Opticallimits

Full Version: Megapixels and limits of glass
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Pages: 1 2 3 4

lightisbeauty

Good day all,



Wanted to approach this forum about any known, or perceived limitations on currently available glass for Canon/Nikon platforms against the quickly escalating megapixel count. I've heard it thrown around that at 15-18MP you're already pushing the limitations of the glass. Anyone have any data or opinions on the matter? As a Canon shooter I'm particularly concerned since they're known to be leaders in high MP count with bodies already in the 20s and a 30+MP 1Ds scheduled for next year.



Thanks again!

-Javier
[quote name='Light Is Beauty' timestamp='1287510592' post='3694']

Good day all,



Wanted to approach this forum about any known, or perceived limitations on currently available glass for Canon/Nikon platforms against the quickly escalating megapixel count. I've heard it thrown around that at 15-18MP you're already pushing the limitations of the glass. Anyone have any data or opinions on the matter? As a Canon shooter I'm particularly concerned since they're known to be leaders in high MP count with bodies already in the 20s and a 30+MP 1Ds scheduled for next year.



Thanks again!

-Javier

[/quote]

It is not correct to think that somehow sensors "outresolve" lenses.



Just take a peek at the tests from photozone, they illustrate this beautifully.



Take any lens tested on the Canon EOS 350D, and 8mp APS-C camera. Then take the same lens tested on the 15mp Canon EOS 50D.



You will ALWAYS see a higher resolution measured on the 50D, no matter how badly the lens performed on the 350D.

That trend will continue for quite some time yet... higher resolution sensors WILL resolve more detail, even with not so sharp lenses.



There should be no worry what so ever about full frame sensors of 30 to 50 mp.

Guest

[quote name='Light Is Beauty' timestamp='1287510592' post='3694']

Good day all,



Wanted to approach this forum about any known, or perceived limitations on currently available glass for Canon/Nikon platforms against the quickly escalating megapixel count. I've heard it thrown around that at 15-18MP you're already pushing the limitations of the glass. Anyone have any data or opinions on the matter? As a Canon shooter I'm particularly concerned since they're known to be leaders in high MP count with bodies already in the 20s and a 30+MP 1Ds scheduled for next year.



Thanks again!

-Javier

[/quote]



Since when is the second best called leader? But anyways, to your question - system resolution depends both on sensor resolution and lens resolution. Only when one is infinite is the other component completely limiting. What this means is that all lenses will benefit from higher MP count sensors - to some extent. Better lenses will profit more than not so good ones.



Additionally, what you are referring to are full frame bodies. If you upscale the MP count of APS-C sensors to full frame size, you end up somewhere close to 40 MP. So, I wouldn't be worried about a 30 MP full frame body at all - unless you need pin sharp corners wide open in 100% pixel view. But who needs that (and remember the paragraph above about system resolution)...?
As Brightcolours and BG_Home mentioned, we're not there yet, not by a long shot.



If I may add something to the discussion, it is really about several things.



1) Lens quality.

The best lenses out there, generally are diffraction limited by about F/4. Think the best Zeiss glass, Canon L primes, the Nikon professional range, the better Sony (Zeiss) glass. This means that these lenses will resolve right to the physical limit possible, which is about 400 lp/mm (line pairs per mm) at F/4, and the criterion being the Rayleigh criterion, which states that this is at 9% contrast, IOW, much less than the 50% MTFs we are looking at from manufacturers, but resulting in higher resolution than on those 50% charts. Do note that "lesser" lenses may not be diffraction limited, but modern prosumer lenses are often close anyway. 250 lp/mm is considered to be excellent.



2) Sensor, or medium (film, sensor, whatever) resolution.

