Starlink Innovation Part 2

The second key SpaceX innovation that makes Starlink possible is an affordable, reasonably sized phased array antenna. Let’s take a look at what that is and why it matters.

In order to make satellite internet competitive with terrestrial internet, SpaceX needs to fly their satellites at a very low orbit. Traditional internet satellites fly very high, at an altitude that allows the satellite to sit in the same spot in the sky relative to the ground. This is good for making the user’s antenna dish as simple a device as possible. Once you properly point the dish at the satellite, the dish just sits there and does its job. It’s bad, however, for a responsive internet because a satellite that sits in one place in the sky needs to be so far away that it takes ages (about half a second or more) to establish a connection between the satellite and the user’s antenna on the ground.

SpaceX flies its Starlink satellites in a very low orbit so the distance from the user’s antenna to the satellite isn’t so great and the connection latency is much improved (about 65 times better than traditional satellite internet.) But this comes with a challenge. At low altitudes, the satellites don’t sit in one place in the sky, they move across the sky, and rather quickly. That means your antenna needs to track the satellites as they move.

One way to do that would be to put some motors on the antenna so it could rotate to follow a satellite’s movement across the sky. But then you run into the problem of swinging the antenna back in the opposite direction to point at the next incoming satellite. You’d need to do that really fast to not have a gap in connectivity and that’s just not practical with motors. Perhaps you could have two antennas then, one tracking the satellite that’s leaving view and another tracking the satellite that’s coming into view. Now you’ve increased the cost and complexity of the system and doubled the number of failure points though — and made the user’s terminal large and cumbersome.

So, how does SpaceX solve this problem? With something called a phased array antenna. Phased array antennas are not new technology but they’ve traditionally been very large and very expensive. How they work is a bit complicated but a simplified description is something like this: you have many small antennas that act as one larger antenna. By powering those antennas individually and selectively you cause them to interfere with each other in such a way that the antenna’s collective beam is bent. This steerable beam can track movement and change direction and location almost instantaneously.

Big and expensive doesn’t sound like such a great solution though. And this is where SpaceX’s innovation comes in. Their hardware team miniaturized the technology, putting hundreds of small integrated circuit antennas onto a single computer board about the size of a typical internet satellite dish. And, they’ve managed to get the cost of building this phased array antenna down to something like $2500. Now, that’s still pretty expensive and so SpaceX is actually subsidizing most of that, charging users only $500 for the $2,500 antenna terminal. They hope future innovation plus mass production will get the cost down further but I suspect that it’ll be some time before they get costs down low enough to no longer need to subsidize. Instead, they’ll lose some money on the terminal and make it back up with the monthly internet service fee. If everyone that signs up for Starlink keeps the service for a couple of years, that will pay for the terminal and then start to be profitable for SpaceX.

If you’d like to read more about phased array antennas, and you should because they’re very cool and if you’ve read this far you’re probably into that kind of thing, start with the Wikipedia article at It’s a good primer and has some nice animations that help a lot to describe phased arrays.

Final Starlink Mount Plans?

Thank you all for your advice and direction. I have or have on order nearly all of the parts to make this build happen.

At the top you see the Starlink user terminal which has a short mast, approximately 16 inches. That mast has a round spring clip for locking into the tripod base that came with the Starlink kit. I’ll remove that clip leaving a half inch hole through the lower part of the terminal’s mast.

The mast slides perfectly inside of a Channel Master J pole that I’ll drill a half inch hole through near the top. Then I’ll place a bolt and locknut to secure the terminal to the J pole. (That bolt and nut are the only parts I’m missing. Will probably have to brave the hardware store.)

The J pole connects to a mounting bracket that’s 1/8″ steel. The connection point is made by two bolts, one through a curved raceway that lets you adjust the angle of the pole.

Up the pole are a pair of Skywalker heavy duty 1/8th inch steel braces that stand off the tree about 8 inches and attach to the J pole with 1/4″ U clamps.

The J pole bracket will be secured to the tree with six 12″ XERATH quarter inch structural screws. Each of the two heavy duty braces will be attached to the tree with two 12″ XERATHs.

I want to make the job as easy as possible for the tree climber while also ensuring it’s secure and won’t wobble in the wind. What do you all think? Does it seem secure? Easy enough to assemble and attach while strapped to a tree 100 feet off the ground?

starlink mount

Starlink Innovation Part 1

The first key SpaceX innovation that makes Starlink possible is rocket re-use. Let’s take a look at why.

