*How do you measure you kerning?

popovich's picture

Okay, newbie or not but after reading several books on typography I can't remember anyone mentioning anything about how to measure kerning. Most of the time would be told "Kerning is important!", "Double-check your kerning!", "blah!", "blah blah!", BUT .... how? Say, If I have a measure X which equals the letter-spacing here "NM", then how do I measure the kerning in the following pairs "NO", "OS", "WS", "PG", etc.. Do you have your techniques? Tell me, please.

mattuk's picture

I tend to only kern large characters in headings, and i only really ever do it visually. Its usually glaringly obvious which characters look ugly together, and realistically not always possible to spend the time going through every font combination in kerning table edit in Quark.

popovich's picture

visually? with eyes? it's seems to me to be subjective - you'll say the kerning is okay, and I would say "hmm... could be better, just move it to the right a bit, just a little tiny bit", you know? I want to believe there are rules for kerning, based on mathematics, geometry or whatever and not only on subjective perception. And I would like to learn those rules...

hrant's picture

There is no math, at least not math a human would be willing/able to do.

So it's subjective, especially for all-caps setting. But for running lc text, assuming your objective is even color, most people's ideas of good spacing converge greatly.

hhp

mattuk's picture

I like to think that design and typography is equally about aesthetics and creative flair as well as hard and fast rules. While i take your point that you can apply sets of rules to certain elements of typography including kerning, I think to a point it takes the fun out if it bowing to mere mathematical calculation.

The kerning of specific characters changes from font to font, but yes there are similarities between pairs of characters. I couldn't tell you whether the kerning between 'ey' is based on any mathematical rule, only that the shape of the characters means that they need to be closer together to appear visually equally spaced.

Someone else might though...

popovich's picture

It's okay to have fun while working - not only in typography. I agree absolutely. But there is so much fun already working with it, that I thought, there are also some definite set of rules at least for some parts of the whole process. Like, Matthew, you say that kerning differs from font to font. True but I wasn't speaking about absolute values. I meant relative rules. Like, NM is 1 measure (1M). OK is 0.8M, WW is 0.5M and so on (hypothetically speaking).
I am surprised there are no rules regarding kerning.
How are kerning tabes made/calculated? I mean the ones supplied for every font. What are those principles?

eomine's picture

>I meant relative rules. [...]

I think this is not possible, because each typeface has particular shapes and metrics. The lack of a priori "rules" doesn't mean anything, you just have to develop your own rules, for each typeface you design.

It's not about fun, it's more about "sensitivity", applied in a objetive way (if this make any sense ;-).

aquatoad's picture

Hi Alexey,

I think you might find the book, Type By Design to be useful to get basic spacing down. Finding a hard copy of Type By Design can be a problem. Somewhere in these forums somebody posted a link to an online version of it. Chapter three is the one you want, I think. I can't find the link for the life of me. Help from the hive mind?

Randy

eomine's picture

Randy, there's a link to Type by Design in this thread.
The spacing chapter (Walter Tracy's method) is great, but I think Alexey is talking about kerning, not spacing.

aquatoad's picture

Ahaaa! So he was.

However, Type by Design still gives a good overview of kerning. I think Alexey, you are looking for the same kind of fomula Walter Tracy gives as a start on spacing, only for kerning.

The book does give a list of important kerning pairs to consider (for english). The reason it'd be tough to make a formula is because the nature of the beast is that each one is an exception to the rule. I dunno, maybe someone's got a masterplan?

I also find it interesting the avg. number of kerning pairs are given for two foundries (at the time the book was originally published) Adobe: 100-200 and Bitstream 200-500. Those may have changed. The key for text though, is the spacing, as kerning has less influence at smaller point sizes.

I'm rambling, sorry.
Randy

graficartist's picture

--self edit--

Diner's picture

Regarding your original question, there are many ways to kern and as mentioned above it's all subjective depending on the type of type you're making . . .

A text face has more predictable forms due to its nature and can actually look good without much additional kerning. In contrast, a display face that contains no repeating character forms in some cases can be riddled with kern pairs.

The key to determine how many kern pairs all comes down to the quality of the spacing job on the font. It is likely a perfectly spaced font could exist without any kern pairs.

