July 5, 2006 notes
Scrubbed VDR Test
We re-signed a lease for space at the Oklahoma Spaceport to
do extended hover testing, since our vehicles have gotten too loud and
obnoxious for our neighbors at my 100 acre site. It is a four and a half hour drive from the
shop to Burns Flat, so a flight test operation just barely fits in a
Saturday. This is definitely
inconvenient, but we don’t have a better solution right now.
We packed everything up and headed out there to test the
VDR, but after we got all set up and went through our pre-flight checklist we
found that our MSD ignition box was dead.
We searched around in the area for a replacement, but finding an MSD
Small Engine Controller wasn’t very likely.
We considered trying to rig up some other spark igniter from random auto
parts, but we decided to just scrub the flight.
We don’t know exactly what caused the box to fail, but our
best guess is that cold soak from being mounted on metal that contacts the lox
tank could have damaged it. The entire
box is filled with potting compound, so I can’t imagine vibration having hurt
it, and other people have commented that welding on vehicles isn’t known to
harm MSD boxes. We are going to
thermally isolate them better in the future, but this is still a worrisome
failure case if you require in-flight engine restarts. Dual plug / ignition box systems may be
necessary if there aren’t completely redundant engine systems.
We had a spare box at the shop, and swapping it out did fix
everything (the coil was fine), but instead of just heading back out the next
week, we decided to concentrate on finishing the Quad before the next test,
since we don’t really need to fly the VDR until after the X-Prize Cup.
http://media.armadilloaerospace.com/2006_07_05/vdr1.jpg
http://media.armadilloaerospace.com/2006_07_05/vdr2.jpg
http://media.armadilloaerospace.com/2006_07_05/vdr3.jpg
http://media.armadilloaerospace.com/2006_07_05/vdr4.jpg
Quad complete
The quad is finished.
We just need to get an engine under it and degree in the gimbal actuators.
This is the first vehicle that came in below my initial
estimated weight – it was 569 pounds with everything on. We have added a few more tabs and things
since then, but it looks like we will stay under 600 pounds. With 400 gallons of tankage
We have done a few water loading tests to look at the
relative loading into the paired tanks, which doesn’t seem to be a
problem. We are still somewhat concerned
about variable depletion rates with roll thruster firings causing minute temporary
differences in ullage pressure, but we will do a lot
of tests to characterize it, and we have quite a bit of margin.
One unexpected thing we had to work around was that water
sprayed on the spheres tended to run down around them and drain directly into
the shock absorbers, which resulted in them corroding and sticking. We cleaned them up and made drip guard plates
that fit between the tank mounts and the phenolic
thermal spacers.
Tanks should be arriving this week for the second quad, but
we aren’t going to do any vehicle assembly until the first one has done flight
tests, in case we have to change something fundamental.
http://media.armadilloaerospace.com/2006_07_05/quad1.jpg
http://media.armadilloaerospace.com/2006_07_05/quad2.jpg
http://media.armadilloaerospace.com/2006_07_05/quad3.jpg
http://media.armadilloaerospace.com/2006_07_05/quad4.jpg
http://media.armadilloaerospace.com/2006_07_05/quad5.jpg
http://media.armadilloaerospace.com/2006_07_05/quad6.jpg
http://media.armadilloaerospace.com/2006_07_05/quad7.jpg
http://media.armadilloaerospace.com/2006_07_05/quad8.jpg
http://media.armadilloaerospace.com/2006_07_05/quad9.jpg
http://media.armadilloaerospace.com/2006_07_05/quad10.jpg
Welding
All of the welding process work we did didn’t seem to make a
bit of difference in the strength of the spheres. The last sphere we burst tested, a 5083 alloy
spinning, broke at 770 psi, just past the weld in the
heat affected zone. The one we tested
this month was a 5086 alloy tank (we had these spheres sitting around for a
long time, and we only settled on 5083 after that order) and it burst at 710 psi, also in the heat affected zone. The 9% strength difference is about what you
would expect from the alloy difference, but I had been hoping that the better
welding process would cause it to break up at the point of maximum thinning on
the spun hemisphere, rather than at the weld.
No such luck. We might have been
getting excessive penetration on the weld, because there was a good bead on the
inside of the tank, which would have been a stress riser. When we tested very thin (0.125”
pre-spinning) 36” hemispheres we were able to get them to fail up on the
sphere, but the thinning ratios may be different with thicker material. It may also be related to the variability of
hand spun
I had ordered a set of 65” diameter spheres early on when
the rules for the lunar lander challenge were just
under discussion, and we thought we might be doing a “vacuum performance correction”,
which was encouraging me to go thinner on the tank walls than I otherwise
would. I am happier with the final LLC
rules that just involve a simple hovering time, but it does mean that we may
not have a great use for these particular tanks. They are 3/8” pre-spun thickness 5086, since
5083 couldn’t be had at that size on short notice.
Just by mass, if the 36” 5086 sphere burst at 710 psi, the 50% thicker 65” 5086 sphere should burst around
590 psi.
