With the poplar IMU + 1/2" padding + doubled PID gain + no flybar weights, she controls attitude like Hillary. Release cyclic & she returns to where she thinks the starting point is. The starting point drifts slightly, but it's roughly what the Corona did when it was hovering autonomously.

P was the key factor, needing to be much higher than it was on the Corona to get the attitude to fight back.

4 U homegamers, the recommended drive belt on Taiwannosauruses is the MX400 drive belt. Intended for a Helimax MX400 but fits the 450.

As for position hold, she definitely holds heading where the Corona could not. Still a disaster with position. Converted the GPS routines from NMEA to Sirf III back in December to try to gain some ms. That involves converting from XYZ to LLH in software. Now GPS seems to report starboard motion when she's really moving port. May need to revert back to NMEA to get something to work, before pushing Sirf III more.

So after extracting Styro-MU from its wrapping, found the Vref wire was not connected. Apparently it wasn't soldered properly back in November https://www.rcgroups.com/forums/showthread.php?t=769983. This only could have affected gyro drift.

Accelerometers were moved from the old RC network https://www.rcgroups.com/forums/showthread.php?t=762150 to the 3.3V bus.

Styro-MU was reverted back to a 0.75" Poplar MU to increase the weight.

Meanwhile, Easystars R cheap but not free. Got the tag for Vicaglider down to around $120. The first program will be flying towards a single point at all times. Then, the search is on for an orbit algorithm.

Some orbit algorithms:

For every GPS reading, determine distance from center of the orbit.
Get error between distance & target distance.
Add port or starboard rudder depending on error & direction of center.

Calculating distance & direction on a PIC is a disaster. Direction accuracy under 45` is unlikely.

Drive belt #1 took only 2.7 hours to break. Probably broke from software tail gyro testing. That will be all for Vicacopter until Tuesday.

Decided software tail gyro wouldn't go any better than it did last time. The success with log curves may have been a fluke or the broken drive belt.

Flight tests with cyclic angles have shown the attitude doesn't always reset when we release the cyclic. There may be more value in PID tuning.

The 3 weapons against vibration:

1) More weight

2) More padding

3) More lowpass filtering

Going by #1, U won't get good results by splitting up the gyros & mounting them on separate foam pieces. A very large, padded IMU is probably necessary.

Have decided removing the flybar could possibly be done, would free up a huge space for padding, & would reduce a huge amount of vibration.

Simply rotate the blade grips so the ball links R on the leading edges & remove all flybar components. Leave washout base & washout control arms in place to turn the swashplate. Connect SF mixing arms directly to main rotor housing with short ends attached to blade grips. Move servo ball links to innermost position.

The blades on the Taiwannosaurus don't teeter like the Corona blades, so removing the flybar gets U a 3" x 3" space on the tail boom. It's all about maximizing cheap airframes so U don't need a 600 to fit enough padding.

Unfortunately, it would take a stupendous software rate compensation. We all remember what happened with the last https://www.rcgroups.com/forums/showthread.php?t=722062 software tail gyro. Concluded ordinary tail gyros use a logarithmic feedback curve whose limits R a closely guarded Chinese state secret.

More software torque compensation (0 min 54 sec)

And closes most dealers. Suspect it's the end of the PT Cruiser & Crossfire. They copied the Crossfire name ever since we made it up. Serves them right, using a fake name for a car. Once the PT cruisers stop rolling off the Mexican assembly lines, it'll truly be the end of Web 1.0.

So the next plan was:

1) Revert to 32 bit lowpass filter.

2) IMU video on the bench with flight RPM & collective.

3) Attitude hold using different accelerometer trust.

Video testing proved high frequency vibration on the Taiwannosaurus was the cause of the accelerometer skew. Not surprising, since accelerometers detect acceleration using vibration. So it's back to finding ways to stuff more padding on the Taiwannosaurus & reduce cable vibration.

That will require large amounts of 9 conductor flex cable. Also went back to the 32 bit lowpass. In 16 bit mode + 1/256 bandwidth, only 8 bits of the analog inputs are actually used.

She actually coughed up 13min flight times out of both 3.3Ah batteries. The reason: battery failsafe stopped working. May permanently disable it and rely on the ESC cutoff, which is quite soft on the RCE-BL35X.

Pointing West wasn't a problem. Definitely looks like the IMU misalignment with magnetometer may have affected it.

Speaking of alignment, the 2.1Ah Align kit battery debutted today with a flight time of 10 minutes, exactly what the 6 month old 3.3Ah batteries R doing. Unfortunately the kit battery rapidly fell down to 9.6V before any battery failsafe could kick in.

Accelerometer trust factor of 0.002 seemed to be the highest before dynamic effects were unflyable, but drift was too high. Vibration is definitely worse than Corona now.

The plan for today's Rain Ramon visit was:

1) isolate the IMU less and hard mount it on the frame. The theory with this is most of the vibration is from the wire & slack in the velcro.

2) Lower IMU bandwidth from 4/256 to 1/256.

3) Make it heavier & pad it more. This assumes most vibration is in the frame. It above all needs test probe wire & smaller size. Cotton would have to enclose it completely.

