Will Mars-bound Starships actually be reused?

[Robotbeat]

Sometimes I think some of you are insistent in keeping the costs at more than $1 million per ticket.

Now this cuts deep. Take it back! 

We share the goal, you're just wrong in your math.

I'll explain in more detail.

Imagine there were two Mars colonization architects.

One designed a ship that's reusable, can fly 10 times, and can do it so fast it's available for reuse on the next synod.  Let's call him Reusable Robert.

The other designed a ship that flies once, and once landed it becomes storage tanks, batteries, pumps, etc. Lets call her Non-reusable Nelly.

Both Robert and Nelly want to launch double as many ships each synod, starting with 1 ship at synod 0, until they reach 1024 ships.

So: 1,2,4....512, 1024  for a total of 2047 outbound trips.

Nelly has to build 2047 ships.

Robert thought he can save 10x, but even though he always has every ship ever launched available for reuse on each synod, his progression looks like:
1,1,2,4...512 for a total of 1024 ships.
He didn't save 90%..  he saved only 50%!!!

But wait. Common practice is that when you make twice as many ships, your cost per ship drops by 20%.

The cheaper ships lie towards the end, so when you average it out, the savings are not 50% but only 37% (says excel)

But wait - Nelly has a lot of assets on the surface - thousands of tanks for all the ISRU fluids for 100,000 people, batteries, pumps, etc.

Robert meanwhile had to build a much more capable spaceport - he has to turn around hundreds of ships per synod, not just let them land.  That's a tough task even on Earth, not to mention on a colony that's still growing.

On the flip side, Robert has a fleet remaining at the end, but after 20 years - is it really relevant?  Because the ships of the first decade are not the ships of the second decade, or the third..

So in fact Nelly might hit a lower cost to Mars than Robert.
   

It doesn’t cost nothing to double how many ships you send every year. By reusing ships, you halve the factory size necessary at each step even if you DO double every synod, so there’s a pretty good case to be made for doing it.

So Nelly has a factory on Earth that's twice as large, but Robert has to build a refueling and relaunch operation that has to match the incoming traffic, pretty much in real time.  That's 100x as large and complicated as Nelly's requirements!

…

This isn’t true at all. Solar and electrolysis & the Sabatier reactor is cheap. Water recovery is pretty cheap, too, once you get started. And again, one such delivery of that equipment, which might cost $5M to build and $10-20 million to land on Mars, will enable the recovery of over 100 Starships in its lifetime, and about 10 Starships just in its first synod. Those starships cost, at a very minimum, $10M apiece, so it’s saving $100 M worth of starships just in its first synod, a payback of like a factor of 4 or 5 in 26 months. 50 over its lifetime.

Solar is just really cheap.

…and I used crazily optimistic values for Starship’s manufacturing cost. If it’s closer to $50-100 million instead of $10 million to build a starship, then it’s even more obvious.

[whitelancer64]

Does anybody have a good feel for the cost of Starship avionics?

Starship's avionics are undoubtedly based on F9 heritage, and the F9 avionics package costs about $10,000.

From Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future by Ashlee Vance:

Musk wanted the bulk of a rocket’s computing systems to cost no more than $10,000. It was an insane figure by aerospace industry standards, where the avionics systems for a rocket typically cost well over $10 million. “In traditional aerospace, it would cost you more than ten thousand dollars just for the food at a meeting to discuss the cost of the avionics,” Watson said. During the job interview, Watson promised Musk that he could do the improbable and deliver the $10,000 avionics system.

...Watson and other engineers built out the complete computing systems for Dragon and then adapted the technology for Falcon 9. The result was a fully redundant avionics platform that used a mix of off-the-shelf computing gear and products built in-house by SpaceX. It cost a bit more than $10,000 but came close to meeting Musk’s goal.

One of the advantages of large rockets is that your avionics don't really get any heavier/costlier, so you achieve better amortization.

From the quote I presume that is about the Falcon 9 v 1.0. Also, that book was published in May 2015 - before they even had a successful landing. Falcon 9 has gotten more complex since then, and has been human rated. Never mind inflation.

The size of the avionics is going to reflect how much input it takes, how much data it has to process. If one rocket engine has 300 sensors, then the Falcon 9 first stage avionics handles 2,700 inputs, but Super Heavy has to manage 9,900 inputs and process all that data, in addition to the other navigation, tank pressures, etc. work it has to do.

[meekGee]

Sometimes I think some of you are insistent in keeping the costs at more than $1 million per ticket.

Now this cuts deep. Take it back! 

We share the goal, you're just wrong in your math.

I'll explain in more detail.

Imagine there were two Mars colonization architects.

