Jack Crossfire's blog - Page 4 - RC Groups
Jack Crossfire's blog View Details
Posted by Jack Crossfire | Aug 12, 2017 @ 03:06 AM | 1,700 Views
With parking selling out just days after the lion kingdom reserved a space, it was clearly going to be good weather. The decision was made to use 5 cameras & try to get a 6th camera from the day job.

gear 360: 4k video from pole

Canon T4I: hail mary timelapse from 9am to 12pm

gopro: shadow bands & temperature

Canon 5D: selfie of lion looking at eclipse from behind

Sanyo: Mount jefferson closeup

Canon 5D mark II: hail mary wide shot with leo

The mane problem with the cameras is setting exposures.

Tests showed the darkest settings which can capture leo are 4s, ISO 400, F2.8. Only Casseopia is visible at 1s, ISO 400, F2.8. This would probably blow out the entire sky. Instead of leo, the lion kingdom will have to settle for Mars & Mercury, the only objects likely to be bright enough to see in 1s ISO 200 F5.6. Tests showed the 28mm gives 37 deg in stellarium, just enough to fit Leo in.

The fisheye will be used for the selfie. The selfie is definitely more important than the Leo shot. Reviving the Sanyo proved difficult. It has some trouble powering up, but was tested to record 45 minutes per battery. It's hard to believe the Sanyo was the mane HD camera for 3 years.

No-one ever talks about their exposure settings for eclipses. The best data is on mreclipse.com, which recommends 1s ISO 200 F5.6 for the absolute brightest. The original idea was to make everything as automatic as possible, but the evidence is the current generation isn't going to get much at all. They're having a real hard time devising shots & using their cameras. What's on the internet is very low resolution.

Pretty much everything is automatic except the star field shot. All the shots are considered hail marys.
Posted by Jack Crossfire | Aug 09, 2017 @ 10:31 PM | 1,791 Views
Tests of focus vs time revealed it did indeed get slightly sharper after 10 minutes.
Posted by Jack Crossfire | Aug 08, 2017 @ 12:04 AM | 1,612 Views
The latest model runs have increased rain chances for Oregon on the day. Only a 500 mile segment may actually see it.
Posted by Jack Crossfire | Aug 06, 2017 @ 10:52 PM | 1,721 Views
Clearly, Samsung learned a 4k camera with fixed focus requires so much tweeking in the factory, an autofocus mechanism pays for itself. The 2nd generation Gear had autofocus. Having decided any serious 360 video would be done after buying another camera, a method was devised to destroy the old one, starting with removing material. It started with the Dremel & continued with the knife. Eventually, the lens was 0.25mm closer to the sensor.

This completely blurred it out. The process shifted to adding material. 3 layers of electrical tape were eventually what sat under the lens. The result was now sharp but slightly smeared. The original was also more smeared than blurry, which leads a lion to believe it's currently farther from the sensor than it was from the factory. At least, having removed material, there's now a way to adjust focus by adding layers of tape.

Important to note, finally left out the battery enclosure, making it just 2 cameras. Not having the battery enclosure pushing on the sensor might have affected it. Running it with no battery & variabilities in the battery enclosure placement might have been affecting it all along.

Also should note these results are the horizon with the camera sideways. Positioned vertically, the horizon is awful. Since no-one cares what is directly above or below, it just needs enough in focus to get the horizon.

Along with this, the shutter bodge wire broke & took an 0201 pullup resistor with it. Having a 1 week lead time for 0201 resistors left putting in a 10k 0402. It had to be tombstoned to fit but made a stronger solder joint.

The mane problem is there's no way to accurately test it besides hiking .7 miles up the mountain in peak daylight.
Posted by Jack Crossfire | Aug 06, 2017 @ 12:35 AM | 1,540 Views
Another test of focus fix #3 in brighter conditions yielded the same results as focus fix #2. Really blurred out on the right edge. Results in the same lighting are in the following posts.

Focus fix #1 was shown here:

Focus fix #2 was shown here:

Focus fix #3 is below.

There wasn't much degradation between #1 & #2, but comparisons are hard because it was rotated. Rotating the camera gives better results in all cases & is much easier to mount. The lighting affected the apparent sharpness. It wasn't temperature related, because the front camera was always sharp.
Posted by Jack Crossfire | Aug 05, 2017 @ 03:57 AM | 1,904 Views
Kind of extraordinary that the 1st Model 3 owners have remaned anonymous & not uploaded any videos. It was that way for the 1st Model X & S owners. They officially only give the 1st cars to employees who are required to keep quiet. After that, they seem to ban anyone affiliated with the media. The media is given specifically vetted cars to test drive.

