|Nikon D2X Digital Camera Reviewed|
|by Bjørn Rørslett|
8. Taking On the Competition
Nikon has stubbornly endorsed their "DX" format as the better approach to digital SLR. They advocate the smaller-sized imager as being the optimal solution for cameras compatible with existing 35 mm cameras and lenses. Other companies, namely Canon and Kodak, have pursued other avenues and designed "full-frame" (FF) cameras. There is a wide-spread belief that FF is not only desirable, but necessary to achieve top performance for "35mm" DSLR models. With the entry of the new top model, D2X, time has finally come to make qualified statements about the wisdom of Nikon's DX-based policy. The currently regarded champion amongst the DSLRs is without question Canon 1 DS Mk.II, sporting a 24x36 mm sensor and having no less than 16.7 MPix imager. So what could be more fitting than pitching these top models against each other.
I decided to test several potentially differential features between the champion contenders. Colour rendition, high ISO noise, wide angle light fall-off, image detail and sharpness were amongst these. Please note this is not a test of the Canon camera as such, just a practical test in which the cameras competed at as equal footing as possible. I called upon my friend and fellow member of NN (Norwegian Nature Phototgrapher's Association), Kai Jensen, which is an avid Canon shooter and uses the 1 DS Mk.II as his bread-and-butter tool. We shot the test targets jointly so there shouldn't be any bias for the Canon caused by user mistakes.
Admittedly, a fair comparison of these cameras is a tall order. Fields of view are quite different of course, so are pixel pitches and pixel density. Canon has the biggest pixel pitch and the lowest density, Nikon the opposite. The 1DS Mk. II is specified for ISO equivalencies from 50 to 1600, with 3200 being "HI". The D2X ranges from 100 to 800 "ISO", with HI-1 and HI-2 settings (1600 and 3200 "ISO", respectively). The frame proportions of the cameras are slight different too, but close enough to 1:5:1 to have no significance.
To make this perfectly clear (already stated below but missed by many people), we selected to use Jpg Fine on both systems, with NO sharpening, NO tonal adjust, and NO post-processing. These files come straight off their respective cameras, and are in Adobe RGB (1998) colour space. We could have used RAW formats, but then differences in post-processing would be inevitable, because no raw converter capable of handling both systems currently exists.
|Nikon D2X||Canon 1 DS Mk. II||Score
N - C
100 ISO, both cameras set to "Flash" colour balance, Abode RGB(1998) mode, no in-camera sharpening, post-processing, or NR, using AFS 300/2.8 and 300/2.8 L lenses respectively. Images shot concurrently under slightly overcast skies (f/5.6, 1/320 sec).
D2X exhibits purer and brighter blue and yellow colours, while its red rendition is a little duller (but more correct according to the GretagMacBeth target). 1DS Mk.II shows very similar colour rendition, perhaps slightly more "off-target" than the Nikon. However, I declare this a draw between the contenders. This is the natural conclusion given that noise levels were negligible for both cameras at this ISO setting
1 - 1
Next test concerns high-"ISO" noise. A lot of people are preoccupied with such features and the current way of thinking is that the smaller "DX" imager would be more troubled with noise than does a full-frame ("FF") camera. To investigate this aspect, I deliberately put the D2X at a disadvantage by chosing to use an "ISO" setting not calibrated by the manufacturer (Nikon specifies 100-800 ISO equivalency only). For the low-ISO shots, no evident difference betwwen the contenders could be found, by the way. Both camera render images clean as a whistle at 100 "ISO".
|Nikon D2X||Canon 1 DS Mk. II||Score
N - C
|High ISO noise (3200
Click each image to get a larger view (NB: File sizes are approx. 250 KB )
Canon 1 DS Mk. II has 3200 "ISO" as its HI setting, and Nikon D2X needs HI-2 to reach the same speed equivalent. This is to say that D2X is more out of bounds than the Canon, and should be expected to be more noisy.
The image crops confirm that the Canon shows less noise than the Nikon. However, there is considerably more chrominance noise on the Canon image than with Nikon, the latter howeve displaying a more gritty and noisy luminance pattern, yet with a lot of detail kept intact. Canon has a much smoother rendition of the colour patches, they are almost "plastic" in texture, but the dividing areas between the patches lack detail compared to D2X and also has more chrominance noise.
The winner here is undoubtedly the 1 DS Mk. II. At 400-1600 ISO, not shown here, the differences are not equally obvious in favour of the Canon, but the verdict would be similar. So 1 DS Mk.II is the low-noise champion, by far.
1 - 2
One key issue addressed by Nikon in defence of their "DX" format, is the need for a normal angle of incidence of light rays to the imager surface. This would largely mitigate light fall-off (vignetting) towards the corners of the image frame. The DX-designated range of wide-angle lenses are constructed with a telecentric optical design, implying that the rays emanate nearly in parallel from the lens. Nikon is adamant that this principle of building wide-angle lenses is by necessity connected to having a smaller imager than "full frame", in order to keep the DSLR camera compatible both with DX lenses and ordinary longer focal length lenses also fitting traditional 24x36 mm cameras. If they were to build telecentric lenses for full-frame systems, then the flange to film/imager distance would have to be increased, thus leading to an entirely new camera system. Canon uses more traditionally designed wide-angle lenses for their DSLR systems, including the top-of-the-range 1 DS Mk.II. The test pictures, exhibited below, show that the Nikon approach is paying off in terms of better light distribution across the image.
|Nikon D2X||Canon 1 DS Mk. II||Score
N - C
|Wide-angle Issue (Equal Horizontal
Field of View, Equal Perspective, Different Image
Magnification, Different Pixel Density)
Nikon: 12-24 DX lens at f/4, 12 mm setting; Canon: 17-40 L Lens at f/4, 17 mm
Full-frame cameras are known to exhibit nasty light fall-off into the corners when wide-angle lenses are deployed on them. This results from light rays striking the imager at an angle, leading to vignetting in the periphery of the recorded image. Colour fringing also may be troublesome, but this issue wasn't elaborated here.
The images show very clearly that "DX" lenses do not display the fall-off found on "FF" systems. Testing at full aperture would exacerate any tendency of vignetting, by the way. When the lens is stopped down, better corner behaviour can be expected, but any colour issues would not be alleviated just by stopping down.
(the obvious colour difference between the shots was visible in the finder of the Canon camera, so might be a lens-specific property. No measure was taken to correct this. And no, there isn't a sunshade on the Canon lens !
2 - 2
The full-frame Canon, given it is placed in the same distance from the subject as the Nikon and both use lenses with similar focal length, will capture a broader field of view. This results directly from image-forming geometry. Such a setup further gives equal magnification of image detail (not equally obvious to people, but nevertheless true). Now, we have an interesting scenario. Since Nikon captures a smaller view, but has higher pixel density at the same image magnification, it should be able to give a sharper detail rendition than the opponent.
|Nikon D2X||Canon 1 DS Mk. II||Score
N - C
|*Image detail (Equal Magnification*, Equal Perspective**, Equal
Depth of Field, Different Field of View, Different Pixel Density)
Click on each image to open a detailed view of the area outlined in red (NB: File sizes are 317 KB and 190 KB)
In this setup, the contending cameras were located at the identical standpoint and both had their brand 300 mm lenses mounted. This gives equal magnification of detail. The field of view will obviously be different. Due to its higher pixel density, D2X will have more pixels within the crop than 1 DS Mk. II.
The winner here is D2X by a very comfortable margin. The clarity of image detail is simply stunning compared to the softness of the Canon version.
