As I discussed here, I have a need to scan a large amount of 35mm film, mainly black and white negatives and color slides. In this article I’ll describe how I use a Nikon D800E to scan 35mm black and white negatives. I’ll cover slides in a later article. I may cover color negatives if I can get the process down where it yields quality results quickly.
With black and white negatives the results are as good as, or better than, the results from a 4000dpi Canon CanoScan FS4000US that I have been using. In fact the picture from the D800E is almost the same size as the 5000dpi result from my ScanMate 5000 drum scanner, and seems to resolve almost as much info.
400 Speed film
Let’s start off with a brief look at the results that are possible. First up is a snapshot of my son William taken roughly 10 years ago. The shot was almost certainly taken with a Nikon N80, and probably using either the 28-105 f/3.5 to 4.5 AFD, or possibly the 50mm f/1.8 AFD lens. Focus would have been attempted on the eyes, but they look a bit unsharp. I scanned this negative on a ScanMate 5000 drum scanner at maximum resolution, about 5080dpi, a Canon CanoScan FS4000US CCD scanner at 4000dpi, and using the Nikon D800E. I then adjusted the tonality of each in Photoshop to match closely. No sharpening was applied to either of the scanned images, but the D800E received a bit of default sharpening from Adobe Lightroom as it was imported.
The results surprised me. I hadn’t really focused on how much resolution the D800 has until I did this scan. The image size is essentially the same as what the ScanMate delivers at 5000dpi. The CanoScan at 4000dpi is smaller by 20% in linear resolution, and 44% in total pixels captured. One question to ask is, does my film hold any extra information that needs 5000dpi to capture. I suspect not for most of what I have shot, and almost certainly not for the images shot on 400 speed film, like this example.
This exercise didn’t start out to be a scanner test, but rather a simple test to see how much I would loose by scanning film with the DSLR. If the resolution captured came close to the lower resolution CanoScan I would be happy. My main goal is to quickly scan many rolls of film to have them available in LightRoom and to post on the web in galleries for the family to view. When I saw that the DSLR seemed capable of out-resolving the CanoScan I was a bit shocked and decided to also scan a few frames on the drum scanner for a point of reference.
All three seem to resolve the film grain well. The Drum scanner has the most well defined grain. This could be softened a bit by opening up the aperture one stop on the scanner, at the expense of some detail. In this case with this film I think that would be the right choice for a pleasing scan. The ScanMate also seems to have resolved a little more detail in the shirt over the other two, but not a huge amount.
The D800E produces an image that’s almost the exact size as the drum scanner. I’m not quoting the exact size because it depends a bit on how I setup the bellows and the film copy attachment. By leaving a bit of room on the edges I assure I can capture the full film frame, at the expense of some resolution and the need to crop the image.
In this image the CanoScan resolved the least detail, but not by much. I think in a 16×24 inch print I couldn’t tell the difference without some close study. That is a print size that’s much larger than I ever print 35mm film.
If this were about picking the ultimate quality I would use the drum scanner every time. But loading the drum scanner takes time, costs money for the scanning fluid and mylar overlay. Each scan also takes roughly 10 to 15 minutes (I haven’t timed it). The CanoScan loads a strip of 6 negatives quickly. The VueScan software I use can also batch scan all 6 images without further user input. Scans still take a long time, especially since I run this scanner over the slow USB connection. The D800E takes about 20 seconds per frame to align and click the shutter. Images can be automatically imported into LightRoom. But the best part of this is the quality appears to be at least as good as the CanoScan, and I can get many rolls of film scanned in an evening. Put on a movie I’ve seen before and just process film.
100 Speed film
So, what are the results like for a finer grained 100 speed film? Surprisingly I don’t have much of this to choose from. I mostly shot faster film in 35mm because I could get good results handheld of fast moving subjects. I am not usually concerned with getting the most resolution out of 35mm shooting. For that I use a larger format film, and usually a tripod. However I did find a roll of Fuji Acros (actually rebranded as Arista Legacy Pro), which is a nice fine grained film. It was shot on an Olympus XA in bright sunlight. I suspect I shot at f/8 given the lighting levels, but I’m not sure. The XA lens is pretty decent, but surely not as sharp as you can get for the format. I’m sure with careful technique and good equipment you can get better results than I am showing here.
Most of the same comments from the other film hold true for this as well. I don’t see a whole lot of difference in the amount of detail each scanning method was able to extract from the film. Based on that I estimate that the film really only contains about 12 to 18 megapixels of information, and it’s probably closer to the 12 mark for the images I have here. But considering that when I set out on this experiment I was only looking to get web quality images out of the scans, I think this is still impressive.