It wasn't that way for the early Apple events. There used to be a race to be the 1st to tear down every new gadget. A famous website like iFixit would buy the 1st.
Posted by Jack Crossfire | Aug 04, 2017 @ 12:29 AM | 1,954 Views
Reviewing photos, it became clear that the gasket was sitting on the standoffs, so it really was functional. With gasket + standoffs, the entire frame was blurry. With standoffs only, the center was only slightly blurrier than the front camera & corners were slightly better. With nothing, the center was sharper than the front camera but the corners were wasted.

Put the gasket back on & tightened as far as they would go. The grinding of the standoffs made it sit a bit warped, but the gasket gave a good chance the lens area was flat. Got a shot in the last moment before the light was gone. It was definitely better than tweek #2. The right edge was still a hair blurrier on the back camera. In this lighting, it was hard enough to differentiate front & back cameras to conclude it was as precise as can be, with the given tools.

The center is a hair sharper in the back, indicating far sightedness. Better focus might physically be possible from loosening the screws, but it's within the range of heat expansion & age affecting it. It's not clear what happens to the gasket over time.

Also noted interference from the pole is now almost gone. There is just a tiny bit in the top right rear.
Posted by Jack Crossfire | Aug 02, 2017 @ 10:14 PM | 1,520 Views
Today's solar filter photo detected a sunspot. 1/320, F5.6 ISO 400 was required to keep it from blowing out. Auto exposure wouldn't work at all. In still photo mode, with enough room to allow a reasonable amount of drift, the sun filled enough pixels for it to actually be worth it. Definitely need to underexpose.

The pole got another rework with the Samsung sideways. It was much easier to fit all the gear on this way. Sadly, the 2nd rework clearly degraded the corners even if it slightly sharpened the center.
Posted by Jack Crossfire | Aug 02, 2017 @ 01:33 AM | 2,098 Views
A quick test of direct driving the motors was another failure. Recycled the G buggy H bridge yet again. Without any position feedback, pulsing wasn't reliable at all. It also shook quite a bit, because the servo wasn't gradually slowing down. Putting brushless motors in the servos would be the next step up. It would be quite involved. In the last 15 years, a few smaller equatorial mounts have emerged at lower prices. The rise of DSLRs making timelapse movies has enabled smaller mounts.

This one is just the right ascension gearmotor. It has to be polar aligned on a tripod.


That ended another bid to make a 2 hour timelapse. A timelapse could still be made like all the others, by hand tracking & drifting. Despite NASA's best efforts, no other timelapses are as solid as what the lion kingdom got from hand tracking in 2012 & 2007. The 2012 solar eclipse was done without a filter, just cropping a 100mm, F32 & 1/8000 without live view frying the sensor.

Annular Eclipse Timelapses (1 min 22 sec)

Despite only 172 views at a time when unknown channels got a lot more hits, it was the only footage which could have made a decent timelapse.
Posted by Jack Crossfire | Aug 01, 2017 @ 01:07 AM | 1,783 Views
The decision was made to start removing material. Upon going back in, 3 plastic standoffs appeared where the gasket was. Removing the gasket didn't move the lens at all but just flexed it slightly. It was decided to cut off the standoffs with the knife. The result was a rear camera slightly sharper & far sighted in the center, while still awful on the edges. That would be all for removing material, since it wasn't making any difference for the edges. Slightly farsighted is good, because millenials in their broken english have explained as best they can that thermal expansion pushes it slightly nearsighted.

They don't seem to be manually focused at all, but rely on having an extremely long depth of field. Left out the speaker & nearfield hardware during this reassembly. It's lighter but can't use the internal battery anymore.

The best solution if your heart is on 360 video, is wait a while & buy another camera for less money. The 2 cameras combined are already less than a single obsolete Gopro. At least the pocketjuice will work for the Gopro. It's still acceptable for 1920x1080, which is what Casey Neistat used.

The disappointment with the Gear 360 made the lion kingdom decide it wasn't worth spending the next $300 to watch the eclipse. That is a conservative price estimate. You can't just drive there & find a parking space under totality. It requires reserving a space for a lot of money.

