Magic Lantern enables 14-bit RAW CinemaDNG on 5D

But that remains to be seen what impact Blackmagic will actually have. Of course, Brawley and Hicks have been going all gangbusters over it because they know what the concept of shooting Raw entails and they have the technical background to appreciate both its power and its complexity.

On the other hand, there is a lot of clamor for 12-bit Raw from people who want to shoehorn it into their AVCHD-like workflow, and then find out the camera is not for them. Thus a lot of the marketplace and eBay listings.

No matter how you slice it, Raw video is a resource hog, and a lot of people don't seem to be able to conceptialize working with it in their mind (probably due to lack of any real-world experience) until it's staring them in the face, and then they realize it's just too much.

And, BTW, I am one of them. For the vast majority of what I shoot: a) I don't want to deal with Raw all the time and b) the people I hand off the footage to aren't going to give it to editors who are ready to deal with it.

EDIT: Yes, I know BMCC shoots ProRes 4444, but even that is pretty big compared to the typical codecs.
 
That's true. I had at one time had a red at my disposal, to do with it as I please, but due also to the kind of work I do, I ended up not using it and used my DSLR instead. BUT as time goes by HOPEFULLY developments in hardware and software advances the workflow and makes working with raw a "generally" pleasant experience.

Adding raw to the magic lanterns bag of tricks for the dslr is merely an added tool or trick that one hopes to use when the need arises. Like variable FPS in ML for instance, it has been useful for me in tricky situations. Its not something one uses all the time.
 
The 5D MKii and III lack bandwidth internally to process 24fps at this data rate. This would, however, be very useful for time-lapse.
 
All of the Canon's can already shoot raw for time lapse, you just need an intervalometer. I have one for my 7D and shoot TL with it all of the time.

As far as raw vs compressed, pick the tool you need for the job. The DSLR for factor is smaller and easier to use than the BMCC, or any of the larger Raw cameras, and if the smaller file format, lower quality of footage, and more limited DR will be ok for the project, then it's the right tool. If you want something of higher quality the BMCC can provide that now, and possibly the Canon's will be able to with further hacks. Seems like win,win, win to me. So many tools, so many options.
To me the BMCC is so close to film in it's capabilities, you can smell the chemicals. I love what possibilities it opens up. For low light, or run and gun I have my 7D as an option with the Technicolor cinestyle.
If the magicians at Magic Lantern can successfully bring true RAW output through live view of any of the Canon cameras, it is a huge deal. It's something not only Canon will have to deal with, but the other big companies will as well. We win again. :happy:
 
I have a question re: the Neumann Video. How do you compensate for all that banding?
It looks like it was fixed before the final image, but I don't understand how.
 
visugeek said:
All of the Canon's can already shoot raw for time lapse, you just need an intervalometer. I have one for my 7D and shoot TL with it all of the time.
Click to expand...

I think the ability to do timelapse without using the mirror is what's interesting people.
 
If you did manage to get DNG raw streams out of a camera HDMI port in live view, your monitoring would be monochrome, since the DNG raw image data is luminance values only, not color.
Color is created during debayer when the color filter matrix values are applied to the raw data. The filter matrix definition is meta data attached to the recorded file, but not part of the raw image data.
But how would you bypass the camera's internal debayer/encode process? This path is likely fixed in hardware.
 
Last edited:
Razz16mm said:
If you did manage to get DNG raw streams out of a camera HDMI port in live view, your monitoring would be monochrome, since the DNG raw image data is luminance values only, not color...
Click to expand...

It's probably more complicated than that. I imagine that any solution using the HDMI to output a raw stream is going to require a full color image, but it is color that would look wildly inaccurate on any monitor. Consider that the HDMI out stream is 8-bit 4:2:2. Using only the luminance channel would leave only 8-bit per sample, which wouldn't make raw recording very advantageous. They would have to store some of the luminance info in the chroma channels to get all of the data to fit. Needless to say, that would look rather messy. Of course, if your monitor supports viewing in monochrome only, then it would probably make that mute.
 
Wesley Byram said:
then it would probably make that mute.
Click to expand...

Like this whole idea in the first place! Whoever started this is just falsely inflating hopes imho... I am sure the heat issue alone would prohibit any kind of sustained transfer. So you can output series of RAW images but can you shoot a ten minute interview? That is really pushing the hardware to scan that huge MP sensor so many times in a row.
 
Bassman2003 said:
Like this whole idea in the first place! Whoever started this is just falsely inflating hopes imho... I am sure the heat issue alone would prohibit any kind of sustained transfer. So you can output series of RAW images but can you shoot a ten minute interview? That is really pushing the hardware to scan that huge MP sensor so many times in a row.
Click to expand...

