Collimating light for more sharpness

Digital Camera Scanning
21

Hi Burkphoto,

Thanks for sharing your experience with the professional Kodak world. It’s striking to see that consumer digital cameras have become so good that they seem to render these professional machines obsolete. I don’t have any experience with professional software such as developed by Kodak, but was blown away by what NLP achieves with just a few mouse clicks. Considering the vast amount of film strips that most of us will have, time is everything when scanning.

My experience with scanning was far from professional: I started out in the early days of digital cameras with attempting to scan with a Nikon slide holder pointed at daylight, a lens, some macro tubes, and an early digital camera, but although the scans seemed okay-ish, it was very time consuming to get anywhere near good color and contrast. I then stumbled across a Minolta San Dual and Vuescan software that gave what looked like proper color out of the box. I then spend quite some long evenings scanning, but in the end gave up because it simply took too much time. In the end I sent everything in batches to Scancafe.com who had a good formula and were affordable. I got offered a Minolta Dimage Scan Multi ii for free at a garage sale that spent years in my closet. Now I’m back to film and tried it, but was again shocked by the time it takes to scan, but this time also by the much, much higher quality from a simple DSLR scan. So the circle is round, and I am back where I started, but mostly because I discovered NLP. Overall, if a scan takes longer than a few seconds to take a picture and a few mouse clicks to process the result, it’s unrealistic for me to consider re-processing my old films, simply because of time constraints. It seems that a smart DSLR setup (whatever shape that may have) plus NLP will enable sufficient short time per scan.

Regarding (tungsten) light sources: Elsewhere on these forums, I stumbled across Solux 4700K daylight sources, which seem to be quite popular, also in museums etcetera. Any thoughts on these?

Perhaps one of the advantages of going with an enlarger head as a light source (either collimated or diffused) is that it’s relatively easy to change the bulb into whatever seems best as a continuous spectrum bulb. You also mention color shift over time with a tungsten bulb, but doesn’t NLP take care of that when the white point is taken off the negative edge? (Not sure of that though)

22

Hi DavidS,

What you write seems to confirm what Digitizer was saying: A collimated light source has no advantages other than picking up more unwanted crud from the negative, and thus making it harder to get clean scans.

I’m not sure I understand why though, nor what optically exactly is going on, other than the empirical comparison with portrait shooting. Need to think about that some more…

A few days ago I picked up the old enlarger that I purchased for a few bucks. Now I need to find some time to set that up and play with that. Will report back what I find.

Cheers,

Arno

23

Last time I looked, Solux lamps are no longer available. They were extremely hot, too.

NLP (actually the eye dropper custom WB in LrC) does eliminate color shift. In optical analog systems, or when multiple tungsten devices are used, it’s a big issue.

I’m writing a full description of my setup with illustrations and will post next week.

24

Hi ArnoG,
Just to give you a bigger picture of what I am doing, I have NLP scanned about 10,000 of about 30,000 color negatives accumulated over a lifetime of photography.

About 20 years ago I purchased a Nikon Coolscan 4000ED for the same project but gave up because it was too slow. Recently I dusted it off and rescanned some NLP scanned negatives (with VueScan software). Wow! the Nikon has far greater control of color and far more information. So I purchased a non-working Coolscan to take apart and study. It has separate RGB and infrared LED illumination, condenser optics, a monochromatic line sensor, auto-focus for each exposure, a special 7 element lens, and Digital Ice for dust and scratches. So my scanning project is on hold while I try to duplicate the Nikon’s quality with NLP and my Sony cameras.

After my experiment with the Omega enlarger condensers I am convinced that condensers were not chosen for image quality. Instead they efficiently used the dim light from the early technology LEDs.

BTW, the biggest happy surprise so far has been the camera objective. I removed the Nikon lens from the broken Coolscan and use it now with the digital camera for NLP (replacing a Leitz Focotar enlarging lens). Now detail in my NLP scans is close that from the Nikon scanner-----with a diffuse light source.

I suspect that I could achieve the full detail of the Nikon scan with more resolution. The Nikon scanner lens only covers the 24 Mpx cropped sensor NEX-7, not full frame. On the brighter side, NEX-7 raw files are much smaller than those from the scanner.

