Print Resolution

We need a figure that will sum up the print quality. One useful number is the maximum print resolution.

The reality with print resolution is that it is often used as a single figure of merit to cover the printer's line drawing capabilities but also as some indication of it's colour capabilities. This may seem an odd connection - what most computer printers do is to use dozens of colour dots to produce one pixel of a picture.

The measure is almost invariably dots per inch rather than per millimeter; its partly a historical thing. The computer printing industry originated in America so their measure became the global standard. Trying to use a measure other than dpi will just cause confusion. Also the number of dots in an inch is large enough to be meaningful, people are likely to understand 100 dots per inch but 3.93 dots per mm just seems weird.

People expect that more dots per square inch will always be better. Historically that was true; a laser printed page at 300dpi looked better than a dot matrix page at 200dpi.

Print resolution varies from about 65 dots per inch using a 9 pin dot matrix printer up to 9,600 dots per inch from inkjet heads intended for fine photographic work.

Bigger numbers are better for print resolution, at least to 600 dots per inch. Beyond that point resolution on it's own ceases to be very meaningful because the dots are smaller than a spec of dust. The finest pens normally used in cartography are 0.1mm and the dot made at 600dpi is less than half that diameter, so it's effectively invisible.

If you want to see how good an image can look at 100 dots per inch you are close to doing so right now - if you are reading this on a screen. The typical computer screen has a resolution of about 85 dpi. TV screens often have somewhat lower resolution - a 40 inch HD-TV has a resolution of 55 dpi.

Printers commonly boast resolutions such as 2400x1200 optimized dpi, ImageREt3600 and 5760 x 1440 optimised dpi. The small dots are used to compensate for the printer's inability to print colour tones. If you use a process that can produce continual tones such as dye sublimation then a printer with a mere 300dpi can produce good looking colour pictures. But dot matrix, inkjet and laser printers aren't good at continual tones so to get a good looking colour picture they use great numbers of microscopic dots which mix and meld to give an overall impression.

Ideally the printer manufacturers would give us details of the surrogates they are using for continual tone, like the size of the smallest dot, level of dot-gain, variation in dot size, number of greyscale levels, density or contrast levels produced by inks. We would also need information about the algorithms used, a picture will often go through several transformations on it's way from camera to printer. Sometimes they drop interesting hints such as microfine toner or more informatively picolitre droplets. However they mostly seem to hope we'll be happy with expressions like ImageREt3600 which roughly interprets as It'll look as good as 3600 dpi (honest, gov) .

As you might gather from the preceding couple of paragraphs the problem with print resolution is that on close examination it turns complicated and perhaps the real details will only be discoverable with a microscope and a colourimeter.

Another problem is that people's evaluation of pictures may not be objective, they react to what they see. Some people like the solid print on ordinary office paper that laser printers give, others might prefer the glossy photographic look of special inkjet paper.

Almost all recent computer printers produce text and basic pictures well enough. Unless the task is producing high quality photographs, print resolution might now be less important than cost and speed.


Screens

An obvious anomaly is that screens typically have resolution below 100 dots per inch whilst printers now commonly have resolutions above 1,000 dots per inch.

Part of the difference is that screens are bright and colourful. Since it costs quite a lot to achieve the limited resolution screens have we quietly forget that they aren't very good. Moving to higher resolution screens would not just be a matter of better LCD technology, a big change in graphics adapter memory and processor power would be needed as well. Two of the competitive pressures in the market are

  • to make TV screens bigger, even though that means lower viewing resolution. People often watch TV at rooms-length to make sharing easier. Since people seem to like very big screens for watching TV there will presumably be a demand for Extreme Definition 1440p displays.
  • to get more information onto smartphone and tablet screens - so some do have resolutions of 300 dpi and beyond. Smartphones are used close-too

There are various anomalies about the way screens and print are used. For instance paper is almost invariably used in portrait orientation whilst screens are landscape.

