[yygomez]

The thing that made the K5 great (and better than other APS-C bodies with regard to dynamic range) was its phenomenal iso 80. For whatever reason, Pentax didn't offer that on the K3, I suppose because it wouldn't have actually improved the dynamic range at all. Still, even back then, I felt that the difference in the measurements between the K5 and the D7000 was not truly significant.

Agreed, They use the same sensor, probably the differences can be explained due to slightly different signal processing and or measurement errors

[Col]

I trust them. It was their measurements that led me not to go from the K20 to the K7.

[drugal]

I trust their measurements as well. When I went from my K-x to the K-5 I noticed a big difference in the dynamic range which is clear in the DxO measurements as well.

Then again, in their explanations of the dynamic range they say that anything over 12 is excellent and differences of less than 0.5 EV are hard to detect. By that reasoning, the difference between the K-3 and the 7100 may not be noticeable, but the K-5 to K-3 difference may be. All of the cameras are excellent, however

[normhead]

A DxO difference of more than 5 points may be noticeable. 12 points is certainly noticeable. Less than 5, it may as well be the same camera. The 7100 from what I understand has a professional flash system. It doesn't have shake reduction, ( worth 2-3 stops on the low light end) or a TAV mode.

[calsan]

If you replace your camera every year, it "may" be worth checking.
Anyone replacing their camera on a 5 year cycle can save themselves the bother of reading tests, unless they have trouble sleeping… zzzz

[Bolt]

"So, we have a single 'weighted average' number, with non-linearity in the calculation mentioned elsewhere in the article, and no information on the actual calculation, what is weighted vs. not, not even confirmation that the same weightings are universally applied. That, IMO, renders the Sensor Score meaningless, a problem further compounded when people erroneously extrapolate that meaningless number to be a measure of overall camera performance."

So, I was having a discussion with a friend of mine who is also a photographer about the accuracy of DxO Mark.

In general, he was in favour of the immediate report and scoring that DxOMark.com refers to when referring to the 'Camera Sensor Ratings' and the scores they show in their ranking list. I disagreed, claiming that the results shown on the immediate report are, in my opinion, inaccurate.

For reasons not mentioned, we were comparing the Nikon D800 and the Canon 1D X. Specifically, the Dynamic Range.

I claimed that the results that the D800 had a Dynamic Range of 14.4 Evs compared to the 1D X's 11.8 is in part true, but an inaccurate representation of the tests overall results. Here is why:


The reported Dynamic Range of the D800 is 14.4 EV. While this is true, it is a representation of the MAXIMUM possible Dynamic Range at only one ISO, not the actual performance throughout the ISO range.

The Dynamic Range of the D800 drops off at an average rate of .91 Evs every increase in ISO.
Compared to the Canon 1D X, which averages an average .46 Evs loss every increase in ISO.
To put it relatively, .46 Evs is less than the smallest loss the D800 sees in an increase in ISO.

After three increases in ISO, the D800 has lost 2.31 Evs.
Compared to the 1D X that has lost 0.16 Evs.

After five increases in ISO, the 1D X has better Dynamic Range throughout the rest of the ISO.

After nine increases in ISO [limit of D800] the D800 has dropped to half its original Dynamic Range.
Compared to the 1D X that has lost as much as the D800 did after five increases in ISO.

The average Dynamic Range throughout the ISO differs by .33 Evs
1D X - 10.60
D800 - 10.93

The deviation between highest and lowest possible levels of Dynamic Range throughout ISO differs by 90%.
1D X - 3.73
D800 - 7.33


In conclusion: [referring to Dynamic Range throughout ISO]
The results displayed on DxOMark's 'Camera Sensor Ratings' scorecards are in my opinion mis-informative of the actual results shown throughout the camera's performance. While in 'perfect' conditions the Dynamic Range of the D800 WILL outperform the 1D X, the average results, standard deviation and high-to-low end loss of performance would suggest that the 1D X is a better overall performer in regards to Dynamic Range throughout the ISO spectrum.

[These results are based only upon those given by DxOMark. I do not believe DxO is an inaccurate or dishonest company. I posted this only to represent a more in depth look at the results given on the immediate Camera Sensor Ratings scorecard. In my honest opinion, this is why I PERSONALLY believe the results shown on the Scorecard are an inaccurate representation of the camera's overall performance]

Now I can wait for the D800 lovers to tell me how much of a moron I am.

Here is a 'graph' of the results:

1DX

11.77
11.81 +.04
11.74 -.07
11.61 -.13
11.18 -.43
10.37 -.81
9.88 -.49
8.99 -.89
8.08 -.91

Deviation 3.73
Average 10.60

-

D800

14.33
13.65 -.68
12.83 -.82
12.02 -.81
11.12 -.90
10.09 -1.03
9.17 -.92
8.16 -1.01
7.00 -1.16

Deviation 7.33
Average 10.93

[derekkite]

What these test tell me is that almost any body produced right now can generate stunning shots in skilled hands.

