Monday, August 08, 2005

Shuttle replacements

NASA is mulling two vehicles to replace the Shuttle. Both are based on Shuttle components. This post is really about three things: one, why the proposed NASA designs are about as good as they can do; two, how the proposed designs are worse than the alternatives, and three, what would have to happen for the U.S. to use a better alternative.

The claimed reason for basing on Shuttle components is reduction of development cost, infrastructure build cost, and time-to-reliability. This last issue is not to be taken lightly -- the flight histories of much of the Shuttle hardware, along with the years of tweaking of that hardware, are irreplaceable. If we're going to launch people on a new rocket, we want to be in a position to say with familiarity that the rocket is safe, and we can only say that about a rocket with a long flight history. Nothing else is going to have a long flight history by 2010. Well, almost nothing, and I'll get to that near my conclusion.

The other reason for basing the new vehicles on Shuttle components is jobs. The Shuttle program currently employs many tens of thousands of people at NASA and its subcontractors. Those people and their employers want to stay in business, and they have gained quite a bit of leverage on their representatives in Washington.

Considerations like this are what make NASA's proposed vehicles so sad in comparison to a clean-sheet design. It's not so much that the NASA vehicles are more dangerous or less capable than the alternatives, but that that NASA will do so much less with them than they could do with the alternatives, for the same amount of money. My current favorite clean-sheet design is the one being pursued by SpaceX, and I think it's interesting to compare the two.

Before I go any further, I should note that unlike NASA and its prime contractors Boeing and Lockheed-Martin, SpaceX hasn't launched any hardware yet (first launch is scheduled for the end of September, 2 months from now). What they do have is 100 or so employees almost all of whom have experience putting stuff in orbit. Assuming SpaceX gets the contract to haul supplies to the International Space Station via their as-yet-unbuilt Falcon V, and assuming they establish regular flights by 2008, they still won't have a long flight history by 2010.

But back to the comparison. Both NASA and SpaceX are pursuing a two stage to LEO design. Both first stages are reusable (they parachute back into the ocean), both upper stages are expendable, both use capsules to recover the crew and have little ability to return large masses from orbit. All of these are good choices.

The limitations of the comparison are that the NASA "Stick" design is supposed to lift 20000 kg to the ISS, and the SpaceX Falcon V lifts 5450 kg. So if NASA were to use SpaceX to send people to the ISS, it would have to be three people at a time instead of six, and the capsule would have to be simpler than the full-blown Crew Expeditionary Vehicle they envisage today. Also, SpaceX hasn't yet announced a crew capsule development for the Falcon V. They are going to man-rate the booster, and I think they intend to use it for people after developing a good flight history with cargo.

Both the SpaceX stages burn liquid oxygen and kerosene. Such engines are well understood, and give predictable, respectable but not spectacular performance. Because the engines are liquid fuelled, they can be throttled and shut down at will. And crucially, kerosene is benign (basically like the gasoline you put in your car) and liquid oxygen is not terribly difficult or expensive to handle. SpaceX uses recently developed materials, manufacturing techniques, and avionics to achieve larger payload per dollar than has historically been possible with this propellant combination. As a result, the SpaceX rocket can be handled by a small team of people and manufactured for small amounts of money.

NASA's first stage is a stretched version of the solid rocket booster currently used by the Shuttle. That SRB costs about as much to recover, refurbish, and refuel as it does to build a new one, so the reusability is cosmetic. The proposed first stage cannot be shut down early (for instance, if the cabin loses pressure). It cannot be throttled, and the thrust is somewhat unpredictable (which means the whole rocket structure has to be built with more margin to sustain higher peak G and aerodynamic loads, and the upper stage needs more delta-V margin to recover when there is less delivered velocity).

