9. D3 against the D300: FX vs DX format

Nikon has stubbornly endorsed their "DX" format as the better approach to digital SLR for a long time. 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, D3 and its FX format, time has finally come to make qualified statements about the wisdom of Nikon's earlier DX-based policy.

I chose to make the comparison with D3 and D300, since they both are 12 MPix models. Intially I thought of using the D2X as the alternative, but since a write-up on the D300 is on my to-do list, I selected the latter model. Admittedly, a fair comparison of these cameras is a tall order. Fields of view are quite different of course, so are angular resolution due to pixel pitches and pixel density. Amongst these two, D3 has the biggest pixel pitch and the lowest pixel density, D300 the opposite. The frame ratio aspects of the cameras are slight different too, but close enough to 1:5:1 to have no significance for my assessments.

Among the claims laid down in favour of either format are finder size, sharpness, depth-of-field (DOF) considerations, "reach" of longer lenses, a "speed" advantage, vignetting behaviour, noise performance, dynamic range, pricing, and size/weight. Since a D2X/Xs (DX) is the same size and heft as the D3 (FX), we rule out the latter directly. The inherent cost of the systems as they now appear would appear to go in favour of the DX, but since Nikon's FX is only just embarking its development this situation could change in the near future. However, all things considered a bigger sensor chip will always cost more than a smaller one. In terms of noise, or rather, a better high-ISO performance, Nikon has provided the answer by equipping D3 with 6400 ISO as the highest calibrated setting plus giving boosted values up to 25600 ISO (equivalent), whilst the D2X/Xs went up to 800 ISO (3200 ISO equivalent boosted), and the D300 is rated up to 3200 ISO (6400 ISO equivalent boosted). The recent crop of DX-format cameras exemplified by the D300 has come a long way towards giving us a better, bigger, and brighter, finder, but undeniably the D3 FX has the upper hand in this department.

In terms of sharpness, pixel density has to be factored into the equation. Thus, given that the D300 has the (approx.) equal number of pixels as the D3, but packed into the half area, it is obvious that D300 potentially can record finer detail. Whether it does come out on top depends on the subject, though. I've seen examples of either camera delivering the "best" image under field testing although the tendency is for the D300 to have the upper hand. But since theory supports the notion that DX potentially can be sharper than the (current models of) FX, I accept this as a fact for now. However, in order to realise the better sharpness potential, you need to be able to focus more accurately and this requirement is not entirely compliant with the finders on these systems.

Whether you consider the cropped DX format to give more "reach" will depend entirely on how you perceive and deal with "magnification" of a framed shot. Using a 200 mm on a DX camera gives smaller magnification than a 300 mm on an FX from the same position and distance to the subject, but you get almost identical field of view. So the same 300 mm lens now deployed on the DX camera will actually increase detail magnification (to the same level as on the FX, of course), but the field of view is smaller. Many would argue this constitutes a "magnificaton", which at best is a direct abuse of terminology. What is happening is that you have effectively cropped away more around the subject. You could equally well use the FX camera and trim the image later, so as to achieve a "DX crop" by indirect means (but the pixel density will be lower since the pixel pitch remains the same).

The alleged "speed" advantage is closely associated with the depth-of-field (DOF) considerations. The reasoning goes as follows: According to current DOF models, with DX you can use a lens with focal length (DX/FX) at 1 stop larger setting and achieve the same DOF extension as on FX. Thus, using a 200 mm lens at f/2 on DX will result in the same DOF (and field of view) as a 300 mm lens at f/2.8 with the FX camera, if both are shot with the same distance to the subject. It's pretty obvious that the "advantage" depends on which camp you consider yourself being a member of. You could twist the argument and say that DX, given that the assertion above holds, will allow you shorter exposure times since you don't need to stop down to the same degree. The entire claim also hinges on the (rather dubious) assumption that the end user won't see a difference in how an FX camera is used compared to a DX system. But for now, we just take this assumption as a parameter to the sequence following.

Much of the debates that one observes flare up on the Internet forums come from people heaping upon their opponents arguments from DOF calculators. Now, "DOF" is not a physical property, it is something perceived and processed by the human mindset, and it is very much an elusive quantity as well. Most DOF models are based upon image-forming geometry, plus assumptions about the viewing conditions to which the final image is subjected. All of these parameters may be physically modelled with high precision yet this tells nothing of the concordance between what is modelled and what can be observed. Thus, typically the image won't clearly break into "sharp" and "unsharp" areas with a distinct border between them, rather there will be zones of increasing sharpness (or unsharpness if you prefer going the opposite direction). Where to draw a line and say "inside" or "outside" a DOF zone is very subjective. Digital images are essentially dimensionless until they are outputted as something tangible, like a print hanging on a wall. Thus, while we can envision the primary stage of optical magnification, namely, the size of the details projected onto the imager, it is more difficult to understand the rôle and amount of secondary magnification (the enlargment stage that only can make detail bigger, but not adding more information on its own). Added complications are the pixel density and pixel pitch, which both interact to limit spatial resolution of the captured image.

