See image below
To realize a 30mm lens, the 'effective' focal length F for the lens must be constructed such that the rays hitting the sensor
d mimic the FOV of a 30mm lens.
With a flange distance of 18mm, that leaves 12 mm for the lens. But this is only true if the lens is a single optic, and most are at least five or six elements (to affect COMA, etc).
You can place the group of lens elements at either left or right side of the 'single lens', but you have to cope with the effects. See discussions wrt/ nodal points of lenses.
If you place the elements towards the right of the lens, closer to the sensor, you have to focus the image on a wider area (magnify). (1)
If you place the elements towards the left of the lens, further away from the sensor, you have to focus the image on a smaller area (retro-focus). (2)
If you want a high quality optic, solution (2) gives much more flexibility to the engineers to create an 'optical sound' solution.
But - this is 2012, and we have in-camera processors !!
Solution (1) can be corrected for in software, IF you know the lens characteristics. E.g. the Canon S90/95/100 camera has optically a 'fish eye' view with lots of barrel distortion. In-camera software compensates for this to make it a high quality picture.
On a fixed lens camera, like the Fuji X100, you have full control on optics and software correction.
Oh - but this is not the entire story.
Leica and Fuji already noticed that light fall-off on the edges is a problem with modern sensor designs, and compensated for this by mis-aligning the micro-lens array with the pixel sensor array. Sony has yet to do this - there are reasons to not do this.
On a Leica or Fuji solution, the camera must always know which focal length is placed in front of it, or the in-camera software will 'mess' up the picture. In the Sony, the picture is wysiwyg - it is optically 'the best' possible.
To create a pancake that has e.g. a 30mm Focal Length and protrude as much as the 16mm lens, you would essentially create a 'inverse fisheye' lens on the rear-side of the lens (the sensor side). This lens will have horrible distortion, barrrel, CA, COMA - all of which have to be corrected for in software.
And, by placing all your optics in the (1) solution, you have a limited space for them - this gives the engineers only so many choices - and quality will be compromised as a result. The faster the lens (larger aperture), the more compound the problems. At a higher f-stop, light bending is a lot simpler than at a lower f-stop - for the different wavelengths of visible light. The 16mm lens is very small, but it protrudes about one inch (2.5cm) above the flange distance of 18mm, for a total length of 4.3cm from the sensor - and it is 'only' f/2.8 (which is easy to do achieve for a 16mm - try doing this for a 32mm lens - then try to increase this to f/2.0, lol).
E.g. the Zeiss 24mm has all its optics elements in front of the focal point (2) and, I believe, actually retrofocusses its image to the sensor. Something that even the newer Leica's do. And some prefer the larger Distagons over the smaller Biogons for the same reason.
So yes, pancakes are possible, but not easy when you make them optically correct. Allowing 'distorted' lenses, to reduce sizes, requires adding lens registration and in-camera compensation software becomes a must-have, and decisions whether to align or mis-align the micro mirrors, thickness of AA filter, and so on are all part of the solution.
And ask yourself: if a lens protrudes 4cm versus a lens that protrudes 2cm (e.g. for a 30mm focal length) - would you insist on the 2cm lens with all the compound problems, or would a 4cm lens be acceptable. (I believe that the SEL30M35 protrudes 4.5cm)?
- Side note: as everything 'scales' with the size of the sensor, a smaller sensor has essentially 'more' space in the same size (pancake) lens. Meaning it is easier to have pancake lenses for a Q or V1 system than for a M43 or APS-C system.
One of the attractions of the Nex system is that it is an 'optically correct' system, allowing adaptation of so many legacy and other name brand lenses. (In 'film' days, there was no 'correction software' possibility).
Going towards a more complex system for 'smallness' is a big step. I am not sure that this is what makes the Nex a Nex. But - as competition gets ahead and shows that it is feasible, I am sure we will get newer solutions.
Disclaimer - I am not an optics designer, not did I study the subject in depth. I merely reflect what I have read in recent articles and discussions around these issues.
- Take this as a guidance only - I may be erroneous in some assumptions
- Repeating my earlier point: keeping the Nex an 'optically correct' system is, imo, a big attraction. If pancakes are a "must have", then maybe the Nex is not the right choice? Look into a Q system?
- Last point - there is 'optics correction' software in the Nex-5N and Nex-7, which is different from the issues discussed above - they deal with lens inefficiencies, such as barrel distortion, vignetting, softness, etc. that are properties of the lens. The in-camera correction software does essentially what the plug in modules do in DxO and LR3 software.
I don't know much about optics, but there must be a reason why there hasn't been a pancake prime with large aperture for NEX. All the existing primes (30mm, 50mm, 16mm) are either large or have small apertures.
My guess is Sony limited their lens size by using the APS-C sensor size.
So maybe it is futile to keep waiting for a pancake prime like the M43 20mm 1.7 lens for NEX cameras? If that is the case, then maybe it's time to accept the inevitable small body large lens aka NEX system.
--
Cheers,
Henry
