Relative Character Widths

j75's picture

Ok, so making continuous optical adjustments to the individual character widths in a typeface I'm working on is taking forever. Is there a more scientific way of doing it than looking at the letters in context (words, text), making (subjective) judgements about their legibility, adjusting accordingly and then repeating the cycle?

There seems to be an (inate?) optimal width for each glyph. If a glyph is too wide it looks wrong and is distracting. This sense of what is ideal is what I'm trying to use as a benchmark against which to make adjustments. But isnt that a bit too subjective?

Where does this sense come from? Have our eyes just become accustomed to reading letters at a certain width, or is there an actual optimal width at which each character imparts its information and can thereby be recognisable? (Perhaps related to the number of vertical strokes in a chracter, for example)

How does you design legibility (as regards charcter widths specifically) into an original typeface? Is it just optical adjustments or is there a more objective way?

je

hrant's picture

This is a very interesting line of discussion.
I would love to see people brainstorm about it.

hhp

nevick's picture

From a total newcomer point of view.

i think the best way is to write out alot of text, pretty soon you will see holes by certian characters and you can adjust acordingly.. then you can see how that change impacts on the whole text.

basically i belive there is just too many subjective factors for it to be some mathematical way to figure out spacing.

hrant's picture

I can't believe nobody is bothering with this thread.
This is where the beef is, people. No appetite for deep thought?

hhp

capthaddock's picture

Jumping in as a relative newcomer:

An important factor might be where in the spectrum from classical proportions to monospaced proportions a typeface lies.

A classical font like Trajan is built on the forms of the circle, the square, and the triangle. The result is that some letters, like O and C (based on the circle) are twice as wide as letters like E and L (based on two squares).

With a typeface like DIN, on the other hand, all letters strive to look the same width.

There's plenty of room in between, but I expect that the classical proportions should give you a good idea of relative widths between letters. Applying this to lowercase might be another matter.

Paul

hrant's picture

Indeed, there seem to be two models of UC width: the Classical with its great divergence*, and what might be called the Modern with its convergence. But I'm not sure how much there is in between. You do see some faces with a letter or two deviating from the Classical (like the wide "S" of Will Carter and sometimes Zapf, and a wide "E"/"F" here and there), but I can't think of any that deviate much from the Modern - I guess that would look totally lousy. What I'm saying is that your caps have to either diverge a lot or not at all**. Makes sense?

* Although I'm not sure how strong the geometric basis is.

** The only exception I can think of is Nick Shinn's fascinating Richler.

So Jos

boole's picture

Interesting question. I'd like to know more about this subject too, having that I'm an absolute newbie.

I'm guessing that as

geraintf's picture

this is what i did when designing a face recently:
made a table showing character widths.
in my case, it was caps only.
the columns contained my face, the trajan inscription*,
and a bunch of faces that were relevant
to my design (eg. johnson sans, optima)

it was a really interesting exercise,
as i had started off with a 'one width fits all' approach,
but doing the comparisons
made me assign some of the
glyphs a unique width


* i used a book which gives an analysis of the
geometry of the trajan inscription.
i would post the ref,
but i don't have it to hand.

geraintf's picture

actually, paul suggested this approach
in the first place. thanks!

geraintf's picture

talking about divergence from a norm/average value:
i wonder how much UC width variation there is among the various surviving roman inscriptions? eg, if you took a dozen Es from a dozen discrete (surviving/recorded) incriptions how much would the widths vary? has anyone done an analysis of like glyphs from these sources?

poeple have always used the trajan column as a model of classical proportions, but surely in 114AD it didn't have that role: it was just one of many public inscriptions...
(off-topic, sorry)

hrant's picture

Although the Trajan style of lettering isn't just on the Trajan column, I share your suspicion that there's more to it than that. More than a suspicion was aired by Cynthia Batty during ATypI's Rome conference, where she glossed over some "competing" styles of lapidary lettering from ancient Rome, and actually took a swipe at Trajan-worship, to the cheer of people like me.

