September 8, 2005 notes
We have tested two new spherical tank sets 18 diameter,
3/16 pre-spun thickness, and 36 diameter, ¼ thickness in 5086 aluminum. Unlike the previous ones, we took these with
as-spun finish instead of the pretty brushed finish that probably contributed
to stress risers. The small ones were
$198 per hemisphere, and the big ones were $650 from www.amsind.com .
The weld bevels werent as deep as we would have liked, but
the thicker hemispheres did line up better than the 1/8 thick 36 diameter
spheres we tested last month. A good
thing, because our method of welding tabs on the inside to force the
hemispheres to better alignment didnt work out at all on the thicker spheres.
The small tank welded very easily, and exceeded our burst
pressure expectations. It had a dry
weight of 16 pounds, and a full-of-water (before any yielding) weight of 125
pounds for a mass ratio of 7.8.
Burst pressure was 1250 psi, and
it failed in the section of maximum thinning from the spinning process, not at
the weld or in the heat effected zone.
The 36" diameter, 1/8" thick tank we previously tested ruptured at
340 psi, which would scale to 1040 psi for this size. We believe the extra 210 psi of strength was from the surface finish, but it might
also be from the better alignment of the thicker hemispheres.
The ¼ thick 36 hemispheres are just about at the limit of
our in-house welding capabilities. With
the relatively shallow weld bevel provided, we need to put a lot of power into
it to get full penetration, and our air-cooled torches get hot fast. We had AMS send some cut off scraps of the
metal so that James could do a lot of test welds to work out best practices,
which turned out to be pretty important.
The 36 sphere was 81 pounds dry and 950 pounds full of
water, for a mass ratio of 11.7. The
first hydrotest resulted in a small leak at the weld
seam at 720 psi.
This was a bit surprising, because usually once any crack starts, the entire
thing unzips. James identified this as
one of his weld stop points. After
re-welding that area, the tank made it up to 760 psi
before completely coming apart along the weld / heat affected zone on one
side. Examination showed that we didnt
have 100% penetration at that point, but this is still good enough for our
We will probably do a bit better on the two production tanks
now that we know what the problem areas are.
James is starting the prep work on the two 36 tanks for the next big
We had been having some annoying troubles with our Esteem wireless
communication system not connecting at startup.
I would go through a voodoo dance of rebooting one or both systems a
couple times until it would eventually link properly. When I finally got sick of it and connected
directly to the units and dumped the log files, it turned out that the wireless
chipset on the flight computer unit was failing to init on boot most of the time. That unit was in the last peroxide vehicle
that had the fall that snapped the catch tether, so it was probably slightly
damaged. Replacing the unit cleared the
For several weeks we had engine test results that were
completely baffling us. It turned out to
have an amusing cause our fuel flow meter hadnt been connected since it had
been removed to weld the flow straightening pipes directly to the meter (we had
chronic fuel leakage from the 1 NPT threads).
The fuel flow A/D channel happened to be read right after the chamber
pressure transducer, so the floating line still had the voltage from the
previous channel, only slightly decayed.
By sheer happenstance, the signal ranges of the two sensors were close
enough to each other that the fuel flow values looked somewhat plausible. Chamber pressure and fuel flow are usually
Plugging it in properly
cleared everything up, but we went through a few design changes with bad data.
In the process of investigating some odd behavior with the
lox pressurization on the test stand, we replaced another check valve on the
test stand that was behaving poorly. We
have seen several problems with check valves over the years, and this is
probably the final straw in my decision to use completely independent
pressurization tanks and regulators on the next vehicle, so we dont need any
check valves. There is some benefit to
having a single pressurization system we lost one regen
engine early on because the fuel pressure was accidentally much lower than the
lox pressure, and it would be more mass efficient to have a unified
system. Dual systems
does completely remove a failure mode, and we might take advantage of
the ability to slightly tweak mixture ratios with different pressures.
