Mother of God - D800 scores 95 DxOMark

Status
Not open for further replies.
Yes the intangibles and subjectives win all at the end of the day, and will vary from individual to individual.

We can argue though about the tangibles and objectives, so we do.

What shouldn't happen is a sacrifice of better subjective intangibles for better objective tangibles.

I am fairly pleased with the 5D3 photos but I can't push shadows in post like I would be able to on the D800. That really is a game changing capability I envy. I don't know which I would prefer subjectively, I would have to do your kind of side by side test.

I think Canon may have the subjective edge in glass; the only lens I envy from the Nikon side is the 14-24. And glass is what lasts and retains value. There is so much EF glass out there the mount's future is secure. Canon will have a sensor I can push soon enough.

loveboxer said:
I have only read a few of the posts in this thread. This is what I have to share. My partner and I have a photo studio in L.A. and we have had working relations with both Nikon and Canon. I shot for almost 20 years exclusively on Nikon and my partner on a Canon for almost 10. We have been shooting for the last 3 years on a 5D2. When the D800 and 5D3 was announced a few months back, we scoured the net for every review and user feedback there is, almost nightly.

We literally spend a few hours every day talking about this subject because we know that the camera we choose will be what we live with for the next three years and perhaps longer because it is a pain to switch. The feedback from the net about a D800 including DXO had my partner switch his entire system to Nikon, to two D700's and Nikons top glass, to buy him time until he could get his hands on D800's.

I could not wait to switch, being a Nikon fan boy but i decided to be prudent and just wait until after I did a side by side test of both cameras, D800 to 5D3.

Finally, last week we found two places that had both cameras available for us to shoot with. My partner was so excited because he was well prepared with all his beautiful Nikon glass and I was so excited just because I was used to my Nikon ergonomics. I know it sounds funny but I just never settled into the way the Canon feels in my hand and I often have to ask him to remind me how to adjust settings because they are not yet intuitive to me with Canon. I can pick up a Nikon and dial it in with my eye's closed.

Well, if you are wondering why the hell I am telling the internet all this in a Canon forum??? Well, I thought you Canon folk would like to know this. My partner and I live and support our families from photography. We are good at what we do. We took a D800 and a 5D3, with their best equivalent glass and shot them side by side, with the same settings. We did this for three days at two separate locations that had different cameras, just to make sure, there was no problem with any of the cameras influencing the outcome or the ambiance of the space affecting the color balance.

Moral of the story..... My partner has already begun selling ALL his Nikon gear. I am not switching back to Nikon...

Let it be known that when virtually every comparative shot was put side by side, the Canon won in sharpness, color balance or just a feeling that would draw us to the Canon image. The Canon shot was picked over the Nikon almost every time even though it was the smaller image and that size seems to usually impress / influence, as it would when we would compare a D700 file to a 5D2.

Moral of the story, no internet review or forum told us what our tests revealed to us, not one... Or if they did, they got shuffled because our minds seem to selectively choose what we want to read / listen to.. :)


All the best...

L.B.
 
Upvote 0
A JPEG file is 8-bit, but that doesn't mean that the maximum scene dynamic range it can encode is only 8 stops, correct?

Hint: yes, the JPEG can show an image with scene DR greater than 8 stops :)

Is there anybody out there who believes seriously his or her poor Canon gear will prevent him or her from showing what his or her creativity is up to?

Shortcomings of systems are real problems for certain photographers/types of photography. There's no use in denying that (if we kept denying it, where would that leave progress?). Usually we try to work around it, but that doesn't mean that new technology isn't welcome. For example, shooting off-center compositions with f/1.4 or f/1.2 primes & a Canon 5D Mark II and praying that you get better than a 30% hit rate on focus for shallow DOF photography... boy was the new AF system on the 5D Mark III a breath of relief.

Interesting post. Thanks for sharing. I was under the impression the Nikon would render a sharper scene, but one that required color correction in post.

It will. That was just the opinion of one poster. Maybe he was shooting JPEG & had the sharpness setting turned up higher than on the Nikon. Certainly not unheard of... this reviewer in the link below made such a mistake & concluded that the Canon image was sharper, but then also provided the full-size JPEGs for comparison which, if you compare closely, show significant more detail in the Nikon image:

http://www.ronmartblog.com/2012/05/comparision-nikon-d800-vs-canon-5d-mark.html

It can be demonstrably shown that the Nikon D800 will always render more detail than the Canon 5D Mark III b/c of its higher pixel density, barring lens limitations.

