May 30, 2004 notes
We changed our A/D from the AccessIO board to a
Diamond-MM-32-AT PC104 A/D board.
Hopefully this one wont give us as many problems. We also went back to a dedicated battery for
the actuators, instead of sharing it with the main power supply. I couldnt fit another big battery on the
electronics board, so I had to mount it remotely.
On Tuesday we did eight hover tests trying to reduce the
back and forth swaying in position hold a bit.
The variables that go into the control equation are angular position,
angular rate, angular acceleration (not currently used), inertial velocity, and
inertial position. After trying
variations on all the parameters, it turns out my initial guess worked
best. Next day analysis showed that
while the initial ratios of parameters were probably pretty close to right, the
total scaling of all the parameters could be increased quite a bit before
hitting the maximum slew rate of the actuators. I added a new parameter that globally scaled all the tipping
parameters, and set it to 1.25 to boost everything up a bit.
The last tests we did on Tuesday were very short boosted hop
tests on a long tether. The first test
was just a 1.0 second boost, which only gave it time to just come up to full
throttle before transitioning to sustain and land mode. This showed up a bug in the code where the
position-hold gains werent active in boost mode, only in hover mode. I fixed that, and we did another test with a
1.3 second boost, which did a nice little boost / sustain / land in place. Unfortunately, when I went to burn off the
remainder of the propellant, I accidentally tapped the boost button on the
joystick again, and the vehicle hopped into the air. Since it didnt have enough propellant to complete the cycle, I
immediately released the control button, but it still dropped from about five
feet up, bending the frame a little.
This pointed out two more changes to make in the code: I now require the boost button to be held
down through the entire boost period, so if you release it early for any
reason, it immediately goes into sustain / land. I also force you to actually exit and restart the flight control
program to do another boost launch.
On Saturday, we went out to our 100 acre test facility to do
boosted hops. We had a break in at the
site recently, with someone taking an angle grinder to our gate and shed. The police said that there are a lot of
problems like that in rural areas because of people looking for anhydrous ammonia
fertilizer for drug labs. Hydrogen
peroxide evidently wasnt useful to them.
The wind was blowing hard, with gusts well over 20 mph, so
we did an initial hover test to see how things went at low altitude. The vehicle lifted off and just hung in
place perfectly, even while the wind was blowing everything else around. The 25% boost in the gains was a big
We loaded up another carboy of propellant and set the flight
profile for a 2.0 second boost. The
boost was perfect with 1.5G acceleration (0.5G above hover), but when it
throttled down to sustain mode, the vehicle started rolling because the reduced
throttle didnt give it enough control authority to counteract the high
winds. After it had rolled past 20
degrees or so, the position hold feedback was now trying to point the vehicle
in a direction that wouldnt help the position, causing it to flip over about
20 feet from the ground.
This is understandable behavior, because the lander is very aerodynamically
asymmetric in many ways. Under high
winds, it wanted to roll-weathercock.
At hover or boost throttle, the vanes had enough control authority to
fight it, but when the throttle was reduced for stabilize, the actuators went
to max displacement (18 degrees minus whatever is needed for tipping control)
and it still couldnt hold roll.
If the software had been doing a full transform of the required
position hold vector into vehicle space instead of just tying north/south to
one axis and east/west to another axis, then the vehicle would have been able
to go ahead and land even with it rolling significantly. A simpler alternative would have been to
just cancel the position hold modifiers if roll isnt well under control, which
would have still soft-landed the vehicle, but with notable horizontal velocity,
so it would have tipped over. In any
case, we really want the vehicle to not roll.
