Dots & Images

Most screens and printers work by putting dots of ink on the page.   In the viewers eyes the dots merge and form characters and scenes, pretty much as happens with a TV picture.

If you are looking at this web page on a desktop computer screen then the display is at about 85 dots per inch. Some recent smartphones, which are used closer to the viewers eyes have resolutions above 200 dpi. On the other hand a 40 inch HD-TV has a best resolution of 55dpi and viewing ordinary channels less than a quarter of that.

Visual Aspect

People intuitively realise that small details further away won't be seen. A rough rule is that human eye's have the acuity to distinguish things when they are separated by about 1 arc-minute so a bill-board on a motorway might look well defined printed at 10 dpi whilst the smartphone screen in your palm only looks good at 300dpi.

Paper is generally intended to be used at arms length and historically 300 dpi has been considered sufficient for text. The large and badly placed dots of fax and dot matrix printers are visually unattractive because the dot size and positioning are a bit distracting. These printers have physical limitations that mean they work at something from 65 through to 200 dots per inch (fax is typically 204 x98 dpi).

All inkjet and laser printers work at 300dpi or better. At print resolutions much beyond 300dpi individual dots are not normally visible even on close inspection. There are two possible needs for higher resolution printers: fine lines for diagrams, and compensation for the print process lacking tonal range or greyscale.

Fine Lines

Although most people probably want their colour printers for photographs a few want them for diagrams. The finest line traditionally used in technical drawings and cartography is 0.1mm which is 250 dots per inch, however that was partly a limit with getting thinner pens to work. A 1 pixel line printed on a 1200 dpi laser printer may only be 20 microns thick but it is visible. People might want high resolution for fine lines.

Photographic GreyScale

Photographs are made of continual tones. A black and white photo is actually black, white and every shade of grey. A colour photograph using chemical processes uses three dyes, cyan, magenta and yellow to effectively produce three photos at the same time, each with a range of tones.

Most print processes can't print a continuous tone, they either mark the paper or not. This is true for most pens, they either write or not, you may be able to make a thin line or a broken line but not a very effective faint line. The same holds for offset-litho presses, ink transfers from press to page or it does not.

Printworks use a process called halftoning to produce colour print. The plate etched for the press has small coloured dots where the areas of colour blend on the paper and/or the viewers eye to produce the desired effect. The print plate for a press uses dots of different sizes to represent the different tones, so a with big dots of magenta and yellow the press produces a vibrant post-box red and with small dots it makes pale pink.

Computer print technologies can't vary the colour or the size of the dot. However they can print great numbers of dots to give a strong colour, letting them merge to give solid colour. Alternatively they print just a few to give pale colours. For instance to get a pale blue sky one position in ten might be cyan, and overprint one in two magenta. The problem is that if those dots themselves are noticeable in any way, big or systematically arranged, then the sky will not look blue so much as speckled. So the first need is to get the dots small enough that individually they will be difficult to see. Dots aren't easily discerned at 600 dpi and below - (which happens to be about 600 dots per square millimeter).

Early colour printers often had 600 dpi or 720 dpi resolution. Whilst the dots were microscopic each dot could only be one of 7 colours, the cyan, magenta, yellow and black native to the printer and the production colours red, green and blue where they are overlayed. When the first inkjets with this sort of resolution appeared people talked of them replacing colour photography. However one close examination the pictures had all sorts of defects: colours that were vibrant but not very like the intended scene, dotted areas that should have been light colours, blurred boundaries.

The smaller the dots the better, great numbers of dots are potentially needed to make up for the lack of colour tone. For instance to give an appearance of grey we might allocate each pixel of an image four dot positions on the page. All dots off leaves the page white, all on turns it black (or nearly so if they don't overlap) and there are two other levels corresponding to two or three dots. Using four dots where there used to be one gives a four -level greyscale - at the cost of needing double the print resolution.

The tonal range from a digital camera is 8 bits for each of the three colour channels, red, green and blue - 24 bits in total. The cameras range for each colour is 8 bits, which gives 256 possible levels. To have 256 possible colour-tones on the page there must be a possible 256 printer dots for each pixel.

Visual acuity suggests that pages need to be printed at something like 300 dpi. Printing at 600 dpi with 300 dpi data will give each colour a 4-level greyscale. At 1200 dpi it multiplies by four to a 16 level scale, at 2400 dpi to a 64 level scale and at 4800 dpi it is finally possible to print at 300dpi but allocate enough dots to give every pixel of the page enough dots that it can range across 256 levels.

4800dpi print can be done, although only by inkjet printers at present. In a sense inkjets could always do it by making very small nozzles, printing slowly an expecting the number of nozzle blockages to rise.