Even with the check valve fixed, the lox pressurization
still behaved a bit strangely. When we
start pressurizing, the lox pressure goes up a bit, then stays almost constant
for a while as the fuel pressure rises, then finally hits a point where it
jumps up rapidly to the same level as the fuel, then
they continue in sync to the final pressurization level. We thought that it was probably the nitrogen
dissolving into the lox, and / or cooling down significantly, so we tried
pressurizing with helium.
With helium, the lox pressure exactly tracks the fuel pressure,
so it was definitely nitrogen dissolving instead of simply cooling. A given regulator flows almost three times
the volume of helium as it does nitrogen, so our pressure rises are much faster
now, and the regulator droop on engine start is much lower. Our engine performance has also gone up,
because we arent sending fluffy lox with dissolved nitrogen into them.
We are going to use helium exclusively from now on. At some point in the distant future, the cost
of helium will become important relative to all of our other operating costs,
but we are paying $390 / six pack of helium, which doesnt seem too bad. I understand it is much more expensive for
the guys out at Mojave.
The one downside to helium is that the roll control
thrusters on the vehicle dump a lot more volume for a given impulse, and purges
drop the pressure very fast.
Partly because of the aforementioned baffling problems with
the fuel flow meter, I purchased a couple cavitating venturis.from Fox Valve for our test stand, calibrated to
flow exactly 3 pounds per second of propellant at 350 psi. I always feel ripped off paying so much for
these, and this time I paid extra for expediting because we are in a bit of a
hurry due to the X-Prize Cup. These have
been helpful, and if I didnt have them, the flow meter issue might have gone
unnoticed even longer.
One smart thing we did on the vehicle recently was to
permanently mount a compass for vehicle alignment on the top of the heat
shield. To use GPS based position hold
the vehicle needs to know which way north is.
For our previous vehicles we have just painted an arrow on the ground at
our two test sites, but since we are going to be flying at a new location for
the XPC, we would have to remember to bring a compass with us. Just gluing the half ounce compass to the
vehicle made a lot of sense.
Related to that theme of try to engineer around needing to
remember something, we are going to be bolting extension handles onto our lox
filling hose so we dont need to carry a big pipe wrench around for popping it
off. The (pretty damned expensive) cryo connector is supposed to just pop off with a twist of
the handle, but it doesnt always cooperate.
Our lox supplier said they have to hit them with hammers sometimes to
break them free, but we found that just some extra leverage does the job
I finally got around to tracking down a supplier for AN sealing caps. I
had seen these mentioned in Caroll Smiths auto
racing tech books years ago, but I never tracked down a supplier. Aircraft Spruce caries
them. They are basically soft
metal foil caps that you put over a male AN fitting
before tightening the female fitting down on top of them. We have had several -16 fittings that always
dripped a bit, and these fixed them right up.
We burned through a number of tube engines, including one
that I really dont think should have. I
am suspecting that the looser fit cooling jacket that we moved to after having
one of the chambers buckle inwards is probably a mistake, giving us too low of
Im having good success at drilling accurate impinging injectors
on our tube engines. Our original
injectors were all just straight towards the center from the outside of the
tube, but by drilling down from a point offset from the center of the tube you
can make different stream angles that result in impingement close to the outer
edge of the tube instead of the middle, giving good spray fans. By using unlike angles, you can make spray
patterns that swirl around the tube. To drill
accurate holes on the slope of a tube, I first mill down a flat with an end
mill by slowly plunging in and milling half a diameter along the length of the
tube (I found a while ago that just plunging with an end mill leaves a slight
bump in the center, which throws off small drills), then spot the hole with a
1/8 spotting drill, then finally drill through with a 3/64 carbide
I want to repeat the test again, but early results show that
swirled fuel and non-swirled lox works well, while counter-swirled fuel and lox
ran fairly rough. This back-to-back test
was with somewhat too-large injection holes (1/16), so the counter-swirl may
still work well with a higher injector drop.
