I’m now going to fundamentally disagree with Elon Musk on a technical detail. Granted, this one is literally not rocket science. Nevertheless, it matters a great deal for people planning to put cargo on the surface.
In his 2017 IAC presentation, Musk said he’d been asked by many how SpaceX intended to get equipment from the cargo hold, high up on a BFR, down to the surface. He pointed at the artist’s conception of a BFR unloading on the Moon, and casually shrugged, “It’s a crane.”
Yes, it was. A very 1920’s style crane. The kind that relies on longshoremen and stevedores with years of experience bringing cargo off a ship. Experience needed to ensure the vessel doesn’t list and the cargo doesn’t unbalance the crane. The kind requiring manual labor to load and unload the crane itself.
Possibly fine once we have crew on Mars, this design is entirely unsuitable for robotically unloading cargo from Constellation and Finity’s End. We need something a lot more like today’s intermodal cranes. They’ll be much smaller and lighter, of course, more like air freight; but the idea is that the crane’s spreader and trolley can connect to our cargo in the hold, move it outside the ship, lower it to the surface, and disconnect, all autonomously. To accommodate this, our cargo will need intermodal-style corner castings on top corner posts which the crane’s twistlocks can lock into without human intervention.
For securing the cargo to the deck, we should look to the air cargo industry’s Unit Load Devices (ULD). These are lightweight containers or pallets (pallets secure their cargo with nets), but they’re loaded and unloaded by rolling them across scissor trucks. Fine for loading on Earth; that obviously isn’t going to work for unloading on Mars, hence the intermodal hybrid. One particular industry innovation we’ll want to implement, though, is the ability to automatically lock the ULD in place on the deck.
Without spending too much time designing at this early stage, we can still sketch out how this would work autonomously. The first thing down needs to be a rover/tractor with a robot arm. That gives us mobility on the surface and the ability to do work. To be autonomous, it will need lots of cameras; that will also be essential when it’s necessary to intervene from Earth. Adding corner posts with corner castings seems reasonably straightforward, as is ensuring the center of gravity is close to the geometric middle of the vehicle.
The second item down is a trailer, preloaded with equipment to set up the solar power field. Without knowing SpaceX’s design for the field, we can only speculate on whether this will be solar panels with integrated power cables, whether the cables will be run separately, and which gets positioned first. Regardless, the tractor connects to the trailer, heads out to the predetermined spot, and starts laying out solar panels and cabling. After the first load is deployed, the tractor trailer returns and spots itself under the crane to receive the next package. Here, the cargo again uses corner castings to connect to the crane and ULD auto-locks to secure to the trailer. Presumably, we have a stack of several solar panels, so we’ll need customized interconnectors to hold the stack together, analogous to “interbox” connectors used for intermodal containers. These are normally placed by stevedores, and we just have a robot arm, so the design will need to be creative. The solar panel stack will also, by necessity, be slightly smaller than the trailer, so our crane’s spreader will need to be adjustable. An advantage of adjustability is that we have more flexibility with center of gravity. The obvious disadvantage is more design complexity. But, there’s a whole industry that does this.
Goal #1 is providing enough power to keep the ships healthy. (We’ll talk about Goals 2+ in future posts.) The primary risk for this design appears to be whether everything we build will work the same in one-third gee as it does on Earth. As we did earlier with landing algorithms, checking this out in one-sixth gee seems both practical and prudent.
My sense is that SpaceX is well aware it will need to do this—design, build, and test the whole system we’ve just described until there’s virtually no doubt it will work reliably. That’s a lot of work—but not compared to designing, building, and testing BFR itself. SpaceX is likely waiting until BFR is demonstrably farther along before it starts serious work for a Mars solar power field, so it’s also reasonable they’ve put “crane” on a list of placeholders they’ll come back to later. Still, the cargo offload system is also the first component they could partially offload to others. That would bring in more people to work on the problem without diluting SpaceX’s core expertise. Tesla doesn’t build tractors, but John Deere, Caterpillar, and Mahindra do. They can also probably figure out how to make their products lighter and run on batteries a lot quicker than anyone else. SpaceX could probably also team with intermodal companies to figure out the best designs for unloading.
