March 25 and 29, April 1 and 5 Meeting Notes

 

Big Tank

 

The big tank was delivered this week.  It is an 850 gallon fiberglass tank from Structural North America, which will be our early  test model for the full size X-Prize vehicle tank.  The plan is that we can have a longer 1600 gallon tank fabricated with carbon fiber, which will have the same mass as the 850 gallon tank in fiberglass.  There is an outside chance that the 850 gallon tank may be the final vehicle tank – if we continue to have problems securing a high volume supplier of 90% peroxide, we may be forced to a bipropellant solution with a lower concentration peroxide oxidizer.  That would be more development work and technical risk for us, but it would make the entire vehicle smaller, and allow the use of the off-the-shelf tanks.

 

I finally got a quote from Alliant Techsystems, the big military conglomerate, for carbon filament winding the full size tank.  I emphasized that it doesn’t need to meet any mil-specs, it just needs to pass a proof test.  I was fearing a completely ridiculous price quote, but it didn’t turn out all that bad.  They don’t have liner tooling for anything in our tank size range, but if we can provide the liners, they could wind them for $15,600 per tank.  Great, I thought!  However, they want $40,000 in non-recurring engineering expenses before the first one is produced.  That put a bit of a damper on the enthusiasm, but if we make two or three tanks, it is still in the ballpark of what I would be willing to pay.  I have a couple more places to try before nailing down an order.

 

The hanging weight of the fiberglass tank was 920 pounds, but 44 pounds of that was in the manway flange, and another dozen pounds or so in our rigging, so the tank is pretty much right on our expectations – they are all about one pound per gallon of capacity.  Our cabin weighs 200 pounds at the moment, but that doesn’t include the hatch, electronics, and air bottle.  The crush cones are only 37 pounds.  The parachute canopies that we were quoted for (before the vendor got cold feet and backed out on us) had a weight of around 150 pounds for a fully redundant system.  The big unknown is large engine thrust to weight.  We are specifying 2400 pounds under the parachutes, which assumes three X-Prize standard passengers, so if we go overweight on our test vehicle, we will just delete some of the ballast for the phantom passengers.

 

Last week, one of the major goals was to finalize the seating arrangements for the other two passengers in the X-Prize cabin, because once we bond the cabin onto the big tank, working on it is going to be much more difficult.  We made a frame arrangement in the cabin with two by fours and plywood, and tested entry and exit for three people, but that was just on our boilerplate 2:1 eliptical tank end.  We weren’t sure if the filament wound tank end would have similar contours, so final fabrication was held off until we could sit it on the tank.  Happily, it still works out fine on the tank – the center dome with the flange goes higher, but the side slope away more, so it still works out.  We will be doing final cabin fabrication with aluminum and honeycomb boards in the coming weeks, then permanently bonding the cabin to the tank.

 

We also did some timings of propellant venting through 2” ball valves, which is data we need to determine the dangerous times in the flight profile where you may not be able to vent enough propellant for a good parachute landing.  A half full 60 gallon tank with 100 psi ullage pressure took five seconds to vent.  A ¾ full tank with 200 psi ullage pressure also took about five seconds to vent.  These numbers should scale linearly to the larger tanks, and by the square root of pressure, so the 850 gallon tank half full with 400 psi would take 35 seconds to vent completely.  We also tried putting one of our little cameras actually inside the tank to see if we could see a swirl pattern as the tank was emptying, but the rapid expansion cooled things so quickly that a dense fog was formed in the tank as it neared empty.

 

http://media.armadilloaerospace.com/2003_04_05/onCrate.jpg

http://media.armadilloaerospace.com/2003_04_05/withCabin.jpg

 

 

Alternate Propellants

 

We are resuming some work on alternate propellant combinations, because it is looking like we will run completely dry of 90% peroxide very soon now.  We expect to have more in two months, but that will leave us with enough of a dead spot that we will want to at least be trying some other options.  We are still focusing on lower concentration hydrogen peroxide as the oxidizer, although we haven’t completely ruled out a switch to a different oxidizer.  Various 50% peroxide / fuel combinations can still run with uncooled nozzles, and should be able to hit around 160s Isp at sea level.  70% peroxide / fuel combinations should get about 20 seconds more Isp, but would require a cooled nozzle / chamber.

 

Our last set of tests seemed to indicate that 50% peroxide doesn’t decompose very well at all through our silver screen catalyst packs, and we were unable to get combustion started with a fuel mix post-catalyst pack.  We had enough problems trying to use solid rocket propellant for an ignition source that upcoming experiments will use electrical means.

