Confused about the low ISO score

[ibangpots]

This a comment for this page on the website

The print graph for SNR is almost identical to the D800 but the 5dIII is rated much lower.  Also looking at picture comparisons I'm surprised to see this low number.

[kinematic]

I'm confused as well... how does someone else who uses the same software by DxO and methodology (as standardized by DxO as well) get far different results?

http://www.techradar.com/reviews/cameras-and-camcorders/cameras/digital-slrs-hybrids/canon-eos-5d-mark-iii-1074186/review/page:5#articleContent

Something doesn't jive. But in my own experience with the three cameras, I would say it matches Techradar's results. I'm not one to think that DxO got it wrong, but something doesn't make sense here.

[Vamp898]

Im not 100% sure but there are maybe two reasons

The 5D Mark III have 2 more ISO Stops (even torough they are completly uselss)

So maybe the 5D Mark III looses some points in case of horrible performance at ISO 50k and 100k

or (what would make more sense)

The D800 have a much higher resolution. So at 36MP the D800 have nearly the same performance than the 5D Mark III at 24MP

But if you downscale the 36MP Images to 24MP the Noise is visible less a the D800 than on the 5D Mark III

So with the 36MP you can compensate the noise that you and with less noise at 24MP, compared to the 5D Mark III --> higher score

As this can be easily seen in every test i think also DxO Benchmark calculates this in

[SLRist]

Techradar are using DxO's software to perform the test. Who do you think is more likely to be using DxO's software correctly - DxO or Techradar?

[dosdan]

The print graph for SNR is almost identical to the D800 but the 5dIII is rated much lower.

This is easy to understand if you consider that:

"The DxO Mark Sports Score is the maximum ISO for which the three criteria (SNR ≥ 30 dB, DR ≥ 9 EV, CS ≥ 18 bits) are still met."

So all 3 criteria must be met.  

http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/%28appareil1%29/795|0/%28brand%29/Canon/%28appareil2%29/792|0/%28brand2%29/Nikon

Since the SNR performance is close, that's not holding the 5DIII back.

Neither is the DR at this ISO level where both sensors are close.

It's the weaker CS (Colour Sensitivity) that causes the Canon's Sports (ISO) Score to be lower. The ISO has to be lowered in this instance until the CS = 18 bits to pass all 3 criteria tests.  

Why is the Canon inferior to the Nikon in this regard? Although DXO has mentioned this in the past (http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Canon-500D-T1i-vs.-Nikon-D5000/Color-blindness-sensor-quality), it's worthwhile bringing it up again: Canon's Red channel filter is not very selective, requiring more signal processing and therefore noisier reds:

Compare the R_raw colour matrix lines (R_sRGB, G_sRGB, B_sRGB):

D800:
1.94    -0.86    -0.08

5DIII:
2.27    -1.42    0.15

As it understand it, the smaller the positive & negative co-efficient values are, the better. A less discriminating colour filter is purported to offer a bit better light sensitivity (more photons reaching the sensel), but that's not reflected in their respective QE differences here:

D800: 56%
5DIII 49%
according to www.sensorgen.info

The superior QE purformance is due I believe to Nikon's lead in micro-lens & coating technology.

Note: I believe, SNR and thus QE, is measured on the Green channel (the colour our eye is most sensitive to), so QE measurements alone may not notice any improvement in Red channel sensitivity.

The D800's Sport Score seems to be coming from its SNR figure.  Consider the SNR18% at a measured ISO2211:  31.3dB.

SNR ideally is inversely proportional to log(ISO).  To compare the "Sports Score" of ISO2853 to ISO2211

= 10 log(2211/2853)
= -1.1dB.

So the SNR at the higher ISO should be 31.3dB - 1.1dB = 30.2dB (approx 30dB).

Dan.

[kinematic]

Techradar are using DxO's software to perform the test. Who do you think is more likely to be using DxO's software correctly - DxO or Techradar?

I said that, but DxO certifies their lab, and they use the same equipment, and the Techradar reflects what I've seen in my own personal tests on all three cameras (5Dmk2, 5Dmk3 and D800).

Something doesn't make sense. I know for a fact that there's far more dynamic range when using the 5Dmk3 yet their charts claim that it's lower than the 5Dmk2. In my own tests with both units, I get far more with the 5Dmk3 (about 8-12% depending on the colour) which matches Techradar's findings.

[IAMDXOMark]

Latest Photoshoot Collaboration of DXOMark and Nikon !

Enjoy:

http://i42.tinypic.com/2q83pes.jpg

[dosdan]

It's the weaker CS (Colour Sensitivity) that causes the Canon's Sports (ISO) Score to be lower. The ISO has to be lowered in this instance until the CS = 18 bits to pass all 3 criteria tests.  

I've just noticed that the same topic was discussed in Horshack's post here with the same conclusion:

http://forums.dpreview.com/forums/read.asp?forum=1032&message=41265241

Dan.

[HenryT2]

OK, I kind of understand why the low score for the 5D. Not sure I agree, but I can accept it.

But I still question some of their conclusions. For instance, I was just looking at 3 prime lenses: the Nikkor 105 micro, the Nikkor 50 1.4G, and the Canon EF 1.8 II. Now, this is basically an unfair comparison as the 105 and the Canon 1.8 are completely different class of lens. In theory, there should be some points for build quality, speed and accuracy of focus, etc. But by the specs they report, the 105 is better in resolution, distortion, and vignetting, but worse by transmission and CA. The 50 1.4 is, on the otherhand, better transmission, but worse or the same in every other category. How is it that they both have better overall scores? I can't see how the math would work unless you add a bunch of points for having the name Nikon somewhere in the equation.

[dosdan]

there should be some points for build quality, speed and accuracy of focus, etc.

These are not considered. The dynamic performance of the lens and AF is not considered either. Everything is measured statically.

Please remember that this isn't a full-featured lens review and probably never will be. It's a by-product of the process of measuring the data to be used in the DxO software to correct lens aberrations.

Dan.

[flash01]

The print graph for SNR is almost identical to the D800 but the 5dIII is rated much lower.  Also looking at picture comparisons I'm surprised to see this low number.

This explains it all...

http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Detailed-computation-of-DxOMark-Sensor-normalization

DXO Marks are normalized to a standard print / display size. So the 36 Megapixels of the D800 actually helps it get a better real-world, non-pixel-peeper, result.