35mm Shift Lens Group Test
Nikon 35mm f2.8 PC v Olympus 35mm f2.8 Shift v Contax Zeiss 35mm f2.8 PC
Comparative data about 35mm shift optics is hard to find, but there's no shortage of options: Nikon, Contax, Olympus and Pentax all have 35mm f2.8, f3.5 or f4 versions branded as shift, PC or Perspective Control. Canon previously offered a 35mm in FD mount, but revised the range with the introduction of the EF system so as to provide wider (24mm T/S), longer (45mm T/S) and tiltier designs. Of the three easily mountable on a Canon DSLR, the Zeiss has a peerless reputation – and price tag – but Nikon and Olympus users equally hail their on-brand contenders, so naturally I was curious to get beyond the hype to see how they really compare when used with full frame digital.
Much has been written about sensors' allegedly sensitivity to angular light, and aberrations potentially caused by rays striking microlenses obliquely – and some tests have shown particular CA problems with certain shift lenses such as the Schneider 28mm f2.8 PC – but in truth these are not generic issues. Rather, they pertain to compromises made to accommodate such 'adventurous' designs that have simply been exposed to wider coverage and more intensive scrutiny in the digital era.
Beyond merely discovering which of the three was the 'best', I was seeking answers to other questions, too: does a shift lens exist at this focal length that can bear comparison with a top-of-the-line fixed focal with a smaller image circle? I hoped it would: one lens less in the bag. What exactly does the Distagon PC bring to the table to justify its stellar price? And can any of these 35mm-format lenses do the job of a medium format lens like the Pentax FA or Mamiya AF 35mm prime deployed via a shift adaptor?
Another question that was easily answered had to do with coverage: if you stitch together a pair of fully left- and right-shifted captures, how wide is that? As it turns out, the answer is about 80°. In other words, although the focal length of each lens is absolutely 35mm, because the image circle is closer to 60mm in diameter than the 43mm required to cover a 36x24mm frame, we really have in our hands something close to a medium format lens – in fact, the conversion of 35mm shift lenses for use on a Mamiya 645 will be the subject of a forthcoming article. And 35mm in 645 terms is pretty wide: roughly equivalent to a 21mm lens designed for the smaller format – but cropped horizontally.
When we reach for a wide angle lens, particularly as architectural and landscape photographers, what we're looking for is commonly horizontal field of view (hFOV). More often that not, the loss of a little sky, or ceiling, isn't critical. Thus, for applications where hFOV matters (interiors, for instance), it may be convenient to think of a 35mm shift lens as a 21-35mm zoom that gets better the wider you use it – or at least it will, providing resolution is maintained in the outer image circle, when shifted. This, too was something I wanted to examine closely.
The reason for the increase in image quality should be evident without explanation. However, to quantify, consider that a single frame capture with a 5D and a Zeiss 21mm is 4,368 pixels wide and will pretty much resolve to the limits of the sensor at f5.6. A pair of shifted captures made by a Zeiss 35mm PC with the same camera and merged into one image will have a similar field of view but measure almost 7,000 pixels in width. However excellent the 21mm prime, it will never come close to the level of detail captured by two overlapping frames, even if the larger lens isn't quite as sharp: we're sampling the same scene with millions more bits of information.
Here's a familiar scene shot at 21mm / f11 with a Sigma 12-24mm at a distance of about 9m, with strips greyed out top and bottom corresponding to the field of view of a 35mm lens. When fully shifted, the 35mm will have much the same hFOV as the 21mm prime. Here we see the Olympus at f2.8. As you'll see shortly, the Zuiko has the longest focal length of the three tested, so you'll note that the coverage isn't quite as wide as the 21mm.
Give the size of the image circle (about 30mm radius), it's impressive – incidentally – to note that vignetting is well controlled by the Olympus shift lens. You'll also see that we've had to extend our usual Zone system to accommodate the extra coverage of these lenses. Bear in mind that the troublesome Zone C (17-23mm radius) that creates so many problems for wide angles is well inboard for a shift lens, creating high expectations of 'corner' performance when movements are not used. A 36x24mm frame is marked here in red.
A few paragraphs ago, I claimed that an OK 35mm lens with a big image circle would run rings around a 21mm prime. Well, here's 100% crops from the above to illustrate the point.
Obviously the f11 centre frame performance of the Sigma 12-24mm is far from being the last word in resolution or colour fidelity, but however exquisitely the 90,000 pixels on the left are rendered, they will never come close to the nicely drawn 263,725 pixels on the right (Canon 5D and Nikon 35mm PC at f11).
Hopefully that's got your attention. Having established why you might want to use a 35mm shift lens – without even discussing the job of perspective correction for which they were designed (which you'll either find indispensable or irrelevant) – we can move on to looking at which shift is the shiftiest of them all.
Colophon: test parameters
All images shot using a 36x24mm 'full frame' sensor: a Canon 5D mounted on a fully locked down Manfrotto/Bogen 058 tripod (minimum extension) with exposed spikes dug into moist turf and a Kirk Enterprises Arca-Swiss ball head. Images were exposed at ISO100 with mirror lock-up via Canon electronic remote release with custom settings applied (Saturation +1, Contrast -5, Sharpness +1) and processed using Phase Capture One v3.7.7. In C1, Extra Shadow and positive Midtone Contrast and Saturation were applied, plus default capture sharpening prior to development into 8-bit ProPhoto RGB TIFFs. In Photoshop, a single pass USM of 100 / 0.4 / 3 was applied prior to profile conversion to Adobe RGB and export using BoxTop ProJPEG. Light came from the sun. Despite grievous provocation, no cats were injured in the making of this test.