25

@burkphoto: There’s still detailed data on solux.net on the 4700K (https://www.solux.net/cgi-bin/tlistore/infopages/4700k.html), but indeed so far I didn’t find a place to buy them except through alieexpress, which I’m reluctant to use…

So far my few experiments with DSLR scanning have been limited to B&W but I might go back and redo my old film stock, which is mostly color, so it’s good to know that WB can be taken care off easily.

Can’t wait to see your setup!

26

@davidS : this is great stuff! You realize that you might trigger a market for broken Nikon scanners this way? :wink:

Based on what we (I) learned so far, perhaps Nikon is using collimated light for IR scratch removal, and diffuse light for the actual scan?

I thought a bit more about the optics of collimated light showing up more unwanted crud and think it does make sense: If one imagines taking a picture of the emulsion side, i.e. with the emulsion side “up”, and a negative having a certain thickness below that, then with collimated light any scratches on the bottom will be clearly projected/shadowed on the top (emulsion) side. With a diffuse source, where light will enter not only perpendicular to the “bottom” but from all sides, these projections/shadows will be much less pronounced. One would still pickup scratches and dust on the emulsion side though, and I can’t yet reason whether these would be less pronounced with diffuse light.

At the same time, if one focuses on the emulsion side, and the emulsion side is “up”, then there should not be loss of sharpness due to the fact that the light is diffuse, in the same way that something is sharp when simply looking at it in diffuse light. Intuitively, I would think that scanning with the emulsion side up should always be better, since the light that contains the details of the image doesn’t have to travel through the film with all the scattering and loss of detail that could result from that, and instead can travel freely into the scanning lens. Does that make sense?

Finally, one could imagine that a limited depth of field would be ideal, so as to only have the thickness of the emulsion being sharp, but in practice this is probably not realistic due to most lenses not having optimal sharpness at their widest apertures.

All in all, taking into account experiences from others as written here, the Callier effect, and picking up as little scratches as possible, it seems to me that diffuse light should give a technically “best” scan, whereas collimated light could be used for “art effects” such as grittiness in B&W.

(?)

27

ArnoG, there already is an active market for broken Nikon scanners. This guy helped to create it with his lens testing.

Actually the Nikon scanner uses condenser illumination for both scratch/dust elimination and the actual scan. It does not use diffuse lighting at all. It has one monochromatic line sensor that moves down the filmstrip 4 times scanning the film 4 times. With each scan a different LED is lighted; red, green, blue, and IR. Your computer combines these separate scans into one RGB file.

I am doing my scans with the Sony NEX-7 camera, the Nikon scanner lens and diffuse light.

28

I hope this worked. I attached a PDF. See link below.

29
1 Like
30

Bravo, @burkphoto :raised_hands: - this is a wonderful guide! :pray:

31

Thanks! I love NLP. This combination of tools certainly isn’t the only one, but it works extremely well for me. I think camera scanning can put a “world of hurt” on the scanner business.

32

Can someone repost a Burkphoto guide? Unfortunately the file is deleted from the server…

33

Paging @burkphoto :slight_smile:
Looks like your guide link is no longer up on dropbox.

34

I made minor edits.

1 Like
35

Thank you! And such a beautiful layout, nice to read!

As an owner of some Creo high end scanners (and a drum one), and based in Saint-Petersburg, Russia i often feel myself, when scanning a big batches, in a kinda sweat shop in an emergency nation )))

To get back on the topic - seen here an interesting opinion, that the scratches and dust on the base side on the film in the collimating light get projected thru, and seen on the image alongside with the all the dirt from the emulsion side. So. Drum scanners use a collimated light source and i do not seen this effect on “projecting” when scanning on the drum scanner. Maybe this is because originals are wet mounted. Meanwhile. I do see a scattering and flaring on the various edges in the image, when scanning on the flatbed, specifically from diffused light. Thats why flatbed scanning always need a sharpening.

36

Even better if I may say so, is the Minolta Dimage 5400 lens. It’s sharper than anything short of a Printing Nikon and behaves like a superchromat. I have this lens mounted with adapters to a Nikon PB-6 bellows and slide copier.

