Magic Lantern enables 14-bit RAW CinemaDNG on 5D

Perhaps slight exposure differences make the 5D look sharper. Though they are very similar to be sure. To my eyes the 5D looks better on an iPad and computer monitor.

It is very clear though that for narrative work the 5D is the preferred choice over the C100/300.


The result depends highly on the workflow. I did this test together with my 2ND D.P. Igor, processing 5dmk3 Raw via Resolve, cinema DNG, in two tests C100 looked better, for highlights handling and detail. There is more color detail with C100 as well. By the way C100 shot was in WDR not Clog.
I will do more testing for public, having no preference, which camera is better. "The truth will save us"! (although there is only subjective truth in this matter)

/I bought 5dmk3 just for RAW, thinking that it will be better than C100. So far in lots of material shot with both cameras it is not. I see what I see. Consider also better ISO performance and less rolling shutter with C100.
 
5d3 vs C100 https://vimeo.com/69304810
I know which one looks better to these eyes.

Hand on a second here.
Are we talking about comparing Canon 5D MIII RAW with the C100's internal codec?
If so then this test is a bit disingenuous, because that is not the best that the C100 can offer.
The internal codec is 4.2.0 with a data rate of only 24MB/s, so of course the 5DMIII RAW would looks sharper with
it's data rate of 90-110MB/s.
Comparing best therefore with worst is not my idea of a fair test.
The C100/C300 really shine when you attach a recorder to them. Far sharper and better color information.
We are talking 220MB/s of info. In some circumstances it's almost too sharp.
On the other hand, if this is indeed the C100's internal codec, then a more apt comparison is with the 5DIII's original unhacked video output.
In which case we know who really wins in that circumstance, and it doesn't start with a 5.
 
Last edited:
The result depends highly on the workflow. I did this test together with my 2ND D.P. Igor, processing 5dmk3 Raw via Resolve, cinema DNG, in two tests C100 looked better, for highlights handling and detail. There is more color detail with C100 as well. By the way C100 shot was in WDR not Clog.
I will do more testing for public, having no preference, which camera is better. "The truth will save us"! (although there is only subjective truth in this matter)
.
Would you mind posting those tests Oleg?
 
Hand on a second here.
Are we talking about comparing Canon 5D MIII RAW with the C100's internal codec?
If so then this test is a bit disingenuous, because that is not the best that the C100 can offer.
The internal codec is 4.2.0 with a data rate of only 24MB/s, so of course the 5DMIII RAW would looks sharper with
it's data rate of 90-110MB/s.
Comparing best therefore with worst is not my idea of a fair test.
The C100/C300 really shine when you attach a recorder to them. Far sharper and better color information.
We are talking 220MB/s of info. In some circumstances it's almost too sharp.
On the other hand, if this is indeed the C100's internal codec, then a more apt comparison is with the 5DIII's original unhacked video output.
In which case we know who really wins in that circumstance, and it doesn't start with a 5.

That's not quite fair either, as the previous 5d3 internal video just has no resolution. But I get your point and I'm sure we'll see more tests with a ninja or the like soon. In any case, I love the look of the C100/300 and in fact used the C300 for my last piece, but I'm just glad to see the 5d can now get to that level of detail. Makes the 5d a great broll cam for a C300 shoot. And I agree, the C300 is so sharp you can push in 130% on a shot and it still looks perfect in a 1080p timeline. All good choices.
 
And in any case, it was argued before (in this thread or the other one abot 5D3+RAW) that the C100 H.264 is much better than DSLR H.264, and that it could indeed be compared with the 5D3+RAW and not end up losing. There were some comparison shots, too, and to me they looked close enough that I didn't know which was which.

So yes, these are so close that it is workflow that will probably determine the winner.
 
So what card do you need for stable 1920x1080? Is 95mb/s Sandisk enough? (the 100mb/s version is pretty insanely more expensive I just noticed...)
 