Certainly, there’s a giggle factor asking outsiders to start seriously considering how to build equipment for Mars. That’s a valid reason to wait until BFR is farther along. But, if we go back to that initial list we made, we see that there are a LOT of things people need to get started on. People like us. After all, that’s the whole point of this website.
Next, let’s look at the one big assumption we already made—the ISRU plants. In “The Case for Mars,” Zubrin describes the ISRU prototype plant he built at Martin Marietta in the early 1990’s (pp. 153-156, and see http://www.marspapers.org/paper/Zubrin_1994.pdf). A 20 kg plant was demonstrated that could generate 400 kg of LOX/liquid methane, using hydrogen feedstock, a simulated Martian CO2 atmosphere, and 300 watts of power over a ~500 day run. Nostromo will need more—a lot more. Specifically, it will need 1,100 tons of propellant per ship, or nearly 3,000 times more than the Zubrin/Martin Marietta prototype plant. Just doing the algebra, it’s actually 2,750 times more propellant, which means the 20 kg plant becomes 55,000 kg. That’s 55 tons, without considering Zubrin’s point that scaling up the plant makes it more efficient, since instrumentation and control equipment become a smaller portion of the overall effort. The 1994 paper concludes 10 kg is a realistic goal for the pilot plant, but does not examine the mass of tank insulation, lining, circulation pumps, refrigeration, and all the sort of things that come into play when you scale up a deep-cryogenic storage tank by a factor of 3,000. Previously, we arbitrarily allocated 100 tons to the ISRU plants, so 55 tons each appears practical. We’ll assume we can shave 5 tons off the mass of each plant. We’re at 100 tons since we need two plants to fill the tanks for both Constellation and Finity’s End in time for their scheduled return. More importantly, each plant will need 825 kilowatts of power, which is why nuclear reactors are often mentioned as a prerequisite for ISRU propellant production.
What sort of solar power field is required to generate 825 kilowatts of power? Round figures, if it were in Earth orbit that would be 8 times the size of the International Space Station’s solar panels (which generate 84-120 kW total). Of course, Mars only receives 43.3% of the solar irradiance of Earth, so this suggests each solar power field is sized with an equivalent to 1.9MW at Earth, or 19 times the ISS capacity.
ISS is a handy reference, but these solar panels are on the ground. So, again the question is posed: what sort of power field is required to generate 1.9MW of power? What’s the mass, what’s the area, what’s the volume?
Visiting https://mitsubishielectricsolar.com, we find a variety of commercial grade solar panels rated down to -40 degrees Celsius. Obviously, we’ll need more than that, but this is a good starting point. The neo solar power (NSP) D6M series provides consistent voltage down to 200 W/m2 irradiance, well below Mars’ 590 W/m2, just reducing the current as the irradiance decreases. The 350W rating applies to Earth irradiance, around 1000 W/m2, which gives about 9 amps. Just to keep the math simple, we’ll use their 350W version, which tips the scales at 23 kg and has approximate dimensions of 1m x 2m x 35mm. At Earth irradiance, it’s spec’d at 38.34V and 9.13 amps, which yields 350.6 W. These values are comparable in similar products.
At Mars irradiance, the spec sheet indicates about 5.6 amps, still 38.3 V, or 214 W. Now that we’re looking at power production with Mars numbers for the equipment, we no longer need the 1.9 MW equivalent; we’re actually interested in what it will take to provide 825 kW with the gear we’ve selected (remembering we need this much for each plant).
To get 825 kW, we’ll need 4,000 of these panels (3,855 to be precise, but let’s go with the round figures and end up with 856 kW). That works out to 8,000 m2 for our solar power field, or about one and a half football fields (end zones excluded). The algebra gives a total mass of 92,000 kg, or 92 tons, for the entire field. And, we need two of these.
At a first approximation, this is alarming, because we only have a remaining budget of about 50 tons, based on 100 tons for the twin ISRU plants. But, these panels are designed for Earth gravity, Earth winds, and Earth snow. We won’t face any of that on Mars, so the design can be much lighter. Can we get it down to 50 tons? A lighter structure is obviously possible, but a large proportion is made up of the photovoltaics and glass. It seems unlikely we can reduce the mass this much.
