Sunday, March 3, 2019

How to Use Starship In NASA's Lunar Transportation Architecture

I'm encouraged by NASA's  broad agency announcement of a lander architecture.  Unlike the the rest of the stuff being done by the Human Exploration & Operations Mission Directorate, it's oriented around commercial launchers, and doesn't seem to require LOP-G to succeed.  That's smart.

But it doesn't mean that all is well.

SLS and Orion are still hoovering up $3.7B of HEOMD funding every year.  LOP-G is eating another $500M.  That's an awful lot of money that could be redirected to building better landers faster.  It could fund a robust lunar base with robust power and robust water mining capabilities.  Mostly, it's money that could have us back on the Moon in 6-7 years instead of 10.

At the same time, in addition to the numerous political problems associated with cancelling all this stuff, one difficult technical truth needs to be faced:  Orion is the only crewed spacecraft that will be certified for operations beyond earth orbit for some time.

CST-100 and Dragon 2 are only certified for LEO operations, and neither has the delta-v necessary to insert into cis-lunar orbit and return back to trans-earth injection.

So, unless we can come up with a credible alternative to SLS/Orion, and to the 3-stage lunar lander architecture being proposed by NASA, we'll likely have to live with them.  And they're expensive.

An Obvious Question:  What About Starship?

For those of you who are still reading a hard-core space geek post and don't know about the SpaceX SuperHeavy/Starship effort, SpaceX is developing what was called, until recently when the corporate flaks wrung their hands enough, the "Big Fucking Rocket", or BFR.  Recently its two stages have been renamed to SuperHeavy (the first stage) and Starship (the combined second stage and interplanetary spacecraft).  Both stages are fully reusable, and touted to have extremely low launch costs.

We used to have fairly good numbers on BFR back when it was a completely paper rocket, but we don't know as much about SH/SS.  I've made some educated guesses that its structural coefficient got a bit higher as design progressed, as well as plugging in numbers for the Raptor engines that have been announced since we had the good numbers.  Here are the parameters I'll be using:

SuperHeavyStarship
Dry Mass (t)280.494.8
Prop Mass (t)2,834.61,090.2
Step Mass (t)3,115.01,185.0
Structural Coefficient9.00%8.00%
Raptor-SL Thrust (SL, kN)1,7001,700
Raptor-SL Thrust (Vac, kN)1,8501,850
Raptor-Vac Thrust (kN)#N/A1,900
Raptor-SL Isp (SL, s)330330
Raptor-SL Isp (Vac, s)356356
Raptor-Vac Isp (s)375375
Number of Raptor-SLs311
Number of Raptor Vacs06
Total Thrust (SL, kN)52,7001700
Total Thrust (Vac, kN)57,35013,250
Delta-v for Landing (m/s)640 + reentry725

Note that, for the time frames we're interested in, we can assume that the Raptor has been upgraded to have a second vacuum variant, which gets higher thrust and has higher specific impulse due to optimal vacuum expansion.

So, why wouldn't NASA want to use Starship instead of SLS Orion?

There are three broad answers to that question, and then a lot of details:
  1. Really nasty politics.
  2. NASA doesn't quite believe that SuperHeavy/Starship is real yet.
  3. Even if SH/SS is real, it's not at all clear that it can be easily crew-certified--or crew-certified at all.
The politics ought to be silly, but they're not.  If Starship is on the table, especially for crew operations, then the giant pork parade that is SLS and Orion comes first into question, then comes to a grinding halt shortly thereafter.  I expect that to happen once SH/SS becomes more real, but it can't even be considered by NASA right now, lest it incur the wrath of its funders.

Objection #2 is just a combination of institutional caution and a certain amount of pig-headedness.  SH/SS is a unique design, with unique features.  It would be silly for NASA to base any lunar architecture on it before it's even off the drawing board, to say nothing of tested.

However, I'm pretty confident that SpaceX will complete SH/SS by at least mid-2022, because they need it by then in order to meet the FCC deadline for Starlink.  Things could obviously go wrong, but necessity, as they say, is the mother of invention.  I don't expect them to fail.  Still, NASA is prudent not to count on Starship before they're pretty sure it's going to work.  My guess is that confidence will be much higher by the end of 2020.

The crew-certification issue is considerably more serious.  Because it's killed crews in the past, NASA is incredibly cautious when it comes to human spaceflight.  The agency criterion for a crewed space system is that it has to have a Probability of Loss of Crew (PLOC) of 1 in 270.  None of SLS/Orion, F9/D2, or Atlas V N22/CST-100 is actually going to make that number, but they'll come fairly close, and the designs are well-enough understood that it's expected to evolve to meet the required PLOC.

Even with well-understood designs, there are strict criteria on the design for human life support and ergonomics.  So the very act of attempting to use Starship, in any flight regime, for human spaceflight comes with a non-trivial burden.  Just allowing astronauts to float around inside SS while sitting passively on-orbit will take some time.

Beyond that, Starship has characteristic that make six things pretty scary:
  1. There is no pad- or launch-abort system for Starship.
  2. If Starship is to be used to send crews directly to the lunar surface and back, it requires refueling--with the crew aboard--in a highly elliptical earth orbit (HEEO), that isn't reachable by the D2 or CST-100, and which transits the Van Allen radiation belts.
  3. Landing a crew on the lunar surface is problematic.
  4. Starship is a methalox system.  For stays beyond a day or so, boil-off of either LCH4 or LOX on the hot lunar surface will become an issue.
  5. Taking off from the lunar surface is a minor issue.
  6. The reentry regime of Starship is novel, untested, and poorly understood.
Let's cover these in more detail.  (TL;DR warning:  This closetful of anxieties going to go on for a while.  We'll return to strategy further down the post.)

Starship Pad- or Launch-Abort

The purpose of pad- or launch-abort is to escape an exploding or out-of-control spacecraft.  To escape such a catastrophe, the incipient failure has to be detected, communicated to the crew module, and the module needs to boost itself out of harm's way before it's damaged by a shockwave or pierced with shrapnel.

