CULTURAL HERITAGE DIGITISATION / DRUM SCANNING
Art & Artefact works with artists, photographic labs, publishers, galleries and museums to deliver the highest quality colour-managed imagery, tailored to your needs and specifications.
Art & Artefact are passionate about making the best possible scans of film. Scanning film is a craft no different to producing a fine print and no less important. As such just as much time and care should go into the scanning of the film as you would take when working on a print in the darkroom. At Art & Artefact we understand the needs and concerns of artists using film, often to produce prints several metres in size, or the importance of maintaining the look and characteristics of a certain film stock. At Art & Artefact every effort has been made to get the best possible scans and we're confident in saying that we offer the highest quality drum scans available, but crucially, at a price that is fair and sustainable for artists.
It's our mission to help and encourage artists working with film.
(This page is a work-in-progress. Please excuse any typos or grammatical oddities!)
What is a drum scanner?
Drum scanners are the oldest form of scanning technology; predating flatbed scanners, and in fact the earliest TV cameras were essentially extra large drum scanners.
The 2015 Federal Agencies Digitisation Guidelines Initiative (FADGI) states that “To this day, drum scanners provide the highest image quality of all imaging devices”.
The reason FADGI describes drum scanners as providing the highest image quality of all imaging devices is because they work in such a different way to nearly every other imaging device. Drum scanners use PMT’s (Photo multiplier tubes); giant analogue vacuum tubes that are incredibly sensitive to light; far more sensitive than any CCD or CMOS sensor. It’s because of this that drum scanners can capture both highlight and shadow detail (along with the most subtle variations in between) that no other scanner can come close to recording. What’s more than that each drum scanner has three of these PMT’s (one for each colour; red, green and blue) this means that a drum scanner records three colours per pixel; while a CCD scanner (for example Epson, Canon, Microtek or Hasselblad/Imacon scanners) only record one colour per pixel in the bayer pattern, then interpolate this data (known as demosaicing) to create a true ‘full-colour’ three-colours-per-pixel images. This is why people will say that a drum scan at a certain PPI will always be sharper and more detail-rich than a CCD scanner at the same PPI; the drum scanner is literally capturing more tonal data per pixel than the CCD scanner, let alone the fact that a drum scanner can always scan at a much higher resolution than any CCD scanner.
A drum scanner illuminates and scans just one pixel at a time; while a CCD scanner will illuminate a line or sometimes the whole film as it scans a whole line of pixels at a time. The benefit of the drum scanners method is increased clarity; as a pin-prick of light is projected through the film into the sensor there is no light coming through the film at weird angles; no loss of sharpness caused by stray light or light bleeding through a thin part of the film into a thicker part of film; because of this drum scanners don’t suffer from the same problems of flaring or softening that CCD scanners do. The data rom the PMT’s then pass into especially precise ‘ADCs’ analogue to digital converters (one for each PMT) which translate the information from the PMT’s into 16-bits-per-channel colour CIELab* data. My scanner creates a linear TIFF (a scan that is made without any curves or levels alterations, without any sharpening or any inverting of negatives; this is a raw scanned file); many older drum scanners automatically invert, apply sharpening, only output to 8-bit or clip highlight and shadow data making them inappropriate for the kind of precise work artists demand today.
As the sensor parts (the PMT’s and ADC’s) of a drum scanner are big, bulky and highly delicate parts, rather than moving the sensors over/under the film as in a flatbed scanner then in a drum scanner the sensors stay put and the film is moved around them; to be exact the film is mounted on a cylindrical perspex ‘drum’ that rotates and corkscrews around the sensors. It’s this drum that gives these scanners their name. Film is wrapped around the drum with special mounting fluid on both sides, that’s then sealed in with optically clear polyester film taped down on all sides. The benefits of this system are twofold; firstly the film is pressed hard against the drum keeping the film the exact distance from the sensor at all times (many people using CCD scanners will know or have battled with the problem of uneven focus due to sagging or not-perfectly flat film) and secondly that the mounting fluid has the effect of working like an optic, making much clearer and sharper scans. It’s common for scientists to ‘wet-mount’ the samples they are looking at under a microscope for exactly the same reason; wet-mounting increases clarity and sharpness tremendously when looking or scanning something at especially high magnification. If you’ve ever been looking through a window into a tank at an aquarium and you’ve bumped your head against the glass because its so clear, then you’re on the right tracks. Wet-mounting film makes a vast difference to the detail and sharpness of a scan.
Drum scanners always outperform CCD scanners in resolution as their resolution isn’t determined by how many ‘pixels’ can be arranged in a row, but is in fact determined by how small the size of the sample area can be on that scanner (how small the area being scanned at any single time can be) along with how precisely the scanner can shift the film to scan from one pixel to the next.
