December 4, 2006 notes:
Matt put together some high-res
photos of Pixel suitable for desktop wallpaper:
I did an interview with c-net after the X-Prize Cup that
covered a lot of ground:
The most obvious thing that we had to improve post X-Prize
Cup was the landing gear. I spent most
of the flight back to Dallas
sketching different landing gear concepts and assessing the tradeoffs. There was a strong temptation to make the
landing gear into sealed pneumatic pistons that use the tank ullage as the reservoir.
A decent sized check valve with a tiny back bleed for each leg would
give a nice linear landing force with no spring return, and with a blowdown system like ours, the shock force would be
somewhat related to the weight of the vehicle, with a 400 psi
liftoff pressure and a 150 psi landing pressure.
The downside would be adding another potential leak path
(with a sliding o-ring seal), which could easily result in a loss of vehicle
failure. With our paired tanks, if ullage gas started leaking out of one tank, propellant
would be pushed over from the other side, and it could quickly exceed the ability
of the gimbal to balance the vehicle. This is the primary weakness of the current
quad architecture, and it didnât seem wise to tempt fate. I considered having an extra tank just for the
shocks, but that adds a fair amount of complexity, and it wouldnât get the blowdown variability benefit.
In the end, we decided to just make extremely sturdy
side-load adapters for a 2â stroke commercial hydraulic shock. The shock also needed to be insulated to
prevent the lox tank from freezing the hydraulic fluid. The sliding piston was made out of bearing
brass, the insulators from garolite, and the fixed
part from aluminum to allow us to weld it to the tanks if we choose. Currently they are just held on with set
screws, and but up against the old shock mount point.
We tested a lot of landing combinations by hoisting Pixel up
in the shop with various water loads to various heights, and optionally tilting
or swinging the vehicle before pulling the Sea-Catch release to drop it. We did drops from over two feet high (3.5 m/s
impact speed) and everything seems rock solid.
Phil sat on top of the vehicle for some tests, and there is a very large
difference in the landing loads. An
empty vehicle dropped squarely on all four shocks lands pretty much like it hit
the concrete without any shocks, but a fully loaded vehicle that lands on one
shock gets a pretty gentle set down. A
computer adjustable gas shock or some kind of sensing shock would be an
improvement, but at our targeted descent rates (2m/s) this will work out fine.
We have replaced the big pin-and-block main thrust u-joint
with a UJNL 20-20 (forged, 1 1/4" bore) from Boston Gear. This is lighter, shorter, and has zero play. http://staging.smartcats.com/bost_root/web/pdf/product_sections/bearing_pp_127_133.pdf
James has been working on mounting a seat on Pixel. We quickly decided that a standard racing-car
seat was not the way to go. After joking
about using a saddle, we settled on mounting a motorcycle seat on a framework
above the computer. The rider will
probably be leaning over onto the seat and gripping some hand holds, rather
than sitting upright.
I suspect we will have our next-generation vehicle(s) flying
before any important manned flight opportunities arise, but we will probably
let Russ ride Pixel under a tether just to see what the experience is like.
While the carbon reinforced graphite chambers are getting
the job done, cracks in the chambers and resulting leakage have been continuous
concerns. The cracking is almost
certainly due to thermal expansion as the 20â long chamber gets really hot and
is restrained by tie rods that donât get very hot. We started using spring washers, but it was
more of a token gesture as it would take an almost absurd number stacked up to
hold the tension we want and still have room to compress enough to absorb all
Our previous plan was to design a chamber that had an
integral flange on it, so it was clamped at the top, leaving the hot part to
essentially dangle unrestrained below the injector. Unfortunately, this wasnât as straightforward
from a manufacturing standpoint as we had hoped. The team at Cesaroni
Aerospace was worried about a lot of possible failure modes, and we dithered
around for a while without a clearly winning design appearing.
We decided to go ahead and try putting a cooling jacket
around a simple graphite chamber. The
cooling doesnât really need to be very good, because the graphite is happy
operating at extremely high temperatures, and we donât have to worry about
making a saddle section around the throat.
Having fuel around the outside also puts the graphite in compression at
all times, so it shouldnât need any additional reinforcement.
