Jan 23 meeting notes:

Location: Long Range Systems

In attendance:

John Carmack
Phil Eaton
Russ Blink

Next meeting on Tuesday at LRS for a VTVL system checkout with water. If the planets align correctly, we may be able to do a tethered power up.

We have a tentative attitude sensing solution.  Gyration (www.gyration.com) will sell their dual axis rate gyro and support chipsets in small quantities.  In lots of less than 10, the gyro is $150 and the support chip is $18.  I got four of each in today, and we had a look over the specs and application notes.  A GyroMouse only retails for $99, so there is a bit of gouging going on, but buying the real parts does give you access to tech support.

The support ASIC turns out to not be doing much of anything special, just 12 bit A/D.  I thought it might be doing some application specific compensation or correction, but it doesn’t seem to be.

I have been testing with a butchered GyroMouse as a two axis rate sensor, and we will probably do initial manual liftoff with that logging data. It has an unwanted radio link, only updates 30 samples a second, and needs one more axis, so we intend to build a custom board for later use.  We will probably incorporate three DC accelerometers as well, giving us a full 6DOF rate sensor.

An interesting side note:  during my testing with the GyroMouse connected as a serial device, the default Windows 16550 configuration caused noticeable problems.  Serial mice send three (or four, in some cases) byte packets, but the default FIFO limits are set so that you only get an interrupt when 12 - 14 characters have been received, or when four character clocks pass without new data.  At 1200 baud, this causes very long delays and getting updates in lumps of three or four at a time, giving very jerky graphs.  Either disabling the FIFO, or setting it to interrupt on the very first character fixes it.  I still need to grunge around in Linux to find how to set this there.

Also interesting, on Win98, the updates come very precisely at 33 or 34 msec apart when I have the scheduling priorities cranked all the way up.  On Win2K, they com at 30 or 40 msec intervals, implying a 10 msec scheduling granularity.  I haven’t tested on Linux yet.
 
The orientation of the sensors is nicer than most of the dual axis accelerometers  one of the axis comes out of the board, so we can get the third axis by just rotating the second gyro chip instead of having to install it sideways on a riser.

It handles 150 degrees a second rate, which is way more than our craft should be doing, and the drift in place doesn’t seem bad at all on the Gyro Mouse.

The biggest limitation is that they are only good up to 2 to 3 Gs of acceleration.  That is completely fine for our first generation VTVL, but not as much as we would want for a high performance rocket.

We will see how bad the drift characteristics are in operation, and see how we can use other sensors, like GPS and magnetometers, to keep it calibrated on longer flights.  For remotely piloted flights, it isn’t going to make a bit of difference, because drift will just move the joystick centering position, which is easy for an operator to compensate for.


I have decided that my initial purchase of an EBX/PC104 form factor computer for our flight system was a mistake.  I liked getting everything I needed for development on a single board, but I am now pretty sure I would prefer a couple pure PC104 boards in a stack instead.  I have a separate document on flight computers that I am putting together as I learn things.


I brought a new NOS pressure guage for the VTVL manifold, so we can leave the other one on the test stand.  It is 1/8 NPT, but the gauge didn’t quite clear the installed hose ends.  I stuck some adapters together to space it out, but in hindsight, I bet we could remove the hose below it, and install it with the proper clocking (gauge facing up) so that the hose can be put back on without clearance problems.

The fill connector still requires an obnoxious chain of adapters.  If we could tap the manifold for 4, we could screw it directly in with an O-ring.

I brought an NOS nitrous filter that we were going to use as a nitrogen filter, but when we looked at it, it didn’t seem all that useful  the screen inside wasn’t as fine as we would like.  I still want to get a good filter, because Bill Colburn has reported that there have been incidents of corrosion inside nitrogen tanks getting blown into a peroxide tank.

I brought a couple NOS blow off valves, but we are going to need to get some adapters to fit them to the manifold.  Russ  the best solution would probably be to get larger taps and just cut the manifold to take it directly.  You might want to check it out and see if it is feasible.  The blow off valves are rated at an incredible 3000 psi for our tanks.  I would like to find some that are more like 600 psi, because the hoses we are using are only rated to 1500psi.


The lander3d.exe program can now deal with the PWM actuation and GyroMouse sensing for the initial VTVL liftoff, and has a bunch of customizations for our first test flight.  It is now configured so that no engine will drive unless the joystick trigger is held down, so just releasing the joystick should shut everything down instantly.  I made a 100’ parallel extension cable for the tethered tests, and it checked out fine.


We cut some light chain for tethers for the first test.  We are only going to give it about a foot of freedom until we have some sense of the vehicle dynamics.  Once I get some attitude and engine logs, I will try and make the simulator model the actual behavior better, then we can think about giving it some more leash.


