Light Fall-Off and Angles on Large Sensors

<p>Recently, I've become fascinated with issue of light fall-off on medium format sensors, particularly, as I understand it, that caused by micro-lens sensors. Having received interesting responses on an earlier, related post, I thought I'd see if I can learn more with this one.</p>

<p>The ad information about the Rodenstock 32mm f/4 HR Digaron-W Lens, designed for full format sensors, says that vignetting is eliminated at f/8. I find this remarkable if it suggests that this lens, which sells new for about $8,000, would vignette wide open, presumably even without tilt or swing, on a sensor that costs about $40,000. That is, I would find it remarkable if the best and most expensive technology available didn't support fast, wide angle photography, such as would be demanded by professional landscape and architecture photographers. And, I suppose, that movements on a technical camera would make the matter worse, perhaps yielding cross-talk even with a retro-focus lens such as the Digaron (which I had previously understood to be the solution for angle-of-incidence problems on large or micro-lens sensors).</p>

<p>As the technology of large sensor photography advances, is wide-angle being left behind to film, older sensors (those without micro-lenses, such as the 48 x 36 Phase One P45+), or smaller sensors (such as the new 44 x 33 CMOS models)?</p>

<p>The cells in digital sensors have a significant depth, much like a well. They are most efficient when light strikes them perpendicular to the surface. Microlenses are used to straighten light impinging at an oblique angle. There is also a finite distance between the sensor and the Bayer filter, so oblique light can cause color shifts away from the centerline.</p>

<p>Lenses also vignette because the nominally round aperture appears elliptical from the edges of the image plane. The smaller the aperture opening, the less the effect. This is simple geometry, not a defect.</p>

<p>Light fall-off is not a big problem for MF reflex lenses. The lenses have a long back focus to accommodate the mirror, hence the light does not impinge on the sensor at such an oblique angle. Technical cameras, on the other hand, are better served using lenses designed for that purpose, but come much closer to the film plane. They are used for a more leisurely form of photography, and generally stopped down and on a tripod.</p>

<blockquote>

<p>Light fall-off is not a big problem for MF reflex lenses. The lenses have a long back focus to accommodate the mirror, hence the light does not impinge on the sensor at such an oblique angle. Technical cameras, on the other hand, are better served using lenses designed for that purpose. </p>

</blockquote>

<p>Right, but this was the point of my post. The Digaron lens I mentioned <em>is specially designed to be used on a technical camera with a full-format sensor</em>--using a retrofocus design not for a mirror but to allow movements and to straighten the light--but seems not to do a particularly good job if the lens can't be shot wide open without vignetting. This is what I found surprising.<br>

<br>

All this said, I've heard it suggested that the best new CCD sensors, though using microlenses, are much more amenable to angled light and so for those spending the money on top of the line equipment, there is a solution there even among new, full frame sensors. The question will be whether as CMOS sensors replace CCD sensors the technology for the required microlenses will continue to allow movements on wide-angle lenses for technical cameras.</p>

<p>Most lenses for technical cameras are not retro-focus. It doesn't surprise me that a 32mm lens is an exception.</p>

<p>Micro lenses will improve the efficiency of light collection and minimize false colors caused by oblique light rays. However vignetting will still occur because edges of the image will still see less of the rear pupil. Miccrolenses are optimized for a centered lens, so the use of movements complicates things. Some lenses employ a circular gradient filter to overcome vignetting, which would apply to camera movements too. Some cameras, such as digital Leica cameras, make a correction in firmware based on a table of lense characteristics. There's no reason a MF digital back, particularly a CMOS sensor, could not do the same. Perhaps MF digital users, being perfectionists, would prefer to do it on their own. Vignetting varies with distance and aperture, so one size doesn't fit all.</p>

<p>Vignetting of this sort is easily corrected in Lightroom or Photoshop. Most of the time I use this correction in the opposite direction, to darken the edges in order to draw your eye into the center of the photograph.</p>

<p>Thanks, Edward. This all makes sense, and I appreciate your taking the time to educate me on this. One, final, simpler question, and perhaps a stupid one (though I'll risk it). Lenses specially designed for medium format digital seem to go no wider than about 35mm, which is about 20mm in 35mm equivalents against a full frame 645 sensor. (There are exceptions, but these seem to be devoted to smaller-than-full-frame sensors and so are no wider.) There are, however, wider lenses (including non-fish-eye) for smaller sensor cameras. So, e.g., Panasonic makes a 7mm lens (14mm in 35mm equiv.) for a micro 4/3 camera. If I'm right, I wonder why it is harder to go wider on larger sensors than smaller ones. Perhaps it's that for small sensors, the image circle need not accommodate movements, and I notice that large format lenses similarly don't offer a wider image against 4x5 film, but I'm not sure I have the geometry right (and why this may be a stupid question). </p>

<p>It's probably a combination of size, weight and cost.</p>

<p>Considering the long backfocus required to accommodate the swinging mirror, wide angle lenses use negative elements in the entrance pupil to provide a retro-focus design. This requires other elements to correct the inherent distortions and aberrations in the lens. On an Hasselblad, for example, a 50 mm lens is not much larger than a "normal" 80 mm lens, but a 40 mm lens is nearly 4" wide across the front. </p>

