June 4 and 8, 2002 Meeting Notes

 

In attendance:

 

John Carmack

Phil Eaton

Russ Blink

Neil Milburn (Tuesday)

Joseph LaGrave (Saturday)

 

Pack Experiment

 

Since we were packing 2” engines for the lander anyway, we tried an experimental low-pressure-drop catalyst pack.  The idea was to skip the stainless screens interspersed between the silver screens, and instead use an anti-channel ring between every pair of silver screens to give it an almost solid outer wall for structural strength, and an air-gap between the screens to expose more surface area, and allow gas to move around liquid.  It still had a spreading plate and 10 stainless screens at the top to spread the flow out, otherwise a high pressure peroxide jet would just cut right through the weak silver screens.   We were only able to fit 22 doubled up 36 mesh screens (44 total) into the catalyst pack because of the depth consumed by the rings, compared to 64 total silver screens in our normal packing arrangement.

 

We had the quite-restrictive 0.070” jet in, which held the flow down to a fairly low level, but it was still an abject failure, barely foaming the peroxide before it got out the nozzle.  We have seen behavior like that before, where some packs just will not cleanly catalyze, no matter how small you jet the flow down, and then, when you pass some magic threshold, they can take all the flow you can push through.

 

Our current formula is serving us well, but we will probably try a few other things over time to improve the large pressure drop.

 

 

Lander Flights

 

We repaired the bent leg from Saturday’s test flights, and finished packing and installing the 2” diameter engines on the seated lander.  With no jet (just the restriction of a –4 AN fitting) at 600 psi, these engines make 100 pounds of thrust, but we had an 0.070” jet and 500 psi pressure, which only gave about 32 pounds of thrust.  This would still be significantly more than the < 20 pounds thrust that the 1” motors would make with 0.060” jets at that pressure.

 

We also carefully leveled the main engine bell with the vehicle hoisted off the ground, so there wasn’t any chance that a bent frame or bent landing gear would give us an unintentional thrust vector.  We also leveled the computer box, so the initial attitude determination from the gravity vector would be aligned with the leveled engine.

 

All the work paid off, as we made a perfectly well behaved five gallon test flight.  The attitude engines had sufficient thrust to bounce it back and forth as needed, but not so much that it was having huge oscillations.  The laser altimeter data was also all good, now that I am discarding the occasional absurd value that gets sent out.

 

We could have flown it a couple more times like that, and had a triumphant day.

 

But no.

 

We decided to go ahead and give the laser altimeter based auto-throttle control a try.  Instead of using the joystick throttle, this is controlled by using the hat switch on the joystick to just signal “up 1 m/s”, “down 1 m/s”, and “hover”.  An initial tap in the up directions should have brought the target altitude to about 2’ off the ground, which would cause the main engine to throttle up until it reached a small acceleration, then throttle down as it approached the target altitude, then modulate the throttle to hover there.

 

We set it all up, I tap the up button, the engine throttles up, and KEEPS throttling up, with the lander leaping off the ground at a quite rapid, and accelerating, clip.  I canceled the auto-throttle by bringing up the manual throttle within a half second of it leaving the ground, and it started coming back down just before it pulled the thethers taut.  Our tether system is much improved from the old one that broke, with solid welded attach points on the frame, overhead lifting-rated shackles, and shock cord wound through the chain links, so we are confident that it would have grabbed it fine, but it probably would have given it a serious attitude adjustment.  As it is, it came down with attitude control and a little bit of main engine throttle, so it landed completely straight again, bending all four legs and banging all the engine nozzles on the ground.

 

Here was my post-mortem the next day:

 

At first, it looked like we had a time lag on the altimeter telemetry, but after looking at it more closely, we found that the altimeter and accelerometers were all working correctly.  The problem was that the ball valve was opening slower than the computer thought it was, so when the acceleration was such that it should begin to throttle down, it started moving the target throttle point down from the current target of 75%, even though the actual ball valve was only at 50%.  This caused the throttle to continue opening for another 200 ms until the target and actual crossed at around 60%.

 

I should have been checking for the target throttle position getting farther and farther away from the actual valve position, which could have bounded the time that it could overshoot.  The only reason that it maintained a target position at all was to allow the auto-throttle to move the throttle at a slower rate than full-open / full-close.  However, simulation showed that the smoothest control was gained with it running the valve as fast as possible, so there really isn't any reason to keep that target throttle position at all, it should just be commanding the valve to open more / close more based on the acceleration versus target acceleration.  I will change the code for this, after which I think that it will actually work.

 

The thing that bugs me is that I DID consider the issue of the main valve lagging the target, and I had tested it on the simulator and found it to not do anything too unusual, it just took a little longer to settle down.  Last night, I made the modifications to the simulator to let it simulate the accelerometer sensor as well as the altimeter, and tried again.  It was just like my previous test, not too exciting.  However, after I changed the simulator engine thrust to be able to give a 1G positive acceleration, and I had the simulator model the ball valve opening curve a bit (very rapid changes around 40% throttle), I was able to get it to behave just like the actual flight: it leapt three meters into the air, then oscillated all the way back down to the ground.

