This field-review is long, so I have divided it into three parts:
Part 1: Introduction and first impressions and handling.
Part 2: Shooting macro, lens selection, diffraction, autofocus performance and modes.
Part 3: Shooting wide angle, lens selection, high ISO, dynamic range and conclusions. Below

WIDE ANGLE
The main reason I planned to snare a D3 for this particular shoot was because I felt that the high ISO capabilities would provide exciting possibilities for wide angle in the dark temperate waters of British Columbia. I also thought that higher ISOs might also be beneficial in Guadalupe, particularly later in the day when the sun typically dips behind the island. Whatever the outcome it would provide diverse challenges for underwater wide-angle photography.

For wide angle I used two lenses, the Sigma 15mm fisheye and the Nikon 17-35mm. I chose the Sigma 15mm fisheye over the Nikon 16mm (of which I own two) because I prefer its closer focus and find it sharper than the aging Nikon. Since I have got used to using the Nikon 10.5mm fisheye, I have been less impressed with the sharpness of the older design 16mm fisheye. The 16mm is not a bad lens, I just prefer the 10.5mm and Sigma 15mm, I don’t recommend everyone rushing out and changing. I selected the 17-35mm over the 14-24mm for two reasons. First, I own one! I was offered a 14-24mm for the trip, but that combined with a D3 is a lot of someone else’s kit to risk below the water. But secondly, I was concerned about the corner sharpness with the 14-24mm particularly because it cannot take a dioptre. So I am afraid there is no test of that lens here.

Ultra-wide angle rectilinear lenses have always been troublesome underwater. Without wishing to run an egg-sucking course for grandmothers, here is an overview of why. Before you skip to the end of this section, understanding the problems of domes and rectilinear lenses is important in understanding why certain solutions work.

Once immersed in water a dome port creates a virtual image that we must focus on, which is both closer to the camera than the true subject and also curved. A dioptre helps the camera focus on this closer virtual image. Land lenses are designed to produce flat images of flat planes of focus. The curved focal plane created by the dome has corners that are closer to the camera than the centre. Typically, we focus on the centre of the image and rely on depth of field to keep the corners as sharp as possible.

Rectilinear lenses are much more sensitive that fisheyes to corner sharpness. One reason is that they have pincushion rather than barrel distortion. This means corner detail is stretched out, rather than squashed in, making flaws more obvious. Furthermore, FX chips are bigger, yet the same distance behind the lens as DX ones, so light rays from the lens are striking the photosites and more acute angles, making light gathering tougher. Perhaps most important, FX cameras have narrowed depth of field at a given focal length and aperture so we struggle more to keep those curvy corners sharp. Theory aside, the proof of these problems is in the photos. Here on Wetpixel there has been no lack of threads with folks struggling to get good corner sharpness with wide rectilinear lenses, both on DX and FX.

Dioptres are useful for two reasons. First, they help the camera focus on the virtual image, which is surprisingly close to the camera (infinity is only three times the dome’s spherical radius). And secondly, and rather fortuitously, single element dioptres actually introduce a bit of field curvature which helps offset some of the curved focal created by the dome. So the big question is would the 17-35mm, with dioptre attached perform behind the dome?