State of the art of medium tele-lenses
What to expect from current designs from Zeiss, Canon and Leica?
Recently Zeiss introduced the Sonnar T* 2/85 ZM and Canon the EF1.2/85 L II USM. Once upon a time the 85 to 90mm focal length were among the most important lenses in the armory of the serious photographer. High-speed versions of this type of lens were preferred tools for photojournalists and portrait photographers. Nowadays this focal length is habitually included in the standard zoom lenses, ranging from 24-05 or 18-200 or 70-300 or 28-90 and so on. The special status of the medium tele-lens has been eroded. So why should one wish to acquire a separate prime lens?
For M-mount users this is not really a question: there is no choice but to buy single focal length lenses in this range of medium tele-lenses.
For Canon EOS users the situation is different: a prime lens would only make sense if it offered additional value. The high speed is one aspect: it is not always fun to use a 5.6/90mm as the fastest tele-lens you have. And there is that somewhat elusive discussion concerning the dedicated use of a single focal length as a tool for creativity. A zoomlens with its range of viewing angles might distract from the premeditated creative ideas.
Around 1880 there was a group of Dutch poets who wanted to deviate from the norm and wanted to renew and invigorate the poetry of that day. The new movement was called 'Tachtigers' or "Movement of (Eighteen-)Eighty", after the decade in which it arose. The 'Tachtigers' insisted that style must match content, and that intimate and visceral emotions can only be expressed using an intimate and visceral writing style. The insistence on beauty and style can be translated to the realm of portrait photography where the medium tele-lenses are the preferred lenses because of the pleasant perspective when using the lens for a portrait filling the whole negative area and the smooth background reproduction that adds plasticity to the portrait. The background blur, in the past a simple attribute of the lens, has become the source of heated discussions since the discovery or, some would say, introduction of the concept of bo-ke(h).
There are also technical/optical issues. A zoomlens is often much heavier and larger than a prime lens and its speed is generally lower and the shortest distance is not that close: many zoomlenses stop at a close focus distance of 1.20 or even 1.40 meter and then the use of a 70 or 90mm lens is not that much fun.
The big advantage of a zoomlens is the flexibility of the moving lens elements that offer enhanced possibilities for correction.
The best primes now also incorporate so-called floating elements to give better imagery at the close focus range. Both the Canon and the Zeiss lenses have this feature. The designation of ' floating' of the element is not correct, however. In fact it would be best to say that the distance between lens elements is floating: in both solutions (Zeiss and Canon) the last element is fixed and the rest of the lens group moves when turning the focus ring. The Leica Apo-Summicron-M 2/75mm ASPH employs a true floating element: the rear element moves relative to the movement of the front group. The goal however is identical: to change the distance between the last element and the rest of optical group.
This report is focused on the new Zeiss and Canon lenses, but I took advantage of the availability of these lenses to compare them to the Leica lenses in the 75-90mm range. Canon states that the 'II' addition in the nomenclature of the 1.2/85 indicates that the lens has been optimized for use on the 5D, which I happen to have. So I could compare the performance of the 1.2/85 in a film-based contest between Zeiss and Leica and an analysis of possible quality differences between film emulsion image capture and the solid-state sensor image capture when the same lens is used with both technologies.
Canon EF 1.2/85 L II USM
Canon has always offered high-speed lenses in this class of focal lengths. The first version was introduced as FD type in 1976 for the F1 camera. With eight elements, an aspherical surface and a floating distance before the last element, which was stationary the designers integrated in one design a host of modern technologies. The image quality was very good (the BJP noted that this was the best lens in its class of high speed lenses). Especially the reduction of flare and secondary reflections was of such a high order that Canon did not offer a hood for this lens. In 1980 the lens was renamed as New-FD 1.2/85 L and the post-fix 'L' signified exceptional quality.
The creation of the EOS series gave birth to a third version of the lens, now with AF added and designated as EF lens. And now in 2006 we have the fourth version with the 'II' added as a Mark 2 version. The lens is said to have been redesigned to exploit the sensor characteristics of the EOS cameras with a 35mm size of sensor. The lens diagrams for all four versions show a great resemblance, with one difference when the lens changed form FD to EF; lens elements 3 and 4 are now separated and some other elements have their radius changed. One can discuss these changes at length, as is the habit in circles of Leica aficionados on the assumption that these changes signify a major change in lens properties.