There don't seem to be great efforts to make ordinary computer screens with significantly higher resolutions. Many computer users sacrificed some of the image they used to see on a screen when they swapped their CRT monitors capable of UXGA 1600x1200 for LCD monitors with SXGA 1280x1024 screens. The LCD image is generally steadier, clearer and the flat screen less prone to reflection, however it is odd that people were happy reducing their screen area to 68% of what it had been.

Text

Text is ordinarily black on white paper; that gives the maximum contrast. White text on black paper is more expensive to achieve (if you use a laser printer for white on black print it costs about 20 times as much - (white toner isn't readily available for black paper).

For ordinary text a resolution of 300 dots per inch is often adequate and was the standard used by laser printers through to the 1990s. If the output is straight-forward text or forms there isn't a great deal to be gained by higher resolution.

Standards for text printing date back a long way and are retained because they work. Typewriters have been producing 10 point text text on lines about a sixth of an inch (12 points) high using characters 1/10th of an inch wide for over a century. Typescript is fairly large print. Book print is typically a bit smaller, 6 or 7 point. Ordinary readable text varies from about 2.5 mm to 3.5 mm high. Character width usually varies with the character, from about 1mm for a letter I to 2.5 mm for an M.

It is normally the the height that matters in rendering text on a computer device. With a couple of exceptions there are more details vertically than horizontally. (The exceptions are M and W of course).

Small print is where any problems will become clear. With 7 point letters 2.5mm 1/10th inch high a 300 dpi has 30 dots height in which to render the print - that's sufficient to give a fairly nice outline of a character with about 6 dots for the tails and risers on lower case and 18 for the main part of the print.

Four-point print which is the smallest normally used has about 1/20th of an inch available, a more limited 15 dots height but still sufficient for a somewhat blocky dot rendering of a character.

The limit for Latin text is usually 7 pixels high, where the middle of the letter B contains a column going 1001001 and the right edge 0110110

Plain black text at 300dpi looks good, sufficiently so that the Apple Laserwriter became the basic technology behind the desktop publishing industry when it was launched in 1985. In fact 300 dpi had been (and remains) the normal standard for professional print on machines like offset litho presses. The idea that 300 dpi is adequate is confirmed to some extent because the finest pens used in cartography and technical drawing are Rotring 0.1mm (250 dpi)

Books and magazines are normally printed at about 300 dots per inch using a half-tone process.

Diagrams

There are a couple of defects in the argument that 300 dpi is good enough for text.

Although small dots are not easily visible very thin lines are. Human hair varies in the range 17 to 180 micrometers and fine hair at about 20 microns (0.02 mm) across is quite visible. Very fine line drawings wouldn't be normal but might be useful for some kinds of diagrams. To get a 20 micron line the printer would need a resolution of 1270 lines per inch.

Wide body printers (plotters) sometimes have an explicitly statement of the finest line they can draw. Laser printers often don't - it is actually quite difficult to tell a laser printer to produce an unadorned, unenhanced fine line because various features of the graphics and driver subystem will try to give it more emphasis.

The other problem is aliasing, a change in the look of something created by the way it is drawn. The most familiar example of aliasing is the stepped effect that can be seen on some screen and print diagonals, what is supposed to be a line looks like a staircase. A more subtle effect is that diagonal lines look a bit thicker than they should. Aliasing can have subtle effects, like giving unintentional emphasis to the letters A and X.

To minimise the distraction of aliasing in text, fonts for screen and printer use commonly use anti-aliasing. Pixels which the underlying information suggest should be part on, part off are not set either completely black or white but to a smaller dot or a level of grey. At the intended distance the viewers eye sees a nicely formed letter although in close-up the result looks weird.

Photographs

The main reason for wanting very high resolution printers is to get the continuous colour tones needed for photographs. Resolution may be useful in itself but mainly it's used as a surrogate for colour.

Colour pictures were an esoteric interest until the early 1990s, when relatively low cost digital cameras like the Kodak DC40 and Apple QuickTake 100 became available. These cameras sold for about £700 in the UK. Equipped with a digital camera, and a PC with Microsoft Word an individual could produce pages looking something like a colour magazine in minutes. The problem was that the printers of the day were disappointing with pictures. For instance the OKI 393 was a fast printer, but a colour photo would take 20 minutes to print.

Photographic printing processes work because they can produce shades of grey. Expose a film to light from a lens and the negative darkens in proportion to how much light it receives, then it is developed by shining a fixed amount of light through the negative at the photographic paper. (OK the colours don't look grey, but they are handled as three levels of intensity.)

There were photographic printers. For instance there were printers based on cathode ray tubes that had output on film, such as through a Polaroid camera. Typesetters are effectively photographic printers, so if the objective really was to produce a magazine (and had £25 thousand to spare) then digital cameras cut out the need to take a photo, get it developed, scan it in and then re-output it through the typesetter. The big use for digital cameras, however, would be to cut out the expense of chemical photography altogether.

Computer screens had little difficulty showing good looking photographs. Computer screens share a heritage with television where the amount of light hitting the Videocon tube becomes a brightness signal which lights or darkens the electron beam in an analogue TV set's cathode ray tube. LCD screens use a completely different mechanism but are plug-compatible replacements for CRTs so they give the same result.

Early digital cameras like the DC 40 and QuickTake 100 had sensor elements that matched the dimensions of a typical screen at the time. The digital camera market took off quickly so that by 1999 there were hundreds of models, but mostly offering resolutions around 640x480 - we'd now call it a third of a mega-pixel.

Photography was a whole new market for printers, but one only a few printer makers were well placed to address.

Most printing processes are binary in nature, ink either transfers to the page or it does not. This is not only true of digital printers, it is what we expect with pens as well, where the pen touches ink transfers. With sensitive handling some pens can write lines of different thickness, but they can't vary the shade of the ink.

The binary nature of print is highly desirable in some ways, when the process is working minor changes won't cause faded print so that the text remains readable. However only a few techniques for producing pictures will work. Through to the early 20th century pictures in printed matter were usually engravings.

The printing industry adapted to the use of photography towards the end of the 19th century by developing the halftone process. An image is projected through optics that divides it up into dots, with the size of dots corresponding to the darkness of the image. By 1890 the technique was common in newspapers. During the 20th century the process developed, although pictures were still often produced as a special print run, not part of the main text.

Computer Print processes are generally not like this, the dot-matrix process provides a good example.

Dot Matrix

Dot matrix printers are a niche product now, used for warehouse work. From the 1970s until the mid 1990s they were very much the standard printer, almost every business and home computer had a dot matrix printer. The mechanism is simple, robust and cheap to run so it seemed as though it would be around forever. Several big names in the printer business were based on dot-matrix technology, guessed that business would go on pretty much as it had in the past, and proved wrong.

Dot matrix print uses conventional typewriter ribbon and paper. For the most part the idea of a typewriter is to press a key which impacts a typebar on the ribbon. The impact produces a neat, dark, printed figure. The dot matrix process is very similar, but the pre-cast type font is replaced with a column of 9 pins which strike the ribbon in a pattern, knocking dots of ink out. After printing 7 positions the pins create a character in dots.

Dot matrix pins are driven by solenoid coils. The darkness of a dot partly depends on how strong the solenoid current is, so a weak current will make a lighter dot. However the process is not at all linear, with half current in the solenoid the resulting dot won't be half as light, chances are there will be no dot at all. There may be some set of current levels where a half-dozen different dot intensities can be relied on but that will be hard to find and as the ribbon fades some of the range will be lost.

Using lots of overprinting there would be a way to get some levels of shade into a picture. That just highlights another of the dot-matrix problems which is that they aren't very fast. If the printer has to make lots of passes then the picture will take ages to emerge. Using a plae ink and overprinting won't give the desired range.

Dot matrix printers can't provide a range of dot intensities, not can they vary the dot size. Since they rely on pins that have to be about 0.2 mm across for mechanical integrity the dots shrink much either, so dot matrix printers are no good for photographs.

Dot matrix printers do many things quite well, but with a few possible exceptions they aren't good at photographs. For the last decade or so printer buyers have not wanted to sacrifice the photographs - even though in practice they don't print many they really want.

