Need feedback for building DSLR scanner

[p6889k]

Hi,

I'm building me a DSLR scanning rig, I made some progress, but still have a lot of questions and thought some of you experienced guys could give me some pointers:

What I have today:

For 35mm:
I use dedicated Nikon Coolscan V film scanner - works great.

For medium format and large format I have now assembled the following:
10.5” iPad Pro as light source, placed inside a custom wooden box that’s 3” tall, top of the box has 8.75”x 10.75” x 3mm ANR glass from Focal Point. The glass sits about 3” from the ipad screen so that the iPads pixels are diffused and don’t show up in photos. Above it is Canon 5IV with Zeiss 100/2 MP lens on a stable tripod. The lens has 1:2 magnification, I’m using one 25mm extension tube to get higher magnification, need to get one more to get 1:1 magnification. Black heavy cloth wrapped around the structure to block light. I place film directly on the ANR glass if the film is perfectly flat, if it’s not I put in film holders or tape it.
Have only used it little bit to compare 35mm positive scans to my Nikon Coolscan and so far results are very very good. The major differences I see is that DSLR scans are coming out with more saturated colors, but otherwise looks the same.

Now I’m thinking about what’s next, so here are the questions:

1) How should I go about profiling the camera to get correct scan colors. For positives, should I purchase IT calibration targets? Which ones would you recommend? I see there are targets for slides/positive film, but don’t see anything for color negative – how would I go about that?

2) If I was to change my light source, I’m thinking about:
a. 99CRI non flickering LEDs from https://store.waveformlighting.com/p...=8190565711974
Have you heard anything about these? The specs look great, high CRIs, tightly spaced.
b. What material would you recommend as a diffusion layer above the LEDs. For the iPad all I’m using is clear ANR glass and it works fine as the iPad pixels are very tight. For LEDs, I’m thinking I will need something else below the ANR glass to diffuse the LEDs. I have no idea where to source 8x11 high quality diffusing glass or plastic. Can you recommend something.
c. The LEDs will be significantly brighter than iPad. How should I regulate light output – by exposure in camera or by regulating the LEDs themselves? The LED manufacturer sells non-flickering matching LED dimmer.
d. How closely should I place LED strips next to each other? Should they be literally touching or is it ok to have say 1” spacing.

3) How do you ensure that you lens is perfectly parallel with the film? Should I just print out some square targets and look through the LCD to see if all the lines are perfectly aligned to the edges of the screen or crop marks on the screen?

For now I’m mostly playing with positives, didn’t tackle negatives and understand that will require additional postprocessing to invert and correct colors.

Thank you,

Pavel

[Ted Baker]

but don’t see anything for color negative – how would I go about that?

No such thing for color negative film, you can use anything you like to calibrate to the colour you like best, for example a color checker chart, but there is nothing that is non-proprietary that is specially made for calibration like a IT8 target. Complex topic.

3) How do you ensure that you lens is perfectly parallel with the film? Should I just print out some square targets and look through the LCD to see if all the lines are perfectly aligned to the edges of the screen or crop marks on the screen?

A mirror where the film sits, adjust the camera so the lens is in the centre seems to work well.

[Peter De Smidt]

There are a number if threads on this in the DIY section. Waveform leds look good. Are you trying to match colors ir get good-looking color?

[Larry Gebhardt]

1) How should I go about profiling the camera to get correct scan colors. For positives, should I purchase IT calibration targets? Which ones would you recommend? I see there are targets for slides/positive film, but don’t see anything for color negative – how would I go about that?
- I shoot raw and for positives I adjust just like any other raw image shot shot on the digital camera. I haven't tried profiling, and I'm not sure there's much advantage unless an exact copy of the slide is desired. For negs there isn't a target and I've seen no support in any color management package for profiling negs.

2) If I was to change my light source, I’m thinking about:
a. 99CRI non flickering LEDs from https://store.waveformlighting.com/p...=8190565711974
Have you heard anything about these? The specs look great, high CRIs, tightly spaced.
- I haven't seen those LEDs before, but they have the best CRI rating of any I've come across.