The mane thing lions were interested in was a high...Continue Reading
Posted by Jack Crossfire | Jul 29, 2017 @ 08:51 PM | 2,278 Views
with forced air requires powering it externally.

It was crazy hard to make this camera take power from the pocketjuice. The cable had to be under 6ft or it would shut down. Only speaker cable had low enough impedance. The fan cable is still a 4 conductor phone cord with 2 unused conductors from earlier in the day. The cooling system seemed only marginally effective. It hit 36C with forced air & the low 40's without forced air. It hit 56C in a hotter part of the day without forced air, but wasn't directly measured with forced air.

It might work better with the fan blowing perpendicular to the battery compartment instead of angled. This gives air an escape path. The cooling system was rebuilt twice to make it not appear in the frames.

It was fortunate to spend the extra money on an 8Ah battery with 3.4A rating over USB. When derated to reality, it's probably barely enough to feed the camera's 1A & fan's .2A.

The next outdoor test had better focus only in the exact center, but still was horrendous near the edges. It might require removing material. Another possibility is buying another camera for a lot less money & interchanging parts or just getting version 2 & not bothering with 30 megapixels. Because neither camera is anywhere near ideal, it's a lot of fuss over nothing. The quality is just barely acceptable for 1920x1080.

Forced air cooling worked. The cooling system didn't get in the shot or cause camera shake. The 2 frames could be...Continue Reading
Posted by Jack Crossfire | Jul 28, 2017 @ 12:30 PM | 1,812 Views
It took a whole commute just to install the board & fix the firmware enough to make it work. It's a very high density board built around the size of the ATMega. There wasn't enough room for a button, so the only way to set the timing is soldering on a 5 pin programming header & flashing it. There are pads for a serial port, so it could be configured by a VT100 interface. It's started by shorting a jumper. Losing the jumper would be fatal.

Of course, any bluetooth module would be the same size & have a wireless interface, but there wasn't enough time to build up the software.

This revealed the Samsung takes .5A when idle & spikes over 1A when shooting a still photo. For video, the current was a constant .8A. When shooting photos at full speed indoors, the heat reached 55C or 131F.

It immediately became clear that the Samsung loses its settings without power. The mane setting is a timer which defaults to 2 seconds. Retaining the settings would require always having a battery in, so terminating the offboard battery connector.

The decision was made to shoot photos as fast as possible. The mane limitation is compression speed rather than SD card speed. UHS 1 or UHS 3 cards only got only 1 photo every 1-2 seconds. It buffers 3 frames while writing.

The debouncing on the shutter button is very forgiving compared to Goog's non deterministic, fuzzy debouncing algorithms. You can make the intervalometer fire as fast as possible while it's still...Continue Reading
Posted by Jack Crossfire | Jul 27, 2017 @ 03:50 PM | 2,064 Views
Cutting a $32 sheet of silver impregnated mylar is stressful. Be sure the CA glue has cured before applying the filter, or it'll fog up.
Posted by Jack Crossfire | Jul 26, 2017 @ 04:54 PM | 2,497 Views

A $250,000 plane which costs $19/hour to fly & can reach useful distances, if you believe the pitch. You could buy 8 & fly to work for the same cost as living near a job. Whether or not this incarnation ever materializes, battery prices may be low enough to cause serious disruptions in population distribution. To be sure, much of today's decisions are cultural rather than economic. The current generation is obsessed with experiences over money, devoting all its output just to live in the middle of a city.

Despite that cultural boundary, cheap batteries have already slightly impacted populations in cities, in the form of electric skateboards allowing commuters to go 5 miles farther. For many years, this market studdered between segways, hoverboards, unicycles, & scooters, but this year seems to be converging on the electric skateboard. While millenials will never choose money over location, they are being allowed a choice.

If few enough people chose to leave cities & fly to work, it could be a massive boom to those who did. If enough people left cities to depress housing prices, the government would most certainly impose drastic measures to stimulate housing prices. It definitely wouldn't be a choice if enough people did it. There would probably be negative interest rates & huge asset inflation until a house in Montana suddenly cost $80 million.
Posted by Jack Crossfire | Jul 26, 2017 @ 03:45 AM | 2,167 Views
The 1st test of the Gear 360 on the pole was a failure. The Samsung recorded a 4k timelapse for only 15 minutes, overheated & stopped, but didn't power down. The Gopro recorded 4k 30fps for the full 28 minutes. Mechanically, the setup worked quite well. Would definitely brace with the big tripod while setting up the tent posts. It wobbled in the wind. It couldn't be extended completely. The wobble was quite obvious in the gopro footage, but the Samsung stabilized all of it out. Sadly, the Samsung's rear camera was out of focus.

