Starlink 12K: Elon Musk’s Big Gamble on Global Satellite Internet

SpaceX CEO Elon Musk’s plan to provide high-speed communications to virtually any location on Earth got a big boost this week when the Federal Communications Commission (FCC) approved the company’s plan to add 7,518 satellites to the company’s Starlink constellation.

The action brings the total number of satellites in Starlink to 11,943 following the FCC’s earlier approval of 4,425 spacecraft last year.

Starlink is Musk’s ambitious entry into the global satellite Internet race. He is gambling big that there is a sufficient market worldwide to make the constellation profitable.

SpaceX launched two test Starlink test satellites into orbit earlier this year. Published reports say Musk wants to launch the first batch of satellites in the middle of next year, with service to begin in 2020.

Starlink is facing competition from OneWeb, which is planning to launch a constellation of 882 satellites to provide similar service. OneWeb plans to begin launching spacecraft next year.

The FCC also approved satellite broadband constellations by three other satellite companies last week. Telesat Canada received approval for an 117-satellite constellation while LeoSat plans to launch 78 spacecraft.

Kepler Communication’s 140-satellite constellation is focused on providing communications for the Internet of Things.

“These proposed satellite systems are expected to enable fixed satellite service in the United States, expanding global connectivity and advancing the goals of increasing high-speed broadband availability and competition in the marketplace,” the FCC said in a press release.

The constellations will greatly increase the number of satellites in Earth orbit. There are currently about 4,900 spacecraft in orbit out of the approximately 8,100 launched since the Space Age began in October 1957. Nearly 2,000 spacecraft are currently operational.

  • envy

    The 17 reflown Falcon boosters would disagree with your opinion of their reusability.

  • Robert G. Oler

    none of them were reused…they were refurbished

  • envy

    The distinction is meaningless if it has no significant effect on the cost reduction.

    Look at it this way:

    A booster costs, say, $45 million. everything else costs $12 million per flight, adding to the current list price of $62 million. A perfect reuse would cost $0, so 10 flights would cost 45+10*12=165 million or 16.5 million per flight

    A “refurbishment” costs, say $10 million, so the total cost for 10 flights is 45+10*(12+10)=265 million or 26.5 million per flight.

    So in this case, refurbishment lowers the cost per flight by something on the order of 60% ($62M to $26.5M), while perfect reuse lowers the cost per flight by 75% ($62M to $16.5M).

    You’re squabbling about a 15% change when they’ve changed cost 60% already (on top of a earlier ~50% reduction from vertical integration et al.). Perfect reuse is not the goal, yet. It’s not necessary to have a huge impact on the market.

  • Jeff2Space

    Here’s a cite on the cost savings:

    From above:

    But SpaceX president Gwynne Shotwell told the Space Symposium conference that the cost of refurbishing the Falcon 9 rocket that originally flew the CRS-8 Space Station resupply mission last year for SES-10 was “substantially less than half” what it would have cost to build a brand new one.

    You’ve been told this repeatedly yet dismiss it out of hand as if the president of SpaceX, Gwynne Shotwell, was lying to the Space Symposium conference.

    The fact is that this first reuse was using a Block 3. SpaceX has standardized on Block 5 which includes many changes to make reuse easier and cheaper.

  • envy

    The descending node to 53 degrees from Boca Chica goes over the gulf, then over Yucatan far downrange. They might be able to fly that, considering that the USAF is allowing overflights of Cuba out of the Cape again.

    Also, BFR doesn’t have to do the RAAN plane change. It can insert them to ~400 km and the birds can raise altitude as the right planes precess over them. This will take several months, but they might be able to patch into the constellation while they are plane-changing and get some useful operations life out of them for that time.

  • Think of it this way: The Cape can’t launch above about 57° because they’ll overfly Newfoundland. The Yucatan is a lot closer–and a lot more heavily populated–than the Canadian Maritimes.

    You raise altitude to insert the birds into the proper slot in the plane, not to change the RAAN. Everything in the plane has the same RAAN. If you want to populate two planes, you need to spend the Δv to get to the second plane.

  • envy

    The issue with high inclinations from the Cape is the lower stages or SRBs falling on the Eastern seaboard. This requires a dogleg.

    RAAN precession depends on altitude. You can thus change planes relative to to the rest of the constellation by changing altitude and waiting a while. Iridium uses this to move satellites between planes all the time.

  • Looks like you’re right on the 57° restriction. The Newfoundland thing is in several documents, but it does look like the main restriction was to avoid dropping things on the Outer Banks. However, I also found a certain amount of language about the extreme poor form of dropping debris on foreign territory. That may be where some of the Newfoundland stuff came from.

    There’s a range safety criterion such that the expectation of casualty is <100 in 10^6 along the entire flight path. (See section 4.3.3.)

    I also found this old diagram from some of the Shuttle documentation, which shows an arc at about 650 miles from launch labelled “SRB impact”, but I suspect that it’s more likely a zone where debris from any launchers in the fleet back then could drop something. (The SRBs impacted about 150 miles downrange from launch.)

    If you used that criterion on Boca Chica, the southerly azimuth that yields a 53° inclination is 138°. 650 miles along that heading is about 40 miles inland from the Yucatan coast, about 50 miles south of Campeche City. It’s quite densely populated. However, I have no clue if the 650-mile arc is relevant for the BFR. It’s way downrange of the BFB separation point, but it’s also well into the part of the range where a premature BFS shutdown would drop.

