As long-time readers (with good memories) already know, network neutrality is a long-time cause that I’ve regularly explored editorially. This post will describe SpaceX’s burgeoning Starlink satellite internet service, specifically focusing on whether it can deliver the necessary competition to enhance net neutrality’s fortunes.
My interest in net neutrality isn’t solely conceptual; where I live (a few minutes from downtown Golden, CO, albeit several thousand feet in elevation above it), Comcast is my only broadband internet option, unless you count geriatric copper-delivered DSL, which I don’t. I’ve generally been pleased with the “Xfinity” service over the past 8+ years that I’ve used it, but at the end of the day, I realize that the absence of credible competition (along with a dearth of government oversight, although hope springs eternal) means that I’m probably paying more per month than I otherwise would.
And given that Comcast is the owner of a conglomerate of companies that also includes plenty of content providers, I wonder how much favoritism (fiscal, speed, response time, packet priority, etc.) that particular content gets, and what other competitive content might be getting downgraded (and downright blocked, if not for what little government regulation does exist) in exchange.
All of this, along with my generally-underwhelming past attempts to get decent internet access in even more rustic climes, has left me admittedly intrigued with the Starlink service that one of Elon Musk’s companies, SpaceX, is now starting to roll out. Understanding how Starlink differs from traditional satellite internet service requires a comprehension of the various orbits commonly used by satellites.
Traditional satellite internet service offered by companies like EchoStar (HughesNet) and Viasat (Exede) harnesses equipment “parked” in geosynchronous high Earth orbit (HEO), approximately 36,000 km (26,000 miles) above the Earth’s surface (said another way, exactly 42,164 km from the center of the Earth). At this “sweet spot,” their orbit matches the Earth’s rotation; from an Earth-based observer’s perspective, they seemingly stay in one place. These particular satellites are located directly over the Earth’s equator, therefore in a geosynchronous “special case” called a geostationary orbit. This means that they don’t vary north-to-south over time, and it also means that they can service an entire pole-to-pole latitude swath of the Earth.
A quick aside: another common satellite orbit region is known as medium Earth orbit (MEO). When placed in orbit at 26,560 km from the center of the Earth (about 20,200 km above the surface), for example, a satellite will cross over the same location on the Earth every 12 hours. GPS satellites are one example of the gear that’s placed here.
Then we get to low Earth orbit (LEO), where the bulk of modern satellite internet activity is focused (along with, for example, the International Space Station and Hubble Space Telescope). Starlink’s constellation of satellites, for example, locates in clusters and “shells” ranging from 540 km (340 mi) to 570 km (350 mi) in altitude (height above sea level). Why a constellation? That’s fundamentally because given their low comparative orbit versus previously discussed alternatives, each satellite will pass overhead multiple times per day (by definition, LEO objects have an orbital period of 128 minutes or less, therefore making at least 11.25 orbits per day).
Any particular satellite will pass into (and out of) a terrestrial antenna’s broadcast-and-reception “beam” fairly quickly. And although satellite-to-satellite laser-based communication has been successfully tested, the bulk of today’s Starlink satellites (1,200 as of mid-March 2021, with up to 60 additional satellites per incremental Falcon 9 launch, and up to 400 per Starship payload, assuming SpaceX eventually sorts out reliably successful takeoff and landing) act as simple relays, transferring bidirectional communications between each customer’s antenna and a SpaceX-managed common “base station” that links to a high-speed conventional terrestrial network. This all means that both the customer’s gear and SpaceX’s array of base stations are required to track and repeatedly redirect communications from one satellite cluster to another.
More than 1200 Starlink satellites have been deployed since 2019.
Falcon 9 launched from Cape Canaveral in November 2019 with 60 Starlink satellites on board. Source: Wikipedia
This photo shows a stack of 60 Starlink satellites in space before they were deployed in May 2019. Source: Wikipedia
There’s another key reason for the multi-satellite redundancy within each cluster in the constellation. The Ka and Ku microwave bands are pretty crowded already; the earlier mentioned geostationary internet satellites also use them, for example, and a lot of terrestrial-only microwave broadcast traffic also resides in those particular spectrum swaths (5G cellular companies, for example, are up in arms about Starlink’s aggressive growth plans, although pragmatically their concerns are likely not only technical but also competitive in nature, as they’d like to be the ones supplying potential customers with high-speed internet service instead).
This means that the satellites are required to dynamically reconfigure themselves to operate in a “clear” area (conceptually like how “smart” ISM band gear “sniffs” the spectrum and picks a channel not already in use by Bluetooth, microwave ovens, garage door openers, neighbors’ Wi-Fi equipment, and the like). And it means that the terrestrial equipment wirelessly linked to those satellites also has to be spectrum-adept.
Starlink satellites can be seen above the town of Tübingen in Germany.