The higher the number of pixels or silver halide crystals or photon counters, or whatever per square unit of area, the higher the resolution. However, there always is a compromise, namely noise levels with sensors, and iso with film. Furthermore, to get the resolution we get with the higher MP sensors these days, we need to shoot low iso B&W film when using analog. As good amateurs we used to get 20 to 40 lp/mm end results in the colour negative, professionals possibly double that, and slightly higher again with B&W, but you need to be prepared to shoot with tripods almost all the time if you really want to realise that sharpness.



Do note that final negative or RAW resolution is inversely proportional to the inverse individual resolutions of sensor and glass, or if said in a different way, 1 / total recorded resolution equals the sum of 1 / lens resolution at a specific f-stop + 1 / maximum recording resolution. You can express this in pixels per image height, as is done on this site, or in lines / mm, or as we used to do in the past, lp/mm, as one needs a minimum of 2 lines in order to depict an edge or line.



Due to the nature of the above formula, it follows that in order to get out the maximum of the one element in that formula, resolution needs to be infinite for the other component. This again is a physical impossibility. The maximum resolution possible is roughly equal to half the wavelength of light, but as we are using the full visible spectrum, this limits us really to the upper end of about 700 to 750 nanometer. Or about 530 nanameter if we use green light, to which the eye is adapted best. 550 nanometer equals about 2 microns in wavelength, which means that maximum resoltion is about 1 micron. As we require 2 of those to see a line, that means that typically maximum resolution is 500 lp/mm just for green light IOW, at F/4 diffraction doesn't have a big impact yet, so makes for a convenient number to work with as a limit.



In principle at smaller apertures (larger numbers) diffraction gets exponentially larger, at larger apertures it gets less, but one really only benefits in those cases from specialist lenses, often when used in monochromatic light. General purpose lenses generally are lens limited due to optical aberrations when at larger apertures than about F/4, one of the reasons why curves with high resolution recording media always show less resolution at large apertures than at F/4, F/5.6 or even F/8 for that matter.



3) Next, there is the issue of image size.

If one looks at the same images from differently sized sensors, provided the technology is the same generation, the higher MP sensors at the same size of sensor will provide sharper images, at the same noise level (! - note: at the same image size or magnification). Newer generations will generally show less noise at the same image size (or print size if you like). Larger sensors will also show less noise, because magnification for the same size image will be less.



We have passed amateur level resolutions for colour negatives already a few years ago, at 6MP APS-C sensors (and that is even when printing large enough to make prints as large as we did with 35 mm film negatives). For professional use this was reached at about 12 MP. We are now quickly approaching low iso B&W film for resolution, at excellent quality even when using high isos, while the sensors are stil a far way from reaching the ultimate limits possible. Note that for a 200 lp/mm end result, we need both the lens and sensor being capable of resolving 400 lp/mm, as follows from the formula above. If you work it out, we are talking here about a ~550 MP sensor in that case. Since we are still rather far from such resolutions, I do think we stil have along way ahead before we even get close. I even doubt we will ever get there, because I do not think that well depth will eb enough to get usable DR and noise levels at those MPs. But then , that isn't really necessary. To surpass general purpose B&W film photography, we only need about 42 MP (110 lp/mm in FF format). IOW, we are nearly there. Also note, that I'd expect for those sensors to give excellent performance even at 1600 iso or more, while we are speaking of 25 iso film here for analog.



Finally, at these resolutions it becomes increasingly impossible to get the ultimate sharpness from the negative, whether RAW or film, for the simple reason that the smallest of vibrations or of deviation from perfect focus WILL influence the end result much more than an increase in resolutions, as is already being experienced by medium format shooters at 50 and 60 MP. You'll need heavy tripods, MLU, remote switches, etc. to get the most out of th esytem.



In short, no we are not at the limit yet by a long shot from an optical or sensor resolution POV, but we are getting rather close to the limit of non-static image taking.



