John Carmack's announcement of a modular rocket that can reach suborbital space for $25,000 per module is revolutionary. Each module can independently reach suborbital space. Group the modules together and any size or shaped payload can reach suborbital space. The cost to get to space is $250 per module in fuel costs.

In a video that John said will be posted to his web site, he showed the modules being hooked together in a square arrays. These arrays can then be stacked for staging.

He predicts that he will produce the Armadillo orbital "Sputnik" which John also referred to as Mitchell Burnside-Clapp's DYANN--Do You All Notice Now?

There are two revolutions here. The first is an open source garage revolution. With a small warehouse and a budget closer to Charlie Farmer's in Farmer Astronaut than COTS winners RpK and SpaceX, Armadillo in a humble, matter-of-fact tone is brashly announcing an orbital program.

The second is the price of the revolution. At $25,000 per module, the capital cost per delta V is unprecedented and substantially lower than RpK or SpaceX.

This revolution was incrementally developed in plain sight and demonstrated in plain sight. No one thought Carmack's Pixel and Texel were minimum concept proofs for a 64-module version. No one thought that by looking at the specifications they were seeing the ultimate cheap first stage and second stage and third stage.

Carmack thinks he can get the mass ratio down from 27 to 15 with some low cost evolutionary modifications. At 15-1, he can loft "Pixel 2" onto a suborbital trajectory with a 64-module first-stage lifter made up of 16 Pixels arrayed in a 4-4 grid or 8x8 single modules. Pixel 2 will be full of fuel and be the second stage. On top of Pixel will be a single module with a 25 lb. payload that will make it all the way to orbit. The cost for this delivery? The capital costs would be about $1.7 million if he can stay under $25,000 per module. If only the first stage is reusable, the cost per flight would be $150,000. If the first and second stage are reusable, the cost per flight would be $60,000. For a three stage system, that is a not very revolutionary price of $2400 per pound to orbit (albeit revolutionary vs. old space of $10,000+ per pound though.)

If they achieve a two-stage to orbit system where the second stage is also reusable, that would deliver a 100 lb payload to orbit for $35,000. That is roughly half fuel and oxidizer and half capital assuming a 100 flight lifetime. $350/lb is revolutionary. If this could be scaled up to Spacex Falcon IX payload size of 22,770 lbs., that's $8 million or $22 million for a Falcon IX heavy sized payload of 62,500 lbs. An array of 100x100 modules supporting a second stage array of 25x25 modules boggles the mind and would cost $265 million in capital costs at $25,000 each. The flight rate assumptions would not be invalidated, however, because the vehicle could be broken up to support the suborbital tourism industry and smaller orbital payloads.

On the optimistic side, this price is before mass production. This mass ratio is before switching to methane (a 10% improvement in ISP over alcohol and a 50+% fuel price drop too). Google revolutionized servers by using modular white box CPUs. Now Carmack is making a bid to do the same thing. Nevertheless, Henry Vanderbilt cautions me that there is a long way to go from a view graph to orbit.

---------Update 3/24/07 7:00 MST---------

A wide plane requires a bunch of successively stronger connectors moving inward and results in very little additional payload delivered by the outside modules. This is especially true with a square grid which require more connections moving in from the corners than a hexagonal one. Other possibilities are more stages so connections are shorter and more vertical and larger, taller modules for lower stages.

Jerry Pournelle had an interesting take on this sort of thing on his Web page in response to the Space-X near failure or near success however you call it.

Atlas blew up a lot of rocket in flight. The idea was that they set up a production line for Atlas rockets so they were cheap enough, and static testing with the right kind of wind tunnels and test cells and vacuum chambers big enough to fire rocket engine were really expensive. So they flew and blew until they got it right.

A lot of the public sense of how dangerous what the Mercury 7 astronauts were going to face was based on all of these flight failures. No one differentiated between rockets destructed by range safety (where you had a few seconds warning before hitting the big red button) and ones which just blew. But the people designing Mercury had a lot of test data to base the tractor escape rocket -- they had a lot of design experience with rocket explosions.

The Saturn program put a lot of money into test cells, stands, wind tunnels, vacuum chambers, and so on. Don't think they ever lost one in flight (one of the unmanned Saturn V's was kind of shakey, and Apollo 13 experienced some close calls on their booster before their infamous "Houston, we've got a problem." All told, all of this ground testing was expensive.

Then there were the Russians with their Moon-race superbooster -- they didn't have the test stands so they flew and blew, but they were strapped for money and time, and some of their explosions did ground damage that set them back.

It is fine to talk about with a mass-produced mass number of cheap micro stages how you will make it to orbit, but there are all kinds of interesting "gas dynamic" things that happen at different phases of rocket flight -- you get a kind of altitude-compensating nozzle effect where rocket exhaust plumes start attaching themselves to flight structures whether you want it or not, and there are all kinds of fluid-dynamic things (Pogo and what not), that it takes quite a bit of engineering to get a workable orbital rocket.

I guess they don't call it "rocket science" for nothing.

Also beware of the vibration. One working engine produces it's own thrust as well as it's self induced vibration. When you add the trust they add as a vector all going in the happy direction you want. However at higher frequencies... they just add as random noise. So if you add 100 boosters together you get 100X the trust but you also get ~+20dB the vibration. (if you don't know much about structural dynamics 20dB = a butt load) Oh and that is assuming you don't add any structural modes with all the stuff you strap these motors together with.
Is it possible? Maybe Yes. But your in for a world of engineering "fun".