Canon 5D Mark III - Resolution Review

[jcs]

Received a 5D3 this morning; quick resolution test: http://cinema5d.com/viewtopic.php?f=30&t=39795

[Stephen Melvin]

Nice conclusion to your post. Glad to see such positive news about the video capabilities, instead of all the "this looks soft" posts.

I don't have permission to view your attachment. I assume it's a video clip?

[jcs]

Images attached. First file (5D3S_Sharpen.jpg) is scaled down still (ideal case), second file is 1080p video frame. Both sharpened @33 (PPro filter).

[peederj]

If this is the cost of ridding ourselves of aliasing and moire, I'm not crying. Sharpness as a concept is overrated in video: since you can't photoshop the imprefections off the talent, you aren't doing yourself any favors shooting sharp.

[Stephen Melvin]

That video grab looks really, really good, with the still as a reference.

[jcs]

It looks like Canon is doing an Unsharp Mask sharpen for video, thus the halos. I'll turning off in-camera sharpen and test again. Note that the still was sharpened the same as the video (and no halos in the still).

For those who want to try this at home: http://www.bealecorner.org/red/test-patterns/ . Comparing a still to a video frame is easier than setting up a proper (and correct) resolution test, especially when your printer can't really print the entire chart at full resolution (pre-made charts are $150+). My Xerox Phaser 6180DN was able to do a good job up to the 4:3 marks on one sheet (could print out 3 sheets in sections). The still frame compare was good enough for my curiosity.

For those with 5D3's: try turning off in-camera sharpen and sharpening with a convolution style sharpen in post. Looks excellent running in real-time in PPro (using the GPU sharpen filter around 33-48).

[Stephen Melvin]

Any other thoughts about the camera, now that you've had it for a whole day? I read a review on one of the video sites, and that guy didn't like it much at all.

I'm getting mine primarily for stills, but it'd be nice to know that it's a potent video camera, too. I know I'd like to have had it at an event I shot about six weeks ago. Lots of girls with fishnets, and the moiré came out to play.

[jcs]

Just tested indoors with the 70-200 2.8 II- looks great. Used "P" setting, "Faithful" profile, and sharpened to 33 in PPro (renamed file .MPG to play in realtime). 100% pixel peeping looks pretty good. Looks excellent playing back full screen. Will test on a real shoot soon (will also test other lenses).

Best results so far:

1. "Faithful" profile (0 everything)
2. IPB mode
3. Rename file .MPG for real-time playback in PPro CS5.5
4. Sharpen as needed, 33 seems to be a good starting point

[peederj]

jcs are you finding IPB is better than ALL-I for tripod shots of nonmoving subjects? I was thinking this would be the case.

And is faithful better than neutral for you? When you say 0 everything, does that mean sliders fully to the left or dead center?

[psolberg]

and a not so great review

http://www.eoshd.com/content/7551/canon-5d-mark-iii-review

the 5DmkIII still falls to the GH2 and just marginally better than the prior, plus noise issues with the All-I codec.

[jcs]

peederj- haven't done a real shoot with people yet, however there is a technical issue with I-frame- perhaps not being deblocked correctly (camera encoder or NLE decoder issue). IPB looks great and uses less memory; not missing I-frame at this point. Haven't tried Neutral yet- should work too. All 0 is the setting for parms. Some params go negative.

psolberg- sounds like the 5D3 is doing better for Andrew after turning off camera sharpening and sharpening in post: http://www.eoshd.com/comments/index.php/topic,449.0.html#d2

[peederj]

Some people have claimed Neutral is a completely different processing than the other modes, so i would prioritize that. Many people have been shooting exclusively neutral (myself included) since Vince Laforet originally recommended it and explained why years ago.

[Axilrod]

jcs are you finding IPB is better than ALL-I for tripod shots of nonmoving subjects? I was thinking this would be the case.

And is faithful better than neutral for you? When you say 0 everything, does that mean sliders fully to the left or dead center?

ALL-I is going to be better than IPB for everything.  As for picture style, 0 is to the far left for sharpness and dead center for contrast, sat, and color tone.  Sharpness needs to be cut down all the way no matter what, and if anything cut down the contrast a bit.  With the Cinestyle profile they suggested 0 for sharpness, -4 for contrast and -2 for saturation. 

[jcs]

In PPro CS5.5, I-only is only good for real-time playback without renaming the file from MOV to MPG (to get around a PPro bug).

I have analyzed the footage: it has more artifacts than IBP (which is more efficient- than I-only). The artifacts look like a deblocking issue (horizontal and vertical lines). Andrew Reid at EOSHD reported this issue as "mosquito noise", which is consistent with a deblocking issue (my guess, could be something else).

[Stephen Melvin]

In PPro CS5.5, I-only is only good for real-time playback without renaming the file from MOV to MPG (to get around a PPro bug).

I have analyzed the footage: it has more artifacts than IBP (which is more efficient- than I-only). The artifacts look like a deblocking issue (horizontal and vertical lines). Andrew Reid at EOSHD reported this issue as "mosquito noise", which is consistent with a deblocking issue (my guess, could be something else).

In your initial comments, you posited that this could be a playback issue, rather than an encoding issue. Do you still think this is a possibility?