The only reason we sometimes think we are limited, is because we pixelpeep. However, if you work it out at human eye resolution, you'll find that that means that we are often looking at ridiculous size "prints", from really close up, and there is no way one can really get a feel for a picture under normal circumstances from these magnifications this close by. Note that picture resolution for the curently used CoCs make for resolutions of about 5-6 lp/mm at a viewing distance of about 25 - 30 cm for a 30 cm (say A4 size) print, which is rendered sharp to the human eye. However, to take in a picture completely, you need to be about twice the diameter away, so it will instantly look a lot sharper than anything closer anyway.



HTH, kind regards, Wim

Guest

VEry good Wim: Allow me to sumemrize you in one sentencefor the 100%-crop-community: Our 100% crops on screen will get less sharper as MP count increases, while our prints will get sharper because an increase in MP equals an increase in print size.

Guest

[quote name='wim' timestamp='1287530403' post='3699']

As Brightcolours and BG_Home mentioned, we're not there yet, not by a long shot.



If I may add something to the discussion, it is really about several things.



1) Lens quality.

The best lenses out there, generally are diffraction limited by about F/4. Think the best Zeiss glass, Canon L primes, the Nikon professional range, the better Sony (Zeiss) glass. This means that these lenses will resolve right to the physical limit possible, which is about 400 lp/mm (line pairs per mm) at F/4, and the criterion being the Rayleigh criterion, which states that this is at 9% contrast, IOW, much less than the 50% MTFs we are looking at from manufacturers, but resulting in higher resolution than on those 50% charts. Do note that "lesser" lenses may not be diffraction limited, but modern prosumer lenses are often close anyway. 250 lp/mm is considered to be excellent.



2) Sensor, or medium (film, sensor, whatever) resolution.

The higher the number of pixels or silver halide crystals or photon counters, or whatever per square unit of area, the higher the resolution. However, there always is a compromise, namely noise levels with sensors, and iso with film. Furthermore, to get the resolution we get with the higher MP sensors these days, we need to shoot low iso B&W film when using analog. As good amateurs we used to get 20 to 40 lp/mm end results in the colour negative, professionals possibly double that, and slightly higher again with B&W, but you need to be prepared to shoot with tripods almost all the time if you really want to realise that sharpness.



Do note that final negative or RAW resolution is inversely proportional to the inverse individual resolutions of sensor and glass, or if said in a different way, 1 / total recorded resolution equals the sum of 1 / lens resolution at a specific f-stop + 1 / maximum recording resolution. You can express this in pixels per image height, as is done on this site, or in lines / mm, or as we used to do in the past, lp/mm, as one needs a minimum of 2 lines in order to depict an edge or line.



Due to the nature of the above formula, it follows that in order to get out the maximum of the one element in that formula, resolution needs to be infinite for the other component. This again is a physical impossibility. The maximum resolution possible is roughly equal to half the wavelength of light, but as we are using the full visible spectrum, this limits us really to the upper end of about 700 to 750 nanometer. Or about 530 nanameter if we use green light, to which the eye is adapted best. 550 nanometer equals about 2 microns in wavelength, which means that maximum resoltion is about 1 micron. As we require 2 of those to see a line, that means that typically maximum resolution is 500 lp/mm just for green light IOW, at F/4 diffraction doesn't have a big impact yet, so makes for a convenient number to work with as a limit.



In principle at smaller apertures (larger numbers) diffraction gets exponentially larger, at larger apertures it gets less, but one really only benefits in those cases from specialist lenses, often when used in monochromatic light. General purpose lenses generally are lens limited due to optical aberrations when at larger apertures than about F/4, one of the reasons why curves with high resolution recording media always show less resolution at large apertures than at F/4, F/5.6 or even F/8 for that matter.



3) Next, there is the issue of image size.

If one looks at the same images from differently sized sensors, provided the technology is the same generation, the higher MP sensors at the same size of sensor will provide sharper images, at the same noise level (! - note: at the same image size or magnification). Newer generations will generally show less noise at the same image size (or print size if you like). Larger sensors will also show less noise, because magnification for the same size image will be less.



