Polaroid SprintScan 35 Film Scanner

I have only the privilege of an ordinary user for the scanner, so I know nothing about the installation. The model I used is the original SprintScan 35. From a page at Polaroid (see link above), several other models of this scanner have been introduced ever since, and pictures and the specifications therein seem to indicate that the model is currently designated as the SprintScan 35/SE. I cannot find the negative carrier, so I never used it to scan a negative. Anyway, most of my images are scanned by this scanner.

Specifications

The scanner captures the image in 10 bits per RGB color and outputs 8 bits each, optical density range is 0 to 3.0 (a.k.a. a dynamic range of 3). The light source is a 4W fluorescent tube. The scanner I used is connected to a Mac, and the main image processing program is Adobe's PhotoShop. The scanner driver is provided by Polaroid as a "plug-in" for PhotoShop.

First Impressions

The scanning process is straightforwardly one-pass. You first preview the image, which has a fixed resolution stored in the scanner's memory. Crop, set up image quality controls, then do the final scan. The scanning speed: about 50 second (not counting the preview) for a highest resolution of 2700dpi (dot-per-inch) and 15 to 20 seconds for 1012dpi. I am quite satisfied with the speed.

There are several controls for image quality:

Resolution: The highest resolution of the scanner is 2700dpi. There are few discrete steps called "scanner's native resolutions" can be used for the "best results": 2025, 1350, 1012, 675, 506, 337, 253, 150. I think this means using pure optical resolutions. Of course, any resolutions below the maximum can be specified via the software.
Filtration: Filtration can be altered in several ways. The first is determined by film type: the software has a predetermined set of filtration built-in. In the second, the filtration can be changed by directly setting the RGB scale for each color channel. Third, by the saturation of each color channel, which can also be optionally linked together to change the saturation without changing the color tone.
Brightness/Contrast/Gamma: The software provides controls for both brightness, contrast, gamma and the exposure. They are actually the same thing. You can bundle all 3 color channels together and change the brightness and contrast, or change it for each individual color.
Dummy Modes: The software's "auto-exposure" used the latter approach and set things up for you based on the preview image. There is also a way in which you define a white and a black color in the image then the scanner adjusts the exposure automatically.
User-Defined Modes: There is a provision for user defined mode for image quality controls, through a setting files. At any point, one can output the current setting to a file, and later import it. There is also a setting that the scanning software will automatically load the custom setting used the previous scan.

In general, I feel the scanner is easy to use, fast, more than sufficient for web-related and newspaper publication needs. I have an image scanned at 1350dpi (B&W) and printed on an MIT's newspaper (The paper printer printed the image in a 800dpi laser printer then used the laser printer's output to make the printing plates for the paper). The image printed on the newspaper looks very decent, although a coarser than those reproduced directly from real photos.

Some Complains

The user's manual is far too simplistic in explaining what does what. It takes a while to figure things out. The technical writers at Polaroid assume everyone who is using their scanners is a dummy, I guess. I wish they know that it takes more than just a dummy to figure out another dummy. :-)
The build-in set of filtration, for specific brands and types of films, seems irrelevant to the designated film types. (The film-type specification is only used for "auto-exposure".) Although the filtration for different films are noticeably different, none of them gives a overall color cast that is close to the original. For Fuji films I used, I found that the "Agfachrome" filtration gives the best results. (The "Raw chrome" actually gives the closest color cast of the image, but not enough color saturation.)
The "auto-exposure", although doing a decent job in adjusting the brightness and contrast, also changes the filtration. So a batch of slides taken in the same lighting condition could end up with very different color casts. This is a very disappointing factor for this otherwise useful feature.
The insertion of the slide into the scanner is a bit tricky, or fiddly to say the least. Basically there is no slide guiding rails or the like. When feeding a slide you can only feel whether it is aligned when it reaches the bottom of the slot. In the landscape mode, the edge of slide occasionally has a yellow strip if the slide is placed very slightly off center, and yet the entire slot has only 0.5mm play for a mounted slide.
Occasionally there are some rainbow-like strips appear vertically across the image. It appears both in the preview and in the final scan, regardless of the resolution. According to Polaroid, this is due to dusts getting into the CCD sensors. But the scanner simply has too many orifices to prevent this in a dusty environment.

This is the only slide scanner I have, in some sense, access to. I'd better stop bad- mouthing it. After all, I am happy just for being able to scan slides.

A Well-Intended But Mis-Programmed Mechanism

Polaroid claims a 3.0 dynamic range of the scanner. I cannot find the exact definitions for the terms dynamic range or optical density range. (Someone informed me that this is a well defined photographic term, see, for example, Ansel Adams: The Negative.) My reverse engineering shows me that 10 bits per color gives 1024 grades, which is about 10 to the 3.0 power. That is, 3.0 = log ( 2¹⁰ ), where log is the logarithm of base 10, or the common logarithm. In any event, I still believe there are some relations between the "dynamic range" and the "bits per color".

