September 1, 2001 Meeting Notes
In attendance:
John Carmack
Phil Eaton
Bob Norwood
Joseph LaGrave
Big Engine Tests
media.armadilloaerospace.com/2001_09_01/BigMotor.xls
We are still trying to smooth out the large engine thrust
performance. It runs perfectly clear,
but the thrust is quite rough.
We noticed last time that we got a good increase in pressure
delivered to the engine by removing the feed hose and one 90 degree
fitting. Today, we arranged things so
that we didnt have any 90 fittings at all, but we had to add a length of hose
back. The delivered pressure went up a
little more, but it was no smoother.
The engine as it will fly will have even better delivered pressure,
because it goes directly below the tank, with only straight-shot manifolds and
valves between them.
We then tried adding a 1/2" check valve right before
the engine, which is how the small engines are configured. Performance dropped off a lot, but it wasnt
any smoother smooth.
I went back and re-read some of the references on feed
system instability, and the general advice is: avoid trapped air spaces in the
plumbing, and have a large enough pressure drop at the injector.
We had been discussed trying out various pressure
accumulator options before the engine, but that would go directly against the
standard advice.
Our injector right now is the micro-etched sheet metal
stretched across the top of the catalyst pack.
The pack is 4.5 in diameter, so even with the tiny holes, it may not
provide enough pressure drop. There is
also a fair amount of volume between the fitting and the plate at the top of
the engine.
We could move to custom chem.-etched injector plates that
have a lot less total open area. ERPS
injectors were done like this, and their 50 pound motor is much smoother than
our 80 pound motor.
We could put a restricted jet at the entrance to the motor,
so the pressure drop happens before it gets to the spreading plate. A small cavitating venturi, probably about
0.15 throat diameter, would be good to test.
The first one that russ made has about a 2.6 throat diameter, which is
probably not enough of a restriction to begin cavitating. As a less efficient, but faster alternative,
we might try just adapting down to a 1/8 fitting right before the engine.
We have also discussed trying to fill up some of the open
volume above the plate by adding some more catalyst discs above the spreading
plate.
Big Frame Hops
http://media.armadilloaerospace.com/misc/sep1_01_short.mpg
http://media.armadilloaerospace.com/misc/sep1_01_long.mpg
The computer power lines are now soldered on, instead of
using the screw terminals, which we believe to be the cause of the computer
resets we have seen on a couple landings.
We didnt have any problems today, but we didnt land very hard, either.
With the added support brace weight, we are near 200 pounds
dry now. With the small solenoid driven
motor in the center, 500 psi tank pressure is enough to lift off and fly
around, but it requires about 3/4 throttle.
The major change is that we are now using a completely
different scheme for attitude control.
The first hop was with five liters of peroxide, and I just
crept it off the ground, because I wasnt sure that the new software was going
to work. I was pleasantly surprised to
see it hovering extremely level, with only a slight vibration back and forth,
and when I pushed the joystick a bit, it responded instantly, instead of in the
lazy way the old control logic would.
The next hop was with ten liters of peroxide, the most we
have ever loaded in any tank. We had
one surprise while doing this. Loading
two and a half gallons through our existing fill solenoids and plumbing took
almost ten minutes. Near the end, a
blurp was heard from the tank, as if the nitrogen solenoid had been turned on
briefly. We believe that this was a
small amount of peroxide decomposing in the plumbing, because the solenoids on
the fill cart can get very hot when held open for several minutes. We will be moving to positive displacement
pumping instead of vacuum loading as soon as our Teflon pump gets here, but if
we do another 10 liter run with the current setup, we may want to replace the
solenoid with the 1/4" manual needle valve that we have.
The flight went perfectly.
The computer does a much better job keeping the attitude than I do
keeping the altitude with the throttle.
Bob mentioned that they have used laser ride height sensors on race
cars, which may be an option for us as a low-altitude height sensor, which
would allow me to add computer controlled hovering.
We still need to figure out how to smooth out our big motor
and get it throttled to a point where it is making a steady 200 pounds of
thrust, but after that, we should be rapidly moving through:
A flight with ballast equal to a person.
A short hop with a person on as a passenger, just holding
on.
A hop with the person controlling the joystick and throttle themselves.
New Attitude Control Software
I had been working the past couple weeks on a different
approach to attitude control, and while I had simulated it under a bunch of
different conditions, I wasn't really expecting it to work out in the real
world on the first try. Under simulation, it would fly a lot of vehicle
configurations (offset CG, under performing engines, etc) that would just fall
over with the old code.
Both the old and new ways start out the same -- they derive
a "current angle" for each axis from the current world space vectors
(you can't just integrate angular rates, you must rotate 3D axis, then derive
angles from there), and a "desired angle" from the joystick
input. Based on the difference between
the current angle and the desired angle, a desired rate of turn is computed.
