Scanner Nikkor 100 mm
This page discusses the lens of the Nikon Super Coolscan 8000 ED and 9000 ED films scanners (hereby abbreviated to Coolscan 8000 and 9000, respectively), and its usefulness in macrophotography. This discussion is part of a small set of pages on scanners and scanner lenses.
The scanner Nikkor 100 mm is housed in a rather large cylindrical barrel made of multiple metal sections. All these sections, and the retaining rings at the front and rear, are locked against unscrewing by thread sealant. One section, located off-center toward the sensor, has a larger diameter than the rest. All specimens I have seen carry numerous white marks, or scratches in the metal, made when calibrating the lens and aligning its optical groups. There are no exposed threads, and no filter mounts. Judging from the weight, the barrel is made at least in part from thick brass.
The lens of the Coolscan 8000 and 9000 has been reported as having a focal length of roughly 100 mm and a speed estimated between f/2.4 and f/2.8. It has no variable aperture. Nikon refers to this lens (as well as the lens in the Coolscan 4000, 5000 and V), as Scanner Nikkor ED. In the following discussion, this lens is referred to as scanner Nikkor 100 mm.
Extracting the lens from the scanner
Instructions for the disassembly of the Coolscan 8000 and 9000 are available from multiple Internet sources, for instance here. However, disassembly is quite straightforward if you do not plan to reassemble the device. The lens is placed at the bottom of a large plastic chassis and hidden by the metal bottom of the case. Except for the lens, none of the other parts of the scanner have any obvious uses in photography. The motorized slider that moves the film is quite lightly built, with numerous nylon sliding parts, and does not appear to be a good choice for focus stacking. The stainless steel rails of the film slider and focuser are precision-machined and polished cylindrical rods, but their non-standard diameter of ?? mm makes them difficult to use.
In most (but not all) specimens, the lens is held in place within the plastic chassis by a metal bar with a screw attaching it to the chassis. After removing this bar and a thin black metal cover hiding part of the lens barrel, the lens is said to just pop out of its retaining groove. In my case, instead, there was no retaining bar and the lens was very tightly seated in its chassis slot. It took a physical effort and leverage with large screwdivers to remove it.
Mounting the lens
Given the weight of this lens and its alignment critical for good performance, I decided to forgo the simplest solution of epoxying the rear of the barrel into a filter adapter ring, and instead decided for a more solid seating within a metal sleeve. I found that a set of extension rings with 52 mm threads in my possession has an inner diameter closely matching the diameter of the narrower parts of the lens barrel, without requiring shims. The section of higher diameter near the center of the barrel length makes a convenient platform for seating the front and rear of these extension tubes and guaranteeing a good alignment of the lens barrel within the tubes.
I epoxied a 14 mm long tube at the front of the projecting barrel section, and a 28 mm long tube at its rear. I am not concerned about losing access to the individual mounts of the optical groups, because I would not be able to realign the elements in any case, after disassembling the lens.
An additional 28 mm long extension tube is screwed at the front of the 14 mm tube, forming a shallow lens shade and providing a female 52 mm mount for filters, lens cap, lens shade, or a reversing ring.
The projecting region of the lens barrel still has a slightly larger diameter than the tube-clad portions of the lens.
Scanner Nikkor 100 mm vs. Printing Nikkor 105 mm A
It has been debated whether this lens uses the same optics as the Printing Nikkor 105 mm f/2.8 A. A direct comparison of this Printing Nikkor and the scanner Nikkor 100 mm puts this rumor to rest. The diameter of the front element (35 mm in the scanner Nikkor 100 mm, vs. 38 mm in the Printing Nikkor 105 mm A), distance between front and rear elements (about 80 vs. 90 mm) and shape of the front and rear element surfaces (strongly convex vs. almost flat) differ in the two lenses. The color of the lens coatings is also different, which may help to explain why the Printing Nikkors are reported to have a better contrast than the scanner Nikkor 100 mm. Only the diameter of the rear element (38 mm) is approximately equal in the two lenses.
The scanner Nikkor 100 mm optical formula consists of 14 elements (13 elements in some prototypes) in 6 groups.
The scanner lens is corrected also in the NIR to about 850 nm for the dust removing function (as reported by Marco Cavina). The Printing Nikkor 150 mm and 105 mm f/2.8 A (and likewise the original Printing Nikkor series preceding the A version) are specified as corrected in the interval 400-800 nm, so these lenses do seem to have similar specifications to the scanner Nikkor 100 mm in the NIR.