Rebuilt the IMU using a 5/8" styrofoam cube & hard mounted it. The hard mounted IMU was greatly improved, but now shows horrible absolute angles. Once she got moving, the IMU still seemed to drift. Flight time is now 11min.

Test 2 was indistinguishable from test 1. Quaternion model was slower to respond with 1/256. Could mostly fly with cyclic angles but she could not return to the starting angle. Need more tests with the bandwidths.

Did 1 attempt on full autopilot just in case and wasn't surprised by the result. U get to settle for more photos on the ground.

Just about ready to replace Minicom with a serial interface specifically for Vicacopter. Typing in rz filenames over and over is almost as tedious as complaining about it on blogs.

Looking things over, we once had to convert all the floats to doubles on the ARM to get AHRS to work. Lowpass filtering on the ARM used doubles before it was moved to the PIC and converted to 16 bit ints. So maybe the PIC should use 32 bit fixed point.

https://www.rcgroups.com/forums/showthread.php?t=769193

When the blades were tracked & AHRS stopped working, we blamed it on aliasing with the averaging stage. That's why lowpass filtering was moved to the PIC.

The pwm loop combined with 32 bit fixed point analog is super slow. Upgraded the radio failsafe mode to shut down all PWM except throttle when contact with the ARM or radio is lost. Also got a nifty flashing LED state for when ARM or radio contact is lost. Makes it easier to test radio failure cases and quiets down the servo jitter in the dumpy apartment.

32 bit fixed point was another dead end. 16 bit frees a lot more cycles for PWM.

Looking at the graph of 32 bit fixed point, we see false detection of nose up when it was really nose down. The accelerometers sensed nose up as expected, but the gyros really drifted up like US inflation. Probably time to think about more mechanical vibration reduction instead of software vibration reduction, but it's already time to start driving back to Silicon Valley.

NASA put this out on Friday and then took it offline, so here is our copy. Sort of like watching a snail climb Everest.

Jan 2008 NASA VSE progress (5 min 22 sec)

After this
nugget of information added to the long list of other nuggets, this dog had X33, JIMO, Mars Scout, Moon program II, III, & IV written all over it.

Next, our edit of the Startrek teaser with all the blackouts removed.

Startrek Begins teaser (0 min 32 sec)

...Continue Reading

The weather forecast now says 5 days of dry weather. Flight time is still pegged at 10min. At 1 hour the Taiwanossaurus is starting to feel more natural, but it's still much harder to control the accelerations.

The high accelerations make any kind of orientation hold unlikely. May be time to start using thermopiles.

With the AHRS, the assumption was drift with quaternions is so bad that the deficiency with Eulers would still be better. Well the Euler integrals didn't work very well either, so that eliminates that + the least squares drift experiment.

The styrofoam + plywood electronics tray works well.

Seeded the main PIC loop with PWM calls and that seemed to reduce the servo jitter slightly. The killer with Vicacopter is communicating with the Gumstix.

Fortunately, fire #2 wasn't something expensive. Never found the cause, though it seemed to be a short in the main power cable. It was extracted from a previous robot built in 2002, flew on Vicacopter I for a year, and probably had its insulation breached somewhere. Extra current from the HS-65's may have melted it.

Secondly, a Xena diode failed & turned into a short. May be seeing static damage from styrofoam. That alone didn't cause the fire.

Built a new electronics tray from styrofoam & plywood. This is to prove it works before building an expensive foam + carbon fiber one.

Found a bug in the last Micromag rework from last year and made yet another rework.

Playing with the straight Euler integration reminded us why we switched to quaternions. It works as long as heading is constant, but it's impossible to lock heading on the bench, so it falls over. The gyro drift is much less with this algorithm, we have much better tail authority than the Corona, so it's worth trying to master it in a constant heading situation.

Since the lunar lander challenge didn't appear to require roll changes, Carmack probably had a much easier problem.

Now some bad news. Now that Russian Heroine is gone, probably going to rename her Coptershyna. Vicaglider would become Glidershyna....Continue Reading

Found the cause of the head bogging. The RCE-BL35X does not default to governator mode. It was a miracle she could fly at all.

Governator enabled, tail rotor authority improved & backwards figure 8's became barely possible. The lack of any kind of horizontal drag is killing the accelerometers. Attitude hold using least squares drift was a complete failure. Heading hold, however, was much better than the Corona.

Reducing least squares history & increasing PID gains R still left. The real danger is inversion under autopilot control. With the lack of horizontal drag, AHRS probably needs to be completely inertial. There's no way we could afford DGPS.

Shooting for the lowest RPM in which the tail gets enough power. The lowest RPM's seem to increase tail oscillations.

And the video of Taiwanossaurus's cyclic feedback test.

Cyclic feedback on the Taiwanossaurus (0 min 24 sec)

Well, going over the sales pitch again, the Easystar has no aeleron for controlling bank, so it probably can be flown by GPS alone. The winglets convert slip into roll.