Low noise at high ISO is more relevant. At higher ISOs, if you believe all the numbers, the 5D MkIII actually has the greater DR anyway.

OK, at ISO 12,800 & above, sure. Which is why I mentioned that the D800 should've offered an 'ISO-less' operation mode. Or, for now, you could just use a different philosophy when shooting w/ the D800: Instead of ISO 12,800, use ISO 3200 & underexpose your image by 2 stops (i.e. use the same shutter speed/aperture as you would on your 5D Mark III at ISO 12,800, but just lower your ISO setting to 3200, or dial in -2 EC). Now your D800 image will have more DR & likely be cleaner than the 5D Mark III image at ISO 12,800.
 
Upvote 0
Neuro has effectively destroyed the credibility of the DxO sensor marks.

Do you mind pointing us to where he did this? And, in the face of all the evidence of the ability to push D800 shadows stops & stops above 5D Mark III shadows, are you really disinclined to believe DxO's numbers that quantify exactly what is being observed & reported all over the internets?

I understand it's less of an issue for you Brian, with your enviable lack-of-banding shadows from your 1DsIII :) That really was a clean dark frame you sent me!

the only lens I envy from the Nikon side is the 14-24

Oh absolutely. Which is why I gave up with Canon ultra-wide zooms & got myself the 14-24 & a Novoflex adapter (costly, but wonderful). I couldn't be happier with edge-to-edge sharpness by f/5.6. Even f/2.8 on the 14-24 has sharper edges than f/11 on any 16-35 or 17-40 I've used.
 
Upvote 0
L

lonebear

Guest
Ellen Schmidtee said:
sarangiman said:
The D800 has 14 bits A/D converters, so how could it have more than 14 bits of dynamic range?

A JPEG file is 8-bit, but that doesn't mean that the maximum scene dynamic range it can encode is only 8 stops, correct?

I'm not so sure.


Think about printing greyscale images with only pure black ink. So you basically have only two colors pure white and pure black, one stop DR only? And it is possible to print a fine greyscale image if the resolution is fine enough. The technique is some kind old.
 
Upvote 0
Think about printing greyscale images with only pure black ink. So you basically have only two colors pure white and pure black, one stop DR only?

Sorry, but that is not at all a valid analogy. You have many shades of grey due to the ability to overlay black dots to get darker areas (when viewed from a distance). B&W film, for example, isn't binary in nature... the size & density of silver deposites within a given area of film determine the dynamic range & tonality for that area... this quickly becomes a very complicated analysis b/c those numbers change based on the size of the area you're sampling. But now we're way OT.

The point is that you can still encode 14 stops of scene data with, say, an 8-bit ADC. Your tonality may suffer though; whether or not that is of any practical importance is still debated. Ideally, though, you want a high enough bit depth ADC such that the read noise of the sensor is not smaller than the quantization step (1 ADU). This ensures an accurate representation of the original data off the sensor. It could be argued that Nikon's D7000 & D800 actually undersample noise (noise in ADU is less than 1 for a dark frame, compared to 6ADU for Canon's 5D Mark III which oversamples noise) & so may benefit from a higher bit depth ADC... whether that would make any real world difference... I can't say at this point.
 
Upvote 0
B

briansquibb

Guest
sarangiman said:
Neuro has effectively destroyed the credibility of the DxO sensor marks.

Do you mind pointing us to where he did this? And, in the face of all the evidence of the ability to push D800 shadows stops & stops above 5D Mark III shadows, are you really disinclined to believe DxO's numbers that quantify exactly what is being observed & reported all over the internets?

I understand it's less of an issue for you Brian, with your enviable lack-of-banding shadows from your 1DsIII :) That really was a clean dark frame you sent me!