 

There are two other mysteries with the results:

 

The telemetry log stopped just as I pulled it out of auto-throttle mode.  I may have clicked the pause button on the joystick when I was canceling the auto-throttle, but there is no way of telling.  The computer lived through the whole thing, but we lost some useful data.  I am going to add logged information about what terminates graphing, but I also want to start having a completely separate computer that does nothing but capture every data packet that comes over the wireless net, allowing us to go back and extract whatever we need irrespective of what the remote pilot system chose to save.

 

We saw over 9 m/s^2 of acceleration on the vehicle, and it wasn't clear that the telemetry we got was even showing peak acceleration.  It moved FAST.  We don't understand exactly why it was capable of moving quite that fast.  The last time we weighed it, the vehicle was 350 pounds dry, and we added 30 pounds of ballast and at least 40 pounds of peroxide, so it would need at lest 760 pounds of thrust.  From our last bench test of the big motor, at slightly under 500 psi tank pressure, it should make 590 pounds of thrust.  The attitude engines made 32 pounds of thrust at that pressure, and two of them are on at all time, giving a maximum thrust of 654 pounds.

 

The four attitude engines running half the time may give more than 2x32 pounds of thrust because of initial inrush, which I seem to recall was the case when we tested 50% duty cycles a long time ago, so that may give another 30 pounds or so.

 

Still, it seems like we have some set of sensors miscalibrated, either the load cell we used for the big engine, or one of the scales we were weighing the lander with.  The big engine's center of thrust on the test stand may also have introduced a pivoting force that caused it to read lower than actual at the load cell.  I am pretty sure the accelerometer is accurate, because I checked the 1G gravity vector both ways on each axis.

 

When we rebuild this vehicle, we are going to experiment with using wire rope isolators from http://www.enidine.com as landing gear.  We are going to move the mount points inboard so they are directly under the triangulated load points.  Extended legs in single shear worked well for the 40 pound lander, but they just don’t cut it on a 400 pound lander.  We are also going to make engine mounts that keep everything high enough that the frame will bottom before an engine nozzle does.  On Saturday, we pulled all the engines off, swapped out the dinged nozzles, and replaced the broken fittings, so the propulsion system is ready to go again.

 

One thing we did notice was that the accelerometer data was very smooth, cleanly ramping up to 9 m/s^2 now that I am smoothing of 8 samples (40ms).  This is going to be a whole lot better for the auto-throttle than the noisy double-derivative (quadratic regression, actually) of the laser altimeter data that I was doing.  The flight computer software is updated to just drive the valve as fast as possible, so we are ready to give it another test in a new vehicle.

 

Tube Vehicle

 

We spend most of Saturday working on the tube vehicle.

 

We pulled one of the tanks off the seated lander frame, and finally got around to measuring it properly.  It weighs 62 pounds, and holds 115 pounds of water, which is 13.8 gallons.  I didn’t measure the length, but it is 14” in diameter.  That is a 3000 psi rated tank.  For comparison, the 45 gallon 150 psi rated fiberglass tank that I am trying to buy only weighs 46 pounds, and is 22” in diameter.  668 psi*gal/lb for the carbon tank is obviously way better than 146 psi*gal/lb for the fiberglass tnak, but I haven’t found any stock carbon tanks built for <= 1000 psi usage.

 

We got centering rings made to hold the tank inside the tube.  We still need to epoxy coat them, and probably add a liner of gasket material to snug the fit down.

 

We got all of our airfoil fins / legs cut and mounted to the bottom bulkhead.  We still need to pin them so they can’t twist, add metal plates for load spreading, and mount the rubber bump stops on the tailing edges.

 

We got the main engine mounted on the top of the bottom bulkhead, so the nozzle doesn’t protrude as far.  We have a Kevlar-phenolic ring under the flange, and alumina-silica insulation clamped around the catalyst chamber, but we still plan on doing a full-vehicle static test to make sure nothing catches fire from the heat inside the tube.

 

We made holes and a mounting strap for the laser altimeter.  We still need to put a piece of lexan underneath the laser altimeter lenses so they don’t get peroxide on them.

 

We plumbed up our full upper manifold with pressure gauge, pressure transducer, temperature RTD, manual purge valve, quick connect, and quick connect bleed valve.  We still need to do a little work on the lower manifold to plumb to the attitude engines on the bottom side of the lower bulkhead.

 

We vented another servo ball valve, because the one on the lander was killed in the crash.  We made a long enough cable to reach from the valve at the bottom to the electronics box on the top bulkhead.

 

We discussed how we want to set up the parachute system and nose cone.

 

We have a number of tests to do before flying the vehicle:

 

Full power captive test to “smoke check” the vehicle.  We will probably flip the tube upside down so the main engine points straight up, and run the propellant from the trailer.  We will run a full five gallons through it like that, and see how hot all the different parts get.

 

Parachute shock cord drop tests.  Joseph can suspend the vehicle off the ground from the shock harness with his tractor, and we can let it come up hard on the cord from progressively greater drops to simulate parachute opening shocks.  We don’t want this big thing pulling apart several thousand feet up.

 

Landing gear drop tests.  If we get wire rope isolators on the bottom bulkhead, it should land nicely from hover tests, but if we only have the rubber bumpers on there, we won’t expect too much from it.  It will be interesting to get accelerometer logs from these tests.

 

Swinging attitude control.  We should give it a good push while it is suspended, then engage the attitude engines to see how much it will be able to stabilize itself while coming down under parachute.  We hope to be able to throttle up the main engine to an on-gear soft landing after coming down most of the way on the parachute.