Fast Telephotos, front element size, and EF-S questions...

So I've heard it before a while ago, both here and CR and other forums, and it's come up again lately in the 1dx vs 5d3 thread. The question is whether an EF-S über-tele would have any size/weight (but most importantly) price advantages over the regular EF-mount lenses (let's just focus on build size, not L-lens quality/sealing/etc).

Now, the "entrance pupil" is what determines aperture. So a 300 f/2.8 has an entrance pupil of 300/2.8 = 107mm, a 600 f/4 has 600/4 = 150mm.
But because of different lens designs, the "entrance pupil" is not (always) at the front element, it's a "virtual" pupil that could be anywhere. The best guess is that it's just behind the front element (as used in panoramic photography), if you rotate around the entrance pupil you get parallax-free stitching.

So here's the thing that's confusing me. I've got a Zeiss Sonnar 180mm f/2.8 lens. Originally made for Pentacon Six, frame size of 56 x 56mm, image-circle diameter is 80mm (compare FF image circle is 43mm, 7D is 27mm). Now, it's got a 180/2.8 = 64.3mm diameter entrance pupil. The front element has a filter ring of 86mm, and the lens barrel is pretty much that diameter for most of the lens until it tapers to the mount (inside diameter of the mount is about 55mm, the rear element is about 35mm dia). The front element, at infinity focus, is pretty much bang-on 180mm from the sensor (ie, it's not a true 'telephoto' in the lens-design sense of the word, it's just a "long focus lens").

But here's the thing.
I put it on my 7D, and I can mount any number of Step-Down rings, down to my smallest of 37mm diameter, and it doesn't vignette at all. But it's still a 180/2.8, so the entrance pupil must still have a diameter of 64mm. Does this mean that the entrance pupil is outside the lens? Or that the pupil is inside the lens, light rays are passing the 37mm step-ring, then being bent 'outwards' inside the lens, to the pupil-size of 64mm? (that doesn't make sense to me, but it could happen)

Maybe that lens is a fluke?
Let's try my other long-fast (ish) tele. Jupiter 250mm f/3.5. Alleged entrance pupil size of 250/3.5 = 71mm. Again, step rings to 37mm and it doesn't vignette, infinity or mfd, f/3.5 or f/22, any combination or anywhere in between. The lens is once again about the same as its focal length, 25cm front-element to the sensor.

OK, lets forget the old East german and soviet glass.
EF 70-300 f/4-5.6L. Front element 67mm.
@70mm, f/4.0, entrance pupil = 17mm. Step rings down to 37mm do nothing.
@300mm f/5.6, entrance pupil = 53mm. step ring to 37mm, again, no vignetting.
This one's a zoom, so a bit different, length at 70mm is about 15cm, at 300mm length is 25cm or so.

How about something faster?
EF 100mm f/2.0, pupil = 50mm, front element 58mm. 37mm step-ring? No vignetting.
FL 55 f/1.2, pupil = 46mm, front element 58mm. Nada.
Samyang 35 f/1.4, pupil = 25mm, front threads 77mm (element a bit smaller). Finally, it vignettes. At 52mm step ring, I could possibly see some, but it may just be the edge-of-sensor phenomenon. With 37mm (i've got no sizes in between), f/1.4 it's a bit worse than my ef-s 15-85 @ 15mm f/3.5. At f/22 it looks like a bad fisheye, the sides are blackened a bit too.


So the results get less valid for the wider focal lengths. But the confusion remains the same. How can I have (effectively, using step rings) a front-element diameter of less than the entrance pupil? If the Zeiss Sonnar was remade as an EF-S long-skinny thing, could it have a front element of 37mm (like maybe filter threads of 52mm)? Do I just not have lenses long enough to test (like 4-500mm+), is the theory of "ef-s long teles aren't any smaller" only hold above 400, 500, 800mm? Does anyone want to post me an über tele to try it out? :) Or test yourself, i know lots of people here would have long-fast lenses and at least one crop-camera. (front elements are so huge you're probably stuck with cutting holes in black paper to emulate front element sizes).

