How to buy an inkjet printer to avoid the resolution pitfalls

I.T. Strategies' Marco raised this skeptical question about surface resolution last fall while mocking the hype tendencies of some inkjet printer vendors about their specifications. I addressed some of the same issues in an article, and while I share Marco's sarcastic view of spec issues, I think it's a matter of looking into Apparent Resolution

The truth behind the quirky word "surface resolution" could help.

Let me try to explain why manufacturers introduced the term and what justifies a particular value. Why the answer to this question is simple: Marketers think potential buyers are naïve — they think that if a printer claims to have higher resolution, then its quality will be higher, as the DPI (dots per inch) value stands for. Well, maybe they don't actually believe this, but they want to make their quality statement as simple (or simplistic as possible). What did Einstein say? "Make things as simple as possible, but not simpler". As many experienced readers know, print quality depends on something more than resolution, and resolution should not be confused with addressability in any way.

Therefore, people writing marketing collateral for printers with a resolution of 360dpi must find a way to make a printer with only 360dpi look better than a competitor printer with 1000dpi. They do this without having to explain the effects of different droplet sizes, the expansion of dots on different paper textures, or the accuracy of droplet placement – it's too complicated. However, if a 1000dpi printer has only a single droplet size, while a 360dpi printer uses "grayscale" or variable droplet size, then the term "surface resolution" provides a simple measure to capture some important research on the perceived quality of photographic images copied at different resolutions and grayscales.

Dr. Joyce E. Farrell, a former HP lab and now Stanford University, has written extensively on the subject over the past three decades. In a 1997 paper, she concluded. "(DPI) is a useful indicator for evaluating the quality of text images. However, it is not a useful indicator for assessing the quality of photographic images. When consumers consider buying a printer optimized for the quality of photographic images, it is best to ask about the number of digits (bpd) of grayscale/dot. In this article and elsewhere, she explores the trade-offs between resolution and grayscale/pixel, but concludes that there is no simple formula for judging that the equivalent quality is quite specific to an image.

This means that "surface resolution" is necessarily a vague and unscientific concept. But here's a partial reason to use traditional halftones as a basis for explanation.

Recall that in traditional offset printing, "AM Plus Net" is used to create halftones to reproduce photographic images. That is, points of different sizes are produced by a high-resolution imager (for example, capable of creating laser points at a speed of 2450dpi). A square point might consist of 1-100 pixels in a 10×10 pixel matrix, each on a fixed-resolution grid of 245 lines/inch — the "screen ruler." The dots on the screen are "amplitated" by the ever-changing density of the photographic image, but their relative positions are fixed by the grid. The pitch or screen ruler is 245lpi, but the size of a single dot varies from 1% (highlight) to 100% (shadow). The most recent printing development is "FM" (or "random") netting, where changes in image density are reproduced by changing the distance between points—frequency—rather than maintaining a fixed point size. The printer is actually printing a collection of highlights that are tightly grouped together in the shadowed areas of the image and scattered in the lighter areas.

This is also how binary inkjet printing works – inkjet printers change the position of fixed-size points to match the different densities of photographic images, often using the "error diffusion" algorithm to place these points. The dots on the paper are made up of fixed-sized droplets of ink.

Grayscale inkjet printing is like a combination of amplitude modulation and FM netting - by changing the volume of the ink droplets, the size of the dots can be changed, and the relative position of the dots can be changed by the error diffusion algorithm. So now we can make an analogy between traditional halftone and inkjet: a traditional (AMPD) image might use a 10×10 pixel matrix to create up to 100 grayscale (0-100% density) screen points, while an inkjet RIP-driven printhead, such as 9 grayscales, can be thought of as using a matrix of 3×3 pixels for each point, placed on a grid with dimensions equal to its advertised resolution. Therefore, the "surface" or "effective" resolution of an inkjet printer is equal to the nozzle density multiplied by the square root of the grayscale progression. As a result, our 360dpi, 9-grayscale print engine can be said to have an "apparent resolution" of 3 x 360dpi = 1,080dpi, effectively matching a 1,000dpi printer with a single drop size. Marketers working for grayscale printer manufacturers have a simple yardstick that matches the resolution claims of competitors, and she can prove this by citing research by Farrell and others who have shown that images copied by grayscale printers look better than images of the same resolution as binary printers.

However, there are a few important caveats – the first being the points that have already been made. "Surface resolution" is subjective and unscientific, although the analogy I have just described gives it a theoretical reason. The second point is even more important – the concept of equivalent resolution applies only to continuous-tone images. Although grayscale printing can reduce the phenomenon of low-resolution line drawings and steps visible in text (also known as aliasing) by filling the ladder with smaller drops, the print quality cannot be compared to high-resolution printers. The "surface resolution" of a 360dpi grayscale printer may look like a 1000dpi printer on a photo, but it cannot be compared on a 4-point font.

So, what is the answer to the question "What exactly does 'surface resolution' mean"? I hope I've proven it —but it's not meaningful enough to justify the purchase decision.

(7860972)