We got some 8â ID aluminum pneumatic tubes for this from: http://www.scotindustries.com/
We had Cesaroni make us a few bare
graphite chambers similar to the reinforced chambers we had been using. They just showed up today, but we havenât
done any fit-up yet. We went with a
shallower converging angle in the nozzle (15 degrees instead of 30), since the
only place we were seeing any erosion at all in the previous engines was on the
converging section, and we believe that the shallower angle will help the film
cooling stay attached better. Lutz
Kaiser had also mentioned that the OTRAG chambers used shallow entry angles
because it reduces organ-pipe combustion stability problems. We havenât had any stability problems, but it
still doesnât seem like a bad idea.
I made all the various manifolds and flanges in the last
couple weeks, and we should be doing water tests soon and hopefully trying to
fly Pixel on the new engine next weekend.
It turns out that the jacket, flanges, and manifolds are almost exactly
the same weight as the tie rods, phenolic insulators,
and retaining plate on the old engine.
We should be able to reduce the film cooling and get a little bit better
Isp, but the main thing we
are looking for here is increased durability.
Our current injector pattern has both good face cooling and
good performance, but we had to modify the design to allow all the fuel to come
in from the sides, and we also wanted to avoid the âburied weldâ in the current
design, where a single internal weld separates part of the fuel and oxidizer
manifolds. This is a classic point of
concern, and we did have one engine that popped the op off on first ignition
that could have been due to a leak there.
The solution was to build the lox manifold as an interlocking
piece that can be welded in from the chamber side of the injector. This involves more machining steps, and the
welding to the injector face must be done halfway through all the operations,
rather than just at the end, but it does resolve the dangerous leak path. Unfortunately, I still have some concern
about those welds cracking, which, while not dangerous, would leak unmetered fuel into the chamber. I initially milled the surface flat before
drilling the injector elements, but we found that the welds were very weak and
prone to cracking. We decided to just
leave the welds at their full height and drill the elements through them (with
spot facing first, of course), but I worry that the spot facing has exposed
some areas below the weld penetration.
We should be doing water tests soon to find out.
We are seriously considering trying out methane as a fuel
instead of ethanol. Methane mixture
ratio by volume is close to 1:1, just like ethanol, so we would get to keep the
identical tank sizes, and we are working on converting the VDR chassis over to
a methane testbed.
Just on raw performance for a ground launched vehicle (for upper stages
it is a pretty clear win), Methane only barely outperforms alcohol because of
the low density, but our reasoning is based on operability issues, not delta-V.
We are considering using lox / methane in a self-pressurized
system, where, instead of using helium to pressurize sub-cooled propellants, we
allow the propellants to reach their boiling point and provide their own
pressurization gas. Although almost all
nitrous oxide hybrids, including SpaceShip One, use
self-pressurizing propellants, it is not common for biprops. The AirLaunch
company with their QuickReach rocket program are
using self pressurizing lox and propane (which must be heated to reach the
desired pressure), but I donât think any significant vehicles have yet flown
with this arrangement. AirLaunch and some old papers term this âVaPakâ for Vapor Pressurization. Gary Hudson was kind enough to answer some
questions for me about their experience, and it does seem worth pursuing.
Getting rid of helium as a consumable would be a big win for
us. The helium costs more than the other
propellants, but it is also bulkier to transport, and we have lost a couple
crucial testing days due to running out of helium. If we can just travel with a big lox tank and
a big methane tank, field operations will be a lot easier. We would only need to connect two hoses to
the vehicle for the filling process, and we could top up the vehicles repeatedly
without venting if desired.
Using methane may also allow us to simplify a few other
systems. We currently have purge ports
so that when the engine is shut off, helium is blown in right after the
throttle valves to force any remaining liquid out of the plumbing and into the
engine. With self-pressurizing liquids,
the fluid in the plumbing should vaporize almost immediately, and an
atmospheric pressure gas mixture shouldnât be much to worry about. If this does still turn out to be a problem,
it will be a pretty big drawback, forcing a gas bottle just for purging. We also think the augmented spark torch
igniter can likely be removed, and replaced with just a couple spark
plugs. Cracking the valves will result
in nothing but gas coming out of the injectors, and I think a spark will light
that without any problems.
Because the propellants boil as liquid is removed, the tank
pressure does not drop as much as you would expect. We did some tests with liquid nitrogen at 225
psi saturated pressure to confirm tank filling
procedures and pressure drops, and the pressure only dropped by a third, from
225 psi to 150 psi, going
from a completely full tank to a completely empty tank. The expelled liquid is also getting colder
and denser as the tank drains, so the actual mass flow is dropping even a bit less
Cooling a chamber with saturated methane will be more
challenging than with a subcooled liquid, but I think
the graphite chambers will work fine even if they get nothing but film boiling
along their surface.
