5D MKIII specs

[HansSteinert]

Yes, but contrast can vary a lot, even with sharp lenses. Low microcontrast can be perceived as soft. Though I agree that the softness we see on MK III out of camera footage is mostly based on in camera processing. What is right on the other hand is that the shallow DOF doesn't always help to match the subjective impression of sharpness comparing the footage to the GH2 for example. Bu I have the suspicion that the GH2, same with the MK II, produces a certain amount of false detail and therefore the impression of sharpness.

I would argue that shallow DOF actually increases perceived sharpness instead of decreasing it since things in focus will pop relative to things out of focus. Plus, shallow depth of field is more efficient in utilizing low bitrates.

I don't have that suspicion with the GH2. The 900 lines of resolution are genuine based on tests.

Microcontrast is a valid point, but L lenses aren't exactly lacking in that department. It's no Zeiss, but it's certainly not guilty enough of causing the sharpness issues we are seeing here.

 

[joe 1008]

Generally I agree that shallow DOF increases perceived sharpness. But it can be confusing if you compare directly deep DOF footage with shallow DOF footage.

 

[dustylense]

Considering we're dealing with 1k and 2k resolutions, I highly doubt the lenses are the primary culprit in terms of sharpness. Those lenses are used professionally for 22mp photographs (I myself am one of these users), so the resolving power of lenses is not the issue of softness when shooting video on a 5D III. The resolving power for 1000 lines, let alone 700 lines, is there with Canon's L lenses. At under 2k resolution, the primary difference between lenses is color rendition, bokeh quality, focus breathing, legnth/smoothness of pulling focus, build quality, and performance wide open. Resolving power isn't a huge variable at under 2k resolution when comparing Canon L lenses (even the 24-105) and higher end film lenses.

I invite you to shoot full resolution RAW's with the 24-105 against the 85 1.8 or the zeiss 50 1.4 and get back to me. Of course they are all able to resolve in stills mode. But my 16-35 Eos vrs my 17mmT/S eos is like night and day! My 85 1.8 vrs my 24-105 EOS is like Bakersfield vrs the milky way. Shooting F4 on the 24-105 vrs f8 is vastly different. I suspect that lots of youtube Mk3 footage is shot wide open with the 24-105. Let's see some primo Zeiss glass alongside some primo lighting with some primo focus (in out in out in out in out handheld in out in out at f/2? Of course it's going to look soft even if you had the Hubble mounted to the camera).

As a stills guy, my point is that if you don't hit focus shooting wide open with your subject 5 ft away from you with a 100mm throw under continuous lighting at 1/50th of a second and expect ultimate sharpness, not gonna happen. That is why we developed strobes. Do this. Take your 200mm lens of any quality and price, go outside and shoot a subject 8 feet away with constant light and the lens wide open at 1/50th handheld. Report back with just how sharp that photo is.

 

[pulpfiction007]

That's more of a testament to how great Shane and his team are at cinematography and storytelling. Not the capabilities of the camera necessarily. He could've shot that on an old SD cam and it would've still been awesome (albeit aesthetic differences of course).

You wouldn't say that a person's cooking is great because they have awesome pots and pans, now would you?

Nope...exactly why I think all the pixel peeping/comparisons with other cameras, saying (insert camera model here) isn't this or it is that muddy's the water. Either a model is CAPABLE of getting a look one wants with the features one wants for the price one can afford....or it can't. I for one feel that most recent cameras ARE capable of extraordinary things and it's just a matter of choosing the one that suits a given style/budget the most. My eyes gloss over reading about codecs/pixels/downscaling figures....I don't really care, although it's important to some people. The audience won't give a second thought to those things either, which is why I chose the short film by Mr Hurlbut and Co- it's both a testament to them AND the fact most modern cameras are CAPABLE story telling tools in the right hands. It's just a matter of finding the one that ticks a shooters personal boxes...if a camera looks bad to someone- fine, move along and shoot with something else.

