This is a topic that has been causing confusion ever since scanners became available for digitising colour transparencies. Things have changed only a little and confusion remains an ever-present blot on the plot.
For pixels, the simplified version runs something like this:
Images are made from pixels. A pixel is an area within an image defined by its size, location and colour. Colour is defined as three values: one each for red, green and blue. The size of an image file, as seen by a computer, is found by multiplying the number of pixels across the width, by the number of pixels down the side, then multiplying that number by three (once each for red, green and blue). That is all that a computer cares about, how many pixels, where they are and what colour they are. The physical size of an image (that is how many inches wide by how many inches deep, only comes into play when the numbers from the file are arranged into a picture, be that on screen or printed to a piece of paper.
That is the first time that ‘resolution’ comes into the frame, we have to know how many pixels to allocate to each inch of length in a picture (despite metrication, resolution is firmly entrenched in the imperial system, it’s always pixels per inch – very rarely pixels per cm).
Our troubles are just starting! We are now in what is technically called ‘device dependent’ territory. Adobe InDesign, Photoshop and Internet explorer all handle pixels and resolution in different ways. If you view, or print, from each program the image will appear in physically different sizes. InDesign honours the physical size – a 6x4 inch print is fitted onto the page at that size; Internet Explorer simply counts pixels then shows as much of the image as will fit onto the screen; Photoshop is something of a hybrid and scales the image according to instructions from other parts of the program.
Printer drivers are more specific, they care about pixel count but only stretch or compress file sizes when asked to do so. The printer driver needs to be given the physical size of print required. It then scales pixels to make the best job that it can; if there are insufficient pixels to do a decent job, the printer carries on regardless!
By now we are firmly in confusion territory and if your eyes have glazed over then here is a simple table which defines how big your file should be (as measured when open in Photoshop but without any layers) to produce a half-decent print.
||Size in MB
What will a Nikon D700
do? 12x8inch at 360ppi, 14x9inch at 300ppi, 24x16 inch at 180ppi
Looking at the numbers tells us that a modern DSLR will make a 16x20 inch print if you accept printing at 180ppi image resolution (more on that later).
Dots Per Inch
So much for image resolution, now what about printer resolution? The numbers take an immediate leap upwards. Typical printer resolutions are 1440 dpi, 2880 dpi and now (Epson R3000
) 5760dpi. The two types of resolution numbers should not be confused or interchanged. Photoshop deals in pixels per inch, printers deal in dots per inch!
A typical inkjet printer might have between four and 11 colour cartridges. Simplistically, if a six-cartridge printer fires two dots per colour to create a pixel then the dpi needs to be 12x the ppi. Sadly the reality is more complex! The printers change the size of the dots (variable droplet technology as Epson call it) using larger ones in areas of heavy colour. Then, to compound matters further, the printer fires the dots in a random (stochastic) pattern to break up the lines of dots and make for smoother transitions (dithering). All we can say for certain is that the dpi required is always many times larger than the ppi, and you should not get them mixed up.
The standard, higher-quality settings on an Epson printer call for 1,440dpi and 2880dpi in the driver settings. For this reason the preferred ppi values are either 180 or 360ppi because this divides into 1,440 eight times or four times. In practice, though, you will be hard pressed to notice the difference between using say 360ppi and 300ppi, the dithering engine accommodates the difference quite well and masks a poor match. At 180ppi things are a bit more of a struggle as there are fewer pixels to play with and, on architectural shots, you might notice the difference on corrugated roof cladding or angled roof lines. For landscape and portrait shots you will be hard pressed to detect any difference.
|The important areas are marked out in red and are typical problem areas when resolution in an image is insufficient.
||The image at the bottom is a size-for-size scan from a print made at 180ppi. Note the disturbance to the corrugated roof lines, they are smooth curves on the building. Also note that the near vertical joint lines of the fascia have been imaged as a series of separate lines.
