Everything We Know About SpaceX’s Starlink Network

When it comes to SpaceX, or perhaps more accurately its somewhat eccentric founder and CEO Elon Musk, it can be difficult to separate fact from fiction. For as many incredible successes SpaceX has had, there’s an equal number of projects or ideas which get quietly delayed or shelved entirely once it becomes clear the technical challenges are greater than anticipated. There’s also Elon’s particular brand of humor to contend with; most people assumed his claim that the first Falcon Heavy payload would be his own personal Tesla Roadster was a joke until he Tweeted the first shots of it being installed inside the rocket’s fairing.

So a few years ago when Elon first mentioned Starlink, SpaceX’s plan for providing worldwide high-speed Internet access via a mega-constellation of as many as 12,000 individual satellites, it’s no surprise that many met the claims with a healthy dose of skepticism. The profitability of Starlink was intrinsically linked to SpaceX’s ability to substantially lower the cost of getting to orbit through reusable launch vehicles, a capability the company had yet to successfully demonstrate. It seemed like a classic cart before the horse scenario.

But today, not only has SpaceX begun regularly reusing the latest version of their Falcon 9 rocket, but Starlink satellites will soon be in orbit around the Earth. They’re early prototypes that aren’t as capable as the final production versions, and with only 60 of them on the first launch it’s still a far cry from thousands of satellites which would be required for the system to reach operational status, but there’s no question they’re real.

During a media call on May 15th, Elon Musk let slip more technical information about the Starlink satellites than we’ve ever had before, giving us the first solid details on the satellites themselves, what the company’s goals are, and even a rough idea when the network might become operational.

Launching the First Generation

Elon reiterated several times that these satellites will be the first of at least three generations of satellites which will eventually make up the Starlink network. They’re closer to the final satellites than the Tintin A and Tintin B technology demonstrators launched in 2018, but still lacking key features which will be necessary for optimum performance.

The first 1,600 Starlink satellites as reported to the FCC.

The biggest omission with these early satellites is the lack of inter-vehicle laser communication links, which means each satellite will have to relay everything through ground stations. In other words, if a Starlink satellite wants to send data to one of its peers, it will have to send it down to a ground station which then routes the information over the terrestrial Internet to another ground station that’s in range of the recipient. This not only increases latency, but requires a large number of ground stations located all over the globe.

To solve this problem, Elon says later versions of the Starlink satellites will use laser communications to form interconnected links, creating a mesh network in space. Data won’t always have to be sent to the Earth and back, and instead can be routed through the satellite network. Of course, ground stations will still be required to ultimately get data to and from the Internet itself, but far fewer of them will be needed and their geographic location will be less critical. This technology, which allows global communications with little to no ground infrastructure, could also be applied to satellite constellations orbiting the Moon or Mars; something SpaceX is almost certainly thinking about on the long-term.

In addition, Elon reiterated that these first 60 satellites don’t feature the “Design for Demise” optimizations which were implemented after the Federal Communications Commission’s expressed concerns over SpaceX’s inability to guarantee that debris from reentering Starlink satellites could be safely contained over the ocean. In their response to the FCC, SpaceX promised that future versions of the satellites would be designed in such a way that they would entirely burn up during reentry, removing any risk that falling debris endangering human life or property.

Experimental Technology

These first generation Starlink satellites might be missing some key features, but they aren’t exactly placeholders either. While we’ll have to wait awhile before we see laser communication or fully degradable construction, they definitely have some unique capabilities that will likely get the attention of other aerospace players if they prove successful.

According to Elon, these satellites are the first vehicles to ever use a krypton-powered ion drive in space. This type of propulsion was experimented with by NASA back in the early 1990’s, but never progressed into operational status as it was found to be less efficient than similar thrusters powered by xenon. That said, krypton is cheaper than xenon, and with low-cost satellites that are only expected to make occasional orbit adjustments during their relatively short lifespans, using a lower efficiency propellant actually made more economical sense.

On the subject of orbital adjustments, these satellites will also be testing an autonomous obstacle avoidance system which is sure to be of interest given the ever-increasing concerns over “space junk” in low Earth orbit. The satellites will receive orbital debris data from a NORAD database and use that information to decide on their own whether they should perform evasive maneuvers. Traditionally such decisions are made by ground controllers, but with tens of thousands of satellites in the final Starlink network, SpaceX reasoned this was a task that could benefit from automation.

Shipping and Handling

Learning more about the technical aspects of the satellites themselves is interesting, but probably the biggest question most people have about Starlink was just how SpaceX intends to hoist 12,000 satellites into orbit within the next couple of years. For reference, since the launch of Sputnik 1 in 1957, humanity has put fewer than 9,000 objects into orbit around the Earth. Viewed in that light, one could argue that a satellite constellation of Starlink’s proposed size would represent a new milestone in mankind’s utilization of space.