 

[hgf4]

Nope...exactly why I think all the pixel peeping/comparisons with other cameras, saying (insert camera model here) isn't this or it is that muddy's the water. Either a model is CAPABLE of getting a look one wants with the features one wants for the price one can afford....or it can't. I for one feel that most recent cameras ARE capable of extraordinary things and it's just a matter of choosing the one that suits a given style/budget the most. My eyes gloss over reading about codecs/pixels/downscaling figures....I don't really care, although it's important to some people. The audience won't give a second thought to those things either, which is why I chose the short film by Mr Hurlbut and Co- it's both a testament to them AND the fact most modern cameras are CAPABLE story telling tools in the right hands. It's just a matter of finding the one that ticks a shooters personal boxes...if a camera looks bad to someone- fine, move along and shoot with something else.

Good point. I remember Richard Linklater used a Pixelvision camera for part of Slacker. How would that camera compare to some of these DSLRs! I think audiences can get past what we might perceive as problems with the image if the story and lighting and other things are well done- and the things I see people getting passionate about, the differences often seem so minor to my untrained eye. I would imagine most people would have the same reaction. But I can appreciate the professionals on here wanting only the best. Anyone know how the current crop of DSLRs compare with what was used to shoot 28 Days Later? Just curious. I suppose they're still inferior in a number of ways.

 

[dustylense]

Good point. I remember Richard Linklater used a Pixelvision camera for part of Slacker. How would that camera compare to some of these DSLRs! I think audiences can get past what we might perceive as problems with the image if the story and lighting and other things are well done- and the things I see people getting passionate about, the differences often seem so minor to my untrained eye. I would imagine most people would have the same reaction. But I can appreciate the professionals on here wanting only the best. Anyone know how the current crop of DSLRs compare with what was used to shoot 28 Days Later? Just curious. I suppose they're still inferior in a number of ways.

ANd to add, remember the house episode shot with the mk2? The big hub-bub and build up? I'm a fan of the show anyway, but when that one hit the air I was so interested in seeing it due to it filmed on the Mk2. 5 minutes into the episode I forgot why I wanted to see it and was just tuned into the story. Same thing happened last weekend when the daughter and I (she likes war films) went to se Act of Valor. I went to see the Mk2 at 40 ft. 5 minutes into it, there I was stuck in the story and could care less about the Mk2 part of the film. Same thing with Red Tails, Black Swan subway scene, all sorts of films. Haven't seen any Gh2 though on the big screen that I paid $8 to see, Um.....
I like resolution. Who doesn't. But I like stories and emotion even more.

 

[Adam J McKay]

A fun anecdote in these trying times. I have recently completed my first feature film ( hold for applause ) Thank you you are all too kind. We shot with the 5D2 running Magic Lantern and shooting cinestyle. I graded the entire thing myself on an macbook using colorista 2. On my macbook it looked alright, there were a couple problem areas, but all in all it was ok. We recently had 3 nights of test screening at a local college, the projector was alright, but it was definitely not a cinema projector by any means. We did a test the night before the first test screening and OH NO, some of the shots looked so bad it made me want to throw up. Moire, ****ing ass 8 bit banding, noise, chunky blacks the whole cast of characters that makes the 5D a nightmare to work with. With the screening less than 24 hours away, we were never going to get another HQ export done in time (thanks Colorista) so we decided to just show the version we had, vomit worthy shots and all.

Now since this was a test screening we had questionnaires that we handed out before the show. We asked everyone to read over the questions beforehand so they could think about it during the film. One of the questions pertained to visual quality and imperfections...350 sets of eyes later there was not a single comment about ANY of the things I was worried about people seeing. There were comments about a boom dip here, a crew member there, but mostly a huge number of comments about how they loved the cinematography, the nature was beautiful, the camp fire scene (mainly lit by the campfire shot at 1600) was such and important moment in the film, it made them feel for the main character. But not a single mention of noise, or moire, or sharpness, or banding.

