Hacker News new | past | comments | ask | show | jobs | submit login
SpaceX loses another Starship prototype as landing sequence fails (arstechnica.com)
71 points by alexrustic on March 30, 2021 | hide | past | favorite | 96 comments



I am beginning to wonder if either:

1. This landing procedure is not viable. Restarting an engine in the middle of a rotation like that, fuel sloshing around is really hard. Just restarting an engine in zero G has always been tricky, because the fuel is floating around and doesn't have gravity pulling it to the pump. This case is even more dynamic, and you have to get it right immediately.

or

2. If the raptor engine is too complex to be reliable.

Then again Falcon9 also failed to land many many times early on, and it has become rather routine. Though still not routine enough that you would want to be a passenger riding one down.

please nobody reply with "prototype" or "rapid iteration" we know, we have seen your posts 20,000 times here already :)_


Well considering we already had a successful landing of sorts with SN10, I think this is a little premature. Yes, it exploded 8 minutes afterwards but it does not seem like that had much to do with the stresses of the landing manoeuvre.

The raptor engine on the other hand does seem like something that they are far from mastering the production of. Being the first full flow staged combustion, methane fuelled orbital ready rocket means that there is less to directly compare to so I'm not surprised that they are having reliability issues with it. I am optimistic that they will eventually overcome these challenges but I'm less certain if they will be able to meet their projected schedule.


It took SpaceX 19 attempts before they landed their Falcon 9 rocket. By all means, they managed to land starship during their 3rd attempt, even if it blew up 8 minutes later due to a hard landing.

Rather fail fast, than fail slow. SpaceX takes this to the next level of space hardware testing with their willingness to fail. Good on them.

Ps: it is a prototype (i said it!). It's expected to fail at some point. Otherwise it would be production hardware...


19 attempts to land it from orbit, with many of those being onto boats. They had lots of "easy" success with low altitude test flights, and a booster isn't ever going to have people on it. this will.

going to be tough, hope it works!


The first stage of the Falcon 9 does not go to Orbit.

These test flights might be low altitude but the flip and landing is realistic.


I don't understand how the belly flop is even possible.

I'm far from a rocket scientist, but I thought that rockets were kind of like soda cans - very strong on the long axis, but not at all perpendicular to that.

If you make them strong in both directions, I wonder how that doesn't make them prohibitively heavy.


Well, sure that is generally what is the case. But of course the designer can design it as he likes depending on the situation.

Generally, of course this is more true when the stages are empty. If the tanks are pressurized, as they are during the whole flight, they actually become very stable in both directions.

Starship is specifically designed to survive a reentry (heating) into the atmosphere at an angle, just like the Spaceshuttle.


> Restarting an engine in the middle of a rotation like that, fuel sloshing around is really hard. Just restarting an engine in zero G has always been tricky, because the fuel is floating around and doesn't have gravity pulling it to the pump.

I was re-watching a series of videos doing a walk-through if the Apollo guidance computer over the weekend and one of the cute 'yes, they thought of that' bits that was mentioned was that before the main firing sequence begins on the LEM descent they perform an 'ullage' burn using the RCS to settle fuel in the tanks.


I think its a bit early to say its not viable. Header tanks are there precisely to solve this. The sloshing in the header tanks should manageable in a tank sized correctly, specially if you are willing add some extra baffling to the header tank.

We should realize that after the first failure SpaceX already had built a couple of prototypes that were 'wrong'. So they had to basically do hacks to figure out how to fly them at all.

I would assume with the next generation they have really thought about how to do the header tanks and redesigned them and how the pressurization of those work. They are seemingly learning about some interesting things about liquid methane and bubble formation in liquid methane. The baffles they had apparently caused a vacuum bubble in SN8.


> fuel sloshing around is really hard

Can't the fuel be stabilized mechanically, e.g. as in a syringe?


In theory yes, and if worse comes to worse that is an option.

But its not what any rocket designer would want to do. Having a mechanical system inside of a deep cryo liquid methane is hugely complex and very heavy. Even if you put it outside, you still have deep cryo liquid inside and that would make it even heavier.