The reusability of the Falcon 9 was a huge piece of the puzzle, but it still wouldn’t be economically feasible unless SpaceX could maximize the number of satellites that could go up in each launch. To that end, they came up with a novel “flat-pack” design for the Starlink satellites. Inside the payload fairing, the satellites are stored in an arrangement not unlike a server rack, and when deployed they unfold their solar arrays and antennas into operational position. Elon admitted there’s a chance a few of them might bounce into each other on the way out, but said that it wasn’t expected to cause any serious issues given how low their relative speeds will be.

With 60 satellites weighing 227 kilograms each, plus the weight of ancillary hardware like the rack itself, this first Starlink launch tips the scales at 18.5 tons; the most mass a Falcon has ever put into orbit. Even still, Elon said it will cost SpaceX more to launch the Starlink satellites than it did to build them in the first place.

So How Do You Sign Up?

Unfortunately, we didn’t learn a whole lot about when and how consumers can actually sign up for Starlink Internet service. In the best case scenario, Elon estimated it would take another six launches before the network could even be activated, and twelve to provide enough coverage for it to be usable. After that, SpaceX will begin looking for commercial partners to actually start selling Internet service and distributing their phased-array terminals, likely with rural customers to begin with. At least for now it seems like SpaceX would rather partner with traditional ISPs than go to war with them, which will probably come as a disappointment to those who hoped Elon would shake up the telecommunications industry.


Satellite coverage renderings taken from Mark Handley’s excellent video about Starlink.

75 thoughts on “Everything We Know About SpaceX’s Starlink Network

          1. Right, the inter-satellite links are going to make huge improvement in coverage in those types of situations. Single-hop is still going to be better than existing options in the rural areas, but things are going to get really interesting once they can start moving data globally.

      1. Can a phased array really compensate for the rocking of the ocean? That seems… challenging at best, to find a sat in orbit while the waves are moving you around.

        1. Phased array can do it better than mechanical as they can move instantly. Most large ships currently have mechanical tracking satellite dishes that can stay locked on geostationary satellites in rolling seas. Their are even small units that can be installed on RV’s and buses that can maintain a link whenever they can see the sky. It’s pretty easy to do with today’s inexpensive IMU’s thanks to cell phones.

        2. A phased array would actually work better and faster because there is no moving parts. Simply having a calibrated gyroscope sensor to offset the roll and yaw the ship makes during transit will keep the array on target with the nearest satellite. A physical dish/patch antenna might pick up more signal but will add cost/weigh more for the gimbal system to hold the reciever and will need servicing as it’s cruising through a gaseous electrolyte. That’s sure to corrode all sorts of things. A phased array in a sealed radome could last for a quarter century, or at least till the caps ‘splode. Plus, phased arrays are just friggen cool. Beam forming is incredible stuff. Newer 5th gen aircraft are using phased arrays in the forward radome (nose) with hundreds of little antennas. It’s pretty expensive but the weight savings and pointing accuracy are incredible.

    1. Well we know it has to be vastly superior to current satellite internet technology, which is already competitive. The satellites are almost 80 times closer to earth and each one covers a much smaller group of consumers.

      1. How many simultaneous users at that speed per satellite? 200? 3Mil globally on 12K Sats? Would be more profitable to sell 3Mbs to 1Bil poor, remote users at $5 per month – $60Bil income a year.

      2. Who backed this project, the military. Who will get first dibs on this project, the military. Who can afford to use this new satellite link, private and military forces. All of you praising this new technology and your new overlords, you guys need a wake up see the bigger picture. Total world domination. These psychopaths don’t give a damn about the world population if anything they want to see more of it and disappear. Shenzhen China anyone?

    2. Nobody? It seems it’s mainly what they’ve been talking about. The whole reason for many satellites in low Earth orbit instead of few in geostationary orbits, is to reduce latency.

      Their low orbits will drastically reduce the distance the signal must travel, which is the latency Achilles’ Heel of existing satellite Internet systems which are bouncing signals up to satellites 42,000 miles away, instead of only 340 miles to the Low Earth Orbits Starlink will use.

    3. Isn’t speed and latency sort of a relative figure? Even if you have the best money can buy, wouldn’t your overall speed be dependent on the website your are accessing, and the number of other people, doing the same thing as you? Ideally, you want everything, and you want it now. But in reality, you need to sit there quietly, and wait your turn…

      1. Don’t forget multiple devices at an endpoint may seek to connect with multiple websites, such as one person watching Netflix and another Youtube. One website may not consume all the link’s bandwidth, but multiple may.