3 - 2
It could be argued that different focal lengths should be
used, to give an equal field of view. In that case, we
forego equal image magnification, change depth of field
to be different, but keep equal perspective
** Perspective is governed by distance camera-subject, not by field of view and hence not by focal length. There is a wide-spread misconception that perspective changes when you switch from a DX-format to a full-frame camera, but as long as the camera isn't moved, this assertion is false.
The final round concerns getting images encompassing equal field of view. I did this in the horizontal dimension for practical reasons. In order to eliminate contribution from the lens, I availed myself of the fact that Canon has a shorter register distance than Nikon, making it feasible to put the Canon on a Nikkor lens. The lens itself obviously doesn't care which body (or recording format, as long as you keep within the standard 35 mm size) it delivers light to. When the field of view is equalized, the Canon imager has the advantage of a higher pixel count and the null hypothesis would be that Canon shows the sharper image, all other factors being equal. The results did not corroborate that hypothesis, however.
|Nikon D2X||Canon 1 DS Mk. II||Score
N - C
|Image detail (Equal Horizontal Field
of View, Different Magnification, Different Depth of
Field, Different Perspective, Different Pixel Density)
Click image to get a larger view (NB: File crop sizes are around 250 KB for the Centre crop)
Both cameras use the same lens, the AFS 300 mm f/2.8 Nikkor for these shots The difference , if any, in detail sharpness within the zone of sharp focus is due to the respective imager. The tripod-mounted setup was moved to give an equal horizontal field of view. Since the Canon imager would cover the encompassed frame with 16.7 MPix vs. the 12.2 of the Nikon, you would a priori assume the Canon image was more detailled. You can examine crops of these images by clicking on them. My conclusion is that the D2X image has the upper edge, despite its fewer pixels. In my point of view, this was probably the most unexpected test result of all. It clearly shows that not all pixels are created equal and that the D2X imager has the higher acuity. It seems the respective makers follow different avenues with respect to how signal processing of the image data should be conducted. I would like to add that this experiment, in hindsight, really should be conducted on a less three-dimensional test subject, so as to minimise effects of perspective differences and depth of field.
Note added 26 February, 2005: To help interpretation, you should note the following variables have changed: perspective (since one camera was moved), image magnification (ditto) and hence depth of field. Because the identical lens was deployed on both cameras, the lens itself is of no influence. I selected a long lens and a slightly distant object in order to minimise the relative changes in perspective (and hence other related variables), but we might perhaps have arrived at a clearer result if another target with less inherent depth had been selected. As a lovely birch stood nearby, I and my Canon friend opted for that scene nevertheless. If you are to compare sharpness in these crops, please concentrate on the areas in which both images are well in the focused zone. Personally I still hold D2X as being a hair better, in particular due to the fact that Canon 1 DS Mk. II for this test has the advantage of 33 % higher pixel count (and thus in theory 16% higher pixel resolution on a planar basis). If you go for a tie instead, the final outcome would be 4-3 instead of 4-2, but if anything, this result shows the cameras are excellent in most areas, with either the one or the other being ahead in a particular field. So it all boils down to which feature(s) one would emphasize the most.
4 - 2
To many people, understandingly more familiar with photography than the intricacies of equipment testing, the test above caused a lot of concern and accusations of it not being fair (implied, to the Canon). What people craved for, was a comparison in which field of view is equal, not done the way I have outlined above, but obtained by putting different lenses on each camera so as to get the infamous "crop" factor, 1.5 in this case. Although this approach might seen intuitive, it is in reality the least satisfactory of all because you now can only get a single variable (perspective) equal, and unless one of the lenses involved is a better performer than the other, you might end up evaluating lens performance instead of the camera. However, having said that, I did this test too a couple of days later, as a result of public demand.
