Also, the size of the rear element of the lens, for say a leica m mount camera, increases in size with aperature. When would this cause problems in the m mount? 25 f1.4? Any equations out there to calculate this stuff?
A lot of this depends on the construction of the lens. Here's a couple of general ideas:
The aperture number is more or less as the focal length divided by the lens' entrance pupil (the entrance pupil of the lens is the apparent diameter of the aperture). How the apparent diameter is related to the physical diameter depends on the lens design. A 50/f1, for example, needs an apparent aperture as big as the focal length. A 200/f1.8 for a SLR needs an apparent aperture of more than 10cm in diameter, which is big. While the apparent diameter can be quite different from the physical diameter, this gives you a very general idea how big a lens has to be diameter-wise for a given speed. (That's one of the reasons why a 21/f4 can be built much smaller than a 500/f4.) It also creates all sorts of interesting construction problems when building things like a constant-aperture zoom, because, say, in a 24-105/f4 lens, the apparent diameter has to change from 6mm to 26mm over the zoom range. That's why there are no constant-aperture superzooms - try the numbers on a hypothetical 18-200/f2.8.
If you have a more or less symmetrical lens, like the typical 50, the exit pupil of the lens needs to be about as big as the entrance pupil. With a more or less symmetrical 50/f1, for example, the exit pupil should be about 50mm in size. Now the M mount only has 41mm or so in diameter, which leads to all sorts of constraints because you can't obviously build the exit pupil of the lens bigger than the throat of the bayonet it's supposed to go through. This is one of the reasons why the Noctilux vignettes wide open. To avoid vignetting you have to modify the lens construction to make it less symmetrical, which makes it more difficult to correct aberrations.
Another constraint is the size of the image circle; a 25/f1.4 lens is relatively easy to build symmetrically (after all you only need an aperture of 18mm or so, which gives you an idea of the size of the lens elements). But because it sits so close to the film plane, it would only cast a relatively small image circle, so the construction would have to become more extreme with bigger lenses. Then you run into a size problem with the bayonet throat relatively quickly. A classic example is the Biogon 35/f2.8 for the Contax, which has a huge rear element that sits close to the film plane to make a relatively fast non-retrofocus lens which covers 35mm film, but which couldn't be built any faster because the rear element would no longer fit through the relatively narrow throat. The only way out is a retrofocus, non-symmetrical construction, which invalidates many of the assumptions underlying these simply calculations, and which also leads to asymmetrical relations between the front and rear elements. That's why lenses such as the Canon 24/f1.4 L have huge front elements in relation to the relatively short focal length of the lens. I have a Flektogon 50/f4 medium format wideangle that covers 6x6 and needs a lot of clearance for the Pentacon 6 mirror box - focal lens is 50mm, lens register is 74mm, and it really needs those 74mm clearance. To accomodate this, the construction is completely asymmetrical with a small rear element of less than 20mm (so no problems whatsoever with the bayonet size), but with a huge front element that needs a 86mm filter ring.
Symmetrical lenses were interesting because they were easy to correct for distortion and aberrations, but in effect nowadays they are largely overrated, because computer-aided lens construction has made it so easy to calculate distortions that there is really not much of an advantage, while there are a lot of drawbacks when building fast lenses. The idea that rangefinders were superior because you can build better symmetrical wideangles was correct up until the 1970s or so, but is now largely obsolete IMHO.
When you start building asymmetrical, retrofocus lenses which use a tele construction (or for wideangles an inverted tele construction), much of this becomes much more difficult to calculate because the construction affects things like the apparent aperture size. However, calculating difficulty is no longer a problem now that everything can be simulated. In principle then you can build lenses arbitrarily fast without caring for the physical parameters of the bayonet. The tradeoff for this is that the construction needs extra lenses and tends to become huge very quickly.
Philipp
