Tuesday, June 16, 2020

You Need a Reusable Starship, But You Might Need an Expendable One Even More

Hans Koenigsmann says in an interview with Der Spiegel that SpaceX is still shooting for an orbital flight of Starship in 2020.  I'm not holding my breath on that one, but it did occur to me that maybe we've got the order of development for Starship all wrong.

If you assume that the first orbital test of Starship will be with a vehicle that's already done successful suborbital entry, descent, and landing, then the end of the year is going to be hard to make.  But what if the first Starships are expendable?  Indeed, what if expendable Starships are about as common as reusable ones?  Then maybe Koenigsmann isn't as crazy as he sounds.

What follows below is to some extent an update to the post I did last year about how to use Starship without subjecting crews to its risks during launch and EDL.  But there are some key differences.  First, we now know that SpaceX is part of the Human Landing System contracts, which means that NASA's going to use a variant of Starship (among other vendors) for the actual Moon landing.

More importantly, however, the economics change pretty radically if you allow for the possibility of expendable Starships and their close cousins, Starships that never return to Earth.

Starship Applications

SpaceX not only wants to revolutionize orbital transport, but they're also counting on Starship to be a huge driver of revenue for the company.  It's worth remembering that there are several different applications in mind for Starship, and that they have fairly different requirements.  Sorting through those requirements will give us an idea about what things they need to attack first.
  1. Starlink.  This is almost certainly the bulk of early Starship traffic.  Starlink has the potential to be a major cash cow, and the payloads are cheap enough that they can be risked early in the Starship program.  Requirements:  Semi-reliable launch, modest SuperHeavy reusability, and launch from Florida (to reach higher inclinations).
  2. Commercial orbital launch.  SpaceX has been advertising Starship as a lower-cost replacement for Falcon 9, but my guess is that most commercial operators will stick with what they know for quite a while.  When commercial operators come calling, SpaceX will have to weigh the opportunity against the risk and the costs of possibly developing features in an order that delays other more lucrative applications.  Requirements:  Extremely reliable launch, good GTO payload performance with no refueling, launch sites dependent on target orbits supported.
  3. National Security Space Launch.  The one group that would be happy to take full advantage of the massive payload capabilities right from the git-go is the DoD.  They'll have a need for both large constellations and unitary large payloads, going to both LEO sun-synchronous orbits as well as GEO.  Note that NSSL launches will be incredibly conservative, though.  Requirements: Extremely reliable launch, excellent GTO performance, direct ascent to GEO, Florida launch site with the Cuban Dogleg to support polar orbits.
  4. Refueling.  While the 3 applications above can all be handled without on-orbit refueling, the remaining ones can't.  So when Starship has BEO missions to perform, refueling is essential.  Indeed, refueling will make up the bulk of the launches needed for any BEO mission.  Requirements:  Semi-reliable launch and semi-reliable EDL.  And refueling tech, of course.
  5. Artemis Human Landing System.  SpaceX has an HLS contract, and they've already committed to delivering a stripped-down version of a Lunar Starship (LSS) for use with HLS.  Requirements:  Reliable launch, refueling capability in both LEO and NRHO, vacuum crew certification (no Earth ascent or EDL required), cargo transfer capability in NRHO, crewed rendezvous, proximity ops, docking (RPOD), lunar landing and ascent capability.
  6. Human LEO and cis-lunar missions.  The loss-leader on this is DearMoon.  However, the real prize lurking here is a replacement for SLS/Orion.  This won't happen right away, but DearMoon will make it almost impossible to ignore the fact that the most expensive version of Starship is still less than 1/8th the cost of an SLS/Orion mission.  The requirements depend on whether you think that you need to subject crews to Starship launch and entry, descent, and landing (EDL), or whether it's OK to board them from a Dragon 2 in LEO.  I think the latter, so: reliable launch, probably LEO refueling capability (it's close), vacuum crew certification, and crewed RPOD, including RPOD after a powered insert back into LEO.
  7. Commercial Lunar Payload Services.  Starship can land lots of cargo on the Moon in one mission.  CLPS is mostly oriented around very small payloads right now, but I'm sure the Artemis folks are thinking up lots of ways of making the crewed missions more productive by pre-positioning large payloads on the surface beforehand.  Requirements:  extremely reliable launch, LEO refueling capability, lunar landing capability.
  8. Robotic Mars missions.  These are needed as pathfinders to prove out methane and oxygen production on the martian surface before sending crewed missions a synod or two later.  Requirements:  Reliable launch, LEO refueling capability, long-duration storage of methalox prop, Mars EDL.
  9. Human Mars missions.  Elon's holy grail, no doubt starting slow and ramping up as quick as he can.  Requirements: crew-certified launch, refueling with crew on board, long-duration transit, Mars EDL, Mars ascent, and Earth EDL at interplanetary speeds.
  10. Big deep-space robotic missions.  Tired of haggling about whether Europa Clipper needs to go on an SLS or take an extra 18 months on a Falcon Heavy with a kick stage?  Just send an entire lander and get it there in half the time.  Hell, send two landers.  Requirements:  extremely reliable launch and refueling.
Note how few times we had to put EDL in the requirements above.  Also note that we can do Starlink even with low-reliability launch, as long as we're willing to lose the occasional payload.

