Photomacrographic lenses, part 3: Shootout at the Zeiss corral:
Zeiss Luminar 63 mm f/4.5
Zeiss Luminar 40 mm f/4
Zeiss Luminar 25 mm f/3.5
Zeiss S-Planar 74 mm f/4
Zeiss Tessovar

On other pages of this site (1, 2, 3, 4), I tested lenses designed for use in photomacrography, as well as lenses not designed for this application, but usable when a specialized lens is not available or affordable. In all these tests, the Zeiss Luminar lenses performed best. By now, I have assembled quite a few pieces of Zeiss photomacrographic equipment, either by design or by taking advantage of unexpected opportunities. Thus, the next logical stage is to pit these Zeiss lenses against each other, and see which one is left standing at the end of the day.

An additional purpose of these tests is to check whether any of the Luminar lenses provides a clear advantage over Luminars of other focal lengths. In other words, since Luminars are not exactly available in every neighbourhood shop, if you can only have one, which focal length should you choose? It is true that each Luminar focal length is optimized for a specific range of enlargements, but these ranges overlap considerably, and at any given magnification likely there are two or three alternatives.

From a theoretical point of view, resolution is ultimately limited by diffraction, and the effect of diffraction is the same at a given magnification and effective aperture, regardless of the focal length of the lens. Therefore, there are no theoretical reasons to prefer a given focal length (as far as resolution is concerned). In practice, however, the higher working distance provided by lenses of longer focal lengths and/or of special designs does make a difference, especially in comfort of use and in the freedom of placement of light sources. Contrast also affects picture quality. Other less quantifiable and more subjective factors, like bokeh, may also play a role. In the end, given the fact that all the Zeiss lenses tested here are excellent, these factors may end up playing a major role in the present evaluation - or none at all, if results should turn out to be visually identical.

The three Luminar lenses tested here (leftmost three in the above picture) are from different versions (or "generations") of these lenses. Luminar versions are discussed here. A complete test would require me to purchase a series of Luminars of all focal lengths for each version, and to compare all these lenses. Since this would be a problematic, as well as rather extravagant, purchase (as the Chief Financial Advisor of this home also reminds me), my small present collection of Luminars and related lenses, assembled for the purpose of using them in practical macro photography and photomacrography, will have to do (quite some time after writing this page, I did have a chance to test the version 1 and version 3 of the 63 mm). In addition, I decided to conduct this test in realistic conditions that I am likely to encounter in practical photomacrography. Although I could assemble a long stack of bellows and extension tubes to test the whole magnification range of these lenses, I decided not to do so, because I would avoid it also in routine photography (instead of using an extreme lens extension, I would rather use a lens of shorter focal length mounted on Nikon PB-6 bellows alone, and live with the shorter working distance).

The diaphragm of Luminar lenses has "a lot" of blades (often more than normal lenses promising a "good bokeh" because they use 7 or 9 blades to achieve a rounded diaphragm shape). My Luminar 63 mm has at least 12 blades. All my Luminars show very round apertures. On the other hand, the S-Planar 74 mm and the Tessovar have only 6 blades, and an unashamedly polygonal shape. I have seen comparable polygonal apertures also in the Schneider Betavaron (probably the most sophisticated enlarger lens ever made, and very expensive in its time, but with only 5 blades), and even in the Canon 35 mm f/2.8 for photomacrography. We will see in the present test if aperture shape makes any difference.

Zeiss Luminar 63 mm f/4.5

This lens (left in the above picture) is optimized for a magnification range between 2x and 10x (as declared by Zeiss). However, based on my calculations of diffraction effects, this lens detectably loses some resolution already at 5x and beyond, even when fully open. This is not to say that this lens should be used only below 5x, but that resolution will be unavoidably lower beyond this limit. In practice, the Luminar 63 mm is rarely used above 4x-5x because of the relatively high focal length, which requires long bellows. I used this lens as a term of comparison in a few other tests (1, 2, 3), in which it always outperformed other lenses. This 63 mm is a Version 1 probably from the 60's.