The two technologies that can adapt to photography are inkjet and laser printers. They can shrink the print dots without too much impact on speed.

Inkjets

Inkjets date back to the 1970s when Siemens produced the PT80, a mono printer with 12 nozzles. The problem with successor products was that the nozzles were too inclined to block and they were based on a piezoelectric principle so they were expensive. Around 1984 HP and Canon began to make printers with simple disposable 12 nozzle printheads, in HPs case based on resistors mounted under a shim. These early inkjet printers were intended to be cheap, lightweight competition for dot matrix machines. Epson took a different approach and developed 24-nozzle piezoelectric printheads intended to compete with daisywheel office printers..

Thermal inkjet printheads like those used by HP and Canon were originally intended to be cheaply made in automated factories so they could be disposable. Within a couple of years the manufacturing process resembled the techniques for making semiconductor chips. A thermal inkjet head is a fairly simple semiconductor chip on a glass substrate with holes etched or punched through it. The heads became a bit more expensive but a lot more capable. HP's Deskjet printer in 1988 had 50 nozzles, a resolution of 300 dpi and claimed "laser quality for under $1000". By 1990 the Deskjet 500 got the speed up to 3ppm and price down to about $700.

The Deskjet 500C launched in 1991 added a colour cartridge. Manufacturers realised that it doesn't take much extra effort to make a colour inkjet printer, its just a matter of adding a cartridge and a little bit of support electronics. Since the customer will pay for any subsequent colour cartridges as consumables the printers actually hold more promise of future sales. Where in 1990 there were few colour printers by 1990 there are no black-only inkjet printers other than for special purposes like mailing.

Canon's BJC-600 series had four separate cartridges and a printhead containing four heads with 64 nozzles each printing at up to 360dpi. . In principle the printhead was changeable, although in practice that wasn't economic. Print control modes included BJ-10 mode (IBM Proprinter X24-E) and LQ mode in which the printer emulated an Epson LQ-2550.

Apple's StyleWriter series sold from 1991 to 1995 were inkjet printers based on Canon innards. For instance the Colour StyleWriter Pro was based on the Canon BJC-600.

Epson's Stylus Colour (aka the MJ-700V2C) launched in 1994 used new Micro Piezo heads. It had a 64 nozzle black head for speed on text. Another head was used as 3 x 16 nozzles for colour. The big selling point was the world beating 720 dpi output which was combined with some level of halftoning - or as it is sometimes called dot-modulation. Piezoelectric heads can adjust to fire a range of dot sizes; thermal inkjets have much more difficulty with this.

Epson maintained an advantage in photo printers for some time. For instance PC-Pro put the Stylus Photo 700 on their A-List in September 1998. Resolution had increased to 1,440 x720 and the colour cartridge now had light cyan and light magenta as well as the conventional colours. Prices were falling as well, the printer was priced at £232 in the UK (about $400).

Epson, Canon, HP, Lexmark, Brother, Samsung, Tally and Olivetti began a specification war with resolution in DPI the prime Figure of Merit. PC-Pro's review of colour inkjets in September 1998 was dismissive about Olivetti and Tally's entries. Neither company made much further showing. Developing printheads and drivers seemed to be beyond all but the major players.

Laser Printer

Who invented the dot-matrix printer is debatable. HP and Canon both claim the inkjet although Siemens clearly had a product before either of them. Laser printers are less arguable, Gary Starkweather at Xerox converted a photocopier into a laser printer by disabling the optical scanner and replacing it with a polygon mirror. The physical modification took weeks. It took a couple of years to develop some formatter electronics to match the capabilities this released. Reaching full potential has taken 40 years.

IBM and Xerox sold a few laser printers in the 1970s but prices were in the $500,000 range. Siemens made machines that burned characters direct onto the page - but it did have a tendency to cut through the page.

HP top management were

Thermal inkjet heads are generally

Resolution is normally given as some number of dots per inch. This sounds straight-forward and objective, something that might be measured with a microscope. However As measures of print quality we might