On the 'repeated failures' piece, that's somewhat new for aerospace. Traditionally, things would be over-engineered, and losing a test vessel would be considered a bad thing. SpaceX seems very content to blow up a bunch of rockets on their way towards not blowing up rockets.

This seems to have caused at least part of their recent issues with the FAA, who seem to be less ok with explosions. It will be very interesting to see if this iterative approach leads to more reliable rockets down the line!

It’s really a throwback to the old days of aerospace. From the 1930s to the early 1960s, the standard method was to iterate rapidly with relatively cheap hardware, to learn and fix problems as quickly as possible. The von Braun and Korolev teams operated this way, among others. They all blew up a ton of rockets. What changed was the necessity of completing Saturn V and landing on the Moon by the end of the 1960s. Apollo program head George Mueller recognized that, with the planned number of iterative tests, they were going to be too late. He made the gutsy decision to launch “all-up”: spend a ton of money up front to overengineer the crap out of the vehicle and cut out most of the iteration. It worked!

But alas, the aerospace world overlearned the lesson and has been cargo-culting it ever since. Mueller’s decision was purely a matter of expediency to meet Kennedy’s goal and beat the Russians. It wasn’t a statement that this was the best way to do engineering, absent those constraints. To make matters worse, all-up only works under the funding conditions of Apollo: a massive spike in funding to cover the up-front cost. Without that spike, the costly development phase has to be stretched out, and the whole thing ends up taking longer than iteration. In other words, you can save time this way when money is absolutely no object, but otherwise you lose time. See Boeing’s SLS: basically 15 years of development and they just static-fired it for the first time. Launch is maybe a year away.

We’ve all gotten used to the all-up style of development, and SpaceX’s rediscovery of rapid iteration makes it seem like a new and untested way of doing things. But I think the dominant narrative—that all-up is prudent and conservative while iterative is risky—is completely backward. All-up carries a massive amount of risk that fundamental design issues won’t become apparent until the tremendously costly development phase is done. Iteration allows assumptions to be tested and modified as quickly as possible. Most of SpaceX’s early ideas about reusability turned out to be wrong. If they had committed all the money and time to those early concepts before flight-testing, they would have never accomplished it. The early rocketry pioneers operated this way, and it’s why they accomplished the things they did. Maybe SpaceX will finally free the space industry from perpetually repeating what worked for Apollo, which after all operated under unique political conditions.

I’m not sure 3200 full scale tests of Saturn V’s F-1 engine, 2000 in an iterative trial-and-error scramble to suppress combustion instabilities is what I would call “overengineered to cut out iteration” during Apollo http://www.yang.gatech.edu/publications/Journal/JPP%20(1993,...

There was of course a ton of testing at the component level, and iteration when necessary. This worked for Saturn V because it could be done for each component in parallel, due to the huge mid-sixties spike in funding. But at the level of the integrated system, most of the testing—and nearly all of the iteration—was cut out, since it would have to be done in series for the entire rocket (or its stages). Thus the design had to be set in stone before the vehicle ever flew. The first and second stages of Saturn V flew just twice before crewed flight began. This was the right process for Apollo’s unique set of requirements and resources, but it is a bad idea when you A) don’t have a hard deadline forcing you to forego system-level iteration and B) don’t have a spike in funding to accelerate the up-front development, which otherwise takes forever. Hence SLS.

Granted, I am biased towards propulsion, but we are not talking about some bolt or bearing as a 'single component'. The propulsion system is a complex thermodynamic cycle, with tanks, turbopumps, preburner, regenerative cooling, injection, engine, nozzle, etc. Seems to be a rather involved 'component'.

That's just one engine - much less than a 5-engine stage even on a test stand, not to say about flying.

And combustion instabilities they were fighting were rather novel at the time, the engineers basically had no other options than to test a lot of variants - and invent debugging techniques with "bombs in the chamber" along the way. Fortunately it worked - low pressure of F-1 helped to mitigate problems of size.

I don’t understand your first paragraph - is this related to the F-1 development having strong iterative components?

Also, Combustion instabilities date back to at least the A4/V2, not quite new, and largely the same engineers.

