Dean M. Chriss
Photography

Film Scanning Optimization and Tips
December 15, 2020

Heilderberg Drum Scanner

I have in the neighborhood of 70,000 images on film that have been properly stored in cabinets since I got my first digital camera in 2003. I also have quite a bit of extra time because I am not willing to catch or spread COVID-19, making most of the things I would normally do undoable. My film images are slated to become landfill fodder sometime in the next few years, so it's long past time I did some digitizing of these images. Like most photographers I bought a film scanner during the years before digital cameras were up to the quality level of film. In fact I had two different Nikon scanners but that was so long ago I can't remember if the first was a Coolscan 1000 or 2000! The problem is that I never scanned very many of my old images because I was usually busy making more of them. The only ones I scanned were those I used online or printed, but there were tens of thousands more. After I got my first good digital camera I very rarely scanned any film. The digital camera images became better than film and and were instantly available. By comparison the film scanning was tedious drudgery and the time it took to produce a single image was ridiculous.

Fast forward two decades and the old film images are still sitting in the same two cabinets, but the world has changed dramatically. Many of those old photographs would be impossible to capture today. Some of the animals are nearly extinct and some national parks have become so crowded that I no longer visit. Photos of the same places during the same seasons would now be full of people. Some wetlands where I used to take pictures have dried up and become indistinguishable from the surrounding grasslands, and the birds that frequented them are gone too. Aside from the time and drudgery involved in scanning film, I suppose another reason I stopped scanning was because I thought I could capture similar images with higher quality using a newer digital camera. I did not realize doing so would take as many decades as it took to expose all of that film, and in the meantime much of what I had photographed would be gone or dramatically changed. So, scanning the old film is now more worthwhile than it ever was.

With so many images there is no way I will scan all of them, and not all are worth scanning. I decided that only the best of the best would get digitized. I want the highest quality scans I can get without spending a fortune because I'd like the ability to make high quality prints at reasonably large sizes in the future. When the image quality on the film is top notch, prints up to around 16x24 inches are possible. If the film images are just the tiniest bit "off", as some are, printing at 12x18 inches is easily possible after processing with modern tools. I know all of this because I've done it. Prints at those sizes can hold their own against those from modern digital cameras if they are done properly, but in order to achieve this the film scans must also be flawless. Obtaining a scan that allows such quality is either a lot more expensive than it sounds or a lot harder than it sounds. In this article I will discuss a few ways to accomplish this. As is typical, the highest quality scans are obtained in a way that is very expensive and does not lend itself to do it yourself users.

The best possible film scans come from drum scans and the best of those are done on Linotype-Hell Tango scanners made by Heidelberg. That model was the best, smallest, lightest, and last they made. These are massive and precise instruments that stand around 5 feet tall and weigh 551 pounds (250 kg). The weight and associated sturdiness is necessary for the precision they achieve, with all of the precision parts mounted on a massive die-cast frame. These scanners can handle any film format, or up to a dozen 4x5 inch pieces of sheet film and can achieve an optical resolution of 10,780 ppi and a DMax 4.2, which is about as good as it gets. Many of these scanners still exist and they are typically used by scanning services. The film is wet mounted to the surface of a transparent drum, thus the name. The film plus drum assembly is then placed into the scanner, which illuminates the film as the drum spins rapidly and the scanning head moves along the cylinder. A single line of pixels are scanned on each revolution of the drum. The scanning fluid aids light transfer through the film, eliminates Newton's rings from forming between the film and the drum surface, fills fine scratches making them invisible in the scan, and greatly reduces the appearance of dust and other imperfections that would otherwise appear in the scan. These scanners originally cost in the neighborhood of $65,000 without the drum and other necessary accessories, but in 2020 they can sometimes be purchased used for around $4,000 - $7,000. You can have your 35mm film scanned on a Heidelberg Linotype-Hell Tango for between $30 and $70 per scan, depending on the desired resolution. It takes a scan of more than 4000 ppi to capture all of the information contained in a well made image on fine grained film, but 4000 ppi will get most of it. Because drum scans are done wet they are a hassle, but they provide demonstrably better results than any other method. Of special note is that your film cannot be mounted in a slide mount and it must lay smoothly against the curved surface of the scanner's drum. Many mounted 33 mm slides are permanently deformed at the film's edge by a heat mounting process used by Kodak.