b. What material would you recommend as a diffusion layer above the LEDs. For the iPad all I’m using is clear ANR glass and it works fine as the iPad pixels are very tight. For LEDs, I’m thinking I will need something else below the ANR glass to diffuse the LEDs. I have no idea where to source 8x11 high quality diffusing glass or plastic. Can you recommend something.
- Last product I used was Makrolon GP (https://www.alro.com/divplastics/Pla...akrolonGP.aspx). But there are now a lot more diffusion products on the market. Search light diffusion acrylic

c. The LEDs will be significantly brighter than iPad. How should I regulate light output – by exposure in camera or by regulating the LEDs themselves? The LED manufacturer sells non-flickering matching LED dimmer.
- You will want a non flickering dimmer. Most PWM dimmers won't work with quick shutter speeds. You really don't need LEDs for this. In many ways good old incandescent with a filter over the lens to daylight balance would be better, and cheaper. Also consider a used color enlarger head. You can adjust the light source to make color neg scanning easier and with less noise in the blue highlights.

d. How closely should I place LED strips next to each other? Should they be literally touching or is it ok to have say 1” spacing.
- That will depend on how close they are to your diffusion plastic as well as the type and thickness of the diffusion sheet. Just photograph the sheet and ensure it's even with no film.
- You could also consider edge lighting with a product like https://www.acrylite.co/led-light-gu...-sample-1.html. I'm experimenting with it now for revamp of my enlarger head. Seems easier to get even lighting than backlit.

3) How do you ensure that you lens is perfectly parallel with the film? Should I just print out some square targets and look through the LCD to see if all the lines are perfectly aligned to the edges of the screen or crop marks on the screen?
- either a laser alignment setup, or photograph the reflection of the camera as Ted said. I use a 3d printed stand with a laser sight on it for my enlarger and things like this.

[Peter De Smidt]

Makrolon works well. I tested some. I just use regular 1/4" thick white Plexiglass, though. If you look at the light source section, you'll see my LED source. Ideally, you'll have at least 2" between the leds and the diffuser, and that's with tightly packed leds. A second intermediary diffuser didn't help in my case. Make sure the light source is about 1" bigger in each direction than the largest negative, or you will have light fall-off at the edges.

Back to the whole idea. You mention scanning at 1x magnification. That will require approximately 25 frames per 4x5" negatives. That's a lot of frames to take manually, and without repeatability and a stitching template, you will run into issues on some negatives. I did 1x because the camera I was using at the time, a D200, needed that level to get to a true 3000spi level. If memory serves, that was a 10mp camera. For a manual system, I'd use a much lower magnification with a current camera. People regularl
overestimate how much resolution they need. Dynamic range, preservation of contrast, and other image parameters can be more important. In my experience, a properly build digital camera system, one that includes very good masking of stray light, does a significantly better job in those other respects than, say, an Epson flatbed.

For color negatives many people use Color Perfect. I don't scan much of that, and so I don't use it.

I've tried a whole bunch of camera profiling systems over the years. I spent a couple of years making my living matching photos to material samples for products in the 10s of millions of dollars. (Gulfstream jets). None of the profiling systems really worked well enough to use. I could never get away from making manual adjustments.

Mirror lock up, shutter speeds outside of the 2second to 1/15th range, where shutter shake can impact the system, all impact the results, with greater impact at higher magnifications.

Here's a link to Daniel's scanner in action: https://www.youtube.com/watch?v=PXy7RJwIBAo&t=47s

[rdeloe]

Some recent thoughts on a camera scanning setup and a design you could try (easy to build): https://www.largeformatphotography.i...ample-approach

For lighting, I tried multiple lights but the coverage was uneven. Remember you're only shooting through a tiny window because you want to mask the rest of the area -- so you don't need a large light source. The LED tablet I'm using (see link) doubles as my light table; I'm only using a small area to illuminate the part of the 4x5 negative that I'm shooting. The light source could be a lot smaller.

I use the approach Ted and Larry describe for getting everything parallel (positioning the reflection of the mirror). It works better than anything else I've tried.

One important thing you need to decide is how much resolution you want. I settled on 2,666 ppi. To get that I shoot 12 overlapping frames each 2.25" x 1.5". The camera has a resolution of 6000x4000 pixels, which is how I end up with 2,666 ppi. With my setup I didn't see an improvement in image quality by going to 3,000 ppi; more pixels doesn't necessarily mean more detail.

You also need to watch out for stitching problems. Peter's cool semi-automatic system has the precision needed to allow reliable stitching with templates in PTGui. I was testing last night with my manual system and I can use templates in PTGui to stitch the image, but I can't (yet) quite get the precision needed. As a result, I'm seeing some misalignments in places in the pictures stitched with the template; I think there's room for improvement but have to see. Note that this issue only comes up when your stitching software can't handle the scene (e.g., if a large area in your negative is continuous tone or mostly undifferentiated). Nearly everything I've tried that has some details in all parts of the frame stitched perfectly in Lightroom.

