Jack Crossfire's blog View Details
Archive for April, 2009
Posted by Jack Crossfire | Apr 30, 2009 @ 04:36 AM | 7,214 Views
It was a bit crazy to replace GPS with sonar, because what U really have is a crummy capacitor sensing position. Designing your own sonar positioning system is like building a GPS system from scratch. It's not much better than starting with a bag of silicon.

After seeing how limited the range was, how bad the gyros were, if U said we'd be hovering a full sized airframe with only sonar & IDG300's, we'd say "My God!!!".

Well here it is doing automated takeoffs, hovers, & landings with sonar & IDG300's. When completely trimmed, in dead calm air, sonar guided takeoffs & landings R a huge improvement over GPS. Unidirectional sonar would make a good takeoff & landing stage for a GPS autopilot. Reflected sonar can't handle the grass & long range, as we showed.

Still have many software bugs. No idea why the wrong position is being tracked. Operator directed position changes don't work either. The best theory is the cyclic gains aren't fast enough to track gyro drift. Other bugs cause her to go into automated landing if U forget to raise the collective. Suspect the idea was to force U to raise the collective in case of emergency....Continue Reading
Posted by Jack Crossfire | Apr 29, 2009 @ 04:58 PM | 7,002 Views
You should have all gotten your social security statements before payback time & discovered U would be getting only $780 of every $1000 of your benefits. What a backwards system in which your social security contributions are more than your 401k contributions & despite being called a retirement plan, social security goes entirely to mortgage bailouts & U get nothing back. This brings us to some more notes.

Space & time R infinite.

Our bodies simultaneously live in an infinite number of different times & spaces, billions of years apart & billions of lightyears apart.

Biology is a game of packetizing the information in the universe so U can't experience the time & space of someone else. The brain with a set of boundaries to its experiences was evolved.

Information can't be destroyed. In a cosmos of infinite universes, the information that defines your existence always exists in at least 1 time & space, no matter how many times U destroy it.

4 U to experience death in your own frame of reference, information would have to be destroyed, but we know information always exists in at least 1 time & space.

If we could experience death, we would already have experienced it. There is always 1 time & space where we already died, but our brains are hard wired to only experience time & space where they are alive.

U never experience death in your frame of reference, but U see everyone around U die. Every marriee can expect to see their spouse die.

Consequences last forever.

U either keep compounding interest & stay retired forever or your government devalues your money to nothing & U have to work again.

Suspect in your frames of reference over billions of years, you'll see the universe keep expanding & scientists will keep pushing back the end of the universe forever. Individual universes may collapse or run out of energy, but U never see it.
Posted by Jack Crossfire | Apr 29, 2009 @ 04:16 AM | 8,093 Views
The answer is yes. U can get 12x31mm brushless motors locally. They're imported from feigao.com. It would cost a lot & not be the smallest, but it would work.

There R no pusher props 3" long, so spin stabilized Vika 3 is screwed.

Have another PID calibration session & your first closeups of Vika 2. It's a matter of finding the 1 spot over the array she can hover best, in addition to the usual constants. Making a computer fly a trirotor with only IDG300's & sonar is almost as mean as making it read the Jack Crossfire blog....Continue Reading
Posted by Jack Crossfire | Apr 28, 2009 @ 02:54 AM | 7,170 Views
Your government has us in high winds until Thursday & another rainy pattern starting Friday, so don't expect any video of Vika 2. In 20mph winds she just flew away, but had 2 short breaks in the wind long enough to get some photos. These flights were without GPS aided heading & were enough to show the IDG300's R hopeless without GPS aided heading.


The mane problem with Vika 3 is the compass. The flytech works because a human always points a controller at the vehicle, providing a horizontal light source to determine heading from. The best Vika 3 can do is a stationary, omnidirectional light source on the floor, illuminating the underside of the fuselage. Sensing ambient light requires AGC & U know how AGC went.
Posted by Jack Crossfire | Apr 27, 2009 @ 12:45 PM | 7,198 Views
3 months of chip flashing, USB hacking, & accoustics reviewing finally paid off with Vika 2 hovering herself. Pretty sure 3D sonar has never hovered a copter successfully before.

Vika 2 did what she was designed to do & proved an autopilot using only 3 crummy gyros & 1 sonar emitter was feasible. It doesn't have enough range to do any of the camera movements we envisioned, but it's definitely enough for a micro copter.

She did indeed use position sensing to correct heading & that actually worked most of the time.

Unfortunately, was so pessimistic about this system ever working, no cameras were present, but we do have some Goog plots.

Mind U, it's a bit much to expect a very large mass with a very slow control system to stay in a 4 cubic meter box directly over the sonar array. When it works, the weather cooperates & everything is aligned, it works better than uBlox5. Did a 6m30s hover & several 3 minute hovers.

We have $40 invested in sonar navigation, so we're still $1930 ahead of Evolution Robotics. As long as U stay in the sonar cone, VikaCopter is the cheapest indoor positioning system & not Northstar.

Because of their 3 lobe cardioid pattern, running the Matsushitas with the shouds on was the lesser evil than losing range. After replacing the shrouds, some brute force hard coding, integrals, smoothing, & reduced expectations, got some pretty nihilistic improvements. Integrating gets U more range. Smoothing the integral gets U the 1" resolution.

Time to begin Vika 3.


3D position sensing using sonar (1 min 0 sec)

The very first decent results using sonar for 3D position sensing. The tests were hand held in a 4 square meter dumpy apartment & scaled up for rendering on Goog Earth.

1:10 Scale Saturn V with stabilization (1 min 36 sec)

The 1:10 scale Saturn V U all forgot about because it's 5 hours old has been stabilized.
Posted by Jack Crossfire | Apr 26, 2009 @ 05:08 PM | 7,199 Views
Take just enough courses to qualify for loans & borrow as much as you can. Use the money to build a business or build something that could be used in a business. Ideally it's part of the degree, but most of the time it can't be & you're more motivated if it's not.

That's how Google, Netscape, Microsoft, Sparkfun & most famous business got started.

In centrally planned economies like US, the money always lags the value. The money goes to what was valuable 10 years ago while the people really want something newer, from China. So instead of waiting for all those congressmen to catch up with what people want, U take those student loans & build what's valuable now. Then when the economy catches up, you're waiting in the door.

Now student loans R free & you have a new age where it's government's responsibility to pay off all debt. Using taxes to finance all debt is going to be as basic as using taxes to fund roads. U need to borrow like no tomorrow.

U think those people working make any money? Of course not. They grow their cash balances by 10% a year if they're lucky & watch the cost of living rise 20%. Meanwhile you're getting student loans for 1% & watching the value of debt fall 20%. You're not making money if you're not borrowing.

Circuits that don't work

Before U completely fall asleep from that, Vbus works in the following configurations.

All cases with sonar plugged in.

Sonar not plugged in & not compiled for sonar.

After busting an RF choke & a diode, decided the problem with Vbus is a low frequency ripple that can only be defeated by a voltage regulator, but since we need 5V, it would take a very expensive DC-DC converter followed by linear regulator. The sonar probably works because it grounds everything to Earth.

The only concern was if a change in the sonar wiring broke this strange behavior or we wanted 1.6mbits without sonar. In the mean time, U've got the comb filter to reject low frequencies.
Posted by Jack Crossfire | Apr 25, 2009 @ 04:47 PM | 7,821 Views
China launched another spy satellite for growing food.


India launched another spy satellite for Israel.


Russia launched another spy satellite for Italy.


US announced its moon program has been... wait for it... delayed by budget cuts.


Ares 1 was reduced from 6 seats to 4 seats to reduce parachute costs. When asked how US would meet its space station obligations, NASA said that's what Russia is for.

Posted by Jack Crossfire | Apr 25, 2009 @ 12:50 AM | 6,809 Views
According to your government, the wind was 20mph, gusting to 29. By the time we got the PID limits increased to where she could handle it, there were only a few minutes of battery power & it was dark.

It was the PID limits & not the PID gain that did it.

In this wind, with the shims reducing vibration, the low RPM motors, U couldn't hear Vika 1 at all.

Didn't bother facing away from the wind.
Posted by Jack Crossfire | Apr 24, 2009 @ 01:19 PM | 6,955 Views
So got Vika 1 all buttoned up for flight, moved all the gear down to Silicon Valley for some daylight photos, & sure enough, the ground controller was dead. Couldn't get it to show on the USB.

Drove all the way back to Rain Ramon & it worked on 1 of 3 AC adaptors & no batteries. So instead of trying to figure this one out, we went to Vbus & it worked.

The answer is yes. The USB has enough juice to power the PIC at 48Mhz, the XBee, the GWS Pico receiver, & the sonar. U just saved $100 on another Lipo.

VBus is really noisy, so you can't get much higher than 256kbits without serious filtering. Decided to try that denoising RC circuit the Goog tends to spit out & it didn't work. We'll be back to what we already know in future commutes.
Posted by Jack Crossfire | Apr 23, 2009 @ 08:18 PM | 6,762 Views

Most LiPos last longer if you land & let them cool off repeatedly, but with some old batteries, they can't fly at all after being used for anything. Even if they only powered a ground station or only flew half their capacity, for the next flight, they're completely dead.

They seem to require a higher temperature to handle current when not at full capacity & flying keeps them above that temperature. If anything puts them below full capacity & they get cold, they can't generate enough current to fly anymore.


U already know how day jobs R going away & being replaced by 1 man online corporations selling limited edition refrigerator pens. Well, what happens when everyone loses their jobs & U have 500 million limited edition refrigerator pen corporations? Are day jobs going to be the new way to get rich?

There's only so much 1 man corporations can do. Once U get beyond the limited edition refrigerator pen, it gets harder & harder to produce something substantial without investment capitol, employees, management of people, accountability, wage slavery.


Managing people sux. Your boss is miserable. U think he treats U like s**t because he's happy? He has to manage idiots who aren't smart enough to be managers & his resume has to say he only managed losers. What manager would say he managed the best? That would mean his subordinates were better than him.

U always think when U get into management, you'll be powerful, rich, & famous. Then when God finally comes down & ordains U a manager, guess what, U now have to be a loser in 3rd person.
Posted by Jack Crossfire | Apr 22, 2009 @ 03:19 AM | 6,604 Views
So, decided to forget about crosstalk & go up to 227khz becuz the Matsushitas have lots of destructive interference & nodes. Trying to do 2 things simultaneously in using a hardware ping generator that can't ping shorter than 0.013 seconds & a there's a standing wave shorter than that ping.


We point 2 point solder because it's free & instant. It would take several revisions & several weeks to get a PCB right. Several months with batch PCB. One question is, with so much historically automated manufacturing now using manual labor in China, will P2P soldered circuit boards return?

Human labor would have to become cheaper than the cost of large numbers of PCB's. It can become cheaper through a sharp rise in oil costs. Petroleum is the basis of all materials & mass production.

Look, there's going to be an oil embargo. U can't expect OPEC to keep making small cuts & holding out for later. They're going to shut it down & demand price controls. You're going to have smaller numbers of each thing being made because raw materials & automation R going to be too expensive. When oil goes higher, humans get cheaper.


You'll be happy to know Calif* was once again as hot as it was in April 1989, so the inland urban heat islands didn't enhance the seabreaze & freeze the day job area as feared. After a scorching 1989, the spring of 1990 was also very scorching, with romantic warm nights, until Mt. Pinatubo erupted in 1991 & froze U solid. The Pinatubo dimming caused a rare rainstorm that summer.
Posted by Jack Crossfire | Apr 21, 2009 @ 03:18 AM | 10,546 Views
There U have it. The fabled Matsushita 0D24K2. The driver is the same size as the Maxbotix, but it has a huge board to resonate 1/2 the frequency. These R the kind used in super expensive motion detectors.

More importantly, finally got the transducers to take op-amp output. Had a lot of theories on why this didn't work, ranging from extremely low impedance to frequency response, but it was dead op-amps. That gives U 20V peak-peak on the emitter.

8V p-p

Is what U get from the 5V bus. Works best within 0.25khz of 24.26khz. The shroud moves that to 24.50.

20V p-p

Is what U get directly from the LiPo. They can go from 24.74 to 23.35khz without shroud.

The Maxbotix's only worked at 38.45khz so U can't combine a Maxbotix & a Matsushita to save weight. No doubt the Matsushita's at 20V R much more powerful than the Maxbotix's at 10V, but you're still a few commutes away from flight test #1.

TODO: Try mounting the emitter a few inches below the fuselage.


If nyquist < signal freq, alias = samplerate - signal freq
If samplerate < signal freq, alias = signal freq % samplerate

With the draconian optimizations, we now capture 3 channels at 66khz. Top speed is 73khz per channel with crosstalk.

Like everyone else, we're probably going to forget about Facebook just like U forgot about myspace & the Jack Crossfire blog. Going to skip Twitter & go straight to Loaner. Loaner is neither single sentences nor blog posts. Loaner has no content. U get paid millions of dollars for future content.


All the friends on facebook are based on real events, not events on social networks. So who's creating current events to base future friends on?
Posted by Jack Crossfire | Apr 19, 2009 @ 04:35 PM | 6,345 Views
After banging on reflected sonar enough, here R your notes.

Get ready for 2ft of deadband if U want to use direct amplitude measurements.

U can reduce deadband by using integrals because they measure the total power of 2 overlapping waves.

Integrals require very high samplerates & reduce the software samplerate to 120khz, so they can't be done in software.

Use a lowpass filter to make a threshold. Hold the transducer at +0V for 1/2 of the deadband, then float it for the rest of the deadband to shorten it.

The shroud made no difference.

A unidirectional system just needs the flight computer & 2 op-amps to generate pings.

Vika III will be a spin stabilized, dual lift fan with no IMU, 1 LED for operator orientation & 1 photodiode for heading. If U stop raising taxes, she may get a pushbroom camera.

We're back to 3 transducers on the ground & much lower expectations.


In the old days, it took years to forget about heroines. Then we met lots of married men & realized it sucked. After Russian Heroine, it's probably back to taking years. There R good days & bad days. One day you're productive. The next day a strange memory or a new nerve connection turns U into an American.

We're at the age where U can expect to fall in love 1 more time, but it's going to be a wrinkled prune. Going over facebook, either all your former heroines became housewives who stay home & update facebook all day...Continue Reading
Posted by Jack Crossfire | Apr 19, 2009 @ 02:19 AM | 6,355 Views
That's right kids. After saving us from OPEC & inventing Sissel, Novay finally pulled off its next hit for us. Now if only Europeans could give up their corny video effects.

It flies itself indoors. It probably uses RTK GPS. It definitely uses actuators for cyclic.

PD-100 Prototype - Mission (1 min 32 sec)

PD-100 Prototype - Flight testing (2 min 28 sec)
...Continue Reading
Posted by Jack Crossfire | Apr 17, 2009 @ 05:12 PM | 6,481 Views
After watching videos, we can tell U exactly how the Wowwee Bladestar works. Its CG is slightly off center, causing it to lean forward if the fuselage is stationary. The fuselage naturally rotates real fast to neutralize the CG & keep it vertical. By pulsing the counter torque fans, the fuselage can hesitate at 1 point in the rotation & cause it to lean in 1 direction.

Avoiding obstacles requires merely tying the fan voltage to the IR reflection voltage. When the IR reflection goes up in the turn, the fans speed up, causing it to hesitate at just the point in the turn where the IR sensor voltage is maximum. That makes the fuselage lean away from the IR sensor.

Now what if instead of counter torque fans, U had counter rotating lift fans that pulsed to make an unbalanced, rotating fuselage lean. What if U went all the way & put GPS on it for position & used a visible light sensor on the spinning fuselage to detect heading by making a 1 dimensional 360` image of the surroundings.

U think U can handle it?
Posted by Jack Crossfire | Apr 17, 2009 @ 03:17 AM | 6,257 Views

Well OK, would probably get more of the aerodynamics lectures if the aerodynamics course was also available. All we do is strap together motors & use software to make it fly, so no-one in this blog ever cared about aerodynamics.

He had some nuggets about how they could fit together baked equations for cones, spheres, & wedges to verify the CFD results.


Got the go ahead to begin reflected sonar altitude. We're only deep into reinventing a Maxbotix from scratch because of the random trigger latency in the product. It's pretty rough because it's using a shroudless transducer, 260khz sampling, pure digital processing, raw amplitude.

Pretty confident the Maxbotix's have enough range to fly a micro copter indoors but not enough range to fly a large copter. U need to show a 3D position with reflected altitude before any money for Vika III starts flowing.

It would cost $50 to build Vika III's airframe, hard wire it to Vika II's proto board to test motor control, but not fly. Then it would be $100 to build a SMT board with the last XBee, which could fly.
Posted by Jack Crossfire | Apr 16, 2009 @ 12:33 PM | 6,303 Views
Reynolds numbers 4 U

Not that anyone outside N. Korea ever listened to them, but we found the 2 aerodynamics lectures pretty useless. Don't think either the students or the speakers had a clue what the speakers were saying. Sounded like the shuttle didn't have any unique aerodynamic problems.

More sound resonance

Seeing huge drops with the tiniest smidgen of offset from 38khz. Such a sharp bandpass response is a crummy situation in the omnidirectional world.

Fortunately, your mortgage bailouts left us enough money for some Matsushitas. The Matsushitas R enormous 1"x1/3" devices & resonate at 24khz, so definitely more hope for range with the large airframe.

The amplifier degrades the radiated power from a Maxbotix.

The western bloc

In the western bloc, April 15 separates the men from the heroines. After inflation & tax increases, we now have less income than we did after finishing college but our boss made enough home equity since Tuesday to build an island.

The Goog loves to say "but outsourcing is lowering prices" & follow it with a mortgage advertisement, but if the price of labor in China rises 50%, is that deflation?