Is the 5D3 SOFT compared to the 5D2?

[PerfectSavage]

I've own both the 5D2 and 7D and have used them for several years (bought the 7D a month after it was out and the 5D2 maybe six months after it was out) and on occasion rented a 1Ds3 for certain shoots as well. A couple weekends ago I got my hands on a 5D3 and as I didn't have a subject to test in studio, nor did I want to pixel peep at tests 100 others out there have already done, I walked an early Saturday morning in SoHo in NYC for a few hours to shoot street and test the touted DR improvements in areas of dark shadow and contrast cast by the low morning sun through the low rise buildings, fire escapes etc. I shot with both the 17-40 f/4L and 70-200 f/4 IS, choosing those over faster primes to allow me a range of focal lengths and eliminate my penchant for shooting shallow DoF and force me to shoot no larger than f/4 to best gauge sharpness, etc. So, in short, I liked it...a lot. It handled very well, just like the 7D and I was pleased overall with the results (RAW in LR3). So as I prepared to buy one this week I started looking at more comparisons out there (5D2 vs 5D3) and then saw a clip on the video performance by Philip Bloom touting the ability to sharpen video content in PremierePro but saying that "...out of camera video is far less sharp and contrast-y than the 5D2 was (at same settings.)" Coupled with some of the muddiness issues in the shadows (if in camera NR not turned off), JPEG issues, and looking at 100% crops in the latest tests I've seen, I'm starting to wonder, is the 5D3's sensor somehow a little "soft" compared to the 5D2. As many of you now have your 5D3 with shipments sent out the past few weeks, I'm curious what you've seen overall given I only shot with it a few hours. In terms of PURE IMAGE QUALITY (not features, no doubt the 5D3 is a completely new camera feature wise over the 5D2), is the 5D3 sensor producing the same quality of images out of camera as the 5D2's sensor does? I'm wondering if I should be buying as many 5D2's as I can find as a short-term investment vs. a 5D3 (will probably outperform FaceBook's stock) ))

 

[neuroanatomist]

I haven't used a 5DIII nor seen valid, side-by-side comparisons. However, Canon has touted the reduced video moiré of the 5DIII compared to the 5DII, and one of the easiest ways for them to reduce the moiré is to strengthen the AA filter. That would lead to 'softer' images, compared to the 5DII.

 

[bdunbar79]

I just did a lot of field testing with my 5D Mark III today and it is not soft vs. 5D Mark II (that was my workhorse the last 2 years). In fact, the images from today are razor sharp with no post-processing yet. Watch the sharpness setting on Standard. I lowered mine to 1 and that actually made them sharper (no oversharpening of fine details). I did notice that problem. Also, it doesn't seem to under or overexpose vs. your subject. I did a lot of spot and center-weighted average metering today shooting reflective objects in a lot of sun, and they look tack sharp. I shot with a 24-105L, 70-200L, and 100-400L and had great success. Play around with it and get a feel for where the camera settings are optimal, and it will not disappoint you. If you need further help with settings, just post as I've almost completely familiarized myself with the features and functions. Thanks, take it easy and get out there and use it!

 

[bdunbar79]

Did some more testing this evening guys and gals. It seems to be, the most crips images were shot when I used Neutral mode (an auto mode). Normally I build a custom mode, but neutral was the sharpest. I DID have to go in post and add exposure, highlights, shadows, and saturation (I used Lightroom which has some other cool features) but setting the sharpness in my settings, in a sunny situation, left the edges of the bright objects, namely lettering, slightly blurred. I think you can program in an sharpness or contrast you'd like, it just looks like sharpness as low as possible is working right now.

 

[LetTheRightLensIn]

I've own both the 5D2 and 7D and have used them for several years (bought the 7D a month after it was out and the 5D2 maybe six months after it was out) and on occasion rented a 1Ds3 for certain shoots as well. A couple weekends ago I got my hands on a 5D3 and as I didn't have a subject to test in studio, nor did I want to pixel peep at tests 100 others out there have already done, I walked an early Saturday morning in SoHo in NYC for a few hours to shoot street and test the touted DR improvements in areas of dark shadow and contrast cast by the low morning sun through the low rise buildings, fire escapes etc. I shot with both the 17-40 f/4L and 70-200 f/4 IS, choosing those over faster primes to allow me a range of focal lengths and eliminate my penchant for shooting shallow DoF and force me to shoot no larger than f/4 to best gauge sharpness, etc. So, in short, I liked it...a lot. It handled very well, just like the 7D and I was pleased overall with the results (RAW in LR3). So as I prepared to buy one this week I started looking at more comparisons out there (5D2 vs 5D3) and then saw a clip on the video performance by Philip Bloom touting the ability to sharpen video content in PremierePro but saying that "...out of camera video is far less sharp and contrast-y than the 5D2 was (at same settings.)" Coupled with some of the muddiness issues in the shadows (if in camera NR not turned off), JPEG issues, and looking at 100% crops in the latest tests I've seen, I'm starting to wonder, is the 5D3's sensor somehow a little "soft" compared to the 5D2. As many of you now have your 5D3 with shipments sent out the past few weeks, I'm curious what you've seen overall given I only shot with it a few hours. In terms of PURE IMAGE QUALITY (not features, no doubt the 5D3 is a completely new camera feature wise over the 5D2), is the 5D3 sensor producing the same quality of images out of camera as the 5D2's sensor does? I'm wondering if I should be buying as many 5D2's as I can find as a short-term investment vs. a 5D3 (will probably outperform FaceBook's stock) ))