One of the main reasons to employ a very high-speed lens (now that you get have noise-free ISO 800 images) is the creative use of the narrow plane of sharp focus. The picture below indicates that the sharpness plane is indeed narrow, but also that the very smooth unsharpness gradient creates images with depth. Picture at f/1.2.
One should approach such changes in a more prosaic matter: different glass types (as example the migration to lead-free glass) or a change in production to hold on to optical tolerances are often the reason for such a change, while keeping the previous level of performance. One can read too much into the lens diagrams!
A look at the MTF diagrams does indicate that the performance differences are small if at all detectable. The FD seems to have the edge in contrast, but the EF has somewhat better corner performance. Canon says that the new mark 2 version has been optimized for use with the digital cameras. One can read everywhere that the Canon sensors with 35mm size are quite prone to strong vignetting in the corners due to the acceptable angle of the micro-lenses in the outer zones of the sensor area. A simple comparison does show that this is one of the many mythical stories floating around on the internet and in the many photographic magazines where the need for actuality overshadows the need for depth in the analysis.
I used the EF 1.2/85 on the 5D (card-loading) and the EOS 33V (filmloading) and set both cameras before a wall and made comparison pictures at all apertures and the same distance. The vignetting at wider apertures was identical in both capture methods. Only in the close distance pictures (around 1 meter) the 5D images showed more vignetting, but even here the differences are not that great.
Pictures Canon EF 85/1.2 left = film capture, picture right = cmos capture
Zeiss Sonnar T* 2/85 ZM
Zeiss and Voigtlander (now Cosina) jointly have developed the new ZI camera: a derivative of the Bessa series of cameras, but with the look and feel of the classical Zeiss Ikon Contax and a number of design elements developed by Zeiss in Oberkochen. I leave it to the aficionados to discuss the true heritage of that body. It is of some historical interest to note that originally Zeiss and Viogtlander were competitors in the period that the German photographic industry was truly global as we would say today. Voigtlander created many fine camera sand lenses, but was in the end merged with the Zeiss conglomerate of camera companies. When Zeiss Ikon closed its doors in 1972, the brand name of Voigtlander was sold to a German businessperson and later acquired by Cosina. Now Mr. Kobayashi is in true Zen fashion waiting for the digital storm to settle down and make his decisions about the direction of the Bessa family lineage.
Most new Zeiss-designed lenses are made in the Cosina factory, but not the Sonnar 2/85mm lens. The design requires a level of manufacturing quality that Cosina does not (yet) possess and this lens (as is the case with the 2.8/15mm) is hand assembled in Oberkochen. Here we see a glimpse of the differences in manufacturing technology and production quality between Zeiss (and Leica) and Cosina. The Cosina lenses are very good and quite reasonably priced, but to keep the costs to a level that profit is still possible, one cannot demand from these lenses the utmost in manufacturing quality and material quality: hold the lens surface tolerances to the smallest possible amount, use the most demanding glass types and assemble the elements in mechanical mounts with microscopic precision.
The products of the German companies are expensive because they add qualities into the product that are costly to manufacture. The production of the Zeiss lenses made by Cosina is supervised by German engineers, an indication that the engineering demands and the manufacturing philosophy are different between the normal Cosina production and the Zeiss manufacturing know-how and tradition. Whether the working photographer needs or benefits from these properties is a personal choice.
The Sonnar is designated as a 85mm lens and this might put off Leica users who with reason assume that the framelines of the 90mm frame will be too narrow for the 85mm to use it with confidence. Comparison pictures between the Apo-Summicron 90mm and the Sonnar 85mm show very small differences in viewing angle and one may confidently use the Sonnar on a Leica body, using the 90mm frame. In fact the designation of '85' on the Sonnar lens is more nostalgia than it seems. The true focal length is more close to 87 or 88mm and thus very close to the Summicron where 90mm may be 89 or 91mm in reality. Real focal lengths may deviate by a few percentage points from the designated focal length in practice, where these small differences do not count.
Leica Apo-Summicron-M 2/90mm and 2/75mm ASPH.