Zeiss Luminar 40 mm f/4

This (second from the left in the above picture) is declared by Zeiss as optimized for magnifications between 4x and 16x. My calculations indicate 8x as the magnification at which diffraction causes a loss of resolution even when fully open. This 40 mm is a Version 2 from the 70's or early 80's. It uses the completely re-designed optical scheme (with maximum aperture f/4) that superseded the f/4.5 Version 1.

Zeiss Luminar 25 mm f/3.5

This lens (second from the right in the above picture) is declared by Zeiss as optimized for magnifications between 6.3x and 25x, or up to 10x according to my calculations. Earlier on, I tested this lens here. The tested lens is a Version 4, probably from the early 90's. It has the dedicated 10 mm extension ring that always accompanied this lens in Versions 2 to 4.

There is also a Luminar 16 mm f/2.5, optimized for 10x-40x. I have none, and it is low on my list of priorities. This lens is notoriously difficult to use. In practice, already at 10x you cannot close its aperture by even one stop without loosing resolution to diffraction, and at magnifications above 20x it is visibly unsharp (if you look at picture crops) even when fully open. A few photographers have reported better resolution and working distance with specialized microscope objectives (e.g., see here) than with this Luminar. This result is supported also by theoretical considerations. In addition, specialist lenses like the Zeiss Tessovar (discussed below) are easier to use, and have much higher working distances at high magnifications.

Zeiss S-Planar 74 mm f/4

This (above picture, rightmost) is an unusual lens. It lacks a focusing helicoid, and its attachment thread (an odd 32 mm x0.5) is located high up the barrel, so the rear of the lens projects about 17 mm behind the mounting flange. This lens is unusually long for its focal length and speed (40 mm between front and rear elements, or roughly twice the Luminar 63 mm f/4.5). It is unlikely you can make it focus at infinity on a DSLR, but the placement of the attachment thread does allow a low magnification factor when attached to a focusing mount or bellows. It is shown above mounted in a home-made adapter with Nikon F bayonet.

This lens is optimized for a 1:1 reproduction ratio, and it says so not just once but three times on its barrel, lest you forget it. It is meant primarily as a copy lens with very high resolution and very low distortion, to be used for 35 mm film duplication. The data sheet does not mention any other enlargement factor, so it is left to my guesses to anticipate that it should work optimally in a range between 0.5x and 2x (and quite possibly 0.25x to 4x), therefore complementing the Luminar series at its low end. The optical design is completely symmetrical, so there is no reason to reverse this lens. The optical formula is simple by today's standards, with just 6 elements in 4 groups, and not unlike the design of most EL-Nikkor enlarger lenses (the EL-Nikkor 63 mm f/2.8 came in second when used as a photomacrographic lens, after the Luminar 63 mm). Luminars are even simpler (typically, 3 or 4 elements), and yet are top performers. This makes me wonder whether the complex designs of current macro lenses (e.g., 10 to 16 elements in Sigma lenses, and up to14 in Micro Nikkors, with floating groups and a couple of aspheric elements and another couple of exotic glass elements thrown in) really are necessary. Perhaps I am too suspicious, but I have a feeling that such a large number of elements is more a way to compensate for mass-production with relatively poor manufacturing tolerances (besides making a selling point) than to provide top optical performance. After all, the sharpest eyes in the animal kingdom only use a single optical element (albeit a very optimized one, accompanied by additional matched optimizations in the rest of the eye).

The aperture of the S-Planar 74 mm as indicated on the barrel is not the nominal aperture (given by F/D, where F is the focal length and D the diameter of the front element). Instead, it is the effective aperture at a magnification factor of 1x, i.e., 2 stops closed-down with respect to the nominal aperture.

