Kerning Pairs; the more the merrier?

Queneau's picture

Hi all,

I've been wondering about kerning pairs in fonts. There seems to be a lot of variation in the amount of kerning pairs per font. A font like Frutiger LT Std 55 Roman (251 characters) by Linotype has 441 kerning pairs. URW's version of Frutiger, Frutus (Regular 243 characters), has 979 kerning pairs, and Cartogothic Std (349 characters), A free Frutiger version (ripoff?) by FontSite has 4726 kerning pairs. What does this tell me about the quality of these fonts, if anything?

Of course it all comes down to the quality of the kerning itself, but still, would one spot the difference in a normal text setting to have 4000 instead of 400 kerning pairs?

Some fonts like Jos Buivenga's Calluna have more than 50000 kerning pairs, but does this make them any better than fonts that only have a couple hundred pairs? What would you consider to be the minimal amount of kerning pairs for proper text setting?

blank's picture

What does this tell me about the quality of these fonts, if anything?

Nothing. Some designs just require a lot less kerning than others. A narrow sans design requires very few. A serif with wide capitals that are kerned to the lowercase requires a lot more. Small caps add even more. And how kerning classes are configured can have a pretty dramatic impact on the number of kerns in a font. A poorly spaced font will have many more kerning pairs than one that was spaced well to begin with. Accurate revivals of metal type require lots of kerning pairs to work with the tighter spacing and kerning used today.

There are some type designers out there who, in need of something good to say about their font, can’t think of anything better to do than tout a large number of kerning pairs. I can’t say that I’ve ever been impressed by those statements.

Nick Shinn's picture

Accurate revivals of metal type require lots of kerning pairs to work with the ... kerning used today.

You need lots of kerning because you need lots of kerning?

Wouldn't an accurate revival of a metal type have no kerning? (Maybe a few two-character "logotypes", if Linotype.)

Alex Kaczun's picture

I feel that "kerning" is somewhat overdone in many fonts.

Yes, it depends on the overall design, but if the type designer has set up the progression of weights properly based on the key control characters (H, O, n, o, 0), then there should be a standard functional set of kerning (approx. 400-600 pairs).

The "functional set" consists of several cap to lower case combinations like cap "T" to several lower case like "a", "o", etc., which obviously need some help with general spacing issues.

However, that being said, it also depends on range of glyphs incorporated into a font.

If the designer includes all the "Eastern European" accented glyphs for example, the "kerning" sets can grow exponentially.

There is also a "size" limit to how many "kerning pairs" a font can have before it breaks. Somewhere around "10,000 pairs".

So, great care should be taken to minimize the "functional set" of kerning pairs in any font.

However, just because a font has a lot of kerning pairs does not necessarily make it a bad font.

If you like the font you are using, why worry about how many 'kerns' are in a font anyway?

And, in conclusion, this whole process of kerning is purely subjective, anyway.

No matter what a type designer incorporates into the fonts, someone else will probably have a different opinion anyway.

I hope this helps and sheds some light on the subject for you.

Grrrben's picture

The better the fitting is, the less kerning it requires.

charles ellertson's picture

The better the fitting is, the less kerning it requires.

This seems to be holy writ in the type design world.

It is true that with the Linotype linecaster anyway, there was no kerning. Monotype could cast a kern, so "what's appropriate" in a revival is a little cloudier than one might think.

If you ask "how much kerning" to those who use type, you're apt to get a different answer. Even if you ignore the pride and craft of the compositor, a few "P.E."s from editors will make you sit up and take notice.

When we get in a new font for setting text, I sit down and look at every letter pair, lowercase to lowercase, and cap to lowercase, and numbers. Included in this is the fit with the most frequent punctuation -- period, comma, hyphen, quotes, slash, and usually colon and semicolon. A lot of kerning comes about with the fit of punctuation.

Now is this the fault of the type designer? I'd say no. The designer cannot tell what size the type will be used at. If you are trying for the best presentation of a type, the value of the sidebearings, as well as the kerning, depends on the setting size.

Another difference from metal, where every size could have its own design.

So from my perspective, the "holy writ" is trivially correct, because it is only the end user who can make the final decisions.

As a side note: I began setting type when "tracking" was available in increments of 1/18th of an em -- 18 units in the 1,000-unit em of today. Never used tracking much, it was just too big. All of a sudden I discovered (late, because of habits formed) that with InDesign anyway, tracking is incremented in 1/1000 of an em. That's useful. Now by in large, if you minus track, say one unit, to set 12-point type that you kerned for, say, setting at 10 points, most of the kerning holds up quite well. The same if you plus track the same type at an 8 point setting. The key word here is "most."

Do I rekern for these? No, it's just an observation. Not quite that compulsive -- or I'm too lazy, take your pick.