Regarding rules for kerning, I have a personal theory. . . If you were to determine the absolute horizontal center of a character and draw a line at that mark then set a long string of text, I would kern anything that exceeded a certain width or percentage minding the width of the glyph.

I think there is a rhythm that can be easily identified in relation to the rest of the glyphs in the typeface and to some degree could become a rule that requires kerning be applied.

Beyond that, another thing to consider is that even if you decide to make as few kerning pairs as possible, if you make let's say a 'vo' kern, expect that will also be required for a 'v

popovich's picture

wow, thanks everybody for taking time and participating.
Randy, I've downloaded the book you mentioned and will surely read the recommended chapters first (and then the whole book ^_^ ).
Stuart, I didn't really get your concept of horizontical center, a line and kerning. Also, the concept of "rhythm" is hard to understand on the fly. I've seen it mentioned in "The Elements of Typographic Style" and other books on the subject, but it's hard to understand it like that. ;/

Also, what about those programs, like HZ-Program now built in InDesign for optical kerning? It calculates something with its formulas. Of course, it would be naive enough to think everybody knows those formulas, but does anyone happen to know the principle? What IS optical kerning? How can a programm be subjective?! O_o

And the word "arbitrary" doesn't make my day shine. See, such an untrained eye as mine will need million years before it can kern right. -_;

aquatoad's picture

The key to determine how many kern pairs all comes down to the quality of the spacing job on the font. It is likely a perfectly spaced font could exist without any kern pairs.

Great point. That's why I immediatly jumped in and started talking spacing. For example, Alexey, if you look at the list of kerning pairs on pages 89 and 90 of the book you downloaded, the essentials are listed at the bottom of p.89. You probably will have to kern (Av, Ay, Aw, AW etc.) But good spacing might just take care of many on page 90 (for example OB, OD, OE, OF, OH, OI, OK, OL, OM, ON, OP, OR etc, a round next to a vertical, they're all going to be really close to the same) One suggestion might be to group the remaining ones into similar groups for kerning (Av, Ay, Aw will be relatively close)

Regarding rules for kerning, I have a personal theory. . . If you were to determine the absolute horizontal center of a character and draw a line at that mark then set a long string of text, I would kern anything that exceeded a certain width or percentage minding the width of the glyph.

I also have personal theory, that guides my spacing/kerning :-) For the latin alphabet my experience is that the top half of the lower case has a larger impact on legibility and spacing. For example if you were to type the word l-i-t-t-l-e or f-i-t-t-i-n-g and cover up the top half a) you'd be hard pressed to read it, and b) the spacing would look really bad! But looking at the letters as a whole, the job might be spaced just right. I guess all that to say, top half of most letters will have the most effect on the spacing/kerning.

One last thought, it's impossible to incorrectly space two letters. When you add the third, that's when it gets tricky. In your metrics window go down the line in groups of three, then check the line as a whole for overall rythm (for spacing and kerning).

I am no expert! Expert methods wanted!

Randy

hrant's picture

> the concept of "rhythm" is hard to understand on the fly.

Or even sitting down. Because it doesn't really exist.

hhp

aquatoad's picture

rhythm (noun; 'ri-[th]m): movement or fluctuation
marked by the regular recurrence or natural flow of
related elements
b : the repetition in a literary work of
phrase, incident, character type, or symbol

It doesn't exist? By saying "rhythm" aren't we talking about
a regularity to the spacing, a cadence that is as even as
possible throught any combination of characters. There will
always be breaks in the rythm because of the shapes of
latin letters. L + anyletter will cause a hiccup in the rhythm.

The bait was tossed and my lip is expecting a sharp pain
any moment now.

Randy

hrant's picture

To be fair, there certainly is rhythm during deliberative reading (like looking at a poster). In fact there are probably many layers of rhythm there.

But in text, there are only two parameters of rhythm:
1) The relationship of saccades to line lengths.
2) The textual content.
But #1 is a typesetting thing, and #2 is an authoring thing - both outside the scope of type design.

So there can be no real rhythm at the letterform level, not even the bouma level. It sure sounds great though - like a kirin in the sky.

hhp

aquatoad's picture

In the end, a good definition of rhythm is not the point (I think ours differ), nor is the existence of rhythm or not. Trying to come up with a good plan to approach spacing and kerning is. I'd be most interested to hear (in detail if you're willing -- or anyone else for that matter) your course of action when setting about spacing and kerning a font.