        As for latency I believe those numbers tend to be “best case scenario” numbers; the minimum time possible for a destination device to immediately respond to the source device’s request.

      2. The question is specific to the satellite link. Current satellites have a lot of limitations. They are geostationary so speed of light propagation alone adds a few hundred ms latency for the round trip. There are not many of them so the bandwidth is shared between all users of the service, which also means that they have a limit of bandwidth you can use each month. Looking at Hughesnet it looks like plans are 10-50gb with a reduced speed after. That’s a lot better than I remember hearing in the past. They also have 25mbps down and 3mbps up.

        I haven’t seen authoritative documentation on the uplink/downlink speeds for the SpaceX satellites, but they are much closer. It’s being said that latency should be similar to fibreoptic links, which is believable because the satellites are only going to be 300-500 miles up, instead of 22,000 miles like Hughesnet. And if there are going to be 12,000+ satellites, then you’re not going to be sharing the bandwidth with as many users. Everyone will be spread out over all those satellites.

        It’ll be interesting to see how good the worldwide coverage will be. It’d be neat to see if ships and airplanes will be able to benefit from global Internet access.

          1. Most of those adjustments are the result of latency and tweaks to slow start (which is also due to latency) so if Starlink can hit their claimed expectations then they should not be needed.

        1. The speed of light in an optical fiber us also about a third less than in a vacuum, so the satellite link may have less latency than long fiber optic cable links. That said as with Iridium the ground station finctionaliity and user cost is critcal.

      3. Latency is actually even more important for perceived speed of the connection.
        It’s nice that you can shovel data at,say, 100MB/s, but if you have to wait 100+ms for a response, most non-stream services will be really slow.

      4. In this context the meaningful measurements are referring to bandwidth and latency from the antenna endpoint to the ground station and back, or the nearest major colo if the ground stations are bottle necked on the terrestrial side. To be honest this number should be calculated as both the ideal best case and after accounting for expected local traffic and interference.

        This number will also be highly dependent on how well they can get their satellite to satellite links working since that allows them to distribute traffic across down-links and select ground stations nearer to traffic destinations further decreasing latency in some cases.

    4. The full constellation will have 240Tbps in bandwidth, and cost US$1Billion per year in capex to maintain. It is reasonable to guesstimate the same again for opex, and a total cost of approx US$10 per allocated mbps per year. Each individual satellite has about 20Gbps (for comparison, an average cell tower can deliver 1Gbps of aggregate bandwidth, and Australia, a country of 24 Million has 12 000 LTE cell towers).

      Experience akin to terrestrial broadband (100+GB per month to unlimited data) may be supplied to between 20 and 50 million subscribers OR a cellular like experience may service 100 million (3.5/4G) up to 250 million (3G) OR 1 Billion subscribers could access narrowband (2G) services. To try and give a sense of scale, there are at least 1.5 million people at sea and 11 million taking to the air each day. To supply broadband just the world’s aircraft would require 20% of starlink’s bandwidth.

        1. Just like your recipe has always been one and a half cups of flour, not one and a half cup. But on it’s own, less than one, half a cup is half a cup, point one of a decade isn’t plural, neither is one, add and exceed one, it’s 1.1 decades.

          1. I have never seen anyone refer to less than 20 years as decades or less than 20 thousand as tens of thousands. Also your one and a half cups is not quite the same as the exact amount is specified in the same sentence. Also, we don’t say one and a half dozens do we?

          2. Wait, we do say 1 and a half decades. But we don’t seem to say decades on its own when referring to times of less than 20 years. English is strange!

    1. ~12K is the target number of active satellites, but they’ll be constantly getting replaced so the actual number of sats built/managed will be much higher.

      Assuming it all goes to plan, anyway. Still years and years out, and that assumes everything works as expected.

  1. Just thought how they’ll be whizzing by all the time, it’s gotta work more like cell technology than you’d first think, but it’s the cell sites whizzing by, not the subscriber barrelling down the highway. In terms I mean of the sats handing the link off to next sat overhead….

  2. I bet one day this will be a free service… if the cost of providing service is lower than the cost of revenue generated by service, then someone can offer it for free – ahem, google – if you buy their devices and use their browser, their app store, see ads, listen to music, rent movies, stream content, etc., etc… and that cost today is around $26/mo to break even I believe…

  3. I understand there are at least 5 more parties who want to put thousands (or tens-of-t whatever) satellites into LEO. I can reasonably say there will be 50000 mini-sats and thousands of regular sats in LEO, by 2025. At this stage, every little miscalculation or even a natural phenomenon, can render the LEO unusable for thousands (tens-of-t) of years. We would be effectively Earth-locked until inventing a giant broom or a giant vacuum cleaner. I very much hope I’m wildly exaggerating ….