Inkjet printhead manufacture resembles semiconductor chip making in some ways. Every couple of years it proves possible to put double the number of nozzles on a chip. Manufacturers can use the increased number of nozzles to print more quickly, at the limit they can use pagewidth heads and print at around 60 pages per minute. An implication might be that in the future all inkjet printers are likely to use pagewidth heads capable of printing at 4800 dpi or above.

Laser printers rely on toner transfer across a voltage gradient from developer to drum. Making the toner particles smaller helps. The laser's pin-point beam also has to shrink. To move from 600 dpi to 1200 dpi there have to be twice as many lines so the scanner mirror has to move faster or have more segments. The laser has to do four times as much work. Many laser printers have achieved 1200dpi print using toners with diameters around 5 microns. There are difficulties going further - toner particles would be very fine, and the memory demand potentially grows rapidly.

There are still some issues and adjustments to be made.

It is important that the dots used don't take on patterns, otherwise even though the dots are tiny the patterns will create distracting moire effects on the page. So the print process can't just allocate smaller and smaller cells on the page, it has to allocate colours and then put a random dither across the dots. What is done with dots near sharp boundaries is likely to be very significant to human perception.

Colour

Camera images are defined in an RGB colour-space. Printers have a CMYK colour-space which prints more brightly but only logically has 8 colour possibilities overlaying the colours but also has that extra black which can be used to darken the image. Some colours like yellows aren't very significant on their own. Human beings can distinguish about a million colours, so the 16 million colours from an RGB camera are actually a bit of overkill.

Cameras and printers both have difficulty with monochrome photographs. People can distinguish about a thousand different levels of grey bit without some adjustment a camera can only record 256 levels and in most scenes there will be very little that is at or near black or white, so if a mono image is wanted there may only be a couple of dozen light levels, this will tend to produce banding - unnatural contours on areas of sky that should be smooth.

Memory

At 300 dpi a page of black and white text needs about a megabyte of memory or 8,280,000 bits (11.5x300x8x300). If that were an 8 bit, 256 level greyscale it would be 8 megabytes. Using four possible colours (CMYK) its 32 megabytes. (11.5x300x8x300x4x8 = 264,960,000 bits / 8 = 33,120,000 bytes)

If the same page were converted to 4800 dpi in the computer and sent to the printer it would be (11.5x4800x8x4800=) 2,119,680,000 or 2 gigabits or 264,960,000 megabytes. If that were colour equivalent to the 300 dpi we should only need to send the printers production colours for each dot. 11.5*4800*8*4800*4 = 8,478,720,000 or 8 gigabits - about a gigabyte for a colour image.

Inkjet printers often have a big advantage over laser printers when it comes to high resolution. The inkjet is relatively slow, and it can stop at the end of a line. Inkjets only need memory sufficient to supply the printhead nozzles with data for the line they are doing. The lines don't even have to be created in the printer and for many printers they aren't, the processing needed for the printer is done in the user's computer. The resulting transfer might be huge, but if it is going across a dedicated USB connection then who is to know or care? If the transfer was across a network or over a 3G network it would be challenging.

Laser printers have a problem. Once the photoconductive drum starts to move it can't be stopped or the image will be ruined. To ensure the page will complete properly the printer ideally needs the whole image in memory before it sets off. A page of mono text is typically just 4,000 bytes and the printer's own "formatter" will interpret that into the page-image. A page of mono graphics is a megabyte - which was a challenge for printers until about 10 years ago. At 1200 dpi that increases to 16 megabytes, a challenge until quite recently. At 1200 dpi with a greyscale it could be 128 megabytes and four times larger still with colour.

Very high resolutions for commercial quality photos will probably only ever be delivered by inkjet technology.

A laser-print technology called wet-process electrophotography could in principle deliver both 4800 dpi and (if it were needed) some measure of greyscale but the problem with it is that the developer and drum have to work in a dielectric liquid like paraffin and some of this inevitably transfers to the page making it smelly.

Inkjets also have an advantage because they can use light-coloured ink. To make a light blue the printer might make something like 20 dots in a square millimeter cyan, and some cyan+magenta=blue leaving more than 500 white. A better looking light blue would be achieved using more dots of lighter cyan and magenta. Inkjets can have separate cartridges for light cyan, light magenta and one or two shades of light grey to improve that issue with monochrome photos. Adding these extra cartridges doesn't actually cost much; to the manufacturer it's just extra carriage positions and a bit of support electronics. To the user it does mean buying more cartridges, but the overall ink use may be no greater. There are inkjets intended for photographic work with ten cartridges. A laser printer with ten cartridges would be quite some size and rather complicated.