I think Stuart was merely articulating some of what i'm (and Alexey's) hoping to hear more of. If I read his part about rhythm with relation to the other glyphs in an alphabet, that's just to say many letters in the roman alphabet have similar shapes, and thus could be spaced/kerned similarly. Forgive me Stuart if I'm off the deep end!

Regards
Randy

BTW Hrant, congratulations on post number 2700.

hrant's picture

> Trying to come up with a good plan to approach spacing and kerning is.

Totally.

Good spacing is good notan (the relationship between white and black). Good notan is where the letters meld into good boumas. When the spacing is too tight/loose, the black gets lost in the discord between intra- and inter-letter white.

> post number 2700

Oh. Time flies. Or something. But you guys were supposed to remind me at 2048 (remember, I'm a binaryboy). So now you have to wait for 4096.

hhp

kentlew's picture

>Good spacing is good notan

Hrant --

Are you making up new words again? This is the second or third time I've seen you use this "notan" word. What the hell does it mean and where does it come from?

-- Kent.

hrant's picture

> This is the second or third time

That's it?! I'll have to ramp it up... ;-)

Notan is an old and highly useful, deep term. So it's not-an invention of mine. :-)

Notan is Japanese, and it means the counterbalancing and relationship of Black and White. Like Yin-Yang.

Without getting too religious, I'd say that Notan is the entire basis of type design, but it has yet to gain proper respect in the West. You hear people say things like "the black has to be balanced by the white", but they don't go any deeper than that, towards the implications, like how lettering and type differ, and how chirography cannot generate optimal notan (because the two edges of the black over-ride the truly necessary relationship, the one between the black and the white). And the fact that the term has to be imported from the East is highly telling.

Notan is the reason spacing is just as important as the letterforms, the reason a font is more than a set of pretty shapes.

--

Recommended reading:
"Notan: The Dark-Light Principle of Design"
by Dorr Bothwell, Marlys Mayfielf
Only US$8, and it includes DIY exercises.

hhp

kentlew's picture

Hmm, borrowed from the Japanese this time. All right, but just this once. ;-) Seriously, though, you might consider putting it in italics -- you know, foreign term and all that.

-- K.

popovich's picture

Thanks for the link Gerald, I do find this intersting.
Hrant, what is a "bouma" then?

hrant's picture

(I can hear the groans...)

http://www.themicrofoundry.com/ss_read1.html
Note: this is highly simplified. In reality, all kinds of strange things happen both at the perceptive level (especially in the parafovea) and most certainly in the higher cognitive areas.

hhp

tidchris's picture

YES! keep it going.

Hrant's deconstructing the 'virtual machine' that processes design - the human perceptual system. Some of it is objective, contextless, pure - some of it is about detecting faces and the difference between a copse of trees and a tiger.

I think the notion of 'notan' invites the designer to start on grey rather than black or white (chiaroscuropsis?) cause in the real world 'figure' is positive and negative variations against ground. In a decade we'll know better how to maximize the channel between visual presentation and the mind. Feedback will play a big part. Until then we'll have to kern by eye, and gut.

-chris

hrant's picture

> chiaroscuropsis?

Hey, that sounds really cool.
Some famous guy once said: "Before I invent anything, I need a nice name for it."

> In a decade we'll know better how to maximize
> the channel between visual presentation and the mind.

Maybe. These days academic funding comes from megacorporations, not philanthropists, so unless you can explain the benefits of unraveling reading to stockholders (somewhat harder than explaining to them the benefits of cupholders...) don't hold your breath. :-/

hhp

tidchris's picture

> don't hold your breath. :-/

Nah. These days I breathe whenever possible.

I am tired of defending every act of art and human mercy by how many dollars it would add to the pockets of the powerful. Megacorporations are not humans, they are not my users and they do not care if I die. It's mutual. Don't trust anyone over 30 ($Billion US).

"Kerning pairs must be as cup and cupholder" - throwing a bone back on topic

kris's picture

>I also find it interesting the avg. number of kerning
>pairs are given for two foundries (at the time the book
> was originally published) Adobe: 100-200 and
>Bitstream 200-500.

