The end of legacy lenses???

[sportyaccordy]

Image software correction is necessary to provide images acceptable even to the uncritical given the fundamentally flawed choice of short flange focal distances in mirrorless cameras.

"Fundamentally flawed?" What the what?

Shorter registration distances are not new. They were common in rangefinder designs, and for decades they produced technically better lenses. Well into the film SLR era, several manufacturers - Konica, Mamiya, Leica, Contax, Fuji, Bronica - deliberately produced rangefinders because they offered better image quality (as well as quieter operation, no blackout period, more than 100% viewfinder coverage etc).

Yes, but they are film cameras, and silver crystals do not care which direction the light comes from. Silicon wells do care.

Not with a BSI sensor. A7RII works great with wide angle Leica M glass. Not going to bother with the rest

Yeah, sure it does. The NEX was fantastic, until it was tested, and the A7 was the answer, until it was tested, and now the A7rII is better than the second coming ... even Michael Reichmann, who lives in Sony's pocket, is not convinced until it is tested.

Neither the NEX nor A7 were claimed to work well with Leica lenses.... and the A7RII has been shown to work very well with Leica M lenses. I would not be surprised if it exhibited less vignetting as well. The BSI tech enables shallower microlenses which are more forgiving with respect to oblique light ray angles. So I'm not really sure what you're talking about here.

Heck, even in your link, Michael Reichmann said this:

I’ll close this section by adding a comment about use of Leica M lenses on the A7RII. I personally have not yet tried this, and therefore can not comment knowledgably. The new BSI sensor on this body may perform differently than previous ones, but at the time of this writing there are only a couple of online commentaries, though these appear to be mostly positive.

So I'm at a loss as to what your point was.

 

[Sante Patate]

Nearly all the testing you see on the internet is done with Imatest software. A basic set of Imatest tools is quite inexpensive and it is easy to get results that look as though you have done a scientific test. You haven't -

Nonsense. A properly conducted Imatest analysis is a scientific test with a higher margin of systematic error than a properly conducted test on an optical bench.

There will be two sources of error involved in the measurements: systematic error and random error. Imatest's use of the whole imaging chain creates systematic errors (to call them that for the moment), but I have never seen an Imatest result with 95% confidence intervals, so I don't know what the random errors are like.

The point about Imatest's "systematic errors" is that they are not, except by courtesy. Imatest tests the whole imaging chain - not excluding the photographer - so calling its results lens tests is misleading. A lot of confusion is caused by folk who think Imatest results are lens tests. I recognise that the more careful Imatest users have discouraged that - to varying degrees and with limited success.

for a host of reasons discussed carefully here.

Which essentially says that optical bench tests have a lower margin of error than Imatest tests. Poorly conducted Imatest tests have a higher margin of error than well-conducted Imatest tests which have a higher margin of error than optical bench tests.

No, there are two key issues that have nothing to do with margins of error:

- Imatest is done at short focus distances - often near closest focus. Lenses, except those specifically designed for close focus, are most often used at distances approximating "infinity", for which they are designed, and many perform quite differently at close focus and "infinity".

- Imatest measures "sharpness" as the modulation frequency at extinction - called, wrongly, MTF. This is nonsense: extinction modulation is not (contrary to what Norman Koren says) a good predictor of perceived sharpness even among the various one-number indices that could be derived from the MTF (and even if it was ignoring astigmatism would make it practically worthless). The fact that extinction modulation is a poor choice for photographic purposes can be easily shown: these two images have identical extinction modulation but obviously different sharpness.


From: Allen & Triantaphillidou, Manual of Photography, 10th edition, Focal Press, 2009.

The most common measure of resolution is the MTF-50 test, which measures lw/ph (line widths per picture height) at 50% contrast. The example you post above clearly shows the left as having a higher MTF-50 than the right, even if their extinction resolution (resolution at a much lower level of contrast) appears to be the same.