Having always lectured that 360 and 180 are preferred but noticed little difference in general practice, we set out to make an accurate comparison. We chose four image resolutions 180, 220, and 360 ppi and then, to really work the system, 354.6ppi! We used the architectural subject shown and examined the indicated areas in full-size prints. It is difficult to put numbers to the findings so a narrative verdict will have to suffice! The findings were as follows:
1. At the two higher resolutions the sloping roof line was cleanly rendered, at the two lower values it was jagged.
2. At the two higher resolutions the shadow noise on the wall was smoother. It was crisper at 180ppi than 220ppi indicating perhaps that the print engine was slightly smoothing the detail, in other words it was slightly softer.
3. The thin joint lines were visually smooth at both the high resolutions but divided up by the lower resolution interpolation into 18 small streaks at 180ppi but only 11 streaks at 220ppi, in other words the lower 180ppi value was slightly better.
4. The corrugations in the roof were true at both higher resolutions but disturbed at lower resolutions, more so at 180ppi than 220ppi. This result reverses the proposed trend.
By now some hard and fast rules are required in the form of practical advice. It is this. Always preserve as much resolution as you can; do not, for example, downsize to 180ppi because it is closer to the ‘Epson optimum’ until you get below 200ppi – above that go for the pixel preservation route! If you choose to use higher resolutions then take account of this in your sharpening. If, for example, you are High Pass sharpening a normal value for Radius is 1⁄100 of the resolution (ie choose 1.8 pixels at 180ppi, 3.6 at 360ppi). So, if you elect to downsize to 180ppi or you are forced down to that value because of limited file size, then back off on the radius in sharpening as well. Even with other forms of sharpening, the theory stays true: higher resolutions can take larger radii than lower resolutions (but not the Amount, this stays the same).
Having confirmed that we should not throw pixels away and that the printer driver does a good job of scaling, things become very simple – you use the printer driver to scale the image to the size that you require. Set the paper size in the driver then check the ‘scale to fit media’ box and your print pops out, scaled, and fitted to the page. If you require a white border around your print, click the scale to fit media button and note the value (say 68% for example). Now uncheck scale to fit media and type say 60% in the scale box and you have 8% of white border around your print.
For the enthusiast/photography club competition, if you need a 20x16” print or, more commonly today, a 50cm x 40cm print, set that size as a ‘user-defined’ page in the printer driver and then scale to that. If you use an oversize piece of paper (A2 in this case) you can set corner crop marks to give you a cutting guide down to 16x20 or 40x50.
Pixels and Dots
The left set shows an idealised model with square pixels from eight cartridges, providing 16 squares in a 4x4 array which mashes up to a mid grey, or with a change of the colours a bit of a flesh tone. The right set is more realistic with different sized dots, randomly placed in the square pixel, some overlapping some not. The reality is a mix of variable sizes of spots, randomly dithered around the nominally square pixel area to create very smooth tones at normal viewing distances.
Surprisingly little is the answer, providing all you wish to do is identify the model! The image to the right shows model, Charlie Edwards, scaled at 9, 18, 36, 72, 144 then 288ppi. For the image above the same images have been up-scaled without any smoothing of the pixels, using nearest-neighbour interpolation. Note the progressive improvement in the rendering of the hair as the resolution rises. If you screw your eyes up even the 9ppi image is recognisable as Charlie but this does not provide anywhere near enough information for detailed printing.
How much resolution do you need?
Borders from the Driver
Here is the simplest way to scale and border. The paper size has already been selected in the printer driver and ‘centred’ was checked there as well. ‘Scale to fit media’ indicated a required reduction to 89% so the box was unchecked and then 80% selected to give a border all round. In the ‘Output’ panel the ‘Border’ button was checked and a 1mm black key line set up. The preview shows exactly what we would get on the print – a centred, bordered and key-lined image. In this dialogue box, if ‘Background’ is selected then the background colour may also be chosen, instead of the default white using the Colour Picker, or even, using the Eyedropper and taking a colour from the image itself.