Jack Crossfire's blog View Details
Archive for November, 2007
Posted by Jack Crossfire | Nov 30, 2007 @ 03:10 PM | 6,165 Views
So much for no Vicacopter money in November. In November, $296 was spent on her. $195 alone was spent upgrading the GPS module. The other $100 was repairing crash damage, the remote control switches, soldering supplies, building materials.

In October, $225 was spent on her. That was just the IMU upgrade.

In September, $113 was spent on her. That was the failed EB-85 GPS upgrade.

Fall was a particularly expensive Vicacopter period. The first successful autonomous hover was the main driver, because the system was suddenly proven enough to justify more investment.

In November we were suddenly paying $300 for food, $100 for PG&E, $300 for gas, another $100 for rent increases, and $3000 rent became standard in nearby cities. Sparkfun, Maxamps, & UPS raised prices significantly. There was no incentive to save money.
Posted by Jack Crossfire | Nov 30, 2007 @ 06:13 AM | 6,129 Views
More effort to improve GPS reception is justified. Attaching the backup battery is the next strategy.

After another horrible test flight, the GWS receiver failed completely. Could intermittently revive it by pressing on the right component, but never make it reliable enough to fly. In keeping with our long term plan for XBee communication, it was time to put it down.

Looking at the budget, did much worse in 2007 than 2006. The end of 401k benefits, runaway rent, food, energy inflation, and addictive robotics hobbies were devestating. BTW, Digikey now charges sales tax in Calif*.
Posted by Jack Crossfire | Nov 29, 2007 @ 04:23 AM | 7,000 Views
When the $30 ANT555 arrived after 4 days of shipping, the uBlox started working despite the proto board. It probably needs shielding but the proto board is enough for rudimentary accuracy.

The antenna enclosure was 20g, but sure enough our $20 Chinese scale stopped working. A new record even for China.

Well, the uBlox only gets 8 satellites in the best of times. Nowhere close to the 11 satellites the EM406 routinely got. It took 40 seconds after every reboot for it to lock on outside. It did lock on once inside, after several minutes. The uBlox has very low sensitivity until it locks, then its sensitivity jumps.

The first uBlox test flights were pretty awful, but she did go autonomous for 9 minutes. There was a definite decrease in the response times. The oscillations definitely oscillated faster and she seemed to follow some operator directed waypoint changes better than the EM406.

We have another round of PID loop & neural network tuning. Maybe even getting rid of the neural network....Continue Reading
Posted by Jack Crossfire | Nov 28, 2007 @ 02:33 AM | 6,557 Views
Realizing that handling the high frequencies of GPS is probably going to require printed circuit boards, decided to finally try the old ironed laser printout trick.

Well, after carefully scrubbing the copper clean, pressing the iron down, soaking the paper, and wiping off fibers until most of the traces were broken, decided this trick must have come from the same internet that produced the brilliant GPS antenna tricks.

Suppose if U bought exactly the right batch of glossy cotton bond paper from Staples during the right phase of the moon, use the right HP Laserjet toner stock and model printer from year 2 of dot com boom #1, and lose your hair pulling out each fiber under a microscope, it might work, but it looks like expresspcb owns us.

Now we're going to get a million spams for PC boards made to order from northern China in 4 weeks for twice the price of rent, but we asked for it.
Posted by Jack Crossfire | Nov 26, 2007 @ 04:35 AM | 6,890 Views
According to the datasheet, the amplifier in the uBlox does 16dB. The
gain in an active antenna is 27dB. The fine print on the uBlox says it's 2dB less sensitive without an active antenna. Certain body positions get more satellites than others.

This nugget http://docwiki.gumstix.org/Frequentl...stix2#Antennae says the uBlox requires an active antenna and the one we got is a complete loss.

For the uBlox to give any usefulness on a passive antenna, it obviously needs a very big passive antenna above 99% of the atmosphere.

All roads lead to another Sparkfun cash infusion. In the interest of avoiding an inductor & capacitor network, the VCC_RF pin looks intriguing but there's no information on how to connect it. "Output voltage RF section" sounds like it has isolation. Guess we're done with position hold until December.

A $150 Gumstix GPSStix would have done everything we need. Despite that, all devices taking the signal from an active antenna need to be shielded or far enough from the antenna to not jam it. That's another problem.

In the midst of the GPS tragedy of 2007, it's time to wallow in self pity and upload GooTube videos of better days.