The Printing Nikkor 105 mm A alone used to cost more than the roughly contemporary Coolscan 8000 or 9000 scanners. The higher price of the Printing Nikkor cannot be explained solely by better lens coatings and a nicer barrel with variable aperture.
The scanner sensor and optical path
The scanner sensor of the Coolscan 8000 seems to have two (possibly three) lines of pixels, spaced a little apart from each other and without color filters. I cannot see any borders between adjacent pixels with a Mitutoyo FS-60 industrial microscope at 400 x, although I can clearly see the conductors leading to each pixel.
The scanned film width of the Coolscan 8000 and 9000 models is specified as 63.5 mm. The sensor chip length is approximately 61.5 mm, and about 4 mm of the length are used by contact pads and multiplexing circuits. The length of a pixel row on the sensor is approximately 57.5 mm. This makes the magnification somewhat less than 1x (more precisely, 57.5 / 63.5 = 0.9x). This may help to explain why the optical scheme of the lens appears to be close to symmetrical, but not completely so. The Printing Nikkor 105 mm f/2.8 A is optimized for 1x.
Another possible difference between the scanner Nikkor and Printing Nikkors is that the scanner Nikkor optical formula should have been computed for use with a flat glass plate (the window that protects the sensor chip) in the optical path, while the original Printing Nikkors were designed for film copying and therefore should be optimized for use without additional glass in the optical path. In general, fast lenses are more sensitive to image degradation than slow ones when a flat glass plate is added in the optical path. The glass cover of the Coolscan 8000 sensor package is thin (about 0.6 mm), but at this level of image resolution, and with an f/2.8 lens speed, its thickness might have a very slight effect on lens performance, including spherical aberration and both types of chromatic aberration, unless the glass plate is included in the calculations of the optical formula.
The thickness of the filter stack covering the sensor of digital cameras varies between roughly 2 mm and 5 mm, with DSLRs tending to have thinner stacks than CSCs (there are good reasons for this difference, but a discussion would be too detailed). The thickness of the piezoelectric "dust shaker" and window of the sensor package must be added to the filter stack. This total thickness is significantly higher than the thickness of the sensor window in the Coolscan 8000 / 9000. It is therefore possible that the lens performance on a camera with a thick filter stack is measurably lower than the lens design specification.
Two of the large-format film holders for the Coolscan 8000 and 9000 scanners do have a glass plate (of thickness unknown to me, but probably thin). I believe in normal operation this glass plate is placed under the film and helps to support its weight. Since the scanner mirror and lens are placed at the bottom of the scanner, this means that the glass plate of the film holder is in the optical path between film and lens. The slight degradation resulting from the glass plate is probably a lesser problem (especially with large-format film) than the effects of film sagging would be in a glassless carrier. Since the glass plate is only present in these two types of film carriers, but not in others, the design of the optical formula of the lens may be a compromise between operation with and without this glass. This seems to indicate that the lens will operate optimally when used on a camera with a thin filter stack, if the lens is reversed and used at 1.1x. With the lens in normal orientation, this advantage does not exist, but tests should be carried out in both lens orientations to see whether this theoretical concern makes a difference to lens resolution in practice.
Any glass in the film carrier placed above the film, on the other hand, is not in the optical path between film and lens, and has no effect on image quality. The Coolscan 8000 (and probably 9000) uses a cylindric perspex rod as a light guide and homogenizer for the LEDs, and a white rectangle painted on one side of this rod reflects and diffuses light through the opposite side of the rod and toward the film. Light coming directly from the LEDs hits the walls of the rod at a low incident angle and therefore is totally reflected back to the interior of the rod (this is called total internal reflection and is a consequence of Snell's law). The portion of the rod painted white, instead, does not reflect light back into the rod like a mirror but diffuses it back across a broad angle. The light then exits the rod along the side opposite the white patch. The film illumination is therefore very well diffused.
Mounting the lens
A stack of additional extension tubes can be added at the rear of the modified lens, ending in an adapter for the camera. Since this lens is optimized for a single magnification, it makes sense to leave these extension tubes and adapter permanently attached to the lens. For use on a large-format sensor, it would be a good idea to place an adapter at the rear of the lens and use extension tubes with a larger diameter, for example 58 mm. Baffles with a clear aperture suitable for the sensor size being used can be added between individual extension tubes, to prevent flare and improve contrast. It is also a good idea to use a narrow lens shade whenever possible.
