Anyone watching the satellite Internet scene in the last few years could not have missed Starlink, the new Low Earth Orbit (LEO) kid on the block. From a researcher’s perspective, SpaceX — Starlink’s operators — are quite opaque, however. There is little engagement with the research community, and official pronouncements often lack detail — they’re clearly not the kind of company that takes people on factory tours to proudly show off how they make and run things. Much of what is being bandied around publicly about Starlink comes from social media influencers whose technical qualifications, insights, background, and access to information are at times questionable, and whose information cannot be relied upon. Very little is peer-reviewed or quality-assured in other ways.
Even what is published in academic journals or conferences often fails to do the complexity of Starlink justice — it’s a mix of clever and agile orbital engineering, radio frequency engineering, optical engineering, manufacturing, logistics, computer networking, and targeted marketing that makes this system tick. Just looking at it from one discipline’s perspective can miss important aspects.
But there is one place where a lot of the specialist information converges, and where correctness matters: SpaceX’s filings with the US Federal Communications Commission (FCC) and the associated documents.
FCC filings: A lament
Alas, FCC filings are not an easy source to work with as FCC applications are voluminous. Large parts consist of boring lists or memos outlining who met when with who and what was discussed. They get amended, often multiple times, objected to, and objections get responded to, revealing hitherto undiscussed aspects. They get approved, but often only in parts, and other parts get deferred or declined. Authorities, once granted, are often subject to modification requests.
You’ll want to read them in chronological order so you become familiar with the background and understand the information subsequent documents refer to. You’ll want to read them in reverse chronological order so you’ll know what the current state of play is without having to wrap your head around constellations that were never pursued in the end. Sometimes, you just want to read them, but while they’re generally not under wraps, they’re behind a fragile online portal that appears designed to make this as difficult as possible by logging you out before you can even say ‘application to launch and operate’.
The filers — both SpaceX as applicant and a plethora of others as petitioners — contribute to the fog by playing, sometimes quite overtly, a game. The rules of the game aren’t published anywhere, but the aims seem to be quite clear — as an established geostationary satellite provider fearing that SpaceX may be eating your lunch, your goal is to throw as many spanners into the works as you can, usually cleverly disguised as technical objections, to delay and suppress.
SpaceX’s goal, on the other hand, seems to be to make you spend as much money on lawyers and other experts as they can, by getting you to comment on system alternatives and options they subsequently abandon. And you’ll accuse each other quite openly of all sorts of things.
I’ll moan a bit more later. For now, let’s get started.
Types of FCC filings and where to find what
The FCC website would be a rabbit warren even if there was no SpaceX, but you’ll want to head to the filing system and, armed with an FCC login (free), you’ll want to search by file number below. File numbers are like folders that contain all documents related to an application. These usually come in two sets: the main application document set and the set of other documents — which includes any order by FCC granting the application, as well as any objections and other correspondence.
Perhaps the most important file numbers for Starlink are the applications to ‘launch and operate’, which start with SAT-LOA-… There are three of them. The first two pertain to the first generation of Starlink satellites and the third to the second generation. These take care of the space segment along with their amendments (SAT-AMD-…) and modifications (SAT-MOD-…).
The ground segment file numbers start with ‘SES’ (an abbreviation of Satellite Earth Station) and fall into the categories of SES-LIC-… (applications for license) and their amendments and modifications in SES-AMD-… and SES-MOD-… There are literally hundreds of them, however, only a small number pertain to end user equipment (Dishys in their various versions). The vast majority is for gateways in the FCC’s jurisdiction (the United States).
Within a SAT-LOA-… application main document set, the most interesting documents are:
- The legal narrative (aka ‘application narrative’) contains a layperson’s description of what they want to do and the justification for why the FCC should approve it.
- The technical narrative (aka ‘Attachment A to supplement Schedule S’) contains the technical description of what is being proposed and the arguments as to why a proposed system will be compliant with all rules or worthy of flouting some. This may contain endless attachments of diagrams showing compliance with limits on interference to competing systems. You will want to understand orbital mechanics, evasive manoeuvres, radio frequency engineering and International Telecommunications Union (ITU) regulations and procedures to read this — or ask a friend.
- The waiver requests — a summary of requests to either be allowed to bend the rules, or to apply in an unusual format (FCC’s forms weren’t designed for LEO mega-constellations, so this is a standard SpaceX request).
- The ‘Schedule S Tech Report’ contains very little prose but lists — for Starlink applications — orbital plane configurations, and receive and transmit beams (including information about antenna gain). Note, by putting the data in a different order here in every amendment or new filing, you can drive up your competitor’s engineering and legal bills as they scrutinize the documents just to find that little has changed.
Some applications also include modelling data for various tools to demonstrate compliance. The set of other documents is like a box of chocolates — you never know what you’ll get.
So what have I learned?
Lots. Starlink’s Generation 1 FCC journey starts with SAT-LOA-20161115-00118 (and yes, that first number is the filing date) asking for 4,425 satellites in 83 orbital planes with five different inclinations at heights between 1,110 and 1,325km, which ended up being approved but never implemented. But it reveals a lot about the beams. Beam shaping ensures that they are quasi-circular when they hit the ground, regardless of direction, and designed to cover just a bit more than one cell.
