Ok, guys, let me muddy the waters even more. Let's talk "sensor" first. CCD stands for "Charge Coupled Device." A CCD array consists of a large number (i.e., an array) of light-sensitive structures that, when struck by light, store electric charge at a rate proportional to the light hitting the regions. The structures also allow the charge to be temporarily stored. CCD arrays also have electronic structures that allow the charge to be moved to the edges of the chip so that it can be measured and converted to bits by one or more Analog-to-Digital converters. The shifting structures take up space on the image arrays. As a result, not all of the array area is actually light sensitive. The Sony link from my previous writeup showed how micro-lenses are used to compensate for the area on the chip surface that is not sensitive to light.
So, when I said "sensor" in my previous post, I meant one individual light-sensitive region, rather than the CCD array as a whole. In this context, I view the the imaging chip as an array of sensors. I believe that the Sony diagram is conceptually accurate and not just a "cartoon" as Bob referred to it. Let's talk about "light" now. The approach that I am describing now is commonly called "ray optics" (as opposed to "wave optics" or the quantum theory of light). "Ray optics" theory has strengths and weaknesses. If you want to debate those, that's a separate topic. I never said that light had to strike the array precisely perpendicular to the array for it to be measured. I said, "...the key is not to make the light purely perpendicular, but rather to minimize the angle." Now, given the size of the light-sensitive portion of an element of a CCD array, the optical properties of the microlens in front of each element, and the distance from the microlens to the element, it's not too hard to work out a formula that will give you the drop-off in measured light intensity as a function of incident angle. For film, the drop-off is something like cos(theta), where theta is the angle away from perpendicular. For a sensor with a microlens, the function drops off faster than cos(theta). Large-format photographers are well aware of this drop-off. That's why they use center filters with wide-angle lenses--to compensate for the drop-off. Now, consider a long-focal-length "simple lens" and a short-focal-length "simple lens." Sketch out ray tracing diagrams for both focused at infinity, and you'll find that the long-focal-length rays hit the image plane much closer to perpendicular than those of the short-focal-length rays. As a result, it seems reasonable to conclude that rays from long-focal-length lenses will have smaller values of theta than rays from short-focal-length lenses and hence will show less light fall-off in the corners of the image plane. I believe that retro-focus wide-angle lenses pass light through to the image plane at much lower angles than simple short-focal-length lenses. Hence, these lens designs already at least partially compensate for light-falloff due to microlens CCD array properties. So let me restate the points that I tried to make in my previous statement: 1) The angle at which light hits the focal plane is more important for a CCD array than for film. 2) Lens designers can control incident light ray angles through clever lens design. Finally, I think that most of the above is reasonably accurate, but I could be completely wrong about something. If you disagree, I challenge you to do a web search, crack a physics book or two, and back up your point of view. --Mark p.s., here are some links that describe different kinds of CCD and CMOS arrays: Generic description of imaging arrays: http://www.howstuffworks.com/question362.htm Kodak description of why the 4/3 size is better--includes a description of "telecentric" vs "non-telecentric" lenses: http://wwwhk.kodak.com/global/plugins/acrobat/en/digital/ccd/papersArticles/ 3-4TypeImageSensors.pdf A student project describing frame readout, frame transfer, and line transfer CCD arrays: http://www.ece.iit.edu/~pfelber/ccd/project.pdf A brief description of "fill factor" in imaging arrays: http://www.ph.tn.tudelft.nl/Courses/FIP/noframes/fip-Pixel.html