The old code would scale the delta between the desired rate
and the current rate (what the gyros are saying) by a gain factor to give a
desired torque to modify the throttle value of appropriate engines to hopefully
move things in the right direction. Each engine wound up with a 0.0 to
1.0 throttle level, which was then expressed as a pulse-width-modulated series
on the solenoids.
The new code changes things around so that instead of
treating the engines as if they were continuously variable (by PWM), it treats
them as the binary devices they actually are controlled as. Each thought
period (adjustable, I set it at 5 msec today), it calculates which of the 16
possible on/off states of the engines will best push all the rates towards the
desired rates, and also balance out the desired total throttle level.
This has a lot of benefits, because it will go all the way to completely leaving
some engines off if necessary to stay level.
This amplitude of the back-and-forth overshoots is sensitive
to complete system latency, which includes the gyro response time, the
processing granularity, the solenoid actuation time, and the actual pressure
building time in the engine. I simulated a wide range of values, but I
was concerned about the pressure building time. It turned out to work
better than I expected -- I assume the chamber blow down acts to smooth out
some of the harsher oscillations that the simulator sees with longer latencies.
The amplitude and frequency are also sensitive to the
control authority of the engines. If we
fly again in short form, with the engines much closer to the center, it will
be a lot smoother.
Upcoming Work
Im adding manual open / close switches to the test stand
electronics box to make working with the ball valves more convenient.
Hybrid engine development will be ongoing.
We still need to drill holes to vent the ball valves
upstream. It is only an issue if we
close the valve during a run with peroxide still in the tank, but that could
certainly happen on the big lander hops.
Ballistic Flight
Of the Small Lander
We will fly it without the leg extensions and landing foam,
but we will add fin material under the crossbars to give it static aerodynamic
stability. If we run the fins all the
way out to the engines, we will need to test them to make sure they wont burn
during the flight.
We need to do some testing to see exactly what happens when
we go out of contact range with the 802.11b communication system. On the initial flight, I want to do
parachute deployment with a manual signal, while we log data from the barometer
and GPS for analysis. I will probably
have it immediately pop the chute if it loses contact for more than a second. This shouldnt be a problem, because it wont
even reach 1000 in the current configuration.
We need to do a couple things to improve the shock tolerance
of the lander:
Better mounting plates for the fiber optic gyros. The time that we flipped the small lander
over, two of the FOGs came loose from the angle-bracket mountings that I had made
for them. We need to make a 4.5
aluminum cube-corner assembly to securely bolt them to.
The little plastic brackets that we are using to secure the
electronics box to the lander wont do for parachute deployment jerk. I would like to use two ratchet straps to
firmly grip the box, but we may need to enlarge the base plate for better grip.
We need to make a damned reliable parachute ejection
system. The small lander with the
electronics box cost nearly $10,000, so we really dont want to drop it from
500 feet or so. At the least, we will
want the best electric matches, dual ejection charges, and a piston deployment
system. Is it worth packing dual
parachutes?
Front or rear parachute ejection? The later vehicles will be rear ejection, but we could go either
way with this one.
Should we add insulation around the engines to reduce fire
hazard on landing? A slit phenolic tube
could be secured to the crossbars around each engine.
We should do another hop soon with the small lander to make
sure everything is still working well after we changed the electronics,
connectors, and plumbing.
Electronics box V3.0
This wont actually be a box, but will be built on a 2
diameter circular bulkhead. This
should, in theory, be the electronics package that we keep up through space
shots. We will probably do initial
testing of it on the manned lander, but it wont fit in the small one. Im planning on reusing most of the
components from the current box, so that probably means the small lander will
be grounded.
The PC104 stack will be contained in a shielded and
vibration isolated container (a Tri-M versatainer), which will eliminate the
vast majority of the electronic noise that keeps our A/D from being very
accurate, and prevents us from mounting a GPS receiver near the electronics
box.
Redundant, diode isolated batteries for the main electronics
box. The solenoids and motor drives will
have their own battery, and the master cutoff microcontroller will have its own
battery, but neither of those will be backed up.
Main power and ground distribution with heavy copper bus
bars instead of bridged terminal strips, which will reduce some other noise
sources a bit.
LCD screen for the pilot to see.
Higher power point to point wireless Ethernet bridges to
replace the PCMCIA 802.11b we are currently using.
We will need to build our own optically isolated H-Bridge
motor drives. The current driver boards
I am using dont use a separate power source for the motor, and we are going to
need four motor drives in the larger vehicles (upper vent, lower vent, master
cutoff, and main engine throttle). The
current solid state relay / power supply board will probably be re-built so
that eight additional digital bits from the computer control the four motor
drives.
High Performance Demonstrator
We will probably build two different 2 diameter tubular vehicles. The first one will just be the tank and main
engine from the manned lander enclosed in an aeroshell with side firing
attitude jets, while the second one will be a proper 1/2 scale model of the high
performance manned vehicle, with full custom filament winding.
We could take the engines from the small lander and use
those as side-firing attitude control jets, but we will probably wind up making
new ones that are shorter.