To make satellite internet responsive, the satellites need to fly much lower than traditional telecommunications/internet satellites which suffer from high latency. But if you fly in low earth orbit, rather than way up at geostationary orbit, your satellite moves across the sky rather quickly, and in a matter of minutes is on the other side of the planet totally out of view of your antenna. So one satellite, or even a few, isn’t enough. Flying in low earth orbit means you need hundreds, even thousands of satellites to ensure one is always overhead of every customer.

A significant chunk of the cost of satellites is the rocket launch to place the satellite in Earth orbit. Commercial rocket launches historically have cost anywhere from 100 million dollars to several hundred million dollars. The high cost of rocket launches is mostly due to their disposable nature. Traditional rockets are one-use beasts that, after their jobs are done, crash into the ocean never to be seen again.

SpaceX has created a partially re-usable rocket in the Falcon 9. Falcon 9 has two main parts or stages. The bottom, and much larger (and more expensive) part is the booster stage. The booster is a giant propellant tank and a bunch of rocket engines, nine in the case of Falcon 9. The booster is used to get the vehicle through most of the atmosphere where drag is a major force working against it. Once its job is done, the booster separates and falls back to Earth which lightens the load for the upper part or second stage to finish the job by achieving enough speed to reach orbit. Falcon 9’s second stage is much smaller and has only one engine. Once the second stage achieves orbit, it drops off the payload and falls back to Earth.

With the Falcon 9, the expensive first stage doesn’t burn up or crash into the ocean. Instead, it relights its engines and performs a propulsive landing and than can be re-used. The second stage, which is much less expensive does crash into the ocean.

To explain why saving the booster stage for re-use is so important to Starlink, let’s look at how SpaceX uses a rocket. Estimates are that it costs SpaceX about 30 million dollars to build a Falcon 9 rocket. 2/3ds of that cost is the booster and 1/3rd is the second stage. For a Falcon 9’s maiden voyage, SpaceX sells a launch to a commercial customer for say 60 million dollars making 30 million dollars in profit. That’s where a typical space launch provider would call it a day and start building its next rocket but not SpaceX. SpaceX lands that booster, refurbishes it, adds a new 10 million dollar second stage and flies it again for either another commercial customer, at a much higher profit margin, or uses it to launch its own Starlink satellites. And then they land the booster again, and can re-use it again, and again. The leading booster in SpaceX’s fleet has been launched 8 times.

So, SpaceX’s cost to launch a batch of Starlink satellites is probably about 1/6th the cost of its retail commercial launches, and 1/3rd the wholesale cost of a new rocket. By lowering this cost so much, SpaceX is able to launch an unprecedented number of satellites for far less than anyone else has been able to do before it. My rough math says that SpaceX was able to launch the first 1,000 Starlink satellites for about the same cost as a single satellite launch from a traditional satellite internet company.

When your plan is to launch 14,000 satellites, though, you need to find even more savings. And SpaceX is working on that. Their next rocket, called Starship, is fully re-usable. Both the first and the second stage return safely to Earth to be rapidly re-used. SpaceX wants rocket flights to be more like airline flights, where you fly, land, re-fuel, and fly again. By making the whole rocket re-usable, and by making it significantly larger and more powerful than the Falcon 9, SpaceX will be able to put the rest of the Starlink constellation up for an unheard of low cost.

Starship is in prototype stages, having made several low altitude flights but SpaceX hopes to achieve orbit with Starship very soon. Oh, and Starship is also the rocket that’s going to be first to put humans on Mars.

Starlink Architecture

Starlink is an interesting architecture. There are three main components. The first is the user terminal. It’s a sophisticated antenna. The second is the satellite, relatively small, mass-produced telecommunications satellites in low earth orbit. The third is the gateway. Gateways are ground stations with 8 antennas each which are connected to the terrestrial internet backbone.

So, the user terminal is the thing that looks like a flat-faced dish and connects to your local network, probably via a wi-fi router (SpaceX provides a router as part of the “kit”.) This “dish” is actually a phased array antenna which is capable of digital steering so it doesn’t have to do physical steering to track the satellites as they move across the sky.