There are two primary performance disadvantages with
saturated propellants: The higher
temperature / pressure propellants are less dense than in their normal
sub-cooled form, and the gas remaining in the tanks at liquid expulsion is very
heavy compared to helium. You can get
graphs of pressure/temperature/density at http://webbook.nist.gov/. At 20 bar,
completely full 36" ID spherical tanks (14.13 cubic feet) would hold 777
pounds of lox and 284 pounds of methane.
At liquid depletion, the gas remaining in the tank at 15 bar is 51 pounds of oxygen and 21 pounds of methane. In expendable booster applications it should
be possible to burn the gas remaining in the tanks all the way down to vacuum,
but the transition point when the first propellant goes to gas is scary from a combustion
flame-out standpoint, and the immediate drop in thrust by a large factor may
make powered vertical landing near that point dangerous.
I had been concerned about free-venting methane, but an LNG
consultant that we spoke with really didnât think it was much of a problem,
especially if we are shooting it up with a couple hundred psi behind it and a nozzle on the end.
We are still tracking down all the things we need to
actually do engine firings with methane.
Apparently LNG (commercial methane, âliquid natural gasâ) dewars are built to a different
spec than LOX / LIN dewars and they arenât very
popular around here. We are probably
going to have to pay to get one made just for us, especially if we want to use
it at 300 psi or so. We need to decide which connectors we are
going to use for filling, since apparently there are three different âstandardsâ.
We need to develop a little resistor
based sensor that can sense cryo liquid level before
the computer controlled vent valve, which will allow us to have both automation
and a good telemetry log of the filling process.
I have started working on a coaxial injector for saturated
lox / gaseous (from the cooling jacket) methane:
We will probably wind up with both internal and external
tapers on the posts for better mixing. The
manufacturing aspects of this are so nice (no welding or angle table work) that
we may try making a coaxial alcohol injector, even though that is usually not
considered a high performance liquid-liquid injector.
We have gone through a few design iterations for our
next-generation, modular vehicle system already, and fabrication will be
starting next month.
The biggest decision is the propellant tanks.
As a baseline, our current 36â ID welded spheres cost about
$2000 each if you include labor costs.
They hold 100 gallons, weight 90 pounds, and burst at around 750 psi.
On the high end, Microcosm gave me a rough quote for 25â diameter
x 59â long linerless carbon fiber lox tanks at $13000
in quantity for a 100 gallon, 55 pounds, 1200 psi
burst tank. For an upper stage, these
would pay for themselves, but probably not in a booster.
I spent a while investigating flowformed
tanks, similar to those used in the OTRAG project. http://www.flowform.com/
has some good information on the process.
Most of the products that are made with flowforming
use two different sets of expensive tooling: a forging tool to create the pre-forms,
and the actual flowforming mandrel that the pre-forms
are spun over. Going from scratch would
be well into six figures, so we tried to find a combination of existing tooling
that could meet our needs. We found that
we could start with 14â sch 40 seamless 304 SS pipe
and machine it to a precision pre-form shape to avoid the need for forging, and
they had an existing 13.305â diameter mandrel that could create 168â long
tubes. The process would allow the ends
to be left thick, so aluminum bulkheads could be held in with snap rings or
threads. The 304 SS would reach about
170ksi UTS when flowformed, so the walls would only
be about 0.038â thick for a 1000 psi burst
pressure. This resulted in a 100 gallon
tank that would weigh about 85 pounds.
The price was $7000 each in quantity.
