[mithrandir]

Theoretically, using a full frame lens (35mm) on a cropped sensor should actually allow for a better image since you will be using more the center (sweet spot) of the lens and not as much of the outer edges where light falls off and image resolution goes soft.

[Marc Sabatella]

No this is simply untrue. Its not a "smaller image". Its a 12MP image.

I'm not sure what you're talking about. I'm comparing the size of an APS-C image of unspecified sensor resolution (although I had mentioned 10MP elsewhere) with a 35mm film image. Where did 12MP come from? And regardless of how many pixels are there, how is the image on an APS-C not smaller than on 35mm film? You seem to be misinterpreting something about what I am trying to say - understandable, since there is so much confusion floating around. So let me try again.

Let me put it this way: say we had a 10MP APS-C sensor and a FF sensor that had enough pixels such that the APS-C-sized central portion was also 10MP. That is, the pixel density is the same. Put the same lens on each and take a picture of an object from a given distance. Now make prints at 4x6". The object appears bigger on the APS-C print - the "crop factor" at work. But clearly, the amount of detail will be the same - it's just bigger on the APS-C print.

On the other hand, if you then put a longer focal length lens on the FF sensor such that the FOV was the same as the original lens was on APS-C, and if that lens had the same resolution as the original, you *would* get more resolution in your final print. That is to say, a 100mm lens on APS-C has the field of view of a 150mm lens on FF, but if the lenses both have the same actual resolution, your print from the latter will shows more detail given sensors of the same pixel density.

But of course, who knows if you *could* find a longer focal length lens with the same resolution, so you may might not actually see this effect in practice.

only if the image resolution is limited entirely by the lens will the overall resolution be related directly to crop factor.
...
Also lens resoltion is not absolute, but simply diminishes in contrast the smaller the detail becomes.

That's a very good point. So it's more complicated than simple formulas would suggest. But there's got to be something to it - isn't this basically the same reason people use medium or large format film cameras?

[Class A]

FunkyMonk, I couldn't find anything wrong with what you wrote (although I haven't checked every calculation). I didn't see anything you don't understand correctly about crop factors.

*isteve, Marc assumes that the lens can limit IQ, which it can in some cases (e.g., sub-par zoom lens, high f-ratios). Certainly, a full frame lens loses resolution when used on an APS-C sensor (for an equivalent image). The question is whether this loss can be resolved by the sensor.

Nikon D2X (12 MP) APSC format with 50mm prime
Horizontal resolution 2400/2800 (absolute and extinction)
Vertical resolution 2000/2500

Canon 5D (12 MP) FF format with 50mm prime
Horizontal resolution 2300/2500
Vertical resolution 2000/2500

Interesting. What unit are these number in? LW/PH?
Where have you got them from?

50mm primes are among the best candidates to outresolve sensors. Other lenses may not do as well and then the sensor size may matter. But in principle I fully agree with you.

[*isteve]

Interesting. What unit are these number in? LW/PH?
Where have you got them from?

50mm primes are among the best candidates to outresolve sensors. Other lenses may not do as well and then the sensor size may matter. But in principle I fully agree with you.

The numbers are from DPreview and are in lw/ph lw/pw depending on dimension.

Actually a 50mm prime designed for FF will give the advantage to the FF sensor because it is designed to cover a larger area. If you used a 35mm APSC lens on the APSC camera, it should show a slightly higher resolution with the same FOV.

---

Last edited by *isteve; 01-21-2009 at 09:07 AM.

[*isteve]

I'm not sure what you're talking about. I'm comparing the size of an APS-C image of unspecified sensor resolution (although I had mentioned 10MP elsewhere) with a 35mm film image. Where did 12MP come from? And regardless of how many pixels are there, how is the image on an APS-C not smaller than on 35mm film?

We are talking about a digital image, the dimensions of the sensor are not relevant. It matters on film only because the grain is always the same size for a given emulsion, so to get more information you need more grain so you need a bigger neg.

With digital, the only thing that matters is how many photosites you use to capture the image. You can use a small sensor with 15M very small photosites, or a larger sensor with 15M very large photosites, but once they are stored on your computer's hard drive, it does not matter what size sensor they came from, you have 15M dots. This is why the Canon G10 (15mp) has the same resolution more or less as a Canon 50d (15mp)!!! Actually slightly better if anything.

