Nikon D800
Busting the Myth of Megapixels

Page 2

(Junej 2012)

     
 

Page 1 (Introduction, The test, Resolution, Chromatic aberration (CA))
Page 3 (Picture Control, ADL and HDR, Vignetting control, Distortion control)
Page 4 (Diffraction and depth of field, Autofocus, White balance (WB) and monitor)
Page 5 (Nikon D800E, What do I find problematic with the D800?, Conclusion, Post Scriptum)

 
 
Noise

I devoted most of my attention to the noise test. I chose six cameras and tested them with different ISO speed settings: D2X, D3, D3X, D4, D7000 and of course the D800. I prepared a standard "studio scene", where a neutral gray cardboard served as background. I used the AFS Micro-Nikkor 60 mm f/2.8 lens and aperture f/8. I tested the entire range of ISO values for each camera, including the "H" (high) setting, but not including the "L" (low) setting. I focused on a bank note (but the bank note is partially masked, since some software applications like Adobe Photoshop refuse to open photos where bank notes are explicitly shown!) and I intentionally placed the box of the GoPro camera outside of the focusing range.

I chose exactly the same image settings in all cameras - (color space: AdobeRGB, color saturation: "Neutral", sharpness one step higher than default, white balance: "Preset"). Only the eight years old Nikon D2X has different (non-standard) parameters for image settings, but I attempted to set it as close as possible to the other cameras. Anyhow, the D2X sticks out of the average because of his obsolescence, so I did not make it a grave injustice with this. 

The noise reduction setting ("high ISO noise reduction") was set to "normal" in the first series of tests and to "off" in the second. In reality cameras adapt the noise reduction according to their ISO capabilities by themselves. For example this means that a "normal" setting while using ISO 100 or 200 actually means "off" with the vast majority of cameras, whereas the "off" setting while using the highest ISO settings with all cameras means "low". When using medium ISO values the noise reduction settings vary from camera to camera and I decided not to consider this nuances in my test. The actual settings can be found in the EXIF files of all the images. Generally speaking, in the years of using Nikon digital cameras I came to the conclusion that the default factory settings are perfectly fitting for most photographic situations, except the sharpness level which is slightly too conservative in my opinion.

By raising the ISO speed and thus increasing the amount of noise, we lose details in the photo. That's why cameras with a high resolution (D800 and D3X) have an advantage. To diminish this, I shot the test scene with each camera from a different distance, so that 10 pixels in the photo represents 1 millimeter in reality. Thus I photographed a 736mm broad scene with the D800, 605mm with the D3X, 493mm with the D4 and so on. Using this method I achieved equal size of details on the pixel level, which enabled me a more objective noise comparison.

 
 

It's common knowledge that image sensors with a high resolution and small pixels have worse ISO characteristics than low-resolution sensors with larger pixels, when comparing the same generation of sensors. When less light falls on smaller pixels, causes weaker electric signals, which show more noise when amplified. All this is happening on a pixel level. However, if two image sensors are of the same size (like the D800 and D4) they actually gather the same amount of light, except that in the case of the D800 it's being fractured into smaller units than in the D4. If we print an equally large picture from both cameras, they should both have a very similar quality at the same ISO setting - at least in theory.

The standard size of a printed picture from the D800 without interpolation at 300 dpi is 62.3 x 41.6 cm and from the D4 it's 41.7 x 27.8 cm. If we printed pictures in the size of 41.7 x 27.8 cm from both cameras, the picture from the D800 would be reduced in size and consequently also the noise would be reduced – the picture's quality would become similar to that of the D4. However, if we printed both pictures in a large format like 62.3 x 41.6 cm, the picture from the D800 would show the entire noise, while the picture from the D4 would be enlarged, which would mean a loss in sharpness. We see that both pictures would suffer a loss of quality, but each one in a different way.

In practice we can reduce the size of JPG photos from the D800 already in the camera (as we can in other cameras, but with the D800 this feature is more important due to its high resolution). Picture size can be set to "L" (large) – default setting, "M" (medium) or "S" (small). When shooting with the D800 in FX format "M" means 5520 x 3680 pixels (or 20.3 MP or a 46.7 x 31.2 cm when printed) and "S" means 3680 x 2456 pixels (or 9,0 MP or 31.2 x 20.8 cm when printed, which is still enough for a magazine cover!). In order to see D800's results at high ISO settings when we don't need the maximum resolution, I shot the test scene with the lower resolution settings "M" and "S" as well. As already mentioned before I compensated the shooting distance in order to obtain equal crops on a pixel level. While shooting in picture size "M" the scene was 552mm broad and in the "S" setting it was 368mm broad.

