08 November 2009
october 2009
09/11/09 22:08
October 2009:
Camera manufacturers keep updating their products. From Olympus with the E-P2 to Canon with the 1D Mark IV and Nikon with the D3s, every company seems to be focusing on ever-shrinking product life-cycles with ever-expanding feature lists. The electronic viewfinder of the E-P2 is smart and the accessory interface certainly deserves an accolade. E-P1 buyers are applauding of course. Amidst this barrage of updated products there is the news of the death of two really great photographers, Willy Ronis in France and Irving Penn in the USA. Both produced stunning pictures with simple equipment and stayed close to the craft of photography. They were not equipment-driven, but meticulously dedicated to perfecting the quality of the image. In another discipline one could refer to Barnack as an engineer who was also focused on perfection. A new book by Ulf Richter (from Concept to Leica) charts the birth of the Leica camera, designed by Barnack over a long period. Barnack was mostly focused on realizing a camera to take pictures with and get excellent results. compared to what was usual in his days. Technique was subordinated to the barely necessary and never an end in itself.
We should also remind ourselves that the Leica camera earned its respect form artists who actually used the camera to create new and exciting photographs.
Film and grain and speed and even tonal curves were facts to live with and the good photographer could internalize these characteristics and use them to get the best pictures that were in his/her mind.
Recently I got several emails concerning the current bias to produce numerical results that are in most cases not comparable with other results and are always not really related to the art of taking pictures or making images. The abundance of tools on the internet has been the cause of an inflation of technical results on every website. Generating and comparing numbers seem to be the only valuable act in the discourse about the quality and performance of a camera model.
What is now happening with the assessment of the performance of a camera, is identical to what happened to lens evaluation some time ago. I remember that Geoffrey Crawley refused to publish numerical results of his lens tests because he was afraid that the numbers would be used without correct reference to their inherent meaning and therefore he relied on carefully worded reports to give an impression of the lens performance. When the MTF graphs became the norm, this approach became obsolete, but the basic idea remains valid. Most persons who look at and compare MTF graphs have only a scant idea of the really important connections between MTF graphs and the actual picture quality and it is safe to claim that he current MTF graphs have only a limited connection to the true performance of a lens.
The same is true of most numerical results available on the internet. In the past, the obsession with resolution figures was, rightly, commented upon as meaningless, but now the same approach, disguised as Lp/ph, is returning and worshipped as important information. We can expand this list with noise analysis, deltaE color bias, dynamic range, Nyquist frequency and a host of other parameters that can be quantified. A good example is the accuracy of a shutter. Many mechanically governed shutters had speeds with 30% or more tolerance. This looks bad: a nominal shutter speed of 1/1000 has an actual value of 1/660! One shudders at the thought! But in practice hardly anyone noticed this difference of about a half stop in exposure as the film was very tolerant and in most situations the dark room could compensate this small aberration.
Nowadays magazines are ranking cameras with a one percent point difference on a scale of 80 to 100, claiming that a camera that scores 78% is vastly superior to a camera scoring 81%, the score being based on an arbitrary act of averaging a range of weighted numerical values.
It has always been my position to focus on a very few meaningful numbers to assess the performance of a lens, camera or film/developer combo and relate these figures to practical picture taking.
In my view the accuracy of the range finder is more important that the color rendition and the definition of a recording medium more important than the final resolution. Grain pattern is important at large magnifications but even then the grain can be used as an expressive component. In my recent report on the color accuracy of the M9 versus the M8 I noted that there are many measurable differences, but that these differences do not have much impact on the ultimate appreciation of the image. I have made very satisfying pictures with the (presumed flawed) M8 without IR filter and also with the (presumed improved) M9. Do we really adopt an approach where we have to say that a certain picture has great composition, important human interest and emotional impact, but the red spot in the background has a DeltaE of 14 and therefore the picture is worthless?
In the glory days of film recording, every one knew that Kodachrome had a different color pallet than Ektachrome or Agfachrome and you had to choose your film according to the required color reproduction. And when you needed a fine grained film with high resolution, you accepted a limited dynamic range. Now we want everything in one camera: accurate colors, high dynamic range, high resolution, high speed and every possible combination without losing anything.
We are in fact in the same position as an optical designer: to design a lens we need to take into account numerous parameters, sometimes conflicting, but always there is a need for balancing and finding a optimum that refers to some set of qualifiers. But we cannot design a lens to suit every demand and every task.