The diameter of its front and rear elements (approximately 24 mm) would be enough for a maximum aperture around f/3, and the "waste" of lens surface with respect to the f/4 nominal aperture suggests design parameters that place the emphasis on top performance even when fully open, rather than price. At 1x, it generates an image circle of 60 mm (if a very slight darkening at the edges is tolerated), or 40 mm with less than 5% fall-out in the corners (i.e., less than 1/8 of a stop, which you will never be able to detect visually). Distortion is practically absent on a 24 by 16 mm sensor (<0.05%), and resolution - from a theoretical point of view - remains highest between fully open and nominal f/11 (or f/22 as indicated on the barrel). The two additional stops available beyond f/11 come in handy when resolution can be sacrificed to extra depth-of-field.

Although the S-Planar 74 mm is advertised at prices exceeding 600 € in second-hand shops, it can be had for less than half of this amount on auction sites, thanks to being more obscure than its Luminar cousins. The only problem is finding one, because they are also much less frequent than Luminars. I have failed to find any historical data, but based on its appearance it may have been manufactured in the 80's and/or 90's, and is in fact a more modern design than the Luminars.

A possible concern is the fact that it does not have a lens shade, nor a filter thread, nor sufficiently recessed front elements, and therefore it could be vulnerable to flare in certain conditions (although this did not emerge during my test). However, a lens shade or lens cap can easily be made out of a plastic 35 mm film canister, which has the right internal diameter (30 mm).

Zeiss Tessovar

I described and illustrated the Zeiss Tessovar in detail here, and also provide a copy of its original documentation. This is a parfocal zoom photomacrographic lens with numerous accessories, and provides a magnification range between 0.4x and 12.8x (when used with a rotating nosepiece containing special add-on lenses). This lens is characterised by a very narrow maximum aperture (nominal f/22), and therefore is difficult to focus at low magnifications. In addition, practical resolution is reduced by diffraction at most apertures and magnifications. The working distance changes with each add-on lens used, but remains constant throughout the zoom range of the lens (as long as the camera is mounted on an extension tube of appropriate length). The working distance is 36 mm in the range 3.2-12.8x, and increases roughly twofold for each lower magnification range, reaching a hard-to-beat 320 mm at 0.4-1.6x.

Test methods

The above lenses (except for the Tessovar, see the setup here) were tested mounted on Nikon PB-6 bellows attached to the table-top focusing stand of the Tessovar and a D200 DSLR.  Illumination was provided by an incandescent lamp for focusing, and a Nikon SB-800 flash in iTTL mode for the exposure (with manual exposure compensation if needed).

In earlier occasions, I carried out tests of these lenses in similar conditions. However, this time all lenses are of very good quality, and therefore, examining the results turned out to be much more difficult. Two main problems became apparent. The first is that the incandescent lamp used for framing and focusing did contribute to the exposure. Since the tests were carried out at different lens apertures, this lamp contributed more when the diaphragm was open, and therefore did affect the colour balance. The solution was simple - just turn off the lamp before exposing.

The second problem is that evaluating the resolution of high-resolution lenses in photomacrography requires the subject to be perfectly focused. This turned out to be especially difficult at the borderline between macro photography and photomacrography (i.e., in practice, between 1x and 2x). At lower magnifications, DOF is high enough to be slightly forgiving. At higher magnifications, DOF is so small that it is easy to see what is in focus, and what is not. In this critical range, however, a small detail appears focused in the viewfinder (even with a 2x viewfinder magnifier) for a rather large interval of movement of the focusing rack, but not when the picture is examined at 1:1 pixel ratio on a large and sharp monitor (while old-fashioned CRTs were a little fuzzy, with pixels slightly overlapping each other, today's LCD screens are as sharp as anything can get). This is a much harder problem to solve. A factor in this problem is that the lenses tested here do not have especially large apertures (f/4 and f/5.6, respectively). Modern macro lenses typically have f/2.8 apertures, and are easier to focus when fully open.