Unfortunately, two of the hemispheres had what appeared to
be cracks in the metal along the spinning lines on the inside. We decided to use those as our test tanks,
and we made good marks where all the cracks were before welding the tank
up. James used a surplus crawler gadget
to rotate the tank instead of the old mill 4th axis, because it
could be controlled with a foot pedal, and the high frequency welder start
occasionally caused the mill axis to jump around (very odd, since we had the
welding stuff completely isolated from it with a couple inches of phenolic).
When we hydrotested the 65”
sphere, it ruptured at 480 psi exactly along one of
the interior cracks, and it had also visibly stretched along the line of the
second deepest crack. I’m going to see
what our spinning vendor (AMS Industries) can do about this issue. They have said that the 65” form is one of
their oldest, and it may need some refurbishment, or they may need to modify
their spinning technique. We could
probably weld up the rest of the spheres and proof test them to 450 psi for operation around 300 psi, but we probably won’t make that decision until
post X-Prize Cup. We might wind up with
OTRAG style tube cluster propulsion units instead.
http://media.armadilloaerospace.com/2006_07_05/tack65.jpg
http://media.armadilloaerospace.com/2006_07_05/crawler.jpg
http://media.armadilloaerospace.com/2006_07_05/crack1.jpg
http://media.armadilloaerospace.com/2006_07_05/crack2.jpg
http://media.armadilloaerospace.com/2006_07_05/crack3.jpg
http://media.armadilloaerospace.com/2006_07_05/crack4.jpg
http://media.armadilloaerospace.com/2006_07_05/crack5.jpg
http://media.armadilloaerospace.com/2006_07_05/crack6.jpg
New electronics
Our next-generation electronics are ready for bring-up. The core differences are:
Newer revs of all major components – CPU board,
IO board, GPS board, Esteem wireless, and IMU.
Moved from a 12Ah to 18Ah battery for the
main computer systems to allow it to run essentially all day without charging.
Combined both the actuator and CPU
electronics into a single box. We
previously had them in their own isolated boxes, but It
was awkward and I didn’t like the required communication cable going between
the boxes. Power and ground are still
completely isolated inside the box.
Current sensors on the actuator drives,
which will be useful for a lot of diagnostics.
More sensors and motor drives, sufficient to run a four
engine differentially throttled biprop vehicle.
Discrete connectors for everything,
instead of giant plugs with wiring harnesses. I can see both sides of this, but on many
vehicles we will be able to connect sensors and actuators directly to the box,
without requiring an additional connector in between, which should be a
reliability improvement, and it is a lot more flexible for trying out different
things.
Sealed connectors.
Connectors that are directly soldered to the board (but
bolted to the case), eliminating all internal wires except for battery power.
The box top is the most complex milling project I have done
so far. I had some initial trouble
breaking 4-40 taps in blind holes with the rigid tapping on the mill, but
switching to thread forming taps made everything work perfectly. We tried thread forming taps for hand tapping
years ago without success, because I didn’t know at the time that you need to
use a larger drill than the standard tap size.
Perfectly obvious – since you aren’t removing metal while tapping, it
has to be removed while drilling.
We are building up two identical electronics boxes, and have
a spare board prepared if we need to make a replacement. We will probably do our first couple flight
tests with the old electronics, then switch over to the new ones.
http://media.armadilloaerospace.com/2006_07_05/electronicsLidInside.jpg
http://media.armadilloaerospace.com/2006_07_05/electronicsLidOutside.jpg
http://media.armadilloaerospace.com/2006_07_05/unpopulatedBoard.jpg
Engine failures
We had a bad month for engine tests.
We decided to try and thread the carbon wrapped graphite
chamber so it could screw on to the injector.
I made a temporary adapter flange that bolted to the current injectors
and left a screw thread for the chamber, but if it worked out, I would change
the injector design to eliminate the flange completely.
We had several reasons for wanting to try this:
The tie rods and clamp flange are 12 pounds that needs to be
swing around by the gimbals.
Adding phenolic insulating tubes
around the tie rods adds another 5 pounds.
I am still worried about the bottom phenolic
insulating plate charring and compressing somewhat on long runs, resulting in leaks
at the top gasket.
Thread on chambers would be great for an OTRAG cluster.
I thread milled an 8 pitch thread in the graphite chamber,
and it looked very nice. We threaded the
chamber on, but as a safety precaution we also added our existing tie rods and
clamp flange around it, but with a quarter inch space between the nozzle end
and the clamp flange, so if the thread broke, the chamber would be caught
instead of launched like a cannon ball.
Good foresight. The engine
fired up nicely, but after three seconds of firing, there was a bit of a puff
by the injector, so I shut it down. It
wasn’t immediately obvious what had happened, but the chamber had broken on the
first thread, and the graphite slid up inside the carbon fiber overwrap. I was
amazed that the failure wasn’t more catastrophic. I probably could have stress relieved the
thread better by ramping in the thread mill, but if it would be that close to
failure it just isn’t a good idea. It
should work great for phenolic engines.
http://media.armadilloaerospace.com/2006_07_05/threaded.jpg
http://media.armadilloaerospace.com/2006_07_05/brokenThread1.jpg
http://media.armadilloaerospace.com/2006_07_05/brokenThread2.jpg
We machined off the threaded part, giving us a chamber that
was just an inch shorter, and prepared for another set of tests.