We are truly the only people that care, especially when it comes to narrative filmmaking. I understand that the commercial world is a little different, but the vast majority of people on here are producing content that goes straight to web where a ton more people are consuming said content on their iphones. At the end of the day, buy the best damn tool you can afford that fits your needs and be done with it. I really don't see the purpose of attacking one another because you they use something or disagree with your assessment of sharpness.

 

[KahL]

ANd to add, remember the house episode shot with the mk2? The big hub-bub and build up? I'm a fan of the show anyway, but when that one hit the air I was so interested in seeing it due to it filmed on the Mk2. 5 minutes into the episode I forgot why I wanted to see it and was just tuned into the story. Same thing happened last weekend when the daughter and I (she likes war films) went to se Act of Valor. I went to see the Mk2 at 40 ft. 5 minutes into it, there I was stuck in the story and could care less about the Mk2 part of the film. Same thing with Red Tails, Black Swan subway scene, all sorts of films. Haven't seen any Gh2 though on the big screen that I paid $8 to see, Um.....
I like resolution. Who doesn't. But I like stories and emotion even more.

Good points here.
But, at the same time, you can't simply discount the evolutionary quality of tools as a cinematographer simply because the story may or may not make up for it. Had this been the case, we'd still be shooting exclusively on 8mm film.

 

[GH2user]

But I have the suspicion that the GH2, same with the MK II, produces a certain amount of false detail and therefore the impression of sharpness.

Ok, all this "false detail" talk is just kind of dumb. These are digital sensors converting light reflected by tiny atoms into digital signals that can be saved and played back... EVERYTHING IS FALSE DETAIL. It's just one of those dumb derogatory "buzz words" made up by the established pro-dudes to squander the initial DSLR trend. Moire and aliasing are just interference problems from the sensor from detail being too close. But sometimes the detail is far enough apart that the interference isn't a problem and it still gets rendered to the image. But this doesn't make it "false". It was still created by some piece of detail the cameras sensor saw. It's not more "real" or "false" than any other pixel in the image. This is why I think those test charts are so stupid. They don't tell you how much detail a camera can see... they just tell how how many stupid really close lines it can render. Detail never looks like this in real life.

As I see it, those line charts made sense for film, because film grain is random and you need a pattern of repeatable lines to discern small detail. However, digital is exactly the opposite, it's already a repeatable pattern of pixels, so to really get an idea of representable digital resolution (again, it's ALL false) you would then logically need a RANDOM-PATTERN chart of grain-like detail at various increments. Shooting a line-pattern test chart with a digital sensor is just a cruel joke that was probably thought up by some "established" guys that wanted to make the looming, and threatening, digital technology look as bad as possible. Now it's just the standard... because "somebody" said so, like everything else...

 

[PappasArts]

Kholi reckons they were post sharpened. They spent some money on them so they probably went to a post house.

Very possible. Something was done that's for sure, or their pre-production camera had a special setting nobody else got.

Pappas
http://pappasarts.wordpress.com/
http://twitter.com/pappasarts

 

[HansSteinert]

I invite you to shoot full resolution RAW's with the 24-105 against the 85 1.8 or the zeiss 50 1.4 and get back to me. Of course they are all able to resolve in stills mode. But my 16-35 Eos vrs my 17mmT/S eos is like night and day! My 85 1.8 vrs my 24-105 EOS is like Bakersfield vrs the milky way. Shooting F4 on the 24-105 vrs f8 is vastly different. I suspect that lots of youtube Mk3 footage is shot wide open with the 24-105. Let's see some primo Zeiss glass alongside some primo lighting with some primo focus (in out in out in out in out handheld in out in out at f/2? Of course it's going to look soft even if you had the Hubble mounted to the camera).

As a stills guy, my point is that if you don't hit focus shooting wide open with your subject 5 ft away from you with a 100mm throw under continuous lighting at 1/50th of a second and expect ultimate sharpness, not gonna happen. That is why we developed strobes. Do this. Take your 200mm lens of any quality and price, go outside and shoot a subject 8 feet away with constant light and the lens wide open at 1/50th handheld. Report back with just how sharp that photo is.

