(Armadillo Aerospace is a small research and development
team working on computer-controlled hydrogen peroxide rocket
vehicles, with an eye towards X-Prize class vehicle development in
the coming years. The team currently consists leader John Carmack,
a bunch of guys, a girl, and an armadillo named Widget. This is
John's story.)
Armadillo Aerospace Log: 08/08/04
On Tuesday we did a very successful set of hover tests with the
big vehicle. I had two changes that I wanted to test: an optional
PWM of the throttle movement to make it change position slower when
it was in hunt-for-an-acceleration mode, and testing a 50% gain
increase which I might enable during high speed flights if it looks
like it is having a hard time controlling the attitude. I had these
set up momentary overrides on the joystick, so I could lift the
vehicle up, engage the change, let go real fast if it isn’t
working, then try the other one, all on a single propellant
load.
When we tipped the vehicle up, several catalyst rings fell out
of the engine nozzle. We looked up the engine with a boroscope and
found that the screen at the bottom had pulled past one section of
the support plate, allowing some rings to escape. This had also
happened on the previous 12” engine after a few runs (you
could see a couple red hot catalyst rings fly out in one of the
static test videos). It didn’t seem to be progressive last
time, so we went ahead and left it alone, expecting the test run to
squash the rings down into an interference fit again.
Because this was set up to be a 25 second hover (tethered),
which would be our longest hover test, we decided to make this a
no-direct-view test, with my flying it from behind a concrete wall
looking at a monitor instead of directly viewing it. The engine
warmed up fine and lifted off and hovered fine. I was about to
engage the first test when the vehicle just set itself back down on
the ground. It took me a few moments to figure out what happened
– I had moved the computer and wireless antenna behind the
wall with me, so the telemetry link was very ratty, dropping quite
a few packets. Eventually it dropped enough in a row to hit the
internal limit and triggered a loss-of-telemetry abort, which is an
auto land. Perfect!
I moved the antenna back in view of the vehicle, and we
completed both of the control system tests without incident. We
used our new propellant disposal burner to catalyze the remaining
propellant, which worked pretty well. The foam coming out was
probably still 10% peroxide or so, but a little water was fine for
washing it away. We might consider adding a spark ignition system
to it so it would completely burn everything away, but that would
be a more complex system, and would leave us with a red hot
propellant burner.
When we set the vehicle back down on the cradle, a few more
catalyst rings came out, but the engine still seemed to be working
perfectly.
Based on these results, I changed the flight control code to use
the PWM valve movement when it is hunting back and forth past a
desired acceleration. If it hasn’t crossed it in 500 msec, or
the desired valve position is fully open or closed, it goes back to
full speed.
We also weighed the vehicle, and surprisingly found it lighter
than we had estimated, right at 1000 pounds.
Complete Loss of Vehicle
Saturday was a perfect day for flying, so we went out to the 100
acres for a boosted hop. We had high expectations for success,
since the vehicle had been operating perfectly on all tests so
far.
After we loaded up the propellant and pressurized the vehicle,
we ran into a problem. When I opened it up to 20% throttle for the
warmup it looked like it cleared up fine, but the telemetry was
only reading 100C, as if the hot pack hadn’t started heating.
We were a long way from the vehicle, so we couldn’t really
tell what was going on. I gave it a bunch of slugs of propellant
until it finally started going up in temperature properly, but we
had blown a lot of propellant out on the ground. Too much.
It finally reached operating temperature and we launched. We had
only been operating this engine at hover thrust levels, so we had
been a little concerned that it might be rough at full throttle. It
was. It flew fine through the roughness, but when it started to
throttle down after the two second boost to a 0.5 G positive
acceleration level for the stabilization phase, the rough pulses
kept passing both above and below the desired acceleration, keeping
the engine from throttling down at full speed, resulting in it
going a lot higher than intended (just under 600 feet high). It did
finally get out of the rough stability zone into clear
stabilization, but a couple seconds later, everything got quiet. It
ran out of propellant.
It had not hit apogee yet, so the unstable vehicle immediately
started rotating, hitting about 50 degrees / second. If the vehicle
had been past apogee when it ran out, it probably would have just
dropped feet first.