I am not debating the issue of the D800 merely pointing out that the DxO sensor scorings are dubious which were explained in a link from Dilbert and taken apart by Neuro. Whilst there is this confusion/debate going on I would suggest that quoting DxO sensor figures to prove a point is fairly meaningless as we cannot be certain of their validity

You are right in that my 1DS3 gives me far better IQ than my 5D2 and also the 1D4 for low ISO. As the 5D3 is not appreciably better than the 5D2 at low iso then I can only assume the 1DS3 is better than the 5D3. It may even be getting close to the D800 at low ISO. Using the histogram in DPP it is clear that I am usually have a DR of 8-10 - which is well within the capability of the 1DS3 (and probably all other modern DSLRs) so what is the benefit of buying a camera for even more DR potential when I am not using what I have at the moment.

I also shoot to the right which means that extracting shadow detail is, for me, fairly rare.

Here is an example which I took today

1Ds3, 135@f/2, 1/60 (yep, no IS needed), iso 100

DR = 7
 

Attachments

  • IMG_8452x.JPG
    IMG_8452x.JPG
    82 KB · Views: 582
Upvote 0
Whilst there is this confusion/debate going on I would suggest that quoting DxO sensor figures to prove a point is fairly meaningless as we cannot be certain of their validity

I was curious enough to test DxO's numbers myself. My 13.2 stop wedge tests pretty much agreed with DxO's numbers for the 5D3 & D7000 within 0.5 stops if you use SNR of 1 as the lower boundary.

I then did side-by-side comparisons of a sunset which reflected my findings in the more controlled ('lab') environment above.

So as far as I'm concerned, DxO's dynamic range numbers are absolutely valid. But I understand that you have no reason to believe my tests until I actually put them up on a blog post & make the RAW files available, etc. I intend to do so I just keep getting swamped with work...
 
Upvote 0
B

briansquibb

Guest
sarangiman said:
Whilst there is this confusion/debate going on I would suggest that quoting DxO sensor figures to prove a point is fairly meaningless as we cannot be certain of their validity

I was curious enough to test DxO's numbers myself. My 13.2 stop wedge tests pretty much agreed with DxO's numbers for the 5D3 & D7000 within 0.5 stops if you use SNR of 1 as the lower boundary.

I then did side-by-side comparisons of a sunset which reflected my findings in the more controlled ('lab') environment above.

So as far as I'm concerned, DxO's dynamic range numbers are absolutely valid. But I understand that you have no reason to believe my tests until I actually put them up on a blog post & make the RAW files available, etc. I intend to do so I just keep getting swamped with work...

I think it is important to have definitive and undisputed measures for the baseline of a discussion.

However on the high DR front I wonder how many get close to DR 10 and therefore would benefit from, say, a DR of 14 body over a DR of 13 body? If it isn't many then there is little point focussing on what is for most a theoretical benefit.
 
Upvote 0
However on the high DR front I wonder how many get close to DR 10 and therefore would benefit from, say, a DR of 14 body over a DR of 13 body? If it isn't many then there is little point focussing on what is for most a theoretical benefit.

Well, photographers are creative people, so I, for one, am looking forward to what capable photographers might do with the expanded DR of the D800. For example landscapes/cityscapes not amenable to grad ND filters or multiple exposures (which become costly when you're doing 30s exposures and the sun is setting/lighting changing constantly). Moonscapes. Or environmental portraits shooting into the sun... even aided with a flash, these could benefit from increased DR.

And let's not forget that sometimes during live event shooting you just get the exposure wrong. Or a flash doesn't fire. Or someone walks in the path of your flash. The ability to rescue such shots should not be overlooked.
 
Upvote 0
L

lonebear

Guest
sarangiman said:
Think about printing greyscale images with only pure black ink. So you basically have only two colors pure white and pure black, one stop DR only?

Sorry, but that is not at all a valid analogy. You have many shades of grey due to the ability to overlay black dots to get darker areas (when viewed from a distance). B&W film, for example, isn't binary in nature... the size & density of silver deposites within a given area of film determine the dynamic range & tonality for that area... this quickly becomes a very complicated analysis b/c those numbers change based on the size of the area you're sampling. But now we're way OT.

The point is that you can still encode 14 stops of scene data with, say, an 8-bit ADC. Your tonality may suffer though; whether or not that is of any practical importance is still debated. Ideally, though, you want a high enough bit depth ADC such that the read noise of the sensor is not smaller than the quantization step (1 ADU). This ensures an accurate representation of the original data off the sensor. It could be argued that Nikon's D7000 & D800 actually undersample noise (noise in ADU is less than 1 for a dark frame, compared to 6ADU for Canon's 5D Mark III which oversamples noise) & so may benefit from a higher bit depth ADC... whether that would make any real world difference... I can't say at this point.

