APO Summicron 50 ASPH review

[helged]

Advertisement (gone after registration)

From LL: The new APO Summicron 50 ASPH becomes a benchmark in optical performance.

[Rick]

From LL: The new APO Summicron 50 ASPH becomes a benchmark in optical performance.

 

Nice review from a guy that sounds like he knows more than anyone...

[Jeff S]

I don't think he mentions flare once in the article. Not that I care, but interesting given all the discussion about it elsewhere.

 

And while he discusses the 50 Summicron (non-APO) and Noctilux by comparison, he seems to omit commentary on the 50 Summilux asph.

 

It is consistent, however, with his love fest with the new M.

 

Jeff

[rramesh]

I am not sure about this comment though "As far as I know, it is the first lens that easily exceeds the resolution of current camera sensors".

[01af]

I am not sure about this comment though "As far as I know, it is the first lens that easily exceeds the resolution of current camera sensors".

Good catch! It's nonsense indeed.

 

Lenses don't exceed the resolution of sensors. Also, sensors don't exceed the resolution of lenses. The limited (i. e. less-than-infinite) resolutions of lenses and sensors both will contribute to the limit of the resolution of the final image—always.

 

So if you switch from a good lens to a better lens then the sharpness of the final image will improve, no matter what sensor you're using.

 

 

EDIT: Now after having actually read it, I must say the whole article is full of errors, nonsense, and misconceptions ... except the conclusion that the Apo-Summicron-M 50 mm 1:2 Asph is a damn fine lens indeed.

[flyalf]

Good catch! It's nonsense indeed.

 

Lenses don't exceed the resolution of sensors. Also, sensors don't exceed the resolution of lenses. The limited (i. e. less-than-infinite) resolutions of lenses and sensors both will contribute to the limit of the resolution of the final image—always.

 

So if you switch from a good lens to a better lens then the sharpness of the final image will improve, no matter what sensor you're using.

 

 

EDIT: Now after having actually read it, I must say the whole article is full of errors, nonsense, and misconceptions ... except the conclusion that the Apo-Summicron-M 50 mm 1:2 Asph is a damn fine lens indeed.

 

I would have to disagree with you. Its possible to quantify the resolution of both lenses and sensors, and its possible to discuss what is the main limiting of these.

 

Also I'm sorry to say hat I find your comments on "full of errors, nonsense, and misconceptions" to be the normal Internet bashing. I urge you to post more exactly what these "errors, nonsense, and misconceptions" are, or even better to make your own analysis correcting all these and get it published.

[01af]

Advertisement (gone after registration)

On the errors, shortcomings, and misconceptions in Mark Dubovoy's article

"Testing Leica's new 50 mm. F/2 APO Summicron ASPH" at Luminous Landscape

 

 

I believe that many of us have periodically wondered what would happen if management of one of our favorite photography companies told the design and manufacturing engineers to make the best product they could, with no constraints and no restrictions. In other words, no cost constraints, no engineering constraints, no manufacturing constraints, no marketing pressures, etc.

 

[...] Leica was introducing a new 50 mm Summicron that came out of an all-out effort with no holds barred ...

While the Apo-Summicron-M 50 mm 1:2 Asph has been developed under much less constraints than most lenses (including most Leica M lenses), it is still far from being not constrained at all. The most severe constraints are size and manufacturability. It's comparatively easy to produce one single prototype lens that blows everything else out of the water, but coming up with a design that does the same over and over when being mass-produced is an entirely different thing. Moreover, the Apo-Summicron-M 50 Asph was supposed to be no bigger than the regular Summicron-M 50 ... in the end, it's only 3.5 mm (1/7 inch) longer but has the same diameter and filter size. If it was allowed to grow to the size of, say, a Summarit-M 75 mm then it could be even better and would vignette less, at a lower price.

 

 

The event that changed all of this was Leica's decision to introduce the M Monochrom. Since the M Monochrom is B&W only, the sensor has no color filters and no Bayer matrix pattern. Therefore, the resolution goes up significantly because unlike color sensors, each photosite produces one pixel in the final image. With the increased resolution of the M Monochrom, the difference between the old lens and a new one could be much more noticeable, provided the new lens was up to the task.

It's a popular misconception to believe that the increased resolution of the M Monochrom was the primary reason why an improved lens was called for. As matter of fact, the M Monochrom's resolution greatly impresses people with any good lens. At the same time, the Apo-Summicron-M 50 mm Asph's resolution also impresses people very much on the good old M9 or M-E. So no, the Apo-Summicron-M 50 mm Asph doesn't depend on the M Monochrom's resolution, and the M Monochrom doesn't depend on the Apo-Summicron-M 50 mm Asph's resolution.

