Jack Crossfire's blog View Details
Archive for December, 2017
Posted by Jack Crossfire | Dec 28, 2017 @ 11:35 PM | 4,852 Views
The Delta IV heavy was a fascinating rocket, when it 1st came out. It looked like a classic giant movie spaceship when it lifted off & pitched over, revealing a row of 3 widely spaced engines. The space shuttle & Ariane 5 didn't look like classic movie spaceships, because they were obscured by solid rocket exhaust or mostly seen with just center engines after booster separation.

The Delta IV heavy never achieved its goal of economic, scalable capacity. Rather than repurposing the same cheap cores from smaller rockets, it probably matched SpaceX's experience of having to completely redesign the center core just for a handful of missions. Hopefully, making the boosters reusable will achieve better results.

Falcon Heavy will look even more massive when it pitches over. 27 engines making a rectangle of light will look like the rear end of a battlestar or a star destroyer only seen in movies until now. Then, when the side boosters separate & the opposing plumes of 3 engine burns interact, it should be a spectacular airshow.

Despite not coming close to the size of the rockets we once had, these triple core rockets are somewhat fascinating because of the potential to be affordable.
Posted by Jack Crossfire | Dec 22, 2017 @ 02:03 AM | 4,413 Views
So what would have happened if Musk said "Picture of FH payload soon" & it was YOUR CAR on the payload adapter? Are the Calif* license plates still attached? It wouldn't stand the loads if it was supported from the rear & the fairing wasn't wide enough to mount it flat, so it got mounted in a dealer pose.

The mane impact from this is that it's his own personal car which he probably drove many times, not something from a customer or a purpose built payload. It's probably going all the way to Mars orbit with the 2nd stage, so it'll be possible to track. It should be quite a sight, as the employees announce the mission milestones with the boss's car.
Posted by Jack Crossfire | Dec 18, 2017 @ 11:09 PM | 4,638 Views
A routine drive ended as a wheel rolled off. The lack of shock absorbers & shifting of payloads forward must have done it in. If it had just a few more stringers, it would last longer. That caused the tires to wear unevenly.
In this age, 10 years after the 3D printer craze, it's hard to believe there's still no way to make custom wheels.
Posted by Jack Crossfire | Dec 14, 2017 @ 10:22 PM | 4,352 Views
Our 1st & only view of the rebuilt LC40 from above. There's no evidence of the explosion, 16 months ago, nor a crater or burn marks. The mane intrigue is how many buildings are packed close to the launch mount, close enough to feel the exhaust heat & the force of the sound pressure waves. Vandypants isn't built up as much, near the launch mount.

Conventional wisdom is to be able to service the rocket as much as possible when it's on the launch mount, but instead of the traditional arrangement of a hangar far away & a mobile service structure, they've moved all the infrastructure as close as possible, weighing knowledge gained about the blast radius over time.
Posted by Jack Crossfire | Dec 14, 2017 @ 05:59 PM | 4,489 Views

Started putting projects on hackaday.io. This is not to be confused with hackaday.com. The hackaday domanes are a confusing mess of illegible grey on black text, but the few readers might find it more useful. It feels like rcgroups has lost attention to the hackaday, eevblog, tested.com properties as the RC world has expanded from toys to industrial robots which control cameras, clean floors, deliver food, drive taxis, & train athletes. An RC vehicle is just as often found delivering food now as it was doing flat spins, 10 years ago.

Steering the lunchbox continues to be a brutal occupation. The stiction algorithm as implemented did nothing. It started becoming obvious that control feedback behaves like an equation. No matter how convoluted the feedback algorithm is, the result converges on the same answer in the long term, like an equation which always outputs a sine wave despite having 1 added & subtracted to it many times. This is why control theorists write their feedback algorithms in equation form. It makes it easier to see what changes reduce to 0.

The stiction problem is specifically caused by the servo saver absorbing much of the motion & less because of the linkages. The servo saver absorbs much motion in left turns but virtually no motion in right turns. Maybe the integral has to be kicked based on swings in the PWM output instead of the heading error.
Posted by Jack Crossfire | Dec 09, 2017 @ 06:05 PM | 4,095 Views
93 miles of driving after the stiction fix, it could once again no longer go in a straight line. The next step would be sending more diagnostics to the GUI, in realtime. It may not be seeing the deviation in heading required for it to kick the integral or it may be waiting for a deviation in the wrong direction. After drifting right for a while, it suddenly kicks left. Showing the realtime deviation in heading, what deviation it's waiting for, & the feedback is part of the solution.

Other ideas were adding noise to the feedback, because it seems to knock itself straight when it hits bumps.

The raspberry pi zero W was too good to be true. After integrating 10 of them without any issues, packed one in a static bag & commuted with it. It no longer detected wlan0.

dmesg|grep brcm went from showing

[ 35.073582] brcmfmac: F1 signature read @0x18000000=0x1541a9a6
[ 35.103366] usbcore: registered new interface driver brcmfmac
[ 35.519555] brcmfmac: Firmware version = wl0: Aug 7 2017 00:46:29 version 78
[ 35.520794] brcmfmac: brcmf_c_preinit_dcmds: CLM version = API: 12.2 Data: 7
[ 42.937355] brcmfmac: power management disabled


[ 35.693070] brcmfmac: F1 signature read @0x18000000=0x41e041e
[ 35.693567] brcmfmac: brcmf_chip_recognition: SB chip is not supported
[ 35.693580] brcmfmac: brcmf_sdio_probe_attach: brcmf_chip_attach failed!
[ 35.693592]...Continue Reading
Posted by Jack Crossfire | Dec 07, 2017 @ 02:30 PM | 3,934 Views
It went only 24.5 miles with the speaker, extra food, & camera. The speaker wasn't much better than silence, because it was 5 hours of the same old song. Audiobooks are a novel concept, there aren't many good ones. Getting the extra mile would be just a matter of another 1Ah battery somewhere. The trick is when connecting a fresh battery to a drastically discharged battery, the inrush current hits 5A at only 11V. Wiring 3 batteries in parallel would fill it with unused connectors most of the time. There's also discharging the 1Ah battery early in the drive, then putting in the 5Ah battery. This wouldn't work because the 1Ah battery would discharge far slower in the beginning. There's also carrying around a terminator & driving just on the 1Ah battery in the end.
Posted by Jack Crossfire | Dec 01, 2017 @ 09:41 PM | 4,253 Views

The 1st running vlog took 1 day to render. 10 hours went to discovering motion stabilization wasn't working. Another 4 hours went to doing it again without stabilization. Not sure why stabilization doesn't work anymore, when it did for the timelapse footage.

Resolution had to be decreased to 1280x640 to finish in a lion's lifetime. A faster computer would be justified, if anyone watched these videos.

Eclipse copter (0 min 34 sec)
...Continue Reading