Macbook Pro, retina screens, hinting, the future

gfrederickk's picture

Hello all,

In light of yesterday's unveiling of its Macbook Pro (with Retina Display), seems like good time to restart this question:
What do you see as the future of TrueType hinting? The short-term value is obvious. I get it. I'm with you.
But when breaking out the checkbook, how do we measure the value of hinting in the context of improving rasterizers, iOS's limited appetite for hinting, screen resolution, etc.?

This is an honest question, not a statement. It's a wonderful, daunting, inspiring, confusing time of possibilities and pitfalls.
Love to read some thoughts on this, in light of Apple's news. Many thanks.

John Hudson's picture

the pixel unit is a "relative" unit

It was, but my understanding of the new definition in CSS3 is that 1 px is now defined as an absolute unit equivalent to 1/96 inch.

gfrederickk's picture

John ---

Correct. From the CSS3 spec (March 2012):

"For print media and similar high-resolution devices, the anchor unit should be one of the standard physical units (inches, centimeters, etc). For lower-resolution devices, and devices with unusual viewing distances, it is recommended instead that the anchor unit be the pixel unit. For such devices it is recommended that the pixel unit refer to the whole number of device pixels that best approximates the reference pixel. "

"The reference pixel is the visual angle of one pixel on a device with a pixel density of 96dpi and a distance from the reader of an arm's length. For a nominal arm's length of 28 inches, the visual angle is therefore about 0.0213 degrees. For reading at arm's length, 1px thus corresponds to about 0.26 mm (1/96 inch). "

Based on the responses, I'd say this is the pure form of the truth, but there are real-world implementation compromises.

Nick Shinn's picture

<"The reference pixel is the visual angle of one pixel on a device with a pixel density of 96dpi and a distance from the reader of an arm's length. For a nominal arm's length of 28 inches, the visual angle is therefore about 0.0213 degrees. For reading at arm's length, 1px thus corresponds to about 0.26 mm (1/96 inch). "

A circular argument.
The stuff about the angle and reading distance is redundant, a pleasantry, methinks.
Also, if the reference pixel is an angle, it’s not precise, as the 28" could be either to the edge of the pixel, or the centre—which are different distances (assuming the pixel has a flat surface).

John Hudson's picture

Nick, 'visual angle' is a measure of object height at a set distance (in this case 28") determined as the angle of arc at which the object is subtended at the eye. It is indeed precise, and of course the height is a measurement from top to bottom object edge, since that is what is being measured.

Té Rowan's picture

@Nick – Nothing circular or otherwise elliptic in there, just clumsy wording. 25.4 (mm in 1in) / 96 (dpi) = 0.26458333... (pixel width in mm). The visual angle presented is correct, too. I just checked both on a pocket calculator.

joeclark's picture

We have device and reference pixels in CSS, which is merely one method of reckoning them. As such, pixels are not immutable objects, I say again.

I believe what is really interfering with our understanding here is original Mac 128/512/Plus/Classic/Color Classic pixels, every one of which was individually discernible to a person with good vision. (Surely you remember the 1984–86 fetish for “woodcut”-like illustrations laboriously clicked into existence in MacPaint.) This 1:1 equivalency tainted our mental image of what a pixel must be. We persist in this misapprehension even as every one of us in this thread type away on machines that use more than one screen dot to represent a pixel.

John Hudson's picture

I don't know what taints the mental image of web designers. As a type designer, I still mostly operate in a world in which the most significant use of the term pixel still relates directly to device dots: pixels per em.

Rob O. Font's picture

"...absolute unit equivalent to 1/96 inch"

Correct! 1 px = 1.333333333333333 points.

"...'visual angle' is a measure of object height at a set distance"

Correct! The 'visual angle' is and angless object.

Nick Shinn's picture

@John: 'visual angle' is a measure of object height at a set distance (in this case 28") determined as the angle of arc at which the object is subtended at the eye. It is indeed precise, and of course the height is a measurement from top to bottom object edge, since that is what is being measured.

If the object size is already known, then the visual angle is not the reference, the object size is.