Starship Variants and Performance to Earth Orbit

Since Starship is a moving target, let's put some stakes in the ground.  I'm going to assume that Raptor vacuum engines with Isp=375 seconds will be available as soon as they're needed.  I'm also going to assume that tankers are just regular Starships with no payload.  That leaves us with two, or possibly three, variants:
  1. Starship "Classic" (SS).  This is a reusable vehicle that launches to LEO, refuels as needed, performs its mission, and then does a direct entry, descent, and landing.  It has three RaptorSL and three RaptorVac engines.  The EDL gear is expensive, because it involves canard and tail aerosurfaces, header tanks to keep the engines reliable on landing, and a lot of thermal protection.  In theory, this can support a crew.  We'll assume that the crew module is payload is not part of the dry mass.
  2. Expendable Starship (ESS).  I'm going to call the version of this that lands on the Moon Lunar Starship (LSS).  Both the expendable and lunar variants have the same architecture:  Same form factor as SS, but with the EDL gear stripped off.  Same complement of engines as SS.   ESS can perform missions where it makes economic sense, and LSS is what SpaceX has to deliver to NASA, crew-certified to land people on the Moon.  In both cases, removing the EDL gear reduces the dry mass.  ESS is a one-shot vehicle.  LSS may be reusable in lunar orbit or even LEO, making more than one run to the lunar surface, but it never comes back to Earth's surface.
  3. The one that I'm not sure about is the Short Lunar Starship (S-LSS).  This is the LSS, but with a few of the cylindrical tank rings taken out to make a shorter, lighter vehicle, with less propellant.  Everything other than the cylindrical height of the prop tanks is the same as a regular "long" LSS:  same crew module, same payload system, same tank domes, same thrust structure.  But, because its dry mass is lower than LSS, it can get away with fewer engines, probably two RaptorSL and two RaptorVac.  If this is easy, we'll see that it's kind of a slam-dunk for being a better answer for HLS.
I'll spare you all the nitty-gritty, but here are my best guesses for dry mass, prop, and unrefueled payload to LEO and GTO.  None of them can make GEO or lunar orbit without refueling.  These are all cargo versions with half vacuum, half SL engines, and assume 9400 m/s of delta-v to LEO.  In all cases, we assume a reusable SuperHeavy boosts whatever version of Starship is listed below:


Dry Mass (tonnes)Propellant (tonnes)Payload to LEO (tonnes)Payload to GTO (tonnes)Can Land on Earth?
Starship Classic (SS)12012001190yes
Expendable or Lunar Starship (ESS or LSS)96120017231no
Short Lunar Starship (S-LSS)6363512826no

The first thing to notice is that the cost of the dry mass for EDL is huge.¹  It effectively renders classic SS unacceptable for any kind of unrefueled launch beyond LEO.  The second thing that's important here is that the S-LSS has performance for both LEO and GTO that aren't shabby at all.  The lower dry mass means that thrust-to-weight for S-LSS is comparable to SS, even for heavy payloads.

The Cost of Launch, and How Reliability Affects It

The point where EDL becomes important is when the cost of expendable Starship missions prices them out of the market.   To find that point, we need to think a bit about the economics of reusability.  Like F9, we should expect Starship to start out with very little reusability, then gradually improve over time.

