Re: Rambling load all right

In article <1127941648.059540.293340@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>, "winhag@xxxxxxxxx" <winhag@xxxxxxxxx> writes 
Kennedy,

A purely educational question. You talk about 'telecentric' lens
designs.
I had never heard of this and looked it up on the web. From what
I could understand, a telecentric lens maps objects of the same
dimension
to the same size on the film plane regardless of object distance from
lens (i.e.
no 'perspective distortion'). Is that what you are referring to or
something different?

Yes.

If you draw out a simple lens then all of the principle rays (the principle ray is the central ray in the bundle) to every point between the object and its image on the focal plane passes through the centre of the lens. This means that there is only one point in the image where the principle ray is perpendicular to the focal plane - the centre of the focal plane. At every other point on the focal plane, the principle ray diverges from the centre. The easiest way to visualise this is just to draw a lens with an object and its image on a *** of paper and then draw in the principle rays from the object to the image, making sure that they all pass through the centre of the lens. Now, if the angle of incidence at the focal plane is not 90deg, there is a chance that some of the light passing through the filter will actually land on an adjacent pixel rather than the pixel some small distance under the filter, and this causes colour distortions because the camera expects all of the light falling on certain pixels to have been filtered to the appropriate colour. Obviously the greater the angle from perpendicular of the principle ray, and the lower the f/# (ie. the wider the cone formed by the bundle of incident rays around the principle one), then the more light that can 'leak' into adjacent pixels this way.#

With a telecentric lens, all of the principle rays reaching the focal plane are perpendicular to it and, since the same f/# must be produced across the focal plane, this means that each point of the image does not use all of the available lens aperture - so the rear lens elements are physically larger than would be required normally. Of course, that means that all of the rays incident at the corner of the focal plane are coming from the periphery of the lens, making it more difficult to control all aberations, not just chromatic. So telecentric lenses tend to have a poorer performance than conventional designs - or they are much more expensive for the same performance. The main reason for producing such a lens in pre-digital days was, as you have found in your net search, to ensure that images did not change size when the system was focus was moved. This would be important in instruments measuring dimensions directly from the image. However, since telecentricity means that the principle ray to every point in the image is perpendicular to the focal plane, there is an argument that it is ideal in digital sensors to ensure that the amount of leakage from one filter to adjacent pixels is fixed - although it does still change with f/# of course.

In most cases though the CA from non-telecentric lenses argument is overhyped and you just need to examine the pixel dimensions to see how spurious it is. Based on the minimum back working distance to clear the mirror, which is proportional to the sensor size, the angle of incidence in the corner of the field for the principle ray from this worst case situation is almost the same in both formats - leaving only the pixel dimension as being the critical parameter. Smaller format sensors usually have smaller pixels and they consequently suffer more from this problem than larger sensors - completely contrary to the hype.

However, putting this all into perspective, the semi-angle of the light cone itself with a fast optic is about the same angle as the worst case off-perpendiularity of the non-telecentric lens. So even a perfect "designed for digital" fully telecentric lens will produce the same colour distortion across the entire field as the worst case simple lens will produce at the corner. Added to which, that worst case only occurs at one particular focal length - shorter focal length lenses than this are partially telecentric because of their inverse telephoto design keeping the principle ray closer to perpendicular even in the corner of the frame, whilst longer focal length lenses produce principle rays closer to perpendicular in any case.

In short, the entire CA argument (and this is one of the cornerstones of the entire 4/3 strategy!) is based on nothing more than hype. 4/3 and APS formats are there because they are cheaper to make, not because they or their optics are intrinsically superior.

If there is CA in the image then it is almost entirely coming from the lens, just in the same way as it did on good old analogue colour film where, incidentally, the distance between the layers of emulsion isn't too different from the distance between the filter and silicon in a CCD or CMOS sensor.

Yes, this issue will eventually become a problem, when pixel sizes get down to 3um or less on these larger focal planes, but obviously the smaller 4/3 and APS cameras will need to be there a long time before full frame cameras do, just on the pixel count wars alone. And to make it worse, these small size pixels will be restricted to faster minimum apertures (as they currently are in P&S cameras) to take full advantage of the pixel count in terms of real resolution, making the CA problem even more significant - telecentric lenses or not.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)
.