When we put for example a 300 mm lens on the 1DS Mk. II and a 200 mm lens on D2X, and shoot the same subject from the same stand point, we will get equal field of view and equal perspective, since shooting distance is the same. We also will have different image magnification and accordingly, different depth of field (DOF). In addition, there will be a difference in pixel density across the recorded frame, in this case an advantage of the 1 DS Mk.II due to its 16.7 MPix compared to 12.2 Mpix of D2X. In terms of file size, the ratio is 1.35:1 implying D2X needs to be scaled up 16.3% to match the Canon on a pixel-by-pixel basis. The aspect ratios of these cameras is not precisely equal, so in order to scale the images we can make them identical in pixel dimension across one dimension only. I selected the horizontal and this means the original 4288 x 2848 pixel frame of D2X is scaled to 4992 x 3316, while Canon keeps its native size of 4992 x 3320. Again, upsizing confers an advantage to the Canon.
The practical test was run using again a 300 mm f/2.8 L IS on a Canon 1 DS Mk.II, and the AFS 200 mm f/2.0 VR Nikkor on my D2X. My Canon friend and I were still on good terms, so we joined forces one more time. In order to assess the importance of the lens in use, I also shot with my venerable 1977-model Nikkor 50-300 mm f/4.5 ED lens, hardly a state-of-the-art design today, but a very good performer nevertheless. I wanted to eliminate this time any possible importance of the tripod and head factors from the previous test series, so this time both cameras (or rather, the lenses with their tripod collars) were secured to identical Burzynski heads atop similar Sachtler ENG 2 tripods. No shots of any of my tests were hand-held, of course. Both systems were shot with mirror lock-up, and we used raw format and jpgs, both with in-camera settings of no sharpening, no tonal adjust, and Adobe RGB (1998) colour space. Afterwards, we processed the raw files along our normal workflow, using Nikon Capture 4.2 for D2X and Capture One 3.6 for 1DS Mk.II, respectively. The processing was done independently on different computers, by the way, I did the D2X, and my Canon friend, the Canon files. No sharpening of the resulting TIFs was performed and the crops from the TIF files were directly stored as jpgs with maximum quality (12).
The reasons for using long lenses in this kind of test are several. Firstly, by shooting more distant subjects we reduce the importance of different DOF, and we have higher image magnification so as to bring details forward. Secondly, and most important of all, these lenses project image-forming rays nearly in parallel onto the imaging surface thus alleviating peripheral vignetting and potential issues with chromatic aberration caused by micro-lenses on the chip surface. Thirdly, the combination of 300 and 200 mm represents the correct 1.5:1 ratio we need to get similar field of view.
So, without more ado, here are the results (f/5.6, 1/200 sec, 100 ISO; f/11, 1/60 for the 50-300 which needs a bit more stopping down to attain peak sharpness). Since I for one will not consider this a sound and fair comparison (NOT because it might favour one of the contenders, but due to the non-exacting nature of this approach), I won't make a score for it either. Read on and draw your own conclusions.
AFS 200 mm f/2 VR
50-300 mm f/4.5 ED Lens
1 DS Mk. II
300 mm f/2.8 L IS Lens
|Equalizing the "Crop"
Factor (Equal Horizontal Field of View, Different
Magnification, Different Depth of Field, Different Pixel
Density*, Equal Perspective)
Click links to get a larger view (NB: File crop sizes are around 250 KB for the Centre crop, 85 KB for the Periphery crop)
*Applies to Native File Dimension, Equalized in Post-Processing by Upsizing, See Text
Centre sharpness of these images show very small differences, we are virtually at the pixel-peeping level here. The Canon combination is possibly just that little better to be declared the "best", with the 200 VR almost neck and neck to it and the old-timer 50-300 a tiny bit behind. I'm not certain these minute differences survive to the web crops, though. The colour balance could have been equalized, but I don't feel this to be important (a note here is that jpgs from the two cameras were much more like in terms of colour, so the difference here indicates different preferences made by one of us during the raw file conversion stage. Remember we did this independently).
