Archived posting to the Leica Users Group, 1998/08/14
[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]- --============_-1309015588==_ma============ Content-Type: text/plain; charset="us-ascii" Please note that I am happy to share my testreports and all the information contained in them to all members of the LUG. The report however is the result of considarable research and must be considered as a document protected by copyright law. Marc can ceratinly help here. So if you cite from, or use information contained in these documents, please respect the notion of intellectual property and refer to the original document or ask permission to copy parts of the text. Before embarking on the lens test itself, it might be nice to state the current thinking about lens design and evaluation. ITHis text might be usefull to all new LUG-members. In the early fifties and sixties the ultimate object of desire for any 35mm photographer, who liked or needed to practise the art of the artless snapshot (HCB-style) was a standard lens with an aperture of 1,4. Any additional photon that could be captured on emulsion while shooting in "available darkness" was most welcome. The slender depth of field at that large aperture added quite often impact and drama to the image. The need for such a large aperture became imperative after the candid pictures of Erich Salomon of Ermanox-fame. The 35mm worker got what he demanded in the early fifties as coated 1,5/50 designs from Zeiss and Leitz became available. These lenses indeed did capture some of the additional photons. But the photons on their way through the many glass elements wandered around (ab errare). Aberrations abounded and the image quality, to be polite, was just acceptable. The Leitz Summilux 1,4/50, introduced in 1959 for the M series camera was the first to offer a higher level of image quality. A redesign , offering very good quality, was introduced in 1962 and is still in production. The Leicaflex user had to wait till 1970 before she could capture scarce photons. Again a redesign in 1978 improved the quality. In the meantime the emulsion technology made some quantum leaps in speed/granularity relationship and the ubiquietous electronic flash lessened the need for high speed optics. Some even predicted the demise of this type of lens. More so as the 50mm fixed focus lens is nowadays often being replaced by a 'standard zoom' of 28/35 to 70/85 focal length. The design of a 50mm high speed lens is quite a challenge. Its sibling, the 2/50mm, offers image quality of the highest calibre (at least in the Leica stable). And the optical aberrations to correct are quite stubborn. Most reviewers of high speed lenses even today will tell you that a 1,4 design is a compromise. What then is the optical problem? Any lens produces a circular image area within which the 24x36mm format has to fit. This circular area can be divided in three parts, the center, the zonal area and the farout zones. The center (or the paraxial zone or Gaussian zone) is quite easy to compute. The zonal areas are more difficult to correct. Optical aberrations have the habit to grow disproportionately if the apertur and/or the field-angle become wider. Many aberrations grow with the square root or the cubic root in relation to the aperture diameter or even more. OK you would say, lets settle for a bit less image quality in the corners. The snag however is this: the zonal aberrations have a strong influence on the performance in the center. Moreover: when stopping down the effect on some aberrations is not reduced. The combined result of all aberrations is always a reduction in contrast: a softening of small details and a low overall contrast. The lens designer's plight however is not over if he succeeds in reconciling all these conflicting demands. Aberrations can be classified as third order, fifth order and seventh order aberrations and so on until the n-th order. (I will explain in a separate post why they are so designated). Third order aberrations are large and suppress all other aberrations in the series. If a designer can tame these third order errors, he will be unpleasantly presented with the next in line. Balancing third order aberrations require often a change in focus position. The well known statement that you can compute a lens for high contrast or high resolution ultimately boils down to this kind of balancing. Fifth order aberrations are mathematically quite challenging. The high quality of Leica lenses is based upon an excellent grip on this group of aberrations. As usual a balancing of conflicting demands and finetuning of parameters is needed to compute a lens to this very high level of correction. But choices are inevitable. If the designer has done her best (actually some of Leicas best optical designers are female) she would be lost in space if the manufacturing department could not support her. If a (hypothetical) lens design needs to satisfy 100 parameters, more than 50% of these would have to be fullfilled by the manufacturing department. Modern Leica lenses