Starlink is a strong uplink with one bad habit
It works where nothing else does. It also drops for a second at a time, and a stream notices that more than it notices slow.
A low-earth-orbit dish gets you a usable uplink from a field. That was not true a few years ago.
The catch is the shape of the failure, not the size of the pipe.
Why it drops
The satellites move. Your dish hands off from one to the next on a regular cycle, and each handover is a chance for a gap.
Add anything in the sky path. A branch, a mast, a roofline. The dish does not need a permanent obstruction to hurt you. It needs one that a satellite passes behind.
The result is not slow service. It is service that stops for a second and comes back.
An encoder handles a lower ceiling. Drop the bitrate and carry on. A one-second outage with no warning empties the buffer, and the viewer sees it every time.
What to check before you rely on it
- Obstruction data. The app maps what the dish can see and logs how much time it lost. Read it over an hour, not a minute.
- Upload, not download. The download figure is the one in the marketing. Upload is a fraction of it and it varies more.
- Your service tier. Fixed-location plans are not the same as roaming plans, and using a dish in motion needs both the right plan and the right hardware.
- Power. A dish draws more than a modem, and it draws more when it is cold or repositioning.
How to use it without getting caught
Treat it as one link in a bonded uplink, never as the only link. A cellular modem alongside it covers the handover gaps. Neither would be enough alone.
Set the protocol buffer higher than you would on cable. Two seconds or more. You are buying time to ride out a gap you know is coming.
Then run it for a full session before you trust it in one. The obstruction that ruins hour three is invisible in minute one, because the satellite that goes behind that tree had not come round yet.
Sources
The Rig Wire reports on streaming technology from public documentation, source repositories and vendor releases. We did not bench-test what is described here. When we have tested something, we say so.