Home · Register · Join Upload & Sell

Moderated by: Fred Miranda

New fredmiranda.com Mobile Site

New Feature: SMS Notification alert

New Feature: Buy & Sell Watchlist

FM Forums | Nikon Forum | Join Upload & Sell

1 2

end

Archive 2012 · DxOMark is out - Nikon D800 new king

S Dilworth
Offline
• •
Upload & Sell: Off

p.3 #1 · p.3 #1 · DxOMark is out - Nikon D800 new king

Rodolfo Paiz wrote:
What I find extremely interesting in the DxO data is how close their Sports (Low-light ISO) ratings are for the D800 (scored 2853) and the D4 (scored 2965). If the D800 does ISO 12800 with quality anywhere near what my D3s used to give me or what I could expect from the D4, I'm going to be one seriously happy camper.

The Sports score, like the other DxOMark scores, isn't very useful. Consider instead the charts themselves. They show the D4 has a roughly 2⁄3-stop advantage over the D800 in dynamic range (i.e. shadow noise, i.e. practical low-light performance at high ISO) above about ISO 1600, rising to about a stop at ISO 12800. And in addition, it has less chroma noise at a given ISO, so more colour will be preserved at high ISO in addition to the 2⁄3-stop shadow-noise advantage.

This means the micro-lenses and/or underlying silicon in the D4 sensor are substantially more efficient than the D800's, and that Nikon traded a small amount of that efficiency for stricter colour filters in the Bayer array of the D4 sensor.

At ISO 400 and below the D800 delivers extremely impressive performance, superior to the D4 in pretty much all respects. At ISO 800 the D4 already offers marginally cleaner shadows. Above ISO 800, its lead is decisive.

(The D3S and D4 are effectively the same over ISO 3200.)

The D800 is noticeably better than the D700 at high ISO (though with a balance of performance orientated towards much more chroma noise), but the gap between the D800 and D4/D3S is greater than that between the D700 and D800.

chuckbernard wrote:
My observation is that their ratings are biased too much towards the peak values instead of overall measurements.
There will be very few real world shooting conditions where you can take advantage of these peak capabilities.

Your observation is correct and I strongly agree with your conclusion. The scores portray a wildly inaccurate picture of the sensors' overall performance. The underlying data is good, but people must pay closer attention to what it really says.

Mar 24, 2012 at 06:01 AM

theSuede
Offline
• • • •
Upload & Sell: Off

p.3 #2 · p.3 #2 · DxOMark is out - Nikon D800 new king

I've run the D800, the D3s, the D3 and the D4 in static tests now.

Below ISO6400 the D800 is better than the D4 - AND the files are easy to handle.
At 6400 the D800 is just as good as the D4, but the files need some tweaking.
Above 6400 (in low WB temperatures) the D4 wins hands down. The files are a lot easier to handle. You can get ALMOST the same end-image quality out of the D800, but it takes a lot more work.

In daylight WB, the D800 hangs on up to ISO12800, easily. But the blue channel is just to weak to cope with incandescent light and ISO6400+

More work for less results? No deal. If you're going to shoot at 6400+ as a rule more than as an exception, the D800 isn't really for you. It does work surprisingly well though, I have no complaints about the image quality per se, more about the amount of extra work you have to put in to get it. At ISO6400+, your resolution is purely noise-limited anyway.
(disclaimer - with the raw-converters we use today. They're not very optimized for large resolutions / oversampling)

Mar 24, 2012 at 09:36 AM

theSuede
Offline
• • • •
Upload & Sell: Off

p.3 #3 · p.3 #3 · DxOMark is out - Nikon D800 new king

The pure scale difference makes metrics like "color resolution" and "color bit depth" very hard to compare. The D800 is just simply BETTER for color resolution and tonality up to mid-high ISO. It's a lot easier to pull out low-contrast detail with the D800 than with the D4. This is purely resolution dependent, even though they measure the same.
One must not forget that the DxO figures are raw ADU measurements, that interact very differently with a raw-converter depending on the resolution.

Mar 24, 2012 at 09:41 AM

snapsy
Online
• • • • • •
Upload & Sell: On

p.3 #4 · p.3 #4 · DxOMark is out - Nikon D800 new king

Suede, thanks for your observations. What do you make of the D800's CFA? The DxOMark color channel matrices imply a bit less selectivity than the D3.

Mar 24, 2012 at 09:53 AM

theSuede
Offline
• • • •
Upload & Sell: Off

p.3 #5 · p.3 #5 · DxOMark is out - Nikon D800 new king

Yes, it has. I didn't do a color comparison setup with the D3, since the D3s, D4 and the D800 were the main subjects of the testing. The D800 is very, very similar to the D7000, but with a slightly better base WB ratio.
D4 and D3s are also quite similar, but not identical.

This is what their "color" looks like if you take raw ADU values and just apply an sRGB profile on top of that.

