March 18 and 22, 2003 Meeting notes
Smallest Armadillo Fan
We have been selling Armadillo Aerospace T-shirts for the
last week or so with PayPal support - previously we hid the fact that we even
had t-shirts for sale due to the fact that were all busy doing stuff (Anna was
willing to take up that task so now we have a paraphernalia page: http://www.armadilloaerospace.com/n.x/Armadillo/Home/Paraphernalia).
The Ward family wanted to outfit their entire crew; including the baby.
Anna is a sucker for such requests (no
more, please!), and had a couple tiny shirts made up. Here are the pics: baby1 baby2
Cabin Pressurization Test
We had intended to delay the cabin pressurization test until
we had mounted the cabin on a full size tank, but I had been thinking about the
consequences of having the bulkhead rupture while the cabin was permanently
mounted on the tank we would never have been able to replace it, and it might
have forced us to scrap both of them.
We went ahead and sealed up the cabin on the boilerplate
tank end, and did some tests.
We used two-part foam to seal between the cabin cone and the
tank end, which worked extremely well, because it poured down, then foamed
up. When we flipped it over, we could
see it had foamed through in the couple spots that we knew were a bit of a
We doubled up the number of bolts we used to secure the
cabin to the tank end, because 20 psi over the entire surface area is 60,000
pounds of force.
We put RTV around all the bolts securing the seat belt
brackets and bulkhead hangers.
We finally got the hatch to fit perfectly by a combination
of hydraulic press work and banging the hell out of the frame with a big mallet
(high tech!). With a door seal gasket
all the way around it, the hatch squashes into place and seals perfectly with
only a fraction of a PSI of internal pressure.
We have gone back and forth with all kinds of schemes for hinges, dogs,
and latches, but I think we are just going to let it be pressure retained,
possibly with a loose fitting chain to hold it when the pressure completely
goes away. We have two removable
handles that screw into the outside of the hatch to allow it to be positioned
and held while pressurizing.
Our early pressurization did show two welds that had some porosity
in them, and we had leaking where we had to grind the flox fillet down to
install the backup bulkhead hangers.
Russ fixed the welds, and we added some more RTV on the outside of the
We put the cabin on the other side of a concrete wall, and
started bringing the pressure up. At 17
PSI, a bubble formed in the RTV by the bulkhead hanger, and popped, giving a
pretty strong leak. It probably would
have been fine as is, if we had let it completely cure, but we went ahead and
put a good bead of RTV around the entire inner flox fillet, because we believe
that it was leaking into the honeycomb from farther away, and only rupturing
out at the point on the back that we had to grind the filet away from the wall.
The structural elements dont seem to mind 17 PSI it at all,
which is all we expect to see during flight (pressurize to 2 PSIG on the
ground, retain full 16 PSIA throughout flight into vacuum). We are going to do some rapid pressure
cycling to 20 PSI next week, but it looks like the cabin sealing is already
nailed down, which turned out to be easier than we expected.
We still need to test the air tank and back pressure
regulator that will maintain a constant, controlled bleed rate for cabin
Another Failed Hover Test
We made a bunch of improvements on the small vehicle.
Russ finished an O-ringed aluminum piston for the drogue
ejection, which works very well. It
seals so well that we have to open the solenoid to allow the piston to be
Phil picked up our custom nomex parachute bag for the main
I removed the potentially-flaky power connector to the
Esteem, directly crimping the lines together.
The strange behavior with dropping TCP packets while
transmitting large UDP packets over the wireless network showed up again, so I
changed the wireless channel on both units, and it immediately went away. There must be another strong transmitter in
our area that is sometimes active.
Russ finished the new power supply board with real terminal
strips, and regulated +12V for all devices.
I also now log the regulated 5v and 12v values, in case that ever tells
The Crossbow IMU was sent off for repair, which looks like
it will cost about $1000 (the FOGs werent damaged, just the interface). Fortunately, our new Crossbow unit was just
about done, and it got shipped out overnight for us, so we were able to test
with that on Saturday. The new
generation (700 series) units have half the noise in the FOGs, and a few other
subtle changes. The zeroing procedure has
changed in a mildly annoying way (no explicit completion signal, just three
minutes of active filter time), the angular rate range dropped a bit, and the
exact timing of updates at different baud rates changed, but overall it was
straightforward to integrate.
One thing we have talked about fixing for over a year is to
replace the excessively long cable that runs from the Crossbow to the serial
port and power source with a proper length one. We decided to go one step farther and make the cable go directly
to the 10 pin ribbon connector on the PC104 module, so the DB9 to DB9
connection went away completely. I
finally got around to buying the proper AMP crimping tools and supplies to make
nice DB connectors (DB15 on the crossbow), so Im not making crappy solder connections
on Radio Shack DB connectors.
Professional crimping tools make all the difference in the world.