We made one test engine based on the theory that it would be
good to let the lox completely vaporize before it gets to the fuel injection
point. To try to do this, we put the lox
injectors six inches above the fuel injector.
From our experience with the old preburner
based engines, we expected the temperature of the lox with just the igniter
spray nozzle to be a bit too high for aluminum, so we made cooling channels in
that section of the chamber for the lox to flow through.
Because there was so little combustion, I didnt think that
we would have problems with the lox vaporizing, but that turned out not to be
the case. Lox flow rates always
plummeted immediately after throttle up, indicating boiling in the channels. We normally leave the igniter spray nozzle
running continuously during engine runs, but I tried turning it off after
ignition so the lox channels wouldnt be getting directly heated.
With the spray nozzle shut off quickly after ignition, the
lox flow rate would slowly creep back up as the top of the chamber cooled back
down, but it was going to take forever.
I tried letting the lox flow longer before starting the igniter to
pre-chill the chamber, which helped a bit.
Then we made a bad decision we tried opening the lox flow
up a lot more before ignition to reduce the heat that went into the chamber
during startup. We had a couple failed
ignitions, then we had an igniter hard start that
peeled the top of the engine like a banana.
It is worth noting that all the fuel for the bang came from a 0.030
spray nozzle. A hard start from main
engine fuel flow is a much more serious issue.
This, of course, killed the load cell on the test stand.
We are about to test a regen
engine with an extra set of film cooling injectors half way down the tube.
We were starting to really stress out about not having a
definitely good engine for the X-Prize Cup demo flights, so we decided to try making
an ablative engine that should be more predictable and for-sure wouldnt have
any issues with throttling.
Initial tests with a 2 ID, 3 OD grade LE Garolite (phenolic linen) rod
from McMaster was promising. We
externally threaded the phenolic with an 8 tpi thread and made an injector head with a matching female
thread and a face seal o-ring. After a
20 second burn it had just started to burn through the threaded area.
For the second test, we tried extending the aluminum
injector for ¾ and internally threading the phenolic
tube, hoping that the swirled fuel would film cool the aluminum enough to
protect it and the phenolic threads, so it could have
the full thickness to ablate through. Didnt work. The aluminum
extension melted off very quickly, making our test stand blast-deflector into a
For the third test, we went to a solid 4 diameter by 12
long rod of phenolic that we bored out to 2 ID and
added a 3 ID exit cone. To retain this,
we used a bottom plate and 8 threaded studs, so it didnt require any thread
machining. This worked well, and we made
several good runs with it.
For the fourth test, we made a new injector head with larger
injector holes to get some more thrust, because we were a little marginal for
flight. We changed the chamber retention
scheme to use a pipe with threads and a metal threaded closure so it could just
be unscrewed and another chamber put in.
This was a bust we made the threads too tight, so after a firing when
it was heated up, they couldnt be turned off.
The bigger injector holes also hurt, only slightly improving thrust, at
the expense of a lot of Isp.
For the fifth test, we used a counter-swirling injector head
in hopes of bringing the Isp
back up with the higher flowing injectors, and we also bored the chamber out a
bit so that it had a real chamber / throat / expansion like a conventional
engine instead of being throatless. We intended to try a looser thread on the
retainer, but the metal for the end didnt come in, so we used the same
retainer as the previous run. This was
also when we switched over to helium pressurization.
On ignition of the first firing of this engine, the bottom
retainer blew off the engine and shot the chamber into the blast deflector
(killing another load cell
). We found
the chamber split in two lengthwise like a piece of cordwood. This is a significant drawback of the solid
rods of phenolic they are cut out of thick
laminated plates, so there are no reinforcing fibers along one diameter
direction. The phenolic
tubes are wound up as you would expect.
At first we through that this was just a result of adding
the bored out chamber area resulting in more down pressure on the retainer,
which we had only tack welded on because the throatless
engines really done have any significant down force. Later testing would show that what probably
happened was that the chamber split first, then the
retainer blew off.