We do have options here. First, Zubrin assumed 500 days; we’re arbitrarily assuming 550. We’ll need to be more precise about which synod and which trajectories we’re really going to use to get any closer on this approximation; besides, we’re only talking 10% difference. Second, if we don’t have enough power to get the job done, then could always just send one ship back. There is certainly some suggestion that SpaceX intends to leave quite a few ships on Mars. Lack of power for the ISRU plants could be a very practical reason.
There’s another alternative, though. The whole reason to work out a six-ship mission architecture was to replace variables with constants. We previously argued for one manned ship and one freighter per cycle, but then eliminated one freighter in favor of a refinery ship. The math now suggests we need to modify this plan, and sent a freighter along with Heart of Gold and Nostromo. That freighter’s sole mission is to carry enough solar panels to fully power Nostromo’s twin ISRU plants. With 150 tons available on the freighter, and 50 tons available on Nostromo, we’re now at 200 tons. That’s more than our 184 tons, which our initial algebra suggests, so we’re well within the engineering ball park. If we can’t get the ISRU plants and solar power field working within these budgets, we might be forced to skip one ship’s return until we can bring more solar panels. But, at a first approximation, this seems to be an achievable engineering goal. There are plenty of challenges, but a reasonable budget to solve them. There will be plenty of time later to argue over kilograms.
Based on a variety of sources, as of late 2018, this architecture is actually a little slimmer than SpaceX’s proposed plan. The press reporting is for two robotic freighters in the first synod (we’re planning for one freighter and one ship with life support) and four ships in the second—two robotic, two with life support. The architecture here requires one ship less, so we’re still very much on the conservative approach. However, these power calculations are specific to the ISRU plants, which is essentially “reserved” cargo. We still need to look at the power for the base and its equipment. That will be for a later discussion.
The change we’ve just made also leads to a new question—the name of our new freighter. To keep things simple, recognizing this is a bit of a one off, but also helping us remember this ship solves a significant problem by bringing Solar panels, let’s call this freighter Serenity.
Back on our reference architecture, Serenity will accompany Heart of Gold on its cycles back to Earth. There’s a nice symmetry here.
Life support, landing sites, water mining, ISRU propellant production…there are dozens, if not hundreds of things that would benefit from prototyping in precursor missions. Of all those, three items stand out as “must do” if SpaceX’s plan is to become real: 1) on-orbit, large-volume refilling, 2) long-term storage of cryogenic propellant in carbon-fiber tanks, 3) reliable operation of large, reusable spacecraft. Everything else is simply expanding what we already know how to do, and tailoring it to Mars.
Let’s take each in turn.
There are some obvious possibilities with regards to on-orbit propellant transfer. ISS routinely transfers fuel from Progress supply ships, but that’s on a small scale in comparison with what SpaceX is proposing. A BFR needs 1,100 tons of LOX and liquid methane. The international standard docking port, which includes fuel transfer lines, simply isn’t going to cut it for this purpose. There are also a lot of questions about keeping the liquid settled over the pumps, etc.
In Musk’s 2017 IAC presentation, the presentation showed a much better solution than the one shown in 2016. In 2016, the ships were side-by-side, requiring all sorts of additional docking mechanisms, pumps, etc. In 2017, they dock back-to-back, and now with thrusters, we can simply use “gravity feed” for propellant transfer, augmenting with pumps if needed. Thrusters let us maintain head pressure at the turbo-pump inlets; pumps let us significantly speed propellant transfer.
We can test gravity feed versus pumps using derivative Falcon 9 gear—two second stages configured with pumps, extra fuel, thrusters, and docking ports. They transfer propellant back and forth under varying conditions of thrust. If we make them reusable, then we can run more excursions. Musk has hinted he finds the idea of reusable Falcon 9 second stages tempting, but doesn’t want to be distracted. Well, here’s the excuse. Once we have a good sense as to how much force we need to apply, we’ll test that on the full-scale tanker and ship, and iterate until we get it right.
