Drum Scan 4X5 to 200 MB or 300 MB?

I have been getting 300 MB drum scans of color 4X5 transparencies (Astia 100F, Velvia 100F) at West Coast Imaging. Recently I have seen some discussions about the amount of information in a 4X5 transparency that has me wondering if any detail would be lost by scanning to 200 MB (which is significantly cheaper). Is the extra resolution just scanning grain, or actually picking up more detail? The largest I print is 30X40.

Any thoughts appreciated.

--John

Scanning at 4000 dpi on my Howtek D4000 gives me a 900mb file @ 8-bit. When I scan Velvia and Astia @ 4000 dpi grain is just barely visible. I think that you should just work backward from the biggest print that you are likely going to make. 30x40 @ 300 dpi on West Coast's Chromira would require a 310mb file. But if you are interested in testing, have a 200mb scan done and get a 30x40 print done. A 200mb scan would give you a 30x40 @ 240 dpi. The Chromira will interpolate up to 300 dpi. The likely hood of you seeing the difference in the print is slim to none, in my opinion, especially at the viewing distance required for a 30x40 print.

John

I have my slides scanned by the Slideprinter in Denver for print sizes of 30x40. The detail and color is stunning, extremely sharp. File size is usually around 90 mb, not more. I would say even 200 is overkill. Scan and proofs run around $30.oo. They use Chromira as well. Hope this helps, Morey.

The image information in film is simply the grain clumps (or dye clouds), and the distribution of the grain clumps (dye clouds).

IMHO, there appear to me to be three major schools of thoughts on scanning film.

First, you should scan significantly "under" the film grain (say, 2000 ppi). The object is to pick up only the detail delivered by the lens. This school assumes that the lens is the limiting factor.

Second, you should scan under, but close to, the film grain size (say, 4000 ppi). The object is to capture some, but not all, of the grain in the scan. This school assumes that the lens/film/processsing is equally matched.

Third, you should scan at least at grain size (say, 6-8000 ppi, depending on the film's grain size). The object is to fully image the grain itself. This school assumes that the film is the limiting factor.

Which school you decide to follow (or make up your own ;-) seems to depend on how you work and what you believe. If, for example, you shoot a lot of 8x10 and really like DOF, you likely spend a lot of time down in f/64 land, and your images are diffraction limited. That would point you toward the first school.

If you shoot 4x5, and you really like sharpness, you likely spend a lot of time at f/16, and you are near the maximum the lens can deliver, which is less than small format, but above being diffraction limited. This points you toward school number 2.

If you are doing small format work and love getting the most you can out of the small film area you have to work with, and only shoot with the latest and greatest equipment, you likely have lenses that can give the film all it can possibly handle. You probably like small grain film, and shoot the slowest film you can, while still shooting at f/8 or above. This points you to the third school.

All that, to answer your question "Is the extra resolution just scanning grain, or actually picking up more detail?" The answer is probably - you are probably picking up more detail in the larger file.

The unasked question is: Is the extra detail worth it? The answer is probably not, if you are really never going above 30x40 inches. IOW, what Eric said.

I suspect that if you have WCI scan the same film back to back (same mount, same operator, all settings the same except scanner resolution) that you won't be able to tell the difference between your 300 MB file and your 200 MB file, because of the 11 micron minimum aperture limitation of the Tango. The Tango isn't the sharpest drum scanner on the block, despite it's hype.

Thanks for the responses. What is the "11 micron minimum aperture limitation of the Tango" and how does it affect sharpness? Are you suggesting that another drum scanner is better (not just sharper)? I don't know that much about drum scanners except what I have read.

--John

Hogarth,

I enjoyed reading your explanation. Very well done. Thank you.

Best regards,

I tried this once this morning and got interrupted and lost it. Sigh... As to the question of how aperture size effects sharpness...

When you drum scan you attach the film to a drum. The drum spins, creating one axis of motion. The lightsource and mirror assemblies travel longitudinally down the drum, creating the second axis of motion. This allows the scanner to scan one pixel at a time.

The light to scan by is made up of collumated light coming through an aperture that varies in size depending on what you ask the scanner to do (or operator override). A large part of the sharpness equation is the size of this spot. The reason has to do with light scattering in the emulsion, Callier effect, and all that messy real-world-laws-of-physics stuff. Basically, you want to match the spot size to the step size for optimal results in a less than 100% efficient system.

For example, let's say you ask the scanner to step in 6 micron steps. If your spot size is 11 microns (Tango lower limit IIRC), clearly when the scanner samples the spot, a percentage of what it sees comes from adjacent pixels. What do you think this is going to do to sharpness? OTOH, if you have a 6 micron spot size, and you are stepping at 6 microns, the scanner sees more or less exactly that pixel, and nothing much else.

If we are talking about a film grain clump, for example, if the grain is 5 microns in size and happens by the wildest coincidence to be in the center of the pixel, you'll see the entire grain in your pixel. If that same grain, in the same place, is illuminated by a 11 micron spot, the sides of the grain are illuminated by scattering off the film emulsion. This tends to make the grain clump look smaller. This is, in fact, one way for a drum scanner operator to combat grain aliasing - by manually selecting a larger aperture. But doing so by giving away some sharpness.

This just points out that the scanner resolution that you tell the scanner to use, is at best only loosely related to the scanner's optical resolution at the particular setting in question. You can only find optical resolution by scanning a resolution target, such as the 1951 USAF resolution target, and examing the resulting file. People have done some quasi-scientific research into this (nothing you could submit to a peer review journal, sadly). You can see some of it here:


http://www.scannerforum.com/

Go through the DIMA 2002 Scanner Roundup presentation. Way off down around slide 22? They show some surprising optical resolution results.

I read up on this stuff quite a while ago and decided to run my own experiment. I picked on a sharp 4x5 Tri-X negative, and scanned it three times (same mount) with all the same settings except for aperture (11x enlargement, which is something under 4000 ppi, IIRC, with no sharpening). I scanned at 6, 12.5, and 25 (my three smallest apertures). I isolated the same size chuck of each image in Photoshop (no sharpening again) and printed each of them straight, so I got about 1 square foot sections of what would have been a 44x55 inch print. I put these up on the wall and asked friends and neighbors which one they liked best without telling them anything about them.

Seven picked the 6 micron one because they seemed to think it was somewhat sharper. Three picked the 12.5 micron one because they thought it was somehow smoother. No one picked the 25 micron one; they seemed to think it was "mushy" IIRC.

They were all right, of course. You trade sharpness for smoothness, and it depends on what you want for a particular image.

So... clearly, YMMV.

Thanks for that contribution Hogarth. Very informative.

Cheers,

One other factor: it's been my experience that grain removal software works much better when the grain is clearly defined. If your grain is resolved at the same spatial frequency as your finest detail you're going to lose both to grain removal. But if the grain is resolved at a significantly higher spatial frequency it's easier for the software to discriminate.