“The good news, is that if we use the Mission Abort System, it’s obvious the mission is well and truly aborted.”
FWIW, Tesla hasn’t authenticated the leaked (and grainy as hell) video of the Dragon capsule undergoing spontaneous self-disassembly. I expect they won’t have much to say until they have some facts to explain it.
Ars Technica summary: Here’s what we know, and what we don’t, about the Crew Dragon accident | Ars Technica
Brian
What exactly is the heritage of the SuperDraco thrusters? I know they had a successful pad abort test in 2015. I assume the Crew Dragon that flew in March this year had a functional set, but they weren’t used?
The link that I posted? The video has been removed from it and no mention is made in the article that it was ever there. Seems like an exploding gun to me.
Dammit! I was just thinking to myself the other day that I’ve come to count on SpaceX–and space stuff generally–for my (increasingly rare) ration of good news!
I’d been thinking about the failure, and felt that the best-case scenario for SpaceX is if they’d been doing envelope testing: that is, the moral equivalent of a test like this. Except that they didn’t quite reach their design margins.
That was pure wishful thinking on my part, but it sounds like this may be close to the truth.
That’s… quite the awkward quote, but if I’m interpreting it correctly, they were simulating the acoustic environment of an abort event, but at twice the magnitude. If so, then this might not be quite as bad as it looks. If it failed at only just below the target design level, then it might not require very significant changes to get it up to that level. Or–and again, this is pure speculation–SpaceX may just ask for a waiver, or specification change, or the like, and say that 150% is sufficient safety margin (assuming it can survive that but not 200%).
Of course they still need to understand the problem. We’ll just have to wait and see on that front.
SpaceX seems to have hired an IKEA designer to pack their satellites.
They have a whopping 60 satellites packed in there. I count 30 repeating layers in the image, so I suspect there are two stacks side by side, though I suppose there could be four stacks of 15 and each satellite has a pair of repeating elements.
There is, apparently, no dispenser. They must have some kind of hardpoints where they can attach in a stack and then release as needed. Presumably they just pop off the top of the stack one by one.
Launch is scheduled for 02:30, May 16 UTC. So, Wed evening US time.
Elon says two stacks of 30
Apparently the shape of the satellites is still some big secret.
It’s impressive how flat they are. And they’ll still unfold solar panels and a phase array antenna, so they’ll be even flatter deployed. I wonder if they’ll use the aerodynamics to their advantage–at the least, it’ll enable fast propellant-free deorbiting. And maybe cheap orbit transfers as well.
And of course I’m curious about the full shape. I wonder if they are not quite as flat internally, and sorta intersect like a stack of folding chairs. We should find out soon… the deployment video should be pretty cool.
Any data/guesstimates on what they’re capable of and how much each costs to make? Is the main performance bottleneck data transfer interconnects (between sats) again?
You likely have insight into this: Sats seem to be getting flatter in much the same way that televisions, monitors, computers, radars, telecomms hubs and likely other stuff I don’t know about have been getting ever closer to Kubrick’s Monolith. What’s that about?
Satellites may themselves form a kind of phase array, when, like individual microlenses in a plenoptic camera or individual antennas in an AESA radar, they form a whole that’s greater than the sum of its parts thru software-controlled interactions.
Any chance of getting solid-state LIDAR any decade soon?
They are supposed to have laser interconnects–I honestly don’t know much about this, but it sounds like that won’t be the bottleneck.
I’d say the main bottleneck is beam spot size. That determines how many users share a downlink, and thus the max population density that it can support. It also determines the power density, which is correlated with bandwidth.
For a phase array antenna at a given distance, the beam size is determined by the aperture–that is, the area of the antenna. So a big flat surface is a better choice than a dense cube.
Just the general progress of miniaturization, I’d say. Things get flatter because that’s how you maximize area while reducing volume. SpaceX’s advantage here is that they can afford to be on the bleeding edge–geostationary satellites are not so state of the art, because they are big and expensive, and intended to last for decades. So they make conservative design choices because they definitely don’t want them to fail early. In contrast, SpaceX’s constellation is designed to handle failures, and they’ll be launching new batches almost continuously. They can roll hardware fixes into the new batches as soon as they know about them.