1 Like
37

Ouch Belinda! I thought that would be true. Do you know if the difference between the Coolscan and the Dimage lenses would be noticeable in color negative scans? I know they don’t really cover full frame so a 24Mpx cropped sensor is pretty much a resolution limit.

The Nikon Coolscan 4000ED lens significantly improved my scans over the same body (NEX-7) with my 50mm Leitz Focotar enlarging lens.

BTW, I also found that vibration was hurting me even more than optics.

Do you know if buying a good macro for my 42Mpx Sony A7R2 would be even better-- or not worth it?

39

Hi David

Sorry, I didn’t mean to rain on your parade!

The Damage 5400 is a beast. It’s not much to look (it’s tiny and entirely without glamour) but it beats almost anything else short of super expensive exotics like the Printing Nikkors. I got mine on ebay for £269 while a Priinting Nikkor costs in the region of £5000. I have attached a number of links for you which will give you lots of information on both your lens and a host of others. You could be reading for weeks! I did.

You may be wrong about the coverage of your 4000ED lens. As far as I know (look at the test) it is designed to cover an FX sensor - as is the Minolta 5400. The 5400’s resolution is so high that it can cope with a 42MP sensor and I’m guessing but the 4000ED is so sharp it can probably work too. I’m happy to hear your view and be corrected as my knowledge on image circles, resolution and sensors size is shaky. On paper the 5400 is quite a bit better than the 4000ED but whether that would translate into a visible difference in real world conditions is debatable. I wouldn’t change your lens - it’s top notch.

For scanning I don’t think it makes sense to use a macro lens or bellows lens for that matter. Scanner lenses are superior in every way - sharper, flatter and with much better CA control but of course they are limited to this one task at a narrow range of magnification. If you want a macro lens for more general use, look at some of the Sigmas on Robert O’Toole’s site. They seem to perform way beyond their price.

I avoid shake by using a Nikon PB-6 bellows and Slide Copy attachment plus flash but otherwise would use a release delay plus mirror locked up and maybe self timer as well.

Hope that helps

https://www.closeuphotography.com/2x-lens-test

https://www.closeuphotography.com/minolta-dimage-scan-elite-5400-lens/

http://www.savazzi.net/photography/scanner_nikkor_40mm.html

https://www.closeuphotography.com/3x-lens-test

https://www.closeuphotography.com/scanner-nikkor-ed-7-element-lens

2 Likes
40

Hi all,

I kinda settled on diffused light as a preferred source, but now this new discussion on lens resolutions made me wonder whether it made sense to take the scanner lens out of my minolta scan multi ii instead of selling it, and more general, what lens (and camera) resolution would be sufficient (for me) and whether what I have is enough. So I started rambling around on the apples and pears comparisons of lines per mm (Units and conversions for resolution) and DPI’s (How to convert pixel size to line pairs per millimeter? | OptoWiki Knowledge Base) and wanted share what I found to see if that made sense.

It seems that consumer films have around 35-50 lp/mm (line pairs per mm) while Kodak introduced at some point higher resolution T-Max at around 100 lp/mm and above (Resolution and MTF curves in film and lenses).

My camera (Nikon Df) has an image resolution of 4928x3280 and a pixel pitch of 7.28 micrometer. What I read is that these pixels are individual R, G, and B, but I assume Nikon does some fancy stuff to retain at least that pixel pitch resolution while doing the color calc. From what I read, 7.28 micrometer would translate to something like 500/(7.28x1.4)= 49 lp/mm for B&W and 500/(7.28x2)= 34 lp/mm for color (the 500 stems from 1000 micrometer in a mm, with a factor 2 for line pairs: one line and one blank). Elsewhere I found statements that lp/mm roughly equals DPI/100, which for my Df would translate to 3280/24x25.4 /100 = 35 lp/mm, which matches the color calc before. Yes, I read that DPI’s cannot simply be translated to optical resolutions, and that the reason lens resolution is spec’d as it is in in MTF charts is to make the lens tests independent on sensor pixel count, but alas, I am only looking for order of magnitudes and trying to be pragmatic on what I really need.

