DxO marks for Nikon D3X are available

[ErikKaffehr]

Hi,

DxO marks for the Nikon D3X are available. Results are quite impressive, especially dynamic range.

Essentially confirms the initial findings of the gentlemen Leping and Chalmers.

http://www.dxomark.com/index.php/eng/Image...abase/Nikon/D3X


Best regards
Erik Kaffehr (in happy possession of a Sony Alpha 900)

[Ray]

Hi,

DxO marks for the Nikon D3X are available. Results are quite impressive, especially dynamic range.

Essentially confirms the initial findings of the gentlemen Leping and Chalmers.

http://www.dxomark.com/index.php/eng/Image...abase/Nikon/D3X


Best regards
Erik Kaffehr (in happy possession of a Sony Alpha 900)

Hi Erik,
That's my impression also. Many people in this forum like to quibble about small fractions of a stop (or EV) of DR. For example, the dynamic range of the A900 at its best (at ISO 100) is 1/3rd of a stop better than that of the 5D2. Big deal! The Dx3 has one whole stop and 1/3rd more dynamic range than the A900 and 1 & 2/3rds stops more than the 5D2, at ISO 100.

I wouldn't be surprised if Michael were to change his mind about buying this camera   .

[Panopeeper]

The Dx3 has one whole stop and 1/3rd more dynamic range than the A900 and 1 & 2/3rds stops more than the 5D2, at ISO 100.

This is unmitigated rubbish. My measurements of the raw data indicate about 0.5 EV advantage of the D3X @ ISO 100, based on the raw files published by Imaging Resources (medium quality for measurement).

I wonder if DxO pubishes the raw files they based their measurements on.

[Josh-H]

wouldn't be surprised if Michael were to change his mind about buying this camera

I know this was lighthearted... But..

I doubt it.. a seasoned veteran photographer like Michael [casting no dispersions on age here!   ] knows full well that there is a lot more to a camera than a published 'score'.

[Ray]

I wonder if DxO pubishes the raw files they based their measurements on.

I was under the impression that DXO results are based on the un-demosaiced files. They are bypassing the different characteristics and performances of individual converters which have caused so much argument and diversion in the past. DXO results therefore relate to what the sensor is capable of, rather than what a particular converter is capable of.

[Panopeeper]

I was under the impression that DXO results are based on the un-demosaiced files. They are bypassing the different characteristics and performances of individual converters which have caused so much argument and diversion in the past. DXO results therefore relate to what the sensor is capable of, rather than what a particular converter is capable of.

Do you see any contradiction? Do you think they can do this without raw files?

I am doing just what you described above, and if I had the raw files they used, I could make the measurements myself. Based on the raw files posted by Imaging Resources, the "1 2/3 stop" advantage of the D3X in DR is the product of some vivid imagination.

[Ray]

I know this was lighthearted... But..

I doubt it.. a seasoned veteran photographer like Michael [casting no dispersions on age here!   ] knows full well that there is a lot more to a camera than a published 'score'.

True! But of course this is not merely a published score but the result of a thorough analysis of the capability of the D3X sensor from a company that Michael appears to have a lot of respect for.

[Leping]

Hi Erik,

No I did not post any findings about D3x, since I don't even own one today -- I merely tried to raise the awareness of the initial reports from the actual D3x owners I was aware, since I thought many would be in the same mindset: weather it worths to bleed the $8000 to get the new body, while these first hand findings appeared in obscure places.

I am glad that so far most of the evidences are pointing to the positive end about the D3x IQ, but I am still not fully convinced (even it happed I managed to offended many, most from misunderstandings of my points), and in the mean while are happy with my 5DII trying to figure out ways to print to 16x24 to my like -- i.e. with no obvious digital flavors or artifacts.

Hi,

DxO marks for the Nikon D3X are available. Results are quite impressive, especially dynamic range.

Essentially confirms the initial findings of the gentlemen Leping and Chalmers.

http://www.dxomark.com/index.php/eng/Image...abase/Nikon/D3X

Best regards
Erik Kaffehr (in happy possession of a Sony Alpha 900)

However, I did make the following post today in the dpreview D3/D700 forum in reaction of people quoting very different dynamic range measurements using software/system other than DxO.  I do understand that the real world is not ideal and linear, just like the micro-lenses are not all truly "gapless"...

-------------------------------------------------------------------------------------

Read the DxO DR definition (RMS noise level = luminance level) before comparing to measures using something quite different:

"3. Dynamic range

Dynamic range is defined as the ratio between the highest and lowest gray luminance a sensor can capture. However, the lowest gray luminance makes sense only if it is not drowned by noise, thus this lower boundary is defined as the gray luminance for which the SNR is larger than 1. The dynamic range is a ratio of gray luminance; it has no defined unit per se, but it can be expressed in Ev, or f-stops. "

http://www.dxomark.com/... .../eng/Technologies/Measurement-definitions/Noise

How could more pixels does not bring down DR?