If the card itself is labelled at 95mb/s you probably wont get more than 65mb/s..... I had a 32 gb Sandisk Pro Extreme, and that was nowhere near enough.

So far the confirmed cards vary, but two keep popping up and that is 64gb Komputerbay 1000x cards (buy from Amazon, test it right away and send it back if it's not fast enough... Kompbay cards aren't quality screened, so it can be a hit and miss experience.)
and the Toshiba 1066x cards which unfortunately are ridiculously expensive :S

People are reporting new cards that work here and there, so best thing is probably to stay tuned to the Magic Lantern forum and follow the thread specifically about CF cards.
 
Sorry if I've missed it, but has a stable *official* release come out yet?

There isn't even yet a stable "official" release for the 5D Mrk III, so unlikely for raw.
You need to check through the forums to find out what build most people are using.
 
What is the easiest and cheapest way to convert the Canon RAW video?
I have a Canon 6D I would like to try this with, but I do no have the Adobe Premiere or some the the other software I have seen mentioned. I also have a Windows PC (i7).
Any suggestions?
 
What is the easiest and cheapest way to convert the Canon RAW video?
I have a Canon 6D I would like to try this with, but I do no have the Adobe Premiere or some the the other software I have seen mentioned. I also have a Windows PC (i7).
Any suggestions?

Use RAW2CDNG to make Cinema DNGs from the ML raw files, then grade and export via resolve lite.

Both are free.

Unfortunately, I have not heard yet of any progress for long and continuous recording for the 6D due to its SD card storage. There are much better results posted from the much older 50D.
 
Last edited:
I have a few cards I used with hacked GH2s and a very high bit rate. I am hoping to try them on my 6D with one of the lower resolutions.
 
I have a few cards I used with hacked GH2s and a very high bit rate. I am hoping to try them on my 6D with one of the lower resolutions.

It's not even the card that is the bottleneck, but the card controller in the camera. Canon wasn't planning on pumping this much data through so they used a slower card controller for SD cards.
 
Took the family out fireworks watchin so I shot some 4th of July fireworks in Teas using ML raw, recorded some audio using h264 video.


 
Hey Ted,

As you seem to the be the MKII guru, I was wondering if you can record a take longer than 1 min in raw?
Another question is using RAW2DNG the prores 444 output seems really good however it seems like the WB is shifting throughout the clip, is this a bug?
Thanks for all your hard work keeping dvxuser updated with ML news.

Cheers,
 
Hey Ted,

As you seem to the be the MKII guru, I was wondering if you can record a take longer than 1 min in raw?
Another question is using RAW2DNG the prores 444 output seems really good however it seems like the WB is shifting throughout the clip, is this a bug?
Thanks for all your hard work keeping dvxuser updated with ML news.

Cheers,

Haha! I'm far from what you'd call a guru, perhaps just the guy with time to spare. :)

To answer your questions,

About the time, it records continuously depending on the resolution you choose and the card you have.
The card must be fast enough, Ive tested 1000x at 24P/fps

For now, 1872 x 936 2:1 is continuous, meaning it will record until your card is full. A 32 gig CF card will give you 7+ minutes.

1729 x 972 16:9 is continuous.

1872 x 1054 (the highest "official" for the 5D2) 16:9 is at 1134 frames ( around 47 seconds )

1880 x 1058 ( the highest "unofficial" ) 16:9 is at 983

"official" is the official ML repository for nightly builds.
"unofficial" is compiled by other developers with their own "flavor".


Resolutions lower than 1729 x 972 are mostly continuous.


As to your question about prores, I can't give you an answer since I have not played yet with that codec.
My workflow is different since I am currently testing DNG or CDNG via AE or resolve, then to Edius.

I use RAWanizer and RAW2CDNG to make the conversions.
 
Last edited:
Hey Ted,

Thanks for getting back to me. I don't seem to be able to record past 1 minute on my lexar 32gb 1000x, do I have to flag something in the ML settings. I'm pretty sure I'm on the latest build.

Cheers,
 
Back
Top