14, 2005 notes
attempt at a hanging hover test showed that we did indeed suffer from the two
problems we thought might show up: the
engine ran rough above 60% throttle, and the GPS lost lock due to vibrations. Interestingly, Elon Musk at Space-X had
mentioned to me that they reproduced the exact same GPS vibration issue we
saw. He was also able to pass on some
relevant advice: bonded rubber
isolators really don't work very well to reduce vibration, but wire rope
isolators do a good job. The same line
of wire rope isolators that we used to use as landing gear scales all the way
down to tiny ones for isolating electronics components in these situations. I ordered some of the small isolators, but
unfortunately our electronics are hanging vertically, so the compact wire rope
isolators were supporting the GPS in shear, which didn't work out well. We wound up going with our previous solution
of wrapping the GPS board in foam and sticking it to the board.
had the bleed solenoid plumbed up to the vehicle to address the roughness by
venting ullage nitrogen into the top of the engine, so I changed the code to
automatically enable the solenoid whenever the throttle exceeded 40%.
finished the isolated voltage signal circuitry, which took us a day to get
working properly. I now have logged
voltage levels from the main computer batter, the actuator battery, and the
master cutoff battery, even though they are all independently isolated. This is very nice. You can see clear voltage drops when the
bleed solenoid actuates on the main battery, and when the jet vanes start
moving back and forth on the actuator batteries. I tried to buy a wattmizer solenoid from Snap Tite to replace the
8 amp nitrous solenoid we are using for the bleed valve, but our distributor
said that Snap Tite now has a 100 unit minimum order for that
the repaired (again) differential pressure transducer back from Omega, so we
spent a full day working with that. One
of the techs said that we blew out the amplifier by hooking the negative output
of the four wire sensor to ground and just reading the positive output on a
single ended channel, so I rewired everything to read the two output channels
independently and subtract them in software.
One thing that tripped us up is that the sensor (PX82) seems to be
sensitive to common vacuum, even though it is insensitive to common
pressure. I'm not sure exactly how that
can be, but it messed up our results until we figured it out. The other thing that was causing grief was
that the combination of our A/D board (Diamond MM32) and this sensor gave
different readings if it was read only once a second versus if it was read 20+
times a second. This caused me fits,
because my "-testADC" and "-testLevel" options were only reading and printing once
a second, and the values I was getting for the sensor were fluctuating as if
one of the lines were floating. I only
clued in to the issue because I looked at the sensor graph from the normal
flight control loop, which runs at 180 hz, and was perfectly smooth.
combination of using a faster A/D sampling rate and always venting the loading
vacuum, it looked like we were getting clean and repeatable numbers as we
loaded two drums of water into the tank, checking every hundred pounds. Unfortunately, the next day we met, the zero
point had shifted significantly on the sensor.
We calibrated again with two drums of water, and the deltas still looked
consistent, but the zero shifted AGAIN after we power cycled everything later
that day. I ordered a second sensor for
backup, which should arrive in a week or so.
We will swap it out and see if this particular unit is just some sort of
times during our testing we froze the loading check valve open during the
nitrogen pressurization. This is a
pretty big problem, because propellant can then just backflow out, filling the
pressurization line and possibly going back into the nitrogen tanks if we
allowed the pressure to equalize. We
could alternatively pressurize on the dry side, but we were concerned that the
venturi effect might harm the sensitive differential pressure transducer (now
that I have looked at the spec sheet, we shouldn't have worried). As a solution, we extended the
pressurization hose with 40' more line and stuffed the line into a drum filled
with water so the cold nitrogen can heat up some before pushing into the check
valve. In the three loadings we have
done since then, we haven't frozen the check valve.
tested the pop-safety relief valve on the tank. It is a big 1" valve that is intended to be able to vent
faster than we could possibly pressurize the tank from our nitrogen tanks, so
there is no chance of us bursting the tank if we accidentally, say, had all six
bottles on the six pack open instead of just one at a time. When we bench tested it we were disappointed
to find that it popped at 250 psi instead of the 300 psi I had ordered. On the tank, we found another characteristic
-- when it pops at 250 it won't close again until the tank drops all the way to
200 psi. We are going to have to be
careful aiming for 240 psi or so for our boosted hops. I have ordered another one of these to see
if we get something closer to 300 psi.
raining again on Saturday, and while we have been doing hover tests in light
rain, we are probably pushing our luck, because our connectors are not fully
sealed. We decided to reschedule for
Sunday, which turned out to be a perfect day for testing.
hover test was done on the blow-out stands and went flawlessly, with a rapid
warmup and a perfect hover for 15 seconds before descent. All telemetry was perfect, the engine ran
smooth with the bleed valve engaged, and the GPS precision didn't change at
all, so we set up to do a ground liftoff with more flight head room.
warmup was again rapid, but just as it lifted off the GPS precision dropped
significantly, causing an immediate auto-land.
The GPS didn't totally lose it, but the PDOP went from 170 to 700, which
isn't a good level to be at. We didn't
want to vent out the entire drum of propellant through the main engine while
the vehicle was sitting on the ground, because the backwash would heat up too
many components and the tank base, so we attached our little waste-catalyzer
engine to the liquid drain valve to dump everything out. We had clearly improved the GPS resistance a
lot, so we opened up the electronics and repacked the GPS with another layer of
softer foam around the initial layer we had used.
We set up
for the third run, and had a problem.
The engine just would not clear up, no matter what I did. The thermocouple was reading plenty hot but
we have found that with a 12" diameter engine, there can be large
variances across the face. All of our
normal tricks with throttling up for a big slug, then putting it back at the
idle level failed to get it to clear up or build significant chamber
pressure. The engine had just run clean
an hour previously, and nothing had really changed. Our current theory is that because we used the burn-off engine to
vent the tank, the engine didn't get its usual nitrogen vent cooling. The engine would have heat soaked a lot back
from the hot pack into the cold pack, raising it well above its operating
temperature of 100 C. We know there are
issues when combustion starts in the cold pack instead of at the flameholder,
and that may have been the case here.
We were running low on battery power, so on Tuesday we will try a repeat
of the ground liftoff test with a cold start of the engine. If the engine is actually dead after only
five firings, we will be extremely disappointed.
lesson from these tests is that bleeding ullage nitrogen into the engine dome
seems to have killed the high throttle chugging roughness dead, which is a Very
Good Thing. In our course of trying
every damn thing we could think of to fix the problem on the test stand, John
Carr was the one that convinced we to try gas bubbling, and I was honestly
quite surprised that it turned out to be the miracle cure. Russ has mentioned that since then he has
found some references to helium injection being used to cure some chugging
problems in the Apollo program, so I guess it isn't such an oddball idea after
configured a pair of high cracking pressure relief valves to replace the check
valves on our LOX test engine purge plumbing.
We have seen enough normal check valves fail in the last few years that
I had paranoid worries about backflow of both LOX and methanol into our purge
system. Doug Jones at XCOR had
mentioned that relief valves were much more reliable than low pressure check
valves, so we set a couple up for our engine.
It was a good thing we explicitly tested them, because the particular
brand I got actually had a reverse-pressure bleed hole, so it didn't function
like a check valve. We soldered the
bleed hole closed, and they seem to work perfectly now.