The next highest quality scans are produced using the no longer made Imacon X5 scanner by Hasselblad. They originally cost in the range of $16,000 and can be purchased used for around $7,000 in 2020. Imacon scanners have a patented method for holding the film at a uniform and precise distance from the scanning head without using any glass, thus eliminating the need for fluid. Since the scans are done dry, using these scanners is far easier. The Imacon Flextight X5 scanner was Hasselblad's best model. It offers a maximum resolution of 8000 ppi and a DMax of 4.9. That is plenty of resolution and the DMax is over one stop better than a drum scanner, which can be a significant advantage if the film is extremely dense. Otherwise the wet scanning of a drum scanner produces a slightly better looking scan because the fluid physically eliminates most scratches and other defects. The fact that the drum and Imacon scanners both hold the film at a consistent and precise distance from the scanning head is key. That keeps the film always perfectly in focus and puts them well ahead of the more affordable dedicated film scanners from manufacturers like Nikon. Most Imacon X5 scanners are in the hands of private individuals but a few of them offer scanning services. In a quick online search I found one doing this for $6 per image on 35mm film. That's quite pricey if you are doing many, but certainly more affordable than drum scanning. The biggest reasons for the price difference is that the scanners were less expensive and the scans are done dry, making them far easier and less time consuming. I suspect the same reasons account for the fact that Imacon scanners now sell used for roughly the same price as the Heidelberg, which when new was over four times as expensive.

Affordable Dedicated Film Scanners

Among the best dedicated film scanners that can be purchased in 2020 for considerably less than $7000 are those formerly produced by Nikon. The tips in this article are based on those but could as well be used with any scanner that can measure the focus distance and be manually focused. The Nikon Coolscan 5000 ED, 4000 ED, and several similar Nikon scanners have the features I mentioned. As far as I know, only the Nikon scanners have 16X oversampling, which essentially averages the results of 16 scans of each pixel to minimize noise. This puts them on top of the list. None of Nikon scanners are still produced but they can be purchased used. Not long after they were discontinued the top of the line Nikon scanners were selling for up to $4200, which is around three times their original selling price. In 2020 the 5000 ED can be found used for between $1,650 and $3000, while the 4000 ED commonly sells for around $700, and the Nikon V ED (LS50) sells for around $1000. These are ballpark prices based on a quick glance at ebay so take them with a grain of salt. Dedicated film scanners are still produced by companies like Pacific Image and Plustek, with the Plustek OpticFilm 8200i SE appearing to be among the best currently made. It is priced at less than $400. I am not familiar in detail with the features of any of the current scanners.

Back in the days when everyone was scanning film lots of people had problems caused by the fact that film is never flat. When left in a strip and not mounted it naturally curves smoothly toward the emulsion side. When mounted in a slide mount, which holds the edges flat, it usually warps in random ways. It still curves toward the emulsion side, but in an uneven way. The amount of warping depends to some extent on temperature. Film subjected to the high temperatures within slide projectors would classically "pop" into focus because of this. Film warp variations of a far lesser degree can even be seen between the exterior and interior temperatures of film scanners as a slide warms up, slightly changing the optimal focus setting. This lack of flatness, and to a lesser extent its change with varying temperature, is the biggest problem with these dedicated film scanners. The primary reason that scans from devices like drum scanners are better is that the film's emulsion always a precisely fixed distance away from the scanning head. This flatness problem in lesser scanners sounds easy to solve but it is not. Putting glass between the scan head (or camera) and the film degrades the attainable quality significantly. The film must be held flat without "sandwiching" it between glass. Glass would create issues in scanning that range from Newton's rings to preventing the infrared scanning that is used to map dust and other defects for later removal. You will notice that no dedicated film scanner scans the film through glass, and that is not an accident. The necessarily small depth of field in scanner optics greatly worsens the fact that affordable film scanners have no way to hold the film perfectly flat.

