Introduction.

A debate often ensues among underwater photographers over which image sensor format is more suitable for underwater photography – the small format (Nikon calls it DX) or the large “full-frame” format (Nikon has termed it FX). The FX (large) format offers numerous advantages for general photography and is widely considered to be the “professional” format. Its two main benefits are a higher quality of image in unfavorable light conditions (when using high ISO sensitivity settings) and the “preservation” of lens’ angle of view. The word “preservation” is deliberately written in quotes, considering the association between the focal length and the field of view is only imagined in our minds, where it has been impressed during the age of 35 mm film. Upon the emergence of first commercially available digital SLR cameras, the small size of the sensor certainly posed a problem for wide-angle photography, as it made the lenses “lose” their wide angle. By now, this problem has been solved well and there are many wide-angle lenses for small sensors available on the market.

I have to agree that the benefits of FX format really outweigh the DX format and use it today almost exclusively for my land photography. On the other hand, I was convinced from the very beginning (since my transition from film to digital technology in 2006) that the DX (small) format is more suitable for underwater photography than the FX format. I have debated this time and again with my colleagues and tried to prove it (more or less unsuccessfully) on the basis of theoretical facts. Then I finally came to a decision to attempt to prove my claims with a practical test (or maybe realize that I was mistaken …). After all, my old D2x has by now become really obsolete and the results of this test will strongly influence my decision about its successor.

Underwater photography is fundamentally divided into two large fields – macro and wide angle, so I have conducted two separate sets of tests. Considering the macro photography, I am sure everyone will agree with me that the small format is the preferred choice over the large format. When taking pictures using the 1:1 reproduction ratio, the small sensor format namely provides a smaller crop of the nature, just as desired. Two additional phenomena must be taken into account in this respect: depth of field and reduced sharpness in the stopping down the aperture diaphragm due to light diffraction on the diaphragm blades. The first one speaks in favor of small format and the second one in the favor of large format. My practical test was designed to determine which of the two phenomena has a stronger effect.

I have devoted a lot more of my attention to wide-angle photography, the subject of most disagreements with my colleagues who advocate the FX format. My arguments were based primarily on the dome port theory, but I was unable to provide firm evidence, backed by experiments.

Equipment.

Conducting these tests, I was especially interested in how the size of the image sensor influences the quality of the image (mainly the sharpness!) created by various optical systems under water. The optical system means either a macro lens behind a flat port or a wide-angle lens behind a dome port (specifically a fisheye lens, used extensively in underwater ambient photography). For this reason I have deliberately chosen two cameras with different sensors and the same resolution (12 Mpix): Nikon D2x (DX or small format) and Nikon D3 (FX or “full-frame” format). If I used cameras with different resolutions (e.g. D2x and D3x), it would be much more difficult to evaluate the test results.

I used the AF-S Nikkor 105mm f/2.8 VR lens for macro photography on both systems because I was interested in the image quality obtained at 1:1 reproduction ratio (or close to it) and not in the difference between the angles of view, which is already known.

I used fisheye lenses (AF Nikkor 10.5mm f/2.8DX on the D2x and AF Nikkor 16mm f/2.8 on the D3) and wide-angle zoom lenses (AF-S Nikkor 10-24mm f/3.5-4.5DX on the D2x and AF-S Nikkor 16-35mm f/4 VR on the D3) for wide-angle photography. The lenses in both pairs of camera setups are comparable and feature very similar angles of view.

For the tests I used the appropriate Seacam underwater housings and, most importantly, I used the same dome ports for wide-angle photography. The size of the dome port is decisive for the image quality, so it is mandatory to use the same dome ports (having the same curvature radius) for equivalent tests. Lighting was provided by Seacam underwater flash units.

I tried to use as equivalent settings as possible on both cameras, similar to the ones I use in underwater photography (the settings on the old-generation Nikon D2s are unfortunately not the same as on the newer D3, but I am pretty sure I was able to come quite close with my selection of image parameters, especially the sharpening level).

