New Mars Forums

From AIAA’s “Daily Launch” email newsletter for Monday,  6-23-2025:
------   
SpaceX traces Starship test-stand explosion to failure of pressurized nitrogen tank
By Mike Wall published 3 days ago  (on Space.com)
"Initial analysis indicates the potential failure of a pressurized tank known as a COPV."

SpaceX thinks it knows why its newest Starship spacecraft went boom this week.

The 171-foot-tall (52-meter-tall) vehicle exploded on a test stand at SpaceX's Starbase site late Wednesday night (June 18) as the company was preparing to ignite its six Raptor engines in a "static fire" trial.

A day later, SpaceX narrowed in on a likely cause.

"Initial analysis indicates the potential failure of a pressurized tank known as a COPV, or composite overwrapped pressure vessel, containing gaseous nitrogen in Starship's nosecone area, but the full data review is ongoing," the company wrote in an update on Thursday (June 19).

"There is no commonality between the COPVs used on Starship and SpaceX's Falcon rockets," the company added. So, launches of the workhorse Falcon 9, which has already flown 75 times in 2025, should not be affected.

The Starship explosion did not cause any reported injuries; all SpaceX personnel at Starbase are safe, according to the update. People living around the site, which is near the border city of Brownsville, shouldn't be worried about contamination from the incident, SpaceX said.

"Previous independent tests conducted on materials inside Starship, including toxicity analyses, confirm they pose no chemical, biological, or toxicological risks," the company wrote. "SpaceX is coordinating with local, state, and federal agencies, as appropriate, on matters concerning environmental and safety impacts."

That said, the explosion did damage the area around the test stand, which is at Starbase's Massey site (not the orbital launch mount area, from which Starship lifts off).

"The explosion ignited several fires at the test site which remains clear of personnel and will be assessed once it has been determined to be safe to approach," SpaceX wrote in the update. "Individuals should not attempt to approach the area while safing operations continue."

Wednesday night's explosion occurred during preparations for Starship's 10th flight test, which SpaceX had hoped to launch by the end of the month. (Static fires are common prelaunch tests, performed to ensure that engines are ready to fly.) That timeline will now shift to the right, though it's not clear at the moment by how much.

The incident was the latest in a series of setbacks for Starship upper stages. SpaceX lost the vehicle — also known as Ship — on the last three Starship flight tests, which launched in January, March and May of this year.

Starship's first stage, called Super Heavy, has a better track record of late. For example, on Flight 7 and Flight 8, the huge booster successfully returned to Starbase, where it was caught by the launch tower's "chopstick" arms as planned.

------   

My take:  if the description “in the nosecone” for the location of the COPV is correct,  then it is located very close to the oxygen header tank (as the version 1 with 1200 metric tons propellant capacity was laid out),  which is also in the nose of the vehicle,  ahead of the “cargo bay” area.  Such a COPV explosion would easily rupture that oxygen header tank.  Compressed gases drive great explosive violence (with shrapnel) when such vessels burst. 

There would seem to be an oxygen header tank transfer piping line down the windward “belly” of the cargo bay section,  based on descriptions I have read.  In the explosion slow-motion video,  the cargo bay splits open through its heat shield,  right where that transfer line supposedly is,  with gush of something white (not fire) bursting through,  followed immediately by an explosion engulfing about the top half of the vehicle,  and a second or so later by a second explosion seemingly centered lower down.   

The main propellant tanks below the cargo bay would be the main methane tank forward,  with the methane header tank located inside,  at the base of that tank,  and finally the main oxygen tank,  just ahead of the engine bay.  The upgraded version 2 has a bigger propellant capacity,  but should be laid out similarly.   

I would hazard the guess that the COPV explosion and bursting oxygen header tank somehow put a large force on the transfer line,  which split open the belly at the cargo bay,  allowing liquid (and vapor) oxygen out through that split,  as well as releasing a few tons of liquid oxygen to fall down on top of the main methane tank. 

My guess is that spilled header oxygen and vented methane vapors are much of the first explosion.  Bear in mind that the impact of a few tons of liquid oxygen on the top of the main methane tank would rupture it as well,  adding some fuel to that first explosion pulse.  That first explosion pulse would massively rupture the main methane tank,  and also likely the main oxygen tank below it.  That’s the second pulse of the explosion,  which was larger and longer,  reflecting the larger mass of reactants. 

All of that scenario is just an educated guess on my part. 

