Are there 39mp & 50mp+ Test Bodies in the Wild? [CR1]

[jrista]

Jrista... I usually gloss over your techno rant...  But the last post (with the DLA MTF table) was well constructed and easy to follow for peeps like me.  Thanks!

Glad to be of service.

[Lee Jay]

You are only thinking pixel size, which I guess is one way to look at it.

Yes, that is the subject we are discussing - is there a point to smaller pixels given the lenses we have available, to which the answer is obviously "yes", as can be shown by simple math and by examples from TCs which show you what the center of the image would look like with smaller pixels.

No, the topic we were discussing is whether adding a TC is exactly the same and just as good as using a sensor with a higher pixel density by the same factor as the TC added.

Yes, that's what I said.

If you are referring to optical system resolution, rather than spatial resolution, then I agree. However you keep applying the units "lp/mm" to system resolution, which feels like a major conflation to me. Assuming the optical spatial resolution of the entire lens setup (original lens + TC) remains the same (which is generally impossible when adding a TC, as it reduces your REALTIVE aperture, which implicitly means your optical spatial resolution of THE ENTIRE LENS SETUP is reduced), the final system spatial resolution will be lower than that of the lens or the sensor, as it is the root mean square of the blur each individual component.

I can't even believe you just said that.  So, adding a TC reduces optical spacial resolution?  Better throw them all out then.

A TC doesn't change optical resolution.

YES, adding a TC reduces optical spatial resolution, because it reduces the RELATIVE APERTURE. Diffraction is dependent on aperture.

Wrong, and right - diffraction is based on aperture, not relative aperture.

http://en.wikipedia.org/wiki/Rayleigh_criterion#Explanation

sin theta = 1.22*lambda/aperture diameter

You're conflating resolution at the sensor in lp/mm (which is largely irrelevant) with optical angular resolution (which is what we care about - what details in the scene can be resolved).

Look, the facts are simple.  If you want to see how a particular lens would perform on a camera with four times as many pixels as your current camera, simply add a good quality 2x TC and see for yourself how the center of that hypothetical sensor would look with the bare lens.  Period.

Sure...but you've just invoked cropping. Who buys a 500mp sensor to crop out the middle 18mp?

I do, as do many others.  If you have a better way to simulate the performance of our current lenses on a hypothetical and not-yet-built higher-pixel-count sensor than using TCs and our current sensors, please cough it up.  And I don't mean buying a $50,000 lens projector, I mean using stuff I actually have.

[jrista]

You're conflating resolution at the sensor in lp/mm (which is largely irrelevant) with optical angular resolution (which is what we care about - what details in the scene can be resolved).

It can't be irrelevant...because the resolution AT THE SENSOR is WHAT THE SENSOR cares about (and therefor, what I care about in the context of this discussion.) Spatial resolution of the real image (the projection of the lens at the focus, no focal, plane...at the sensor's surface) is exactly relevant when discussing what a sensor is actually resolving and registering, and what is converted into a digital image by the ADC, as compared to what a TC is doing to that very same real image. The ability of a lens to discern detail at the focal plane, in the virtual image, is not in the context of the discussion here. The focal plane could be twenty feet away, it could be a thousand feet away, but that won't change the spatial resolution of the real image projected by the lens onto the sensor.

I think your inverting the problem, and thinking about everything exterior to the camera. Spatial Resolution of the image projected by the lens onto the focus plane (which exists INSIDE the camera AT the sensor) is quite explicitly what I am referring to. I don't see how you could logically discuss anything else when comparing the effects of magnification by a TC in relation to increasing sensor spatial resolution...all of that exists inside the camera, behind the aperture, not outside in the real world relative to the front element of a lens or its entrance pupil.

[TheSuede]

You are seriously talking past each other now, and things are mixed up beyond belief.

Disregarding the real-world effects of a TC (increased reflection and absorption losses, decrease in sharpness due to optical imperfections) from now on through this entire post: Yes, of course a TC magnifies the diffraction circle by exactly the same amount as the rest of the image. But that is also the point; the object referred diffraction is already determined at the front of the optical system, by the entrance pupil (as long as we're within reasonably Gaussian systems, for microscopes and other applications with very high magnification you need to look at angular aperture in stead of numerical aperture). A teleconverter will magnify both this object referred diffraction and target detail, a wide-converter will decrease magnification on both diffraction and target detail. It varies the projected image magnification and not the angular object referred diffraction, which is what optically limits your target resolution.

