Click on any of the images above to go directly to that page.
Resolution is a top issue for many digital photographers. Ever-higher megapixel counts persuade many to “upgrade”, though the truth is that improvements in areas other than resolution can be more important. Examples include focus accuracy, lens performance, color accuracy, noise, high ISO performance, buffer size, etc. Fortunately for those who are after higher resolution, increased megapixel counts are often accompanied by improvements in these other areas.
The “dirty little secret” of the megapixel game is that digital SLR resolution is fantastic under some conditions such as high-contrast black and white resolution charts, but only fair to good under many real-world scenarios, such as low-contrast but detailed scenes, mixed colors in fine detail, etc.
One need only compare results to scanning backs to see that digital SLRs, as good as they are, drop the ball on the potential clarity of a digital image. This is a direct result of interpolation, or in laymen’s terms: “intelligent guessing”. The actual color of a pixel is based on this interpolation of the Bayer Pattern sensor found in all color digital SLRs (and medium format backs), excluding a few low-resolution (but high-quality) also-rans, such as the Sigma SD10.
It is disappointing that neither Canon nor Nikon has incorporated technology similar to the Foveon X3 chip, which offers the prospect of up to twice the resolution of a Bayer Pattern sensor, as found in the D2X and EOS bodies reviewed here. The claim of double resolution is exaggerated, but certainly 50% is very defensible when real images are studied.
Challenges in comparing cameras
To fairly compare results between cameras is more difficult than it might seem.
First, exposure should be very close between cameras. This is not as simple as choosing the same aperture and shutter speed, since the light transmission of the lens, aperture accuracy, shutter accuracy and actual ISO all affect the exposure. Each camera has its own idea of what the tonal curve should be, which means the histograms will never be exactly the same—and in the field it is often hard to determine how close a match has been achieved, due to the limited resolution of the screen.
Second, comparable lens optical quality must be determined, so that tests between cameras are about the camera, not the lens. Otherwise, the comparison shows as much about different optics as it does about a camera’s prowess. Using the same optic on all cameras solves this problem, but the situations in which this can be done are limited.
Third, pinpoint focus is required to fairly compare the resolving power of these very high resolution cameras. Especially on resolution charts, a barely perceptible change in focusing can make a stellar lens look just “fair” or even poor. Even on “3D” subjects, there is often little margin for error for a valid comparison. On a resolution chart, curvature of field can also be a problem; a sharp center may yield fuzzy corners and vice versa, leading to an incorrect conclusion about corner performance (assuming the center is focused perfectly). Autofocus usually does not focus precisely enough to reliably achieve accurate focus—see Focus Accuracy at diglloyd.com.
Fourth, lighting should be the same between cameras. A truly strict testing regimen (and a boring one) would confine all tests to a studio environment using strobes. But in practice, so long as the lighting doesn’t change color, direction or brightness significantly, the comparison is still legitimate.
The “Anti Pixel-Peepers” club
The term “Pixel Peeper” has become fashionable lately. It is a pejorative term for those who would zoom in to “actual pixels” and critique the fine points of image quality (resolution, noise, etc), the assumption being that doing so is without photographic merit.
Thus, a convenient label is available that in one stroke dismisses concerns about the merits of a particular camera’s imaging quality, and simultaneously denigrates the person expressing those concerns. The author knows of at least one photographer who is fond of the term, but at the same time is also fond of buying the latest high-resolution digital camera or medium-format back! Of course, those who accuse others of “pixel peeping” could instead spend their time making new images.
The author would rather be a Pixel Peeper than in the aforementioned club, but is in actuality neither, having both pixel-peeped and chosen the Nikon D2X, which has lower resolution compared with the Canon EOS 1Ds Mark II.
To finish this droll topic, let us mention the following points which bear on it:
1. Prior to digital cameras, photographers had no aversion to comparing different filmstocks, developers, etc. A digital camera is akin to a particular type of film, with its characteristics of noise (grain), dynamic range (tonal curve), sharpness (resolution), color accuracy, etc. There is a natural transition from comparing films to comparing digital cameras.
