totm nov 2023 SpaceX Discussion Thread

[sh1pman]

5 minutes ago, tater said:

This.

As I was trying to say up the thread, the mere existence of the GSE tanks implies SS/SH production is absurdly inexpensive, not matter how you do the math. The only assumption required is that SpaceX would not spend more than some multiple of off the shelf propane tank cost. Pick a number and work from there? Think SpaceX would happily spend 100X off the shelf tank cost? Then a SS costs ~$10M. Would they not burn money in that way, and only spend 10X tank cost? $1M. Etc.

I think SpaceX will always go for the cheaper option. If you can buy a GSE tank for cheaper, why make it yourself? One possibility is that they’re at a point when the bulk of their tank costs is worker salary and machine maintenance, so there isn’t a huge financial difference between building a tank or not. Might as well build something. 

[Spaceman.Spiff]

2 minutes ago, sh1pman said:

the cheaper option.

I sure hope so now that tax dollars are partially paying for them.

[tater]

7 minutes ago, sh1pman said:

I think SpaceX will always go for the cheaper option. If you can buy a GSE tank for cheaper, why make it yourself? One possibility is that they’re at a point when the bulk of their tank costs is worker salary and machine maintenance, so there isn’t a huge financial difference between building a tank or not. Might as well build something. 

I'm open to them thinking spending more is OK... but not past a certain amount.

[tater]

[Spacescifi]

 

Um...why cannot Elon just run computer simulations of what could go wrong BEFORE it does?

 

Might save some metal from becoming scrap?

 

What? Some stuff you cannot predict? Cannot pressure and propellant flow be calculated for? That's what computers do best!

No offense really, RUD's are entertaining, but watching Starship do what it was meant to do would be more so.

[Spaceception]

From SpaceX updates: https://www.spacex.com/updates/starship-moon-announcement/index.html

Quote

SpaceX is rapidly advancing Starship development, drawing on an extensive history of launch vehicle and engine development programs. Since January 2020, SpaceX has built 10 Starship prototypes, with production and fidelity accelerating on each build. SpaceX has manufactured and tested more than 60 of Starship’s Raptor engines, accumulating nearly 30,000 seconds of total test time over 567 engine starts, including on multiple Starship static fires and flight tests. We have conducted six suborbital flight tests, including two 150-m hops and four high-altitude flights. SpaceX has also built a full-size Super Heavy booster as part of a pathfinder effort, and currently has five vehicles in production.

Italics mine. That's impressive. It seems as though they've slowed down Raptor development from last summer, and are taking their time with the more recent Raptors. If last year's tweet updates were anything to go by, it sounded like they were moving through ~10 engines per month or so, and would likely have hit SN50 by early fall. Since that's supposed to have been a major milestone for the Raptor, could they have started to slow down around the same time?

[tater]

47 minutes ago, Spacescifi said:

What? Some stuff you cannot predict? Cannot pressure and propellant flow be calculated for? That's what computers do best!

You actually think they don't do simulations?

They do BOTH.

If simulating real world complex systems for a single SN# hop took 2 months to code, and run, and took 99% of all possible real world issues into account, you'd still miss that 1%, but hey, it's easy to change and fix, right? What if it took 6 weeks to build the thing and actually fly it? Faster to fly the real thing.

52 minutes ago, Spacescifi said:

Might save some metal from becoming scrap?

Metal is cheap.

[Spacescifi]

1 minute ago, tater said:

You actually think they don't do simulations?

They do BOTH.

If simulating real world complex systems for a single SN# hop took 2 months to code, and run, and took 99% of all possible real world issues into account, you'd still miss that 1%, but hey, it's easy to change and fix, right? What if it took 6 weeks to build the thing and actually fly it? Faster to fly the real thing.

Metal is cheap.

 

I see. I was beginning to think all the crashes were due to stuff he could have caught.

 

But it seems he has a deadline so he is going ASAP so if crashes cannot be avoided and that is the price of fast learning so be it.

 

Did not know comp testing was more tedious than actual spaceship flight tests LOL.

[tater]

4 minutes ago, Spacescifi said:

 

I see. I was beginning to think all the crashes were due to stuff he could have caught.

But it seems he has a deadline so he is going ASAP so if crashes cannot be avoided and that is the price of fast learning so be it.

Did not know comp testing was more tedious than actual spaceship flight tests LOL.

There is no deadline other than actually doing it in a time frame that is not decades.

Testing is not necessarily more difficult than actual testing, but you have to then simulate everything. Look at Starliner—they simulated everything on Starliner, but in the real world it failed because they didn't simulate the 2 vehicles talking to each other (though they simulated each alone).

The very first flip for Starship looked... exactly like the SpaceX simulation video I presume you have seen—and that was from a couple years prior.

What that does not tell them is about the unknown unknowns. You can't simulate things you have not thought of already.