The depth of field (DOF) is significantly different, as indeed should be the case since the 200 mm lens has less image magnification. The foreground of the test scene shows this clearly. We cannot compare image qualities here due to this fact, it wouldn't be fair. What we can do is following the plane of focus out to the edges of the image, and if the crops from this area are examined, the differences become very obvious. The 300 mm lens now operates close to the periphery of its image circle and its performance declines accordingly, while both Nikkor lenses perform much better with far more image detail. Even the 50-300 lens, with its ancient optics harking back to 1977, can resolve fine detail such as single branches and spruce cones, all of which simply do not resolve on the Canon image. On the DX format, these lenses project just the centre of their image circles and this area has the best quality. On the full-format camera, you do get a tendency of darkening of the corners simply by law of optics, and the result on the 1DS Mk.II might indicate this tendency is exacerbated in a digital system compared to film.
The "fair" test thus really isn't fair at all, it recoils in an unexpected fashion, and it does show a little appreciated weakness of the full-frame approach. True, the Canon might have at least as good performance as D2X in the centre of the frame, but is let down by poorer performance towards the peripheral areas of the image. Almost no existing lens has equal performance and equal illumination across the entire imaged circle, only specialized "macro" lenses are candidates for an exception. Note that this applies to the circle "seen" and projected by the lens, not the one recorded by the camera, which may be in case of the DX format, significantly smaller. Thus, even though the bigger file size of the Canon would be considered to give a potential for larger prints, the lower sharpness outside the image centre conflicts with this. We might end up with the paradoxical situation in which both systems can deliver prints up to the same size in practice. Of course, the imaging qualities of the lens in use and technical command of the user set the end limits, here as elsewhere. It is still the photographer who takes the picture, not the camera.
Another less intuitive difference is that you will get shallower depth of field with the FF camera, because a longer lens is employed. While the entire landscape scene shot as the test subject was in sharp focus at f/5.6 with D2X, it wasn't with the 1 DS Mk.II and you would need to stop down at least one extra stop to get equal DOF. Unfortunately, this means longer exposure and a longer lens, both contributing to increase the risk of vibration and camera shake. You could turn the argument inside out and regard shallower depth as a benefit, of courese. But again, the DX-format lens could be opened up a smidge more than one stop and again we would have equal DOF, but now the DX camera employs even faster shutter speeds again leading to less risk of camera shake. This is a hidden benefit of any smaller format. As a long-standing view-camera user I'm acutely aware of this when I switch back and forth between my 6x9 cm cameras and the 4x5" and 8x10" systems. The benefit of the smaller format is clearly constrained by the need for having sufficiently fast lenses available, so not all demands for very shallow DOF can be met with the smaller formats. The opposite, getting more DOF, always can be achieved.
To underscore the potential of the D2X further, I could add features such as higher shooting speeds, cropped mode, and probably much better long-exposure noise control. On the other hand, the 1 DS Mk.II almost certainly would outperform D2X with respect to dynamic range, but this was not tested here The Canon also has smoother rendition with high-ISO images, but shows slightly more chrominance noise than the D2X within some areas of the test targets (colour patches). You could throw in the diversity of the massive and impressive lens lines offered by the respective makers, which system has the most advanced lens features, and so on, ad nauseam. Any way you look into this, there is something to suit you in both camps.
The inevitable conclusion is that both cameras are top-flight performers, but Nikon D2X takes the lead over Canon 1 DS Mk. II in several areas of performance. Any prospective purchaser of a top quality DSLR system should detail his/her needs, and find the camera candidate which fills these in an optimal fashion. And you should carefully evaluate whether or not the full-frame approach is the one you really need.
Note added 26 February, 2005:
No part of my review has been more hotly debated than this shoot-out. Sadly, much of the debate concerns wrongly applied photographic concepts and people are mixing different issues. A surprising number of people think "perspective" has something to do with field of view, "crop" factor, or focal lengths, all of which are false assertions. I have elaborated some of the test conditions to make the set-up rationale clearer.