You are misunderstanding what a TC does. A teleconverter is a magnifying glass. It simply enlarges what the original lens projects. It magnifies everything....including diffraction. You cannot add a TC to a lens an not increase the effects of diffraction, despite the facts you just described above.

What YOU forget to mention in your little maths excursion under the qouted text above is that the f/# number also implicates that you have a reproduction scale. This indicates both that you're within Gaussian optics rules (as opposed to in microscopy, where angular aperture is the metric used) and that you have an Airy disc diameter that is constant with f/# on the image plane. A 50mm used on f/11 will give the same Airy disc size on the sensor as a 100mm f/11 used on teh same sensor.

But the target magnification (reproduction ratio) is twice as high on the 100mm option, so you have:
-same Airy disc size on the sensor
-twice the reproduction ratio!

This means that if you shoot a side view of say "a car" from a distance where 50mm would give you a car length of 500 pixels on the image and a diffraction effect of maybe 3 pixel widths, using the 100mm lens at f/11 would give a car size on the sensor of 1000 pixels, but still a diffraction effect of 3 pixels - and that means that you've halved the diffraction effect on the car, i.e halved the angular object referred diffraction. Doubled the optically limited usable target resolution - since you doubled the entry pupil size.... (50/11 = 4.5mm pupil, 100/11 = 9mm pupil)

Diffraction and Airy disc size are linearly scaled by the same constant since the angular diffraction in front of the lens depends on the entry pupil diameter and nothing else (until you hit the Gaussian model limit, see angular aperture). If the airy disc covers, say, a one inch detail on an object far away, the SAME one inch detail will be covered by the exact same relative Airy disc, no matter what magnification/resolution you inspect the projection with in the image plane.

This isn't actually very hard to see in reality (most of us do have a zoom lens available I suppose?)
Take one shot at say 100mm and F11 of a distant object. With the same camera, directly after that take another shot aimed at the same target from the same distance, but now with 200mm and F22. They both have the same entrance pupil diameter (9mm), and they're both into diffraction limited range on most modern sensors.

Which of the two will have the highest target resolution?
Since one has twice the target reproduction ratio or magnification we need to either downsample the 200mm image or upsample the 100mm image to compare them at equal size presentation. And unless your aperture calibration is seriously off on the lens you use, the 200mm F22 shot will have equal or better target resolution!

Don't doubt, try for yourself.

In astro (which is a purely Gaussian limited application with ordinary systems, with infinity focus targets) this is extremely important, since the angular resolution in front of the lens is determined by the entrance pupil. NOTHING you do behind that can make things better, in any way. It doesn't matter what focal length you use, the entrance pupil determines how small the (infinity distance) details you can accurately resolve optically is. Within practical limits of course, but this depends more on lens manufacturing and smallest available pixel size with good performance. You won't find many spectacularly detailed shots of the moon taken with a 24mm lens.

Keeping the entrance pupil constant: If you use a shorter focal length you get a smaller reproduction ratio, and you need smaller pixels to accurately resolve the optical projection image. Use a longer focal lens, and  you can use larger pixels. It will not in any way have an effect on the object space angular resolution of the system, you just adapt the sensor resolution to fit the optical resolution.

So: You get the same optical far-field target resolution (again using the elusive "perfect" TC) if you use a 400/4.0 with 2x TC, as if you use an 800/8.0 on the same camera, or indeed as you do if you use a 400/4.0 on a half size (quarter area) sensor with the same amount of pixels. Diffraction limitation of the target does not change, light energy per pixel captured per second does not change.

[jrista]

You are seriously talking past each other now, and things are mixed up beyond belief.

Disregarding the real-world effects of a TC (increased reflection and absorption losses, decrease in sharpness due to optical imperfections) from now on through this entire post: Yes, of course a TC magnifies the diffraction circle by exactly the same amount as the rest of the image. But that is also the point; the object referred diffraction is already determined at the front of the optical system, by the entrance pupil (as long as we're within reasonably Gaussian systems, for microscopes and other applications with very high magnification you need to look at angular aperture in stead of numerical aperture). A teleconverter will magnify both this object referred diffraction and target detail, a wide-converter will decrease magnification on both diffraction and target detail. It varies the projected image magnification and not the angular object referred diffraction, which is what optically limits your target resolution.

