In theory, a desktop scanner dedicated to 35mm slides should be easy to
manufacture. In practice, no maker has yet produced one scanner which
combines the best qualities found in specific models.As a brief run-down, here's the history of 35mm scanners. The first widely-used model was devised by Siemens in Germany in 1986, jointly with Hasselblad, and used Zeiss optics with a condensor point-source light path. A large and heavy device, it was marketed as the Barneyscan, the Hasselblad 35 and Dixcel, and the Linotype T35.
This model had several disadvantages - the sensor was just 2048 pixels long, and took three sluggish passes, with dense RGB colour filters and a heavy infra-red cut-out, to acquire a scan. It also needed regular recalibration to prevent uneven results. The use of digital signal processors and direct memory access inside the scanner was very advanced, as was its ability to create downloadable film type look-up tables for any new film which came along. But its rectangular image window meant that portrait-format scans had to be rotated after acquiring, and it only worked well with unmounted slides.
Its greatest disadvantage was the condensor light path, necessary to get the high level of illumination the early CCD sensor required. This produced very high contrast which could easily turn into blocked up shadows, burned out highlights and emphasised grain, dust and scratches.
The Nikon 3500, rapidly followed by the 3510AF, used a sensor twice as long but did not achieve twice the resolution, opting instead to cover a 40mm square so that landscape, portrait, Superslide and correctively rotated originals could all be scanned straight in. A special halogen light source using a condensor-diffuser system gave much better control of contrast, and Nikon optics ensured good quality. Autofocus made scanning mounted slides more reliable.
We are now approaching the 1990s, and the LeafScan LS35 arrived with the highest resolution so far - 4,000 dpi. A discharge tube provided fairly cool and well-balanced low contrast, high intensity light and Wratten filters gave sequential RGB exposure. Unlike the bigger LS45, the 35 did not have autofocus but used strip and mount carriers to locate the film plane accurately. With a Rodenstock Apo Rodagon 1:1 75mm f4 taking lens to ensure sharpness, it offered far more 'bit depth' and remains the first desktop slide scanner to penetrate shadow detail properly.
The Microtek 1850S arrived at the same time at a fraction of the price of the Nikon or Leaf. With a fluorescent lamp and lower resolution of 1850 dpi, partly as a result of covering a 36mm square instead of a 24mm strip, it was only 8-bit and showed very little detail in dense slides. However, it was by no means poor and could match the Hasselblad with normally exposed originals.
The 1850S was replaced by the similar three-pass Microtek 35t, which offered 1800 dpi interpolated to 3,600 dpi, a filmstrip capability, and improved shadow detail. The Nikon range was expanded with the drive-bay sized Coolscan, at a similar price to the 35t, which was very compact but proved extremely slow - and despite its cold light source did not give any better shadow detail.
The Kodak RFS2035, aimed at Photo-CD production labs, showed that scanners with this kind of 8-bit quality could be made much faster. It did in seconds what the Microtek and Coolscan took minutes to tackle. As Photo-CD users found, it shared their inability to record concurrent shadow and highlight detail in long range slides like Velvia or Panther.
Then, in 1994, Kodak enhanced the density range of this unit. Polaroid announced the SprintScan 35 re-engineered from the Microtek 35 with a higher resolution sensor and new electronics. In 1995, Nikon introduced the Super Coolscan with a giant leap in both scanning speed and density range - Minolta launched the QuickScan 35 which outpaced everyone but failed in our tests to yield acceptable shadow detail from slides - Nikon upgraded the regular Coolscan to model II - and Polaroid announced the 35ES version of the SprintScan, entirely based on a software upgrade.
When you consider there are no other significant players in this entire market, this brief history is underpopulated in the extreme. The sheer quantity of 35mm film shot or in existence makes it a huge market. Despite this Canon, Umax, Epson, Fuji and others have steered well clear of it to date.
Today, you can still buy a Hasselblad scanner or a LeafScan 35 at surprisingly good new prices, and though Leaf's distributors believe that no LeafScan 35 has ever been installed in the Britain (everyone buys the 5 x 4 model) I suspect this does represent the best dedicated 35mm scanning money can buy. I say this because the slower, clumsier Leaf 5 x 4 model outperforms all dedicated 35mm scanners for tonal range and colour accuracy and the 35mm model is simply a very much more efficient single-format version of it.
