Starlink’s high-speed satellite Internet: What’s the catch?

By on 22 Apr 2021

Category: Tech matters

Tags: , ,


Blog home

Image of Starlink satellites taken in April 2020, by Jud McCranie on Wikimedia Commons.

You may have heard about Starlink from SpaceX. You may even have been lucky enough to be outside at the right time to see one of these ‘string of pearls’ events.

What you’re seeing below is not the final mesh of satellites, its what you see after a rocket ejects a cluster of satellites as they are on the way to their low earth orbit (LEO).

It’s causing concern among the sky-watching community because of the high impact it has on the visibility of stars in the night sky, astrophotography and the pleasure of watching the heavens above.

But I’m not writing today about the concerns of star-watchers. Instead, I’d like to discuss some potential challenges ahead when LEO satellite Internet services (such as Starlink’s) are rolled out more broadly.

Yes, it’s fast, and it’s potentially a great solution for remote areas

This isn’t some future delivery, this is a viable, fast, and most importantly, low-round-trip-time (RTT) method of doing high-bandwidth Internet from almost anywhere.  If you are somebody who lives outside a core city high-speed broadband footprint, this is good news. 

When satellite broadband was discussed in the past, it was not an attractive choice. It was sometimes the least-worst option for people who lived beyond the footprint of a wired (or fibred) or wireless land-radio link.

The reason was the delay — traditional satellite services came at a huge delay (half a second, or 500ms RTT), which caused all kinds of problems, especially when using voice transmission and interactive networking. LEO satellite Internet providers like Starlink get around delay issues thanks to their low orbit.

Why is Starlink faster than traditional satellite broadband?

Starlink is different to the typical satellite broadband options of the past. It’s LEO. The distance has been reduced so much from Geosynchronous Earth Orbit (GEO), which is 35,000km out in space, to around 550km. The effect on delay is transformative because this is all about the speed of light — you can’t defy the laws of physics about the speed of light in space.

GEO ends up taking 500ms, while LEO ends up with around 50ms.

If you’re trying to do the maths here and compare the distances with the speed of light, bear in mind that it won’t add up perfectly.

The 500ms is overwhelmingly the RTT to space. The 50ms, on the other hand, includes other processes such as encoding overheads, multiple hops, and various things that are included in the bigger figure for GEO.

The bandwidth provided by these LEO satellites is really very good… for now. Starlink, unfortunately, runs the risk of being a victim of its own success.

The catch: The more people using it in an area, the slower it gets

Surprisingly, in both Australia and New Zealand, people on the fringes of cities are hooking up too. It’s not just being used for remote areas.

Starlink is now being seen as a way to get high-speed Internet when the normal terrestrial high-speed Internet is within view, but alas, not quite at your door.

What’s the problem? If it’s high-speed Internet, why should just out-of-the-way places get it? People on the fringes of cities who don’t have good fibre options would understandably want to access to high-speed satellite Internet.

The catch here is that the bandwidth is at its best while you are the only person in the footprint using it at the time. It’s a classic ‘early adopter’ situation. The more people who take up service in your location, and take a beam from the LEO system at the same time as you, in the same place as you, the more it has to be chopped up into smaller units to share out. This is called ‘multiplexing’ and the only fair outcome in the end is for roughly equal shares. So, if you get 100mbit throughput now, by the time five of your neighbours come on stream, it probably won’t be 100mbit any more. It might still peak at 100, but it may also only be 25. Fair’s fair.

This means the suburban fringe of cities is where the nuances of signing up to Starlink need to be considered. If you live just outside the city, on the fringe of the broadband deployment area, perhaps in the suburban outskirts of Bangkok, or Sydney, or Auckland, or maybe on a hobby farm in the countryside in America (but not too far from the nice deli and the supermarkets) you are going to be closer to your neighbours than the people who live in really remote areas like farms and villages in isolated communities. It’s these isolated areas that get the biggest advantage from Starlink.

If this was about intermittent use by mobile devices, this might not matter.

But if this ends up being about a semi-permanent hookup to the Starlink network in urban fringes, then two things are going to happen:

Firstly, you won’t get the bandwidth you hoped for initially, because you’re too close to your neighbours.

But this also flows on to a second, very serious thing to consider. What happens if the people who make decisions on building infrastructure decide “well, there’s Starlink, so do we need to build broadband to these fringe areas?”.

In that scenario, the capital investment needed to bring terrestrial broadband to your door may not arrive because it may be assumed you have good enough bandwidth from the LEO satellite. In which case, lacking broadband, more people in this fringe area would sign up to get Starlink, thus making it slower.

It’s a catch-22. The better that LEO satellite Internet is, the more this risk could grow. 

Rate this article

The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.


  1. Nick F

    Guess George is forgetting that starlink is planning to offer a gigabit speed, not 100. There’s been already proof that even higher specs exist from the military aircraft testing, so no, starlink is just fine with capacity.

  2. Mark Smith

    You’ve completely missed the point of the article.

    A transponder on a satellite is a shared resource, and the more people who share it, the less they each get.

    This issue happens on terrestrial networks all the time. 5G, HFC, and all links that are carrying aggregated traffic.

    In terrestrial networks when there isn’t enough capacity to share such that everybody gets enough, we add capacity – fatter or just more links.

    Can the do the same with LEO networks?

  3. ZZ

    what you see after a rocket ejects a cluster of satellites

    This video is two years old and the phenomenon depicted can no longer be seen due to mitigation efforts.

    The catch: The more people using it in an area, the slower it gets

    Uh, yeah… all transport mechanisms work that way. How is this news?

    What happens if the people who make decisions on building infrastructure

    The consumer ultimately makes those decisions, even if just indirectly. When adequate service is not provided, regional co-ops will take up the slack.*

    decide “well, there’s Starlink, so do we need to build broadband to these fringe areas?”.

    As time progresses, those areas will be covered by a combination of fiber and municipal wireless networks.* Also, Starlink can tell where you live, what cell you are in, and what sort of competition exists at that location. It can therefore tailor pricing and bandwidth according what the market will bear, and will not compete directly against terrestrial broadband providers, according to Musk. I don’t see how anyone could sincerely construe this as a losing value proposition for Starlink or consumers. I wonder if you are another Musk hater, or just weak on economic theory.

    @mark: You’ve completely missed the point of the article.

    It’s the article which misses the point.

    @mark: When there isn’t enough capacity, we add capacity – Can they do the same with LEO networks?

    Of course. Just add more satellites and lasers. (Plus competitors will exist on different orbital planes)


  4. David Dawson

    Hi, APNIC comms here. I had a chat with George on this. He clarified that the simple arithmetic of “five people divide a link by five” is not applicable.
    The model comms engineers use is more like “we can put ten people on a link where the sums might say three, or six”. They have a statistical risk-management approach.

    So essentially, it’s rarely exact and they can often squeeze a bit more out of it than the basic mathematics would indicate. Hope this clarifies!

  5. Robert

    How or who do I contact if my country is not listed even though many of the nearby islands are listed.

    I am speaking specifically about Sint Maarten


Leave a Reply

Your email address will not be published. Required fields are marked *