June 27, 2004 notes
We finally found out exactly why we have had the computer
reboot on the big vehicle a couple times.
We were looking at various possibilities with the valves
over rotating in case there might be a shorted spot in the feedback
potentiometer. The pot never caused a
problem, but we did occasionally see the computer crash right when the valve
hit the shaft limit switch, and we found that we could also get the computer to
crash by manually shorting or triggering the limit switches on some of the
valves. It isnt a current draw issue,
because the actuator battery is separate from the computer battery, and
manually shorting one of the powered actuator lines can burn the transistors,
but not hurt anything else. Even with
the actuator drivers and battery completely isolated from the computer power,
the abrupt interruption of power to the motors would cause enough of an
inductive kick in the other electronics to kill the computer. Russ put a scope on the computer power and
found that there was a short +/-2 V buzz at very high frequency when the biggest
valve hit the limit switch and crashed the computer. Using the driver board to switch directions didnt kill things
because the transistors provided a more gentle switching transient than the
manual limit switches.
The two motor drive wires were run in the same cable as the three
potentiometer feedback wires, and they were even twisted together, which made
it easy for the voltage kick to couple into the sensor ground, which is shared
with the computer ground. This is
probably also what kept frying the Access IO A/D boards. The smaller vehicles had smaller valves with
less inductance, and shorter cable runs, which is why we didnt have the same
problems on them. It all makes good
sense now. I hate analog electronics.
We rewired all the motor drive circuits to be separate from
the sensor circuits, which fixed the issues for the vane motors, but the master
cutoff valve motor was still giving us some problems, probably due to the
larger size of the valve and the way it was integrated into the main
electronics board. We separated it back
out into a private enclosure, and strapped it down on the base of the vehicle
so it had a short run to the valve, leaving only serial communication on the
long run. On our next redesign, we will
move all the motor controllers away from the rest of the electronics, and fully
isolate the batteries. Currently the
actuator battery powers the spark ignition, which has to have the engine
grounded for the spark plug, and there are some voltage paths through our
pressure transducers which keep us from completely isolating them. If we use a separate driver for the ignition
system, we can keep all the motor drives completely isolated.
This was frustrating to track down and resolve, but it is
still really good news, and should clear the way for high reliability.
We also solved one of our other mysteries with the KZCO
valve actuators. The pot feedback
ranges seemed to occasionally take permanent shifts for no obvious reason. It turns out that internally the gears
connecting the shaft and potentiometer are not equal sizes, so if the valve
ever rotates past the electrical limit switches, the pot will be at a different
point when it cycles back into range on the next cycle. This is easy to do with the vanes by hand,
or while rebuilding a valve, or on occasions like our last test when the valve
had exhaust force on it while already at the limit switch. It doesnt hurt anything, and we can just
turn it back the other way if the range gets outside our valid A/D range.
I have started adding an automated self test of the jet vane
actuators at startup, as well as all the other sensors. If something isnt functioning, we have to
explicitly override the test to do anything.
Our 450 gallon tank arrived. At 48 diameter, it looks very small compared to the 63 diameter
tanks, but it is much more the right size for doing non-launch-license test
flights, which can only use 150 gallons of propellant. The tank has the same flanges as the big ones,
so the propulsion section and electronics section will just bolt right on to
the smaller tank. All we would really
have to do to get it in the air is make cables, but we are probably going to
get a conical adapter section made for the top so we can bolt down one of our
thin cones for good streamlining.
It turns out there is a commercial supplier of pre-mixed
sodium permanganate solution for use by the electronics industry. It goes under the trade name Liquox from Carus
Sodium permanganate has about 10 times the solubility in
water that potassium permanganate has, and it is really nice to just get it
pre-mixed in five gallon containers. The
price was $3.14/lb in five gallon quantities, so if we used it at a 1:20 ratio
with propellant, it would add $0.16 / lb to our current cost of about $1 / lb
of propellant. We should be able to use
a smaller ratio than that for steady state, but throttling would use it at an
increased rate. We did drop tests with
the liquox on 50% peroxide, and it is clearly a lot more reactive than the 5%
potassium permanganate solution we last tested with. Interestingly, it seemed to make a pop when we dropped it on some
mixed-monoprop propellant, indicating the sodium permanganate we be strong
enough to actually react with the methanol at the elevated temperature. We are expecting to still need a spark plug
and flameholders, but if it was actually completely hypergolic by itself, it
would certainly be a bonus. We will
probably fire some of this in our little liquid catalyst test motor next week,
but we are also waiting for a new type of static mixer from Sulzer to replace
the laminar flow helical mixer we are currently using. If we see good performance and throttling,
we will probably try building a 12 motor with liquid catalyst to compare with
our existing motor designs.
We received a set of 7 machined tubes, so we now have all
the pieces to make completely fresh and uniform engines without cannibalizing parts
from old engines (except the nozzles).
We built up a new engine with all of our current thinking on best
practices, and we plan on basically testing it to destruction, doing boring run
after boring run until it stops working.
This engine has a fairly high open area spreading plate, so there is a
chance it may not run smooth, in which case we will have to cut off the top and
replace it with a different one.
Unfortunately, Global Stencils laser is giving them problems right now,
so the rest of the laser cut spreading plates are delayed.
The engine assembly procedure for the test engine was:
10 long by 7 ID by 0.20 wall thickness 316 tube. Rolled and welded, then machined both inside
2.1 nozzle throat, expansion trimmed down to roughly 2:1
ratio. The low ratio is because I still
worry about flow separation while deep throttling on the vehicles.
0.5 gap at the bottom of the tube to make space for the
thermocouple and pressure tap.
0.5 thick water jet cut support plate, deeply beveled for a
3 x 8 mesh 316 stainless screens
1200 grams of ring catalyst. This is slightly increased, because we felt that we didnt have
enough depth with 1000 grams after compressing.
1 x 8 mesh 316 stainless screens
0.25 thick perforated metal retaining plate, welded in with
the catalyst under 4000 psi gauge pressure (about 8000 pounds)
0.25 thick perforated metal plate for flameholding
0.75 gap for spark plug
0.5 thick water jet cut support plate
2 x 8 mesh 316 stainless screens
2 x 900 cpsi x 1 thick catalyst monoliths. We have previously used the 900 cpsi ones
with only 1 thickness, but we are trading reduced thrust for increased life
20 x 20 mesh 316 stainless screens for flow spreading
1676 x 0.022 hole laser cut spreading plate. I have three different hole counts on order,
this is the one with the maximum open area.
Brazed in under 2000 psi gauge pressure.
9 x 0.375 spacers
Top plate welded in under moderate pressure. This is a new option, designed to hold the
spreading plate rigid. With our
previous flanged designs we couldnt control the height tightly enough to do
The Scaled Composites team deserves huge congratulations for
the 100km flight of Space Ship One on Monday.
They probably have the X-Prize in the bag now, but just in case, I did go
ahead and place orders for all the long lead time items we still need. If their flight had been flawless, I
probably wouldnt have bothered. We can
still have our final vehicle assembled this year, but it isnt clear that we
have time to recover from the inevitable setbacks during testing.