I have a Sigma 85mm f1.4, which is sharper than the Canon 85mm f1.8, and I also have the 24-105. I also have Canon's 100mm f2.8 L macro, which is arguably their sharpest lens. The sharpness comparison between the zoom and prime are very noticeable in photography when shooting full resolution. It isn't very noticeable if you were to take the photograph and resize it to 700-800 lines.

 

[dustylense]

I have a Sigma 85mm f1.4, which is sharper than the Canon 85mm f1.8, and I also have the 24-105. I also have Canon's 100mm f2.8 L macro, which is arguably their sharpest lens. The sharpness comparison between the zoom and prime are very noticeable in photography when shooting full resolution. It isn't noticeable if you were to take the photograph and resize it to 700-800 lines.

It would be if your were not resizing and capturing at 2megapixels.

 

[joe 1008]

Ok, all this "false detail" talk is just kind of dumb. These are digital sensors converting light reflected by tiny atoms into digital signals that can be saved and played back... EVERYTHING IS FALSE DETAIL.

You are plain wrong. I own a MK II and I can tell exactly when false detail because of line skipping occurs. It is there and sometimes it looks even fine, it can give a take a crispness that makes it look sharper - until it turns into rainbow moiré and aliasing. That has nothing to do with established guys or digital vs. analogue. I suppose the GH2, to a certain extent, suffers from the same artefacts. At least I saw aliasing in some footage online. But I can't tell to what extent, I'm not a GH2 owner nor did I ever shoot with it.

 

[HansSteinert]

It would be if your were not resizing and capturing at 2megapixels.

At the end of the day, it comes down to whether or not these lenses can resolve the resolution you are shooting at. Canon's 5D III shoot at 1080p. In reality it's more like 700, maybe 850 (but I'm not convinced). If the lens is capable of resolving that amount of lines, how can you blame the lens for a soft picture?

 

[jcs]

Hey Lammy- download his original MP4 from vimeo and add a Sharpen of 50-70 to the wide angle shot by the river. The 5D3 footage will have similar perceived sharpness to the GH2. However, there is indeed a great deal more information in the GH2 shot. Since the shadows are the same, it would appear it was shot at the same time. The 5D3 I-frame codec completely tossed detail from the river's small waves caused by wind. That whole region is smooth! Also look at the tree bark- wiped clean on the 5D3. However, edge detail looks similar. Again, I don't think the I-frame codec is ready for primetime. Would be helpful to see what the IPB codec looks like, as well as a still frame shot. So, if edge detail is still present, but other detail is gone, looks like an issue with the DCT quantization matrix and/or macroblock system: instead if 91Mbps, it looks much lower. Ask Andrew what the actual bitrate was for both cameras. From my own I-frame test, it wasn't anywhere near 91Mbps. For his test, if it's not up near 91Mbps, something is broken: much detail in that shot.

I think Andrew's point is he wants sharp, detailed images straight from the camera. From our tests with line charts, it could be the 5D3 is actually capturing more detail than the GH2, but the I-frame (perhaps both) codec(s) are tossing too much information. Would really help if someone did the GH2 chart test in the same way I did the 5D3.
Again, until Canon tells us how they are processing the Bayer data, we won't know if quality can be improved with a firmware update. My gut feeling is that it can, based on the issues with the I-frame codec and low actual bitrate.

 

[joe 1008]

Jcs - excuse me, but this thread became so monstruos big and I missed your chart. Could you hint me where to find it?

OK, found it... :

http://www.dvxuser.com/V6/showthread.php?276379-5D-MKIII-specs/page108

 

[squig]

I can tell you my Leica R glass smokes my 24-105 Eos. It's twice the optic and they are 20 years old. Lot's are judging mk3 footage on youtube and vimeo (I'd say most are on this thread. I have the mk3, do you?) What glass are they shooting through. THe 24-105 is a fine lens, but it's not perfect. The 50 1.4 Eos is good, but my Leica 50 kills it. Lots of variables out there to consider before passing judgement.

cinema5D have a poll on glass. 50% of people shoot with EOS glass. 19% Zeiss. And us Leica freaks don't even get a mention, we're in the other 10% category. Anyone who knows anything about glass knows that Leica-R glass smokes the competition. I bought mine when I read Shane Hurlbut compare them to panny primo's. Mine are 20-40 years old. I used to shoot with Nikon ai which are nice but the Leica's are stunning. Anyone complaining about IQ should look at their glass.

