Shop our Airplanes Products Drone Products Sales
Jack Crossfire's blog View Details
Posted by Jack Crossfire | Jun 11, 2009 @ 12:33 PM | 4,968 Views
Going over the Walmart specials again, a spin copter would take buying lots of toy copters & finding out if any have enough power. There's no other way. $100 was the price of finding out 12mm brushless motors R too heavy to land.

The Havoc Cyclone looks like it might work if chopped up. Lift junk hardware first. Etch a single sided board with no voltage regulator. Single sided is lighter than double sided. That's July's budget.

Our double sided boards cost 2x as much & weigh 2x as much because they need 2x the etching board & the vias R too big. Their only advantage is the form factor & the wow factor for our Iranian readers.

SSME nugget 4 U

The answer is yes, the SSME has 1000 tiny tubes on the inside of the nozzle, not the hatbands U see on all the media photos. The outside rings R called hatbands & just keep it from falling apart. Didn't know this, so didn't get a high resolution photo of the tubes. Only got a wide angle photo of the nozzle.

The tubes R 1/4" in diameter on the aft end & shrink to pin size as they approach the forward end. Liquid hydrogen flows from the aft end to the forward end to keep the nozzle from melting. The tubes R braze welded to the nozzle wall. Finding leaks is real hard.
Posted by Jack Crossfire | Jun 10, 2009 @ 03:30 AM | 3,190 Views
The answer is yes, etched boards can fly a copter. Her only major problem is a nasty throttle falloff which happens between every ESC reboot. It doesn't matter what max throttle is set to. The only explanations R PWM glitching high, rate damping not clamped.

In case U don't know what standby power on a sonar guided copter is for, the IDG300's normally don't initialize properly, so once U get them going, U need to keep them on. The IDG300's also put out a lot of heat, so it stabilizes the IMU temperature.

Flying on sonar again is a perfect time to remember Jim Anglesey who died of Leukemia in 2005. We wouldn't know as much about audio if it wasn't for Jim. We would know something, but the whole sonar navigation idea may not have happened. So when you're watching Vikacopter flying by soundwave sensing, there's a bit of Anglesey in there.
Posted by Jack Crossfire | Jun 09, 2009 @ 04:19 AM | 3,195 Views
The answer is yes. U can etch a PCI card if U have enough money & R obsessed enough. There's a pale fat guy in Thousand Oaks who probably did it. Not sure what U would do on PCI anymore. It's not portable enough for videography. No video U get in the apartment is uncompressed.

Board etching would allow sonar samplerates in the 500khz range instead of the current 180khz. Not sure there's any more data to extract from the transducers.

Future Vikacopter autopilots will be double sided. Vika 2 will get extra 12V pads. Standby power is a mess because of the lighting. The solution is to power lighting from the main plug instead of the autopilot pads. Main diode can't handle the Vika 2 load either.

gEDA can do double sided components. The trick is to toggle the "far side" layer before picking components from the library. They're always placed face up on whatever side is visible.

It's day 3 of Vika 2 configuration but month #3 of sustained 20mph winds in the midnight hours.

The IMU needs more padding & another round of rate damping configuration. That's why we don't mount them on main boards, U all in one freaks.

Choked up tail horn seems to be a disaster. Go back to outer hole & throw in more right rudder.

Throttle range needs calibration.

Nose mounted sonar is getting destroyed by landings but range seems improved.
Posted by Jack Crossfire | Jun 08, 2009 @ 12:05 AM | 3,552 Views
So the board gets too hot with the padding. U couldn't have the IMU on the same board. The PIC locks up again. PWM doesn't work again. U need a full radio test. All the telemetry is coming in at least.

Every time U build a new computer, U think it can fly in 1 day when in reality 40,000 lines of code & dozens of systems have to be calibrated & work perfectly.
Posted by Jack Crossfire | Jun 06, 2009 @ 04:43 AM | 3,282 Views
Just when U thought U had a temporary through hole aircraft to fill the blog posts from now until Monday, the miniature parts arrive early. So that means you're ripping up yesterday's work & installing new electronics again.

Now look at the damage:

4 gyros: $53.92
More fleet upgrade parts: $66.74
Gyro & fleet upgrade parts: $44.21
--------------------------------------
$164.87

That's surface mount commitment.

Now the etching consumables:

Etch board #2 & flux remover: $21.93
Transparencies: $32.76
FeCl: $10.96
Draino: $6
Etch board #1: $13.16
Black light disaster: $5.67
--------------------------------------
$90.48

$255.35 in total for fleet modernization & Vika3. It better have an advantage over any Arduino or gyro breakout board U could buy.

