squig

16 minutes ago, mercer said:

Well, it only stopped recording once with the recording limit warning asking me if my card is formatted as ExFat.

Can using a card not formatted as ExFat cause any other issue other than recording length?

I'm using the March 30th builds. I just noticed the issues. I was planning on doing more research tomorrow, just thought it might be a common issue.

The only thing I can think of is, even though I could not format the card as ExFat, I did erase the files on my computer with my previous tests... whereas with this test I reformatted in camera. 

If the card's not formatted exfat you can only write up to 4Gb per recording. Best to format your cards exfat to determine whether that's the issue. I'm still using yosemite as my main system, exfat formatting works.

Alternatively, you could try an earlier build.

1 hour ago, mercer said:

Has anyone else experienced any of this?

Nope, never. Is it happening on recordings under 4Gb?

What's the date of the build you're using?

17 hours ago, SullyCortez said:

I'm just gonna go ahead and say it..... who cares??

Well no one clearly.

As for the rest of your post: If you're gonna masturbate, don't do it in public.

1 hour ago, Jimmy said:

I don't wanna sound like i'm moaning... but is there a reason the HFR modes aren't being tested, is it not implemented yet?

They are being tested, but they're quite buggy. I sold my A7s too, the MK3 raw is so much better.

Actually, looks like rolling shutter is a bit higher: 1080p 24p standard 1x crop raw is 18ms. 3K 16:9 is 31.9ms according to ML. a1ex may be able to reduce that a bit if and when he enables FPS override.

3 hours ago, cpc said:

So how's the rolling shutter in the higher resolutions?

3K 24p rolling shutter is about the same as 1080p, <20ms

7 hours ago, silvertonesx24 said:

That would have been a waste of money. Since Adobe removed multi-processing after CC 2014, good luck getting it to use more than 200% (out of 2400% possible) CPU. Maybe on a good day it'll use a third of the total CPU power.

Na, I'm still using CC 2014 for that very reason.

They're doing a pissy update after 4 years. LOL. So glad I decided against getting the trash can and bought a second hand 2010 6 core 333Ghz tower. I put a GTX 980, 12 terabyte raid 5, and 2x terabyte SSDs in it. It screams, and cost me half what the trash can and external raid array would've cost. My only regret is that I should've spent a bit more on a 12 core, After Effects rendering is CPU based. The trash can is cool, but it should've been introduced in addition to the towers.

16 minutes ago, SigurdW said:

Have been curiously n00bing with the experimental builds the last few days and still trying to get my head around the actual crop factors as RAW video displays 1.00x, no matter what I change in the Crop mode module. Judged by the suffering live view there is a difference, but I might be wrong.

Will the crop factor change accordingly to the resolutions or is it constant ≈1.6x from 1920 up to >3k?

(Btw, if there is any magic lantern moderators here, I would have loved to get my account approved :D)


Can not express how much I look forward to the day its included in the nightly builds!

Cheers

The crop factor reduces as you increase the resolution. Select 3x crop. Click the trash button to get out of the ML menu. Click it again to get back into the menu. Go into the raw video menu. Select aspect ratio. Use the joystick thingy to scale up the resolution. You'll see the crop factor when you click out of the aspect ratio screen.

1 hour ago, independent said:

Low light? Can somebody post a high ISO / low light still or video? Aside from the resolution penalty (who cares), I'd like to see what the noise patterns are like. 

Focus is a bit off, but this is 3K scaled down to 2K @ 1600 ISO. If you wanna know what the noise pattern looks like just look at any raw 5D MK3 photo. Things tend to get pretty noisy about 1600 ISO, that's where the digital gain kicks in, but with the right exposure up to 6400 ISO is doable.

11 hours ago, cpc said:

The idea is that a 3x3 binned image draws from 3*3*1920*1080 photosites (if it is true binning), whereas a 2.7k image (say, 2700*1520, can't remember the exact resolution) would be based on ~4.25 times less photosites, which should affect SNR significantly. But yes, if you are consistently going to shoot 2 stops lower ISO sensitivity, it should be fine.

I did some tests: I shot standard 1080p raw vs 2.8K raw at the same ISO and aperture, taking a few steps to get close to the same framing; the 1:1 2.8K crop appears to be about half a stop brighter than the 3x3 binned 1080p. Somebody else might wanna test that to confirm.

5 minutes ago, mercer said:

So is this something you can utilize for your feature? If so, how will you handle synching your audio without the sound... I guess just visual cues from a slate... or does ML have a timecode feature?

Yeah if a1ex gets the live view sorted out. Pixel binned 1080p is a little softer than I'd like. 2.7K gives a little bit of wiggle room for the softness of the Iscoramas, and for denoising. Also the crop means I can shoot the low light stuff at lower ISOs, which is cool. I've already got my eye on that new Sigma 14mm f/1.8; that plus a Sigma 18-35mm will cover the wide end nicely. Audio's no problem, I've got a clapper board, and I dual system record anyway. Plus ML can beep at the start of a recording. I think ML will incorporate audio in time, I've just asked about whether they can include card spanning into the build, that would give us another 20MB/s to an SD card

2 minutes ago, cpc said:

So is the resolution gain from 1080p to 2.7K actually worth the (probable) loss of sensitivity and wide angle abilities? I'd figure binned full frame 1080p is still the best bang for buck? Maybe a direct comparison of noise levels will shed light on this.

Well I've got an ultra-wide lens so I can still get a 20mm equivalent FOV so that's not an issue. Loss of sensitivity? I don't see any. Noise wise for an equal FOV and DOF to full frame you're shooting at f/2.8 100 ISO instead of f/5.6 400 ISO, so I don't see how it could be noisier. You are zoomed in on the noise though, but to my eyes it's the lesser of two evils. It's definitely worth it for the extra detail, it's now has the resolution of an Alexa!

22 minutes ago, Andrew Reid said:

Why is 1.1.3 better?

It's slightly faster.

I tried my Lexar 1066x card, it's a little faster than the Toshiba. 3K 2.39:1 is repeatable at 100 ISO. I had to go back to 2.7-2.8K at 1600 ISO. That works for me, I'll shoot 3K 100 ISO for wide outdoor establishing shots, 2.7K for everything else, and render to 2K scope.

1 hour ago, mercer said:

I actually don't have a choice but to shoot with it. The only other camera I own now is the D5500. And although I love the image out of that camera... I'm not going to shoot a short on it when I have a Raw capable camera lying about. I'll just make sure I have 4 or 5 batteries, 3 or 4 64GB cards and my MacBook Air with me to offload the files if needed. It's only going to be a 3-5 minute, mostly silent, short. So I am preparing it in such a way that I am shooting a 3:1 shot ratio. 

You've used ML on the 50D so it shouldn't be a problem. Use the raw histogram and the ETTR hint. I have my LCD brightness 1 notch from the minimum setting which gives a good exposure preview for ETTR shooting. You can use shutter fine tuning to get a 180 degree shutter. Set FPS to exact. If you're using an external monitor you may have to switch the camera on before you turn on the monitor.

59 minutes ago, Alex T said:

Does the 14bit lossless compression work for 2k or 1080p full frame? Also what kind of card are you using?

14bit lossless works on any resolution. I've been using a Toshiba 1066x.

35 minutes ago, hyalinejim said:

I can hit 3264 x 1366 continuous on ISO 100.

What cards are you using?

1 hour ago, hyalinejim said:

OK, I'm pretty sure that the variable compression is due to ISO. In crop_rec 3K (3072 x 1286 ) I can get continuous recording at ISO 100, but if I whack the ISO up to 6400 I get eleven seconds. Can you check if it's the same for you?

100 ISO - black screen ~49%
6400 ISO - black screen ~60%
6400 ISO - busy scene, fast movement ~69%

Explains why you're getting 3K and I'm getting 2.7-2.8K, I've mostly been shooting at between 1500 and 6000 ISO

2.7K scope 14bit lossless. Debayered in ACR, graded in AE with filmconvert. Rendered to 1080p. Sigma 20mm f/1.8 @ f/2.8 1500 ISO

39 minutes ago, Andrew Reid said:

Max height is 1920 but of course depends on bit depth and card speeds

Still some questions for you early mavericks out there -

Does the lossless 58% JPEG compression mode work in 3K and 4K?

How's the quality of 3840×1600 at 10bit? Any DNGs to share?

You can pop the plastic back off to make most EF-S lenses fit the 5D and 1D C. I have used the 18-135mm EF-S on my 1D X Mark II with no issues.

The lossless compression isn't a fixed ratio, it varies; the highest I've seen it is 51%, the lowest was about 80%.

The way it works is you have to go into the crop mode menu to select the sensor crop, 3K, UHD, etc, then you have to go into the regular raw video menu to select the resolution you want to record. 10bit @ 2.7K 2.39:1 is around 90MB/s, roughly the same as 14bit lossless. As I said, there is no 4K 24p recording, just a 4K sensor crop which no available CF card can record, the media is just too slow to write that much data. I've recorded up to 3.3K 2.39:1, but because the compression size varies you can't expect to record every single shot at 3.3K. I'm trying to figure out what resolution it will do 95-100% of the time, so far around 2.7K 2.39:1 24p with a 2.1x crop is pretty solid.

8 hours ago, Andrew Reid said:

Before the 14bit RAW with ML was pixel binned 1080p

This is like cropping a full res RAW still, 1:1 pixel readout

It is not just going to step up the sharpness and detail.... The dynamic range and noise performance should improve as well.

Astounding work.

You should mention in your article that there's no CF card that's fast enough to record 4K 24p 14bit lossless, somebody might buy a 5D Mk3 thinking they can. The highest resolution I can record continuously is 3.3K 2.39:1, but continuous doesn't mean endlessly repeatable. The lossless compression needs a bit of headroom, so 2.7-2.8K is a more realistic repeatable recording target. I don't know what the situation is with lossless compression for 10 and 12bit, that would certainly get us closer to 4K.

21 minutes ago, tweak said:

What's the upper limit now for Anamorphic ratios like 4:3 or 5:4?

You can record around 3.3K scope: 2272x1364 with a 1.5x anamorphic lens; maybe a little higher, but best to leave some headroom for the lossless compression.

Around 2.7-2.8K is looking like the ballpark for sustained 2.39:1 24p 14bit lossless shooting with a 2.05-2.1x crop. Higher than that isn't repeatable 100% of the time in my tests.

@hyalinejim did you shoot that with the 1st or 2nd build?

8 minutes ago, hyalinejim said:

https://vimeo.com/211239982

 

Great, you saved me the trouble. I'd rather spend my time looking for bugs than cutting and grading. If people wanna know what it looks like; it looks exactly like a 5D MK3 photo cropped to 3K, there's millions of those around.

Are you having any issues with this: http://www.magiclantern.fm/forum/index.php?topic=19300.msg182227#msg182227

7 hours ago, jpfilmz said:

I'm in 1.13 and getting pink colored fames using MIRaw Viewer....but it's only with the 14bit lossless setting enabled.   

Use mlv_dump https://drive.google.com/file/d/0B-7nnVbwzGVkUmpHLW5kbEU3OWc/view

8 hours ago, Andrew Reid said:

Try the 3840 x 1600!

This is the upper limit for continuous raw, beyond this you'll get x amount of frames before the buffer fills.