.

Build a useful measurement plan

Choose a nearby provider endpoint and the actual application endpoint. Log wired tests at busy and quiet hours, keeping terminal obstruction events and WAN utilisation beside latency samples. Calculate median and 95th-percentile delay and the longest loss burst. A low average can hide a short outage that disconnects a control session.

Design decision

Place safety-critical control and its immediate feedback loop locally. Use satellite for supervision, synchronisation and remote access with explicit interruption handling. Before changing providers, distinguish orbital delay from Wi-Fi contention, an overseas application region, VPN detours and upload bufferbloat. The best remedy depends on which term dominates.

From concept to acceptance

Worked design exercise

Suppose a remote desktop takes six sequential request–response exchanges to open a view. At 600 ms RTT, the serial waiting component alone can approach 3.6 seconds; at 60 ms it is about 0.36 seconds. Parallel requests, caching and server processing change the observed result, so measure the actual application rather than simply multiplying its packet count. The useful architectural question is whether those six dependencies can be combined or moved closer to the user.

Accept the service only after measuring response time under representative upload load and several handover periods. Save both application timings and network measurements so future troubleshooting can distinguish a changed route from changed software.