Satellite Latency: GEO vs MEO vs LEO

Satellite latency is not a single number assigned to an orbit. It is the sum of propagation, transmission, scheduling, processing, queueing, and terrestrial-routing delays along a specific path. Orbit altitude sets a physical floor, but gateway placement, elevation angle, inter-satellite routing, congestion, and the application endpoint determine the result a user measures.

GEO, MEO, or LEO: which has less latency? With the same number of space legs and comparable geometry, the shorter paths of lower orbits reduce propagation delay. A single transparent GEO hop contributes about 240–280 ms one way, or 480–560 ms propagation RTT before network overhead. The illustrative MEO and LEO paths below have lower physical floors; their service figures depend on the named network, endpoint, route, and load.

Modeled Floors vs Operational References

The floor column assumes a transparent single-satellite path with two vertical space legs each way and no network overhead: RTT = 4h/c. The next column contains different source-defined metrics, not a uniform set of measurements to the same endpoint. Do not compare one system's physical floor with another system's service result. The calculations below show the assumptions; the measurement checklist shows how to test actual service RTT.

Glossary: GEO, LEO, Latency | End-to-End Architecture

Define the Path Before Calculating Latency

Three terms are often mixed together:

• Space leg: one earth-to-satellite or satellite-to-earth segment.
• Satellite hop: an earth-station-to-satellite-to-earth-station path containing two space legs.
• Round-trip time (RTT): the time for traffic to reach the destination and for a response to return. A symmetric single-satellite path contains four space legs in one RTT.

This distinction fixes a common GEO error. Dividing 35,786 km by the speed of light gives about 119 ms. That is only one vertical ground-to-satellite leg. It is not the one-way delay between a user terminal and a gateway, and it is not RTT.

The propagation calculation is:

t = d / c

t     propagation time, seconds
d     total signal-path distance, kilometres
c     299,792.458 kilometres per second in vacuum

The value of c is exact in SI units; see the BIPM SI Brochure, Section 2.2. For a transparent satellite path with user slant range r_user and gateway slant range r_gateway:

one-way space delay = (r_user + r_gateway) / c
space-only RTT      = 2 × (r_user + r_gateway) / c

Use slant range, not altitude, when site coordinates and satellite longitude or ephemeris are available. Altitude equals slant range only in the simplified overhead case. For a fixed GEO satellite, the GEO look-angle calculator can check azimuth, elevation, and idealized horizon visibility. It does not calculate slant range or operational service latency.

Worked Propagation Example

The following calculation intentionally uses a simplified boundary condition: the satellite is directly above both ends of an idealized short ground path, the same satellite carries both directions, and processing, queueing, terrestrial backhaul, retransmission, and inter-satellite links are zero. It is a lower-bound comparison, not a service promise.

Example for GEO:

one-way space delay = 71,572 km / 299,792.458 km/s
                    = 0.23874 s
                    = 238.7 ms

space-only RTT      = 2 × 238.7 ms
                    = 477.5 ms

The result is close to RFC 2488's 239.6 ms ideal one-way value. Real GEO sites are not both directly below the spacecraft. Longer slant ranges explain why the same RFC gives up to 279.0 ms for one satellite hop at the edge of view and notes that an RTT may reach at least 558 ms before other delays.

Boundary Conditions That Change the Result

The simple 4h/c model is useful for checking units and order of magnitude, but it does not model a deployed network. Replace it with full path distance when any of these conditions applies:

• the user or gateway operates at a low elevation angle, increasing slant range;
• traffic crosses one or more inter-satellite links;
• a star network sends remote-to-remote traffic through two satellite hops;
• the return route differs from the forward route;
• a distant gateway adds a long terrestrial backhaul segment;
• demand-assigned access makes a terminal wait for a transmit slot;
• congestion, retransmission, encryption, or application processing is material.

GEO Satellite Latency

A geostationary spacecraft has a nominal altitude of about 35,786 km above the equator. It appears fixed to a ground observer, allowing fixed-pointing antennas and persistent wide-area coverage.

For a single transparent GEO hop, use 240–280 ms one way as a documented space-propagation range, not 120–140 ms. The corresponding propagation RTT is approximately 480–560 ms. RFC 3135, Section 5.1.1 independently describes 240–280 ms sender-to-receiver propagation in geosynchronous VSAT networks and notes that queueing and shared-channel access can add much more.