The fact that something is common does not make it correct, as gay people point out.

MTF stands for "modulation transfer function". It is well explained technically here .

As the use of "function" tells you, it is how modulation transfer varies as a function of something - for images, modulation frequency. For an aberration free lens modulation transfer falls nearly linearly with modulation frequency, until a limiting frequency is reached which is set by the aperture and the wavelength of the light (Figure 3 in the Florida State link). Absolute resolving power has always been defined as the frequency where the MTF is zero or very low (<10%) - ie, the Rayleigh criterion.

"Sharpness" is a subjective impression (Ansel Adams, in his book The Camera, puts the word in quotes - at p73 - for this reason). The key point is that it has very little to do with resolution. If you look at the images on p17 of Nasse's article about MTFs you can see the same point made in the image I copied: macro contrast has little or no connection to micro contrast.

From a psycho-physical point of view the question is whether a single index can be obtained from the MTF that correlates with viewers' judgements of "sharpness". There is not, because perceived sharpness depends on all sorts of characteristics of the image, such as the subject (portraits are perceived as sharper than landscapes), B&W vs colour, and contrast and viewing distance (as viewing distance increases higher contrast images are perceived as sharper).

Resolution is measured as line pairs per mm. Only target reproduction tests (eg, Imatest) use lw/ph because they cannot separate the sensor + processing algorithm from the lens.

 

[Great Bustard]

Nearly all the testing you see on the internet is done with Imatest software. A basic set of Imatest tools is quite inexpensive and it is easy to get results that look as though you have done a scientific test. You haven't -

Nonsense. A properly conducted Imatest analysis is a scientific test with a higher margin of systematic error than a properly conducted test on an optical bench.

There will be two sources of error involved in the measurements: systematic error and random error. Imatest's use of the whole imaging chain creates systematic errors (to call them that for the moment), but I have never seen an Imatest result with 95% confidence intervals, so I don't know what the random errors are like.

The point about Imatest's "systematic errors" is that they are not, except by courtesy. Imatest tests the whole imaging chain - not excluding the photographer - so calling its results lens tests is misleading. A lot of confusion is caused by folk who think Imatest results are lens tests. I recognise that the more careful Imatest users have discouraged that - to varying degrees and with limited success.

for a host of reasons discussed carefully here.

Which essentially says that optical bench tests have a lower margin of error than Imatest tests. Poorly conducted Imatest tests have a higher margin of error than well-conducted Imatest tests which have a higher margin of error than optical bench tests.

No, there are two key issues that have nothing to do with margins of error:

- Imatest is done at short focus distances - often near closest focus. Lenses, except those specifically designed for close focus, are most often used at distances approximating "infinity", for which they are designed, and many perform quite differently at close focus and "infinity".

- Imatest measures "sharpness" as the modulation frequency at extinction - called, wrongly, MTF. This is nonsense: extinction modulation is not (contrary to what Norman Koren says) a good predictor of perceived sharpness even among the various one-number indices that could be derived from the MTF (and even if it was ignoring astigmatism would make it practically worthless). The fact that extinction modulation is a poor choice for photographic purposes can be easily shown: these two images have identical extinction modulation but obviously different sharpness.


From: Allen & Triantaphillidou, Manual of Photography, 10th edition, Focal Press, 2009.

The most common measure of resolution is the MTF-50 test, which measures lw/ph (line widths per picture height) at 50% contrast. The example you post above clearly shows the left as having a higher MTF-50 than the right, even if their extinction resolution (resolution at a much lower level of contrast) appears to be the same.

The fact that something is common does not make it correct...

True, but MTF-50 tests do measure resolution in the form of line widths per picture height at 50% contrast.

...as gay people point out.

Not sure what gay people have to do with it.

MTF stands for "modulation transfer function". It is well explained technically here .

OK.

As the use of "function" tells you, it is how modulation transfer varies as a function of something - for images, modulation frequency. For an aberration free lens modulation transfer falls nearly linearly with modulation frequency, until a limiting frequency is reached which is set by the aperture and the wavelength of the light (Figure 3 in the Florida State link). Absolute resolving power has always been defined as the frequency where the MTF is zero or very low (<10%) - ie, the Rayleigh criterion.

"Sharpness" is a subjective impression (Ansel Adams, in his book The Camera, puts the word in quotes - at p73 - for this reason). The key point is that it has very little to do with resolution. If you look at the images on p17 of Nasse's article about MTFs you can see the same point made in the image I copied: macro contrast has little or no connection to micro contrast.

From a psycho-physical point of view the question is whether a single index can be obtained from the MTF that correlates with viewers' judgements of "sharpness". There is not, because perceived sharpness depends on all sorts of characteristics of the image, such as the subject (portraits are perceived as sharper than landscapes), B&W vs colour, and contrast and viewing distance (as viewing distance increases higher contrast images are perceived as sharper).

Perceived sharpness is another matter all together. I am talking about your test target above and how resolution relates to an MTF-50 test.

Resolution is measured as line pairs per mm.

Resolution on the sensor is measured in lp/mm (line pairs per millimeter), but, again, the contrast needs to be specified. MTF-50 tests show resolution (lw/ph -- line widths per picture height) at a given contrast (50%). Manufacturer MTF curves show contrast for a given resolution on the sensor (typically 10 lp/mm and 30 lp/mm although Olympus correctly uses 20 lp/mm and 60 lp/mm for mFT).

Only target reproduction tests (eg, Imatest) use lw/ph because they cannot separate the sensor + processing algorithm from the lens.

It is true that the sensor plays a significant role in the scores for an MTF-50 test. However, so long as the same sensor and processing is used on the MTF-50 tests, they test results will correspond to the innate differences between the lenses.

So, for example, the MTF-50 score for a lens on the 50 MP 5DsR sensor will be higher than for the same lens on the 50 MP 5Ds sensor which will be much higher than for the same lens on a 20 MP 6D sensor.

 

[dahod]

Imatest themselves, on their website, recognize that they do not test lenses but rather camera systems.

From their website here comes the statement

"You cannot measure a lens in isolation. It is a part of an imaging system that includes the camera’s image sensor, RAW converter (which may sharpen the image), and signal processing pipeline."

That's been my point all along - we have no idea how good these new lenses really are - we only know how good they are on the camera you're using at the moment with the software it currently has loaded. Many are quick to point out that's all it's really about anyway and are likely frustrated and exasperated with my thread - sorry.

Olympus is replacing their top of the line SGH lenses with their PRO lens line at roughly half the price. My guess is that they're not getting the cost reductions through manufacturing improvements but rather through taking some of the lens's job into the software. Not a bad idea necessarily - it doesn't sound like many users really care and we certainly benefit from the lower costs - but it does have implications for new lenses that we didn't see a few years ago.

Have a good day

Dave

 

[Sante Patate]

Nearly all the testing you see on the internet is done with Imatest software. A basic set of Imatest tools is quite inexpensive and it is easy to get results that look as though you have done a scientific test. You haven't -

Nonsense. A properly conducted Imatest analysis is a scientific test with a higher margin of systematic error than a properly conducted test on an optical bench.

There will be two sources of error involved in the measurements: systematic error and random error. Imatest's use of the whole imaging chain creates systematic errors (to call them that for the moment), but I have never seen an Imatest result with 95% confidence intervals, so I don't know what the random errors are like.

The point about Imatest's "systematic errors" is that they are not, except by courtesy. Imatest tests the whole imaging chain - not excluding the photographer - so calling its results lens tests is misleading. A lot of confusion is caused by folk who think Imatest results are lens tests. I recognise that the more careful Imatest users have discouraged that - to varying degrees and with limited success.

for a host of reasons discussed carefully here.

Which essentially says that optical bench tests have a lower margin of error than Imatest tests. Poorly conducted Imatest tests have a higher margin of error than well-conducted Imatest tests which have a higher margin of error than optical bench tests.

No, there are two key issues that have nothing to do with margins of error:

- Imatest is done at short focus distances - often near closest focus. Lenses, except those specifically designed for close focus, are most often used at distances approximating "infinity", for which they are designed, and many perform quite differently at close focus and "infinity".

- Imatest measures "sharpness" as the modulation frequency at extinction - called, wrongly, MTF. This is nonsense: extinction modulation is not (contrary to what Norman Koren says) a good predictor of perceived sharpness even among the various one-number indices that could be derived from the MTF (and even if it was ignoring astigmatism would make it practically worthless). The fact that extinction modulation is a poor choice for photographic purposes can be easily shown: these two images have identical extinction modulation but obviously different sharpness.


From: Allen & Triantaphillidou, Manual of Photography, 10th edition, Focal Press, 2009.

The most common measure of resolution is the MTF-50 test, which measures lw/ph (line widths per picture height) at 50% contrast. The example you post above clearly shows the left as having a higher MTF-50 than the right, even if their extinction resolution (resolution at a much lower level of contrast) appears to be the same.

The fact that something is common does not make it correct...

True, but MTF-50 tests do measure resolution in the form of line widths per picture height at 50% contrast.

No, they measure MTF-50, in one direction.

...as gay people point out.

Not sure what gay people have to do with it.

They invented the slogan "Straight's not normal, just common".

MTF stands for "modulation transfer function". It is well explained technically here .

OK.

As the use of "function" tells you, it is how modulation transfer varies as a function of something - for images, modulation frequency. For an aberration free lens modulation transfer falls nearly linearly with modulation frequency, until a limiting frequency is reached which is set by the aperture and the wavelength of the light (Figure 3 in the Florida State link). Absolute resolving power has always been defined as the frequency where the MTF is zero or very low (<10%) - ie, the Rayleigh criterion.

"Sharpness" is a subjective impression (Ansel Adams, in his book The Camera, puts the word in quotes - at p73 - for this reason). The key point is that it has very little to do with resolution. If you look at the images on p17 of Nasse's article about MTFs you can see the same point made in the image I copied: macro contrast has little or no connection to micro contrast.

From a psycho-physical point of view the question is whether a single index can be obtained from the MTF that correlates with viewers' judgements of "sharpness". There is not, because perceived sharpness depends on all sorts of characteristics of the image, such as the subject (portraits are perceived as sharper than landscapes), B&W vs colour, and contrast and viewing distance (as viewing distance increases higher contrast images are perceived as sharper).

Perceived sharpness is another matter all together. I am talking about your test target above and how resolution relates to an MTF-50 test.

Whether a detail in an image will be visible at any given level of contrast reproduction depends on its contrast. For a scene with lots of detail made up of very subtle shades of grey you would need higher modulation transfer than for a scene with equally sized detail made up of black and white. The catch is that the graph of modulation transfer vs frequency is not a straight line for real lens: it sags below and to the left of the aberration-free line in a complex way. The frequency associated with 50% modulation transfer does not tell you what the frequency associated with 90% or 10% transfer will be.

As long as there is some white and some black between the lines it does not matter how wide the grey fringes are: modulation transfer is still 100%. So the middle image of the three on p17 of Nasse's article means high modulation transfer at low and medium frequencies with a sharp drop off at some high frequency (same as the image above). The image of the contrast transition un-sharp but accurate (the squares, or the lines, are black and white). Conversely, if the modulation transfer is medium at all frequencies you get the right-most of Nasse's images: a very sharply but inaccurately imaged contrast transition.

To illustrate the photographic effect of this, here is another image:




Allen and Triantaphillidou, Manual of Photography, 10th ed, Focal Press, p347.

