Jump to content
Andrew Reid

Canon 5D Mark III - 3.5K and 4K raw video with Magic Lantern

Recommended Posts

 

Available now! The new EOSHD 5D Mark III 3.5K RAW Shooter's Guide

With the new Magic Lantern experimental build for the 5D Mark III, we have an amazing 3.5K 10bit lossless raw mode with Super 35mm continuous recording.

Read the full article

Share this post


Link to post
Share on other sites
EOSHD Pro Color for Sony cameras EOSHD Pro LOG for Sony CamerasEOSHD C-LOG and Film Profiles for All Canon DSLRs

Nice work. With all of the, deserved, attention to the GH5, it's refreshing to see you testing and reviewing other cameras for us non-GH5 users.

This camera continues to amaze me. Without a doubt, purchasing it was the best decision I have made for my filmmaking. 

I'm sure you go over this in the guide, which I plan on purchasing, but a quick question I hope you'll divulge... is the image, at least, centered in live view? I may buy your guide today and test this build tomorrow if it's centered. 

Also in your review you mention you have a color live view in 3.5K but it reverts to grey scale when you hit record. Is there any benefit to having grey scale enabled the entire time?

Sorry, I assume you go over this in the guide, but just a few questions before I decide to test it this week.

Share this post


Link to post
Share on other sites

Hello:
        I congratulate your initiative, I will buy the guide, just like I did with the previous guide. I have a 5d3 and I use a atomos blade, when I record in raw in full hd (1920x1080) I do not have any problem to see the images in the monitor, When I record to 3,5k, do the images look good on the external monitor?

Katrikura

Share this post


Link to post
Share on other sites
5 hours ago, Katrikura said:

Hello:
        I congratulate your initiative, I will buy the guide, just like I did with the previous guide. I have a 5d3 and I use a atomos blade, when I record in raw in full hd (1920x1080) I do not have any problem to see the images in the monitor, When I record to 3,5k, do the images look good on the external monitor?

Katrikura

That's covered in the guide, you see the answer in there :)

8 hours ago, mercer said:

I'm sure you go over this in the guide, which I plan on purchasing, but a quick question I hope you'll divulge... is the image, at least, centered in live view? I may buy your guide today and test this build tomorrow if it's centered. 

Yes it's centred and the framing is 100% accurate in live-view for 3.5K.

8 hours ago, mercer said:

Also in your review you mention you have a color live view in 3.5K but it reverts to grey scale when you hit record. Is there any benefit to having grey scale enabled the entire time?

Nope. The grey scale is a bit lower res than the colour live-view.

8 hours ago, mercer said:

Sorry, I assume you go over this in the guide, but just a few questions before I decide to test it this week.

Test it. It's damn fun.

Share this post


Link to post
Share on other sites
9 hours ago, Andrew Reid said:

I will be uploading some of the original ProRes clips and Cinema DNGs. They amount you can push them in post without them falling apart is incredible.

If I may embellish.  I suspect many people are confused about 10-bit video.  When we look at a colors on our 8-bit screen we're looking at 256 shades of each primary color, R,G,B.  The dynamic range is generally around 6 EV between darkest and brightest.  In 10-bit compressed video, like that of the GH5, We have 1,024 shades of colors, but STILL IN THAT 6 EV dyamic range.  What this means is that if you shoot a wide DR shot of the sky and some people under and umbrella, you can't bring back more detail from the clouds in 10-bit then you could in 8-bit.  The only real benefit of 10-bit compressed video, that I could see, is banding  where there are fine gradations of color.  And that benefit, is almost impossible to see in a moving image.

In 10-bit DNGs, that Andrew is mentioning above, you're getting 1,024 shades OUTSIDE that 6 EV gamut, so you can recovered highlights or shadows.  If all ML did was get 10-bit RAW to work on the 5DIII, that alone, is worth the guide right there!!! :)  

As soon as they port this stuff to the 7D I'll buy that guide immediately!

Share this post


Link to post
Share on other sites
10 hours ago, Andrew Reid said:

I will be uploading some of the original ProRes clips and Cinema DNGs. They amount you can push them in post without them falling apart is incredible.

Dear Andrew,

I'm thinking to buy a 5D MkIII for shorts where slow motion is not required.
Do you think it performs better then GH5 (which I love, especially with V-Logl)? Is it reliable in long shots?

Thank you very much! :)
 

Share this post


Link to post
Share on other sites
3 hours ago, kidzrevil said:

Looks like Canon's been playing us this whole time

If you only read the title, it does.

If you actually use the camera, you'll know that it's not 4k, it's not 16:9, it has no usable live view and requires CF cards that were very expensive when the 5D3 was launched.

Share this post


Link to post
Share on other sites

@Jimmy thats good to know. Soooo....nothing to see here ? 😅

16 hours ago, Mattias Burling said:

I thought the 5Dmkiii was a real winner before the 3.5K... this just certifies its spot imo.
Nice video, some really awesome looking shots in there.