Consider that 1kg extra weight on the second stage reduces payload to orbit by 1kg and payload to Mars by much more overall.


How about rotating the rocket so the fuel is pushed outward? The gyroscopic effect could also have a stabilizing effect on the rocket itself.


Is there some kind of material that remains elastic and strong at that temperature that could be used to create a bladder? I assume no, since that seems like something that would have already been done in place of helium pressurant if it were viable. All the rubbery or fabric-like things I know of that are stretchy like that become brittle at those temperatures.


I don't know if such a material exists. The problem is SpaceX operates with Deep Cryo fuel (only company that does so as far as I know). They are literally going to 66K on the LOX and methane.

Helium was just temporary as well so once the actually design system is back with SN15 it should be better already.

The idea that Starship has is to use the fuel and oxidizer in its gas form. So the methane tank contains liquid methane pressurized by gaseous methane.

The benefit of this is that if the engine ingests some of that gaseous methane its less of an issue in terms of engine thrust compared to helium and its possible more recoverable of a situation.


Do you mind if the material is only elastic and strong at that temperature? Suppose that perfluoroheptane is rubbery at those temperatures. (the melting point is -78 °C, which is 195 Kelvin, and SpaceX oxygen is kept at 66 Kelvin)

It works for Mars rover tires too. Want rubbery tires? Use something like perfluoroheptane. Keep the vehicle cold during the trip, or mount the tires in disposable molds that wear away during use.


Not your line of thinking, but I believe ferrofluids were developed to provide control over fuel in zero-g (i.e. a material that is elastic and strong, but that gets mixed with the fuel instead of encasing it)


Doesn't it then get burned? Seems like it would deposit shit in the engine, which is bad especially for a reusable engine.


I believe it's meant to be fully filtered out of the fuel before burning (being magnetic, I think the filtering is a bit easier than some other non-reactive contaminate would be). Most of this is only half-recalled from someone's recent project that showed separating ferrofluid from some liquid on Earth...but I don't recall any other context :/


I've heard of using a gas driven plunger in fuel tanks, but I can't remember the name of the vehicle that did this. Starship uses small "header tanks" that have a similar effect. Weight and reliability are big priorities in this kind of system.


Isn’t helium already used to push fuel into the pumps?


I think it's more to maintain pressure in the tank, in cases where the remaining fuel doesn't vaporize sufficiently on its own (autogenous pressurization). Otherwise, I guess atmospheric pressure or compressive force from the engines/payload could crush the tank. In the case of Starship, the fuel is moved by turbo pumps, which are integrated with the engine, and are powered by burning some of the fuel (there's a diagram of this at https://en.wikipedia.org/wiki/SpaceX_Raptor).


I think that was part of the problem with the last launch. it started to burn some of the helium which caused issues.


I thought they used surface-tension / capillary-action to move the fuel. Like with passive geosatellite heat distribution.

Think a radiator, lots of tiny fins, but all inward pointing in a tube. Coat the fins so they are '-philic' to the liquid. Get creative with the geometries of the fins. Surface tension / capillary-action spread over a lot of the fins in the tube will move the liquid to a sink, like the combustion chamber.


I think they picked a very hard and complex task, and the hype around it is so high that everyone just assumes they will succeed.

The procedure might be unworkable, the engine unreliable, or not reusable enough, tons of things. I think such an ambitious project is always a “maybe” until it’s done. But not in the general perception.


Or (3) the first full-flow staged combustion engine to have ever really flown will require more iteration time to become reliable.

Seems totally reasonable to me. It's a new design of a new type of engine.


It's only been a few months since the first landing attempt, so it's not surprising that it's not working yet. Number of attempts is not as relevant as time since time is needed to find and implement fixes.


Darn! Certainly a set back, if a brief one— they're astonishingly building new Starships every few weeks now. It was almost exactly 15 years ago that SpaceX attempted their first (albeit unsuccessful) launch of the Falcon 1.

Also, as of their 25th Starlink mission last week they now have 1320 satellites in orbit. Approximately ⅓ of all operational satellites are now Starlink relays, which is kinda mind boggling.