Theoretically this is sufficient for almost all computer displays, which are typically 8 bits per color (and called 24-bit video, or 16 million colors). But my experience is that most scanning I've done do not contain as fuller as possible a dynamic range: the shadow details are not sufficient, and the image manipulations afterwards tend to worsen the situation as shadow details turn out in wrong colors. Occasionally, even the preview image shows the dark area in funny colors. In such cases, I think the best we can do is to take as large a dynamic range the scanner can offer during the scanning, then fine-tune the image in post-scanning image processing.

From this perspective, I think the white-point and black-point mechanism would ideally serve this purpose. However, in the actual use, the selection of white-point and the black-point is done on the preview image. It is a pity that the preview image has a very low resolution and cannot be enlarged to a size that is comfortable for working with pixels. It becomes very tricky to pick the correct pixel, and in the event the chosen point is not the brightest or the darkest point of the image, the image look weird.

Even I am lucky enough to have picked the correct (desired) point, guess what: the color cast can become very different. The software has a provision to maintain the color cast. However, I believe Polaroid software engineers have not had their minds straight. The software works this way: if the picked dark point has the RGB triplet as (10,10,15), the program subtracts 10 from each to give (0,0,5). So the dark cyan becomes dark pure blue. This might not be significant since the point is declared "black", but if the same method is used to scale all other colors, it is simply impossible to maintain the color balance. I think the correct way should subtract the each element in a proportional way to make it (2,2,3).

In any event, the white-point/black-point mechanism is a well-intended mechanism but practically useless as it is currently implemented. Using the programs "auto-exposure" usually is the best compromise between the speed and the quality of results. But, as mentioned before, the color cast again becomes an issue.

Undocumented Goody and Nightmare

I used the "auto-exposure" almost exclusively for quite a period of time, and found an undocumented feature: the auto-exposure is actually related to the selected area of the preview image. The selected area is the portion of the images that will be scanned. Afterwards, the selected area can be changed before final scan. This feature allows me to play with the image quality adjustment through "auto-exposure" and choose the best one.

The "auto-exposure" changes the color tone. Luckily, this information is displayed, so I can keep an watching eye on this issue, and correct the color if necessary. On the other hand, I found that is it almost impossible to obtain consistent results. This is the main reason that turned me away from using the "auto-exposure".

Actually, what I really want is to have faithful representations of my slides. All I want is to have the scanner scans the slide according to a set of predefined parameters, and I later use an image editing software to do a procedural fine-tuning. I had imagined that this would be a humble and actually quite a mechanical procedure.

With this scanner, things actually turned quite nasty. Even with a custom setting predefined in a file, scans still turn out different colorcasts. I suspect the reason of this hopeless situation is its calibration process. The scanner uses a cheap fluorescent tube as the light source. The designer of the scanner is aware of the quality (of the lack of) of such tubes, and thus the color calibration is done in every scanning. This calibration becomes a nightmare if one wants to obtain consistent results.

The Way I Do My Scanning

After using this scanner for almost a year, I have finally settled with the following way of doing my scanning:

I always use the portrait mode. This avoids placing the slide outside the CCD sensor array and creating a yellow strip at the edges. I only use the center portion of the CCD arrays. Doing a image rotation is a piece of cake in image editing stage.
I define a set of custom settings based on a set of well-exposed slides. One for high contrast, one of low contrast, and one for normal contrast. In this set, I have all the color filtration nullified, and the only thing that is different in settings is the gamma value. (The different gamma values enable me to capture as wide as possible as a dynamic range of the slide.) This makes the color correction in the later stage of image processing much easier.
As an added benefit of using the custom setting, most of the time I can skip the preview by letting the scanning software keep using the same setting file from the previous scan. All I need is to ensure the proper insertion of the slides. This increases the throughput of the scanning job quite a bit, and pushes the job toward the mechanical procedure: yeah, I am a mechanical engineering major :-).
I am trying my best to stick with one particular computer to do the post-scanning image processing. Different monitors have vastly different color displaying characteristics.
I save the image in JPG format (factor of 8 in PhotoShop), and I keep the original scan. (JPEG image shouldn't be opened and then resaved. Every saving in JPEG format degrades the quality.) After a year, I just realize how much disk space all those photos, especially those high-resolution scans, take!

Conclusions

Easy to use, fairly fast, and gives very good quality image. But it could be really hard to obtain consistent results for serial shots with same color tone. More than enough for web publishing needs. If printing (using a high-quality photo printer) is the final goal, I think the dynamic range of this scanner is minimally usable: it would do a fine job for a properly exposed not-too-high-contrast image. Otherwise, it would be tough to get a satisfactory print.

If I am in the market for a scanner for myself, I'd probably look for a higher model in the line: Polaroid SprintScan 35 Plus, a 12-bit-per-color model, or Nikon's equivalent Super CoolScan 1000.

Back to my Digital Darkroom Page
First written: May 13, 1997. Last updated: March 12, 1998.