After multiple credit card rejections & alerts, the bang good transaction went through. 4 days later, the goods arrived with a bang. The quest for a handle began.

1st, it needs another control board to generate roll, pitch, & mode PWM signals. The PWM generator is required whether it's on a handle or a vehicle.
Only 6 signals go in: power, GND, roll, pitch, mode, video There is a 0.176V diode on the shock absorber board to protect against reverse voltage.

The USB connector on it is overkill. It really takes a standard UART TX & RX signal on the same pins as the roll & pitch, so it can't be programmed when the PWM source is running. The 5V from the USB connector is floating, so it needs to be powered from the PWM generator while no PWM is being generated.

The original plan was to remove the enclosure & bolt the handle directly onto the motor. The problem is a control board is bolted to the enclosure, which is bolted onto the motor. You can't remove the enclosure without removing the control board. Hacking the board to operate outside the enclosure would prevent the original shock absorbing board from being used. It also uses high speed I2C to communicate with control boards attached to all 3 motors & the IMU. That won't work outside the enclosure. There aren't any bolts long enough to go through an outside plate, the enclosure, board & motor.

The shock absorbing board can be removed, yielding a much more compact setup suitable for a...Continue Reading

WE'RE MOVING!!! (10 min 54 sec)

Romantic Night (10 min 10 sec)

...Continue Reading

Feiyu mini 3D handheld gimbal DIY (2 min 27 sec)

Economical 3-Axis SteadyCam (DIY) (13 min 34 sec)

...Continue Reading

Washington DC Timelapse run (7 min 33 sec)

After making that, a better stabilization system became a lot more desirable. So basically, the brushless gimbal made 3 years ago ended up too big & cumbersome to be useful for most running videos or traveling. There's no way without the power of Chinese manufacturing to miniaturize it. Either that or the cost would be more than a commercial version.

In the future, 360 cams will replace all brushless gimbals, but there is currently no incentive for anyone to make a 360 cam work like a brushless gimbal when it's cheaper to borrow money. There might be a way to hack it. Years ago, there was an attempt to use a phone's internal IMU to stabilize the video from its own camera. The problem was they had to use an off the shelf phone & only the software APIs included. The focal length was too long to give enough overscanning area. The IMU wasn't fast enough to track fast movements. It only worked if the API gave time stamps which could synchronize the video with the IMU data.

The new idea would be capturing IMU data externally from the camera & synchronizing it by reading the VSYNC pulses from the sensor. New cameras may not have an exposed VSYNC signal. It's not likely the clocks would stay synchronized tightly enough to pull it off.

The raw IMU deltas would be recorded on the microcontroller's flash. They would generate offsets for Cinelerra's motion tracker. They would have to be exactly synchronized...Continue Reading

Air Hogs took another swing at the spin copter game again with the Hoverblade. It looked a lot more like a monocopter than their previous efforts, but once again, it just had up & down control.

Spin Master Air Hogs 360 Hoverblade Unboxing (13 min 15 sec)

The lone gootube video of it showed a bad horizontal oscillation. It's almost like they tried a monocopter, but after getting nowhere with it, went back to the standard vertically controlled spinner.

While somewhat surprising no-one has ever commercialized Marcy 1, miniaturizing it enough to make it practical for the toy market might be too difficult. The motor size might be the limiting factor. There may not be materials strong yet light enough to achieve the same slow motion in a smaller size.

It was the 1st computer a fake test pilot could afford by his own doing. There will never be another 1st.

It was a monster for its day. 200 whole megahertz. 512 k of "cache". Surely it would open up serious audio reverb, all done in the "cache". The cache was physically the same as the DRAM in those days. It was just addressed faster, but all the memory had to be accessed through another chip. What a joy to finally have a living breathing UNIX workstation that would never crash, with someone else paying the electric bill. It finally crashed when the heatsink filled with dust.

It didn't take long after mastering its audio capabilities for video to arrive. 1st came playback of 320x240 MPEGs. Then came recording 320x240 at 4fps or 640x480 at 1fps. Still didn't know enough about video compression to get the most out of it, but it was a new world. There will never be another new world in video or audio. The new world nowadays is the physical world.

You can't afford to keep a lot of the stuff you once had, especially with rent heading for a 3 handle. After 19 years of waiting for some kind of embedded repurposing that would never come, it was time to put it down. It might have actually had some use, right until 10 years ago when the Gumstix arrived. By 2014, the CPU heatsink was repurposed for the contact lens agitator. In 2015, the last of the PCI video cards was discarded, so it would never run again.

7m15s mile (8 min 4 sec)

Snagged a 7m15s mile from the Ruckus. The image stabilization had its 1st material improvement in 18 years. There were previous attempts, but nothing which gave a net improvement. Now, much higher resolution can be stabilized in translation & rotation in much less time. This one took only 3 hours on the Macbook. On the dinosaur PC, it took 5 hours. Previously, the dinosaur PC was slightly faster than the Macbook.

The mane focus was using mipmaps instead of the full resolution image. The original algorithm stressed memory access while the new algorithm stresses the cache more. It shows the Macbook may have slower memory bandwidth, but faster cores. Memory may only be slowed down by the virtual machine.

There's still no solution to rotation drifting. Stabilizing a flipped video showed it's not a rounding error.

Then came the reality that using the vehicle as a pacer over a full 2 miles resulted in a slower total time than running without it, because of all the flips & navigation through mobs. The vehicle still allows higher speeds when maneuvering isn't required. Surprising how much a phone screen scratches after many slides on the pavement.

Steering oscillation remaned pretty bad. Steering was also too slow to respond, symptomatic of high D constants to overcome the oscillation. These factors made it very hard to navigate obstacles.

After roughly 110 miles & 2 months, the SPMS401 died. While disappointingly short lived, the failure was software induced. When full time heading hold arrived, it made the steering build up when sitting still. After drifting far enough, the servo saver jammed well within the servo's range & the motor stalled. That's what happened when it was left powered on the bench after a successful drive. It worked until the end, but the next day, the servo was fried.

It was another case of software costing money & a servo saver destroying a servo. Ideally, the heading hold would time out after 30 seconds of no throttle. Replaced the SPMS401 with a Futaba S3102. While it didn't fit the servo saver, it worked fine with a standard horn. The standard horn didn't jam in the servo's entire range. Pity the fool who bought a replacement SPMS 401. It's a standard micro servo.