To get an idea of what we're looking at, let's make a very crude model for the manufacturing cost of both Starship and SuperHeavy.  The model uses three parameters for each stage:
  1. The cost of a Raptor.  Musk has tweeted multiple times that he expects the cost for even the first version of Raptor to be no more than $1M, so we'll use that everywhere.
  2. What percentage of the cost of each stage is engines.  I've fooled with a lot of F9 cost models, and the various constraints seem to indicate that the F9 core stage cost is about 33% engines, and the second stage engine is about 15%.  As the number of engines goes up, I'd expect that their share of total stage cost also goes up, so I'm going to assume that Starship is 20% engines and SuperHeavy is 50% engines.
  3. The number of reuses.  I expect this to evolve over time, just as it has with F9. We need to estimate how many reuses are necessary for a particular mission to be economical.  These are just guesses at this point. 
Let's put 'em all together, using what I consider to be a plausible evolution of reuse for Starship:

Component Manufacturing EstimationNumber of Reuses Expected For Each Component
Engine PctEn-
gines
Manu-
facturing Cost
Early SSEarly ESSEarly LSSMature SSMature ESSMature LSSCur-
rent F9
F9 / D2Mat-
ure F9
SLS / Orion
SuperHeavy50%3162.01045100100100
EDL Starship20%630.0320
Non-EDL Starship23%626.11315
Falcon 9 Core33%927.35510
Falcon 9 S215%16.7111
Falcon 9 Fairing6.025
Dragon 250.02
SLS1,000.01
Orion1,000.01
Per Launch Cost ($M)16.241.621.12.126.75.815.137.110.62,000

Note that I've also included some costs (not prices) for F9, F9/D2, and SLS/Orion.  The F9 numbers I'm pretty confident in, the D2 number is a SWAG, and the SLS/Orion numbers are, sadly, likely optimistic for what NASA actually will pay.

One other note:  I'm assuming that the short and long versions of the LSS cost the same to manufacture.  I'd expect the short version to be a bit cheaper, but not by enough to matter.

Economics of the LEO High-Runner Case: Starlink

Before heading off to the fun stuff, let's remember that Starship's first reliable gig is getting bajillions of Starlinks to a fairly high-inclination LEO.  Today, that's done with a reusable F9, which can launch 60 Starlinks, each weighing 260 kg, using a single packaged stack that fits into its 4.6m diameter x 6.7m high cylindrical fairing section.

Starship will have an 8m diameter x 8m high cylindrical fairing section.  That will allow each stack to contain 72 Starlinks, and the wider diameter will allow 3 separate stacks, for a total of 216 Starlinks per launch.  To do more than this will require a pretty exotic use of the nose section of the payload bay, and I'm going to ignore it for now.

We know the masses, we know the performance, and we know the launch costs of the vehicles.  So let's look at launch costs on a per-Starlink and per-kg basis:


VehicleNumber of StarlinksCost Per Launch ($M)Cost Per Starlink ($)Mass Per Launch (t)Cost Per kg ($)
Reusable F9 Today6015.1252,02015.6969
Expendable SS, Modest SH Reuse21641.6192,53256.2741
Mature F9 Reuse6010.6176,56615.6679
Reusable SS, Decent Reuse21621.197,66556.2376
Reusable SS, Mature Reuse2162.19,81556.238

It's no surprise that reusable Starships are cheaper than expendable ones.  But the real eye-opener here is that an expendable Starship, with a modestly reusable SuperHeavy launching it, is a better deal than an F9 with today's reusability.  In addition, launch slots are a scarce resource, and you get 3.6x as many birds to orbit per expendable Starship launch as you do on an F9.  That makes for a pretty compelling case that, in the short term, SuperHeavy reusability is more important than Starship reusability.

Weasel words, though:  How compelling this case is obviously depends on whether the F9 actually gets all the way to 10 core reuses and decent fairing reuse.

Right now, we see SpaceX diligently making preparations to do hop-testing of Starship, which will likely lead to a high-altitude test of the "skydiver" descent profile and the "swoop-n-slam" landing maneuver.  These make a lot of sense for two reasons:
  • Raptors are currently thin on the ground, and you can get a lot of testing done with 3 of them per test article.
  • If something bad shows up in either skydiver or swoop-n-slam, it's back to the drawing board for Starship.  They need to know that sooner rather than later.
But once the basic architecture of Starship is proven, the next thing that's going to happen is that Raptor engine production is going to increase rapidly.  The question then becomes:  what's the next step?  We've assumed that SpaceX would proceed to testing Starship by itself in downrange suborbital trajectories, to get experience with the lifting reentries that it'll need for orbital reuse.