It only needs 2 HS-81's so that leaves only the XBee. Just correlate elevator & rudder to turn & climb rate. Another idea is to use turn rate & climb rate to center the gyros instead of an accelerometer. Is Easy a long term fixed wing solution or is it another Corona, doomed to be replaced by much more complex & expensive solutions?

Well unlike a Corona, U could probably build an Easystar.

Is it worth supporting a more ambitious fixed wing future beyond the Easystar? Well the experienced programmers always say don't overdesign for future goals which may never materialize.

The house Wehrner von Braun lived in from 1953 -
34.729055, -86.56355

which he bought at the age of 41!

Ground view:
http://huntsville.about.com/library/...icvonbraun.htm

As valuable as copters are for squeezing around Calif*, space is owned by the fixed wings.

Most of the spaceflights by people other than CEO's will probably be on autonomous fixed wings, probably in China. The pilot will be in the spacecraft and the fixed wing booster will glide home autonomously.

Maybe spaceships will ride on the routine airline flights. The airline would hop to 50,000 ft to launch the spaceship and continue to its ground destination.

No-one takes NASA and its moon programs seriously. Even China has lately been pushing a space station more than a moon program. The only meaning in VSE or USSEP is that's what the next failure is going to be called.

Imagine being wealthy enough to rent a 2 bedroom apartment, let alone a house by the age of 41.

Finally got good tail rotor effectiveness and without any software heading hold. It is indeed an RPM phenomenum. Not enough RPM and the tail rotor is worthless even if U have enough collective. There's not enough tail rotor pitch to compensate for 13` collective. Furthermore, fixed pitch flying is impossible because of the lack of tail rotor margin.

Current skid mount is proving worthless. Anything but the lightest landing and it breaks off. Need to think about carbon fiber & bigger skid mounts.

Significant head bogging & battery heating. Much harder to fly than the Corona. The Corona would stop rolling beyond a certain point. The Taiwanossaurus keeps rolling.

Need a few hours of basic nose-in landing, nose-in takeoff, backwards figure 8, forwards figure 8, climbing, decending, & dead battery before any realistic autopilot tests can happen. Software heading hold would probably work around the head bogging problems.

Vibration on the IMU appears similar to the Corona.

Last vibration study with the Corona:

https://www.rcgroups.com/forums/atta...mentid=1505552

Well, the BEC handles the HS-65's easily. The inductor still gets hot, but not even close to the LM7805. Took 2 hours to fit the BEC on the main board & rebuild everything and then it was time to start driving back to Silicon Valley.

PWM constants were rewritten to try to reduce jitter. Another round of engine mode tests is required. Then the IMU has to be calibrated because yaw is now 100% software.

Getting both stable placement & cushioning for the IMU seems impossible on a Taiwannosaurus. There's not enough room for foam like there was on the Corona....Continue Reading

Well the local guy had the Jesus BEC for $22 + tax so the choice was easy. Despite being 1/2" thick, it really is tiny compared to the other options. The servo connector shown is not enough for the rated current. Probably need to pile it on the main board.

In other news, Russian Heroine moved on. She started getting more reticent around when her new car arrived in November and finally decided it was over on Sunday. In an area like Rain Ramon with 3 men for every heroine, U can't live for emotional support. After an adjustment phase, our full attention should be back on machines.

So metal gears arrived after only 2 days. Unfortunately, the same story from 1/20/08 happened. Clearly the metal gears needed more current and the LM7805 promptly heated over 200F once the servos started moving. Then it just shuts down and we lose servo authority. Another point in favor of the pilot on the 1/23 crash.

Now alternatives are $25 for a commercial BEC, plug it into the BL-35X ESC, get a separate battery, build a buck converter, or just write blogs about it. Remember what happened the last time we used an ESC for servo power.

According to this story the HS-65's need up to 1.3A so that's 6A, almost as much as a second rotor. At least Hillary's not flying it.

Ultimate BEC 4 U:
http://www.castlecreations.com/products/cc_bec.html

HS-65 metal weight: 10.1g
HS-65 karbonite weight: 9.3g

Should be noted our boss 2 bosses ago spent a year trying to get buck converters to work. They never did really work, and then he quit....Continue Reading

With Vicaglider's GPS going at 4Hz, the PWM resolution dropped to 18usec. U can see the effect.

Servo jitter at 18usec resolution (0 min 10 sec)

That's why we only soldered 1 servo header. A small, second PIC for PWM looks almost certain. Another trick is seeding PWM calls throughout the main loop but that never worked well.

Endless rain & Vicacopter downtime has us leaning towards sooner Vicaglider acquisitions.

For Vicacopter, decided to assume any current GPS position is worthless & start hard coding the altitude waypoint in absolute coordinates. Then, under autopilot it should gravitate towards the hard coded point as GPS improves instead of flying away because the point U thought it was reading was really wrong.

Found the conversion from LLH to meters was wrong. Derived a new one by plotting points in Google maps. That may have been the root of our west pointing malfunctions. The conversion just scales latitude & longitude. U have to recalibrate if U ever leave Calif*, probably never going to happen.

Software yaw stabilizing is a bit harder. U need to get a center stick position & a second yaw PID loop for manual mode....Continue Reading