If the object is flat, then its distance from the eye is not exact, because it is further away at the edges than the centre.

Té Rowan's picture

Forgot to mention it, but a pixel's visual angle is arctan( pixel_width / distance_between_screen_and_eye ).

Oh, and @dberlow, your number is correct for a 54ppi screen.

Rob O. Font's picture

It was backwards day:) Correct! 1 point = 1.333333333333333 pixels.

But how come we need anything at all between font master ppm and device dpi?

Té Rowan's picture

Sodomy non sapiens...

aluminum's picture

"It was, but my understanding of the new definition in CSS3 is that 1 px is now defined as an absolute unit equivalent to 1/96 inch."

Well, that's the reference pixel browsers should (but may or may not) adhere to. And one based on a desktop screen. Which is hardly the norm anymore, so likely not all that useful. Alas, W3C recommendations take quite a while to flow through the pipeline.

That said, if it were widely supported, that'd be great as it'd move us away from resolution dependance.

John Hudson's picture

David: But how come we need anything at all between font master ppm and device dpi?

The CSS3 pixel isn't between font ppem and device dpi (at least, no more than the point or the millimetre etc. can be said to be between them). In fact, it isn't a pixel at all in any physical or raster understanding of the term: it is just a new reference unit of measurement, like a point or a millimetre, by which the font outline is to be scaled and then rasterised to the closest approximation given the ppem and the device resolution. The only reason it is called a pixel is that the goal of the new definition is for user agents to be able to take all that old web content spec'd in px units and make it big enough to see on higher resolution displays. It is easier for W3C to make the px unit device independent than to require web developers to rebuild their sites using existing device-independent units.

Rob O. Font's picture

"The CSS3 pixel isn't between font ppem and device dpi..."

It isn't!? I thought it says, "Hi, I'm a pixel but not a device pixel. I'm a pixel that's some number of device pixels. Nice to meet you."

I'll be sad if it isn't who it says it is, 'cause I've wrapped my head around it, and now I'm going to hold my breath until it dies.

John Hudson's picture

"Hi, I'm a 'pixel' but I'm not actually any kind of pixel at all: I just have a backwards compatible name for a non-backwards compatible unit equal to .75 point. My relationship to actual pixels is the same as that of .75 point. If, every time you see me, you remember to mentally substitute .75 point and ignore what I'm called, we'll get on fine."

gfrederickk's picture

Here's a couple of goodies that have helped me wrap my head around this. In addition to the sage counsel provided in this forum, of course.

http://www.youtube.com/watch?v=4wscVOXjIzQ

http://blog.cloudfour.com/pixels-are-ruining-my-life/

Nick Shinn's picture

Ideally, device screens will have a rangefinder with iris-recognition software which computes the viewer’s distance. That would enable accurate (default) specification of image sizes in terms of angle-of-vision*.

From this, page designers would figure out what kind of resolution they would like, in terms of matching image pixels to various screen pixel dimensions.

Rather similar to spec’ing a photo for print, taking into account how sharp the original is, what screen value the publication is, and thence deciding how big the final image can go before reproducing original-image artefacts such as film grain; the digital equivalent would be compression and anti-aliasing artefacts.

*This principle of constant optimum size is not presently employed when one rotates an iPad thru 90°, and image and font size change.

JamesM's picture

If any of you are using a Macbook Pro with retina display, you might find this useful — Apple has updated a FAQ page about retina displays.

http://support.apple.com/kb/HT5266

Rob O. Font's picture

Thanks James. I'm particularly fond of the "larger text - just right - more space" concept.

"http://www.youtube.com/watch?v=4wscVOXjIzQ" this was really interesting. I'm quite font of speakers who "hope I don't confuse you" 4-5 times an hour. ;)

But in the end the speaker in this presentation doesn't know the actual size of anything and just hopes his sites start out reasonably well and end well for the user via resolution, pan, zoom and chewing gum.

Did anyone ever mention to the web consortium that type works best when it scales non-linearly? and that at the bottom and the top of the actual size scale this non-linearly is relatively precipitous? Just a thought:)

John Hudson's picture

David: Did anyone ever mention to the web consortium that type works best when it scales non-linearly?