A much harder question than on-orbit refilling is how the carbon fiber tanks of the ship will stand up to years storing LOX and liquid methane. This is hard because ultimately, the only way to be sure is to expose the tank material to the cryogenic liquids—and temperature cycles—for a long time. I wouldn’t be surprised to learn SpaceX already has subscale experiments in progress. Ultimately, we need to be sure the tanks will handle the cryogenic liquids indefinitely. It isn’t enough that they last an entire mission, or even two or three. For the ships to be reusable for 30 years, their tanks must handle LOX/liquid methane for 30 years, because the only time they won’t be full of propellant is during their six-month transits. The rest of the time, some tank will be slowly filling with liquid oxygen or liquid methane—whether that tank is on Nostromo or one of the cyclers.
While on Mars, the propellant needs to be liquid, but not necessarily deeply cryogenic. Additional refrigeration can occur shortly before launch, if necessary. (It’s unclear whether SpaceX’s figures show deep-cryo methalox is absolutely necessary for Mars-Earth return, but that features prominently in the discussion.) Ultimately, this is a materials problem, driven by the physics of cryogenic propellant storage.
Oxygen’s boiling point is -183C, but SpaceX’s desire is to use LOX just above the melting point, which is -219C. Mars rarely gets below -150C, so additional refrigeration will be required, either way. Methane’s boiling point is -161C and its melting point is -183C. So, deep cryogenic LOX could be used in a heat exchanger to drop methane to just above its melting/freezing point, but we’ll still need liquefaction equipment. (Check “Hampson-Linde” cycle on Wikipedia; also “liquid air.”)
On Earth, liquid natural gas is kept cold by allowing a small amount to boil off. The energy of evaporation transfers heat out of the liquid. The same technique can be used here, with active refrigeration then used to re-cool and liquefy the gas, and then return it to the tanks. I leave it to cryogenic fluid experts to recommend the best technique for returning the re-cooled liquid, and whether it’s best to keep the liquids static or mixed. Of note, it appears the Amos-6 mission was lost because the liquid oxygen actually froze, so we should spend extra time examining the implications of this. In particular, what long-term temperature do we want to use, and what do we do when the time comes for extra cooling prior to launch? Another obvious question is what temperature the propellant fluids should be at when they are transferred from Nostromo to the return ship. Given the scale of this industry on Earth, it seems a safe bet these are easy questions to answer. Only the tank material interaction is in question.
The third major problem almost sounds a bit absurd…operating giant reusable spaceships? That’s all? But in all seriousness, this isn’t just “make the science fiction work.” This is a very specific problem: SpaceX is going to build very big ships. They will have lots of new systems. Establishing what works, what breaks, and what the malfunction scenarios are will be much like the early days of trans-oceanic flight. Early on, designers wondered if long-range airliners should be designed to float if they crashed. The answer was that a better solution was to ensure they didn’t crash. Early on, overwater flights required four engines, in case one failed; even today, there’s a special name for long-range twin engine operations overwater—ETOPS (Extended range, Twin-engine Operations). Rocket engines have a tendency to explode when they fail; jet engines only fling their fan blades into the wing and fuselage. By comparison, getting hit by a shredded engine blade is minor. But jet engines are still designed to contain their debris, and they are tested for containment using deliberately damaged blades. We need the same concept for rocket engines.
Disasters happen. Airliners crash. We won’t have to worry about the most common problems—hitting another aircraft in the vicinity of an airport, controlled flight into terrain by a distracted crew, even spatial disorientation. But, we will need to establish what fails, how often, and how catastrophically. Multiple engines shutting down can be handled with software. One engine exploding may well destroy the vehicle. We need to figure out how and why that happens, and take ruthless steps to ensure it doesn’t or protect the rest of the ship if it does. So, heavily instrumented equipment with very conservative shutdown protocols seems the right approach—fail safe, in this case, is also fail-op. And the only way to be confident we’ve thought of the likely failure scenarios is to fly those big ships over and over, off Earth and back to Earth, preferably unmanned. Losing a $200 million ship is terrible; losing a $200 million ship and a 4 person crew could be mission-ending. Of course, once you have that part figured out, then a 4 person crew saving a $200 million ship gives you heroes with the Right Stuff. That won’t be a landing scenario, but could well occur on orbit or in deep space.
