Why are the ClearType font so small?

ralf h.'s picture

All the ClearType fonts are significantly smaller than the old Windows Core Fonts. I guess there is a technical reason for that. Does anyone know why that is?

The problem is: this makes it almost impossible to use the CT fonts for body copy texts on websites. If you declare the fonts in a CSS like:
"calibri, Verdana, Arial, Helvetica, sans-serif"
you will have to choose the right type size for either the ones who have the CT fonts or the ones who doesn't. If you focus on the old Core Fonts, the CT fonts will be rendered much too small and if you focus on the CT fonts, users without it will see much too large type. So as a web designer you are really in a pickle here …

There were plans for a CSS attribute that would scale the fonts according to the x-height (font-size-adjust), but it has been dropped in the current CSS specifications.

Any ideas or solutions on this subject?

Ralf

kuroneko's picture

When you open a CT font with Fontlab you can see that all the letters are two times bigger than as usual, maybe it's linked…

David Sudweeks's picture

Somebody tell me there's a better way to do this. You check what browser and version they are using, and by that—you guess whether or not that means the CT fonts are being used—and serve up the appropriate CSS. (I believe some browsers have implemented full x-height specification support) Yours is a great question Ralf; my guess is that the answer is most likely to first come from a different community than this since its effects reach far beyond that of type design and use.

ralf h.'s picture

Well, one could:
* Use Flash to retrieve the users font list
* Pass the list to Javascript to check if the ClearType font is installed
* use DOM scripting to change the CSS class definition. (or better: store the result in a cookie to use the CT class defition for every following page)

I made an example page:
http://www.fonts.info/cleartype/
It checks for Calibri and replaces the paragraph's CSS class from "nocleartype" to "cleartype".
(Doesn't work in IE yet. The javascript call from Flash seems to be ignored. Don't know why.)

But all this trouble for one font replacement!? So I wonder if there was a good reason to make the Cleartype fonts so small.
Are we not supposed to use them for websites?

Ralf

blank's picture

I’ll just be pedagogic and say that consistent font sizing should not be a consideration in XHTML/CSS sites. Disabled users (or guys like me who can’t handle screen type under 12 points) need to be able to scale fonts sizes to something that they find readable; given this, a good layout should be flexible enough to accommodate both Calibri and Verdana.

ralf h.'s picture

I totally agree about the flexibility. But most users will always use the website's default settings, so these values have to be reasonable in the first place (And that's the problem with CT fonts). Whether the layout can deal with font-size changes is a different topic.

blank's picture

I see what you mean now, and you’re right, it’s a problem. I’d say web designers will just have to use workaround or avoid the Cleartype fonts until Vista is widespread and Apple has licensed the fonts.

Uli's picture

> All the ClearType fonts are significantly smaller

Not all ClearType fonts are significantly smaller. For instance, Constantia has correct metrics, but the other fonts have incorrect metrics, i.e. the other ClearType fonts are technically faulty.

The top height of capital letters of Latin typefaces should be 70% (+/- 5%) of the font grid. PostScript fonts have 1000 units per em, and hence the top serif of the uppercase "I" should be 700 units in the case of PS and OTF fonts.

For TrueType and also for the ClearType fonts, the grid is 2048 units, and 70% of 2048 is slightly above 1400 units (i.e. 1433 units, to be exact).

In the case of Constantia, the height of the capital "I" is exactly 1400 units, and this is perfectly correct. A deviation of +/- 50 units (i.e. the range below 1400 down to 1350 and above 1400 up to 1450) is tolerable for TrueType fonts, but if the caps height drops below 1350, the metrics must be regarded as incorrect, because such fonts can no longer be mixed with other fonts in the same line of text, since the letters begin to dance (see below).

The metrically worst ClearType fonts are Calibri, Candara and Consolas, all three of them with a caps height of only 1300 units. Cambria is tolerable with a caps height is 1365 units.

If we mix Constantia (which has correct metrics) with Calibri (with has faulty metrics) in the same line of text, the faulty metrical design becomes apparent by "dancing letters", whose top capital heights do no longer align:

see http://www.sanskritweb.net/temporary/calibri.pdf

It is hard to believe at first glance that all dancing letters in the file calibri.pdf are typeset in exactly the same point size (here 48p), but this is the case, as anyone can find out himself or herself by typesetting a line of sample text using and mixing Constantia and Calibri.

Kudos to Mr. Hudson, who was able to design his Constantia font with correct metrics (and non-kudos to Mr. de Groot, who was not capable of designing Calibri with correct metrics).