One possible approach is to support its barrel in a V-shaped cradle (above). This is especially suitable in horizontal setups, since the lens weight helps to keep it aligned in the cradle. It is necessary to use multiple V-shaped supports because, on the surface of the extension rings, smooth sections alternate with machined hand grips in slightly higher relief. The V-shaped supports should only touch the flat portions, to minimize the risk of misalignment. For this reason, it is not possible to use a single, long V-groove cradle of the type used for drilling and machining pipes, which would make the construction of the support simpler. The figure shows multiple V-shaped lens supports of the type used on 12 mm rail rigs for movie cameras, attached to a long Arca-compatible plate. This solution leaves the possibility open to small misalignments, because of compression of the rubber pads, as well as loose tolerances between screw shafts and the Arca-compatible plate.
For vertical setups, a V-shaped cradle would require the lens to be tied tightly to the cradle. The risk of sliding within the cradle and ruining the focus settings is always present. A better solution for vertical setups is to employ a system that incorporates a lens collar for tripod attachment. Thorlabs markets a few reasonably priced tube holder rings in different sizes. I did not test them but, based on specifications, the SM2RC (dead link) might tightly fit around the 52 mm extension tubes, possibly without shims.
Using the lens, in theory
Like many highly-optimized lenses, the scanner Nikkor 100 mm appears to be designed for use at a single magnification (0.9x, or 1.11x reversed). Slightly different magnifications may work well, but need to be tested. The Nikon Printing Nikkor 105 mm f/2.8 A, for example, has been rated as outstanding from slightly more than 0.5x to slightly less than 2x for what concerns image resolution. My own tests on Micro 4/3, however, suggest that a narrower range of magnifications should be used with this lens (ideally, only 1x).Shooting at 1x in 50 Mpixel mode and cropping provides a better overall image quality than recording the image in 20 Mpixel native sensor resolution while pushing the lens to 1.5x by increasing its extension or using a focal length multiplier. Using a Metabones Speed Booster to reduce the magnification was likewise unsuccessful, but for different reasons (vignetting).
Similar uncertainties apply to add-on lenses mounted in front of the scanner Nikkor 100 mm to increase its magnification. Only a test can give reliable answers.
Lens resolution in theory
The scanner Nikkor 100 mm is rated at 4,000 dpi (2,000 lppi) on the subject side across the image circle. This translates to 2,222 lppi (line pairs per inch) on the image side at 0.9x. It is quite possible that center resolution is actually higher, but we cannot know without testing, because this parameter was not relevant to Nikon's intended use of the lens. For Nikon's purpose, it was enough for the lens to produce a minimum resolution of 4,000 dpi across the image circle.
The Printing Nikkor 105 mm f/2.8 A is rated at 240 lines per mm at the center, and 180 lines per mm across a 60 mm image circle (on both subject and image sides, since this is a 1x lens). There is no indication of the source of this measurement in the above paper, but it is possible that the author quoted a figure from Nikon literature. My tests indicate instead a resolution of at least 230 line pairs per mm as recorded on a small central portion of image circle in 50 Mpixel mode, i.e. roughly twice the resolution rated above. Is it possible that the resolution stated in the above source was mistakenly reported as lines per mm, but was instead measured as line pairs per mm? The omission of this key word may have been introduced anywhere along the chain from measurement lab through internal product documentation, marketing, printing and proofing. If we consider that resolution is typically measured as line pairs (not lines) per unit of length, and assume that this was the original measurement unit, my measurements essentially agree with the resolution reported by the above source.
As discussed elsewhere, a 20 Mpixel Olympus E-M1 Mark II has an approximate resolution of 3,045 lppi (line pairs per inch) in native (20 Mpixel) mode, and very roughly 6,000 lppi in 50 Mpixel enhanced resolution mode (interestingly, the latter figure is equivalent to 236 line pairs per mm, see the preceding paragraph). On paper at least, this camera therefore outresolves the scanner Nikkor 100 mm, and is therefore a reasonable test platform for testing this lens.
Lens resolution and image quality
See the detailed tests at www.closeuphotography.com. The scanner Nikkor 100 mm is extremely close in performance to the Printing Nikkor 105 mm f/2.8. In practice, the main differences are the lack of variable aperture in the former lens and an optimal magnification of 1.1x with the scanner Nikkor 100 mm reversed, while the Printing Nikkor 105 mm uses an apparently symmetric design and is optimized for 1x. The Printing Nikkor 105 mm f/2.8 A also has a modest edge in contrast and resistance to flare.
The scanner Nikkor 100 mm is potentially cheaper than the Printing Nikkor 105 mm, and for all practical purposes displays virtually the same high image quality of the latter lens. However, the scanner Nikkor 100 mm has no aperture diaphragm, and has quickly become very hard to find, and requires modification or adaptation for mounting on a digital camera.