More 6D sample images - with RAW files.

[x-vision]

Opened with capture one 7.0.1 (preliminary support)

Many thanks.
How about the same but both the 6D and 5DIII at ISO-6400.

There's no doubt that the 6D is better than the 5DIII at super high ISOs.
I'm also interested to find out what 'happens' at ISO-6400.

[Area256]

Very nice. I would say the difference is about 0.7 stops better for 6D (comparing ISO 12800 to 25600). Yummy!

I just had a look again, and the 6d images are softer - that maybe because...

• the pre-release raw converter is not up to it
• the (preproduction!) 6d trades in a little sharpness for less noise, maybe because Canon doesn't expect 6d owners to shoot with sharp lenses anyway, but in low light :-p
• some problem with the lens (50/1.8...)
  

The video samples of the preproduction model were said to suffer from the same thing. Btw it isn't as obvious at higher iso levels since the noise creates an impression of artifical sharpness, for the softness look at iso 800...

I guess we have to wait for raw samples from the *final* retail model with a more recent raw converter, I hope it shouldn't take long since the the 6d is already sold.

From what has been learned from the lensrentals posts, I'm guessing the 50 f/1.8 copy they used just doesn't preform as well on the copy of the 6D they got.  The other issue is the light has moved locations between the tests, so it's possible that is effecting the perceived sharpness.  It could even be that they didn't nail the focus as well, or it could even just be the effect of extra MP on the 5D3.  It may also have nothing to do with the sensor or lens and just the anti-aliasing filter being a little on the strong side.

Overall though, the difference really isn't big enough for concern.  If you really need the sharpness, the D800 is really the best option now.  For me, the difference is too small to matter in my applications.

[Marsu42]

For me, the difference is too small to matter in my applications.

Same for me - but it makes you wonder what a €2000 24-70ii is supposed to do on a 6d if the sensor is in the habit of making it look like a Tamron 24-70 for half the price. That's because I'm on a budget so that €1000 does matter and €2000 for a camera body is a lot, so too much "it'll do somehow" wouldn't be worth it...

[Area256]

It may just be me, but I think Canon has improved their banding issues. 

The images on the left hand side have no NR, and have been pushed by 4EV.

The images on the right hand side have base 50/50 NR, and have been pushed 4EV.

5D3 images are on top, 6D images on the bottom.

Done using Capture One.

Notes:  The 5D3 images were slightly darker than the 6D images with no exposure compensation.  The difference was small enough however that I don't think it would make a huge difference to the banding.

I'm looking for some 6D RAW files with a black sky to really push this test, but I haven't found any yet.

Of course this is still a pre-production camera, but I think things are looking good.

What do others think?

Edit: It's easier to see the difference if you download the full image.

[Marsu42]

It may just be me, but I think Canon has improved their banding issues. 
The images on the left hand side have no NR, and have been pushed by 4EV.

I wouldn't be surprised about an improvement because Canon (5d3, esp. 7d, 5d2) has still very bad banding in comparison over Nikon so there's big room for improvement.

But raising shadows +4ev is not a realistic scenario given the dynamic range of the sensor, the more interesting thing is raising +1ev or +2ev at high iso (i.e. less dr) which might often happen in real life postprocessing... a shot that needs to be raised +4ev is either broken in the first place or was with a lens with extreme vignetting, so only the borders have the issue.

[LetTheRightLensIn]

Quick and dirty tentative result:

Looking at the edge of the frame, the black outer masked part it, looks (SNR as measured in Iris) better than the 5D3 and perhaps even a trace better than the 1DX (5.0 vs 5.3 ADU and 6D also has just a trace more MP) but not close to an Exmor. It might have about the best DR at low ISO that Canon has made so far, maybe about like the 1Ds3. Seems like the 6D,1Ds3,1DX will be Canon's best for low ISO DR most likely. Still a far cry, from Exmor/Aptina/etc., as expected. Rough results though, sometimes the main frame is worse than the masked area, that didn't seem to be the case so much for the 1DX but was for the 5D2/5D3 so perhaps the 6D might end up a touch worse or the same instead of a touch better than the 1DX, probably doesn't realistically matter too much either way.

So not a bad result by any means but hardly cutting age in this era.