You are misunderstanding what a TC does. A teleconverter is a magnifying glass. It simply enlarges what the original lens projects. It magnifies everything....including diffraction. You cannot add a TC to a lens an not increase the effects of diffraction, despite the facts you just described above.

What YOU forget to mention in your little maths excursion under the qouted text above is that the f/# number also implicates that you have a reproduction scale. This indicates both that you're within Gaussian optics rules (as opposed to in microscopy, where angular aperture is the metric used) and that you have an Airy disc diameter that is constant with f/# on the image plane. A 50mm used on f/11 will give the same Airy disc size on the sensor as a 100mm f/11 used on teh same sensor.

But the target magnification (reproduction ratio) is twice as high on the 100mm option, so you have:
-same Airy disc size on the sensor
-twice the reproduction ratio!

This means that if you shoot a side view of say "a car" from a distance where 50mm would give you a car length of 500 pixels on the image and a diffraction effect of maybe 3 pixel widths, using the 100mm lens at f/11 would give a car size on the sensor of 1000 pixels, but still a diffraction effect of 3 pixels - and that means that you've halved the diffraction effect on the car, i.e halved the angular object referred diffraction. Doubled the optically limited usable target resolution - since you doubled the entry pupil size.... (50/11 = 4.5mm pupil, 100/11 = 9mm pupil)

Diffraction and Airy disc size are linearly scaled by the same constant since the angular diffraction in front of the lens depends on the entry pupil diameter and nothing else (until you hit the Gaussian model limit, see angular aperture). If the airy disc covers, say, a one inch detail on an object far away, the SAME one inch detail will be covered by the exact same relative Airy disc, no matter what magnification/resolution you inspect the projection with in the image plane.

This isn't actually very hard to see in reality (most of us do have a zoom lens available I suppose?)
Take one shot at say 100mm and F11 of a distant object. With the same camera, directly after that take another shot aimed at the same target from the same distance, but now with 200mm and F22. They both have the same entrance pupil diameter (9mm), and they're both into diffraction limited range on most modern sensors.

Which of the two will have the highest target resolution?
Since one has twice the target reproduction ratio or magnification we need to either downsample the 200mm image or upsample the 100mm image to compare them at equal size presentation. And unless your aperture calibration is seriously off on the lens you use, the 200mm F22 shot will have equal or better target resolution!

Don't doubt, try for yourself.

In astro (which is a purely Gaussian limited application with ordinary systems, with infinity focus targets) this is extremely important, since the angular resolution in front of the lens is determined by the entrance pupil. NOTHING you do behind that can make things better, in any way. It doesn't matter what focal length you use, the entrance pupil determines how small the (infinity distance) details you can accurately resolve optically is. Within practical limits of course, but this depends more on lens manufacturing and smallest available pixel size with good performance. You won't find many spectacularly detailed shots of the moon taken with a 24mm lens.

Keeping the entrance pupil constant: If you use a shorter focal length you get a smaller reproduction ratio, and you need smaller pixels to accurately resolve the optical projection image. Use a longer focal lens, and  you can use larger pixels. It will not in any way have an effect on the object space angular resolution of the system, you just adapt the sensor resolution to fit the optical resolution.

So: You get the same optical far-field target resolution (again using the elusive "perfect" TC) if you use a 400/4.0 with 2x TC, as if you use an 800/8.0 on the same camera, or indeed as you do if you use a 400/4.0 on a half size (quarter area) sensor with the same amount of pixels. Diffraction limitation of the target does not change, light energy per pixel captured per second does not change.

I don't disagree with anything you have said. I believe I explained the effect of magnification (reproduction ratio, same thing) in my answer at #78 (twice the diffraction but 4x the subject size, so 2x the detail.)

