March 7, 2004 notes
We finally started logging thermocouple data to help us get
to the bottom of our engine warmup issues. What we did was drill the
swagelok fittings on the engines all the way through, so a 1/8" pipe can
pass completely through it instead of butting up against the internal
stop. We then took an immersion thermocouple with a 1/8" shaft, put
the swagelok ferules around that, inserted it all the way into the motor, and
tightened the swaging nut to lock the ferules on the TC shaft.
We can now interchangeably put either a pressure transducer
or a thermocouple at any port, which is very handy. Conveniently, a 1000C
thermocouple (K type) and amplifier exactly fits our engine operating
temperatures. We get up to about 1800F under well mixed conditions, but
it does go higher under the flameholders where the water isn't mixed well with
Our data acquisition system is now logging eight channels:
load cell, tank pressure, between catalyst pack pressure, chamber pressure,
valve position, flowmeter current (which we still have never gotten working),
upper thermocouple, and lower thermocouple.
None of our engines have ports for both the pressure and temperature simultaneously,
but they would be easy to add.
We have learned some good stuff with the temperature
logs. When the spark hasn't caught anything in the flameholder, the
exhaust is a steady 100C, exactly what you expect from decomposed 50%
peroxide. This answers one question we were always a bit unsure about
the methanol is all completely vaporized at this flow rate, but the water is
still almost all liquid. At higher flow
rates the temperature probably goes down somewhat. When we do get a flame started, the temperature starts going up
fast, but there are some interesting changes of slope during the
progress. When the thermocouple is below the final cat pack, it sometimes
shows a rapid rise due to flameholding below that bottom plate. We have seen this in many of our warm-ups,
where we get a popping flame out the bottom of the engine, but it doesnt
actually heat the pack very rapidly.
When the flame is held above the catalyst, the top probe warms very
quickly, with the bottom lagging by several seconds.
Once the flame catches, it takes about 20 seconds at 15%
throttle for the bottom of the pack to reach about 950C. If we let it get that high, throttling up
has no clouds at all, it just makes perfect thrust. However, the flame still doesnt catch reliably from the spark
plug when the engine is fully assembled.
When we increased the plug gap to 0.125, it seemed like we got
perfectly consistent ignition in open pack tests, but with a catalyst and
nozzle below it, we had more difficulties.
Several times now, we have seen a situation where we make a
change that seems to result in reliable ignition, only to have it fail on the
second starting of the engine. We are
theorizing that this may be due to different behavior after the actual
flameholder has reached a higher temperature.
Because we are right at the boiling point of water, if the engine is
cold, the liquid water hits the flameholder angles and slides off the side,
while the oxygen and methanol vapor more rapidly curl up underneath,
effectively being a gas /liquid separator.
When the engine is hot, the flameholder may be quite a lot hotter, so
the incoming water may vaporize on hitting it, which then lets the water vapor
curl around with the oxygen and methanol, making it difficult to light. We have found that if the throttle is
briefly opened higher, that the large inrush of propellant initiates a
temperature rise, and if it can be backed off to 12% - 17% throttle, it will
then continue warming reliably. This is
tricky to do with manual control, but we can easily do it with the computer if
necessary. Our MSD-10 ignition system
has also just arrived, so the 5x stronger spark may let us bypass the
I had a theory that our high thrust instability might be due
to the fact that our flameholder engines have directly supported the foil
monoliths, which act as excellent flow straighteners, possibly giving more
laminar flow past the flameholder bars that would be more susceptible to
disruption. All of our previous engines
(that didnt have stability problems) had
supported the monoliths with a perforated metal plate, which induced lots of
turbulence. We built up another
production engine with the catalyst over the perforated metal plate, and a
flameholder cross in the 7 section below that.
The combination of the perf plate, the angle cross, and the
5.5 to 7 step section made a SPECTACULAR flameholder in open pack tests, and
with the wider plug gap we were able to light it every time we tried. However, when we welded it into an engine,
it was difficult to get the flame started again, and once we got it going and
were waiting for the bottom temperature to reach fully warmed, we saw some of
the dreaded sparks coming out the nozzle that usually mean we are melting
something internally. We went ahead and
ran the engine, which still demonstrated the high thrust instability. In later tests we found that it would run
stable with our cavitating venturi inline, but that only produced 480 lbf.