Edited April 17, 2021 by tater

[cubinator]

57 minutes ago, Spacescifi said:

 

Um...why cannot Elon just run computer simulations of what could go wrong BEFORE it does?

 

Might save some metal from becoming scrap?

 

What? Some stuff you cannot predict? Cannot pressure and propellant flow be calculated for? That's what computers do best!

No offense really, RUD's are entertaining, but watching Starship do what it was meant to do would be more so.

Simulations aren't perfect.

So many equations, especially in aerospace engineering, have pieces that aren't actually mathematically derived. Jet engines have a "manufacturer's constant" that goes in an exponent to calculate thrust. All kinds of numbers are decided on by real-world testing because there's no way to derive them mathematically. Even when you plug it all into a simulation, you have to check and compare with real world data to make sure the simulation is predicting things correctly, not in some alternate universe where drag or flow works slightly differently. So simulation will never totally match a real system when making something as complicated as a rocket.

[tater]

 

[magnemoe]

8 hours ago, Ultimate Steve said:

There's a lot of speculation that this award is because the SpaceX option was the only one they could afford, and even then they had to talk SpaceX down in price a bit.

There is further speculation that, because Congress probably won't be happy with this choice, that this also serves as a play to say "We'll have two landers, we'll give you what you want, if you properly fund us."

SpaceX has the benefit that they only need their hot gas landing engines who they need anyway for accurate landing on earth. Yes they need new legs but they don't have to retract or be hidden by heat shields,  an nose docking port. 
In short this will be pretty cheap for SpaceX, Main cost is probably the procedures for the operation

[Spacescifi]

7 minutes ago, cubinator said:

Simulations aren't perfect.

So many equations, especially in aerospace engineering, have pieces that aren't actually mathematically derived. Jet engines have a "manufacturer's constant" that goes in an exponent to calculate thrust. All kinds of numbers are decided on by real-world testing because there's no way to derive them mathematically. Even when you plug it all into a simulation, you have to check and compare with real world data to make sure the simulation is predicting things correctly, not in some alternate universe where drag or flow works slightly differently. So simulation will never totally match a real system when making something as complicated as a rocket.

 

I see...so real flight testing is the ultimate proving ground for success or failure.

 

Kind of like life. One can guess all day about what COULD happen or just do a real test and write out the test results for future reference.

 

The scientific test method could be useful in life for all kinds of unknowns one may wish to try.

 

Data derived from resl tests never lies. That you can bank on.

[mikegarrison]

So many shake-my-head issues here. Let's start with one:

15 minutes ago, Spacescifi said:

Data derived from resl tests never lies. That you can bank on.

Please meet my little friend, Mr. Bar. First name Error. Error Bar gets a lot of work in science, because data derived from real tests abso-f'ing-lutely can lie. For all sorts of reasons.

Next is the question of computational engineering (or even physics). The results of any model are only at best as good as the model. Model has a bug in it? Bad results. Model is perfectly coded but makes a simplifying assumption that wasn't true? Bad results. Model is great at jobs X and Y, but you use it for Z? Bad results. You see where I'm going with this?

Now real-world testing. What happens when you have some experimental error? Same as "model has a bug in it". What happens when your test fails to match real-world conditions? Same as "wrong assumptions". How about when you take this test data and use it for something it wasn't designed to test? Same as "X and Y but not Z".

What happens when you run a test and get 10 data samples, but the part you are testing has a low-cycle fatigue issue that makes it break after 20-30 cycles? Well, that's a problem, right? What happens when you tested it 10,000 times in the lab, but never when it was -15C, and yet you need to fly on a cold winter day?

There are reasons why computational CFD has mostly replaced wind tunnel testing, but there are also reasons why there are still wind tunnels.

[Spacescifi]

20 minutes ago, mikegarrison said:

So many shake-my-head issues here. Let's start with one:

Please meet my little friend, Mr. Bar. First name Error. Error Bar gets a lot of work in science, because data derived from real tests abso-f'ing-lutely can lie. For all sorts of reasons.

Next is the question of computational engineering (or even physics). The results of any model are only at best as good as the model. Model has a bug in it? Bad results. Model is perfectly coded but makes a simplifying assumption that wasn't true? Bad results. Model is great at jobs X and Y, but you use it for Z? Bad results. You see where I'm going with this?

Now real-world testing. What happens when you have some experimental error? Same as "model has a bug in it". What happens when your test fails to match real-world conditions? Same as "wrong assumptions". How about when you take this test data and use it for something it wasn't designed to test? Same as "X and Y but not Z".

What happens when you run a test and get 10 data samples, but the part you are testing has a low-cycle fatigue issue that makes it break after 20-30 cycles? Well, that's a problem, right? What happens when you tested it 10,000 times in the lab, but never when it was -15C, and yet you need to fly on a cold winter day?

There are reasons why computational CFD has mostly replaced wind tunnel testing, but there are also reasons why there are still wind tunnels.

