Tradeoffs Affecting Scanner Performance

Today, Apr 2004 by Cashman, Paul M

1. Introduction

Comparison-shopping for a mattress is a notoriously hard consumer task, since the manufacturers go out of their way to provide a large number of proprietary and whitelabeled products, all at different price points, coil counts, sizes, coverings, etc. Comparing scanners can be just as confusing, unless you know the basic tradeoffs among the key variables that determine scanner performance. This paper describes those variables and the simple arithmetic relations among them.

2. Simplified scanner structure

To begin with, consider the simplified structure of a scanner. A camera is positioned over a moving transport, and documents pass along the transport and go under the camera. The camera takes a picture of each document, and the resulting image is fed to an image-processing board which massages the image by removing speckles, straightening (deskewing) it, cropping the part of the image which doesn't contain the paper image, etc. The image is then output to a file for later processing, such as offline OCR.

The camera consists, at its heart, of an electronic device known as a CCD (charge-coupled device). Every CCD has a number of elements, each of which captures one pixel. As the paper passes under the camera, the CCD "fires" its elements, so each element captures one pixel across the width of the document'. As the transport moves the paper, the next "scan line" of the document comes under the CCD, and the CCD captures that scan line. Each scan line succeeds the previous scan line by one pixel in the length dimension. Clearly, it takes time for the CGD to fire its elements and capture a one-pixelwide scan line. So the transport cannot move the paper too fast, or the CCD will miss some scan lines.

How fast can the transport move without causing the CCD to miss a single scan line? That depends on the optical resolution required. Suppose, for example, that the resolution required is 200 dpi. Each dot is one pixel. So the CCD must be able to make 200 sweeps across the entire width of the document in the time it takes the transport to move the document one inch farther down the track2. If 400 dpi resolution is required, the CCD would either have to scan twice as fast as in the 200 dpi case (at the same track speed), or the transport would have to move only half as fast (at the same CCD scan rate as in the 200 dpi case)3.

3. Field-of-view and optical resolution

CCDs come in a number of different sizes. For example, IBMCs Image Trak has a low-resolution camera with 2096 elements and a high-resolution camera with 4096 elements. Some CCDs have more elements than are used in actual processing. This is useful in cases where the scanner manufacturer wants to allow room for higher-resolution capture than is offered in standard models, without having to redesign the camera. Unused CCD elements do not affect the time the CCD needs to capture a scan line.

How many elements4 are needed in a CCD? That depends on two factors, the field-oi'-view (FoV) and the optical resolution required.

FoV is the maximum width the CCD can "see." This is equivalent to saying that FoV is the maximum number of pixels in the width dimension that the CCD can capture in one scan line. Let's consider a CCD with 4096 elements. At 200 dpi resolution, one scan line can capture (4096 pixels)/(200 pixels per inch) = 20.48 inches of width. At 300 dpi, the same CCD can capture (4096 pixels)/ (300 pixels per inch) = 13.65 inches of width. And at 400 dpi, its FoV is 10.24 inches.

Conversely, suppose we needed to capture at 400 dpi a document that is 11'' wide. That means we would need a CCD with at least (400 pixels per inch) × (11 inches) = 4400 pixels (elements).

So we can sum this up as:

Figure 1 shows the maximum optical resolutions possible, given different sizes of CGDs and different fields-of-view. Clearly, the bigger the CCD (i.e., the more elements used), the greater resolution is possible at any given FoV But this will have an effect on transport speed, as we will see in the next section. There is no free lunch.

4. CCD elements and transport speed

Now that we have the relationship between FoV optical resolution, and number of GGD elements, we can go back to the question in section 2, namely, how fast can the transport run?

Suppose we have a 7500-elemcnt CGD and when we use only 4096 elements, the CGD is capable of performing 10,000 scans/ second. (This is a consequence of its circuitry, and is not related to any other factors we've been discussing.) Let us also assume that 4096 elements give us a sufficient FoV at the resolutions we're interested in.

If 200 dpi resolution is required, then the maximum transport speed is (10,000 scans per second)/(200 scans per inch) = 50 inches per second (ips). Note that we've changed "200 dpi" to "200 scans per inch". This is because we're talking about speed in the length dimension: before the transport can move to the next scan line, the CCD has to complete one full scan in the width dimension. So to get 200 dpi in the length dimension, it has to make 200 scans per inch. We can sum this up as: