March 11 and 15, 2003
Manned Drop Test
We had two crush cones prepped for testing today. They are essentially the same as the first
one tested, with the addition of two 2 diameter pressure vents. The cabin bulkhead has also been reinforced
with welded retainers and a much better bond, in case the pressure rise is
Both tests were with the nose 10 off the ground, and a
total drop weight of 2400 pounds.
The first drop landed straight and square, but our secondary
restraining line came off, so it still tipped over. We were a little concerned that it might crush all the way,
because the last test had a significant air-spring effect with the un-vented
cone eventually popping the main bulkhead off.
However, this time it only crushed a little bit farther, and without the
overpressure induced buckling, it actually ended up taller than the last
one. The accelerometer traces showed a
longer deceleration, without the increase and spike from the air spring effect. The accelerometer data is still vibrating
back and forth hugely, but if I filter it way down, it shows a steady rise to
4G of deceleration, which is about the behavior we expect crushing the wider
part of the cone requires more energy.
For the next test, Russ got in the cabin and strapped
down. The pilot accommodations include
a five point racing harness with strap pads, a 2 thick hard foam plate that
covers all the seat belt and reinforcement brackets on the bulkhead, a 4 thick
memory foam pad under his back, 4 thick memory foam pads on his sides, and a
couple other foam pads positioned wherever he wanted. He wore a helmet and neck brace.
The drop went perfect 0.8 seconds of zero G, then about 0.25
seconds of deceleration. We finally got
the catch rope right, so it didnt even tip over. Russ reports that it was absolutely no problem at all, it felt
like you were just standing on a big aluminum can that someone poked in the
As an experiment, we had left one of the vent holes
completely open, and one covered. The
covered one did still blow out, which confirmed our calculations that a single
2 hole would still generate several PSI of pressure. We intend to go to permanent louvers in the cabin flange section
in the future, so additional fabrication work doesnt need to be done on the
In practice, under parachute the vehicle will be coming down
at some angle, and with some horizontal velocity, so it is highly unlikely to
work out this nicely for real recoveries, but we have tons of margin in both
pilot comfort level, and maximum energy absorption, so it will probably be
ok. We will learn from the parachute
flights of the 2 diameter vehicle, and then a helicopter drop test of a full
(warning 16 megs)
Failed Hover Test
We got the 2 vehicle out under the lift for its first
attempt at hovering. We dont have
landing gear on this vehicle, so it needs to be supported from above if it isnt
going to come all the way over and land nose-first under parachute.
The engines are all brand new, only having been fired once
each. We spent a long time trying to
carefully warm them up, but because we were at 73% humidity, they never did
clear up. I slowly brought the engines
up to warm them all together while the vehicle was still on blocks, and it unexpectedly
hopped off the ground a little bit. I
killed the throttle, so it came back down and banged one of the legs into a
cinder block, but nothing seemed to be hurt.
However, the Crossbow IMU had stopped updating. The computer and telemetry were still fine,
but the Crossbow wouldnt reset until the entire system was power cycled. This was worrisome.
This showed up an issue with the flight control software
that I then fixed if the crossbow was dead, my flight control software would
automatically exit, my original thinking being that you cant possibly fly it
without the IMU, so you shouldnt be allowed to try. However, that overlooks the valid use of the manual flight
control mode for venting the tank and engines.
We have a manual switchbox that also serves that purpose, but it is much
more convenient to use the computer when it is functioning.
We loaded more peroxide, and, after verifying that the
engines were still hot and producing thrust, we pulled the blocks out from
underneath the legs to let the vehicle just dangle, then throttled up for a
liftoff. Very shortly after liftoff,
the crossbow died again, leaving the vehicle to careen off course, then the
telemetry ceased another second later. The
vehicle swung like a big pendulum under the lift, but didnt suffer any damage
except for a dent in the cone. This was
close to the behavior at the Oklahoma crash, but is probably unrelated, because
the main voltage was not collapsing.
This did show an error in the new abort logic that hadnt
been tested yet the engines should have gone into immediate throttle down on
losing the crossbow, but they didnt. One
of the changes I made when we moved to the four servo engines is to drive the
fundamental tic rate directly from IMU update packets, rather than the system
timer. This removes an average of ½ the
update period in sensor latency, but it means you have to fall back to a
different way of cycling when the IMU is dead.
The code was in there for that, it just didnt work right. All it took was moving one line, and it seems
to do the right thing now.
The crossbow didnt come back after power cycling, and now
seems to be really dead. It was
probably fatally damaged in the Oklahoma crash, and only happened to be barely
working in static situations since then.
I have had a new model ordered for three months, but Crossbow is having
delivery problems. I hope they take
that into account and put a priority on fixing this one when I ship it back to
We have a few other anomalous things that we dont
understand: The feedback calibrations on
the valve pots has shifted. The valves
got fairly hot while we were slowly pushing all the peroxide through the
engines at ground level, but they still work fine, so it seems odd that the
resistors would have been damaged. The
main parachute chain releases were fired at some time during the flight. The vehicle did tip past horizontal, but
that should only be a trigger in parachute mode, not hover mode, which it
was definitely in. A crashing computer
can often trigger random actuators, which is probably what happened, but I wish
we knew for sure.
On looking inside, we found that the wireless transmitters
power cable wasnt very secure, so it is conceivable that the computer didnt
actually die, and we just lost telemetry, but that doesnt fit with the fired
parachute releases. I want to find a
small VGA screen we can put on the vehicle to let us sort those cases out, but
the smallest analog VGA LCD screen I have found is 10 so far.
There are at least two intermediate connections on the
electronics board that can be removed the Esteem power connector can be
replaced with a crimped connection, and the Crossbow cable goes to a DB9 /
ribbon connector on the PC104, which can be replaced by making a custom cable
that goes directly from the DB15 on the Crossbow to the power terminals and the
We have been planning to replace the current power supply
board, because it doesnt have proper ring terminal connectors on it, and we
suspect it of some possible failures.
We are going to do this before flying the vehicle again. We will also be using a regulated 12v power
supply for the Esteem and pressure transducers, instead of letting them run
directly from the battery. They should
be fine down to 10v, but we want to completely remove the uncertainty. The Crossbow is rated for 15v to 30v
operation, and we currently have a 15v power supply for it, but we have been
cautioned by a reader that running at the minimum input may not be the best
thing for reliability, so we are looking into possibly increasing this. The next-gen Crossbow FOG IMU takes 8v-30v,
which will be nice, but they have slipped several months in delivering them so
This is discouraging, because we had made several
reliability improvements on the current electronics board over our previous
ones. If the Crossbow can be repaired
quickly, we are going to change a few more things and try again.