Procurement conversations about satellite connectivity spend most of their time in space. Orbital altitude, constellation size, spectrum band, beam architecture. All of it matters, and all of it is genuinely interesting to discuss.
Then the service goes live and the problems that show up are almost never orbital. The application feels slower than the latency figure suggested. Throughput collapses during a specific two-hour window. A failover works in testing and behaves strangely in production. In most cases the cause sits somewhere in the ground segment, which is the part of the system nobody asked about during evaluation.
What the ground segment actually includes
The term covers more than an antenna. A single user session touches a chain of components, and each one imposes its own limits.
At the site there is the terminal itself: the antenna and its pointing or beam-steering mechanism, the block upconverter and downconverter, and the modem. On the network side there is the teleport or gateway with its own antennas and RF chain, the hub equipment that terminates the waveform, the network functions that handle acceleration and traffic shaping, and then terrestrial transport carrying traffic to a point of presence where it meets the public internet or a private connection into a customer network.
Every one of those elements can be the constraint. In practice the constraint is rarely the space segment, because capacity in orbit is the part everyone measures and provisions carefully.
Where the ground segment costs you performance
Point of presence geography. This is the most common and the least discussed. A service can deliver an excellent satellite hop and still feel slow because the teleport lands traffic in one region while the application the user needs sits in another. Terrestrial mileage adds latency that has nothing to do with orbits, and it compounds when traffic backhauls to a central breakout before reaching a nearby cloud region. Operators running workloads in a specific cloud region should be asking where their traffic physically egresses, and whether a direct interconnect at that location is available.
Gateway diversity and weather. Ka-band gateways experience rain fade, and the standard mitigation is site diversity, meaning a second geographically separated gateway that can take over when conditions degrade at the primary. Whether a provider operates gateway diversity in your region is a concrete question with a concrete answer, and it correlates strongly with availability figures during storm seasons. Adaptive coding and modulation handles moderate degradation gracefully by trading throughput for link robustness, which means a service can remain technically available while delivering a fraction of the contracted rate. Availability and performance are separate commitments.
Hub contention. Capacity is shared somewhere. The location and depth of that sharing is a design decision by the provider, and it produces the recurring pattern of a service that performs well at three in the morning and poorly at seven in the evening. A committed information rate is meaningful. A stated peak rate under a contention ratio the provider will not disclose is considerably less so.
Modem and waveform generation. Terminals age. A site running an older modem may be locked out of efficiency improvements available in newer waveform standards, and the extended DVB-S2X specification in particular introduced modulation and coding schemes that meaningfully improve spectral efficiency over the earlier standard. Replacing modems across a large fleet is a capital exercise that tends to get deferred, and the result is a network where the space segment has been upgraded and the terminals have not.
Virtualisation is changing the economics
The ground segment has spent years catching up to the software-defined transition that the rest of networking went through earlier, and the shift is now well underway.
The technical enabler is digitisation of the intermediate frequency signal. Traditionally the RF chain and the modem were connected by analogue cabling, which physically tied a given modem to a given antenna at a given site. Digitising the IF signal and carrying it over standard IP networking breaks that coupling. The DIFI Consortium, formed in 2021 to standardise this interface on top of the VITA 49 packet format, exists specifically to make equipment from different vendors interoperate across that boundary.
Once the signal is packets on a network, several things become possible. Modems become software running on commodity servers, which can be scaled and updated like any other workload. An antenna in one location can be paired with baseband processing in another. Capacity can be allocated dynamically across sites rather than provisioned as fixed hardware. Ground station as a service offerings, where an operator rents antenna time on shared infrastructure instead of building teleports, depend on exactly this decoupling.
The limits are worth stating plainly. Virtualised baseband has demanding requirements on the underlying network, since digitised IF traffic is high-rate and intolerant of jitter. Timing and synchronisation are harder in software than in purpose-built hardware. And the interoperability the standard promises is still uneven in the field, so a multi-vendor ground segment requires more integration work than the architecture diagrams suggest.
The terminal side is moving too
Electronically steered flat panel antennas have moved into commercial deployment across mobility applications, where a mechanically steered dish is impractical for reasons of profile, weight, or vibration. Their tracking performance and their ability to hold multiple simultaneous beams determine how cleanly a terminal can hand over between satellites, which in turn determines whether a multi-orbit service feels seamless or produces a visible gap during transition.
Cost per unit remains the constraint on wider adoption, and the gap against a mechanical antenna is still substantial for fixed installations where nothing is moving. Power consumption is a second consideration, particularly for installations with limited generating capacity. For a fixed remote site, a well-installed mechanical antenna often remains the more sensible choice, and the flat panel argument is strongest where motion is involved.
Questions to put to a provider
The following are answerable, and the quality of the answer tells you a great deal about the infrastructure behind the service.
- Where does my traffic egress geographically, and can I get a direct interconnect into my cloud region rather than backhauling through a central breakout?
- Do you operate gateway diversity covering my area of operation, and what was the measured availability at that gateway over the past twelve months?
- What contention ratio applies to my service tier, and what is the committed rate as distinct from the peak rate?
- Which modem generation and waveform standard will my terminal use, and what is the upgrade path?
- What does failover actually do to an established session, and how long does re-establishment take under normal conditions?
- Is any part of the baseband virtualised, and if so, what happens to my service during a software update?
None of these questions are exotic, and a provider with well-run ground infrastructure will answer them without hesitation. The evaluations that go wrong are usually the ones that stopped at the orbital diagram.
