Loris,
Your patience with me is very much appreciated. Having Attention Deficit Disorder but not being diagnosed until I was over 40 can sometimes create frustrating situations for me and other people. It might take two or three times repetition for me to understand but, rest assured, once I have learned I have a mind like a steel trap.
It is my understanding, here, that we are not necessarily talking about a particular "DPI" resolution to scan at but we are talking about some resolution to scan at which does not violate the Law of Diminishing Returns. In other words, we ask the question: "What is the point where my scanner will reveal no more detail in my image but will only produce larger and larger files.
Overall, we are asking what is the "State of the Art" in scanners which will produce a good image at a particular resolution.
In specific, I am asking, "How well should I expect my scanner to perform, given a frame of 35mm film under average conditions?"
I am a cinema technician by trade. I work with 35mm cinema film on a daily basis. Although the dimensions of the two versions of film are virtually identical, the emulsions might not be. However, the late John Pytlak, former Senior Technical Specialist for Eastman Kodak has said that a frame of 35mm cinema film can produce the equivalent of approximately 4000 lines of resolution per frame. A image frame on cinema film has different dimensions than photography film. Cinema film's image is .980 inches tall. It can produce the equivalent of 4096 pixels, vertically. That equals 4179 ppi. (4096 / .980 = 4179.59)
Number one: I know that cinema film emulsions and photography film emulsions are not necessarily the same. It is possible to use the same emulsions and they are sometimes used interchangeably but they are not always the same.
Number two: Developing of cinema film and photo film may differ. Again, they do not have to be different but they are not always the same.
Number three: Conditions under which the film is handled, exposed, processed and put to end-use may differ greatly.
Given those constraints, it is easy to see that it is not very likely that an average photographer will be able to achieve 4000 dots resolution from film. If we were talking about Mr. Pytlak, working in a proper laboratory with the full force of Eastman Kodak's resources behind him, I would expect this to be true but the "Average Joe" would be lucky to get that kind of performance. I, myself would feel lucky if I got half that resolution from film.
Now, with this premise in mind, I would expect 2400 dpi to be the maximum resolution that I could scan with my scanner before coming face-to-face with the Law of Diminishing Returns. I could scan at progressively higher resolutions... 300, 600, 1200 or more... and each time I would be able to reveal more and more detail in my image until, I hit that point at which no more detail would be revealed. At this point, I would be spinning my proverbial wheels in the sand.
It is your supposition that film produces detail of 10 m at the smallest level. I believe that to be true. It is also your supposition that the highest resolution that one can scan at which will reveal better detail, up to the limit of film's ability to produce detail, is approximately 2400 dpi. If I also understand you correctly, you are saying that, even if film was able to easily produce infinitely small detail, an average flatbed scanner could not produce a true representation of more than 1,200 or 2,400 pixels per inch.
I can see that to be true from your photos of the boat. The enlarged image, I can see where the pixel noise starts to blend with the grain noise in the film to produce a "speckled" effect in areas that should either show detail or should be continuous tone. (For example, the freeboard area of the boat, just below the outer gunwale.)
We haven't even covered the issue of "Zones of Confusion" whereby the resolving power of the lens may or may not produce sharp detail on the film, regardless of its resolution expressed in dpi. That even begs the question that the zones of confusion produced within my cheap, $200.00 scanner can even resolve the detail that is produced on film.
Now, I also propose that scanners, being producers of raster images in the Cartesian system can not truly reproduce a random-grain image as is produced on a piece of film. This is similar to what happens when we try to send our Photoshop images to a printer who will screen them for halftone: One pixel, or any number of pixels, do not necessarily correspond to one rosette in a halftone image. Neither will any number of pixels in an image correspond to a grain of silver (or halo of dye) in a film image.
What is the typical (although not necessarily correct) way to deal with images you know will be halftoned? Produce your image at double the screen resolution, edit then scale down to the final resolution for output. Correct? If this is true, why should we not do the same for scanning film but in reverse? Could we not scan at double what we think we need then scale down to the final output?
Yes, we would be going above the ideal resolution of film and we would be producing large files which do not contain 100% faithful reproductions of the original image, as it is laid out on the film but we have to mash film grains into a Cartesian pixel space or vice-versa. This task is going to take extra work.
I do not do this all the time. I believe you when you say the practical resolution at which one can scan is approximately 1,200 or 2,400 dpi, depending on the scanner. I follow your guidelines most of the time. But, when I come upon an "extra special" image that needs my full attention, I am not afraid to go as high as 2,400 or even 4,800. Yes, I know I am in that gray area when I am pushing the Law of Diminishing Returns to its limit but I have a fast computer with lots and lots of hard drive space and the latest operating system and the latest software.
I use a Mac Pro, 8-core computer which runs OS X 10.6 and Photoshop CS 4 and I have to 1.0 TB hard drives installed in addition to the boot drive. The scanner (Canoscan 8800F) will produce a full-frame scan of an Ektachrome slide at 2400 dpi in just a little more than a minute. The file size is approximately 22 MB. Raising the resolution to 4800 dpi produces an 85 MB image in two and a half minutes. These are limits I can live with. In fact, I would be comfortable with going to double that resolution but I know that would be silly. I truly would be spinning my wheels to do that.
I am not trying to be a pariah, here, but I feel it is important to explain myself in detail so that you understand me well. With my Attention Deficit, this is a difficult thing to do so I am trying to be as clear and detailed as I can. Then, once I have made myself understood, you can tell me where I have gone wrong. You are likely to have much more experience with photography than I have. I am only a technician for a movie theater.
The bottom line is that I do often follow your guidelines. I usually scan film at 1,200 or 2,400 dpi and leave it at that. However, on occasion, if I need to, I am not afraid to go as high as 4,800 dpi and I am confident that I am doing productive work even if I am treading the razor's edge of the line between productivity and diminishing returns.