The satellites are very cool. They are designed to be flat-packed so that 60 of them can stack inside one rocket nose cone. They’re rectangular in shape, a slab 9 feet on one side and 4.5 feet on the other with a 30 foot solar array that deploys to give the satellite an L shape. There are currently about 1,000 of them orbiting Earth of a planned 14,000. They fly at super-low orbits and move across the sky in just a few minutes. Their orbits are spread out to ensure there is always a satellite overhead.

The gateways are strategically located to provide overlapping areas of coverage. Each gateway is a collection of 8 antennas that connect the satellites to the ground-based internet at major fiber optic connection points. There are about 40 of these gateways already providing nearly full coverage to the U.S.

So, as a customer, my computer makes a request for something on the internet. It sends that request across my home wi-fi network to the Starlink user terminal which beams it up to a Starlink satellite that’s moving across the sky at 17,000 mph. The satellite beams my signal down to a Starlink gateway where it connects to and races across the fiber optic backbone, eventually finding its way to the requested resource. Then everything happens in reverse as that resource is delivered back to my computer.

But this is just how it works today, with the first generation of Starlink satellite. The second generation of satellite, some of which are already launched, have laser interlinks. Instead of simply relaying the signal from my user terminal directly to a gateway where it travels to its destination on the terrestrial internet backbone, the satellites will talk to each other, becoming a part of the internet backbone itself, in space. This will mean far fewer, but larger gateways will probably become the norm as Starlink coverage expands.

(Perhaps not practically useful, but nonetheless interesting to me, for some connections, especially those that are from one side of the world to the other, the Starlink latency could be lower than the terrestrial fiber optic network’s. That’s because light travels faster in space than in glass.)


We live in a 100 year old log cabin in a beautiful redwood forest about 15 miles from the heart of Silicon Valley. There’s a lot to love about it, the trees, the creeks, the quiet, and more.

One thing that’s not so great is the internet connectivity. Cable doesn’t make it out here. The trees are too tall for traditional satellite. We’re too far from the nearest teleco office for DSL. The closest cell coverage is about 3 miles away. And forget about fiber ever making it here.

So, for the last 11.5 years, we’ve relied on a small wireless ISP for our internet connectivity. Even that was a bit of a chore, requiring us to mount the radio and antenna near the top of a 200 foot tall redwood. We still weren’t able to get line of sight and the signal moves through the tree tops so our speed is limited to about 1.5 Mbps.

Before the pandemic, this slow speed was tolerable because if I needed a fast connection, I could go to the office. But now, working from home full-time, with lots of videoconferencing, the slow speeds are starting to be a real problem. I have to ask Deanna to get off the computer when I take a meeting and that’s usually several hours a day 🙁

This is why Starlink is so exciting. Going from 1.5 Mbps to 150 Mbps is going to be a game changer. We’ll be able to stream television in HD! We’ll be able to videoconference and surf the web at the same time. I won’t have to drive up the mountain to a cell signal to do software updates for our computers and phones.

If you see me posting a lot about Starlink over the coming weeks and months, that’s why.


The Starlink antenna is still sitting on the ground so connectivity is very intermittent but when it is connected, I’m seeing some pretty encouraging speeds and latency.

Hopefully soon we’ll have it mounted up high enough to be usable.

I plan to maintain our old connection for a while and set up a dual WAN router with failover so that when the high-speed Starlink connection isn’t available, we’ll fall back to our slow connection.

In case you’re wondering why I’m posting this, it’s because Starlink is approximately 100 times faster than our current internet connection. That excites me!

speed test showing 160 Mbps

It’s Alive

Starlink is alive! I’ve temporarily set up the antenna on the ground to make sure everything works and this is the first sign of life. Can’t wait to get it mounted way up high where it has a full view of the sky. (For comparison, my connection for the last 11 years is about 1 Mb/s)

speed test results showing 91 Mbps

Starlink Summary

I got a couple of questions from friends about what exactly Starlink is. Here’s a summary. (I think I posted one of these before, but here it is again 🙂

Starlink is a new kind of high-speed satellite internet service from the rocket company SpaceX.

I’ll fist explain what’s wrong with traditional satellite internet and then I’ll share how SpaceX is solving those problems to provide a service that’s competitive with terrestrial internet.

