LS Foto Home  /  technical  /  12 vs. 14 bit image test

Go to the update 06.11.2007 (on this page)
Go to the update 08.11.2007 (on this page)
Go to the second update 08.11.2007 (on this page)

Date 07.10.2007
This is a test to try to see the difference between 12 and 14 bit images (converted from RAW of course):

  From RAW From jpg
Canon 40D 14 bit RAW Click to view the full 10Mp image Click to view the full 10Mp image
Canon 5D 12 bit RAW Click to view the full 12.7Mp image Click to view the full 12.7Mp image
 
Both images shot using a tripod, mirror lockup, cable release, ISO 100, Auto WB, manual mode, f/8, 2.5sec. shutter, noise reduction off.
The lens used on both tests was the EF 130mm f/2.0L. For the 40D shot, the tripod was moved back 44% to compensate for the pixel
density (the 100% pixel size from the 5D is 4% larger than the 40D in this test). Images on this page are 100% center crops.

Click on the images to view a full 10/12.7Mp image.

Both files was batch processed from RAW using DPP Ver 3.1.0.0.

The files seems very equal. See below for a 400% crop of the two files:


  400% crop from RAW
Canon 40D 14 bit RAW
Canon 5D 12 bit RAW
 
Hard to see any difference even at 400%. Of course there should be many situations where the 4X bit depth of the 14bit file
is better, but I haven't seen any examples yet. The only thing I can see (again!) is the fine details from the 5D!

Date: 07.10.2007
Update 06.11.2007:Back to the top
The first test (at the top of this page) didn't show any difference between 12 and 14 bit images (as I could see). But this updated test really shows that the 40D shows a lot more details, especially in the highlight areas (both shot at ISO100 and the highlight priority was turned off on the 40D).

It could be that the image processor of the 40D is a lot better, but I was pretty surprised of the result here:


Both images shot using a tripod, mirror lockup, ISO100, f/8, 1/250sec. 40D + EF 50mmL, 5D + 85mmL.
100% crop from the center, no post processing, converted from RAW using DPP.

To push it even further, I tried to overexpose and underexpose the images in DPP with 2 stops both ways:



I think it's pretty clear that the 40D's 14bit image is better both in the highlights and the shadows.
Looking forward to the next 5D with 14bit...

To check these images yourself, please click here to download the 5D RAW file (11.0Mb), and here to download the 40D RAW file (9.83Mb).

Update 08.11.2007:Back to the top
Somebody suggested to look at the histogram, and it seems the 40D exposure is a bit different than for the 5D. Both images used the same unmodified Picture Style, so it's still a bit strange (to me) that the images are so different.
Below are the histogram for the actual shots from DPP and CS3:


Histograms of the full images (DPP)
Histograms of the cropped areas (CS3)


From the DPReview foums, I learned that the 5D exposes appr. 1/3 to 2/3 stops brighter than the 40D (because of the larger sensor?). Below are the histograms with the 5D corrected. Seems more similar now:



The same images now with the 5D corrected:

Except from the very bright wall, they are more similar now. Seems like the curve on the 40D is different, and this could explain why the shadow and highlight areas are better. The midtones seems a bit better on the 5D.

Back to the top
Another update 08.11.2007: I did the test again using a zoom lens to be sure the lens did not change the image. Used the EF 24-105mm f/4L at 50mm for the 40D and 80mm for the 5D. The two upper images are the same exposure, and the next two images are from the 5D with -1/3 stop and -2/3 exposure:


40D 50mm, f/8, 1/320sec, ISO 100, ±0
5D 80mm, f/8, 1/320sec, ISO100, ±0

5D 80mm, f/8, 1/400sec, ISO 100, -1/3
5D 80mm, f/8, 1/500sec, ISO100, -2/3


Below are the histograms for the 4 files:

40D 50mm, f/8, 1/320sec, ISO 100, ±0
5D 80mm, f/8, 1/320sec, ISO100, ±0
5D 80mm, f/8, 1/400sec, ISO 100, -1/3
5D 80mm, f/8, 1/500sec, ISO100, -2/3


Interesting to see that the 5D indeed exposes different from the 40D, and -2/3 stop seems very close from these histograms.
All images was shot using AV mode, and in the ±images from both the 40D and the 5D this shutterspeed was the correct exposure (no manual mode).