First day of 2000 (1 min 17 sec)

Construction timelapse in FL 2000 (3 min 28 sec)
...Continue Reading
Posted by Jack Crossfire | Nov 23, 2007 @ 07:27 PM | 6,630 Views
So U can get GPS satellite positions from http://www.nstb.tc.faa.gov/Full_WaasSatelliteStatus.htm

Getting the antenna to work is a real buster. It hasn't worked at all since Wednesday. Have some conclusions on the homemade antenna blogs:

1) No-one writes a blog about it when it doesn't work.
2) The ones that work are rare cases of luck.
3) None of the posters ever gets what they read to work.

Can plugin the old EM406 and it reports 12 visible satellites indoors from the Indian ocean to Argentina.
Posted by Jack Crossfire | Nov 22, 2007 @ 05:45 AM | 7,251 Views
Good news: removed the SMA cable and soldered the uBlox directly to the passive antenna section. No tank circuit or anything. Got 4 satellites indoors, 3D position and velocity at 4Hz.

Bad news: Like the EB-85, there's no capacitor to retain settings between power cycles and it can't acquire indoors. U have to go outside for it to acquire, then go inside.

Finally, the answer is yes. The PIC oscillator was kicked out in the crash and forgotten in the golf course for 3 days, getting run over by lawn mowers, sprayed on by sprinklers, and stepped on by middle managers. We'll see how accurate it still is in the air.
Posted by Jack Crossfire | Nov 21, 2007 @ 02:04 PM | 6,445 Views
So the uBlox module & antenna arrived. uBlox charged $38 in shipping. Unfortunately the antenna requires a tank circuit to supply power to it. There are anecdotes of running these without power but we haven't succeeded. Tank circuits are notoriously hard to get working, so this $180 paperweight is going nowhere for a long time. Did manage to get a time of day out of it, using a 22pF + a random inductor. Since the uBlox has an amplifier already, it would be best to run it without an active antenna.

FYI, the uBlox is based on an Atmel GPS engine.
Posted by Jack Crossfire | Nov 20, 2007 @ 07:10 PM | 7,899 Views
The 4Ah 11.1V we got last year for $94 is now $114. The 3.3Ah 11.1V for $84 in July is now $94. Annually the MaxAmp batteries are going up 20%. Combine that with Sparkfun's 10% price increases and U need to think in terms of buying now and installing later.
Posted by Jack Crossfire | Nov 19, 2007 @ 04:28 AM | 6,727 Views
So had another test of nose cam underway to document that mount in flight and test higher shutter speeds. Documenting auto hover requires low altitude. Couldn't get any auto hover of any kind going. Then she drifted too far away at low altitude. Seemed to lose radio contact momentarily, she pitched way way down and could not recover. Managed to kill the throttle but the crash was the hardest ever....Continue Reading
Posted by Jack Crossfire | Nov 18, 2007 @ 03:30 AM | 6,657 Views
The answer is yes. U can strap on a 40,000 candlepower light to Vicacopter and have the most expensive flare ever made.

The value in this is stationary lighting from any position for outdoor video and blinding criminals. Because it's very hard to fly with a super bright light attached to your copter, autopilot is mandatory.

Previous research showed the current drawn by the light degrades flight performance even more than it already is, so you need a lighting battery in addition to the flight battery.

So how much will the lighting package cost?

Lighting platform: 150g
3.3Ah lighting battery: 270g

Total: 14.78oz

A new record when combined with the autopilot....Continue Reading
Posted by Jack Crossfire | Nov 17, 2007 @ 08:13 PM | 6,797 Views
Not enough light for good results, but U can see the effect of vibration on the Aiptek's rolling shutter. Nothing to photograph pointing Northwest, but at least managed autonomous hover pointing Southeast.

Now the weight of the complete autopilot system: 7.5oz

Electronics board: 109.6g
GPS: 23.7g
IMU: 32.4g
Bolts: 8g
Electronics tray: 38.6g
Posted by Jack Crossfire | Nov 17, 2007 @ 06:03 AM | 6,576 Views
Well finally put in the solid 3 hours required to straighten the main shaft. The shaft was 1mm off center, obviously the result of hours of unbalanced heads. The only way to straighten it is muscle & pliers. Got it really close but it can never be perfect.

The IMU saw huge reductions in vibration but the HD cam can always be counted on. Still worthless video. Vicacopter never got vibration free video unless the camera was on a mount so heavy it soaked up all the vibration. Such a mount is too heavy now that the autopilot is on board. Looks like HD is out with this copter.

She pulled off a Northwest & West facing hover again. Never found a smoking gun on that one. It's probably a matter of difficult wind direction, magnetometer calibration, orientation drift. From Nov - May most wind comes from the Southwest. From June - October it comes from the Northwest.