Traditional satellite internet has a few problems that make it slow for the price. First, the satellites are super high up there. They sit about 22,000 miles up where their speed matches the Earth’s rotation. This is how they seem to always be in the same place in the sky and your dish can point to that one spot and not have to have a motor to track the satellite’s movement. But there’s a big downside to having a satellite so high up. They’re really far away. Even though the signal moves at the speed of light, light has a finite speed, and 22,000 miles means at least half a second round trip. This makes the internet feel “laggy”.

The second problem with traditional satellite is that the technology is often kind of old. Historically it cost a lot of money and took a lot of time to build these communication satellites. This means that the traditional providers don’t upgrade their satellites very often with the latest and greatest technology so even though technically they could be faster, they’re often relying on years-old tech that’s just not up to speed.

The third problem with traditional satellites is that the service providers over-subscribe. They simply have too many people connecting to a single satellite which has limited bandwidth and so all the customers suffer as a result. They can get away with this because there’s little to no competition for rural broadband. When it’s your only option, you’ll take it even though it’s not great.

Starlink tries to solve all three of these problems and it does so on the back of SpaceX’s very affordable rocket launch service. Where every other major rocket launch company throws their rockets away with each launch, SpaceX lands its boosters and re-flies them many times so their costs are ridiculously low (perhaps as low as 10-20% the cost of a new rocket.)

What can you do with super-low-cost rocket launches? Well, you can put a lot of modern satellites up quickly and affordably and replace them with new ones as soon as you’ve got something better. SpaceX has a massive assembly line that builds small, cheap, flat-packed satellites, and then they rely on those used rockets to launch the satellites in batches of 60 at a time. Starlink currently has about 1,000 satellites orbiting Earth and will have many, many thousands more in just a few years.

The Starlink satellites fly at very low altitudes, about 340 miles up. This has two implications. First, round trip time to the satellite is dramatically reduced compared to those flying at 22,000 miles. SpaceX plans on sub-20ms latency so your internet won’t feel laggy. (As we speak they’re directing the the whole constellation to lower itself a bit to further improve latency with a target of 16ms.)

The second implication of flying this low is that the satellites are moving really fast to maintain orbit. This means they don’t sit in one place in the sky but rather move across the sky pretty quickly . And that means you need a lot of them in order for one to be visible overhead at all times and you also need an antenna that can track their movement.

As I mentioned earlier, SpaceX already has over 1,000 orbiting satellites on their way to many thousands, even tens of thousands. But what to do about needing an antenna that steers to follow the moving satellite. SpaceX solved this problem not with mechanical steering but with something called a phased array antenna.

A phased array is a collection of many small antennas and when power is supplied to each one in just the right amount, their individual signals interfere with or reinforce each other in just such a way as to bend or steer the beam. This is digital steering, not mechanical steering so the user’s terminal can be a simple solid state device and doesn’t need a motor running all the time which would wear out and consume a lot of power.

OK, so that’s how Starlink addresses the traditional satellite problem of latency. What about the other two problems? The second problem with traditional satellite internet was stale technology. Because Starlink can launch satellites so cheaply, and because they build small, cheap satellites at scale, they can iterate their designs and put up improved versions of the satellites as soon as their engineers have made those improvements. So the constellation is always up to date with the latest and greatest technology while the earlier generations of satellites are simply de-orbited. (Note: they burn up fully on de-orbit.) A Starlink satellite is only expected to be useful for a few years. They’re almost disposable.

Then there’s the oversubscriber problem. This has more to do with business than with technology but both play a part. Traditional satellite internet services only have a few satellites to support all of their users and those satellites can be on 5 year old technology. SpaceX’s Starlink has thousands of satellites to support their users and those satellites are becoming more and more capable every year. So, as long as SpaceX is careful with not selling service to too many people in a geographic area (which is why they’re targeting rural users and not city dwellers right now) and they meet their target for how many satellites they want to have up there, then bandwidth should be good for all of their subscribers.

So, Starlink is a constellation of thousands of cheap, low-flying satellites and a technologically advanced user terminal/antenna that today can provide about 150 Mbit/s internet speeds with 20-30ms latency for $99/month and in a few years should be gigabit with even better latency.

Most of this is made possible by SpaceX’s innovation in rocket re-usability.

Right now, SpaceX is beta testing their constellation with over 10,000 customers. I’ve just joined that program. If you’re interested, you can try to get into the beta program or reserve your spot for when they become generally available at