As I understand it, what the ML devs have done is found a way to access the video stream that the camera is already creating for its live-view function, and that stream is "pristine" in that it hasn't been manipulated by the encoder yet. It isn't scanning the sensor any differently than the camera normally does. Someone feel free to correct me if I'm mistaken about that.
 
That would be grand if the full sensor was already being used but the video I watched was talking about DNG RAW files at 14 fps but for only a few seconds. A series of still frames rather than a video stream. Which is how we could get a true RAW editing situation. This is a blend of the photo and video world and the terms can get mixed up a bit like RAW and uncompressed. I guess more information would help the situation but it does seem like HDMI will not be the method to get this new stuff out of the camera as you would lose the RAW editing capability.
 
Bassman2003 said:
That would be grand if the full sensor was already being used but the video I watched was talking about DNG RAW files at 14 fps but for only a few seconds. A series of still frames rather than a video stream. Which is how we could get a true RAW editing situation. This is a blend of the photo and video world and the terms can get mixed up a bit like RAW and uncompressed. I guess more information would help the situation but it does seem like HDMI will not be the method to get this new stuff out of the camera as you would lose the RAW editing capability.
Click to expand...
HDMI is just a pipe. You can transfer any form of digital data via the pipe. The Digital Bolex is using 1080p HDMI out to monitor the monochrome raw image dump from the sensor. But as a monitor signal it is encoded, not raw, without debayer for color processing.
HDMI 1.4 separately passes digital video within defined standards, digital audio, IR and serial control codes, and 100 Mbps bandwidth Ethernet. Doesn't mean any particular device will support all of these functions though.
 
There's a saying that this development makes me think of:
"The mind is willing but the body is not able."

Short of essentially engineering a new camera, this can't be done.
 
Can the HDMI installed on the 5DMKIII transmit RAW files for later editing is the real question. Because the whole 14bit thing is based upon RAW image processing, not a true 14bit video stream which would be huge anyway. Can CF cards keep up with a RAW image stream at 24fps or 30fps? I have not done the math but I would guess not at this time.
 
Wesley Byram said:
It's probably more complicated than that. I imagine that any solution using the HDMI to output a raw stream is going to require a full color image, but it is color that would look wildly inaccurate on any monitor. Consider that the HDMI out stream is 8-bit 4:2:2. Using only the luminance channel would leave only 8-bit per sample, which wouldn't make raw recording very advantageous. They would have to store some of the luminance info in the chroma channels to get all of the data to fit. Needless to say, that would look rather messy. Of course, if your monitor supports viewing in monochrome only, then it would probably make that mute.
Click to expand...
There is no color data within the raw image itself, to get color info of any kind you have to use processing power to apply filter matrix meta data and debayer the raw stream. That can be a quick and dirty process, but then you would not have a raw file anymore if it was applied to the HDMI output and not useable for recording high quality images.
So far I don't know of any external raw CDNG recorders either.

To see any kind of image on a monitor it has to be encoded to a video format. Monitors can't interpret or display raw unencoded data.
 
From what I gather, the sensor is still being sub-sampled as it is read out in live-view mode, which is the data source for this hack. So, it isn't as if they are trying to read the whole 22 megapixel sensor continuously at 24 frames per second. That would almost certainly be impossible. However, the video data in the live-view buffer that they have accessed has 14-bit precision and is apparently higher res than what is normally sent to the encoder. So, the ML devs are examining ways of making use of this higher quality image. One of the ways to do this could be to slightly shuffle the bits around and just send it out the HDMI port, and then use another program on the user's computer to reverse the "shuffling" and get a sequence of DNGs as a result. There's already a camera that does this, albeit via HDSDI instead of HDMI, but the principle is the same: http://www.camadeus.com/?page_id=2388

In principle, that shouldn't be very computationally expensive. In practice...well, I guess we'll see.
 
In the end, we are all very appreciative that ML is diving so deep in this camera. I for one would be happy if they could just get more detail out of the image in an 8bit 4:2:2 video stream, or even the internal CF recording codec. 14bit RAW seems like overload for the hardware but detail is just being thrown away in the whole existing sensor to video stream process.

The ML team probably has to be careful to not let too much out of the bag until the major firmware update is completed as Canon might make it more difficult to make changes if they see the 5MKIII making a better image than their product line plans. The MKIII is a camera that has a very pleasing image but is lacking in some areas. A little bump and it would be the class of the field.
 
Last edited:
You must log in or register to reply here.
Back
Top