I have done some macro lens experiments this weekend. As I've said earlier, my personal goal is not so much a more efficient way to scan, but to be able to digitize to get film transparencies into my digital workflow for fine art prints. Highest quality is thus of key importance.
I started off with a 24x36mm transparency. I decided quite quickly that 5Dmk2 resolution corresponding to ~3900 ppi at maximum mangnification (1:1) was a bit on the low side for prints that is enlarged so much that the film grain becomes visible. To get the analog look requires significant outresolving of the film, I changed to a Canon 7D, corresponding to 6000 ppi, although a bit softer pixels due to pushing the lens resolving power. I just for fun tested with a 2x teleconverter to get 12000 ppi, but I could not see any significant advantage.
Focusing is a challenge. The film is never perfectly flat (I did not wet mount it though), but I don't think the focal plane is perfectly flat either, even if this lens (a new Sigma 150mm/2.
is specifically designed towards having a flat focus plane. I could adjust precisely with a leveling head screws, but pulling one corner into focus would put another out of focus, so one have to settle with a compromise, all parts of the picture being somewhat in focus, but not perfect focus peaking over the whole surface. But this is at f/2.8. I tried lots of apertures to find the best DOF vs sharpness/diffraction compromise, and f/8 was the best. Slight diffraction onset (a little less microcontrast than for f/5.6 or f/4), but then enough DOF to bring the whole surface in focus. At f/8 getting the film in focus is not too hard. With perfectly mounted film f/5.6 would probably be the best choice.
Since one looks only at a 22x15 mm area at a time with the APS-C camera (stiching required!), I'm not too worried about film flatness for larger transparencies, at f/8 it only needs to be within 0.3mm oin that area, if you just verify and possibly readjust when moving the slide for the next area.
I did stiching in Hugin and it works well, so resolution-wise I think this works.
I was planning to post examples and stuff, but I got stuck in an unexpected area. Dark parts of transparencies are dense but still contain detail which means that some dynamic compression in post is often interesting, meaning that the capture must be made at a high dynamic range, which at least not a 7D or 5Dmk2 is up to -- shadows get too noisy when pushed for me to be satisfied. I only want to see film grain in the digitized picture, not camera noise. I knew this from earlier experiments, so I was going to do HDR.
When I do digital photography I always do my HDR manually, blend in a bright sky etc, rather than using HDR software, so I'm new to this. I tried many different HDR software (including what is builtin in Hugin of course), but not a single one produces satisfactory results. The problem is that the HDR programs does not properly understand digital camera exposures so they blend in blown highlights from the bright exposures instead of ignoring them, thus reducing the highlight quality (improving shadow quality works though, shadow noise problem disappears). The worst software reduce highlight quality by much, the better by less but all noticably and unacceptable to me. I was surprised that HDR software has not come farther in merging exposures, but I guess it is because HDR software is rarely used just for quality improvement, they are designed for "cool effects", not for fine art printmakers.
Without solving the HDR problem this method will not produce dynamic range compared to the better film scanners. Concerning resolution I think it is competetive, with APS-C camera probably better than 4000 ppi 135 film scanners, and most certainly better than 3200 ppi medium format film scanners. I have not made that side by side comparison yet though, I'm not really up for it until the HDR showstopper is solved somehow.