This was a feature of some, but not all, spacecraft.  The Apollo, Orion, CST-100, and Dragon 2 systems all have abort systems that can:
  1. Come up to full thrust in about 200 ms (i.e., they either have to be solid rocket motors or pressure-fed hypergolic systems).
  2. Generate accelerations capable of outrunning the bad stuff that's happening below them on the stack.
For reference, here are four abort systems:

SpacecraftAborted Mass (t)Abort Motor Thrust (kN)Abort Acceleration (g)
Apollo5.068914.1
Orion18.01,77910.1
Dragon 211.15835.4
CST-10013.07125.6

Note that the accelerations vary by quite a bit.  The reason for this is that the rockets underneath them also vary in size quite a bit.  The bigger the rocket, the bigger the shockwave if it blows up.

Starship is bigger than any of the rockets on this chart.  I can't imagine a safe abort occurring at less than about 8-10 g's.  So let's figure out what we have:

Dry Mass (t)Min Viable Crew and Hab Payload (t)Min Prop to LEO and Back to Land-ing (t)Aborted Mass (t)Raptor Max SL Thrust (kN)Number of RaptorsTotal Abort Thrust (kN)Abort Acceleration (g)
94.810.0473.0568.01,900713,3002.4

So the Starship has only about a quarter of the abort acceleration it would need.

On top of that, its Raptor engines are turbopump-fed, which means that they take at least a couple of seconds from start command to full thrust.  And on top of that, the Starship interstage is fixed and solid, so there's nowhere for the rocket exhaust to go before the Starship is pulled away from the core stage.

Of course, NASA had a crewed spacecraft system that didn't have a viable abort mode:  the Space Shuttle.  But it also killed two crews.  Much of NASA's obsession with small PLOCs comes from its experience with the Shuttle.  I think it's highly unlikely that NASA will accept Starship, ever, as an acceptable vehicle for putting crews into LEO.

This risk can be mitigated, however:  If it blows up with no people in it, it's no big deal.  So we can simply launch the crew on a D2 or CST-100 and have them rendezvous with Starship once it's been launched.  But that implies that the place we rendezvous is reachable with the commercial crew spacecraft and their launchers, and that it's safe--which brings us to our next problem.

Refueling Starship and Crew Safety

One of the most innovative features of Starship is that it's designed to be refueled in orbit.  This means that Starship can use all of its propellant just getting into LEO.  Then, after several additional launches of SH/SS have rendezvoused, docked, and refueled the original mission, it has enough delta-v to perform a wide range of lunar or interplanetary missions.  It's revolutionary.

But there's a fly in the ointment here when it comes to going to the lunar surface directly from LEO:  Even if Starship is refueled completely in LEO (a process that takes 11-12 launches, BTW), it still can't quite get to the lunar surface and back with a crew from LEO.  Instead, it needs to boost into an HEEO to reduce the amount of delta-v needed on the trip to the Moon.

We can figure out how much extra delta-v we need simply by working backwards.  First, we need about 2874 m/s of delta-v back to TEI (which will allow us to reenter) + let's say 725 m/s to land.  Normally I don't add safety margins into these kinds of back-of-napkins, but we need this number to be pretty precise, as we'll see in a moment.  If we add in 10% margin, we're looking at 3711 m/s back from the Moon.  With a 10 tonne payload of crew and crew module, that gives us 166 t of propellant required for the return trip.

Gross mass at lunar landing will therefore be 270.8 tonnes.

After that, we can find out exactly how much delta-v we have for the outbound trip: 375 * 9.807 * ln(1185/(94.8+10+166)) = 5429 m/s.  But we need about 6010 + 10% = 6611 m/s to get from LEO to the lunar surface with adequate safety margins.  So we need 1182 m/s pumped into an HEEO before refueling.

There are two problems with that.  First, an HEEO with the right amount of energy to soak up that amount of delta-v (when executed from a reference 200 km x 200 km x 28.5° orbit), winds up being about 6850 x 200 x 28.5°. An altitude of 6850 km puts the spacecraft pretty close to the highest intensity radiation in the first Van Allen belt--roughly 100,000 times that of LEO.

That means that the Starship will have to be refueled with no crew aboard, because we'd fry them if they hung around through multiple trips to the center of the first Van Allen belt.  According to my numbers, that's doable with a reusable F9/D2 with 500 kg of crew aboard, but they still have to do a rendezvous operation in a hostile radiation environment.  At the very least, they're guaranteed two extra trips to the center of the belt, but that assumes a highly precise rendezvous, and expeditions crew transfer, and a very quick checkout before heading off to TLI.

Another possibility is to refuel in a circular orbit in between the first and second belts.  But that would require about 2450 m/s of delta-v, which is beyond what the F9 and D2 can muster.

It's possible that NASA might deem boarding the crew in LEO, then boosting up to the safe zone.  But that requires refueling with a crew aboard.  I'm pretty sure that this is something that can be certified as being safe eventually, but it's not as safe as a rendezvous and transfer to a pre-fueled Starship.

Unlike the launch abort issue, this isn't a showstopper, but it's definitely going to cause some concern from NASA, and will take longer to certify.  If we can do refueling and crew ops inside the radiation belts, it's a big risk reduction.

Landing On the Moon With People

For the moment, let's assume that we've managed to get our crew on and refueled the Starship in an HEEO, so we can actually get to the lunar surface and back.  Now let's see what's involved with landing.  The requirements are to give the crew the proper tools to evaluate and commit to an exact landing site, and to land safely without tipping over.  There are several problems:

Ergonomics.  However the crew module for Starship is constructed, it's unlikely to be a very good design for looking down to visualize the surface during landing.  I suspect that SpaceX will have to prove to NASA that some kind of binocular camera system will give the crew the tools necessary to see what they're about to land on.

