Eclipse Preparations (EOS-M edition)
9/28/2023. The little Canon EOS-M worked beautifully as a video camera for the 2017 eclipse, and I intend to use it again next year. To that end, I've bought a cinematic cage for it ($10 on eBay), a dummy battery with USB-C power connection (Amazon, $20), and a tiny 10,000 mAh power bank (Amazon, $23). That should address the one weakness I've encountered with the camera: it's got a tiny battery and can run through it fast enough to make me a little paranoid about not-to-be-missed moments. The last thing I want to think about is battery power while the Sun goes out. Being able to mount the battery on a cage will make the camera a self-contained rig with plenty of run time.
I need to update the Magic Lantern firmware and practice with it. I should add a bigger, faster SD card. The R6 will share the equatorial mount behind the 65mm or 92mm APO or the 500mm F6.3 mirror lens. I expect the R6 to be set to auto-bracket and to run with its electronic shutter throughout totality either unattended or with minimal attention. The EOS-M will simply take video from a few seconds before 2nd contact to a few seconds after 3rd. It's point is to get the contact phenomena, so if needed I could start and stop it twice. Practice is in order.
9/28/2023. Later that same day... the battery pack and cabling are here. Seems to work fine. The cage will arrive tomorrow.
There's now something called "Crop Mood" in the Magic Lantern depository. This latest version seems to bring deep bit-depths to the full-resolution cropped modes that I adore with the EOS-M. But there's more to work through. 14-bit anyone? 12? 10-bit? Maybe forget the Dual ISO features if I can video in sufficient bit depth (or use them with the extended depth for really extraordinary dynamic range). I've downloaded the August 2023 build and the dedicated video editing code. Some hints, including notes about settings prior to invoking ML for maximum stability:
https://www.youtube.com/watch?v=JPrt-PV1wn8
But also search out and study Zeek's videos about the EOS-M.
9/29/2023. Crop Mood firmware installed! There have been many, many changes since I installed Magic Lantern back in 2016 or 2017, so there'll be some relearning to do and instruction sheets to work up. S'OK, I was way out of practice with the old stuff anyway. The new stuff looks extraordinary. Figure out focus peaking because it might save the temptation of trying to fit a larger monitor.
Fumbling through the capture and output software, I've discovered that Magic Lantern (aka Crop Mood) produces .MLV files which need to be refined through MVLApp to be turned into video accessible by standard editing tools like, for example, DaVinci Resolve. I have v15 of Resolve, and it works, at least sort of, on this machine with the Codecs on hand, so let's not poke it. I can also output straight to AVI, or to a myriad other formats.
Obscure notes I may or may not need later: DaVinci 15 runs with my GPU preference set to CUDA and the GPU on the NVDIA GeForce 960 selected manually. All other settings (so far) produce "error -59" messages.
Here's a sketch of the workflow: open the imported clip (select the folder with the MLVApp output, then select the tons of .DNG files within) and, I dunno, drag it into the right panel? (Is that *the* way, best way, only way, a silly way?) Instead of looking for an "export" or "output" or "save as" button, use the "deliver" button at bottom right. This will add the "project" to the "render queue" which you will then need to execute, or render or whatever.
Presumably, there are a shit-ton of options (think Photoshop with motion) to explore prior to pressing "Deliver" as well as seemingly endless options in Magic Lantern and in MLVApp to be tried out, adopted, and made routine.
Keep in mind:
- With today's settings (export as DNG uncompressed) each 1920x1080 .DNG frame is 2.533 MB. It's nice to be able to load them individually in Photoshop, but count on generating ~76MB/sec at 30 fps. A 4m30s clip (enough for 2024's totality or a complete Moonrise) would therefore involve 270s x 76MB = 20GB of data. That's not as far out of line as I expected -- it's not too different from the 1,000 frame 16-bit .SER files I use routinely for solar photos. (Note that "DNG Lossless" output produces files of about 575kb and still includes RAW corrections; "DNG fast pass" does not include at least some RAW corrections.) Photoshop appears to open everything with more than 8 bits as 16-bit. These DNGs are actually 10-bit, not the richer 12- or 14-bit files I would prefer for totality. Experiment and be not intimidated. At worst, deeper files might be almost twice as large. Today I was using a 1-second portion of a 5-second clip just to find a through line from camera to viewable video. That's not an ambitious plan, but it was a crucial step, and that much is done.