First let's take a look at the results with noise reduction set to "normal", which is the most commonly used setting.
 
 
 
 
 
 
 
 
 
 
 

When using the lowest ISO speeds (100 and 200) there was practically no difference between the cameras, even when compared on a high-quality monitor. But already a quick glance at the test photos with slightly higher ISO speeds shows, that one camera (D2X) sticks out of the average, negatively of course. The difference starts to show already at ISO 400 and becomes more apparent at ISO 800, at ISO 1600 and 3200 it becomes seriously disturbing. Henceforth I shall not compare the D2X anymore with other cameras, since any further comparisons are meaningless.

At ISO 800 the D3X starts to slightly loose quality, while changes in quality among the other cameras are practically undetectable. At ISO 1600 the noise on D3X's pictures is quite obvious and we start noticing noise in all other cameras as well. At ISO 3200 the D3X has the strongest noise again, followed by the D7000 and D800, which are almost equal, the D3 is better and finally the best results are with the D4. Still, the differences between cameras are minimal. ISO 6400 is the highest setting for the D3X, even though I don't find its pictures at this setting bad at all and I believe that it could easily reach one level higher. At this setting all cameras show noise, but it's least present with the D4, followed by the D3 and close behind it the D800 and D7000. At ISO 12.800 the D4 clearly leads the way. An undesired chromatic noise appears in the darker parts (black turns towards blue hues) in D7000's photos. Both cameras, the D7000 and the D800 start losing details on the bank note, while they remain relatively clear with the D3. The D800 can afford somewhat stronger noise reduction algorithms due to its high resolution. At ISO 25,600 only the D4 manages to obtain a picture without the undesired chromatic noise, while all other three cameras exhibit bluish tones in dark areas. Details in the bank note are quite blurred already, a decrease in sharpness can be noticed even in the D4. Only the D4 enables higher ISO values. Even though this is not a test of the D4, I cannot but notice that the last two ISO settings (102.400 and 204.800) are completely useless. Interestingly the ISO 51.200 setting causes less chromatic noise than ISO 25.600 in the D3, however according to my opinion it contains less details in the bank note.

I compared photographs taken by the cameras with noise reduction set to "off" at ISO speeds between 800 and 25,600. I believe the differences should be even more obvious here.
 
 
 
 
 
 
 

At ISO 800 the D2X already displays strong noise, with the D3X it's noticeable and even with the D7000 and the D800 we can notice it slightly. The D3 and the D4 exhibit practically no noise. At ISO 1600 we can detect noise also with the D3 and barely noticeable even with the D4. Photos made with the D7000 and D800 show a very slight chromatic noise. Even though the noise with the D3X is stronger, it's not chromatic. At ISO 3200 the D4 clearly gives best results, yet the D3 is close at his heels. Photos made with the D3X show the strongest noise (except those made with the D2X, which was already excluded from the commentary) but the noise is not chromatic and reminds us on a film grain. At ISO 6400 the noise in D3X photos is fairly strong, yet still not chromatic. We see a drastic increase of D3's noise, while the D7000 and D800 clearly show chromatic noise. The D4's lead in quality has just increased. At ISO 12,800 the D3 exhibits first signs of chromatic noise, while the chromatic noise in photos made with the D7000 and D800 becomes very disturbing. A high degree of noise can be noticed in photos made with the D4 as well, however the noise is not chromatic, similar as with the D3X. At ISO 25.600 things look even worse. The D4 starts to show chromatic noise, while all other cameras have extraordinary strong and ugly chromatic noise, especially disturbing in photos made with the D800.

