January 23, 2008 notes:
Film cooled engine
The big lesson from the X-Prize Cup last year was that our
engine start sequence wasnt reliable enough.
One of the issues we were fighting with was that our cooling jacket held
quite a bit of volume, so starting the engines at idle required a 1.6 second
igniter operation time, and we ran into issues melting the igniter. We detuned the igniter to the point that it
didnt melt, but then we had some problems with reliable ignition. We believe there were also separate problems
with pushing hot gox back into the filling cooling
jacket, and also possibly an assembly problem on one of the engines.
We had plans of attack to resolve these issues with the regen cooled graphite chamber engine designs, involving
reducing the jacket volume and adding purges during startup, but we decided to
try another approach that might resolve it more definitively.
Making a chamber out of stainless steel and just adding
extra film cooling has some significant benefits, although it is going to
suffer some penalty in Isp. We had some data points from the radiatively cooled carbon reinforced graphite chambers that
we were using in 2006. We had one engine
that glowed orange hot before failing, but the engines that we flew at the 2006
XPC werent even glowing red hot on the outside after 90 second flights.
We recently had a chamber fabricated by spinning seamless
pipe down onto a two part tool. The
first article did not come out perfect, with the throat not getting down to the
desired radius and being somewhat non-concentric, but it was good enough to
start our testing with, and we expect future ones to be better.
I debated for a long time whether to make the injector out
of aluminum and either bolt to a flange on the chamber or slide it in and
retain with a snap ring. I finally
decided to just make the entire thing out of stainless, and weld it directly to
the chamber. At least some of our
problems can be traced to o-rings, so completely removing them has some
benefits. It also opens up the
possibility of welding the main propellant valves and igniter solenoids
directly to the engine, removing any possibility of loosening fittings and
leaks. We have checklist procedures to
check various things for tightness pre-flight, but it is better to make it just
not possible. Considering that a fitting
leak brought down the first Falcon I launch, it is a real concern worth taking
what steps we can to avoid.
I hate machining stainless, and I did break a few tools and
trash a couple parts as we worked up the first prototype engine. The trick is to just be really patient. I finally got down to one quarter the spindle
speed and one eighth the feed rate that I use on aluminum, and changed the
drill bits before each part, rather than trying to get multiple parts out of
The stainless chamber is narrower in diameter than the
aluminum cooling jacket, so our injector design couldnt be used as-is. We are trying to go back to a single ring of
injector elements with the igniter on the side of the engine, which gives us
some fabrication advantages, but in the past we only got our injector face
cooling completely resolved when we moved to a fuel ox fuel manifold arrangement,
so this may not work out.
When we finished the prototype engine, we were pleasantly
surprised to find that the assembly was 19 pounds lighter than the old engine,
and when we go to welding the valves directly to the injector, we will save a
couple more pounds.
On the first flight test we had a perfectly smooth startup, but
we saw some burning metal coming out of the nozzle, so I aborted the flight
early for inspection. Shutdown was fast
and clean. The cylindrical part of the
chamber wasnt very hot at all, with no
discoloration. There was discoloration
on the nozzle, but it wasnt uniform.
The imperfect concentricity and contour on the first-article spun
chamber probably caused the unevenness.
Still, the chamber was in great shape.
The injector had eroded under the lox manifold fairly deeply, but we
decided to go ahead and run it some more to get more startup / shutdown data.
On the second flight, I skipped the normal lox pre-chill
that we do, and it still started up smoothly.
Somewhat surprisingly, there wasnt any more metal burning on the second
run, and I flew the vehicle for over a minute before setting it down. Shutdown was again very clean. It appears that the injector may have reached
a thermal steady state thickness after burning some of the deck away.
We were going to do some more runs, but we were getting low
on our fire suppression water after putting out some grass fires from the first
two runs. Since we werent sure that the
brand new engine design wasnt going to blow up, I had been flying it at a
lower altitude than normal, which results in a lot more flying hot rocks that
can reach both sides of the runway from our test spot.
We are going out again next Saturday to horizontally fire
the engine at full throttle to lean on it a bit more than the hover testing
does. The following week we will have a
second article chamber, and a modified injector with a smaller lox-wetted area
and a thinner deck. Hopefully, that can
become our new standard engine, and we can run off a big batch of them.
We probably wont wind up flying the two-module,
differentially throttled configuration.
I wanted to fly it as part of our phase I SBIR for the Air Force
Research Laboratory, to demonstrate the fundamental nature of the modular
rocket concept take two separate flyable rockets, bolt them together, and you
have a bigger rocket, but time has run out and apparently the phase II
decisions have apparently already been made, but not announced yet. We should find out in a couple weeks if we
are getting it or not.
We have four module tank sets mostly complete now, so we
will probably be flying the fixed engine, differentially throttled
configuration next. There has been some
debate in the community about exactly how fast the valve actuators need to be for
differential throttling a vehicle, with both Masten
and Unreasonable Rocket believing that faster actuators are needed. Paul Breed went so far as to bet me $100 that
we wont be able to make a stable flight of our four module vehicle with our
KZCO actuated valves.
I had recently updates our flight computer too log all the
data at the full 200hz internal tick rate for several seconds after startup to
give us higher fidelity data than what we normally get over telemetry to help
develop engine start sequences. We were
able to get a good look at the responsiveness of the engine to valve commands
as the vehicle went from idle to liftoff thrust. Each vertical bar in the graph is 100
msec. The vertical cursor is on the frame where the computer first sets
the bit to start driving the valves open. The valve pot feedback has
moved on the very next frame, 5 msec later. The
chamber pressure has moved 50 msec later, and is
substantially changed 100 msec from first command.
These are 0.5s KZCO medium torque valve actuator being driven at 16.0V by
Russ's custom motor drive board. I think
this is going to work just fine, and Paul is going to lose the bet. J
Once we have flown the four module system and convinced
ourselves that it either works or doesnt work, we will break the modules apart
and start flying them higher and faster in Oklahoma.
While we dont have the permit in hand yet, it looks like AST has agreed
in principle to let us fly our vehicles to 4000 with our current safety
systems. By light loading the vehicles
and accelerating harder, we should be able to hit the same max-Q that our
proposed suborbital vehicles will see, since we intentionally fly rather slow due to the wide, draggy
nature of our modular vehicles. I expect
we will wreck one or more of the modules in flight testing, but we have four of
them, so it wont be that big of a deal.
Once Spaceport America gets their final permit, we
will take any remaining modules out there and see how high we can go. With the legs on, the modules dont have a
chance of getting to 100km, but if we learn all we need with normal flights, we
might make a potentially-sacrificial flight of a module lifting off from a
stand without legs. The landing wouldnt
be very pretty, but it might not be any worse than the return leg of our XPC
flight last year.
Assuming the differential throttling works out on the four
module system, our commercial vehicle plan is a six module triangular configuration
with engines on the side between the tanks, using the base of the bottom spheres
as landing pads. This configuration
gives us full module-out redundancy, easy vehicle CG determination before
launch, no landing gear weight, and it travels in flight orientation without a
wide-load permit (just barely). For a
cabin, we are going to use a 5 diameter transparent sphere. Ill see if I can get Matt to make a
rendering of the vehicle for the next update.
A couple pics of the integrally
machined fittings discussed last month: