Jack Crossfire's blog View Details
Posted by Jack Crossfire | May 21, 2009 @ 12:47 PM | 4,674 Views
Spent $60 on an SCP1000 breakout board over a year ago, but the cost of buying something & not using it all this time is less painful than the cost of having to put it back in. It means doing the same thing as everyone else, adding more weight, more wires, surrendering to the new normal of free SUV's & broken GPS.

Our previous data started on May 06, 2008.


Pressure sensors suck. For the amount of taxes U spend on GPS, they're like spending 75% of your income for a car & walking.

The SCP1000 locks up at random. It starts sending the same
reading over & over even though DRDY is toggled. Pretty sure this happened last year, but we never bothered blogging about it. It only makes sense in terms of some electrical interaction on the airframe mounting.

If the budget for a Vika 3 tri rotor ever made it, you'd have 1 micro copter & 2 full size copters to read about. After the sonar derating & the loss of GPS altitude, operating 2 full size copters doesn't buy much anymore. Vika 2 would have to use the Maxbotix EZ1 for altitude.
Posted by Jack Crossfire | May 20, 2009 @ 12:53 PM | 4,584 Views
GPS had a pretty big loss of performance recently & it seems due to the loss of satellite 5 on Mar 26. Your communist government doesn't have enough money to keep all 34 satellites going, so it's paired satellites which it thinks R about to die with newer satellites in the same plane, & that's 5 & 12, 3 & 6, 9 & 27, 7 & 25, 30 & 1. Unfortunately, 5 & 12 was the pair used during our normal flying time.

The effect is greater sensitivity to interference, degraded altitude precision & that means you're probably getting back the SCP1000.

All the media is reporting is "an increased likelihood that in 2010, as old satellites begin to fail" & "the first replacement satellite is expected to be launched this November, some three years after the original launch date." Whatever it takes to keep your mortgages solvent, but we know they're already dying.


While Arrow shipping only took 1 day, the passives R still a week away, so instead, how about the highest resolution photos of the LISY300AL ever shown on the internet.
Posted by Jack Crossfire | May 19, 2009 @ 12:22 PM | 4,292 Views
All because of those gyros doubling in price, now your bank needs double the bailout. Now your government needs to print double the money.

Ghetto IMU damage:

Passive components: $44.21
Gyros: $53.92
Transparencies: $32.76
Draino: $7
Ferric Chloride: $10.96
Photoresist board: $13.16


U know, it would have been cheaper to buy $90 of $parkfun breakout boards. Hopefully this'll come out cheaper in a few more boards. We can now build lighter boards for everything.

The actual cost of ghetto IMU consumables in 5/17/2009 dollars:

Gyros: $30
Passive components: under $1
Photoresist board, transparency & chemicals: under $2

Obviously much more in 5/18/2009 dollars, but at $33 it's not your normal IMU.

Then there's the Feigau spin copter disaster:


How R U supposed to recover that in a tri rotor?

Cost to use 2 more Feigau's + Chinese ESC's in a tri rotor:

Feigau 17g 150g thrust $33 * 2
Micro servo: $20
Tax: $8.6
Supersimple 10A $8 * 2
Chinese shipping $30


Cost to buy ESC's locally:

Feigau 17g 150g thrust, 3" propellers $33 * 2
Micro servo: $20
Castle 9A $25 * 2
Tax: 11.6


Cost to start over with chinese parts:

AD-C5 Micro Motor 4.6g 90g thrust, 2.5" propellers $18.20 * 3
Supersimple 10A $8 * 2
Chinese shipping $30
Micro servo: $20
Tax: $2


Batteries are the same. Hobbycity says those AD-C5's can't make as much...Continue Reading
Posted by Jack Crossfire | May 16, 2009 @ 07:55 PM | 4,663 Views
Exposed it using the 2 transparencies, 26W fluorescent light, distance of 18" for 5 minutes.

Dunked it in the most diluted draino possible. Had just enough draino to wet the bottom of the container & the rest was water.

Fluid flow affects the developing rate. Wherever U grab on the board is where the fluid is most turbulent & that develops the fastest.

As it developed unevenly, dunked just the parts which needed more developing. Eventually should use q tips. Slowly dunked & rinsed to remove the tiniest amounts.

Photoresist stood up better against ferric chloride than it looked. Even though it was faded, got solid traces. Concentrated, 80F ferric chloride was good enough to etch.