But I no longer think that's right.  Instead, I think they'll cut over to building out and testing SuperHeavy as quickly as possible.  Sure, Starship tests will continue, but not at the expense of SuperHeavy testing.  Because as soon as SuperHeavy can take a Starship to LEO, the numbers above say that it's likely worth it, even if Starship doesn't survive EDL.

In other words, we'll have a situation not unlike how F9 got tested:  You launch a real payload, and you test reusability after each stage has done its job.  As SuperHeavy becomes reusable (which should be easy; it's just the F9 on steroids), the cost of tests drops precipitously.  Then each Starlink launch becomes a Starship EDL test.  If they fail, no harm done, because SpaceX got a bunch of Starlinks to orbit for less than they would have on an F9.  Eventually, as EDL starts to work and Starship becomes reusable, other classes of missions will open up, at lower price points.

However, to do this for Starlink, SpaceX would have to be able to build a SuperHeavy/Starship pad in Florida.  Without it, the high inclinations for Starlink aren't really accessible from Boca Chica, because they can drop debris on Louisiana or Yucatan in the event of an accident.  SpaceX has plans to put a Starship launch pad near the current LC-39A pad, but we've just recently witnessed what happens when a methane-fueled rocket explodes.  I'm not sure NASA would be wild about witnessing the same thing with 4x as much propellant, right next to the pad where they plan to send crews to the ISS later this year.

Beyond Earth Orbit and Accounting for the Cost of Refueling

The first class of missions that desperately needs a reusable Starship is a "lift tanker" (LT) to deliver prop to LEO. To get to BEO, refueling is required.  Refueling can occur in different orbits, at different times, so accounting for it at the point of refueling is confusing.  However, for now at least, all propellant is launched from Earth to LEO.  If we know how much is required there, then we have a clean cost estimate for the prop that any mission needs.

Let's assume that a Starship tanker is, for now, just a Starship Classic with no payload.  The first tanker of a series will become the "accumulation tanker" (AT), receiving prop from subsequent LTs until there's enough prop to accomplish the mission for the payload Starship.  This minimizes risk to the payload Starships, and keeps them on the ground until the AT is full in LEO.  Note that it's always most efficient to refuel at the lowest orbital energy possible.

In some cases, even after fully topping off the payload Starship in LEO, missions will require that a second AT boosts itself into a high orbit to top off the payload Starship before it can fulfill the mission.  I've been calling this a "high orbit tanker" (HOT).  There are even cases where you need to do "HOT-to-HOT" refueling, so that one of the HOTs has enough prop to top off the payload Starship.  I'm going to assume that all HOT tankers top off payload Starships in the lunar non-rectilinear halo orbit (NRHO), where NASA plans to put the Gateway and also to do pre-Gateway rendezvous between Orion and HLS.  It's not as efficient, but it's much saner operationally.³

Obviously, the more reusable the LT is, the less the flight costs, the less propellant costs to get to orbit.  From the tables above, we can expect an LT to deliver 119t of prop to LEO.  The minimal amount of reuse is probably something like the "early SS" reusability shown above, which is $16.2M per flight.  It's tempting to just convert these two numbers to a cost per tonne and have done with it, but until we have a very high cadence of BEO missions, using the same inclinations, you might not get to use any non-integral remainder from the last LT for one mission as the beginning of the AT for the next mission.  For this reason, we're going to assume that the cost is in integral LT flights.

Note that the cost of LT flights is the single biggest determinant on mission cost.  As time goes on and we approach the "mature" cost of LT flights ($2.1M), propellant costs become less and less important.  We'll look at this in some detail below.

What reliability do we need out of our LTs?  If we lose one, either on ascent or EDL, we're out some hardware, but presumably refueling ops require multiple tankers with some redundancy in the supply chain, so a loss during the LT-to-AT refueling phase costs money, but isn't a big risk to the mission.  HOT rendezvous accidents are a bigger deal, but we can assume that reliability for on-orbit operations is higher than that for ascent or EDL.