There are other issues, and we’ll go through them in turn, but these are the deal breakers.
SpaceX originally planned for a number of precursor missions using Red Dragon. Now that Red Dragon has been cancelled, is there a need for something smaller before going all the way to BFR? NASA had Gemini, which was absolutely critical to the success of Apollo. On the other hand, SpaceX cancelled the Falcon 5 as unnecessary, going directly to the Falcon 9. Or, have we actually settled on the “medium” sized ship instead of ITS? That seems plausible, but now we don’t have a “small.”
The biggest problem we face without Red Dragon is characterization of Mars’ atmosphere before we send the first BFR. Red Dragon would have researched Mars’ atmospheric dynamics, to support reentry and landing algorithms. This data was considered so valuable NASA was willing to trade mission support in order to get the data SpaceX would collect. What does it mean for our architecture that this is no longer an option?
Nothing even remotely BFR-sized has ever set down on Mars. For that matter, nothing BFR-sized has ever set down anywhere. An obvious first question is how we verify the guidance algorithms for other than one gee, and how we establish that they’ll work during Mars descent?
If we assume some degree of linearity, we have an opportunity to check the guidance algorithms before heading to Mars—we can land on the Moon. If the guidance works in one gee, one atmosphere (Earth), and it also works in one-sixth gee, zero atmosphere (Moon), then that should at least boost our confidence that it will work in one-third gee, 1% atmosphere (Mars). Atmospheric dynamics is critical here, since that’s the primary mechanism BFR will use to match velocities with Mars, before igniting its engines for final landing.
Frankly, this is a little scary. In our current plan, Constellation (or Finity’s End) will arrive at Mars at the slow end of a Hohmann transfer, smash into Mars’ atmosphere to get a big delta-V, and then land on the surface under rocket power. There will be no opportunity to determine how the BFR aeroshape actually responds to Mars’ atmosphere at reentry velocities before we try it for the first time.
If Constellation succeeds, we’re fine. If either Constellation or Finity’s End fails, we have serious concerns, but we may be okay, depending on what went wrong. If both fail, for any reason, then Heart of Gold isn’t going in the next launch window. There’s no way we’ll risk a manned ship when the unmanned predecessors failed.
Without Red Dragon, do we have a way to gather data without risking a ship?
One possibility would be to send some type of BFR to Mars with no intention of landing. Instead, it would aerobrake into Mars orbit. Based on what we know about the SpaceX design, this would leave lot more room for error. In fact, if we send a fully-fueled tanker as the first ship, we buy ourselves a tremendous amount of leeway. If everything goes exactly according to plan, the end result is a tanker in orbit around Mars with a fair amount of propellant on board. If things go completely wrong, the tanker has enough propellant that it can correct for a lot of problems. Too shallow in the atmosphere, we have propellant to get us into Mars orbit from an interplanetary transfer. Too deep, we have propellant to raise the orbit to whatever degree is necessary. Either way, we now have a vehicle in orbit that can send the data it collected back to Earth.
A tanker is probably the least expensive second-stage ship SpaceX will build, so it’s a good candidate for a precursor of this type. The tanker could also provide a way to get propellant to later manned ships if everything about ISRU propellant production goes wrong. (One tanker won’t do it, but it’s still a proof of concept.)
Extending this thought a bit more, the idea of sending tankers to Mars orbit enables a variety of other options. For starters, the Lockheed Martin orbital “Mars Base Camp” suddenly seems less farfetched (https://www.lockheedmartin.com/content/dam/lockheed/data/space/photo/mbc/MBC_Updates_IAC_2017.pdf), although SpaceX tankers aren’t designed to support a hydrogen economy. Orions attached to a Mars Base Camp suddenly start to look like the orbital equivalent of rovers, able to explore Phobos and Deimos and then refuel. The Mars Base Camp itself could probably be launched fully assembled from Earth.