Arduinos: 3 * $33.95
LISY300AL board: $30
IDG500 board (1x1 in): $75
No-one sells a 3D sonar board yet.
--------------------------------------
$206.85

The IDG500 board doesn't meet the size requirement for Vika 3. This number doesn't include tax & shipping. We're probably breaking even by etching.
Posted by Jack Crossfire | Jun 04, 2009 @ 03:10 AM | 3,268 Views
In case you're wondering, the IDG300/LISY300AL IMU does work. It might be more stable than previous IMU's because of the improved soldering. Gave up on the gyro Vref output because it was seen as 1 more source of noise in addition to Vdd & Vss.

Mane problem is the uBlox4 still can't handle the 25Mhz clock. Probably going to be into sonar until the Vika 2 autopilot is built. That one can be moved around & shielded.

While simplifying the Vika 2 autopilot & finding more bugs, discovered modern autopilots R just microprocessor breakout boards. It would have been a lot cheaper to buy Arduino stamps for weight reduction than to etch all these new boards. The stuff that makes it an autopilot hangs off the board either way.

Most Ardupilot buyers probably use it as a microprocessor breakout board & not an autopilot. Except for the strange failsafe circuit, there's no active component which is specific to an autopilot.
Posted by Jack Crossfire | Jun 03, 2009 @ 03:47 AM | 3,258 Views
Since your knowledge is doubling every week from reading the Jack Crossfire blog, U might as well figure out this week that U could have used the IDG-300's U already have & your heat gun reflow oven to make 2 flat IMU's & save a few hours of fabrication & a few sq ft in payload. Didn't think of that one either.

IDG-300's for pitch & roll don't matter because those axes don't care about drift. Only the yaw is drift sensitive.

The heat gun is really a reflow oven in your hand. It heats up fast. It solders & desolders anything that doesn't blow around. It heat shrinks tubing. It can solder QFN's in a couple minutes. It probably can do BGA's.
Posted by Jack Crossfire | Jun 01, 2009 @ 03:29 AM | 3,309 Views
In the interest of increasing flight time & justifying the cost of board etching, it's time to modernize the fleet. Took $65 just to begin etching boards. Going to take another $60 to upgrade Vika 1 & finish Vika 3's board.

As predicted, keep finding mistakes in Vika 3's etch board. Already time for desoldering & etch #2.

Got it down to 2 breakout board autopilot designs. The Vika Elite board supports the 6 DOF IMU, compass, GPS, & barometer. The Vika Lite board supports only sonar & a 3 DOF IMU.

Each connects to the sensors using cables. Most of U try to put every sensor on a single board, but that's very expensive to change & every sensor has required moving around. Barometers have to be away from propellers. Compasses have to be away from power. IMU's have to be shielded from vibration. GPS has to be away from XBee. Sonar has to be visible from a wide angle.

Using GPS instead of soner can give a much smaller board than Vika Lite, but U need GPS.

Etching surface mount PCB's from scratch is hell, but when you're not selling millions these things for money, you have no choice. To spend years making stuff that has no chance of making money, U can't forever plow zillions of dollars into etching services & $parkfun proto boards.

Look at what commercial proto boards are being used for & it's basic, boring, tried & true stuff. People build just an IMU, just a basic fixed wing, or make some Google phone widget run on a servo. That's the stuff which can afford outsourcing.

If U must have 2 layers, U can stack 2 of these 1/32" boards on top of each other instead of trying to etch 2 sides. The result is .05mm thicker than a $parkfun proto board. Drilling vias is another matter.
Posted by Jack Crossfire | May 30, 2009 @ 09:44 PM | 4,079 Views
Something 4 U to think about

A micro paraglider. Use throttle for altitude & a coil actuator for yaw. Fold up the parachute & put it in your pocket. Never been documented before. Someone in China should have one ready in a couple months while...

Calif* is out of money again

2 months after U voted for higher taxes again, BARF ran out of money so U raised BARF fares another 6%. A day after raising BARF fares again, U, wait for it, ran out of money again.

http://cbs5.com/politics/california....2.1017293.html

Guess 10% wasn't all the sales tax U would ever need. Maybe 15% is all you'll ever need. It's a temporary increase of course, because next month it'll be 17%.
Posted by Jack Crossfire | May 30, 2009 @ 04:17 AM | 4,203 Views
The answer is yes. A heat gun can desolder SMT chips. Good news if U plan on many revisions.