We're going to look at two different missions:
  1. Uncrewed payload delivery to the lunar surface, using 30t of payload as a baseline, with no return mass.  We'll look at both expendable and reusable options here.
  2. Landing crews on the lunar surface, using a 20t crew module and a 10t surface-deployable payload (30t total), with the crew module and 1t of mass returning to Earth.  There are a lot of different architecture options to explore here, and we'll look at several.
Here are the numbers for lunar cargo missions.  The LSS and S-LSS versions are one-way, and therefore assume truly expendable Starships.  The classic Starship mission is two-way, with no mass returned to Earth.  "Early reuse" corresponds to using the "early SS" assumptions from above for refueling and the two-way mission, and the "early ESS" for the payload missions themselves.  The "mature reuse" uses the assumed end-state scenarios for the same things.


Lunar Cargo Launch CostsCargo to lunar surface: 30t. Cargo from lunar surface: 0t
Mission CharacteristicsEarly ReuseMature Reuse
Mission ConceptLaunchLEO - NRHO TransitTo / From Lunar SurfaceEDLHigh Orbit RefuelTotal Prop to LEO (t)LTsTotal Mission Cost ($M)Early Reuse ($/kg)Total Mission Cost ($M)Mature Reuse ($/kg)
One-way LSSSH/LSSLSSLSSnono3754106.4$3,54647.9$1,598
One-way S-LSS
SH/S-LSSS-LSSS-LSSnono287390.2$3,00641.6$1,386
Two-way SSSH/SSSSSSSS1 HEEO1,54513231.7$7,72327.6$919

Here are the same results in chart form:



There are a few things that are notable in these results:
  1. Until Starship reusability is close to reaching its mature form, with 20ish Starship reuses and 100ish SuperHeavy reuses, it does not make sense to send cargo on a reusable Starship.  The reason for this is that returning the Starship requires massively more LT launches, which swamp any savings on reusing the Starship itself.
  2. The short LSS does better than the long version.
  3. Note that neither of the expendable options require refueling other than in LEO, while the reusable option not only requires a HOT refueling, it also requires one in HEEO; there's not enough headroom to refueling NRHO.  See note #3 at the bottom for why I'm not fond of HEEO.
  4. If you're really getting mature reusability out of the SH/SS system, then returning cargo Starships makes more sense than sending expendable LSS or S-LSS systems.
Crewed missions are a lot more complex, there are more options, and there are likely the political constraints that require SLS/Orion to be used for launch, crew transit to NRHO, and return of the crew direct to EDL.  I see this happening in four phases.

First, SpaceX will do what NASA is paying it for:  A semi-reusable lunar Starship that is refueled and reprovisioned in NRHO, with the crew also transferring to and from it in NRHO.  Here's a little cartoon of what that looks like:



Note that we have a combination high-orbit tanker and cargo Starship that receives a series of refuelings before setting out to NRHO, where it transfers cargo and fuel to a reusable LSS.  In this picture, we're assuming two things:
  1. This is a short LSS.  If it's a long LSS, you need a second HOT and a HOT-to-HOT transfer in HEEO to get enough propellant to NRHO.  (You can also just send two HOTs to NRHO, but that's more expensive.)
  2. The S-LSS is reusable, and already on-orbit in NRHO.  If this is the first trip for this particular S-LSS, it can be launched with cargo pre-loaded, but it needs to be refueled in LEO before setting out for NRHO.
Once the S-LSS is refueled and cargo is transferred (by magic--the lack of a clear mechanism to do this is a big deal), it's ready for the crew.  The crew is launched on an SLS/Orion, and the Orion docks (presumably nose-to-nose) with the LSS, where crew is transferred.  Note the limitations of the diagram I'm using; the little astronauts with the arrows indicate crew transfers (space walks are unnecessary!).  Once transfer is complete, the S-LSS undocks from the Orion, goes to the surface, returns and eventually re-docks with the Orion, the crew transfers back, and the S-LSS is left in NRHO for the next mission while the Orion goes home to EDL.