Dunk it in a container of water to stop etching. It avoids splashing pure ferric chloride around the sink.

After etching, dunk it back in the draino to remove the last photoresist.

The Nate Siedle technique of running a solder ball over it sort of works. U need to constantly melt fresh solder, let it bridge, then suck the excess. Apply the solder with pulling movements, like a felt pen.

Expect to destroy most of the boards in the draino. This is really a matter of luck.
Posted by Jack Crossfire | May 16, 2009 @ 04:26 AM | 4,584 Views

So the laser printer isn't dark enough. Even 2 transparencies aren't enough.

Did learn U need to dilute the draino 10:1 or it'll dissolve everything instantly. Keep the draino off your skin or else. The best result used ordinary copier paper & 10 minutes of baking. Need to find someone with a super dark printer.

One other thing. No-one sells transparencies anymore because everyone plugs their laptops into projectors. So that's transparencies & inkjet printers in the list of extinct requirements for DIY photo etching. This is yet another one of those myths which could only work for the 1 guy in outer Gainesville who still uses transparencies & inkjet printers for presentations.


Got it locked indoors & swapped the aluminum in & out. Didn't notice a single difference. Between changing satellite coverage, microprocessor interference, & fuselage conductivity, even if there was a change, U couldn't say "Bingo! It's the aluminum."

When the GPS locks, its sensitivity goes up to the maximum, so any change in signal strength blows up.

Note that our uBlox5 lost its configuration after 3 weeks of no flying. Obviously the settings are what's backed by that battery & not the ephemeris. U better have initialization in software.


Ideally U have an SCP1000 away from rotor wash. We just hate copying everyone...Continue Reading
Posted by Jack Crossfire | May 15, 2009 @ 12:22 PM | 4,781 Views
Photoresist board: $11
Incandescent Black light: $5
Tax: $100000000000000000000

$5 incandescent black lights don't do the job. Ordinary 26W spiral fluorescent lights actually seem to work.
Posted by Jack Crossfire | May 14, 2009 @ 01:56 AM | 4,710 Views
After overcoming the shock, here R the mane budget drivers:

12mm brushless motors R too heavy for spin copters without landing gear.

7mm brushed motors don't have enough power for the smallest spin copters.

U don't have anything which can move around using sonar.

U have to recover your investment in 12mm motors & sonar before buying something else.

All roads lead to an indoor, IMU guided, sonar guided, autonomous, tri rotor home movie camera.

Fortunately, take a look at this.

The LISY300AL is mountable on a single side of 0.75x0.75" board. U can probably make an IMU for $30 + chemicals.

Originally got the laser printer to iron on toner masks. That didn't work. Now they're finding U need an inket printer for the photoresist method because the new laser printers aren't dark enoguh. Useless as always, Google.

Now today's video.

External tank with stabilization (1 min 20 sec)

No, they cannot be harvested & turned into a space station. The external tanks fall right back down into the Indian ocean after separation. They are completely drained to the pipes but it still isn't enough to get the shuttle into orbit. The shuttle uses the OMS engines to get into orbit.
Posted by Jack Crossfire | May 13, 2009 @ 03:06 AM | 4,523 Views
Spin stabilization had no way of landing. That should have killed it on day 1. Thought a blanket would be soft enough for crash landings but it wasn't. Too heavy for any kind of crash landing without destroying the propeller. Just didn't see it coming without test flights.

Total weight without sonar: 104g Only 17.5g was our crummy through hole board.

Also, even short bursts of descending power got the motor too hot. Based on motor heating, no power system capable of lifting itself with a 3" propeller is light enough for crash landings.

At full power, got enough thrust to lift itself without spinning. Didn't get enough flight time to calibrate the photodiode, but it appeared to have huge amounts of noise from either propeller reflections or motor induction.

For the spin copter to work would require the same design U see in malls, the giant styrofoam stator to land on & the tiny rotor inside. Otherwise, it would have needed a much larger propeller & much lighter power system to survive crash landings like a bladestar. Fuggedabout a world record.

Spin copter had but 1 purpose, to be the world's smallest UAV. It would have been hopeless to operate. The battery was fixed. It would have needed a pushbroom camera. It didn't excercise all the cost saving IMU measures we developed on Vika 2.

