Remember film? Even if you don't, there's a really good chance a dark corner of your house is packed solid with old slides. In my case, not only do I have slides I personally shot 30+ years ago, but I also inherited larger, older, slide collections from my dad and my uncle. It's dozens of Kodak carousels of 35mm slides, a couple dozen boxes of 6x6cm slides (some in glass mounts), and untold many frighteningly unsorted slides in boxes. Fortunately, I also have a Nikon CoolScan 9000, which is capable of making awesomely great scans... in just minutes per slide. Taking 25 minutes for a top-quality 6x6cm scan is fine for the most critical photos, but not for scanning lots of slides. There are cheap film scanners that are faster, but quality is much poorer, and most don't handle 6x6cm slides.
Instead, how about using a slide scanning attachment with a high-quality digital camera I already have? I actually already had a commercially-made one from the 1970s for copying 35mm slides, but I needed to make one for 6x6cm. I did that by creating a 3D-printed slide copy attachment that fits on a macro lens -- and many kit zooms can focus close enough so that a macro lens isn't needed.
So, this Instructable explains how to get high-quality scans of slides really fast, using nothing more than the expensive digital camera and 3D printer you already own. ;-) In particular, I used a Sony A7 camera with an ancient Minolta Macro Rokkor 50mm f/3.5 lens and a MakerGear M2 3D printer.
Step 1: Slides
Let's start with a little discussion about slides. We'll ignore negative (print) film until I get around to adding an unmounted film holder to the copy adapter design....
Slides come in various sizes, the most common of which are 35mm and 6x6cm. The slide mount dimensions are fairly well standardized for each format, but 35mm slides are much smaller than 6x6cm slides -- about 50x50mm vs. 70x70mm. Thus, you need different slide holders for the different film formats. In fact, the projectors used for viewing slides each tended to support only one slide format.
There were many different types of slide film. Chemistry for processing color slides could get quite complex, but the E6 process most commonly associated with Kodak's Ektachrome films was one of the easiest to use, and many color slide films could be processed with it yielding vaguely similar results. Ektachrome colors are particularly familiar to anyone who has seen lots of National Geographic photos. Unfortunately, over time the cyan dye in Ektachromes tends to decompose faster -- taking image content with it. Kodachrome, which requires more complex processing, has famously brighter colors and much better dye stability, with yellow being the first to fade. The point is that old color slides probably have color issues, and those issues are highly dependent on the film type -- so any restoration of color in scans will be too. Even severely discolored slides usually can be made to look good, although often the colors are credible guesses rather than accurate restorations.
Independent of format and film type, old slides are likely to collect dust, dirt, scratches, and other defects including pink splotches or even branching fungus infections. Wiping the slide (as opposed to a glass mount) is more likely to cause damage than to clean the slide, but a puff from a hand blower is worth trying. Professional film scanners like my Nikon use NIR light to image many defects so that software can automatically remove them, but "heal selection" tools in image editing software can do a pretty good job synthesizing credible patches to cover the bad spots you identify.
You'll often hear people claim film has very high resolution, and my Nikon scanner can deliver 20MP 35mm scans and 80MP 6x6cm scans. However, that doesn't mean they resolve that well. In fact, film has flatness and thickness issues as well as grain problems that limit resolution. Realistically, good 35mm slides can be expected to be at least as good as clean digital images of 1.5MP and 6x6cm is good for about 4MP. Occasionally, you'll get up to about 4X that, but scanning at much higher resolutions is primarily about ensuring Nyquist sampling constraints are met.
Basically, you need to have realistic expectations. If you are expecting perfect color, freedom from defects like dust and grain, and awesomely high resolution out of your 30+ year-old slides, think again.
Step 2: Introducing the DupliHood
Remember Spiratone? Back in the days of film, that was Fred Spira's store in New York that sold the most amazing collection of photographic equipment and accessories. Well, not surprisingly, they sold a bunch of slide copier attachments. There was an awkward Bello-Dupliscope using a bellows, a Vario-Dupliscope with variable magnification, and the simple Dupliscope that was fixed around 1:1 magnification. It was that last version that I mounted on my Sony A7 to immediately give me a quick way to digitize 35mm slides.