and their exquisite quality would be unthinkable without the control in the production line. This is a not well known fact: the designer is nowhere without support from the manufacturing guys. But the designer is restricted in more ways. A lens has certain physical dimensions. If you would free a designer from the physical constraints he can perform wonders. The famous lenses for the Contarex series followed this paradigm: whatever the physical dimensions, the optical performance may not be jeopardized. Result: ergonomically the lenses were often hardly usable. So the designer needs to serve very many stern masters. The relative neglect of the 1,4/50 since more than 20 years might be the consequence of all these considerations. Now Leica has introduced a brand new 1,4/50 for the R-series. Any tester will be challenged to assess its performance against the background I just outlined. The ASPH riddle. Leica has introduced in recent years several lenses with one or two aspherical surfaces. Generally the image quality of these designs is quite high, to say the least. Some observers of the Leica scene have erroneously concluded that the equation aspherical=high image quality now has universal validity. Some even went further and deduced that any lens design without an aspherical surface can not be designated as a modern design. Occasionally one will hear or read the statement that for example the current Summicron 50 lens for the R series is an old design and needs or will be superceded by a newer design of invariably aspherical signature. It is easy to be charmed by such reasoning. This assumption,however, is not correct. First some facts. One: the new Summilux-R 1,4/50 (subject of this report) has been designed with conventional means. We can then conclude that the Leica designers could realize the required level of image quality without recourse to aspherical surfaces. If the use of asphericals would have been advantageous for the state of corrections and the production requirements, Leica certainly would have incorporated them. Two: asperics are not always the best way to go. The Ricoh 28mm uses two aspherical surfaces but its image quality is below that of the Leica 28mm and the Zeiss 28mm for the G-series, both without asphericals. Three: all Zeiss lenses for the G-series are quite recent designs and none of them has any aspherical surface. The image quality of these lenses is beyond any reasonable doubt. One of the main reasons for employing ashericals is the correction of spherical aberration and attaining a flat field free of astigmatism. But the use of asphericals may also affect other aberrations in a dangerous way, especially if the distance between diaphragm position and asperical surface is relatively large. The design of an optical system must always try to balance many demands and variables, some of which are optical and some of which are manufacturing oriented. It could be that a designer tries to incorporate an aspherical surface only to find out that the strain on production tolerances is too heavy. He also could note that given the overall configuration of his/her design, the aspherical surface has no added value, or even will enlarge or introduce other aberrations. Note that any optical system must be regarded as a delicate whole of carefully designed and matched components. Note also that all aberrations act on every image point in conjunction. Note further that 'image quality' is not a fixed set of parameters. Zeiss will adjust the balance of corrected aberrations and the magnitude of corrections according to different rules than Leica does. Leica may conclude than given the required correction in some cases asperical surfaces are justified and in some cases not. The question of old designs. I have no idea when and why a lens can be designated as 'old'. The current Summicron and Summilux 50 and 75/80 lenses (include also the current Zeiss Planar lenses) are all variants of the double-Gauss type of optical classification. This design is now almost 100 years old. The general formula of a lens may be 'old'. Important is however not the general design but the state of corrections. A small change in curvature, different location of the diaphragm, different glass types and small changes in the distances between lens elements may alter the image quality quite substantially. Important is not the age of a design, but its optical performance. If a certain design has state-of-the-art image quality, it is a good design whatever its original optical formula or its year of introduction. If a lens fails to deliver, however old recent its design, it is a bad lens. It is really hat simple. The current Summicron-R delivers (close to) state-of-the-art image quality and it is questionable if a new design would be substantially better without augmenting the selling price severely. Who then, given the current level of Leica prices would be willing to buy it? The very high level of corrections of the current Leica Summicron 50mm lenses is a tribute to the excellent quality of the designer