And before the great defenders of L glass start attacking me let me clarify I didn't say L glass isn't sharp it just doesn't have the warmth, character, and magic fairy dust of a Leica in my opinion...but it's just my opinion.

 

[k.wasley]

Hey Lammy- download his original MP4 from vimeo and add a Sharpen of 50-70 to the wide angle shot by the river. The 5D3 footage will have similar perceived sharpness to the GH2. However, there is indeed a great deal more information in the GH2 shot. Since the shadows are the same, it would appear it was shot at the same time. The 5D3 I-frame codec completely tossed detail from the river's small waves caused by wind. That whole region is smooth! Also look at the tree bark- wiped clean on the 5D3. However, edge detail looks similar. Again, I don't think the I-frame codec is ready for primetime. Would be helpful to see what the IPB codec looks like, as well as a still frame shot. So, if edge detail is still present, but other detail is gone, looks like an issue with the DCT quantization matrix and/or macroblock system: instead if 91Mbps, it looks much lower. Ask Andrew what the actual bitrate was for both cameras. From my own I-frame test, it wasn't anywhere near 91Mbps. For his test, if it's not up near 91Mbps, something is broken: much detail in that shot.

I think Andrew's point is he wants sharp, detailed images straight from the camera. From our tests with line charts, it could be the 5D3 is actually capturing more detail than the GH2, but the I-frame (perhaps both) codec(s) are tossing too much information. Would really help if someone did the GH2 chart test in the same way I did the 5D3.
Again, until Canon tells us how they are processing the Bayer data, we won't know if quality can be improved with a firmware update. My gut feeling is that it can, based on the issues with the I-frame codec and low actual bitrate.

The images of the water were at different times. If you watch the video the shadows have moved. I'm surprised you thought the 5D could smooth that much detail!

 

[squig]

Now since this was a test screening we had questionnaires that we handed out before the show. We asked everyone to read over the questions beforehand so they could think about it during the film. One of the questions pertained to visual quality and imperfections...350 sets of eyes later there was not a single comment about ANY of the things I was worried about people seeing. There were comments about a boom dip here, a crew member there, but mostly a huge number of comments about how they loved the cinematography, the nature was beautiful, the camp fire scene (mainly lit by the campfire shot at 1600) was such and important moment in the film, it made them feel for the main character. But not a single mention of noise, or moire, or sharpness, or banding.

I've been saying this forever but it's been falling on deaf ears. Now I have proof!

The only reason I've been hesitant about shooting my feature with a DSLR is because I would see the vomit worthy shots that I know nobody else would see, but being someone who strives for perfection it would kill me. Now with the MKIII it seems that there's a lot less chance I'm going to vomit and I can get on with far more important aspects like story and performance.

congrats on the film.

 

[dustylense]

At the end of the day, it comes down to whether or not these lenses can resolve the resolution you are shooting at. Canon's 5D III shoot at 1080p. In reality it's more like 700, maybe 850 (but I'm not convinced). If the lens is capable of resolving that amount of lines, how can you blame the lens for a soft picture?

If you put it this way.

Soft+Soft= Softer. Soft+(-)Soft= Less Softer. It's not something I made up. It's something I'm seeing with my 24-105 vrs my 50 R Leica.

Leica R shot
http:// Jaz mk3 85mm by christophermayerphoto.com, on Flickr

Nikor 85mm F2

http:// Jaz mk3 2 by christophermayerphoto.com, on Flickr

Look at the thumb in the second shot. Is that not have detail? That is 600 lines?