If you're a hybrid film shooter like me, you are a) a masochist and b) probably insufferable company, lonesomely chewing your gin-soaked olives in the darkest corner of the cocktail lounge of life. Or at least, this is what I gather from any number of pointless debates between (presumably) old men on photo.net about the usefulness or resolving power of the Epson V750 flatbed scanner, or any flatbed scanner.
I'm going to independently add to this mound of increasingly irrelevant and arcane dirt with this blog entry, mainly to comfort myself with the knowledge that if my beloved and ancient and increasingly un-supported Microtek ArtixScan 120tf dedicated film-scanner were to bite the dust, goddess forbid, I could very probably very happily continue scanning at a similar (or perhaps even better!) quality using the V750.
First for some quick, poor-man's theory with numbers and science and things. Or not. Basically, I defer entirely to Norman Koren on just about any technical question relating to film, film scanning, and resolution, presented in a format almost comprehensible to the lay reader. However, my goal isn't to find the resolving power of the V750 using USAF resolution targets. What I want to do is show, with side-by-side examples of the same piece of film scanned on both machines, that the overall image quality and resolving power is equivalent between them.
Without any further ado, here are the results. This is from an image taken at Jacob Riis beach last summer on my Canonet QL17 II on Kodak New Portra 160. This is a reasonably sharp lens and an incredibly smooth-grained (if not incredibly sharp) film.
Before I tell you which is which (no peeking at the file names!), let's just examine the superficial qualitites. I quickly adjusted colors and sharpening on each so that they were as subjectively equivalent as possible. More on this later. In no particular order:
Alright. Which do you prefer!? Who cares. They're really quite similar and I'd be happy using either scan. I suppose I would say scanner "A" if pressed only because of the reds. However, this could really be a matter of color profiling. (I used different IT8 targets for each scanner, so maybe they'd come closer into line if I profiled each using exactly the same target. Hell, one might even be a LUT and the other a Matrix/Shaper... dunno. Color science is HARRRD.) It does also seem to have a bit more detail, but honestly it could just be more "texture" (grain? aliasing?) which doesn't really contribute to the overall sense of image quality. Honestly, the practical difference between these two is really neglible (EXCEPT for the amount of time it'd take me to spot/de-scratch these for printing!)
Alright, are you ready to know which is which? Scanner "A" is the Microtek 120tf at 4000dpi and Scanner "B" is the Epson V750 at 6500dpi, bicubically downsized to 4000dpi. Using the "Smart Sharpen" filter in Photoshop CS2 set to "Lens Blur", I gave the Microtek one pass using a radius of 1.0 at 125%. The Epson got a radius of 1.9 at 225%. So definitely, the Artixscan is subjectively about "twice" as sharp straight away, but this can be due to any number of things that doesn't necessarily mean that the Epson scans are less detailed. To wit, I did a very minimum of fussing with the focus height of the Epson: literally, I set the holder to the highest "+" setting, saw that it was resolving grain reasonably similarly to the Artixscan (which has an autofocus mechanism) and haven't bothered changing it since. So possibly, I could get the Epson even closer in raw quality to the Artixscan by fiddling a bit more with the focus.
(I suspect that the scanner glass must contribute in some way to a loss of contrast and sharpness. It would be cool if somebody adventurous and mechanically handy enough created a jig to allow scanning with the V750 with the glass removed.)
There's a lot of poo-pooing about scanner manufacturers "inflating" the DPIs of their scanners. Basically, people scan USAF resolution targets, determine a resolution of something like 2200 dpi and say "it is unnecessary to scan above this". However, if you read Norman Koren's website above, you'll note that "overscanning" plus sharpening has definite benefits, both in terms of detail as well as for avoiding aliasing, provided that there is detail beyond the resolving power of the lens. With regards to the latter, I would say that it makes sense for scanner manufacturers to outfit their systems with CCDs with approximately twice as much resolution as the resolving power of the lens for this reason, and to scan at this resolution if the limiting factor is indeed the lens (and not, say, the film/grain/lens on your camera).
I've already spent too much of the afternoon on this, so I'm going to wrap this up here. Leave comments if you have any questions or comments. Here's the final, full-frame.