One designed a ship that's reusable, can fly 10 times, and can do it so fast it's available for reuse on the next synod.  Let's call him Reusable Robert.

The other designed a ship that flies once, and once landed it becomes storage tanks, batteries, pumps, etc. Lets call her Non-reusable Nelly.

Both Robert and Nelly want to launch double as many ships each synod, starting with 1 ship at synod 0, until they reach 1024 ships.

So: 1,2,4....512, 1024  for a total of 2047 outbound trips.

Nelly has to build 2047 ships.

Robert thought he can save 10x, but even though he always has every ship ever launched available for reuse on each synod, his progression looks like:
1,1,2,4...512 for a total of 1024 ships.
He didn't save 90%..  he saved only 50%!!!

But wait. Common practice is that when you make twice as many ships, your cost per ship drops by 20%.

The cheaper ships lie towards the end, so when you average it out, the savings are not 50% but only 37% (says excel)

But wait - Nelly has a lot of assets on the surface - thousands of tanks for all the ISRU fluids for 100,000 people, batteries, pumps, etc.

Robert meanwhile had to build a much more capable spaceport - he has to turn around hundreds of ships per synod, not just let them land.  That's a tough task even on Earth, not to mention on a colony that's still growing.

On the flip side, Robert has a fleet remaining at the end, but after 20 years - is it really relevant?  Because the ships of the first decade are not the ships of the second decade, or the third..

So in fact Nelly might hit a lower cost to Mars than Robert.
   

It doesn’t cost nothing to double how many ships you send every year. By reusing ships, you halve the factory size necessary at each step even if you DO double every synod, so there’s a pretty good case to be made for doing it.

So Nelly has a factory on Earth that's twice as large, but Robert has to build a refueling and relaunch operation that has to match the incoming traffic, pretty much in real time.  That's 100x as large and complicated as Nelly's requirements!

…

This isn’t true at all. Solar and electrolysis & the Sabatier reactor is cheap. Water recovery is pretty cheap, too, once you get started. And again, one such delivery of that equipment, which might cost $5M to build and $10-20 million to land on Mars, will enable the recovery of over 100 Starships in its lifetime, and about 10 Starships just in its first synod. Those starships cost, at a very minimum, $10M apiece, so it’s saving $100 M worth of starships just in its first synod, a payback of like a factor of 4 or 5 in 26 months. 50 over its lifetime.

Solar is just really cheap.

…and I used crazily optimistic values for Starship’s manufacturing cost. If it’s closer to $50-100 million instead of $10 million to build a starship, then it’s even more obvious.

Forget cheap..  suppose they are free!

Look at Boca Chica though. Look at the amount of non-Watt infrastructure.

It's not the theoretical power calculations...  It's the sheer amount of logistics and equipment necessary.

You're talking about a spaceport that can turn around X ships in several months, but built at a point in time where in the last synod you've only landed X/2 ships.

E.g. you've only landed 1 ship, and with that you want to build a spaceport thay can turn around 2...  And still have bandwidth to build a colony..

And for what reason?  As I've shown, the savings even on Earth are just a small fraction of the total effort required, and being spendy on Earth beats being spendy on Mars.

[Twark_Main]

From the quote I presume that is about the Falcon 9 v 1.0. Also, that book was published in May 2015 - before they even had a successful landing. Falcon 9 has gotten more complex since then, and has been human rated. Never mind inflation.

The size of the avionics is going to reflect how much input it takes, how much data it has to process. If one rocket engine has 300 sensors, then the Falcon 9 first stage avionics handles 2,700 inputs, but Super Heavy has to manage 9,900 inputs and process all that data, in addition to the other navigation, tank pressures, etc. work it has to do.

There are a lot more controllers involved, both for sensors and actuators.  Those hook up to more engines, more valves for more COPVs, weird aerosurfaces, copious temperature and pressure sensors, etc.

So what, maybe $50,000 instead of $10,000?

My inflation calculator says 30% (actually 29%) since 2015, so we're at $80,000.

[Twark_Main]

This talk of hundreds of Starships per synod makes me cringe .

...Starship will morph into something bigger

How can you be so certain Starship isn't already the 'goldilocks' size? Bigger isn't always better. See: the 747 and A380.

"In nature, the optimum is almost always in the middle somewhere. Distrust assertions that the optimum is at an extreme point."


I suspect that in general, finding a use for a lot of Starship-sized ships on Mars would be easier than trying to repurpose some massive vehicle.

[MickQ]

This talk of hundreds of Starships per synod makes me cringe .

...Starship will morph into something bigger

How can you be so certain Starship isn't already the 'goldilocks' size? Bigger isn't always better. See: the 747 and A380.