<p>A similar problem arises at the other end of the scale. A 250 mm lens for the Hasselblad is over 7" long, despite telephoto design, and for 350 mm and 500 mm (the longest offered), the dimensions increase on steroids. Similar issues confront smaller formats, but lenses can have a much shorter or longer relative focal length and still be manageable.</p>

<p>Large format cameras are limited as to how close the lens can approach the film plane, or how long the bellows can stretch. On 4x5 cameras, lenses shorter than 75 mm be mounted on a recessed lens board. Lenses longer than 200 mm often have a telephoto design to shorten the back focus. You would be considered well-equipped to go afield with three lenses - a 75-90, 135-150, and 180-225.</p>

<p>Lenses for rangefinder and mirrorless cameras can be much smaller because it's seldom necessary to employ retro-focus or telephoto designs. An Hasselblad Superwide has no mirror, and the dedicated 35mm lens is a symmetrical Biogon design, much smaller than the 40 mm for the reflex camera. Rangefinder cameras are limited by the viewfinder to a range of 28 mm to 135 mm. By accommodating the wide end, it becomes very difficult to focus or accurately frame lenses longer than 50 mm. Mirrorless cameras with electronic viewfinders may represent the future of cameras for amateurs and professionals alike.</p>

<p> </p>

<p>Interesting. Thanks so much.</p>

<blockquote>

<p>Lenses specially designed for medium format digital seem to go no wider than about 35mm, which is about 20mm in 35mm equivalents against a full frame 645 sensor. (There are exceptions, but these seem to be devoted to smaller-than-full-frame sensors and so are no wider.) There are, however, wider lenses (including non-fish-eye) for smaller sensor cameras. So, e.g., Panasonic makes a 7mm lens (14mm in 35mm equiv.) for a micro 4/3 camera. If I'm right, I wonder why it is harder to go wider on larger sensors than smaller ones. Perhaps it's that for small sensors, the image circle need not accommodate movements, and I notice that large format lenses similarly don't offer a wider image against 4x5 film</p>

</blockquote>

<p>The larger film and digital formats are not really that devoid of ultra-wideangles. Like the little Panasonic, the widest rectilinear medium and large format lenses also tend to be equivalent to around a 14-15mm lens on 35mm format - eg. a 35mm APO-Grandagon on a 6x9 cm backed Alpa (a tech-cam type of setup). </p>

<p>It's true that few medium format <em>SLR</em> systems approach that rectilinear ultra-wideness, but the record holder is probably the first version of the Pentax 645 24mm, which covers 645 film, and is equivalent to a 15mm lens on 35mm format. On digital medium format SLRs, the widest lens is the Hasselblad 24mm, which covers their bigger Kodak sensors with an equivalence of 17mm on 35mm format; next comes the Mamiya 28mm, which covers a "full frame" Dalsa 60MP or 80MP sensor with an equivalence of 18mm on 35mm format; and then the Pentax 24mm is slightly behind that, at 19mm on 35mm format, due to the smaller sensors in the 645D and 645Z.</p>

<p>Tech cams can push even wider on digital, thanks to the Rodenstock 23mm which covers a "full frame" Dalsa 60MP or 80MP sensor with an equivalence of just under 15mm on 35mm format.</p>

<p>Meanwhile, on 4x5 film, the Super Angulon 47mm XL is equivalent to a whopping 13mm lens on 35mm. </p>

<blockquote>

<p>The ad information about the Rodenstock 32mm f/4 HR Digaron-W Lens, designed for full format sensors, says that vignetting is eliminated at f/8. I find this remarkable if it suggests that this lens, which sells new for about $8,000, would vignette wide open, presumably even without tilt or swing, on a sensor that costs about $40,000.</p>

</blockquote>

<p>As far as vignetting goes, there are two types which arise in lenses: mechanical and optical. Stopping down the lens doesn't do anything for optical vignetting, because it is a geometrical effect which is unaffected by the pupil size. But stopping down usually improves mechanical vignetting, which is the partial obstruction of off-axis light traversing the lens. </p>

<p>A feature of retrofocus wideangles is that they normally vignette less than "optically true" wideangles of the same focal length. This is because their design reduces the optical component of vignetting. </p>

<p>So you may indeed wonder why this expensive retrofocus Rodenstock still suffers vignetting from f/4 to f/8. It's basically mechanical vignetting. As a tech-cam lens, it must have its own leaf shutter, and that is one mechanical bottleneck (it seems that only small size 0 shutters are used nowadays). Oversizing the lens elements, shutter, and barrel components would alleviate mechanical vignetting wide open, but that would make the lens even more expensive, and increase its weight and bulk.</p>

<p> </p>

<p>Hey, Ray, so you'll know the 35 Apo Grandy can be used on a 2x3 Crown or Century Graphic. Who needs an Alpa?</p>

<p>Cheers, Dan. That Alpa/35 Apo Grandy combination is repeatedly referred to by Roger Hicks & Frances Schultz, so that's why I had it in mind.</p>