It's pretty wild to see trains of these Starlink satellites pass overhead. I've been using this tool to see them travel over our area every few days: https://james.darpinian.com/satellites/


What a ridiculously useful and easy-to-use site, thanks!


Very cool, I've always wanted to watch the NSA watching me!


The NSA has proper camera satellites. These are for WiFi only, the 4 hubbles are much bigger and for view. If weather permits.


I don't think it is a setback. They are building them all the time and they are iterating very fast. I don't think they are close to reusing them so they are just interested in seeing what works and what doesn't and then iterate on the design in subsequent launches.

I wouldn't be surprised if this Starship didn't incorporate all the improvements they learned about during last few launches (and booms).


The next one (in a few days) is SN15, which is a major upgrade with "hundreds of design improvements across structures, avionics/software & engine.", according to Elon.

"Next major technology rev is at SN20. Those ships will be orbit-capable with heat shield & stage separation system. Ascent success probability is high.

However, SN20+ vehicles will probably need many flight attempts to survive Mach 25 entry heating & land intact."

Surviving the orbital entry and landing is probably the most difficult challenge yet. After that the next major challenge is landing on Mars.


According to the article, they're probably going to skip over SN12, 13, and 14 in order to try with a later design. That makes sense, since it's likely that several crashes happen between when construction starts on a new rocket and it finishes. Once it's done, they may have design might already be out of date.

> Fortunately, as part of its iterative development program, SpaceX is building a Starship vehicle at a rate of every two to three weeks. So the next prototype—likely to be SN15 as the company skips ahead toward a more advanced version—may be ready to fly in several weeks. Still, SpaceX would have liked to have gotten this one back in one piece.


Reminds me of the movie October Sky. I always laughed at the launching sequence where the rockets keep exploding.


A big difference is that at that point, they weren't going for the rapid iteration and we know we'll fail some.


I don't see how this is viable the way it lands right now. One failure and you have total loss. With a parachute or wings you can have some failures and still land safely.

Is there going to be an ejection system in case of failure?

EDIT: Instead of just down voting my valid question you could please provide me with an answer so I can understand how such issues are going to be resolved. If the flip maneuver fails I don't want to die.


But isn't there some redundancy of the parts at least? Of course you can't recover from a failed flip, but having more engines than necessary means some failed ignitions can be tolerated.


There is not sufficient atmosphere on Mars to use either a parachute (not enough drag) or wings (not enough lift), so neither of those options work.

Vertical landing of the rocket seems to be the only rapidly reusable architecture for landing on other planets - the primary purpose of this rocket (eventually).

I'm not sure about other failure handling, you ask a good question and I don't have a great answer. But the mission requirements suggest that vertically landing the rocket is the only feasible landing method I think.


What was the backup for the shuttle?

What was the backup for landing on the moon? What about Mars?


The shuttle has wings, so it doesn't just drop out of the sky.

The moon lander didn't perform a complex flip maneuver.


The moon lander had many complex things, including at the last minute. Remember in space they had to connect up, and separate. Their computer got overloaded when approaching the landing site, and the astronauts had to decide what if anything to do about it. Then they were going to land in a boulder field and the pilot had to take over and go to another place - they had to decide what to do at the last moment. And they almost ran out of fuel before landing (or had little margin left). Each of these things could have lead to loss of the mission.


And neither will Starship when it lands on the moon... it'll do the exact same thing as the LEM. The flip is due to Earth's atmosphere. And there are no runways on Mars, so you can't make a shuttle for it, so you don't have any options left.


I think the current plan with Starship is to not have an abort system, but maybe they'll change their minds between now and when they're ready to start thinking about putting humans on board.

https://everydayastronaut.com/starship-abort/


It only needs one or two of its three engines to land.


I think the flip maneuver is a big issue as it happens near ground and if it fails there is not time to recover irregardless if there are other engines that could perform it.


The flip starts with 3 engines and should work even if 1 of them doesn't start.


It still failed on SN9.


So a single failure of a prototype means its never ever ever gone work? What kind of logic is that?

Also you are wrong. SN9 actually used 2 engines for the flip, not 3.