As for the nitrogen tank,  said to be a “COPV”,  or “composite overwrapped pressure vessel”,  maybe that is not the right choice this early in the flight test program.  Such a design is a metal shell that is simply too thin to hold the pressure,  overwrapped by a yarn or fabric-reinforced composite material,  to bring it up to strength at a lighter weight. 

Here’s the problem:  no composite material has a large plastic (post-yield) strain capability.  If the COPV over-pressures for any reason whatsoever,  failure will be sudden,  without any warning!  Maybe a heavier all-metal nitrogen tank,  one with much more plastic strain capability,  would be a better choice until the other bugs all get worked out.  At least you could see it stretch before it explodes.  You do not want to fly even experimentally,  with too many possible failure modes! 

Lots of things look good “on paper”,  but there are a lot of other things to worry about,  many of which cannot be put on that paper.  This is where the “older hands”,  with many years of school-of-hard-knocks experiences,  can be effectively very much wiser than youngsters fresh out of school.  SpaceX has no “old hands” on its staff:  they hire no one over about age 40 or 45.  There’s no gray heads visible anywhere in that organization.

GW

I thought Scott Manley did a good job in that link,  primarily saying "we don't know,  must wait and see".   He did raise some issues with the header tank piping causing the initial splitting-open on the heat shield side.  That slow-motion he has is better than anything else I've seen up to now. 

It shows cold gas erupting through the split opening up through the heat shield,  splitting the thing open in the cargo bay zone before ever there is any combustion.  That's a rather important clue!  And I don't know what it means,  either.  And right now,  I doubt even SpaceX's team knows.  And rightly so. 

"Build it,  break it,  build another" only works if you don't make very many changes from one build to the next.  Too many things changed between versions 1 and 2 of the Starship upper stage.  None of the 2's has been at all successful yet,  while some of the 1's were.  I think that outcome makes clear that too many changes were made jumping from 1 to 2,  something only easily recognizable after-the-fact.  There's no point to "you should have known better" blame games.  No one could know better,  especially without any "old hands" on the staff to provide any "hard-knocks" wisdom to a crew of newbies.  Newbies being pushed to do it too fast by the head honcho (that being Musk,  who as both Bob Clark and I have both previously pointed out,  is not a qualified engineer in any sense of that word).

Myself,  I would not finish ship 37 (a version 2),  I'd build another version 1,  just putting the revised heat shield and the current version 2 Raptor engines in it.  If that should actually fly all the way to an ocean landing,  then you know that the trouble lies in the other version 2 changes,  and more likely among those made trying to cut down inert weight.  You inevitably lose structural margins when you do that. 

We also did not see as many upper stage engine problems with Starship version 1 (unlike with the booster),  whether it was Raptor version 1 or 2.  Therefore I suspect not so much the engines themselves,  as I suspect the feed plumbing in the vehicle bringing propellants to them.  Yes,  the engines were probably too vulnerable to the unanticipatedly-intense engine bay fires from leaks.  Leaks less likely at the power heads of the engines themselves,  and more likely in the supply plumbing and maybe the tankage in the vehicle design,  independent of which version of the engines was being flown.  Just wrap that stuff up against the heat of the leak fires.  Find and fix that first,  and only then try to reduce inert weights without screwing up that prior gain.

500 bars is about 7250 psi.  That's high enough to be quite challenging,  but it is something attainable.  I'm not sure that anything other than SS304/304L has the plastic strain capability to handle system-type problems at pressures like that.  That stuff will stretch almost an order of magnitude further before failing,  than any other inert gas container material candidate,  that I can think of.

Right now,  rectifying a failure mode is way more important than saving inert weight.  I think most other "old hands" would agree with me about that.

GW

PS -- and even after saying all of that,  I still feel (too low a frequency to hear) some sort of thrust/pressure oscillation in the Raptors they are testing at McGregor,  including the new Raptor 3's.  They haven't recognized that risk factor yet,  much less fixed it. So there still are engine problems to fix,  as well.

Getting accurate reports about this is not as easy as it should be.  Some of the things I read early on said it was Starship stacked atop a Superheavy.  I did see two explosions about a second or two apart.  The video quality was too low to properly identify what blew up.  Now most of the reports say it was Starship only,  no booster.  That's a change in the reporting in only 1 day. 

I guess it doesn't matter.  This time it was not an engine failure,  because none of them were running.  I do note that every Starship flown since they went to version 2 has failed in some way.  The only 2 or 3 that actually made it down to the Indian Ocean were version 1. Making the changes clearly has not turned out so well. 

