Selecting A Flatbed Scanner

Picking a scanner can be a tough job. Find out from a prepress specialist how to sort through the different categories and specifications.

The opening years of the new millennium offer a remarkable array of scanners to choose from. Consequently, making a decision—much less gathering all the information necessary—can be a tedious and frustrating exercise.

Scanners can be more or less grouped into three categories: high-end, midrange and entry level. The boundaries between these categories are blurry, and exceptions abound. Still, they will help bring some order to the scanner selection process.

Also keep in mind that drum scanners are often rated independently of flatbeds, and that desktop "film and transparency" scanners are frequently treated as a separate category. There are also the document and halftone film scanner categories for creating digitized workflows from old hard copy or analog jobs—like the CREO Renaissance II copydot scanning system. These specialty scanners will have to be left for another day and another discussion.

Rotary vs. Flatbed

Rotary drum scanners have dominated high-end color markets in spite of challenges from flatbeds since the mid 1990s. But because flatbed scanners are generally considered more user friendly, require shorter learning curves and usually cost less than comparable drum scanners, they have been selling faster.

Before looking at the categories, let's consider some of the risks involved in relying too heavily on hardware specifications as a way of selecting a scanner.

The Problem With Scanner Specs

Regardless of your particular circumstances, technical specifications are going to be an important factor in choosing a scanner. Among the important specifications are things like:

1) The number of elements in the CCD array.

2) Dynamic range.

3) The size of the scanner bed.

4) The color channel bit depth.

5) The pixel capture resolution.

These specifications seem straightforward enough, and just a few years ago they could be used to neatly categorize scanners into low, midrange and high ends—along with price, the bundled software and a few other things. It's not quite as easy today, as some low-priced scanners now rival far more expensive machines as far as the published specifications are concerned.

A few savvy professionals have suggested that some manufacturers intentionally muddy the waters by inflating their specifications, further complicating the selection process. (Since the mid 1990s, and no doubt as a result of criticism, many manufacturers seem to be more careful about how they state their specifications.) Let's look at just a couple specifications to see how terminology issues can further complicate an already challenging selection process.

Dynamic Range vs. Maximum Density

Dynamic range (also called density range) translates into the ability of a scanner to capture detail in the deep shadows of an original. This is not such a big concern if all you're scanning are photo prints for pleasing color, but it's critical for transparencies and high-quality printing. It is therefore an important specification; there is an enormous difference between a dynamic range of 3.0, for instance, and one of 3.8. Like most important quality-related specifications, it is the result of many contributing factors, such as CCD sensitivity, bit depth, level of unwanted system noise and the quality and resolving power of the scanner's entire optical system.

According to Robert Gann, a scanner design engineer for Hewlett-Packard and author of Desktop Scanners: Image Quality Evaluation, the dynamic range specs given by some manufacturers do not accurately reflect true performance. For instance these numbers often do not take into account system noise, which can significantly reduce dynamic range. In other words, the numbers you read on some scanner spec sheets may not reflect that machine's actual performance capability.

Dynamic range (DR) is also complicated by the fact that some manufacturers don't actually give the 'dynamic range' specification, but rather refer to what they call "maximum density," which is going to be a couple points higher than DR.

For instance, to give a good example of how specifications should be published, a Heidelberg site states that its older high-end Topaz model has a "maximum density" of 4.0, while its "resolvable density range" is 3.7, a significant 0.3 difference. "Resolvable" here suggests that it is actually achievable in production. Some marketing literature mentions only 'maximum density,' presumably in the hopes that potential buyers will read it as density 'range' due to ignorance or inattention.

Resolution vs. Resolution

Pixel capture resolution is another specification that some manufacturers do a poor job of communicating accurately. (Resolution here is not to be confused with the resolving power or quality of the scanner's optical system.)

Pixel resolution represents the number of pixels per inch (or centimeter) that can be physically sampled from an original, and is usually called 'optical' resolution. It is directly tied to the number of CCD elements resident in the scanner. Being one of those specs that directly affects image capture quality, it is particularly important for some line art workflows, as well as for extreme enlargements of small-format transparencies.

A further complicating factor is that many scanners are capable of delivering on their published optical resolution only at specified maximum copy sizes, causing optical resolution to steadily decrease as the size of the originals increases. The new high-end X-Y scanners are not restricted in this way.

Another factor that complicates scanner resolution has to do with the fact that software can enhance resolution through interpolation. In other words, the scanner software can create a higher resolution than it is physically capable of creating. It does this by creating additional pixels based on the information gleaned from already sampled pixels.

The problem is that the numbers don't represent actual sampled pixels. In the past, advertising literature sometimes didn't make this critical distinction. This isn't to say that interpolating on the scan is not a wonderful feature for some workflows.

