Headlines > News > Armadillo Aerospace: Pixel flies again, Modular design, Tank welding, Insurance

Armadillo Aerospace: Pixel flies again, Modular design, Tank welding, Insurance

Published by Sigurd De Keyser on Tue Feb 6, 2007 4:10 pm
More share options

Pixel flies again

We made three good long flights with Pixel this month. There are three significant changes since the X-Prize Cup: new landing gear, regeneratively cooled engine, and new gimbal actuators.

The landing gear is plenty strong now, but we have had them fall off a couple times when we lift the vehicle up after a landing loads them enough to upset the set screws holding them on. We will drill explicit holes for the set screws so they act as pins even if they loosen. The gear on the lox tanks is also fairly sticky when cold, only very slowly extending after lifting away from the ground. We had put some krytox lubricant on the sides, but we may be better off leaving it just metal-to-metal. We are considering adding some desiccant inside the gear to reduce the chance of ice forming.

The first set of tests on the cooled engine resulted in the graphite chamber cracking when it got hot. We had carefully measured everything to give us some room for thermal expansion, but the welding of the bottom closure to the tube had caused enough warping that we were still torquing the injector down directly onto the graphite chamber, instead of only to the outer aluminum flange. We machined 0.030” off the end of the next chamber for clearance, and it is still pristine after a 100 second and a 70 second flight. We are strongly considering going to a retaining ring based assembly to avoid the weld warping and save the weight of the flanges.

The o-ring between the graphite chamber and the aluminum injector is causing some trouble. We have frozen it with lox pre-chill or purge, and we are getting four spots of burning on it between the spokes in the injector pattern. We are going to try a helicoflex metal seal from http://www.helicoflex.com/assets/pdfs/helicoflex.pdf.

I am still a bit concerned about heat soak after shutdown, since we have 25 pounds of fairly hot graphite. The o-rings would be the first thing to go, but it isn’t out of the question that some of the aluminum might get hot enough to lose temper. I am letting the entire fuel tank ullage bleed down to 50 psi through the engine to provide some cooling after shutdown, but we have the option of dumping some lox on it for additional cooling if necessary. I always had fears of the lox igniting the graphite like a hybrid grain, but it has never been a problem.

We changed our purge plumbing so that it only blows the injector manifold dry, not the feed pipe and cooling jacket. This avoids dumping a fairly significant quantity of flaming alcohol out of the engine on shutdown, and has the nice benefit of leaving the alcohol in the cooling jacket to absorb heat after shutdown. You can hear it boiling after the purge has shut off, which is a pretty good indicator that the temperature isn’t too terribly high, or it would have been gone a lot sooner. We probably have plenty of margin with the current injector, but I fear that if I removed the film cooling to increase Isp, the chamber would be a whole lot hotter, and we might have a problem.

We have changed our gimbal linear actuators to “bugs” from Ultramotion: http://www.ultramotion.com/products/bug.php.

The Electrak units that we had been using were cheap, but we had to swap the motors for custom units to avoid the thermal cutoffs, and we had to make internal modifications to reduce the backlash. They were probably still fine for our old, small engines and vehicles, but I have been pretty uncomfortable with them on the big motors. One of our flights had a data trace that showed a temporary drop in speed on one of the actuators that looked like it was nearly stalled, and one of the actuators had broken internally after the last fall-over-on-side flight.

I had originally selected the geared motor, but Ultramotion pointed out that the non-geared motor, combined with a different lead screw pitch and pulley, provided the same speed at greater force and efficiency, as well as weighing slightly less. The pulley drive arrangement is quite convenient, we swapped pulleys for a faster speed after we did our initial testing.

The units cost quite a bit more than the Electraks, at just under $1000 each, but we are now faster and stronger than before, and we have some growth room if we need to add more speed or power in the future. The hardware design is a lot higher quality, and the extra mounting flexibility is nice. Ultramotion is more than happy to work on rocket projects, and gives fast turnaround time.

