Was Exploding Starship SN8 a Failure?
The 8th prototype of Starship built by Elon Musk’s rocket company SpaceX, exploded spectacularly on 9th of December. Does that mean Elon Musk screwed up?
The launch of SN8 was quite dramatic as SpaceX sent Starship to a higher altitude than ever before and performed the belly flop maneuver, which will be absolutely crucial for SpaceX to master if they are to land Starship on Mars or on Earth. Naturally this event caused quite a lot of spectacle, especially since the rocket exploded in a massive pyrotechnic show at the very end.
Judging by the media response this was some kind of failure. What was worth reporting seemed to have been that SN8 exploded, but was that really the story?
Prototypes are Not Final Products
Whether a rocket explosion represents failure or success really depends on what sort of rocket it is an what the goal of the launch is. SN8 was not a finished rocket but a prototype. A prototype isn’t merely an early version of something. A prototype is something designed to test one or more specific things of the final product.
E.g. the first prototypes of the Palm Pilot handheld computer was simply blocks of wood. How is that useful? Because the first prototype was going to answer the question: What dimensions should the Palm Pilot have? The prototype testing then involved multiple employees walking around with blocks of wood in different dimensions. This test would answer what dimension was most comfortable and practical.
Personally I have spent a lot of type prototyping and testing user interfaces just drawn on paper. Users would pretend to be pushing buttons on the piece of paper. The important part to get is that a prototype is not made to be the final product. A prototype is made to answer specific questions.
Thus to judge the Starship SN8 launch, we need to instead ask ourselves what questions did this prototype seek to answer?
- Does the design work for high altitude launches. Previously Starship had only done short hoops at low altitude. This was a test of whether it could go really far and whether engines and everything else would hold up.
- Would the famous belly flip maneuver work?
- Could they land?
Question 3 is not really that important as we know from SN5 that they can land. It was far more important to answer especially question 2. And this question was beautifully answered by SN8. They where able to gimbal (move and orient) the rocket engines so that the rocket could first flip over and they later straighten itself up for landing. They also got to test whether falling belly down was as stable as they had hoped, reduced velocity as expected etc.
These where the questions SN8 was made to answer, and it answered those questions beautifully. You don’t want to test these things on a fully working and far more expensive rocket only to realize that your approach doesn’t work and a far more expensive rocket blows up.
That is why we build prototypes. They are much cheaper to fail with.
But this brings me to the next issue. If everything is fine, then why do we keep seeing SpaceX rockets blowing up, but not rockets from other companies? Surely there must be something more to this? And indeed there is. One of the most obvious answers is that SpaceX is far more public and transparent about their tests and results. They show their tests and failures live. A more secretive company like Blue Origin only gives you footage of successful tests after they happened. Still that is not the whole story. To answer that we need discuss the different approaches and philosophies of rocket building.
Different Rocket Building Philosophies
Broadly speaking, we have what I would call the Werner von Braun approach and the Soviet approach to rocket building. Werner von Braun which was behind the American Saturn V, the former head of the Nazi V2 rocket program. A bit like a German caricature he followed a very meticulous approach where every component of the rocket would get built and tested separately before assembled into a larger piece which again gets tested.
This approach leads to much fewer prototype launches, which tend to have less chance of blowing up.
Read more details: Lean Software Development and Soviet Rocket Design
To Soviet approach in many regards also fit a Russian stereotype about winging it. It means rapid production of prototypes which would get launched quickly. This approach meant lots of launches and lots of explosions. It was a move fast and fail fast approach. E.g. there was 14 unmanned launches of the Soviet N1 moon rocket, and they got to blowing up four of them before canceling the project as the US got to the moon first.
Case in point is that the Soviets blowing up a lot of rockets was not an indication of failure or that they did not know what they where doing. It simply meant a different approach to development. Soviet approach was basically on integration tests rather than unit-tests. The idea was that you would learn more from testing the whole system, than by only testing individual components in isolation.
And in fact this approach proved superior as the Soviet had an early strong lead in the space race. In fact there is some chance they could have won if not for the fact that the Americans abandoned the Werner von Braun approach midway through the Apollo program.
NASA, Boeing, ULA, Blue Origin and many others still seem to be trapped in this old Werner von Braun way of building rockets. Blue Origin’s New Glenn rocket began development before Starship yet there has been no prototypes flown. NASA’s Space Launch System (SLS) began officially in 2011, but actually started much earlier as it was mostly reuse of technology from the cancelled Constellation program. Yet despite having been under development for longer time than the Saturn V, and is based on well established already developed technology, it still hasn’t made a single flight. Not even a prototype flight.
The SpaceX Development Approach
SpaceX follows a bit of a hybrid approach. In fact they do extensive testing of individual parts. However SpaceX emphasize testing actual working hardware over lengthy certification processes. There are several examples of how this differs in practice.
When SpaceX was building a facility to produce heath shields they did that by building a small model factory first. Once they got that working, they built a larger model factory but still not at full scale. In fact they iterated on several prototypes, one larger than the previous sorting out problems discovered on the way until the final factory was built.
This is what we in the software industry or car industry would call a lean process. You iterate quickly and fail fast.
The rest of the industry tends to follow, what we often call a water-fall approach. That means meticulous planning, writing lots of thick specification documents. Lengthy reviews and discussions of every step before the full sized heath shield factory is built. In other words there are no experimental stages in between. There is no iteration. There is just lots of meticulous planning.
The problem with this approach is that reality is never that neat and there are very valuable lessons from getting your hands dirty by dealing with physical reality. SpaceX e.g. is very specific when they hire people. They primarily hire people who have hands on experience building model rockets. They don’t want people who can only do theory and write specs.
SpaceX does in fact not produce a lot of specs and documentation compared to the competition, but in contrast they produce a lot more physical hands on hardware that is tested in the real world.
This is a long winded way of stating that: Rockets blowing up is part of the SpaceX development process. They are launching incomplete rockets which are meant to test one or more ideas or designs they are trying to nail down. This feeds into the build process and design of the next prototype.