Now a tweet has surfaced from Musk saying it was most likely a bursting inert gas container that triggered the event.  I suppose that could be true.  But it sure is early for someone to be saying anything about cause. 

GW

This from today's AIAA "Daily Launch" email newsletter:

-----   
article image   
CBS News
SpaceX Starship upper stage explodes during ramp-up to expected engine test
A SpaceX Starship upper stage exploded in a spectacular conflagration during ramp-up to an expected engine test firing at the company's Starbase manufacturing facility on the Texas Gulf Coast late Wednesday, destroying the rocket in what appears to be a major setback for the Super Heavy-Starship vehicle Elon Musk says is critical to the company's future. Video from LabPadre, a company that monitors SpaceX activities at Starbase, showed the Starship suddenly exploding in a huge fireball just after 11 p.m. CDT, 10 to 15 minutes before the anticipated engine test firing, sending flaming debris shooting away into the overnight sky from a churning fireball that engulfed the test stand.
----- 

It's not clear from this that any of the engines were even running. 

GW

PS -- After looking at more reports plus a video of the explosion,  I can say for sure that no engines were running.  I can say it is likely the Starship upper stage exploded first,  followed a second or two later by the booster exploding.  There are a lot of possible fault trails,  but it would appear on the surface that those that might be responsible for this explosion lead to the vehicle itself,  not to its engines.

The problem with asteroid protection is that you have to find it well before it will hit you,  in order to do anything about it.  These things are black-on-black in the dark,  very hard to see.  If you are trying to see them in sunlit conditions,  they re impossible to find. They are easier to see against cold black space by infrared than visible,  because being near the sun,  while cold they are warmer than the 3-4 K of deep space.  Even so,  if you look anywhere near the direction of the sun,  its IR signature swamps what you are trying to find. 

The net upshot of all those physics is that you need to be located sunward of any potential threats,  so you can look outward with your detectors and not be swamped by the sun's output.  You will have an easier time spotting asteroids with IR than visible.   Which in turn why the B612 Foundation proposed many years ago an IR satellite located at about Venus's distance from the sun,  looking outward with IR to hunt asteroids.  They never got to do it,  but NASA did,  except that it is to be located at Earth's inner LaGrange point,  not the orbit of Venus.  It has not yet been built and launched,  and may not ever be,  if Trump's attempts to gut NASA science are successful.

With the discovery of asteroids ahead of and behind Venus in its orbit,  the same detection physics apply,  it is just that you need your surveillance satellites nearer the orbit of Mercury,  in order to look outward and see these things near Venus,  as well as the ones near Earth's orbit.  That's quite a bit more demanding and expensive mission,  but if you are serious about asteroid protection,  it has to be done.  There literally is no other reliable way to detect these threats. 

GW

The shock wave moves circumferentially around and around.  Flow ahead of it is supersonic,  but flow behind it is subsonic,  because it is a normal shock.  It has to be.  So this technology is inapplicable for anything claimed to be "supersonic combustion".  The confusion factor with terminology here is with "supersonic" ahead of the wave.  That is NOT due to the flow speed coming into this thing from the inlet,  it is due to the speed of the circulating wave itself,  which must be inherently supersonic in its circumferential speed. 

Shock wave-induced combustion is a pressure-gain kind of combustion that we call "detonation",  as opposed to merely explosively-fast "deflagration" that occurs with a slight pressure loss.  Those are the only two kinds,  and achieving stable and controlled detonation combustion has been a very difficult thing to do.  For the same pressure gain from the air inlet hardware,  the detonation-wave combustion will result in a higher-pressure subsonic chamber condition behind the circulating wave,  than could possibly be produced by deflagration in a more conventional combustor scheme.  That higher pressure results in a different nozzle throat size,  too. 

Higher chamber pressure is simply higher thrust potential,  which for the same fueling rate is higher specific impulse (if a rocket).  That definition of thrust and specific impulse gets a lot more complicated if you do this as an airbreather.  But the nozzle thrust term in airbreather thrust really is bigger. 

GW

The AI in post 10 gets it wrong because the problem posed to it is set up wrong.  You fed the AI bad numbers.

If you come off a Hohmann transfer at average planetary distances,  the difference between your 2-body apohelion speed and the 2-body speed of Mars about the sun will fall in the 2-something km/s range. It supposedly is your speed with respect to the planet.  However that difference does not include the 3-body effect of Mars's gravity upon the craft as it approaches Mars orbit distance just as Mars arrives there,  too.  2-body analyses can be done with actual equations,  in a spreadsheet.  3-body problems can ONLY be figured with finite-difference computer odes.  There are no equations,  other than the differential equations of motion.

