Starlink Speed Test

Measure what a satellite connection actually does: speed, latency, the latency you get while the link is busy, and how steady it stays across several satellite handovers.

Latency: why 25-60 ms is the floor, not a fault

A Starlink satellite orbits about 550 km up. Every packet climbs to it, crosses to a ground station, travels the ordinary internet to wherever it is going, and comes all the way back. That geometry is why SpaceX quotes 25-60 ms rather than the 5-15 ms a fibre line manages, and no amount of router tuning will beat it.

What you can act on is the spread. A dish with a clear view holds a steady figure; one looking through a branch shows the same median with spikes and gaps around it. That is what the stability scan on this page is looking for.

Why your upload is a fraction of your download

Starlink publishes 5-20 Mbps upstream against 25-100 Mbps downstream, and the ratio is deliberate: the dish transmits with far less power than the satellite does, and far more of the shared spectrum is pointed downwards because that is what people use. An upload result inside that band is the system working as designed, not a fault to report.

Latency under load is the number that decides your video calls

Run a big download while someone is on a call and the call breaks up, even though the speed test just said the connection was fast. That is bufferbloat: a queue builds in front of the saturated link and every packet waits behind it.

This page measures it by probing latency during the download and upload phases rather than before them. A link that adds tens of milliseconds under load stays usable while it is busy. One that adds hundreds does not, and the fix is queue management on your own router rather than more speed.

The 15-second rhythm

Starlink reassigns dishes to satellites on a 15-second cycle aligned to UTC, so a brief rise in latency at :00, :15, :30 and :45 past the minute is the constellation handing you over, not a problem with your setup. The stability scan spans several whole cycles and reports how many of the spikes it saw landed on those boundaries - spikes that ignore the cycle are usually something closer to home.

Before you blame the dish

Test near the router or over Ethernet. Wi-Fi across a house drops more speed than the satellite link usually does, and a speed test run three rooms away measures the Wi-Fi.

Check the obstruction map in the Starlink app. Even a couple of seconds of blocked sky per hour shows up as dropped connections rather than as a lower speed.

Try again at a different hour. Cells get busy in the evening, and the same dish can measure very differently at 21:00 and at 07:00.

Look outside. Heavy rain and wet snow absorb the signal, and snow sitting on the dish does it more thoroughly than weather ever will.

SpaceX publishes 25-100 Mbps download, 5-20 Mbps upload and 25-60 ms of latency as typical for Starlink Residential. A result inside those bands is a healthy dish. Well below them is worth investigating: check the obstruction map, test again at a quieter hour, and make sure you are measuring the link rather than your Wi-Fi.

Because capacity over your area is shared. Every dish talking to the same satellite beam divides the same spectrum, so the busiest hours are the slowest ones. Running this test at two different times of day tells you whether you are seeing congestion or a permanent problem.

A satellite handover. Starlink reassigns dishes on a 15-second cycle aligned to UTC, and the switch costs a moment. The stability scan checks whether the spikes it measured fall within 2 seconds of those boundaries, which is how you tell a handover apart from an obstruction.

Yes, and the difference matters. The app measures to Starlink’s own ground station, which is the best case for your dish and stops short of the internet. This page measures the whole path from your browser to a public test server out on the internet, which is the path your video calls and downloads actually take.

The measurement works on any connection. The comparison bands on the results are the published Residential figures, so a Priority or Maritime line may sit above them and a congested Roam session below them. The raw numbers are still yours; only the band you are compared against assumes Residential.

It asks the internet’s phone book. Starlink gives its customer addresses a reverse DNS name under starlinkisp.net, so one reverse lookup of the address your request arrives from recognises the network and, usually, the ground station you are routed through. There is no IP-intelligence service involved, nothing is stored, and nothing on the page depends on the answer.

Because three of the four things it measures need time. Speed can be measured in seconds, but stability cannot: the stability scan runs for 21 seconds so that it covers several whole 15-second satellite handover cycles. Untick it and the test finishes in about 21 seconds with the speed and bufferbloat results intact.

Comments

Loading comments…

Be civil and stay on topic. Comments are public, so do not post passwords, network keys, addresses or anything else private.

  • Loading comments…