Which is mine.

The fact that the roll problem seems to get worse as the burn progresses could be a result of:

1. higher acceleration as the vehicle gets lighter
   
2. less atmosphere as altitude increases, with correspondingly less aerodynamic control
   
3. increasing thrust asymmetries as the nozzle erodes
   

I'm guessing that it's a combination of all of the above and that the vehicle doesn't have enough RCS control authority to muscle past the (unplanned) thrust asymmetries. The pilot is probably fighting to keep the nose pointed forward, and as Brett Buck suggested, yaw and pitch moments are getting coupled into roll.

If so, this is a problem that could be solved with a better engine nozzle design, thrust vector control on the main propulsion (a more expensive fix), more powerful RCS jets, or all of the above.

As I said previously, though, this shouldn't necessarily prevent them from winning the prize, as is.

Burt says it was caused by a "known deficiency":

The unplanned corkscrew maneuver Wednesday was characterized as a "spin-stabilized" roll. Rutan said there’s a "known deficiency" in SpaceShipOne that caused the roll.

Wind that hits an airplane from the side causes the craft to roll as a corrective technique, Rutan explained. The same thing applies to SpaceShipOne as it pierces the upper reaches of the atmosphere, although the rocketship "rolls much too much to correct for that," he said.

That still doesn't make much sense to me. Perhaps a better explanation will be forthcoming.

I believe that there is a well known aspect of rocket engines wherein exhaust gasses swirl and impart torque to the vehicle. Perhaps the RSC muscle as you speculate is inadequate to handle this torque.

I'm not aware of anything that would induce swirl in a pressure-fed hybrid, or that would get worse over time.

I know Burt wanted a simple design, but the next machine should be fly-by-wire. It could catch the roll just as it happens and would most likely provide a smoother flight than the current stick and rudder design. I honestly think that the 3G ascent is just too much for almost any pilot to control 100% manually.

I don't really have a grasp on how roll coupling works so I don't know if this would apply, but--

If the payload isn't distributed to keep the vehicle vertical center of gravity on the thrust axis, you'll get pitch torques steadily increasing in magnitude over the course of the burn. Same for yaw torques if you have a lateral CG error. Either way the fix may be to have some ballast weights on pulleys that you can use to adjust the CG location in flight.

I'm starting to see reports that an actuator failed and that might have been the cause of the roll. Was the roll in the same direction as the last one and did it start at the same time in the burn as it did in the initial launch?

I'm not a rocket guy, but just thinking about the raw physics of it, I have a hard time seeing how thrust asymmetries could produce a roll torque. Pitch and yaw, certainly -- this is what I thought had been the problem with the previous SS1 flight, where the horizon was pitching all over the place, but Mellvil now says that it was actually pilot-induced-oscillation. The fact that we didn't see any of that pitching and yawing on this flight tends to support that, and in my mind rules out the nozzel erosion / thrust asymmetry hypothesis. (Of course I'm sure it didn't hurt that the vehicle was spin-stabilized on this flight...it may actually turn out that this "undocumented feature" is a desirable attribute!)

My theory goes something like this: SS1 rapidly loses control authority as it ascends. Thus a pilot may be able to accidentially initiate a roll which they are unable to cancel out just a few seconds later. Momentum and oxidizer sloshing could carry on the roll for some time.

This doesn't really explain why the roll seemed to keep accelerating, though. Perhaps a stuck RCS thruster, ala Gemini 8?

I'm not a rocket guy, but just thinking about the raw physics of it, I have a hard time seeing how thrust asymmetries could produce a roll torque.

As stated, through coupling of the axes of rotation, in which the pilot was fighting the pitch and yaw.

Above Mach 1, the stick forces are so high that
the pilot is forced to use electric trim to
control pitch (which, after the pull-up, should
require only subtle inputs). Since this thing
has two trim motors, any differences in rigging,
motor output, freeplay, ect will give a differential elevator => roll.

> I honestly think that the 3G ascent is just too much for almost any pilot to control 100% manually.

The gee forces are not a problem. Mike Melville pulls five gees in an aerobatic plane on a regular basis. We flew Maching Bird One to six gees on three flights in one day, with the same pilot. Three gees is baby stuff.

Something I haven't seen: what about the changing aerodynamic environment? Is it plausible to think that as the Reynolds number blows up the damping of small weirdnesses in the aerodynamic flow gets too slow? It just seems like it ought to be an important factor that the aerodynamics is changing somewhat rapidly for this puppy at those altitudes and vertical speeds.

Pournelle (both the Dr. and his son at XCOR) and the XCOR guys think it's uneven erosion of the rubber component of the hybrid, combined with the nozzle erosion. That's the 'best guess' from the hanger next door.

And of course any mismatch in the electric trim units wouldn't help any with that!

Speculation is that that might not use hybrids in the commercial design Virgin just ordered 5 of (and I'm sure XCOR hope for a design win there).

Ack, hybrids.

Karl Gallagher's probably right (not surprisingly) - it looks to me like a classic case of inertia coupling.

Rand (and others)-

Have you seen these videos yet? They're the ones that were posted today on Scaled Composite's website, but they got Slashdotted, so this is an alternate mirror:

http://www.css-auth.com/ss1/

The second one has interior footage from the attempt; the spinning looks nauseating.