Yes, it is valid (and it's not analogy but industry standard way to handle it). Beside printing industry, this technique had been used in the earliest raster displays as well. The original doubt was whether we can generate higher DR given a limited single pixel DR, and the answer is yes. The example I given is an extreme with only two colors. If given 8 bit pixel formats I should be easy to double the DR using 2X2 dithering.

When the resolution goes high beyond certain threshold, human eyes will do the blending automatically. That is how the current printing technique works. Ink printers will blend some droplets together to some degree (and thus more preferable in photo printing), but tuner printers have to rely on the remaining white spots among the black array to create different shades. Putting the print-outs under microscope (kind of pixel peeping), the uneven particles should be easy to see. Film is different as it is not only with much higher resolution (to the size of cluster of molecules) but also with various silver deposit density. However, if you can place the finest film photo under electronic microscope (that's the king of pixel peeping in current technology ;-) ), the most smooth tonal changes will reveal their uneven dithering instantly. Basically, I think you know what I mean.

Back to the original DXO DR score, I don't think DXO will take the dithering into consideration. No normal photographers will EVER sacrifice IQ that much for some occasionally needed extra DR. However, given a fixed formula of DR calculation, it may allow some unintended dithering scenario passing through, and generating a higher score. Of course, this is only one explanation among others.
 
Upvote 0
the result becomes a system overall performance.

... which is exactly what we should care about.
So the overall performance should include the entire camera, ie. noise, ISO, FPS, AF, lens, etc., not represented by DXO only.

It has been well known in the astrophotography that Nikon does NOT output real raw data, some degrees of NR is applied to their RAW.

This is a good point, something I didn't want to get into earlier b/c of the complexity of the issue. But since you brought it up... what one can do is actually fit the quantitated data from the RAW file (of the wedge shot) to the theoretical luminosity curve of the wedge. You can then find the point at which the fit deviates significantly from the model, & set that as your lower end. To standardize results, maybe set a 'maximum deviance' criterion. From this fit vs. model you can also detect 'shadow crushing'... that is, if data is significantly clipped on the lower end (as some suspect is the case for Nikon RAWs), the quantitated data will deviate from the model quicker (instead of continuing to be linear on a log scale, the quantitated data will level out). I've done these measurements for a D7000 vs. my 5D II & 5D III; both cameras show this 'toe' on the darker end, & even if you pick the lower acceptable end based on this toe, the D7000 is still ~2 stops better than any of the 5D series bodies.
This hokey stick type of curve looking like a check mark will tell you what the minimal DN vs Electron of the sensor is, it is again at least the overall of the QE+Opamp gain+ADC gain+processor algorithm. The processor can do a magic trick here to minimize the noise. I do agree, by inspecting Nikon vs Canon photo, that Nikon is better than Canon on the algorithm of this, they used another method. However, this does not mean the two sensor is of that much difference. DXO score does not indicate the sensor performance, as they claims, and this is my point. I think many of us are mislead by DXO score. Nikon, or Canon, or any one can design their camera targeting the DXO score if they want, but with bad actual picture, refer the following notes.

Let's put another grey color filter, there is no maximum white to be represented by the sensor any more

