Dots Per Inch
tbray.org
tbray.org
Look at his blog. Look at his photography. Look at his outputs. He is a competent writer with technical knowledge , but when he starts making aesthetic arguments I think he's out of his depth. Especially when he uses other people's informed decisions to justify his uninformed ones (e.g., JDD's choice of monitor being good enough for him).
That'd be fine if this article was a researched technical argument, but it's not. This article is a "120ish seeems good enough for me, and I got this stuff from JDD and I seem to remember old printers being good at..."
> I suspect that unless you’re preparing high-quality print work, there’s no appreciable benefit for a photographer, and perhaps some downside, in an ultra-high-res display.
Count me among (amateur) photographers who would just love to have a 300 DPI display on which to work on photographs.
Here for example is one quite obvious appreciable benefit: I can see at full resolution what my prints will look like with varying amounts of sharpening applied, in a roughly accurate simulation. Previously, if I opened an image up at 100%, the proper amount of sharpening for print would make an image look extremely “crunchy” on screen, because 100% on screen has pixels three times as big as a 300 PPI print resolution. If instead I looked at the image scaled to 50% on a typical display, I could no longer notice any difference from fine-resolution sharpening, because the screen display would be interpolated.
More than that though, I’d love to have a 300 DPI tablet-scale device to show off work on. Photographs would be absolutely gorgeous.
So, I guess the obvious question is, how much DPI do various prints come out with? And since almost all electronic displays have far less than 300 PPI, would you agree that web designers do themselves a disservice if they get a high-resolution display?
As for web designers: (a) they presumably aren’t the same as “photographers” that Tim Bray was explicitly talking about so I’m not sure they’re really relevant to my comment, but (b) it’s a lot easier to simulate lower-resolution displays than higher-resolution displays. That is, if you had a 300+ DPI monitor, you could approximate the appearance of both a traditional (say, 100 DPI) computer screen, and an iPhone, Android Phone, high-resolution netbook, or whatever else.
DPI has to do with the printed image - how many of the smallest dots of ink can fit along an inch on paper. The resolvable image detail depends on this as well as other factors like the dithering algorithm.
PPI defines how many groups of pixels (each normally made up of an R, G, and B sub-pixel) fit in a linear inch. Even this is complicated by things like some of the current OLED screens that have twice the number of green sub-pixels than red and blue.
A 300ppi screen that fits 300 pixels into an inch is exactly the same as a 300dpi printer that can fit 300 dots of its laser into an inch of drum.
The difference comes because screens can directly vary the colour of an individual dot. Printers can't -- ink is either present or it is not. The way you get around that is by varying the size of the dot, in a process called halftoning.
Halftone dots are very small, and are arranged in patterns you'll recognise if you look at newsprint photographs close-up. You measure the size of the screen in LPI -- lines per inch, the number of lines of dots (at max size) that can fit in an inch.
Because of the need to vary the size of the dots, on a 300dpi printer you end up with a very coarse line-screen of about 60 LPI. This looks rotten, and so printers go up to resolutions of 1200+dpi. This means you can have a linescreen of around 150lpi, which is good enough for most printing. Each line roughly equates to 2 pixels, and hence a 300ppi image is sufficient for printing, regardless of the resolution of the image or platesetter.
But this doesn't mean there's a difference in ppi or dpi: they're still exactly the same.
* DPI is print resolution, a property of prints and displays
* PPI is image resolution, a property of printed and displayed images
* LPI is halftone minimum printing resolution, a property of printers
PPI is taking a digital image and saying "I want it yay big", LPI is saying "I'll do it by drawing at least this many differently sized dots of entirely different colors", and DPI is going "This is how many dots I'm counting you did it with".
edit: Apparently, some manufacturers of small LED displays like to report DPI as the number of LEDs per inch - that is, 3x the native resolution for images. That's a good example for people like me, who know little about printers.
There's no technical difference, however.
I don't care about the dot density, as long as there are enough of them; I care about the dots/pixel, and deeply about the actual inches. I ran into this problem as a kid, printing some images that had dpi metadata that the printer took literally, but I didn't know how to fix it back then so I ended up with a thumbnail print. Attaching ppi (!) metadata to images is stupid anyway, since the actually relevant information would be the inches (or the dots/inches, for designer decrees).
There is a distinction, and I'm all for precision in language, but it's nothing more than a language distinction for the sake of clarity.
It is not a technical distinction, is my point, and people like Bray aren't wrong to call this a 300dpi display. The worst you can say is they are slightly less precise in their speech than is optimal.
It seems to me that a lot of people do draw a line between pixels and dots, and that a lot of people don't (but usually mean pixels). I don't know if there are proper authorities for this, but I consider the distinction important enough to warrant specific language. My dictionary here says that pixels are onscreen dots - I used to think the same, but now I consider pixels to be the samples in memory. I think you are with the dictionary, since you used the term "physical pixel". My adopted distinction meshes well with my ability to change the resolution of my desktop, measured in pixels, but not the resolution of my display, measured in dots.
I agree that Bray can call it a 300dpi display - I didn't criticize him for it - and I explicitly translate that to a 300ppi natural resolution display. As noted above (according to Wikipedia), some manufacturers are now exploiting this by reporting such displays as 900 dpi, counting the subpixels as dots. Therefore I consider it important to state natural resolution ppi numbers for devices with natural resolutions, such as LCDs, and dpi resolutions for devices that don't have a natural representation of a pixel, such as CRTs. Representation numbers, ppi and dpp, I have yet to see anyone provide (maybe because there are too many resolutions already).
tl;dr: It's confusing people, so make the distinction if you can.
You yourself show the interchangeable nature of dots and pixels when you say you measure the size of your display in dots, even though (assuming it's not CRT), it's very much measured in pixels, regardless of the resolution you've got your desktop set at. (This whole thing came about because of people saying Bray was wrong to measure his screen in dots, after all).
tl;dr: you can insist on a distinction between dpi and ppi, but it doesn't really get you anything, and it definitely doesn't help people understand printing!
So dpi is equivalent to 3ppi for LCD under my definition, and both 1ppi and 3*ppi dpi counts have been sighted in the wild. Using my definition is like using SI mibibytes; when you see dpi you don't know which is being measured, but when you see ppi you can be sure, since ppi is only used by people like me.
As such, I do not consider dots and pixels interchangeable; pixels are what I'm trying to represent, dots are what I'm representing them with. If I'm dealing with a printer, I acknowledge that I'm not going to get neat square pixels no matter how hard I try; instead, I'll end up with round, scattered dots. So I look at the dpi of the printer and the ppi of the ideal representation of my image (in inches), and get something like 6 dots/pixel. I consider this sufficient to render a close approximation to a tiny square, and proceed (actually, I don't usually look at that number, since printer resolution is never a problem for me).
This lets me focus on the one aspect of printing that actually matters to a computer: megapixels. As long as I keep the ppi ratio of my layout high, and trust that the printer has high dpi and proper scaling, I get to ignore stupidities like jpeg dpi metadata. In theory, anyway - occasionally print houses request specific dpi, which can be considered a ppi hint and used to do rescaling with the algorithm I want.
Disclaimer: I've never done any print or layout work.
Once you surpass the limits of human vision, where individual pixels can't be resolved at 12-24" distances (whether this is 300dpi or 400dpi, or whatever), there simply is no benefit to going higher. There are significant costs to increasing DPI, such as memory + GPU bandwidth. The hardware/manufacturing limitations are probably costly as well (does _anyone_ manufacture displays in volume with > 350 dpi?)
But I think the psychophysics get more complicated when you throw in time and movement.
Yes, it would be laughable if it was 6 inches from your face during normal usage.
Also, it would be interesting to know at which point, as screen resolutions get higher, font smoothing becomes pointless.
A 300 ppi screen is a 1200 sub-ppi screen. Those subpixels are not as versatile as ink dots in some ways (a green subpixel can’t suddenly turn red) but more versatile in other ways (as you already said, a ink dot is either there or not – it can have a few different sizes but it’s arguably easier for a subpixel to change its brightness).
Of course, if there are enough dots in particular physical area, the human eye is incapable to see a single dot from the closest possible focusing distance. This is what I would call a "retinal diplay". Not sure if 326 DPI is enough for that, but is surely enough for the iPhone's screen.
[1] http://www.w3.org/TR/2008/REC-CSS2-20080411/syndata.html#len...
So recent examples have been seeing adverts in Standard Def for High Definition devices or services. It seems to involve lots of slow-mo and highly intricate images that by their very existence on your SD screen mean you don't need an HD screen to view them.
Similarly DAB digital radio was advertised on standard FM radio with lots of space-age swooshy noises.
For high-dpi screens it seems water droplets are all the rage.