Lack of hover capability.  When an F9 core comes in to landing on a drone ship or one of the shore-based landing pads, it performs what is now called a hoverslam: its velocity goes to 0 very close to where its altitude goes to zero, but there's a fair amount of upward acceleration at that moment when the engines get turned off.  If the engines weren't turned off, the core would go flying back up into the air and there'd be a big mess.

This isn't going to work on the lunar surface.  For the first missions, the surface will be unprepared, and navigational aids will be minimal. The cargo and crewed missions will therefore need to be able to hover, evaluate the landing site, potentially move laterally to find a better site, hover again, and finally touch down.

Let's compute the hover criterion for a crewed Starship.  Up above, we computed the gross lunar touchdown mass to be 270.8 tonnes.  In lunar gravity of 1.62 m/s², the weight of the vehicle will be 438.7 kN.

A single Raptor engine has a maximum thrust of 1900 kN, and it's expected to be able to throttle down safely to 20% thrust, so the minimum thrust for Starship in vacuum is 380 kN.  So a single Raptor can allow Starship to hover at the lunar surface.

But two Raptors can't.  If NASA wants a fault-tolerant landing profile, Starship can't provide it.  To do so, it would have to have some combination of prop and payload that had at least 208 t more mass than our minimum landing.  That will require an HEEO that absolutely can't be orbiting in the Van Allen belt safe zone, and it will require more refueling launches, which will raise the cost/mass delivered to the surface--especially if the mass is just ballast.  (There's nothing to say that it can't be real payload, but I'm not sure how NASA is going to feel about massive payloads flying with crews, and I'm also not sure how much a massive payload is going to cost.  It may be that NASA has a hard time filling a Starship up with useful stuff to its full 100 t payload limit.)

Damage Mitigation.  One of the nice things about a two-stage lander with an expendable descent module is that if you damage the descent engines on landing, the ascent engines are still fine and capable of getting the crew back to orbit.  (If you think this is silly, it actually happened on Apollo  15.)  Since nothing is expendable about Starship, NASA would have to be convinced that the bottom of the vehicle was tough enough to survive any landing scenario without damage.

Last But Hardly Least: Tilt Stability.  Below is a diagram of the base of Starship, assuming that it's 9 m wide and has 5 m wide fins/landing legs.  The base is simply the outline connecting the three points of contact with the ground.  Because there are three legs, the closest point from the center of mass to the edge of the base is 4.75 m.



Now, the criterion for something not tipping over is that the center of mass has to stay inside the prism (in this case a triangular prism) made by extending the base upwards.  If it strays out of that prism, the torque that wants to push the center back upright will be smaller than the torque that wants to tip the Starship over, and it will indeed tip.

I did a crude model assuming that we land with the 166 t of prop and a 10 t payload, and came up with a center of mass that's 17.9 m up from the bottom of a 52 m (with landing legs) Starship.  To remain upright, the tilt angle can't exceed arctan(4.75/17.9) = 14.9°.



Let's now think about the conditions under which Starship will have to transport crews.  If it's the pathfinder mission, it will likely be landing in the following conditions:
  1. The landing area will be an unimproved area.
  2. The area likely have at least small boulders, so the crew will likely have to fiddle with the touchdown point to avoid those boulders--and may only be successful at avoiding the really big ones.
  3. Soil conditions will be somewhat uncertain (although this risk can be minimized by the robot cargoes that precede any human landing).
  4. Most of the landing sites we're interested in are near a Peak of Not-Quite-Eternal Light: a small set of regions near the poles where the rims of craters are in almost-continuous sunlight, allowing a lunar base that can be solar powered without having huge amounts of battery storage for the 2-week lunar night.  The angle of repose of most of the PONQUELs doesn't exceed 30°, but we have a system with a maximum tilt angle of 14.9°.
This is a bad problem.  I don't know if it's a showstopper or not.  SpaceX could increase the width of the landing leg system through various forms of funky mechanical engineering.  They could conceivably also find a way to land Starship horizontally, although that's not something that's ever been discussed.

My guess is that NASA will be a bit worried about landing cargo with Starship, but the cost advantages will eventually outweigh the hand-wringing.  But the Brown Trouser Brigade will be out in full force if any attempt is made to land crews without a lot of risk mitigation.

Supporting Long Missions on the Surface

Suppose we find a way to land crews safely.  What limitations does Starship impose on the length of the crew mission and the safety of the crew on the surface?

For the very first mission, there's a decent likelihood that the crew will be living in the lander, at least until they can get a pre-positioned hab set up.  This is a fairly small nit, but getting out of and into a Starship will involve a fairly elaborate mechanism to get to the surface.  The crew module has to be at least 32 m above the surface.  Lunar gravity is weak, but a fall from that height will have your astronauts hitting the ground at 10.2 m/s.  That's almost 23 mph--it's not recommended.

By far the biggest risk here has to do with Starship's cryogenic propellant.  The lunar surface is hot during the day, because it not only has solar thermal radiation from the sky hitting stuff, but radiation is also reflected off of the surface onto any object.

Starship is designed to keep methalox propellant cold enough in space to achieve a full Mars transit, so we're assuming that SpaceX has a plan for how to do this.  But the Moon is a lot hotter.  Boil-off under surface conditions will be higher than in space.  This may impose a maximum mission limit.

I view this set of surface risks to be manageable in fairly short order.  But they do need to be managed.

Lifting Off From the Moon

Compared to landing, lifting off is a breeze, but there is one thing to be aware of, and it's a variant of the problems Starship has doing launch aborts:  Turbopump-fed engines don't come up to full thrust instantaneously.

When we launch from Earth, this isn't a problem, because we have hold-down hardware to keep the rocket from lifting off before we've come to full thrust.  But that won't be available on the Moon.  So the engines will have to be sufficiently reliable and predictable in their startup behavior so that an uneven start or an engine failure doesn't wind up tipping the vehicle over.