- What are the little red dots I see in the DNG output files in Photoshop? [See note below.] They are pixel for pixel identical in two almost adjacent frames, so I'm thinking that they are actually hot pixels. In that case, I may need a dark frame to control them, but I've just started guessing. There may be an option or parameter somewhere that will get rid of them. Since they are single pixels, noise reduction works well, but that would take a while to apply to a usefully long clip.
- Figure out how to get MLVApp exposure, contrast, color temperature adjustments to be reflected in the DNG output files. (Maybe that's for AVI etc only? Changes made in that menu do stick to AVI output.)
- My current Magic Lantern-loaded SD card (64GB Lexar, 150 MB/s, UHS-2) does not permit me to record fast enough for more than 10-bit video despite its having a speed rating considerably faster than the preferred card. I found the recommended card (SanDisk Extreme Pro, 95MB/s, UHS-1) on eBay ($26 for a 128GB card, compare to $70+ on Amazon). The UHS-1 thing is apparently important since, evidently, that's what the EOS-M uses to write. Faster speeds with UHS-2 or -3 mean nothing to it. I see YouTubers having success at 14-bits with EOS-M's and this card. Even 12-bit data would be a big improvement for eclipses. Get the rest of the workflow down and be ready to try for better bit depth next week. I keep trying to buy 95 MB/s and 170MB/s card and vendors keep "helping me out" by sending faster cards. Zeek recommends the 170, others the 95. Turns out that the SanDisk ExtremePro 200MB/s UHS-1 cards work well, too. If anyone ever actually sends me one of the cards I order, I'll let you know how it goes. (Note that a 32GB 95MB/s ExtremePro that I had in the Fuji100T doesn't work particularly well at all.)
- Stop messing around with halfway measures as far as the rig goes. Don't tape, rubber-band, or velcro the powerbank to the frame, use a Small Rig clamp made for the purpose. It's $23 which seems high until I start thinking of the reduced functionality and convenience of kludged "solutions." Such things tend to compound and grow toward mediocrity. Buy the good stuff when you can.
Note about (2) above. The red (and some other color) dots arranged in a complex pattern across the frame, more visible in bright, strongly stretched areas, are "focus dots," an artifact, presumably, of the EOS autofocus firmware. MLVApp offers the option of fixing them, but its configuration was not obvious. Not only do you need the files defining the focus dots for particular camera models on board for MLVApp to find, you must drag them into the work area of MLVApp prior to processing the video. This works! The fix appears to be durable; these files do not need to be loaded each time MLVApp is fired up.
Also (2), MLVapp has provision for a dark frame. Shoot some and play.
Is the 300mm Rokinon mirror lens sharp enough? The Moon shots last night were abysmal, but they were handheld. Their fuzziness may just be a focus issue. Put it on a solid mount and give it a try. The 300mm Nikkor did great in 2017; it would have done better if I had actually remembered to use it at F5.6 or F8 instead of F4 where I was focusing it. Focus with the Nikkor would be significantly more confident. How much larger is it, really, than the 300mm Rokinon mirror? (Not horrifically...)
10/01/2023. I filmed a Moonrise for the first time in ages. I began with the old 8-bit 3x crop mode for reference then grabbed several raw clips using 10- 12- and 14-bit formats to see what the limits and framing looks like using the suboptimal SD. At 24 fps, I can do 12-bits at almost 1080p (1000+ in any case) and the framing would be fine for a total solar eclipse. 14-bits would be tight. Really, all a faster SD will get me is room for drift and make 14-bit data more comfortable. It might also get me 30 fps. All that is worthwhile, but clearly a RAW totality is in reach right now -- barely so -- with the card already in hand. Taken all together, the advantages of speed are considerable. Now work on dark frame methods and focus etc while waiting for the preferred SD to wander here from California. Tonight I just cranked the focus ring over to the infinity limit, tweaked it slightly as an airplane flew through the field (no bright stars being really handy), stopped down to around F8, and let it rip. The powerbank worked fine taped to an L-bracket just for the moment. I didn't record exact settings -- this was only my intro to RAW video, a plausibility check. The details will change with experiments, practice, and faster memory.
