Nikon D2H Reviewed. Image Quality

Nikon D2H has, pixel by pixel, the best image quality of any Nikon DSLR so far available to the public. Do not let the pixel count mislead you into thinking "small in numbers, small in quality", because the exact opposite is true. From the D2H emerges digital images with a clarity you should see to believe, provided you manage to get the exposure correct. The camera itself exposes virtually perfect when the lighting is flat or the contrast range within the frame is moderate, so under these conditions absolutely no user intervention with regards to the exposure settings is necessary. However, you do need to be a little more careful under high-contrast situations or when flash is used, because most shots then will benefit from a judicious additional increase in the exposure. Anything from +0.3 to +1 EV can be appropriate.

To exploit fully the potential of D2H imagery, using raw (/NEF) files is the optimal approach. However, as only Nikon Capture 4 at this time of writing supports D2H NEFs, we don't tknow yet if other (third-part) software might improve the image quality even further. Nikon Capture 4 itself has a user interface marred by a slightly vulgar "XPish" quality, and although the program runs smoothly on my dual-CPU machines, it isn't a superfast performer in any way. However, image output quality is superb both from D1-series as well as D2H NEFs, so I won't complain too much (if they only could improve the interface, I might even take a liking to this program).

Setting up valid comparison tests shots for film is quite straightforward, because recorded image magnification and the order of secondary magnification after processing and enlargement are identical, provided the setup incorporates a stationary subject and a camera position which is fixed at a givewn distance to the subject itself. My test subject typically is the brick wall outside my office premises, and this part of the world is little troubled with earthquakes so the stationarity criteria should be fulfilled.

With digital cameras, things act a little different. Thus, you can easily get a setup giving identical magnification of the image projected onto the CCD/CMOS/LBCAST chip inside the camera. However, because chip size may differ so might angle of view for the end image, or chip pixel counts may differ to result in non-identical angular resolution of the pixels. A sound and basic principle underlying scientific tests is that you should be able to have full control of the variables involved and preferably change only a single variable independently of the others.

For tests with D2H set against D1H and D1X, I ran a whole gamut of procedures to ensure valid comparisons between the contenders. When image input magnification was kept constant, D1X and D2H images had about equal quality in terms of final sharpness, albeit the graininess from D2H was slightly higher because of its lower pixel count necessitated greater secondary magnification to compare with D1X. The images from D1H showed as expected less detail.

To compare pixel quality directly, I ran a separate setup in which the subject-camera distance was changed to give the pixels equal angular view. This approach is valid given that the performance of the test lens is constant across the distance range involved, which I'm very certain applies to my AFS 300 lens between 3 and 5 m. Since this lens has a focal length which gradually shortens as the lens is focused closer, and the alteration of focal length itself is undocumented from Nikon, it is frustratingly difficult to get a perfect match in pixel angular coverage, but I got within a few percent of the target anyway so that'll do for this time.

D2H

A small, approx 470 by 650 pix crop of a D2H frame shot with the AFS 300 mm f/2.8 ED-IF lens at f/6.3, 200 "ISO". The crop is from an unadjusted NEF frame and no sharpening of any kind has been applied. Full frame comprises 2634 x 1624 pixels.

The difference to the same crop from D1X, using the same lens and f-setting, but "ISO" 125, is quite pronounced. Full frame is 4018 by 2624 pixels.

The D1X crop includes (approximately) the same number of pixels as that from D2H, so angular view of each pixel should be similar.

Both images shown at100% magnification, NEF files processed in Nikon Capture 4 with no sharpening or other adjustments applied to them.

I was quite surprised by the results as depicted above, so repeated the test several times also with other subjects, but always with an identical result. Nikon D2H indeed has a much improved digital clarity to its recorded image, in comparison to its predecessors in the D1-line. This improvement simply means that Nikon's new technology, the JFET LBCAST sensor, lives up to expectation and should be regarded as a step forwards in imaging technology.

Disbelievers might scorn this test result and saying that D1X at 10.5 MPix is just "interpolated" resolution, so is no valid guideline to judge D2H quality against. Let them just believe this and let them be spared the comparison shots with D1X at 3008 x 1960 pixel output, or D1H.

Also I've already seems comments from people who fail to understand that this test addresses the possible and real quality differences between pixels in various cameras. Just pay attention to the fact that the recording pixels - with this setup - each "see" the equal area and thus, if quality of their recorded detail is similar, so should the end images be. Thus, we are ensured that visual differences between test images really are differences in the way the pixels capture image detail.

I'd address other aspects of the performance of the LBCAST chip in following chapters on noise and long exposure performance.


Vividly saturated, sharply delineated colours result from the LBCAST technology	© Bjørn Rørslett/NN