July 28, 2001 Meeting Notes
Todays Flight Test Video
Gyro Bias Drift
We definitively solved a significant problem today.
On Tuesday, we noticed that the gyro rates drifted
noticeably when all four solenoids were open, causing the attitude sense to
drift somewhat in proportion to the throttle position even when the vehicle was
flat on the ground.
We were able to see in the logs that the bias point voltages
moved when the solenoids were drawing their 35 amps of current, which also
dropped the main voltage level a couple volts, so I programmed the computer to
read both the rate signal and the reference level each frame, and subtract them
for software differential A/D. This
made a minor difference, but the problem wasnt easy to replicate on the bench.
If I dropped the bench power supply from 12v to 9.5v, the
bias moved slightly, but not nearly as much as it appeared to when the
solenoids were operating. If I dropped
the voltage below about 8.5v, the gyros went into a very fast roll, much more
noticeable than the problem we saw. I
checked the specs, and the gyros take 9v 18v, so it isnt too surprising
dropping below that caused problems.
Those problems were not addressed by the software differential A/D.
I added tracking of the minimum voltage seen at any time
during a frame, so we could see if it was ever dropping down to the 8.5v level
that causes major gyro problems.
I started with a freshly charged battery today, and was puzzled
to find that the rates still drifted when the solenoids were firing, even
though the minimum voltage was still over 10v.
It clearly wasnt just a problem of voltage, although I still worried
that with a little bit of battery discharge, we might drop under 9v with all
the solenoids on.
We moved a few wires and cables around and retested, but to
We finally decided to run the solenoids from a separate
battery from the rest of the electronics.
I had wanted to avoid multiple batteries, but it was worth trying.
Luckily, we had everything we needed on hand to rework the
power supply, and the battery that was intended for the master cutoff computer
became the main electronics battery, while the larger battery was left to
solely power the solenoids. We still
had a common ground, but we had two completely separate +12v sources.
We tested it again, and while the voltage minimum logged by
the electronics box didnt change at all during firing, the gyro rates still
moved when the solenoids went on.
We took it back apart, and separated the grounds. The solid state relays are optically
isolated, so there is now no connection whatsoever between the solenoids and
their battery, and the rest of the electronics.
We tested it again, and it now worked perfectly. No change whatsoever in the rates for any
combination of solenoid activations.
Lesson learned: keep
your electronics COMPLETELY separated from any high-amperage solenoids.
Damn spooky analog crap.
We taped a new little inner tube on top of the lander tank,
because the last one sacrificed itself for the cause on the last flight with
the power outage.
I made two changes to the flight control logic since last
flight: the gain has been turned down a
bit so it shouldnt kick itself back and forth as hard, and all the sensor
values are averaged over an 8 msec time to reduce the noise we currently get
from our A/D board.
We did a couple hover tests, and I would say that it is in
general a slight improvement, with the oscillations being reduced to about
three a second, and of somewhat lower amplitude. The filtering adds some additional latency, and it is clearly a
crutch for a not-working-well A/D system.
Compare against the July 7 tests:
We did our first untethered flight (the mpeg at the top),
involving flying over a little median in the parking lot.
It worked fine, and was straightforward to control.
I had purchased a force-feedback joystick with the intention
of using a stick vibration as the propellant timer warning, because beeps from
the computer just werent audible when the rocket was flying. I really just wanted something crude like
the Nintendo rumble packs, so the Microsoft Sidewinder force feedback joystick
is massive overkill in terms of sophistication, but it was all I found. I havent written the code to actually use the
special features yet, so I was just using it as a normal joystick today. I find that it doesnt offer as good of self
centering as a normal joystick, and it is easy to pull it a little off center
without knowing it. We switched back to
the other joystick to let Russ fly a hover test.
Russ is going to cut all the engines down so we wont need
the spacers in the chambers, and also fix the threads on the top closure that
is visibly leaking in the video.
Our next flight test will involve a new joystick interface
for piloted roll, allowing 360 degree or greater rotations, instead of just
+/-20 degrees like the other axis have.
There will be a couple subtleties to doing this, but it shouldnt be
I am also ready to test a flight control option that
recalculates the engine levels every millisecond, instead of every PWM frame,
allowing it to change an engine level without waiting for the next frame in
some cases. This should have a slight
benefit in reducing latency related oscillations.
I may also try the accelerometer-as-inclinometer initial
angle setting soon.
We got the foam pads under the frame today. They are a bit stiffer than we were
expecting, but they should work fine.
Russ and Phil finished the rest of the attitude (70 pound
thrust) engine machine work today. We
will make cat packs, injector plates, and retaining plates, then test fire each
of them on Tuesday.
We need to get slightly smaller clamps to hold the engines
to the frame.
We made connectors and cables for all the solenoids to the
Russ has the metal stock for the main engine.
We are going to wait until we have all the plumbing in place
and everything nicely routed and tied down, then Bob is going to take it back
and add brackets for the throttle and joystick, and get it powder coated.
We may be hopping the frame on just its attitude engines
(without a person) in three weeks or so.