The F1 is a bigger (!!) version of the same engine. It's got the same crucial features:

1. turbo-pumps

2. the nozzle is cooled and the fuel warmed by making the nozzle out of tubes through which the fuel passed

3. holes drilled in the tubes so fuel leaked into the combustion chamber to provide boundary layer cooling

I think the pogo-ing was solved the same way, too - putting baffles in in various places.

By that logic essentially all cryogenic engines are “the same”.

In a similar vein, all modern jet engines can trace their core features back to the Ohain engine (not the Whittle engine).

I agree that Apollo set the stage for a lot of things in aerospace. It's worth noting that the death of the entire Apollo 1 crew actually led to less iteration in the late stages of the program.

The acceleration of aerospace development costs started long before the Apollo program.

That is definitely true. I do think Apollo was the turning point when rocketry changed from a development process focused on rapid iteration (seen as late as Redstone-Jupiter-Juno) to one focused on a lengthy design phase (and component testing) with late and limited integrated testing. The roots of the shift are probably in the ICBM world, but Apollo was when civilian rocketry (over)learned the lesson.

The FAA seems to be concerned with exploded bits falling on innocent people. You are welcome to blow up your rockets all day long if you aren't putting other people at risk.

The SpaceX rockets don't fly anywhere near people. Nobody is underneath any part of their flight path or within any possible destruction radii.

The early days of the Soviet space program were similar - fast construction and iterative improvements with each successive test.

Soviet rocketry was developed in the artillery department. There iterations are normal. US rocketry was developed under Air Force - here more experienced pilots, lots of training...

Plenty of American rockets failed, especially early on! https://www.youtube.com/watch?v=13qeX98tAS8

And SpaceX has blown up ten Falcons on landing attempts.

After accomplishing their primary mission, which is a key distinction. If you blow up on landing that just means you lose your reuse discount on a future flight.

That's no good for Starship, which is meant to land with passengers and hardware on board.

Yeah, which is why they're making Starship better than the Falcon 9 in being able to land safely even when things go wrong (i.e. Starship can throttle down enough to hover). They're still in the early prototyping stages. You can't judge potential long-term reliability based solely on what we've seen so far.

*While building the first space rockets in the world, literally making it up as they went.

That's not quite the same situation SpaceX is in right now.

Actually it is the same situation. The launch was fine. What blew it up is something nobody has attempted before.

It's totally different because only the government is allowed second chances! Everyone else has to get it right the first time. Especially if they're competing with <insert defense/aerospace contractors here>!

Never attempted what? Relight an engine? Engines are routinely re-lit, especially on upper stages, yet Space-X gets it wrong with regularity. The Russians and Arianespace do it just fine.

I'm sorry, what? SpaceX are the undisputed champions of re-lighting engines. They've landed orbital rockets 73+ times, each of which involves multiple engine re-lights while falling through the atmosphere. Let alone gets it wrong "with regularity" — the Falcon 9 Block 5 is a remarkably reliable rocket.

Anyway, landing Starship involves a bunch of things never before attempted: landing a fully-reusable second stage, landing a vehicle of this size, relighting and running engines during the belly-flop-and-flip maneuver, and flying and relighting full-flow staged-combustion engines. Probably some other firsts too.

Today's crash was due to failure of one engine to restart. Briz and Fregat Russian upper stages are routinely restarted, with ~ 98 % success, same with Ariane EPS and Chinese Yuanzheng. It's doable, just not for Space-X.

You realize the Raptor engine is different from the Merlin used in Falcon rockets, right?

You're comparing a prototype against decades-old, non-reusable rockets.

Briz and Fregat start much smaller - 20 kN - engines with no time constraints, while Starship re-lights 2200 kN engine and rocket immediately does rotation by 90+ degrees, and time is of essence.

Where do thrust and chamber pressure even come into the equation? It's not that the engine exploded, the thing failed to ignite.

> Where do thrust and chamber pressure even come into the equation?

Thrust given to show difference in engine scale. Pressure isn't shown - what do you mean? :)

The point is that re-light of Briz and Fregat is very different and so hard to compare.

Saying SpaceX gets it wrong regularly is LITERALLY false. SpaceX Falcon 9 is fully qualified for all DoD Orbits and has reached highest qualification for NASA Scientific Missions and Human mission for. All require demonstration of reliably relight with incredibly high reliability.