 

Got it, so a combo of forethought and real test result review is vital to see what worked and what did not.

Compare notes as it were. DO let your right hand know what your left did...at least when it comes to evaluating unknown potential.

[kerbiloid]

1 hour ago, tater said:

You actually think they don't do simulations?

Yes, but do they have installed RO? RealFuels? RSS? Principia?

We know they have MechJeb or TCA (cuz they land), and Hangar Extended, but what about realism?

1 hour ago, tater said:

If simulating real world complex systems for a single SN# hop took 2 months to code, and run, and took 99% of all possible real world issues into account, you'd still miss that 1%, but hey, it's easy to change and fix, right?

And Kraken.

1 hour ago, tater said:

Metal is cheap.

Metal itself. But the hardware manufacturing, including precise equip and staff salaries.

1 hour ago, tater said:

but in the real world it failed because they didn't simulate the 2 vehicles talking to each other (though they simulated each alone).

Iirc, Starliner had problems with clock and so on before it could fail also the meeting.

Anyway, if the talking was not simulated, then it raises a question, how did they test the equipment for that. Kinda, one hand clap.

1 hour ago, Spacescifi said:

Did not know comp testing was more tedious than actual spaceship flight tests

Of course, it is.
For the comp testing you first need code the code, debug it, run beta-testing,
While the rocket is just a pipe with flammables which flies on its own. It either bursts, or not.

[magnemoe]

32 minutes ago, Spacescifi said:

I see. I was beginning to think all the crashes were due to stuff he could have caught.

But it seems he has a deadline so he is going ASAP so if crashes cannot be avoided and that is the price of fast learning so be it.

Did not know comp testing was more tedious than actual spaceship flight tests LOL.

As I understand much of the problems with the landings has been sloshing in the header tank. First they used pressurized methane but the sloshing created an pressure drop. 
Next try with helium pressurization and you got bubbles in the  methane.  Now they know this they can simulate it at least with helium and design baffles who reduce sloshing but don't trap gas while flipping. 

[Elthy]

What i wonder when seeing the pictures of Dragon ontop of the Falcon 9: How exactly is it fixed to the service module? Any hardpoint would be a gap in the heatshield while the umbilical arm that reaches around the heatshield surely isnt enough to hold it in place...

[magnemoe]

5 minutes ago, Elthy said:

What i wonder when seeing the pictures of Dragon ontop of the Falcon 9: How exactly is it fixed to the service module? Any hardpoint would be a gap in the heatshield while the umbilical arm that reaches around the heatshield surely isnt enough to hold it in place...

Its bolts trough the heat shield as I understand. this is not an problem with say an 1 cm bolt, yes it will be hot but the heat would spread out pretty fast.  Cabling and cooling pass the umbilical arm on the side to avoid heating as this is stuff who can not handle heat and its also an much larger hole. 
Pretty sure all capsules use an variant of this. 

[kerbiloid]

All of them are attached by piercing the shield, afaik.

Gemini for MOL, TKS, and Soyuz VI also have successfully tested man-sized hatches in it.

And Shuttle together with Buran both had at least three huge doors for wheels underneath.

And it was never a source of problems.

[tater]

 

[kerbiloid]

Lunar missions became commercial?

[CBase]

17 hours ago, Spacescifi said:

Um...why cannot Elon just run computer simulations of what could go wrong BEFORE it does?

 

Might save some metal from becoming scrap?

Actually you are solving the problem that NASA always did: How to get a single rocket up there.

That was never the intention of Elon Musk: He wants a Mars colony with hundreds of rockets flying, so his vision is a rocket and a factory. He takes a lot of expierence from automotive engineering where a lot of test vehicles are built while building up the factory.

Although these prototypes are work intensive and do cost more than final versions, their cost is less than it seems because you need to build them anyway to master the production process. And even if they are not perfect for orbital flight and landing, they might be good enough for some hop tests. So build them early and they are useful to engineering. Build them later and they are just scrap.

[tater]

51 minutes ago, kerbiloid said:

Lunar missions became commercial?

All US space efforts are "commercial." NASA builds nothing, contractors do. Most look at the USAF and NASA as the customer, then design capability to exactly service that customer in a "commercial" setting as we commonly talk about it. Others design specific thing to a NASA spec, bespoke—SLS for example.

Mueller (designer of the Merlin engine) was referring to the case for SpaceX. If they built it, aside from Mars (which doesn't generate revenue, it costs money), what is it good for? One use case is selling lunar missions to someone with deep pockets... the US government. SpaceX was gonna build SS anyway, if they can sell a couple flights for $2.9B, that's great for them.

Note that the contract is for 2 flights, one uncrewed demo, and one crew mission.

The 2 losing submissions are entirely free to build their landers anyway—the difference between them and SpaceX is that minus the government throwing money at them, they won't do it.

[tater]