The new Apo-Summicron-M 1:2/50 mm ASPH has not yet rolled off the production line, but there is already a queue eagerly waiting its official arrival. Note that the pre-production samples that are now being commented upon, might not bring the results of the production run. Therefore I wait with comments till I have the official version. By the way: what is big and exciting news in Leica-land is hardly noticed outside its borders. The new Canon Pancake lens 1:2.8/40 mm, priced at about $200 gets probably more attention.
One can prepare for the serious analysis of the ASM50 however by setting some pickets that involve other 50 mm lenses by Leica. Such an analysis can also set the stage for the testing equipment. It is now almost universally the case that the digital camera itself constitutes the test equipment with Lightroom as the arbiter. From a practical point of view this approach seems reasonable: if the camera + software is the limiting factor why search for more? There is still a considerable number of Leica photographers who use film and they have more options to chart the final frontiers of the ASM50.
As a start I used the M9-P with the following three lenses: Summarit-M 1:2.5/50 mm, Summicron (IV) 1:2/50 mm and Summilux-M 1.4/50 mm ASPH (FLE). The test objects were placed at 2 meter distance. I know that the favored distance setting of Henri Cartier-Bresson was 4 meters, but at that distance the sensor cannot record all the fine differences I want to study.
The first test is a picture of the classical resolution chart. The numbers have to multiplied by 40 to get the true resolution. I made pictures at ISO 160 and the apertures 2.5-2.8-4-5.6/2-2.8-4-5.6/1.4-2-2.8-4-5.6. In all cases (all three lenses at their range of apertures) the limiting resolution was between 50 and 56 linepairs/mm (numbers 1.25 to 1.4). There was of course an increase in contrast when stopping down (the SX at f/1.4 obviously had a lower contrast than the other lenses at their widest aperture), but overall the resolution figures were the same. This is not a surprising fact. The sensor of the M9-P has a Nyquist limit of 73 linepairs/mm and I have reported several times that a useable/factual resolution limit should be set at a level that is 20 - 25% lower. And on the other hand all modern Leica-M lenses have optical resolutions way above 50 linepairs/mm, even at the widest aperture. The new Monochrom might possibly constitute the decisive factor to differentiate between these lenses.
A more critical test is the test pattern from that old 1950 test chart. The chart consists of 24 alternating black/white rings around a sharply defined point (good for accurate focusing) and a range of characters (black on white and white on black) in different sizes. This layout is rather critical for digital capture because of the many rounded figures.
The SX at f/1.4 has a medium contrast mage showing 15 to 16 rings and has difficulty with the small white-on-black characters. At f/2 contrast improves to a high level and 16 rings are visible. From f/2.8 to f/5.6 this performance hardly changes, but the increased recording capabilities show moiré-effects in the pattern of the smallest circles.
The Summicron (IV) at f/2 is almost indistinguishable from the SX at the comparable apertures. Apertures f/4 and f/5.6 of the Summicron show the same recording capabilities as the SX, but the moire-effects are slightly more visible at an equally high contrast.
The Summarit lastly has almost identical performance to the Summicron: this is not surprising as the Summarit has a comparable six-element design as the Summicron.
The last test pattern is the well-known Siemens star, used by many imaging programs. The overall conclusion is the same: even the SX at f/1.4 can record the smallest details of the Siemens spoke pattern, but the finest details are blurred with color-artifacts.
The graphical results of the MTF analysis are very important for the final conclusions.
Below are two results: f/1.4 and f/5.6 of the SX 50 mm. The results are based on the raw files without any manipulation. This represents the status for the original negatives in the AgX technology. Note that the differences are most conspicuous in the mid-frequencies as it should be.
Below SX 1.4
Below: SX 5.6
From these preliminary results it becomes clear that a serious comparison between the 50 mm lenses in combination with the recording media is quite a daunting task, that goes a bit beyond the casual comparisons one sees currently on the internet. There is a real danger that the true characteristics of the 50mm lenses are not covered. Results produced by technical means that can rectify subjective impressions are necessary to provide the correct perspective. And one needs comparisons with high-quality AgX emulsions to make an in-depth study of the most intriguing and subtle differences between the Apo-Summicron-M and the other 50 mm lenses attached to the M9 and M Monochrom and the MP or M7.