The Leica lenses are not an integral part of this testreport, but are always in the background for comparison purposes. Both lenses are the ultimate in their class and it is most interesting to see whether the new (Sonnar) or redesigned (Canon) lenses are a challenge for or even surpass the Leica performance. In previous reports I noted that the ZM lenses offer a performance that is close to or even better than the Leica lenses, where 'better' is of course to be taken with some perspective in mind: resolution is not the issues where the lenses differ most. It is the flare reduction where the designs do really differ and the size of the lens.
Most users of the Leica and Zeiss lenses (but the same is true of the Canon and Bessa users, but they are not so nitpicky about it) ask themselves about the real reasons for the differences in performance and relate these to what is known in the public domain. One of the notions that seem to explain quality differences between lenses is the use of aspherical surfaces and these have assumed mythical proportions. Any lens with one or more aspherical surfaces must always be better than any lens without these intricate surfaces, seems to be the common denominator of lens discussions.
The facts show a more complicated situation: aspherical lenses as such do not imply by default a better design. One has to add some parameters. When one compares the Leica M and the Zeiss ZM lens designs, there is one important difference: size! Size has two dimensions length and thickness or diameter. Leica lenses by tradition are designed to be very short physically in relation to the true focal length. A few millimetres already make or break the design. The compact build of Leica lenses in combination with superior performance does require the employment of aspherical surfaces. The Apo-Summicron has a length of 78mm and the Sonnar measures 82mm. For a user the mere 4 millimetres may not be significant, but for a designer it is important.
As a general we may say that the requirement of a compact design necessitates the use of aspherical surfaces and if one can be more lenient in the physical dimensions an aspherical surface is less important. The often cited rule that the employment of aspherical surfaces reduces coma and other unwelcome aberrations is true, but only to a certain extent: the EF 1.2/85mm has more coma than the non-aspherical Sonnar, but here one has to relate to the maximum aperture. The upshot is that one needs to see all aspects in a meaningful relation and not isolated.
The performance (on film and on memory card)
The Canon lens has at full aperture a high overall contrast and a slightly soft definition of quite fine details, especially in the outer zones of the images (from image height 12mm). Night pictures of scenes with strong light sources show coma effects and some halo effects around bright lights. Vignetting is more than one stop in the corners. The small jump to 1.4 does not bring visual improvements and the exposure meter does not react to this half stop closure of the diaphragm, itself an indication that the difference between the nominal 1.2 and 1.4 are effectively smaller than indicated. At f/2 the contrast increases visibly and now very fine detail is crisply rendered in the centre of the image and a slight softening at the edges. At f/2.8 the crispening extends to the corners and the lens as it were powers up and from f/4 the quality is not distinguishable from the colleagues with a slower speed. The close up performance is very good and even at f1.2 the detail definition is excellent. Vignetting is more pronounced now, but gone at f/4.
The Sonnar at f/2 has more punch than the Canon at 1.2, which is not surprising, but at f/2 the Canon is still a bit softer. It is not so that the Canon has lower resolution, but the fine details in the case of the Sonnar lens show a more powerful delineation. It is in the mid-frequencies where the Sonnar has the edge with a high edge contrast. Coma cannot not be detected in the Sonnar lens wide open. Vignetting is in the same order of magnitude. Close up performance is again excellent with crisp definition of fine detail, but with a less pronounced reproduction of very fine detail (the Canon lens is not better in this respect).
Stopping down does improve the overall performance and aperture for aperture you can see more fine details emerge from the background noise of the film grain structure. The Sonnar then is the best lens in this respect, at least up to f/4. After that aperture both lenses perform equally well. The old rule that a very high-speed lens cannot be as good as lesser speed lenses at smaller apertures is again a myth of the past.
The Apo-Summicron 90mm wide open brings imagery that is equal in definition of very fine details compared to the Sonnar, but the fine details have more depth and vibrancy. Here the apochromatic correction weighs in the reproduction quality. This correction covers the whole visible spectrum, where the Sonnar has a less pronounced correction in the blue part of the spectrum, visible when you take pictures of details against a clearly differentiated background colour: small blue fringes may be visible in the Sonnar case. Even more so in the Canon case. In the close range area, the Leica lens is less good and here the Zeiss and Canon show the better imagery. The Leica lens is a bit soft as can be seen in the enlarged parts of the models eye. I did not apply any colour corrections to the pictures and the scans came out quite reddish, based on the Fuji Velvia film used.
