Jack Crossfire's blog View Details
Archive for September, 2013
Posted by Jack Crossfire | Sep 27, 2013 @ 11:48 PM | 7,129 Views
After many weeks of hacking gearboxes, the gearbox which finally had enough torque & hackability was the pololu 25D 75:1 gear motor.

Despite being enclosed, it actually unbolted. The pinion gear was removable without a pinion puller.

...Continue Reading
Posted by Jack Crossfire | Sep 27, 2013 @ 01:38 AM | 6,838 Views

It's every bit as crash prone & fragile as expected. Every day of flying requires a bag of propellers. The propeller mounts tend to break off.

Not sure exactly how to repair the motor mounts. Glue isn't very robust. You can't replace the motor mounts without replacing the entire frame, which takes 1 week to rebuild. They may eventually need plywood doublers. Discovered the propellers can sometimes be bent back into place, after a mild crash.

Microquad Crashes (1 min 1 sec)

A sudden tendency to flip over in normal flight appeared. A motor+ESC was replaced but found not to be the problem. Increasing the I2C speed to support the MS5611 was not the problem. The ESC's were not malfunctioning. The turboPWM wasn't too fast. The autopilot wasn't overflowing or byte wrapping.

The problem was tracked down to the attitude control feedback. The PID limits have to be extremely low, compared to brushed motors. It took recording the PID values & PWM values on the ground to notice at what offset a flip would occur.

Since the microquad was too small to mount on a test stand, the only way was to fly as low as possible, wiggling the cyclic to try to get it to flip, & accepting propeller damage.

The px4flow seemed to do a lousy job separating angular motion from horizontal motion. It uses a lowpass filter to get the gyro centers. It reads only 1 noisy gyro value for each frame. Then it has a complicated set of rules for...Continue Reading
Posted by Jack Crossfire | Sep 26, 2013 @ 09:22 PM | 6,799 Views
3D Robotics Announces $30 Million Series B Financing

Amazing how far the good old quad can go. That's probably the most any UAV startup has ever gotten. When Rotomotion, Adaptive Flight & Neural Robotics were in their prime, they were lucky to get $100,000 in contracts in a year. The Kamen & Procerus buyouts were only in the $5 million range. $30 million is probably more than any hobby store's value.

Jordi isn't nuking his credit cards for parts anymore. Now the trick is staying in the game until their shares vest. So many company founders get booted before their shares vest, the vesting time is specifically designed to get free labor in most cases.

The baby is out of the bag & now in the hands of the corporate managers. That kind of money is going to have to go to higher end products, just as makerbot had to shift to higher end machines. It's hard to see a future without 3D Robotics supporting some military programs.

Like it or not, quad copters are now the big leagues, just like Linux in 1999. You either ride the wave all the way to the top or let DJI take all the money & live in a world of DJI products. In the Linux craze, the key open source developers were compensated nicely with pre-IPO options. The key arducopter & PX4 people should make out well.

To be sure, if I knew where 3D Robotics was going to end up & how much the founders were going to be worth, I would have quit the day job in 2007, gone...Continue Reading
Posted by Jack Crossfire | Sep 24, 2013 @ 09:06 PM | 6,422 Views

The attraction to this gearbox was that it was open, allowing you to disassemble it more easily.

It turned out extremely tiny.

The motor was bolted on, but the gearbox doesn't come apart.

...Continue Reading
Posted by Jack Crossfire | Sep 24, 2013 @ 02:06 AM | 7,570 Views

The jesus barometer goes on.

& gets earplug foam insulation. Earplugs are decent sources of breathable foam.

The temperature compensated MS5611 output

...Continue Reading
Posted by Jack Crossfire | Sep 22, 2013 @ 05:36 AM | 6,111 Views
Bay bridge run (6 min 13 sec)

10 miles of a 21 mile day. The brushless gimbal fell over after 2 hours, making 90 minutes the longest it can go on 5 Ah without destroying the battery.
Posted by Jack Crossfire | Sep 20, 2013 @ 01:13 AM | 6,458 Views
Posted by Jack Crossfire | Sep 19, 2013 @ 05:46 AM | 7,252 Views

Sonar on the micro quad was decidedly worse than the Syma X1. The motors are noisier & there's more airflow. Attempts to move it closer to the center failed. It got stuck at 0.3m. EMI from the power distribution board could be jamming it or it could be unlucky airflow. Foam wasn't rigid enough to keep it pointing down. There is hope office carpet yields better results than apartment carpet.

