August 7, 2006 notes
NVIDIA is Armadillo Aerospaces official sponsor for the
X-Prize Cup events this year.
I was very reticent to talk sponsorship with any of the
companies that I work with in my capacity at Id Software, because I didnt want
there to be any hint that there is a sense of obligation involved (either way) based
on our working relationships in software, but I casually mentioned the
possibility to NVIDIA a little while ago, and they jumped on it with a We
would love to be involved with that!
This works out well, because I wasnt sure if it would be
worth it to me to put random sponsor logos all over our program just to defray
costs. NVIDIA is a company that I have
worked closely with for a long time, and I have wholeheartedly recommended their
products on technical merit since the very early days (if we skip NV1 and
). Ill be happy to have our
vehicles flying NVIDIA colors this year.
There is a chance at this point that I may have written the
last personal check I need to for Armadillo.
This sponsorship will cover all of our expenses through the X-Prize Cup,
and I expect to come away from that with some prize money. If we run the table with first prize wins in
both categories, we will be doing great.
It still wouldnt cover all the money I have put in over the last five
years, but it should be plenty to take us through the commercialization of the
current technologies. We have several
customers ready to pay for flights even below the man to 100km goal of the suborbital
space tourism market, so I dont think we are too far off. If we dont pocket the big prize, we will still
carry on at our current pace, but if we get the extra money we will be stepping
things up by bringing more of the team on full time.
We have two experimental launch permit applications (the VDR
and the Quad) in process with AST, both of which have been deemed substantially
complete. While the legally mandated
time period for official response is actually beyond the date of the XPC, AST understands
the importance of the event, and is making all efforts to have things fully
approved in time. There are other teams
that have submitted initial applications, but we are the first organization to
ever reach this milestone in the new experimental permit regime. This particular submission is specifically
for the VTVL flights in front of the XPC crowds, and we are being run through
the wringer even more than we would be for flights all the way to 100km in desolate
areas. After the Cup event, we will be
submitting modified permit applications for the envelope expansion flights to
higher altitudes, then moving on to full launch license applications for
Neil Milburn has been working full time on this for a
while. His day job is as a physics
teacher, so we decided that it would be worth while to have him become
Armadillos first full time (although temporary) employee instead of teaching a
summer session this year, primarily to concentrate on the regulatory work. This doesnt qualify as fun by any stretch
of the imagination, but it is vital for our progress moving forward.
Flying the Quad
The quad has been up in the air four times this last month,
as we sort out the issues involved in the new vehicle.
I finally got around to making an o-ring sealed leak testing
plate that we can bolt to the bottom of the engine to allow us to leak check
all the post-valve connections on the engine chamber: main fuel, main lox,
igniter fuel, igniter lox, spark plug, chamber pressure port,
chamber-to-injector. Lox leaks aren't
particularly troublesome, but fuel leaks can cause fires, and hot gas leaks can
cause catastrophic burn through. The one
problem with this is that the existing chamber retaining flange flexes a little
bit as it is torqued down, causing the tie rods to splay out a bit, making it
difficult to fit the leak test sealing plate on afterwards. I need to make a thicker bottom flange to
avoid this in the future.
We intended to use 1.5" diameter stainless flex hose
for the main propellant lines, but it really wasn't flexible enough. One axis is fine, but if it needs to twist at
all, it is far too stiff. We went back
to 1" lines, which may be a bit of a performance limiter. The valves are 3/4".
We had originally mounted the valves on the vehicle, and ran
the flex lines directly to the engine, as we did with last year's X-Prize Cup
vehicle, but when we did ignition tests on the vehicle we got really messy
flamey shutdowns as the purge threw out a quart or so of ethanol. We made some 45 degree adapters that allowed
us to mount the valves back on the engine, which helped a lot.
When we set up for doing the first hover test, we became
aware that there were a lot of things on the vehicle that the tethers could
catch on. We had a couple bad instances
of that last year, so we quickly threw together a crude vehicle cover out of
spare parts around the shop. We still
have bungee cords on the tethers to keep them from drooping for the first
several feet, but once you are over five feet in the air, they will still go
We are using 4 long sections of pipe to support the vehicle
and blow out on liftoff, which is working fine.
We did the little hop tests with 10 gallons of fuel, and
rather more lox than necessary based on weighing the vehicle during
loading. If we go to depletion, we want
it to go to gox instead of throwing burning ethanol all over the place. We are using 200 psi pressurization, which is
about what we would have remaining at the end of a fully loaded flight starting
from 400 psi.
When we pumped all the fuel into the center fitting with our
small peristaltic pump, we found that it mostly went into one tank. When we previously tested it with high flow
loading of water it went pretty evenly, but the slow loading didn't work
well. We put quick-connect fittings on
the bottom of each fuel tank and loaded five gallons into each tank
The lox seems to flow in evenly with the moderate pressure
loading. We are currently using around
40 psi to push the lox in, but we will probably increase that when we are going
for full loads..