The first thing that melted was one of the thermocouples,
but when we cut the engine open we found that there were various other parts of
the hot pack damaged. From the
discoloration, it was clear that the center of the angle cross flameholder got
the hotesst, and that the 3 flameholding section was nowhere near long enough
for a flame that size to evenly mix with the surrounding flow, resulting in a
very hot cross impinging on the hot pack and damaging things. We cant afford to make the engines longer,
so we are going to have to deal with smaller flameholders. It may turn out that the best thing to do is
just reliably get the flame started on the catalyst support plates that we have
always had, since those never showed any high thrust instability like the cross
bars seem to produce.
Our big order of new ring catalyst just shipped, so we will
thankfully be able to replace the parts we burned before that holds us up.
Matt has been too busy at Fountainhead the last couple weeks
to work at Armadillo, so we dont have any pics or video of the recent work. We have also had a new face around the past
two weekends -- James Bauer, a local welding instructor, has been giving us
some tips and doing some welding for us.
We have never had a failure due to welding, but since Russ is self-taught,
it is nice to have someone with more experience point some things out to us.
Powered Landing Sim
I finally got the right algorithm for the powered landing from
altitude on the simulator. It is
trivial if you have sensors with no noise or latency, and you have engines that
respond instantly, but I planted quite a few simulated vehicles while
experimenting with realistic constraints.
Decision #1 is what altitude to transition from stabilized
falling mode to landing mode. Eventually, the decent wont have the
engines on at all (or just barely at idle), but our early tests will have them
operating at 25% throttle to maintain attitude stabilization after burnout all
the way to landing. The conservative
point for this is based on what altitude you exited boost mode, with the
important caveat that "exited boost mode" is when the throttle
reached stabilize level, not when you started to throttle down. I kept
smacking into the ground too hard until I figured that little bit out.
Eventually, when we are going fast enough that drag at the end of the boost
time is significant and our available acceleration at landing is significantly
higher than at launch due to propellant weight, the landing altitude will get
much lower than the exit-boost altitude, but we will have to conservatively work
our way towards those numbers with test flights.
The constant decision during landing mode is whether to
throttle the engines up or down. I tried having it mimick the boost
acceleration curve or aim for a constant decelleration, but the finite movement
speed of the valves made it always lag what you wanted, and hence hit the
ground harder than you wanted. What worked well was to take
the current altitude, velocity, and acceleration, and calculate what the
velocity at ground level will be (if it is even going to hit the ground under
those conditions), and throttle up or down based on if that velocity is above
or below our target 1.5 m/s landing rate. Works great.
I still have to do a little work to make it aim for the
landing rate 3 meters above the ground to account for gps inaccuracies.
The 1.5 m/s landing rate still gives the vehicle a pretty good bounce, because
even though the engines start to throttle down to sustain level as soon as they
detect a ground bounce, the vehicle is nearly hovering when the isolators kick
back. The GPS velocity is so accurate that we could aim for a
feather-soft touchdown if we had a more accurate altitude signal, but we need
to trade it off against what the GPS can be off by. A laser or rader
altimeter, even with a fairly low update rate, could give us the necessary
data, but I still have concerns about lasers and landing engine dust.
Differential GPS, with the vehicle basically landing very near the differential
sensor might also do the trick, with no vehicle modifications.
On long burns, the vehicle can pick up a fair amount of
horizontal velocity, landing off horizontally by maybe 5-10% of the
altitude. We can use the GPS to aim for zero horizontal velocity, but to
do this we will need to align the vehicle with true north at the launch pad, so
it knows which way to tilt to modify velocity. There will still be some
drift, because it isn't aiming for an absolute position, but it should come out
very close. Aiming for zero horizontal velocity on landing will also
allow us to use lighter landing gear in the future.
With the sub-one-G acceleration that the current engines are
going to give, the vehicle isn't going to go very high at all. A single
drum of propellant will not even get it to 1000 feet in a 30+ second total
flight, and if we are conservative on the warmup propellant and boost times, it
is only going to go a few hundred feet up. We could creep off the pad
with two drums of propellant and still stay within the 3500' box at Burns
Flatt. We will need the big engines to go high and fast, but we will
probably still want to reduce their nozzle size somewhat so we don't have
ultra-low chamber pressures and possible flow separation during final landing,
or 4G accelerations at launch. I haven't done sims yet, but with the big
engines and a launch-license-limit propellant load, it should go in the
neighborhood of 20,000' altitude.