In this case of course, the lens resolution is a factor but in this case its the APS camera thats limited vs the compact. Compact lenses easily outresolve SLR lenses because they have a much smaller registration distance and a much smaller image circle. Some compact zooms manage well over 250 lppm at MTF50 - four or five times better than any SLR lens.

You seem to be misinterpreting something about what I am trying to say - understandable, since there is so much confusion floating around. So let me try again.

Let me put it this way: say we had a 10MP APS-C sensor and a FF sensor that had enough pixels such that the APS-C-sized central portion was also 10MP. That is, the pixel density is the same. Put the same lens on each and take a picture of an object from a given distance. Now make prints at 4x6". The object appears bigger on the APS-C print - the "crop factor" at work. But clearly, the amount of detail will be the same - it's just bigger on the APS-C print.

So if you crop the FF image to 10MP you will get EXACTLY THE SAME PICTURE, which is a 10MP image. Assuming a good lens, both 10MP images (the APSC image and the FF crop) will have exactly the same resolution because its the number of dots that matters, not the size of the sensor or the lens.

On the other hand, if you then put a longer focal length lens on the FF sensor such that the FOV was the same as the original lens was on APS-C, and if that lens had the same resolution as the original, you *would* get more resolution in your final print. That is to say, a 100mm lens on APS-C has the field of view of a 150mm lens on FF, but if the lenses both have the same actual resolution, your print from the latter will shows more detail given sensors of the same pixel density.

You only get more resolution because if the the FF camera can be cropped to 10MP, its actually a 20MP sensor. You have 20M dots compared to 10M dots in your image so of COURSE it has more resolution!!!

But of course, who knows if you *could* find a longer focal length lens with the same resolution, so you may might not actually see this effect in practice.

You will see it of course, you are using a higher resolution sensor!!!

But what if you take a 12MP FF camera with a 75mm lens and a 12MP APS camera with a 50mm lens. The image in both cameras has the same field of view. Now unless the lenses are limiting resolution (which is rare at this size) then both images will have THE SAME RESOLUTION. Both contain 12M dots and can print on A3 paper at the same DPI.

That's a very good point. So it's more complicated than simple formulas would suggest.

No its exactly as simple as the formula suggests.

1/r (system) = 1/r(sensor) + 1/r(lens).

If r(lens) is high compared to r(sensor) then system resolution relies more on the sensor than the lens. Even if they are close, then the lens only accounts for some of the difference.

But there's got to be something to it - isn't this basically the same reason people use medium or large format film cameras?

Yes, but you are comparing the wrong thing. On film, the grain has a fixed size for a given ISO and emulsion. So a 6X4.5 negative will have more than 3X the number of grains as a 35mm negative for the same ISO. So, effectively it has more "dots" hence I can blow it up larger.

This is your analogy with a 10MP APS camera with a 20MP FF camera. The latter has 20m dots so has twice the resolution. With the same pixel density a Leaf back would have 56 MP. What matters is just the number of dots.

But with digital we dont compare ISO we commonly compare MP. In this case ANY 12MP sensor will have roughly the same resolution and enlargement potential with a good lens. The difference is noise and DR with larger pixels. I can push the ISO higher with larger pixels, just like high ISO film has much larger grain.

[falconeye]

This topic is scattered with traps.

For a second, let's assume we are beyond all the standard traps now.

In some other thread, I have derived a universal fact valid in the following situation:

(A) We compare two cameras, with same pixel count but different sensor size, and the camera with the smaller sensor has a focal length smaller by the same factor, too. However, both cameras use a lens of exactly the same physical diameter in mm.

Then, both cameras, have:
- Identical field of view (FoV)
- Identical depth of field (DoF).
- Identical level of noise (the camera with smaller sensor would use a lower ISO figure).
- Identical resolution limit for diffraction, when measured in LW/PH.
- Different resolution limit for ray-optics, when measured in LW/PH (*).
(*): with an advantage for the larger sensor.

BTW, this four property equivalence will go beyond everything you are able to find elsewhere in the web.

One way to see the fifth point would be by comparing two lenses with identical construction, only that the one with shorter focal length would need thicker glass for stronger refraction, in general leading to larger optical defects. On the other hand, the image circle is smaller but the required deviation size is smaller, too. Some deviations are independent from the distance from optical axis (advantage large sensor), some linear, some increase with their square (advantage small sensor). Overall, this is a non-trivial issue. Still, with an advantage for the larger sensor. Which is more of a problem at the wide angle end.