My solution was to take pictures of a subject with a slightly convex surface, so that, no matter where the focal plane was located, I could always find an area that was completely in focus. The drawback of this solution is that I am unable to show crops of exactly the same spot for comparison among different lenses. However, the subject (a sea shell with a lot of small-scale surface detail, see Fig. 1) is sufficiently uniform to provide similar detail in different areas. Also, DOF is so low that even in a 400 by 400 pixels crop some spots may be out of focus. Thus, in evaluating the results, concentrated on the sharpest areas and disregarded unfocused ones. Also, reddish colours are predominant in the chosen subject. Since red light has a longer wavelength than other colours, diffraction shows its effects (i.e., loss of resolution) in this colour more than in others. A green or blue subject would have a higher sharpness (in an objective sense, regardless of our perception or the sensitivity of the camera sensor), and would allow closing the diaphragm between half and one more stop. Nonetheless, this test object is quite representative of  real-world photomacrographic subjects. 

The present tests differ from earlier ones also in testing all lenses at the narrowest aperture that still provides a resolution visibly unaffected by diffraction. This aperture can be computed as a function of magnification, sensor resolution and circle of confusion (see here), and is independent of focal length and lens design. Of course, it is entirely possible that, in a real lens, aberrations and other design faults will cause a loss of resolution at this and/or more open apertures. The computed value is a theoretical limit that no lens can exceed, i.e., beyond this limit, visible diffraction does arise no matter what the lens designers may choose. Thus, the aperture computed in this way is a theoretical "sweet spot" in photomacrography, because it provides the maximum possible DOF and resolution at the same time. It should always be used, unless DOF becomes more important than fine resolution. A truly excellent lens will be designed to perform at its best when used at this aperture. As the main target for the current test, this criterion ensures that excellent lenses can be separated from poorer ones.

The above criterion, of course, is very severe. It is based on a circle of confusion as small as 3 pixels (20 μm) on a D200 sensor, i.e., the smallest distance that theoretically can render two points as separate in an image. It results in a recommended range of apertures and magnifications narrower than those specified by Zeiss for the same lenses. However, this is intentionally a severe test. If you allow a larger circle of confusion, you will end up with maximum magnification ranges similar to those specified by Zeiss.

Results

The following figures are either full-frames reduced to 600 by 400 pixels, or crops of areas of pictures with an original size of 400 by 400 pixels, displayed at 1:1 resolution on the screen. Both types of pictures are compressed for web publication, and therefore may display compression artifacts and are of lower quality than the originals. The main data relevant to each picture is indicated in the caption below a picture.

The above pictures show that the S-Planar 74 mm is extremely sharp throughout the range of magnifications allowed by the PB-6 bellows. The reduced full-frame in Fig. 1 shows excellent contrast and colour saturation. The detail in Fig. 2, at 0.62x, shows a slightly lower resolution than at higher magnifications (Fig. 3-4). Resolution improves by opening the aperture one stop. Fig. 2-4 are taken at the narrowest aperture settings that are still unaffected visibly by diffraction (as calculated on a theoretical basis). These apertures provide the best DOF and resolution at the same time. Fig. 5 shows a very slight loss of resolution (the smallest white dots are more diffuse) by closing one stop further than in Fig. 4. Fig. 6 shows the result of closing the diaphragm all the way (5 stops from the optimal), which produces unacceptably fuzzy pictures even when substantially reduced. In other words, diffraction is not to be trifled with. In Fig. 6, closing the diaphragm also produces a large DOF on the side of the film/sensor plane, which causes microscopic dust particles sitting on the antialiasing filter of the camera to become visible (the darker round "doughnut" at center-right, surrounded by diffraction haloes).

