Scanning technique on 4870

[Brian Ellis]

Ted says: "At 2400 ppi you are scanning at a higher resolution than the scanner may really be delivering. It is in the right ballpark but you might want to try a scan or two at say 1800 dpi."

paulr says: "One trick is to scan at the full 4800 ppi . . . If you just scan at 2400 ppi the software throws out every other pixel and every other scan line."

These two suggestions from two people who seem to know what they're talking about apparently are in direct conflict. Who's right here? Scanning with the Epson 4990 at 4800 ppi takes me close to a half hour from beginning (pushing the "start" button on the scanner) to end (having the photograph saved in a file) so scanning in any volume at that ppi is difficult. I'd pretty much have to devote an entire day just to scanning 16 sheets of 4x5 film. And that's using Vuescan with a gig of RAM and a 60 gig external hard drive. OTOH, throwing out every other pixel and every other scan line sounds horrible.

I don't know whether the apparent discrepancy between the two quoted suggestions bothers anyone else but does me. I'd appreciate some reconciliation of the two (i.e. an explanation of why they really aren't in conflict) or some further elaboration.

[Kirk Gittings]

Brian,

All of this is a little like asking what is the best film and developer combination. Both do different things well. In terms of sharpness, scanning at the true native resolution of the scanner is superior. For noise reduction though Pauls method produces less shadow noise. For my taste noise reduction is more valuable in moderate size prints, because any lack of sharpness is easily made up with proper sharpening in prints up to 16x20 but noise is a sticky problem, though I think the new "studio" multipass upgrade to SF AI does a better job than what Paul suggests.

For those that are interested, there can still be some registration issues on the SFAI Studio multipass on humid days as the neg heats up, dries out and stretches after the first pass. This is wwhat I do.
I do a batch scan with the same negative. The first scan is with just one pass to heat the negative up and the second is the 4x,8x or 16x scan. The neg is then preped for the real MP scan by the first single pass scan. It works very well.

[Ralph Barker]

Scott - after Silverfast does its preview scan, by default it sets black and white points that are often well inside the maximum output tonal range of 254 shades. My preference is to manually set those via the little triangles on the histogram slider to their maximum, and then set the black and white points in Photoshop. I think I get better tonal range in the scans that way. Also, in Photoshop, I snug the black point up to the beginning of the image data in the histogram (to get a true black in the scan), but set the white point slightly outside the image data. Doing so retains better detail in the highlights. I'm gathering that Photoshop's white point is white with no detail.

[neil poulsen]

There's a debate going on as to whether it makes sense to profile scanners. If you're getting good color to your eye without a profile, that you think correlates well with the scene, consider not using a profile. The point is, if you typically adjust color in Photoshop, whether or not you're using a profile, what's the point in profiling? Try it with and without and see if it makes for improved color.

I comment on this, because I was told that the IT8 target by Kodak isn't that good. (Gretag-Macbeth IGA, in a four-day color management workshop.) I don't have data on this, but they told us that the IT8 target wasn't even designed for profiling scanners. They recommended one of the Hutchcolor transparencies that comes with a custom reference file. This is a very expensive solution. But, it highlights the inadequacy of the Kodak target.

Scanners are capable of broad digital color gamut, kind of like digital cameras. As soon as one uses a profile, it limits the gamut to the boundaries of the target used in developing the profile. Again, you can try it both ways. If you don't see any improvement, don't use a profile.

Another related point. Consider converting to the ProPhoto working space when you first bring a scanned image into Photoshop. I had been using Adobe RGB 98. But at the advice of Seth Resnick, a well-known digital person who advises Adobe on Photoshop, I started using ProPhoto. I've always had trouble getting good browns from Adobe 98, which are important for landscapes. They always appear more megenta than brown. As soon as I began using ProPhoto, I got good browns. The ProPhoto working space has a larger gamut than Adobe 98, particularly in the red region. Some of this red region not included in the Adobe RGB 98 gamut will print on an Epson printer. This makes for better browns. And under any circumstances, never use sRGB for your working space, except for Internet images. It's gamut is limited, even when comparing to Adobe RGB 98. Unfortunately, sRGB is the Adobe default, at least it is for Photoshop 7.

Try different things and see what you think. While there's logic to all this, it really boils down to what looks and works the best for you for your paricular images.

[Ted Harris]

Scanning, especially with the 'consumer’ class Epson 3200/4870/4990 and the Microtek i900 scanners, is as much an art as a science in terms of trying to squeeze the best possible image out of the machinery. One of the things that I find more and more is the importance of using the best hardware/software combination and using it to the maximum advantage. in the case of the Epson 4870/4990 (or the Microtek i900) that will mean using the manufacturers supplied software in many instances. Silverfast is also a good. robust substitute. At the risk of starting a war, I have to question the use of VueScan. It is fine software if you want a quick and dirty adequate scan but it does not offer you the flexibility of the manufacturers software offered by either Epson or Microtek or of Silverfast Ai. All of these packages allow a wide range of manipulation but all have rather steep learning curves to master the manual settings. Time well spent if you are doing a lot of scanning and want the best possible results.

Remember that these scanners are the entry level for scanning film, not the ultimate. They do an adequate job when carefully used but there are issues. Focal point of the lens has been mentioned and inaccuracy of the stepping motor alluded to. Remember too that they are relatively light machines so even a slight vibration from heavy footsteps or a light bump of the table can have a noticeable impact. I am not necessarily suggesting you get a better scanner (although you can do a lot better for slightly less than twice as much money) just that you use these machines very carefully.

Brian, just a note that the scans you are talking about would take well over an hour each on an Optronics Colorgetter or similar drum scanner. The limiting factor is the speed of the scanning motor not the computer in this instance.

[paulr]

"Ted says: "At 2400 ppi you are scanning at a higher resolution than the scanner may really be delivering. It is in the right ballpark but you might want to try a scan or two at say 1800 dpi."
paulr says: "One trick is to scan at the full 4800 ppi . . . If you just scan at 2400 ppi the software throws out every other pixel and every other scan line."

These two suggestions from two people who seem to know what they're talking about apparently are in direct conflict. Who's right here?"

I don't think Ted and I are in conflict. Ted is pointing out that the actual optical resolving power of the scanner may be below 2400 ppi. Most people who have tested the scanner would agree.

My comments added a couple of points: one, if you maximize film flatness and find the exact focal plane, you'll possibly come very close to 2400ppi of useful optical resolution. And two, there's a reason besides detail resolution to scan at a higher sampling frequency--noise reduction. If you don't see the difference between downsampling from 4800, then it's not worth the extra time. But you may find that it helps.

And as someone already pointed out, if you're finding high res scans take you an hour ... well, welcome to high res scanning.

[Brian Ellis]

Kirk, paul and Ted - Thanks for the responses, I think I get it.

[Scott Rosenberg]

ralph, thanks for the info. i'm between scanners right now, but will likely pick something up in the next couple of weeks, as my pile of stuff to be scanned is getting out of control. i appreciate the tips and will be trying them very soon.

wonderful thread all around,

scott