20 minute exposures still aren't bulletproof. Need to dilute only enough draino to cover 1/2 of the bottom of the container & keep using the cotton swabs. Only took 1 hour to expose & etch the last board.

Remove contact lenses before soldering. That gives objects 20 microns away the same sharpness that normal people see 20 ft away. It's like going to atomic force microscope mode.

After another 2 hours banging on the soldering iron & getting nowhere again, decided to blast it with the heat gun to remove it. It automatically centered itself on the pads, joined all the solder balls, & started working!

That probably won't work for anything that blows away & U need guess when to stop blasting.
Posted by Jack Crossfire | May 29, 2009 @ 12:32 PM | 4,253 Views
Lost 1 to the draino. The draino was the same concentration as yesterday but real active for some reason. Then for the next 2 hours, had board after board expose unevenly from 5 minute exposures. After giving up on having any blog material by the next commute, exposed 1 for 20 minutes & it was the most beautiful photoresist in the history of photoresist board torture.

No-one ever used 100W fluorescent spiral lights before so there was no way anyone could guess 20 minutes. Unfortunately, couldn't get it to work before commuting time.

There it is. The most complicated board U can make in gEDA without going more insane than U already R. PCB has the 2nd worst user interface ever made, matched only by Windows itself. Got it down to 28x30mm. It could be much smaller if it was double sided, but we have to subsidize your mortgages.

To program the XBee, we have sort of a header. The PIC has to be in reset mode to override its serial output. The PIC programmer uses the same pins as PWM, a high risk design.
Posted by Jack Crossfire | May 28, 2009 @ 12:16 PM | 4,109 Views
60 minutes into the draino, realized there was no way to do more than 1 gyro per commute. It takes 2 hours to etch a board. The yield on these boards is too low to combine multiple designs in a single etch. 3 hours later, Gyro #2 started working. The mane problem is there's no way to test the connections on a leadless package. U just keep reheating all the connections over & over until it works.

There's no budget for a hotplate to test the Nate method & everyone besides Nate says it just detaches the pads. Assuming we keep getting time off, you'll have an IMU on Friday & the computer board will have only leaded packages. If U can figure out an optimum layout, U'll have 1 more day of driving to Santa Clara for parts before that board is fabricated. Didn't have enough money to buy it all.
Posted by Jack Crossfire | May 27, 2009 @ 12:01 PM | 4,352 Views
There it is. The LISY locked & loaded. Took 4 hours to get it working. Chees n crackers that thing was hard to solder. U SMT worshippers R crazy. Mutilated those SMT components nicely, but only needed 1 rework, an SMT world record.

Unfortunately, before U can build Vika3, U need to fabricate all the electronics & prove they work on Vika 2. That means a complete flight computer needs 2 B made out of SMT parts & soldered to Vika 2. The flight computer has no leadless parts like the IMU but it has many vias. Probably going to fake the vias with wrapping wire.

The Vika3 budget is limited by LCROSS. Really wanted to find a telescope in September for photographing LCROSS & that means you're lucky to get $300 for Vika3.

Focused on mirror lenses. 800mm at F/8 may sound horrible, but digital cameras R heading to ISO 20000. Would expect it to mainly do astronomy. Mirror lenses being what they R, U'd only want the cheapest.

In lens school, we learned all consumer lenses R made by Tamron. Opteka, Rokinon, Bower, Phoenix, & Vivitar R probably all the same. U ever go to lens school?
Posted by Jack Crossfire | May 26, 2009 @ 12:20 PM | 4,864 Views
With the SCP1000 altitude sensor, we're using very aggressive landing & takeoff programs to get these missions done as fast as possible between commutes. That means U need stronger landing gear. Contact lens bottles R on the scene again.

PICOC TAKES COMMAND

Well OK, we did integrate picoc as a runtime thread in VicaCopter. It locks up, it doesn't do floating point, it crashes your entire autopilot when it encounters an internal error, but after working around all the problems, managed to get some nifty flights out of it.

MOVE HOME+10 HOME HOME
HOVER 0
MOVE HOME-10 HOME HOME
HOVER 0
MOVE HOME HOME HOME
HOVER 0
MOVE HOME HOME+10 HOME
HOVER 0
MOVE HOME HOME-10 HOME
HOVER 0
MOVE HOME HOME HOME
HOVER 0
LAND

is now

#define RANGE 10

move(HOME + RANGE, HOME, HOME);
hover(0);
move(HOME - RANGE, HOME, HOME);
hover(0);
move(HOME, HOME, HOME);
hover(0);
move(HOME, HOME + RANGE, HOME);
hover(0);
move(HOME, HOME - RANGE, HOME);
hover(0);
move(HOME, HOME, HOME);
hover(0);
land();

What a revolution. It's like turtle graphics with a very very expensive turtle. U can build up incredibly complex flights out of reusable maneuvers. U can repeat maneuvers. U can nest maneuvers in maneuvers. Too bad none of U R old enough to remember LOGO.