Would never spend $100 on spin stabilization if all of it couldn't be used in a backup. The same hardware can go into a micro tri rotor. It would need...Continue Reading
Posted by Jack Crossfire | May 11, 2009 @ 07:35 PM | 4,357 Views
We blew $100 on a copter so it's time for the 401k blog.

It is cheaper to pay the 10% penalty for early 401k distributions than to pay full income taxes. That's right kids. Calif* income tax is now high enough that U can move to a state with no income tax & do better with the early distribution penalty than U would taking full paychecks in Calif*. Since every other state has a lower cost of living, your tax bracket would also be lower.

It's a stupid idea of course, not because you're ever going to retire, but because paying in cash is stupid. You're supposed to be borrowing home equity. That's what mortgage bailouts R there for. Your government doesn't give free loans to first time home buyers & people with the right skin color so they can buy houses. It gives loans to make your houses appreciate 20% every year so U can buy stuff.


In the future, cash is meaningless. Everything is $100,000 & paid for with home equity. You're awarded a certain size mortgage bailout every year, based on party loyalty, race, skin color, religion, sexual preference & that becomes home equity U can borrow for some electronic root boots.


Calif* feeds on outsourcing. U lose your jobs in Montana because if it can be done in Montana, it can be done in Mongolia. U move to the main office in Calif* which can't be outsourced. The faster the internet gets, the more of U have to move to Calif* to find jobs. The more...Continue Reading
Posted by Jack Crossfire | May 10, 2009 @ 10:22 PM | 3,713 Views
So U want to use timers instead of peak light searches to determine heading. It needs less memory & it could be applied to random pattern searches. The challenge is in #1 the insanity & #2 the memory constraint. Vika 3's fully completed flight computer had only 6 bytes of RAM left.


Another board & resurfacing of the clock jitter problem & it's a bit more robust. U forget the difficulty in synchronizing 2 clocks within .2ms of each other because it normally just works.

We have 5ms jitter in the USB timing. That would cause 65" of sonar position error if it wasn't defeated. But how do U know if your starting times are the 5ms offsets or the true offsets? That's the key with the clock jitter problem.

Crystals have to be hand calibrated a few thousand Hz for best results with sonar. Any mass production would require detecting the true clockspeed & storing it.

Vika 3 flies by measuring time alone. Position is derived from timers. Heading is derived from timers. It's the ultimate Einsteinian conversion of space into time.

Now to compare reality with


XBee with headers: 3.7g
XBee without headers: 3.2g
Main board: 17.5g
Sonar: 2.9g


The avionics came in over budget because of the voltage regulator & the soldering. There was no way a balancing tap hack with resistors would never have a ground loop & explode. Now we just have to burn up $100 on taxes & maybe a motor if you're lucky.
Posted by Jack Crossfire | May 09, 2009 @ 06:50 PM | 3,818 Views
The MKV inspired us to develop 3D sonar navigation. We wondered how that thing could be so stable indoors, decided GPS wasn't the end all, & decided it was probably sonar. After fully funding 3D sonar, discovered it was probably UWB positioning. 3D sonar wouldn't have happened if it wasn't for the MKV.

No surprise that Oboobs cancelled it to fund the welfare revolution. Korea is still developing its own MKV, but they don't have 0% credit cards baby!
Posted by Jack Crossfire | May 08, 2009 @ 12:44 PM | 3,937 Views
20 years ago, we envisioned Copter Soccer, a 3D soccer game using copters. Got it mostly working on a Commodore 64, but it was unplayable. Didn't know enough at that age to control rates of movement. It may still be on a 5.25" disk somewhere under Milpitas.

Copter Soccer could be done in real life nowadays. It would have 2 750mm copters & a micro UAV for the ball.

The ball would have to be an active, GPS guided vehicle to slow it down to a playable speed. It would detect the copters & actively bounce off without touching anything. It would detect invisible boundaries & actively bounce back into play.

The 2 copters would be computer controlled. A human would control the outer loop of 1. The computer would keep them from crossing the invisible boundaries.

Unfortunately, don't remember most of what we wrote back then. The first program we ever wrote was a character piston animation in 1985 in Commodore Basic. Then came a character triangle animation, very hard if U don't know algebra. Then came sprite animations, character set bitmaps, 1 bit audio. Spent a lot of time figuring out the least number of custom characters required to generate the largest number of images.

