January 5, 2002 Meeting Notes

 

In attendance:

 

John Carmack

Phil Eaton

Russ Blink

Joseph LaGrave

 

A new volunteer is also joining our team: James Monzel. He is bringing aerospace composite fabrication experience, which is going to be very useful.

 

More Crash Follow Up

 

I have swapped joysticks and updated my laptop from WinME to Win2K as speculative improvements. Win2K still has RTS and DTR on at all times like WinME did, which causes problems with my test stand hardware. I think I finally figured out why they did this – serial mice need those lines on to provide power, so to allow Windows to autodetect serial mice when they are plugged in, those lines need to be on by default. They should still have a driver preference to leave it off…

 

I also received some very interesting information from a USB developer: the behavior we saw looks very much like a USB timeout. I had been assuming that the driver sent out poll requests to the devices on a regular (120hz or so) basis, but it turns out that if a response doesn’t arrive (missed interrupt, hardware glitch, etc), the system will wait for a significant timeout period, on the order of seconds, before asking for another update! While there is a good chance that the dropped data could have been due to the general suckiness of WinME, it is important to note that it can still occasionally happen on any system. I am probably going to write a custom linux driver that explicitly polls the joystick, or see if I can just cut the timeout down to something reasonable, like 50 msec.

 

On thing I fixed recently was that the default linux joystick driver has the bad behavior of giving the joystick throttle a “dead band” around 50%, as if it was supposed to be self centered. I finally tracked down how to turn that off with an ioctl.

 

 

Altimeter Flight Control

 

I have placed an order for this laser altimeter:

http://www.laseroptronix.com/flyg/altm400.html

 

It has a 1000hz update rate, a short enough minimum distance, and good enough accuracy for us.

 

The Advantage from http://www.laseratlanta.com costs the same amount, but only has a quarter the update rate.

 

The Impulse from http://www.lasertech.com/ is cheaper, but I couldn’t get good enough specs to be sure it would work for us.

 

http://www.rieglusa.com has a range that includes models with up to 12khz update rates, but their 200 hz model was over twice as expensive as the laser-optronix model.

 

I have been simulating different automatic flight modes based on the error range of the ALTM400, and the flight control is going pretty well. It was going great until I finally started simulating the real speed that the motor valve opens and closes, at which time the simulated vehicle started going into fairly large up/down oscillations when it was hunting for a given altimeter velocity. I’m not sure yet if I want to use a filtered accelerometer signal, which we pick up a lot of vibration on it right now, or the double derivative of the altimeter for halting throttle movement.

 

I am working on two different augmented control schemes, both using the joystick hat switch for control instead of the throttle. One scheme just assigns one of three target velocities: +2m/s, 0, -2m/s. This is a fine scheme for flying around the parking lot, but you can’t go very high. The other scheme still has the middle position for hovering, but has the up position go to full throttle, and the down position, go to the minimum throttle that results in a landing with a given impact speed. This lets you reach the maximum altitude and descend with the minimum propellant use. However, because our attitude engines are running constantly, just coasting up and down with attitude control takes about a quarter of the peroxide required for hovering, so it still can’t reach very high altitudes and still be able to land under power. For high altitude, we really need passive aerodynamic stability in the coast to apogee and the descent towards the ground. Getting both requires either letting the vehicle flip over, or some form of deployable aerodynamic modifier.

 

 

Catalyst Pack Experiments

 

Before our engine testing today, we experimented with a different tank and feed system. The small tank on our test stand had enough turbulence through the bottom manifold that we were seeing rough pulsations in the flow of even just water coming out of a valve hooked directly to the manifold, and it was making it difficult to tell what roughness was the fault of an engine, and what was the fault of the plumbing. We hooked one of our big, dual ended tanks up today, with one of the nice tank thread to 1/2" NPT converters that Russ made for the vehicle frame three plumbing on the bottom, and the old manifold on the top. The flow is acceptably steady now.

 

We continued our experiments with pure silver screen catalyst packs today. We also got a quote back for getting more foam plated and (to improve the life) sintered, and it was surprisingly high. The 20x20 mesh pure silver screen is a whole lot less active than the plated foam, but it is actually relatively inexpensive. Russ found an Indian supplier, and the screen price is just barely above the going rate for raw silver. We are probably going to want the very fast acting foam in the attitude control engines, but we might move the central engines to screens. The rotor tip engines will probably have to be screens, due to the high pressures across the pack.

 

The motor we bought from Juan Lozano (http://www.tecaeromex.com/) back when we first started uses plated screens of different materials. We only fired it a half dozen times or so, but it worked fine. It takes about 4x as long as the foam packs to warm up, but he has since developed a “pentametalic catalyst pack” which is supposed to be much faster reacting, not needing a warmup pulse at all. I need to remember to accurately measure the pack dimensions on that motor for an update sometime. Basically, it was about twice the depth as our foam packs. Juan said that he has run his packs for up to two and a half hours without losing efficiency, which is a heck of a lot better than our foam has done.