Optimizing Depth of Field

The better Nikon Super Coolscan models can all automatically focus at the center of the image area when a slide is inserted, but you can also use a cursor to make them auto-focus anywhere within the image area. The resulting focus setting is displayed, and useful, because the scanners can be manually focused by manually changing that number. The units associates with this setting are presumably tenths of a millimeter, but the units are irrelevant. Through much experimentation I have determined that the scanner's maximum depth of field is around 10 units or 1 mm between the smallest and largest focus distance setting.

If the important parts of an image lie within the central area of the film, Nikon's automatic autofocus can be good enough. If the area where maximum sharpness is required is localized but not in the center of the image, manually focusing in the middle of that localized area will probably be good enough. If neither of those situations is true, I recommend measuring a variety of different points within the important parts of the image to determine if the extremes are within 1.0 mm (10 units) of one another. If so, averaging the numbers and manually setting the focus to that average will provide a good sharp scan, but that sharpness may not extend into the image corners.

Focus Stacking

If you truly need corner to corner sharpness you will find that most slides go well beyond the 1.0 mm depth of field limit. In the old days nothing could be easily done about this because focus stacking was a difficult manual process that could not be accomplished if there were even tiny geometric changes within the image, like those due to "focus breathing" in camera lenses or variations of warp in scanned images. Today things are different.

Making a focus stacked composite of images captured at different focus settings is relatively easy today. It works due to the magic of automatic content aware fill in Photoshop that seamlessly joins pieces of the multiple images that do not exactly fit together. This holds true for film scans made at different focus settings. I have found that only two scans are needed, and going to three or more offers no improvement.

To do this I typically take measurements at roughly the locations shown below, find the midpoint between the smallest and largest measurements, and then average the measurements above the midpoint and average those below the midpoint. Those two averages are the focus setting I manually enter for each of the two scans that will be focus stacked. But there is a catch.

Focus measurement locations
Focus Measurement Locations

Remember those thermal issues that can change the film geometry? Those are very real and can significantly affect the focus setting even during a scan. Allowing the film to bake in the scanner for some time while it is powered on helps a little. What seems to help more is periodically (perhaps every 30 seconds) autofocusing on the center of the image until the focus setting stops changing. Even after doing that you may sometimes find that the focus settings have changed significantly during the first scan. Because of that I repeat the initial procedure of finding the high or low average focus setting (whichever one I haven't just used for the previous scan) and proceed to make the second scan using that new average. We aren't dealing with 500 pound pieces of laboratory grade equipment so these hassles are necessary to extract the ultimate quality from the less capable equipment. The good news is that although this may seem inexact, has never failed to produce a very good scan. I think the "extra" DOF gained in the process more than makes up for the shift in focus caused by temperature changes. If your purpose is online sharing or small prints, none of what I am describing is necessary.

Last but not least. Nikon stopped supporting their scanner drivers and software almost as soon as they stopped manufacturing the devices. In the meantime operating systems and computers have changed. You can run the original software in Windows 10 but you need to go through some contortions to make the drivers work. It is actually quite easy if you have some experience with things like this. The following two links will be extremely useful for Windows users in this endeavor.
https://www.sevenforums.com/drivers/44994-getting-your-nikon-coolscan-work-w7-x64.html
https://www.sevenforums.com/drivers/44994-getting-your-nikon-coolscan-work-w7-x64-60.html#post3376729

Mac users must purchase third party scanning software to make these scanners work. Windows users can do the same if they don't want to get involved with making the original drivers work on new operating systems. The primary makers of such software are Silverfast and VueScan.

