Jack Crossfire's blog View Details
Posted by Jack Crossfire | Mar 20, 2009 @ 09:02 PM | 3,092 Views
The Goo Tube had another nugget 4 U. The latest in indoor position sensing is UWB positioning. UWB position sensors work exactly like VicaCopter's sonar concept, except these pseudo random pulses go from 3-12Ghz & repeat at 10Mhz. They use UWB because it's immune to reflections.

UWB position sensors R already sold commercially. Just take out your home equity line of credit & go to


Now have an indoor copter using UWB position sensing.

JJ Autonomous Flight (1 min 58 sec)

With UWB positioning, the spider drones in Minority Report should be no problem.
Posted by Jack Crossfire | Mar 20, 2009 @ 02:37 PM | 2,802 Views
Look kids, let's get 1 thing perfectly clear. UAV hobbyists posting discoveries on RC Groups don't aid the enemy. If they ever had the brain power to produce military results, they would already be in the business instead of doing it as a hobby. If they R doing the military level as a hobby while the enemy is using it for defense, then the enemy deserves to win.

Who wants to live in a world where the state of the art is RCGroups while the trash is Boeing? That would be a system which punishes the best & rewards the worst. By all means, let the enemy win & give US a system which rewards the best instead of suppressing them.

Fortunately, we've never seen anything on RCGroups anywhere near what could aid the armies of the world. Take a look at the AIAA journals & you'll be satisfied the system is still rewarding the best. The armies of the world R all above that & that's just what the Google corporation allows you to see. Google filters out all the crazy research. All those results in turn R above the RCGroups hobbyists.
Posted by Jack Crossfire | Mar 20, 2009 @ 12:55 PM | 3,319 Views
So the optocouplers R the best solution for the given budget. PWM adds a lot of noise. The gain is half wave only. That may be a DC offset problem. It takes a lot of CPU time to correlate 157khz audio. The packets get out of order when using 2 endpoints. Most of all, the gain oscillates.

Someday there is going to be a 157khz sampling test with the Maxbotix's to see if software can get any omnidirection out of them. That sonar system would require 2 discrete A/D converters but would be silent & deadly accurate. They're so directional, we expect nothing except reflections.
Posted by Jack Crossfire | Mar 19, 2009 @ 02:00 PM | 3,401 Views
So U want more bandwidth but U don't have enough memory for ping pong buffers. U need smaller audio buffers but Linux has too much latency. U want to send different streams of data without having to multiplex packets by hand. U need USB endpoints.

The pros don't call blocking read calls on a file handle like U do with UARTs. For USB they allocate many file handles & request simultaneous I/O on all of them in the background. That way the device always has virtual buffers to fill & doesn't overflow waiting for Linux. Also, each endpoint can be a different stream, so U don't need to multiplex anything by hand. It's a great way to stream telemetry.

With that out of the way, VicaCopter shows up like this in /proc/bus/usb/devices.

T: Bus=02 Lev=02 Prnt=20 Port=01 Cnt=02 Dev#= 34 Spd=12 MxCh= 0
D: Ver= 2.00 Cls=00(>ifc ) Sub=00 Prot=00 MxPS= 8 #Cfgs= 1
P: Vendor=04d8 ProdID=000b Rev= 0.01
C:* #Ifs= 1 Cfg#= 1 Atr=80 MxPwr=100mA
I: If#= 0 Alt= 0 #EPs= 8 Cls=00(>ifc ) Sub=00 Prot=00 Driver=(none)
E: Ad=01(O) Atr=02(Bulk) MxPS= 32 Ivl=0ms
E: Ad=81(I) Atr=02(Bulk) MxPS= 64 Ivl=0ms
E: Ad=02(O) Atr=02(Bulk) MxPS= 64 Ivl=0ms
E: Ad=82(I) Atr=02(Bulk) MxPS= 64 Ivl=0ms
E: Ad=03(O) Atr=02(Bulk) MxPS= 64 Ivl=0ms
E: Ad=83(I) Atr=02(Bulk) MxPS= 64 Ivl=0ms
E: Ad=04(O) Atr=02(Bulk) MxPS= 32 Ivl=0ms
E: Ad=84(I) Atr=02(Bulk) MxPS= 32 Ivl=0ms

That's a lot of endpoints. There's 2 for sonar, 1 for the Xbee, 1 for 72Mhz, 1 for commands,...Continue Reading
Posted by Jack Crossfire | Mar 18, 2009 @ 12:55 PM | 2,905 Views
Got the PIC up to 157khz. That's high enough to record raw Maxbotix transducers. The mane limitation is it only has 384 bytes of memory. U need to service USB requests in 1/400s or the buffer overruns. It definitely overruns depending on the Linux load.