Speaking of competition, although SpaceX is seemingly way ahead right now with Starlink, other companies are also participating (or planning to soon) in LEO-based satellite internet service. There’s OneWeb, for example, which entered (and exited) bankruptcy last year and strives to differentiate from Starlink via an enterprise customer focus (ironically, OneWeb has used SpaceX rockets as launch platforms for its own satellites). Amazon, with Project Kuiper, and Telesat, with Lightspeed, are notables among the others.
One oft-mentioned presumable competitor is L band– and LEO-located Iridium with its satellite telephone service, although reality doesn’t really match the hype. Iridium’s focus is fundamentally (one might even say solely) on voice communications; while (using legacy analog modems as an example) it’s possible to push data through a voice-tuned channel, doing so won’t be particularly speedy or reliable. Conversely, while one could certainly run VoIP on a data channel, SpaceX has to date shown no ability (or even inclination) to squeeze a Starlink transceiver into a handset.
Whether data or voice (or both), why are these companies all pursing seemingly more complicated LEO-based approaches? The answer comes down to physics fundamentals. A LEO-based satellite is less than 2% the distance from the Earth’s surface of its geostationary alternative. A “few” years ago, when I tested HughesNet at a RV park, I regularly experienced multi-second “ping” responses, assuming I got an “ack” at all, and I’m pretty sure the root cause of the problem was not terrestrial in nature. Conversely, SpaceX confidently claims a sub-100 msec worst-case latency (a necessary prerequisite to it being a candidate for rural broadband government funding) and typical 20-40 msec delays, ironically less than many terrestrial-only broadband alternatives.
Starlink is also a whole lot faster than conventional satellite internet service in both the downstream and upstream directions, although why is a question that I have yet to find a clear answer to in my research (I suspect the inverse square law, which would notably attenuate signal strength to and from a HEO satellite versus its much closer LEO counterpart—all other factors being equal—plays a part, as do more modern multi-channel bonding, modulation, and coding techniques analogous to those in the initial 802.11a/b to latest-generation 802.11ax evolution).
To-date test results published last October reported 79.5 Mbps average downstream speed, versus 50-150 Mbps Starlink guidance to early users at the time, and 13.8 Mbps average upstream bandwidth. Since then, SpaceX has collected more Starlink customers, of course; the public beta launched one month later (last November). But the company has also further built out its satellite constellation since then. Whether average speeds will further increase or degrade over time, driven by these contending variables, will be interesting to observe; incremental evolution of the network itself (such as the earlier mentioned satellite-to-satellite laser links) may further tweak the situation.
Starlink and its LEO peers are not without criticism, of course. Cost is one question mark; beta testers have to put down a $99 (refundable) deposit on a $499 hardware kit, along with $99/month in service fees. How this will all evolve over time is anyone’s guess; keep in mind, too, that the company will need to build out base stations in/near all countries in which they hope to cultivate a customer base. SpaceX executives and engineers vow no service “blocks” (which ironically is of no small concern to China, home of the Great Firewall) and at the moment don’t foresee any bandwidth-usage caps, either, although they reserve the right to evolve this latter position based on future network-usage trends.
The Starlink hardware kit includes a user terminal, shown here being held by SpaceX board member Steve Jurvetson. Source: Wikipedia
Other concerns are more fundamental. There’s the earlier-mentioned spectrum corruption issue voiced by 5G cellular providers and other telecom and datacom operators. Astronomers are also understandably up in arms, both because satellite clusters will periodically block views of astronomical objects “behind” them and because the sun-and-other illumination reflecting off the satellites will generate conflicting light pollution. SpaceX is experimenting with both albedo-reducing DarkSat coating and “sunshades,” neither of which has been shown to appreciably alleviate the issues to date.
This 333-second exposure image, taken from the Blanco four-meter telescope at the Cerro Tololo Inter-American Observatory, shows signal pollution from Starlink satellites. Source: Wikipedia
And then there’s the ever-increasing clutter of objects in LEO, which could impact other LEO-based operating equipment (potentially even creating a Kessler Syndrome situation, which viewers of the movie Gravity are already familiar with) and with rockets passing through LEO (either to higher orbits or beyond Earth’s gravity entirely). Estimates of Starlink satellite operating life, driven by eventual on-board electronics failures and other factors that I referenced in my research, range from 3-7 years.
One of those other factors is the faint effect of the thin atmosphere at LEO; the satellites contain weak ion (Hall effect, krypton-based, believe it or not) thruster engines that enable them to be precisely positioned within the cluster/constellation and at a given altitude post-launch, but are insufficient to counteract this ongoing atmospheric drag. After operating demise, a satellite will need to be replaced with a successor, but it’ll still remain floating dead in space until drag effects finally send it back to Earth (burning up in the atmosphere along the way, mind you), which could take up to 5 years.
There’s lots more that I could write, but I’m nearing 1,700 words at this point, and as I noted earlier, ongoing Starlink evolution will inevitably compel me to update and expand on my coverage in the future. So for now, I’ll put down the cyber-pen and hand it to you for your comments!
—Brian Dipert is Editor-in-Chief of the Embedded Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.