DxOMark on D3x

[ejmartin]

How come the DXOMark dynamic ranges are so different from most other estimates. The D200 is said to have 11.5 stops DR, compared to estimates by others of 8 stops.

One has to be careful about what measure of dynamic range is being discussed.  DxO uses the most liberal definition, that of sensor saturation level devided by minimum recordable signal (signal level equal to the noise, S/N=1).  Others use a higher standard of minimal acceptable S/N, or use converted jpegs, or some other measure.  The DxO choice is the standard engineering definition of DR.  Whether that is the most useful one for photography is a separate issue.

[Panopeeper]

The sensors ARE "the same" - at least the portion that does not include the AA filter and also micro-lenses

The sensors ARE NOT the same (Nikon D3X and Sony A900). These sensors include the analog/digital conversion, and that is obviously very different: the D3X sensor creates between 14800 and 16384 different levels, in 14bit mode, while the A900 sensor creates about 4000 levels, supporting only 12bit.

Added: not even the pixel dimensions are the same.

[barryfitzgerald]

Looks like we have another real world test for Michael to do

A900 v D3X v 5d MkII dynamic range..

[NikosR]

I guess this is not in dxomark scores, for there is no such thing. The color reproduction of the sensor is totally independent of the ISO.

Of course you are right. But color separation during demosaicing is affected by noise.

[BernardLanguillier]

Looks like we have another real world test for Michael to do

A900 v D3X v 5d MkII dynamic range..

For such a test to be really interesting:

- It would have to include a 22MP and 39MP MFDB digital back also, let's say a P25+ and P45+,
- It would have to be based on using a best in class lens for each mount (let's say a Zeiss ZF 100mm f2.0 or 60mm macro lens on FX bodies) at f5.6 - f8
- It would have to focus on ISO100 (or the best ISO for each camera),
- It would have to be based on conversions from C1, at least for the D3x and Phase backs, both with optimal capture sharpening (I can provide the best settings for the D3x at least),
- It would have to be made from the top of a sturdy tripod with MLU and cable release,
- It would have to use Live view for accurate focusing.

Based on what I have seen so far, I expect the result to be P45+ > D3x = P25+ > A900 > 5DII

The draw between the D3x and P25+ being based on:

- similar micro detail at any print size (A1 would be a good test size) even if the P25+ still shows a tiny bit more micro detail when viewed at 100% on screen
- Slightly more DR for the P25+
- A lot less artifacts for the D3x

Cheers,
Bernard

[ErikKaffehr]

Bernard,

Sorry, I don't think Michael is the DSLR/MFDB testing business  

It's nice to hear that you are satisfied with your D3x, by the way!

Best regards
Erik

For such a test to be really interesting:

- It would have to include a 22MP and 39MP MFDB digital back also, let's say a P25+ and P45+,
- It would have to be based on using a best in class lens for each mount (let's say a Zeiss ZF 100mm f2.0 or 60mm macro lens on FX bodies) at f5.6 - f8
- It would have to focus on ISO100 (or the best ISO for each camera),
- It would have to be based on conversions from C1, at least for the D3x and Phase backs, both with optimal capture sharpening (I can provide the best settings for the D3x at least),
- It would have to be made from the top of a sturdy tripod with MLU and cable release,
- It would have to use Live view for accurate focusing.

Based on what I have seen so far, I expect the result to be P45+ > D3x = P25+ > A900 > 5DII

The draw between the D3x and P25+ being based on:

- similar micro detail at any print size (A1 would be a good test size) even if the P25+ still shows a tiny bit more micro detail when viewed at 100% on screen
- Slightly more DR for the P25+
- A lot less artifacts for the D3x

Cheers,
Bernard

[BernardLanguillier]

Bernard,

Sorry, I don't think Michael is the DSLR/MFDB testing business  

Erik,

Well, if he doesn't do it, somebody else will...

Cheers,
Bernard

[Ray]

For such a test to be really interesting:

- It would have to include a ...........

Bernard,
Keeps things simple. I understand you have both the D3 and D3X. You are in an ideal position to compare these two cameras and check your results with those from DXOmark.

According to DXOmark, at the pixel level the D3X produces almost one stop more DR than the D3 with 100% more light reaching the sensor. Sounds reasonable to me. The lowest ISO for the D3 is ISO 161 (camera's ISO 200). The lowest ISO for the D3X is ISO 78 (camera's ISO 100). On an 8x12 print, the D3X dynamic range is claimed to be almost 1.5 stops greater than the D3's. At the pixel level, or on screen at 100%, it's slightly less than one stop greater.

I'd have no problem finding a scene with a very high 'subject brightness range' in Australia, because the light is so bright over here. If you are having trouble finding such a scene in Japan, send the cameras over to me   .