I plan to do a very careful comparison between my 5D2/7D and 5D3 next week before I sell off my 5D2 and 7D of this very thing. The 7D is a trace soft at 100% in part because it uses split green filters, if you raw convert to get max detail then you also get nasty mazing artifacts.

As for the soft video, perhaps it is because the AA filter is aimed at 22MP stills and video is created from 3x3 blocks so they need to anti-alias it a bit and that softens it up? Or maybe they do too much NR of low, super fine contrast detail? With RAW development, in ACR, you can add some pre-raw conversion, small radius sharpening to bring out fine detail and fight the AA filter, I don't think Canon does that for internal jpgs and movies. They use large radius poor quality sharpening for JPG with no apparent pre-demosaic sharpening for fine details and they seem prone to NR away fine low contrast details. Personally I hate the waxy look and wish they'd focus more on natural looking images.

 

[LetTheRightLensIn]

Did some more testing this evening guys and gals. It seems to be, the most crips images were shot when I used Neutral mode (an auto mode). Normally I build a custom mode, but neutral was the sharpest. I DID have to go in post and add exposure, highlights, shadows, and saturation (I used Lightroom which has some other cool features) but setting the sharpness in my settings, in a sunny situation, left the edges of the bright objects, namely lettering, slightly blurred. I think you can program in an sharpness or contrast you'd like, it just looks like sharpness as low as possible is working right now.

Interesting, I heard some people claim that neutral mode made video slightly sharper, even at ISO100 with NR, so it seems they do some nasty bits of NR even at lowest ISO even with NR off and that in neutral it does it the least. I have wondered if custom modes would end up softer then, like marvelcinema etc. since will the camera treat it like neutral processing or standard and use mroe NR?

Anyway you'd think they could offer a fine tune option for this in firmware.

 

[bdunbar79]

Did some more testing this evening guys and gals. It seems to be, the most crips images were shot when I used Neutral mode (an auto mode). Normally I build a custom mode, but neutral was the sharpest. I DID have to go in post and add exposure, highlights, shadows, and saturation (I used Lightroom which has some other cool features) but setting the sharpness in my settings, in a sunny situation, left the edges of the bright objects, namely lettering, slightly blurred. I think you can program in an sharpness or contrast you'd like, it just looks like sharpness as low as possible is working right now.

Interesting, I heard some people claim that neutral mode made video slightly sharper, even at ISO100 with NR, so it seems they do some nasty bits of NR even at lowest ISO even with NR off and that in neutral it does it the least. I have wondered if custom modes would end up softer then, like marvelcinema etc. since will the camera treat it like neutral processing or standard and use mroe NR?

Anyway you'd think they could offer a fine tune option for this in firmware.

It is disappointing for sure. In addition to portraits, landscapes, events, and some weddings, I also have an odd hobby of railfanning photography. I like to shoot a small short line railroad from Mansfield to West Salem, OH. My point here is that I have shot that line many times with 5D Mark II, 24-105L, Standard mode, and now I did the 5D Mark III, 24-105L, Standard mode, same settings, same lighting (since they "run" the same time every Wednesday) and my Mark II photos are sharper than my Mark III photos, noticeably, until I put the camera in Neutral mode, at which point the 5D Mark III blows it away.

You are right, it is frustrating that you pay $3500 and have to deal with this. Oh well, I hope others do other extensive testing like you are doing. Btw, I did the same thing as you, sold my Mark II and 7D for the Mark III upgrade and might buy a 1D Mark IV as a "backup" camera. I still think, despite that, the upgrade will be worth it overall. Let us know what you find out.

 

[wickidwombat]

when i was testing my first copy and posted all the raw comparisons up here I came to the conclusion the 5D2 file were sligthly sharper than the mk3 files the mk3 files had slightly less moire (these are raw only)
however the difference was only really noticable at 200% and not that significant so for most purposes when focus was on target the IQ was basically the same.