[list type=decimal]
[*]First of all, that's an extreme case.
[*]Secondly, of course there's still maximum white... you adjust the exposure. Like I said, you take exposures right around the exposure that blows the brightest patch. An ND filter, e.g., simply does not trip up this methodology whatsoever.
[/list]
What you do is to find the max and min of the electrons (photo) the sensor takes. You method can eliminate the effect of the optics, and I hope you did NOT use camera's shutter/aperture/ISO in this test. But, the problem is explain as my direct quote from another forum dvxuser.com:
"trez 04-25-2012, 04:06 AM
Speaking about DR, people often mix two things - the ratio of the brightness and the ratio of the coded values (digital numbers). These are not interchangeable terms.
Imagine 1-bit codec. The value can only be 1 or 0. Now, this doesn't mean much, unless we know which brightness levels that "1" corresponds to.
In general case, the camera has to squeeze its DR (or part of it) somehow into the available bits. There are various ways to achieve that and none is perfect - one of the reasons we have so many gamma curves. Assume 8-bit codec - even linear gamma curve can squeeze more than 8 f-stops into them - all it needs to do is to 'compress' the brightness range, so that, f.ex. digital 255 represents the brightness of 12 f-stops above what's represented by digital 0. It's just the mapping.
The problem is that the more we compress, the worse the tonal resolution becomes in the mids (and shadows), which is of great importance, f.ex. for skin tones - there's risk of severe banding, when there are not enough distinct (coded) levels to smoothly represent skin gradation. That's why non-linear gamma curves are used - they assign more coding space (distinct values representing various brightness levels) to where it makes more sense, taking into account things like logarithmic tonal perception of human vision, anticipated viewing conditions, noise, gradability, being able to use standard displays for monitoring etc. F.ex, LOG-based gammas sacrifices easy monitoring in order to improve on other areas.
So, while 8-bit codec doesn't necessarily limit the DR it can represent, there's no point in squeezing too much DR into it, as severe banding will occur in post, if we're trying to stretch it to get back to the original scene DR. Of course, some compression is fine - we don't need too much tonal resolution in highlights so we can squeeze them - this is similar to what film does, known as "shoulder".

I just don't see how you can say the DXO score IS the sensor only properties.

I never claimed anything of the sort.
Good, we are on the same boat then.

to downsize the picture? What a joke to these downsize from nikon, first, why not you downsize both 5D3 and D800 to 800x600, I bet they will be the same. To be extreme, we can downsize any photo to 1x1 size and all picture will be same for same the scene and exposure. Second, are you buying a 36MP camera and use as a 22MP each every time?

Since this has been covered extensively in other threads, I won't belabor the point... but, in a nutshell, the point is simply that the 36MP camera offers you the advantage of more resolution if you want it, while giving you just as clean images at the resolution of the 5D Mark III.

I will say though that it would've been nice for Nikon to include mRAW functionality for lower resolution RAW images right off the camera to ease the workflow of photographers. But that has its own issues, since mRAWs require demosaicing prior to downsizing (and so you don't benefit from the freedom of choice of demosaicing algorithm, or the evolution of such algorithms in the future). So perhaps it behooves RAW converters (ACR, Aperture, etc.) to offer this option in their software as we get higher & higher megapixel cameras... but that raises other issues -- e.g. do you save the original RAW file or not?
I don't clip/reframe my photo too much, I don't need THAT much MP unless I am a spy. You finish your composition before the shutter, do you? if not, you are really not a photographer. BTW, any MP greater than 10MP is fine for me already. You may say my lens is not a tele so I have to shoot than enlarge. Well, if you can no see, you can not shoot, simple as this. Again, I am not a spy.
BUT, I do need higher ISO and FPS, such as D7000, 7D, 5D3, etc.
 
Upvote 0
sarangiman said:
Whilst there is this confusion/debate going on I would suggest that quoting DxO sensor figures to prove a point is fairly meaningless as we cannot be certain of their validity

I was curious enough to test DxO's numbers myself. My 13.2 stop wedge tests pretty much agreed with DxO's numbers for the 5D3 & D7000 within 0.5 stops if you use SNR of 1 as the lower boundary.

I then did side-by-side comparisons of a sunset which reflected my findings in the more controlled ('lab') environment above.

So as far as I'm concerned, DxO's dynamic range numbers are absolutely valid. But I understand that you have no reason to believe my tests until I actually put them up on a blog post & make the RAW files available, etc. I intend to do so I just keep getting swamped with work...
Thank you for you good work to verify DXO numbers, I have no doubt about the numbers. That is science and mathematics, and they don't lie. What is wrong is that we, human beings, make wrong assumption and wrong judgment. As I said, DXO score does not indicates the D800 is the best . So do not worry and be serious, especially don't be mislead by DXO.
You can put a 4-cylinder or a 16-cylinder engine into a Lexus and tune up to as same as the 8-cylinder. Then it is the personal taste issue.
My personal taste: I think the Nikon DR curve at shadow is more usable than Canon in the real word, compare to Canon DR curve at high light, which is their tradition.
As you said on another post, more DR or better tonality, is still in debate.
 
