If you are asking why is my satellite internet slow, the answer is usually contention, not weather. Your terminal shares capacity with other sites in the same beam, and at busy hours the pool runs dry. Next comes a fair access policy that throttled you after a data threshold. Weather and antenna pointing sit lower than most people assume.
Slow satellite internet is rarely a mystery once you know the order to check things in. If the link was fine last week and is unusable now, something changed on the ground or in the shared pool. What follows is our satellite internet troubleshooting order on a support call: the symptom, a five minute confirmation, then the fix.
Latency and bandwidth are two different complaints
Bandwidth is how much data you can move per second. Latency is how long one packet takes to get there and back. On a geostationary link the round trip runs roughly 560 to 650 ms, because the satellite sits about 35,786 km up and your signal covers that distance four times over a full round trip. No amount of extra megabits changes that figure. On LEO networks the round trip is typically well under 100 ms, though it varies by constellation and by gateway routing, and MEO usually lands between 120 and 180 ms.
The two need different fixes, so ask which one before anyone touches a setting. Pages that hesitate before loading, or a voice call with a beat of delay, point at latency, which on a geostationary link is mostly physics. A file transfer that crawls points at throughput, and throughput is usually fixable. It helps to know which beam covers your site: our satellite coverage maps show which regions fall under which footprint.
Ten causes of a slow link, ranked by how often they are the real one
1. Contention and oversubscription
Symptom. Fast at 04:00, unusable between 20:00 and 23:00. The morning speed test matches your plan and the evening one gives you a fifth of it.
Confirm it. Run three tests, one at a quiet hour and two during your local peak. Then put two questions to your provider in writing: what is the contention ratio on this plan, and what committed rate am I guaranteed? A service sold as 20 Mbps at 1:20 contention means 20 is a ceiling, never a floor. As a rough guide, an office of light users lives happily at 1:20, while a site running video calls and remote monitoring down one pipe wants something nearer 1:5 or 1:10.
Fix it. Certainty costs money, and a committed information rate layered on the shared pool buys it. For a camp, rig or vessel that cannot tolerate an evening collapse, dedicated SCPC bandwidth costs more per megabit and holds the same number at every hour.
2. A fair access policy or data cap that kicked in
Symptom. Normal for the first two or three weeks of the billing month, then a cliff. Slow evenly all day, including at 04:00.
Confirm it. Compare usage against your allowance. A throttle looks different from congestion on a graph: flat and constant, and it does not lift overnight.
Fix it. Either raise the allowance or find what is eating it, and the culprit list is short. Operating system updates and cloud photo sync on crew phones account for a great deal of it, along with a CCTV recorder pushing footage offsite and the one laptop quietly backing up 40 GB a night. Traffic shaping at the edge stops the bleeding, which is what our SD WAN service does on customer sites.
3. Local network faults wearing a satellite costume
Symptom. One user complains loudly while everyone else is fine, or the accommodation block is slow and the office untroubled.
Confirm it. Plug a laptop into a LAN port on the router and test. A healthy wired result clears the satellite link. Then scan the 2.4 GHz band and count how many access points are fighting over one channel.
Fix it. Push clients onto 5 GHz and split the crew SSID from the operational one. Add an access point where coverage is thin; winding the transmit power up rarely helps. Then check whether your quality of service rule was ever applied. Broken QoS turns up constantly: the rule exists, it reads correctly, and a firmware update unbound it from the interface. Sensible Peplink router configuration takes an afternoon and saves months of arguments.
4. DNS
Everything feels sluggish to start, then downloads at full speed once going, and the speed test looks perfect so nobody believes the users. Time a name lookup, then ping the address directly: on a 600 ms link every uncached lookup costs a full round trip, and a page pulling assets from thirty domains pays that toll thirty times. Run a caching resolver on site and point every client at it. Then hunt down any lookups still going to a resolver on another continent.
5. TCP behaviour over a long latency link
Symptom. One large file transfers at a little under 1 Mbps on a link that speed tests at 20 Mbps. Start eight transfers and they all run, filling the pipe.
Confirm it. Test one stream, then test with parallel streams. Throughput on a single connection is roughly the TCP window divided by the round trip time, so a 64 KB window over 600 ms works out at about 870 kbps.
Fix it. Check that acceleration is enabled at the hub and in the modem, keep window scaling on, and use tools that open several connections. One warning: a full tunnel VPN encrypts the headers the accelerator needs to read, and on customer links we have seen that cut usable throughput substantially, sometimes by half or more. Split the traffic so only what needs the tunnel goes through it.
6. Rain fade
Degradation or loss of lock during a heavy squall, with recovery minutes after the cell passes, is the signature. Watch Es/No or SNR on the modem while the rain falls: a drop of several dB as the cell arrives, followed by a climb back afterwards, answers it. The remedies are more link margin, adaptive coding and modulation, uplink power control, or a different frequency band, and the band choice is a tradeoff between cost per bit and availability.
7. Pointing and skew drift
Symptom. On a fixed site, a slow decline over weeks or months. On a vessel, a step change after a rough passage or a yard period.
Confirm it. Compare today's SNR and transmit power against the commissioning record. If you do not have one, that is the first problem to fix.