Digitizing Film Images with a Digital Camera

What I have said about film scanning so far may have convinced you to digitize your film by photographing it with a digital camera. Lots of people have started doing this. I thought about trying it, then I thought in detail about the precision and other intrinsic issues that would be encountered in trying to obtain the best possible results. Then I decided not to try it. I am sure it can work well enough and without much fuss for many purposes like online image sharing and making small prints, but making high quality prints at sizes like 16x24 inches from 35mm film requires the absolute maximum detail and quality attainable from a well made image on 35mm film. To accomplish that you need to get the equal of a great 4000 ppi film scan, all the way into the corners, from a camera. If that's possible it certainly is not easy. On top of that, this is not a plug it in and use it scenario. You have to source very good components and possibly build one or more of them. Then you need to set them all up with the utmost rigidity and precision. An optical bench would help greatly, but at the price of a small car that's not very practical. Many online articles exist describing the required camera setup, so I won't reinvent the wheel. What I will do is discuss some problems that I think are common and that must be overcome to get the highest possible image quality. I will also mention some possible ways to solve them in case you decide to go down this path:

1. Camera: If you do the math, a full 35mm film frame at 4000 ppi produces about 21.3 MP, so you need a good camera with considerably more than 21 MP resolution. That's because the resolution of the entire optical system (camera plus lens) is never equal to the camera's image sensor resolution. It would be with a theoretically perfect lens, but those do not exist. This is not just just about the number of pixels in the camera. It's about whether the entire system, at the shooting aperture and focus setting, can resolve 4000 ppi from the center of the frame all the way into the corners. I don't know if there are any camera and lens combinations capable of that at the corners of the frame. That's one reason films are normally scanned in lines of pixels, capturing one pixel at a time and keeping resolution absolutely constant across the entire frame.

2. Lens: You need a true macro lens of outstanding quality with an extremely flat field at its largest apertures, and it must be capable of 1:1 magnification at a minimum, and ideally more. As already mentioned, the corners of the field have to resolve 4000 ppi, and extreme corners are always softer than the center, so the center has to resolve much more. Lenses typically attain their highest resolving power at the center of the frame when they are wide open, or nearly so, but they also have the softest corners and most aberrations at those wide apertures. As the aperture gets smaller corner resolution increases, center resolution decreases, lens aberrations are reduced, and depth of field increases. We need the best of everything, from the center of the frame to corners, to happen at those wide apertures. Few if any lenses of any type can do that. Adding extension tubes to attain the 1:1 magnification results in field curvature, so a true 1:1 macro lens is needed. Some candidates include the Canon EF 100mm f/2.8L Macro, Sony FE 90mm f/2.8 Macro, Nikon AF-S 105mm f/2.8G ED IF VR Micro, Sigma Art 70mm F2.Macro, and Tamron SP 90mm f/2.8 Di Macro. Some or none of these may be acceptable and there are undoubtedly other possibilities that my quick search did not find. It's just a starting point.

3. Setup Rigidity: At the wide apertures and magnifications required, the slightest change in camera position will put the image out of focus. The slightest vibration will blur the image. We're talking about less than pixel-sized movements here, so you need a good, heavy, copy stand or similarly rigid and vibration free support for both the camera and film.

4. Film and Image Sensor Parallelism: Making the camera's sensor plane perfectly parallel with the film must be done more accurately than you probably imagine and is a lot more difficult than it sounds. With the paper thin depth of field involved, the smallest lack of parallelism makes parts of the photographed image un-sharp. As I discussed in item 2 above, using smaller apertures to compensate will degrade the captured image. Depending on the details of your setup, stopping down the lens can also bring elements of the backlight into the depth of field, producing artifacts in the brightest (most transparent) parts of the captured image. See item 3 below. Lack of parallelism will also distort the geometry of the captured image. The easiest way to make the plane of the camera's image sensor parallel to that of the film is by using a mirror placed where the film would be. Use your camera's Live View or a mirrorless camera and turn on the grid lines feature. When the reflection of your lens is perfectly centered your camera's image sensor is parallel to the mirror.

5. Film Flatness: Film is never flat, and it needs to be flat if you are to get a good capture of the image it holds. This problem plagues the camera "scanning" method as much as it plagues all of the affordable film scanners. Remember that the depth of field is paper thin and if you stop down to increase it you lose resolution. Ideally the film should be held flat without any other surface (like glass) touching either side of the film's image area, and without photographing through any other material, even the film's base layer. The film emulsion should be imaged directly. Many people sandwich the film between a tablet or laptop screen or other backlight and a piece of glass to "solve"  the flatness problem and provide even illumination. There are several issues with this approach.

