Canon 5D III (Magic Lantern Raw) VS Sony a7R II

QuickHitRecord

QuickHitRecord

Veteran
My friend and I did a quick (and unscientific) side-by-side test between these two cameras with a single LED source shooting in through the window. I focused on matching the framing, focal length, and background rendering as much as possible. In a perfect world, the 5D could have used a little more light because it doesn't "see" as much light as the Sony, and it also needs to be exposed to right even more than the Sony for optimal results.



There's more of a write up over on Vimeo if you want to know more about my workflow.

Personal notes: The a7R ii is easier to use in almost every way, but I greatly prefer the image from the 5D3 shooting raw. SLOG (in this case, SLOG2) is well implemented in the higher-end FS7, but I struggle to get passable skintones from it in their mirrorless hybrid cameras, especially with fair-skinned subjects. Bumping up the saturation of gray just gives you more gray. I know it can be done, since I've seen in other examples on these boards. But I'm assuming that it takes a lot of work to get it there, and probably multiple windows. The irony is that the raw footage from the Canon is very flexible in terms of grading, but it needs very little (if any tweaking). I think that's worth a lot.

I'll be hanging onto my 5D3. It's a 2012 camera, which makes it a dinosaur, but with Magic Lantern's incredible hack, there's a lot of life in it yet.
 
QuickHitRecord said:
My friend and I did a quick (and unscientific) side-by-side test between these two cameras with a single LED source shooting in through the window. I focused on matching the framing, focal length, and background rendering as much as possible. In a perfect world, the 5D could have used a little more light because it doesn't "see" as much light as the Sony, and it also needs to be exposed to right even more than the Sony for optimal results.



There's more of a write up over on Vimeo if you want to know more about my workflow.

Personal notes: The a7R ii is easier to use in almost every way, but I greatly prefer the image from the 5D3 shooting raw. SLOG (in this case, SLOG2) is well implemented in the higher-end FS7, but I struggle to get passable skintones from it in their mirrorless hybrid cameras, especially with fair-skinned subjects. Bumping up the saturation of gray just gives you more gray. I know it can be done, since I've seen in other examples on these boards. But I'm assuming that it takes a lot of work to get it there, and probably multiple windows. The irony is that the raw footage from the Canon is very flexible in terms of grading, but it needs very little (if any tweaking). I think that's worth a lot.

I'll be hanging onto my 5D3. It's a 2012 camera, which makes it a dinosaur, but with Magic Lantern's incredible hack, there's a lot of life in it yet.
Click to expand...

The 5D3 is 14 bit in raw ML. That's pretty incredible. Even the FS7 is only 12 bit. So while it has less DR, it's files can really be pushed around a lot more. The color is truer out of the box as well, which is down to Canon's color science. The only thing to be careful of is not overexposing, which can be easy to do with 11.3 stops fo DR and a sunny day for instance. All in all, it might be a 2012 camera, but it is a great workhorse with ML.


One thing to note, if you look at the window in both shots, the A7RII has at least a half stop more exposure than the 5D3. So that might need to be taken into account, especially when looking into your blacks as a result.
 
Last edited:
You can't beat the colors of the 5D3+RAW, but sgamut on the a7 bodies is particularly awful unless you use the perfect workflow that nobody seems to use, try these settings instead, they are designed for good skin tones (you can dial down the saturation in-camera if it's bothering you, but I like what I get when I set it that high).
 
Yeah, the 5D could have used a half stop or even a stop more exposure to match the a7R ii. If I'd had my NDs handy, I could have made that happen.

Unfortunately I don't have access to that a7R ii. It was a friend's (the guy in the video) and we only had about a half hour to set this up and shoot it. Otherwise I would try the Flaat settings. But maybe I'll send that link to him. The settings we were using (SLOG2) sucked the color out of his skin. There just wasn't anything to "grab onto" in the grade, even when I boosted the saturation. It was like trying to trying to take a white plaster wall and coax it to be peachy-pink. As mentioned above, I haven't had these problems with the higher-end Sony cameras. I know that there are ways to address this, but man, with Canon cameras it's never an issue. That's one of the big reasons why they still have my business.
 