Moonrise, October 2. When I couldn't get the RAW modes to behave on my second night out, I messed up a lot of options trying and then ended up just shooting some stills in original EOS mode. Here, a long exposure (300mm, F4, 1s, ISO 6400) catches the Halloween-colored Moon lighting up the terrain. A bonus is the Pleiades rising over Barrett Ridge, 18 miles east of the community lot. Click to biggify.
10/04/2023. The powerbank mount is here and mounts very nicely on the cage wrapped around the EOS-M. I corrected the error I introduced into the previous Magic Lantern installation during Moonrise, and all seems solid (again). I also tried moonshots using the R6, the 500mm F6.3 Rokinon, and a 2x Nikkor teleconverter. Sharpness looks good. The tripod foot I hacked onto the Nikon-EOS-R adapter works much better wrapped around the teleconverter instead. I think this 1000mm configuration will be the eclipse combination for the R6. It's slow, but with tracking, stacking, and modern CMOS sensors, slow is OK -- it's two to three stops slower than anything I used in 2017, but I was not short on light in any case. I can go up to 1/2 second in electronic shutter mode (1/2 second, F13, ISO 800 or faster should do, especially since I expect to stack lots of images).
Framing with the 500mm Rokinon and 2x converter on the R6.
Also a plausibility and sharpness check of the combination. 1/3200, ISO 6400, handheld.
Here's the beginning of 2017's totality presented at the same "plate scale" as the previous image.
It was re-scaled from a file acquired at approximately 402mm in an attempt to include a starfield.
10/05/2023. I put both the EOS-M / 300mm and R6 / 500mm x 2 on the SWSA. The sheer mass is not an issue (it's a fraction of the TMB solar or deep-sky kit). I hoped each body and lens would counterweight the other, and that works, though the rig needs to be adjusted carefully. Framing will be critical for the EOS-M / 300mm working in a cropped RAW mode, so I put it on the geared declination arm and left the R6/500mm on a small ball head. Balance requires that the R6/500 be set up with the camera mounted "outboard." Pictures will make this description clearer, but I want to see it all work before bothering to document it. Also, while working out the balancing act, I've got the SWSA mounted on a heavy Linhof tripod. For airline-based eclipse-chasing, it will need it to perform on something considerably smaller, probably my Siri carbon. And I'll try a heavier ball head under the 500mm; the small one there at the moment looks insufficient but feels solid.
Now then, back to EOS-M concerns. The faster SD card should be along in a couple of days. [I received a SanDisk ExtremePro 200MB/s -- not what I ordered, but it does work well.]
10/06/2023. Balance, tracking, alignment, damping. I can more or less control all that. What part does wind play? I moved the kit onto the for-real eclipse tripod, the bargain carbon fiber I took to Alaska in '17. It feels and looks like it should work really well. It is and could remain very low, so wind-screening in Texas (typical April wind, 15mph) could be done with a sturdy umbrella.
10/07/2023. Another night, another Moonrise. For Moonrises, I'm going to need to use a time-lapse configuration. That's neither here nor there for the eclipse, but it'll be fun anyway. On the YouTube, Zeek knows his stuff! Study his videos, hardware, settings, and build for Moonrise, for the eclipse, for everything. Here's a still shot at the end of the session around 1:30AM on 10/8 (Raw, 1/4s, ISO 6400, 200mm F2.0 Nikkor). Apparently, I could do this or something very much like it as part of a raw 4k timelapse all from inside Magic Lantern settings. I do not, however, intend to sit up until 2:30 and try to figure it all out tonight. I can practice on clouds by day.
Mundane lessons from last night: (1) do NOT forget to pack the Vixen to Arca adapter. It's OK to hurry back from the community lot for it but not OK to do so from Thunder Hill, let alone from San Antonio. And (2) the neat little 10,000 mah battery pack mounted to the EOS-M rig seemed essentially eternal. It took a long, long time to go from 100% to 50%, but it then went to zero in the space of a few minutes while I tried out a few modes this afternoon. I've ordered some USB-A to USB-C adapters so I can use other USB power sources already on hand [these may work for data but they apparently do not transfer power from the original USB ports; a 20,000 mah battery pack was on deep discount during Prime Days and will be here soon]. Also, a couple more Extreme Pro cards are enroute: a refurb from Adorama and a new one from Amazon recommended by Zeke. The Amazon-supplied card is a current model, so new ones are available all over the place in all sorts of sizes (Amazon supplied a 200MB/s version, but it is UHS-1 and turned out to be OK). Other maybe-worth-its include a cage for the R6 if it would let me use the geared-down timing-belt contraption focuser I put together last year.
10/09/2023. After three days lost in the Magic Lantern labyrinth, I have finally managed to get ~4000x2200 12-bit RAW 24fps video out of my EOS-M. The crop factor is only 2.03. The key was choosing the 1:3 anamorphic mode and letting MLVApp sort it out. This mode "runs green" with the SD cards on hand. Better may be in the offing, but this will do. This mode should have plenty of resolution, bit depth, and sensor real estate for the eclipse. 4m30s of video in this configuration will come to something over 20GB, so be prepared.
I spent some time shooting sample videos to get a really good sense of the scale this mode presents and decided that the eclipse really needs about twice the 300mm focal length of the optics I have on hand. I tried a 2x extender behind my 80-200mm Canon, and the resulting 400mm scale wasn't enough. The obvious solution is a 2x teleconverter behind the 300mm Nikkor, but mine will be deployed behind the 500mm Rokinon to produce an EFL of 1000mm for the full-frame R6 (see above). I shopped for a duplicate in Japan, Canada, and all over the US before finding a good deal at Adorama ($40 even after tax and shipping). The doubled 300mm Nikkor will give me a 12-bit (or better) solar image about 1200 pixels across which is only slightly smaller in proportion to the 2200-pixel frame width than what I got in 8-bit compressed video mode in 2017. True, it will be F9 rather than F4.5, but there's plenty of light at 2nd and 3rd contact, which is what this is really all about. The R6 can take care of mid-totality. Beads, we got 'em.
You're going to need to know this: a 30 arcminute image is FL/115 wide at the focal plane. Watch your units: that's 600mm/115 = 5.22mm. The EOS-M's pixels are 4.3 microns across, hence 5.22/0.0043 = 1,210 pixels. The R6, behind a 1000mm EFL, will be dealing with an image 8.7mm in diameter. Its pixels are 6.56 microns, so the Sun will be 1,325 pixels wide. That sounds like a duplication of effort, but the R6 offers a larger field of view and will be operating across a much wider exposure range. After I get the EOS-M options nailed down, I'll get started on the hands-off sequence for the R6. The EOS-M is all about beads, diamond rings, and the chromosphere; the R6 is for the corona.
Experiments with Dual ISO. Remember you can set the info display in Photo Mode rather than in Movie Mode in case you want to assign toggling Dual ISO mode on and off using the Info button and still use the same button to tailor the display. Just switch over to "photo" mode and change display styles there.
I just messed around using 200/800. I would rather do 100/800 or maybe even wider (I think I had ISO set to 400; if I set the ISO to 200, DualISO uses 100/800 -- I haven't sussed out all the rules and options yet).
Productive processing steps, wide gamut and "log" output options in MLVApp will all take their own experiments. My first tries suggest this is promising but very slow: 5 second clips took about 5 minutes to render. So a 270-second totality video would take about 4.5 hours, give or take. And the output is huge. Five seconds, 346MB of MLV video yielded an AVI of 1.8GB. It's a start. The output IS beautiful. The size and lack of speed is not down to the Dual ISO feature; videos of this mass simply have their demands.
The output needs to be graded. In Photoshop, import the file, make it a smart object, adjust, and render. Which with curves and levels applied (does that matter?) is about as slow as the MLVApp step. Premier Elements is worth a try, but if I recall previous experiments correctly, these demands are enough to choke that software on this machine.
Remember to uncheck the Dual ISO module if you want the INFO button back the way it was. I think you can simply toggle the DualISO mode off and then set higher ISOs as well as 100.
I've updated my backup SETUP directories to this configuration just in case things go to hell. Note that if you want to make a backup card, you need to format it in the camera and then do the ML firmware update again with the second card. Then copy the /ML/Settings files from the instance you want to clone into the new /ML directory, and Voila!
10/10/2023. Got a look at the Moon while I walked Gemma this morning. It was almost exactly as high in the sky as April's eclipse will be. I came home and ordered a bunch of cheap mirror tiles to see about rigging up a way to see the back of the EOS-M when it is aimed up at that steep angle while mounted close to the ground. I intend this mainly for composition, but could also use it for focusing if focus peaking works when pointed to the Sun. Plenty of time to try that. Maybe things won't be as cramped as I suspect; it will still be good to be able to look over casually and check composition without getting flat on the ground.
With the 200MB/s ExtremePro cards, 12-bit runs green; 14-bit runs green sometimes, orange and red at other times. The subject matter matters. Bright, high-contrast subjects flood the buffer and recording eventually stops ("eventually" means in 2 - 30 seconds, for example). Unless something changes before April (a genuinely fast card, a slightly less demanding video size...), I think it would be foolish to use 14-bit for an event where a second take is not possible. A solid 4k 12-bit 24fps solution is not bad at all. There might be usefully similar modes that would take just a little less bandwidth. For the record, 14-bit frames export as 2.8MB DNG files which open to occupy >55MB in Photoshop. That's over a gigabyte per second for the rendering machinery to chew on.
In DualISO mode, processing takes about three seconds per frame and a little less than half that in single ISO mode.
UHD mode (standard 4k) is a slightly tighter crop than the 4.2k mode used to investigate the options above. UHD is much more amenable to running in 14-bit mode. Play with it some. That might be the ticket for April.
10/12/2023. Good lord! Can't anything be straightforward? The TC200 arrived and would not mate to the 300mm F4.5 for anything. It wasn't the focal length or optical compatibility per se, but a mechanical meter-coupling fiasco. That is something for which I have no need at all. I ended up taking the telextender apart, removing the meter coupling hardware (more extensive than I ever imagined) and putting it back together as a stupid tube. Eventually, on the third try to get everything aligned as it should be, it looks, feels, and works exactly as I need it to do. That was much more exciting than it needed to be.
10/13/2023. Actual solar images prove the framing of the two cameras in the modes I intend to use in April. The weather forecast for tomorrow says I might get to see some of the partial eclipse after all. I have ideas.
Here's a still frame from the R6 with the doubled 500mm Rokinon F6.3:
And here's one frame from a 4.2k, 12-bit, dual-iso video from the EOS-M with the doubled 300mm Nikkor F4.5:
Nevermind the settings. This is just a plausibility check. There'll be a lot more experiments before the big day. These were casually focused looking through Baader Solar Film, handheld, juggling the camera and filter. The color has been tweaked in post to make sure it can be. One lesson from these trials: clean the sensor before the show, and consider making some flats, just in case. I worked out an easy and effective mounting scheme for the timing-belt based focusing rig for the 500mm. It could be easily transferred to the 300mm, but I think that lens is pretty well focused at the infinity stop, and besides, I can stop it to F8 for 2nd and 3rd contact (light is not in short supply) which would just about guarantee good focus.
Today, Adorama delivered a genuine SanDisk 95MB/s ExtremePro. This card comes so highly recommended for recording EOS-M RAW video that I expected it to smoke the 200MB/s cards I've been working with so far. It didn't. Like the 32GB version I already had on hand, the much-anticipated 64GB 95MB/s card is notably slower than the newer 200MB/s cards. (And don't say, "D'uh..." There were reasons to think otherwise.) With the newer cards I can do 12-bit video any time and 14-bit sometimes. With the supposedly ideal older cards, I can do neither at any time. I'll keep messing with this, but I'm happy enough with what I have.
And that is more than enough for this long page. [There's more at Page 117 et seq.]
:: top ::