=SpaceX =technology =communication =radio



SpaceX has launched over 1600 satellites to provide internet service, and plans to launch many more.



Traditionally, satellite dishes communicate with geostationary satellites, because then the dishes don't need to move. But geostationary orbit has some disadvantages:

- getting there is more expensive
- the number of slots is limited to 1800
- transmission over greater distances is harder and has higher latency


The idea of Starlink was to use satellites in low earth orbit (LEO) and use phased array dishes that can instantly switch between satellites as they come and go. This avoids the above problems, but has some disadvantages compared to geostationary satellites:

- Phased arrays are obviously more expensive. I don't know what the manufacturing cost of Starlink dishes is, but they've been priced at $1000.
- Geosynchronous satellites can be kept over useful areas. Satellites in LEO spend a lot of time over oceans.



The original plan of Starlink was to use laser links between satellites. This is certainly possible, but lasers spread out quite a bit over the distances involved, so losses are higher, and satellites can only receive one signal at a time. Realistically, the beam diameter at a receiver would probably be >10m. This makes the transmitters and receivers considerably more expensive than fiber optic transceivers: not only would a reflector need to be very precisely mechanically steered, but the usual approaches for high-bit-rate fiber optic transceivers only work for low power, so the data rate would probably be less than 1 gigabit/second, limited by the switching speed of high power laser diodes. That capacity is then divided by the number of hops and the number of users per satellite.

So, instead, Starlink satellites just act as relays between users and ground stations, which can then have fiber connections for backhaul.

Elon Musk didn't believe his engineers when they said that approach wasn't practical, so it's still planned for "when" they figure it out. In fact, Starlink launches have now been paused until laser links can be added to all the satellites, which is supposedly going to happen in a few months. I'm skeptical of satellite laser links with cost-effective bandwidth happening on that kind of schedule. On the other hand, if you're willing to pay much higher prices for data transfer in places without ground stations, perhaps for, say, military UAVs, then the economics change.



Overall, if I compare the Starlink approach to geostationary satellites for high-volume communications, Starlink seems significantly better. The problem is, that's not the right comparison; a better comparison is satellites vs towers and aerostats.

What if those ground stations just communicated directly with users? This eliminates the need for satellites. Because towers don't move, dishes can be fixed, which is much cheaper than phased arrays. And because towers are closer than satellites, less power is needed; Starlink satellites orbit at 550 km, and will never be directly overhead.

Because the satellites are far away, Starlink dishes use >100x the transmission power of WiFi. If you used the same power and dish size with towers, the distance would be limited by the horizon.



The amount of air between the ground and space is equivalent to about 10 km of air at sea level. Suppose we're pessimistic, and only allow a 10km range for our towers. In that case, the economics of towers are still better. In theory, a satellite can cover a larger area, but the number of users a single satellite can serve is limited, so even in places with low population density, it's better to use towers, assuming that building a tall enough tower to get line of sight is legal.

If getting line of sight with towers is a problem, it's also possible to use aerostats instead of towers. Project Loon was a dumb idea, but tethered balloons with fiber and electrical lines running up to them are feasible; they've just generally been more expensive than towers. For example, the military has used stuff like this, and Altaeros is trying to commercialize aerostats as an alternative to cell towers.



Companies that use rockets to launch things to space already existed. SpaceX did the same thing but with less bureaucracy, and is thus outcompeting ULA. Maybe it should focus on things that already work but are expensive.

For example, if SpaceX can make cheap phased arrays, then maybe it should be selling military AESA radar. I guess the market is smaller, but the prices are much higher.

If I was running SpaceX, instead of Starlink, I would have gone for competing with Boeing. Aircraft work pretty well and there's a big market for them, and like ULA, Boeing has a lot of institutional problems, so there's room for competition.

If you can get enough government money for internet service, and you can launch satellites, it's tempting to do internet service with satellites, but I don't think Starlink will be able to meet the program requirements, for basically these reasons.



You might think that SpaceX would have people who considered these things more deeply than I have, but my track record of mocking Musk-related things is pretty good. For example:

- I mocked Tesla for using induction motors for electric cars, which was a choice that it seemed rather proud of, and eventually it switched to typical modern permanent magnet motors.
- I mocked the Hyperloop paper. So far, everything novel about it (air bearings and turbocompressors) has already been scrapped by its fans, and development isn't going anywhere.
- I disagreed with SpaceX's choice to use a gas generator cycle, and now they're moving away from it.
- I mocked the aerodynamics of the Tesla Cybertruck, and mocked Musk for saying its drag coefficient would be 0.3; here's an analysis that seems about right.
- I agreed with Vern Unsworth, and Musk's response made me lose some respect for him.


More debateable are some choices that I argued were suboptimal, but never said weren't viable. For example:

- methane fuel (yes, I know they want to minimize coking for planned reusability, but even in that case, ethane is better)
- vertical landing instead of flyback boosters (yes, vertical landing was easier to add to an existing design, but wings give better overall performance, but the comparison can get complicated)
- I strongly disapprove of the way SpaceX overworked employees, but obviously a lot of people think long hours make companies more successful


Yet, if you compare SpaceX to, say, Blue Origin, it's done a lot better. Maybe it doesn't matter that much if you do some silly new things, and what's really important is doing things. Or maybe the bad ideas are a side effect of Musk understanding technology well enough to choose relatively competent engineers. Or maybe Musk is just better at running a company than Bezos, and would have outcompeted Amazon if he'd been in direct competition. I like the way Bezos banned PowerPoint, and I also like this from Musk; they both have a focus on actively countering and disrupting typical American bureaucratic managerialism. Tory Bruno isn't dumb, but perhaps the anti-bureaucracy of Bezos plus half the technical understanding of someone like Tory Bruno is a winning combination.


Anyway, I'm not a huge fan of Elon Musk, but I'd never mock the people placing hope on him; people need hope, and a lot of American leadership is moribund these days, so he doesn't have much competition.



back to index