August 13, 2007 notes:
Two 180 Second Flight Tests
We had three flight test outings this month, with somewhat
We are in a little bit of an unfortunate situation. Cessaroni
Aerospace, the company that has been machining our graphite chambers, informed us
that their preferred grade of graphite for this application will not be
available for four months due to the supplier being obligated to provide their
complete production capacity to a major customer, leaving nothing available for
the smaller customers. We are going to
have to live with their second choice supplier for the rest of this year. They consider the exact supplier and grade of
graphite a trade secret, so we dont know the details, but the machining
characteristics are definitely different on the new graphite, and we have
suffered two failures with it.
We had one engine with the old graphite that had made ten
consecutive long duration flights on Pixel without a problem. We moved it
to Texel, and
cracked the graphite chamber on startup due to factors discussed and addressed
last month. We replaced the chamber with
an old, eroded one, but I didnt mention in last months update that when we put the chamber into the engine, we
found that it sat at a slight angle, and we had to wedge a small spacer on the
side to get the injector in. We couldn't recall if it had been like that
originally, and the tilt was due to some weld distortion, or if it had been a
result of either all the thermal cycling on the engine, or the last start that
cracked the chamber. We went ahead and
used it for the single module flight tests, and we didn't have a problem, but
that was a light vehicle, each hop was less than 20 seconds, and it was the old
We received a new chamber with the alternate grade of
graphite, and installed that in the same engine. It still sat at a bit of an angle. Our next test was to attempt back-to-back
180+ second flights under tether with Pixel. This was a full load, with a
2200 lb liftoff weight, so the engine operated at a high chamber
pressure. The startup was perfectly smooth, and the vehicle was
performing flawlessly until 23 seconds into the flight, when the vehicle just
sort of tilted over in place and triggered the automatic abort.
The chamber had cracked and ejected only half of the nozzle expansion cone,
which resulted in a side force at the bottom of the engine in the couple
hundred pound range. The telemetry was interesting to look at. When
it happened, the gimbals started to get pushed over by the side force, but they
ramped up to full force and stopped the gimbal
motion, and even began to slowly push it into a correcting direction. It
wasn't enough to catch it, so the vehicle hit the tilt shutdown point. As
the valves closed and the chamber pressure dropped, the gimbal
actuators were able to push the engine around again.
The sobering thing is that this is the first engine failure we have had this
year that would have resulted in a loss-of-vehicle accident in a free
flight. All of the other times we have had a graphite chamber crack, the
engine was perfectly happy to continue operating at a reduced efficiency with
fuel leaking into the chamber, and we could safely land it. This time,
the abrupt failure would have happened at 60 meters altitude in a free flight,
and it would have resulted in a catastrophic crash.
This could be taken as an argument for extremely strong gimbal
actuators, but we already have 600 lbf actuators on
there (they have a six inch lever arm from the engine midpoint, but the engine
bottom is about 26 inches down from the pivot point), and I don't want to
revisit that development decision. If we had enough gimbal
force we might, maybe, have been able to keep the vehicle stable, but I don't
think it is a sure thing.
The obvious lesson is that if an engine doesn't seem to fit together right, it
probably has something wrong with it. We checked one of our other cooling
jackets, and the engine wasn't tilted, so we do think the hard start probably
bent this particular chamber.
As an extra corrective measure, instead of welding the fuel
manifold at the bottom of the tube, we now use an o-ring and spiral retaining
ring, just like we do for the injector side.
Since we cut the ring grooves and trim the tube on the mill, this gives
us an extremely accurate chamber length, and perfect flatness. The warping may have been due to the hard
start instead of welding, but we had been considering going this route anyway,
and it seemed like a good time to make the move.
We got a second chamber with the new graphite in, and fit it
up with the new engine tube. It sat
perfectly flat now, but it also revealed that the injector top and bottom faces
werent exactly square. Again, we werent
sure if it was due to original manufacturing, or part of the last hard
start. We faced it off, and got the
motor put back together with everything now nice and square.
There is one other problem that we have been trying to
address: the igniter exit nozzles in the
injector have been eroding since we moved to the four-way angled exits and the
pre-drilled metering orifices in the igniter mount. I can see it happening in the telemetry as
the igniter pressure drops significantly after the first second on the first
fire, and somewhat less on each subsequent fire. It isnt clear if it just chews it out until
it reaches an igniter pressure that it can live at, or if it will keep going
until the igniter pressure is too low for our ignition interlocks. We know we can make four runs in a row with
it, so it isnt absolutely critical, but we want to get it fixed. We have been welding it back up each time we
take the engine apart, but we decided to try threading the injector for a
stainless steel insert for this test.