Our ignitions have always started with a bit of a bang, so
we tried adding a 0.020 reducing jet in front of the 0.030 fuel spray
nozzle. This has made the ignitions
much, much softer.
The next test run did not split the chamber on startup, but it
was running rather rough. We dont know
for sure if it was because of the counter-swirled injector or the bored out
chamber ahead of the throat.
For the next test, we went back to the #3 engine with the
3/64 injectors, fuel swirled, lox non-swirled.
We were out of new phenolic rods, so we took
one of the older chambers that we had already burned, bored it out to match the
previous test, and ran it. Performance
was good, so we rule out the chamber and the helium as causes of the roughness
in the previous engine, leaving the large counter-swirled injector as the
When we got more phenolic in, we
repeated the run and had the chamber split again a few seconds after startup. We decided to stop boring the chambers out,
going back to a throatless design that should have
less stress, but lower Isp. We will probably be better off using the
thinner wound tubes instead of the thicker laminated rods, even if it limits us
to 15 second flights, because they wont crack.
We went ahead and added gimbal attach points and
mounted the engine under the vehicle for flight testing.
Ablative engines are not what we want to use, but this is a
prudent step to hit our XPC flight obligations.
A side benefit is that the ablative engines give a much more visible
exhaust plume, with the bright burning phenolic and
cotton to go with the clean burning ethanol.
A few weeks ago we did an initial test with the vehicle and
one of our regeneratively cooled engines. We modified our blow-out stands to support
the vehicle 4 off the ground, and used four climbing ropes on two eye-bolts to
suspend the vehicle about 20 under the lift.
After snapping the tether on the last peroxide vehicle, we wanted more
redundancy in supporting the vehicle.
A software error related to the auto-hover control and the
new pulse width modulated valve speed control caused the engine to only
throttle up to 50% and basically stay there.
This was not enough thrust to lift off, and the low fuel flow caused the
engine to burn through.
Because of this, I finally got around to updating my
simulator for the gimbaled engine configuration, and got the auto-hover and boosted
hop logic working properly with all the new changes.
Last night, we finally got the vehicle up in the air
(briefly) on an ablative engine.
On the first test, I briefly lifted off with manual throttle
just to see if we had enough thrust. We
did. However, the attitude data was a
noisy mess. It turned out that when the
flight computer was moved over from the test stand, it wasnt bolted down
securely, so it basically rattled around on top of the vehicle.
For the second test, I disabled GPS position hold so it
would only try to do angle hold without velocity / position gains added in,
which makes some analysis easier. I lifted
off with auto-hover, and it worked fine.
The vehicle isnt balanced very well, so the angle hold had a constant
tilt to it, but it corrected back and forth properly until I shut it off.
I turned on position hold for the next test, which we
expected to work perfectly, but when we were checking the vehicle out we
noticed that a short crack had begun in the chamber at the nozzle end. We went ahead and swapped the chamber out,
but when we fired everything back up we were surprised to find that the vehicle
was no longer making sufficient chamber pressure to lift off. It just nudged off the supports and swung on
the tether with the engine burning full throttle.
It took a little while to figure out what had happened, but
I am 90% sure that the problem was decreased lox density. It was 45 minutes from the time we loaded the
lox to the time of the third liftoff attempt, and it was under pressure the
At the boiling point at 15 psi,
lox is at 92k and has a density of 1.13.
When lox starts boiling at 400 psi, it is at
140k and has a density of 0.81.
Lox at atmospheric pressure is almost 40% denser than saturated 400 psi lox! The flow through the plumbing goes up a
little for the lower density, but it is still probably 30% less lox mass going
into the engine, which is going to be 20% less total mass, and a much richer
mixture ratio. This can easily account for the lower thrust on the last
We are implementing many little changes in hardware and
software based on the experiences, and will be testing extensively in the coming
Combined telemetry graphs for the engine-active portions of
all three tests:
Video of the three tests:
A bunch of images from the last month:
( Matt was doing training for his pilot
instrument rating for a week )