We did have some issues with the pot feedback. The feedback is from a linear pot, which is generally a good thing, especially if you have looked at the flimsy geared arrangement on the Electraks, but the resistance went slightly out of our range. We have 12V excitation for all sensors, but we can only read 10V on the A/D channels. We have always been ok on our previous sensors, but we had to add a resistor to get the Bugs in range. More seriously, one of the units we got had a short-to-case at one point in the travel on the pot feedback. Ultramotion sent us a new one next-day delivery, but we also found another issue: when the motor is driving the screw into the bump-stop one direction, the pot seems to “fall off the end” into undefined space. The range is valid over the entire normal travel, but compressing the bump stop must push it over the edge. The other one didn’t do that, so it is probably a result of having the stop set a few thousandths too far.

An unexpected result of the high efficiency ball screws is that the engine assembly won’t hold its position without active control, it slowly slides back to a position slightly off center due to the tension in the feed hoses. We had to make a minor change in the way we calibrate the actuators due to this.

In addition to the vehicle changes, we are continuing to improve the ground support hardware and procedures. The pressurization hoses have been modified with some different vent valves and check valves, allowing everything to be done by one pad crew member, instead of requiring two people to open two tank valves at the same time.

This ground liftoff flight was intentionally cut short to 60 seconds, because we still didn’t know if the regen engine was holding together properly. Cracks in the chamber just result in higher fuel consumption as fuel leaks into the chamber without going through the injector, so I would hate to have the vehicle drop to the ground at 87 seconds because it ran out of fuel. It turned out not to be a problem, and the residuals were as good as ever. I landed the vehicle off to the side intentionally, so it wouldn’t be on the concrete that had been cooking for so long. I am investigating the oscillations that picked up later in the flight, it could be either the faster gimbals, the new roll thruster locations, propellant slosh, or some pull on the tether bungees. It settled down before landing, which is odd, and perhaps points to slosh.

Many of you are probably sick of quad flight videos, but this one has the on-board camera pointing at the engine, which is an interesting view. It also shows we can land without breaking now…


We have a number of things to be testing and improving with tethered flights this month, but we hope to do untethered flights in Oklahoma next month. I’m happy to hear that a number of new teams are springing up looking to compete in the lunar lander challenge (I know of at least six prospects, outside of the 2006 entrants), but I think it is going to be a long shot for anyone to beat us to first place. The once-yearly nature of the competition makes for a much livelier competition, but I will bitch and moan about it when we do fully qualifying test flights without being able to collect the prize. I do encourage everyone involved to consider keeping a blog of their progress. Even if you aren’t comfortable doing it “live” where your competitors can see, it would be cool to read a history start-to-finish at some later point.

Modular design

We are probably going to wind up with a fuel feed pipe going through the lox tank in the new design. There are certainly concerns with that, but we can arrange to have a completely seamless tube through the tank, with no weld joints separating the lox and fuel. This arrangement gives us a very convenient arrangement of all our gear under the tanks, and should allow us to change to a different tank technology (filament wound or flowformed) with very little pain.

In theory, the strains on a through-pipe welded at both ends would be the same as anywhere on the sphere, but we fabricated an aluminum bellows to allow the tank to grow during pressurization without straining the pipe. Bending aluminum like this is certainly not good from a fatigue standpoint, but the total number of pressure cycles isn’t going to be all that high, and a failure would only be about like a big burst disk letting go.

Our first design had an 8” diameter flat bulkhead on the bottom, allowing us to mount the engine u-joint directly to the tank center, but it bowed so much that the bellows bottomed out and the inner tube broke during hydrotest at 790 psi. We have since redesigned to allow the through-pipe to go down the exact center, and mount the engine below the propellant manifolds.


We tested several different pipe configurations for the impact on ethanol in a cryo temperature tank. We filled a Styrofoam cooler with liquid nitrogen, and made “test tubes” in different configurations filled with ethanol (90%). The bare pipe obviously froze first, followed by the 1/8” thick polyethylene liner, followed by the 1/8” thick phenolic liner, followed by the ¼” thick pneolic liner. By far the best was a 1/8” air gap held by a couple o-rings, separating a tube-within-a-tube. It took about 40 minutes to really freeze it, and at 20 minutes it was still completely liquid (although < -100F !). We would be having worries about low lox densities before the fuel froze.