They only swapped to 3 for SN10 and the flipped worked perfectly there. The problem with SN10 was the temporary helium system they put after SN8 in so the engine was ingesting helium.

The point is that you have engine out capability both on the flip and during landing. Of course over time SpaceX need to test some of these profiles in more detail, just as they did for Falcon 9 where they tried various ways of landing.


Of course not. But you can't assume that anything will work 100% of the time. So what will be done if in a rare case the flip maneuver fails? Engines fail but a plane can still land with just one engine. If we are going to be doing intercontinental flights it needs to be as safe as flying.


What happens to a plane if the tail falls off?

What happens if all the engines fail?

What happens if it flies into a mountain?


The Wright Model A crashed, and actually killed someone. That's why aircraft are so unsafe.


they weren't trying to light all 3 for SN9.

now they light all 3 and turn 2 off when one lights.


They can do the flip at a higher altitude to provide more time for recovery.


I just realized I don't know how Starship, or the Falcons for that matter, "know" their position prior to landing.

Orientation, I expect, would be determined by some kind of gyro (ring laser gyro? or super precise accelerometer?). How about altitude? What is the feedback mechanism(s) to know where the ground is in the moment before touch down?


In addition to radar, gps, gyro.. they likely also have terminal guidance radio very similar to what airports use for guiding airplanes to the runway in fog/cloud. See Instrument Landing System. They may use active radar reflectors, and if they do, I'd say that's a different form of the same basic idea (using fixed radio sources to calculate an expected landing position.)


>What is the feedback mechanism(s) to know where the ground is in the moment before touch down?

Radar


I'd imagine a wide variety of inputs sources are all being fed into a Kalman filter or something similar. Certainly they're going to be using gyros but also radar, GPS, and probably other things.


> SpaceX loses another Starship prototype

Not true. They know exactly where most of the pieces landed!


Can anyone break down the economics of this? How can they afford this over and over again?


Stainless steel is really cheap, and they adapted fabrication methods from the grain silo industry.


The stainless steel is irrelevant.

They lose 3 raptor engines per crash and that hurts. A lot.


Not sure if they are at v1.0 yet, but "Raptor cost is tracking to well under $1M for V1.0" [0]. Assuming they are, 3 million dollars worth of engines per test isn't that much.

[0]: https://twitter.com/elonmusk/status/1179107539352313856


I would like to know too. When sn8 crashed I googled and saw a quora article saying it costs $180 mil for each prototype. But this article says ~10 mil. Im assuming to ~180 mil number is the production version?

https://qr.ae/pGRZv0

It makes sense at $10 mil since they are pumping these out like mad.

Also NASA awarded $135 mil for the starship design so it can’t cost $100 for each of these sn’s right!?


> Im assuming to ~180 mil number is the production version?

The best assumption would be that the author of the $180M figure has no idea what they're talking about and derived at this number from a lack of information and through bad analogies and misapplication of scaling laws.


A quick google search says that SpaceX charges 62 million for a typical Falcon 9 launch. Let's say they have a 50% profit margin (no idea if that's true, but let's go with it). They had 26 launches in 2020, so that's 806 million in profit they can redirect to R&D on Starship (in addition to loans or VC funding rounds or loose change found in Elon Musk's couch).

The Starship prototype cost is a big variable, but assuming it's somewhere in the (log scale) middle of the 10 to 180 million range it's not unreasonable to think that SpaceX can afford to keep producing these things at a rapid rate for quite a while without going broke.


They must be willing to foot the bill for testing beyond a certain amount of the NASA award (40 million of the 135?).


I don't believe the $10 mil figure for one second.

That'd mean a Raptor engine would cost less than $1 mil and that just can't be true - that'd be about half the price of a much smaller and less complex Merlin engine...


They're heading towards that $1m per engine, but that price includes development effort. The cost to make 1000 engines is not going to be 10 times the price of making 100 engines, the more that are made, the better they get at making them


Heading there doesn't mean the engines they're currently using are already at that point.