At some point, I expect that SpaceX will rub NASA's nose in the fact that any LSS or S-LSS capable of landing on the Moon can also go from LEO to NRHO and back to LEO, rendering SLS/Orion silly.  NASA will no doubt want dual sourcing on this, and will therefore keep the Artemis architecture largely intact, so there will still be a transfer from the "transit" LSS to the "HLS" LSS:


Note that I'm assuming that getting Starship to be crew-certified for launch and EDL takes a long, long, long time, and that the transit LSS will therefore have crew boarded via an F9/D2, with the D2 waiting in LEO for the transit LSS to return to take the crew home to EDL.

Again, this is an S-LSS picture, at least for the HLS LSS, with the long LSS requiring a HOT-to-HOT transfer in HEEO to get enough prop.  However, the transit LSS can be either short or long.

If you only need to take a Starship to lunar orbit for sightseeing, you don't need any high-orbit refueling, and an LSS-like vehicle can be pre-fueled before the crew arrives.  This sounds an awful lot like #DearMoon.  It's what I expect #DearMoon will actually be, rather than a direct launch and EDL mission.  Killing eccentric Japanese billionaires is bad for business, but D2/F9 is pretty well understood, and making an LSS crew module that won't kill anybody in vacuum for a couple of weeks is pretty easy.  As an added bonus, Maezawa will get a chance to orbit the Moon for however long he wants to.

Note that this is really freakin' complicated, with two (or three) prop accumulations, four crew transfers, and an NRHO fuel and cargo transfer.  No doubt SLS fans will harp on this incessantly, and they won't be wrong.

So the next step is to stage the HLS right from LEO, eliminating a pair of crew transfers.  Here's what that looks like for a long LSS:


Note that here I've put in the HOT-to-HOT transfer that I left out of the previous two schemes.  The big drawback here is that this is the first case where we have to refuel an LSS with a crew onboard, which incurs some extra risk.

This is a somewhat bigger deal in the S-LSS case:


While the S-LSS eliminates the HOT-to-HOT transfer in HEEO, the cost of having smaller propellant tanks is that the S-LSS can't go straight from the lunar surface to LEO without a second refueling in NRHO.  A single HOT can do both refuelings, though.  We now have two refuelings with crew on board.⁴

At some point, Starship will be capable of launching crew directly from Earth, and recovering them directly to EDL.  This is much simpler, and it's even cheaper than the schemes above, but I'm still counting on crew-certification of the whole profile taking a long time.  If not, this is clearly the way to go:


Much simpler!  However, note that this requires refueling a crewed Starship in HEEO, with its attendant orbital mechanics issues and radiation problems for the crew.  This can be avoided by moving the refueling to NRHO or some other lunar orbit.  However, you need a HOT-to-HOT transfer in HEEO to get enough prop there, or a significant increase in total prop to do two HOT refuelings in NRHO.  Both of these are a small price to pay to simplify the mission, though.

Here's summary of all of the missions we've discussed above, along with the actual propellant and hardware costs:


Lunar Crew Launch Costs
Crew and cargo to lunar surface: 30t. Crew and cargo from lunar surface: 21
t
Mission CharacteristicsEarly ReuseMature Reuse
Mission ConceptLaunchLEO - NRHO TransitTo / From Lunar SurfaceEDLHigh Orbit RefuelTotal Prop to LEO (t)LTsTotal Mission Cost ($M)Early Reuse $/kgTotal Mission Cost ($M)Mature Reuse $/kg
Two LSS for LEO-LS-LEO
F9/D2 Launch and EDL
F9/D2LSS #1LSS #2D21 NRHO2,16319350.0$11,66652.0$1,732
Two S-LSS for LEO-LS-LEO
F9/D2 Launch and EDL
F9/D2S-LSS #1S-LSS #2D21 NRHO1,55214269.0$8,96641.4$1,378
Single LSS for LEO-LS-LEO
F9/D2 Launch and EDL
F9/D2LSSLSSD22 NRHO2,01717296.5$9,88341.9$1,396
Single S-LSS for LEO-LS-LEO
F9/D2 Launch and EDL
F9/D2S-LSSS-LSSD22 NRHO1,57114247.9$8,26335.5$1,184
Single Starship Launch, Transit, Lander, EDLSH/SSSSSSSS1 HEEO1,69915264.1$8,80333.9$1,131
SLS Launch
Orion Transit
LSS Lander
SLS/OrionOrionLSSOrion1 HEEO
1 NRHO
1,622142,247.9$74,9302,035.5$67,851
SLS Launch
Orion Transit
S-LSS Lander
SLS/OrionOrionS-LSSD21 NRHO1,159102,183.1$72,7702,027.0$67,568

Here are the results in chart form.  Note that I've left off the Orion missions, because their cost is so off the scale that the interesting stuff is no longer readable.