The Luminar 63 mm is also extremely sharp, but loses a little of its legendary performance at 0.9x and aperture 15 (opening to 8 improves the results). Since this magnification is quite outside its design range, it can hardly be regarded as a fault. It attains maximum sharpness at 2x, and is still at top performance at 3.5x. Also in this case, the pictures shown above are taken at the optimal aperture and DOF. The S-Planar 74 mm has a slight edge on the Luminar 63 mm below 2x, but the two lenses are essentially equal between 2x and almost 3x.

At 1.75x and the theoretically optimal aperture 8, the resolution of the Luminar 40 mm is slightly degraded, but improves if the aperture is opened to the 4 setting. This is the most consistent deviation from theoretically expected behaviour I encountered in this test. It is not very important, because this lens is optimized for a magnification range starting with 4x. At 2x, performance at the "sweet spot" aperture is already improved. At 3x and above, performance is virtually undistinguishable from the Luminar 63mm.

The high magnification provided by the Luminar 25 mm forced me to choose a different subject (the juvenile portion of the same shell, which is green rather than reddish), as the original one was running out of useful detail at this scale. The Luminar 25 mm performs well also at magnifications lower than its optimal range according to Zeiss specifications (6.3x-25x), and is fully usable at least from 4x (which is the lowest magnification I tested; Fig. 16). For optimal resolution, the aperture can be closed only one stop (or none at all above 7x), which makes this lens difficult to use on non-flat subjects (Fig. 17-19). Albeit, this is true of any lens in this magnification range. At 11x, diffraction slightly reduces resolution even with the aperture fully open (Fig. 20), and becomes more visible after closing it one stop (Fig. 21), although DOF slightly improves as a consequence, making the effective resolution more difficult to detect. Therefore, I do not recommend using this lens above 10x.

The Tessovar was tested only with turret rotated at positions III and IIII, and at the minimum and maximum magnifications obtained by zooming at these positions (Fig. 21-26). Zeiss recommends letting the aperture be regulated by the interlocking mechanism with the zoom knob. This varies the aperture from 4 (fully closed) at the lowest magnification of all ranges, to 1 at the highest magnification. However, it is possible to manually override these values, and I tested all magnifications at 1 and 4.

It is evident that diffraction is affecting resolution at all tested magnifications, even with the aperture fully open. However, roughly in the first half of the zoom range, resolution is acceptable when fully open, if you don't need very fine detail. In fact, resolution is qualitatively better than I expected from the horrifically small effective aperture. Beyond this magnification, diffraction becomes quite visible even when you reduce the whole picture to fit onto the screen (about 40% of the original picture resolution on my monitor). The other side of the coin is a higher DOF than in preceding tests (Fig. 24). Working distance is also quite high.

It is a pity that the Tessovar is so difficult to focus at low magnification by using the camera's viewfinder. At magnifications below 1.6x, the effect of diffraction is lesser, and resolution higher. It approaches that of the Luminars below 1x.

Summary of settings

The following table lists several measurements useful if you need to choose or use one of these lenses. All measurements are based on lenses mounted on Nikon PB-6 bellows with the front standard moved to the front of the rail, and the rear standard at the specified distance, as read from the scale on the rail of the bellows at the back of the rear standard (since there are two scales, it felt natural to use the one that increases with increasing magnification). Magnification factors may be slightly approximate. Apertures are as indicated on the barrel. Luminars and the Tessovar indicate apertures as exposure factors relative to fully open, while the S-Planar 74 mm indicates the effective aperture at 1:1. Non-optimal apertures (framed in red) are those at which resolution is visibly degraded by diffraction (but still may be used if a higher DOF is worth a lower resolution). A few empirical tests show that the values framed in red correspond very precisely, at least for the lenses tested on this page, with the apertures at which diffraction effects become visible in test images. I am using this table as a practical companion when shooting with these lenses.

Notes:

(1): values in red are experimentally verified, and slightly different from theoretically computed values.
(2): Topmost range is according to my tests and calculations, second range (in parentheses) is specified by Zeiss.
(3): Position of the Tessovar rotating turret.
(4): I recommend this range because it is difficult to focus at lower magnifications, but resolution is high also at magnifications below this range.