DR is determined by the sensor's full well photom capacity and thermal and read noise level.

It is obvious according to the science of physics, but many would just not accept. If the micro-lens(es) are "gapless", and you devide the one pixel area into four. End results? When you binning the 4 pixels together, or properly downsample to 50% size, you get back exactly the same SNR, since the level of noise is proportional to Sqrt(N), where N is the number of photons hit the single low resolution sensor pixel area. You lose SNR by a factor of 2, or 6dB, by checking the high resolution pixels at 100%, since only 1/4 of the photon is hitting, but after bining everything comes back exactly the same (just devided into four parts and individually read and properly combined), since the total number of photons, and hense the level of flctuations are the same.

However, by going 4 pixels instead of one, you GAIN the obth ways:

1. The option to use the highre resolution, in this case doubling;

2. The option of downsizing to 50% to half the noise. If the sensors are not Bayer type, that's it, and you do not lose or gain anything. Nevertheless, if the sensors are Bayer, which introduces another 3dB true resolution loss, you actually gain here again, because of the much less demosiacing artifacts, better color definition and smoother tonor transition after downsizing, all the higher resolution brings in in the first place for Bayer sensors, even when your lenses can not resolve to the single sensor size level, with or without CA, out of focus, and out of DOF.

These has been adequately and confincely tested and confirmed in Lloyd Chamber's DAP reviews, with these free side-to-side comparison 100%/unsampled/downsampled samples:

http://diglloyd.com/diglloyd/2009-01-blog....0090109NikonD3x

In the D3x vs D3 case, the area difference is not 4 but close to 2, so that the noise level differs by 3dB in the ideal (linear) world.

Go back to the DxO site, and checkout the following conclusions under the Insights tab:

"Contrary to conventional wisdom, higher resolution actually compensates for noise:

This Insight uses currently-available DSLRs to demonstrate the technique for objectively comparing noise for cameras with different levels of resolution. Such comparisons conclusively show better results overall for high-resolution sensors, despite the increase in noise. ..."

Thanks for reading.

-------------------------------------------------------------------------------------

[Ray]

"Contrary to conventional wisdom, higher resolution actually compensates for noise:

 

This is now conventional wisdom. Only flat-earthers disagree.

[ziocan]

Differences of 1/3 of stop and more on DR, may vary from one lens to another.
In any case, images published by decent photographers so far, do not show any of this kind of advantage on the d3x compared to the a900. Au contraire, we have seen better DR, High light roll off and colors on raw files from the a900.

[Panopeeper]

Differences of 1/3 of stop and more on DR, may vary from one lens to another

Such a lens would be called a "thick color correction filter".

In any case, images published by decent photographers so far, do not show any of this kind of advantage on the d3x compared to the a900. Au contraire, we have seen better DR, High light roll off and colors on raw files from the a900

1. Published images never reveal the dynamic range of the scenery. It is the vivid imagination of some photographers, when they want to judge from a JPEG, how high the DR of the scenery was.

2. There is no "highlight roll off" in raw files, except if compressed lossily.

[Ray]

Do you see any contradiction? Do you think they can do this without raw files?

I am doing just what you described above, and if I had the raw files they used, I could make the measurements myself. Based on the raw files posted by Imaging Resources, the "1 2/3 stop" advantage of the D3X in DR is the product of some vivid imagination.

Well, clearly this matter needs more investigation then. You can't both be right   .

[ErikKaffehr]

Hi,

I guess to really measure DR you need something like a well executed photograph of a "Stouffer edge". DR in pictures says probably more about the ACR (or whatever) skills of the photographer.

I have been impressed how much shadow details that can be extracted from Konica-Minolta and Sony "raw" images. I have no experience with other cameras.

Erik

Such a lens would be called a "thick color correction filter".


1. Published images never reveal the dynamic range of the scenery. It is the vivid imagination of some photographers, when they want to judge from a JPEG, how high the DR of the scenery was.

2. There is no "highlight roll off" in raw files, except if compressed lossily.

[Panopeeper]

I guess to really measure DR you need something like a well executed photograph of a "Stouffer edge". DR in pictures says probably more about the ACR (or whatever) skills of the photographer

I have now a Stouffer transparency strip. I made suitable shots with my 40D (this is not an easy task) and spend hours by collecting several hundred measurements.

Now I start looking for fellow photographers from the Vancouver, BC area with different cameras, who would not mind a session of such boring shooting; then I can build a strong basis of objective and documented comparisons.

[Daniel Browning]

Now I start looking for fellow photographers from the Vancouver, BC area with different cameras, who would not mind a session of such boring shooting; then I can build a strong basis of objective and documented comparisons.

Vancouver? Sure! I'll be right over. Oh, wait, wrong country. I'm in "Little" Vancouver (in Washington).  Next time I visit the Real Vancouver, though, I'll be sure to have my camera in tow.

[douglasf13]

 Iliah is also saying he doesn't see near the DR from the D3x that Dxo Mark is reporting (and he owns the D3x and A900.)  Strange.