Unfortunately, when multiplexed for 3 microphones, it comes down to 52khz. Next, 8 bit audio is pretty worthless without AGC. U can make a voltage controlled variable resistor out of a pair of optoisolators. It's the simplest voltage controlled gain possible. Just 6 optoisolators required.

Now that the Goog was online for a few minutes, U get the sound recording.

Home made soundcard (0 min 33 sec)

Posted by Jack Crossfire | Mar 17, 2009 @ 02:16 PM | 3,457 Views
Luckily 4 U, the Linux kernel can be a software bus analyzer. Go into linux/drivers/usb & use printk's to dump the buffer contents. U need to stop searching the Goog for bus analysers.

5 days of insanely difficult assembly language debugging & the home made soundcard finally started working. The USB driver ended up being under 1000 bytes, quite a bit less than the host we did before. There R no interrupts for USB. U have to use polling, which slows it down.

It can do 150khz before running out of clockcycles. That's 1.2 megabits coming out of a PIC. It can only do 23khz before the sound quality starts deteriorating. U need 150khz for sonar. The Goog is offline, so all U get is a picture of the waveform.

GPS aided heading only works in hovers. It falls over in horizontal movement. Going to acceleration doesn't work. That leaves washing out velocity or forgetting about it when translating.

Must remember to try adapting the sonar ping rate based on velocity. Maybe 1Hz when moving slowly & 10Hz when moving fast. It could either make it more bearable or make it sound radioactive.
Posted by Jack Crossfire | Mar 16, 2009 @ 01:54 PM | 2,314 Views
Been thinking about making a fake H1-B. Unlike American citizens, H1-B's can work anywhere & do anything. H1-B's R like MIT degrees. When U want to hire the best, H1-B is a key word. Most resumes get filtered out just by not saying H1-B. H1-B's get pushed to the front of the line because they have a limited time before they get shipped back to India. If U want the best, U have to beat the 3 month auction deadline.

Citizens using fake H1-B's R going to be an epidemic in the future. The heads of all your government agencies R H1-B's & they're going to protect their buddies. There will be raids on rich mansions & aerospace startups to bust illegal American citizens. Americans R going to be rounded up & deported back to the tent cities & bread lines. To get a job, you'll need a DNA test to prove you're not a son of the revolution. Getting caught with a social security card is going to send U to Gitmo.

Yes, Gitmo is going to reopen to house the flood of illegal American citizens.
Posted by Jack Crossfire | Mar 15, 2009 @ 07:12 PM | 2,147 Views
Was shocked to see climb rate lagging altitude change once again. This occasionally happened with GPS in the past. The first uBlox 5 test had equivalent climb rate.


Airborne dynamic models R busted. Been using them because they said "airborne." It's like having diodes which say "diodes" but are really transistors.

The Pro Power once again returned 12 min.

Unfortunately, sonar is dead. USB doesn't work. It gets a descriptor request from the host but the host can't receive the response. Without a bus analyzer, U can't really optimize the reference software as U need to.
Posted by Jack Crossfire | Mar 14, 2009 @ 12:45 AM | 2,820 Views
What happens when your day job is near a hobby shop? U drive down thinking you've got things under control & drive up wondering what happened.

Pro Power, 3 DT700's, & 3 12x6's on a tri rotor is pretty insane power. Turns VicaCopter into a bullet. A few more evolutions like the Pro Power should have us up to ballistic spaceships. U can break props with a slight twitch of collective. Unfortunately, when it dies, it dies hard. 10.4V is dead. Flight time till death was around 12 min.