Generally, we can say that all the cameras are quite similar to each other, with the exception of the D2X, which is several classes worse. Obviously the best results were made with the D4, but according to my evaluation it's only 1 EV ("exposure value") better than the runner-up - the D3. The D7000 and D800 have practically identical ISO characteristics. This makes perfect sense, since their pixel size is also almost exactly the same. Over the thumb I'd say that they lag behind the D3 for about 0.5 EV. The D3X pleasantly surprised me, considering its humble upper limit is set at ISO 6400 (all other camera's limits are set to ISO 25,600, except the D4), I expected its ISO characteristics to be worse. In my opinion it lags behind the D700 and D800 about 0.5 EV or 1 EV behind D3. Of course these are just my subjective perceptions. On the other hand I believe that the highest two ISO settings of the D4 are fully useless. Probably it's a matter of company policy since the competition offers similarly meaningless high settings, which are most likely just as useless.

The upper evaluation was based on the pixel level. I've photographed the test scene from different distances, so that the objects on the pictures would be of equal size, even though they were recorded with different resolutions. Had I photographed the test scene from the same distance with all cameras, the D800 would have a great advantage, since it would capture the objects with a greater amount of pixels and would thus despite the stronger noise reduction algorithms capture more details than other cameras. Similar goes for the D3X.

As I've already mentioned in the introduction, picture size reduction theoretically reduces the amount of noise. Therefore, I've tried to compare different photos made with the D800 (in size L, M and S straight from the camera) shot at a high ISO speeds (1600 to 25,600) with photos made with the D3 and D4 in full resolution.
 
 
 
 
 
 

At the highest ISO speeds where differences are greater, I'd say that the 20 MP "medium" picture from the D800 is pretty equal or even better than the 12 MP photo taken with the D3, while the 9 MP "small" photo from the D800 is almost equal or considering the noise even slightly better than the 16 MP photo taken with the D4! However the treacherous color deviation from black to blue hues, which starts to show at ISO 12800 and becomes very apparent at ISO 25600, reduces the quality of the reduced photograph from the D800. The photo taken with the D800 at size "small" captured great details of the bank-note at all ISO settings, which are considerably better and sharper than in the photos made with the D3 and D4!

In practice we usually shoot in the highest resolution and reduce the pictures later in the computer to a desired size. Precisely for this reason, I carried out an additional test, despite the fact that I expected to find similar results. I shot the test scene with the D800 again in full resolution from exactly the same distance as I previously shot with the D3 and D4. I then reduced the size of the photos in Photoshop, so that the leading side measured 4256 pixels (same size as a D3 photo) and 4928 pixels respectively (same size as a D4 photo). This comparison seems most fair to me, since the resolutions are identical and so is the shooting distance, which means that all details in the photos are of the same size.
 
 
 
 
 
 

In this mode of comparison I can say without a doubt that the D800 clearly leads before the D3 at least up to ISO 12.800. At this setting a slight bluish tinge starts to show in the dark parts of the D800 photo, while at the ISO 25.600 setting the D800 is leading again, since the blue colored chromatic noise appears in a large scale in the D3 photo as well.

When reducing the size of photos taken with the D800 to 16 MP and comparing them to photos taken with the D4, we can see that there is practically no difference up till ISO 6400. At ISO 12.800 and 25.600 a blue chromatic noise appears in the dark parts of the D800 photo, so the D4 scored a convincing victory in this field.

Considering all results I graded the cameras (subjectively) according to their ISO characteristics (on a pixel level). The Nikon D2X gets one point and then I add one point for each EV step (i.e. a camera that shows the same noise at ISO 800 that the D2X shows at ISO 400, would get one extra point). According to this grading scale I compiled the following chart:

D2X     1
D3X     3
D7000  3,5
D800    3,5
D3        4
D4        5

I need to emphasize again that I'm talking about noise on the pixel level. If we were to print photos from all cameras in the same size, cameras with a higher resolution would have a great advantage. That way the D800 would easily pass the D3 in quality and come very close to the D4!

Considering the noise level of the D800 I reckon, that shooting up to ISO 800 at full resolution is possible without any loss of quality. But when smaller picture sizes than 36 MP suffice the D800 can be fully used up to ISO 6,400. Only at the ISO 12,800 setting a degradation of the reduced picture due to chromatic noise (black starts to turn to bluish tones) can be spotted, therefore I recommend to avoid this setting and higher.
 
 
Page 1 (Introduction, The test, Resolution, Chromatic aberration (CA))
Page 3 (Picture Control, ADL and HDR, Vignetting control, Distortion control)
Page 4 (Diffraction and depth of field, Autofocus, White balance (WB) and monitor)
Page 5 (Nikon D800E, What do I find problematic with the D800?, Conclusion, Post Scriptum)