There is a conservation of mechanical energy approximation which will adjust the 2-body-based difference of 2-something km/s into the 3-body-affected speed as you reach the desired low altitude about the planet.  You need to know the escape speed at that altitude.  For 300 km at Mars,  it is right at 4.9 km/s.  Pick any number you want between 2 and 3 km/s for the 2-body difference and call it Vfar.  The number you seek we will call Vnear.  The approximation (which gets very good results!!!) is Vnear = sqrt[vfar^2 + Vesc^2]. 

It almost doesn't matter whether you use 2 or 3 km/s for Vfar.  Using Vesc = 4.9 km/s,  you will get a number pretty close to 5.4 km/s for Vnear,  which is ALWAYS inherently larger than Vesc,  as the plus sign inside the square root in the approximation formula so very clearly indicates.  You WILL ALWAYS be moving at a speed greater than local escape speed,  as you pass by Mars at 300 km!  YOU WILL NEVER CAPTURE INTO AN ELLIPSE, THEREFORE!

You can capture into an ellipse if you slow to any speed less than local escape speed.  The closer that value is to escape rather than circular,  the more extended your ellipse will be.  For,  say,  4.8 km/s (100 m/s less than escape),  the min dV = 5.4 - 4.8 = 0.6 km/s.  For going straight into circular,  it is 5.4 - 3.4 = 2 km/s.  Once in orbit about Mars,  everything is a 2-body problem again. 

If you go into the ellipse using 0.6 km/s,  you still have to make a periapsis burn to circularize at 300 km,  if that is what the mission requires.  That would be dV = 4.8 - 3.4 = 1.4 km/s.  The initial 0.6 km/s burn to get onto the ellipse,  plus the 1.4 km/s burn to circularize,  adds to THE VERY SAME 2 km/s burn if you just decelerate straight into circular!!!

He did not seem to realize that the spacecraft as it reaches Mars's orbit,  is moving slower than Mars.  He said it was moving faster,  and that is entirely wrong.  Mars literally runs over you from behind!  All these speeds are with respect to the sun.

Him saying you have to slow down to arrive at Mars was correct,  although I believe his understanding of that is flawed.  There is a change of coordinates involved,  before you can determine speed relative to Mars instead of the sun.  Your speed with respect to the sun,  minus Mars's speed with respect to the sun,  is your relative speed with respect to Mars.  That will be a negative number,  because Mars is faster. 

Doesn't matter,  that difference is BEFORE the effect of Mars's gravity accelerates you toward the planet to an even higher speed relative to Mars (still a negative number).  The negative numbers do not matter,  when you square them to compute kinetic energy.  That's what my little correction equation for "Vnear" vs "Vfar" does.  And that higher speed close-in is what you have to "kill" to one extent or another,  either to land directly,  or to go into orbit about Mars. 

Off of Hohmann,  that Vnear higher speed is actually a little above Mars surface escape,  at about 5.4 km/s (escape 5.0 km/s).  Off a faster trajectory,  it is higher still,  and requires vector math to compute,  since the orbit arrival point is not a collinear tangency,  but a crossing at some significant angle.  Off the 2-year abort ellipse,  it is closer to 7.4 km/s.

Since Mars is literally running over you from behind for a Hohmann transfer,  you cannot use its gravity to "catch up the planet",  it is catching up to you (!!!),  and it is pulling you in ever faster as it tries to run over you from behind.

GW

Those who know me know that I understand the "build it,  break it,  build another" approach works,  but only so far as the problems to be solved are simple,  not subtle.  For subtle problems,  you need the wisdom of the "old hands" to guide what needs to be done,  because identifying the real trouble can be quite tricky. 

Combustion "instabilities" in the form of unwanted thrust and pressure oscillations (those go together,  if you have one,  you have both),  have always been rather tricky problems to solve.  They have no "old hands" on their teams,  and that's their weakness.  They hire no one over age 40-45,  because older folks cannot tolerate 70-80 hour weeks chronic for 52 weeks a year.  That kind of workforce abuse is why they have high turnover,  too.  It is Musk that drives that.  Shotwell should know better,  but has to do what he says.

