scanning with dslr? part II

[boris]

because of the complexity of the project i would like to split it in a few part and try to solve them step by step with the help of the forum.
first question RESOLUTION
i would suggest to start with a comparison of a 1 small piece of a negativ and leave all the other problems aside.
do somebody have a dlsr with a good macro lens AND a drumm-scan of the same negativ? i would ask this membre to make a picture with the dslr of a sharp fraction of the negativ (30x20mm and 60x40mm) and than compare the resolution with the drum scan.
if this first question is clear we could proceed to solve the others.
boris

[Nathan Potter]

Boris, I think a test is almost academic. With most any Dslr all you need is adequate magnification with a good macro lens to see all the detail available in a section of film. One of the issues is that the higher the magnification the smaller the field of view of the film so the greater the number of scans required. The optical resolution, at high enough magnification, will pretty much beat any existing scanner.

Think about the limit of optical microscopy. You'll resolve detail down to roughly a wavelength of light (say 0.5 um). Of course the field of view at such magnification will be small.

For your example above, say close to a full frame Dslr, (say 20 Mp) the pixel pitch is in the vicinity of 7 um. But to adequately sample a transparency you'd want to sample at least twice that pitch, say 3um at the transparency. So it sounds like the magnification is about 2X using a good macro lens. So field of view referenced to the transparency would be half of your sensor size or 15 X 10 mm. All rough estimates.
This kind of resolution would be equal to a good top of the line Aztek scanner. The dynamic range is another matter since the dynamic range of the photomultipliers one finds in high end drums are quite spectacular. Also one must remember that flatness of field and focus at wide copying apertures for the Dslr approach are pretty critical.
But the mechanical issues are solvable problems.

The example above assumes that the detail to be recovered from the film is on the order of the pixel pitch of the sensor, so is an extreme example. The reality of 4X5 film, for example is that seldom does one achieve distinct detail of less than say 20 um (25 lp/mm) so that a magnification factor of say 1:1 would usually be adequate.

Cheers

Nate Potter Austin TX.

[Ben Syverson]

Depending on the camera, I don't think you even need 1:1... The 5D Mark II has a resolution of about 4000 DPI on the sensor, so a 1:1 photo of 4x5 would correspond to a 4000 DPI scan.

You could get away with half that (2000 DPI or 0.5:1), which is my goal. That way you only need 3 x 3 images to capture the 4x5.

[VictoriaPerelet]

do somebody have a dlsr with a good macro lens AND a drumm-scan of the same negativ?

I updated older page that had comparison between Drum & Flatbed - Added DSLR Macro of same negative:

http://victoriasphoto.com/Notes/Drum_Flatbed/

[sanking]

Victoria,

Thanks for posting. From here it looks to me that DSLR Macro is better than the flatbed, but the big difference in contrast makes the comparison difficult to evaluate.

Sandy King

I updated older page that had comparison between Drum & Flatbed - Added DSLR Macro of same negative:

http://victoriasphoto.com/Notes/Drum_Flatbed/

[Nathan Potter]

Victoria, many thanks for the 1:1 macro comparison. About what I might expect given the many variables involved. The flattening of the sensitometric curve seems reasonable with the 1:1 macro and dominates the other effect of copying which is that of increased gamma. But there are additional opposing variables at play here (scattering from the backlighting technique, possible light scattering from stray front lighting etc.).

Boris this should give you some early hints of what is possible and suggests that the technique is not crazy.

Nate Potter, Austin TX.

[VictoriaPerelet]

Sandy, Nate. CCD flatbeds tend cut contrast severely because of very small CCD, lens & mirror system. CCD scanners with large optics aka Imacon are pretty much "DSLR" systems with line scanning CCD. PMT drum scanners have inherently wide Dmax with rather strong cut off on both sides - makes problem with C41.

PS. I used studio flash to light negative. Light table gives bad color cast (no good for E6 or C41), it can be adjusted with DSLR white balance but that will stretch effectiveness of camera sensor ....

[boris]

thank you victoria!
the exampel is quite impressiv. i would like to direct the attention on the aperture used in the comparison.
the original negativ was made with f: 64 and the dlsr shot was made with f:20 at 1:1 . this would equal a f stop of 40! at infinity. f 64 is alredy pretty small for the 8x10 inch format, but f 40 for a 15x22mm sensor is a real resolution killer.
if we presume a lf negativ shot with f 16 and the dlsr shot made with f 4 at 1:1 resolution SHOULD be much higher.

guys, i have to admitt that i know how to carry my 40 pound linhof gear on a bike
BUT have no clue of dslr, scanning technics, software and so on.
i will need some help to continue this thread to go on because of the lack of knowledge and language...
having said this i think that drum scan resolution is possible with ideal f-stop and appropriate technique. ( a volunteer to prove this assumption would be much appreciated.)
how to manage the weak contrast?

[Darin Boville]

Hey Victoria,

Cool tests.

I took the drum scan image and the DSLR image and did a simple levels/apply s-curve in photoshop.

The Drum has better tonality in this simple processing but it has a weird grain--it looks artificial in many parts of the image. Doesn't look like film grain. In addition the drum scan is noticeably blurry in the top half--the DSLR has much more detail.

I'm not criticizing just pointing out the differences--I think I prefer the DSLR version of the three.

Very interesting.

--Darin

[Ben Syverson]

I did some tests today too... I think the aperture you want to be at is more like f/8 or f/11... By f/20, you're quite diffraction limited (especially on a crop sensor).