My first series of FX/DX DOF comparison were conducted in the field using fairly long lenses, such as 1200/800, 300/200, or 125/85, applied to landscapes. While one might have an indication of differences, the actual DOF for such shots is great enough to make the delineation of sharp vs unsharp zones quite difficult, not to say a little meaningless. I have posted some of these observations earlier and the interested parties can consult them here.

In the FX/DX comparisons following below, I tried to do a precisely defined setup that could clarify differences FX/DX in more detail. I availed myself of the fact that the 70-180 mm Zoom-Micro-Nikkor has its nominal and effective aperture similar along its focusing scale. This makes for elimination of potentially additional error since we can use the same lens for either format and we are certain that the apertures are compatible. I set up the test with a target oriented precisely at 45° to the lens and shot at a distance of 0.72 m with the Micro-Nikkor set to 122* mm and 180 mm for DX and FX formats, respectively.

* although the EXIF reports 122 mm, I had the exact field of view for both formats so the actual length should be very close to 120 mm. It seems the focal lengths reported by the 70-180 are not continuous but have discrete steps, so one could either get 116 mm or 122 mm, but not 120 mm although the lens was set between these two values.

First we test the basic hypothesis that the format as such doesn't influence the depth of field when the final magnifications give the same apparent size of details. I have given supporting evidence for this earlier but it doesn't hurt to repeat the test with a different experimental layout.

Does this comparison corroborate that using the smaller "DX" format in itself will impact the depth of field, a claim that so frequently is encountered? Honestly I don't think so. Within the error of the experimental setup, all crops appear to be identical in terms of their distribution of detail sharpness. What we can observe, albeit just barely, is that the lower pixel density (but identical pixel pitch) of the DX mode of the D3 renders the final image just a tad less crisp when it is brought to the same apparent size as the others.

In the case above, the primary magnification (concerning detail at the film plane) is kept similar. Now, we venture into the dire straits favoured by the "equal FOV" fans. This is to say that you deliberately put the smaller format at an disadvantage in order to make the captured field of view identical. In this process, if we choose to keep perspective equal, then picture angles will be different and so will detail magnification by a factor of the ratio of FX/DX diagonal (1.5 in this case). You could offset this drawback by being able to open up 1 stop more so as to give a shorter exposure time. All again per the parameter set I outlined earlier. Now, let's check how reality fares against the predictions.

What can be learned from the panel above is the stark real-life existence and behaviour of the "DOF" issue. There is no such thing as "real" phenomena which can be measured and plugged into a nice, quantitative model. Essentially, perceived DOF is a qualitative variate. We can compare the manifestations of it in a given picture under a set of viewing conditions and say "more" or "less, but not "twice as much" or "half the amount". This is the data to debate and not the output from a DOF model.

The visual differences obviously are small, almost down to the nit-picking level. So even a full stop variation of the aperture setting gives negligible changes in the perceived DOF. Feel free to quantify the "speed" advantage of the DX format here. Is it one stop as claimed? Or is the better question rather: does the format matter? Or are we back to an apples vs oranges discussion again? I leave the decision to the reader. If you perceive the matter differently, just fine with me, we're talking about subjective impressions here.

So, what about the FX vs DX statement presented in the opening of this page. Nikon has ruled in favour of supporting both formats for the foreseeable future. As of now, if you wish to have the smaller, most portable camera, with an image quality capable of surpassing the D3 in resolving power, well, then the D300 is the answer. If you need the utmost ruggedness, the extreme high-ISO performance, or the ultimate firing rates, well, then the D3 is the better option. Or if you do close-up work and just demand the best finder, again the D3 comes to mind. If you are into wildlife photography, a D2X or a D300 might be a "better" camera" than the D3 unless you also have to have the ultimate high-ISO performance and call upon the D3. For close-up work, being in position to use longer lenses is a plus for D3 and having the better finder handy won't hurt either. And so on, ad nauseam.

The list of pros and cons for these two formats is literally endless. You simply have to define your requirements and find the format and camera model that slots in the best. From this point of view, the dual support of FX and DX format is a very wise move.

Nikon D3 Reviewed