February 11, 2009 notes:
Trashed a vehicle
Its been a while, but we pretty much trashed a vehicle last
month. We were doing the first test of
the super mod with completely full propellant tanks and an external high
pressure helium tank with a computer controlled high pressure valve for tank
pressure regulation. The goal for this
design is to get enough performance from one of the modules to do the level 2
lunar lander challenge without having to use Pixel,
because we still worry a lot about slosh and propellant balance on the quad
vehicles. Moving to external
pressurization is also one of the major performance growth paths for us in the
future, so it is a useful development direction.
We didnt really expect the first cut to be able to hover
for 190 seconds, because the propellant capacity is still less than what we put
in Pixel, and the vehicle is overbuilt in a couple ways the legs weight as
much as one of the propellant tanks since they were designed to be ok for a
four module cluster, and the high pressure tank we are using is DOT rated for
3000 psi, even though we only load 2000 psi into it.
Cascade loading the high pressure bottle on the vehicle was
a new operation for us, but it went very smoothly. We loaded fuel first, then helium, then lox
While it would have been possible to use a big regulator for
this amount of flow, using a computer controlled high pressure ball valve
offers a lot of advantages. It is easily
scalable to an arbitrary size. It allows
us to tailor the tank pressure curve to minimize the range of throttle valve
movement, so instead of holding, say 400 psi until the high pressure bottle is
empty, then transitioning to blowdown, we can have
the pressure start at 400 psi, and decay by 1 psi for each second of
flight. It also allows us to leave the
propellant tanks unpressurized until just before firing.
We had tested the servo regulator over a good range of flow
conditions, but we hadnt tested it at high flow on a tank completely full of
liquid with almost no ullage space. When the engine throttled up for liftoff, the
tank pressures overshot the target value due to a very noisy pressure
velocity signal as the propellant valves and regulator valves were filling and
sinking from the negligibly small ullage volume. This was the same engine design used for all
of Pixels 180+ second flights, which would have the chamber glowing bright red
to orange early on, before dulling down as the chamber pressure decayed. Pixel was initially pressurized to 425 psi in
both tanks, but by liftoff time the tank pressures were usually down to around
400 psi in the lox and 410 in the fuel due to cooling of the ullage gas and the lox chilldown
dump, and they dropped steadily from there.
This time, both tanks were at 435 psi when the engine went to full
The vehicle jumped in the air rapidly at this high thrust
level, but almost immediately it started to burn through the side of the
chamber. It is possible that there was a
manufacturing difference between this engine and the engine that Pixel did all
the long flights with, but I suspect the issue is just that we were so close to
the limits that the slightly leaner and slightly higher pressure mixture was
just too much for it. We were only
aiming for 350 psi, which would have almost certainly worked out fine.
Normally, this wouldnt have been a big deal. I would have shut the engine off, and the
vehicle would have bounced on the tethers.
However, this happened to burn through right next to one of the gimbal attach points, and a second after the flame started
shooting out of the side of the engine, the gimbal
let go, and the engine shot over to about a 45 degree side angle, sending the
vehicle into a vicious cartwheel. The
valves started to close as soon as the vehicle hit the 20 degree tilt abort,
but it still flipped completely over and came down hard on a single tether
The tether mount broke.
If it had been a normal, half-full blowdown
module load, it would have been fine. If
it had come straight down and loaded both tether mounts, it would have been
fine. Now here is the really painful
part what actually broke was the bolts holding the strap cylinders to the
tank mount points. They werent the
right bolts. The legs and tether mount
points are all set up for a bit of a hammer fit for 7/16 bolts. The other flight module has those everywhere,
but this module had 3/8 all-thread holding the tethers on, because we didnt
have the right bolts when it was assembled, and we never went back to replace
them. That probably only had half the
shear strength, and it very easily could have loaded and failed the
loose-fitting bolts independently.
The second tether mount point then failed as well, and the
vehicle crashed to the ground. We have
always used four tether straps, and the quad vehicles have four attach points
directly on the frame, but the upper leg mounts on the modules made it
convenient to only have two tether mounts and double up the straps. A clear mistake in hindsight.
None of the tanks ruptured, but some of the plumbing around
the engine broke, and pretty much everything at the base of the vehicle got
burned before we could get the fire out.
We had to reposition our fire truck once while fighting the fire due to
wind conditions, which was an unexpected complication.