Reentry and Landing

Again, there are multiple problems:

Novel Thermal Protection.  Recent tweets from Elon indicate that Starship will be forgoing a more conventional heat shield and using a stainless steel skin with "transpiration cooling", where a small amount of cryogenic LCH4 will be pumped into a jacket surrounding the hot parts of the ship, to be expelled and evaporated through small pores in the outer skin.

Needless to say, this hasn't ever been done on a crew-certified spacecraft--especially one that's coming in at TEI speeds.

Aerodynamics and Acceleration.  The reentry profile will have a high angle of attack.  The Space Shuttle had an angle of attack of 40°; SpaceX has described the entry profile of Starship as like a "belly flop", with an angle of attack of almost 90°.

The idea is to use as much surface area as possible to generate friction and slow the vehicle down.  But that of course requires limiting accelerations to something that won't turn the crew into goo.  Working out a human-tolerable profile may be tougher than doing so for a cargo ship with a bit of down-mass.

Powered Landings Only.  Once we get through the scary parts of the reentry, there are some mundane but still potentially lethal issues.  NASA is very fond of parachutes and not-so-fond of powered landings.  Parachutes to slow a 100 tonne vehicle simply aren't going to happen.

Landing Off-Course.  There's also the possibility of large navigation errors.  We've already discussed Starship's problems landing on the Moon; the same issues apply if it has to put down somewhere other than its prepared pad.  (Note, however, that the tilt stability will be better, because almost all propellant will be gone on landing.)  And of course most of Earth's surface is covered with water.  The possibility of a tail-down water landing doesn't sound great to me (although it could probably be made survivable--rockets with empty prop tanks float great, as long as the tanks don't rupture).

More Bad News

All in all, the Six Scary Things (and their Numerous Other Scary Sub-Things), represent a daunting challenge to getting Starship crew-certified.  If all of these have to be tackled before crew certification, we'd be looking at a longer than usual certification task.  To get some feel for what "longer than usual" might mean, let's looks at post-Apollo history:

SpacecraftEvolved FromDesign StartOpera-tionalYears to Certify
Space ShuttleClean Sheet1968198215
Orion MPCVOrion CEV2004202421
Dragon 2Dragon 1201220209
CST-100Clean Sheet2010202011

The vehicle that Starship most closely resembles is the Space Shuttle.  It benefits from SpaceX's experience making reusable engines and spacecraft, which would likely make it faster to certify.  On the other hand, it has lunar landing modes and high-energy reentry requirements that the Space Shuttle doesn't have.  I wouldn't be surprised if certification across all the flight regimes discussed above took about 10 years.  Start that from, say, last year, and you're looking at 2028 before Starship is available to completely replace the suite of NASA vehicles that will be used for the lunar architecture.

And Now, Finally, Some Good News

There's one flight regime we haven't talked much about: transporting crews from LEO to some kind of cis-lunar orbit and then back to LEO.

If the Starship launches without a crew, doesn't require an HEEO with an altitude greater than 600 km, doesn't land on the Moon, and doesn't reenter with a crew, then the only sub-problem from the Six Scary Things is to manage methalox boil-off while Starship waits in cis-lunar space--which is much simpler on-orbit than on the lunar surface.

The idea here is that Starship launches with no crew but with a crew module (CM) to a very minimal 600x200 HEEO.  (This is about 100 m/s above 200x200.)  After however many refueling flights get Starship topped off, an F9/D2 or Atlas V N22/CST-100 brings the crew to HEEO and transfers them to the CM in Starship.  Note that at 600 km, radiation from the inner Van Allen Belt isn't a problem.  The D2 or CST-100 then loiters in this orbit, waiting for the crew to return.

Starship then does TLI and inserts itself into NRHO.  From there, the crew does whatever it came for and, when they're done, they return to the CM in Starship (if they ever left it--more on this in a moment).

Starship then does TEI but, instead of doing a direct reentry, it does a powered insertion back into the same HEEO it left.  Note that this takes considerably more delta-v than direct reentry, which limits the payload substantially.  Once back in HEEO, Starship can rendezvous with the D2 or CST-100, the crew transfers back, and uses the commercial crew capsule to reenter and land.  Starship may or may not have to refuel at this point.  Once it does, it also reenters and lands, with no crew to endure Scary Thing #6.

Let's spend a moment on the crew module.  If done properly, the CM is merely another payload for a cargo Starship to take to HEEO.  The CM is self-contained and capable of sustaining a crew in comfort for at least two weeks.  It has no propulsion, although it may have attitude control.

An unaddressed detail concerns letting the crew see out (which is pretty important for tourism!).  This could be a big problem, but I'm assuming that Starship will wind up swapping the Pinto Hatchback half-clamshell fairing shown in the slideware for something resembling the Space Shuttle payload bay doors.  If that's the case, you just put windows in your CM and open the PBDs when Starship isn't maneuvering.

Note how we've collapsed the process of crew-qualification for Starship to two simple requirements:
  1. Have a CM that can sustain the crew.
  2. Crew-qualify the CM and Starship to make maneuvers that occur only in space, and only expose the crew to radiation on the way to and from cis-lunar space.
This isn't a slam-dunk.  There are innumerable requirements to certify power, ECLSS, thermal management, MMOD damage mitigation, etc.  But this is all stuff that SpaceX is doing with Dragon 2.  I can't imagine this taking more than 1-2 years after getting cargo Starship into service.

Why NRHO?  Why Not LLO Instead?

Crew-certifying Starship and the CM would allow it and commercial crew to replace SLS/Orion as a (much cheaper!) way of getting crews to NRHO.  But if we're going to use Starship to ferry crews from LEO to cis-lunar space and back, why should we settle for NRHO?  Why not go all the way to LLO, allowing simplification of the architecture?

There are three big reasons:

First, NASA is going to use NRHO.  It has to at least send a couple of crews via SLS Block 1/Orion to save congressional face, and NRHO is as deep as Orion can go into the lunar gravity well and still get home.