Measured IP RTT may be higher because the ping target is beyond the satellite gateway, queues form under load, or the access protocol waits for capacity. A transparent bent-pipe transponder does not justify adding a generic 5–15 ms processing allowance; equipment and waveform delay must be taken from the actual modem, codec, scheduler, and payload specifications.

GEO delay is noticeable in conversation, remote desktops, multi-round-trip authentication, and short transactional requests. It does not make every interactive application unusable. Voice can operate over GEO when echo control, jitter management, packet loss, and user expectations are handled, while bulk transfer, broadcast, telemetry, and asynchronous workflows are often less sensitive.

MEO Satellite Latency

MEO covers a broad altitude range between LEO and the higher orbital regimes. The exact boundary depends on the classification system. This page follows the ESA Space Environment Report 2026, Table 1.2: its LEO class has both perigee and apogee from 0 to 2,000 km, while its MEO class has both from 2,000 to 31,570 km. The report classifies boundary-crossing orbits separately. A service's actual orbit must therefore be used instead of a generic “MEO altitude.”

At an illustrative 8,000 km altitude, the idealized space-only RTT is 106.7 ms. Slant range and network overhead move the operational result above that floor. SES describes O3b mPOWER as providing 150 ms satellite latency across large regions in its official system overview. Its 2023 Cruise mPOWERED + Starlink service brief, page 3, explicitly labels the MEO figure 150 ms round-trip satellite latency and lists traffic termination at an SES gateway or defined point of presence. These are named operator and product references, not universal MEO latency or a current service guarantee to an arbitrary internet endpoint.

MEO can support interactive cloud access, voice, and video with substantially less delay than GEO, while using fewer satellites per coverage area than a lower LEO constellation. Terminals still need tracking and handover arrangements because MEO spacecraft are not fixed in the sky.

LEO Satellite Latency

LEO extends up to 2,000 km under the ESA classification used here. A 550 km overhead example produces only 7.3 ms of space-only RTT, but that figure omits the moving slant geometry, gateway path, routing, processing, queueing, and any inter-satellite links.

Operational service figures are therefore more useful for application planning. Starlink's published specifications give 25–60 ms latency on land and more than 100 ms in some remote locations. Treat these as published service references, not measurements made for this article or a guarantee for your site. Gateway availability and route geography can outweigh the few milliseconds in the ideal altitude-only calculation; verify the applicable plan's current specification and test the intended endpoint.

LEO delay also varies over time. The serving satellite changes, slant range changes during a pass, and the chosen ground or inter-satellite route may change. For engineering acceptance, record latency distribution and handover behavior, not only a best-case ping.

Satellite Beam Handover Explained | Hybrid Satellite Networks

Application Impact

Voice and Video

ITU-T G.114, Section 4 does not define 150 ms as a universal pass/fail limit. It says that most applications experience essentially transparent interactivity below 150 ms one way, while highly interactive tasks can be affected at lower values. For general network planning, it recommends not exceeding 400 ms one way and recognizes exceptional cases such as an unavoidable double satellite hop.

Voice quality also depends on codec delay, jitter buffer, packet loss, echo control, and the two parties' conversational behavior. Use the ITU-T E-model when planning speech quality rather than judging a service from propagation delay alone.

Web, Authentication, and Remote Desktop

Interactive workflows often require several dependent request-response exchanges. High RTT makes each serial exchange visible even when bandwidth is ample. DNS caching, connection reuse, modern transports, edge caching, and local breakout can reduce the number or distance of these exchanges, but they cannot remove the orbital propagation floor.

Remote desktop and command-line sessions expose delay immediately because user input waits for visual or textual feedback. Test the actual application against representative endpoints; a ping to the satellite gateway does not include application-server delay.

TCP Throughput and Bandwidth-Delay Product

The bandwidth-delay product is the data volume that must be in flight to fill a path:

BDP = bottleneck bit rate × RTT

For a 100 Mbit/s path with 600 ms RTT:

BDP = 100,000,000 bit/s × 0.600 s
    = 60,000,000 bits
    = 7.5 MB

The sender needs roughly 7.5 MB of effective in-flight data to fill that ideal path, before allowing for loss and protocol overhead. This is why window sizing, congestion control, and loss recovery matter on high-capacity GEO links. RFC 2488, Section 2 identifies the large bandwidth-delay product and long feedback loop as separate satellite-channel constraints.