I mentioned it to them, complete with illustrations (CSS3 briefing, TypeCon Los Angeles, following webfonts working group meeting), and everyone agreed that my non-linear, size-specific type looked very crisp and legible, but that having text reflow when zoomed is a non-starter. So unless you're going to make a distinction between scaling and zooming -- i.e. a distinction between what happens to the type depending on how it is made bigger/smaller -- there's going to be zero uptake for non-linear biggering and smallering. Put it another way: user experience of blocks of text at variable sizes trumps user experience of reading text at particular sizes.

Once again, I'm not endorsing this view, just reporting it.

k.l.'s picture

Did anyone ever mention to the web consortium that type works best when it scales non-linearly?

I don't think that this is true, or rather, I think that this is dealing with a problem by way of one specific approach when there are other (and less 'destructive') approaches too.

hrant's picture

Repeat: WYSIWYG is a succubus.

hhp

quadibloc's picture

@Hrant:
WYSIWYG is a succubus.

I would like it if one could use word processors in a mode that allowed the body copy to flow automatically, so that one could change the point size of a whole document and just check that everything worked out right, instead of reformatting each page...

Think of old-style text formatting programs - such as the one that still survives in Unix, troff.

However, for a word processor to let you see what your printout is supposed to look like, so that you can minimize your use of paper, instead of making multiple printouts during the revision process, is, I would still say, a good thing, not a bad one, even if there are technical challenges in bridging the gap between the screen and the printer.

Recently, it was the lower resolution of the screen that was the problem (requiring the problematic technical solution of hinting), in the past, when many people used daisywheels instead of printers, or when laser printers had their own fonts (actually still true of specialized ones today) WYSIWYG (of course, that phrase has an offensive origin, so in the very early days, they risked insulting secretaries by using it in their presence), at least when proportional spacing was used, had problems in the other direction - the computer had to be informed about the metrics of the font used within the printer, as it couldn't impose the screen font it was using on the final copy.

John Hudson's picture

Hrant, when it comes to reading on screen, as opposed to previewing print, WYSIATI -- What You See Is All There Is -- and WYSINWSES -- What You See Is Not What Someone Else Sees. The ideas that informed the succubal WYSIWYG are irrelevant. The gap we need to mind has moved.

Rob O. Font's picture

Really? I think the mind we need to move has a gap.

Té Rowan's picture

WYSIAYG – What You See Is All You Get.
WYSIWYGAINGW – …And It's No Good Whining!

@quadibloc – I think that most instances of troff nowadays are as part of groff. Mind, the Heirloom Toolkit has a free-standing one that can deal with TrueType fonts.

Rob O. Font's picture

Hmmm, John?

“…non-linear, size-specific type looked very crisp and legible…”

Oh no, that’s not the point.

“…having text reflow when zoomed is a non-starter.”

Text reflow, having it and stopping it are entirely up to the developer and/or user.

“…So unless you're going to make a distinction between scaling and zooming…”

Resizing, is a lot like reflow in that its ways of happening and whys are up to the developer and/or user.

“…there's going to be zero uptake for non-linear biggering and smallering…”

It’s already taken-up big-time by the public in both old and new media.

So, in general, we’re reading a 3-yr old report that says “if non-linear scaling from a single font file were allowed as an option, no one would be safe.”

The problem comes with “responsive media” vs. “Write HTML pages, post HTML pages, view HTML pages. Adjust. Repeat.”…

…(on either end of the point size spectrum), without “responsive type” as the resolution spectrum is attenuating, according to this thread...

Chris Dean's picture

A similar typophile thread: Will hinting be necessary when 300dpi high-res screens are the standards?

Really? This conversation again? Yeesh.

Posted elsewhere:

http://prometheus.med.utah.edu/~bwjones/2010/06/apple-retina-display/

“So, if a normal human eye can discriminate two points separated by 1 arcminute/cycle at a distance of a foot, we should be able to discriminate two points 89 micrometers apart which would work out to about 287 pixels per inch. Since the iPhone 4G display is comfortably higher than that measure at 326 pixels per inch, I’d find Apple’s claims stand up to what the human eye can perceive.”