Any viable architecture must also rely on multiple launch service providers.  Going with Starship as a single-source provider is politically impossible, even if NASA could become comfortable with the current risk associated with Starship.  Here's a list of the various contenders:

Launcher SystemCapacity to LEO (t)Capacity to TLI (t)Capacity to NRHO @ Isp=375Capacity to LLO @ Isp=375
SLS Block 185.026.523.37.5
SLS Block 1B105.039.034.311.0
Falcon Heavy Expendable63.820.017.65.6
FH 2-Core Reusable55.016.914.94.8
FH 3-Core Reusable39.510.99.63.1
New Glenn45.012.511.03.5
Vulcan Centaur Heavy34.910.49.12.9
Good data, educated guess, sheer wild-ass guess

From this, you can see that the most likely non-SpaceX commercial options will be New Glenn and Vulcan Heavy (this consists of the Vulcan core, 6 solid rocket boosters, and the Centaur 5 long second stage).  As we'll soon see, the payloads barely acceptable when NG and VH go to NRHO; going to LLO makes them tiny.

Finally, things aren't exactly rosy for Starship, if we're to keep to the crewed profile we've been discussing.  Here are three different Starship missions:

Starhip to Various Cis-Lunar Desinations and Back
ManeuverVehicleDelta-v + 10% (m/s)IspDry MassPayloadStart Prop MassEnd Prop MassRefuel Launches
HEEO to NRHOStarship386537594.8100.01,090.2254.411
NRHO to HEEOStarship366537594.813.2254.425.8
HEEO to LLOStarship442837594.832.01,090.2238.311
LLO to HEEOStarship442837594.87.3238.30.0
LEO to LLOStarship453837594.8100.0692.063.47
LLO to Reentry/LandStarship182937594.83.663.40.0

Note that there's plenty of mass for a generous CM and even a lander to send it to the surface from NRHO.  But, because going to LLO takes so much more delta-v, the payload has to be smaller, which makes any CM and associated lander tiny.

Lander Architectures

Can Starship do anything to improve the 3-stage lander architecture that NASA is proposing?  Well... sorta.

The rationale for NASA's lander architecture is sound:  by having 3 stages, each stage is small enough to be launched on a variety of commercial launchers, reducing the need for SLS Block 1B launches to NRHO.  Based on an arm-wave for the sizes of the various stages, here's an example of something that would work:

NASA 3-Stage Lander Architecture Components
ComponentRe- useDry MassTarget Prop MassReturn Prop MassPosi- tioning Prop MassTotal PropPay- loadIspDelta-v + 10%Delta-v to Stack Point + 10%Step MassStack Gross Mass
Ascenderyes5.19.00.00.09.00.53052,860014.114.6
Landerno1.714.10.04.118.214.63752,29147119.934.5
Tugyes1.28.80.30.09.134.5375803010.344.7
Tug for Refuelno1.22.10.02.44.514.137504715.719.7

The tug has two purposes:
  1. It ferries propellant from TLI to NRHO, to refuel either itself or the reusable crew ascender vehicle.
  2. It acts are the first stage of the lander stack, taking the lander and ascender from NRHO to LLO, before returning empty to NRHO.
NASA expects the tug (aka the "transfer vehicle") to be reusable, and relies on two expendable "logistics resupply" flights to bring prop for both the transfer vehicle and the ascender.  It makes more sense to me to send two tugs, one containing its own prop for the transfer to LLO, and the other to refuel the ascender.  It saves having to develop two different stages.
Once the ascender/lander stack has been left in LLO by the tug, the expendable lander takes the ascender to the lunar surface.  Note that the lander is exactly 20 tonnes, the heaviest mass that an expendable Falcon Heavy (FHE) can take to TLI.  Also note that it's a methalox stage, because it doesn't matter if the residual propellant boils off.  It's staying behind on the surface.

Finally, after the surface mission is over, the ascender goes straight back to NRHO to rendezvous with the crew's ride home.  Note that the ascender stage uses storable prop, so boil-off doesn't affect mission life.

Here's a sample mission using the NASA architecture.  It uses SLS Block 1/Orion to send a crew to NRHO, refuels the reusable ascender and transfer stages (assumed to already be on-orbit), and flies out a new expendable lander, before transferring the crew and taking them to the surface and back.

NASA 3-Stage Lander Example Mission
Payload DescriptionManeuverVehicleDelta-v + 10% (m/s)IspDry MassPayloadStart Prop MassEnd Prop Mass
Prop for Ascender+TugEarth to TLINew Glenn12.3enough
Prop for AscenderTLI to NRHOTug4713751.29.61.50.0
Prop for Tug+TugEarth to TLINew Glenn12.4
Prop for TugTLI to NRHOTug4713751.29.71.50.0
LanderEarth to TLIFHE19.9
EmptyTLI to NRHOLander4713751.70.018.215.8
PropellantRefuel AscenderAscender03155.10.00.09.0
PropellantRefuel TugTug03751.20.00.09.1
NoneStack T+L+ATug03751.231.69.19.1
Orion+CrewEarth HEEOSLS Block 1enough
Orion+CrewHEEO to TLISLS Block 1empty
CrewTLI to NRHOOrion47131616.30.59.35.6
CrewTransfer in NRHOAscender03153.80.56.66.6
Crew+Ascender+LanderNRHO to LLOTug8033751.232.19.10.8
EmptyLLO to NRHOTug8033751.20.00.80.4
Crew+AscenderLLO to SurfaceLander22913751.714.615.80.9
CrewSurface to NRHOAscender28603155.10.59.00.2
CrewTransfer in NRHOOrion031616.30.56.66.6
CrewNRHO to EarthOrion45331616.30.56.63.4

Can Starship and the CM participate in this architecture?  Or, better yet, can it improve on it without having Starship land on the surface and triggering a couple of the Six Scary Things?

Starship can absolutely haul components and prop from LEO to NRHO.  If it made any sense, Starship could probably also act as the transfer/tug stage, hauling the ascender and lander down to LLO.  But the NASA architecture is sized the way it is to allow other commercial launchers the ability to get in on the action.  If you change the size of the components to better utilize Starship, that goal goes out the window.

