On Thursday, SpaceX launched its first full version of the “Starship” rocket from a pad on the coast near South Padre Island, Texas. Four minutes into the flight the rocket was deliberately destroyed at an altitude of about 35 km, with debris raining down about a dozen kilometers out into the Gulf.
In many ways, the flight was a triumph. It was the largest rocket ever constructed. It had more engines than any other booster in history, edging out Russia’s failed N-1 Moon rocket. It was constructed largely in the open, in full view of fascinated space fans, instead of behind closed door in near “clean room” conditions. Observers watched as the Starship was welded together from rolls of stainless steel, rather than constructed from carbon fiber, aluminum, or other more traditional rocket materials. The engines for the Starship were developed from scratch in conjunction with the rocket, trading more usual fuels for methane and reaching extreme levels of efficiency.
Everything about the Starship was the result of a series of decisions designed to make spaceflight cheaper. The methane fuel. The steel structure. The method of construction. Even the rocket’s enormous size. All of it was a gamble to create a system that is fully reusable, bringing the cost of getting to orbit down to a small fraction of what it is today and making space almost infinitely more accessible.
However, one decision in the process didn’t just result in the destruction of the rocket, it generated a cascade of failures, one that’s likely to set the program back by a least a year, erasing the chance of NASA’s scheduled return to the Moon in the process. That decision is 100% on Elon Musk.
At both Tesla and SpaceX, Musk has frequently repeated his basic philosophy of design.
“The best part is no part. The best process is no process. It weighs nothing. Costs nothing. Can’t go wrong.” — Elon Musk
It’s certainly an approach that seems to bear some serious fruit. You can see if in Tesla’s cars, where a single touch screen has replaced all the buttons that would normally operate the radio, air conditioner, and everything else that is crammed onto the dash of other cars. Doing it that way greatly simplifies the construction of a Tesla, though it frustrates some potential owners, and is part of why right now that company is able to engage in a price war with other car makers while still maintaining a relatively high profit margin.
Ford loses money on its Mustang EV. Tesla makes money on the Model Y. And the Model Y is cheaper. It’s also much more successful. Right now, the Model Y is the best-selling car in almost every nation where it’s sold, including the United States, outselling such longtime stalwarts as the Toyota Corolla. (Though it’s still outsold by the F-150 and Chevy Silverado. Because this is America, where everyone needs a truck.)
However, you can also see the downside of Musk’s “the best part is no part” philosophy at Tesla. In its constant efforts to simplify construction, Tesla removed the radar that was part of the traffic-aware cruise control in earlier models, making them completely dependant on how their software interprets images from cameras. Can you say “phantom breaking?” If you drive a Tesla, you can. They followed up by removing even the ultrasonic sensors that were used to help detect nearby obstacles.
As a result of this, for some months buyers of new Teslas found themselves without park assist warnings, and even those who paid the ridiculous $15,000 Musk wants for “full self driving” found that their cars couldn’t do so much as the automatic parking that is available on many cars. Some of this has since been fixed, but there are remaining restrictions when compared to the older cars with radar and sensors.
In fact, the full self-driving, which is still far from complete years after Musk first said it was “two weeks away,” is likely a victim of this philosophy. Other car companies pursing this goal have employed both radar and laser-based lidar to characterize the world around the car. Musk has insisted from the beginning that cameras were enough and resisted any attempt to add other systems to the car (though it now seems that radars may make a return). Considering the effort, and dollars, that Tesla has applied to their autonomous vehicle efforts, it’s hard to say where they would be if Musk hadn’t crippled the effort on the hardware end.
And it was just such a stubborn refusal to add necessary hardware that not only doomed the initial launch of Starship, but stands likely to derail the whole project for months. If not more.
What went wrong?
Just over two minutes into the flight, Starship was reaching the point where the massive 33 engine booster stage should have shut down and handed over to the 6-engine upper stage (also, confusingly, known as Starship). In most rockets, a hydraulic or mechanical system is used to push the two stages apart. Not on Starship. Instead the rocket just pitches over slightly. That’s supposed to allow the two parts to gently separate, after which the second stage fires its engines and heads for orbit while the booster flies back to earth.
That didn’t happen. Instead, all of Starship stayed in one piece, rotating end over end, until the ship actually began to bend in the middle and SpaceX controllers on the ground were forced to press the flight termination button on the falling rocket.
So was that it? Was the missing system for separating the two stages the issue?
It may turn out to be an issue in future flights. But it wasn’t what went wrong on Thursday. To understand the problem requires going back to the first seconds of the flight, and to decisions that were made months ago.
SpaceX has referred to its elaborate launch table and tower as “stage zero” for their rockets, and it’s actually more complex, and more costly, than the rockets themselves. It has some special features, because it has to.
