May 8, 2004 notes
We did several more flight tests to track down the GPS
loss-of-lock issue. We replaced the
internal 802.11b antenna, which was very close to the GPS antenna, with an
external one several feet removed, but that had no impact. We actually moved the GPS antenna off the
vehicle to a point ten feet up the tether to the lift, and it still lost lock
almost immediately after liftoff, so we conclude it has nothing to do with the
antenna. We then took the GPS circuit
board off of its mounting standoffs and wrapped it in foam and loosely wire
tied it down. We made two perfect
flights in a row with this arrangement, so it seems conclusive that some kind
of vibration is hurting the RF processing on the board itself.
We then tried an auto-hover with the GPS, and it worked
fine. The vehicle was bouncing up and
down a lot in place, because the control authority was a lot higher than the
responsiveness of the 10 hz GPS update, but it worked properly.
I had another theory on the GPS problems that sounded
plausible: while the entire electronics board is mounted on a foam isolation
ring that dampens any vibrations from the frame, it is mounted right above the
engine, and the base is directly exposed to the ground reflections from the
engine. Acoustic vibrations might be
getting transmitted directly to the board.
We remounted the GPS on the standoffs and glued a foam pad on the bottom
of the electronics honeycomb and did more tests. This seemed to be better, as I was able to get a full flight off,
but it still had a cutoff once right as it was leaving the ground, so we went back
to the loose mounting of the GPS. I am
fairly appalled that this $8000 GPS system intended for solid fueled missile
applications is this sensitive to vibrations.
It sounds like a cracked trace on the board, but Im not sure if I want
to send it back to Thales for another $500 service check that will take two
weeks and probably result in them just flashing the bios again and saying it is
To make up for the relatively low 10hz update rate of the
GPS (I wish I had gone ahead and ordered the 20hz option) and to provide some
more graceful failure modes, I combined the inertial position and velocity
sensing with the GPS updates, so it gets reset every time a valid GPS packet
comes in, but will coast with pure inertial data if the GPS is failing, and
provide useful data between GPS updates.
This isnt as good as a truly integrated GPS / IMU system that can use
the IMU data to smooth the selection and balancing of different satellite
signals before generating a GPS output, but it does several positive things for
us. The upwards position / velocity is
easy to use the IMU for, because it can auto-orient from the gravity vector
while on the ground, but to get north / east data, we now have to orient the
vehicle correctly before launching. I
have a magnetometer that we could use for an automatic roll orientation, but I
havent plugged it in for a couple years.
I also updated the GPS baud rate, which doesnt give me any
more samples per second, but decreases the latency in getting the updates it
The auto-hover was somewhat smoother than before, but nor
dramatically so. The overshoots are
proportional to the control authority times the sum of the sensing latency and
the actuating latency with the current control algorithm, so I may just
intentionally slow down the valve movement.
I may convert it over to a gain based control system like the attitude
control, but it is trickier because the vehicle is constantly changing weight
and tank pressure as propellant is depleted, so there isnt a reasonably point
on the throttle that a default actuation position could be based on. I dont want to waste much time on it,
because hovering isnt actually an important part of what the vehicle is
supposed to do.
(two angles plus on-board camera looking out)
Horizontal position hold, however, would be very useful for
us. Ideally, we want the vehicle to
land on the same pad that it launched from.
There was quite a bit of testing and adjusting to get all the axis
transformations correct in dealing with inertial space, GPS space, and jet vane
angles. Just like in computer graphics,
you almost always wind up getting something backwards the first time, then just
throwing in a negation or swapping the order on a cross product. One point of note is that until we get our
differential GPS base station and correction factors working, I am using
integrated GPS velocity instead of direct GPS position data for the position
hold. The idea is that a slow drift is
much superior to the vehicle instantly thinking itself ten feet off the target
spot when the GPS chooses different satellites.
We had a couple exciting runs where the correction factor
was backwards, making the vehicle tip into the direction it was moving instead
of away from it. Those runs completely justified
the expense of hanging the vehicle off the ground with a lift truck.
Finally, I got everything in the right order, and we made
two successful tests with the computer modulating the vanes to hold
position. All I did was say go up,
and the computer managed the throttle and the joystick, then came down eight
seconds later. The position correction period
was about two seconds, and the amplitude was about one foot in Y and only a
couple inches in Z. This being a first
random guess at parameters, Im sure we can cut this down to an extremely
smooth position hold with a little tuning.
I am looking forward to doing untethered flights like this at our remote
test site and moving on to accelerating boosted hops.
Unfortunately, the engine on the little vehicle was starting
to give us problems by the end of the day, with the exhaust staying cloudy on
the ground, and spitting some liquid in the air. It has over twenty runs on it, so it is probably time for us to cut
it apart and see how it looks. We have
plenty of ring catalyst, so if the hot pack is just worn out, we can quickly
replace it. If the monoliths in the
cold pack (which were scavenged form an even older engine) are worn out, we
will have to build a new engine for it.
We are extremely thankful that we only have a single engine to worry
Big jet vanes
We got lots of work done on the big vehicle conversion.
The tank manway came off, and I cut out the old parachute
mounting peg and milled the central hole wide enough to take a 2 pipe. We also added anti-swirl baffles to the
inside of the manway. We still have
separate pipes through the manway for fill and drain, but the next one we make
will only have a single (probably 4) pipe coming from the tank, with fill and
drain branched off of that below the manway so there are less weakening holes
cut in it, and no pockets of trapped propellant in the vertical orientation.
We are going to try mounting the landing gear directly to
the manway instead of to our bonded mount points at the tank periphery. We have some concern about hard side loads
on the manway flange, but it makes a lot of fabrication things easier, and it
will allow us to just transfer the propulsion and landing gear to any other
tank easily. We needed to add almost
two feet of height to the vehicle to clear the big engine and vanes and leave
us room for growing to a larger engine.
The big jet vane system is coming together rapidly. The vanes and actuators are mounted to a
stainless plate that is welded directly onto the big nozzle.
We cut off the plumbing and mounts on the new big engine
that we had set up for the test stand and made custom flanges for the 2 KZCO
valves so we could stack the throttle and master cutoff valves as closely as
possible. The manway feed pipe slides
into the top flange with a radial seal, so there is no stress on the plumbing
when the engine is mounted. We still
have to build the mounts that will hold the engine to the manway.
We have a pretty good shot at doing a tethered hover of the
big vehicle on jet vanes next Saturday.