Meanwhile, even if Starship can't land people on the surface because of Scary Things 3-5, is there a lander architecture that could take advantage of Starship but still be Less Scary?  There the answer is yes.

What if we take the CM and, instead of just using it as a living space on Starship, we stack it as the crew module for some kind of lunar ferry?  Now we have the CM doing double duty:  It could just be an Orion replacement to get crews to NRHO, but the same module could also be the crew cabin that takes people to the lunar surface.  In other works, the crew would get on the CM in LEO, fly in Starship to NRHO, then the CM+Ferry--with the crew still on it--would be deployed from the payload bay to take the crew to the surface and bring them back up.

Above, we discovered that getting 100 t from HEEO to NRHO and then some small amount of payload back to HEEO was doable.  You may have wondered where that weird 13.2 t number for return mass came from.  It turns out that it's the dry and crew mass of the following ferry system:

Crew Module and Ferry from NRHO to LS and Back
ComponentRe- useDry MassProp MassPayloadStep MassIspDelta-v + 10%
Crew Moduleyes9.10.00.59.1#N/A0
Ferryyes3.686.69.690.23005,954
Gross Mass99.8Return:13.2

The payload to the lander is the mass of the CM and its payload (aka crew).  The lander is intended to be reusable, and is returned to Earth after every mission.

Since Starship can carry a full 100 t to NRHO and still get back to a powered HEEO insertion, this CM/Ferry combo is a perfectly good replacement for the NASA architecture.  I don't think that that means that the NASA architecture goes away, but it does mean that this architecture is capable of supplanting it eventually, even if Starship can't be certified for surface landings.

I've set the Isp on the lander to be quite low.  Most MMH/NTO engines can get about 315 s.  However, if SpaceX wanted to build their own lander (they might not), using a couple of SuperDraco engines would come out with about the right T/W for lunar landing (2 SuperDracos throttled to 20% has a T/W of 0.53).  They'd have to be modified to be fully expanded, so the Isp is a guess, but 300 s is nice and conservative.

It's an open question whether a CM dry mass of 9.1 tonnes is adequate to a crew of 4 on a two-week mission.  Remember, this isn't just a way to get to and from the lunar surface; it's mostly to support the crew for the trip to and from NRHO, which takes 10 days total, on opt of the 2 days to get from NRHO to the surface and back.  On the one hand, the CM's a bit smaller than a D2 (9.5 t) and an Orion CM (10.4 t).  On the other hand, it doesn't need a heat shield, and it has zero need for resistance to dynamic pressure.  My guess is that it'll turn out to be a bit more spacious than either the D2 or Orion, with plenty of budget for consumables.

Here's a full-up crew mission to the lunar surface, using F9/D2, Starship, and our lander+CM:

Starship to NRHO with Reusable MMH/NTO Ferry and Crew Module
Payload DescriptionManeuverVehicleDelta-v + 10% (m/s)IspDry MassPayloadStart Prop MassEnd Prop Mass
CM+Ferry+propEarth to HEEOSH/Starship99.31,090.243.8
CM+Ferry+propLEO to HEEOStarship11037594.899.343.836.8
Methalox PropellantRefuelStarship1053.436.81,090.2
CrewEarth to HEEO/StarshipF9/D20.5plenty
Crew+CM+Ferry+propHEEO to NRHOStarship386537594.899.81,090.2254.5
Crew+CMNRHO to SurfaceFerry30943003.69.686.621.7
Crew+CMSurface to NRHOFerry28603003.69.621.70.0
Crew+CM+Ferry(Dry)NRHO to HEEOStarship366537594.813.2254.525.8
CrewHEEO to EarthD20.5enough
CM+AscenderHEEO to EarthStarship72535694.812.725.80.8
This takes 11 refueling flights

It seems to work!

Another obvious question to ask:  How would this ferry do as a cargo lander?  We've previously assumed that NASA would grumble about Scary Things 3-5 for cargo, but ultimately allow things to be delivered directly to the surface, as long as the things weren't human.  But that assumption might not be right.  After all, the equipment we need to land on the surface will be insanely expensive, so reducing risk to it is pretty important.  Maybe NASA would be OK launching it in Starship, but not landing it.  (Note that NASA doesn't care about the reentry, because the mission of delivering it will have been accomplished.)

For this mission, we can deploy the same ferry in LLO.  It doesn't need as much prop, so it can carry more cargo.  Here's what we get:

Cargo and Ferry from LLO to LS and Back
ComponentRe- useDry MassProp MassPayloadStep MassIspDelta-v + 10%
PayloadN/A0.00.019.20.0N/A0
Ferryyes3.677.119.280.63004,348
Gross Mass99.8

We can get away with the LLO mission because we can launch and reenter directly.  Here's what the mission looks like:

Starship to NRHO with Reusable MMH/NTO Ferry and Crew Module
Payload DescriptionManeuverVehicleDelta-v + 10% (m/s)IspDry MassPayloadStart Prop MassEnd Prop Mass
Cargo+Ferry+PropEarth to HEEOSH/Starship99.81,090.243.8
Methalox PropellantRefuelStarship621.843.8665.6
Cargo+Ferry+PropHEEO to LLOStarship442837594.899.8665.663.5
CargoLLO to SurfaceFerry30943003.619.286.615.4
NothingSurface to LLOFerry20573003.60.015.45.9
Ferry (Dry)LLO to Reentry and LandingStarship182937594.83.663.50.0
This takes 7 refueling flights

This is dramatically less cargo than a full-up Starship landing could provide, but two things to note:
  1. It's also dramatically fewer refueling launches than a full-up Starship launch.
  2. 19.2 tonnes of stuff is a lot, especially since NASA is going to be sizing the surface architecture to land with the 3-stage architecture.  Note that if you swapped the 14.6 tonnes of ascender for cargo, this architecture comes out slightly better, but not enough to get everybody's panties in a twist.