The Kodak scanner is also surprisingly affordable, and very fast in use. In practice, however, most photographers will choose between Nikon Coolscan II or Super, Polaroid SprintScan 35ES, and the Microtek 35t. These have a street price range of around �700 (Microtek) to �1,900 (Polaroid) with the two Nikon models about equally spaced in the gap.
Discovering that Microtek scanners were not as widely available as I had thought was a surprise. We've used them with great results ever since they were the first colour scanners made - from the original 300ZS onwards. As there's no 35mm neg and slide scanner made by Umax or Epson, or coming close to the 35t in price, it's surprising that every major computer dealer doesn't stock this item.
Left: removing
the neat front cover reveals not only the carrier mechanism, but a really
neat way of storing a spare flourescent lamp (supplied).Our task was to see whether early problems with the Polaroid, which could produce unpleasant colours too readily, had been solved with the new software release.
The software is comprehensive and excellently written as Photoshop plug-ins go. A large preview window is accompanied by a small, separate control window. Reasonably logical and intuitive entry boxes to set percentage magnification or final image size and resolution work almost the same way a designer does, while pull-down menus hide a wealth of additional controls.
The preview is in live colour. If you make adjustments - with the white, black or neutral point tools to over-ride auto density and colour balance setting - you see them on screen before scanning. You can even see the effect of changing from one film type to another (say from 'Fujichrome' to 'Pro Fujichrome') unless the difference is too great, in which case you need to do a new pre-scan.
Even sharpening, which includes a very useful setting for smoothing out grain and dust but sharpening edge contrast, is visible in the preview window and handled on the fly during scanning. It is also fully adjustable, just like Photoshop unsharp masking.
In fact, the main Photoshop post-processing controls - gamma, levels, colour saturation, sharpening or softening, exposure, white point, black point, neutral balance - are all available in the plug-in for real-time application as the scan is acquired. Nor does their use appear to slow down the process, with a full page cover pictures typically being grabbed in around 90 seconds. Most scans seem to take around 30 seconds to preview including 'getting started' time, a few seconds to adjust and crop or scale, then as little as 15 seconds to scan plus the same to save.
Unless dust or scratches need removing, these ready balanced and sharpened scans can be saved and placed for printing without any Photoshop work. It would be reasonable to assume a productivity of around 30 scans an hour, and we did achieve this when tackling typical magazine pages.
The results
Despite the provision of a long list of film types in the Polaroid plug-in
menu, many are missing. There's no Velvia or Panther; no distinction is
made between Kodachrome 200 and other Kodachromes, despite its entirely
different colour and density; all Ektachromes share one generic type - and
so on.
Colour negatives (slightly easier to scan well because of their lower density) fare much better. In practice, we found that the preconfigured slide film types simply didn't work properly. A sheet of perfect Kodachrome 25s would yield awful over-saturated colours with emerald greens if set to the Kodachrome default, but perfect neutrality using the Polaroid Chrome setting! One Fujichrome would be well rendered using the Fujichrome default, the next would have an unpleasant colour cast and need setting to Ektachrome, while that setting gave all Ektachromes dirty yellow whites.
The film settings are not always ideal: top left; the Ektachrome
setting with Ektachrome 64, top right, with Ektachrome 50 (amatuer);
center left, with Ektachrome 200; center right, with Panther 100.
Bottom
left, the Fujichrome Pro setting used with Provia 100 - right, used with
Fujichrome 400 II
The film types appear to include in their instructions to the scanner minimum and maximum density, gamma and colour saturation variations. The most common errors were in posterising the shadows, losing shadows to solid black, and boosting colour saturation. Checking the pull-down menu after an 'auto' prescan would reveal the saturation sliders all turned up to 14, when a far superior and more natural result could be obtained with them zeroed.
Unfortunately for this scanner, the Raw Chrome setting did not do as expected and turn on maximum d-min to d-max range, zero saturation change, neutral colour and neutral gamma. It yielded results showing more clipping or colour shift than any randonly picked film type like Ektachrome.
Most worrying was that an inexperienced operator, opening slide mounts, often found only a film type number (especially on Kodak films) and no word saying 'Kodachrome' or whatever. The list of film types encouraged opening plain mounts to try to find what the film was - and that's something to be discouraged. The LeafScan needs no such action as it treats all slides to a maximum range prescan, then sets the range according to prescan's values.