This was a very interesting test, because alcohol doesn’t freeze like water does. It just gets more and more gelatinous, rather than clearly crystallizing. When it got very cold it took on a honey like consistency, and just got thicker and thicker. When we removed the tubes from the liquid nitrogen and let their outside warm slightly, we could dump an “alcohol slug” out of the tube. Also of interest, the solid slug was pretty much invisible in a bucket with alcohol in it, not showing the clear refraction difference that water does. This test leads me to doubt the viability of subcooled propane / lox, a combination sometimes considered for launch vehicles because of the better density. While it may technically still be a liquid at lox temperatures, it probably isn’t very fluid, and an uninsulated wall common with a lox tank would likely have quite a bit of residual propellant clinging to it, especially if it was a completely submerged tank.


We are going into production on five individual modules (two tanks + one engine) for our first run. We intend to fly each module as an independent unit, and we will probably fly a two module system in Titan II differentially gimbaled form, and a four module system in differentially throttled form with the gimbal actuators replaced with tie rods. We may also fly a four module system with four gimbaled engines, controlled by two independent electronics boxes for full system redundancy. We want one module as a spare, in case we mess one up.

This means that we are producing parts in lots of five through forty parts each, depending on how many go on each module. This is a great design exercise for us, and motivates us to do everything CNC, instead of just cobbling things together with a band saw / drill press / welder. Russ is working on the computer mounts, and James is starting to learn CNC programming so all the load isn’t on me for part programming.

We went through several different designs before settling on a simple bracket arrangement at the tank equators for holding the modules together and acting as leg attach points.


Tank welding

We hydrotested our first 5383 (Sealium) tank to failure, with good results. The tank was 36” ID, ¼” prespun thickness, like most of our previous tanks. The source blank to spin these hemispheres is just under 48” diameter, so there is a fair amount of thinning in some places as the 1800 square inch blank deforms into a 2034 square inch hemisphere. The spheres are 81 pounds dry and 950 pounds full of water, for a mass ratio of 11.7.

The burst was at 860 psi, significantly better than our previous results. There is likely still some room for improvement, as the initial burst point was at one of the spots where James had a hole on the root pass, which meant that it had a broader area with nothing but filler material when it was closed back up. We are going to consider moving to either 5083 or 5556 welding wire (from the current 5356) to increase the strength of the raw weld that doesn’t alloy any of the base metal in.


I have been working on a document with all the various trades in aluminum tankage, but the bottom line is that for tanks used in the as-welded condition, 5059 (Alustar) is the clear winner, and we will probably be moving to that in the future, although there is a fairly large minimum order. 7039 or 7005 can deliver similar strength, and possibly a bit more with post-weld aging (somewhat countered by the higher density), but the welding is more challenging, and the alloys need to be painted / coated for corrosion resistance. After that, 5383/5083/5086, in that order, are the best bets. 2219 looks best if you can solution heat treat the entire assembly after welding, or if you can provide thickened weld lands. If you can friction stir weld, 2195 lithium-aluminum wins big. I do not think there are any prospects of us FSW our tanks.

Several other teams are also working with hemispheres from www.amsind.com. They are great to work with, and the prices are good, but the spinning is not CNC, so there is a fair amount of variability to deal with.

Paul Breed reports at http://unreasonablerocket.blogspot.com/2007/01/normal-metal-tanks.html on a smaller and thinner stainless steel sphere. I suspect that the welding was the weak point of the test, and with practice they could do better, but with just that data point, the aluminum tanks look significantly better.

Kevin Sagis reported in private e-mail:

John, tested two 20″ 5086 tanks today. Material was 3/16″ prespun thickness.

One failed at 925 psi, other about 850 psi. The 850 failure appeared to fail in weld, the 925 failure definitely failed in HAZ.