How Much Does Refueling Cost?

If we assume that each Starship tanker launch can get exactly about 100 t to HEEO, it'll take 10-11 launches to get the Starship completely full before heading to cis-lunar space.  However, this is a bit of a tricky computation to make.  If you had 100 t of prop and tanks in the Starship payload bay, you'd have to subtract the mass of the tanks from the total prop delivered.  However, if we just launch the Starship with no payload, we actually don't wind up with 100 t of prop left.  It's peskily non-linear, both in terms of the mass ratio (0 payload has less delta-v than 100 t payload)  and in terms of the prop that needs to be reserved for SuperHeavy to return to the launch site (lighter gross weight means more prop needs to be reserved to kill the launch speed after staging).  This remains an open question--and probably a moot one, given the number of assumptions we've baked in as the design has changed.

SpaceX Scheduling Issues

How confident should be we that SH/SS will be ready in about the time that Elon has been arm-waving?  He is famously optimistic, so early 2021 seems unlikely.  But I have a pretty good idea that Starship will be ready by mid 2022.  If it goes much later than this, SpaceX won't have the launch capacity to deploy the Starlink constellation in time to meet its FCC deadlines.  So they're going to be highly motivated to deploy on time.

Above, we assumed that the CM could be ready within 2 years of first cargo deployment.  That would bring us to mid-2024--about the same time that the first mission to NRHO is scheduled for SLS/Orion.  However, this will be heavily dependent on SpaceX's capital and cash flow situation by then.  Starlink could easily soak up all SH/SS launch capacity--and a lot of SpaceX's R&D capability--for a couple of years.  To be conservative, let's push the CM availability to mid-2025.

I have no clue whether the ferry is a viable strategy or not.  It subsumes the entire 3-stage architecture.  That sounds like a slam-dunk advantage, but it's not.  Anything that drives the architecture toward a single source will be viewed as a threat.  If I had to guess, there will be no financial support for a reusable ferry from SpaceX.

It's entirely possible that SpaceX would pass on this.  They would much rather work on burning down the Six Scary Things, because achieving that goal gives them a position of complete dominance in the cis-lunar market.  SpaceX may be perfectly happy to let the other vendors go after the 3-stage architecture, confident that it's ultimately a blind alley.

However, a reusable ferry operating from LLO would make a mighty nice project for a third party.  If it happens, I'd guess that it would take another 2 years beyond the CM, which would take to to 2027.

If SpaceX passes on the ferry, then getting crewed landing and takeoff from the surface are next on the agenda.  That will start with cargo landing, offloading, and takeoff from the surface, along with direct reentry and landing.  That's likely a risky set of missions, and likely to incur some failures.  If work on this starts right after the Starlink frenzy dies down (call it 2025),  it's likely a 1-2 years before NASA will give them real payloads to land.  Call it mid-2026.

Finally, SpaceX should have the Six Scary Things Resolved by 2029.  NASA crewed missions to the surface using Starship can commence.

NASA Scheduling, Budget, and Putting It All Together

Meanwhile, the NASA lunar architecture is basically a bolt-on to the SLS, Orion, and LOP-G programs.  It's an encouraging step because it really doesn't rely on SLS, Orion, or LOP-G, which means that it can continue even if those programs are drastically scaled-back or even cancelled.  But that's not to say that priorities and budget of the offending programs has been reallocated or even reduced.

I'm pretty sure that the lander architecture is actually the first substantive salvo in a war to scale back SLS and LOP-G, but we should look at what the plan is today, and how much it'll cost.

The non-TL;DR version goes something like this:
  1. Do automated scouting and science missions to the lunar surface ASAP, using commercial launchers and landers. (2019 - indefinite.)
  2. Get Block 1 SLS/Orion up and running. (2019-2022)
  3. Get Block 1B SLS/Orion up and running and start to build LOP-G (2024-2027)
  4. Build and test 3-stage lander architecture. (2020-2026)
  5. Presumably land some cargo on the Moon. (Not scheduled!)
  6. Land a crew on the Moon. (2028)
  7. ???
We haven't talked at all about equipment needed for the lunar surface, mostly because there isn't a prayer of any substantial budget to build it.  Right now, that "???" is a placeholder for yet another unsustainable set of lunar landings, followed by... nothing.

But let's get a baseline of what the current plan will cost in some detail.

First, we have to assign costs to the various components of the NASA mission, so we can compare them to those of a replacement of some of them with Starship-based components.  Costing is always a black hole, but these come out reasonably well.

Some notes:
  1. SLS and Orion costs are derived by taking the current NASA budget for human deep space exploration, extrapolating it out to 2028, and then dividing by the number of missions in the current plan.
  2. LOP-G component costs are derived sorta-kinda the same way.
  3. I'm going with $20M per launch for SH/SS.  That seems conservative.
  4. I'm close to clueless on the lunar surface components.  At the absolute minimum, we need a surface habitat and a power system.  I have other stuff listed here, but it's outside the 2028 window.
Here are the launch costs:

ComponentCost Per Use ($M)
SLS Block 12290
SLS Block 1B2290
SH/SS20
F9/D2150
Atlas220
FHE180
FH2R150
FH3R120
New Glenn80
Vulcan Heavy150
Good data, educated guess, sheer wild-ass guess

Here are the payload component costs:

ComponentNumber of UsesCost Per Use ($M)
Orion11,560
SS CM2020
SS Ferry1015
SS Ascender108
SS Lander1100
NASA Lander1200
NASA Tug1010
NASA Ascender1020
Refueling Payload130
Lunar Scout180
LOP PPE1514
LOP Esprit1514
LOP US Util Mod1514
LOP Node/Airlock1514
LOP International Hab1514
LOP US Hab1514
LOP Airlock1514
LOP Docking Node1100
Surface Power350
Surface Hab3133
Regolith Mover2100
Water Miner1300
Surface Hydrolox1300
Surface Resupply130
Random Payload130