The colors at 0:43 are amazing

Share this post


Link to post
Share on other sites
6 hours ago, maxotics said:

 The only real benefit of 10-bit compressed video, that I could see, is banding  where there are fine gradations of color.  And that benefit, is almost impossible to see in a moving image.

In 10-bit DNGs, that Andrew is mentioning above, you're getting 1,024 shades OUTSIDE that 6 EV gamut, so you can recovered highlights or shadows.  If all ML did was get 10-bit RAW to work on the 5DIII, that alone, is worth the guide right there!!! :)  

As soon as they port this stuff to the 7D I'll buy that guide immediately!

Banding is easily seen in 8 bit moving images but not necessarily due to them being 8 bit but because of the high levels of compression on internal codecs. 8bit files recorded via HDMI to an external recorder in a good codec like ProRes can be very smooth and banding free even when pushed around in grading. 10 bit allows even more manipulation and opens up the use of LOG without fear of banding.Some cameras on the other hand can't record a blue sky without showing banding even straight out of the camera in a non log profile. 10 Bit RAW is probably just about OK for a Canon 5D3 which tops out at a theoretical 11ish stops of DR but in practice due to it not being able to ETTR perfectly on every shot means 10 stops or less as a practical limit. From this you can generate good 10 or 8 bit files with the benefit of a RAW conversion done outside the camera where quality rather than speed can be prioritised. We know for instance that the RAW converters in NLE's like Resolve are not as good as those for stills in say Adobe Camera RAW or Capture One due to the CPU overheads required  so imagine what a tiny camera CPU does with RAW sensor data in real time in order to spit out an HD 8bit 4.2.0 file at 25mbs........

Share this post


Link to post
Share on other sites
2 hours ago, kidzrevil said:

@Jimmy thats good to know. Soooo....nothing to see here ? 😅

There's plenty to see.... ML is stunning, I love the 60p mode and the 3.5k mode can be handy, at times.

The title is just a bit clickbaity. You could take literally any camera and, if you had the source code, you could run it at full capacity at breakneck speed/heat and on the very fastest CF cards and get more out of it that the given specs. It comes at a price though, dropped frames, occasional messed up frames, no usable live view and a time limit.

Share this post


Link to post
Share on other sites

You're not quite understanding the way the hardware works Jimmy.

The sensor delivers that 3840 pixel wide image under the normal Canon firmware, every time you hit the focus magnification button and it's capable of sensor windowing with any full pixel readout crop of the full frame sensor, like 3.5K, 3.3K, 2K, whatever you like. Magic Lantern are extracting the existing capability of the camera within the normal spec and clock speeds of the hardware, they are not running it at full capacity or increasing clock speeds.

That 3.5K image goes into the buffer as normal and then Magic Lantern take it from there and put it onto the card. The only heat increases come from a harder working card and battery due to the data rates. The processor and sensor do not get 'worked at breakneck speed'. In fact the processor is relaxing if anything because it doesn't need to think about compression or H.264.

The only reason the live-view isn't as nice as in the normal 1080p mode, is that Canon's code is designed to show the 5x punched in display for focus checking, so Magic Lantern had to hack it to give us accurate framing for the entire resolution, which is very difficult as they lack any kind of documentation from Canon or an SDK to tailor camera operation to their choosing.

The hardware is 100% capable of 4K video and likely with the same MJPEG codec as the 5D Mk IV and 1D C. The reason it is not in there is due to marketing. They wanted to reserve the feature for the 1D C and target the 5D3 predominantly at a stills audience and consumer crowd for whom 4K would be somewhat overkill or wasted and result in more support costs, distracted focus in the marketing messages, confusion on the shop floor, etc. The 5D3 also wasn't allowed to compete against any of the pro Cinema EOS cameras which were fresh on the market at the time it was released, because the margins would be much greater on these cameras and video shooters should not have been tempted by something cheaper offering such a recording high spec... That's another reason the 1D C cost $15,000 on release.

Plenty of compact flash cards in 2012 were fast enough for MJPEG 4K, otherwise they wouldn't have released the 1D C the same year doing MJPEG 4K to compact flash cards. You talk like the cards didn't even exist or would drop frames.

Also in 3.5K there are no dropped frames or occasional messed up frames, it is continuous recording with perfect image quality, like you see in my video. Sometimes a clip may stop due to the data rate being on the limit of the card, but there are no corrupt frames...and we always did have these occasional stoppages with the original uncompressed 14bit 1080p as well... that didn't stop a ton of AMAZING work being shot with it. For many things, shorter takes are enough. And if you want longer ones just decrease the resolution slightly and shoot 10bit lossless instead of 12bit.

11 hours ago, maxotics said:

If I may embellish.  I suspect many people are confused about 10-bit video.  When we look at a colors on our 8-bit screen we're looking at 256 shades of each primary color, R,G,B.  The dynamic range is generally around 6 EV between darkest and brightest.  In 10-bit compressed video, like that of the GH5, We have 1,024 shades of colors, but STILL IN THAT 6 EV dyamic range.  What this means is that if you shoot a wide DR shot of the sky and some people under and umbrella, you can't bring back more detail from the clouds in 10-bit then you could in 8-bit.  The only real benefit of 10-bit compressed video, that I could see, is banding  where there are fine gradations of color.  And that benefit, is almost impossible to see in a moving image.