2. The 35mm format has co-existed with 645, 6X6, 617, 4X5, 8X10, etc. Prior to digital, there was no talk of Pixel Peeping simply because a photographer chose a 4X5 camera over a 35mm camera, yet resolution is one oft-cited reason 4X5 is chosen over the smaller 35mm format!
3. The cost/quality issue will not go away. Today’s digital photographer may choose anything from a sub-$1000 digital SLR all the way to a $30,000 medium-format digital back. The idea that this choice should be made without some degree of Pixel Peeping is ludicrous, since ultimately the choice comes down to image quality versus price—why spend $30,000 if $8,000 or $1,699 will suffice? Only a critical evaluation of image quality can provide the facts to allow a rational choice.
The image dimensions for the images produced by the cameras are shown below:
|Camera||Image dimensions (pixels)||Sensor Size||Linear Resolution relative to D200||Pixel size (microns)|
4288 X 2848
23.7 X 15.7 mm
3872 X 2592
23.6 X 15.8 mm
Image dimensions indicate the potential to resolve image detail, but other factors come into play, namely a variety of optical limitations which include lens resolution and contrast, lens aberrations such as chromatic and spherical aberration, and depth of field. Of course, resolution in the final image can also be reduced by camera or subject movement.
Strictly speaking, resolution is somewhat vague without also stating the contrast of the resolved detail (MTF). For the purposes of this article, the term “resolution” is used to mean detail that a human can discern in the image, regardless of contrast. Contrast is more subjective, making an image “pop” or look “flat”, though contrast can be increased in post-processing with excellent results in most cases.
Physical limits on resolution and contrast
Diffraction, an optical phenomenon caused by the bending of light around an object (such as a lens aperture), limits resolution to approximately (1600/f-stop) line pairs/mm. For example, the theoretical resolution limit of any lens at f16 is approximately 100 line pairs/mm, or a spot size of 5 microns (millionths of a meter). There are other varying characteristics of the resolved “spot” beyond the scope of this article.
The Nikon D2X, with its 5.5-micron pixels, suffers when the lens is stopped down beyond f16, because diffraction rapidly expands the spot size resolved by the lens. At f22, the spot size is already about 7 microns, quite a bit larger than the D2X pixels.
However, even if lens resolution is adequate, stopping down past f8 begins to rapidly diminish the contrast of the resulting image, which degrades the image by making it appear softer and more diffuse, even though actual resolution may be the same.
In practice, significant degradation in image quality begins when stopping down beyond f8 in the D2X, and f 9 in the D200. The limit for a maximum-quality image (with a slight loss of contrast) is f11 for the D2X, and f13 for the D200. That is not to say that f22 should never be used, but the marginal increase in depth of field comes at a high cost in image contrast and sharpness, particularly on the D2X. Note that two stops are required to double/halve the depth of field.
In addition to not stopping down too much, there is the question of how far to stop down to achieve sufficient resolution for the sensor. No lens from any manufacturer (that I’ve tried) provides optimal resolution and contrast on these high-resolution cameras when used wide open, except perhaps at the center of the image. Typically two to five stops are required from maximum aperture to deliver the best a lens is capable of.
Matching field of view
For a fair comparison, the field of view must be as close to identical as feasible. The sensors used in the D2X and D200 are for all practical purposes identical in size, so comparisons are straightforward with regard to sensor size.
However, there is another practical difficult: simply swapping cameras on the same tripod does not yield the same composition due to the height difference of the cameras, and the slightly different angles of the mounting plates on each camera. It was not possible to swap the two cameras and achieve identical framing; the camera always needed at least minor repositioning.
The test shots
All comparisons were done using a Gitzo 1325 tripod with Really Right Stuff BH-55 ballhead or the Burzynski Protec ballhead. Mirror lockup was used and all files were shot in RAW (NEF) format. All results were carefully checked for inconsistencies and errors; in some cases the test was redone.
The Nikon D2X is the resolution champ, with the D200 close behind.
Lenses matter. The author conjectures that there is simply no lens available that can fully exploit these sensors across the entire image. Still, even with these optical limitations, results of astonishing clarity are achievable that justify the abandonment of traditional film.