Nick Shinn's picture

The top height of capital letters of Latin typefaces should be 70% (+/- 5%) of the font grid.

This is not a technical requirement.
The idea that cap heights should be fairly consistent between faces is spurious, as there are many factors that may come into play when one mixes typefaces on the same line, and often relative x-height is more important.

However, without trying out combinations of Cleartype fonts in a page layout, I wouldn't want to stick my neck out further.

Are the Cleartype fonts intended to be used with one another? How far was this a criterion of the development process?
It doesn't seem to be much of a factor, given the format of "Now Read This" which, like the vast majority of type specimens, shows new faces in isolation.

How do the bolds of Calibri and Corbel (as sideheads, for instance) work as contrast faces with Constantia and Cambria (for the rest of a text blocks)? Show that in a page layout, and you would have a more meaningful comparison than a row of letters.

crossgrove's picture

Oh, Nick, that would be boring. Then Uli can't categorically condemn a designer's abilities.

Uli, from what rule book did you get your pronouncement about the cap height of typefaces? There's not much call for lines of all caps, in mixed fonts of any kind; there is common use of different styles of the same design in the same line. X-height is indeed a more useful alignment. Have you taken up grunge typography? Sounds like another pointless opportunity to whine. And you didn't address the actual question, which pertains to web use. Since the CT fonts were made much later than Georgia/Verdana, and some of them have much larger, more diverse character sets (see Cambria math), perhaps there is something you are missing. For starters, I suggest you look beyond the designer's "capabilities" to the client's specifications. None of the CT designers acted in a vacuum.

blank's picture

Uli, from what rule book did you get your pronouncement about the cap height of typefaces?

The Elements of Martian Typographic Style.

Uli's picture

Uli, from what rule book did you get your pronouncement about the cap height of typefaces?

The Elements of Martian Typographic Style.

The title of the book is somewhat different. The book is entitled "Type 1 Font Format" published by a company called Adobe. Did you ever hear of this outfit? They claim to have to do something with fonts. On page 24 of this book, they claim that the typical value of the caps height is 700, which is 70% of 1000 units.

And there is another person claiming to have knowledge of fonts. He is called Peter Karow. Did you ever hear of him? He published the voluminous 347-page book "Schriftstatistik" containing thousands of statistical metrical analyses about thousands of fonts (and he incidentally also confirmed the above 70% thumb rule).

Nick Shinn's picture

they claim that the typical value of the caps height is 700

They also state that x-height is typically 450.
Yet there are genres of type that differ greatly from this proportion.
Might not the same be true for capitals as for x-height?

You are claiming that what is atypical is incorrect, and stating that designers who decide to use atypical proportions are "incapable of designing with correct metrics".

Spire's picture

A typical American male is five feet eight inches tall.

Non-kudos to the parents of Kareem Abdul-Jabbar, who were not capable of producing offspring with correct metrics.

(His metrics must be regarded as incorrect, because he can no longer be mixed with other humans in a line of airplane seats, since the knees begin to collide into backs.)

k.l.'s picture

The Adobe PST1 Specs say "typical", not more. And Schriftstatistik contains exactly that: statistics. It is data provided "for your consideration", not more.
The great thing about digital type is that it freed designers from restrictions as regards the body, and thus ascender and descender heights.

The problem is not fonts, but applications. OT and TT fonts already include x- and H-height values, so application could well allow a typographer to define type size by x- or H-Height (or ascender or descender height, if you like). If applications or CSS don't do this, it is not the fault of font formats or type designers.
The x- and H-height ratio is a design decision. Imagine Bickham with "correctly" dimensioned ascenders/descenders and according vertical metrics -- would be a nice caricature.  :)

Uli's picture

The question of Mr. Herrmann was whether you can substitute Arial etc. by Calibri etc. in as CSS, i.e. in a website cascading style sheet.

The answer is NO, because the rendered point size of Calibri is much smaller than the rendered point size of Arial etc. due to the incorrect metrics of Calibri.

I enlarged my initial calibri.pdf document by two additional examples:

see new file http://www.sanskritweb.net/temporary/calibri.pdf

The first new example shows mixing of Constantia with an Indic font with 300 units caps height.

The second new example shows mixing of Constantia with Calibri with 3oo units caps height.

These examples make it clear that you cannot use two fonts in the same text, if the two fonts do not have the same metrics.

Mr. Lücke claims that "applications could well allow a typographer to define type size by x-height or H-height". This is the typical "frog in the well" attitude shown by most Typophilers. Since most of them never learnt Indic scripts etc., they do not know that most non-Latin fonts do not have lowercase and uppercase letters. They behave like the Japanese fairy-tale frog in the well who know nothing of the rest of the world.

k.l.'s picture

The answer is NO, because the rendered point size of Calibri is much smaller than the rendered point size of Arial etc. due to the incorrect metrics of Calibri.