The problem with this argument is that the resolution limit for ray-optics is zero w/o economical constraints. So, what the fifth point really means is this: A large sensor camera with a given, high enough resolution can be cheaper than an equivalent small sensor camera.

Dedicated APS lenses usually have higher resolution because they have a smaller image circle to cover.

Here comes the caveat.
In the above equivalence, the lens for the larger sensor would have a higher f-stop number. Which is ok.

But another way to compare would be an exact scaling miniaturisation of every single aspect of the camera. Smaller image circle. Same f-stop number. Same resolution. Independent of the crop factor. What *isteve says.

So, let me rewrite my comparison table for this situation:

(B) We compare two cameras, with same pixel count but different sensor size, and the camera with the smaller sensor has a focal length smaller by the same factor, too. Both cameras use a lens of exactly the same f-stop number.

Then, both cameras, have:
- Identical field of view (FoV)
- Different depth of field (DoF) (*).
- Different level of noise (both cameras would use same ISO figure) (*).
- Different resolution limit for diffraction, when measured in LW/PH (*).
- Identical resolution limit for ray-optics, when measured in LW/PH.
(*): with an advantage for the larger sensor.

So, we get a two property equivalence, only.

Therefore, same ray-optical limit, correct, but different diffraction limit. Which means that the lens for the smaller sensor would have to become diffraction-limited at a wider aperture which may be impossible or very expensive to achieve in practical terms (more lens elements and of a more complex shape) (cf. above).

A note on this: Diffraction-limited f/0.8 lenses exist, with a focal length of ~1mm, in microscopy. The image circle may be about 1mm, too. On the other hand, 35mm lenses which are diffraction-limited at f/2.8 exist (Zeiss). The image circle is about 43mm. Taking the larger diffraction into account, this would still be more than 100x the resolution of a microscope, in number of pixels.


So, in summary:

The possible resolution is independent of sensor size, as long as we aren't approaching the limits of diffraction. This has already happened for P&S. Not yet for DSLR (requiring 5µm only where best prime lenses deliver 1.5µm).

However, and this is my personal opinion, a forthcoming 100 MPixel resolution will be possible with FF sensors (or larger), not with APS-C.


So, an even shorter summary:

The possible DSLR resolution is independent of sensor size today, but will become dependent at some moment in future.

[*isteve]

This topic is scattered with traps.

For a second, let's assume we are beyond all the standard traps now.

In some other thread, I have derived a universal fact valid in the following situation:
We compare two cameras, with same pixel count but different sensor size, and the camera with the smaller sensor has a focal length smaller by the same factor, too. However, both cameras use a lens of exactly the same physical diameter in mm.

Then, both cameras, have:
- Identical field of view (FoV)
- Identical depth of field (DoF).
- Identical level of noise (the camera with smaller sensor would use a lower ISO figure).
- Identical resolution limit for diffraction, when measured in LW/PH.

BTW, this four property equivalence will go beyond everything you are able to find elsewhere in the web.



Here comes the caveat.
In the above equivalence, the lens for the larger sensor would have a higher f-stop number. Which is ok.

But another way to compare would be an exact scaling miniaturisation of every single aspect of the camera. Smaller image circle. Same f-stop number. Same resolution. Independent of the crop factor. What *isteve says.

But, and this is a big but, this line of argument ignores the equivalence above. DoF, noise, and most importantly, the diffraction limit, would be worse for the smaller sensor.

Which is what I said....
Actually the diffraction issue is not a problem if you stick to your "same diameter" principle. Most lenses are sharpest about 2 stops below max aperture, and the FF lens needs to be stopped down more for equivalent DOF so for max resolution and the same DOF neither FF nor APS will approach the diffraction limit.

Same ray-optical limit, correct, but different diffraction limit. Which means that the lens for the smaller sensor would have to become diffraction-limited at a wider aperture which may be impossible or very expensive to achieve in practical terms (more lens elements and of a more complex shape).

Only for wide angle lenses and only because the registration distance is the same between APSC and FF for engineering reasons. If you scaled the cameras exactly, there would not be any problem.

Or, in terms of my four property equivalence, I could add this fifth point:
- Different resolution limit for ray-optics, when measured in LW/PH, with an advantage for the larger sensor.
One way to see this would be by comparing two lenses with identical construction, only that the one with shorter focal length would need thicker glass for stronger refraction, in general leading to larger optical defects. Which is more of a problem at the wide angle end.

The problem with this argument is that the resolution limit for ray-optics is zero w/o economical constraints. So, what the fifth point really means is this: A large sensor camera with a given, high enough resolution can be cheaper than an equivalent small sensor camera.

Whoa there. This is totally ignoring sensor fab costs and everything you just said.

Whatever the resolution, FF sensors and APSC sensors will maintain the same price difference of about 10X. FF cameras will always cost more.

Now if you insist on equivalence in all things, then you will need F2.8 lenses on APS and F4 lenses of 1.5X the focal length on FF. Since a 200 mm f2.8 lens is about the same price, size and weight as a 300mm F4 lens, then I am still left with the higher cost of the FF camera.

If you want wider angle, then as long as you use APS dedicated lenses, you are still not getting much more value out of FF. Compare the list of Nikon DX lenses - there is a nice selection around the 12-24 and 17-50 mark. Same for Canon, if not moreso.

So, in summary:

The possible resolution is independent of sensor size, as long as we aren't approaching the limits of diffraction. This has already happened for P&S.

Not yet for DSLR (requiring 5µm only where best prime lenses deliver 1.5µm).

However, and this is my personal opinion, a forthcoming 100 MPixel resolution will be possible with FF sensors (or larger), not with APS-C.

So, an even shorter summary:

The possible DSLR resolution is independent of sensor size today, but will become dependent at some moment in future.

True, although in fact we are getting to the point where lens resolution is starting to limit APSC and FF. A useful limit in terms of DR/noise/resolution tradeoff is about 20MP on APSC and 35MP on FF.

However, I struggle very hard to understand why most photographers think they need that much image information when they generally shoot hand-held shots and never print larger than A4.

I frequently print A3 sized and 19" X13" prints from a 6MP *istD and they look great from a normal viewing distance. An A2 print from my K20D looks equally good. In exhibits, I have seen A1 prints from even lower res cameras (Nikon D2X) which look amazing if you dont have your nose against the poster.

I have 2X 1TB RAID enclosures just to backup my image files. Soon I will need another one. This is a lot for an amateur photographer to cope with.

Now if I was a pro landscape photog or I shot advertising posters for a living then fair enough, but 90% of the folks I know seldom do more than post web images.

[Marc Sabatella]

With digital, the only thing that matters is how many photosites you use to capture the image. You can use a small sensor with 15M very small photosites, or a larger sensor with 15M very large photosites, but once they are stored on your computer's hard drive, it does not matter what size sensor they came from, you have 15M dots.

True, if all else was equal with regard to the field of view and the resolution of the respective lenses, sure, the images would be indistinguishable at any print size. But things are *not* generally equal between field of view and lens resolution when dealing with differing sensor sizes, and that's my whole point. Assuming you want to keep the same field of view, the smaller sensor needs correspondingly higher resolution from its lens in order to take full advantage of those 15M pixels. You seem to be thinking about a world where lens resolution is always greater than the sensor can record, but that's simply not the world most of us live in today.

Compact lenses easily outresolve SLR lenses because they have a much smaller registration distance and a much smaller image circle. Some compact zooms manage well over 250 lppm at MTF50 - four or five times better than any SLR lens.

OK, but we're not comparing DSLR's to compacts with entirely different mounts - we're comparing APS-C to FF on the *same* mount. At least, that's what *I'm* talking about. And I wouldn't assume that a given 70mm lens on APS-C is necessarily giving resolution comparable to that of a given 105mm lens on FF (for example). It might - it's going to depend on the specific lenses. But in order for that 70mm lens on APS-C to provide the same usable resolution as the 105mm lens on FF, it's going to need to resolve *more* lines per inch, because it's going to have to pack the same number of lines of resolution into the smaller size of the APS-C sensor.

Now unless the lenses are limiting resolution (which is rare at this size)

This is specifically what I am challenging. I think lenses *are* limiting resolution in the vast majority of cases. If they weren't, why on earth would people be paying the big bucks for the betters lenses? And why would there be noticeable improvement in sharpness as you stop down down even the best lenses?