SCADA and Control Systems

“SCADA” is not one latency requirement. Periodic telemetry may tolerate GEO delay, while protection, teleoperation, or closed-loop control may not. Define the maximum command-response time, jitter, packet-loss behavior, retry policy, and safe failure state for the actual system. Do not infer suitability from orbit class alone.

SCADA over Satellite | QoS over Satellite

How to Measure Satellite Latency

Use a repeatable test plan instead of quoting a single screenshot:

1. Define the endpoints. Test the satellite gateway or provider edge separately from the real application endpoint.
2. Record route context. Note terminal location, service plan, orbit or constellation, gateway region, interface, and VPN state.
3. Measure a distribution. Report sample count, median, 95th percentile, 99th percentile, minimum, maximum, and packet loss.
4. Test unloaded and loaded conditions. Queueing under upload or download load can dominate RTT.
5. Use more than ICMP. Some networks deprioritize or filter ping. Add TCP or application-level measurements on the ports that users actually depend on.
6. Run long enough to observe path changes. For non-GEO systems, include multiple satellite and gateway handovers.

A useful decomposition is:

measured RTT = space propagation
             + access scheduling
             + modem and payload processing
             + gateway and terrestrial routing
             + queueing
             + endpoint response behavior

Ping can measure the total to its chosen target. It cannot, by itself, prove how much came from each term.

Common Calculation and Procurement Errors

• Counting only one GEO space leg. 35,786 km / c ≈ 119 ms is ground to satellite, not end-to-end one-way latency.
• Mixing one-way delay and RTT. State which metric every number represents.
• Using altitude as slant range. This understates delay away from the sub-satellite point.
• Comparing a physics floor with a service SLA. The floor excludes most elements present in an operational measurement.
• Treating all LEO or all MEO systems alike. Altitude, gateways, ISLs, routing, load, and handovers differ.
• Calling 150 ms an ITU hard limit. G.114 provides planning guidance and an interaction-quality context, not a universal application cutoff.
• Adding generic payload-processing numbers. Use equipment and waveform documentation for the deployed system.
• Reporting only the minimum ping. Procurement decisions need percentile latency, jitter, loss, test endpoints, and load conditions.

Choosing an Orbit for a Latency Requirement

Start from the application, not the marketing label:

1. Define acceptable one-way delay, RTT, jitter, and packet loss.
2. Identify the real traffic path, including gateway and application-server locations.
3. Calculate the propagation floor from slant ranges and the number of space legs.
4. Add documented equipment, scheduler, backhaul, and processing delays.
5. Validate with percentile measurements during representative load and weather conditions.
6. Recheck resilience: a lower-latency path may still fail the requirement if handovers, congestion, or gateway outages are not controlled.

GEO remains appropriate where fixed infrastructure, coverage, predictable geometry, or broadcast efficiency matters more than conversational response time. MEO offers an intermediate path length. LEO offers the lowest space-delay floor but requires a dynamic constellation and route. Multi-orbit designs can select different paths for interactive and delay-tolerant traffic.

Key Takeaways

• Satellite latency must be calculated from the complete path and reported as either one-way delay or RTT.
• A single GEO satellite hop contributes about 240–280 ms one way, producing roughly 480–560 ms propagation RTT before other delays.
• The idealized space-only RTT is 7.3 ms at 550 km, 106.7 ms at 8,000 km, and 477.5 ms at 35,786 km when vertical geometry and zero network overhead are assumed.
• Operational LEO and MEO latency is higher than the altitude-only floor and depends on gateway, routing, load, and handover behavior.
• ITU-T G.114's 150 ms figure describes generally transparent one-way interactivity; it is not a universal hard cutoff.
• Engineering acceptance should use declared endpoints, assumptions, median and tail latency, jitter, loss, and loaded tests.

Continue the Latency Investigation

• Reduce non-propagation delay — Diagnose queueing, scheduling, transport, and application effects after establishing the physics floor.
• Count satellite hops from the network topology — Distinguish star, mesh, hybrid, and remote-to-remote paths.
• Trace how satellite internet works — Follow the terminal, satellite, gateway, and terrestrial segments that contribute to measured latency.