theoretical question

hi all,

i was wondering
**IF** indeed Canon will release a MP "monster" (for the sake of this question, let's say 30-50MP), how many lenses out today could match the resolution of the sensor?

I mean, for the 22MP 5dIII with the highly touted 70-200 II, would it be as "great" if the sensor was of a 2-2.5X resolution? Same thing could be asked from a lot of the top lenses: 20-700 f4 is, 135 f2, 85 II, 100 L IS, sigma 50 art/Otus 55, 17 tilt, etc....

How is this match even calculated?

thanks for any ideas on this thought experiment

Well the current crop of lenses it would seem are perfectly capable of out resolving the current Canon crop sensors. If they wern't Canon would not be fitting those sensors with analogue AA filters. Your 50 Mpix sensor if FF would have a resolution about the same as the current crop sensors. 50-60 Mpix on a Canon crop sensor might just get to the point where you could really get away with removing the AA filter. That is for resolutions capable by current lenses, who knows what new lens technology might come along that could improve the resolution presented to the sensor.

Digital sampling theory (it's more of a law really) says that you have to have two samples per period or you will suffer from aliasing effects, which cannot be distinguished from real signals so cannot be removed computationally. What this means for the photographer is that if for example you have a lens that can resolve one line pair per mm (one black line next to one white line) then you need to have at least 2 sensels per mm to correctly record them. For our colour images produced by a Bayer Colour Filter Array, which groups sensels into squares of 4 sensels with one red, one blue and two green filters you would idealy want to count each group as a single data point when calculating the optimum sensor resolution. Over sampling is never a problem for the engineers.

The issue with over sampling will come from the photographers who just have to pixel peep. Although we feel that a subject with a supposedly sharp edge should go from one tone to the other in a single pixel this is really impossible in the real world. In the real world edges will always show as less than perfect. Even if you had a perfect lens (which you could never sample digitally), a sharp edge between two areas of different tone will also introduce diffraction, just as the aperture blades do.

Personally I think that given no major breakthrough in lens resolving power then we still have a long way to go in sensor resolution, but that while we are using any sort of digital system and people willing to use 100% pixel views on output devices with resolutions of around 100 PPI we will still get moaning and groaning about soft images. Now introduce large sized displays (30"+) with 300+ PPI resolutions (at least then it will still be displayed at 100%) we might be happy with our 50-60 Mpix crop sensors.

I know that I love the way images look on my new Galaxy Note 3 at around 380 PPI.

Alan

Big Al,

THAT was a pretty amazing "real" answer to the OP's "theoretical" question.

Some of it was lost on me, but if I understood correctly, today's top lenses are still better than today's top processors? Or did I get it backwards?

BigAl007 wrote in post #16882908
Well the current crop of lenses it would seem are perfectly capable of out resolving the current Canon crop sensors. If they wern't Canon would not be fitting those sensors with analogue AA filters. Your 50 Mpix sensor if FF would have a resolution about the same as the current crop sensors. 50-60 Mpix on a Canon crop sensor might just get to the point where you could really get away with removing the AA filter. That is for resolutions capable by current lenses, who knows what new lens technology might come along that could improve the resolution presented to the sensor.

Digital sampling theory (it's more of a law really) says that you have to have two samples per period or you will suffer from aliasing effects, which cannot be distinguished from real signals so cannot be removed computationally. What this means for the photographer is that if for example you have a lens that can resolve one line pair per mm (one black line next to one white line) then you need to have at least 2 sensels per mm to correctly record them. For our colour images produced by a Bayer Colour Filter Array, which groups sensels into squares of 4 sensels with one red, one blue and two green filters you would idealy want to count each group as a single data point when calculating the optimum sensor resolution. Over sampling is never a problem for the engineers.

The issue with over sampling will come from the photographers who just have to pixel peep. Although we feel that a subject with a supposedly sharp edge should go from one tone to the other in a single pixel this is really impossible in the real world. In the real world edges will always show as less than perfect. Even if you had a perfect lens (which you could never sample digitally), a sharp edge between two areas of different tone will also introduce diffraction, just as the aperture blades do.

Personally I think that given no major breakthrough in lens resolving power then we still have a long way to go in sensor resolution, but that while we are using any sort of digital system and people willing to use 100% pixel views on output devices with resolutions of around 100 PPI we will still get moaning and groaning about soft images. Now introduce large sized displays (30"+) with 300+ PPI resolutions (at least then it will still be displayed at 100%) we might be happy with our 50-60 Mpix crop sensors.

I know that I love the way images look on my new Galaxy Note 3 at around 380 PPI.