 

We tried using a finely atomized spray nozzle with a 50% mixed propellant to see if it could be made to decompose or light in an open air situation, but that didn’t work either.

 

The next step was to try 50% peroxide with separate injection of a catalyst fluid dissolved in a fuel.  Calcium permanganate solution was used as a catalyst in some of the German peroxide mechanisms, reportedly because it is much more soluble in water than other catalysts.  We tried to track down a supplier, but the two quotes we got were for several hundred dollars for TEN GRAM quantities!  No thanks.  We have manganese acetate and potassium permanganate on hand, which seems plenty reactive enough.

 

We know from previous tests that catalyst dissolved in fuel will autoignite with 90% peroxide, but not with anything lower, so we installed a glow plug between the two spray nozzles as an additional ignition source.  The open-air tests didn’t do anything exciting.  You could hear the peroxide mist decomposing in the air, but it evidently didn’t happen fast enough, close enough to the glow plug to get anything started.

 

The next step will be to try the same tests, but move to lower flow / finer atomization for the ignition system, and contain everything in a chamber.  If we get that working, then we can stage in higher flow.  We are currently experimenting with high flow spiral nozzles from Bete, but we will move to impingement nozzles for the ignition system.

 

 

Hover Test

 

We tried yet again to fly the 2’ diameter vehicle, but we still had problems.

 

The telemetry problem we were having, that I thought I had fixed by changing the RF channel, came back.  The problem behavior is that sometimes when the flight computer starts transmitting the UDP telemetry packets, the telnet TCP connection locks up, and pings stop getting through.  I thought for sure I had found the root cause when I noticed that there was an IP conflict between the laptop and one of the internal bridge IP addresses, but that didn’t change a thing.  Cutting down the rate or size of the telemetry packets didn’t fix it either, but when I changed the telemetry packet from a UDP broadcast to a UDP packet only addressed to the laptop, the problem did disappear, and it didn’t show up again the rest of the day on Saturday.  This is probably a firmware problem in the Esteem bridges, but if the workaround holds, I can live with it.

 

We got the damaged engine bulkhead fixed up and back in the vehicle.  All of the bolts that broke off came right out with an easy-out extractor.  I still found it surprising that the bolts broke instead of zippering through the 1/8” thick filament wound tube, but it sure made the repair easier.  One of the legs is still bent a fair amount out of shape, but it doesn’t hurt anything.  We replaced all four manifold-to-engine hoses.  We were very careful not to kink the short hoses during installation, but we still have some concern about that.  I ordered a set of convolute core hoses that should be much more flexible (at the expense of more flow losses), but they were a three week lead time item.

 

We tested a little 7” LCD from Xenarc with the flight computer, but we didn’t work out a good mounting position for it.  We replaced the A/D board again, and removed the currently unused four port serial board from the PC104 stack to reduce the number of things we need to worry about.

 

I tested new abort modes for the things that failed the last two times – crossbow failure, and A/D converter failure.  I also added an abort if the vehicle tips more than 10 degrees.

 

Because it had always felt like one of the engines wasn’t making nearly as much thrust as the others while warming up, we manually set all the valves to a specific position by using a drill bit as a feeler gauge before we installed the hoses and put everything back together.  When we powered on the flight computer, I collected some A/D samples of the current feedback values.  I had to do some averaging and conversions, so I just saved the data off and we went about the testing, but I really wish I had done it right then, because the data didn’t make sense later – three of the valve positions weren’t even in the calibrated open / closed range.  My current theory is that the servo valves don’t have any manual stops, relying only on their limit switches, and we turned them the wrong way when setting them.  This may also have caused some of the problems we experienced later.  This may also explain one of the other mysteries with the current setup: one of the valve feedbacks is in a higher range than the others.  It may have been rotated past its limit switch at some point.  I currently clamp the converted throttle positions to the 0.0 to 1.0 range, but I will start letting them go out of range, so we will clearly know what happened if we see something in the –1 to 0 or 1 to 2 range.

 

We pulled the vehicle out and lifted up fairly high for the water test of the new plumbing, but after loading some water and pressurizing, we found that we had left the master cutoff valve closed again.  I had added a warning light for this, but I hadn’t recorded the proper range values to make it accurate yet.  Since we knew this position was closed, I noted that, then we hooked up the manual valve switchbox and opened the master cutoff, and I noted that value.  The warning light now functions correctly.  However, when we hooked everything back up to the computer and I ran water through all the engines, engine 0 wouldn’t move at all.