The Summicron 75mm has the best performance in this department, but compared to the 90mm lens from the same stable has the advantage of two additional elements and a true floating element construction. Seen from this perspective the Apo-Summicron 90mm with a mere five elements shows an amazing quality, but it is definitely designed as a lens optimized for the medium distances where one would deploy a 90mm reportage lens.
right first row left to right: Summicron 75, Summicron 90, second row left to right: Canon 85, right, Sonnar 85
Of course we can hear the muttering of the photographers who claim that a slight softness at close distances is best for portrait photography where razor sharp images are not required. I humbly dare to disagree: clean and crisp imagery is always the best option for photographic purposes. Portraits made at closer distance with wide apertures thrive on the clear differentiation of sharp and unsharp areas and this is best seen when the sharpness plane is as clearly depicted.
The comparison pictures do show the differences in background blur between the Sonnar and Leica designs. The Sonnar shows a more pleasing (that is less distracting) background. The Summicron picture has the better definition of background details, thanks to a higher level of correction in the sagittal plane. It is a personal choice which one is preferable. Here analysis must stop and subjectivity reigns.
The control of flare and secondary reflections and veiling glare is notoriously difficult to test in a meaningful way. It is not easy to find a situation where these aspects can be compared and the results are often unpredictable. And a lens that shows good flare suppression in one situation may be quite bad in another situation.
The following series of pictures gives a good impression of what the photographer can expect when taking pictures in backlit scenes. The Zeiss Sonnar and the Leica Apo-Summicron 75mm are best, with the Canon in second place and the Apo-Summicron 90mm showing a fair amount of reflections in the lower part of the scene. This lens on the other hand shows very good control of veiling glare as can be seen in the background contrast of the farm: here the retention of contrast is excellent. One could debate the rank of the Canon EF 85 lens. The secondary reflection is not that strong, but quite large and the contrast in the background part of the scene is very low. All pictures made at f/2.
The Canon is not absolutely flare free and now the company provides a huge hood, that should be used whenever possible. Especially when you take pictures in the sun when the sky functions as a large uniformly lit dome, there can be a strong veiling glare, that damages the picture. It is an illusion to assume that any lens is flare-free under all circumstances. Given the wide aperture and the big front element, the Canon EF 85mm is commendably useful in backlit situations.
The medium tele lenses are not only well suited for studio portraits, but they excel in street scenes too. The next series of pictures are designed to show several aspects. The Zeiss 85 and Leica 90 pictures are quite comparable with a small advantage to the Zeiss lens because of the better eyelash detail. These pictures do already show the deterioration of the scan process: the highlights are not clean and grain is colouring the white part to a muddy tone. The original slides produce a picture with much more sparkle and punch. Both lenses keep the contrast eve in this backlit situation.
The Canon 5D picture is shot at f/1.2 and ISO 400. The use of the 1/6000 of a second improves the detail definition as you can freeze the movement of the walking model. Note too that the texture of the skin is very smooth: this is the advantage of digital capture or its disadvantage; depending on the position you take. Some claim that the exceptional smoothness of the digital file detracts from the realistic impression that film provides. Others say that the smoothness brings medium format quality to the 35mm worker.
It is interesting that in this type of scenery, the much discussed wider dynamic range of the digital capture is not visible. The highlights in the Canon scene are as washed out as is the case with the filmbased pictures.
Definition of fine detail and the impact of the low-pass filter.
When using film as a recording medium the lenses reviewed here perform within the limits of the imaging chain. The film is the limiting factor and the lens can not go beyond this limit, even there is a higher optical potential for definition. When using the same film, in this case the Fuji Velvia 100, the ranking is Apo-Summicron 75 and very close on its heels the Sonnar 85. Then we have the EF 85 and the Apo-Summicron 2/90. It is a bit unethical to compare an f/2 with an f1.2, but I was surprised how good the EF lens 1.2/85 holds itself compared to the Leica 2/90mm. The fact that both these lenses can be compared without a clear loser says something about the quality of the EF lens. On the assumption that this level of performance is identical to that of the 1976-version, we have to say that in those days the image quality should have been superb. But then we did not have the films to exploit it.