Upvote 0
I hope you did NOT use camera's shutter/aperture/ISO in this test

I believe I mentioned in my initial posting re: this that I do NOT use the same shutter/aperture/ISO in my test. Instead, I take a variety of exposures & select the one that is just short of clipping. Which eliminates effects of optics (within reason, of course).

So the overall performance should include the entire camera, ie. noise, ISO, FPS, AF, lens, etc., not represented by DXO only.

I agree. DXO is only addressing certain features of the camera. Important ones for some people. And they do it in a very objective manner. I was suspicious of their numbers as well until I did my own tests.

This hokey stick type of curve looking like a check mark will tell you what the minimal DN vs Electron of the sensor is, it is again at least the overall of the QE+Opamp gain+ADC gain+processor algorithm. The processor can do a magic trick here to minimize the noise. I do agree, by inspecting Nikon vs Canon photo, that Nikon is better than Canon on the algorithm of this, they used another method. However, this does not mean the two sensor is of that much difference.

Yes, it's a measure of sensor + downstream electronics + DSP. And in the end, that's what matters to the photographer, not sensor only performance. The 'hockey stick type curve', even with DSP tricks to minimize noise, allows you to select a lower end that still manages to distinguish between the darkest patches of the wedge. So you, the evaluator, can select what you deem is acceptable in the face of shadow crushing due to tricky signal processing. And yet, still, the Nikon D7000 will distinguish many more dark patches than the 5D sensor will (while SNR>1)... and this translates to my real world tests as well. So I respectfully disagree that it's just a magic algorithm in the Nikon that allows for this. Their sensor+electronics are simply cleaner and, therefore, do not much with your lower signals much.

Another way to evaluate this is to do the test of ISO-less capability. Take a proper exposure at ISO 6400, then, without changing shutter speed or aperture, shoot at ISO 3200, 1600, ... 100. Boost exposures properly in post to match ISO 6400 image (ISO 3200 + 1stop | ISO 1600 + 2stops | ISO 800 + 3stops | ISO 400 + 4stops, etc.). The 5D series shots fall apart by ISO 1600 + 2 stops compared to the ISO 6400 image. The D7000 ISO 200 + 5 stops looks virtually identical to the original ISO 6400 shot. That speaks volumes as to how clean the sensor & downstream electronics are.

And, as I said, this is easily reflected in real world shots of high dynamic range scenes.

Again, I am not a spy. BUT, I do need higher ISO and FPS, such as D7000, 7D, 5D3, etc.

That is totally valid. In fact, the awesome AF & higher FPS of the 5D Mark III is what's making me hold on to it. Once I get my D800 (if ever, ha!), I'll compare the AF & if it's just as good as the 5D III (I doubt it, b/c it lacks side cross-type AF points... but only experience over a number of shoots will tell me how much this matters to me), I may take the hit of lower FPS for better image quality.

My only point was that it was unfair to call DXO biased. I see no evidence to support this hypothesis. My only gripe about them is that they don't fully publish their protocols/methodology.
 
Upvote 0
sarangiman said:
Again, I am not a spy. BUT, I do need higher ISO and FPS, such as D7000, 7D, 5D3, etc.

That is totally valid. In fact, the awesome AF & higher FPS of the 5D Mark III is what's making me hold on to it. Once I get my D800 (if ever, ha!), I'll compare the AF & if it's just as good as the 5D III (I doubt it, b/c it lacks side cross-type AF points... but only experience over a number of shoots will tell me how much this matters to me), I may take the hit of lower FPS for better image quality.

My only point was that it was unfair to call DXO biased. I see no evidence to support this hypothesis. My only gripe about them is that they don't fully publish their protocols/methodology.
Yes! I also want to know how DXO does their test. All scientific studies have to publish how and what, not just the result. Giving a number is just commercial ad and marketing.
When you compare the AF, please notice the AF lag time also.
BTW,
1. the shutter/Aperture/ISO meant not the setting on the body b/c they might be not reliable.
2. by bias I am saying the assumption/pre-condition DXO sets up might be biased. In other words, Nikon un-intentionally take advantage of it.
3. I agree with you on the D7000 low noise performance, although I don't know if it comes from the circuit or the DSP. I doubt Canon can not do this. They just don't do as Nikon. Canon's approach seems different and hurts the result at low light which is not appreciated by me.
 