- Lifting arms
Because of the way Starship is constructed out of very thin steel, it’s structurally sound only when upright. It’s not possible to lay the rocket over on its side as is done with other rockets, including SpaceX’s Falcon 9 and Falcon Heavy. Getting the upper stage onto the booster also has to be done in the upright position. At NASA, the upper stages of the Saturn V were all stacked inside the Vehicle Assembly Building and rolled to the launch pad on the very slow, very massive, very expressive crawler. SpaceX brings the pieces to the pad one at a time, then lifting arms—using a motor designed for oil rigs—lifts the booster onto the pad, then lifts the second stage onto the booster. The pad is the assembly building for Starship.
Those same arms are also referred to as “chopsticks,” which is a pun on a scene from the movie “The Karate Kid” in which the wise karate teacher instructs the main character on how to capture a fly from mid-air using chopsticks. Unlike the Falcon boosters, which land on legs, the Super Heavy Booster from Starship is intended to eventually fly back to the pad, hover, and be caught by the same arms that lift it. So they have a whole system designed to allow the arms the range of movement necessary to catch the booster. This was done so that the booster didn’t have to carry the weight of legs and their associated systems. Is this the part that was removed which caused the problem? Nope.
The table on which the booster sets is quite elaborate, with a series of small water jets inside to help cool and protect systems, and a ring of hydraulic clamps that can hold the Starship down during test firing of the engine. That system is actual related to what went wrong. But we’re about to get there.
Put it all together, and “Stage Zero” is more costly and much more time consuming to construct than than even a half dozen Starships. The rocket is, after all, designed to be super cheap. Part of that individual cheapness comes from moving functionality to the pad. The booster that flew was actually booster 7. The upper stage was ship 24. SpaceX has rolled through that many prototypes getting to this point. Most of them have simply been cut up for scrap. But it has only built one launch tower.
Now, let’s walk through the actual flight.
The short life of Starship
After a hold at T-40 seconds, SpaceX seemed satisfied that all issues had been addressed and walked down to the launch. However, right from the beginning there was a bit of weirdness. The engines should have fired up at around T -6. They didn’t. Instead, it wasn’t until about one second before the clock hit zero that the first flames appeared. Then the clock actually hesitated for a moment before moving the other way. Five seconds later, Starship was still sitting on the pad. It would be 15 seconds before it cleared the tower.
What’s also visible on close inspection of these image is some absolutely enormous pieces of debris. It’s not just ice falling from the sides of the tank, its massive chunks of concrete sent flying into the sky, some of it coming up right alongside Starship. Images from other angles show large chunks of concrete flying to the waters on all sides of the tower. In fact, the reason that the cloud here is brown rather than the white usually seen at Cape Kennedy launches is because it’s not smoke. Most of it is sand, rock, and shattered concrete being hurled up from the ground.
Much of Stage Zero was terrifically damaged at this point.
Sixteen seconds in, just after clearing the tower, SpaceX rolled out a graphic that provides a lot of nice information about the progress of the flight. But one of the first things that’s visible is that three of the Raptor 2 engines on the flight are already out. That’s likely because they were damaged by rocks or concrete chunks flung upward while Starship was still on the pad.
Remember those clamps? On a rocket like the Saturn V, they would actually hold the rocket in place for a couple of seconds while the rocket came fully up to power. Then the clamps let go.
SpaceX did it another way. They opened the clamps before the countdown even began. Then they slowly throttled up the rocket on the pad. That’s why it took Starship, with twice the power of a Saturn V, almost twice as long to clear the tower. It just sat there for a lot longer, blasting away at the ground. That was all part of the plan — but it was also part of what doomed the flight.
A minute into the flight, Starship was approaching “Max Q,” the point of maximum stress on the airframe. The announcer certainly sounds happy in the video of the flight at this point, but the truth is that Starship is gaining velocity more slowly than expected, and is several seconds behind in reaching this critical point. Looking at the infographic, it’s not hard to see why.
That graphic shows four engines out, but there are good reasons to disbelieve the infographic by this point. Here’s a closeup on the rear of Starship just a few seconds later.
It’s not missing four engines. It’s missing six.
Starship is still going, but it is rising more slowly and gaining speed with more difficulty than it should. Which isn’t surprising, with 18% of its engine power missing. Not only have there been a number of engine failures, when the first images come from the rocket, they show that it is both rotating and also repeatedly changing its attitude. The first probably represents a loss of some flight control, possibly damage to a second stage fin or the ability of the booster to gimbal its engines. The second is probably software struggling to put the rocket onto the designed flight path, but failing because of those missing engine.
Also, the engine exhaust to this point has been very yellow — not the expected color for a methane rocket. Starship is likely not getting the correct balance of fuel and oxygen, possibly because of damage to lines or valves while it was still on the ground. In rocketry terms, it looks as if it is burning “engine rich,” eating up the metal of its own systems as it goes.