The worst SprintScan results were very poor, with strong green or cyan casts and excessive saturation combined with dense shadows and dirty highlights. The best results, either fortuitously or by hard work over-riding the scanner's decisions, were really excellent. I would class myself after six years of scanning trannies to be reasonably expert, but I still found some originals which I could not persuade the SprintScan to handle well with any of its preset film types, and had trouble adjusting manually.
Any other complaints are limited to a handful: some slide mounts did not sit well in the spring-gripping carrier slot, and ended up a degree or so off square; the scanner is fixed focus and tended to be sharper at the top of the slide than the bottom on full page scans; and when switched on, it made an annoying undulating hum, barely audible but ever-present. (continues on page 42)
Finally, the 2,700 dpi maximum resolution would not manage a true 300 dpi bleed A4 reproduction. The actual scan area is restricted by the mount or carrier and the highest resolution I could achieve on a 303 x 213mm scan from 35mm was 283 dpi.
My conclusion is that the SprintScan is capable of results which are superior to the Minolta QuickScan, Nikon Super Coolscan or II, and Microtek 35t. Its best shadow detail and contrast control outperformed all these. Yet the very clever software, designed to ensure this benefit of 10-bit depth in the hands of the new user, consistently shot itself in the foot.
This is exactly what the old software used to do with even greater panache. Readers of our contemporary (rival?) The Photographer will recall the famous 'trip to Mars' issue in which the then president of the British Institute of Professional Photography, Gil Cox, had recorded an official visit to China on film. Every single picture was reproduced with a virulent green overtone and, much to Polaroid's discomfort, proudly attributed to the use of a SprintScan.
While it could be argued that a well-adjusted computer monitor will prevent such accidents, I find that the eye is very accommodating. Only by direct comparison with other files was I able to spot a similar set of scans before they went to print in our Minolta Image magazine here at Icon - and even after that, my adjusted versions were far from perfect.
Polaroid has the best colour scientists around (OK, maybe Scitex and Kodak would disagree!) and they really should be able to devise a better solution. The old Hasselblad scanner method of scanning a calibration slide including totally clear film and solid d-max is a much better route to 10 or 12-bit accuracy. It calls for a little photographic know-how, but if the user could just create and save named maximum range calibrations like this for present and future films, the scanner would be proof against mismatches.
I have tried DCR (Microtek) and Agfa FotoTune calibration using these target slides, and invariably ended up with a parody of accuracy, colours being exaggerated or suppressed in different parts of spectrum. Apart from this, the Photoshop post-processing needed is tediously slow. My opinion of ColorSync 2.0 and Kodak CMS transformations is little different.
The 35t, in fact, seems to have one advantage over the Polaroid - it can prescan raw, unadjusted, full range scans. Their new ScanWizard plug-in is as much an upgrade to the original 35t as the new Polaroid plug-in is to the SprintScan. Its live preview is pitifully small, and its visible adjustments are far from accurate - the scan never actually looks like the preview - but it offers a whole set of on-the-fly controls similar to Polaroid's. The key difference is that if you apply these, scan times are slowed right down because the host computer does all the work. SprintScan's built-in DSPs take that load instead.
Because the 35t never throws up a violent cast or contrast problem, it is actually far safer to use. But, because it also does not accept preset film types, its negative scans are nowhere near as good. It is in scanning negs that these special look-up tables really start working.
The main limitation of the 35t is that it only scans up to a half-page litho reproduction, and yields barely enough data to make a good 10 x 8 dye sub print from a neg, without interpolation. Still, it will interpolate up to 3,650 dpi, which makes it able to produce a 300 dpi full bleed A4 scan - which the SprintScan can not.
Don't be deceived, however. The SprintScan at 2,700 dpi is much sharper - down to grain level. There is little real point in letting the Microtek interpolate data, as a clean scan at maximum optical resolution will look just as good and print faster.
If you are entirely new to the field, work with uncalibrated systems and have no way of seeing a final print without paying a lab or bureau for output, approach the SprintScan with respect and caution. It has been given a mind of its own, and has ended up with artificial stupidity rather than artifical intelligence. The controls for over-riding its enthusiasm are complex for the beginner to master.
In short, a fine bit of hardware with a plug-in that benefits from experienced control but tempts you to leave it to its own devices. Understand this before you buy, and you will probably never make that unwanted trip to Mars.
-David Kilpatrick
Return to Photon July 1996 Contents