When we pressurized, one of the new fuel quick connect
fittings started leaking. We threw a
plugged socket on it as a temporary measure.
This reinforces one of my biggest reliability mottos: "beware
things with springs". Check valves,
quick connects, solenoids, relief valves, and regulators. We have seen them all fail at one time or
Ignition and liftoff was fine, but the vehicle started
tilting. The gimbal went to maximum
displacement trying to catch it, but it still slid to the limit of the tether,
which caused a tilt shutdown.
The gimbal movement as a result of flight control gains was
capped at +/-20% for no particularly good reason, so I increased that to
+/-40%. The linear actuators don't have
internal limit switches, so I do want to keep them at least a little bit away
from their stops.
When we drained the remaining fuel out, it was all in one tanks. It wasn't clear how much was before/during
flight, and how much was from swinging on the tether.
Hovering is only 70 psi chamber pressure, but with a mass
ratio of 4, that is about where we expected it when propellant is almost
depleted. Engine efficiency is very
poor. This particular engine isn't very
good even at higher pressure, and at this low pressure it pretty much
sucks. It could still make the 90 second
flight, but we know we need to do better for the 180 second flight.
While increasing the gimbal range would probably let it fly
fine as is, we wanted to try to get it to fly with the gimbal in the center of
the range, so we have the maximum amount of margin both ways. We weren't sure how much of the angle problem
was due to propellant transfer moving the CG versus the initial gimbal angle
not being set right, so we took steps to improve both.
We permanently mounted bubble levels on the vehicle for use
during loading, and also as a reference for re-centering the gimbals. There was a fair amount of slop in the gimbal
assembly when we centered everything before, so we could have been off in our
We did the following to tighten up the gimbal actuator
Replaced actuator pin/block u-jopints with needle bearing
u-joints (ball/socket u-joints had worse slop than the pin/block ones).
Made the actuator to u-joint bolt a press fit.
Internally shimmed the linear actuators to remove slop.
The remaining slop in the engine assembly is now in just the
heavy duty main thrust u-joint, which we aren't going to try and do anything
Quad Hop 2
Lifted off straighter, but went into a roll, triggering a
roll abort and a soft auto-land.
The roll thruster had come on full blast, but it didn't seem
to be able to stop it. The vehicle still
flew at a tilt, so we assumed that the CG was still shifting, and the gimbal
motion was generating roll torques.
We decided to put a valve between the tank ullages so they
could be completely separated after loading, which would pull propellant back
into the original tank if it temporarily transferred out. It would also be somewhat self-correcting for
minor consumption differences between the tanks, because if one tank drains
faster, its ullage pressure will decrease, and the opposite tank will drain
faster. We moved the roll thrusters
around a bit so that each direction pulled equally from a single pair of
tanks. If we had a continuous roll
torque, it would mean that our mixture ratio would change, and we wouldn't get
even depletion, but that wasn't high on my list of worries.
Because the vehicle hopped up a little faster than I would
have liked for testing under the lift, I changed the hover velocity profile to
go slower for the first couple meters.
I was also making some software changes based on simulating
the full Lunar Lander Challenge flight profiles. Several GUI changes were made, and the
dead-man switch was changed to an explicit shutdown button. Holding a trigger is fine for 15 second
flights, but not something you want to do for 180 seconds. During one of the simulator test flights, my
finger slipped off the trigger, causing the vehicle to shutdown and drop from
50 meters. If that was a real flight, I
would have felt like jumping off a bridge.
Loss of telemetry is still an automatic abort, so switching to the
explicit shutdown doesn't give us any new failure modes.
Quad Hop 3
We found that when pressurizing, the extra pressure drop of
the isolation valve (while open) on the ullage connecting pipe caused all of
the propellant to get pushed over to that side.
It only takes 2 psi of pressure differential to force liquid up the 4
siphon tube and back over to the other tank.
We were able to get it even again by closing the isolation valve and venting
a little bit of pressure out of the opposite side. Once the fuel pipe was full, it would gravity
level. This is still clearly not a good
Liftoff was gentler this time, but we still had roll abort /
auto-land. We noticed that a recent code
change had made the gimbal not auto-center until engine ignition, but that didnt
have any impact on the flight. I fixed
I was beginning to think that there might be a real flaw in
our configuration, but we decided to boost our roll torque to see if we could
get a short level flight, even if it had to continuously fight a roll torque.
We moved the roll thrusters to posts at the tank girth weld
instead of the top of the tank, for a 50% torque increase. We drilled out the roll thrusters from
3/16" to 1/4" for some additional torque (the solenoids are only
1/4", so it won't be a full 16/9 increase).