That first body had a faulty AF and because the focus was never accurate all the images seemed very soft
when manually focused tethered in live view and tweeked they were basically the same
I returned the first copy and have a new body now which seems pretty spot on out of the box no issues yet

so if you are getting noticably soft images i would seriously have a look at the AF on the camera

try AFMA first if that doesnt give any joy you might want to look at taking it in
my first copy also gave an error when attaching kenko teleconverters which my new copy now has no problem with.

Of course I had to weather a storm of retarded troll's telling me a bad workman blames his tools and I should learn how to use a camera and read the manual... :

 

[Christian_Stella]

While I have been taking a break from my work as a food photographer, so I have not done any studio shots on a tripod with Zeiss glass on the 5d3 yet... I can say definitely say that I am getting super sharp images from this camera handheld.

I shot almost exclusively prime lenses on the 5d2 until buying the 5d3 kit and I'm getting tack sharp photos with the 5d3 and 24-105 handheld. Whether or not this is just the improved AF system giving me more and better keepers remains to be seen, but I am extremely happy either way.

I'll be working on a food project in the next two months, so I'll see how it performs there, but I honestly can't imagine a noticeable difference, other than what I feel is much better color rendition in the 5d3, especially for vibrant colored food.

I also did a reshoot with the 5d3 of a scene for a feature film I shot on the 5d2. Video sharpness looks identical out of the camera to me as I was not shooting a scene that would have a lot of aliasing. Any perceived extra sharpness in the 5d2 was most likely aliasing on fine details like brick and grass. I am quite glad this is gone as the aliasing and moire can really kill the filmic look. With the ability to slightly sharpen in post, the 5d3 definitely edges out the 5d2 in sharpness (but not amazingly so). The high ISO video quality is absolutely astonishing though. Yes, the video can start to get waxy looking, but I can push the ISO so much higher now. I would have to use Neat Video on footage on the 5d2 above ISO 800 which would make the footage very waxy anyway and add a ridiculous amount of time to the render.

There are two tricks you can use to get rid of this waxy look of noise reduced high-ISO video... 1. Light your scene better and use a lower ISO! 2. Add film grain in post. My last stage of video processing is adding real 35mm film grain scans by Gorilla Grain over top of the video. While it may seem counterproductive to reintroduce noise, I think that ISO noise is ugly and film grain is not. Of course your taste may vary... but even at only 50% opacity, this film grain will "dither" your video and effectively cover over that waxy look.

 

[LetTheRightLensIn]

While I have been taking a break from my work as a food photographer, so I have not done any studio shots on a tripod with Zeiss glass on the 5d3 yet... I can say definitely say that I am getting super sharp images from this camera handheld.

I shot almost exclusively prime lenses on the 5d2 until buying the 5d3 kit and I'm getting tack sharp photos with the 5d3 and 24-105 handheld. Whether or not this is just the improved AF system giving me more and better keepers remains to be seen, but I am extremely happy either way.

I'll be working on a food project in the next two months, so I'll see how it performs there, but I honestly can't imagine a noticeable difference, other than what I feel is much better color rendition in the 5d3, especially for vibrant colored food.

I also did a reshoot with the 5d3 of a scene for a feature film I shot on the 5d2. Video sharpness looks identical out of the camera to me as I was not shooting a scene that would have a lot of aliasing. Any perceived extra sharpness in the 5d2 was most likely aliasing on fine details like brick and grass. I am quite glad this is gone as the aliasing and moire can really kill the filmic look. With the ability to slightly sharpen in post, the 5d3 definitely edges out the 5d2 in sharpness (but not amazingly so). The high ISO video quality is absolutely astonishing though. Yes, the video can start to get waxy looking, but I can push the ISO so much higher now. I would have to use Neat Video on footage on the 5d2 above ISO 800 which would make the footage very waxy anyway and add a ridiculous amount of time to the render.

There are two tricks you can use to get rid of this waxy look of noise reduced high-ISO video... 1. Light your scene better and use a lower ISO! 2. Add film grain in post. My last stage of video processing is adding real 35mm film grain scans by Gorilla Grain over top of the video. While it may seem counterproductive to reintroduce noise, I think that ISO noise is ugly and film grain is not. Of course your taste may vary... but even at only 50% opacity, this film grain will "dither" your video and effectively cover over that waxy look.

yeah i've been experimenting with real 35mm grain scan overlays

for wide landscape details, it wouldn't hurt to have a bit more not blurred away though, even ISO100, it does something a touch waxy to all video and jpgs, especially out of neutral setting

yeah it probably has 1.5-2 stops better SNR than 5D2 video and less banding at the high isos so a dark scene at 12,800 on 5D2 was often not even watchable, sometiems even 6400 but on 5D3 it can be fine for some things and it's certainly watchable even in the worst lighting scenarios, less noise and no ugly banding mess

 

