6D - sensor measurements and comparison with other Canon FF

Hi, from files of a friend's 6D I have measured read noise, dynamic range and banding intensity, in comparison with 5DII, 5DIII and 1DX: http://legault.perso.sfr.fr/6D.html

In summary, the 6D holds out on 5DII and 5DIII and is not so far from the 1DX (they have similar read noise at high iso). Here is the dynamic range chart:

--
Thierry Legault - www.astrophoto.fr

how are you reading the lines and are you taking into account column and row black point offset data?

There is some decent DR improvement on 6D (although not in EXMOR territory).

I wish this sensor was in 5D3 as the extra 0.6 EV DR will be a wellcome improvement as I find noise in shadow area on 5D3 unacceptable.

What I was expecting to see!

For what I see the 6D is superior to the 5DIII by a bit...The 5DII held up well! Is it me or the 5dII is better than the 5DIII in some aspects?

--
http://www.flickr.com/jdfeos

Yes, I too want to know how to make sense of the 5DII graph, which is higher than that of the 5DIII.

I expect 5D2 and 5D3 graphs to overlay each other. Also, 1DX outperforms 5D2 by quite a bit even at low ISO, this was not shown

The 5DII has a higher dynamic range than the 5DIII at low iso because of its lower read noise (dynamic range is directly proportional to read noise). My measurements gave, at 100 iso, 25 electrons for the 5DII and 34 electrons for the 5DIII (34 electrons too for the 1DX).

This was the first time that read noise did not improve from one generation to the next, and it was all the more disappointing since Sony Exmor sensors (used in most Nikon and Pentax cameras) already have a very low read noise at low iso (under 10 electrons), giving them a much higher dynamic range. Fortunately, all Canon cameras have a very good read noise at high iso, especially the 1DX and 6D (under 2 electrons at 6400 iso).

--
Thierry Legault - www.astrophoto.fr

I'm afraid that "quite a bit" doesn't make much sense on this kind of chart and even in the real world 

I have measured a second 6D and found a read noise 10% higher, leading to a dynamic range lower by 0.15 Ev. Even if those cameras are produced by controlled industrial manufacturing processes, there are variations from a batch to another and even from a camera to another. Take a look at the DxO dynamic range charts for the Nikon D800 and D800e. In theory, they should exactly overlap since it's the same sensor (the low pass filter has no influence at all on the dynamic range): you'll see a difference of 0.2 Ev at 200 iso between those two cameras!

--
Thierry Legault - www.astrophoto.fr

I use Christian Buil's software (Iris) to open raw files and make raw data measurements. Offset black point is directly visible through the raw pixels values, in addition is it visible in an exif field in the files (if you know where to look).

--
Thierry Legault - www.astrophoto.fr

I meant: inversely proportional, of course 

--
Thierry Legault - www.astrophoto.fr

How do you explain the noticeable difference between your measurements and those of DxO?

I don't know what you mean by "significant" but I have found no difference higher than 0.4 Ev and generally they are within 0.2 Ev (make sure that you display the chart in "screen" mode). Since dynamic range measurement is very straightforward, I explain the differences by the fact that different cameras are measured. After all, DxO found a difference of 0.2 Ev at 200 iso between the D800 and the D800e, so we could say that DxO finds differences...with itself! 

--
Thierry Legault - www.astrophoto.fr

Didn't say "significant" - said "noticeable". Of course, by that, I didn't mean noticeable in final images!

So, in effect, the difference is within the tolerance of measurements and variability of different bodies?

you're right 

Yes I think so. I see no other explanation (you can't really make a mistake in measuring dynamic range, it's just a ratio between two figures, one is noise and the other a fixed value). I have measured variations of 0.2 Ev between different bodies, therefore even a difference of 0.4 Ev between two extreme ones is not impossible. Imo, the second decimal in the DxO chart means nothing, and even the first decimal shall be taken with caution. Even if people sometimes expect the Holy Truth (with capitals) from those charts, they are not destined to be watched with a magnifier! 

--
Thierry Legault - www.astrophoto.fr

Thanks for the answers Thierry - really appreciate your honesty and pragmatism regarding these measurements (which you have obviously gone to some trouble to gather).

This "tolerance and variability" would probably also go a long way toward explaining what many see when comparing these cameras (that is, very little difference) and also the continual arguments about the performance of one versus another. Hehe.

Anyway - thanks for your work. I appreciate it.

From what I see ,at ISO 100, 5DMK3 is slightly better than 6D ? (Less horizontal and vertical lines are showned ). It appeares that way !!!

I'm talking about banding !

But what about DXO Labs ? They say different ! They say 5D MK3 is better than 5D MK2 at all ISO values for noise and only with 0.2 DR worse between iso 100-1200 and better after iso 1200 than 5D MK2. i'm very curious to see what they has to say about 6D.

Crops showing banding are taken at random on the image (anyway it's not a fixed pattern, the lines move from an image to the next even if the average "noise" intensity remains the same). What they show is that actually all recent Canon FF are not very different, but figures (chart) show that 6D is better than 5DIII vertically and similar horizontally (except at 100 iso).

--
Thierry Legault - www.astrophoto.fr

thank you thierry for the explanation and your hard work! it's always nice to see such objective data.