There are no easy answers here.  Before launch and EDL crew-certification for Starship, the S-LSS has a significant cost advantage over its long cousin, although that advantage comes at the cost of some refueling complexity in NRHO.  The other big takeaway here is that as SuperHeavy and Starship reusability go up, the cost of propellant all but disappears.  So the choice and phasing of Starship work is highly dependent on both reusability progress and crew-certification speed.

I suspect that the short LSS will turn out to be a winner.

The Cost of Building Short LSS

So let's talk about the difference between a short LSS and a long one.  Why would you want Starships in two different form-factors?  It's one thing to strip off EDL gear, but quite another to build vehicles of different lengths.

However, the construction methods used on Starships make it easy to build things with different sized tanks.  Each LOX and LCH4 tank is has a dome and a common bulkhead.  You don't want to fool with those.  But the rest of the tank is formed by stacking and welding rings of stainless steel sheeting.  Removing part or all of those rings is easy.  The limiting factor becomes the domes and bulkhead in the LCH4 tank:  you need a small amount of cylinder to which to weld the flanges for the domes, and that's the minimum volume for that tank.  That's how the propellant size for the S-LSS is derived.

The real cost is not in building the other variant of the ship.  Instead, it's in payload processing and ground support equipment for that ship once it's built.  You need different sized high bays to service the different sized ships, along with alternate umbilicals and strongbacks for the different ships on top of the SuperHeavy.  Again, the cost of doing that work depends on how expensive prop is in LEO.

What About NSSL?

Because this post isn't quite agonizingly long enough, there is one last issue.

We talked briefly about how unrefueled Starships simply can't reach GEO.  That's not germane to a lunar program, but it's highly important to the military.  One of the key baseline missions for NSSL is the ability to get a payload on station in GEO within a small number of hours.  These "direct to GEO" profiles are very difficult for many launchers, and it's a pretty big advantage for the current Falcon Heavy.  But Starship can't do this.

Now, Starship can absolutely do this if it's refueled in LEO.  I get something like 12t of payload to GEO with three LTs' worth of propellant, which dwarfs any other platform.

But Space Force simply isn't going to allow refueling on a multi-billion dollar, one-of-a-kind spacecraft.

However, up above, we saw that an expendable Starship can get 31t to GTO.  Unfortunately, that's doesn't solve the "direct to GEO" requirement.  But SpaceX has already revealed that they're making a version of the Dragon 2 spacecraft, the Dragon XL, as a logistics module for resupplying the lunar Gateway.

If you look at the D2 architecture, you'll discover that the guts of the vehicle are concentrated in a ring of equipment that circles the bottom of the pressure vessel.  It looks like what they've done with the DXL, which can only be launched inside a fairing, it to keep the guts but swap the crew pressure vessel for a larger--and flimsier--pressurized cargo container.

It occurred to me that swapping in yet a third system would be pretty interesting:  If you put one or two SuperDracos in the place where the DXL docking ring is currently, mount some fairly large extra tanks of MMH and NTO above the "Dragon guts", and then put a payload attachment on the front, you have yourself a mighty fine GTO-to-GEO tug.  This would be effectively a third stage for the Starship, deployed in GTO from the Starship payload bay, and capable of circularizing to GEO as soon as it reached the first GTO apogee.  Time to GEO:  maybe 12 hours. 

I did a little figuring and came up with a "Dragon tug" with a dry mass of about 4t and 16t of MMH/NTO propellant, which can take about 15t of payload to GEO on an expendable Starship.  No refueling needed.  That's something that Space Force would really like.