Hence, would I want a higher resolution camera with a smaller pixel pitch? Yes, perhaps to get an ultimate sharpness from my scans, but whether it makes sense (for me) is a different question. Back in film days, I did not use fancy film, nor could I afford fancy high resolution lenses, nor was(is…) my technique not likely anywhere good enough to achieve laboratory resolution in real-world photography. Hence, for me at least, my camera should be sufficiently good for the resolutions that will be present on my old films.

With respect to lens options, I have a Zeiss 100mm Makro Planar, a Nikon EL 50mm 2.8 enlarger lens that I often use for macro, and the minolta scanner lens, and a blob of other lenses that can be used with extension tubes and bellows. So what gives me the 35 lp/mm that my camera can achieve, and what should roughly be present on my old films?

The Zeiss is very good (Zeiss Makro-Planar T* 100mm f/2 ZF (FX) - Review / Test Report - Analysis) and virtually without errors when used around f/5.6 with about 4000 lw/ph (line width per picture height) center and “planar” with little field curvature (corner sharpness around 3500 lw/ph). A MF negative is 55 mm square, which the Zeiss (1:2 max enlargement) achieves without extension tubes, so for MF 4000 lw/ph would translate to 4000/110= 36 lp/mm, so comparable to what my camera can achieve, i.e good enough. For 35 mm I need to push the lens out by 50 mm, effectively enlarging the projected image by roughly a factor two and thereby halving the resolution: Now 4000/2 = 2000 lw/ph. Since the picture “height” for 35 mm is now the width, i.e. 36 mm, this gives 2000/72 = 28 lp/mm, which is less at 1:1 compared to 1:2 due to the extension, but should be still okay for my purpose.

The Nikon EL 50mm 2.8 I have is lower resolution than its newer “N” version. The latter measures 2500 lw/ph when used pushed out to 1:1 if I interpret the charts in El-Nikkor 50mm f/2.8N lens test correctly, yielding 2500/72 = 35 lp/mm, but I have the lower resolution “pre-N” version, and the lens, despite being an enlarger lens, does not at all seem to have a flat field, loosing ~50% of resolution in the corners. Hence, for me, the Zeiss is better since I do need a flat field for negative scanning.

Minolta specifies the optical resolution of the scan multi ii lens in their manual at 2820 dpi for 35 mm and 1128 dpi for MF. Dividing that by 100 to get (roughly) to lp/mm, it is not really worth to pull it and try it.

Overall, with the Zeiss I would get close enough to my camera resolution and what I can realistically expect to be present on my old films. Would a 4000ED lens or Minolta 5400 lens in combination with a smaller pixel pitch camera yield higher resolution scans? Sure, but whether I would get indeed sharper pictures from my old film stock is doubtful. Hence, my conclusion from this exercise is that what I have, and can presently build together, should be sufficient for my purpose, and it doesn’t make sense (for me, and at least for now) to upgrade my camera or lens solely to scan my old negatives. Whether that remains true now that I am shooting film again remains to be seen though…

Does this make sense or am I wrong in calculating or missing some key things?

Cheers,

Arno

41

Let me try to answer this question by another few questions:

  1. How large are you going to print?
  2. Will you look at the print at normal viewing distance or use a loupe to count pixels?
  3. If you don’t print, what can your screen do?

In physics, there’s always a theory and a bunch of experiments to prove the theory right or wrong.

  • now that you have established a ranking of your gear
  • and considering that you have all possible ways to set up a rig
  • why not put the most promising things together, run a few tests and see what comes off it?

A typical RAW file contains a preview (maybe also a thumbnail) and raw image data look like this before demosaicing and colour correction (a few things have been done below so that we can see things):

Bildschirmfoto 2021-08-09 um 14.35.22
Bildschirmfoto 2021-08-09 um 14.35.222574×2170 568 KB

The"fancy stuff" takes place when the camera creates the built-in preview, creates a jpeg file or when the image is “developed” in Lightroom or any other app, than converts raw image files. A 24 Megapixel sensor records a 12 Megapixel green image and a 6 Megapixel red and blue, shifted, image each. This means that the fancy stuff has to create 12 missing green Megapixels as well as 18 red and blue Megapixels each. Lightroom and others strive to best in “inventing” the missing pixels based on surrounding existing pixels. No matter how this is done, errors are introduced, not visible as individual pixel errors, but as a general loss of perceived sharpness.