[ziocan]

Such a lens would be called a "thick color correction filter".


1. Published images never reveal the dynamic range of the scenery. It is the vivid imagination of some photographers, when they want to judge from a JPEG, how high the DR of the scenery was.

2. There is no "highlight roll off" in raw files, except if compressed lossily.

I beg to differ on the "thick color correction filter" nonsense you just wrote. With some lenses, contrast is so high that make a difference. like the canon 85mm 1.8 or most of the 300$ 50mm 1.4 from any maker. Not mentioning most of the crappola zooms that the majority of the photographers are using.

1. there have been enough raw files to judge colors and dr from all the new 20+ mp cameras. maybe you missed them.
2. what are talking about....

[ErikKaffehr]

Hi,

In real world shooting situations we have always lens flare to consider, and this of course does affect practical DR. In the sense we discuss DR here it is a sensor/signal processor thing, so optics don't really play a role.

I would not say that we have a strict definition of DR. DR is AFAIK limited by the noise floor and I have never seen any definition of acceptable noise. Regarding high light roll of the "raw" data is supposed to be linear, in the sense proportional to number of captured photons, so highlight roll of is not supposed to be possible. Would it be possible it could be as well (or better) fixed in ACR, LR, Photoshop or what we happen to use.

Gamma encoding of "raw" data is possible and I think that the Leica M8 does just that (as it's supposed to have 8bit DNG as "raw" data). It seems to be a fact that some image corrections are best done on linear data so I guess that any non linear representation of "raw" data is a bad thing.

I happen to have an A900, it's a fine camera. Nikon may achieved a better performing camera based on what is seen basically the same sensor chip, that's fine with me.

Erik

Differences of 1/3 of stop and more on DR, may vary from one lens to another.
In any case, images published by decent photographers so far, do not show any of this kind of advantage on the d3x compared to the a900. Au contraire, we have seen better DR, High light roll off and colors on raw files from the a900.

[ErikKaffehr]

Hi!

OK! This may be pure non sense, but I mention it anyway as I think it is an interesting idea.

Someone (on LL-forum) suggested that it would be possible to take a dual readout from a Sony sensor (you supposedly can do that on CMOS). The idea would be that first a readout would be done with low preamplification and than another readout with high amplification, like combining 100 and 800 ISO. That could be possible.

The Sony chip has internal 12-bit AD-converters, one for each column. A dual readout could be used to achieve 14-bit data with the on board 12-bit converters.

Erik

Hi,

In real world shooting situations we have always lens flare to consider, and this of course does affect practical DR. In the sense we discuss DR here it is a sensor/signal processor thing, so optics don't really play a role.

I would not say that we have a strict definition of DR. DR is AFAIK limited by the noise floor and I have never seen any definition of acceptable noise. Regarding high light roll of the "raw" data is supposed to be linear, in the sense proportional to number of captured photons, so highlight roll of is not supposed to be possible. Would it be possible it could be as well (or better) fixed in ACR, LR, Photoshop or what we happen to use.

Gamma encoding of "raw" data is possible and I think that the Leica M8 does just that (as it's supposed to have 8bit DNG as "raw" data). It seems to be a fact that some image corrections are best done on linear data so I guess that any non linear representation of "raw" data is a bad thing.

I happen to have an A900, it's a fine camera. Nikon may achieved a better performing camera based on what is seen basically the same sensor chip, that's fine with me.

Erik

[ziocan]

Hi,

In real world shooting situations we have always lens flare to consider, and this of course does affect practical DR. In the sense we discuss DR here it is a sensor/signal processor thing, so optics don't really play a role.

I would not say that we have a strict definition of DR. DR is AFAIK limited by the noise floor and I have never seen any definition of acceptable noise. Regarding high light roll of the "raw" data is supposed to be linear, in the sense proportional to number of captured photons, so highlight roll of is not supposed to be possible. Would it be possible it could be as well (or better) fixed in ACR, LR, Photoshop or what we happen to use.

Gamma encoding of "raw" data is possible and I think that the Leica M8 does just that (as it's supposed to have 8bit DNG as "raw" data). It seems to be a fact that some image corrections are best done on linear data so I guess that any non linear representation of "raw" data is a bad thing.

I happen to have an A900, it's a fine camera. Nikon may achieved a better performing camera based on what is seen basically the same sensor chip, that's fine with me.

Erik

My definition/perception of all the above, is related on how I see the results on the final printed photo. All the rest, is stuff that I can leave to who likes to be concerned with such matters.
I can only care how lens flare/contrast can affect DR. the way you discuss DR here in LL forum, it will not affect the images I take at the same extent my lenses will do in real life.
The same apply for highlights roll out. the way that is represented on printed paper or on a monitor, may vary from camera to camera and from lens to lens. Than that technically is not possible to have a roll out, again it does not matter to me when I still can have/give the perception of achieving it on printed paper.
What do you think really matter at the end?