Now; where drum scanning becomes a little more complex is the relation between the scanners aperture and the sample size. Drum scanners have a range of apertures; really tiny apertures (measured in microns) and just as in photography a smaller aperture will be sharper than a larger aperture. The real art of drum scanning comes in the selection of the most appropriate aperture in relation to the sample size (the resolution or PPI) the grain of the film, and the resolution of the image rendered by the lens on the film; with the perfect combination it’s possible to scan at a high enough resolution to really record every last detail that the lens has rendered on the film, but have selected an aperture just the right size so not to over-accentuate the film grain (effectively we want to chose an aperture large enough that each sampled area overlaps, but still small enough that the image remains sharp and every detail is recorded). This aspect of drum scanning is what makes scans look so beautifully clean, free of noise or ‘grain-aliasing’ yet still spectacularly sharp and detailed.
So if drum scanners are so good, why aren’t we all using them?
Drum scanners were hugely expensive (and sometimes equally huge themselves) machines that were never designed for easy use. Drum scanners were generally sold to labs or pre-press companies along with a service contract and training. It’s taken me years to fully understand and get the best from these machines. Wet-mounting film; getting no bubbles and a super-tight leak free mount isn’t the hardest thing in the world but is certainly fiddly. The software is far from intuitive and the scanners and software are for the most part unsupported. In short, drum scanners can be an enormous pain, but the results make up for it and more.
The real catch to drum scanners is the time they take to scan – a typical scan can take anything between 30 minutes to 4 hours. This will be producing between a 128-700 million pixel file, but the scanner is effectively scanning just one pixel at a time (it is doing so very fast as the drum spins at up to 1600rpm) but it still takes a lot of time to capture all of this data.
The golden age of drum scanners was in the 80’s when the vast majority of professional photographs were shot on transparency film and scanned on giant drum scanners made by Heidelberg or Crosfield. These scanners allowed a photographic lab or prepress company to fill a giant drum with transparencies, set these scanning and leave the scanner working for a few hours at a time until the next drum was loaded. Typically these were reasonably low-res 8-bit scans made on machines that would auto-apply sharpening and colour adjustments so the images were ready to go straight into the workflow of a printing press. These early drum scanners are fantastic pieces of engineering but are totally unsuitable for modern high-quality drum scanning; they were never designed for precise high quality super-high resolution scans. Luckily towards the end of the 90’s a few scanner makers (Heidelberg, Howtek, Screen, ICG and Scanview) produced a new breed of drum scanner that was generally smaller or designed in a more space-efficiant upright-standing format but more importantly produced 16-bit colour. Crucially these scanners were built with the far more challenging task of getting excellent scans from negative film in mind. While pretty much all of these scanners use old SCSI cables (no need for data transfers faster than SCSI as the scanner is only scanning 1 pixel at a time) and run on OS9 or early iterations of OSX its still hard to argue that these are fantastic bits of kit.
Are drum scanners better quality than flatbed scanners?
Yes, in every single way apart from ease-of-use and speed. Todays Epson or Canon flatbeds are relatively basic mass-produced low-precision machines compared to the truly professional-grade flatbed scanners made by Heidelberg, Scanview, Fuji, Scitex and Screen in the 80’s and 90’s. However even in comparison to these superior flatbed scanners drum scanners have greater resolution, clarity, gamut, sharpness, focus, and see far more detail in the shadows and highlights than any flatbed scanner.
Are drum scanners better quality than specialised film scanners such as those made by Imacon/Hasselblad, Nikon, Noritsu, Fuji and Pakon? Or using a camera to scan film?
Imacon/Hasselblad, Nikon and Pakon have made some truly wonderful specialised film scanners. These scanners are generally fast and easy to use and produce exceptional results. I think a scan of medium format film on an Imacon scanner used by someone who knows how to get the best out of them is spectacular quality considering how quick and easy they are to use. However there are limits to these machines, for example, while a Imacon scanner does a very good job of getting nearly al of the data from a medium format negative the sensor is simply far too low resolution to retrieve all of the data from a 5x4” negative. While these more specialised scanners generally do a far better job than flatbed scanners at holding film critically flat and focusing accurately on the emulsion they suffer from many of the problems inherent to CCD sensors (just as with the flatbed scanner) of just not having the superior clarity, gamut, absolute 3-colours-per-pixel details and shadow/highlight data as a drum scanner can resolve. So while you can make spectacular scans of medium format on a Imacon, or spectacular scans of 35mm in a Nikon or Pakon scanner, a drum scanner will always have the edge in absolute quality.