"In nature, the optimum is almost always in the middle somewhere. Distrust assertions that the optimum is at an extreme point."


I suspect that in general, finding a use for a lot of Starship-sized ships on Mars would be easier than trying to repurpose some massive vehicle.

Probably a poor choice of words on my part, thinking back to the original 15mt diameter plans.

I tend to agree with someone upthread that the system will change to big orbit to orbit transports that are serviced by Starships, or similar vehicles, at either end.  I just cannot see the physical logistics required to turn around any significant number of ships and return them to Earth in a reasonable timeframe being possible on Mars in the foreseeable future.

My opinion.

[Robotbeat]

I think it might actually be the main vocation in Mars…

They have to develop SOME ability to return Starships. Might as well get good at it.

[Coastal Ron]

If you do lay down a ship, slice it in half from nose to tail, lay it on a flat surface and cover it with dirt, you could just inflate a pressure barrier inside of that space - sort of like a big plastic bag. So I'm not thinking it would be difficult to use a Starship in that way.

A non-rigid pressure barrier will just inflate to form a cylinder, so your quonset hut won't sit flat on the ground.

A rigid pressure barrier with a flat floor will have sheer stress, which makes it highly inefficient (ie not a thin-walled pressure vessel).

Nah, it can be designed to account for that. That is just an engineering problem to be solved with pretty straightforward solutions. For instance, laying down a truss floor underneath the pressure barrier will keep the forces pushing flat against the floor.

If you want to lay down a pressure vessel, don't try to cut a cylinder in half. It's a classic mistake.

Classic? This has already happened on Mars? 

If you want a flat floor, just build one (or more) inside a full cylinder.

That is a possibility. You and I aren't deciding what method will be used, we are only theorizing what the options are.

The same method could be used to line tunnels that are bored into the underground, so it shouldn't be a technique that would be unknown to the inhabitants.

It's not really the same.

In tunneling the liner doesn't hold any air pressure load.

Yeah, on Earth. However on Earth we don't use boring machines to build living spaces, but on Mars that may be an option, and we'll need pressure barriers of some type to seal the tunnels.

With a quonset hut, the lifting force of the internal atmosphere would be far greater than the weight of the shielding regolith above.

Luckily the solution to that is pile more regolith on top, which provides more radiation protection - one solution solves two problems... 

But my hope is that there will be better options that using decommissioned Starships for living spaces of any type, other than temporary housing in vertically standing Starships early on in the establishment of the first colonies.

[Brigantine]

To kind of summarize what I've got from this thread...

Era 1: Starship-powered exploration/science/institutional bases. Growth is modest, re-use happens (on legs) to mitigate cost
Era 2: Migration becomes viable for individuals. Ships are a product not a service, return/re-use mostly stops
Era 3: Martian industrial capacity increases and out-competes re-purposed ships. Re-use starts again (on chopsticks)

The bit people are most interested in seems to be era 2.

[Robotbeat]

To kind of summarize what I've got from this thread...

Era 1: Starship-powered exploration/science/institutional bases. Growth is modest, re-use happens (on legs) to mitigate cost
Era 2: Migration becomes viable for individuals. Ships are a product not a service, return/re-use mostly stops
Era 3: Martian industrial capacity increases and out-competes re-purposed ships. Re-use starts again (on chopsticks)

The bit people are most interested in seems to be era 2.

I think era 1 and era 2 should be merged. It'll actually be the earliest missions that won't return the Starships. Return of Starships happens regularly (not just for returning occasional crew) after 100 starships land.

[Twark_Main]

My spreadsheet takes into account:
- increasing launch cadence of 2x per synod
- reduction of fabrication coat of 0.8x per doubling
- reuse lag time of either 1 or 2 synods

Just curious, how does this change if the learning rate (and "doublings") are calculated w/r/t the total number of rockets SpaceX has manufactured, F1 and F9 included?

It seems like most of those lowest-hanging discoveries would be found in these very early stages, hence the extremely rapid learning rate. By only counting the total number of Starships in the "doublings," there's a risk overestimating the true (Wright's Law) rate of learning.

I suspect the real learning rate is going to fall somewhere in the middle, basically just an all-time-units curve superimposed on a current-gen-units curve, with step discontinuities in between generations. Does that make sense?

Did I miss the spreadsheet? Sorry catching up

[Robotbeat]

Fabrication of Starships has limits on how low the cost will ultimately be. I don't expect it to go below $100/kg, and even below $1000/kg is very ambitious. Airliners have a list price of about $1000/kg, and they're made about 1000 per year for the most mass-produced model.