It is a bit of a shame that they could not have at least given the 5D3 at least their best effort if they can manage it in the cut down low cost model.... the 5D3 they actually gave a bare trace worse DR than the old 5D2 at low ISO (better at high ISO though). Pretty weak that the 6D will have cleaner low ISO and basically the 5D3 will have the least clean low ISO DR of any body release for the year. That said the 6D is definitely not using some new uber line, not that I expected it, so it's not WORLDS better than the 5D3 in this regard by any means, but it does seem like it will be better and just enough to be noticeable in the real world, looks like you will be able to push it a bit more. At least for low ISO, the sensor seems like it is probably modestly better than the 5D3 sensor.

Doesn't make their 5D3 effort look like they tried so hard and if they can do better in a cheap model then it means they could have done better in a pricey one.

Hopefully this is not off of the rumored new process line otherwise they will still be far behind Exmor for years it would seem, but I bet it is more just a modified 1DX line. Maybe the 7D2 or the new high MP FF later next year uses the new line??? assuming that there really is one

[Woody]

Seems like the 6D,1Ds3,1DX will be Canon's best for low ISO DR most likely. Still a far cry, from Exmor/Aptina/etc., as expected.

I am about to cry

[LetTheRightLensIn]

But raising shadows +4ev is not a realistic scenario given the dynamic range of the sensor, the more interesting thing is raising +1ev or +2ev at high iso (i.e. less dr) which might often happen in real life postprocessing... a shot that needs to be raised +4ev is either broken in the first place or was with a lens with extreme vignetting, so only the borders have the issue.

a shot that needed +4ev raised in the low tones was not necessarily broken by any means, it just means that it had lots of DR, enough with that nonsense

[Area256]

a shot that needed +4ev raised in the low tones was not necessarily broken by any means, it just means that it had lots of DR, enough with that nonsense

Agreed. I shoot a lot of landscape/architecture at ISO 100, and there are lots of times I need to push shadows by quite a few stops - maybe not 4 stops all the time, but pushing by 4 stops gives a good idea of how it will break down, or not.  Plus if you are shooting any photojournalism type stuff, sometimes you miss the getting the right exposure even in good light, and have no choice but to push the image.  That's what make's the Sony/Nikon sensors so great at the moment - they have more DR, and don't "band" when you push them.  I'm still sicking with Canon for the time being, but I would love it if they could improve this weakness of banding + low DR.

[Marsu42]

a shot that needed +4ev raised in the low tones was not necessarily broken by any means, it just means that it had lots of DR, enough with that nonsense

Thanks for your input, but I think you're confusing available sensor dr with postprocessing. If a sensor has rather limited dr (like Canon) raising shadows a lot won't help much because the resolution in the shadows is very low - in these cases, a real hdr should would be required (or get a d800). So at least with whatever I have been shooting, +2ev was the max to leave enough resolution, and imho only in these cases it's important that banding doesn't kick in - except for emergency cases of course.

Btw, this has been discussed all over when the 5d3 was new and it was discussed if tests like this have any real world meaning: http://a2bart.com/tech/allcamdknz.htm

[LetTheRightLensIn]

a shot that needed +4ev raised in the low tones was not necessarily broken by any means, it just means that it had lots of DR, enough with that nonsense

Thanks for your input, but I think you're confusing available sensor dr with postprocessing. If a sensor has rather limited dr (like Canon) raising shadows a lot won't help much because the resolution in the shadows is very low - in these cases, a real hdr should would be required (or get a d800). So at least with whatever I have been shooting, +2ev was the max to leave enough resolution, and imho only in these cases it's important that banding doesn't kick in - except for emergency cases of course.

Btw, this has been discussed all over when the 5d3 was new and it was discussed if tests like this have any real world meaning: http://a2bart.com/tech/allcamdknz.htm

Yeah but the whole point of what he was doing was to find out if there is a point to bothering with being able to raise that much or not.

[baltmin]

Some more comparisons including 1Dx (original files first, followed by resized files - all cameras 18mp)

[baltmin]

resized files

[Marsu42]

Some more comparisons including 1Dx (original files first, followed by resized files - all cameras 18mp)

My suggestion is to do chroma noise reduction and then compare again. Chroma noise is easy and artifact-free to remove in my experience, and if the 5d3 only shows more of this type and just needs a higher setting it isn't much of a difference - it's luma noise that's the main problem.

Comparisons with zero nr are interesting, but what counts is the ability to do (small to moderate) nr without introducing artifacts or killing the details.

[baltmin]

There is noticeable difference in luminance noise as well