Mar 24, 2012 at 10:53 AM

theSuede
Offline
• • • •
Upload & Sell: Off

p.3 #6 · p.3 #6 · DxOMark is out - Nikon D800 new king

In the D4 the reds are brighter/more saturated, and the blues are slightly less "contaminated" by the reds and greens. This necessitates less saturation increase in the raw conversion, and hence also less chroma noise in the developed image.

It's very hard to find any significant difference between the D800 and the D7000 except for the white balance ratios. All cameras were linearized with the gray patches as reference.

Mar 24, 2012 at 10:57 AM

snapsy
Online
• • • • • •
Upload & Sell: On

p.3 #7 · p.3 #7 · DxOMark is out - Nikon D800 new king

theSuede wrote:
In the D4 the reds are brighter/more saturated, and the blues are slightly less "contaminated" by the reds and greens. This necessitates less saturation increase in the raw conversion, and hence also less chroma noise in the developed image.

It's very hard to find any significant difference between the D800 and the D7000 except for the white balance ratios. All cameras were linearized with the gray patches as reference.

Thanks. In my previous discussions with Iliah about the D7000's colors it was my opinion that the sensor's noise is so low that one can boost the channels in post to match the saturation of a denser CFA while still producing good results. Then the discussion turned to cases where one can't precisely reproduce the hues of a denser CFA and also of metamerism failures, although I haven't seen a good demonstration of these two cases in practice. What's your view?

Mar 24, 2012 at 11:11 AM

lou f
Offline
• • • • •
Upload & Sell: Off

p.3 #8 · p.3 #8 · DxOMark is out - Nikon D800 new king

nice, looks like the d800 needs a tad more contrast...

Mar 24, 2012 at 11:33 AM

theSuede
Offline
• • • •
Upload & Sell: Off

p.3 #9 · p.3 #9 · DxOMark is out - Nikon D800 new king

At the levels we're discussing this, it's mostly academic. A metameric failure can affect either hue, saturation or brightness - or any combination of the three.

It actually makes more difference how you build the camera color profile, and how the color profile affects the noise reduction algorithms in the raw converter. A more dense (more peaky, more discriminating) color filter array isn't necessarily good in all situations. If you look at medium format formulations, they're typically just CRAP if you use them in fluorescent lights. Totally impossible to correct.

You have to weigh color noise against color accuracy against color discrimination. There is no "correct" solution.
Trying to mimic the human eye exactly is quite useless in a real camera. I know a few multispectral solutions that can do this, but they're only useful in very good light at base ISO. The M and L cones in the eye are really to tightly spaced in wavelength sensitivity to give good color discrimination, but your brain makes up for this with some really heavy processing over both angle and time (humans don't see in still images, we see "shape in continuous video"). If we tried to mimic that in a camera the color noise already at ISO800 would be horrible, and affect the color accuracy way, way more than a small amount of metamerical failures.

As long as you can go from camera raw ADU to standardized XYZ in a reasonably linear way, and the filters are good natured (no double peaks, smooth and few - preferably only three - derivate nulls on the wavelength sensitivity) you can get color that's "good enough".

The black magic is in finding a filter formulation that works well and easy with the type of calibration that you use. Sony had a good thing going with the A900/850, that seemed to hit a sweetspot regarding this, even though they weren't any measurably better than many other examples.

Canon has chosen a more "human-like" filtration, that gives them their trademark magenta/green color noise, and good resistance to brightness variations in skintones in difficult lighting. This does sacrifice color accuracy and hue discrimination though. Canons are notoriously hard to get "just right" in the more saturated yellow-orange part of the spectrum.

Mar 24, 2012 at 11:54 AM

snapsy
Online
• • • • • •
Upload & Sell: On

p.3 #10 · p.3 #10 · DxOMark is out - Nikon D800 new king

Thanks again Suede. Yet another one of your posts I've bookmarked for future reference. I promise, no more questions for a few days

Mar 24, 2012 at 12:21 PM

nugeny
Offline
• • • • •
Upload & Sell: On

p.3 #11 · p.3 #11 · DxOMark is out - Nikon D800 new king

What about D800e, that is supposed even better for landscapes. Dox score 105?

Mar 24, 2012 at 02:35 PM

SloPhoto
Offline
• • • •
Upload & Sell: Off

p.3 #12 · p.3 #12 · DxOMark is out - Nikon D800 new king

nugeny wrote:
What about D800e, that is supposed even better for landscapes. Dox score 105?

Given the way they test, I would wager that they will have near identical scores.

Mar 24, 2012 at 02:57 PM

aztwang1
Offline
• • •
Upload & Sell: Off

p.3 #13 · p.3 #13 · DxOMark is out - Nikon D800 new king

Mar 24, 2012 at 04:01 PM

mshi
Offline
• • • •
[X]

p.3 #14 · p.3 #14 · DxOMark is out - Nikon D800 new king

Thom Hogan has predicted D800E will be constantly out of stock at major places. I guess he is probably right.

Mar 24, 2012 at 05:32 PM

1 2 3 end
FM Forums \| Nikon Forum \| Join Upload & Sell 1 2 3 end

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.