    1. At the orbital heights they are talking about any debris will re-enter and burn up in less than 10 years. There’s also a lot more room up there than you seem to think, and a few thousand mini-satellites aren’t going to take up noticeable amounts of that space.

    2. It’s important to understand how much smaller these are than what we’d traditionally think of as communications satellites, and the far lower orbit than normal.

      Also worth mentioning that the FCC had to sign off on the whole plan, and SpaceX had to address issues like this to their liking before they would grant them the license. A lot of thought has gone into this arrangement.

    3. Another tidbit is that the StarLink satellites have krypton Hall thrusters, star trackers and an autonomous avoidance system. They get NORAD TLE’s uploaded and maneuver accordingly.

    4. It’s easy math. Surface area 500km above the Earth is aprox. 593 266 411 747 461 square meters. So if one sat has about 1 square meter, than you can cover up the surface area 500km above the Earth with 593 266 411 747 461 satellites. There is planty of space.

  4. Correct me if I am wrong, but I thought I heard that these first 60 sats will be able to talk to each other, but using radio links. And that this simply means lower inter-sat bandwidth compared to the laser connects that will be used. Am I wrong?

  5. As a citizen of an government-restricted connectivity monopoly (Venezuela), I can’t wait to have an alternative like this one. I hope it is affordable enough …

  6. There’s something that i don’t understand, say i request access to some internet page, i will have to fetch it from some server on earth and it will be transmitted to me via the cables.
    Now, how does this work when we are using the satellites? do all these servers have to buy the devices that connect them to this wireless network?
    if that’s the case say i’m a big firm, like Facebook and i don’t wanna use this technology, how will the people access the data when using the satellite internet.

    1. The satellite network will have ground stations that connect to the internet backbones. For nodes just on the satellite network there would conceivably be no need for the backbone connections except maybe for DNS.

    2. You will use a small satellite modem connected to your computer or phone and talk directly to a satellite. It will forward your session to a large ground station (probably in a city) and connect to the Internet backbone there. From there to the web site, it is traditional Internet technology.

  7. A few observations from a non-affiliated techie:
    1) A traditional LEO bird orbits from 500 to 1400 KM altitude. The orbital period ranges from 90-100 minutes, with a fly-over dwell time of only 7-12 minutes over one spot on the ground. There are some Amateur radio satellites you can experiment with communications over the 2 meter band. However, conversations must be short, and the Doppler effect means you have to constantly tune the frequency and simultaneously point the antenna as the bird passes over head. It’s difficult to do manually, but would be impossible at the lower altitudes for this Starlink constellation. I’m impressed with the Starlink technology to pull that off.
    2) In the satellite world, bandwidth is directly related to power. These birds will be power limited, and there will be many hops of varying bandwidth along any path. Therefore, it’s impossible to predict what the end-to-end bandwidth will be. In fact, bandwidth and latency will probably vary widely and constantly during each session. Eventually, the infrastructure will need to support torrent-style (redundant packets), multi-path flows to get the required bandwidth. Single path flows will be too constrained. That will require packet reordering in the receiver. Amazing stuff.
    3) Inter-satellite laser links have been used in other constellations in the past. The trick here will be a custom MANET routing protocol, in order to point and re-point the lasers, that doesn’t overwhelm the bandwidth with routing overhead. That is also amazing technology.

    Bottom line: The deeper you get into this technology, the more difficult the problems are. Overall, I’m amazed that this team will be able to pull this off. Congrats to them!

    1. I enjoyed your comment. Do you think they’ll use a protocol like the relatively new Disruption Tolerant Networking (DTN) then? Single path flows will definitely not cut it.

  8. This seems like an admirable and ambitious goal, but I am deeply concerned about the consequences for ground-based astronomy. It’s a little sad that the man who dreams of colonizing Mars may end up ruining the night sky for those of us still on Earth.

    1. While I am greatly supportive of efforts to bring additional, competitively priced Internet Service Providers to market to challenge the many monopolistic ISPs, I believe it will be a while before Starlink is competitive at a large scale–mostly due to capacity, power, and RF interference/latency constraints. With all the hype over climate change in recent years, a web of RF energy orbiting the planet is likely not going to exist without unintended consequences. I hadn’t considerrd that all these low orbit satellites would impede an amateur astronomers ability to view space from their home telescope. It will also be interesting to see how Starlink performance pans out if the net neutrality regulations (Obama era) are reinstated in the future.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.