[BernardLanguillier]

I'd have no problem finding a scene with a very high 'subject brightness range' in Australia, because the light is so bright over here. If you are having trouble finding such a scene in Japan, send the cameras over to me   .

I assume that you would need a set of lenses also?

Cheers,
Bernard

[Ray]

I assume that you would need a set of lenses also?

Cheers,
Bernard

I've already got one very fine Nikkor lens, the 14-24/2.8   .

[BernardLanguillier]

I've already got one very fine Nikkor lens, the 14-24/2.8   .

That's only a pretty average lens in the Nikon world...

Cheers,
Bernard

[Slough]

One has to be careful about what measure of dynamic range is being discussed.  DxO uses the most liberal definition, that of sensor saturation level devided by minimum recordable signal (signal level equal to the noise, S/N=1).  Others use a higher standard of minimal acceptable S/N, or use converted jpegs, or some other measure.  The DxO choice is the standard engineering definition of DR.  Whether that is the most useful one for photography is a separate issue.

Thanis. I had a suspicion that DxO used a more theory based definition rather than a real world one. On that basis it is not even clear is we can compare differences, as even those may differ from real world measurements.

[Panopeeper]

According to DXOmark, at the pixel level the D3X produces almost one stop more DR than the D3 with 100% more light reaching the sensor

They must have hired a drunkard to make the measurements.

[Tony Beach]

I had a suspicion that DxO used a more theory based definition rather than a real world one. On that basis it is not even clear is we can compare differences, as even those may differ from real world measurements.

http://forums.dpreview.com/forums/read.asp...essage=30703740

[Panopeeper]

That's only a pretty average lens in the Nikon world...

LOL, then what is an "over the average" lens in the Nikon world, among the wide ones?

[cecelia]

LOL, then what is an "over the average" lens in the Nikon world, among the wide ones?

I was thinking the same thing!  I agree wide lenses are never as good as mid-to-tele lenses from a Strehl/CA... perspective, but it is hard to find any fault with this lens beyond weight and lack of filters...I love mine for when I need the focal lengths.

-Cecelia

[ErikKaffehr]

Hi,

I didn't really get the stuff with 100% more light reaching the sensor. I'm eagerly awaiting your evaluation of the D3x using your "Stouffer edge" before discussing drunkards. Got the impression on what's written on these forums that Nikon did a good job on the D3x. I'd point out that under most conditions all cameras will produce good pictures. IMHO we have to much focus on how horrible pictures get at 3200 ISO? In my film days Velvia (at 50 ISO) was OK and not much else ;-).

Best regards
Erik

They must have hired a drunkard to make the measurements.

[marcmccalmont]

That's only a pretty average lens in the Nikon world...

Cheers,
Bernard

Please list your excellent lenses! I thought the 14-24 was outstanding? was thinking of getting one to replace my Canon 17-40L
Marc

[Panopeeper]

I didn't really get the stuff with 100% more light reaching the sensor

I guess you are referring to Ray's post. Half the number of sensels on the same size of sensor means twice as large area for the light collection (roughly).

I'm eagerly awaiting your evaluation of the D3x using your "Stouffer edge" before discussing drunkards. Got the impression on what's written on these forums that Nikon did a good job on the D3x

If it is 0.5 EV better in noise than the 5D2, then they made a very good job. The 1.5 EV is in the sphere of hobby fishers.

[Ray]

That's only a pretty average lens in the Nikon world...

Cheers,
Bernard

Bernard,
For a comparison of noise in the shadows, the sharpest lens available is not necessary. For a practical test in the sort of real-world conditions where most photographers would find the additional DR of the D3X useful, I had in mind something like a rainforest scene with bright sunlight streaming through the trees onto a waterfall surrounded by dark, impenetrable undergrowth. You would take a series of exposure bracketed shots of the waterfall with both cameras (1/3rd stop interval), not only to ensure that you get as close as possible to a full ETTR, but to ensure that you have a few overexposures in order to guage the difference in shadow noise between between the pairs of images being compared.

You would then compare any two images with equal ETTR exposure and examine the darkest shadows. According to DXOmark, the D3X would produce cleaner shadows in these circumstances. But how much cleaner is the question. This is where you would use the overexposed D3 shots to find out. If DXO is right, a 2/3rds stop overexposed D3 image should be almost as clean in the shadows as the D3X ETTR shot, and a 1 stop overexposed D3 image should be very marginally cleaner, in the deepest shadows.

You could then downsize each pair of images under comparison and use the same method to determine if the difference has increased to somewhere betwee 1 & 1/3rd and 1 & 2/3rd stops.

We could then all sleep peacefully.