Scanning Velvia is a demonstration of the difference in scanning at different resolutions and with different types of scanners.  

The first two  scans were done with an Epson 1640SU.  The rest were from a commercial drum scan done at 12,000 ppi.  Note that not only the resolution improves with the drum scanner but also the dynamic range.  This is because the flat bed scanners uses a solid state sensor and the drum scanner uses a photo multiplier tube.                

 Velvia advertises a resolution of between 60 lpm (line pairs per millimeter) at low contrast and 160 lpm at high contrast.  It should be noted that at 1600 ppi scanner resolution, we are no where near getting all the detail off  the film.

 The scans shown in the presentation are a  small section of  the 4"X5" transparency shown above.  The crane in the presentation is on the far right of the photograph and about one mile from the camera. On the transparency, the word "Morrow" is 1/10th of a millimeter tall.  Sometimes people will assume that because they don't see additional detail in a slide or negative after increasing the scan resolution, above 2000ppi, that there is no more information on the film.  They will then use this to calculate artificially low digital/film comparisons.  I.E. 10 meg is equivalent to 35mm. The problem is what they are actually measuring is not the resolution of the film, but the resolution of the scanner.  Due to optics and mechanics, a 4000 ppi scanner is only capable of resolving about 55 lppm.  Any good photographer with a tripod and a decent lens should be able to get detail on most film which would measure 80 to 100 lppm.  This means that a good 35mm camera  is more like a 20 to 30 meg digital camera.

It should be noted that this image however, was not made with a 35mm camera but a 4x5 which most people say is not capable of producing as much detail in a localized area.  However, if you consider that the letters in the word "Morrow" are 1/10th of a millimeter tall and would take at least ten to twenty rows of pixels to resolve, you can see that it would take hundreds of megs of pixels to replicate the entire transparency.

 The photo was taken with a Schneider 360mm Tele-Xenar made in 1950.  It is a single coated lens which has a ding in the filter ring where it was dropped.  The camera was a 30 year old Linhof Technica .

If you have gotten this far, you just might be interested in an explanation of line pairs per millimeter.  In the past people measured lpm by photographing test targets (One theory is that the first crop circles were test targets for spy cameras.).  However, that gave you a measure of how much detail you had but not how clear the detail was so film and lens designers developed a new automated type of test called "Modulation Transfer Function" patterned after the same types of tests that they used to spec the frequency response of audio equipment.

Schneider optics explains it better than I could:

This is a comparison showing how much resolution you need to get all the information off of B&W film. 

In the photo the wording on the construction crane "Morrow" is less than 1/10th of a mm tall.  Significant research has been done on what resolution it takes to scan documents for archiving and still maintain legibility.  For 1mm characters the recommended minimum resolution is 300 ppi it follows that for 1/10mm it would require 3000 ppi.  Therefore it becomes quite easy once we know that film will resolve letters less than 1/10mm tall, to calculate the number of pixels it takes to duplicate a given format film.

This it a copy of a resolution chart from a Tmax negative.  The first picture is the resolution taken from a microscope photograph.  The second is a direct flat bed scan at 3200 dpi (max advertised optical resolution of Epson 1640su)  Note: you can not use a mid range flat bed scanner to evaluate film. (note some of the grain that you see in the microscope photograph is from the drug store processed color film that I used to make the photograph of the negative, it is not the Tmax grain.)

This link will take you to a Kodak web site that explains everything there is to know about Linear sensors and has a schematic that shows what all the areas in my picture do.  

Special thanks to Mike Collette at Better Light for supplying me with an old sensor to photograph.

The following photographs demonstrate the difference between CCD image sensors, film and photo diodes in liner array scanning backs. On the left is a microscope photograph that I took of the die clouds in slide film.  The photograph is from an area on the back of the woman's neck.  The photo next to it is of the CCD image sensor out of a video camera.  The pixels are 9 microns in size. Both photographs are at the same magnification.

The Second photograph is of a two rows of a three row linier sensor out of a digital scanning back for a 4x5 camera.  The square areas are the photo diodes and are also 9 microns in size.  However instead of using a Bayer matrix and interpolating color for each pixel, each area of the scene is measured three times, red blue and green.  Also the full 9 micron area senses light where the on the CCD sensor each pixel must do several tasks..

The    

This

T

The following links will take you to the FBI's Forensic Practices site which among the following: does not recommend the use of digital cameras for crime scene documentation, and equates 35mm cameras with 18 to 20 meg digital cameras.

Crime Scene Documentation..."Silver-based film cameras are recommended for use as the primary image capture device. The minimum recommendation is a 35 mm (SLR) camera capable of manual override, interchangeable lenses, off-camera flash, and tripod mount"