May 22, 2005 notes
The design that we have been following so far uses a
separate chamber to vaporize all the lox with a small amount of burning fuel
before directing it out holes at high speed to atomize the alcohol. The problem with this is that you basically
have a small rocket engine on top of your main rocket engine, and we have
managed to hard-start the preburner twice, resulting in some bent metal. One rather obvious thing we found is that
watered alcohol has worse ignition characteristics than pure alcohol. It still isnt really difficult, but if you
have marginal ignition with straight alcohol, adding water (for better cooling)
may push you over the edge to the point where you have enough ignition delay to
cause a bang.
We decided to build another preburner that didnt have a
bottom distribution plate at all. There
is a central tube with the preburner spray nozzle and spark plug on top, and a set
of radial lox injector holes near the bottom of the tube. This just exits directly into the main
chamber, while the fuel still comes in radially from the outer edge of the
chamber at the top of the cooling jackets.
An advantage of this arrangement is that there are no problems with
melting the gox injection plate, since it isnt there any more. All surfaces are cooled by lox, so we can
run the burner injector at a higher rate and we dont need to worry about throttling
that down separately when we reduce the lox flow. We had just finished setting up a tiny 1/8 ball valve for
throttling the burner fuel, but now we can go back to just using a solenoid
instead of a closed-loop actuator.
Ignition has been nice and reliable with the new burner (we
also made changes in the spark plug position and gox flow for the burner), but
unfortunately our Isp with the existing cooled chamber has gone from not-good
to basically crap. The hot gox jet and
splash action had been contributing a significant amount to the combustion
process, and the new design is just less efficient (but safer). Another sort-of obvious thing we realized
was that for any given O:F ratio rich of stoichemetric, the lower your Isp, the
more heat you are going to be pumping into everything outside the rocket
engine, because more of the fuel is burning outside the chamber instead of
inside, and the thermal energy doesnt have an opportunity to be turned into
kinetic energy by the nozzle. We
started to melt some wires on the test stand during our runs.
Just about any injector will give high performance if you
can give it enough chamber volume for complete combustion. We built a very large stainless steel
chamber to run under the preburner and a couple different options for fuel
injection. The theory is that there
should be a range of constant temperature from the bottom of the chamber up to
some point above it that corresponds to full temperature being reached in the
chamber. With a steel chamber, you
should be able to tell this from the glow colors. Our test didnt go perfectly due to a leak past our improvised
fuel injector ring resulting in some burning at the top flange, but it looks
like the current injector needs an L* (characteristic length, the ratio of
chamber volume to throat area) of about 60, while the current cooled chamber
only has an L* of about 35.
That would be about 4 longer on the chamber section, which will
be pushing our in-house fabrication capabilities, but more importantly it will
be increasing the total amount of heat that the coolant needs to absorb. I suspect that the combination of higher
total temperature plus more area to cool will push us beyond what we can manage
at this thrust level. Load on the
coolant gets lower as engine thrust increases due to volume / area scaling, so
we might have to go up to a single bigger engine and roll control thrusters for
the test vehicle instead of two smaller engines.
The alternative is to build a better injector. We may try an unlike impinging injector with
our preburner torch just acting as a continuous igniter instead of a lox
vaporizer. The conventional wisdom is
that you cant throttle such an injector very much, and it will be more prone
to combustion instability problems if the injector pressure drop gets low, but
we may just give it a try anyway. We
will have a large pressure drop at the throttle valves upstream, so it may not
be too horrible to let the pressure drop across the injectors fall down. We also dont care all that much if the Isp
suffers while throttled, as long as the performance remains smooth.
Another little lesson:
we were rebuilding our blast deflector, but we ran out of the furnace
cement we used to keep the high density fire bricks in place, so we put some in
with RTV on the back. That worked for a
while, but on the third run after things had heat soaked, most of the bricks
flew off the stand, except the one directly under the exhaust plume, which was
held in place until the run ended, at which point it slid down to the base of
the blast deflector. The backs had the
powdery white surface of burned RTV. I
got a five gallon bucket of the furnace cement for the future.
The new actuator linkage (u-joints and hinges instead of
ball joints) and PWM software control work great:
In general, I have been anti-pump for low cost rockets. They add complexity, and if you optimize for
them you wind up with a much more fragile vehicle than if you were pressure fed. Especially when we were using peroxide and
had access to the incredibly cost efficient, robust, and pretty good mass ratio
fiberglass tanks we were using, I didnt think pumps made any sense for us. Still, as I observed the work that
Flometrics and XCOR were doing, I began to think that it wouldnt be all that
hard to put one together.
Now that we are working with lox, I am having a harder time
finding large pressure vessels for tanks that are anywhere near as convenient
as the ones we were used to. I have
only just started the search, so Im not too worried yet, but using a pump with
a trivial tank is less of a stupid idea for us than it used to be.
For a year or so, I have been saying that we might just go
off and build a pump to get the experience.
I recently had to buy some piloted ¼ solenoid valves to increase our engine
purge capacity over the direct acting NOS solenoids we were using, so I went
ahead and picked up some big ½ piloted solenoids that would be suitable for
pump chamber controls. Supplier
note: we had been getting these lines
of solenoids from Snap-Tite, but it turns out that they were just the US
distributor for Jefferson, who has now opened a direct office. Im still ticked at Snap-Tite about the 100
piece minimum order for Watmizer solenoids, so I am happy to take my business directly
to Jefferson. All good so far. $110 for the ½ piloted solenoids, $86 for
the ¼ piloted solenoids, and $47 for direct acting 1/8 solenoids.