Conclusions

So I've droned on for a long time here.  What does it all mean?  Here are the big takeaways:

  1. Expendable Starships for GTO missions are a no-brainer.
  2. Expendable Starships for lunar cargo missions are a no-brainer.
  3. Everything else depends on how quickly crew-certification of the entire Starship flight profile, from launch to transit to lunar landing/ascent to EDL, takes place.  My bet:  launch and EDL are really hard, but everything that happens in vacuum is pretty easy.  Similarly, how quickly reusability comes online, and the extent of that early reusability, will dictate a lot of decisions.  My bet here is shakier, but I'm guessing that there is a substantial advantage in concentrating on getting SuperHeavy reusability first, which allows a variety of expendable Starship missions to be cost-effective sooner.
  4. Without the full profile crew-certified, all crewed lunar Starship missions are pretty complex, but they're so much cheaper than doing things with SLS and Orion that investing in these profiles where the F9/D2 gets crews to and from LEO seems worth it.
  5. There's a fairly simple play to get heavy stuff to GEO for NSSL.
It'll be interesting to see what happens.


¹A bit of a caveat here:  In the interests of getting things somewhat orderly, I've assumed that it takes 9400 m/s of delta-v to get to LEO.  That's likely a bit conservative.  Also, as payload mass gets smaller, the higher thrust-to-weight reduces gravity drag a bit, which might give you slightly more payload to GTO.  (This is offset partially by the fact that SuperHeavy gets faster and further downrange, and requires more prop reserve for its return to the launch site.)  So it's possible that Starship Classic could get a couple of tonnes to GTO.  But bidding a giant, unproven launch system for less payload than an F9 can put in GTO seems kinda silly--unless the cost is ridiculously low.  And it's not--at least not for a while.

²I think the way to do this is to use tail-mounted payload pods and transfer them from the logistics SS to the LSS during tail-to-tail refueling.  Seems like something that a robot should be smart enough to manage.  Otherwise, you're flinging stuff through hatches, which sounds like a terrible idea.  But deploying tail-mounted payload on the surface will require some interesting modifications to the LSS skirt.  I suspect that, until this problem is addressed, LSS will be an expendable vehicle, sent to a disposal orbit after the crew have transferred back to their NRHO-LEO or NRHO-EDL transit vehicle.

³SpaceX has talked about using a "highly elliptical earth orbit", roughly similar to GTO, to do the most efficient refueling.  And it is true that the optimal refueling point is always at the lowest energy to complete the mission.  So if you need a mission that takes more than the Starship's capacity to complete, you first fill up the SS in LEO, then proceed to the lowest energy above LEO that will let you add just enough prop to finish the mission.  For a lot of lunar missions, that energy turns out to be 2000-2500 m/s above LEO, so NRHO refueling is not the most efficient thing to do.

However, efficiency is not the only consideration:
  1. HEEO is bad for crewed missions, because you'll do a pair of passes back through the Van Allen belts for each orbit in HEEO.  If something delays refueling, you don't want to abort a mission for radiation exposure.
  2. HEEOs are hard take a long time to phase.  Minor errors at insertion require rotating the apside line of the spacecraft, which grows more expensive as the eccentricity and major axis increase, and takes longer if you want to do it optimally near perigee and/or apogee.
  3. If you want to do an optimal rendezvous (and you do to avoid radiation exposure--see #1), then you've got your payload Starship sitting in LEO, burning for the HEEO, at exactly the same time that the HOT is coming up from behind it at the highest-speed part of its orbit.  Space is very big, but I can't say I'm wild about putting myself in the path of an object that's intentionally going to kiss my orbit at exactly the time that I'm trying to burn to match its orbit, when the closing speed is 2000-2500 m/s.  Things are gonna get dicey if my burn fails to occur for some reason.
  4. Because of the eccentricity, optimal departure windows for TLI are going to be in a fairly narrow window.  If you miss that window, you'll have to rotate your line of apsides to adjust for a new insertion, which is expensive and time-consuming.
In contrast, NRHO is a nice, stable orbit, well outside the Van Allen belts, with well-understood rendezvous characteristics.

⁴This may be a good argument for a not-quite-so-short LSS.  I haven't done the math on this, but if you can eliminate the extra crewed fuel transfer in NRHO, something a little less efficient might be worth it.

1 comment:

David Richards said...

I'm curious this weekend about re-use/expending of LSS. I starting asking about LSS over here: https://forum.nasaspaceflight.com/index.php?topic=51736.msg2143772#msg2143772

A few folks addressed my question, but your writeup is more what I was looking for.

I wonder which (if any) of these makes the most sense to you?

1. LSS is designed to return to Earth
2. each LSS is expendable
3. LSS returns to Earth-orbit, for possible use in future missions to lunar surface or lunar orbit