On the threaded engine test we had also tried another
experiment – mounting our chamber pressure transducer directly to the engine
igniter with only a sintered metal pressure snubber
to temperature isolate it instead of a 6”+ length of 1/8” stainless tube. Bad idea. Cooked transducer. We went back to the isolation tube with a new
transducer, but the new sensor contributed to some of our problems that day.
The way our ignition interlocks work is that the igniter
chamber must first show > 80 psi of pressure to
let us know that there is flame, then, after the main valves have throttled up
to idle and the igniter has been shut off, we must see at least 15 psi of pressure from the main chamber, or we shut
everything off. This relatively small main
chamber pressure can be problematic, because 500 psi
transducers may have a 10 psi zero bias. We ran into this with the new transducer,
with several failed starts because the pressure wasn’t quite high enough. I finally went in and raised the idle
throttle point a bit. There is a safety
trade off here – the higher the idle throttle point, the more propellant gets
thrown into the chamber before you check to see if you have combustion. I’m not sure it is really possible to fail to
light main combustion if the igniter torch has already been determined to be
functioning, but it is a bit of a concern.
I almost chased a phantom issue, until I remembered that we
had seen the exact thing before. We were
seeing a chamber pressure rise just after the engine was shutting down due to
lack of chamber pressure, implying that if I had just checked 200 milliseconds
later, it would have looked like combustion was going well. I recalled on a previous test day changing
the sense point three times, and the pressure rise always moving farther away,
until I realized that the pressure rise was due to the manifold purges on
shutdown pushing the propellant out faster than the cracked idle throttle,
resulting in a brief burp of higher thrust.
After several failed starts, we always worry about putting
too much heat into the igniter, but when we checked on the engine we had
another issue: there was a small puddle of alcohol in the chamber from the
manifold purges after the failed shutdowns.
We have the engine canted down 10 or 15 degrees, but the converging
section is 30 degrees, so it is possible for puddles to form, and it appears
that our purge gas isn’t sufficient to entrain that out of a 3.75” diameter
throat. I hate horizontal testing. We may be able to add a bit more angle, and I
may upsize our purge system.
After increasing the throttle setting to guarantee that it
made enough chamber pressure to bypass the automatic shutdown, the engine fired
right up and I throttled it up to a perfect looking burn. It was making 2500 pounds of thrust at
relatively low feed pressures, and with the smaller test stand valves, which
was right on track. After a few seconds,
there was a white flash in the exhaust as some aluminum started burning, and a
half second later the engine just fell down off the test stand mount and
sprayed combustion gasses and burning aluminum up through the igniter port all
over everything.
The injector was obviously toast, but it also managed to
ruin every single actuator (main valves, igniter solenoids, purge solenoids),
and the wiring harness was torched. The
graphite chamber was undamaged.
Our first thought was that all the failed starts had put
enough heat into the igniter to get it to melt, but a little more reflection
produced the most likely failure method:
Our igniter threads into the main injector body, and also serves as the
main engine mount. This was a
convenience to allow us to swap out different injectors with the same
igniter/mount, but we have had a couple cases where we have felt the screw
thread get loose. We didn’t explicitly
check that for tightness on the test day, and all the handling, or possibly the
breaking of the threaded chamber, could have easily loosened it up. A loose thread there would let combustion gas
flow up through the igniter port, melting everything just like we saw. We’ll just weld them on in the future.
I had the foresight to make the test stand wiring harness
extra long on the last rebuild for just such occasions, so I chopped off the
last two feed and attached new connectors.
It did turn out that we blew one channel on the driver board when the
purge solenoids shorted out as all the insulation melted away. For test stand work I temporarily moved purge
to one of the roll thruster channels, but I will have to move it over next week
to a different connector where we still have a spare channel that was
originally for the servo regulator.
Next test day, we had a brand new injector with a welded on
igniter, a new set of actuators, new connectors on the wiring harness, and a
completely new failure.
Before starting the engines I always let some lox flow
through it to pre-chill the lox manifold, because otherwise the startup is
extremely rich with just gox mixing with the
fuel. Since we had problems the previous
test building chamber pressure at idle, I let this pre-chill run for longer
than usual. Combined with the fact that
we used the larger vehicle valves, this meant that quite a lot of lox flowed
out.
When I started the engine, the igniter lit immediately, but
as soon as it opened the main fuel valve, the top of the fuel manifold let go
with a bang. This resulted in some
alcohol spraying around, but we got that extinguished without harming anything.
Pretty clearly we managed to fill the fuel manifold with
oxygen during the horizontal pre-chill.
Lox just rolled down out of the lox injectors into the 45 degree fuel
injectors. Did I mention that I hate
horizontal testing? I am going to add an
automatic helium purge to the start of the ignition cycle now.
Russ managed to actually hammer the peeled up fuel manifold
back down and re-weld it, so we are going to try again on Saturday.
News