[Christian_Stella]

yeah i've been experimenting with real 35mm grain scan overlays

for wide landscape details, it wouldn't hurt to have a bit more not blurred away though, even ISO100, it does something a touch waxy to all video and jpgs, especially out of neutral setting

yeah it probably has 1.5-2 stops better SNR than 5D2 video and less banding at the high isos so a dark scene at 12,800 on 5D2 was often not even watchable, sometiems even 6400 but on 5D3 it can be fine for some things and it's certainly watchable even in the worst lighting scenarios, less noise and no ugly banding mess

I definitely don't disagree that detail isn't even there at ISO100, but I don't think that's anything new to the 5d3. Almost the entire movie I shot on the 5d2 is shot wide in fields and woods and the detail is definitely lacking. On the opposite side of the coin, some of our shots were so cramped that I don't think I could have shot with any other camera.

For filmmaking, I can't go back to before the 35mm grain overlay! Everything looks like video to me without it. The Gorilla Grain I am using has a very soft flicker and silver sheen that I really love too... though I must color grade with the overlay turned on, because it slightly brightens the video.

 

[wickidwombat]

i would post the raw comparisons again but i cant be bothered since rapid share has deleted the files due to inactivity of 30 days but at 200% they are pretty close with the 5d2 slightly sharper if you are getting noticably softer mk3 images i suggest your camera might have a problem. This is based on my experience and i think the problem lies in the AF not in the sensor

 

[PerfectSavage]

All of these responses are very helpful (and educational). Thanks a lot.

 

[PerfectSavage]

While I have been taking a break from my work as a food photographer, so I have not done any studio shots on a tripod with Zeiss glass on the 5d3 yet... I can say definitely say that I am getting super sharp images from this camera handheld.

I shot almost exclusively prime lenses on the 5d2 until buying the 5d3 kit and I'm getting tack sharp photos with the 5d3 and 24-105 handheld. Whether or not this is just the improved AF system giving me more and better keepers remains to be seen, but I am extremely happy either way.

I'll be working on a food project in the next two months, so I'll see how it performs there, but I honestly can't imagine a noticeable difference, other than what I feel is much better color rendition in the 5d3, especially for vibrant colored food.

I also did a reshoot with the 5d3 of a scene for a feature film I shot on the 5d2. Video sharpness looks identical out of the camera to me as I was not shooting a scene that would have a lot of aliasing. Any perceived extra sharpness in the 5d2 was most likely aliasing on fine details like brick and grass. I am quite glad this is gone as the aliasing and moire can really kill the filmic look. With the ability to slightly sharpen in post, the 5d3 definitely edges out the 5d2 in sharpness (but not amazingly so). The high ISO video quality is absolutely astonishing though. Yes, the video can start to get waxy looking, but I can push the ISO so much higher now. I would have to use Neat Video on footage on the 5d2 above ISO 800 which would make the footage very waxy anyway and add a ridiculous amount of time to the render.

There are two tricks you can use to get rid of this waxy look of noise reduced high-ISO video... 1. Light your scene better and use a lower ISO! 2. Add film grain in post. My last stage of video processing is adding real 35mm film grain scans by Gorilla Grain over top of the video. While it may seem counterproductive to reintroduce noise, I think that ISO noise is ugly and film grain is not. Of course your taste may vary... but even at only 50% opacity, this film grain will "dither" your video and effectively cover over that waxy look.

Thanks for the response. One question since you mention the video and there are references to the changes in profiles, etc. (potentially). Since combining footage from both, did you use the same profile(s) when shooting video on the 5D3 as the 5D2? Any noticeable differences, especially in post if color grading etc.? I've been wondering how they might work together, i.e. an A/B cam scenario, with respect to the profiles (custom, CineStyle, etc) as I don't plan to sell either of my 5D2s given the price drop.

 

[Christian_Stella]

To PerfectSavage on the issue of color profiles... I did not shoot with the technicolor anyone else's custom cinestyle. I personally found these way too hard to bring color back into in post, especially skin tones.

I shot on Neutral with the contrast and sharpening all the way down. Because it was only one scene and the whole scene being reshot, I wasn't too concerned with the color matching previous footage, as I am doing extensive color grading anyway. I can say though that I've shot this exact scene with the same lighting 3 times now, twice on the 5d2 and once on the 3 and they all look very similar, though there seemed to be more color in the skin in the 5d3 version while still remaining very flat. Though this could have been a lighting placement thing because I did position the lights somewhat differently.

I'd say you'd see a bigger difference in color and contrast changing lenses than from the 5d2 to 3 on the same settings though. I switched from the Zeiss 21mm 2.8 to the Zeiss 35mm f2 for shooting additional footage and everyone noticed a big difference in color. The 35mm is beautiful and a perfect video lens. The Zeiss sharpness of the wide 21mm is lost on 5d video anyway... But I needed it to shoot in a cramped place.