Initially we built a Flometrics style (www.rocketfuelpump.com) pistonless
pump. They use closed loop control with
level sensors to cycle the valves, but I just used fixed timing, allowing
liquid to fly out the vent valve without trying to shut it just-in-time. Initial results were pretty good, pumping
garden hose flow up to 200 psi.
On normal 12v power, these solenoids are rated for 225 psi,
but they open up to 350 psi. They are
continuous duty at 12v, so you could get higher pressure switching ability by
driving them at 24v. Since the duty
cycle in a pump is inherently periodic, they could probably still run
forever. The NOS direct acting solenoids
are an example of doing this, they run 6v (someone said they are actually 3v)
coils at 12v, allowing them to open much larger orifices against 1000+ psi than
they otherwise would be able to. The
downside is that they draw 8 amps and get pretty hot after 30 seconds, and they
burn up if you leave them on continuously for a few minutes.
The inlet splashing issue concerned me for a pistonless pump. Our initial chambers were a bad choice for
this design, being fairly squat, but even with a narrow chamber you are always
going to be either wasting pressurant gas by closing the vent early, or splashing
some propellant out by closing late.
You also have swirl issue on pushing the fluid out, and you need to
worry about pressurant compatibility / dissolving in the working fluid. I really didnt see why you wouldnt want a
free-floating piston in the chambers. I
then realized two additional benefits of using a piston: The vent and pressure valves on top could be
combined into a single three way valve, because it is safe to leave the piston
pressed up against the top of the chamber.
You also dont need closed loop control with pistons at all you set
your cycle times based on the time required to fill a completely emptied piston
chamber. If the flow is throttled
downstream of the pump, the chambers simply dont empty all the way before they
need to refill. Avoiding any level
sensors looks like a large advantage to me.
I also realized that it would be a good idea to have a
cavitating venturi after a chambered pump to prevent an excessive flow that
could drain one of the chambers dry, resulting in a momentary total flow stoppage. With a properly sized CV, pressure would
just start dropping progressively if too much flow was being allowed.
We upsized some check valve plumbing and made version 2.0 of
the pump, with larger chambers containing polyethylene pistons. Each chamber holds about half a gallon
The parameters I use to control the valves are: Pressurize
time: the time allowed to push the fluid out of the chamber under
pressure. Isolation time: the delay
between closing the vent and opening the pressure valve and vice-versa, to
prevent pressure from rushing right out the vent. Crossover time: the time that both chambers are under pressure to
insure a smooth transition. The total
cycle time and fill time are derived from those values.
It turns out that crossover time isnt needed at all with
free pistons or pistonless pumps. In a system
with pistons connected to each other that operate in lock step, when one piston
finishes its power stroke, there is no pressure at all on the output, so you
need another piston to provide some push to avoid a severe drop in
pressure. With a free piston (or no
piston), when you close the pressure valve a chamber just becomes a pressure
accumulator, decaying in pressure instead of having the pressure go to
zero. In fact, we get upwards spikes in
outlet pressure when both pistons are pressing, due to outlet flow losses being
divided between two plumbing paths. Even
with zero commanded crossover time, we still get positive thrust kicks due to
the actuation time of the valves. I
will probably try some negative overlap values to actually put a gap between
the cycles later.
Some isolation time is necessary to keep from getting a
burst of high pressure out the vents just as they are closing. This behaves as expected without a piston,
but with a piston the piloted solenoids have a bit of a problem if the piston
gets all the way to the top and any residual pressure is allowed to completely bleed
down before the vent solenoid closes, it wont close because there is no
differential pressure to operate the pilot.
As soon as the pressure valve opens, the vent will close, but it still
lets out a puff of high pressure gas. Using
actuated ball valves instead of piloted solenoids would fix this. You can see/hear this behavior clearly in
Water hammer effect is visible on the feed hose at high flow
rates. I like flometrics idea about
submerging the entire pump inside the propellant tank and just having open
check valves for inlet.
The issue we were left puzzling over is that there seems to
be different flow rates between the two chambers. At 8 gpm you can see it a bit, but as the flow rate goes up to 20
gpm, the outlet pressure differences get larger and larger. At first we thought it might have been the
plumbing after the outlet check valves, which had an extra set of fittings on
one side, but replacing the outlet plumbing with a big box manifold didnt
reduce it at all. We added pressure
taps on each chamber, but the battery ran down on the flight computer as we
were about to test it. Either the
outlet check valves are different, or the piston is binding a bit in one of the
The other obvious thing is that as the flow rate goes up,
our regulator cant keep up with the flow.
We have been planning on making a servo regulator that uses closed loop
control on a high-pressure ball valve to regulate pressure directly from a high
pressure tank. We may try that next
Some pics of the vehicle fabrication:
We also finally bought a hydrostatic tester of our own, and
we had to break something to test it out:
This was a 316 SS tube of initially 1.65 ID with 0.036
wall. It failed at over 3500 psi (after
yielding quite a bit), which would be 80 ksi stress at the original wall
thickness, or 103 ksi at the thinned 0.028 wall at time of rupture.