Two and a half minutes into the flight, Starship is nearing what should be main engine cut off and stage separation. Only this is where things go finally, terminally wrong. Because the main engines don’t cut off.
By this point, the infographic shows five boosters out, but as we’ve already seen, that graphic isn’t accurate. A look at the flame pattern suggests that Starship is now down by eight engines — it’s lost almost a quarter of its thrust.
Even so, Starship does begin the pitch over maneuver that should separate the two stages. That camera view on the left should show the first stage dropping away and the second stage kicking in. Only it doesn’t.
That’s because, even as Starship pitches, and pitches, and pitches, eventually going through a full 360 degree loop, the main engine just keeps firing. Down by eight engines, the rocket hasn’t reached the speed or altitude it was supposed to attain. It also hasn’t burned through the amount of fuel that it should have consumed. Some part of the booster’s software seems to be insisting that it has to keep going, even as another part has signalled time for separation.
That continued thrust from the first stage keeps the two parts of the ship pressed together. They can’t separate, because the first stage keeps on pushing and won't stop. The first stage keeps pitching, and thrusting, and by now Starship is in a full head over heels tumble. It’s also stopped gaining altitude, somewhere around 37km, and has begun to fall back. The first stage is still burning.
SpaceX lets it keep tumbling until it has fallen about a kilometer, then it finally opens the plastic cover no one wants to open and presses the big red button.
Here’s the TL;DR version of all of the above
- The no-clamps slow throttle-up meant Starship stayed on the pad for a long time, throwing up concrete, rock, and sand all directions, damaging the pad, nearby facilities, and Starship itself.
- By the time it left the pad, that debris had already destroyed three of Starship’s engines and likely damaged valves and systems that would lead to additional engine failures as well as an incorrect fuel mixture.
- Starship was slow to reach every point in the flight plan, suggesting that other engines were not able to throttle up to compensate for the lost engines.
- At what should have been stage separation, either software errors or more smashed hardware kept the main booster firing long after it should have shut down.
- The result was an uncontrolled spin that required Starship to be destroyed.
Why this is 100% Elon Musk’s fault
Starship is the work of hundreds of talented engineers and thousands of employees who put their best into making this thing go. The design is extremely daring, and something of a wonder. The engines are amazing, even if they have demonstrated that reliability is currently lacking. The whole system of construction promises to revolutionize the space industry.
But are two parts that were left out of Starship that absolutely doomed this flight and the decision not to include them falls right with the guy at the end of the first row at “Star Command.”
Those parts were not parts for the rocket. They were parts for the launch pad.
For some reason, Musk became convinced early on that he did not want the launch tower to have:
- A flame-diverter or flame trench to redirect the blast from the booster’s engines
- A water deluge system to dump a massive amount of water around the launch tower during liftoff
The launch facilities at Kennedy have both of these. Even the launch pads used for the much smaller Falcon 9 have both a flame trench and a water deluge. They help to protect not just the launch pad, and the surrounding area, they also help to reduce the noise. Which sounds trivial, but that noise is energy. That’s what broke up the concrete under the Starship Stage Zero, not the fire. That’s what sent car-sized chunks flying in all directions.
A flame diverter and a water deluge would have greatly reduced, or even eliminated, the damage to the area around the pad. They would have prevented the blow back of debris that damaged Starship before it even left the ground. It might have headed off the whole cascade of events that resulted in that button being pressed 4 minutes into the flight.
We don’t have to guess about whose decision it was not to implement these systems, because Musk already said he decided to skip these systems over the recommendations of his engineers. Musk even had a preview of what was going to happen, as past test flights of the upper stage also resulted in significant spalling of concrete structures and damage to at least one of the ships. He just made them try different kinds of concrete.
The parts for a water deluge were actually on site, ready to install, but Musk decided to forego that installation—likely so he could enjoy the pun of launching his super-joint on 4/20. Which was something Musk had joked about doing months ago.
Hopefully he enjoyed the joke, because the EPA and FAA are going to be thinking long and hard before they authorize another flight from Boca Chica. All those engineers, and all those workers, and all their good work, is held hostage to Musk’s whims.
Also a victim of Musk’s decision to leave these vital pieces off the table? The Artemis Program at NASA. Musk has already been awarded the contract to create the first lunar lander for the new program, but that lander is absolutely dependent on Starship. It’s a sure bet that Musk won’t have his part of the program ready on schedule. It’s going to be some time before we even so another test flight.
In the meantime, SpaceX can repair the damage, build a flame diverter, install that deluge system, clean up the software, and ditch the whole “pitch over” means of stage separation for something simpler—like using the second stage engines to push the stages apart with an unignited shot of methane.
See you in 2024, Starship. Maybe.
A thread of images showing some of the destruction on the ground.