Another factor that was unexpected was that there were large
surges in the acceleration after boost while it was in the stabilization mode. After boost, the vehicle tries to maintain a
0.3 g acceleration or 0.20 throttle level until it is time to throttle up for
landing, whichever is greater. The 0.3
g level is the greater on this vehicle, so it was in acceleration-hunt mode
after the boost. Our hovers have always
had a pretty good up and down bounce to them because of the latency from moving
the valve to actual engine thrust changing, but at hover thrust levels we have
always been around the midpoint of the ball valve travel, while at 0.3 g thrust
levels we are near the opening point of the ball valve where the flow changes
are very non-linear, giving surges of over twice the amplitude. That probably wouldnt have affected the
vehicles ability to land, but it could have caused it to come down harder than
desired, depending on the phase of the surge.
I am going to have to do some work to reduce the surges in hover, which
should also take care of them in stabilize mode. A combination of prediction and a slowing of the valve movement
on overshoot should take care of it.
Stripping the vehicle down showed that the frame was a
wreck, a couple of the wire rope isolators were destroyed, and the electronics
board had ripped its foam isolation ring apart, but everything else seems to be
ok. We are mindful of the fact that the
last time we crashed a vehicle from altitude (much higher than this), the
electronics seemed fine on the bench, but gave us serious problems in
vehicles. The combination of the much
lower drop and the fact that the electronics board absorbed a lot of energy
tearing the foam ring off gives us much hope that everything really is ok this
time. If not, I have backup gear on
hand, although I would hate to go back to the older Crossbow FOG with the worse
drift rate and the +15V supply voltage.
We decided to convert the old 2 diameter tubular vehicle
over to jet vanes for the next test vehicle, which will give us a streamlined,
axisymetric profile. Rockets really
ought to be axisymetric. We pulled out
the four small throat 90% peroxide engines and cut everything off the aluminum
bulkhead. We took one of the 7
diameter engines from a differentially throttled quadrant of the old big
vehicle (the engine on the lander needed to run at over 500 psi for boost
thrust, which the tube tank cant handle, so we needed a bigger engine), and
made mounting brackets for the jet vane board from the crashed lander. Since the position hold seems to be working
very well, we are mounting linear shock absorbers instead of wire rope
isolators for landing gear, which may give it a chance of actually staying
upright after landing, even with the very narrow base. We are going to add ground contact sensors,
since the soft landings and shock absorbers have been too soft to reliably
trigger my acceleration based.
The current electronics arrangement cant be connected up in
the tube, so I am taking this as an opportunity to rewire most of the
board. I am going to mount the batteries
remotely, which frees up enough space to mount the A/D breakout board and the
GPS flat instead of vertically, and I can orient the various pieces to require
less wiring now. The connector for the
three batteries will serve as the main power switch, and the charger will be
modified to accept the same plug.
Pieces of four previous vehicles are going into this new
one, and it looks like we are going to have it back in the air for hover tests
next Saturday, which will work well, because it is going to rain a lot this
week, making the remote test site unusable.
We should be able to try another boosted hop the week after, weather
permitting. The roll-from-wind problem
should be gone by the nature of the vehicle shape, but I am also going to up
the stabilization throttle minimum to 0.5 G, just in case tipping from higher
speeds becomes an issue. This will
waste propellant on the way down, but the new vehicle has a 60 gallon tank, so
we can load up plenty of extra. If we
can get a burn time waiver, it should be capable of flights over 60 seconds.
Dropping this vehicle doesnt really set us back at all,
because going to a streamlined subscale vehicle will let us test a lot of
things more representative of the big vehicle before we put it at risk, and it
looks like we are only going to have one complete shop day lost to fabrication. We have also finished the rework on the big
vehicle with the 1/8 / ½ vanes and shafts replaced with 3/16 / ¾ vanes and
shafts, and extra containment panels bolted behind the shaft slots in the cover
to keep anything from blowing past. We
also covered the foam blocks with aluminum insulation protection, which should
prevent the exhaust from ablating them so badly. I am going to change a couple things in the wiring to sync up the
big and little vehicles, but we may also hover test the big one next Saturday. We will probably be ready for boosted hops
of the big vehicle as soon as we prove everything out on the small one.