team more than 20 years ago (Dr Mandler as example). One must stay realistic: without any doubt it will be possible to improve on these designs. Whether the improvement will be visible enough for the user to justify a much higher price is a BIG Question. The question of comparisons. As said earlier the designer will encounter many high order optical aberrations of increasing compexity as the aperture and/or the field angle become wider. A 2,8/100mm lens as example is much 'easier' to correct to a high level than a 1,4/35mm. The design-complexity might be a factor of 10 higher. It is a unwritten rule that only classes of lenses with the same order of complexity may be compared directly. Testprocedures. This is again a difficult topic. Generally speaking Leica designs are quite advanced and its image quality goes often beyond what can be found by commonly used evaluation procedures. The 'classical' resoluton testchart or any of its popular derivatives will not do justice to most Leica designs. The well-known testresults from the French magazine 'Chasseurs d'Images' are very difficult to interprete and often contradictory. The basic of the CdI-test is a kind of MTF testing, the results of which are 'translated' to generate the bars. As these bars do not directly refer to the original MTF graphs, the translation may or may not be adequate as representing the true image quality. In my view they do not. The MTF graphs as delivered by Zeiss and Leica are very informative, but only if you understand the theory behind it. I will post a document on this topic someday. The question of contrast versus resolution. The definition of image quality has changed over the last three or four decades. Parly because we have better understand ing of the eye and its vision and partly because we have better knowledge about optics. In reality contrast and resolution are two sides of the same coin. If we have high contrast we also have high resolution. The confusion is in the other direction: we can have very high resolution but low contrast. Good clarity of fine image details (as needed for HQ 35mm photography) however must have high contrast till the cut-off frequency (see below). That is at most 40 to 60 lp/mm and at this level contrast and resolution are in fact interchangeable. Popular testing however often lags behind and uses the expectation profile for any optical design as formulated twenty or even thirty years ago. In popular testing light falloff and corner resolution (or sharpness or contrast) figure prominently as 'bad'. Now strong vignetting is certainly bad. Slight vignetting and also slight drop of contrast in the far corners actually might improve the overall image quality. The desigher can balance the conflicting design issues to a higher level if he does not have to pay that much attention to what might be called cosmetic flaws. A topic quite relevant for the Summilux test is the so-called cut-off frequency. It has been first established by Zeiss that the maximum resolution and the contrast at that figure are not really important for assessing image quality. As example the seven element Summicron 50mm from 1954 has a resolving power of far beyond 100lp/mm, but the contrast is below 5%. Not exactly visible therefore. But the lowcontrast noise that is being produced by this state of affairs impairs the visible quality severely. In most picture viewing situations (transparancy projection and enlargements) we are looking at the image from a certain distance. If you look at a projected transparancy from about one to two meters it is impossible to see the 40lp/mm. The eye simply has not enough resolving power at that distance to perceive this fine level of detail. Most MTF graphs give results for 5, 10, 20 and 40 lp/mm. It can be proved experimentally that the 5 and 10 lp/mm are responsible for the overall impression of image quality. The 40lp/mm refer to extremely fine detail in the original object. And the 20 lp/mm define the limit of details than can be usefully recorded on film. It is also the limit of what we refer to as the clarity of fine image detail. I a way it is the cutoff frequency. Above this limit we find the optical properties that are mostly responsible for image impact. Below this limit we get an unfavourable signal-to-noise ratio and we need quite sophisticated detectors to record even finer details with good clarity. On the optical bench it is easy to demonstrate that contrast is more important than resolving power. I conducted the following experiment. I focused the Summilux with maximum resolving power in the center. The outer zones dropped dramatically in contrast and the whole image became soft. Then I refocussed with maximum contrast at the 20lp/mm. The overall image quality improved as expected. The image now has very good contrast and excellent clarity of fine to very fine details. Any designer then has to define his own mix of components of overall desirable image quality and balance the optical design accordingly. The testreport will appear in one week. Erwin - --============_-1309015588==_ma============ Content-Type: text/enriched; charset="us-ascii" <fontfamily><param>Palatino</param><bigger><bigger>Please note that I am happy to share my testreports and all the information contained in them to all members of the LUG. The report however is the result of considarable research and must be considered as a document protected by copyright law. Marc can ceratinly help here. So if you cite from, or use information contained in these documents, please respect the notion of intellectual property and refer to the original document or ask permission to copy parts of the text. Before embarking on the lens test itself, it might be nice to state the current thinking about lens design and evaluation. ITHis text might be usefull to all new LUG-members. In the early fifties and sixties the ultimate object of desire for any 35mm photographer, who liked or needed to practise the art of the artless snapshot (HCB-style) was a standard lens with an aperture of 1,4. Any additional photon that could be captured on emulsion while shooting in "available darkness" was most welcome. The slender depth of field at that large aperture added quite often impact and drama to the image. The need for such a large aperture became imperative after the candid pictures of Erich Salomon of Ermanox-fame. The 35mm worker got what he demanded in the early fifties as coated 1,5/50 designs from Zeiss and Leitz became available. These lenses indeed did capture some of the additional photons. But the photons on their way through the many glass elements wandered around (ab errare). Aberrations abounded and the image quality, to be polite, was just acceptable. The Leitz Summilux 1,4/50, introduced in 1959 for the M series camera was the first to offer a higher level of image quality. A redesign , offering very good quality, was introduced in 1962 and is still in production. The Leicaflex user had to wait till 1970 before she could capture scarce photons. Again a redesign in 1978 improved the quality. In the meantime the emulsion technology made some quantum leaps in speed/granularity relationship and the ubiquietous electronic flash lessened the need for high speed optics. Some even predicted the demise of this type of lens. More so as the 50mm fixed focus lens is nowadays often being replaced by a 'standard zoom' of 28/35 to 70/85 focal length. The design of a 50mm high speed lens is quite a challenge. Its sibling, the 2/50mm, offers image quality of the highest calibre (at least in the Leica stable). And the optical aberrations to correct are quite stubborn. Most reviewers of high speed lenses even today will tell you that a 1,4 design is a compromise. What then is the optical problem? Any lens produces a circular image area within which the 24x36mm format has to fit. This circular area can be divided in three parts, the center, the zonal area and the farout zones. The center (or the paraxial zone or Gaussian zone) is quite easy to compute. The zonal areas are more difficult to correct. Optical aberrations have the habit to grow disproportionately if the apertur and/or the field-angle become wider. Many aberrations grow with the square root or the cubic root in relation to the aperture diameter or even more. OK you would say, lets settle for a bit less image quality in the corners. The snag however is this: the zonal aberrations have a strong influence on the performance in the center. Moreover: when stopping down the effect on some aberrations is not reduced. The combined result of all aberrations is always a reduction in contrast: a softening of small details and a low overall contrast. The lens designer's plight however is not over if he succeeds in reconciling all these conflicting demands. Aberrations can be classified as third order, fifth order and seventh order aberrations and so on until the n-th order. (I will explain in a separate post why they are so designated). Third order aberrations are large and suppress all other aberrations in the series. If a designer can tame these third order errors, he will be unpleasantly presented with the next in line. Balancing third order aberrations require often a change in focus position. The well known statement that you can compute a lens for high contrast or high resolution ultimately boils down to this kind of balancing. Fifth order aberrations are mathematically quite challenging. The high quality of Leica lenses is based upon an excellent grip on this group of aberrations. As usual a balancing of conflicting demands and finetuning of parameters is needed to compute a lens to this very high level of correction. But choices are inevitable. If the designer has done her best (actually some of Leicas best optical designers are female) she would be lost in space if the manufacturing department could not support her. If a (hypothetical) lens design needs to satisfy 100 parameters, more than 50% of these would have to be fullfilled by the manufacturing department. Modern Leica lenses and their exquisite quality would be unthinkable without the control