As for "blowing up a lot of rockets",  the ignorant reporters are too young to remember what space flight was like in the 1950's and 1960's.  A lot more rockets blew up than actually flew,  especially in the 1950's,  and up to about 1965-ish.  An awful lot of that was due to thrust/pressure oscillations feeding into structural or plumbing vibration modes.  And an awful lot of it was inadequate flight control knowledge.  Another big part of it was cooling design failures with regenerative engine cooling designs,  although that usually showed up in ground tests before you tried to fly. 

We are a bit better at flight controls and cooling designs today,  but the thrust/pressure oscillation thing is still there,  and it is aggravated by the complex combustion cycles intended to for zero bleed flow dumped overboard.  That's just the nature of the technology.

And then there's the obviously-amateur choice of relying on tank pressure for cold gas attitude thrusters,  when you have a demonstrated history of leaks and fires,  with the leaks dropping tank pressures.  That lack of "wise old hands" at SpaceX really shows,  does it not?

GW

As an update to what I put in post 5 above,  and what Tom put in post 6 above,  the same applies to laser range finders as to radar rangefinders.  If the surface is rough,  you get conflicting signals,  leading to enormous uncertainty in your estimated altitude.  That is inherent!

Now,  if you decelerate to zero vertical (and horizontal) speed at something like 100 m up,  you can slowly descend to the surface at just about thrust/weight = 1.  That makes the landing burn and its dV longer and larger.  Inherent.  But you have a chance to set down,  if the boulders aren't too big,  and your vehicle is not too tall and spindly.

If on the other hand you are planning your descent to decelerate to zero speed just as you reach the surface (to save on landing burn dV),  you are screwed landing in rough field conditions,  because of the surface-roughness-induced uncertainty in your altitude,  NO MATTER HOW MEASURED!

Apollo-11 proved that,  beyond a shadow of a doubt,  way the hell back in 1969.  The computer was going to crash them among house-sized boulders,  too big even for the short,  squat LM design,  just because it was ignorant of what was in front of it (not even radar,  just accelerometers).  Armstrong and Aldrin had to take manual control,  stop and hover at a few dozen meters altitude,  and then hover-fly out of that boulder field.  They touched down with a single handful of seconds of propellant left.  They actually needed an even bigger dV!!!

So why is it that nobody seems to actually learn from demonstrated history?  THAT I cannot understand!

GW

Didn't I already tell everyone that spin is stable only about the min and max moment of inertia axes?  This has been in engineering dynamics texts for over a century now.

There,  I said it in 2 sentences.  The AI needed a whole page.

GW

You completely missed my points.

There is some kind of 10 Hz pressure oscillation in the engine chamber.  I can feel those effects from 6 miles away,  and they are there whether or not there is any sort of plumbing around the engine. 

In the ship, but not the test stand in McGregor,  there are particular forms of plumbing that might be susceptible to being driven by that 10 Hz oscillation in engine pressure,  via the mechanical shaking of the engine in its mountings in the stage.  That is because of the lengths of feed piping in the ship (that are not there at McGregor).  You have to pipe propellants from the tanks to the engines.  One of the header tanks has been in the nose,  almost 50 meters away.  The lengths are there to hit one of the first couple of harmonics to that 10 Hz oscillation.

If the plumbing responds to the 10 Hz shake by its contained fluid resonating with it,  that could affect the propellant flow,  which literally is the energy release rate in the chamber,  so there is the possibility that there could be an energy feedback from engine to plumbing and back to engine.  The engine shake drives a wave in the plumbing,  which affects flow rate,  which increases the intensity of the vortex shedding oscillations in the chamber.  If the wave in the plumbing gets strong enough,  the pressure nodes of it it break a tube or a fitting somewhere.  And THAT effect is exactly what they have been describing with their talk about oscillations and fuel leaks. 

It does not matter one tinker's damn that there is no ship or ship-like plumbing at the McGregor test stands!  The initiating oscillation in the engine is still there!  Which is what I have been feeling from all three versions of Raptor. There is nothing outside the engine for it to interact with.  Put the engine in the ship,  and there is.

Merlin does not seem have this problem with pressure oscillations in its chamber,  at least not at a strength that I can observe.  But it burns kerosene,  not methane!  That's inherently a sooty flame,  and that soot cloud is a draggy inertia on the gas oscillations,  acting to damp them down.  So,  even if the propellant injection geometry is identical in both engines,  and so cause the same exploding-vortex shedding,  the two engines are just not very likely to respond in the same way,  as one has a damping cloud in it,  the other does not.  Methane burns virtually soot-free.

GW