Since the announcement of the first Summicron design in 1953 there have been seven identifiable main versions: the collapsible original version, the rigid mount version, the close-up version with special goggles (the Nah-Summicron or Summicron-DR), the redesign from 1969 and again a redesign in 1979, that were preceded by the Summicron versions for the R-mount from 1964 and 1975. Disregarding the various mounts and only looking at the optical cell, there are two main design types: the seven-element original design by Kleineberg and Zimmermann and the six-element design by Mandler. It is very well possible to start a discourse about the subtle differences between the many changes in the six-element design (curved versus plane surfaces, changes in glass, improvements in close-up performance, better micro-contrast). The original Mandler construction can be regarded as one important design type.
Summicron lenses have always been designed against the background of the main photographic Zeitgeist. The first Summicron design was created when color slide film was the preferred capture medium. The design was carefully optimized for the rendition of the subtle color shades and nuances that the slide film was capable of registering. Because of the disturbing light fall off when projecting slides the Summicron had a low amount of vignetting. The design was restricted by the then-available type of glasses, the state of the art of optical design knowledge and experience and limits of the manufacturing technology. The famous debate between the proponents of German lenses (high resolution, good color correction) and the Japanese lenses (high contrast, stark color rendition) may be directly traced back to the fame of the Summicron lens.
The second main type, the six-element design responded to the demands of the new generation of black-white emulsions that combined high acutance with extremely fine grain. Computer-assisted design programs had vastly extended the possibilities of optical designers to explore many different design constructs. Modern statistical and tolerance techniques made it possible to fine tune the design to the limits and requirements of the production engineers. One could draw upon an extended catalogue of glass types to find the best solution.
The Summicron design from 1964 was one the very first high-contrast acutance optimized lenses that incorporated the new insights based on MTF analysis where it was found that a sharp cut-off at the limiting frequency gave the best image quality.
The current Summicron-M 1:2/50 mm and its sibling, the Summarit-M 1:2.5/50 mm are the direct descendants of that 1964-design. The modern Summicron lens has capitalized on the improvements in manufacturing of precision mounts, the new methods of coating of anti-reflection layers and lens grinding and polishing. One can see however that its design is deeply rooted in the optical possibilities of silver-halide emulsion technology.
The new Apo-Summicron-M 1:2/50 mm ASPH. (ASCR50) designed by Karbe is the apt answer to the different and profound challenges posed by current and future digital capture technology and the highly evolved silver-halide emulsions, from microfilms to the newly upgraded Kodak T-Max 400.
The classical benchmarks for lens performance, like resolving power, vignetting, color temperature are rapidly losing their relevance and defects like color fringing and focus shift are moving into the foreground. The reason for this shift is the replacement of ‘thick’ (15 to 20 microns) emulsion layers by zero thickness solid-sate surfaces of light sensitive elements. The precise location of the optimal sharpness plane of the lens could be established with a margin of some 20 microns (0.02 mm). The circle of confusion of 0.033 mm has to be added to this value and all tolerances in the camera system (rangefinder accuracy, film register and distance from film gate to bayonet flange) could be related to this sum (about 0.05 mm).
These two ‘safety margins’ have disappeared in the digital capture technology. The sharpness plane of the lens has to coincide precisely with the location of the surface of the sensor in order to exploit the inherent image quality of the optical system. The higher the performance, the more critical the match between sharpness plane and sensor surface. The surface plane is a ‘plane’ only by convention. In reality the lens will project a curved surface onto the capture medium (film and sensor). The lens elements are curved (even the aspherical ones have curvature) and this property implies that the image projected on the medium will be curved too. The combined effects of (sagittal and tangential) astigmatism and curvature of field can be smartly used to counteract the tendency for curvature, but it will in most cases not be perfect. Note for instance the MTF graphs for the Summicron-M (SCR50) and look at the dip in the curves in the range 12 - 16 mm of total image height.
