Jack Crossfire's blog View Details
Archive for January, 2013 - Page 2
Posted by Jack Crossfire | Jan 09, 2013 @ 04:26 AM | 4,755 Views
The baby Vicon continues. The great difficulty with it is the extremely laborious calibration. It has USB instead of ethernet cables. It's not clear if the real Vicon uses the internet protocol or just uses the same cables. It has a much easier job than a real Vicon.

The great task with the copter was creating the right voltage for white LED's. A BJT with fixed PWM has produced a reasonably stable voltage of 2.6-2.9 for all battery voltages from 3.4-4.2. The BJT takes 0.6V, leaving only 0.6V for the PWM to drop. More voltage is lost in the wiring. The LEDs aren't nearly as bright as possible, but good enough. So the PWM doesn't really do anything besides a last bit of protection.

The mane problem with vision has emerged as velocity detection. Doppler shift in the carrier frequency allows GPS to get very up to date velocity readouts, almost before the vehicle is even moving. All vision can do is differentiate between 2 frames. You can go to 200fps like Vicon, to get the velocity readout as close to 1/200s after the present time as possible, but it's never going to be a velocity readout at the speed of light that doppler shift can get.

The ages old problem of conversion between forward flight & hovering has emerged as a glaring issue. It didn't matter outdoors, because there was enough space for wobbling & GPS was good enough to compensate for it. It didn't matter in the apartment, since there wasn't enough space for translation. Flying...Continue Reading
Posted by Jack Crossfire | Jan 08, 2013 @ 01:15 AM | 4,834 Views
In 2011, a bag full of $1 LED flashlights arrived. $1 seemed too good to be true for 5 LED's, so maybe they were really a single LED split by a lens. The batteries quickly died in the last year, so it was time to tear down.

There were indeed 5 LED's, crimped but not soldered together. That explained the tendency to glitch off. There were no resistors, showing they probably ran on the direct voltage of the batteries, at least 3.6V....Continue Reading
Posted by Jack Crossfire | Jan 05, 2013 @ 03:44 AM | 5,287 Views
In about 17 seconds, the preamp reverted to its original, analog pot. All the work on the digital pot took no time to rewind to the time before & the humming was gone. There was also some noise from electronics in the apartment. Unbalanced audio is hard.

Extremely relaxing mane brushing (11 min 19 sec)

Meanwhile the methuselah of localization systems begins. 2 cameras & 4 servos.
Posted by Jack Crossfire | Jan 04, 2013 @ 06:21 AM | 4,879 Views
Stereo CAD GXL-2400's (2 min 26 sec)

Imagine the astonishment to find after the blog post on Dec 22, a dedicated blog reader sent a 2nd CAD GXL-2400 to make a stereo pair. More impacting than finally having stereo large diaphragm condensation, was the feeling of summer 2011 returning, opening up the gold box again, revisiting the same questions of how to mount it, how to connect it, whether the pre-amp would work. That summer is going to be remembered as 1 of those unique times & a lot more positive than when it actually happened.

The pre-amp basically worked in stereo, but it can't amplify the balanced signal in stereo. There was a lot more noise, electrical buzzing, buzzing that faded as the capacitors filled. Some of that could be the Zoom H2, but the transient nature of some of it pointed to the pre-amp's digital pot. The digital pot needs -2.5V from a charge pump.

All the switching circuits run at multiples of the same frequency, to minimize resonant humming, but it never completely worked. There are also transient current spikes in the microcontrollers. Compared to other recordings with consumer gear, the noise is unacceptable.

The proper solution is using 2 more amplifiers to amplify the balanced signal, then either record 4 channels or combine the amplified signals with op-amps.

The reality is to go back to a purely analog pot, maybe even use 11V instead of 48V phantom power or use an off the shelf boost converter. Off the shelf boost...Continue Reading
Posted by Jack Crossfire | Jan 01, 2013 @ 08:56 PM | 4,627 Views
Manual flying with a tablet (9 min 19 sec)

Is a lot harder than sticks. Maybe it needs some user interface tweeks, but it's also a very small room. It's hard to find the level point. This was followed by autopilot flying between waypoints. The mane problem is wifi dropouts.

All the work hardening the 900Mhz chip radios & conformal coating the boards is undone by wifi. Have had dropouts with both the home made networking stack flashed on the STM32F4's & the raspberry pi running the stock networking stack. It could be why the RTL8192's were originally only $2, but any 2.4Ghz radio should be capable of the same reliability of a chip radio.

The 802.11 standard requires lots of arbitration, which seems built into the hardware. Bluetooth should be better, but would take another long driver port.