First, photographing through glass or any other material will degrade the captured image. The surfaces produce unwanted reflections that degrade DMax, contrast, and dynamic range. Optical quality glass with an anti-reflection coatings would help, as would preventing all extraneous light from entering the optical path between the film and camera lens. Any extraneous light falling on the film or the camera lens will reduce DMax, contrast, and dynamic range. A flocked black tube might help with that. When digitizing film by any method the film's emulsion side should be imaged directly, not even through the film's base layer, much less a sheet of glass. There are very good reasons that the best film scanners do it this way.

If the film is mounted in a slide mount, sandwiching it between a tablet screen or glass and another piece of glass will not keep it flat enough. The thickness of the mount still allows the film to warp. Removing the film from its mount and sandwiching it for flatness is not a good idea either. The biggest problem is that contact between the film and other smooth surfaces will produce Newton's rings. This is one reason film scanners do not use glass to flatten the film. Newton's rings are unavoidable unless the glass is dirty or you slather scanning fluid on your tablet or backlight, the film and any other glass or transparent material used to make the "sandwich". There is a type of glass called "anti-Newton ring glass" with a very slightly roughened surface that prevents the formation of Newton's rings. This was used in some high end slide projection mounts to flatten the projected film image for better focus and sharpness. Photographing through anti-Newton ring glass would be a terrible idea, but using it behind the film against the base layer might work. Then, holding the film flat with an open window frame, something like the mat on a framed print, might work. If that frame was made of magnetized metal, and a similar metal frame was glued to the opposite side of the anti-Newton ring glass, you might have a great setup! No matter how it is done, the film should be photographed with the camera facing the film's emulsion side to allow photographing the emulsion directly without any intervening layer.

6. Even Illumination with High CRI: Evenly illuminating the film is critical, and it must be done with illumination that has a very high CRI (color rendering index). 100 is the highest possible CRI, and that can be provided by common, old fashioned, incandescent light bulbs. Other sources typically have a lower CRI. Unfortunately, incandescent bulbs also produce a lot of heat, and as discussed above, heat changes the geometry of the film. It is possible to use these if the lamp is above everything else, since heat rises. In particular, some use old enlargers as a perfect 100 CRI light source. They were designed to very evenly illuminate film for analog printing so they are ideal if the normal configuration lamp is retained, with the lamp on top.

Also as mentioned previously, most people use something like a tablet or laptop screen to illuminate the film. iPads are said to work well, but there is no reason to think the CRI of such illumination is high unless someone measures it and shows that to be true. Perhaps that information is buried somewhere online. It is also important to remember that the light produced by these devices emanates from pixels that are large relative to film grain and the finest image details. Sufficient distance between the tablet screen and the film is needed to assure that the pixels cannot be partially resolved by the camera, especially in bright (nearly transparent) areas of the film's image. This is another reason that using small apertures are a bad idea when "scanning" film with a digital camera. Other light sources designed for digital camera film scanning can be found here, though I do not know whether they are as good as they claim to be.

7. Dust and Scratches: The best digital camera "scans" will contain plenty of dust, fine scratches, and other nits. Performing such scans in ways that require any contact with areas of the film containing the image are bound to damage some film with fine scratches at a minimum. The film scanners I mentioned have nothing that touches the image area and can do a separate scan with infrared light to map dust particles and scratches, which allows the scanner's software to automagically remove them. This does a fantastic job and saves hours of retouching if there are more than a few images.

8. Oversampling: The film scanners I mentioned can oversample an image up to 16 times for the purpose of maximizing DMax and reducing noise. That means they scan every line 16 times, averaging each pixel on each scan to cancel out the noise, which is random. This helps greatly in extracting fine details from shadow areas within the image. You can't do that with a camera. Well, you actually could, but I wouldn't want to.