The one that failed at 925 had an equatorial thickness of only 0.125″ in about 6″ of circumference (the other tank had a minimum thickness of 0.150″) so we clearly did a better job on it.

Our tanks have done somewhat better than that by direct pressure/volume comparison, so he is corresponding with James on exact welding procedures.

XCOR has also bought some smaller hemispheres from them for tank ends, but I don’t know how they turned out.

Here is James’ welding document with lots of details:


We have been operating with a fairly standard business liability insurance policy for our shop (which was still quite a challenge to get), but for us to perform flights under an experimental permit or launch license, we need to have explicit insurance for third party damages, and one of the sites that we are considering for larger static / tethered tests is also requiring us to have additional insurance. The X-Prize Cup covered the insurance at the event last year, but they weren’t required to, so we may need to cover our own operations there this year as well.

Phil did all the work on this, here is his update:

Procuring insurance for Armadillo operations has not been as horrible as it could have been. A little research led us to Ralph Harp of Falcon Insurance in Houston.

Ralph has been doing “Odd-ball” insurance as he puts it, for 20 years. His specialty is Aviation, balloons, Smoke jumpers, and other various flying things or activities. He likes Aerospace and Rockets, and has insured several sounding rocket launches.

Ralph and his underwriter Chris Barnett with Houston Casualty Company are excited to support the budding industry of commercial space access. They understand the industry, speak the language, and are willing to accept a certain amount of risk while providing a reduced cost premium for the AST required coverage for 3rd party liability as well as operational liability for every day testing.

Through the process of negotiating the rate and assessing the risk, the more information we provided the more they reduced the premium. The video and description of past failures helped them quantify their actual risk based on history. Something that is seriously lacking for them use as justification for reducing premiums.

We have finally ironed out a deal to cover an unlimited number of free flight tests under AST issued Launch Permit or Launch License, as well as coverage for static testing and tethered flights for a 12 month period regardless of the location. We are not able to disclose the exact premium amount, but I can say it is under $50K and over $40K. Your mileage may vary depending on test locations, total liability, and how much information you are willing to share with the underwriter.

The coverage amount is $1M for static and tethered vehicle testing and $3M for free-flight testing under Permit or License. This represented the minimum amount of coverage AST would accept for the Permitted flights at the OKSP location.

If we are to apply this to another location we will have to mitigate risk to the same level, which just so happens to be a single 3rd party casualty. This is the minimum level of coverage AST will allow at this point, and they appear to be unwilling to move on this for now.

The insurance agent and the underwriter request that we keep them in the loop and treat them like a member of the team. The better we build this relationship moving forward, the easier it will be to increase the limits of liability for much larger endeavors such as orbital flights.

Ralph said he would be happy to speak with anyone regarding coverage of their test programs. They can provide single event coverage as well as yearly policies.

Ralph B. Harp
Falcon Insurance Agency of Houston, Inc.
Ph: (281)540-8822 or 1-800-880-8822
Fx: (281)540-8222

Lessons learned:
If you are developing a product that is intended to fly, you are in the aviation industry. That makes your insurance more expensive!

If you want to go over seas to Lloyds, they won’t even talk to you unless the premium amount is over $50K. They wanted upwards to as much
as $100K or more for what we needed.

If there are only 1 or 2 people willing to accept the risk, the policy will be expensive.

Don’t forget that the guy doing the underwriting has someone looking over his shoulder asking him to justify the numbers with data, and we don’t have much by comparison to other aviation endeavors.

Sharing your failures and lessons learned is a good thing! They figure if you don’t share those things you must have something to hide, and the premium goes up very fast.

AST’s Graphical Risk Matrix is a helpful way to quantify risk level of certain elements of your flight systems. It’s easy for the underwriter
to interpret that kind of data.

Again, they like information overload. It helps them place odds on weather you “can” fail in a fashion that causes 3rd party damage.

No comments
Start the ball rolling by posting a comment on this article!
Leave a reply
You must be logged in to post a comment.
© 2018 The International Space Fellowship, developed by Gabitasoft Interactive. All Rights Reserved.  Privacy Policy | Terms of Use