Practical learning rates for physical goods are usually not that extreme beyond a certain point. Plus, learning rate depends on sufficient demand. There's just not going to be a need for thousands of starships per year. Who's paying for these?? Who's using them??

So again, I'd put a floor on the cost of starships, not just use raw (optimistic) learning rates.

But also... learning rates would apply to propellant production capacity.

[Robotbeat]

The idea that there's a 20% reduction in fab cost for each doubling in production is ridiculously optimistic. If we believed it, then SpaceX should stop reusing Falcon 9.

Reusing a Falcon 9 booster 16 times might halve the launch costs, but that's 4 doublings in manufacturing that you're leaving on the table, you'd be just as well off if you "just" mass manufactured more stages, as you'd reduce costs to just 0.8^4 = 41%...

An over-optimistic learning rate undermines the argument for any kind of reuse. It's also not congruent with reality, IMO. An 85% learning rate is more typically used for aerospace. 80% vs 85% doesn't sound like a lot, but after 10 doublings, it's about a factor of 2 difference.

(The biggest issue is demand... where the heck is the demand for this many Starships coming from?? Even Elon doesn't have enough money for this even if he started liquidating stock, once you get beyond like 100-1000 Starship per year.)

10 doublings is all you're going to get. Boeing and Airbus combined produce 1000 jets per year, and the vast majority of those are more like 737 sized than Starship, which is 747 sized.

[meekGee]

My spreadsheet takes into account:
- increasing launch cadence of 2x per synod
- reduction of fabrication coat of 0.8x per doubling
- reuse lag time of either 1 or 2 synods

Just curious, how does this change if the learning rate (and "doublings") are calculated w/r/t the total number of rockets SpaceX has manufactured, F1 and F9 included?

It seems like most of those lowest-hanging discoveries would be found in these very early stages, hence the extremely rapid learning rate. By only counting the total number of Starships in the "doublings," there's a risk overestimating the true (Wright's Law) rate of learning.

I suspect the real learning rate is going to fall somewhere in the middle, basically just an all-time-units curve superimposed on a current-gen-units curve, with step discontinuities in between generations. Does that make sense?

Did I miss the spreadsheet? Sorry catching up

True, but the conclusion stands even without a cost learning curve.

Just assume a doubling of the outbound fleet size every synod, and then even a super fast 1-synod turn around ship (which IMO is somewhere between plain impossible and just prohibitive) is not that attractive.

If it's a 2-synod turn-around, it's basically worthless.

This is before you factor in Wright's Law, so if you're suspicious of it, just let it be.

It's also before factoring in the added work required on the Mars side, and the lack of landed ships to cannibalize.

I'm generally rabidly pro-reuse, since the early days of F9 amd through Starship, and think reuse is key to the solar system, but in this very narrow case, everything is stacked against it.

Oh one more thing - I don't think the very early reductions in cost that we see during development follow Wright's Law. They're kinda unpredictable and can even go the wrong way...  It's fair to start any such discounts when the configuration kinda stabilizes, as much as it ever will at SpaceX.

What's more likely to happen is that after 6 years they'll start making 12 m ships or something like that, and the cost will drop because of that, and now a 10-cycle reusable ship really starts looking silly.

[Coastal Ron]

Fabrication of Starships has limits on how low the cost will ultimately be. I don't expect it to go below $100/kg, and even below $1000/kg is very ambitious. Airliners have a list price of about $1000/kg, and they're made about 1000 per year for the most mass-produced model.

Yes, I also don't think that the cost of the Starship will decrease by too much, since there are material costs that don't reduce over time, like stainless steel, and facility costs that will only increase over time as they add capacity and capabilities.

Practical learning rates for physical goods are usually not that extreme beyond a certain point. Plus, learning rate depends on sufficient demand. There's just not going to be a need for thousands of starships per year. Who's paying for these?? Who's using them??

No doubt there will be some Earth-local demand for Starship transportation services, and Musk has been talking about point-to-point transportation for a long time. Plus Starlink requires Starship in order to meet their service cost goals, and Musk has stated that he wants to phase out Falcon 9/H as soon as practical and replace it with Starship.

But most of that demand could be met with ten's of reusable Starship ships, and less than ten boosters.

SpaceX, however, was created to colonize Mars, and Starship is how Musk is trying to make that affordable. Hard to know how much Musk and/or SpaceX will fund a Starship transportation system to Mars, and we probably won't know until they perfect landing ships on Mars and returning them - which should be the final test before colonization efforts start in earnest.

The colonization of Mars will essentially be a humanitarian project, and maybe one of the biggest humanity has ever attempted, but the big question is who will pay for it?