We have passed amateur level resolutions for colour negatives already a few years ago, at 6MP APS-C sensors (and that is even when printing large enough to make prints as large as we did with 35 mm film negatives). For professional use this was reached at about 12 MP. We are now quickly approaching low iso B&W film for resolution, at excellent quality even when using high isos, while the sensors are stil a far way from reaching the ultimate limits possible. Note that for a 200 lp/mm end result, we need both the lens and sensor being capable of resolving 400 lp/mm, as follows from the formula above. If you work it out, we are talking here about a ~550 MP sensor in that case. Since we are still rather far from such resolutions, I do think we stil have along way ahead before we even get close. I even doubt we will ever get there, because I do not think that well depth will eb enough to get usable DR and noise levels at those MPs. But then , that isn't really necessary. To surpass general purpose B&W film photography, we only need about 42 MP (110 lp/mm in FF format). IOW, we are nearly there. Also note, that I'd expect for those sensors to give excellent performance even at 1600 iso or more, while we are speaking of 25 iso film here for analog.



Finally, at these resolutions it becomes increasingly impossible to get the ultimate sharpness from the negative, whether RAW or film, for the simple reason that the smallest of vibrations or of deviation from perfect focus WILL influence the end result much more than an increase in resolutions, as is already being experienced by medium format shooters at 50 and 60 MP. You'll need heavy tripods, MLU, remote switches, etc. to get the most out of th esytem.



In short, no we are not at the limit yet by a long shot from an optical or sensor resolution POV, but we are getting rather close to the limit of non-static image taking.



The only reason we sometimes think we are limited, is because we pixelpeep. However, if you work it out at human eye resolution, you'll find that that means that we are often looking at ridiculous size "prints", from really close up, and there is no way one can really get a feel for a picture under normal circumstances from these magnifications this close by. Note that picture resolution for the curently used CoCs make for resolutions of about 5-6 lp/mm at a viewing distance of about 25 - 30 cm for a 30 cm (say A4 size) print, which is rendered sharp to the human eye. However, to take in a picture completely, you need to be about twice the diameter away, so it will instantly look a lot sharper than anything closer anyway.



HTH, kind regards, Wim

[/quote]





Wim........ That's one of your better efforts. This Forum would be the poorer without you





Jim

PuxaVida

Some months before, the same issue was discussed and as I already mentioned there also, I personally believe that this post must be saved somewhere in PZ so that people can make use of it. It's a brilliant post Wim...



Regards,



Serkan
[quote name='jenbenn' timestamp='1287554914' post='3702']

VEry good Wim: Allow me to sumemrize you in one sentencefor the 100%-crop-community: Our 100% crops on screen will get less sharper as MP count increases, while our prints will get sharper because an increase in MP equals an increase in print size.

[/quote]

Thank you, jenbenn,.

And almost right... <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/biggrin.gif' class='bbc_emoticon' alt='B)' />. They will get noisier mostly at 100%, and sharper at equal print sizes, and with more beautiful rendering (more pixels for the same detail).



Kindest regards, Wim
[quote name='AAC7man' timestamp='1287572155' post='3705']

Wim........ That's one of your better efforts. This Forum would be the poorer without you





Jim

[/quote]



[quote name='PuxaVida' timestamp='1287572418' post='3706']

Some months before, the same issue was discussed and as I already mentioned there also, I personally believe that this post must be saved somewhere in PZ so that people can make use of it. It's a brilliant post Wim...



Regards,



Serkan

[/quote]

Jim, Serkan,



Thank you very much for your kind words. Much appreciated!



Kindest regards, Wim

Guest

Very informative post, thanks, Wim!



All discussions of resolution aside, I feel that with a Canon 5D MK II I am at a level that is perfectly adequate (prints above A2 are very rarely called for) - but what I really want is more dynamic range. Maybe I'm wrong here, but I feel that the industry is selling their new technology on the strength of megapixels because it's easier than improving dynamic range.
Pages: 1 2 3 4