Aside from the fact that I need to be able to copy 6x6cm slides, not just 35mm, there are a couple of issues:
- The Dupliscope's lens is not only fixed focus, but fixed aperture -- around f/16, presumably to ensure that slightly cupped slides would still be completely in focus (or maybe just to cut cost?). Unfortunately, that really ensures any dust on your camera's sensor is crisply imaged along with any unevenness in the light source or diffuser backlighting the slide you're copying.
- The slot to hold the slide is too thick and has a significant gap between the back of the slide and the black metal mask (which isn't larger than the film viewing area) and frame to hold a diffuser. This often results in shadows and light strike marks around the frame edges.
Those are not difficult issues to fix in a custom-made slide copier adapter... and I have a 3D printer. :-)
In honor of the Dupliscope, we'll call my new thing the DupliHood. After all, it looks a lot like a big lenshood and screws-into the filter thread of a lens that can focus close enough. The 3D-printed adapter is large, but weighs less than 100 grams, so you don't really need to worry too much about the load on your lens.
What lens should you use behind your DupliHood? I used a Minolta Macro Rokkor 50mm f/3.5 manual-focus macro lens, which happens to have been a spectacular choice. For 6x6cm film, there is a good chance your camera's kit zoom will suffice. However, even an APS-C camera's sensor will not be filled by a 1:2 macro image of a 35mm slide. I used a short extension tube on my Sony NEX-7 behind the same lens I used without a tube for the 6x6cm on my Sony A7. Confusing, eh? I also was able to use a Canon FD 50mm f/1.8 on an extension tube with the NEX-7 for 35mm slides, but image quality was somewhat poorer than with the macro lens despite stopping down to f/8 or so. Try making images of a slide hand-held to see which of the lens options you have is the best choice.... You want the imaged area to be slightly larger than the slide's film area so you'll be sure to get it all (you can tightly crop later). You also want the distance from the slide to the front of your lens to be short enough for your 3D printer to be able to make it -- long focal length lenses don't help here.
Step 3: Making Your DupliHood
My design is Thing 431624, DupliHood: a 3D-printed slide copy attachment at Thingiverse.
The DupliHood consists of three parts that are friction-fit together as shown in the yellow exploded assembly figure: tube, spring mount holder cap, and optional diffuser. If you have a diffuse light source (e.g., a sheet of white paper on a sunny day), you don't need or want the diffuser. The parts are easily constructed out of PLA using a consumer-level 3D printer, such as my MakerGear M2. You want precise dimensions, so I'd avoid using ABS and keep fill below 40% to minimize distortion during cooling. They should be printed in the orientation in which they are shown. I recommend opaque black for the tube and white or clear for the other parts -- to avoid light leaks and color casts.
However,if you print the parts using the default design parameters, it's unlikely to work with your lens and camera... which is why the design is actually an OpenSCAD program designed to use Thingiverse Customizer for setting the parameters:
- Thread Diameter: the filter thread diameter for the lens you'll be screwing this onto.
- Thread Pitch: the filter thread pitch -- usually 0.75mm.
- Thread Tolerance: this setting is allowing for inaccuracy in your 3D printer. You've probably never been crazy enough to try to print a 0.75mm pitch thread before, but this thread design is tweaked so it should work. It is still a very tight tolerance part... and this parameter lets you pick how loose the fit should be to allow for slight placement errors in the extrusion. Around 100-125 microns (0.1-0.125mm) seems sufficient for my MakerGear M2.
- Thread Rotate: supposedly identical screw threads have a nasty tendency to stop at different rotational positions. There are even lenses with the nasty habit of the thread rotating as focus is changed -- avoid them for this project. We need the tube's film mask to line-up with the camera body's notion of horizontal; you can probably force it at least 10-15 degrees, but if you're really unlucky.... Tuning this parameter rotates the printed thread to align this; if the first one you print doesn't line-up nicely, at least the second one can by tweaking this parameter. You will probably want to print the "test part for thread rotational alignment" first to check Thread Tolerance and determine Thread Rotate; this part can be discarded after confirming parameters are correct.