After cleaning up, we stripped everything down and proof
tested the tanks to 600 psi again, and they still turned out fine. However, most of the gear on the vehicle will
need to be replaced. We are going ahead
and building lightweight legs for it now, which will probably give it all of
the performance margin we need A new
wiring harness has been built, and most of the other little parts are on their
way, but this is a low priority project for us until the new Lunar Lander
Challenge rules are announced.
Put hard limits on the servo regulator behavior, such that
it will never throttle up when above the target pressure, no matter what the
pressure velocity it.
Test the servo regulator with completely full tanks.
Use four independent tether attach points.
Use the right bolts.
Consider moving the gimbal
mounting points to the top of the engine instead of down on the chamber, so
they cant get burned off.
Set up our big 1600 gallon fire tank with two fire hoses so
we can cover both sides of a fire simultaneously.
Self Pressurized Methane
We have successfully flown a module on self pressurized lox
/ methane a few times now. It looks like
any other module flight, but the fact that it has worked fairly smoothly is
Vapor pressurized propellants, or VaPak,
systems have some very tempting attractions.
You can fill your tanks completely full, yet still have 80% of your
initial pressure when the liquid is fully expelled from the tanks, giving the
mechanical simplicity of blowdown, but the mass ratio
and thrust of externally pressurized systems.
Getting rid of helium can as much as halve flight costs in some
situations, especially during testing with partial loads. It may also be possible to simplify or do
away with torch igniters and purge systems.
Air Launch LLC http://airlaunchllc.com/
was the most recent proponent of this with their QuickReach
rocket, using lox / propane propellants.
They fired some large engines for significant durations before their
development contract ran out. Almost all
nitrous oxide hybrid rockets, including Space Ship One and presumably Space
Ship Two, also use vapor pressurization for the oxidizer.
There are a few downsides:
It doesnt work very well for higher pressures, because
density drops fairly precipitously as the saturation pressure increases. I believe AirLaunch
settled at 250 psi, which seems about right to me. This is generally not a problem for an upper
stage or an air launched vehicle, but it is a lower than ideal pressure for
ground liftoff, where you would tend to choose a somewhat heavier tank for
higher chamber pressures and Isp. Nitrous oxide is often used self pressurized
at higher pressures, but that is more due to the convenience of room
temperature operation than any particular performance merit.
At liquid depletion, your tank is still full of a lot of
cold, dense gas, which has a significant impact on mass ratio when compared to
helium. With upper stages this can
potentially be turned into an advantage by allowing the gaseous propellants to
burn in the engine at a reduced blowdown thrust (and
presumably reduced efficiency), allowing your stage to burn the tanks to
vacuum, which is even better than anything you could do with helium. That isnt so helpful for reserve landing
propellant on a VTVL, where a major drop in thrust as you are
coming in for a landing is a problem.
Propellant conditioning is an issue for repeatability. Propellant can stratify into different
temperature regions in the tanks, especially with slosh baffles. We hoped that since the engine feed hoses
would cause more boiling than the tanks, the convective cooling would stir
things well enough, but it doesnt work out that way. We currently deal with this by shaking the
rocket under the crane to stir the propellant as it warms up, but that isnt a
very scalable solution.
Our first test was using the exact configuration we flew
with helium pressurized lox / methane, but allowing the tank pressures to come
up by themselves with temperature, instead of adding helium. The computer individually relieves pressure
in the tanks as necessary to let them both arrive at the target pressure. With the same injector that we used for the
other flight tests, the engine made less than half the chamber pressure at full
throttle, which was not enough to lift off.
The propellant density was less than 25% different at that pressure, so
there was clearly two phase flow in the injector elements, reducing the total
We made another injector with significantly bigger holes and
got the vehicle up in the air for a few flights, but Isp was miserable.
We made another injector with more holes of the smaller size, and it
improved somewhat, but it was still worse than the helium pressurized one. I had been hoping that the self-atomizing
nature of the propellants would make things better, but that seems to not be
the case. We are experimenting with
other unlike-impinging designs now to try and get the performance back up. The self-pressurized propellants will
hopefully not have the same combustion stability problems we had with the
Once we knew that this was basically working, we stripped
off all the insulation on the methane module so it would self pressurize faster.