I am trying to get some bonded rubber isolators to mount the
entire electronics board with, but I have been playing phone tag with a sales
rep. I had avoided doing this in the
past, because I didnt want to hurt the bandwidth of the IMU, but I realize now
that there are a lot of vibrations over a couple hundred hz that would be nice
to get rid of, even if we dont touch the low frequency ones.
Several readers pointed me at the Xenarc 7 LCD VGA screen,
which I have placed an order for.
When we got everything hooked back up for testing, I noticed
that the recovery tank (drogue ejection) pressure transducer was reading a ridiculous
off scale value (-200 psi). Russ opened
up the bottom access panel and found that the transducer itself was warm to the
touch, indicating that it was likely internally shorted. I went back to the data logs from last week,
and found that the sensor did die about a quarter second before the computer
did. We may have had some kind of a
short that spiked the power bus, although the battery voltage did not appear to
drop in the telemetry.
The timeline for last weeks test was:
0.00s begin throttle
1.00s crossbow stops
updating, code doesnt properly recognize this, so valves that were opening
continue to open
pressure transducer dies
We just cut the pressure transducer away for now, but we
will replace it with a different brand soon.
This might be an argument for fuses, but I have heard people make the
case that they can add more unreliability than the problems they solve.
We pulled the vehicle out under the crane and loaded up 2.5
gallons of peroxide for the first test.
150 psi tank pressure is plenty for hovering the vehicle, even with the
small nozzles and partial throttle. The
first thing we ran into was that Russ had accidentally closed the master cutoff
valve with the manual switchbox (we have the hardware installed, but no
microcontroller running it yet), so none of the engines would fire. I have added a warning light for that
condition on the laptop console now, so we know about it before we load any
It was raining all day, so the engines were very cloudy, but
they were making enough thrust to lift off.
We really should have done more thorough test stand testing of each of
the four engines before integrating them in the vehicle, but we had hoped that we
would just be fortunate and have an all up test work out well in the vehicle.
Current questions about relative engine
thrust and decomposition efficiency probably wont be able to be answered until
we take it all apart and test each of them individually, but we dont have
enough peroxide on hand at the moment to do that.
The first liftoff pitched over about 10 degrees, but it was
catching it and correcting as I throttled it back down due to lack of
visibility. It looked like it probably
just needed some adjustments to the control gains. When we tried a second load, we experienced another failure. During throttle up, just as it registered
some positive acceleration, all the voltage signals on the A/D board
dropped. The battery and power supply
voltages only dropped a few tenths of a volt, but the valve potentiometer
feedbacks dropped by about 0.75 volts, which put them all the way off the
bottom of the scale. This caused the
computer to throttle them up, sending the vehicle into the crane again. The telemetry ceased about the time it hit
This time, it came down hard enough that the line stretch
and boom flex allowed one of the legs to impact the ground, causing some damage
that we will have to repair. In the
future, we will take it up at least another foot. When we restarted the computer, all the signals on the A/D board
were dead, but we could still read them with a voltmeter, so that board seems
to be toast. We will replace the board
next week, but we dont know yet if a bad board was a case or a result of the
problems. I am going to double check
the IO and IRQ usages of the boards in our PC104 stack (ampro core module 400,
Winsystems A/D16, Winsystems Com4, Winsystems IO48), because this is a very
different stack than we used on our previous systems.
The fastblock-800 thermal coating on the engines continues
to impress. After five seconds of
engine firing, you can go up and grab the engine with your bare hands. There was also some concern that the coating
might be sensitive to water, because the raw coating is water based, but after
curing, it doesnt matter if it gets wet.
The heat on the base plate from exhaust gas blowback is the main thermal
issue now. It would probably be a good
idea to coat the entire base plate, because tail section heat may be one of our
contributing factors to strange sensor readings.
Hanging on a rope does cause the roll angle to drift quite a
bit before liftoff, which skews the attitude control a bit. We had originally intended to pull blocks
out from underneath the vehicle at liftoff, but with the big clouds of smoke,
we cant do that accurately. On the
plus side, from the little one second airborne times we are collecting, the
angular rates are drastically smoother than we used to see with the solenoid
based attitude control systems. When we
finally get it all worked out, it should look a lot more stable than the old
binary control system.
Upcoming small vehicle work:
Repair the bottom bulkhead
Replace the A/D board
Figure out what the hell is causing our reliability problems
Figure out why the engines arent running very well
Cut down cone for longer crush cones
1,000 flight locally
Assemble TV transmitter equipment bulkhead
3,000 flight at Oklahoma Spaceport
10,000 flight somewhere else in Oklahoma
Upcoming big vehicle work:
Pressure cycle test
Test propellant venting times for aborts
Finalize three person cabin arrangements
Install cabin on 800 gallon tank
Fabricate tank end closure for parachute and engine mounts
Helicopter drop test for parachute / crush cone interaction
Test fire the 12 engine
Install four 6 engines and one 12 engine for test flights