I took the chance to compare the performance of the Canon lens when recording the same scene on film and on flash card. There is some discussion that the employment of the low-pass filter in the Canon 5D which theoretically does limit the maximum resolution (definition) cannot keep pace with the true resolution of film where fine detail can be captured without restrictions
Some additional comments are needed here. The low-pass filter functions as a signal processor where the high frequency range of a signal is cut off. This is done to remove the aliasing effect where the high-frequency components of a signal are transposed to low-frequency components. As long as the information content of a signal is not reduced, this is acceptable practice. In addition we should also note that even if the information content is not reduced, the removal of the high frequency range of the signal will reduce the edge contrast (micro contrast) of the detail as the square edges of a signal are rounded off. Part of this reduction can be compensated in the post processing of the signal where the original content can be reconstructed. This is the main difference with film where the loss of detail through light scattering in the emulsion cannot be regained.
I took a series of pictures of a stone wall with quite small detail in the texture of the concrete between the stones. All pictures on Velvia 100 at f/2. Canon with AF, all others manual CRF.
Top row left = Sonnar, right = Summicron 90
Bottom row left = Summicron 75, right EF 85
The results give the same ranking as before: Sonnar 85 and Summicron 75 show the best definition and the S75 slightly ahead with higher edge contrast. The information content is the same however and some might say that the Sonnar definition is a bit more pleasing. The lower contrast of the Canon picture may be caused by the AF function, but generally the AF on the Canon is quite accurate. But there is that soft voice whispering in your ear: I told you! manual CRF focusing is more accurate.
The Velvia slides were scanned with the Nikon Coolscan 5 ED and every slide was individually focussed. The scan resolution is 400 dpi/inch or a six-micron pixel size. The Canon 5D sensor has an eight-micron pixel size, so the level of fine detail should be higher in the scanner case. Visual inspection of the slides under the microscope revealed somewhat finer detail to be seen and also the overall contrast was higher with more sparkle in the slide than can be seen on the screen. The film operates at its own resolution limit here and I doubt whether a better scanning device could extract more information form the film. The scanned files (images) had some post sharpening applied, but the same amount for all slides.
The same scene was photographed with the 5D and the Canon EF lens. Here the Raw images were processed with PhaseOne Capture software with some sharpening applied in the developing stage. No other manipulations were done. Below are the results for the film scene and the sensor scene at f/1.2. In addition there is the scene at f/2 with the solid state sensor capture and the film recording.
The differences between the images at f/2 are quite visible, a clear indication that the post-processing software can extract more detail definition from the scene than film can. At f/1.2 we see a small improvement in detail definition, but here the optical capability of the lens sets the limits and even good software cannot create detail where there is none.
This comparison covers a wide range of topics and lenses. The obvious candidates for comparison are the Zeiss Sonnar T* 2/85 ZM and the Leica Apo-Summicron-M 2/90mm ASPH. The conclusion is not as simple as it seems to be. The frugal design of the Summicron (five elements and one aspherical surface) brings outstanding results of apochromatic signature in the medium distance range till infinity. Here the lens is unbeaten and probably unbeatable.
The Zeiss design is at infinity very close to that level of performance, but has somewhat more colour fringing. The Sonnar excels in the close distance range and has the better flare suppression. It also has lower astigmatism, giving the outer zones of the image a more contrasty look when looking for fine structural details. But the Sonnar has more bulk and does not feel that rock solid as the Leica lens. Horses for courses, I would say. The Apo-Summicron-M 2/75 ASPH shows what Leica can do when all stops are pulled. It is a general-purpose lens of outstanding quality with some very slight weaknesses that hardly interfere in the quest for best imagery.
The Canon EF lens has amazing performance when the high speed is taken into account. The 1.2 aperture can be used with confidence and at smaller apertures the lens is as good as any high quality medium tele lens and even better than the 1.8/85 from the Canon stable.
The bulk and weight are really over the edge and this does limit its usefulness in handheld situations. You get tired after a longer shooting with this lens.
That same lens brings additional quality when used in the digital workflow. In scenes with extremely fine detail as seen in the stonewall picture the advantages are not that big, but where the demands are less exacting as is the case with the portrait (see the eye samples) the advantages are evident.