Tungsten Set
As we saw in the Canon shoot, the tungsten set causes trouble on a couple of fronts. Tungsten light, for one, is a little more of a challenge for any digital camera to balance properly. If a camera system has a weakness, it’s going to show up under tungsten light, and add to that the fact we’re shooting at ISO 800. Higher ISO effectively turns up the gain of the sensor, similar to turning up the volume on your sound system. If there’s noise, or interference, you’re going to turn that up too. A lot of the leaps that have been made in file processing in the last few years are addressing specifically the noise issues, and we’re seeing a huge boost in file quality at unbelievably high ISO ratings. The D3s, for example, in Hi-3 Extended ISO mode, gives a whopping ISO equivalent of 102,400. Typically, though, ISO processing compromises resolution, and the more a file is processed, the more you lose apparent resolution and sharpening.
Camera JPEGs
These shots are from the JPEGs out of the camera, ISO 800, tungsten light, all camera settings on Neutral.
Nikon D300s - Camera JPEG, full composition |
Nikon D300s - Camera JPEG, 100% crop |
Nikon D700 - Camera JPEG, full composition |
Nikon D700 - Camera JPEG, 100% crop |
Nikon D3s - Camera JPEG, full composition |
Nikon D3s - Camera JPEG, 100% crop |
Beyond exposure, the results show a surprisingly closer rendering of skin tones between the D3s and D300s. The saturation of the D700 that we got a hint of for the strobe shots now shows dramatically, along with a pronounced yellow shift.
Nikon Capture NX-2
The files processed in Nikon Capture NX-2 weren’t much better in overall color rendering, although the density was slightly better.
Nikon D300s - Capture NX-2, full composition |
Nikon D300s - Capture NX-2, 100% crop |
Nikon D700 - Capture NX-2, full composition |
Nikon D700 - Capture NX-2, 100% crop |
Nikon D3s - Capture NX-2, full composition |
Nikon D3s - Capture NX-2, 100% crop
|
When you look at the 100% views you’re starting to see the contrast in noise processing and resolution. The D300s file, on the left, is grainier, or noisier, but appears crisper. The D700 shot now reveals the resolution we know is there, but is less crisp. The noise processing has softened the file somewhat, but overall it’s a cleaner file than the smaller sensor. Look particularly under the eye, in the D300s shot, to see what we’re talking about.
Adobe Camera RAW 5.6
Finally, let’s look at what Adobe Camera RAW does with these files. We’re getting a much more even result with color mapping, a very similar strategy to what we saw with the strobe shots. We are, however, getting some significant variation in rendering.
Of the three, probably the D700 is the most pleasing rendering of the skin tones, and the D300s has a pronounced magenta cast. The D3s is far better in Camera RAW than the in-camera JPEG, but still seem like it’s struggling.
Again, the results are startling, enough so that we, once again, double-checked all of our tungsten files. All of the files we shot back this up. The apparent sharpness in the D300s was seemingly better at high ISO and in tungsten light than its higher-priced cousins.
We really have to ask why this is the case. Comparing the three cameras, you have one that is ultra-high speed capable, at 9fps, and ultra high-ISO capable as well. You have, at the other end of the spectrum, a fairly routine camera with an interesting balance of specs, by the current standards: a maximum ISO of 3200 and a fast shooting speed of 7 fps. In the middle you have a large sensor, a higher ISO of 6400, and a slower frame rate of 5fps.
We can only guess that some of the tradeoffs necessary to get these cameras doing what they do, a camera shooting a full frame at 9 fps and ultra high ISO, or a slower frame rate with the same sensor, required the design to take a hit on the apparent resolution. Remember, all of these cameras running 12mp, but the two that fell short are running a larger sensor, indicating larger pixels. Somewhere in all of the math, somebody made a tradeoff.
The whole story really won’t be told until we see the bench testing, but, in any case, this is just the kind of thing we wanted to know, and what the bench doesn’t always tell.
Just a confirmation: here are two more shots, same conditions, with the D700, first, and then with the D3s. The focus is on-spot, there’s no camera movement. The hairs are being resolved, the details “mush out” as soon as anything drops into the midtones: a classic example of noise processing’s hit on resolution. In fact, in the second sample you’re starting to seen the strange shadow banding that you get with a heavily processed file.
Nikon D700 |
Nikon D3s |
Banding
Let’s take a closer look at that banding. Here’s a detail of ACR processed files, but this time concentrating on shadow areas, showing some disturbing color shifting and banding.
Of the three cameras, the D3s shows the effect the least, but you’re seeing it pretty clearly in all three samples. This is, compared to the rest of the market, a pretty significant showing of shadow noise, but again, this will show up on our bench testing in a more controlled way.
Nikon D300s - Banding |
Nikon D700 - Banding |
Nikon D3s - Banding |