We have no idea what the marginal cost of the engine is, but it's far lower than the total cost of developing an engine, and the more they build the closer they get to $1m, $500k and towards the $250k target

But even if it were $2m, there's 3 engines, that's $6m, leaving $4m for a bit of steel out of that $10m budget.


Let me do some guesses.

Well, they are shooting at a per Raptor cost of 100k long term 250k short term. Even assuming it is 4x as much now, that is about 3m for the engines.

The stainless steel only cost a couple 100k$ at most.

Add in some bits and pieces, like flight termination system, thruster system and tanks, sensors and so on, add another couple 100k$, lets say 1 million. All of these are already produced by SpaceX in pretty large numbers.

Then manufacturing cost, is a bit hard to calculated, they are producing one prototype every couple of weeks and most of the labor is in welders and most of the machines are welding robots. Labor cost being the dominate cost.

A large part of the cost would likely be the launch site construction itself, and how you want to amortize to an individual prototype is hard to say.

Then there is another bunch of cost for operations and fuel add another couple 100k.

So I would assume a cost of 3 million to 20 million. I'm giving this a pretty large range because its a guessing game. Over the course of a year they tested about 10 increasingly complex prototypes. That would at the high end be a cost of 200 million for one year of this process, if we ignore the design cost.

Consider that NASA SLS has a budget of about 2.5 billion a year including ground system (and that has been going on for 10+ years and will go on for many more), SpaceX is basically nickel and dimming a launch vehicle of that size.

Elon Musk has a couple of saying along the lines of 'if its long its wrong' regarding development. Spending 50% less time while blowing up prototypes regularly is cheaper then spending longer on the project overall. A lot of the things SpaceX does is novel, so no matter how much design work, nobody has ever flipped a deep cryo methane tank and started a full flow staged engine with torch igniter.

Working the perfect design for this on paper would take incredibly long and even then you might have to start over again and once you finally test it.

This is certainty an expensive program, but compared to Starlink is seems less insane in terms of cash burn. Doing both together is a bit extra insane, but on the other hand if you have Starship, the cost of Starlink drops to pretty absurdly low level.


> but compared to Starlink is seems less insane in terms of cash burn

Starlink has an insane cash burn during deploy, indeed, but after that it's a cash cow. People are paying through their noses for shoddy satellite internet, and Starlink is orders of magnitude better quality and way lower priced...


> Well, they are shooting at a per Raptor cost of 100k long term 250k short term. Even assuming it is 4x as much now, that is about 3m for the engines.

That's some super optimistic numbers you chose there. There's no reason to believe that at this point in time a Raptor engine is only half as much as a less complex Merlin engine.

The loss of three engines per test is by far the most costly part of the operation. How much an engine costs right now, I don't know of course, but I'm very confident it's more than $2 million a piece.


Musk himself said the cost was <1M$ in 2019: https://twitter.com/elonmusk/status/1179107539352313856

And in the next tweet, he mentions his concern is about "secondary structure", which I take it to mean all the infrastructure that you would need to build and operate to make that rocket fly.


He wrote that they were well on track for <$1 M for V1.0.

Find me a tweet where he actually states that Raptor is at V1.0.


I do not believe the $2 million for Merlin. That number was der At peak production rate Merlin was less then $2 million, I think $1 million or maybe less be now. Or maybe a little more again since production rates have likely declined a bit.

Raptor already seems to have a production rate comparable to Merlin.

And btw, in my final estimate I gave a larger range. Even assuming 3 million per Raptor, that would fit within the 20 million I mentioned.


It will be interesting to see if they push Superheavy/Starship into production before they've completely mastered the landing/re-use of the second stage (or even the first stage)? They might as well get paid for some of these launches while they continue to refine their design, like they did with Falcon 9. Starship is supposed to be able to carry 240 Starlink satellites at a time, that alone might be worth pushing ahead.


They recently raised $1.9B - that will buy a whole lot of grain silos... and quite a few fancy rocket engines, too.


So when will it be regarded as working ???.

10 successful landings or 20 or 50 ???.

Apollo did 6 launches (bar the first one - cabin fire) before it was manned.


Starship is being developed in a very different way to Apollo. Apollo program's first launch was October 27, 1961, the first crewed launch was October 11, 1968 - 7 years later.