See also:

Curcio, C. A., Sloan, K. R., Kalina, R. E. & Hendrickson, A. E. (1990). Human photoreceptor topography. Journal of Comparative Neurology, 292, 497–523.

Long short, there’s only so much detail the human eye is physically capable of detecting. Sure, if we get in an look at the type 1 cm away we’ll see a difference, but at 12 inches, ~300 DIP is our threshold. Beyond this neither resolution nor hinting make a significant difference.

Take the hint. It’s pretty much dead (and before you say it, yes, there will always be a crappy ATM in a third-world country airport).

It’s also important to point out that the people arguing for the importance of hinting are quite possibly the most biased population you could ask. The bulk of them stand to profit from its significance. You’d be hard pressed to get a type designer who gets paid to hint to admit what they do is makes no significant difference aesthetically, in measures of human performance, and will eventually become extinct.

Si_Daniels's picture

"Sure, if we get in an look at the type 1 cm away we’ll see a difference, but at 12 inches, ~300 DIP is our threshold."

Yep, but the screen in question is 220 DPI.

John Hudson's picture

What Si said.

I've no problem with the notion that hinting gets less relevant at higher resolutions and will eventually be either unnecessary or easily automated for the kind of resolutions we're looking at in an increasing number of mobile devices. But we're not there yet.

I'll also point out that hinting isn't just a technology for making type look nice in otherwise friendly conditions, which seems to be all you're thinking about. Hinting is also a tool for addressing frankly unfriendly conditions, such as inherited vertical device metrics. We spent a lot of time over the past year trying to squeeze ten Indic scripts into the vertical metrics of a Latin UI font -- don't ask! -- and hinting was essential to the process of making the forms legible at target ppem sizes.

Its not like we're going around the place selling hinting services to people who don't need them, Christopher. Clients come to us because they discover problems that need hinting solutions.

PabloImpallari's picture

I'm not concerned about hinting.
I'm just happy that retina displays will allow us to add much more detail to our fonts.

Rob O. Font's picture

"I'm just happy that retina displays will allow us to add much more detail to our fonts."

And how much detail did you need to remove before?

And, i hint, for longer than anyone else and i admit it — where it makes no significant difference aesthetically even in unmeasured measures of human performance, it will eventually become extinct.

How hard was that press pressed?

PabloImpallari's picture

"And how much detail did you need to remove before?"

For example:
In standard screens, concave stems becomes visible at about 40px or up.
In retina, you see the concave effect at 16px.

You don't need to remove anything, but below a certain threshold details becomes invisible. In retina, instead, that threshold is much smaller and much more details are now visible (for good or worst :)

hrant's picture

Christopher, your data is flawed, as is your interpretation. One thing consistently ignored by the "300 dpi is enough" people is that we have two eyes! And this could very well translate to (and anecdotally does seem to translate to) much more than a mere doubling. See the end of this page for an example:
http://www.bbc.com/future/story/20120719-awoken-from-a-2d-world/2

hhp

quadibloc's picture

I would tend to think that 300 dpi is enough for a display, because we know from our experience with laser printers that while 60 dpi is crude dot-matrix printing, 300 dpi seemed indistinguishable from the product of fully-formed character printers.

Of course, it is also true that this initial impression was flawed, because 300 dpi was still a tad fuzzy, and certain types of subtle detail in some typefaces - notably the flared stems of Optima, for example - were lost.

But in a display, unlike with a laser or ink-jet printer, one can control the color of each pixel exactly. So if one combines a technique like Microsoft's ClearType with a 300 dpi display, which, unlike hinting, can be carried out in an automated fashion, I think the results will meet any reasonable standard of adequacy.

That doesn't mean that there's anything wrong with pursuing higher resolutions when the technology gets to the point that they will be easily available, but it does mean that at 300 dpi, there will not be any pressing need to seek further improvements in display resolution... and, thus, there will be a disinclination on the part of consumers to pay premium prices for higher resolutions beyond this level.

Syndicate content Syndicate content