- Tube Length: take your camera and lens and measure how far the front of the lens filter thread is from the slide when the film area just fits in your camera's finder. That's the number that goes here. Try t o avoid lenses with very long focal lengths because they'll need very long tubes.
- Film: which film format is this for, 35mm or 6x6cm? This determines the mask sizes, spring design, and even the diameter of the parts.
- Part: which part do you want to make? You'll need at least the threaded tube and spring cap, which you can request separately... or you can just make the exploded view and separate the pieces in the generated STL file. You will almost certainly want to make the alignment test part first -- it's a small part that has the thread and is marked with the rotation alignment value it was made with, so you can screw it onto your lens to check where the thread stops rotationally.
Not too bad... but of course I can't guarantee it will work for you, because it does depend on details of your lens and 3D printer. The way to find out if it does what you want is to go to Thing 431624, DupliHood: a 3D-printed slide copy attachment, set the parameters, and print your DupliHood.... It's cheap to try -- it uses less than 100 grams of PLA, which means just a few dollars worth of material.
Step 4: Using the DupliHood
It seems pretty simple to use your 3D-printed DupliHood: put it on your lens, check that the masking is aligned, insert a slide, point the camera at a diffuse white light source and click. Well, sort-of. Here are some details using my Sony A7; similar tricks apply on many mirrorless or DSLR cameras:
- Focus and aperture. You've got plenty of time, so use manual focus with a magnified live view or EVF. What's more, do it with the lens at the taking aperture so you can see that the whole slide is sharply in focus. What aperture should you use? Use the smallest f/number that allows everything to be in focus and gets peak performance from your lens, usually about f/8 (not f/16 or higher, they'll just show worse diffraction and sensor dust). Take your time with this because you probably only need to do it once for copying lots of slides.
- ISO setting. It's ISO limbo time -- how low can you go? Slides can have a surprisingly large dynamic range, and nearly all digital cameras have the dynamic range recorded dramatically drop at higher ISOs. Color fidelity and noise also tend to be much better at lower ISOs.
- Other camera settings. Let the shutter speed fall where it may, probably using "A" mode; you don't even need a tripod for multi-second exposures because the entire copy rig moves as a unit with the camera. You're probably thinking this is a time to shoot raw... maybe. Most cameras have some form of shadow enhancement; Sony's DRO works pretty well and the built-in multi-shot HDR works even better. Underexposed slides and mutated dyes can hide detail, and I've found that a 6.0 Ev HDR JPEG pulls out noticeably more than I can get from a single raw. Don't worry about getting that artificial HDR look -- if your slide didn't have it new, the HDR copy will not either. Of course, using camera JPEGs limits postprocessing, so you'll also want to get the camera's color balance close (perhaps via custom white balance) and consider using Adobe RGB instead of sRGB if you will still need to do significant postprocessing of the colors.
As for the postprocessing, well, here are two suggestions:
- Ektachromes tend to lose cyan. In your favorite image editor, a good start is individually auto-leveling the R, G, and B channels, playing with the gamma on each (mostly, turn down R gamma), and then perhaps turning saturation down a bit. The particular example shown here wasn't too bad, so simply setting white balance did most of the work. You can probably do much better in general with more complex color tweaking, and maybe sharpening and grain reduction, but most fancy manipulation can be scripted and automatically applied to a bunch of similar slides.
- Your image editor's "heal selection" tool is generally the right way to remove localized image defects (dust, dirt, etc.), but it can take a long time to select all of them by hand. Sometimes, you can get most of them by simply selecting the color of one (near black), growing the selection a few pixels, and manually subtracting the (relatively few and large) non-defect areas from the selection. It's also been observed that most defects make the image darker, so median filtering and then combining layers using "lighten only" can take out a lot of the smallest marks without destroying too much detail -- often good enough for things like sky.
Given all of that, I think the results I've gotten with my DupliHood are fairly spectacular. I'll still use my Nikon CoolScan 9000 for the most critical slides -- the ones where I'll let it take an hour or more sampling to get the ultimate quality -- but the quickie copies from my A7 with this rig are shockingly close in overall quality. So, if you didn't happen to spend thousands of dollars on a Nikon CoolScan 9000 back when they were still available, now you've got a good option.... :-)