You've put it purely in the context of astrophotography, which is effectively hyperfocal or "infinity" distance only. (Something I already disputed before as a very small niche of photography that has unique circumstances not generally experienced in normal photography.) Lee Jay has implied that what occurs in astrophotography applies everywhere, and that using a TC is just as good or effectively the same thing as using a higher resolution sensor, or that use of a TC can always be measured in terms of "full-frame sensor megapixels" of resolution (another highly misleading concept...do you see my pattern yet? ;P)

Lee Jay has the tendency to evade rather than debate head on as well...so I get a little irked when I have to keep corralling him. Apologies to both of you.

In astro (which is a purely Gaussian limited application with ordinary systems, with infinity focus targets) this is extremely important, since the angular resolution in front of the lens is determined by the entrance pupil. NOTHING you do behind that can make things better, in any way.

Well, that would only be true if a sensor with larger pixels is the limiting factor in terms of spatial resolution. If, assuming an astro context, and diffraction is only 5.8 microns (the 173lp/mm spatial limit of an f/4 aperture), but your sensor uses 6.95 micron pixels (such as the 1D X)...switching to a sensor with 5.8 micron pixels (a hypothetical 26mp sensor...a change in something behind the diaphragm) would indeed improve the detail and quality of the RAW image actually produced by the camera. Would it not? 


Let me ask. Do you believe adding a 2x TC to a 400mm f/4 lens (800mm f/8) is the same as using just the 400mm f/4 lens on a sensor with half the pixel pitch, in a general photographic frame of reference (vs. just the astrophotography frame of reference)? Or would you agree that using the 400mm f/4 lens with a sensor twice as dense will produce just as detailed output that encompasses a wider field of view (greater total area) than the TC setup?

In the former case, an 800mm f/8 lens on a FF sensor with say a 5.8 micron pixel pitch. Diffraction won't affect resolution enough to matter on that sensor, as the pixel is the same size as the airy disc.

In the latter case, a 400mm f/4 lens on a FF sensor with say a 2.9 micron pixel pitch. Again, diffraction won't affect resolution enough to matter on this sensor, as the pixel is the same size as the airy disc.

Assuming you use both setups to photograph a landscape of some kind...a small waterfall at some distance. Let's assume the entire waterfall fits on the FOV of the 400mm lens. Would you agree that the 800mm f/8 5.8um setup would capture only 1/4 of the total area of the waterfall? Would you agree that the 400mm f/4 2.9um setup would not only capture the entire waterfall, but that it would also capture the same 1/4 area as the 800mm setup in nearly the same detail?


My primary key point here is not so much that the 800mm f/8 setup is capable of reproducing that 1/4 area of the waterfall in high detail. I've never disputed that (I believe my post at #78 entirely agrees with you on that point, actually.) My point is that the 800mm f/8 5.8um setup is capable of reproducing only 1/4 the area of the waterfall, while the 400mm f/4 2.9um setup is capable of reproducing the ENTIRE waterfall, with roughly same amount of detail in that same 1/4 area, as well as roughly the same amount of detail in any other 1/4 area that you could crop from the original frame.

My second key point here is that no matter what you do with any number of TC's...the spatial resolution of the real image at the plane of focus (the sensor) is intrinsically limited by the spatial resolution of the sensor you are actually using. Saying that a TC added to a lens on an 18mp sensor suddenly gave you the same "resolution" as a 369mp sensor of the same dimensions is a fallacy.

(At least, in the frame of reference of sensors, who's resolutions are always measured in terms of spatial resolution. If you wish to move to a different frame of reference and use a different measure of resolution such as angular resolution, you need to make all of that very clear, and make sure you transform EVERYTHING, all numbers and units for all participating elements of the discussion, into the same frame of reference...I'm not really sure how you measure a sensor in terms of angular resolution. Additionally, it is the sensor that "sees" in a camera, not something external, not even the front lens element that is gathering the light...it is the sensor that sees and records an image. So it seems logical to me to remain in the original frame of reference: Spatial Resolution at the Sensor).

To keep things consistent, if the discussion continues. Can we use the following sensors, cameras, and lenses?

Sensor A with 5.8 micron pixels (25.6mp FF)
Sensor B with 2.9 micron pixels (102.7mp FF)

Lens A is 800mm f/8 (400mm f/4 lens with 2x TC)
Lens B is 400mm f/4

Camera A with Sensor A and Lens A
Camera B with Sensor B and Lens B