The lox goes up in pressure faster than the methane, even though it has
almost 3x the mass, since the difference in specific heat and boiling
temperature more than make up for it. We
load the methane first, but the lox still winds up getting up to pressure and
venting first. It takes about 40 minutes
for the preopellants to come up to 200 psi in our
current configuration. A lot of steps on
the checklist went away without helium pressurization, but we now have a 40
minute hold between loading the propellant and firing. The bulk propellant temperature does not rise
evenly -- when the tanks first reach 200 psi we lift the vehicle up in the air
with the crane and shake it around a bit to mix things up, which usually drops
the pressure back down to 150 psi. It
takes another ten minutes to get back up to a more uniform 200 psi. I want to try letting some dewars get up to 250 or 300 psi
for a direct feed-in to a 200 psi controlled tank relief, which would be
immediately usable and consistent, at the expense of wasting propellants in boiloff.
We somewhat inadvertently tested one of the major benefits
of vapak propulsion -- when we were struggling to get
enough thrust for liftoff we started short loading the lox, and we made one
flight that went to liquid lox depletion.
The thrust dropped a lot, so the vehicle started to descend from its
hover even after throttling up to max, but it continued burning smoothly as it
transitioned to burning on the cold gox ullage. It may be
possible to set up a VTVL so that it is normally landing at a fairly deep
throttle on liquid propellants, but can still maintain a constant descent for a
little while by going to full throttle if it happens to transition to gas
flow. You want to make sure that your
run out of liquid lox before liquid methane, both because it is a lot heavier,
and because going to lox / gch4 would almost certainly fry a film cooled engine
in short order.
We have been able to light the engines with just a spark
plug instead of a torch igniter. We had
bangs doing it with the main propellant valves on the unlike impinging
injectors, but using the manifold gas purges has been reliable in most
cases. We hope to be able to completely
do away with solenoid valves after getting back to a well-mixed unlike
impinging injector, but we have decided to stick with torch igniters for the
Starting up and shutting down without manifold
purges has worked fine.
Opening both propellant valves identically has worked
fine. With lox / alcohol, we always
pre-chilled the lox manifold by briefly opening and closing the lox valve,
which prevented us from physically linking the valves together (without adding
a dedicated purge valve). Especially
after our recent experiences with valves getting out of sync, physical linkage
is looking a lot more attractive.
We are evaluating using larger 1 V-cut ball valves from
AVCO instead of our current ¾ reduced port ball valves. This should give us a little more flow at
full throttle, but more importantly it should allow a lot more precision at
very low throttles. We have some hope
that we will be able to run the engines at a deep enough throttle, almost a
pilot light, that we could do full 100km suborbital flights without having to
shut the engines off and relight them.
This would be a Very Good Thing, although it would make the vehicle
unsuitable for true microgravity work.
We are building a linked set of main propellant valves,
using one of the Ultramotion Bug linear actuators to
drive them. The KZCO rotary actuators we
have been using for years dont have the torque to turn two 1 valves at the
speeds we want, but the Bugs have enough power to handle even larger valves if
we ever need them. Having independent
valves gave us a few advantages in the past:
it was easier to plumb things up, we could open the main lox valve to
chill the lox manifold without opening the fuel valve, and we could vary the
mixture ratio dynamically, although those tables have been identity transforms
for all of our actual tests. The issues
we had last year with one actuator moving and not the other, especially the
resulting burn-through on Pixels engine, are making me think that the
advantages arent worth it, even though we have since resolved those
problems. For lox-alcohol, we would need
to add a dedicated chill valve, but for lox-methane, we could just use it
This first-cut test has some strength and play issues, but
the second version has a longer isogrid plat that
also serves as a mount for our spark box and pressure transducer.
Another possibility that linked valves enable is the ability
to fly a differentially throttled vehicle with no required position sensing
feedback on the actuators. The control
system currently uses vehicle position / orientation, rates, and actuator
position to determine a desired actuator position that the motor is driven
towards. This works fine, but we have
seen problems with the position feedback on both the KZCO and Ultramotion actuators.
It has never caused us an in-flight problem, but the danger is
there. If we could use angular
acceleration as one of the inputs, the output value could just be a motor
current, and you wouldnt need to really care about the actual position. You would certainly log it for analysis, but
it would be like chamber pressure, not flight critical. With independent valves, this would lead to
drastic mixture ratio skews almost immediately, but linked together it would
probably work fine. You would still have
to worry about the possibility of completely shutting off an engine, but there
are strategies for that.