We had telemetry all the way to the time of impact, which
matched the video perfectly, landing eight meters from the launch
point. The vehicle hit the ground basically sideways, a little tail
first. The bottom manway flange broke off the tank, and the 450
pound tank with 180 psi pressure still in it got punted about 200
yards away by the gas release. $35,000 of rocket is now a whole lot
of primo Armadillo Droppings. There are a few pipe fittings that
survived, but that’s about it. Amazingly, even though the
on-board camera was destroyed, the tape did survive with only some
scuffed sections. It’s a good thing Doom 3 is selling very
well…
From analyzing the telemetry (integrating the chamber pressure
during the flight), it looks like it wasted two thirds of the
propellant on the warmup. If it had lifted off with a normal
warmup, it would have landed ok even with the rough throttling, but
we would have been in violation of the 15 second burn time limit by
the time it landed. There was twice as much propellant loaded as
this flight should have required, which I thought was enough to
cover any off-nominal conditions, but we obviously should have
scrubbed when the warmup didn’t catch after the second or
third try. We are going to look into getting a continuous
capacitive level sensor next time so we can have a firm no-go line
for liftoff. If anyone knows of a peroxide compatible (316 SS /
Teflon / viton / eetc) capacitive sensor that runs off of 12v or 5v
DC and can handle 300 psi (we may be willing to run past rated
pressure if nexessary), let me know. Ideally we would want a 5V or
10V analog out, but we could live with a current sensor, or (with
some begrudging) a serial port. We would like to mount it on the
bottom of the tank instead of the conventional top location, but we
don’t think that will be a problem.
The failure did give us some demonstration data that we always
sort of wanted to get (but not that bad). The vehicle is
absolutely, positively, NOT going to continue flying nose first
when it loses active control. This should be blatantly obvious from
the CG, but we had a WSMR engineer pushing us towards a NASA
consultant to prove it. When it fails in the air, it just drops
like a rock, landing very near the launch site. Rupturing a
fiberglass tank doesn’t produce shrapnel, but it does drop
kick the tank pretty good. This looked pretty close to an optimal
45 degree launch angle for the tank, so we have a pretty good idea
what our safe distances should be.
We probably would have been able to save the vehicle if we had a
rocket drawn parachute on board, but we are trying to have a
pyro-free vehicle. A pneumatic drogue cannon might have been able
to deploy a chute fast enough, but it would be a lot more
debatable.
We cut the engine open with the plasma cutter to do a
post-mortem, and found what had been causing the engine issues. The
combination of the bottom catalyst retaining plate bowing down
because it was only welded on the bottom and some catalyst escaping
both out the bottom and some out the top (the top screen was burned
through in a couple places) left the bottom catalyst not even
completely covering the diameter of the engine. When we had the
nozzle and cold pack cut off and the engine on its side, you could
see right through the hot pack at the top.
This explains the apparently clear exhaust at the start while
the thermocouple was still reading only 100C, because the
thermocouple was fairly short (we used to use a longer one, but the
bowing of the retaining plate forced us to use a shorter one so we
could still insert it) so it was in a stream around the edges that
bypassed most or all of the hot pack catalyst (driving down the
highway probably also settled the catalyst on the opposite side
from the sensors), while much of the main flow was still being
burned. The loosening catalyst is also almost certainly why this
engine “got rough” after we had been using it for a
while.
The support plate bowing can be fixed by either making a full
depth angle on the sides of the plate so the weld gets full side
coverage, or actually weld the plate between two chamber sections,
instead of inside a single chamber section. We are making new
plates that are made with 1300 quarter inch holes instead of large
water jet cut squares that are bridged by screens. This will let us
completely avoid the screens altogether, and we are also going to
tie the top and bottom plates around the hot pack together by
putting quarter inch bolts through some of the quarter inch holes,
and welding them together as a unit with the catalyst in between.
This should fix the engine behavior.
Everything else operated perfectly, so we still feel good about
the general configuration, but we have a number of improvements for
robustness and operability that we will be making in the next
vehicle we put together. A couple of the necessary items are fairly
long lead times, so we are probably grounded for five weeks.