 

The small motor (a rotor tip motor) we are testing has a 0.65” diameter (0.332 square inches) by 2.55” long catalyst pack chamber, and a 0.25” diameter nozzle throat. A small amount of the length is consumed by the two anti-channel rings and a retaining plate. This is up from the 1.29” length in the motor tested last weekend.

 

We found last week that too much compression can close the silver screens up just as badly as the foam, so we went lighter on the compression today. The first test had 96 screen in the pack. When we ran it at 400 psi, it was briefly clear, but then clouded up somewhat as it got rougher.

 

We opened the engine back up, and while the pack wasn’t visibly loose, when we put it back in the press, it compressed quite a bit farther for the same amount of effort. We put 46 more screens in, for a total of 140 screens, which is a fairly restrictive pack. The 400 psi test run was finally a clear one, the first sustained clear run we have seen with the pure silver screens. Performance was only smooth at the beginning, then it stayed rough the rest of the run. We did another run at 650 psi with similar results.

 

To smooth out the roughness, we probably need to add a restrictive injector plate ahead of the screens. That has always improved smoothness in our other motors, but Juan’s motor is still reasonably smooth running with only a very coarse distribution plate on top. If the motor had stayed running smooth, the 96 screens probably would have been sufficient, because a rough running motor pulls in disproportionately large amounts of peroxide during the pulses.

 

We expect this to make around 20 pounds of thrust, which is 1/6 of a pound of peroxide a second at 120 Isp, or about half a pound a second per square inch of catalyst pack, or 30 lb/sq in-min. The Rocketdyne peroxide book states “Most silver screen catalyst beds are operated at bed loadings of ~20 lb/sq in-min”, so this is probably too small of a diameter. They also mention that typical packs use “15 to as many as 100 catalyst screens”, and that the active catalyst screens are usually alternated with inert screens, and the pack is usually terminated with several inert screens..

 

We will probably test the following changes next week:

 

Make a larger diameter / shallower cat pack. I am going to make a 1” diameter test engine next.

 

Add a restrictive injector before the catalyst bed. I’ll make a couple different hole counts.

Alternate silver screens with inert (SS or nickel) screens to make the pack stronger. The current pack may still function with 50 silver and 50 inert screens.

 

Things we might try later on:

 

Try the samarium nitrate “activation” treatment. There are strongly mixed reports on this. Mark said that it significantly helps pure silver screens (but not plated screens), Juan said that it is a detriment with organically produced peroxide, and there are published papers taking both sides.

 

Sandblasting the pure silver screens to increase their surface area.

 

 

Rotating Tip Engine Testing

 

http://media.armadilloaerospace.com/misc/FirstSpin.mpg

 

Even though the tip engine was running roughly, we went ahead and spun it on our rotating test fixture. The single engine is (roughly) counterbalanced, but because only one side is thrusting, it does generate side loads on the bearings as it spins. We will probably be using matched pairs of engines, but the side loads aren’t really all that bad, far less than what, say, a crankshaft sees.

 

We tested the rotary joint and plumbing initially by hooking a water hose directly to it. We have some little garden hose to NPT adapters that we were intending to use for a test stand water cooled hybrid nozzle, and they work well for this. The rotary joint has quite a bit of friction, so it only slowly rotated around. One disadvantage of this was that unlike pushing water out of the tank, the lines weren’t purged dry after the water was in them, so when we started with the peroxide, there was still a lot of water in the system. We probably should have blown it dry with nitrogen or compressed air.

 

We set it up with the tiny nitrous solenoid we were using on the test stand, which was probably a mistake. At low tank pressures, that valve isn’t going to flow much at all. The solenoid feeds into the rotary seal, which feeds into the hardened bearing shaft, which feeds into the stainless side pipe that feeds into the engine. This is a fair amount of volume after the valve, and if the shaft is rotating, it will all get pumped out to the engine, so I had some concern about how much it would continue thrusting after I closed the valve.

 

The first test we only pressurized the tank to 150 psi. The first couple presses of the fire button didn’t do anything, but eventually we got the water pushed out of the system, and got some little puffs of peroxide decomposition. With the spokes at rest, the engine behavior was very odd, with the dribbling in peroxide causing it to make little intermittent chuffs for about ten seconds. Slow experimentation used up the rest of the peroxide.

 

The second test we increased the tank pressure to 300 psi, and I let it run a little harder. It worked pretty well, spinning up to about 200 RPM on a couple quick pulses of the fire button. Once it was spinning reasonably to pump everything out to the engine, the response was a lot better.

 

The seal and bearing shaft are not made of good peroxide materials. They are after the solenoid, and vent through the open engine, so it isn’t a safety issue, but they did wind up depositing a visible amount of rust on the top of the catalyst pack.

 

We are going to balance it out better, add a check valve before the engine, change the top solenoid, and hook up the tachometer on the next test session.

 

We need to make real legs for the frame that we can securely bolt down.

 

Some interesting rotor blades that would make setting up integral engines really easy:

 

http://www.vortechinternational.com