A less critical yet potentially confusing practice for the uninitiated is the use of such acronyms as dpi, spi and ppi to refer to optical resolution (ie. dots per inch, samples per inch and pixels per inch, respectively).

Also misleading is the practice of designating optical resolution by listing both the 'x' and 'y' dimension resolutions of the scanner. The across-the-bed 'x' dimension resolution is limited by the number of elements in the CCD array. The lengthwise 'y' dimension resolution is usually a function of the stepping motor and necessarily involves pixel creation by interpolation, and thus does not represent true optical resolution. Therefore, it is best to ignore the higher number altogether whenever there's a difference, since Macintosh pixels are square and the operator cannot, therefore, take advantage of anything greater than the lower of the two numbers.

Another area of confusion is the use of different terminology to describe the same specification. For example, one manufacturer refers to color bit depth as "signal resolution." Similarly, another refers to its optical resolution as "hardware resolution," while referring to what is usually called interpolated resolution as "scanner resolution." All these factors combined make it more likely that the average buyer will be somewhat muddled.

The fact that most of these specifications are important selection criteria doesn't help matters when the uninitiated buyer finds similar sounding specifications appearing on spec sheets for entry level, midrange and high end scanners. This is partly due to inconsistent terminology, but it is also due to the fact that some specifications have become very similar across all price categories in the past few years.

For instance, it doesn't cost a manufacturer much to incorporate software enhancements (initially intended for high-end machines) into entry-level or midrange models (ie. pixel depth). In a similar fashion large "interpolated" resolutions—such as 7,200 ppi on the entry-level LinoScan 1200 , or 9,200 ppi on the PowerLook III ($1800)—can make it seem as though a $1,200 or an $1,800 scanner is more or less equivalent to a $15,000 or even a $50,000 scanner.

Entry-level Flatbed Scanners

There is, of course, a very low-end category of scanners for the home and office computing markets—often referred to as "consumer market" scanners. Pricing for these models, which is usually for reflective copy only, ranges from less than $100 to several hundred dollars. Although I have one on my desk and use it regularly for instructional purposes, these are not intended for professional print media.

For the purposes of this review we'll begin with the low end of what are typically considered midrange scanners; those considered entry level for the graphic arts industry. These range from as low as $700 up to $1,500 or $2,000, and are often used by graphic designers, within graphic arts educational programs and in some print shops that service less-demanding markets. (These price ranges are very rough estimates based on current marketing information. Pricing and specifications are rapidly moving targets.)

In the past, scanners in this entry-level category handled only reflective copy and had noticeably lower bit depth, interpolated resolutions, and software capabilities than those in the more expensive midrange. But these limitations have now been passed down to the consumer line of scanners. Entry-level models now boast some specifications that rival models with price tags 10 or even 20 times greater—hence the reduced significance of some specs, such as bit depth and interpolated resolution.

Nearly all scanners in this price range now handle reflective and film originals, and have optical resolutions of 1,200, some going as high as 2,000 or 2,400 ppi, with impressively higher interpolated resolutions. Examples representing this category are Agfa's DuoScan 1200 ($700), Hewlett-Packard's ScanJet 4c ($900), Heidelberg's Linoscan 1400 ($1300) and Microtek's ScanMaker III ($1,500). All prices listed are as of April 2001 and represent manufacturer's recommended pricing rounded to the nearest $100. (Special thanks to Midwest Sales Manager Paul Caruso of London Litho for current scanner pricing. Contact him at 800-949-4721.)

Although dynamic range (DR) is less of a factor than in previous years, it can still be a useful indicator when trying to distinguish between scanner performances. For example, the entry-level Heidelberg LinoScan 1400 reports a 3.4 maximum density (the DR being closer to 3.0), while the Agfa Arcus comes in at 3.2 DR. On the other hand, the ScanMaker III and Duoscan models claim a 3.4 DR. Needless to say, most of the more expensive midrange scanners, and all of the high-end machines will now exceed these numbers.

Pricing is by far the biggest indicator of positioning because it directly reflects the quality of the hardware (i.e. the CCD chip, light source, optical system, stepping motor, etc.). These have an enormous impact on image resolution quality and system noise levels, according to Hewlett-Packard engineer Robert Gann. Entry-level scanners may also come with more limited color control and fewer productivity enhancements via bundled software—the trade-off for price savings being ease of operation and decent color for those whose background in color is somewhat limited.

Midrange Flatbed Scanners

Full-fledged, midrange flatbed scanners, for the purposes of this article, fall roughly between $3,000 and around $10,000. (The considerable gaps between the three categories are necessary to allow for inevitable overlapping and exceptions to the rule.) Examples of scanners in this category are the Microtek Artix AS2500 ($4500), the Agfa DuoScan T2500 ($5,000), the DuoScan T2000 XL ($9,000), the Heidelberg LinoScan F2400 XL ($8,000), and the U-Max Powerlook 3000 ($7,700).