The answer is NO, because CSS does not "scale the fonts according to the x-height" as Ralf Hermann wrote in his original post. That's all.

... shows mixing of Constantia with Calibri ... These examples make it clear that you cannot use two fonts in the same text, if the two fonts do not have the same metrics.

Therefore it is still up to typographers to combine typefaces that match. Either one decides that matching x-height is more important (text), or matching H-height is more important (all caps setting).
And if all metrics should match, type designers sometimes make the decision to design super-families with a serif and sanserif family, see Peter Bilak's thread over here. But this involves more than just vertical metrics, of equal importance are the typefaces' color (weight), style of curves, etc.

This is the typical "frog in the well" attitude shown by most Typophilers. Since most of them never learnt Indic scripts etc., they do not know that most non-Latin fonts do not have lowercase and uppercase letters.

My brother is into indoeuropean studies, so I am aware of this and other problems (non-standard encodings, use of strange tools to get at Indic scripts' ligatures, etc) which make it a pain to typeset even a simple scholarly text.

Rob O. Font's picture

"I guess there is a technical reason for that. Does anyone know why that is?"

The "CT Collection" are smaller on the body than Verdana and Georgia because they were made differently, i.e. not for use specifically as screen fonts, not maximized for use at small sizes in low resolutions. That's all I know for certain.

Cheers!

Uli's picture

> The “CT Collection” are smaller on the body than Verdana and Georgia

Constantia must be ruled out, because the caps height of Georgia is 1419 units and that of Constantia is 1400 units, so that both fonts have more or less identical caps height metrics.

Cambria must also be ruled out, for its caps height of 1365 units is sufficient.

So, the main culprits are Calibri, Candara and Consolas, whose caps heights are insufficient and untolerable.

Nick Shinn's picture

I'm working on a super family at the moment, where the x-height of the sans and serif regular fonts are different.
My decision is based on the way people use type in layouts.
Generally, different weights and styles are more likely to be contrasted in adjacent paragraph blocks than in adjacent words in the same line. So the important thing is to give your types proportions which create harmony in that kind of usage.

But as I've said, I haven't worked with the Cleartype fonts.
I don't think that one can make meaningful statements about the way types interact on the page by comparing a few dimension -- you have to see it in context as well.

Thomas Phinney's picture

Don't feed the troll.

Nick Shinn's picture

I may occasionally disagree with Uli but I like his attitude.

John Hudson's picture

Cross-family consistency in the CT fonts was modelled on having identical vertical metrics in the OS/2 table, so that fonts from the different families could be used together on the same line of text without the interline spacing being distorted by the presence of a font with larger OS/2 metrics within a block of text. Since one of the prime targets for these fonts was MS Word, and we knew most users would not manually set interline spacing values but accept default linespacing, I still think this was a sensible decision.

Cross-family visual consistency was aimed at by coordinating similar x-heights not similar cap heights (as Uli seems to want). Most text is in lowercase letters, not all caps, so coordinating cap height seems to me pointless and guaranteed to create inconsistent text colour unless all fonts have an indentical cap-to-x height proportion. If I recall correctly, the Constantia x-height is slightly shorter than the other CT fonts, and the Consolas x-height is slightly taller. But the others are very close.

Uli's picture

Cross-family consistency in the CT fonts was modelled on having identical vertical metrics in the OS/2 table, so that fonts from the different families could be used together on the same line of text without the interline spacing being distorted by the presence of a font with larger OS/2 metrics within a block of text. Since one of the prime targets for these fonts was MS Word, and we knew most users would not manually set interline spacing values but accept default linespacing, I still think this was a sensible decision.

I agree that this was a sensible decision. However, there is the problem that the glyph metrics differ noticeably among the CT fonts (e.g. Calibri versus Constantia) and therefore only by voluntarily fixing identical OS/2 values this identical interline spacing could be achieved.

Cross-family visual consistency was aimed at by coordinating similar x-heights not similar cap heights (as Uli seems to want). Most text is in lowercase letters, not all caps, so coordinating cap height seems to me pointless and guaranteed to create inconsistent text colour unless all fonts have an indentical cap-to-x height proportion. If I recall correctly, the Constantia x-height is slightly shorter than the other CT fonts, and the Consolas x-height is slightly taller. But the others are very close.