As this work was going on, I was going back over all the
data, and I realized that the first flight, which didn't roll abort, didn't
fire the roll thrusters at all. The
continuously increasing roll rate on the other hops implied a nice continuous
torque, which didn't seem to be related to the gimbal angles, which moved
It looked rather like the roll thrusters were backwards.
They were hooked up as they were documented to be, and as
the manual test wanted them, but it turned out that the manual test (hat left /
hat right on the joystick) was backwards.
This was our problem. The roll
thruster changes were probably still a good thing to do.
As I was working through all this, I realized two positive
things about our current configuration.
A conventional gimbal below the CG has a couple problems that aren't
immediately obvious: There is a control inversion when a leg hits the ground,
causing gimbal angles to do the opposite of what they do when the vehicle is in
the air. When you want to move to the
side, you need to point the gimbal in the opposite direction to rotate the
vehicle to point the direction you want to go, causing a temporary movement in
the opposite direction. Both of these
issues are gone with the gimbal point above the CG, which should make this
vehicle a much better VTVL than our previous configurations.
Quad Hop 4
Increased propellant load to 20 gallons of fuel, 1000 pounds
Steady flight, but increasing oscillations eventually caused
a tilt shutdown. Roll thrusters fired in
both directions, behaving properly.
This type of divergence is usually a simple matter to fix
with a gain change, and I was able to replicate the behavior in the simulator
by tripling the polar moment of the simulated vehicle, which makes sense
comparing the quad to last year's X-Prize Cup vehicle. Correcting the behavior in the simulator was
just a matter of doubling the angular position gain (and leaving the rate gain
The vehicle was still flying at a tilt, even though it was
holding position. It turns out that in
the process of bumping the propellant back across the siphon tubes to equalize
them, we had left one of the equalization valves open for the flight. The funny thing is that as we were planning
to break the ullages apart, I had commented that an isolation valve could be
left in the wrong position, while requiring completely separate connectors on
the ground support equipment would make that impossible.
I was feeling great about it at that point, but Matt noticed
that we had burned through the injector face on the engine.
(we probably should have removed the lifting / tether bars for the shots)
(camera 2 crapped out on this one)
We took off the ullage connecting pipes completely, moving
the pressure transducers and purge valves off to the side tanks. We made an exactly equal T split on the ground
support pressurization hose, which does indeed seem to allow us to pressurize
without pushing propellant from one tank to the other. We seem to have this resolved.
The drip guards over the shocks were not effective, they
still rusted up from the lox frost.
Water may be wicking water up from the ground when they sit in puddles. We are going to try using some synthetic grease
on the shocks.
The engine injector that had the burn-through was a
like-impinging, broad-fan-impinging design.
It appeared that some fuel was getting past the lox fans and burning in
the central recirculation zone.
We tested a new engine on Saturday that was a similar design
with 38 doublets each in fuel and lox (already machined before we had the burn
through), but with a machining change to allow the fan impingement points to be
much closer together, and the holes were all reamed, 0.0650 for fuel, 0.0715 for lox. The film cooling holes were changed from 10 x
1/16 holes to 19 x 1/32, so the total film cooling mass flow was about halved,
but more evenly spread. Another modest
change we made was going to a square igniter / engine mount post instead of a
round one, which has some fabrication and repeatability benefits.
The new injector (the 10 degree film cooling relief went a
little too deep, and left a frizzy edge):
Matt finally found a location for the camera that could get
a broad elevated view of our tests without getting doused in water:
The reamed holes seem to have gotten rid of the occasional
stray spray jet we would get on previous engines, but fast exposure photos show
that the spray pattern is still a bit gloppy.
I believe that our atomization is hurt by the shallow impingement angles
(45 degrees) that we have, but that is about all the angle I can get while
still directing the fuel and oxidizer fans towards each other in this
arrangement. We may try a different
The engine had better chamber pressure for a given feed
pressure than the one that burned through, implying better combustion, but it
ran rather rougher. The film cooling
still seemed more than sufficient. We
are going to go ahead and hop the vehicle on this engine to test out the gain and
loading modifications, but it wont be the final engine, and it may well burn
through just like the last one did.
My theory is that the greater roughness is due to the
machining that brought the impinging exits closer together shortening the L:D
of one leg of the impinging injectors to only 2:1. I will probably be making another test with a
changed internal contour that allows us to make all the holes at least a 4:1
I also started machining an injector out of 303 stainless
steel. Depending on the conditions, this
may either prevent the burn through, or burn through faster. I havent decided which hole pattern to try
with it yet. I was pleased to find that
I could run the same CNC program I use for aluminum, and just cut the speed and
feed down to 20% to machine stainless.
Hopefully the injector holes wont be any more troublesome. We may wind up going with copper for the
injector if we continue to have heat problems, but we prefer to weld stainless
over brazing copper, and copper is a good deal more expensive.
I also want to machine some manifolds for the OTRAG injector
that Lutz Kaiser left with us to at least do some visual comparisons of spray