In many respects, these scanners share the same specifications as the significantly higher-priced models, as well as with the lower priced entry-level scanners. The difference between an entry-level and a beefed-up midrange scanner, or between a midrange and a high-end model, will not necessarily be seen unless an operator knows something about scanning, workflows and the subtleties of color reproduction.

Needless to say, that knowledge is of more limited value on an entry-level scanner, while in the case of a high-end model and the demands of hypercritical color or high-volume scanning it could be worth a king's ransom.

One of the advantages of most midrange scanners is the added value related to controlling color quality and productivity that an operator can potentially take advantage of—but here again only if he or she knows something about color. No matter how good the operator, though, there will be potential differences in quality and productivity for certain types of work done on midrange versus high-end models.

High-end Flatbed Scanners

High-end models range from around $15,000 to upwards of $50,000, though scanners between $15,000 and $25,000 can hardly be expected to compare favorably with those coming in at $50,000. Until the mid 1990s, this category was occupied exclusively by rotary drum machines.

The high-end flatbed challenge began with the Linotype-Hell Topaz and the Scitex Smart line of scanners, which have been superseded by the Heidelberg NexScan ($37,500 to $43,000) and the CreoScitex EverSmart Pro ($40,000) respectively. The latter models utilize the new X-Y image capture technology, which is exclusive to the high end. They also benefit from other innovations, such as advancements in CCD arrays and editing software, enhanced dynamic range, dynamic thermoelectric cooling systems for noise reduction, multiple pixel sampling, and improved color management.

These enhancements have truly put flatbed scanning in the high end (with some skeptics continuing to resist, of course). Other flatbed scanners in this league are Agfa's XY-15 Plus ($22,500), Fuji's Lanovia C-550 Sprint ($39,500), Screen's Cézanne Elite ($36,400), and Purup-Eskofot's F10 ($26,300)—formerly of the ScanView product line. ScanView and Imacon also produce a high-end flatbed scanner.

The September 8, 2000, Seybold Report on Publishing Systems featured the results of a study called "Pixel Perfect II: The Scanners," which reviews top-of-the-line professional graphic arts scanners. A book of printed samples can also be purchased at the Seybold Publications Web site (www.seyboldreport.com). Needless to say, printed samples with scanning speeds and other test results are valuable resources in the selection process.

It is interesting to note that few of the judges in Pixel Perfect, all of whom are considered highly qualified, agreed in their rankings. This shows how subjective color and quality are, and how difficult they are to quantify.

New high-end X-Y scanners can achieve the optical resolution they claim anywhere on the scanner bed and for all sizes of originals. On the other hand, midrange and entry-level models can achieve optimum optical resolution only on a relatively narrow band, called the sweet spot, that runs down the center of the scanner bed. High-end X-Y scanners also enjoy more direct light paths that further ensure the integrity of image quality and detail resolution.

So Where To Begin?

The scanner decision should ultimately boil down to what is actually needed—or will be needed—in any given production workflow. These needs should be articulated at the beginning of the selection process. The bottom line in purchasing a scanner is what it has always been: You get what you pay for. In other words, there is no way that a $1,000 scanner can compare favorably with an $8,000 scanner, much less a $40,000 model.

Narrowing your focus to one of the three categories is a reasonably sane way to begin. At that point comparing the more important specifications can further narrow the field—keeping in mind the risks and ambiguities of published specs. Since software can have an enormous impact on image quality and workflow productivity, as well as on learning curves, examine it carefully. Opinions of colleagues and others in the field are often helpful, as well.

After your options have been narrowed within a given category, it is generally considered good practice to consult actual users doing similar work. Visiting a plant to see how a particular scanner performs in a production environment is preferable to trusting in marketing samples and relying on second- or third-hand information. But even this can be misleading if the skill and knowledge levels within your company are significantly better or worse than those in the company you're visiting.

The investment of this much time and attention can hardly be justified if an entry-level scanner is going to fit the bill. But if your in-plant serves a clientele that demands hyper-critical color, or workflows that require a high volume of scanned images, then this kind of front-end research and planning is indispensable.

-by Richard Harmsen

Richard Harmsen is a faculty member in the Department of Printing and Imaging Technology Management at Ferris State University. Since 1989 Rick has been teaching digital scanning and imaging, photography, and quality management. Contact him at: H_Harmsen@ferris.edu&012;

Thanks to Danny Cox, president of Bay Digital in San Francisco, for pricing on Screen's Cézanne, and to Robert Fine of Systems Automation of Atlanta for pricing on the Agfa XY-15 Plus. Pricing for all other models compliments of Midwest Sales Manager Paul Caruso of London Litho.