It seems *obvious* to me that the glass is limiting resolution with "most" lenses on most DSLR's, but it's just as obvious that you are pretty knowledgeable on the subject - so could you explain what makes you say lenses are not limiting resolution - and how you claim is not at odds with the phenomena I just describe (eg, the fact that we can see a difference in resolution between different lenses and between different apertures on any given lens)?

[falconeye]

Actually the diffraction issue is not a problem if you stick to your "same diameter" principle. Most lenses are sharpest about 2 stops below max aperture
[...]
This is totally ignoring sensor fab costs and everything you just said.
[...]However, I struggle very hard to understand why most photographers think they need that much image information

*isteve, I'm with you in most points. Let's clarify a few, though...

First, the diffraction issue is an issue. To clarify, just assume that your sensor is 1µm x 1µm only (Gedankenexperiment...). Diffraction-limited to one pixel only. Whatever else you do. This trivial fact MUST show up somewhere in a sound argument. A lens for such a system would be sharpest at max aperture -- because it would be diffraction-limited at any aperture. Your rule of thumb "best 2 stops below max aperture" is not general enough and not exact enough to be useful in our argument. Besides this, most zooms are sharpest at f/8, good zoms at f/5.6, good primes at f/4, regardless of max. aperture.

The more we head for max. resolution, the more this kind of consideration becomes relevant. Not today (*). But tommorrow.

(*) Today: The Zeiss 35mm f/2 which is sharpest at f/2.8!! The only lens I know of which is sharper at f/2.8 than at f/4!

And I am with you on the question of "who needs this"? Nevertheless, I foresee a future application: video walls capable of displaying 100 MPixel+. Wouldn't you like to be able to display your photos in such a forthcoming standard, when video walls act like additional windows?

[*isteve]

*isteve, I'm with you in most points. Let's clarify a few, though...

First, the diffraction issue is an issue. To clarify, just assume that your sensor is 1µm x 1µm only (Gedankenexperiment...). Diffraction-limited to one pixel only. Whatever else you do. This trivial fact MUST show up somewhere in a sound argument. A lens for such a system would be sharpest at max aperture -- because it would be diffraction-limited at any aperture. Your rule of thumb "best 2 stops below max aperture" is not general enough and not exact enough to be useful in our argument. Besides this, most zooms are sharpest at f/8, good zoms at f/5.6, good primes at f/4, regardless of max. aperture.

The more we head for max. resolution, the more this kind of consideration becomes relevant. Not today (*). But tommorrow.

I agree it becomes a serious issue if the pixel pitch drops too much further, or at least it becomes pointless adding more pixels.....I just meant we are not quite there....yet. I think we agree on that one.

(*) Today: The Zeiss 35mm f/2 which is sharpest at f/2.8!! The only lens I know of which is sharper at f/2.8 than at f/4!

And I am with you on the question of "who needs this"? Nevertheless, I foresee a future application: video walls capable of displaying 100 MPixel+. Wouldn't you like to be able to display your photos in such a forthcoming standard, when video walls act like additional windows?

How close do you want to stand to it?

Maybe one day when I have a laptop with a 200 TB solid state drive but I feel all such an image will do is show the nasty inherent issues with 35mm lenses.

Also, unless you use a very heavy tripod, the slightest movement will cause visible blur.

More likely I will buy a window +100 nice movie loops (with sound) to make it look more realistic

[falconeye]

How close do you want to stand to it?

I forsee my breakfast coffee table besides it. So, let's say 50 cm / 2 feet away? And by wall, I mean wall -- at least 4m wide...

I agree, an ambitious vision. But what else makes the world go round?

BTW, the blur thing can be resolved in software. I already do it now for my tripod-free night vision shots. 100 MPixel is just 2.6x more blur than today

[*isteve]

It seems *obvious* to me that the glass is limiting resolution with "most" lenses on most DSLR's, but it's just as obvious that you are pretty knowledgeable on the subject - so could you explain what makes you say lenses are not limiting resolution - and how you claim is not at odds with the phenomena I just describe (eg, the fact that we can see a difference in resolution between different lenses and between different apertures on any given lens)?

I never said that lenses had no effect, I said that unless its a cheapy lens, its not the primary contribution to resolution - at least not yet - and that the sensor dimensions dont tell you anything about enlargement size because sensors do not all have the same size pixels.