Finally, GPS heading is probably as good as we can get it. It stays within 45deg of the right heading no matter where it's pointing. When it starts losing position, the movement pulls it back to the right heading, as designed. Facing South is a bit more accurate. Over 5 minutes, when facing South, GPS can actually tune in a more accurate heading than a magnetometer. Facing West & East is down to +/-45 deg.
Posted by Jack Crossfire | Mar 13, 2009 @ 03:47 AM | 2,450 Views
Since GPS heading got ahead of sonar, you're going to be reading about sonar 4 a while. The circuit which blew up 3 years ago is going to do 1,200,000 bits/sec of sonar capture. This time, U know it needs a 4700uF capacitor. Unfortunately, it still requires an external power supply. Fortunately, unlike your last USB operation, this is a simple client.

USB is much easier to connect than RS232 & if U use the bloated reference software, about as hard to get working as RS232.
Posted by Jack Crossfire | Mar 12, 2009 @ 02:22 AM | 2,697 Views
Another commute & she hovered without a compass again. The algorithm is showing signs of life. High rudder gain & no turns seem to help. U can't get a cold heading from attitude & velocity. U need to hedge the heading with a priori information from the yaw gyro. There's a lot of position oscillation & drift 2 B sure.

So much for that. Now let's go to usaspending.gov & see who's in the money.

Lite Machines (2008): $2,497,182

Don't count on the Coronas coming back.

Procerus (2008): $828,025

Flying wings aren't as in demand as flying suppositories.

MLB Company (2004): $263,676

The more expensive the rent, the smaller the contract.

Adaptive Flight (2008): $50,001

Time to give up on the T-Rexes & upgrade to tri rotors.

Jack Crossfire: zzzzzzzeeeeeeerrrrrrooooooooo

Realistically, usaspending.gov has hardly anything useful. There's nothing on Baldwin Technology.
Posted by Jack Crossfire | Mar 11, 2009 @ 02:23 PM | 2,707 Views
A software comb filter drastically improved the sonar. This gets U in the 6in range at over 14ft. Was hoping to use the laptop microphone in flight test #1 but it sux.

Every commute brings a new pulse, a new indoor test, a new filter. Pink noise is better than constant frequencies. Shorter pulses R better than longer pulses. Multiple nodes R required. Highpass filtering isn't as effective as the comb. Comb filter -> correlation -> integral -> smoothing seems to be the common currency of sonar.

GPS heading got busted in flight #2. Back to Detroit with that disaster. A full IMU & compass is a lot more stable but no fun.
Posted by Jack Crossfire | Mar 10, 2009 @ 03:23 AM | 2,689 Views
The current champion in soundcard sonar is the highpass filter + correlation filter + integral + smoothing. Still pretty awful, but still improving.

Heading without magnetometer is the holy grail. U can't sell a product using magnetometers because they're prone to misalignment if stored near any metal objects or high current.

The main problem is U can't simulate GPS heading corrections. It's never going to be as good as a hand tuned magnetometer. The available movement in a hover is too small to sense very small angles. We're probably 50 years behind China with it.
Posted by Jack Crossfire | Mar 09, 2009 @ 01:33 PM | 2,380 Views
The DT700 12x6's suck 20% of your flight time just like the 2410-9 12x6's. At least Vika 1 is now completely silent & has a lot more power.

The best audible sonar has been the traditional short sine wave pulses. The correlation filter & amplitude sensing take turns with better results. There is no escape. You're going to die from audio torture.
Posted by Jack Crossfire | Mar 08, 2009 @ 11:18 PM | 2,487 Views
Correlation filtering is pretty bad. It's not much better than straight amplitude. It doesn't reject noise well enough to have a constant jet engine sound. U get better results with short bursts of high frequency pink noise. With this grotesque soundcard sonar test, we're looking at maybe 1m of accuracy.

Propeller lightning round

APC 12x6's on DT700's R hopeless. They make a good shake table. The SF 11x4.7's work. 4mm drilled SF 12x6's work & look better.
Posted by Jack Crossfire | Mar 08, 2009 @ 04:31 AM | 2,600 Views
Swapping an 18A Suppo for a 30A Suppo made no difference. It's down to the motors & vibration. We're leaning pretty hard towards running the 2410-9's on 30A ESC's & getting rid of the DT700's.

Only got the DT700's because Chinese shipping is outrageous. The Goog is displeased with DT700 vibration & weight. It doesn't see anyone sticking with them after a few blog posts. The 4mm shafts R a real problem since most parts R for 3mm shafts & no local parts have 3mm shafts. Once U drill out 4mm holes, you're stuck with a vibration machine.