The great gift of digital recording is the fact that the resulting product is a file of numbers that can be manipulated at will. A digital image is nothing more than a finite string of values in the form of a matrix of numbers from zero to 256 or 16384. But this string of numbers can be manipulated at every stage of the process of image transformations. The results can be analyzed and presented as graphs or numbers. But makes it sense? I doubt it: the color reproduction evidently differs between several color spaces and one can spend a life time comparing all results from all camera makers. Popphoto notes that a difference of Delta E 8 is an upper limit to get a tag of, but they also note that this difference is less than the color shift in color negative films. But they fail to notice that the perception of color shifts differ from color to color.
To bite the bullet: no one ever claimed that Kodachrome with its specific color reproduction was better than Velvia that had different colors. The fact is that most viewers liked the Velvia more than the Kodachrome and therefor photographers adopted Velvia.
In the same vein I would not dare to claim that the M8 is better or worse than the M9 because the average Delta E value is higher or lower.
In the digital world you can correct colors and shift color spaces in a more liberal fashion that was ever possible in the AgX period. But this fact makes color analysis less important.
In a previous blog I said that the value of the standard MTF graphs can be questioned, because a more sophisticated analysis of lens performance is required now that lens quality is approaching a plateau. The same is true for camera analysis: many of the current numerical analysis options need to be re-calibrated to do justice to the high quality of the camera performance. It is amazing that a small compact camera like the M9 can deliver results that a medium format film loading Hasselblad would be envious of. With this statement as background one wonders why reviewers are so much pre-occupied with the noise level at ISO 2500 or the color reproduction in the ProRGB color space or the limiting resolution at the Nyquist limit to assess the profile of the camera.
I would like to return to the images of Willy Ronis or Irving Penn. They worked with cameras and films that were much less potent than current materials. But they produced masterpieces not worrying about color reproduction in sRGB or AdobeRGB or noise at ISO2500.
The new Canon 1D Mark IV has a crop factor of 1.3 and Canon does not see this as a problem, but as a solution to a specific set of challenges. The idea that only a full sized sensor with 24 x 36mm is an acceptable proposition, is a bridge too far. The M8 sensor size is not as big as the one found in the M9. But this fact alone does not disqualify the M8 as a serious photographic tool.
Numbers are needed to assess the performance of a tool. I would draw attention to the ideals of the Arts and Crafts Movement in England where meticulous standards are adopted to assess the quality of the product. These standards are rooted in the craft itself, and not based on the fact that there are tools with which any kind of measurement is possible.
Camera manufacturers keep updating their products. From Olympus with the E-P2 to Canon with the 1D Mark IV and Nikon with the D3s, every company seems to be focusing on ever-shrinking product life-cycles with ever-expanding feature lists. The electronic viewfinder of the E-P2 is smart and the accessory interface certainly deserves an accolade. E-P1 buyers are applauding of course. Amidst this barrage of updated products there is the news of the death of two really great photographers, Willy Ronis in France and Irving Penn in the USA. Both produced stunning pictures with simple equipment and stayed close to the craft of photography. They were not equipment-driven, but meticulously dedicated to perfecting the quality of the image. In another discipline one could refer to Barnack as an engineer who was also focused on perfection. A new book by Ulf Richter
We should also remind ourselves that the Leica camera earned its respect form artists who actually used the camera to create new and exciting photographs.
Film and grain and speed and even tonal curves were facts to live with and the good photographer could internalize these characteristics and use them to get the best pictures that were in his/her mind.
Recently I got several emails concerning the current bias to produce numerical results that are in most cases not comparable with other results and are always not really related to the art of taking pictures or making images. The abundance of tools on the internet has been the cause of an inflation of technical results on every website. Generating and comparing numbers seem to be the only valuable act in the discourse about the quality and performance of a camera model.
What is now happening with the assessment of the performance of a camera, is identical to what happened to lens evaluation some time ago. I remember that Geoffrey Crawley refused to publish numerical results of his lens tests because he was afraid that the numbers would be used without correct reference to their inherent meaning and therefore he relied on carefully worded reports to give an impression of the lens performance. When the MTF graphs became the norm, this approach became obsolete, but the basic idea remains valid. Most persons who look at and compare MTF graphs have only a scant idea of the really important connections between MTF graphs and the actual picture quality and it is safe to claim that he current MTF graphs have only a limited connection to the true performance of a lens.
The same is true of most numerical results available on the internet. In the past, the obsession with resolution figures was, rightly, commented upon as meaningless, but now the same approach, disguised as Lp/ph, is returning and worshipped as important information. We can expand this list with noise analysis, deltaE color bias, dynamic range, Nyquist frequency and a host of other parameters that can be quantified. A good example is the accuracy of a shutter. Many mechanically governed shutters had speeds with 30% or more tolerance. This looks bad: a nominal shutter speed of 1/1000 has an actual value of 1/660! One shudders at the thought! But in practice hardly anyone noticed this difference of about a half stop in exposure as the film was very tolerant and in most situations the dark room could compensate this small aberration.
Nowadays magazines are ranking cameras with a one percent point difference on a scale of 80 to 100, claiming that a camera that scores 78% is vastly superior to a camera scoring 81%, the score being based on an arbitrary act of averaging a range of weighted numerical values.
It has always been my position to focus on a very few meaningful numbers to assess the performance of a lens, camera or film/developer combo and relate these figures to practical picture taking.
In my view the accuracy of the range finder is more important that the color rendition and the definition of a recording medium more important than the final resolution. Grain pattern is important at large magnifications but even then the grain can be used as an expressive component. In my recent report on the color accuracy of the M9 versus the M8 I noted that there are many measurable differences, but that these differences do not have much impact on the ultimate appreciation of the image. I have made very satisfying pictures with the (presumed flawed) M8 without IR filter and also with the (presumed improved) M9. Do we really adopt an approach where we have to say that a certain picture has great composition, important human interest and emotional impact, but the red spot in the background has a DeltaE of 14 and therefore the picture is worthless?
In the glory days of film recording, every one knew that Kodachrome had a different color pallet than Ektachrome or Agfachrome and you had to choose your film according to the required color reproduction. And when you needed a fine grained film with high resolution, you accepted a limited dynamic range. Now we want everything in one camera: accurate colors, high dynamic range, high resolution, high speed and every possible combination without losing anything.
We are in fact in the same position as an optical designer: to design a lens we need to take into account numerous parameters, sometimes conflicting, but always there is a need for balancing and finding a optimum that refers to some set of qualifiers. But we cannot design a lens to suit every demand and every task.
The great gift of digital recording is the fact that the resulting product is a file of numbers that can be manipulated at will. A digital image is nothing more than a finite string of values in the form of a matrix of numbers from zero to 256 or 16384. But this string of numbers can be manipulated at every stage of the process of image transformations. The results can be analyzed and presented as graphs or numbers. But makes it sense? I doubt it: the color reproduction evidently differs between several color spaces and one can spend a life time comparing all results from all camera makers. Popphoto notes that a difference of Delta E 8 is an upper limit to get a tag of
To bite the bullet: no one ever claimed that Kodachrome with its specific color reproduction was better than Velvia that had different colors. The fact is that most viewers liked the Velvia more than the Kodachrome and therefor photographers adopted Velvia.
In the same vein I would not dare to claim that the M8 is better or worse than the M9 because the average Delta E value is higher or lower.
In the digital world you can correct colors and shift color spaces in a more liberal fashion that was ever possible in the AgX period. But this fact makes color analysis less important.
In a previous blog I said that the value of the standard MTF graphs can be questioned, because a more sophisticated analysis of lens performance is required now that lens quality is approaching a plateau. The same is true for camera analysis: many of the current numerical analysis options need to be re-calibrated to do justice to the high quality of the camera performance. It is amazing that a small compact camera like the M9 can deliver results that a medium format film loading Hasselblad would be envious of. With this statement as background one wonders why reviewers are so much pre-occupied with the noise level at ISO 2500 or the color reproduction in the ProRGB color space or the limiting resolution at the Nyquist limit to assess the profile of the camera.
I would like to return to the images of Willy Ronis or Irving Penn. They worked with cameras and films that were much less potent than current materials. But they produced masterpieces not worrying about color reproduction in sRGB or AdobeRGB or noise at ISO2500.
The new Canon 1D Mark IV has a crop factor of 1.3 and Canon does not see this as a problem, but as a solution to a specific set of challenges. The idea that only a full sized sensor with 24 x 36mm is an acceptable proposition, is a bridge too far. The M8 sensor size is not as big as the one found in the M9. But this fact alone does not disqualify the M8 as a serious photographic tool.
Numbers are needed to assess the performance of a tool. I would draw attention to the ideals of the Arts and Crafts Movement in England where meticulous standards are adopted to assess the quality of the product. These standards are rooted in the craft itself, and not based on the fact that there are tools with which any kind of measurement is possible.