The SRB is a low performance first stage. It is heavier than the SpaceX hardware: The aluminum-lithium SpaceX booster has a mass fraction of 94%, where the steel SRB has a mass fraction of 85%, and the SpaceX booster has a higher exhaust velocity, 2980 m/s versus the SRB's 2636 m/s (measured at sea level air pressure). The result is that the SRB has to be much bigger per kg of payload as the SpaceX design.

There is a school of thought in the rocket design community that espouses the Big Dumb Booster. Their point is that a larger rocket is just fine, so long as it costs less. Cost is mostly related to things other than size, like the number of people necessary to build, transport, and launch the rocket. But the Shuttle SRB is so big that its size implies complexity. Pouring single grains of propellant that large requires enormous facilities to mix and cure the components. The entire SRB is too large to pour or transport as a single piece, so it is poured as four pieces, seperately transported to Cape Canaveral, then assembled there. And while the clevis and tang joints between the sections have now been sufficiently engineered to be safe, they impose handling restrictions (NASA cannot leave the Shuttle out in the cold), and cost more money.

As a result of all this complexity, each Shuttle SRB costs $40M per flight (that's half of the first stage, with no guidance or communication system). A two-stage SpaceX Falcon V sells for $15.9M (that's everything but the capsule on top) [number has been updated - thanks, Jon]. For a sense of how cost scales with size, note that the Falcon V lifts almost ten times the payload of the Falcon I, but costs a little more than three times as much. Each SRB has a bit more than 7 times the thrust of the Falcon V first stage.

NASA faces a difficult choice for the upper stage. There is only one man-rated engine suitable for upper-stage use in production in the U.S., and that is the Shuttle's SSME. That engine is complete overkill for an upper stage engine, because it's designed to work in the lower atmosphere too, which makes everything more complicated. Throwing it away each flight will cost a lot. Any other engine (i.e. the in-development Cobra or the in-production RS-68) will require at least some development money and time to certify as man-rated, but will cost less per launch.

All the engines NASA is considering for the second stage burn both liquid hydrogen and liquid oxygen. This decision alone pushes the cost of the NASA solution far away from nearly anything else. LH2 will give better performance than kerosene (4300 m/s exhaust velocity versus 3330 m/s, measured in a vacuum), and thus a smaller rocket, but the relevant difference is cost. NASA can't go with a lower cost propellant combination because there are no US-manufactured man-rated engines which burn anything less costly.

Now that we know why NASA is mulling an expensive crew vehicle, and what a less expensive vehicle looks like, and interesting question is, how could we end up with the cheaper alternative? I assume, in particular, that the SpaceX honchos think about this problem relatively often. I also think that Michael Griffin, the NASA chief administrator, thinks about this fairly often as well.

It's a political problem. NASA has to get crews to the international space station after the Shuttle is retired. The Russians have complained about being solely responsible for launching crews for the last few years, but that is because we weren't paying enough for the launches. If we were willing to pay, the Russians would be happy to provide launch services (they recently asked for $63M per launch). They have a very dependable launcher, which costs a fraction of a Shuttle launch. Buying time to design and build a better, cheaper U.S. launcher would be cheaper than building the stopgap crew launcher currently envisaged.

We would then be faced with the smaller problem of launching the ISS portions originally intended to go up on the Shuttle. Both EELV boosters have the mass capability to launch these cargoes (though NASA would have to rework each piece to sit in the EELV payload fairing rather than in the Shuttle payload bay). EELV launches are expensive (around $254M), but they're half of a Shuttle launch, and don't endanger crews.

Michael Griffin is going to take the expensive but politically necessary route. He'll pay the big NASA prime contractors many billions of dollars to develop a stopgap launcher, while ensuring that SpaceX and perhaps t/Space and Kistler get sufficient funding to get launch histories. Eventually, perhaps after his tenure, it will become clear that the startup U.S. launch companies are safe enough and obviously cheaper enough to mothball those horribly expensive launchers. This comparison will have to be painfully real, immediately obvious, and undeniable. Only then can the NASA prime contractors be fired and the money formerly used for their support used instead for space exploration.