To the Drawing Board

On 28 September, 2019, Elon Musk gave an update on “Starship,” formerly BFR, in what has become something of an annual tradition. After n hour-long delay due to high winds, the outdoor presentation took place in front of the best possible backdrop--an assembled Starship Mk 1 suborbital test vehicle. A thousand miles away, a second suborbital Starship was taking shape at Cape Canaveral. Following a major design change in October 2018, the vehicles now sport 301 stainless steel construction. As Musk pointed out, SpaceX found that stainless steel has a number of advantages. Cost is one--$2,500 per ton, vs. $130,000 for the previously planned carbon fiber—making the raw material for the ship (minus engines) about $275,000 vs. $14,300,000, just using the “about 200 tons” figure. Obviously, both versions will include wasted material, and this figure doesn’t include the Super Heavy booster. Ease of fabrication is another advantage—traditional welding techniques can be used, reducing labor costs and letting the assembly team experiment with manufacturing techniques. Musk pointed out two additional advantages—because stainless steel melts at 1500C, the heat shield requirement is considerably reduced, again reducing material and labor cost. And, because stainless steel is strong (end even stronger at cryogenic temperatures), very little is required, counterintuitively making it a very lightweight solution. The new design is expected to tip the scales at 110 tons, once prototyping is done and a variety of small design improvements are implemented. Musk hinted at a stretch goal of 99 tons.

For our design, the change from fins as landing legs means we need a new place to connect ground power, but this seems fairly trivial. More landing legs (six) frankly seems like a good choice, particularly since Mars landing sites aren’t going to be perfectly level.

Much more important, though, is that we might suddenly be able to build and send far more ships than originally planned. When the ITS/MCT version of the ship was expected to cost perhaps a billion dollars, we could consider one—maybe two—per launch window. The suborbital prototypes are taking about four months to assemble. Just making the math simple assume teams of 10 working 20 hours a day (3 shifts, but including some slack) for 120 days, making $100 per hour. That’s $2.4M. The design work is the high cost investment, and the Raptor engines are going to expensive. But the hulls, not considering heat shields or pricey actuators, are low-cost in terms of material and labor. So low, in fact, that there’s a significant advantage to amortizing the design cost over lots of hulls. If you’re only hauling cargo to Mars in a bare-bones freighter, about the only part of the vehicle you’d like to get back are the six Raptors. You can leave the hull there to be scrapped and recycled into the million things you can build with steel.

At this point, it’s too soon to rework our entire mission reference architecture, but we can see the shape of an incredibly more ambitious plan—imagine ten bulk cargo ships per launch window, in addition to a crewed MCCS and a high-end MFR. The MFR’s primary job on the return might be hauling Raptor engines back to Earth.

What else might we do? For starters, we might put a thermal jacket on one of those freighters and send a load of LH2. Or a load of drill pipe. More on that later.

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.

Up to now, we have assumed mining in-situ water, which is probably the most difficult part of this process. Allow me to recommend two online resources: http://www.nasa.gov/sites/default/files/atoms/files/mars_water_isru_planning-hangout-5-24-16.pdf and https://planetaryprotection.arc.nasa.gov/file_download/90/29-Sanders.Mars.ISRU.PP_Sanders.V2.pdf as a homework assignment, and then let’s talk about Zubrin’s Mars Direct fuel plan. Both of these resources walk through variety of research done by NASA on what it will take to mine water on Mars. If you’re wondering what sort of technology development today’s NASA is capable of, grab a big cup of coffee (if it’s morning) or a big bourbon (if it’s evening) and relish this work.

The summary version from both presentations is that if we’re going to rely on surface water, then we need to be absolutely sure we can access it and get it to the ISRU plant. This will require orbital reconnaissance, ground surveys, and precursor missions to verify technology… all of which requires time, technology development, and money. We’ve already concluded that’s very problematic, leading us to the uncomfortable situation of putting manned ships on Mars with no proven way to get them home.

The primary problem isn’t LOX; that’s trivial. Power plus CO2 plus equipment equals oxygen. The problem is methane, CH4. There’s plenty of carbon available on Mars (see: CO2), but hydrogen…not so much. Hence the need to mine water from regolith or grab it directly from Mars’ atmosphere, both with a single objective: hydrogen. Hydrogen as feedstock for our ISRU plant; hydrogen to make methane.

Now, consider an alternative. What if we could put off mining water, and instead take our own hydrogen to Mars? As Zubrin points out, the mass penalty for bringing your own hydrogen still results in an 18:1 improvement over no ISRU. What would this look like?