Now we get to pricing out missions in the architecture.  Here's the strategy

The "NASA" side of the chart below includes planned missions, using SLS Block 1B co-manifesting and cargo as planned, with the full-up LOP-G architecture.  There's a little bit of guesswork here, but I've stuck as close as possible to what NASA has published.
The "Starship" side assumes that we do the following:
  1. Never develop SLS Block 1B.
  2. Use Block 1/Orion to transport crews only until the Starship/CM system is available.
  3. We continue to build the 3-stage lander architecture, and we continue to use as many commercial vendors as possible.
  4. Trim LOP-G down somewhat, but basically keep it.  (I want this to be as apples-to-apples as possible.  We can fight the LOP-G budget battle for another day, and the 3-stage architecture actually needs at least a small LOP to help with stacking and refueling.)
  5. We phase in Starship lunar cargo landings late in the process, and use it to deliver the hab and power modules.
  6. This assumes that crewed Starship missions to the lunar surface occur outside of this time frame.  My guess is that they could happen by 2029 or so, which might make the whole 3-stage lunar architecture a hideous waste.  I don't see a way around this, either from a rational NASA risk reduction standpoint or from the much more important political standpoint.  We'll see.
Neither side includes robotic exploration missions.  Those come out of a different pot of money and presumably occur in the next 2-4 years, so it's unlikely that the plan--such as it is--will change very much.

Here we go:

NASA Plan of RecordStarship Crew to NRHO ASAP, Some Lunar Surface Cargo
YearMissionComponentsCost ($M)YearMissionStarship Alternative ComponentsNum Need- edCost ($M)
2020EM-1 Orion Unmanned Test FlightSLS Block 122902020EM-1 Orion Unmanned Test FlightSLS Block 12290
Orion1,560Orion1,560
2022EM-2 Orion Crewed Test Flight in Lunar Free-Return TrajectorySLS Block 122902022EM-2 Orion Crewed Test Flight in Lunar Free-Return TrajectorySLS Block 12290
Orion1,560Orion1,560
2023PPE launch for LOP-GAtlas2202023PPE launch for LOP-GAtlas220
LOP PPE514LOP PPE514
2024EM-3 First LOP-G Crewed MissionSLS Block 1B22902024EM-3 First LOP-G Crewed MissionSLS Block 12290
Orion1,560Orion1,560
LOP ESPRIT514SH/SS480
LOP US Util Mod514LOP ESPRIT514
FHE180LOP US Util Mod514
NASA Lander200NASA Lander200
Random Payload30Random Payload30
2025EM-4 Deliver International Hab ModuleSLS Block 1B22902025SS Crew Test and Deliver International HabSH/SS9180
Orion1,560SS CM20
LOP International Hab514LOP International Hab514
F9/D2150
2025EM-5 Deliver US Hab ModuleSLS Block 1B22902025Finish LOP-GNew Glenn80
Orion1,560LOP Docking Node100
LOP US Hab514
2026EM-6 3-Stage Lander Uncrewed Test Mission (Crew on LOP-G)SLS Block 1B22902026SS-Assisted 3-Stage Lander Uncrewed TestSH/SS9180
Orion1,560SS CM20
NASA Ascender20NASA Ascender20
FHE180FHE180
NASA Lander200NASA Lander200
New Glenn80New Glenn80
NASA Tug10NASA Tug10
2027Cargo Pre-Positioning?SLS Block 1B22902027SS Landing TestSH/SS15300
NASA Lander200Random Payload30
Surface Hab1332027Cargo Pre-PositioningSH/SS13260
FHE180Surface Hab133
NASA Lander200Surface Power50
New Glenn80
Surface Power50
2028EM-8 Crewed Mission to Lunar SurfaceSLS Block 1B22902028SS-Assisted Crewed Mission to Lunar SurfaceSH/SS9180
Orion1,560SS CM20
LOP Airlock514NASA Lander200
Vulcan Heavy150Vulcan Heavy150
NASA Tug10NASA Tug10
Refueling Payload30Refueling Payload30
New Glenn80New Glenn80
NASA Tug10NASA Tug10
Refueling Payload30
FHE180
NASA Lander200
TOTAL:34,97716,809

If you add up the HEOMD deep space exploration budget, it comes out to about $37.1B through 2028, so the $34.9B doesn't match perfectly, but seems kinda in the right ballpark, once you account for the Second Law of Thermodynamics as applied to large sums of government money.  Close, as they say, for government work.

But the big, big, big result is that, even with all of the 3-stage nonsense, and most of the LOP-G nonsense, this shows that replaceing SLS/Orion with Starship/CM, and adding just a little bit of Starship direct lunar cargo landing, cuts the cost of the architecture in half.  

Let me repeat that:  IN HALF!!!!

Note that we've done something as conservative as possible with Starship, namely jam a habitable tuna can into its payload bay and use it as a glorified tug between LEO and NRHO.  We're sticking as closely to the NASA architecture as possible.

And we still cut the cost in half.

I can't even imagine the things we could do with an extra $17B over the next 10 years.  But here are some fairly mundane suggestions:
  1. Accelerate the whole project.  Seems like we can lop off 2-3 years from the 3-stage lander architecture.  I'd like astronauts prancing around on the lunar surface in 2026 much more than the faint possibility of 2028.  (I'm getting old; I want to jump up and cheer those guys without breaking a hip.)
  2. Build the lunar base up quickly, and go for the ISRU lunar water operation as soon as possible.
  3. Maybe we can spend a fair amount of money getting ready for Mars.
I hope I've conveyed just how skeptical you should be of going whole-hog into a lunar architecture that consists of only Starship.  But even with that skepticism, if SpaceX comes in even vaguely on time and vaguely in the right cost per launch, this is revolutionary.

As always, the fly in the congressional ointment is trimming back SLS and Orion.  But all we really have to do is to hold off a couple years on the Block 1B work and the rest pretty much takes care of itself.

This all seems like very good news.

No comments: