Jump to content

MTF Curves ~ Sensor Resolution


k-hawinkler

Recommended Posts

Advertisement (gone after registration)

I am interested in quantitatively comparing the resolution of Leica lenses with the resolution of digital sensors. So, I would like to know how one can determine whether a given lens outresolves a given sensor and vice versa, or whether the resolution of lens and sensor are comparable.

 

For example, provided I didn't make an error, the

 

• Leica M9 has about 145 pixel/mm linear resolution, whereas the

• Nikon D800E has 205 pixel/mm, and the

• Sony NEX-7 has 255 pixel/mm.

 

MTF curves for some Leica R-lenses can be found here: Leica Camera AG - Downloads

 

For example, the

 

• LEICA APO-TELYT-R 280 mm f/4 has this MTF graph wide open, and the

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

• LEICA SUMMILUX-R 80 mm f/1.4 has this one wide open:

 

 

How would one interpret these two types of data, namely linear resolution and MTF curves, and compare the two?

Or is there other type of data available that is more suited for the task at hand?

 

Many thanks for your explanations.

Edited by k-hawinkler
Link to post
Share on other sites

I am interested in quantitatively comparing the resolution of Leica lenses with the resolution of digital sensors. So, I would like to know how one can determine whether a given lens outresolves a given sensor and vice versa ...

I wrote so many articles in this topic here in the recent two years ... obviously it was entirely futile. :(

 

So let me repeat it once again: Lenses don't outresolve sensors, and sensors don't outresolve lenses. Better lenses are better on any sensor and better sensors are better behind any lens.

 

 

... or whether the resolution of lens and sensor are comparable.

No, they are not.

 

 

Do Sensors “Outresolve” Lenses? by Rubén Osuna and Efraín García on Luminous Landscape.

This useless article is full of layman's errors and misconceptions. The maximum useful sensor resolution is not a matter of absolute limits but of return on investment. How much are you willing to spend for further (small) improvement if your current system already is very good?

  • Thanks 1
Link to post
Share on other sites

Basically 01AF is right, but there is of course an interaction. If you carry this to extremes, it makes no sense to use a 500 Lp/mm lens on an 1 Mp sensor, and the other way around, a Coke bottle bottom on a 24 Mp sensor will not be very impressive either.

The basics are explained in the old Erwin Puts compendium, in the article where he compares MTF curves of film and lenses. That can be expanded to sensor resolution, if you wish. (Which EP does not do)

The old rule of thumb that in general a lens resolution of over 80 Lp/mm (40 according to Zeiss) has no photographic relevance is still more or less valid, though.

  • Like 1
Link to post
Share on other sites

Advertisement (gone after registration)

I am interested in quantitatively comparing the resolution of Leica lenses with the resolution of digital sensors. So, I would like to know how one can determine whether a given lens outresolves a given sensor and vice versa, or whether the resolution of lens and sensor are comparable.

 

For example, provided I didn't make an error, the

 

• Leica M9 has about 145 pixel/mm linear resolution, whereas the

• Nikon D800E has 205 pixel/mm, and the

• Sony NEX-7 has 255 pixel/mm....

.

 

Without entering on details, regarding standard photographic sensors one has always to keep in mind that, owing to Bayer filter, "one" pixel is indeed recorded as an interpolation of four (a 2x2 square of) , which, basically, halves the linear resolution.

Link to post
Share on other sites

... one has always to keep in mind that, owing to Bayer filter, "one" pixel is indeed recorded as an interpolation of four (a 2×2 square of), which, basically, halves the linear resolution.

This is not quite true. The 2×2 Bayer cluster of photosites is the basic picture element in terms of colour but not in terms of luminance recording. So overall, linear resolution is not reduced by a factor of 2 but rather a factor of, say, 1.4 or 1.5.

Link to post
Share on other sites

This is not quite true. The 2×2 Bayer cluster of photosites is the basic picture element in terms of colour but not in terms of luminance recording. So overall, linear resolution is not reduced by a factor of 2 but rather a factor of, say, 1.4 or 1.5.

 

Right... I said it in simple terms, but within math terms (theory of information) a value around 1,5 is more correct.

Link to post
Share on other sites

Lenses don't outresolve sensors, and sensors don't outresolve lenses. Better lenses are better on any sensor and better sensors are better behind any lens.The maximum useful sensor resolution is not a matter of absolute limits but of return on investment. How much are you willing to spend for further (small) improvement if your current system already is very good?

I rather like this, my condensed version of 01af's post, which creates a rather succinct statement and it makes extremely good sense to me. I would suggest though, that starting with a better lens first probably makes more sense than the other way around......

Link to post
Share on other sites

How would one interpret these two types of data, namely linear resolution and MTF curves, and compare the two?

 

One really can't. MTF charts graph CONTRAST at several specific resolutions, that usually are no higher than 40 lppm. Since any decent 35mm-format lens usually resolves more like 60 lppm, and really good ones hit well above 80 lppm at some apertures, the MTF charts "clip" or dumb-down the data.

 

I.E. you can't tell what the max resolution of a particular lens is from an MTF chart - only that at the moderate resolution level of 40 lppm it has high or low contrast.

 

For example, in this particular test sample of the Canon vs. Leica-M 35 f/1.4 lenses from the image corners (baby images), the Canon has higher edge contrast and might well deliver a higher MTF curve. but the Leica lens shows more resolution, albeit with duller contrast:

 

http://www.l-camera-forum.com/leica-forum/customer-forum/199670-35-summilux-asph-vs-canon-35-a.html#post1824983

  • Like 1
Link to post
Share on other sites

Gentlemen,

 

Thanks for the feedback. However unfortunately I may have formulated my question, nevertheless, I am glad I started this thread as I already learned a few things. To keep on track, please permit me to frame the discussion.

 

It looks to me, that the subject I had in mind, can be broken down into several pieces, that can be analyzed separately. Here we go (and I am sure any mistakes I make will be corrected):

 

 

• How to translate linear sensor resolution [pixel/mm] into line pairs per mm [lp/mm]?

 

Based on the above feedback it looks like one divides linear sensor resolution by 1.5. So we get:

 

• Leica M9: 145 pixel/mm corresponds to 96 lp/mm.

• Nikon D800E: 205 pixel/mm corresponds to 137 lp/mm.

• Sony NEX-7: 255 pixel/mm corresponds to 171 lp/mm.

 

Question: Can someone please briefly indicate the derivation of the 1.5 number or point to a reference that does explain it? Thanks.

 

 

• Lens resolution limits and where to find the information?

 

In the LEICA COMPENDIUM by Erwin Puts, 2011, ISBN/EAN: 978-94-91089-02-2, page 562, he states that the

 

APO-Telyt-R 280mm f/4 is diffraction limited and has a resolution limit of of a breath-taking 500 lp/mm at full aperture.

 

Additional lenses of interest to me include:

 

(1965) Elmar 65/3.5

 

(1970) Summicron v2 90/2

(1960) Elmarit 90/2.8

(1965) Tri-Element Elmar 90/4

 

(1979) Elmarit-M v3 135/2.8

(1970) Tele-Elmar 135/4

 

as well as:

 

(1963) Telyt-V 200/4

 

(1965) Telyt-V v2 280/4.8

(1975) Telyt-V v3 280/4.8

 

(1967) Telyt-V 400/5.6

(1969) Telyt-V 400/6.8

 

(1968) Telyt-V 560/5.6

 

In http://www.l-camera-forum.com/leica-forum/2021715-post15.html Mike (michali) has listed a number of R-lenses that he had leitaxed. I think these lenses would also be of interest in the current context. The lenses are:

 

19mm Elmarit vers.2

28-90mm Vario Elmar

105-280mm Vario Elmar

80mm Summilux

100mm APO Macro Elmarit

280mm f4 APO-Telyt-R

1.4x APO-Extender-R

 

I would like to add the following R-lenses:

 

2x APO-Extender-R

180mm f2.8 APO-Telyt-R

180mm f3.4 APO-Telyt-R

280mm f2.8 APO-Telyt-R

 

I plan to go through the Puts' Compendium to see whether I can find more lens resolution limits but would appreciate if folks listed here those limits that they know about.

 

 

• Lens resolution limits, sensor resolution

 

It's comforting to know that:

 

Lenses don't outresolve sensors, and sensors don't outresolve lenses.

Better lenses are better on any sensor and better sensors are better behind any lens.

 

Question: Can the last sentence be quantified? If so, how? What is considered an optimal match?

 

 

• Errors and misconceptions

 

01af states that the article Do Sensors “Outresolve” Lenses? is full of layman's errors and misconceptions.

 

Question: Could you please elaborate and briefly describe the 2 or 3 most egregious mistakes?

 

 

Thanks again for the feedback already given and hopefully more to come.

Edited by k-hawinkler
Link to post
Share on other sites

After post #12, my head hurts. ;)

 

However:

 

Question: Can someone please briefly indicate the derivation of the 1.5 number or point to a reference that does explain it?

 

It is an approximation. We know the resolution of a sensor is not 1:1 with the spacing of the pixels, for a variety of reasons - Bayer pattern color distinctions; AA filters, or the aliasing artifacts one gets without AA filters, etc. We also know that it is often better, depending on subject detail, than the 1:2 relationship known as the Nyquist limit (at least 2 samples for every captured value - i.e. line of lens resolution). Therefore it is some number between 1 and 2, and 1.5 is the shorthand for "somewhere between 1 and 2".

 

One might also note that since image details don't necessarily line up with the grid of pixels, in some cases the effective "resolution" will be between the centers of pixels with touching corners, which will be root 2 (1.41) times the sides of the pixels.

 

"So we get:

 

• Leica M9: 145 pixel/mm corresponds to 96 lp/mm.

• Nikon D800E: 205 pixel/mm corresponds to 137 lp/mm.

• Sony NEX-7: 255 pixel/mm corresponds to 171 lp/mm.

 

... so long as one does not take into account any AA filtering, or other blurring. Some say there is noticeable loss of per-pixel resolution between the M8 and the M9, simply because the M9 has a thicker cover glass/IR filter (0.7mm vs. the M8's 0.5mm - I don't see it myself). The Nikon D800E uses a "reversing" AA filter - one blurring layer that supposedly cancels out the effect of the other blurring layer - which I doubt works perfectly, but no doubt restores some sharpness. The Sony Nex-7 has - who knows what level of AA filtering (fairly little, from the samples I've seen, but probably not zero).

 

It seems to me you are trying to come up with hard, quantitative results based on data that is going to, by necessity (and literally, when it comes to AA filters) always be "fuzzy." Lots of things that Sony or Nikon or Leica know, but are not public domain.

 

You can probably find Modern or Pop. Photography reviews** of some of your earlier lenses that gives the resolution the magazine measured on an optical bench (they used to do that, pre-1980s). But if the review reports "87 lppm at f/8" - well, the next lens off the assembly line might have tested at 83 lines, or 90 lines.

 

**My college library had MP or PP back-issues, in hard copy, from 1945-1980 at one point.

  • Like 3
Link to post
Share on other sites

Interesting thread.

Andy, what do you think of the difference between measured and calculated MTF-curves?

Those lenses with measured results can be better than that, because this is the tolerance for the

quality assurance.

Calculated ones are the wishes of the marketing people, these lenses must not be as good as these curves. These curves are seldom published.

Jan

Link to post
Share on other sites

After post #12, my head hurts. ;)

It does? So does mine :rolleyes:

 

 

Question: Can someone please briefly indicate the derivation of the 1.5 number or point to a reference that does explain it?

 

It is an approximation. We know the resolution of a sensor is not 1:1 with the spacing of the pixels, for a variety of reasons—Bayer pattern color distinctions; AA filters, or the aliasing artifacts one gets without AA filters, etc. We also know that it is often better, depending on subject detail, than the 1:2 relationship known as the Nyquist limit (at least two samples for every captured value—i. e. line of lens resolution). Therefore it is some number between 1 and 2, and 1.5 is the shorthand for "somewhere between 1 and 2".

Oh dear. Where to start? Both you and 'k-hawinkler' are badly confusing things like lines, line pairs, Bayer patterns, Nyquist-Shannon, and a few more and mixing them up into a melange of myths, factoids, and misconceptions which is just short of impossible to sort out.

 

Among other things, line pairs are not lines. The Nyquist-Shannon theorem does not say 'at least two samples per cycle;' actually it says, 'more than two samples per cycle.' And 1.5 is not shorthand for 'somewhere between 1 and 2.' The number quoted was '1.4 or 1.5,' and that's what it meant.

 

 

So we get:

 

• Leica M9: 145 pixel/mm corresponds to 96 lp/mm.

• Nikon D800E: 205 pixel/mm corresponds to 137 lp/mm.

• Sony NEX-7: 255 pixel/mm corresponds to 171 lp/mm.

Divide these results by two, then you're closer to what we actually get—in theory, in the general case (special cases not withstanding). After all, it's line pairs, not lines. This will be still not necessarily the same what we get in real life, because what we really get depends on many factors, including (but not limited to) the minutiae of the in-camera image processing as well as contrast, orientation, spatial phase, and colour of the structures to reproduce.

 

And then, that's only what the sensor can do. This still is different from what you'll see in an image cast by a lens on the sensor. For the image resolution to approach the sensor's resolution, the lens must approach infinite resolution ... which is impossible obviously. So image resolution will be always less than the sensor's theoretical limit, and there is no such thing as a lens that 'outresolves' a sensor.

  • Like 3
Link to post
Share on other sites

After post #12, my head hurts. ;)

 

However:

 

Question: Can someone please briefly indicate the derivation of the 1.5 number or point to a reference that does explain it?

 

It is an approximation. We know the resolution of a sensor is not 1:1 with the spacing of the pixels, for a variety of reasons - Bayer pattern color distinctions; AA filters, or the aliasing artifacts one gets without AA filters, etc. We also know that it is often better, depending on subject detail, than the 1:2 relationship known as the Nyquist limit (at least 2 samples for every captured value - i.e. line of lens resolution). Therefore it is some number between 1 and 2, and 1.5 is the shorthand for "somewhere between 1 and 2".

 

One might also note that since image details don't necessarily line up with the grid of pixels, in some cases the effective "resolution" will be between the centers of pixels with touching corners, which will be root 2 (1.41) times the sides of the pixels.

 

[u]"So we get:

 

• Leica M9: 145 pixel/mm corresponds to 96 lp/mm.

• Nikon D800E: 205 pixel/mm corresponds to 137 lp/mm.

• Sony NEX-7: 255 pixel/mm corresponds to 171 lp/mm.[/u]

 

... so long as one does not take into account any AA filtering, or other blurring. Some say there is noticeable loss of per-pixel resolution between the M8 and the M9, simply because the M9 has a thicker cover glass/IR filter (0.7mm vs. the M8's 0.5mm - I don't see it myself). The Nikon D800E uses a "reversing" AA filter - one blurring layer that supposedly cancels out the effect of the other blurring layer - which I doubt works perfectly, but no doubt restores some sharpness. The Sony Nex-7 has - who knows what level of AA filtering (fairly little, from the samples I've seen, but probably not zero).

 

It seems to me you are trying to come up with hard, quantitative results based on data that is going to, by necessity (and literally, when it comes to AA filters) always be "fuzzy." Lots of things that Sony or Nikon or Leica know, but are not public domain.

 

You can probably find Modern or Pop. Photography reviews** of some of your earlier lenses that gives the resolution the magazine measured on an optical bench (they used to do that, pre-1980s). But if the review reports "87 lppm at f/8" - well, the next lens off the assembly line might have tested at 83 lines, or 90 lines.

 

**My college library had MP or PP back-issues, in hard copy, from 1945-1980 at one point.

 

 

Andy,

 

Thanks for your post. As always very informative. Sorry if I gave you a headache with this thread.

 

We got quite a bit of wet snow over night. It brought down a power line in the neighborhood and we were without power until a few minutes ago, so my computers were cut off the Internet.

 

Maybe I should explain briefly what I hope to get out of this thread. Basically, at a minimum, a rough estimate of Leica lens resolution limits as they developed over time, and hopefully accurate numbers for some lenses. Despite power loss, I was able to read on my iPad some older material of Erwin Puts that I had downloaded before and found some more useful information about Leica lens resolution limits. I plan to summarize that later.

 

If I understand you correctly, then the corresponding line pairs per mm numbers I computed for the three different sensors should be viewed as upper limits due to the effects you describe.

 

And 01af seems to imply to cut back the numbers even further by another factor of 2. Interesting.

 

With regards to the D800E, a Nikon manager stated in an interview, that for cost reasons in order to have the same register distance for the D800 and D800E, they left the first part of the filter in the D800E. This splits the signal in two. So, instead of splitting it again subsequently (in the other direction) as in the D800, they just recombine the split signals, effectively canceling the first split. I doubt Nikon will release the D800E to market unless they convinced themselves of a satisfactorily implementation.

 

Thanks also for the pointers to Modern or Pop. Photography reviews. I will see what my local library has.

 

Many thanks again. I appreciate your help.

Link to post
Share on other sites

Jan, Jaap,

 

Thanks. I found a statement by Erwin Puts about Leica and MTF curves that might be of interest.

In https://docs.google.com/viewer?a=v&q=cache:UyaxiCgv5JwJ:en.leica-camera.com/assets/file/download.php?filename%3Dfile_1742.pdf+Two+methods+are+being+used+to+get+these+MTF+figures.+One+method+does+measure+the+values+by+using+equipment+for+the+determination+of+the+contrast+transfer.+An+illuminated+target,+almost+always+a+very+narrow+slit+of+0.02mm+is+projected+through+the+lens+on+a+detector+(a+sensitive+surface,+nowadays+a+CCD+chip)+,+and+the+brightness+difference+at+the+edges+of+the+slit+(illuminated+and+non-illuminated+part)+is+recorded.+This+can+be+done+because+the+scanning+slit+is+much+smaller+than+the+target+slit+and+we+are+able+to+record+the+brightness+variation+along+the+edge.+Ideally+we+would+see+a+square+shape+as+the+transition+from+dark+to+white+is+abrupt.+Due+to+aberrations+we+see+a+slightly+rounded+off+top+with+a+gentle+slope.&hl=en&gl=us&pid=bl&srcid=ADGEESg2E9st9OCmxU9HX4SWdukJ97IUX1QQiUezK7g7i3Wc2jvZXAVYlicx5bkifF3amAK_JXMGqyYFt619SOugG4v4w8K0KdGBWqxC65JiZxEd64NZkt7GsUBmf7YMexea0Kz0pbI8&sig=AHIEtbRgpAGxxoKYi_6tDKmGDHkK5Pgq8A page 16, he states, quote:

 

"How are MTF measurements

actually obtained?

 

There are two methods: one method

computes MTF data, the other method

measures MTF values. Basically, there

are no differences, and Leica uses ei-

ther method, whichever is most appro-

priate: the optical design department

computes the MTF values, and the

manufacturing department uses an

MTF-measuring instrument to obtain

MTF data (see the diagrams). Both

methods are based on the same theo-

retical principles, so that their results

should not be different from one an-

other. A variance between the two val-

ues only occurs when the lens assem-

bly department can not conform to

calculated tolerance values."

  • Like 1
Link to post
Share on other sites

It does? So does mine :rolleyes:

 

 

For the image resolution to approach the sensor's resolution, the lens must approach infinite resolution ... which is impossible obviously. So image resolution will be always less than the sensor's theoretical limit, and there is no such thing as a lens that 'outresolves' a sensor.

 

Thumbs Up :cool: Mathematically RIGHT (*) and briefly said...

 

(*) paradoxically, even in a limit case as a 500.000 pixel 24x36 sensor with a superb Summilux 50 asph in front... ;)

Edited by luigi bertolotti
Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...