The problem is that when I tried differentiating the rate
gyro signals on a test flight, the resulting rate acceleration signals just
looked like noise. I decided to try some
direct-reading angular accelerometers form http://www.cfxtech.com/ to see if
the signal looked clean enough to use. It was interesting playing around with
these. They are rated for 25 rads / sec^2, or 1400 degrees / sec^2, and if you roll one
of them between your fingers, they are constantly going full scale as you move
them around, because they are so easy to accelerate at very high rates. However, when bolted onto a heavy box, the
values you get when the box is swung around stay a lot lower. On the vehicle in flight, they were even
lower. One ton vehicles dont change
rotation rates all that quickly (our cartwheeling module
What we discovered was that the wire rope isolators that we
mount the electronics box with shake around at something like 50hz. We got a much
cleaner signal when we hard mounted the angular accelerometers directly to the
vehicle frame, with no isolation. When I
went back and looked carefully at some of our previous flight data, I can
nicely see that the gyros are tracking the box shaking, which was why the derived
angular rates were so wacky.
Unfortunately, hard mounting the IMU or the entire electronics box isnt
an option, because the accelerometers pick up too much vibration, and that
kills our hover control.
We did a flight test with the hard-mounted angular
accelerometers, and the signal does look good enough to consider for control
use. Interestingly, the roll axis was
quite a bit noisier than the other two axis,
presumably because the micromachined sensing element
was in line with the engine thrust. Probably the really right thing to do would be
to hard mount a set of fiber optic gyros directly to the vehicle for accurate
body rate and derived angular acceleration, and have a separate comfortably
isolated set of accelerometers. I have
been meaning to try http://www.fizoptika.ru/
gyros, which is what Crossbow uses internally, but they didnt return my last
inquiry, and dealing with a Russian company directly may have some
We are building up five new electronics box lids. At this point, we dont have a single
virgin electronics box they have all been through a vehicle crash at some
point. They still work, but we dont
want to rely on them much longer. We are
only making modest changes in the systems, like bringing out three more analog
inputs, adding three dedicated digital inputs for panel switches, and using
different relays for the watchdog cutoffs.
We are going to build two complete brand new boxes, but we will hold off
on assembling the other three until we need them, since there is almost $20k of
parts in each box (dominated by the $11.5k Crossbow FOG IMU).
For years, I have had some enthusiasts extolling the virtues
of rapid prototyping to me. It sounds
great build anything you want directly from CAD drawings in only a few
days! However, the reality isnt quite
up to the promise yet. A few years ago I
tried getting a regen cooled chamber fabricated with
SLS, but all the vendors returned a will not quote on it. Recently, we did get pretty good results on
some cast chambers from http://www.proivc.com/ , but the surface finish is
still rather poor, and the thin wall structure we had made showed a bit of ovaling. In
addition, two other parts I tried to have fabricated there werent feasible for
In the last month, we have been working with
http://dyna-tool.com on some pintle designs
fabricated with a new direct-laser metal system from
http://www.3dsystems.com/products/sls/sinterstation_pro_slm/index.asp. Until recently, most metal rapid prototyping
work was selective laser sintering, which didnt produce a completely solid
part. You could bronze infiltrate it as
a secondary operation, but that is a poor substitute for stainless steel in
most cases. The latest machines can
directly produce non-porous parts in stainless steel, aluminum, and other
Dyna Tool is still learning all
the tricks of the process, which is complicated by the fact that the original
developers are in Germany,
and the initial parts have been somewhat flawed. There is still a lot of promise here, and we
are going to stick with it for a while.
I do think that we will eventually get to the point where we send a CAD
file out Monday morning, and we get the part that we want back on Friday for
testing n Saturday. Build time and expense
is fairly proportional to part mass in a given material, so it probably isnt
going to be making major structural pieces any time soon, but I can see a lot
of detail pieces that would be convenient to make small runs of. Several of us are practicing up on solid
modeling tools now.
We will be taking one of the methane engines to White Sands
Test Facility to test nozzle extensions in a vacuum chamber soon. We have both a traditional bell extension,
and a dual bell to test.
http://www.richardinspace.com/ stopped by for a visit last weekend, which was
fun. The early Ultima
games were significant inspirations for my game development career, but despite
being in the same industry for over fifteen years, we rarely bump into each
other. His first hand experience with
both the Russian and American space programs in valuable, and I learned a few
new things just from chatting with him.
I hope we can pick his brain more in the future, and he likes the idea
of skydiving off one of our vehicles in the future. He already has his own space suit
Now that the new FAA amateur rocket regulations are in
effect, we expect to do some higher speed, untethered
free flights before the next update. We
should be able to work our way up to nearly 8000 essentially in our back yard,
which is going to be very, very convenient.