 

We had to take a break to wait for a few tornados and a hail storm to move past our section of Dallas, so we pulled the vehicle back inside to do some other work.  We initially thought that we must have bent a pin hooking up the manual switchbox for opening the cutoff valve, but when we hooked the manual switchbox back up, that couldn’t move valve 0 either.  Eventually, Russ cranked it a little bit with a wrench, and it went back to working.  Seems like a limit switch issue.  Today, I went back over the data logs (one of the smartest things I have done this project is make the remote pilot application archive absolutely every telemetry packet, even if it isn’t in flight or manually saved, which makes this type of investigation a whole lot easier) and was able to determine that valve 0 had stopped working before the master cutoff had been opened.  It operated fine for the tests we did indoors, but when it closed after the final test, it ramped down to a below-zero level on the pot feedback.  This seems to have caused some problem with the limit switches.  I tried to replicate this behavior with an older valve servo that I have taken apart at home, but it didn’t seem sensitive to it.  The internals have changed a bit in the more recent KZCO valves to allow a visual position indicator on top, so the new ones might have an issue the old ones didn’t.

 

After the nasty weather cleared off, we took it back out and water tested again.  Everything worked properly, but we couldn’t tell clearly if the engines were all flowing correctly.  We loaded up 2.5 gallons of peroxide, hoisted it well off the ground, and slowly warmed the engines up.  Engine 2 was clearly not making as much thrust as the others during the warmup pulses, but the warm-ups only go to 30% throttle, which is just barely opening for ball valves, so it might have been a minor difference in initial calibration.  After finally clearing all the water from the packs and getting them up to operating temperature, I slowly throttled up for a liftoff.  Unfortunately, the peroxide ran out just as it was getting light.

 

When I looked at the telemetry, it showed engine 2, which we thought was weak, throttling a lot higher than the others, which I took as more evidence that we had a problem with one of the engines.  Now that I look back more closely at the data, it turns out I was wrong.  It wasn’t just one engine going higher, it was a pair of engines going higher to attempt to counteract the roll of the vehicle as it was hanging on the line under the crane.  I didn’t notice the other engine, trace because it was almost on top of the other one, and I was drawing that number in yellow, which is very hard to see on the white graph background.  I have since changed yellow graphs to purple, which shows up a lot better.  My current view of the data is that during throttle up, all the engines were moving close to the same rate, and there isn’t actually any real problem, which helps explain the rest of our tests.

 

We did full-system measured water flow tests at both 100% and 40% throttle.  There was no radical difference in flow rate, but engines 0 and 2 flowed about 20% less than engines 1 and 3.  We could not come up with any reason why engine 2 felt week, but I now believe it to just be an initial position slightly more closed than the others, making it seem week at the 30% warmup level.

 

After the 40% throttle test, I thought that two of the valves stuck open at 40%, which had us highly confused, suspecting a blown motor or blown driver board, but when we got it inside, it still responded fine to the manual control, and when I talked to it on the computer again, it seemed to be working fine.  The really odd thing is that the stored telemetry shows all the valves closing just as they were supposed to, so now I am suspecting we may have hallucinated the problem.

 

In summary, I think that I now understand all of our issues, and if we had just tried to fly it again on Saturday, it probably would have worked.

 

The valve with the weirdly out of range feedback position almost certainly overrated past the limit switch.  This could happen on a shock landing, because we default to leaving the valves commanded to close, relying on the limit switch to stop them.  The limit switches are reed springs, so if the were banged hard enough to bounce off their contact, the motor would have started rotating underneath it.  We are using the very fastest valves available, which may contribute to being able to run away from a bounced limit switch.  I bet if we manually crank it a few degrees past close, it will cycle back to the same range as the others.

 

It seems to be possible to stick the valve at the closed limit switch.  A reasonable argument could be made that we should completely remove the limit switches, because the computer can stop it properly with just the existing pot feedback.  However, this would make controlling the valves with the manual switchbox very difficult.  Instead, I am going to change the code to have the computer only go from 5% to 95% of the range of motion, so it should never actually hit the limit switches in operation, except when we calibrate the valves.

 

The weak feeling engine is most likely an initial cracking calibration issue.  Instead of taking everything apart and doing the feeler-gauge job, I will make a little calibration utility that lets us creep the valves open with the computer, and someone can just listen for when pressure starts escaping from the tank through the engine.  If we took out the limit switches, we could also find the final closing point, which would let us accurately determine the wide-open point halfway between.  Instead, I think I am going to just assume that only the initial calibration varies from engine to engine, while the total range of motion is the same, which should be the case, given that they use identical pots, and the valves are inherently 90 degree devices.