Even if you start from the first Starhopper launch, that gives SpaceX until 2025 to be technically ready for the first crewed launch.


Is it my imagination but it seems so many of these failures are not recorded (or have really bad video quality / cut in-and-out).

Is it correlated to the success of the mission or just a coincidence.


No, it's just you. SpaceX and its fan community has pioneered real time coverage of their program. There were a dozen publicly available cameras on today's launch (almost all were covered with fog of course, but we still got to see bits of spacecraft raining down), and spacex show live high quality camera feeds from all their missions -- that's why so many people are so excited by spacex.


Starship launches are documented very well. Tons of live coverage from dozens of cameras, both from SpaceX and third parties. There's plenty of great footage of the prior explosions. Today happened to be foggy, but SpaceX doesn't control the weather. In fact they wanted to launch yesterday but couldn't because the FAA didn't have its act together.


According to the Washington Post's space reporter it's not really the FAA's fault. https://twitter.com/wapodavenport/status/1376668877699047424


I'm not surprised that's a Washington Post reporter's take, but given the same information I have a different take: this is very much the FAA's fault. This is a billion dollar development program and the FAA ought to be able to get someone in state for oversight, or allow remote oversight in cases where they can't get their guy there promptly (because, don't forget, of a different government bureaucracy, not SpaceX). Also they should be easier to reach.

By saying that SpaceX "must provide adequate notice of its launch schedule", they are basically saying that it's unreasonable for anyone to try a fast-paced space development program. SpaceX has to slow down so the FAA isn't inconvenienced by having to hire a local inspector, or having the inspector wait idle a few days, or keep their phone on and travel on short notice, or figure out remote inspections. Which is just the epitome of government bureaucracy.


Seems SpaceX are doing a mea culpa after dumping debris over the town in the explosion:

https://twitter.com/elonmusk/status/1376903913564160002


Fake news. Debris was not dumped over the town. SpaceX wants to make Brownsville nicer because they are hiring thousands of employees who will live there.


I saw a video of someone grabbing a piece of freshly landed insulation but I guess that could fly a lot farther than the more dangerous stuff.


Isn't it the FAA's fault for requiring an inspector on scene before every launch? There's simply no way to iterate quickly with a constraint like that, especially if the inspectors are in Florida and need to be flown out each time.


The logical thing would be for SpaceX to fund a full time on site inspector working for the FAA

I have no doubt spacex would love to do that so why isn't it possible?


> Washington Post

We'd do well to keep in mind that the Washington Post is owned by Jeff Bezos, who also owns Blue Origin, a rival spaceflight company. That doesn't necessarily mean that the reporter's information is false, but he's potentially biased.


There is no evidence that this is the case. Can you cite an example that is not Trump saying the "Amazon Washington Post"?


Conflicts of interest cannot in general be proved or disproved to have influenced any person's behavior (short of an explicit quid pro quo agreement). For that reason, the safe assumption is generally to assume that if money is flowing from one person or organization to another that a conflict of interest exists and that it is very likely to be influencing behavior. In this case it means that it's reasonable to regard what the Washington Post says with some degree of skepticism. The best way to resolve conflicts of interest is to not have them in the first place. Unfortunately, if the Washington Post's continued existence depends on backing from a rich patron/owner, then their credibility is a little bit less than it would otherwise be on subjects that overlap with Jeff Bezos's interests. That doesn't mean the Washington Post is wrong, just that they deserve extra scrutiny.

At the risk of letting politics intrude into the discussion: the Washington Post also exhibits a conflict-of-interest pattern of behavior elsewhere. Bernie Sanders is a vocal critic of Amazon and their treatment of workers, and the Washington Post is often harshly critical of Senator Sanders. Again, there's no proof that Jeff Bezos is influencing the editorial content of the Washington Post, but there's a pattern of behavior and a financial incentive.


> There is no evidence that this is the case. Can you cite an example that is not Trump saying the "Amazon Washington Post"?

Huh? What does Trump have to do with Bezos owning the WaPo?


> but SpaceX doesn't control the weather

Yet




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: