Magic Lantern enables 14-bit RAW CinemaDNG on 5D

Thanks. I was trying to lower the card speed requirements by maybe shooting 720p but it looks like that is not a great approach. I would like to keep the 16:9 aspect ratio while filming, so is 1080p (or 720p60) the only way to get this?
 
If you want to use the full sensor and with 16:9 aspect ratio, the lowest bitrate is achieved by shooting 1080p24 (83MB/s). From what I've read around the web, a Transcend 1000x card should do it, they are around $300 for a 128GB card that will hold 25 minutes of footage.
 
Thank you. Seeing all of the non-standard resolutions mentioned with ML is quite confusing. What happens if you do not use the full sensor for a lower resolution setting, I assume focal length is affected? Will you see black bars? Lenses not being as wide etc...?

The MKIII with h.264 will be improved with any RAW setting imho but I would like to retain a normal filming proceedure.
 
If you want to save bandwidth, you can shoot 1920x800, which is 2.40:1 and requires 62 MB/s instead of 83 MB/s. You can use that footage as is (2.40:1) or enlarge it 35% and paste it on a 1920x1080 timeline, filling all the vertical space and losing a bit on both sides of the original image. It is equivalent to shooting with an APS-H or S35 camera (1.35x crop factor).

But you already spent a lot of $$$ on that camera, I would go the extra mile and get a fast card (if 25 minutes is enough for you, as it would be for me: I hardly ever shoot more than that in a half-day).
 
Bassman2003 said:
Thank you. Seeing all of the non-standard resolutions mentioned with ML is quite confusing. What happens if you do not use the full sensor for a lower resolution setting, I assume focal length is affected? Will you see black bars? Lenses not being as wide etc...?

The MKIII with h.264 will be improved with any RAW setting imho but I would like to retain a normal filming proceedure.
Click to expand...

Using the 5D Mark III with ML raw: As long as your horizontal width is 1920, you are using the full sensor width. You can choose to record in various aspect ratios, if you want. You can also record in less than 1920 width, and then you start cropping the sensor. In all cases the recorded raw file will not have black bars, it will be the exact resolution you specify.

For example, you can shoot:

1920 with 3:2 aspect raw will be 1920x1280
1920 with 16:9 aspect. raw will be 1920x1080
1920 with 1.85:1 aspect. raw will be 1920x1038
1920 with 2:1 aspect. raw will be 1920x960
1920 with 2.39:1 CinemaScope aspect. raw will be 1920x804

1728 with 16:9 aspect. raw will be 1728x972, picture will be cropped some and not full frame, you would have to upsize in post to 1080 for 1080p

Again, in no case are any black bars ever recorded. The exact size you specify will be the size of the DNG files.
 
O.k. How much can you dial in the resolution numbers? Are they set or just any number you want to dial in? With your 1728x972 example, what would stop you from dialing in a 1280x720 resolution? Or is there a limit to how far you can get away from 1920?

If I don't need 1080p, I could see a benefit to using a smaller frame size and to making life easier on the CF card.

Thanks
 
The way it works is: there are a bunch of preset horizontal resolutions, such as 1920, 1856, 1728, 1600... on down. Yes 1280 is there also. You cannot just set an arbitrary resolution. Then you choose the aspect ratio you want to shoot in, such as 3:2, 16:9, 1.85, 2.39, etc and that decides the actual recorded resolution. So 1920 at 16:9 ratio will record at 1920x1080.

Yes, you can record 1280x720 23.976p. The CF card will need to be able to handle 36.8 MB/s for continuous recording. The 1280x720 will be a crop from the center of the full frame sensor. The focal length can be calculated. Full sensor width size=1920, x=crop factor when the cropped width is 1280.
 
Thank you very much. Makes sense now. Good to know you can still use some form of RAW with slower CF cards.

Any visible signs to the image (defects) as the resolution goes away from 1920 full frame? I would imagine it all beats h.264
 
Try 1728 x 972 thats a 16:9 ratio, not drastically different in terms of FOV for your lenses. This is the max width for the 7D and its still great than h264 in terms of image quality. using that on the 5D could let you use your slower card.

This will also be straightforward in post as some NLEs like EDIUS will automatically stretch it on the timeline youre using.

But of course I'd still suggest you maximize your camera's potential by getting a faster card as others have suggested.
 
I know that I am supposed to do due diligence and research boards everywhere to find the latest info- but I get so confused and lost trying to keep up with the current progress over at magic lantern boards. I would love some solid information on which build to use. I am using oct 30 now, 5d3. anyone have solid info on which build is the most solid? thanks in adavance.
 
dustwaterwindfire said:
I know that I am supposed to do due diligence and research boards everywhere to find the latest info- but I get so confused and lost trying to keep up with the current progress over at magic lantern boards. I would love some solid information on which build to use. I am using oct 30 now, 5d3. anyone have solid info on which build is the most solid? thanks in adavance.
Click to expand...

I'm not so updated to the mkIII developments as I only have the mkII and the 7D, but I do get my mkII nightly updates here: http://builds.magiclantern.fm/#/

It's the official nightly and some innovations made by other devs have already been ported so I'm assuming this is your safe bet as well. Just browse to your camera model and it will give you the latest "official" nightly build.

A mkIII user could chime in for a better insight.
 
If you have solid performance with your October build, just stay with it because there really isn't an answer to your question. To use Magic Lantern you have to "Cowboy" it a bit. That is, get on the horse and see if it throws you off or not. The latest build I have tried is Nov 18, 2013. No problems that I have found. I don't change builds every day or even every week, I check the nightly build changelog to see if anything I use has been updated and occasionally check the raw video sub-forum to see if any problems have come up.
 
I got to played with a ML raw enabled mk3 for a few hours and the raw image is apparently 2/3 stops (or more) underexposed in Resolve (with both rec709 and BMDfilm curves) and ACR (embedded profile; a bit less apparent underexposure with VisionLog). This is compared to any picture profile applied in-camera. Apparently Canon pushes the values up a bit during processing. Dxomark lists actual isos as 1/3 stop below reference, so this may have something to do with the underexposure.
So I am curious how do you guys do your metering with ML raw? Simply overexpose by a fixed amount (1/3, 2/3...)? ETTR is obviously going to give best noise characteristics on per shot basis. But this means inconsistent noise between shots in a sequence, not to mention the need to level every shot in post.
 
Did you set iso according to real stops.... The only functional isos are fx 100, 200, 400, 800, 1600 and so on.... So setting 1000 will appear right on the lcd but your raw is exposing to 800....
 
Yes, everything was shot at true iso's. I did some googling and it seems this is normal. Just something to have in mind while exposing.
 
So what is the latest developments then of ML with the 5DM3? Any audio yet with RAW? Is RAW completely safe or could it mash up the sensor/processor??
 
Aduio is now available though not in the nightly builds via the magic lantern MLV file format.

With this you can edit DNGs with audio only in Premiere CC at this time.

(no worries about damaging camera sensor. I have shot a short film and days and days of footage with no issue)
 
You must log in or register to reply here.
Back
Top