[Peter De Smidt]

One thing to be careful of with a small light source, is that if there is any unevenness, this can lead to a pattern in the final scan. To avoid that, I chose to make a light source that illuminates the whole negative as evenly as possible. As long as your source is very even, I'm sure that a stationary small source would work ok. Light has to hit each part of the negative from all directions from bellow. Other wise it will do so in the center of the illuminated area but there will be less light at the edges.

[rdeloe]

Peter makes a good point above: "People regularly overestimate how much resolution they need." For my purposes I've landed on 2,666 ppi as a good balance between somewhat future-proof scans and usability. However, I was curious to see whether or not more resolution produced more actual information. To that end, I made a 7,917 ppi "scan" and compared it to my standard 2,666 ppi scans. In this picture, 7,917 ppi is on the left, and 2,666 ppi is on the right. These are pictures of the spines of Compact Disc cases. The full height of the CD case on the negative is 11.1% of the long side of the negative.



Even in this tiny comparison image it's possible to see that, yes, the higher resolution picture resolves more detail. In other words, it's not just adding pixels with no gain in image quality. Importantly though, all the text that's readable in the higher resolution image is also readable (but softer) in the lower resolution image. So my "low" res scans are still really decent.

The largest size I can currently print on my Epson 3880 is 17" wide (or more realistically, 16" wide to leave a bit of margin). At 2,666 ppi, I'm already getting 667 ppi on a print that's 16" on the short edge from my 2,666 ppi scans. On the matte paper I use I find it very difficult to distinguish a 720 ppi print from a 360 ppi print -- so my "low" resolution 2,666 ppi camera scans are already overkill for my needs. My computer would grind to a halt processing a 7,917 ppi scanned file. I haven't calculated how many frames I'd have to shoot to make a 7,917 ppi scan, but it would be in the neighbourhood of 80. Now someone printing pictures that are 44" on the short edge would enjoy 720 ppi from the high resolution 7,917 ppi camera scan I can currently make, but that's not me.

EDIT: After posting this I made a 4,000 ppi camera scan of the same area. The 4,000 ppi is better than the 2,666 ppi, and the 7,917 ppi scan is better than the 4,000 ppi scan. However, it's quite likely that the real maximum resolution before it's just adding pixels rather than image quality is less than 7,917 ppi and more than 4,000 ppi. It would take a lot more testing, ideally with a USAF 1951 resolution test chart, to confirm.

[Henry Ambrose]

1. I don’t think you will be able to profile the camera. I’d set it to a daylight setting and provide it with as close to daylight as possible.

A deeper light box with multiple layers of diffusion will get you the most even light across the frame. Its easy to test by making a picture of the blank light source without film.

Adjust the light intensity so that you are making exposures at the best aperture for your lens over the whole frame. Find the best combination of sharpness and even illumination. 5.6 or 8 might be fine, but refer to the lens maker’s MTF diagrams if you have them or can find them easily. Use more diffusion in the light box or a neutral density filter on the lens to get there. More diffusion is probably cheaper, easier and better.

2. Look at the spectral distribution of any light source. CRI does not tell the whole story. Ideally, you want a spectral distribution that mimics daylight. Many manufacturers supply this information, but you have to know to look for it.

3. You will probably be square enough when the four corners are all sharply in focus. If you want to test a set up, make a tiny scratch in the corners of a piece of film and inspect the capture in PS. If the four tiny scratchs are all sharp its good. A copy stand for the camera might be a good investment to make set up easier and faster.

[p6889k]

Thank you everyone for your feedback.

Nice advice for using mirror to align the lens.

Regarding Makrolon for diffusion...What thickness do you recommend? I also see that they have a choice of "paper mask" or "film mask" - do you know what the difference is? They also have color options: clear and grey - I have no idea how clear the clear one is - which one do you use?

The LED lights I'm looking at have an option for 5000K or 6500K color balance, I'm going to go with 5000K, thoughts?

Eventually I want to build out motorized unit just like in Daniel's video, but that will be later, want to take it step by step. Certainly am not planning on doing manual 1:1 large format multiscan stitching, but will manually attempt 4 shot stitch for medium format.

For profiling of my scanner with positive film, I ended up ordering a set of IT8 calibration targets from Wolf Faust at http://www.targets.coloraid.de/.

Will need to go and reread the giant DSLR scanning thread, I'm sure reading it again I may pickup on advice I missed.

What material do you recommend for creating a mask around the film to block light? I assume it needs to be something soft/weak enough to cut with a blade, yet heavy enough and sturdy to stay flat and hold the negative down.

Thank you.