Ironically, picoc was written to fly airplanes but its author never flew it in an airplane. VikaCopter is the first airborne use. His only application was that Surveyor rover.

If VikaCopter had any more flight time, she'd probably get an interactive interpreter.
Posted by Jack Crossfire | May 24, 2009 @ 08:41 PM | 4,475 Views
Chinese proverb #1: Even motor windings can break

Chinese proverb #2: There R people in this world who wind their own motors.

Chinese proverb #3: It's a good idea if U buy Chinese motors.

Did horrible in motor theory class. Just learned to recognize which equation to use for which problem, but learned nothing about picking wire guage, winding count, poles, phases, termination.

Reading through

https://www.rcgroups.com/electric-mo...struction-361/

kV determines winding count which determines wire guage. Use the largest wire which can fit on the stator. Anything U rewind 4 a trirotor has to be done for 3 motors.

PICOC

So U want something more powerful than VikaCopter's UAV programming language. Your best option is to keep using VikaCopter programming language & compile a higher level language into it. The other option is to run picoc in a thread & have it send 1 VikaCopter assembly language command at a time. Have those function send commands block until VikaCopter advances the command pointer.

How about some Picoc notes.

Run a C program by calling picoc() with a pointer to a string of the source file. It returns when the program finishes.

Create C library functions by adding lines to struct LibraryFunction CLibrary in clibrary.c. Then create native functions in a platform_*.c file. These can just print out lines of VikaCopter assembly language.

Well, now U have 2 languages: VikaCopter pico C functions & VikaCopter assembly language.

If U go with the preprocessor, there's no way to translate branches & conditions in picoc to branches & conditions in VikaCopter assembly language. It would expand everything based on compile time conditions.

U really need a function which scans a frame of video & determines the next navigation command by return value. The next step is remote control by body gestures instead of ground stations.

Vision activated commands would have to be branches in VikaCopter assembly language & not native autopilot functions.

The payoff is loops, complex expressions, variables, nested functions.
Posted by Jack Crossfire | May 24, 2009 @ 05:06 AM | 4,715 Views
Time for your first SCP1000 demo in 20mph wind. It's also your chance to see the latest in navigation tuning, fulltime GPS power. U know how most VTOL demos slowly crawl through their waypoints. Well take a look at how fast VikaCopter screams through these missions.

SCP1000 altitude control (4 min 1 sec)


According to our unbiased gyro biased panel of tri rotor judges, this is the optimum barometer placement & the barometer bandwidth should be down to 0.125. With 11" propellers, it's really solid in the calm air. The T-Rex never had enough space to mount a barometer.

Dug up our old waypoint missions from last year & ran through a few. There's no problem with vortex ring state & automatic takeoffs anymore. Automatic landings R probably slightly faster than GPS.

Mind U, we're hard coding collective to get her into the first bit of ground effect, then ramping. It keeps U from dragging on the ground. The barometer is so sensitive, the collective hardly ramps at all before going into altitude seeking mode.

Ublox5 could do just as well, but not consistently & certainly not with today's satellite outages. Sometimes we power cycle it to see if full time power makes any difference, & it does. Had a pretty serious GPS glitch which sent us into a tree.
Posted by Jack Crossfire | May 22, 2009 @ 08:11 PM | 4,687 Views
The day job gave Friday off so we could get some work done instead of pay off your mortgages. The SCP1000 on the tri rotor was as advertized on the T-Rex. Lots of noise in the rotor wash. Lots of drift in the sun. We have more placement options than the T-Rex so hopefully those problems can go away. Certainly no lag in the pressure changes.

It's not as delayed as last year because we're using high speed mode instead of high precision mode. Clearly a lot more magnitude in air pressure changes, but is it just rotor wash?

SHUTTLE VS. U

The shuttle takes 200lbs of paper to orbit, mainly cue cards & reference manuals. Computers aren't reliable enough to do the job. U take no paper to the flying field.

DECOY COPTERS IN COMBAT

When fighting a war, U need to avoid giving away your position by launching UAV's. Every man portable fixed wing must be hand launched & give away your position. What about copters?

U can stash a copter far away from your position & have it take off outside of line of sight, unattended. That's a good idea for a demo. Just need to get some Americans to play the communists.