The following settings were used:

*Image sharpening settings in the Nikon D2x are between -2 to +2, and 0 to 9 in the Nikon D3, wherein 9 yields stronger sharpening than +2, so the sharpening levels were approximately equivalent judging from my experience.

In Photoshop I only adjusted the levels of photographed images and did not do any sharpening or other modifications.

The following dome ports were used in the tests: a fisheye dome port, a small fisheye dome port and a modified Seacam “Superdome”.

Three different dome ports were used in testing the wide-angle lenses:

The first one was a modified Seacam “Superdome.” It was modified in such a way to slightly reduce the glass diameter, leaving the curvature radius unchanged. The quality of the dome port depends on the curvature radius and the diameter determines the maximum angle of view. Because the curvature radius has not been changed, the quality of my dome port is identical to the original Seacam “Superdome.” The dome port was attached to the underwater housing using a combination of different extension rings in order to position the lens at an optimal distance to the dome port glass. I used a 45 mm wide extension ring for the 10-24mm f/3.5-4.5 lens and an 85 mm extension ring for the 16-35mm f/4 lens, which is physically considerably longer. The tests with this dome port and zoom lenses were conducted in a swimming pool only.

The second port was a Seacam “fisheye” dome port: a glass hemisphere with approximately 80 mm curvature radius (approx. 160 mm diameter) optimized for 180° fisheye lenses. To perform the tests in the sea simultaneously, I used two identical dome ports, each mounted on its own housing. This dome port was used to conduct tests in the swimming pool and in the sea.

The third dome port was a small “fisheye” dome with 50 mm curvature radius (approx. 100 mm diameter), a piece of equipment that is lately becoming increasingly popular for extreme close-up wide-angle images. This dome port performs well with lenses that are able to focus very closely (e.g. 10.5mm f/2.8). However, the optical quality of images taken with this dome port is inferior compared to the ordinary large “fisheye” dome port due to the small curvature radius. The tests involving this dome port were conducted in a swimming pool.

Chromatic aberration (CA).

Nikon D2x is an outdated older-generation camera and is unable to automatically remove the CA introduced by the lenses themselves or in combination with other optical elements (e.g. flat ports in underwater photography). To obtain a clean image without CA, we have to take the photos in the raw (NEF) format and then process them with a suitable program (the best choice here is the original Nikon Capture NX). All photos in JPG format obtained directly from Nikon D2x exhibit CA. Nikon D3 as a new-generation camera automatically removes CA using the same algorithms as Capture NX (this function is now provided by all Nikon cameras, even the cheapest entry-level models). JPG photos from Nikon D3 therefore exhibit no CA and cannot be compared directly to JPG photos made by D2x. My goal was to squeeze the best possible quality from both systems. For this purpose, the image quality was set to RAW + JPG large fine on both cameras. After processing the RAW (NEF) photos with Capture NX, I have compared the images without CA. In some cases, I deliberately processed the images in such a way to leave the CA untouched (even on the D3 images!) and then compared the “raw” optical results.

Macro photography.

The main and most obvious advantage of small (DX format) sensors compared to the large (FX format) sensors is their performance in macro photography. When taking pictures using a macro lens and 1:1 reproduction ratio, the DX sensor makes a 24x16 mm crop (or even a bit smaller!) in nature, while the FX sensor makes a 36x24 mm crop.

This feature is recently becoming especially important with the increasing popularity of so-called “super macro” photography. In the times of film, 1.4x teleconverters were used to provide magnification (yielding approximately a 26x17 mm crop, which is lower magnification as the crop made by a bare lens attached on a DX sensor camera!). However, the use of teleconverters reduced the general sharpness of the lens and reduced their speed for one full f-stop. This impaired the reliability of autofocus and made it substantially slower. The underwater macro photographers have therefore profited from the advent of DX image sensors, as we were once again able to use “bare” macro lenses (with better sharpness and quicker and more reliable autofocusing), with the function of an 1.5x teleconverter provided “for free.”