At a first approximation, round figures, 61 tons of hydrogen would be enough feedstock to completely fill Nostromo’s 240 ton fuel tank. (Our friends at http://www.wolframalpha.com can help with this one, “hydrogen in 240 tons of methane.) That doesn’t seem like very much, considering that our ships are all capable of transporting at least 150 tons of cargo. Unfortunately, hydrogen presents a serious design problem—even in liquid form, it is extremely low density. In fact, if you take the entire pressurized volume of a BFR (825 m3, about the same as a 747-8 freighter), that’s what it takes to store 58 tons of liquid hydrogen. Sure, 58 tons and 61 tons are close, but the point is that liquid hydrogen takes 14 cubic meters per ton. That means that to fill a ship’s tanks with methane for the return voyage, the hydrogen required will consume the entire available payload of another ship. Not its cargo mass, not the volume of its holds—it needs a specialized tanker whose sole purpose is to transport LH2, replacing the entire pressurized compartment.

On the one hand, this is a real problem. An obvious workaround to the problem of in-situ water mining has always been the idea of bringing along the necessary hydrogen. Now, we realize that doing so actually requires another ship whose sole purpose is bringing that hydrogen. But, with the downsized SpaceX plan, we’re actually just looking at a specialized tanker travelling with Nostromo and Heart of Gold. In fact, Musk even mentions in his 2017 Reddit AMA that some of the optimized tanker designs look pretty weird…suddenly, bringing along hydrogen seems feasible, at least for the first manned landing. It brings up a number of problems, of course, starting with the infrastructure needed to transfer hydrogen from the special tanker to Nostromo. But the risk it buys down is tremendous. And, if we need to continue the pattern, additional hydrogen tankers in our architecture seem a small price to pay to ensure we can get the crews home.

Of course, sending an extra hydrogen tanker one-way for every MFR or MCCS going to Mars doesn’t seem sustainable long term, but it certainly gives us some valuable planning flexibility in the near term. The first couple of manned missions are where we really need to buy down risk.

One hundred fifty tons of cargo per mission, one thousand five hundred tons by the time we complete ten missions.  The biggest thing about Musk’s idea is just that…it’s big.  With refilling in orbit and reusable ships, all the previous calculations about how much mass we can take to Mars simply go out the window.

In the early days of space exploration, there was a rule:  don’t do anything on the satellite you can do on the ground.  Mass and computing power were too precious to waste.  In the late 80’s, that began to change.  Now, the rule is:  don’t do anything on the ground you can do on the satellite.  Ground operations are costly; the satellite on orbit is a sunk cost.  Anything it can do, you don’t need someone else to manage.

Much of Mars preparation today is focused on low-mass operations, trying to figure out how to squeeze every ounce out of a given capability, and every second out of astronauts’ time.  In the process, though we stay trapped in a paradigm that looks very much like ISS science operations.  Ironic, considering how much many Mars evangelists dislike the ISS and NASA’s approach to it.

We now have a capability that lets us stop worrying about every ounce, and start thinking about big ideas.  It also opens the door for more people to think about big ideas.  I no longer need to spend ten years getting my brass board prototype down to a perfectly-optimized flight weight.  Instead, I can spend two years getting to a flight-safe configuration, and launch it.  I’ll learn in a real Mars environment what works and what doesn’t, and I’ll have a robust system with a significant toolkit that will let me make adjustments on site.  I wasn’t being rhetorical earlier with the front end loaders, dump trucks, and drill rigs.  We can afford to try things, and have redundancy in what we try.

It’s a bit like the difference between aircraft development in the 50’s and 60’s versus aircraft development today.  Today, we have 787 Dreamliners, F-35 Lightning II’s, and MV-22 Ospreys.  Amazing capabilities, but expensive, and they take decades to develop.  In the 50’s, we turned out new aircraft models like cord wood.  None of them had the efficiency of a Dreamliner; none of them had 1% of an F-35’s capability, none of them approached what an Osprey can do.  But, we learned at an incredible rate, because there was so much to learn.  Today, we would call it, “Build fast; fail fast; learn fast.”  We’ve never had that option in space exploration before.

This brings up another item…there are plenty of critiques that can be made for SpaceX’s overall plan.  Those are all fine, but irrelevant for our purpose.  The focus here assumes the SpaceX plan works out (more or less) the way Musk envisions.  That’s not fandom; to make progress in planning, you have to accept certain assumptions as valid, or you just spin your wheels.  Big, reusable ships are the whole point of this architecture.

In fact, the single most important difference between this and other projects (such as Mars Direct) is that this effort assumes all reusable ships.  So, the effort we’re undertaking assumes SpaceX solves the obvious problems (zero gee cryogenic propellant transfer, storing deep cryogenic methalox, building and launching a large spacecraft, etc.), because the whole point is that SpaceX is saying, “We’re going to do X, and we’re leaving it to others to solve Y.”  We’re the others.  We’re solving Y.

We’re talking about X, but that’s just to set the stage.

On September 17, we’ll see whether SpaceX tells us anything more about X.