We again set out to try to do back-to-back 180 second
flights with Pixel. Everything started
out fine, but at about 30 seconds something was ejected from the engine. The plume was a bit more orange than usual,
so we were quite surprised to see that it still made a complete 182 second
flight, and had pretty good propellant reserves remaining. Initial inspection showed that we had ejected
about a quarter of the nozzle expansion cone.
Looking at the telemetry, we could clearly see that this time the gimbal actuators were able to overcome the side force and
hold it as necessary to compensate. The
stainless steel igniter nozzle insert was completely gone, leaving the injector
face a bit chewed up. We were able to
find the slagged piece of igniter on the ground,
rather to my surprise.
It is possible that the remains of the stainless igniter
plug might have caused a crack in the chamber that led to the nozzle fracture,
but we currently believe that this grade of graphite just isnt as strong. To compensate, we are going back to using a
solid block of graphite without a step to submerge the fuel manifold. This is stronger for two reasons, it keeps
all of the graphite under compression from the cooling jacket, and it avoids
the stress riser of the step. We were
considering doing this anyway, but this was the clincher. To keep our same engine length, this does cut
about 7 off of our L*, but we are still quite generous at around 64. I am a little worried about overheating the
fuel manifold at high altitudes where the plume will expand to directly contact
it, but we will worry about that later.
We decided to go ahead and build up a new injector, rather than repairing the
existing one with the failed stainless igniter plug experiment. To help the igniter problem, we reduced the
lox jetting significantly, which should give both a lower chamber pressure and
a cooler flame. We were looking a lot
closer at flatness this time, and we found that very few of our machined engine
components were completely flat, even before welding. We trued things up as necessary, but I am
probably going to start using ground blanks exclusively for engine components
that need to be completely flat. Just
machining both sides doesnt necessarily solve it, because clamping down a
warped plate lets it bend back to its original shape after you machine it and
relieve the clamping pressure. We are
going to go back and investigate some other injector geometries post XPC that
may also reduce the number of welds in our engines.
For the third time, we loaded up for a back-to-back 180
second test. The first 180+ second flight went fine, and the chamber did not crack, but the
igniter was burned out worse than before.
Looking at the telemetry, it was surprising to see that the chamber
pressure was nearly the same, even though we had cut the lox (actually gox during startup) orifice size in half. It clearly started cutting through the
igniter throats almost immediately, so it seems we went from very lean to close
to stoichemetric in the igniter. Going back and looking at our earlier tests
before we changed to the mill-machined igniter orifices, I saw that our igniter
pressures were less than half of what they are now, which is a significant
factor for the melting. We are adjusting
the igniter throats to bring the igniter pressure back down, and changing the
machining to reduce the length of the throats.
We are going to repair this igniter, and move this entire
engine over to Texel
to use for the 90 second level 1 runs, then build up
another one for Pixel and the 180 second runs.
With ten more weekends between now and the X-Prize Cup, I am pretty
confident we will be in good shape.
James has been working hard on fabricating the rest of the
modules, so we can hopefully have a full four module vehicle at least for
display at XPC. All of the fuel tanks
were hydrotested this week, but we got a surprise
when one of them failed at 590 psi. A good tank bursts at over 800 psi, and we hydro each of them to 600 psi. James marks all the imperfections he finds in
the hemispheres as he is going through the cleaning / fitup
/ welding procedures, and this one let go right where he had marked a flat spot
on the tank and subsequent but weld misalignment. This was the worst of the issues on the
tanks, with about 0.080 mismatch over five inches or so, and it took over 30%
of the strength out of the tank. We have
stepped up our investigation into options to reduce the critical nature of the
tank welding. The tanks arent the
dominant cost in a module, so it is worth it for us to explore some options to
get better tolerances and a thickened weld land, even if it costs a decent
We worked out all the nosecone attachment hardware for the
module, and managed to get it fabricated and installed in time to take the
module and one of the quads down to http://www.quakecon.org/
this year for display. It was fun to
have both of my (John Carmack) endeavors present in
one place, and a lot of the quakecon crowd was pretty
well informed about Armadillo. James,
Joseph, and Phil spent a lot of time talking with the gaming crowd about the
rockets. The younger kids especially
seemed to get a kick out of it.
We fabricated a new, more robust roll thruster mount that
hangs them completely inside the legs for easier handling.
We have purchased our own crane truck, which we will be
picking up in two weeks. This is going
to make our flight testing go a lot smoother, because we can make
semi-permanent mounts on the bed for all the dewars, tanks, and vehicles, and we can make
dedicated stowage areas for all our hoses and toolboxes. All signs point to us continuing to make two
to three flight test outings a month, and more will start happening at the
Oklahoma Spaceport and eventually Spaceport America as time goes on.