The wind problem should go away when the uBlox arrives. Nuked the credit card on the uBlox ahead of schedule. Unfortunately Switzerland is on another vacation.
Posted by Jack Crossfire | Nov 16, 2007 @ 05:31 AM | 6,721 Views
So $30 & 1 week later, the scale arrived for blade balancing. Had .3g of imbalance. Did balancing improve vibration? Of course not. Now we're down to straightening the shaft, probably not possible beyond its current straightness.

Now some useful weights.

Canon TX1 + shock absorbing mount: 12.72oz
Canon TX1 + nose mount: 10.54oz
Aiptek + nose mount: 2.23oz
3.3Ah cube battery (MaxAmp): 9.07oz
3.3Ah rectangle battery (Hyperion): 9.49oz

Now a word from Mike Fouche himself:
> We first experimented with a Garmin GPS unit with a 1Hz update rate.
> We learned - the hard way - that it really had a 1/2 Hertz update rate
> (updated once per second but the data lagged by 2 seconds). Since we've
> gone to the Novatel OEM4-G2 GPS boards (20 Hz update rate) our
> navigation screams.

So it's a kind of miracle to be flying on 1Hz GPS at all and further work on it is a waste of time. After 2 days of stable air between storms, strong winds are forecast indefinitely now as the Aleutian low has finally arrived for the Winter.
Posted by Jack Crossfire | Nov 15, 2007 @ 02:52 PM | 6,535 Views
So the magnetometer was significantly out of calibration again in the X & Z directions. The Y direction was untouched. The dumpy apartment is packed with very large magnets from our magnetic levitation days.

uBlox is supposed to have a new GPS module in December. This one uses 50 satellites, a factor which will be useful when Europe has its own constellation. Accessing the European constellation requires the $120 module.

Looks like success with the 0.003 accelerometer weighting may have been luck. Need to focus on acceleration from GPS.
Posted by Jack Crossfire | Nov 14, 2007 @ 06:33 AM | 6,782 Views
According to the flight recording, the actual highest point yesterday was 250ft AGL. Only 327,350 left to reach space.

So the lowpass filtering was once again moved to the PIC, 8" of trim tape was stuck on the flybar, and results were better. The gyro drift seems lower than its ever been. We actually did a very slow 360' pirouette in operator directed mode. She still had trouble pointing West but she didn't lose it.

The HDCAM can always be counted on. Still major malfunctions when facing West but South is quite stable. Slight improvement in vibration. The first 21 seconds of today's movie were at 147ft AGL. The last 34 seconds were 131ft AGL

Vibration with balanced flybar, PIC lowpass filtering (0 min 56 sec)

The textbook blade balancing procedure is worthless. It's best to just add tape randomly and spin up the motor to watch for vibration. The head speed may be too slow to ever eliminate vibration.
Posted by Jack Crossfire | Nov 13, 2007 @ 06:00 PM | 6,961 Views
Well OK. Past performance is a good indication of future results. Thought the higher shutter speed in daylight would eliminate the vibration we saw at night, but it didn't. The only difference is in daylight the vibration has a beat frequency U can use to get sharp images in between the smeared images.

Today her weak heading was West. She could not maintain orientation at all when pointing West. Suspect different headings cause different vibration modes because of where the wind is hitting the blades. Some vibration modes cause more gyro drift than others.

Position hold was once again pretty good against actual air movement. She stayed in an actual ballpark using 1Hz GPS. After verifying bug free hovering at low altitude, commanded the autopilot to 200ft for the first daylight view from real altitude. Timelapsed the climb to altitude for Goo Tube.

HD Cam onboard autonomous copter (2 min 38 sec)
...Continue Reading
Posted by Jack Crossfire | Nov 12, 2007 @ 05:11 AM | 6,224 Views
Managed to revive the Micromag3. Looks like the PORTD interface on the PIC completely died. The only explanation is debris was pulling SS on the mag high, which caused PORTD to die trying to pull it low.

In any case, the voltage upgrade and a new IMU sampling rate of 90Hz finally got us hovering autonomously again. She even did a waypoint store + recall. Given our lack of space, it amounted to rotating 180' and flying 5 feet.

A new behavior for the waypoint switch is now installed. In waypoint store mode, store a new waypoint when autopilot is engaged and whenever operator directions happen. In waypoint recall mode, fly to the waypoint but don't store new waypoints at all.

It's only 1 waypoint, but when U only have 13 minutes of flight time, 1 is all U need.

Our biggest goal was finally achieved. She finally hovered with the HDCAM on autopilot, facing South. 7m15s of autonomous hover to be exact. There was too much vibration for any night time shooting but maybe it's good enough for a daytime attempt.

We're now using a 4 foam wheel camera mount. It's the most successful at damping vibrations of any mount.

Unfortunately she didn't do so well in an operator directed turn.

Successful autonomous hover with Canon TX1 (0 min 40 sec)