It could also support manned missions doing telerobotic operations. Without reusability and orbital refilling, this is borderline fantasy. With both, suddenly we exchange fully autonomous surface operations for remote control surface ops. That seems very doable. We might be able to set up water mining operations entirely from orbit. We even extend our “resupply until we get it right” concept without being stuck on the wrong end of a gravity well—with a Mars Base Camp in place (presumably supported by a MCCS), we would just need to send tankers until the manned ship has enough propellant to return.
Backing up, for the price of one tanker, we get lots of atmospheric data, an opportunity to refine our algorithms before committing to a landing, and a propellant depot in Mars orbit. It can also be done without any of the surface operations gear, autonomous deployment technology, or life support. That means it could be done before any of that is ready. Constellation and Finity’s End could launch in the next window, with much greater confidence in their ships’ handling in Mars’ atmosphere. A manned ship could even go and remain in orbit (Constellation?) to conduct telerobotic surface ops with the MFR’s cargo.
This doesn’t appear to be in SpaceX’s plan at the moment, so for planning purposes, we would need consider this a branch plan. But it seems to offer a lot of benefit. Enough, perhaps, that this branch should become the base plan.
Assuming there’s room on Nostromo, what cargo should it bring? What about Heart of Gold? Constellation? Finity’s End?
Mars launch windows open every 26 months; Musk’s 2016 IAC presentation clearly indicates a high-speed trajectory for the original passenger MCT, so this presumably holds for the scaled-down ship. As we discussed, for a cargo version, a trajectory closer to a Hohmann transfer will increase the allowable mass.
For simplicity, if we assume a 6-month near-Hohmann transfer, the travel time means that when a ship arrives at Mars, there are about 20 months until the next launch window from Earth opens. (A pure Hohmann transfer of 8 months en route gives 18 months on Mars before return.) Assuming 30-60 days on Mars to test out any ideas/technology and give feedback to the engineers on Earth, then the lessons learned from “this” launch window have 6 months to be further researched and 12 months to be applied before the next ship launches from Earth in the next window. That’s a very short interval, but it also means that we have a very fast feedback loop to get improvements into the next cycle. We also have a lot of optimization available for mass, travel time, and launch opportunity, once we reach that stage of design.
This suggests that when in doubt, we should try to take as many different cargos to Mars as practical. Rather than trying to determine the best technical solutions, if there are competitors, it will be better to take smaller versions of each, or one version of each versus two or three.
We also want to take as many different “things to do” as practical. In other words, we want to take different types of water mining equipment as soon as possible, long-range rovers as soon as possible, inflatable domes as soon as possible, and so on. The sooner we get new technologies on Mars, the sooner we’ll get feedback to the engineers building the next generation.
We do have an important question about how much cargo is actually available. In a master’s thesis evaluating the original ITS’s life support system, available at https://mediatum.ub.tum.de/doc/1388335/, Bernd Schreck questions whether that cargo is actually intended to be life support capacity and consumables for the Earth-Mars voyage. The SpaceX presentations make a fairly big show of unloading the cargo at destination, and the early missions suggested that these cargo holds might even be unpressurized, which suggests these aren’t enroute supplies. Schreck’s analysis of trip duration also doesn’t consider Musk’s statements that the early ships would also serve as surface bases. That said, a 307 page detailed life support analysis isn’t something to be discounted casually.
A reasonable compromise position is that delivering cargo is a primary purpose of the MFR’s. Even if the math eventually shows we need all the MCCS cargo for life support and consumables, we still have the MFR cargo. So, if that’s the case, then our next three ships after the architecture described will probably need to be another set of freighters.
This is actually a fairly positive discovery. Even if it turns out we have a significant miscalculation in our mass requirements, correcting it is simply a matter of adjusting the MFR/MCCS ratio. Since SpaceX will be building only BFR-class ships at this point, adding BFR freighters is much more economically feasible than extra ITS ships would have been.
But, before we go too far forward, we should take a step back and consider what we did during the precursor missions before that first journey. That will help us establish what things we can assume have been proven out, or at least demonstrated.