The C64 wasn't fast enough to do anything but static images in the frame buffer. The frame buffer was 1000 8x8 cells. Each cell was 8 vertical bytes. So the pixel at x=9 y=2 was bit 6 of byte 10. Very expensive to draw any kind of polygon that way, so the only...Continue Reading
Posted by Jack Crossfire | May 07, 2009 @ 12:31 PM | 3,372 Views
Dreamed Bill Murray ran some kind of scam operation in Maryland & he was flying us to NY in some kind of beat up copter to commit some kind of crime. Russian Heroine was there.

There was a lot of running around airports. It was taking some doing to escape from New York after committing the crime. Ended up eating in a restaurant with Russian Heroine while Bill Murray was looking for an escape. What a joy to be back with Russian Heroine.

Then someone pulled us out of the restaurant & left Russian Heroine there. Bill Murray took us to an elevator heading down to the airport.

"I found a way out," he said.
"But what about Russian Heroine?" we said.

"Look! You don't want to waste your time with her. She's just going to
domesticate herself," Bill exclaimed.

Started missing Russian Heroine again, how she would go on a separate path in the restaurant, probably find another way home, completely forget about us.

Then the elevator stopped. The cops were waiting downstairs. The only way out was to go back through the restaurant & wait for the copter on the roof. Found Russian Heroine still there & got her.

Not a literary masterpiece, but it beats the usual copy of a NASA press release all those other blogs have.
Posted by Jack Crossfire | May 06, 2009 @ 12:29 PM | 3,383 Views
8Mhz is not enough to do communication, photodiode, PWM, & sonar. U need a crystal. Didn't get as much code reuse from the tri rotor as hoped. Really hoped to avoid a crystal because PC board area is heavy.

Since most of U owe $50 billion houses, surely U can design an algorithm to determine heading from scan line video.

Bladestar uses a flashing light as a heading reference, but we want to be hands off.


1) Buffer 128 pixels of 1 scan line & send it down the XBee in packets. Drop samples to throttle scan rate.

2) When XBee is sending buffer, don't buffer but keep scanning.

3) Calculate period of a revolution from the buffer.

4) Adjust scan rate until enough revolutions per 128 pixel buffer are sensed.
- Search space of all ADC rates, period sizes for best pattern correlation.


5) Use peak light as heading reference. Microprocessor can sense this easily.

6) Calculate time after light peak to offset throttle phase
- use relative position of light source & copter


5) Use arbitrary light pattern as heading reference.
- must reverse engineer heading with trial cyclic
- assume RPM is stable
- use time of scan line buffer as offset of throttle phase
- measure time as number of samples in software
- assume RPM is constant for a few revolutions after the buffer
- add periods to buffer time to get throttle phase

Not enough memory on the through hole version to do the hard algorithm very easily, but not enough money to start with the SMT version. Use easy algorithm to prove it can get off the ground.

If people stop reading blog, consider a POV led system.
Posted by Jack Crossfire | May 05, 2009 @ 12:46 PM | 3,440 Views
Had the mission control dream. U ever have the mission control dream?

We were manning some navigation console during a day of simulations & Wayne Hale was the flight director. As the simulations began, our console always broke. It had some ancient Apollo era belt drive & the belt was always breaking. We were also supposed to configure tons of parameters in a real short time before each simulation but couldn't figure them out.

As the simulations wore on, for some reason, never figured out how to configure all those parameters in time & Wayne Hale was getting pissed. He kept polling, "Navigation". We'd say no. He'd grumble, "Umm kay," & just continue the simulation without us.

Everyone else was having no problem getting their consoles configured, we were supposed to be the navigation expert but were completely blanking. The pressure of being under Wayne Hale in mission control, on mankind's greatest adventure completely destroyed us.


You ever have the shuttle nightmare? By any chance? It's the one where the astronaut core is running dry because they keep dying, so they draft U into the astronaut core. The day of your launch, you're full of dread, suited up, sitting on a bed in a room with a view of the shuttle, waiting for the weather to improve, looking outside the window at the machine that killed more than all the home owers in Calif*.


Most people in Silicon Valley do 1 thing. They write software to move bits from point A to point B & that's all they ever do.