    Ultimately, the expectation of casualty is an integral over the path that takes account of the instantaneous probability of debris landing at that point times the population density at the same point. I suspect the reason that the dogleg Cuba overflight is being considered is that the path goes over a fairly narrow spot on the island, where the population density is low. You’d have to do the calculation for the BFR in some detail to see if you could get away with the launch to the south.

    It’s a lot closer to feasible than I expected though. And it sure would solve a whole bunch of launch capacity problems for SpaceX.

    Looks like the Iridium RAAN precession is a result of the equatorial bulge affecting near-polar orbits. I’m not sure whether that precession would even work at 53°, or how long it would take to cover the needed angle. But if you actually launched to the south, the 725 m/s needed to do it the old-fashioned Keplerian way isn’t so bad.

    Thanks for the info; very informative.

  • envy

    That’s a good summary.

    Also, all orbits precess, even equatorial ones. The difference between precession rates for 400 km and 550 km orbits at 53 deg is small, though, probably a few degrees of RAAN per month. Still, drifting 7 or 8 degrees between planes would not take that long compared to the 5+ operational life of the satellite.

  • OK, I worked it out, using the horrible formula. For 53° inclination and circular orbits at 550 and 400 km, I get 0.36°/day difference. For 24 planes (per the application modification), that would be 7.5° of RAAN separation, which would only take 3 weeks.

    That’ll work. I’m sold.

  • Jeff2Space

    While you have a point, sometimes Robert’s comments come across as the old guy who really hates change (I can relate because we’re going through a lot of change where I work and I’m not as young as I used to be). Robert’s repeated assertion that refurbished Falcon 9 first stages have not proven that they save money is quite simply false. The only way that it could possibly be true is if Gwynne Shotwell straight up lied when she made the statement that the first reuse’s refurbishment costs were “substantially less than half” of what it would have cost to build a brand new booster. I’ll take the word of Gwynne Shotwell when she makes what is a financial statement about SpaceX. She’s one of the biggest reasons that SpaceX is doing as well as it is, IMHO.

    I absolutely do respect Robert’s experience in the aerospace field, but this is one instance where he absolutely should admit that he was wrong. If you accept the fact that reuse of launch vehicle hardware can save money, it opens up a lot of possibilities for the future of crewed spaceflight.

  • rod57

    What limits the launch rate on a range? Matula above says “if it takes 3 days to recalibrate between launches.” what is being recalibrated ?
    Why can’t there be a launch a day (off multiple pads) if mission launch windows allow and pads are available, and could there, one day, be two simultaneous countdowns progressing on one range ?

  • Here are the things I know of that are required to clear the range:

    1) Notification that the pad operations are ready, or likely to ready, and that flight termination systems are ready.

    2) A maritime notice to clear down-range. This may actually be the biggie, because it takes a while to re-route shipping.

    3) A NOTAM to the airlines and civil air community. There are fairly intense discussions ongoing about coordinating between the air traffic system and the range, but it’s gonna take a while to work this out and put the communications in place.

    4) Some amount of maintenance, calibration, and testing of the radars and comm links required.

    I suspect the hard limits on range capacity depend on how often it’s economically feasible to clear the airspace and sea lanes downrange. Currently, the Eastern Range has a goal of 48 launches per year, up from about 20 right now. But that’s a multi-year goal. The Western Range has been below 10 historically, and I have no idea what its actual capacity is.

  • rod57

    Thanks, I can understand the aviation notifications – but with a launcher with a very low risk of dropping anything in the sea – perhaps one day they wont need to clear the sea for a launch that plans to recover boosters.

  • The risk isn’t so much staging as it is malfunctions. It’s entirely possible that controlled landings mean that you need a smaller area cleared downrange, but it’s not likely to be a zero area. And since we’re interested in how fast the range can scale to support Starlink, “perhaps one day” isn’t exactly the timeframe we’re looking for.

    The thing that the aviation community is complaining about is the length of time that the airspace needs to be closed. My guess is that that will also apply to the maritime folks. There are three factors in determining that time period:

    1) The size of the zone based on the uncertainty in the trajectory.
    2) The uncertainty in launch time based on holds within the launch window.
    3) The time it takes for something in the zone to leave.

    These are a bit a cross-purposes: Fast things on a constant course need to have the zone closed sooner than slow things to avoid entering it if there’s a hold. But, once in the zone, fast things clear out sooner than slow things do. I don’t know how the range folks manage this, but it’s likely to be more complicated a problem than you might think, with some fairly subtle requirements that feed into how aviation and maritime operations proceed.

    However, it’s pretty clear that reducing hold times and narrowing the distribution around which they occur, as well as narrowing the range of of deviant trajectories (which is exactly what AFTS does) will result in smaller exclusion zones with shorter durations. That’s stuff that’s not easy, but is at least in the control of the launch operators to improve.

  • rod57

    Thanks. Prompted by your mention of AFTS I found which says “Automated systems should enable faster turnarounds from one launch to another. Monteith said SpaceX could, in theory, launch twice within hours when both of its Cape pads are available.”, also interesting that AFTS/AFSS reduces personnel by 150 (60%).

  • The 45th Space Wing (which operates the Eastern Range) has a project underway called “Drive to 48”, which aims to do 48 launches/year by 2023. This has an overview, as well as a bunch of things about range operation that I hadn’t thought about.

    But with Starlink in the offing, 48 launches/year is a serious constraint, especially if that cadence isn’t reached until 2023.