Flatbed Film Scanning

I have never done this and will not say much about it, but from a bit of research indicates the drum scans always come out on top. The difference between flatbed scans and the digital camera method depends on how the scans are performed. The Petapixel website has a great comparison video that puts the flatbed scanner ahead of the digital camera method, but the flatbed scanner uses an inverted fluid mount. That is uncommon unless you are a scanning aficionado, but a little research shows it can increase flatbed scan quality significantly. I'm not getting into all of this but if it's something you want to pursue there is more information here and here.

This independent website puts the digital camera method ahead of the flatbed scanner, using a dry Epson V550 Perfection flatbed scanner. His comparison of the flatbed scanner and digital camera results are fair enough. But he also shows that the digital camera results are comparable to a "lab scan".  My problem with this review was the lack of data regarding the "lab scan", but he did state the lab name and that a “large/corrected” JPG scan was requested. After looking these up online, and then looking up the two Nortitsu scanners they use it appears that the scanner is a commercial grade device designed for high throughput of either 600 or 1200 negatives per hour at 1002x1512 pixels. That's a very low resolution of just over 1000 ppi, so the reviewer has effectively shown that the digital camera "scan" produces the equal of a 1000 ppi film scanner. That is certainly not indicative of the results from a good film scanner at 4000 ppi, which would produce pixel dimensions of 5640x3776 pixels using 35mm film. In addition, at that high throughput there is obviously no oversampling, and the compressed JPEG file is not what I'd prefer either. Given the large deficiencies of the "lab scan" in this review it seems obvious that a good dedicated film scanner would outperform the "lab scan", which in this review was roughly equal to the digital camera method.

 

Conclusions

You already know what's coming. Taking everything above into account, ranked by the output quality, the drum scanners come first, followed by the Imaxon X5 (and perhaps high end flatbed scanners using inverted fluid mounts?). The more affordable dedicated film scanners come in second, "scanning" with a digital camera comes in third, and dry flatbed scanners come in last. For me, the hassle of sourcing all of the right components for the digital camera method, including a good light box and copy stand, setting everything up with the precision required, and the space needed for the setup, is also a big disadvantage of the digital camera "scanning" method.

Considering the high cost of the drum and Imacon scanning, I think the affordable dedicated film scanners are the best practical solution if scan quality and cost are primary concerns. If none of this is appealing there are cheaper scanning services that do a good job, but you usually have to pay considerably more than their advertised prices to get the scans at 4000 ppi (if that is even offered) in a wide gamut TIFF file instead of the more normal 3200 dpi or lower resolution scans output as sRGB JPEGs. I have not seen 16X (or any) oversampling to reduce noise and increase dynamic range available from these services, probably because doing it takes 16 times the amount of time required by a normal single pass scan. On the other hand I'm sure someone would do it for the right price, which is likely 16 times the normal price! The fact remains that larger formats are far less demanding on everything from capture to scanning for comparable quality at the same printed sizes. That means they are more amenable to the other scanning methods, like using a digital camera, with better results than could be gotten from 35mm film.

All of the fuss and crazy precision I discussed is required due to the microscopic size of the smallest details on the film and the enormous magnifications involved. To produce a good 16x24 inch print from 35mm (1.41x 0.944 inch) film it must be magnified 288 times! A 4000 ppi scan gives roughly the 16x24 inch size at 360 ppi, which is a good resolution for input to most Epson printers. Of course the image on the film also must be essentially perfect to get good results. The fuss and precision is also necessary because we are trying to approach the scan quality of a  laboratory grade device like a drum scanner with a cheap consumer grade dedicated film scanner. Performing the focus averaging or focus averaging plus stacking I have described certainly takes a lot of time. If you have images on the film that you can never photograph again I think it's very worthwhile, but the same can be said of paying to have scans made on a drum scanner or Imacon scanner. It all depends whether you have a greater excess of time or money, and what you are willing to do. As always, your mileage may vary.

Happy digitizing!

Dean

Copyright 2020 Dean M. Chriss

email