Alan

That pretty much covers it. Just wanted to point out that "theory" in science is a "body of facts" and not as it is more casually used when people really mean "hypothesis". Kind of a pet peeve of mine. This misuse of the word is why BigAl felt the need to say "(it's more of a law really)"

bigal - a really useful explanation. Thank you.
I am not sure that I understand everything you wrote (at least not the depth of it), but I appreciate the detail and the seriousness of the answer!

Many thanks

Maybe off topic but I think for most photographers the limiting factor is not sensors nor lenses but perhaps technique. Ability to get that shot that pushes everything to the limits. Gene

LOL..... that is definitely **MY** biggest liability for sure!

gasrocks wrote in post #16883452
Maybe off topic but I think for most photographers the limiting factor is not sensors nor lenses but perhaps technique. Ability to get that shot that pushes everything to the limits. Gene

Yep

Theres a bunch of people out there shooting with ancient gear and junk lenses, and producing incredible work
But many more people with all the gear, and no idea - struggling to produce good images, and bouncing from new lens to new body to new lens again- in the hopes that purchasing this new gizmo will fix all their problems.

Equipment is not the answer. Gene

Echo63 wrote in post #16883486
Yep

Theres a bunch of people out there shooting with ancient gear and junk lenses, and producing incredible work
But many more people with all the gear, and no idea - struggling to produce good images, and bouncing from new lens to new body to new lens again- in the hopes that purchasing this new gizmo will fix all their problems.

Have you been spying on me?

I would look at what lenses(EF) work well on the new crop cameras.

Echo63 wrote in post #16883486
Yep

Theres a bunch of people out there shooting with ancient gear and junk lenses, and producing incredible work
But many more people with all the gear, and no idea - struggling to produce good images, and bouncing from new lens to new body to new lens again- in the hopes that purchasing this new gizmo will fix all their problems.

Yeah, and they make for excellent sources for used equipment.

gasrocks wrote in post #16883452
Maybe off topic but I think for most photographers the limiting factor is not sensors nor lenses but perhaps technique. Ability to get that shot that pushes everything to the limits. Gene

I would agree with that Gene, I know I'm in the Nikon camp but I shoot with a D800, I do use very good glass that resolves excellently but you do have to give more attention to technique, that big sensor and good glass does pick up flaws in your technique.

First thing that popped into my mind -- these lenses are not designed for DSLR cameras only, but mainly for SLR cameras, film and slide until recently were used exclusively with film and slides. So, basically all or at least most are already capable of much more than our digital sensors can 'sense' ...

DSLR camera has not really come of age until ten years or so ago -- lenses have been around A LONG time. It's the senors that are behind, IMO ...

monkey44 wrote in post #16884703
First thing that popped into my mind -- these lenses are not designed for DSLR cameras only, but mainly for SLR cameras, film and slide until recently were used exclusively with film and slides. So, basically all or at least most are already capable of much more than our digital sensors can 'sense' ...

DSLR camera has not really come of age until ten years or so ago -- lenses have been around A LONG time. It's the senors that are behind, IMO ...

I mostly now print 16×12 these days using a lab that is printing on Fuji Crystal Archive paper. Even my 300D at 6 Mpix is producing results that are at least as good I could get from Kodachrome 25 and printing at 10×8 on Ciba-Chrome paper. It's not that the sensors are bad, they've been better than even slide film for a long time now, it's that top end lens resolutions are just very good.

As Canon do not actually publish maximum resolution information on their lenses it is difficult to know just how much resolution they can support. The published MTF charts are based on 30lp/mm which could be resolved at the Nyquist limit (where the sampling rate is 2× maximum recorded frequency) by a 3.1 Mpix full frame sensor, although I would prefer to double the linear resolution to get two full sets of the Bayer FCA in each line pair giving a FF resolution of 12.4 Mpix. The 70D has 243.2 sensels/mm so can record a maximum resolution of 121.6 lp/mm at the Nyquist limit. Although I think 60 is a better practical limit. This is Canon's highest resolution DSLR sensor on sale, and still needs an AA filter, so the lenses must be doing better than this in terms of resolution.

One note I would not rely on DXOMark for lens resolution data. Although it has been possible to record lens resolution limits using analogue means in the lab for many years they actually use a digital recording method (actual cameras) to determine lens resolution. The Nyquist limit for each camera sensor limits the maximum possible value for lens resolution. So on a 70D for example no lens could achieve more than 121.6 lp/mm, and the AA filter is likely to limit that to around 60 lp/mm.

Sorry for all the technical stuff, it is not really possible to discuss this subject without it. I have tried to keep it as simple as possible, and have really limited the math as much as possible.

Alan