I’m not so sure about that. There’s a principle, called the thinned-array curse, which more or less states that sparse arrays are pointless as a means of increasing power. It’s true that you can use them to decrease the spot size, but the extra power leaks out to side lobes and so the density stays unchanged.
Sparse arrays are great for things like radio astronomy, which can afford to have long sampling times and so aren’t as power limited. They benefit from the extra resolution. But for a comms satellite, it’s really about power density on the ground since that’s what determines the signal-to-noise ratio and thus the bandwidth by the Shannon-Hartley theorem.
It’s coming. But I’m in agreement with Elon here when it comes to self-driving. You can’t solve self-driving without solving the vision problem, even with LIDAR. And if you’ve solved vision, you don’t need LIDAR. This argument is independent of the actual difficulty of solving the vision problem, which is probably being underestimated. But whether it’s on the horizon or decades away, it doesn’t change the picture–LIDAR only gives you information you already had, given that you’ve solved the other required problems.
Would you say that the lowest hanging fruit to overall satellite telecomms performance improvement lies in reducing spot size (irrespective of power density) or increasing power density (irrespective of spot size)? Or is it a situation where many small users would be better served by reducing spot size (irrespective of power density) whereas a few large users would most benefit from the increasing power density (irrespective of spot size)?
SpaceX’s more experimental, iterative approach, which Must must have learned thru writing software, seems to work pretty well whether it’s applied to real or virtual space rocket building.
Radar for range/closure rate/Doppler signature fused with IR/visible spectrum cameras seems to offer similar or better capabilities with less cost and more readily developed tech. But lasers are as great as rockets so I’m conflicted.
Depends on your target market. Is the goal to provide broadband internet access to every rural place on Earth? Or is it to replace Comcast, etc. in existing dense urban areas? If it’s the former, then power is what you’re after–the population densities are pretty low in most areas, and so you just need enough power density to provide decent bandwidth. But in dense areas, every user in a given spot is sharing bandwidth. So a small spot is very beneficial here.
It is a refreshing change from the usual slow development cycle in aerospace. I’ll note that Musk isn’t the only one taking this approach to satellite development. Planet Labs has a constellation of 150 cubesats and has a similar iterative development approach. Crazy that those 150 satellites will be peanuts in comparison to StarLink once SpaceX really gets going on launches. I wonder if they’ll have cameras on them…
StarLink will also be a great way to keep the Falcon 9 launch rate up. A huge part of the costs in aerospace comes from having such a low unit count. A known high volume customer means they can keep their factory at capacity. They can also take more risks with their own launches than they might with others–as evidenced by their desire to reuse the payload fairings on a StarLink mission (not sure if it’ll be this first one or some later one).
Standard LIDAR doesn’t even give you velocity information, though there are variants that do (not sure if there’s solid-state Doppler LIDAR yet). But anyway, there’s still the plastic bag problem. Is that obstacle in the lane ahead of you a plastic bag or a rock? You need vision to decide.
Depth from vision is one of the earlier problems there is. I wouldn’t quite say it’s 100% solved yet either, but I’m confident in saying it’ll be solved before any of the really hard problems in vision.
The telecomms aspect seems to be the less interesting part; Connecting with rural Sub-Saharan Africans, Saudi Nomads or the Kentucky backwoods may matter a lot to them but it’s not going to advance humanity much apart from bringing up the laggards. The major centers where civilizing happens are already well-connected. E.g.: If you wanted to get a tech start-up going, would the lack of telecomms access be a major obstacle to you compared to the other challenges?
The main potential seems to be gathering data and perhaps sending intel. For example, you could have not just cameras but every kind of potentially interesting sensor, from the lowest frequency radio waves to Gamma rays, including magnetic and temperature sensors. If it might pick up anything potentially interesting, gather data and ask questions later.