In fact, in over 100 flights, SpaceX did not have a single failure of a Second Stage engine not starting or not re-lighting. This is unlike Arianespace who just had a failure when they tried to start an engine on one of their upper stages.

Outside of that Merlin is easily the engine that can be re-lit more then any other on the planet and its not close.

This however is a non-commercial prototype, where they are testing new technologies. Its a completely new engine of a type that has never flown before and is not finished developing. The vehicle has been newly designed and it does things no other vehicle has done before in a way no other vehicle has done before.

You honestly just sound like an incredibly petty hater. I can grantee you that Russia and Europe have plenty of issues with engine starting on their development engine.

Why is SpaceX making the same mistakes 70 years later?

I don't recall NASA ever doing a rocket landing so I'm not sure which mistakes you are seeing repeated.

I was surprised to learn recently that NASA did do some rocket landings in the 90s: https://en.wikipedia.org/wiki/McDonnell_Douglas_DC-X

Ok, so which mistakes is SpaceX making that NASA also made?

NASA and the USSR did many powered rocket landings, actually. NASA even had a program for reusable first stages via powered landings that was scrapped because automatic guidance wasn't feasible yet.

Ok, so which mistakes is SpaceX making that NASA also made?

Failing to restart an engine in a retrograde trajectory. A difficult task that both the Soviets and NASA (and Arianespace and the Chinese) had issues with, but worked out.

It's very difficult because of the turbulent airflow in an engine bell pointed against the airspeed vector. Despite this, many rocket engines can do it reliably.

> Failing to restart an engine in a retrograde trajectory

That's not a failure, that's the effect of some upstream failure.

What failure are they repeating? Pumps sized too small? Improper torquing? You talk like you know the details, so share them.

There are a lot of reasons for which an engine would fail to start in such a scenario. They all boil down to the turbulent airflow and high pressure in the nozzle inhibiting the ignition process.

How exactly you decide to fix it depends on many things. It's impossible to tell. But they are all downstream of the same failure, which is overcoming the high pressure and turbulent flow.

NASA did rocket landings on another planet with 1970s technology: https://en.wikipedia.org/wiki/Viking_program#Entry,_Descent_...

>At an altitude of about 1.5 kilometers (5,000 feet), the lander activated its three retro-engines and was released from the parachute. The lander then immediately used retrorockets to slow and control its descent, with a soft landing on the surface of Mars.

Ok, so which mistakes is SpaceX making that NASA also made?

The same outcome does not require the same cause.

>> On the 'repeated failures' piece, that's somewhat new for aerospace

Unless you go back to the early pre-Mercury program days of NASA - they had lots of failures, and that was just trying to go up without re-usability, not land and reuse. There's all kinds of fun failure reels (e.g., [1]) and lists (e.g., [2])

[1] https://www.youtube.com/watch?v=g79K-R7xTFo [2] https://en.wikipedia.org/wiki/List_of_spaceflight-related_ac...

You're basically watching the CI/CD workflow for the best aerospace organization in the world, live on YouTube.

> that's somewhat new for aerospace

Because NASA does not launch anything anymore. If they were doing hundreds of launches per year you can bet you'd see failures the whole time.

SpaceX themselves made a failure montage for Falcon 9

https://www.youtube.com/watch?v=bvim4rsNHkQ

It's easy to forget how many failures that had, but look how far it has come, at this point I don't think they've had any major failures in 5+ years. Maybe a couple sea drone ship booster landing issues but zero mission failures in a very long time.

This is one of my favorite videos. On the SN9 explosion today my partner speculated that SpaceX should make one like this for Starship one day. :)

Neat, seems pretty much any mistake results in a fireball.

Scott Manley and daughter have a fun song about other failures, "You will not go to space today", https://www.youtube.com/watch?v=Ayu0GsrvKQA

I'm guessing the title is probably an allusion to this https://xkcd.com/1133/.

I was half expecting:

"If an O-ring gets a bit too cold, you will not go to space today"

...but I guess that's a bit too dark, eh?

"If your rocket runs at all engine rich, you will not go to space today..."

And after every crash I have to read this comment.