Kent Gittings wrote: [Lots of stuff on CCD imaging.... Quotes and comments interspersed below]
Hi Kent, Just a few notes, interspersed with quotes from your earlier message.... > The common type of CCD/CMOS array currently being used is > front illuminated. That means the part that illuminates the pixel is in > front. This means the light getting to it must pass through and > around this part of the array. As a result I'm pretty sure in the case > of front illumination the angle of incidence (angle it strikes the array) > is very critical to minimize distortion from the hardware in front. It's certainly true that the field of view of a single pixel is limited. This is as it should be, since you want that pixel to see only the tiny speck of light directly in front of itself, rather than having it respond to the speck of light in front of its nearest-neighbor pixel. However, this is not the same thing as saying that the pixel will "see" only those rays of light that strike the focal plane in a perpendicular fashion. The image formed at the focal plane array -- in ~front~ of the pixels -- can be formed by rays coming in from a very large cone of light. A simple thought experiment will help illustrate this. Imagine setting up a very small CCD array -- perhaps only a few tens of pixels wide by a few tens of pixels long -- on a gigantic telescope at an observatory. The telescope is moved in a direction where the image of only a single star is focused onto the sensor array. The telescope mirror focuses light from across its very large aperture onto this single pinpoint star image, and the rays striking the array form a cone with a large angle -- far from the perpendicular condition you describe. You'll agree that the star image is exceedingly bright in this case, right? Now imagine stopping down the aperture of the giant telescope so that only those rays close to perpendicular are allowed through. Let's say that to make sure the rays are very close to perpendicular, we stop down the aperture to only a couple of inches -- perhaps a hundred times smaller in diameter, and ~very~ close to the perpendicular condition you describe, right. What happens to the brightness of the image of the star? Certainly, you'd agree that the image formed at the focal plane is ~much~ dimmer now, and that the signal level measured by the corresponding pixel is ~much~ lower as well, even though the intensity contributed by the ~perpendicular~ rays hasn't changed at all. > The angle of incidence (angle it strikes the array) is very critical to > minimize distortion from the hardware in front I know I just quoted this same sentence above, but it bears further comment. Distortion doesn't come into play here in the way you suggest. Remember that the optic creating the image is the macroscopic-sized lens mounted on the front of the camera (or in my example above, the huge astronomical mirror). Any distortion in ~this~ optic will of course degrade the image formed at the focal plane. Once you've broken this real image into quantized bits, or pixels, there's no more optical quality to protect, and there's essentially no more imaging taking place. Each pixel measures the total integrated intensity of light in a pixel-sized chunk of the original image. This integration is by its very nature a "distortion" of that tiny portion of the original image. Whether that tiny portion of the image was formed by a narrow cone of light from a slow objective or a large cone of light from a fast objective has nothing to do with the ability of the individual pixel to properly integrate the signal level. What does matter, of course is that the field of view of the ~pixel~ is very narrow -- otherwise, it will pick up light intensity from adjacent pixels. > The reason they are transitioning to rear illumination is that the > definition of each pixel improves without this front hardware distorting > the light path. That's not correct. The reason one uses back-thinned, rear-illuminated CCDs is to get enhanced blue sensitivity. All that "front harware" really eats up blue and near UV signal intensity. AFAIK, the back-thinned arrays are only used for monochrome CCDs. Some of the more expensive astro imaging devices you've probably seen in Sky & Telescope use back-thinned arrays. > I'm also of the opinion that due to the differences between how the film > lies and how the array lies that lenses for digital cameras also needs a > flatter field than is necessary with film. Because the only cameras > where the film lies absolutely flat is ones with vacuum backs on them. > So there is likely a little more leeway in field curvature in film cameras > than in digital ones where the array is absolutely flat to some fraction > of a wave of sodium light. Interesting thought. I wouldn't hazard a guess on flatness of the CCD array -- perhaps it's flat to better than a wavelength, as you suggest. Certainly, film isn't flat to this extent, and I'd bet that a vacuum back wouldn't bring film into this regime of flatness either. However, I can't imagine that a lens designer would try to take into consideration the curvature of the film. I'd think that he/she would instead figure the necessary lens curvatures to produce a rigorously flat image plane. To do otherwise seems like it would involve a lot of very poorly characterized assumptions on film curvature. Even if film were held perfectly flat, the gelatin emulsion is certainly not an optically flat medium. Also, even if CCD arrays are made rigorously flat, the manufacturing tolerances for the body holding this array in position are certainly not going to be good to a precision of a few tens of nanometers. Thus, I'm not convinced that a designer would calculate curvature for lenses for a CCD camera any differently than curvature for lenses for a conventional film camera. It will be interesting to see what the article says that Jaume was looking for. Bill Peifer Rochester, NY - This message is from the Pentax-Discuss Mail List. To unsubscribe, go to http://www.pdml.net and follow the directions. Don't forget to visit the Pentax Users' Gallery at http://pug.komkon.org .