What would that tell us? We can’t know yet. If there are systems that involve tens or hundreds of factors all interacting with each other non-linearly (like a planet), it becomes impossible for a human brain to track so many relations and discern patterns through the noise. But a computer has practically infinite memory and the strongest case ever of Asperger. A computer with neural networks might be able to figure out systems so complex that whatever algorithmic patterns it discerns look like alien gibberish. But we might still be able to use that gibberish to make predictions, like when your intuition tells you not to do something even if you’re not really sure why. AI/neural networks would transform previously overwhelming amounts of raw data into correlated, integrated information.
The next step would be figuring out the best ways to use that relevant, timely, accurate information as actionable intelligence. I guess satellites could be useful there too to communicate with robots/automated systems or designate spots. I’m guessing more could be done than that but my imagination fails me. Does anyone have ideas?
Is it like fab building where the capital costs are astronomical because everything has to be both custom-made and cutting edge? If so, once a critical threshold is reached, average costs will get low enough for more users to jump in, further lowering average costs which will bring more users in and so on. If we knew what that critical threshold was and what to do to bring it about, things could go much faster.
It now appears that Musk has not one, but TWO separate groups building his stainless steel spaceships–one in Boca Chia, the second at Cape Canaveral.
The thing with Falcon 9 and Falcon Heavy is that as the boosters are reused while new ones are being manufactured, SpaceX’s inventory of rockets grows. This was intentional, as Musk’s plan has been to build up a large enough inventory of Falcon 9’s that they can then halt production and have lots of rockets for customers while they re-purpose production for the Starship and its booster.
Starlink is very useful in this regard because the timelines are fuzzy. If Starship production is delayed, they can keep making Falcon 9 boosters. Also, if it’s *really delayed, they’ll need more boosters than they thought. So this strategy risked SpaceX either running out of Falcon 9 cores before Starship can take up the missions, Or building too many Falcon 9 cores. But Starlink gives them a ‘backup’ set of missions that they control the pace of. If they get more commercial bookings than they expected, they can back off a bit on Starlink launches. If they get fewer, or for other reasons find themselves with more cores than they need, they can accelerate the Starlink deployment and use up their cores that way.
If you’ve got more rocket cores than you need or can sell, then if the choice is mothballing them or using them for Starlink, Starlink satellite launches become almost free to SpaceX - especially if they can reuse the fairings for Starlink. Now we’re down to facilities, fuel and labor for launch costs. It’s sort of a synergistic relationship, and puts SpaceX way ahead of anyone else in terms of launching such a vast constellation of satellites.
In other news, the Trump administration is requesting a supplemental funding of 1.6 billion dollars for NASA to implement their moon-in-2024 program. They’re still planning on using SLS/Orion for the astronauts, but everything else is open to bid - landers, satellites, fueling stations, surface buildings, whatever.
Also good news: They’ve scaled back the ‘lunar gateway’, which is a stupid idea anyway, but at least it’s less stupid now. They’re putting it in a lunar-only orbit for one thing, and scaling its size down.
I think the gateway exists almost exclusively as a SOP to the international ISS partners and to keep all the ISS-related workers in jobs. There’s actually no real use for it. It can’t be manned permanently because of radiation exposure, and in the last incarnation was in a crazy orbit that would severely limit the windows for going to the moon and back and actually increase the Delta-V requirements for a moon landing. So the more they scale that back, the happier I am.
A list of highlights from a StarLink media call with Musk
Musk: Each Starlink satellite has "about a terabit of useful connectivity
I’m sure he means a terabit/sec of connectivity. Pretty good, even if some of that has to go toward multi-hop routes.
The ISS has in the ballpark of 100 kW of solar capacity. So, each satellite here is probably ~1.5 kW.
Decision to use Krypton. Elon makes Superman joke. Real answer “Costs less than xenon.”
Neat. Xenon is the usual choice for ion drives since it’s the highest performance, but krypton is probably pretty close and I’d guess a few orders of magnitude less expensive.
Each Starlink costs more to launch than it does to make, even with the flgiht-proven Falcon 9. #Starship would decrease launch costs of Starlink by at least a factor of 5
Considering that SpaceX’s launch costs are probably <$30M, and they’re launching 60, that means each satellite probably costs <$0.5M.
Some other good stuff in the link above.