The MB1240 runs at 20V & the highest you can go without serious electronics is 30V. A dual transducer at 30V might do better.

Then, compass interference from the power distribution board has returned. There is a config parameter which records the mag readings to a file, during a hold down test. That can be graphed & used to set mag offsets in the config file. The hold down test doesn't give the true offsets during a flight, so the numbers need some trial & error scaling.

Only moving the mag far away would completely eliminate the interference. That's not going to happen with the MPU9150.

Finally, the altitude given to the PX4flow is degliched but giving worse results than glichy altitude, so it needs to be tested. The PX4flow does a lousy job subtracting angular movement from translational movement, yielding roll oscillation.

It would probably give better results if the pixel flow was completely processed on the flight computer instead of relying on the PX4flow to convert the flow to meters. The sonar forwarding would go away.
Posted by Jack Crossfire | Sep 17, 2013 @ 10:07 PM | 6,757 Views
Comca$t finally closed the door on getting local channels in the digital format they are aired in. For now on, they'll be encrypted & only available via an official Comca$t box after paying for the digital service. If you don't pay for the digital service, you'll be reverted to the analog low definition quality last seen in 2001.

It's a horrendous regression in service. So many people must already be paying the $100+ for digital service that no-one complained. The good news is there wasn't anything worth watching on TV in years. It might have been amusing to see some of the recent shows or a really important sporting event in HD, but they have no value.

The 3rd party DVR's & home theater PC's keeping Comca$t's CEO out of the $5 billion club have finally been squashed in 1 swift stroke. Your only recourse is to drive within line of sight of sutro tower with a portable recording device.

The goog automatically searches for "BCM7002 datasheet" for you, because so many guys have already tried hacking in. There's no unencrypted transport stream exposed on a pin for you to capture. There's no way to rip off the keys from the chip. It probably does a handshake to ensure your account is paid for, then receives a new key.

The transport stream is probably buffered in the SDRAM. Some crazy Russian could probably extract it with a really fast logic analyzer or SDRAM sniffer.

After all the years of local channels being...Continue Reading
Posted by Jack Crossfire | Sep 17, 2013 @ 04:38 AM | 7,916 Views
PX4Flow micro quad (1 min 16 sec)

Micro quad flies some waypoints, then a position hold test over plush carpet, using a PX4flow modified to use an MB1240 sonar module.

Based on the previous video of the TCM8230 over carpet, the PX4FLOW is a huge improvement. It actually holds position quite well over plush carpet & below 1m. Replacing the MB1043 with an MB1240 & translating the serial protocol from cm to mm is required.

The sonar still has issues above 1m. It glitches to 5m when it can't detect a reflection. Sometimes you want it to send 0 when it reads 5m & sometimes you want it to repeat the previous reading when it reads 5m. It still glitches, despite this algorithm.

It's very heavy, so it handles the wind quite well. Crashes require a lot of repairs & propellers.
Posted by Jack Crossfire | Sep 14, 2013 @ 12:52 AM | 6,594 Views
Heroinesapien walking on carpet (2 min 15 sec)

Eons ago, back when Goo Tube was only 320x240, the NYSE was below 9000, & phones still had keypads, I got a robosapien for $50. It was the 1st "learning toy" of any note. Manually manipulate the arms & legs while recording & it could play back the manual motions with some level of accuracy.

Actually got it to walk around on carpet, way more than its design. It needed fully charged batteries. It under extended the motions because the pots wouldn't completely follow the limbs.

It had a timer which made it record a waypoint after a certain period of not moving. That recorded a lot of wrong positions if you didn't hit the right position on the 1st try. It had a fairly high threshold for being considered stationary, making the training motions ridiculously fast.

These factors made training it a very cumbersome affair. The training was an extremely exaggerated bunch of movements that resembled none of the playback. It ended up pretty boring & tedious, compared to programming movements from a computer. The novelty was supposed to be the fluid, biomorphic movements it could produce during playback, I suppose.

It did look pretty fluid compared to previous toys, but just like graphics in a computer game, the eye candy only went so far before the logic of the game & the ordeal of teaching it by hand took precedence. There are much less visually appealing games which...Continue Reading
Posted by Jack Crossfire | Sep 13, 2013 @ 01:02 AM | 8,482 Views
Posted by Jack Crossfire | Sep 10, 2013 @ 05:02 AM | 7,369 Views
Posted by Jack Crossfire | Sep 08, 2013 @ 11:27 PM | 7,003 Views

It's single sided PC boards, conventionally routed. No 3D printing, laser cutting, or anything else.