I will not discuss here whether caps heights are "pointless", because the fact that for instance in German texts capital letters are used much more frequently than in English texts can be ignored here, since the CT fonts were designed for Americans and not for Germans etc.

The problem with Calibri is not the coordination of caps heights, but that ALL glyphs of Calibri, i.e. both lowercase and uppercase glyphs, do not sufficiently fill the glyph grid with the consequence that the glyphs look too small. The grid is the same, but there is too much white space around the black glyphs (if printed in black color on white paper), and hence they look too small as compared with Arial, Helvetica and similar sanserif fonts.

Let's look at

http://www.sanskritweb.net/temporary/calibri-arial.htm

In case you don't have Calibri and Arial installed on your computer, you can instead look at the PDF file

http://www.sanskritweb.net/temporary/calibri-arial.pdf

It is obvious that Calibri in the left column looks much smaller than Arial in the right column, although the sample texts are typeset in the very same point sizes. How come that Calibri looks smaller than Arial? The answer is that the glyphs of Calibri do not sufficiently fill the 2048-grid of this TrueType font, as compared with Arial and Helvetica and similar fonts.

Let's look at

http://www.sanskritweb.net/temporary/grid2048.gif

To the left we see the uppercase "I" of the TrueType Helvetica (version 1.0), which has a caps height of 1470 units, followed by the uppercase "I" of Calibra, which has a caps height of 1300 units.

To the right we see the lowercase "x" of the TrueType Helvetica, whose x-height is 1071 units, followed by the "x" of Calibra, whose x-height is 956 units.

This means that the absolute values of BOTH caps height AND x-height of Calibra are too small, and hence there is too much white space around the glyphs in the 2048 units TrueType grid, and hence Calibra looks too small.

Since the caps height occupies usually two thirds (2/3) of the grid, i.e. roughly 70% of 2048 units, it follows that the caps height measured in grid units or in percentage can be used to determine whether a font looks too small or not. This can be proved by the following test:

Let's look at

http://www.sanskritweb.net/temporary/caps-height.pdf

In the first row (H1470-100%) we see the original TrueType Helvetica with a caps height of 1470 units. This original vertical size is regarded here as 100 per cent for comparison.

In the second row (H1176-80%) we see the down-scaled Helvetica, where all glyphs (both uppercase and lowercase) have been reduced to 80% of the original size of all glyphs.

In the third row (H882-60%) we see the down-scaled Helvetica, where all glyphs (both uppercase and lowercase) have been reduced to 60% of the original size of all glyphs.

Now look at the three "Hamburgefonts" sample texts, all three of which were typeset in the very same point size of 36 points on 36 points body:

The H1176 sample text looks smaller than the original H1470 sample text, and the H882 sample text looks still smaller than the other two sample texts.

This proves: Whenever the glyphs of a font do not sufficiently fill the 2048-grid of TTF fonts (or the 1000-grid of PS fonts), the glyphs look too small, because there is too much white space around the black glyphs.

That is why Calibri looks much smaller than Arial, Helvetica and similar sanserif fonts, since the glyphs of Calibri (both lowercase and uppercase) do not sufficiently fill the 2048-grid.

Nick Shinn's picture

Perhaps the smaller size wasn't done for the benefit of American readers, but Slovak, Vietnamese etc., -- to accomodate diacritics.

I don't think that being "small on the body" is inherently wrong for a face -- different conventions exist. For instance, news faces are large on the body, and many Emigre fonts are small on the body -- and generously spaced -- to create an open look straight "off the shelf".

In fact, as the CT fonts are targeted at Word, the built-in open leading is part of an appropriately less strident effect than, for instance Verdana, which for all its merits, is a strong cup of tea.

So what we're seeing is a more sophisticated look from Microsoft.

John Hudson's picture

It is obvious that Calibri in the left column looks much smaller than Arial in the right column, although the sample texts are typeset in the very same point sizes. How come that Calibri looks smaller than Arial? The answer is that the glyphs of Calibri do not sufficiently fill the 2048-grid of this TrueType font, as compared with Arial and Helvetica and similar fonts.

The OS/2 WinAscent value in Arial is 1854 units. The tallest glyph in the Arial font is /Aringacute/, the top of which is at 2060 units. That means that more than 200 units of visible glyph in the Arial font falls beyond the WinAscent height, which means it will be clipped in all GDI applications and many printers. The uppercase Vietnamese diacritics in Arial are only kept within the vertical metrics by horribly squishing and distorting their shapes.