How To in Theory
In short all you need to do is to take a picture of your negatives and invert the image. But for best results you want to make sure the camera is perfectly square to the negative. This will prevent distortion and assure that focus is even across the entire negative. You will also want a lens that will work in the macro range of one to one (1:1) reproduction ratio. In other words the subject will be projected on the camera sensor at original size of the film. This assumes a full frame sensor. If you are using a DX sensor camera the ratio will be closer to 1:1.5, depending on the exact camera.
You can make any lens work by adding extension tubes of the right length. But lenses designed for the reproduction range you are using will give you noticeably better results. Most camera systems have macro lenses that will do a fine job at 1:1 ratios. You will sometimes still need to add extension tubes to allow the lens to focus close enough.
Another helpful, but not necessary, piece of equipment is a bellows system. These act as variable extension tubes, and some allow movements like a view camera (this isn’t needed here). I’ve found that a bellows is by far the easiest method to adjust the the framing and focus in this operation. Many bellows also support a film or slide holder to hold your film and a piece of diffusion plastic to create a smooth light source, which is convenient. These were originally intended to copy slides, or make internegatives. But they work just as well for “scanning” film with your DSLR.
The simplest solution is probably something like the Nikon ES-1 Slide Copy Adapter, which I have not tried. But if I didn’t already own the bellows that’s what I would try. I have also seen home made versions of this online made out of what look like cardboard tubes.
A tripod and lightbox will also work, but I found that very difficult to keep aligned over the long run. But it was certainly good enough for a proof of concept. The important part is the lens. The rest is just convenience.
How To in Practice
I use a Nikon PB4 bellows with a Rodenstock 75mm f/4 Rodagon-D lens. This lens was designed for 1:1 reproduction of medium format film. It is very sharp across the whole frame. But I have also used a Nikon 55mm f/2.8 AIS macro lens on the bellows and gotten excellent results, especially with the D7000 camera (DX format, so 1:1.5 ratio). For a light source I have used a desk lamp, my light box stood on it’s edge, a flash, and as pictured here an Omega color enlarger head. For black and white negatives all gave similar results.
At these ratios a small adjustment in distance is very visible, so the magnified LiveView image is very helpful. I usually frame and roughly adjust the setup using the viewfinder, but I focus using LiveView at 100% magnification. Focus only needs to be setup one time. With the PB4 bellows I have found you need to make slight adjustments to the tilt to get the plane of focus perfectly parallel to the sensor. With LiveView it’s pretty easy. Just check that the sides are as sharp as the center.
I set the lens to f/5.6 which seems to be a good compromise between diffraction and compensating for minor film curvature or warpage.
I shoot in manual mode since the light is constant. I adjust exposure by examining the histogram in a sample shot and adjust the exposure time to fully expose the image to the right of the histogram. By exposing to the right, but not clipping you will minimize the noise from your camera. This is especially important since the image will be reversed and any noise that would normally be buried in the shadows will now be in your highlights. Exposing to the right keeps the image nice and clean.
I tether the camera to Lightroom so the images are imported as I take them. This makes it easier to check them as I take them, but shooting to a card and importing works just fine.
During the import process I apply a User Preset to each photo. This preset does a few things. First, it inverts the image with a custom Tone Curve. Second it sets the image treatment as Black and White.
Notice in the top histogram the the plot is pushed to the left. This is because this histogram shows the results after adjustments. Since we have inverted the image using a very basic Tone Curve things are looking a little dark and low contrast. In the Tone Curve you can see the original histogram, which is exposed close to the right.
As it is now this Tone Curve is very safe in that no information is clipped, but it produces a very flat (see the histogram bunched at the one side). When I am scanning a lot of film at once I will tweak the curve to give better results as a default. This has the advantage of making decent looking images in LightRoom, which is really what I want for most of my scanning. Note that the curves are reversed from the normal way of working in PhotoShop because we are dealing with a negative.
You can save many setups as Presets and have Lightroom apply these as it imports files, or after the fact just by clicking on them.
If all of this seems too complicated, it’s really not too bad in practice. The hardest part will be finding a good way of holding the camera and film at the correct distance, and possibly finding a good lens if you don’t have one. But the results are as good as or better than any 35mm scanner on the market. If you don’t own a scanner this will let you digitize your old archive of negatives. Once you factor in the speed this method becomes even more compelling.
But it must cost a lot right? Even if you don’t have any of the equipment besides a DSLR you can pick up a 55mm AIS macro lens for about $150 and a Nikon PB-5 Bellows for about the same. The ES-1 is about $70. So for about $200 to $300 you can have a very decent scanning solution (less if you shop carefully on ebay). You don’t need a fancy autofocus lens, or image stabilization. In fact the older lenses with an aperture ring are better for this. I don’t know what’s available in other brands, but with Canon at least you can use Nikon lenses with a cheap adapter. If you own a higher resolution DSLR you will get much better results out of this setup than you will with a low end scanner, and you will have a macro lens and bellows to take cool pictures of small things on top of that.