We did a lot of engine tests today, with somewhat mixed results.  We got good power and perfect catalyzation out of every run, but many were rough.  What we do seem to have found is that compressing the catalyst packs more always increases the smoothness of the run, but drops the power output a bit.

A large pressure drop would not be acceptable for a high performance vehicle, but since we are a long way from caring about optimization, we have two workable paths:  either just run much higher tank pressure, or just open op the throat more and live with a low chamber pressure.  Our tanks are good to 3000psi, so we could certainly go that route, but I like staying around 400 psi for safety, and I don’t mind the Isp drop.

From our previous tests with highly compressed packs, it showed that we don’t need to fill the engine with them, but if we cut the size down, I do want to add spacers so we keep the pack all the way to the inlet fitting.

A useful result we got was that the front injector plate can be omitted entirely if you have a dense enough cat pack.  Having catalyst right up to the inlet fitting prevents a puddle of peroxide from forming in the engine at the end of a horizontal test run, which I believe was the cause of the thrust bumps at the end of our previous runs.

The new modifications to the PWM static test program (which we use even when doing 100% duty cycle runs) worked great.  It now clips off some of the initial cruft, allows comments in the scripts, and can continue to vent the tank after it has stopped logging.

Results are at: media.armadilloaerospace.com/misc/jan23.xls
Run 1: Uncompressed pack of cut disks, new check valve in place of pipe union, no spreading plate.  This run was extremely ragged.

Run 2: Go back to the old check valve.  Still extremely ragged.

Run 3: Add spreading plate, slightly compressing the pack.  Smoother, but still not as good as we have seen.

Run 4:  Compress pack a quarter inch, add more disks, keep the spreading plate.  Nice and smooth.

PWM run:  After the initial warm up pulse, this run two seconds of 10 ms on / 15 ms off, then two seconds of 15 ms on / 10 ms off, then two seconds of 20 ms on / 5 ms off.

The small duty cycle seemed to ramp into the medium duty cycle.  I need to extend the test program to log the duty cycle by the thrust curve to see better what is happening/.  The medium duty cycle was decent.  The high duty cycle intermittently failed to operate the solenoid inside the 5 msec, making it very ragged.

My current theory on the PWM is that there may only be a narrow range of “off” times that do not either have solenoid opening problems, or allow the chamber pressure to blow down to the point that the nozzle flow separates.  We may not be able to deeply throttle smoothly, but we should be able to get high percentages by stretching the total cycle time instead of shrinking the off time more.  More experimentation is needed.

Run 5: No spread plate.  This started out the smoothest run, but got ragged at the end.

Run 6: 450 psi instead of 400 psi.  This stayed ragged the entire time.  Something may have happened to the pack at the end of run 5.  We don’t know.

Run 7: Test one of the new engines with a new pack compressed a half inch and no spreading plate.  This has exactly the same throat area and expansion ratio as the old engine, but has a smoother radius in the nozzle, and a 15 degree expansion cone instead of the 30 degree that was on the original (because Russ didn’t have a small enough tool at the time).  The nozzle does seem to make a bit of a difference.

This run also seems to give some evidence that my load cell meter is point sampling at 12hz instead of doing any averaging, because the warmup pulse is completely missing.  This is likely contributing to making the PWM runs look worse on the graph than they really are.  I need to get a high frequency load cell meter.


Stuff to do:

Order more peroxide from X-L.  We have enough to do a bunch more engine tests, but the VTVL will chew through our supply rapidly.

Adapter for blow off valves, or drill out the manifold.

Possibly tap one of the manifold ports to 4 for the quick connect.

Cinder blocks for the tether chains.

Finish the rear plates for the new engines.

Cut catalyst for all the new engines.  I think we should go with highly compressed (at least the half inch we ended up with), but we could probably use half the total depth or less if we can get a good spacer and want a little more chamber pressure. 

Bring in the pressure transducer and specs so I can see about getting the logging integrated.

Check my glassware box to see if we have something suitable to replace the broken beaker, or buy a new one.

Is it possible to make a chamber pressure tap in the current test engine?

Go through Linux From Scratch on the embedded system.

Allen screws and nuts so we can get the swivel 90 fittings on the engine mounts.

Various female pipe and AN fittings, which we always seem to need to double up other fittings to simulate.

Find a big viton O-ring for the bottle necks.

Protective plate on the bottom of the VTVL manifold to protect the hose ends.

Mount the multiple solid state relay board on VTVL.

Build a couple sets of batteries and work out mounting on the VTVL.

Mount the butchered GyroMouse on the VTVL.

Put everything in a Tupperware bin in case we want to hose down the VTVL?

Fix up the original single solid state relay for the test stand so we don’t need to pull the multi board from the VTVL.

Build the 6DOF rate sensor board.