The mass ratio would only be a little better than the
current aluminum spheres, but for a vehicle with many modules it would have
much better aerodynamics due to the aspect ratio. Conversely, the aspect ratio would make testing
small, four module configurations as vertical landers
much harder, requiring broad spacing interconnects and/or extremely wide legs. Flowforming may yet
be a good technology when coupled with a more exotic material, like
aluminum-lithium or maraging steels, which would give
mass ratios competitive with the composite pressure vessels, but making the
forging tools for it and sourcing the materials would be significant up-front
The fiberglass pressure vessels from Structural http://www.structural.com/base_pages/composite.htm
that we used for most of our peroxide vehicles are so tempting for propellant
tanks, because they are dirt cheap, about $10 / gallon across the entire
product range from 60 gallons to 1600 gallons) and rather high performance,
with the double-flanged designs holding over 1000 psi
without bursting (the single flanged and screw on tanks will fail at 600 psi), at a weight of about 10 pounds per gallon. The liners are cross linked polyethylene. We have several of these in the shop, so we
finally just said âwhat the heck, lets see what happens if you fill them with
liquid nitrogenâ. First we tested some
small plastic containers to get a sense of what might happen. The first test was discouraging â an HDPE
quart jar cracked at the base just from having liquid nitrogen poured into
it. A PET water bottle didnât have any
problems holding LiN.
A piece of one of the tank liners we had around did seem to hold up fine
after being submersed in a cooler full of LiN, and
still seemed to have decent flexibility, so we thought it was worth testing the
full size tank.
We did the test with remote controls at our 100 acre site,
because if the vessel did rupture at high pressure, it was going to be a pretty
big kaboom due to the LiN
having soaked up at least some heat and being at a saturation level above
ambient pressure. It turned out to be
pretty anti-climatic, with the liner cracking at only 60 psi
and letting the LiN leak out. This was probably for the best, because it
would have been a bit sporty to put LOX in a PE lined tank anyway, and these
tanks do make the most distressing crackling sounds when you pressurize them
above their 150 psi rated pressure the first
time. Still, I wonder if they can rotomold the liners out of PETâ¦
The last option was to improve on our existing spun tanks. AMS industries http://www.amsind.com/ has been great to work with, and Iâm happy to hear that
XCOR and Paul Breed (follow his work at http://unreasonablerocket.blogspot.com/)
have also ordered tank ends from them based on our experience. We have gone over a lot of different options for
improving the mass ratio.
The 36â ID x Â¼â thick tanks have all been bursting in the
heat effected zone by the girth weld, so we could achieve better performance if
we started with a thicker plate, and machined it down everywhere except a
thicker band at the weld zone. It always
seemed to me that you should be able to do that while the part is still chucked
up on the spinning machine, but AMSâs machines donât
have the ability to do that. Spincraft http://www.spincraft.net/index.html
is the high-end aerospace tank spinner, and I bet they could do it, but they
never return my calls or emails. I
realized recently that we could probably pre-machine the spinning blank to
different thicknesses while flat, then spin it. AMS thought that this would probably work, but
getting the plate completely flat before machining might be challenging. We may give this a try at some point. A first step would be to just thicken the
weld band, but later optimizations could try to correct for the thinning that
happens at the point of maximum curvature as well. A 36â ID hemisphere is spun out of a 48â diameter
flat circle, so there is certainly some additional room for evening everything
Our current tanks are made out of 5083 aluminum, but we have
been considering other options. One of AMSâs supplier suggested that we look at 5383 aluminum, a
relatively new alloy. This goes under
the trade name âSealiumâ, because it is designed to
replace 5083 for aluminum boat building, and it does look like a very good
alloy for us. It is basically just like
5083 for forming and welding, but the welded joints are 15% stronger in the
HAZ. The alloy is available in the sizes
and thicknesses we need, so we went ahead and chose this alloy for our next
batch of 24 hemispheres. 5059 aluminum
is an even newer alloy aimed at the same market, with a >25% strength
increase across the board, but the availability isnât as good. The shipbuilding industry is actually a lot
closer to what we do than the traditional aerospace industry. http://www.aws.org/wj/feb04/anderson_feature.html
After a lot of searching, we did finally source 2219
aluminum (the material used for the space shuttle external tank before they
moved to an aluminum-lithium alloy), but the price was very high at $1800 for a
single sheet (that was the cheaper quote).
If we are willing to heat treat the entire tank after welding, 6013 may
also be a good alloy to consider.
Elon Musk put me in touch with their
contact at Alcan that is supplying the Al-Li metal
for Space-X, and I have a meeting set up with their rep this week. Apparently the minimum order is around 3000
pounds, and the lead time is six months, but that might not be out of the
question, depending on the price. They
are trying to find us some material to make a test article out of.
Availability is a bigger issue than most people think, and
even as we start building honest space vehicles, we are still more likely to
use a less-optimal material that we can actually get our hands on quickly and