When I bought the K5 a couple years ago the sales guy at London Drugs, an old Minolta hand, described the time it took for film to mature compared to the time it has taken for digital to get to this point. The differences in sensors from here on, unless there is some major breakthrough, are marginal, and can be more described as flavors as opposed to better or worse. Some may be better for a style of shooting. But the big differences are, as in film days, handling and lenses.

[rawr]

Here is a 'graph' of the results:

Here's a better one [a more apt comparison BTW would have been between a Nikon D4 and the 1Dx, since they are direct competitors]

[derekkite]

Is the EOS 1Dx usable at those high isos? Compared to the D800? I'm not familiar with either, just curious.

[audiobomber]

I put a lot of faith in DXO's methods, but it's important to read the background info and realize what the scores and graphs mean and what they don't mean. Two points difference in overall score is considered meaningless. Resolution is a good example of something the scores do not show. Clearly there's more to image quality than the DXO metrics. If the DXO score is a true measure of IQ, how would you explain the K-5 beating the 645D?

In conclusion: [referring to Dynamic Range throughout ISO]
The results displayed on DxOMark's 'Camera Sensor Ratings' scorecards are in my opinion mis-informative of the actual results shown throughout the camera's performance. While in 'perfect' conditions the Dynamic Range of the D800 WILL outperform the 1D X, the average results, standard deviation and high-to-low end loss of performance would suggest that the 1D X is a better overall performer in regards to Dynamic Range throughout the ISO spectrum.

The Dynamic Range score has relevance. It tells the maximum DR for the camera, which is an important measure for landscape photography. You don't normally use anything above base ISO for shooting landscapes on a tripod. They state what the number means quite clearly, it is the maximum DR achievable. To get the whole story, you need to look at the graph, which shows how DR changes with ISO.

[rawr]

Is the EOS 1Dx usable at those high isos?

I've shot alongside a 1Dx user at several night events, and seen the results. Most times he just kept his ISO at 12800, his shutter at 1/500 and aperture at f2.8 the whole night. The results looked OK for small web / Facebook size images - noise was well controlled, but detail suffered, and all the colour and tonality had bled away, so the images looked as flat as cardboard. I don't think they would print very well past 4x6 size. I believe I often produced better output with my K-5 going up to ISO 6400, in combination with careful shooting and good PP.

[civiletti]

I do find the DXO measurements very useful, specially the dynamic range, and for those saying they are biased remember the K5 not long was on their list the best APS-C, and right below some FF and medium format bodies

A small difference in DR can be important. It can mean the difference between recoverable and nonrecoverable detail in some images. The great DR of the K-5 is a great asset in landscape photography.

[azimuthal]

The results displayed on DxOMark's 'Camera Sensor Ratings' scorecards are in my opinion mis-informative of the actual results shown throughout the camera's performance. While in 'perfect' conditions the Dynamic Range of the D800 WILL outperform the 1D X, the average results, standard deviation and high-to-low end loss of performance would suggest that the 1D X is a better overall performer in regards to Dynamic Range throughout the ISO spectrum.

It's a better performer by what criteria? Even you mention it averages out in the D800's favor?
"high-to-low end loss of performance" and standard deviation (which in this case seems to be the same thing?) aren't necessarily bad here...

For instance, if two cameras followed exactly the same curve except at the base ISO setting one had 5 ev better performance, it would have a higher "high-to-low end loss of performance" but few (besides, maybe you) would argue that this is a bad thing...most would say the higher peak performance gives up nothing and that would be true in this scenario.

Like someone else already said the "landscape" score reflecting the peak DR performance makes a lot of sense, but if you wanted to include "typical" usage patterns you would probably want to look at iso 100 to 6400, and I wouldn't mind seeing an average of those in their sensor reviews, in this case the average is .7 EV in the D800's favor.

The Canon does not make up for its poor peak performance at base iso (2.5 EV deficit) anywhere in its iso range, at best it beats the D800 by ~1 EV.

I think you did a great job of pointing out that the landscape score does not reflect performance across the iso range, you did a poor job of convincing me that the 1DX is a better overall performer in this context, it's just not.

[screwdriver222]

Check out their scores for lenses on the K5 and you get strange results. The cheap DA 50 F1.8 is second highest overall and second highest in sharpness.

Regards

Jeff

[normhead]

Not only that, the scores are not in any way standardized. A lens on a k5 gets a lower rating than the same lens on a D7000 even though the k5 has a slightly higher sensor rating. And there is no mention of the deviation that exists from one lens to the other. I trust their ratings to be what they are, an inadequate attempt to provide information about cameras and lenses. They can help you pick a direction to investigate, but they can't predict whether you'll get a good or bad copy of a lens, whether or not your lens will need warranty work, whether or not you'll like the way the lens handles and focuses, or whether or not you'll use it. For Pentax, they test very few of the available lenses and hardly any 3rd party lenses. Possibly a better resource for Canon and Nikon users. So I'm happy it's there, and I'll read what they have to say. Someone who has actually tested a bunch of lenses and come up with an opinion is worth listening to. But I believe their rating system exaggerates the small differences between cameras. The best policy is still the user reviews. FInd someone who's written a review, who shoots what you'd like to shoot in the way you'd like to shoot it, and see what they have to say. That's still the most useful information.