Which is why a fleet of 1,000 Starships seems like a challenging task, but I think it is only achievable if Starships are reused multiple times. That requires Mars propellant production in vast quantities, but I'm not sure the colonization of Mars can be achieved - or made affordable - by single-use Starships.

[Robotbeat]

There's a pretty strong case for reusing Starships in the long term: it makes the Martian energy and industrial base MUCH stronger.

Think about it. A Mars settlement will have to have some basic capacity to launch starships. Luckily, launching starships is a lot easier than the full superheavy starship stack, particularly in the low gravity and low pressure. More like the suborbital stands SpaceX used for hop tests, so will require a lot less infrastructure.

BUT once you have some of that infrastructure, scaling out that infrastructure means you can get two-way trade occurring. The first export will be Mars samples and astronauts, then Starships for reuse. And if you're able to make Starship reuse make sense, you don't have to improve the costs much for actually *fueling up the Starships with Mars ISRU for sending people to Mars*.

This is especially true as the launch rate becomes ludicrous and thus potentially constrained by NIMBYs and even some legitimate large-scale environmental concerns (which start at around 10MT IMLEO/year). So not only will Martians be forced to develop refueling of Starships as a core skill just to enable roundtrip astronaut missions, but there will eventually be scaling issues with launch that would encourage off-loading some launch capacity away from Earth and to Mars. That'd provide much-needed income to Mars.

Because let's face it, there's not a lot going for Mars, economically-speaking. Freedom from anti-solar (and anti-nuclear, for that matter) NIMBYs and the lower gravity (which means a single stage is fine) may help a lot, and Mars will need all the help it can get.

Plus, having massive capacity to refuel Starships means your Mars city will have bountiful energy to act as backup in case of sandstorms and stuff (and just economy-of-scale making energy cheaper).

Reusing Starships will be the first major industrial activity of Mars. It would be a shame to not invest in it.

[Robotbeat]

Boeing and Airbus each only produce around 200 widebody aircraft per year. Of those devoted to cargo, that's about a tenth that.

I can imagine SpaceX producing 100 Starships per year... after all, they produce that many Falcon upper stages per year right now (roughly). But much beyond that seems very unlikely. The good news is that 100 starships per year, if reused around 10 times, is ~2000 Starships per synod, maybe half of those available for Mars (and passengers), so you get your 1000 Mars Starships per synod, around 100,000 passengers per synod. That's enough for a sizable Mars city, maybe a million people.

(There are clever ways of moving cargo around that don't require putting all the cargo into a single Starship fairing for the duration of a Mars transfer... So vast majority of the Mars Starships could be devoted to passengers in the long term.)

[meekGee]

Boeing and Airbus each only produce around 200 widebody aircraft per year. Of those devoted to cargo, that's about a tenth that.

I can imagine SpaceX producing 100 Starships per year... after all, they produce that many Falcon upper stages per year right now (roughly). But much beyond that seems very unlikely. The good news is that 100 starships per year, if reused around 10 times, is ~2000 Starships per synod, maybe half of those available for Mars (and passengers), so you get your 1000 Mars Starships per synod, around 100,000 passengers per synod. That's enough for a sizable Mars city, maybe a million people.

(There are clever ways of moving cargo around that don't require putting all the cargo into a single Starship fairing for the duration of a Mars transfer... So vast majority of the Mars Starships could be devoted to passengers in the long term.)

That's not how it works though, because Starships only fly once every 2 or 4 years (unlike jetliners)

So if you have a fixed production capacity of 100/synode (for example) then it'll take 20 years to make that fleet.

Are you really thinking that over 20 years, production won't increase?  Or Starships won't evolve to create larger, better, cheaper ships? And then the value of reusing older ships will drop in comparison to just using them on Mars.

Hell, over 20 years even cost-of-money considerations become dominant.  People go through most of their career... 

Again - it's the long cycle time that kills the "divide ship cost by reusability count" type of thinking.

[Robotbeat]

How are you going to get more than 100,000 people to fly to Mars every synod?

[meekGee]

How are you going to get more than 100,000 people to fly to Mars every synod?

You'll need thousands of outbound ships over many (tens of) synodes, and because of the very long time span involved, you can't treat the manufacturing capacity or ship design or ship size as constants to be divided by 10 or 20...

For those time spans, fabrication capacity will always increase, either by number of ships or size of ships, and 10-year-old designs might still work, but they won't provide the kind of cost savings you're imagining through reuse.

Doubling the number of ships produced every synod just means that you keep building more factories.  More high bays. Trains more people.  Tesla did it, right?

This wil be the same, but the cars will all be going to this place that's 2 years away.