in the production line. This is a not well known fact: the designer is nowhere without support from the manufacturing guys. But the designer is restricted in more ways. A lens has certain physical dimensions. If you would free a designer from the physical constraints he can perform wonders. The famous lenses for the Contarex series followed this paradigm: whatever the physical dimensions, the optical performance may not be jeopardized. Result: ergonomically the lenses were often hardly usable. So the designer needs to serve very many stern masters. The relative neglect of the 1,4/50 since more than 20 years might be the consequence of all these considerations. Now Leica has introduced a brand new 1,4/50 for the R-series. Any tester will be challenged to assess its performance against the background I just outlined. The ASPH riddle. Leica has introduced in recent years several lenses with one or two aspherical surfaces. Generally the image quality of these designs is quite high, to say the least. Some observers of the Leica scene have erroneously concluded that the equation aspherical=high image quality now has universal validity. Some even went further and deduced that any lens design without an aspherical surface can not be designated as a modern design. Occasionally one will hear or read the statement that for example the current Summicron 50 lens for the R series is an old design and needs or will be superceded by a newer design of invariably aspherical signature. It is easy to be charmed by such reasoning. This assumption,however, is not correct. First some facts. One: the new Summilux-R 1,4/50 (subject of this report) has been designed with conventional means. We can then conclude that the Leica designers could realize the required level of image quality without recourse to aspherical surfaces. If the use of asphericals would have been advantageous for the state of corrections and the production requirements, Leica certainly would have incorporated them. Two: asperics are not always the best way to go. The Ricoh 28mm uses two aspherical surfaces but its image quality is below that of the Leica 28mm and the Zeiss 28mm for the G-series, both without asphericals. Three: all Zeiss lenses for the G-series are quite recent designs and none of them has any aspherical surface. The image quality of these lenses is beyond any reasonable doubt. One of the main reasons for employing ashericals is the correction of spherical aberration and attaining a flat field free of astigmatism. But the use of asphericals may also affect other aberrations in a dangerous way, especially if the distance between diaphragm position and asperical surface is relatively large. The design of an optical system must always try to balance many demands and variables, some of which are optical and some of which are manufacturing oriented. It could be that a designer tries to incorporate an aspherical surface only to find out that the strain on production tolerances is too heavy. He also could note that given the overall configuration of his/her design, the aspherical surface has no added value, or even will enlarge or introduce other aberrations. Note that any optical system must be regarded as a delicate whole of carefully designed and matched components. Note also that all aberrations act on every image point in conjunction. Note further that 'image quality' is not a fixed set of parameters. Zeiss will adjust the balance of corrected aberrations and the magnitude of corrections according to different rules than Leica does. Leica may conclude than given the required correction in some cases asperical surfaces are justified and in some cases not. The question of old designs. I have no idea when and why a lens can be designated as 'old'. The current Summicron and Summilux 50 and 75/80 lenses (include also the current Zeiss Planar lenses) are all variants of the double-Gauss type of optical classification. This design is now almost 100 years old. The general formula of a lens may be 'old'. Important is however not the general design but the state of corrections. A small change in curvature, different location of the diaphragm, different glass types and small changes in the distances between lens elements may alter the image quality quite substantially. Important is not the age of a design, but its optical performance. If a certain design has state-of-the-art image quality, it is a good design whatever its original optical formula or its year of introduction. If a lens fails to deliver, however old recent its design, it is a bad lens. It is really hat simple. The current Summicron-R delivers (close to) state-of-the-art image quality and it is questionable if a new design would be substantially better without augmenting the selling price severely. Who then, given the current level of Leica prices would be willing to buy it? The very high level of corrections of the current Leica Summicron 50mm lenses is a tribute to the excellent quality of the designer team more than 20 years ago (Dr Mandler as example). One must stay