While flatbed scanner technology stalled about 20 years ago, one technology that is pushed further every year is camera sensor design. Digital camera sensors, be that in SLR’s or camera phones have been pushed to do things we didn’t think would ever seem possible thanks to a highly competitive market. While I’m yet to see someone using a camera to scan in a properly scientific and controlled way I’m actively experimenting using high-resolution Phase One 100 million-pixel CMOS sensors, macro-lenses, flash light boxes and anti-newton glass holders to create linear-tiffs that are characterised with accurate ICC profiles. I’m yet to see a camera come close to what a drum scanner can produce but I look forward to the day when I can say I can make scans just as good on a camera-scanner as my drum scanner.
Why drum scan?
Because you want the absolute best in quality. You want the best in colour-reproduction, clarity, sharpness and resolution. You want a file that pulls all of the data from the film into a digital file so you have an archival surrogate of the film. If you want a single scan so good that you will never need to scan the film ever again.
Because the difference is visible. Because you are interested in your film looking as good as possible. Because you want your film to look like film, not like a standard digital file. Because you want to create an especially large print and need to get all of the data out of your film.
Do you scan at resolutions other than those listed on this site?
I do. But my price list and listed resolution settings are a guide that I advice you not to ignore. I’m interested in making the best possible scans and my philosophy is to make a single scan that is so good that you will never have to make another scan of that image ever again. To provide you with a master file that you can than make multiple output files; from giant print, to web jpeg to image in a book etc…
Quite simply I don’t think its worth my time mounting film in my scanner if I’m only going to make a quick relatively low-res scan. The resolutions I list are carefully chosen after years of testing scanning different films at different resolutions and scanner aperture settings. Quite simply these resolutions are what I consider the optimal resolution for that given format in that they should capture all of the data that the lens projects onto the film, but are not so high resolution to be wasting valuable space on your hard drives.
Do you scan odd sizes of film?
Yes, unlike other scanners that require a custom film mount for different sizes of film I wet-mount my film to the drum, so it can be any size from sub-minature to 10x8”.
Do you drum scan reflective material such as prints?
I can, but chose not to. Drum scanners are exceptional at pulling near-microscopic high-resolution details out of film, rendering the most subtle variations in colour. When reproducing reflected material there is nearly always no reason to need such high resolution capture so instead I reproduce reflective material using a high-resolution digital medium-format camera that is custom ICC characterised producing results within two DeltaE (2000) colour accuracy.
Film is also an especially hardy material that is happy to be wet-mounted on a scanner, while more often than not prints aren’t too happy to be taped round a drum. By photographing reflective material (maybe putting a sheet of low-iron glass over the top if it’s not particularly flat) using a specialised camera and workflow I achieve results just as good as any drum scanner.
Other labs list their scans by the MB, why don't you?
Pricing scans by the MB makes sense when we’re talking about drum scanning as it reflects the fact that a drum scanner could make a 1.5gb file from a 35mm film just as easily as it could make a 1.5gb file from a sheet of 5x4” film. While we know that the 35mm format holds far less detail than a 5x4” sheet of film the drum scanner can still pull out 1.5gb of meaningful data from the much smaller 35mm frame.
However I don’t believe in making scans that are too low-resolution to record all of the data from the lens onto the film or scans that are too large to just be filling up your hard drives with useless data. When I’m thinking about this balancing act I’m effectively thinking about how many pixels per inch it takes to resolve all of the information a superb lens can project onto that given size of film, so for this reason I generally describe my scans in pixels-per-inch (PPI), I do however also list the rough size of the resulting file in MB’s. Notice that while some labs charge extortionate amounts for scans any larger than several hundred MB’s my scans start at around 500mb. Firstly this is because most labs are still using the same mindset as the pre-press companies scanning back in the 80’s thinking that drums should be packed as full as possible with film and scans should be as quick and automated as possible, and secondly it’s because most labs are using the same scanners as they were using in the 80’s with poor quality software and 8-bit-per-chanel colour output.
Do I need 16-bit-per-channel (48-bit) colour scans?
Yes! While early scanners either didn’t have 16-bit output as an option or scanner operators knew that their computers simply wouldn’t be able to handle manipulating the large file sizes their scanners could produces our computers have moved on a fair bit and we can now enjoy all of the benefits of 16-bit scans.
We all know the benefit of working on digital camera raw files and working on 16-bit scans is just the same, it provides us with ALL of the information from the camera sensor for us to then play with, to retouch, to add contrast or work on the colour. Retouching 8-bit files is bad practice and leads to banding.
Most importantly when scanning negative film it’s crucial to understand the way that data is processed to go from negative to positive image. A negative is an inverted extremely tonally-flat image that contains a huge amount of exposure latitude and sadly it’s not at all a case of just inverting your negative (If you have a smartphone trying inverting the screen and switching on the camera, hold up a negative to some light and point your smartphone camera at it – you’ll see exactly that inverting a negative is far more complex than just making the black point the white point and the white point the black point etc..). To cut a long story short inverting a negative digitally involves a huge amount of pushing and pulling of data and I would never do this unless I had a proper 16-bit file.
© Art & Artefact 2017