Upvote 0
Jul 21, 2010
31,088
12,851
sarangiman said:
So as far as I'm concerned, DxO's dynamic range numbers are absolutely valid.

I agree. Allow me to clarify. I see no issues with DxOMark's measurements - they are useful and accurate, and they disclose their methods to generate those measurements. (FWIW, I personally prefer and use DxO Optics Pro as a RAW converter.)

The problem is the DxOMark Sensor Score. Here's what Luminous Landscape states:

Documentation about the way the final DxOMark Sensor score is computed from Dynamic Range, Color Sensitivity and Low-light ISO scores is not currently available.

That same article (linked in a thread started by dilbert) indicates there is some 'non-linearity' in the way the Score is calculated. So...a weighted average, but undisclosed weighting factors, no confirmation that those unknown factors are even consistently applied, non-linear calculations, etc., to me, that makes the DxOMark Sensor Score a meaningless and useless number (not measurement, just an irrelevant number).
 
Upvote 0
neuroanatomist said:
to me, that makes the DxOMark Sensor Score a meaningless and useless number

While not transparent, the score isn't meaningless to me if it usually reflects the iq differences that a multitude of reviews express in words sooner or later. You can see it this way, too: Google's search engine algorithm is not transparent, too, but does that make Google useless?
 
Upvote 0
B

briansquibb

Guest
Marsu42 said:
neuroanatomist said:
to me, that makes the DxOMark Sensor Score a meaningless and useless number

While not transparent, the score isn't meaningless to me if it usually reflects the iq differences that a multitude of reviews express in words sooner or later. You can see it this way, too: Google's search engine algorithm is not transparent, too, but does that make Google useless?

IQ deduced from a sensor score that reflects numerous measurements not relating to IQ?

Google's algorithm does not give a score that influences spending habits nor does it pretend to be 100% accurate.
 
Upvote 0
Ok, I get you, Neuro. Yeah I pay no attention whatsoever to DXO's 'Sensor Score'.

2. by bias I am saying the assumption/pre-condition DXO sets up might be biased. In other words, Nikon un-intentionally take advantage of it.

That's an interesting conspiracy theory and one, I must admit, I didn't discount earlier on much like yourself. But, if this were true, then:

[list type=decimal]
[*]Nikon shadows raised in post would appear blurry due to noise reduction. They do not.
[*]Whatever processing you suspect Nikon DSP may be doing can be replicated on the RAW Canon file in post. Clipping, noise reduction, you name it. Varying the clipping point (e.g. 2047 for 5D3) doesn't change the point at which the response curve on a log scale 'toes' (bottom of hockey stick), so that irrelevant IMHO. It does 'clean up' the shadows by making the signal go to 0 (black) quicker and the noise drop (standard deviation goes from 5 to 1.4 for the darkest patch as you go from 2047 to 2057 for the black clipping point; I wrote some code to quickly analyze these types of questions so I've tried a number of things but cannot replicate the cleanliness of the D7000 dark patches)... but that doesn't magically increase the SNR or change the point at which the curve toes. Meanwhile, the darkest patch on the D7000 still has SNR>1. Maybe what we could do is apply noise reduction to shadows on the Canon file to see if we can better SNR in those dark patches; however, again, I don't think that's magically going to change the point at which the curve toes (i.e., the point at which the sensor no longer accurately record the scene luminance changes). But I'll try this in my analysis just to ease such concerns.
[/list]

3. I agree with you on the D7000 low noise performance, although I don't know if it comes from the circuit or the DSP. I doubt Canon can not do this. They just don't do as Nikon. Canon's approach seems different and hurts the result at low light which is not appreciated by me.

Like I said, I doubt this to be true b/c if it were then with enough image processing you'd be able to make the Canon shadows look like the Nikon shadows. Fred Miranda's, or any of a number of other reviewers', comparison demonstrate that there's just no way in heck you could clean up the Canon shadows to get the levels of detail/cleanliness present in the D800. You can't magically create detail. If the transitions from patches 35-42 on my 13.2 stop wedge are completely lost in noise in the Canon file, you can't magically get those transitions back with noise reduction (i.e. create detail).

But I guess to convince you I could try heavy noise reduction & then quantitation to see if it helps mitigates the deviation from the model (of what patches 35-42 should be). I can almost guarantee right now it won't.
 
Upvote 0
Status
Not open for further replies.