My take on DxOmark.
Take w a spoon of pepto but I now prefer the SCREEN comparison on DxO rather than the PRINT version.
I print big, their test print is 12x18 inches i think, at about 180 or 200ppi to = 8MP.
18x12 is the smallest print size I make, which matches up well to the 10MP output of my old 40D, for example.
If you look at the screen comparisons for a lot of cameras what you'll find is that this pixel-level type of result is very good at showing you how various camera systems can perform when you're viewing on screen at 1:1 size. Go ahead, check your favorite cameras... I'll wait...
OK. notice how much more similar SCREEN results on most of them are at:
- 18% SNR?
- Tonal Range?
- Color Sensitivity?
- Dynamic Range? ( more on this one later )
What's that telling us?
To me it looks like most camera system mfrs are doing a pretty good job pressing the limits of physics and electronics.
If you look at cameras that are at or below about 10MP, SCREEN and PRINT numbers are closer than they are for higher rez cameras of a similar type. It's once you get above about 10 MP sensors that the S vs P numbers start to show more of a difference, with PRINT pulling ahead.
Likely because the merging of pixel data from higher resolution sensors into few effective pixels for print tend to average out noise and other pixel-level inconsistencies.
I think if the test print was 30x20 inches at 200ppi we'd see different results until sensor resolutions exceed 24MP.
SO, if you're using more than 8MP worth of sensor data to print a similar 18x12" print at 200ppi you're likely getting dimishing returns from higher resolution gear but your results may align better to the PRINT results and actually look a little better.. however it is that they actually measure that.
If you're printing huge sizes where individual pixel data starts to become visible at nose-to-paper distances then your results may be more closely related to the SCREEN results.
Back to the Dynamic Range spec. This is where Canon's gear is reaching its limits at iso settings of 800 and lower compared to the competition. At the limited print size used for the DxO test this is still not likely to show very much of the low iso banding problems some of us hardware-pushing types complain about. If you only print little postcards sizes like 8x10 or 11x14 (
) you're not very likely to notice banding except perhaps on screen at 1:1 size.
I've used a 7D that showed significant low-iso banding that's sometimes so bad it's very visible on screen with only a slight tone-curve tweak to bring up some shadow area details. There are vertical bands 8 pixels wide, 8 pixels apart, across shadow areas of low iso images. Even if making an 18x12" print that's at 300ppi, that's less than 20 line-pairs of noise per inch, something that can quite easily be noticed.
I also shoot with a 60D, the DxO scores for which are virtually identical to the 7D with the latter tending to rate slightly higher.
Funny thing about the results tho, the 60D suffers considerably less low-iso banding issues than the 7D. They're not showing up in the shadows of my larger prints or on screen, certainly not to anywhere near the extent they are from the 7D. If I were to look at the DxO results, the 7D looks better by a hair. Real life is not the case in my experience.
I don't know exactly what criteria or methods DxO is using for their SNR and DR measurements, whether SCREEN or PRINT. I'm not sure if they take into account this sort of pattern noise that some cameras add too much of to an image. If averaging out the noise data, the banding will have less of an effect on the measured specs yet still show up as a problem in images.
Still, the better the SNR and DR numbers are then the less likely this is to be a problem.
I'm looking forward to seeing how much of DxO's reported 2.5 stops of base-iso DR lead for the D5100 translate into real-world usability over a 60D or 7D. And maybe over the 5D2 as well.
Would that make the $650 marvel a better camera than the 5D2? heck no! but it makes it worth using in situations I now know some of my Canon gear will not be able to handle as well.
Just trying to use the best tool for the job.