And yes you can see the resolution differences between different lenses on the same sensor, but the question is what happens when I put the SAME lens on a higher resolution sensor? The answer is.....I get more resolution!!! I may not get ALL the extra resolution I expect, but for most current lenses and sensors I will get a fair proportion of it, especially as I can pull a few tricks with a higher resolution sensor like reducing the AA filter strength.

As things stand, you are not going to see much difference in resolution between a 12MP APS and 12MP FF camera, and this is entirely borne out in lab tests including the results above for the Canon 5D and Nikon D2X.

Closer to home, how about the Nikon D300 and D700 on DPReview? Note that the lens was used at F8 which is more likely to blur the D300 shot, and yet....

Nikon D700 Review: 31. Compared to (Resolution): Digital Photography Review

Almost no difference.

Note, the A900 is compared as well, which gives you some idea of how important the number of megapixels is compared to sensor dimensions (A900 vs D700 and Canon 5D) using the same lens.

Are we reaching the point of diminishing returns on APSC? If you build a range of dedicated APS lenses, probably not. However the use of old FF lenses will start to show up resolution limitations if you were to build a 24MP APS camera.

However the main issue would be noise and DR and diffraction and thats why I dont think APSC will grow much denser in the short term.

[Marc Sabatella]

I never said that lenses had no effect, I said that unless its a cheapy lens, its not the primary contribution to resolution

But remember, we aren't comparing randomly different sensor sizes and resolutions - we're comparing APS-C to FF with a given line of DSLR, and yes, of course we're assuming similar sensor resolution. You seem to be wanting to let all variables vary and then asking would would be most likely to happen; I'm trying to keep everything as similar as possible and looking at the effects of sensor size *only*.

If it helps, my argument is most clear if we assume sensors of essentially infinite resolution. Then the effect of the lens resolution is most clear. And my assumption is that any time the sensor is outresolving the lens, the lens will still have an effect. And while sensor resolution at 10-15MP may still be the *biggest* factor, it still seems clear lens resolution has *an* effect.

And yes you can see the resolution differences between different lenses on the same sensor, but the question is what happens when I put the SAME lens on a higher resolution sensor? The answer is.....I get more resolution!!! I may not get ALL the extra resolution I expect, but for most current lenses and sensors I will get a fair proportion of it

Right - and it's the part you *don't* get that I am talking about. This is the part that is dependent on sensor size.

As things stand, you are not going to see much difference in resolution between a 12MP APS and 12MP FF camera, and this is entirely borne out in lab tests including the results above for the Canon 5D and Nikon D2X.

OK, but presumably such tests were not able to account for lens differences.

In any case, I'm not saying it's going to be a *big* difference, and it's not one I'd ordinarily care about - I'm the one usually pooh-poohing the need for FF and vowing to stick with APS-C as long as they'll let me. But the question was asked whether the smaller size of the sensor could have any effect on resolution, and it still seems the answer is yes, if all else is equal. Maybe not a big difference in practice, but I still cannot see how could could fail to me *some* difference in the cases where the lens is not completely outresolving either sensor.

[*isteve]

But remember, we aren't comparing randomly different sensor sizes and resolutions - we're comparing APS-C to FF with a given line of DSLR, and yes, of course we're assuming similar sensor resolution. You seem to be wanting to let all variables vary and then asking would would be most likely to happen; I'm trying to keep everything as similar as possible and looking at the effects of sensor size *only*.

.... and it's the part you *don't* get that I am talking about. This is the part that is dependent on sensor size.

I can do no more than point you at real data which I have done twice. If you can see a real difference between a D300 and D700 in resolution then fine, but neither I nor Phil Askey can, so what are you arguing about? Your theory is not backed up by the evidence. How do YOU explain the numbers?

[newarts]

I think the following is true:

A perfect lens' resolution is fixed by diffraction at something like its Rayleigh limit of
r=1.22(f-stop)(wavelength). For greenish light this is about (2/3 micrometer)(f-stop).
So for a practical f-stop like 6, the resolution limit for a perfect lens is fixed by diffraction at:
r=4 micrometer.

There must be at least detector two pixels at this spacing, so pixel spacing for this resolution should be 2 micrometers.

Pixel Pitch - Camera
5.0um - K20D
6.1um - K200D
7.8um - K100D

In theory, the K20D does the best job of matching the resolution of a perfect lens at F/6.

Dave