I’ve known type designer and font hinter David Berlow for many years. We first met in Boston in 1995 at the Spring Seybold publishing conference. I was the obligatory Microsoft sacrifice in a roomful of Macintosh users at the auto-da-fe known as the “Font Free-For-All”; David was the “moderator” – something of a misnomer, since anyone who knows David also knows “moderation” is not really his forte – he could start an argument in an empty room…
But I guess that was the point of the Font Free-For-All anyway; the attendees came to see blood – preferably Microsoft blood (it was the time of the Font Wars).
At this point, most of you will be scratching your heads and saying, “Fonts? People went to war about Fonts?”
A newspaperman once asked Scottish football manager Tommy Docherty,”Surely you don’t think football’s a life-and-death affair?” To which Docherty replied,”It’s more important than that!”
Many font people feel that way about fonts…
Q: What do you call two font experts in a room?
A: An argument…
Q: What do you call 12 font experts in a room?
A: You call 911…
(One day, perhaps, I’ll tell the behind-the-scenes story of the “Font Wars Truce” that led to OpenType, which began with about 12 Adobe and Microsoft font experts in a room…)
Anyway, David’s become a good friend over the years. His company, Font Bureau, does great work, and David’s one of the handful of really great “font hinters” in the world.
Writing on Roger Black’s blog (see link on this page) about fonts on the screen, rendering technologies etc., , David says:
“144 dpi is the first uncontested resolution at which all users will be satisifed exclusively by subpixel rendering to the smallest sizes, and that happens regardless of underlying “technology” i.e. at 144 dpi “greyscale” will be the thing users choose, because it adds no color at any size when defining what should be by definition, a monochrome definition.”
He’s talking about different anti-aliaising technologies, including ClearType.
Despite being one of its inventors, I can see a time when ClearType becomes obsolete. It solved a real problem of too-low display resolutions when we invented it about nine years ago, and I estimate it still has a long useful life in front of it. But it definitely has a limited lifespan. How long that will be depends on how quickly displays move towards higher resolution. That might take longer than anyone realizes. In fact, maybe a combination of ClearType and hardware resolution will get us where we need to be – in which case it could be around forever.
The human visual system has a vernier acuity of about 1/600th of an inch. We can’t see anything smaller than that. So printing on paper or viewing a screen with a resolution of 600 dots per inch is the most we need. The only reason high-end imagesetters for the print industry need higher resolutions (all the way up to 2500 dpi or thereabouts) is because of the “lossy” nature of the printing process itself – ink that spreads, paper that stretches, and so on.
One day we may have 600ppi screens, but we may never need to go that high. Display manufacturers have shown small screens with a resolution of 300ppi, and resolutions of ~200ppi are becoming common in the small screens used in cell phones, digital cameras and PDAs.
It’s not that these displays can’t be manufactured in large sizes. I have an IBM display in my office that’s 22 inches and change, and has a resolution of 204ppi, and I’ve had it for at least eight years. It still works perfectly, BTW – a tribute to its engineering – and I’m currently driving it with a Dell PC running Windows Vista.
The problem isn’t the manufacture – it’s the math that’ll kill you. Let me explain…
Most screens today are around 100ppi. To go from there to 200ppi isn’t a x2 jump in processing power required – it’s n-squared, or 4x (twice the number of pixels in each dimension). To go to 600ppi is 36x graphics processing. You can handle the graphics processing required for a small 200ppi screen, but a larger one is quite another story.
For instance, my 204ppi display – with a fast graphics card – will still only refresh at 13Hz (13 cycles per second). On an LCD display, that’s actually acceptable and usable. On a CRT it would be instant epilepsy… And don’t try watching video on it!
Why so slow? Well the screen has 3840 pixels x 2400 – about 9.2 megapixels. With 24-bit color, that takes 3 bytes per pixel. So we’re close to 28 Megabytes per screenful of pixel data. Now imagine you’re trying to refresh that even as slow as 60Hz (which you need for full-screen, full-motion video), and you’re up to a whopping 1.6Gb per second of graphics processing! You need a monster graphics card, and a graphics bus that won’t bottleneck data throughput.
ClearType helps a lot here. Since it uses the RGB sub-pixels, it effectively triples the resolution of the display you can address in one dimension. Of course, it does nothing in the other dimension (although later versions of the technology actually do some work on that), so it’s not a 3x resolution multiplier. What the true resolution multiplier is, I’ve no idea. People have made guesses and suppositions, but no-one really knows for sure because there’s no agreed way to work it out. However, it IS a significant benefit for most people.
Fact is, we need a major leap forward in graphics processing and driver technology to be able to support high-resolution displays at normal sizes. And it’s not just about the speed of the chips – it’s also about the power required, and the heat they generate. It’s no accident that all the high-power graphics cards on the market today have their own inbuilt fan.
Dell has made more strides in the area of high-resolution support than any other PC company. Dell shipped the first 133ppi display more than a decade ago, and then went on to ship 147ppi displays in their Inspiron series of laptops. I’ve been running one for many years. Dell did seem to have figured out how to drive that many pixels (1920 x 1200, or one-quarter the number in my IBM display) at an acceptable rate, although those laptops were pretty power-hungry. I always used to take a spare (huge) battery on a plane ride. And yes, the laptop ran pretty hot, too.
But therein lies my dispute with David. I’ve tried grayscale on that laptop, and it’s nowhere near as good as ClearType. The pixel’s still too big. Or is he suggesting that we all switch to monochrome displays but with the 3x (RGB) sub-pixel pattern in place? I can’t see that flying – we all like our color far too much.
I use ClearType even on my 204ppi display. You can still see the difference. And at that resolution, color fringing is completely gone.
As you increase in resolution from today’s ~100ppi towards 600ppi, you of course see a difference in clarity and sharpness. But in my view it’s not a linear change, it’s a flattening curve. The big improvement comes between 100 and 200ppi. After that, you’re in The Law of Diminishing Returns, where the rate of perceived improvement flattens out pretty dramatically, no matter how much more resolution you throw at the problem.
So here’s my contention: somewhere between 150 and 200ppi with ClearType is enough. We really don’t need to go farther; there’s little point. We could wait until there’s some new major breakthrough in graphics processing technology. That could happen, for example, by using a parallel graphics processor, with each core of the chip handling one segment of the display, if we can figure out a low-power technology and a way to run without a huge heat problem.
But 144ppi is still not quite good enough. Especially when you think beyond Latin-based languages to Japanese, Korean and Chinese, where even 144ppi still does not give you enough pixels at normal reading sizes of 9-13points to portray all the strokes of some of the more complex characters.
I suspect, given where most of our hardware is manufactured, that it will be the requirements of these languages which will drive high resolution on larger displays.
As an illustration of the heat problem, the Toshiba TabletPC on which I’m typing this is a convertible model; it doubles as a laptop. But that’s a real misnomer. During operation, it gets so hot you’d never put it near any area as sensitive as your lap. We’ve hit a limit in the technology that the hardware and driver guys have yet to solve.