The improvement from elevating the camera 11 ft was only obvious nearby, while the horizon was virtually unchanged. It might get over people's heads. Combined with the wobble, a quad copter would have been a better solution.

Important to note the Samsung only records timelapse in 3840x1920. Only single photo mode takes 7776x3888, in which case the broken rear camera was horribly out of focus & the edges of the FOV on both cameras were worthless. Photo mode does extract more detail from the center of the cameras. Making a timelapse would require cracking it open & wiring a remote control to it.

The most reliable way to start recording was decidedly pressing the buttons with the pole retracted & then extending it. Use a mirror to view the Samsung's display. Any wireless method is more risky.

The Samsung's long rehabilitation began with repairing the focus. It required digging all the way down to the...Continue Reading
Posted by Jack Crossfire | Jul 26, 2017 @ 01:33 AM | 1,908 Views
It's the 1st graphics card a lion paid for in 19 years. The last one was a $50 AGP with no 3D, which could barely play low definition video. For 17 years, lions got by with leftover day job parts & donations. After 10 years, a version of Linux which could decode H.265 no longer supported the Quadro FX 3400. It probably would be cheaper to use Windows & have Linux in a virtual machine, nowadays.

This one was $135 & way overkill for the ancient CPU. It uses far less power than the 3400 & doesn't need the extra power connectors lions long lived with. It was all about getting the most driver longevity. The 1050 came out in 2016, so the drivers should last until 2025. Suspect PCI Express won't be around by then. The current CPU is 7 years old, so an upgrade could be within 3 years. There's marginal improvement in 4k video.
Posted by Jack Crossfire | Jul 22, 2017 @ 02:24 AM | 2,556 Views
Then it arrived. It was a brick compared to the gopro. Something would have to be done to protect the lenses on a vehicle. It didn't output JPEG photos. Its timelapse movies were all H265 files. Its MTP implementation barely worked. It didn't work with any phones in the apartment. Fortunately, all the required modes except remote control were accessible from the hardware.

The lack of video editing software was disappointing. It'll take reinstalling Linux merely to compile an H265 decoder, which is about as bad as reinstalling Windows. The decision was made to have the Samsung make timelapse photos on a tall pole, with the gopro making 4k video right below it. Starting them up remotely & cooling them is the tricky part.

Things didn't go so well in the narrow angle video department either. The t4i doesn't support Av or Tv in video mode, only fully automatic or fully manual. Fully automatic might actually be enough. Narrow angle video isn't the mane priority.
Posted by Jack Crossfire | Jul 18, 2017 @ 01:18 AM | 2,999 Views
The absolute best eclipse movie possible with today's technology would be a realtime 8k in hydrogen alpha. This resolution could reveal individual mountains of the moon overlapping prominences of the sun, without zooming in. The only videos the goo tube comes up with are extremely poor quality. Thus, the lion kingdom expects similarly poor quality of the 2017 eclipse.

Unfortunately, the idea of rigging the servocity to track the sun is not looking good. Testing with the trusty signal generator showed PWM to not be accurate enough. Direct driving the servos with H bridges doesn't look good either. They have to get above stall speed & wait for inertia to slow down. The minimum speed it can step is too fast. These problems didn't happen when using wide lenses, but are deal breakers with the 200mm.
Posted by Jack Crossfire | Jul 16, 2017 @ 04:44 PM | 2,179 Views
Based on the research, every section of totality near a road is going to be packed so there won't be any quad copter flying. There are very few roads & all the land owners have already rented out parking spaces for a lot of money. A bathroom is required for such a mission, but all the bathrooms are going to be near said parking spaces. It's burning man on a nationwide scale. It won't be possible to book a parking space at the last minute, based on weather. Polecams are looking like the best move, but there are going to be a lot of tall RV's. Would say the protocol is pay a huge sum just to park the vehicle & then hike somewhere.

Having said that, there are a lot of do or die events lions never saw besides totality. Lions never saw a space shuttle lift off or got married. The mane attractions are the chance of a shot no-one else is going to get & keeping up with the Joneses. The best chance of someone else getting the shot is given by the 360 craze, but this craze has run its course & there are now very few 360 cams around, let alone operators interested enough to bother mounting them on a pole.