Two major technical problems in macro photography are the depth of field and diffraction of light on the diaphragm blades. The depth of field is tiny as a result of the small reproduction ratio. In order to increase the depth of field, we have to stop the aperture down, in turn causing the problem of diffraction that reduces the image sharpness.

The depth of field depends only on the reproduction ratio and the f-stop value. In macro photography, it can be calculated using the following simplified formula:

DOF = 2 . k . f . (1/X + 1) . X2

where: DOF = depth of field k = allowable circle of confusion (0.03 mm for FX format and 0.02 mm for DX format)* f = f-stop X = reproduction ratio

• The allowable circle of confusion for DX format is 1.5x smaller, because the photo must be printed 1.5x enlarged due to the sensor being 1.5x smaller.

If we take a photo so that the larger side measures 36 mm in nature, the reproduction ratio for the FX format sensor (36x24 mm) is 1:1 (X = 1). If we want to make the same picture using a DX sensor (approx. 24x16 mm) and the same lens, we have to move back to 1.5x the distance, and the reproduction ratio is 1:1.5 (X = 1.5). Using the same f-stop (e.g. f/22) and taking into account two different allowable circles of confusion in the formula, the depth of field is 2.64 mm for the FX format, and approximately 3.30 mm for the DX format (about 25% larger).

Unfortunately, stopping the aperture down in order to increase the depth of field we are limited by the diffraction of light on the aperture diaphragm blades. The diffraction appears at some limit f-stop value and then reduces the general sharpness and contrast of image in a practically linear manner. Very simplified, the limit f-stop value can be calculated by multiplying the pixel size (in mm) by 2000. Let us look at our two test cameras (Nikon D2x and D3), both having 12 megapixels: The pixel size in Nikon D2x (DX format) is 5.6 µm (0.056 mm), and in D3 (FX format) it is 8.4 μm (0.084 mm). The limit f-stop for the DX sensor in Nikon D2x is therefore 0.056x2000 = 11.2 or f/11, while the limit f-stop for the FX sensor in Nikon D3 is 0.084 x 2000 = 16.8 or f/16. This means that the images will start losing sharpness when the aperture is stopped down to more than f/11 on Nikon D2x and to more than f/16 on Nikon D3. It must be emphasized once again that the diffraction depends on the pixel size and not on the sensor format. Considering both cameras have an approximately equal number of pixels, the pixels in Nikon D3 are bigger due to a bigger sensor. On the other hand, the pixels in Nikon D3x are approximately 40% smaller than in Nikon D3, making it 40% (about one f-stop) more sensitive to diffraction, and also very similar to Nikon D2x in this respect.

The optimum aperture values for macro photography would therefore be f/11 for Nikon D2x, f/16 for Nikon D3 and f/11 for Nikon D3x. In real life however, all these f-stops unfortunately offer too small a depth of field at 1:1 reproduction ratio, so we usually stop the aperture down for one or two further f-stops (thereby sacrificing the image sharpness!). I believe the best trade-off between depth of field and diffraction in real-life conditions is given by the f-stop values f/16 to f/22 for Nikon D2x and f/22 to f/32 for Nikon D3.

The macro images were tested using a special clipped aluminum block I have made in the time of film photography for tests and to find the optimal relation between the depth of field and diffraction. The block is mounted on a special carrier that can be attached to the housing in an arbitrary distance. This setup guarantees a constant distance and crop. A label with a millimeter scale is pasted to the block and the block surface is deliberately rough, so the sharpness or fuzziness of the photo can be easily identified. The right half of the block is clipped to an 11° angle and is designed for studying the depth of field. The tests were done in a swimming pool and the lighting was provided by two flash units. The photos were taken with f-stop values ranging from f/5.6 to f/45 in steps of 1. Shutter speed was 1/250 s to eliminate the influence of ambient light and the vibrations. The photos were taken using the 1:1 reproduction ratio with Nikon D3, and both the 1:1 and approximately 1:1.5 reproduction ratios with D2x, producing the same crop as with the D3.