realistic: without any doubt it will be possible to improve on these designs. Whether the improvement will be visible enough for the user to justify a much higher price is a BIG Question. The question of comparisons. As said earlier the designer will encounter many high order optical aberrations of increasing compexity as the aperture and/or the field angle become wider. A 2,8/100mm lens as example is much 'easier' to correct to a high level than a 1,4/35mm. The design-complexity might be a factor of 10 higher. It is a unwritten rule that only classes of lenses with the same order of complexity may be compared directly. Testprocedures. This is again a difficult topic. Generally speaking Leica designs are quite advanced and its image quality goes often beyond what can be found by commonly used evaluation procedures. The 'classical' resoluton testchart or any of its popular derivatives will not do justice to most Leica designs. The well-known testresults from the French magazine 'Chasseurs d'Images' are very difficult to interprete and often contradictory. The basic of the CdI-test is a kind of MTF testing, the results of which are 'translated' to generate the bars. As these bars do not directly refer to the original MTF graphs, the translation may or may not be adequate as representing the true image quality. In my view they do not. The MTF graphs as delivered by Zeiss and Leica are very informative, but only if you understand the theory behind it. I will post a document on this topic someday. The question of contrast versus resolution. The definition of image quality has changed over the last three or four decades. Parly because we have better understand ing of the eye and its vision and partly because we have better knowledge about optics. In reality contrast and resolution are two sides of the same coin. If we have high contrast we also have high resolution. The confusion is in the other direction: we can have very high resolution but low contrast. Good clarity of fine image details (as needed for HQ 35mm photography) however must have high contrast till the cut-off frequency (see below). That is at most 40 to 60 lp/mm and at this level contrast and resolution are in fact interchangeable. Popular testing however often lags behind and uses the expectation profile for any optical design as formulated twenty or even thirty years ago. In popular testing light falloff and corner resolution (or sharpness or contrast) figure prominently as 'bad'. Now strong vignetting is certainly bad. Slight vignetting and also slight drop of contrast in the far corners actually might improve the overall image quality. The desigher can balance the conflicting design issues to a higher level if he does not have to pay that much attention to what might be called cosmetic flaws. A topic quite relevant for the Summilux test is the so-called cut-off frequency. It has been first established by Zeiss that the maximum resolution and the contrast at that figure are not really important for assessing image quality. As example the seven element Summicron 50mm from 1954 has a resolving power of far beyond 100lp/mm, but the contrast is below 5%. Not exactly visible therefore. But the lowcontrast noise that is being produced by this state of affairs impairs the visible quality severely. In most picture viewing situations (transparancy projection and enlargements) we are looking at the image from a certain distance. If you look at a projected transparancy from about one to two meters it is impossible to see the 40lp/mm. The eye simply has not enough resolving power at that distance to perceive this fine level of detail. Most MTF graphs give results for 5, 10, 20 and 40 lp/mm. It can be proved experimentally that the 5 and 10 lp/mm are responsible for the overall impression of image quality. The 40lp/mm refer to extremely fine detail in the original object. And the 20 lp/mm define the limit of details than can be usefully recorded on film. It is also the limit of what we refer to as the clarity of fine image detail. I a way it is the cutoff frequency. Above this limit we find the optical properties that are mostly responsible for image impact. Below this limit we get an unfavourable signal-to-noise ratio and we need quite sophisticated detectors to record even finer details with good clarity. On the optical bench it is easy to demonstrate that contrast is more important than resolving power. I conducted the following experiment. I focused the Summilux with maximum resolving power in the center. The outer zones dropped dramatically in contrast and the whole image became soft. Then I refocussed with maximum contrast at the 20lp/mm. The overall image quality improved as expected. The image now has very good contrast and excellent clarity of fine to very fine details. Any designer then has to define his own mix of components of overall desirable image quality and balance the optical design accordingly. The testreport will appear in one week. Erwin</bigger></bigger></fontfamily> - --============_-1309015588==_ma============--