I'm not a lens designer, i've got more chance of building my own diy sensor than building a lens, so maybe i've missed something completely obvious. But still, could we (theoretically) have long-lenses in EF-S mount that are smaller, lighter, cheaper than the EF-mounts? (It's not going to happen from a marketing perspective, at least until the 7D2 or 3D come out with APS-C 45pt f/8 AF, ie never), but still, i'm intrigued (and confused)...


(And if i've learnt one thing from this exercise already, it's that I do up my step-rings too tight, and if you pay $2 for filter wrenches off ebay, they last just as long as you'd expect them too...)

The entrance and exit pupils are not something that actually exist as physical objects. Only the aperture stop itself is a real object. The pupils are simply the images of the aperture stop as viewed from the corresponding end of the lens, and as such, they can appear at any position behind the end of the glass you're looking at. The entrance pupil can, for instance, appear behind the front element at a distance that is greater than the physical length of the lens.

That said, the entrance and exit pupils are in fact optically constrained by the diameter of the glass in front of them. That is to say, an entrance pupil of 100mm cannot be produced if the front element is only 90mm. It is impossible, in as much as it is impossible for the entrance pupil to appear in front of the front element.* If you reduce that front diameter through the introduction of a secondary stop, you will reduce the amount of light entering the lens, and therefore, the f-number of the system will change. Whether or not this results in noticeable vignetting depends on how the lens is constructed and where you put the field stop. In some cases (frequently, telephoto lenses), what you would see is simply a change in exposure. In others (e.g., wide-angle lenses), you would see hard vignetting whilst the central portion of the image remains unaffected.

Something else that you should consider when conducting your tests is that the f-number of the system pertains to the behavior of the lens when it is focused at infinity. If you take your telephoto lens and look through the front element from a few feet away, the entrance pupil will appear a certain size. If you set the lens down and walk away from it, the entrance pupil will seem to grow in size until it appears to take up the entire front element's diameter. This is because when you look at the lens, the light rays that form the entrance pupil correspond to the effective f-number of the system, not the relative f-number. When you are close, the light rays emerging from the lens and reaching your eye are not parallel and the actual light-gathering ability is less than what the nominal aperture would suggest. When you are far, the light rays are nearly parallel, and this corresponds to the case where N = F/D. The reason I mention this is because if you were to conduct your tests by imaging a subject at a short distance, you would not notice a significant change in exposure with the step-down ring because the effective f-number has already been increased. If you shot the lens at infinity focus, the camera absolutely would see less light if you blocked the periphery of the front element with step-down rings.

As for your question regarding the relationship to sensor format size, a smaller sensor could make part of a telephoto lens smaller but not the front element. With a shorter back focus distance and a smaller image circle, the lens could be made a bit more compact, but not drastically so, if the f-number is to remain fast.

*When I speak of the pupil diameter, I am speaking of the size of the pupil as it appears to us (or the camera sensor), not as it actually is. For example, a pupil that is 100mm in diameter at a distance of 50cm away looks the same size as a pupil that is 200mm in diameter at a distance of 100cm away. The light cone they produce has the same angle. A camera sensor cannot tell the difference between the two. Thus, you can have a pupil that is infinitely large yet infinitely far away, but it still appears to us to have a finite size (e.g., telecentric lens) and thus a finite light-gathering capacity. As the pupil diameter pertains to this capacity, we are therefore interested only in the angular size of the pupil--how large it appears to us, not how large it actually is, in relation to its distance from us.

Did adding the step-down rings make the image darker? It doesn't have to be vignetting, stopping down a lens the regular way doesn't add vignetting either, just reduces the amount of light that passes through. I expected the image to get darker, so, with all manual settings, including ISO, the image should have been underexposed with the step down rings.