MIT man says systems engineering is the future of everything, because the problems of the future require all the skills instead of just moving bits. May be true for stuff you buy, but Silicon Valley tends to be a self sustaining economy. The jobs will focus on moving bits long after the rest of the world is moving control surfaces.
Posted by Jack Crossfire | May 04, 2009 @ 08:15 PM | 3,635 Views
The shuttle has delta wings because of the Air Force. It was originally supposed to launch from Calif*, put a spy satellite in polar orbit, & land in Calif* in 1 orbit. If it stayed aloft for more than 1 orbit, the Soviets would have tracked the satellite orbit. Fortunately, it's not that bad anymore.

It is impossible to launch & reenter exactly over Calif* in 1 orbit. Calif* rotates 1500 miles to the East, putting U in the ocean. U need to reenter over the ocean & glide 1500 miles East. If it wasn't for the Air Force requirement, it would have had much smaller wings or parachutes.

The payload bay came from the need to launch a Centaur upper stage. The Centaur was 10x30ft & the payload bay was 15x59ft.

They got all the way to having a live external tank in Calif* before giving up.


Your imagination is writing checks your body can't cash, isn't it. U don't own that credit card, the taxpayers do.

Didn't get very far with integrating delay calculations in the microcontrollers due to lack of processing power, but U could definitely make a sensor network out of many sonar transducers spread around a flying area. They would calculate the delay when they got a ping, so you could use a slow wireless network.


Sonar guided helicopter 4 U (0 min 37 sec)

There it is. A tri rotor flown by only 3 gyros & a sonar transducer.

After 4 days of constant rain & wind, we had a 12 hour window before the worst May weather ever recorded in Calif* continued, so it was now or never for video.

3 systems failed to get that shot. A sonar solder joint broke, the 72Mhz link failed, & the yaw actuator horn broke. U see the 72Mhz fail in the shot.

Just too many cheap systems with too many hours on them in too much wind. After 90 minutes of problem after problem & 10mph wind, it was time to get back in the car & start driving back to the day job.
Posted by Jack Crossfire | May 04, 2009 @ 12:45 AM | 3,966 Views
Because using tri rotor parts in a spin stabilized prototype would be too dangerous & require a new battery, your only choice is to spend $100 on brand new parts & build a complete vehicle. That gets U

FeiGao 5300kV 17g 141g thrust: $33
800mAh 7.4V 40g: $20
10A ESC 7.1g: $20
Tax: $30


A lot of the thrust goes to spinning the fuselage. It might have enough thrust to lift a through hole board & prove the concept before building the SMT board.

xbee: 4g
sonar: 3g
through hole chips: 4.7g
peg board: 8g


Sadly, this brick would never sell. The only reason anyone flies micro copters is money. Given a choice, they'd rather fly a full size airframe for $100 than a micro brick for $100. The Jump Jet is $100 & it doesn't sell because U can build full size tri rotors for that price.


The idea with this was to find the minimum number of parts to do the job, attach it to a drill, spin it around like the real thing, & design methods of optical heading sensing. There's no way this is going to spin on a drill. It's limited by the rate a human can spin around.

It does indeed sense heading in a room with only 1 light source & handles changes in tilt if the light source is a tower. It's too nauseating to do more than 1 test per commute. Need a better way.
Posted by Jack Crossfire | May 02, 2009 @ 09:49 PM | 3,764 Views
Running out of reasons to support sonar. If it isn't practical for large copters, that leaves indoor copters. To justify an indoor copter, it has to be the world's smallest & cheapest. If the motor technology isn't up to it, you're done.

Vika 2 is the world's cheapest 750mm autonomous copter. Never mind that she can't do anything but hover. U can't do better than 3 gyros, a sonar transducer, & a scratch built tri rotor.

Spin stabilization problems:

Payload must be purpose built for spin stabilized fuselage.
Can't fly at all without erecting an LED tower & sonar array.
Altitude limited by LED tower.
Requires pushbroom camera for AP.

Spin stabilization advantages:

Will be world's smallest UAV when motors improve.
Cheapest UAV possible.
Highest indoor maneuverability.
Could actually make money.


It doesn't have a geared lift fan. It's 1 piece spun by the 2 pager moters. Still stand behind the use of throttle modulation to control translation.


flapping hinges damp translation. If it had rigid blades, it would need constant inputs to hover in place. Minimum RPM for spin stabilization, precision of LED heading sensor, & throttle modulation time give it just enough translation authority to get moving.

We have never touched a bladestar. This is just from videos.