...Continue Reading
Posted by Jack Crossfire | Sep 06, 2013 @ 09:01 PM | 8,046 Views
How the U.S. and Its Allies Got Stuck with the World’s Worst New Warplane

Another piece criticizing the F-35's suitability. It was originally appealing, because of its simple mechanism for hovering compared to the Harrier. As the conspiracy theories have mounted, other factors have to be considered.

It never looked very maneuverable, but modern flight by wire can make anything maneuverable. It was tasked with doing many things: hovering, carrier landings, bombing, dogfights. Once again, flight by wire was advertized as making everything accessible to a single airframe.

1 obvious downside is the pilot can't see backwards, instead being expected to rely completely on radar & the opponent not being stealthy.

Another downside is the internet has increasingly decided there are limits to how much flight by wire can compensate for aerodynamics. By the time the F-117 was retired & destroyed, it was unanimously considered suffering from very lousy aerodynamics which the software couldn't compensate for.

Finally, there's the issue of why 3 variants of the F-35 are required. Why can't all of them hover, land on a carrier, & do whatever the air force does? If taking out 2 of the 3 options makes them better at 1 mission, wouldn't they be better if the rest of the airframe was optimized for 1 mission too?

So it's probably not as good as the A-10, F-16, F-18, & F-14 were at their jobs. It's probably better than the harrier for its...Continue Reading
Posted by Jack Crossfire | Sep 06, 2013 @ 08:19 PM | 7,121 Views

After 5 years, the mighty HP has finally been outdone by television standards. The HP did 2560x1600 at 30", which was way beyond any TV resolution for many years. No matter how big a TV was, it could never do more than 1920x1080.

You might say it made day jobs involving TV standards a turn off. It was a vacuous size for opening many windows, having giant fonts for text editing, & viewing photos at nearly their full resolution.

But now, China has unleashed the next generation of TV's on the world. 4096x2160 at 85". 60% more horizontal & 35% more vertical resolution than the HP, with enough size to completely immerse you. It's the next must-have status symbol of women my age.

Even if you're not a jet setting 38 year old blond hottie with $40,000 to burn, 4k is what every TV is going to use, right down to the HP's $1300 tag. Of course, modern 30" monitors are $800 & have much better color. The HP always had garish colors.

A TV would be a buy for me, if it was bigger & higher resolution than a computer monitor. Everything used TV's in the 1980's but TV's later suffered from being locked into low end standards while computers got ever higher resolution. 4k might be the resolution to tip the scales back towards a TV for everything.
Posted by Jack Crossfire | Sep 05, 2013 @ 01:11 AM | 7,841 Views

The PX4flow is liberated from its inferior sonar module & useless headers.

...Continue Reading
Posted by Jack Crossfire | Sep 04, 2013 @ 04:20 PM | 7,654 Views

The 1st spill on a body part in 25 years of chemistry experiments. That was a fast run to the shower. Never saw someone use the showers in college, but it must have taken quite a cleanup.

Also noted the boards don't float unless they're dry.

Unfortunately, no-one wanted to pay for a radar experiment, so it's 1 more shot at a better optical flow system.

...Continue Reading
Posted by Jack Crossfire | Sep 03, 2013 @ 03:36 AM | 7,927 Views

The STM32 can do a 1024 point FFT in 5ms. There's also the unknown matter of computing magnitude, normalizing, making it probably 10ms. At 10 sweeps per second, that gives only 10 FFTs per sweep.

That comes with a 8192 byte tag.

2048 bytes: analog capture window
2048 bytes: imaginary value window
2048 bytes: simultaneous capturing of next window
2048 bytes: normalization table

Then, the 1st 256 samples from each window are stacked on a polar projection. Only the 1st 1/4 of the window has had useful range. That comes with a 2560 byte tag for the current projection & 2560 bytes for the previous projection for 10 FFT's per sweep. 50 FFT's per sweep would take 12800 * 2 bytes.

Motion tracking in a polar projection would take some doing, but be the most efficient solution. An additional 5120 bytes of ROM might be required for a lookup table to convert the polar pixels to square pixels for motion tracking.

Then, some 1024 bytes of scratch memory might be required, bringing it to 60416 for the 50 FFTs per sweep or 19456 for the 10 FFTs per sweep.

To rotate the FFT's & motion track a square projection would take 256kbytes to retain the full resolution. It would involve a similar lookup table to only track filled pixels.

No matter how limited the microcontroller, the resolution & speed of the sampling can always be scaled down to fit. The difference is only in the resolution of the output, not whether it can...Continue Reading