The ARCR50 had to fulfill four very critical requirements: a truly flat image field, an accurate location of that field when attached to a camera body, no shift of the image plane when stopping down (focus shift) and minimal shift of the image plane when changing focus from infinity to closest distance. The first, third and fourth requirements are part of the optical prescription and explains the elaborate layout with eight elements, aspherical surface and floating group. The second requirement, the incorporation of the floating group and the very high image quality explain the very tight tolerances and the high precision of the engineering and manufacture that are needed to hold the tolerance.
The performance of the lens is flawless. For what it is worth, the ASCR at full aperture (f/2) is able to resolve 160 linepairs/mm on microfilm, developed in Spur chemicals. The lens therefore has a large quality margin for use with the M8/9/M-E/Mono with the 6.8 microns pixel size (70 lp/mm) and the M with 6 microns pixel size (80 linepairs/mm). The excellent flatness of field can be seen in the homogeneous performance from centre to the corner. The SCR50 has a visible dip in performance in the area from 12 to 16 mm of total image height (21.6 mm) as predicted by the MTF.
Color fringing (again quite visible in pictures made with the SCR50) is impossible to detect when using the ASCR50. Even in strong and steep contrasts between dark and light (as dark branches against the clear blue sky) the subject delineation of the ASCR is crisp and clean. The often noted purple fringes at dark/bright borders is in many cases the result of overexposure and the reaction of the processing software to this phenomenon and not the result of the existence of the secondary spectrum.
There is no focus shift detectible when using the ASCR50. And the coincidence between exact rangefinder focus and the optimum image plane is spot on. The M with the Live View option and the 10x magnification provides an easy check and quite elaborate testing with the M, M9-P, M8.2 and M Monochrom did not reveal any discrepancies over the full range of distances on all cameras used.
The performance of the lens hardly improves when stopping down and stays on an equally high level from f/2 to f/11 with a slight drop in contrast at f/16.
There is a slight vignetting in the corners at the wider apertures which may be a bit surprising as the lens detection on the M Monochrom and M do recognize the ASCR50
The close-up performance (remember that close distance focusing starts at about two meters and is not restricted to the 0.7 meters distance) is outstandingly good. Performance over the full image field is very high and focus shift (visible in a severe drop of contrast) is not detectible.
The SCR, while in itself a high contrast lens, has a certain propensity to flare. The ASCR50 on the other hand is not only flare-free, but does suppress the internal reflections and the veiling glare on small areas totally. This behavior is not only the result of the mechanical mount and its internal baffling structures, but mainly to the reduction of residual aberrations. Indeed these two characteristics of a highly and unusually sophisticated reduction and control of optical aberrations and the hitherto unseen level of engineering precision and accuracy of manufacture (the main reason for the delay in the delivery of the ASCR50) give the lens an unique fingerprint. On casual inspection the SCR50 and the SX50 may perform rather similar at comparable apertures, but when looking at the ASCR50 pictures with a sagacious and disciplined gaze the pure clarity of the overall image and the crisp rendition of fine detail, even in the near-unsharp areas are stunning and provide a dimensionality and life-like rendition that is unequalled.
The experience of the lens can be compared to tasting an outstandingly good Scotch whisky: once you notice the quality and superior flavor, you are lost and you do not want to settle for less. This will surely happen to anyone using the ASCR50 for some time.