We learn that the satellites adjust beam transmit power to ensure a uniform signal at the Earth’s surface. Plus, this signal, across all satellites and beams, must never exceed an ITU-imposed limit to prevent interference to ground-based receivers of other services. You may also have heard about Starlink’s famous requirement for a minimum satellite elevation of 25° above the horizon — it’s still 40° here, but we learn that this is to protect terrestrial microwave links.
SAT-LOA-20170726-00110 was the next iteration, approved in parts, and represents the authority that the Gen. 1 (V1 and V1.5) satellites of Starlink operate under, with new Ku- and Ka-band frequencies appearing in the application. Three modifications to the application (SAT-MOD-20181108-00083, SAT-MOD-20190830-00087, and SAT-MOD-20200417-00037) bring it to the present final constellation with shells at 53°, 53.2°, 70° and 97.6° inclination at orbital heights between 540 and 570km. The inclination of shells matters because it determines the northern- and southern-most latitudes that satellites in the shell can provide service to.
The FCC journey of Starlink’s Dishy terminals started with SES-LIC-20190211-00151, with SpaceX requesting a blanket license for up to 1M of them, to communicate with satellites above 25° elevation on Ku-band, later above 40°.
Enter the Generation 2 space segment with SAT-LOA-20200526-00055, which was granted in parts only on 1 December 2022 after being amended in SAT-AMD-20210818-00105. SpaceX requested to use Ka-band for communications with user terminals at the time (something that has yet to be reflected in Dishy license applications) and use of E-band for gateway communication.
The part-grant relates to 7,500 satellites in 525km / 53° inclination, 530km / 43° and 535km / 33° shells, where SpaceX had requested 3,360 satellites in each of these shells. Interestingly, the grant of authority does not say which of these shells should take how many of the 7,500 satellites — that appears to have been left to SpaceX’s discretion, who so far seem to have concentrated on the 43° shell, perhaps given that they already had Gen. 1 shells at 53°. SpaceX’s wish to deploy at shells around 340-360km for use in communication with mobile phones has so far not been granted, despite repeated pleas over time, as late as March 2024 in SAT-AMD-20240322-00061.
The Dishy journey continued with SES-LIC-20210309-00698 for ‘Earth Stations in Motion’ (ESIMs), which for the first time mentions 64QAM as the intended modulation scheme. Soon thereafter, SES-LIC-20210708-01019 asked for a blanket license without a limit on numbers for V2 Dishys. SES-LIC-20210803-01360 extended this to ESIMs, and SES-LIC-20220125-00081 to a ‘high-performance fixed earth station’. All were granted. No mention of Ka-band use in any of these.
SES-LIC-20230228-00228 and SES-LIC-20230228-00232 apply for the latest user terminal versions but likewise don’t mention the Ka-band. This is reasonably interesting as Ku-band user downlinks under SAT-LOA-20200526-00055 don’t really allow more than about 20 Gb/s of downlink capacity to a single cell (in practice a lot less in most cases at this assumes zero capacity in adjacent cells) regardless of what satellite(s) serve it. Using Ka-band downlinks could in theory bump this up by another 25.2Gb/s. And then? Good question!
I learned a lot more, much more than I can fit in this post, from Starlink tickling the regulatory and competitor crocodiles’ tails in terms of geostationary orbit protection and arguments over ITU methodology to FCC requirements for ITU certification due to ITU backlog endangering SpaceX’s launch schedule. Oh, and there are numerous astronomers petitioning FCC that Starlink photobombing their telescope images prevents them from seeing that killer asteroid that will wipe us all out one day.
Meanwhile, the filings also show SpaceX reporting to the FCC about the number of near misses and evasive manoeuvres their satellites had to conduct in space, spiced with plenty of reassurance that long satellite life in orbit isn’t really a SpaceX design goal and that what goes up has to come down sooner rather than later, so the constellation can stay technically up to date.
But it’s a lot of paperwork by anyone’s standards. One wonders whether SpaceX’s recent ITU application through Tonga’s authorities for a constellation with nearly 30,000 satellites represents an attempt to circumvent FCC scrutiny by launching under a flag of convenience. Or is it just a red herring designed to scare competitors or keep tongues wagging? Never a dull moment!
Dr Ulrich Speidel is a senior lecturer in Computer Science at the University of Auckland with research interests in fundamental and applied problems in data communications, information theory, signal processing and information measurement.
This work was carried out while the author was on research and study leave at the IIJ Research Lab in Iidabashi, Tokyo, whose support the author wishes to acknowledge with gratitude.
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.
wonderfully useful. thank you. Just a note that the Gen2 addition of Ka band to the user terminal links will increase downlink spectrum to 3850MHz, and using dual polarisation as opposed to single pol for Gen 1/1.5 will double it again. Hence you get effectively 7700Mhz user downlink vs 2000Mhz currently and assuming same spectral efficiencies around 4 x capacity increase which gets you to the 80Mbps for Gen2 vs 20Mbps for Gen1