The right-hand image has higher resolution but lower contrast, but the left hand image has higher contrast but lower resolution. In the terms of Nasse's chequer-boards the left-hand king is the central one, and the right-hand king is the right-hand chequer-board. If you look from close-up the right-hand king is perceived as sharper, but if you get further away, where the detail in the right-hand image is no longer resolved by the eye, the left-hand king is perceived as sharper.

Only target reproduction tests (eg, Imatest) use lw/ph because they cannot separate the sensor + processing algorithm from the lens.

It is true that the sensor plays a significant role in the scores for an MTF-50 test. However, so long as the same sensor and processing is used on the MTF-50 tests, they test results will correspond to the innate differences between the lenses.

Yes, but the MTF-50 will still be a poor metric.

So, for example, the MTF-50 score for a lens on the 50 MP 5DsR sensor will be higher than for the same lens on the 50 MP 5Ds sensor which will be much higher than for the same lens on a 20 MP 6D sensor.

Yes - but that tells you precisely nothing.

 

[Great Bustard]

Nearly all the testing you see on the internet is done with Imatest software. A basic set of Imatest tools is quite inexpensive and it is easy to get results that look as though you have done a scientific test. You haven't -

Nonsense. A properly conducted Imatest analysis is a scientific test with a higher margin of systematic error than a properly conducted test on an optical bench.

There will be two sources of error involved in the measurements: systematic error and random error. Imatest's use of the whole imaging chain creates systematic errors (to call them that for the moment), but I have never seen an Imatest result with 95% confidence intervals, so I don't know what the random errors are like.

The point about Imatest's "systematic errors" is that they are not, except by courtesy. Imatest tests the whole imaging chain - not excluding the photographer - so calling its results lens tests is misleading. A lot of confusion is caused by folk who think Imatest results are lens tests. I recognise that the more careful Imatest users have discouraged that - to varying degrees and with limited success.

for a host of reasons discussed carefully here.

Which essentially says that optical bench tests have a lower margin of error than Imatest tests. Poorly conducted Imatest tests have a higher margin of error than well-conducted Imatest tests which have a higher margin of error than optical bench tests.

No, there are two key issues that have nothing to do with margins of error:

- Imatest is done at short focus distances - often near closest focus. Lenses, except those specifically designed for close focus, are most often used at distances approximating "infinity", for which they are designed, and many perform quite differently at close focus and "infinity".

- Imatest measures "sharpness" as the modulation frequency at extinction - called, wrongly, MTF. This is nonsense: extinction modulation is not (contrary to what Norman Koren says) a good predictor of perceived sharpness even among the various one-number indices that could be derived from the MTF (and even if it was ignoring astigmatism would make it practically worthless). The fact that extinction modulation is a poor choice for photographic purposes can be easily shown: these two images have identical extinction modulation but obviously different sharpness.


From: Allen & Triantaphillidou, Manual of Photography, 10th edition, Focal Press, 2009.

The most common measure of resolution is the MTF-50 test, which measures lw/ph (line widths per picture height) at 50% contrast. The example you post above clearly shows the left as having a higher MTF-50 than the right, even if their extinction resolution (resolution at a much lower level of contrast) appears to be the same.

The fact that something is common does not make it correct...

True, but MTF-50 tests do measure resolution in the form of line widths per picture height at 50% contrast.

No, they measure MTF-50, in one direction.

They measure resolution (lw/ph) at a given level of contrast (50%). I'm not sure if they test in only one or average multiple directions, however.

...as gay people point out.

Not sure what gay people have to do with it.

They invented the slogan "Straight's not normal, just common".

Ah.

MTF stands for "modulation transfer function". It is well explained technically here .

OK.

As the use of "function" tells you, it is how modulation transfer varies as a function of something - for images, modulation frequency. For an aberration free lens modulation transfer falls nearly linearly with modulation frequency, until a limiting frequency is reached which is set by the aperture and the wavelength of the light (Figure 3 in the Florida State link). Absolute resolving power has always been defined as the frequency where the MTF is zero or very low (<10%) - ie, the Rayleigh criterion.

"Sharpness" is a subjective impression (Ansel Adams, in his book The Camera, puts the word in quotes - at p73 - for this reason). The key point is that it has very little to do with resolution. If you look at the images on p17 of Nasse's article about MTFs you can see the same point made in the image I copied: macro contrast has little or no connection to micro contrast.

From a psycho-physical point of view the question is whether a single index can be obtained from the MTF that correlates with viewers' judgements of "sharpness". There is not, because perceived sharpness depends on all sorts of characteristics of the image, such as the subject (portraits are perceived as sharper than landscapes), B&W vs colour, and contrast and viewing distance (as viewing distance increases higher contrast images are perceived as sharper).

Perceived sharpness is another matter all together. I am talking about your test target above and how resolution relates to an MTF-50 test.

Whether a detail in an image will be visible at any given level of contrast reproduction depends on its contrast.

Resolution and contrast are flip sides of the same coin. You measure resolution at a certain contrast (e.g. MTF-50) or measure contrast at a certain resolution (e.g. manufacturer MTF charts).

For a scene with lots of detail made up of very subtle shades of grey you would need higher modulation transfer than for a scene with equally sized detail made up of black and white. The catch is that the graph of modulation transfer vs frequency is not a straight line for real lens: it sags below and to the left of the aberration-free line in a complex way. The frequency associated with 50% modulation transfer does not tell you what the frequency associated with 90% or 10% transfer will be.

Sure -- no one said or implied otherwise.

As long as there is some white and some black between the lines it does not matter how wide the grey fringes are: modulation transfer is still 100%. So the middle image of the three on p17 of Nasse's article means high modulation transfer at low and medium frequencies with a sharp drop off at some high frequency (same as the image above). The image of the contrast transition un-sharp but accurate (the squares, or the lines, are black and white). Conversely, if the modulation transfer is medium at all frequencies you get the right-most of Nasse's images: a very sharply but inaccurately imaged contrast transition.

To illustrate the photographic effect of this, here is another image:


Allen and Triantaphillidou, Manual of Photography, 10th ed, Focal Press, p347.

The right-hand image has higher resolution but lower contrast, but the left hand image has higher contrast but lower resolution. In the terms of Nasse's chequer-boards the left-hand king is the central one, and the right-hand king is the right-hand chequer-board. If you look from close-up the right-hand king is perceived as sharper, but if you get further away, where the detail in the right-hand image is no longer resolved by the eye, the left-hand king is perceived as sharper.

OK.

Only target reproduction tests (eg, Imatest) use lw/ph because they cannot separate the sensor + processing algorithm from the lens.

It is true that the sensor plays a significant role in the scores for an MTF-50 test. However, so long as the same sensor and processing is used on the MTF-50 tests, they test results will correspond to the innate differences between the lenses.

Yes, but the MTF-50 will still be a poor metric.

DxOMark used to publish the resolutions at various levels of contrast. I never saw, not even once, a situation where System A had a higher MTF-50 than System B, but a lower MTF-XX. Now, the differences in resolution at 50% contrast and other contrast levels varied, of course, but, as I said, if System A recorded a higher MTF-50 than System B, then it was higher at all other levels of contrast.

So, for example, the MTF-50 score for a lens on the 50 MP 5DsR sensor will be higher than for the same lens on the 50 MP 5Ds sensor which will be much higher than for the same lens on a 20 MP 6D sensor.

Yes - but that tells you precisely nothing.

It tells us that more pixels result in greater resolution regardless of how sharp the lens is, all else equal. It also tells us, based on the relative numbers, that the sharper the lens is, the greater the resolution increase, and the less sharp the lens is, the lower the increase in resolution.