Once More, DPI is not a Quality Measure

In product briefings, HP’s product managers are loathe to enter into “dpi wars”. The Color LaserJet 4500 is a 600-dpi printer. Other color lasers on the market print at 1200 dpi. HP now calls its image enhancement technology ImageREt 2400, hoping that, through innuendo, it is not left behind in specmanship battles. Since innuendo may not be enough, HP is sometimes more specific in its product collateral. A headline in one of HP’s product brochures says, “HP Color LaserJet 4500 Series printers give you fast, hassle-free, 2400-dpi-class color laser print quality.”

At a product introduction briefing session, Norman Burningham, an Imaging Scientist in the Advanced Technology Section of HP’s LaserJet Solutions Group, provided a brief explanation of why HP is reluctant to use dots per inch to describe its printers. A principal problem, it seems, is that although the dots-per-inch spec does describe printers, it does not describe prints, at least when one is discussing print quality. “There is zero correlation between dpi and perceived quality,” said Burningham. “DPI is not on the image.” Burningham provided the following definition of image quality: “Image quality is the weighted combination of all the visually significant attributes of an image when considered in its marketplace or application.” He explained, “Image quality is defined by characteristics of the image, and it is defined by all of the characteristics of the image.” In a 40-minute color technology briefing in which Burningham shared the podium with two other speakers, he could do little more with the complex topic than to provide a basic definition, and list a few of the measurable visual image characteristics (see chart). It is important that the characteristics be measurable, even if print quality is subjective, because programs to improve print quality have to be able to “track progress, identify weaknesses, and focus on them,” said Burningham.

Examples of Visual Image Characteristics

Despite the subjective nature, it is possible to obtain data and measurements of perceived quality, using psychometric scaling. Quality performance should be judged over a set of images, since intended use (marketplace or application) is one of the variables mentioned in Burningham’s definition. When one is designing experiments, Burningham recommends that one “create realistic documents—avoid ‘junk’ images.” Burningham had a particular “junk” test image often used by a consulting company in mind, but he was too diplomatic to name names. Burningham provided an example of how participants in a print quality scaling study judged the performance of the HP Color LaserJet 4500, considering five test prints. The Color LaserJet 4500 out-performed all five printers (none identified) on three of the five documents reviewed by the study.

When all is said and done, the spec sheet for HP’s Color LaserJet 4500 does, indeed, include a resolution specification: 600 by 600 dpi, with HP Image Resolution Enhancement Technology (ImageREt 2400). An HP Technology Brief explained, “ImageREt 2400 provides 2400-dpi-color-laser-class quality through a multi-level printing process.” HP can blend to four colors on a single spot or pixel, and can vary the amount of toner of each color on the spot. “In contrast, the single-level printing process found in the default mode of other color lasers does not allow the colors to be mixed within a single dot. Thus, the printer is limited to creating colors by placing various colored dots (limited to eight colors) close together within a cell to create the illusion of another color. This process, called dithering, significantly hinders the ability to create a side range of colors without reduced sharpness or visible dot structure.” The result of ImageREt2400 is that HP avoids image artifacts that are the result of dithering. Images therefore have smoother-looking gradients, and produce solid colors without visible dots. HP says that ImageREt 2400 uses the extra addressability of a 600-dpi printer to deliver prints with “even smoother fills, finer photographic details, more colors, and better transitions between colors.” The two principal techniques employed in ImageREt 2400 are blending four colors on a single spot and varying the amount of toner in a given area. We understand that the imaging algorithms that direct four dots to be placed on a spot are proprietary to HP, but we doubt that the ability to vary the amount of toner on a spot is available only to HP. How different is ImageREt 2400 from contone printing? HP expained that, generally speaking, digital devices do not have continuous control over the amount of toner placed on a spot. “There is a set number of possible increments. When these increments are not perceivable by the eye, vendors use the term ‘contone’ to describe the capability.”

Canon has announced the Canon 460 PS, based on the same print engine it has provided HP for the Color LaserJet 4500. We only have a couple of prints from each printer. Looking at the image construction techniques through our hand-held 30-power microscope, we can see that Canon accomplishes shading by making diagonal lines of color wider or thinner. In areas of light shading, the diagonal lines of color disintegrate into dots or specs of color. Up close, shading techniques in HP’s images are more subtle. One can find places, as image content moves from dark to light, where HP uses wide and thin diagonal lines, too. But as the image gets lighter, the diagonal lines maintain more of a pattern, becomming almost a dotted line. Up close, the effect reminds us of a waffle, with regular light and dark areas on a grid. According to HP, “Image REt 2400 utilizes technology in the engine and on the printer formatter. By using the same engine, Canon does not automatically offer ImageREt 2400 print technology.” However, based on the samples we have seen, Canon doesn’t need ImageREt 2400 to produce good-looking prints.

We also compared prints from the Color LaserJet 4500 to prints from Tek’s new Phaser 740, a 1200-dpi machine. Prints from both machines are impressive, but the HP wins out, in our opinion. On the Tek print, one can see with an unaided eye some machine-jitter horizontal and vertical bands in certain areas. One print from the Tek Phaser 740 contains a band going across the page blending from light orange to dark orange. As the band gets darker, the dark toner produces just-visible dark vertical lines, subtle enough to be missed by most recipients of the document, but clear enough to those who look closely. If the prints are comparable or better, HP feel it has an advantage; according to HP’s ImageREt 2400 white paper, “virtually every other competitor will ask you to trade off performance or purchase more memory to achieve its printer’s optimal print quality.” So what does the trade-off add up to? The Color LaserJet 4500 configured as a network printer has an expected street price of $2,950. The Tektronix Phaser 740N, configured with 32 MB for 1200-dpi printing, is expected to sell for $2,695. (A non-networked version of the Color LaserJet 4500 is expected to sell for $2,499, which might be compared to a 600-dpi network-ready version of the Phaser 740 for $1,995.) It appears that, even after making the trade-off and beefing up memory on the Tek 740 to get the highest addressability, the Tek is still $255 less expensive. So one pays a little more for the HP Color LaserJet 4500, but, based on the samples we have seen, one gets better-looking prints.

Color Business Report is an information service of Blackstone Research Associates. Information and analysis presented in this publication are based on the best information available, but cannot be guaranteed for completeness or accuracy. Copyright © 1998 by Blackstone Research Associates. Reproduction without permission prohibited.

Return to Home Page