Righto. Why do some foundries boast that they have
thousands of kerning pairs, yet Bitstream et al have
hundreds? I read somewhere that more = worse
because that would mean more data. Is this necessarily
true? By spacing, do you people mean metrics?

Thomas Phinney's picture

I just noticed this thread with Kris's quote.

Sure, more kerning = a bigger font and more data. But frankly, that's not much of an issue on modern computers and with most current-edition software.

Lots of kerning may result from a designer trying to compensate for poor initial spacing. But even given ideal spacing, a well-made single-byte font may be able to make good use of several thousand kern pairs.

Although Adobe's earliest fonts in the mid-80s typically had 100-200 kern pairs, by the time Adobe stopped making Type 1 fonts at the end of the 90s, 500-3000 was a fairly normal range of kerning pairs for an Adobe Original design.

I'll note that the 100-200 count typically involved not kerning accented characters, so it was basically inadequate for most Latin languages beyond English.

Today, the whole point has become moot for those of us doing OpenType fonts. The character sets may be larger (sometimes vastly larger), which means more kerning is needed. Further, the opportunity to use class-based kerning means that "pairs" and the linear relationship between total kerning and data size become irrelevant.

If you took the kerning classes in Minion Pro Regular and expanded them into pairs, you'd end up with >70,000 pairs (we did a test at one point in development; the final total would doubtless be larger). But the font also has something like 1600 glyphs.

We don't generally "boast" about the quantity of kerning in Adobe fonts. It's not something we advertise on our Web site. But I will say that we spend a great deal of time and effort trying to get spacing and kerning right.

Cheers,

T

kris's picture

Thomas, I realised after I posted that contemporary
technology pretty much renders file size irrelevant. But
does "compensat[ing] for poor initial spacing" suggest
that the intial spacing effort would be poorly 'crafted'?
If technology allows for many kerning pairs, does this
render metrics largely irrelevant? I know it is a poor
comparison, but I realised that metal type has fixed
spacing, which I presume means no kerning pairs,
except ligatures. Was metal type spaced well, or was
a little bad spacing here and there accepted because
of the limitations?

Thomas Phinney's picture

Kerning, even large amounts, is no substitute for good spacing in the first place. It rather complements good spacing.

Metal typefaces were often well spaced, though sometimes poorly. They were probably better spaced on average than digital typefaces, because of the amount of work involved and metal type creation being a more specialized craft. In particular, the role of the punch-cutter in spacing was crucial.

T

kris's picture

>Kerning, even large amounts, is no substitute for good
>spacing in the first place. It rather complements good
>spacing.

aha! I suspected as much. So it goes back to Hrant's
notion of notan

kentlew's picture

The raw number of kern pairs in a digital font cannot be used as a measure of the quality of the spacing. You must look at the nature of the pairs included.

A font may have thousands of kern pairs because the base spacing is poor and has been compensated for. For example, does the font have more than about 150-200 lc-lc pairs?

Or it may have thousands of pairs because more combinations have been considered and included. Does the font include kerns for accented characters (as Thomas mentioned above) or infrequent combinations?

A font may have fewer than 500 pairs because the design and the base spacing don't require more. Or it may have fewer because the designer just didn't bother to refine the spacing.

It's often a matter of how thorough a designer chooses to be, balanced against a number of considerations. Code size used to be more of an issue in the early days; less so now. A commercial venture needs to decide how much additional time to spend reviewing pairs of lesser statistical value. A line needs to be drawn somewhere. Different foundries or designers will draw it at different places.

For example, my Whitman fonts contain on average about 1700 kern pairs. Some may consider that high. But about 20-25% of those are for accented characters -- both UC-lc accents (V

John Hudson's picture

Well said, Kent.

Thomas Phinney's picture

Yes, that was an excellent summary.

T

hrant's picture

Kent, great post.

What I would say myself is that:
- A font with "too many" pairs is usually better. Somebody who makes his base spacing bad to the point of needing many corrective pairs is unlikely to spend the time putting those pairs in! The habit of looking down upon large numbers of pairs (which de Groot has interestingly called an "Americanism") might be nothing more than an excuse not to provide good kerning. Gobs of auto-kerning notwithstanding.
- A font with too-few pairs is always bad! Unless it's to make an ideological point by having no pairs at all, like some of Dan Carr's work.

BTW, don't forget Ypres, an important name in military history... as well as in Monty Python.

hhp

kris's picture

Kent - thank you, excellent post. That is exactly the
answer I was looking for with my rather wobbly
questions!

kentlew's picture

As we've said, the raw number of kern pairs is no measure of the quality of a font's spacing. Ultimately, the best evidence is to set text and look at it. A well-spaced font will be, well . . . well spaced.

Generally speaking, though, a truly refined font will have a greater number of pairs. That said, however, a well-crafted font should still set adequately with kerning turned off. Kerning is a refinement of spacing, not a substitute.

I'm no expert in the subject of spacing, but I have given considerable thought and study to the matter. Further to the turn this thread has taken, I thought I'd share the following analysis.

Let me contrast two excellent text typefaces with two that often get used for the task but are, in my opinion, not built for the job.

Matthew Carter's Miller is one of the most excellent text faces I've had the pleasure to work with over the years. His Fenway is also a contemporary masterpiece for text.

In comparison, Mrs Eaves and Filosofia are two typefaces which recently have begun to show up more and more for text in books (heaven forbid). While I admire Zuzana's designs and like both of these in display, I think it is no secret that they are not up to the task of extended text, spacing-wise.

I spent a little time a while back sorting some AFM files to gain a little more insight into the inner workings. I broke them down into the following categories:

Uppercase-Uppercase
Uppercase-lowercase-- including the odd lc-UC, like cT or cV
Uppercase-punctuation -- in either order, e.g., "A and A"
lowercase-lowercase
lowercase-punctuation -- in either order
punctuation-punctuation -- e.g., various points with various quotes, etc.
numerals -- involving figures as either element, in combination with anything: other numerals, punctuation, lowercase (e.g., 1t, 7t, etc.)
accents -- involving any accented character (cap or lc) as one of the elements, including Oslash, Ccedilla, dotlessi, etc. (which aren't accented, strictly speaking) but not the dipthongs

Here is a breakdown of the number of kern pairs for each.


MillerFenwayMrs EavesFilosofia
UC-UC122679385
UC-lc2002775878
UC-punct2853191026
lc-lc1569775
lc-punct24528920
numerals6412900
accents3313700
Total11281359170194



It is instructive to look at the particular pairs in all these instances. I can't go into them here, but I recommend all those who wish to learn more about kerning to look closely into the AFM files of several fonts.

-- K.

hrant's picture

Another killer Kent post!

> I think it is no secret that they are not up to the task of extended text, spacing-wise.

It's not a secret as much as it's just ignored...
The good news? InDesign's "optical spacing", which turns those two "loose bycicles"* into usable text faces.

* Robin Kinross's qualification of Mrs Eaves.

hhp

Thomas Phinney's picture

This is really interesting. Thanks for taking the time to put that together. Yet, given the spacing problem with Mrs Eaves--a design I quite like otherwise--I'm not so concerned with its kerning.

T

hrant's picture

I just concocted a bitchen kerning test: "T

John Hudson's picture

good sweet wine

Oxymoron.

hrant's picture

:-)
I prefer wine ultra-dry myself (same with beer - Asahi rules).
But there are always exceptions - one day I'll make you my trademark Sangria...

hhp

hrant's picture

And you guys were silently thinking you'd never actually need to kern "Yp":
http://news.bbc.co.uk/2/hi/europe/3256553.stm

hhp

eomine's picture

Just FYI, the "Basic Kerning Pairs" file (available
right here at Typophile) already contains the "Yp"
pair. No "Yq", though.

fonthausen's picture

Another good example is the Trinit

jfp's picture

And why the MS core set seems to doesn't have any kerning? Does this will change in the future? (I think about the future OpenType support of Longhorn)

hrant's picture

Probably to better control the matching between screen and print, especially in terms of linebreaks.

hhp

John Hudson's picture

And why the MS core set seems to doesn't have any kerning?

Georgia shipped without kerning -- an error, apparently --, but as far as I know most of the other MS fonts have 'kern' tables and some of the more recent ones have OT <kern> feature. Which core fonts are you looking at, JFP?

hrant's picture

> Trinit

Syndicate content Syndicate content