 

Instead, the reasons why the Apo-Summicron-M 50 mm Asph was launched now and not any earlier are these two: (1) The overwhelming success of the expensive M9, the expensive Noctilux-M 50 mm Asph, and the expensive Leica S system has convinced Leica Camera's management that even an outrageously expensive lens could still be an economical success if it only is good enough. (2) It is not possible to eliminate or reduce the effects of chromatic aberrations in the post-processing of monochrome images. Hence, the M Monochrom can make good use of lenses with extra-low chromatic aberrations.

 

 

The Best Lens In the World Syndrome. Are We At A Crossroads?

I am convinced that we are at an exciting crossroads in the history of photography.

This mindless statement is holding as much water as, for example, "Ferrari just launched the new F458 Speciale—we are at an exciting crossroads in the history of traffic and transportation" ... i. e. nothing but a hollow cant. Ernst Leitz Wetzlar, now Leica Camera, keeps producing improved lenses every couple of years for decades now, and they probably will continue to do so for some time to come (hopefully). So the introduction of a new and improved lens is always an exciting event for a photographer but not exactly "crossroads" for the history of photography. Get a life!

 

By the way, other lens manufacturers are also working on Best Lenses of the World ... look at Sigma, Olympus, or Zeiss, for example. Today's lenses are better than yesterday's, and tomorrow's lenses will be better still. Just the usual inevitable consequence of progress ...

 

 

In general, our camera systems today are lens-constrained. What I mean by this is that the best lenses we use are very close but not quite as good as the sensors in our cameras. One example that comes to mind is that we can compare lenses and see and measure the resolution differences in our cameras. This means that the sensors have better resolution than the lenses. If all our lenses had significantly higher resolution than the sensors, we would not be able to see or measure these differences as the sensor would be the limiting factor.

This is simply not true—as any mathematician or physicist will tell you. It's the usual layman's misconception of the weakest link in the imaging chain that will allegedly determine the final image's resolution. But in fact, an image's resolution is always determined by the individual resolution limits of all components in the imaging chain—in particular, lens and sensor. So with any sensor, switching from a good lens to a better lens will improve the image. And the opposite is also true—under any lens, switching from a sensor with a lower pixel count to a sensor with a higher pixel count will improve the image (real-world side effects like increased noise not withstanding).

 

So the lenses we were happy with in the past will continue to delight us even on improved sensors.

 

Except when we want to print bigger than we ever did before. In the past, a print magnification of about 20× was considered the maximum. Starting with a 35-mm-format negative, 20× means a print size of approx. 50 × 75 cm (19 × 28"). Back in the film days, such a print would have been considered huge, and hardly anybody actually printed that big—not from 35-mm negatives. Today, looking at a digital image file from the M9 at 100 % view on a regular 1,920 × 1,200 pixel 24" screen means a magnification of 40×, or a print size of 1 × 1.5 m (40 × 60"). A Leica M (Typ 240) file on the same screen at 100 % is 45×; a Nikon D800 file even is 55×, or 1.3 × 2 m (4 × 6 ft)! And people innocently believe that at these insane kinds of magnifications, lenses should be basically aberration-free.

 

That's the actual reason why lens manufacturers are working on better lenses.

[120]

...

This is simply not true—as any mathematician or physicist will tell you. It's the usual layman's misconception of the weakest link in the imaging chain that will allegedly determine the final image's resolution...

 

 

He has it right. If one of the terms in 1/x + 1/y is low res and the other varies, the sum doesn't change much...which is what he says informally.

 

This is an empirical relationship in photographic science...why would a math or physics PhD study it? (But by coincidence the author is a physics PhD.)

[jaapv]

On the errors, shortcomings, and misconceptions in Mark Dubovoy's article

"Testing Leica's new 50 mm. F/2 APO Summicron ASPH" at Luminous Landscape

{...}

 

While the Apo-Summicron-M 50 mm 1:2 Asph has been developed under much less constraints than most lenses (including most Leica M lenses), it is still far from being not constrained at all. The most severe constraints are size and manufacturability. It's comparatively easy to produce one single prototype lens that blows everything else out of the water, but coming up with a design that does the same over and over when being mass-produced is an entirely different thing. Moreover, the Apo-Summicron-M 50 Asph was supposed to be no bigger than the regular Summicron-M 50 ... in the end, it's only 3.5 mm (1/7 inch) longer but has the same diameter and filter size. If it was allowed to grow to the size of, say, a Summarit-M 75 mm then it could be even better and would vignette less, at a lower price.

 

{...}

 

It's a popular misconception to believe that the increased resolution of the M Monochrom was the primary reason why an improved lens was called for. As matter of fact, the M Monochrom's resolution greatly impresses people with any good lens. At the same time, the Apo-Summicron-M 50 mm Asph's resolution also impresses people very much on the good old M9 or M-E. So no, the Apo-Summicron-M 50 mm Asph doesn't depend on the M Monochrom's resolution, and the M Monochrom doesn't depend on the Apo-Summicron-M 50 mm Asph's resolution.

 

Instead, the reasons why the Apo-Summicron-M 50 mm Asph was launched now and not any earlier are these two: (1) The overwhelming success of the expensive M9, the expensive Noctilux-M 50 mm Asph, and the expensive Leica S system has convinced Leica Camera's management that even an outrageously expensive lens could still be an economical success if it only is good enough. (2) It is not possible to eliminate or reduce the effects of chromatic aberrations in the post-processing of monochrome images. Hence, the M Monochrom can make good use of lenses with extra-low chromatic aberrations.

 

{...}

 

 

This mindless statement is holding as much water as, for example, "Ferrari just launched the new F458 Speciale—we are at an exciting crossroads in the history of traffic and transportation" ... i. e. nothing but a hollow cant. Ernst Leitz Wetzlar, now Leica Camera, keeps producing improved lenses every couple of years for decades now, and they probably will continue to do so for some time to come (hopefully). So the introduction of a new and improved lens is always an exciting event for a photographer but not exactly "crossroads" for the history of photography. Get a life!

 

By the way, other lens manufacturers are also working on Best Lenses of the World ... look at Sigma, Olympus, or Zeiss, for example. Today's lenses are better than yesterday's, and tomorrow's lenses will be better still. Just the usual inevitable consequence of progress ...

 

{...}

 

]This is simply not true—as any mathematician or physicist will tell you. It's the usual layman's misconception of the weakest link in the imaging chain that will allegedly determine the final image's resolution. But in fact, an image's resolution is always determined by the individual resolution limits of all components in the imaging chain—in particular, lens and sensor. So with any sensor, switching from a good lens to a better lens will improve the image. And the opposite is also true—under any lens, switching from a sensor with a lower pixel count to a sensor with a higher pixel count will improve the image (real-world side effects like increased noise not withstanding).

 

So the lenses we were happy with in the past will continue to delight us even on improved sensors.

 

Except when we want to print bigger than we ever did before. In the past, a print magnification of about 20× was considered the maximum. Starting with a 35-mm-format negative, 20× means a print size of approx. 50 × 75 cm (19 × 28"). Back in the film days, such a print would have been considered huge, and hardly anybody actually printed that big—not from 35-mm negatives. Today, looking at a digital image file from the M9 at 100 % view on a regular 1,920 × 1,200 pixel 24" screen means a magnification of 40×, or a print size of 1 × 1.5 m (40 × 60"). A Leica M (Typ 240) file on the same screen at 100 % is 45×; a Nikon D800 file even is 55×, or 1.3 × 2 m (4 × 6 ft)! And people innocently believe that at these insane kinds of magnifications, lenses should be basically aberration-free.

 

That's the actual reason why lens manufacturers are working on better lenses.

Spot-on. I might add that the concept and initial design of the apo-summicron lies more than a decade before its introduction, long before anybody thought current megapixel counts would be possible in the near future - or even make sense... It was not a design decision to introduce it, but a marketing one.

[CaptZoom]

I seem to recall the head lens design guy at Leica commenting on the S system lenses being designed to out resolve current sensors, including the one in the S camera. The interview/discussion was recorded as part of the promo/marketing for the lain h of the S system. So it's dated by now, but the interesting bit is the head lens designer did not think the question or statements regarding lenses out resolving sensors were meaningless.

[jaapv]

Any improvement in the image chain, be it sensor or lens, will improve image quality. In that sense it is certainly not meaningless, just one-sided. I think Mr. Karbe was indicating a game of leapfrog between the two.

[01af]

... but the interesting bit is the head lens designer did not think the question or statements regarding lenses out-resolving sensors were meaningless.

He was trying to put it in a way even laymen can understand. You cannot try to educate your audience about the finer points of the maths and physics of information transmission in an interview that's supposed to promote your latest products.

[jankap]

He has it right. If one of the terms in 1/x + 1/y is low res and the other varies, the sum doesn't change much...which is what he says informally.

 

This is an empirical relationship in photographic science...why would a math or physics PhD study it? (But by coincidence the author is a physics PhD.)

 

This is general knowledge. IS one of the factors very low?

Sometimes also PhDs write too general stuff, of course one must not read the space fillers.

Jan

[pgk]

But in fact, an image's resolution is always determined by the individual resolution limits of all components in the imaging chain—in particular, lens and sensor.

O1af

Am I correct in thinking that this could be expressed by cascading MTF graphs of the components within the imaging chain (although MTF data for sensors doesn't seem to be too easy to obtain). So what may be shown is that any better lens, with a higher MTF, will, when combined with a sensor's MTF result in a cascaded MTF reduced by a lower amount than it would be by a poorer lens? (Its a very long time since I was involved with MTF testing!)