Well, got a sonar ping which U humans can stand. We have uploaded a recording from the electret microphone preamp 4 U sonar freaks. It now completely depends on implementing a suitable correlation filter.

Sonar & U (0 min 20 sec)

Posted by Jack Crossfire | Mar 07, 2009 @ 04:14 AM | 2,711 Views
High dielectric under the pins got us to receiving on the airframe, but running the PIC at 25Mhz jammed it completely. Running the PIC at 40Mhz was crummy but it worked slightly. Alternative power supply? Fuggedaboutit. It's RF from the PIC.

While ghetto rotor chews on her crummy GPS, it's time for another propeller mount on Vika 1. The propeller mount worked but attitude control was gone. It's oscillation + constant control losses. The 30A Suppo's R the main suspect followed by vibration.

Well, the mane problem with sonar is we don't see a way to pull it off for less money than GPS. U can build GPS modules that weigh less than sonar. With GPS, U can skip the ground station completely & use 802.11 instead of XBee's. With sonar, U need a ground station.

Not sure exactly how audible sonar would work. U need a true USB microprocessor to move the audio fast enough. U need a super steep bandpass filter with under 0.1s phase shift. Then, U need to adjust hardware gain from software.

Another idea is to synthesize the known soundwave on the ground station & correlate it with the received sound wave to get the phase.
Posted by Jack Crossfire | Mar 06, 2009 @ 01:22 PM | 2,782 Views
Ghetto GPS didn't work on the airframe. Never got a thing in the field, with all other systems disabled, at 100ft AGL, or any of that other stuff. The latest theory is foam touching our beloved upside down pins.

No problem flying ghetto rotor manually, using IDG300's. Just barely enough power to go up, even with the new motor mounts. Slightly tail heavy according to the PWM recording. Ghetto LEDs were probably visible enough.

Now lets get one thing perfectly clear, for everything U know & everything U've done, there are millions of people who know & do the same thing. In a planet of 7 billion people, it can't not happen. They don't all show it on the internet. Even fewer get on The Goog.

There are millions of people with the skills to put VicaCopters together, millions of UAV's of 1 kind or another in the world, & thousands of autonomous tri rotors. There are more UAV's than all the full time soldiers in the world. A world war fought purely by the current UAV fleet would be broader than anything humans could pull off.

1 autonomous tri rotor is produced somewhere in the world by 1 Jack Crossfire every day. All the defense contractors & hobbyists of the world produce at least 1 UAV every second.

Compared to the human birth rate of 2.5/second, humans have probably 2 years before their birthrate is outstripped by UAV's. In 30 years, humans can expect to be outnumbered in total number by UAV's.
Posted by Jack Crossfire | Mar 05, 2009 @ 03:19 AM | 2,732 Views
Got a sheet of copper in the mail last year but only figured out a use 4 it recently. It's now the ground plane of our uBlox 4 ghetto GPS module. The tri rotors don't have enough flight time to lift the Altoids, so it was time to give passive antennas another try.

The passive antenna mounted right above the module actually worked when connected directly to a PC. Now that old uBlox4 was a better investment than AIG.

Does ghetto blox receive anything fully integrated on the airframe? Fuggedaboutit. We have a few more commutes to go with this one.
Posted by Jack Crossfire | Mar 04, 2009 @ 05:12 PM | 2,478 Views
For all your heroine worshipping, it's time U faced facts. Heroines caused great depression II.

In the old days, the media called heroines investing geniuses because they were experts in housing, the CEO of the house. They knew all the housing & domestication stocks & those stocks were doing great. Their stereotypical credit card charging made them richer through the power of housing inflation.

Where are the CNN stories praising heroines for their domesticated genius now? When the unlimited credit disappeared, the CEO of the house was yet another bimbo with a stack of credit cards.

CEO of the house before. Now CEO of the depression. The mortgage bailouts are bailouts for heroines.

Men wanted gadgets. Heroines wanted real estate. How many of your heroine coworkers bragged about how they made their husbands buy a house instead of a model copter back in 2005? The depreciation on average Bakersfield mansions would cover 300 model copters, if taxpayers weren't covering it.

The fact is if heroines thought more like men & wanted gadgets instead of kitchen remodelling, Micron wouldn't have laid off all of Idaho. We'd probably have humans exploring Mars instead of plunging their total economic output into mortgage bailouts.

Heroines caused great depression II (0 min 40 sec)