Samuel H said:
You can't beat the colors of the 5D3+RAW, but sgamut on the a7 bodies is particularly awful unless you use the perfect workflow that nobody seems to use, try these settings instead, they are designed for good skin tones (you can dial down the saturation in-camera if it's bothering you, but I like what I get when I set it that high).
Click to expand...
Samuel,

I'm debating between a 5DM3 + ML vs an A7SII. Do you think your flaat profile for the Sony makes the A7SII winner? I love the skin tones of the 5DM3, but I also love the low light performance of the A7SII. Although I do wonder if the 5M3 + Neat Video could come close to the Sony in terms of low light.

Thanks so much for your help!
 
BobbyMurcerFan said:
Samuel,

I'm debating between a 5DM3 + ML vs an A7SII. Do you think your flaat profile for the Sony makes the A7SII winner? I love the skin tones of the 5DM3, but I also love the low light performance of the A7SII. Although I do wonder if the 5M3 + Neat Video could come close to the Sony in terms of low light.

Thanks so much for your help!
Click to expand...

The A7SII will give you better skin tones if you use an external recorder.
Internal it is only 420 in FullHD. External it is 422. The compression scheme does make a big difference.
 
yoclay said:
The A7SII will give you better skin tones if you use an external recorder.
Internal it is only 420 in FullHD. External it is 422. The compression scheme does make a big difference.
Click to expand...

So, in other words - you could almost buy 2 Canons by the time the Sony is ready to be competitive.

Now you're 3/4 of the way to an URSA Mini 4.6k.
 
BobbyMurcerFan said:
Although I do wonder if the 5M3 + Neat Video could come close to the Sony in terms of low light.
Click to expand...

Yes and in many cases even better. I did an indie film that was finished four months ago and I used several cameras. The 5d iii raw was my A camera but I did use a lot of RX10 ii shots (slowmotions), A7s shots and also a couple of A7s II and A7r II shots (but only a couple)

The 5d raw helds up really well in low light, but you gotta keep it at ISO 6400 max. Funnily I consider that ISO to be about even with the A7s at 25,600. The 5d has WAY more texture and grain but also sharpness and no motion artifacts at those ISO's. The a7s starts to smooth it's own image and the colors turn to s*%t.

Here's the short itself (shot mostly during the evening / night). The last sword fight is the 5d at ISO 6400 (with a 50mm f1.2 and a 20mm f1.8) and I had a backup A7s at 25,600 ISO shooting the same scenes. I thought the A7s didn't look as good so I mostly used the 5d raw material. I didn't even touch Neat Video. The ISO of the 5d actually maxes at 3200 so after that you have to boost it up in post. The a7s will go further once you get past ISO 25,600 but after that it's kinda terrible too.

https://vimeo.com/149755248

I sold the A7s not long after.
 
For me, it depends on what you care about the most.

The 5D3+RAW has great color, but the workflow can be painful (if you ask me, Resolve is the only viable option right now) and even though it's RAW it has less DR than a7S+slog2. Also, you have to think about media and storage.
a7S and a7R II have more DR and better low light abilities, higher fps, and 4K. But color can be awful if you don't use the proper settings (or the proper workflow), and even with good settings and workflow it's not as good as 5D3+RAW. It's also smaller and lighter.

Personally, I care most about DR, then color, then size. I'm very happy on the Sony side. YMMV.
 
maarek said:
Here's the short itself (shot mostly during the evening / night). The last sword fight is the 5d at ISO 6400 (with a 50mm f1.2 and a 20mm f1.8) and I had a backup A7s at 25,600 ISO shooting the same scenes. I thought the A7s didn't look as good so I mostly used the 5d raw material. I didn't even touch Neat Video. The ISO of the 5d actually maxes at 3200 so after that you have to boost it up in post. The a7s will go further once you get past ISO 25,600 but after that it's kinda terrible too.

https://vimeo.com/149755248

I sold the A7s not long after.
Click to expand...

Hardcore! Great work. There's a little bit of fixed pattern noise but nobody except us would notice it. It's nothing like the fixed pattern noise that made it into that Dredd movie.
 
maarek said:
Yes and in many cases even better. I did an indie film that was finished four months ago and I used several cameras. The 5d iii raw was my A camera but I did use a lot of RX10 ii shots (slowmotions), A7s shots and also a couple of A7s II and A7r II shots (but only a couple)

The 5d raw helds up really well in low light, but you gotta keep it at ISO 6400 max. Funnily I consider that ISO to be about even with the A7s at 25,600. The 5d has WAY more texture and grain but also sharpness and no motion artifacts at those ISO's. The a7s starts to smooth it's own image and the colors turn to s*%t.