[120]

O1af

Am I correct in thinking that this could be expressed by cascading MTF graphs of the components within the imaging chain (although MTF data for sensors doesn't seem to be too easy to obtain). So what may be shown is that any better lens, with a higher MTF, will, when combined with a sensor's MTF result in a cascaded MTF reduced by a lower amount than it would be by a poorer lens? (Its a very long time since I was involved with MTF testing!)

 

see e.g. Price, W. (1976) Scientific American 235:72, redrawn elsewhere

[01af]

If one of the terms in 1/x + 1/y is low-res and the other varies, the sum doesn't change much ...

That's right—but "not much" still is a lot more than "not at all."

 

 

... which is what he says informally.

No ... I'm afraid this isn't what he's saying, not even informally.

 

Actually he's saying that being able to test lenses on current digital cameras (i. e. being able to tell good lenses from better lenses from excellent lenses) was proof that current sensors "outresolve" lenses and hence, better lenses were called for. But that's wrong. In fact, we can look at the differences of various lenses also through the window provided by a low-res sensor. So the fact we can see differences between lenses is NOT proof that sensors "outresolve" lenses. To the contrary—when accepting low contrast, even mediocre lenses can transmit way more line pairs per millimeter than current sensors possibly can resolve.

 

By the way, the same is true in stereo systems. Even when using poor speakers, you can still assess the differences between good, better, and excellent amplifiers (or any component in the sound-reproduction chain). Of course, with a good pair of speakers, this would be easier to do, and, beyond testing, it would be more fun to listen to the whole chain ... but the characterics of amplifiers, pre-amplifiers, and sources can be heard through low-end speakers, too.

 

 

Am I correct in thinking that this could be expressed by cascading MTF graphs of the components within the imaging chain [...]. So what may be shown is that any better lens, with a higher MTF, will, when combined with a sensor's MTF, result in a cascaded MTF reduced by a lower amount than it would be by a poorer lens?

Yes, exactly.

 

 

... (although MTF data for sensors doesn't seem to be too easy to obtain).

Well ... lenses and sensors both have limited resolving powers; that much they have in common. But the characteristics of these limits are very different. Sensors have a hard limit clearly defined by their pixel pitch. Lenses basically have no real limit; instead, they will just roll off smoothly for increasing spatial frequencies. That's why you won't find data about a lens' resolution limit—it simply doesn't exist. Instead, you'll find the contrast transmitted at a given spatial frequency or the maximum spatial frequency for a given minimum of transmitted contrast. So the means to express the resolutions are very different for lenses and for sensors.

 

Still, the basic idea is just as you've outlined above.

[CaptZoom]

Any improvement in the image chain, be it sensor or lens, will improve image quality. In that sense it is certainly not meaningless, just one-sided. I think Mr. Karbe was indicating a game of leapfrog between the two.

 

I understand the limiting factors in a chain. Horse carriages are as fast as the slowest horse, etc.

 

He was trying to put it in a way even laymen can understand. You cannot try to educate your audience about the finer points of the maths and physics of information transmission in an interview that's supposed to promote your latest products.

 

We are in agreement regarding your statement about education and promotion. Nonetheless, if he was putting things in terms for laymen to understand, doesn't that imply some basic truth to what he was saying? If so, what is wrong with someone posting an online review written for laymen using terms and statements that laymen can understand?

 

To date, there hasn't been a single online or print review that has been the basis for purchasing photography gear. In some cases (say the CV 50mm f/1.5) an aggregation of opinions helps. but in almost all cases seeing how others are using things and the results they are achieving helps me make decisions (sorta like "what one man can do, another can repeat"). Technical information usually isn't helpful. Doesn't matter to me if the lens is out performing the sensor, vice versa, or both, or neither- but it is interesting to read about such things.

 

On an unrelated note, do you do any videography? If so, can I PM you?

[otto.f]

Well if there was a Steve Huff Price of Writing Style in Lens Reviews this man deserves the 2013 Price! Better than any TelSell spot. Compared to him Steve Huff is just whispering.

[01af]

I understand the limiting factors in a chain. Horse carriages are as fast as the slowest horse, etc. [...] Nonetheless, if he was putting things in terms for laymen to understand, doesn't that imply some basic truth to what he was saying?

Sigh ... I've written so many words here, and all for the birds

 

No, the "slowest horse" analogy is exactly how it does NOT work. So don't expect Peter Karbe (or anyone) to bother you (or anyone) with elementary truths.

[Advertisement]

Hallo,

Look here: Leica M

[CaptZoom]

Sigh ... I've written so many words here, and all for the birds

 

No, the "slowest horse" analogy is exactly how it does NOT work. So don't expect Peter Karbe (or anyone) to bother you (or anyone) with elementary truths.

 

Ok señor. I suppose I should get you in touch with my university physics professors. Their teaching methods and knowledge apparently needs to be updated.