Summary

From the above tests, the following results may be summarized. They deal with all lenses tested on this page, except for the Tessovar. Since the latter is a very different system, it needs to be discussed separately.

- All tested lenses are exceptionally sharp, and free from aberrations. They all out-resolve the camera, when used at an appropriate aperture. There is no absolute winner, although each lens clearly has an optimum magnification range. In all cases, within the optimal magnification range, diffraction is the only factor that lowers resolution to any detectable amount (i.e., to a point where the lenses do not out-resolve the camera sensor). Since there is no workaround that eliminates the effects of diffraction (short of using frame stacking or similar multi-image methods), we can safely conclude that all the lenses tested on this page are "perfect" (i.e., essentially, they cannot be improved to provide visibly better results). They differ very slightly in color rendition and contrast, probably as a result of different lens coatings, which is to be expected of lenses manufactured over a time-span of about 30 years.

- For the tested Luminar and S-Planar lenses, the optimum range of magnifications starts at or below the lowest magnification specified by Zeiss (see above table). At the opposite (highest) end of the magnification range, the maximum specified by Zeiss causes visible diffraction even with the lens fully open. I provide more conservative values of maximum magnification for each lens in the above table. Of course, the limits at both ends of the range can be exceeded, as long as you are aware of the consequences. At low magnifications, these lenses may need to be opened one stop, with respect to the computed optimal aperture.

- The S-Planar 74 mm is specified only at 1:1, but performs exceptionally well throughout the tested range (0.62-2.9x). At magnifications lower than 1x, it should be opened one stop with respect to the computed optimal aperture.

- Working distance, ease of use and the optimal range of magnifications are the only factors to consider when choosing one lens among those tested here. Aside from this, you cannot go wrong with any of these lenses. Within their optimal magnification and aperture ranges, they all out-perform your camera in terms of resolution. Just as a reminder, DOF does not depend on focal length, but only on aperture and magnification, so DOF cannot be a factor in your choice (unless you are considering a special-purpose lens with a fixed and/or very small aperture).

- Any of these lenses will need numerous accessories, including:

• bellows
• a lens mounting adapter
• a precise focusing rack. The one built into the base of the Nikon PB-6 bellows is not adequate to shoot with the bellows in a vertical position. For this test, I used the focusing stand of the Tessovar.
• a light sourcefor framing and focusing
• a right-angle viewfinder, if you use the camera mounted vertically, which really is the most sensible choice.

- The Tessovar suffers from diffraction to a larger extent than the other lenses, because its small maximum aperture causes diffraction to be visible also when used fully open (and as a consequence, also provides a higher DOF than the other lenses). Aside from this, it is every bit as good as the other lenses, in spite of being a 4:1 zoom with a total 32:1 range of magnifications. The Tessovar is a specialized system. If you want maximum resolution in absolute terms, as well as maximum freedom and versatility, do not choose a Tessovar. However, the Tessovar is generally faster and more comfortable to use for long periods of time, and also far sturdier and more durable than bellows, especially if you take photomacrographs for hours on end and/or on a daily basis.

- Focusing at and below 1x is more difficult with all these lenses, than with a modern macro lens. Focusing becomes easier at higher magnifications, as long as you have sufficient light. In general, I would prefer to use a good macro lens, rather than one of the specialist lenses discussed here, up to 1x. I may choose to use the S-Planar 74 mm at 1x, if top performance is an issue. Above 1x and up to about 10x, the unquestionably best choice (as far as picture quality is concerned) is to use the S-Planar 74 mm or one of the Luminars, depending on the required magnification. The only case in which I could see myself using a normal macro lens above 1x is in emergency situations and/or in the field. I would use the Tessovar if a high DOF is needed at the expense of resolution, and/or if I had to take hundreds or thousands of pictures in a relatively short time.