BTW, can we see your data?

hhp

geraintf's picture

>BTW, can we see your data?

sure, although it'll have to wait a few days, because i'll be away. also, my table was compiled with the development of my typeface 'playford' in mind, so perhaps i should post it over at its critique thread*

*http://www.typophile.com/forums/messages/29/12398.html?1059754435

j75's picture

Maybe typefaces take so long to design and go through so many modifications when passed down the ages (when transfered between technological contexts and passed from foundry to foundry, cut & recut etc), that it seems they just evolve the right dimensions over time...

hrant's picture

Such sentiments of "evolution" are not uncommon in type design, but in the end, since fonts are not animals created by Nature roaming the wilderness for food and mates, they seem to me too metaphysical to be reliable.

Not to mention that there are two schools (as mentioned above): the Classical and the Modern - both of which are valid.

hhp

bieler's picture

"... took a swipe at Trajan-worship, to the cheer of people like me."

Hrant

Don't know if this was brought up but Matthew Carter wrote an article in Printing History a while back about how the lettering on the columns was size optimized. The letters are modified at various heights on the column so that they look all the same from the perspective of the viewer. Something like that. If true, not bad for "ancient" Rome, huh!

bieler's picture

I think this is the correct reference:

Printing History. Whole Number 26-27 (Volume 13, No. 2 and Volume 14 no. 1) 1991-1992

Matthew Carter, Theories of Letterform Construction. Part 1.

hrant's picture

> Our Latin alphabet has been designed and redesigned

Actually, it's only been designed (successfully) once: the Carolingian standardization. To me the rest is too subconscious to be called design. What we do design is letters, but not the alphabet per se.

Alphabet design is very rare, although in one case -Korean Hangul- it resulted in a writing system that makes all others look like village idiots.

> the most successful types of all time are are drawn very closely on the same character widths.

Do you mean the same as the Trajan model, or the same as each other?

> study what has gone before.

Great advice indeed.

--

Garald, I have that article. And I'd recommend it highly because it contains references to all significant [historical] "construction" efforts (which I agree are generally misguided, for text at least). And I wish Carter would write part 2 already!

> not bad for "ancient" Rome, huh!

Except they got that -and most everything else of value- from the Greeks.

hhp

bieler's picture

"Except they got that -and most everything else of value- from the Greeks."

And the Greeks got it from?

hrant's picture

Good question!

hhp

hrant's picture

Back to the original issue:

> There seems to be an (inate?) optimal width for each glyph.

It seems that the optimal width for a given character is a value that can be deviated from safely, upto a point. This point depends on the sensibilities of the viewer (which is subjective). Furthemore, the deviation can be controlled to make the font more dynamic. However, the set of deviations might have to be consistent, otherwise the harmony is too deep, and it will be seen as discord by practically everybody. What I mean is you can't make half the UC Classical and the other Modern.

As for what the "ideal" is, I would say it depends on the structure of the glyph. One obvious example is the "B" versus the "D": the former is narrower because you can't make half-size bowls look nice that wide. But it's not so simple because a "B" with bowls of [roughly] equal height can't be as wide as one with a large lower bowl.

So there might be a way to quantify the optimal widths of characters by first quantifying their structure, but taste is subjective, so you could only do that for a given person (and really only at a given instant). So what you could do is make a set of caps where the widths please you personally, then make variations of certain characters (like make a "B" with a lower/higher middle) and see how that affects your ideal width, and formulate some numbers for future use. So the next time you make a font, you can draw a single character and "propagate" the ideal widths of all the others, based on their other structural determinants. Once you have the set of ideal widths, you would work inwards.

To avoid making the solution too personal, you could do this to a "classic" face. For example, you could take the Trajan "B", lower the middle, see how its ideal width changes, and use that dataset as the guide for a new font.

--

This is exactly the sort of thread I think type design needs most now. So I hope many people brainstorm about it and delve deeply towards its warm core.

hhp

capthaddock's picture

How about when language-specific subjectivity comes into it? Would Germans prefer it if the CH and CK digraphs used narrower, tighter glyphs (like the old blackletter ligatures) than lone C's, H's and K's? How about the Dutch with IJ, or the Spanish with CH, RR and LL?

Paul

.00's picture

Duke Ellington's comment about music "if it sounds good, it is good" is how I feel about character proportions. You can intellectualize it all you want, but if they look good, they are good.

hrant's picture

To an individual, sure. But when it comes to pleasing others (which I think is the heart of type design), you have to think about it some more. For example, some people like the Classical scheme while others prefer the "Modern". So you have to use such context to improve your design.

hhp

.00's picture

I think there are enough typefaces available in the "Classical" and "Modern" theme.

Time for something else.

hrant's picture

> Time for something else.

I'm all for that!

But just to make sure: we are talking about the widths
that glyphs conform to, right? Not letterform style.

What do people think of Richler's caps,
especially in terms of the widths scheme?

--

Something new, eh? Hmmm... What about this:

Since letters are put together to form words, and we can appreciate the way a word sets as much as the way a letter sits, the set of optimal widths in a writing system depends on linguistics. If the Classical widths are seen as optimal for ancient Roman inscriptions (in Latin), they simply can't be seen as optimal for a contemporary novel. So maybe we need a new set of widths?

That said:
1) There are -as I'm fond of stating- two strands of reading: deliberative (where you can consciously appreciate shapes) versus immersive (where your reading "firmware" takes over, and doesn't care how sexy that finial is). As a result, "optimal widths" can be [at least] two things...
2) I've only thought about caps so far...

hhp

hrant's picture

I found something. An article in the 1968 Penrose Annual (which I actually mentioned in the "Trapping" thread) by Louis Rosenblum, where he's devised a system to standardize the measure of the relative widths of text fonts. The apparent aim being to be able to predict layout and/or save on storage (remember: 1968).

Rosenblum's style is very terse (he's an engineer by training), so I'm not exactly sure, but it looks like he's taken measurements of many "classic" fonts and arrived at average relative widths:

Lowercase:
5: i
6: f j l
7: r s t
8: z e c
9: q g x a o
10: n h u d p b k y v
13: w
15: m

Uppercase:
8: I
9: J
10: S
11: Z
12: P L F
13: Q O A V B T E C
14: G U Y R
15: N H X D K
18: M W

Numerals: 9 units.

One strange thing is that for a given width the letters listed aren't alphabetical. But there must be a reason behind their orders. My guess is that they're listed in increasing statistical divergence from the given width.

Rosenblum also ties the TTS width system into the discussion, but I couldn't follow that well enough.

One last thing
In pointing out that metal fonts have smaller sizes that are proportionally wider and darker than the larger ones (optical scaling), he says that's due to practical manufacturing concerns! I'd never heard that before.

hhp

aquatoad's picture

Hi All.

I posted this resource is a crit a while back. It is the proportions for modern capitals as laid out by Morris Fuller Benton and copied/refined by Ed Benguiat (as it

aquatoad's picture

Double Post disease... attacks... out... of... nowhe...

kentlew's picture

>Rosenblum's style is very terse (he's an engineer by training), so I'm
>not exactly sure, but it looks like he's taken measurements of many
>"classic" fonts and arrived at average relative widths:


Hrant, this looks suspiciously like a Monotype matrix layout.

The other possibility is that it is the actual TTS width specification.

-- K.

hrant's picture

> this looks suspiciously like a Monotype matrix layout.

But he specifically says it's different than any existing system. He did however use a lot of Monotype fonts in his calculations. One thing he points out is that a system limited to 15 widths (as opposed to Monotype's 18) fits nicely into 4 bits of data.

TTS: It seems like he's treating TTS as just another "font", in terms of calculating the relationship to his work.

hhp

Syndicate content Syndicate content