The Apo-Summicron-M 1:2/50 mm ASPH. is the best standard lens for the Leica rangefinder camera, film-loading and sensor-equipped. It is easy to become convinced of this claim when one does base it on looking at pictures and examples on the internet. We should however be aware of the mechanics of the psychology of vision. We assume that seeing is believing, but in fact the reverse is true: believing is seeing. What we want to see is that what we see. And we should address the very tricky problem of visual comparisons and the manifold ways that a setting intended for comparisons introduces errors or looks at the wrong aspects. When comparing pictures made with the ASCR50 and another lens, both mounted on a digital camera (the preferred option currently) we are not looking at the characteristics of the lens, but at the system performance of the lens/camera combination. Here many conditions that may influence the result may stay unattended. As example: if you focus on an object with two different lenses (stationary object and camera on tripod) the chance that the distance is accurately set in both attempts is zero. A 5% margin of tolerance is always possible and that difference may ruin the comparison.
The MTF measurement is accurate and objective and most importantly evaluates the lens in itself. It is still the best measure for the optical performance of the lens. The practical implications of the MTF results are a different matter. The importance of the MTF results has been a bit exaggerated over the years as the correlation between MTF and practical imagery (a word that Mike Johnston hates) can only be established indirectly.
As a benchmark for evaluation purposes and optical quality, the MTF graphs are still unrivaled, but one should be careful not to read too much out of the numbers.
The MTF results are always presented for the tangential and sagittal image planes (better curved surfaces). As can be seen from the image below, the idea of a flat plane surface is a theoretical construct. In reality there are three differently curved surfaces that join each other at the central axis of the lens. (P = Petzval surface, S = sagittal surface, T = tangential surface.) A small detail of the object will be focused more or less unsharp, depending on the surface it is projected onto. Because most objects in real space have depth, we will have a different depth impression when looking at the reproduced detail. The MTF graphs show nicely the difference in these curves, but one should realize that we are mostly looking at a mix of both representations.
Modern Leica lenses are very precisely manufactured with tight tolerances because of the fact that the plane of sharp focus is very precisely defined and because of the fact that the throw of the focus ring is quite small: a slight shift of the focus ring will have its effect on the sharpness.
The MTF diagram below compares the two best standard lenses on the market. The difference in performance is not that great, but it is clear where the ambition of the designers lies: the highest possible contrast transfer at the 40 linepairs/mm limit. Zeiss has set the target very high: a contrast transfer of 70% for the 40 lp/mm for a high speed lens at f/2 is rather ambitious. The Zeiss designers did succeed as the graphs show, but at a price: the lens mount is very substantial in size. Attach it to the M and the camera is dwarfed.
The performance of the ASCR50 at f/2 is equally good. Note in particular the very homogeneous quality over the whole image area. A difference of 10 percentage points on this level is most likely not noticeable. Because the ASCR50 will be used preferably on the sensor-equipped Leica M bodies, the behavior at the 80 lp/mm is quite interesting. The contrast transfer is above 20% overall and mostly above 30%. This is a major accomplishment and well above the Summicron-M 50 mm.
The size of the ASCR50 is quite small and here lies the true difference between this lens and the Distagon. Leica has always claimed that making a good lens is not a problem nowadays, but making a good small lens is. Indeed, it has to be noted again and again: the superior performance of the Leica M lenses in combination with the s all size is the true accomplishment. The graphs show how well they have succeeded.
A very important characteristic of a new lens is its correction of the secondary spectrum. Normally lenses are best corrected for the green and red part of the spectrum, but the blue and purple part is neglected. Below is a table of the wavelengths:
380 - 440: purple
440 - 483: ultramarine
483 - 492: iceblue
492 - 542: sea green
542 - 571: foliage green
571 - 586: yellow
586- 610: orange
610- 780: red
780- … : IR-A
Because of the relative insensitivity of silver-halide emulsions for the blue region, most designers in the past assumed that one did have to pay attention to this part. This attitude has been referred to as the 'cursed 435 wavelength'. Even a talented designer as Dr. Mandler evaded this wavelength. In the diagram below one can see the strongly under corrected blue part of the spectral graph. The modern approach is to pay much attention to this wavelength as it is responsible for the blue color fringing, so visible in digital photography. The Apo-Summicron-M shows the much improved correction of the blue region. The correction is four times better than what one should expect normally. As a comparison, note that the new Distagon is equally well corrected.