Here's the short itself (shot mostly during the evening / night). The last sword fight is the 5d at ISO 6400 (with a 50mm f1.2 and a 20mm f1.8) and I had a backup A7s at 25,600 ISO shooting the same scenes. I thought the A7s didn't look as good so I mostly used the 5d raw material. I didn't even touch Neat Video. The ISO of the 5d actually maxes at 3200 so after that you have to boost it up in post. The a7s will go further once you get past ISO 25,600 but after that it's kinda terrible too.

https://vimeo.com/149755248

I sold the A7s not long after.
Click to expand...

Thanks so much!! Can I ask, you say that the 5DM3 maxes out at 3200, but you also say that you have to keep at no higher than 6400, so which ISO did you set the camera to? Thanks so much again!
 
BobbyMurcerFan said:
Thanks so much!! Can I ask, you say that the 5DM3 maxes out at 3200, but you also say that you have to keep at no higher than 6400, so which ISO did you set the camera to? Thanks so much again!
Click to expand...

The RAW files max out at 3200. But during shooting you can still see it go up to 12800. The files itself will be 3200 which you then have to boost up.

I do remember going over +1 stop when I did the postwork so I guess some shots were between 6400 and 12800 on the 5d raw.

I didn't do any scientific testing but the feeling was that the 5d RAW was not that far off the A7s (except the pattern noise was stronger on the 5d). A7s doesn't get grainy in the same way, it starts to exhibit noise reduction artifacts and at 25,600 iso motion artifacts. There's a lot of processing to the a7s image that the 5d lacks. A7s looks like it already went through Neat Video.
 
yoclay said:
That's totally incredible. I wasn't aware that ML built in these fix choices. Things have really moved on with them.
How's the stability of ML these days ?
Click to expand...
This short thread might be useful: http://www.magiclantern.fm/forum/in...b17&topic=16971.msg165237;topicseen#msg165237
I use Lexar 1066x cards. That said, I don't use the 5D3 with ML professionally as I get paid to do other stuff. I only use it on personal projects, but that is me.

Thanks for posting these tests QuickHitRecord. Really useful information. Was seriously considering picking up a A7Rii, but can't let go of the 5D3 just yet. Just such a great, specific look.
 
yoclay said:
That's totally incredible. I wasn't aware that ML built in these fix choices. Things have really moved on with them.
How's the stability of ML these days ?
Click to expand...

Here's more on MLVFS, which I find an essential part of the workflow.: http://www.magiclantern.fm/forum/index.php?topic=13152.0

It's a little outside of the box (you select your options in a web browser, and then hit reload to implement them), but you can open the MLVs directly in Resolve, tweak color balance and exposure, and export to ProResHQ or 4444. No DNG conversion necessary.

As for reliability, I haven't had a single pink frame since using one of their 2014 builds -- and I usually leave Global Draw and Digic Peaking on. I'm using 1.13 vs 1.23 because it better suits my needs.

Also, I am using a Komputerbay 256GB card, which can now be purchased for about $180 (it was $400 1.5 years ago) and gives me about 48 minutes of recording in 1080P. Mine is 1066x, but the reviews on Amazon suggest that there may be some quality control issues (there have been some duds reported) so you might look at the 1200x cards instead. I think they cost about the same.
 
QuickHitRecord said:
Click to expand...

For those who don't feel like watching four minutes of ISO noise, here were some quick takeaways from this exercise:


  • If I don't hit refresh on the web interface (even with none of the boxes checked), I seem to get crazy dancing bad pixels and just more noise in general.
  • The 5D Mark III seems to be a native ISO 400 camera. There is a lot of additional noise introduced at ISO 800.
  • Bad pixel fix works great. Aggressive bad pixel fix didn't seem to yield any different results. Maybe this is useful for the Rebel and EOS-M lines that seem to have more bad pixel issues.
  • Fix pattern noise helps a lot, and seems to eliminate vertical stripes as well.
  • Vertical stripe fix doesn't seem to do anything.
  • Chroma smoothing adds a lot of noise. I don't think that I will be using this.
  • RawMagic displays vertical stripes, even with vertical stripe correction on.
  • Footage from RawMagic is about a stop brighter by default, and therefore is not a good fit for Cinelog or possibly other LUTs.
 
You must log in or register to reply here.
Back
Top