Many of the world's orthographies need greater vertical space than the inherited scaling and metrics of Anglo-American and western European font development. If you are making fonts that you anticipate being extended to support additional languages, the last thing you want to do is to make them too large on the body, since vertical metrics usually cannot be increased due to backwards compatibility and document reflow issues. This is the trap that Arial and many other core fonts fell into.

Uli's picture

I wonder why both Mr. Shinn and Mr. Hudson mention Vietnamese, because Calibri does not cover Vietnamese diacritics. For instance, Vietnamese diacritics with móc are missing in the Calibri font (and these móc diacritics are also missing in the Constantia font, for that matter).

Uli's picture

Mr. Shinn:

> “small on the body”

Could you please define exactly what you mean by this technical term.

In the old German letterpress printer's parlance we had two completely different terms so that I am not sure which term exactly you mean by your above English expression, dating back to the foundry era, in view of digital fonts under discussion.

Nick Shinn's picture

I mentioned Vietnamese and Slovak because their diacritics take up a lot of space.
But in fact, the most vertically challenging accent to squeeze in, at least in my experience as a type designer, is capital A ring.
So perhaps CT fonts are Scandinavian-friendly.

The "body" is indeed a metal type term, otherwise known as the "shank". In a piece of type, it's the rectangular column of metal of which the letter shape forms the top part.
So "small on the body" means "small on the em square", or, as you might say, "small on the grid".
Yeah, I know, it's an archaic term, but I like the suggestion of a tangible space it gives to the em square -- for me, digital line spacing that deals with the distances between parts of glyphs just doesn't "get" what typography is, and is responsible for a lot of difficulties.

Please, call me Nick, unless you're 16 or younger.

John Hudson's picture

I wonder why both Mr. Shinn and Mr. Hudson mention Vietnamese, because Calibri does not cover Vietnamese diacritics.

It doesn't in the current release version, but you can expect that it will in future versions, and not only Vietnamese. The point I am making is that maxing out the size of the glyphs in the em space without giving thought to the future extension of the typeface for other languages is a mistake. Basing such decisions on what characters a font supports in its first development phase is short-sighted, especially when the font is being made for a company that is dedicated to internationalisation and that has a long history of extending their fonts to support additional languages. The phase 1 CT font development was aimed at modern European languages, but it would be naïve to think that this would be the limit of their support.

Uli's picture

1)

> It doesn’t in the current release version, but you can expect that it will in future versions

I am pleased to learn that Vietnamese diacritics are to be expected in future font versions. Since many Americans do not regard Vietnamese as human beings, but as "weed" belonging to the "axis of evil", and therefore annihilated millions of them as "weed" using herbicides ("Agent Orange" etc.), it is laudable that today some Americans plan to offer fonts in the future for those who survived the herbicidal annihilation.

2)

> it will be clipped in all GDI applications

For those who have not yet experienced this phenomenon, I have prepared an illustrative example.

see http://www.sanskritweb.net/temporary/l-candrabindu.pdf

> the most vertically challenging accent to squeeze in, at least in my experience as a type designer, is capital A ring.

This is a challenging case indeed, but the worst case of all in my experience is not ordinary capital letters with ring etc. diacritic, but lowercase ascender letters such as "l" and "k", which often extend the ordinary capital height, especially in almost all old-style typefaces, as shown by above ascender letter "l" with candrabindu on top of it.

Si_Daniels's picture

My advice would be to avoid "the weed" while posting to typophile. Also the "Americans" in question are Welsh, Irish, Pakistani, Engish and Canadian.

Anyhow, in the process of adding Vietnamese to these fonts we did have issues with Cambria, whereas for Calibri and Consolas things went pretty smoothly.

With respect to clipping it's been our experience that printers do not clip, and that increasing leading mitigates the issue. I feel that for a document font clipping is preferable to having marks that crash together or crash into the base.

Cheers, Si

Uli's picture

The “body” is indeed a metal type term, otherwise known as the “shank”. In a piece of type, it’s the rectangular column of metal of which the letter shape forms the top part.
So “small on the body” means “small on the em square”, or, as you might say, “small on the grid”.

Thanks for the explanation. Now that I know how you define the term "body", I made three experimental fonts. Here is the result of my tests.

http://www.sanskritweb.net/temporary/em.pdf

In case someone wants to experiment too, I can upload the three Em1.ttf, Em2.ttf, Em3.ttf experimental fonts too for download.

Nick Shinn's picture

the worst case of all in my experience

Right. I forgot that I had wrestled with l-acute and h-circumflex!

crossgrove's picture

"My advice would be to avoid “the weed” while posting to typophile. "

And possibly also excessive drink. ;)

Syndicate content Syndicate content