Fix it. Repeak the antenna, check the mount bolts and the grout under the pedestal, and verify polarisation skew against the current figure for the satellite you are on. Stabilised marine antennas need gyro input and tracking calibration checked too, routine work on Intellian stabilised systems and across maritime VSAT fleets.
8. A blocked or wet radome
Fine in the dry, poor after the rain has stopped, and on a ship poor on one heading only. Go outside and look at the thing: salt crust, bird mess, standing water in a seam, a delaminated panel, a new container in the line of sight, a crane boom parked across the arc. On flat panel and LEO terminals, including Starlink and Peplink deployments, open the obstruction view in the management app before you climb anything. Cleaning, drying, resealing a seam or moving the terminal are cheap jobs, provided somebody on site has been made responsible for the antenna.
9. Cable and connector loss
Intermittent trouble, often worse in the afternoon heat, sometimes with transmit errors and no obvious receive problem. Measure DC voltage at the antenna end of the run, open the connectors and look for green corrosion or trapped water, and check the coax type against the length installed. Then replace the connectors and weatherproof them to spec, and shorten the run or move to lower loss coax if someone pulled 90 m of thin cable through.
10. A satellite low on the horizon
Consistently weaker signal than a comparable site further south, sharper rain sensitivity, occasional loss of lock at high latitude. Look up your elevation angle: below roughly 15 to 20 degrees your signal pushes through far more atmosphere and picks up more ground noise. Move to a satellite with a better look angle, or stop fighting the geometry and add a LEO or MEO service, which does not depend on a single fixed look angle.
How rain fade differs between Ka band, Ku band and C band
| Band | Typical frequency | Rain behaviour | Typical planning availability, before extra margin | Where it earns its place |
|---|---|---|---|---|
| C band | 4 GHz down, 6 GHz up | Barely affected, often under 1 dB in a real storm | 99.9% and better | Monsoon and equatorial regions, broadcast, anything that must not blink |
| Ku band | 11 to 12 GHz down, 14 GHz up | A few dB in heavy rain, seconds to minutes of degradation | 99.5% to 99.7% | Enterprise and maritime across Europe and the Middle East |
| Ka band | 19 to 20 GHz down, 29 to 30 GHz up | Strong, 10 dB or more under a tropical downpour | 99.0% to 99.5% | High throughput where cost per bit matters and short outages are tolerable |
Those availability figures are planning numbers used in link budgets, and no operator contracts to them. The real answer for your site depends on its rain zone and its elevation angle. It also depends on how much margin the operator designed into the beam, and that figure is rarely published. Our page on Ka band capacity sets out where it stops making sense. Plenty of sites keep Ku or C band as the floor and add a second path for bulk traffic, which our page on hybrid links that combine two paths covers.
Symptom to likely cause
The table below compresses all ten causes into the shape you will recognise in the field.
| What you see | Most likely cause | Confirm in five minutes |
|---|---|---|
| Fast overnight, crawling by mid evening | Contention in the shared pool | Speed tests at a quiet hour and at peak |
| Fine for three weeks, then slow all day | Fair access policy throttle | Month to date usage against your allowance |
| One device slow while everyone else is fine | Wifi channel clash, or that device syncing | Wired laptop test at the router |
| Pages slow to start, downloads fast | Uncached DNS lookups over a long link | Time a lookup, then ping the address |
| One big file slow, many files fast | TCP window against high latency | Single stream versus parallel streams |
| Slow only while a squall passes over | Rain fade | Watch Es/No on the modem |
| Steadily worse over months | Pointing or skew drift | Compare with the commissioning figures |
| Worse after the rain has stopped | Wet or dirty radome | Inspect and dry it |
| Intermittent in afternoon heat | Cable or connector fault | Check connectors and DC at the feed |
| Weaker than a comparable site further south | Low elevation angle | Look up the look angle for your position |
A diagnostic checklist you can run in fifteen minutes
- Ask the person complaining what slow means to them, and which application. Write the answer down.
- Note the time of day it appears and whether it arrived suddenly or crept in.
- Test from a laptop wired to the router. In our support experience this single step clears a large share of reported satellite faults.
- Record modem statistics: SNR or Es/No, transmit power, error counters and current modulation. Compare against the commissioning sheet.
- Run a single stream transfer, then a parallel stream test, and note the difference.
- Time a DNS lookup for a domain nobody on site has visited today.
- Check month to date usage against the allowance in your contract.
- Watch the top talkers on the router for ten minutes and see which device owns the bandwidth.
- Walk outside and look at the antenna, the radome, the cable entry and the horizon.
- Only now call your provider, with everything above attached.
When the link really is undersized
Sometimes the service was undersized for what the site became: a drilling camp that grew from 20 people to 90, or a vessel that added remote engine monitoring and crew welfare to the same 10 Mbps pipe. A clinic that starts uploading imaging does the same thing to a link almost overnight. Tuning will not rescue any of them. The honest advice is to resize the service, or to put a committed rate underneath the shared pool and let policy decide what runs over which path. Our professional services team handles that as survey and design work. Our 24/7 NOC watches for the drift cases above and often flags them before the site raises a ticket.
What to do next
Run the checklist first. A quarter of an hour will usually name the culprit. If the wired test is slow, or the modem statistics look wrong against commissioning, send the numbers through our contact page and we will tell you which of the ten causes it looks like and what fixing it is likely to cost.