In 10-bit DNGs, that Andrew is mentioning above, you're getting 1,024 shades OUTSIDE that 6 EV gamut, so you can recovered highlights or shadows.  If all ML did was get 10-bit RAW to work on the 5DIII, that alone, is worth the guide right there!!! :)  

As soon as they port this stuff to the 7D I'll buy that guide immediately!

10bit has more advantages than just less banding my friend :)

In rec.709 your dynamic range is limited in 8bit because you simply run out of room.

We know the look that comes with trying to compress more dynamic range into a rec.709 8bit, colour suffers, it looks like HDR puke.

That's because with 0-255 you have regions of large variation in brightness crammed into the last 5 or 10 shades, so things like a sky looks like crap, no accurate colours in it.

LOG is a way around this to a certain extent but is also a form of compression

The reason the rec.709 image looks so good on the GH5 is the entire image processing pipeline is 10bit this time round... things have room to breathe and when you do bring up the blacks or pull down the highlights there's 100x more information in there and they don't fall apart.

So in short, 10bit = more dynamic range, especially on a 10bit HDR display using Hybrid Log Gamma.

That's why the HDR standards all demand 10bit or more. Dynamic range!

Also 10bit raw = more dynamic range, plus more control over colour and white balance as it isn't baked in.

Also lossless compression = 50% smaller file sizes with no loss of visual quality, so it is almost like having uncompressed raw, rather than compressed H.264 10bit.

In addition the bump to 12bit and 14bit on the 5D3 with Magic Lantern gets you a bit more dynamic range, but close to the ceiling of the sensor.

I'd say the image quality at 10bit is very close to 14bit... 12bit even closer as to makes barely any difference. 14bit has a bit more accuracy in the highlights. Shadows still look AMAZING in 10bit, like the Blackmagic raw cinema cameras but less noisy.

Share this post


Link to post
Share on other sites
3 hours ago, Shirozina said:

Banding is easily seen in 8 bit moving images but not necessarily due to them being 8 bit but because of the high levels of compression on internal codecs. 8bit files recorded via HDMI to an external recorder in a good codec like ProRes can be very smooth and banding free even when pushed around in grading. 10 bit allows even more manipulation and opens up the use of LOG without fear of banding.Some cameras on the other hand can't record a blue sky without showing banding even straight out of the camera in a non log profile. 10 Bit RAW is probably just about OK for a Canon 5D3 which tops out at a theoretical 11ish stops of DR but in practice due to it not being able to ETTR perfectly on every shot means 10 stops or less as a practical limit. From this you can generate good 10 or 8 bit files with the benefit of a RAW conversion done outside the camera where quality rather than speed can be prioritised. We know for instance that the RAW converters in NLE's like Resolve are not as good as those for stills in say Adobe Camera RAW or Capture One due to the CPU overheads required  so imagine what a tiny camera CPU does with RAW sensor data in real time in order to spit out an HD 8bit 4.2.0 file at 25mbs........

I didn't mean to suggest 8-bit is what causes banding, only that the benefit of 10-bit in an internal compression reduces it.  I take issue with "10 bit allows even more manipulation", more manipulation how?  I feel there's  lot of misinformation out there, with flmmakers believing that "manipulation" is more than it is.  Whether there is banding or not, a filmmaker will grade to their taste and will NOT see a real-world difference between 8 and 10-bit compressed video.  That's my experience.  10-bit doesn't improve grading.  It only makes for small improvements at for some smooth color patches.  If you have some sample where you can see a real difference in "manipulation" I'd love to see it.  Otherwise, my fun-sucker opinion is 10-bit compressed video is more marketing pitch than real-world benefit.

Of course, you say as much with "so imagine what a tiny camera CPU does with RAW sensor data in real time in order to spit out an HD 8bit 4.2.0 file at 25mbs.."   I agree that 10 bit RAW is probably "just about OK", but again, compared to internal compressed video, it is MORE than OK.  It's giving you 10 stops from the PHYSICAL IMAGE, which is different than 10-stops of gray bars from a severely color-compromised 10-bit compresssed data stream.

Anyway, we're saying same thing I think.

Share this post


Link to post
Share on other sites

There's a lot of misinformation out there because there's a lot of fake 10bit.

The whole camera, colour processing, colour profiles, image pipeline and codec need to be built around 10bit for it to be an advantage, like on the GH5, or in raw on the 5D Mark III with Magic Lantern (where the colour profiles and image processing is done in POST, bypassing the camera's 8bit processing limitations).

The 10bit GH5 Rec.709 image is night and day better than the GH4 10bit HDMI output.

And don't get me started on 8bit "uncompressed" HDMI... It is compressed as hell! No point in recording it. Especially not as 10bit ProRes.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...