One final observation: I find often enough that with letters having similar shapes and similar sidebearings, I'll still use slightly different kerning values. For example, even though c, d, e, o, and q share a common left sidebearing and appearance, when I kern them, I'll wind up with slightly different values. So I don't much care for class kerning. If you want to call that whim, fine, I disagree. I also don't think it is the type designers responsibility.

* * *

Every once in a while, a book (interior) designer winds up choosing a display font for setting the text. Sometimes it is a good choice for the book, but it is a lot of work. I'm currently working on such a font, a sans-light, where I'm sure the font designer assumed a 30-point or larger setting. It looks pretty good at that size, but looks just dreadful with a 10-point setting. The book designer noticed that, and tracked the font 20 units. Guess what, the ligatures broke, as they should. I've gone into the font and added 9 units to the left and right sidebearings, which solves the tracking/ligature issue. But the fit of the letters changes with such a drastic size change, and once again, I can see first hand that spacing values that work quite well at 30 points are too tight when set at 10 points, even with the extra space from increasing the sidebearings.

None of this work is brought about by a poor job by the font designer -- but I did have to point that out to the book designer.

Queneau's picture

Thanks for all your comments, it's very interesting to hear what the pro's think about this. I guess there are many different opinions on this topic...

On a related note; I recently rediscovered Kernus on my URW Typeworks collection. Unfortunately I did not as yet get it to function (it runs on mac OS classic, but somehow it cannot find the right data to be able to calculate the new kerning tables, I've read the manual, but it did not bring me much further; any help here would be appreciated) Has anyone here have any experience with this program, is it still relevant to use it with today's layout software (InDesign in my case)? Would be great to have some feedback on this.

William Berkson's picture

IIRC Kernus was the basis for what became "Optical Kerning" of Adobe, which is very problematic.

Charles, can you name a few fonts that are in your view well kerned, and would serve as models of excellent kerning for font designers?

Nick Shinn's picture

There are different styles of letterspacing/kerning that may be applied to the same typeface -- either at the same size (for different circumstances or by different typographers), or for different sizes.

I added positive kerning between "rounds" in Figgins Sans, as the default. This is traditionally a display look, "tight but not touching", but I figured it would be practical at text size also.

Here you can see the difference between the three modes of kerning available in InDesign.
Top, with no kerning.
Centre, with foundry kerning ("Metrics").
Bottom, with the godawful "Optical".

blank's picture

I also don't think it is the type designers responsibility.

That’s exactly why I like class kerning. Finely tuned kerning is dependent on both size and context, so I use classes to get something acceptable and leave it to the end user to polish it up as appropriate.

charles ellertson's picture

Charles, can you name a few fonts that are in your view well kerned, and would serve as models of excellent kerning for font designers?

I suppose it should be no surprise -- Matthew Carter's fonts always seem well kerned to me -- that is, the values chosen are very good. I add kerning pairs, but rarely change what's already there. The most recent of his that I remember working on was Miller, which I got from Carter & Cone. I don't know whether Font Bureau has changed things.

There are probably others. I remember Miller because of a conversation. As I remember, Kent Lew's Whitman was also quite well kerned (in my opinion).

That’s exactly why I like class kerning. Finely tuned kerning is dependent on both size and context, so I use classes to get something acceptable and leave it to the end user to polish it up as appropriate.

Makes perfect sense to me.

Just remember that within the EULA, you allow this. It has always seemed a terrible notion to me to prohibit things that can be done, albeit with more effort, with the layout program. Of course, those "modifications" by the layout program can't be prohibited.

What is key for the designer/owner is that all copies of the font are paid for. You can let somebody modify it, but if they have a license for only one copy of the font, that just became their one copy. If they want to keep a copy with the original values, say for setting display, they should pay for a license that lets them have two copies.

I suppose it's your business if you feel that your font is perfect as is, and want to prohibit all modifications. From my perspective, even if you're right, it just became like a metal font -- a one-size design. Well, sort of. Actually, metal fonts got modified.

dezcom's picture

As is usual for most things, numbers don't tell much of the story. The issue is really the quality of the combination of fitting and kerning. The right "number" is whatever it takes to make that particular font work best.

Queneau's picture

Yes, Chris, of course you are right here. The problem is that I (A graphic designer) unfortunately was never given any proper education in stuff like kerning and spacing. I guess my education was more focused on expression and concept. I thought myself quite a lot afterwards, but still I sometimes get the feeling I need to know more to be able to judge the spacing, kerning, fitting etc. of a font. As I buy my fonts almost exclusivily via online retailers like MyFonts, or via a foundry itself, one mostly has to be able to judge the typeface on screen. Printed specimen help, but one needs a very good printer to get the proper quality required to judge the fonts. So any objective values or criteria are always welcome :-)

charles ellertson's picture

unfortunately was never given any proper education in stuff like kerning and spacing.

That's a curious statement. If the only purpose of the type is to be read, the smooth, uninterrupted flow of the letters is the proper "kerning and spacing." Yes, letterfit will interact with word spacing, leading, size, and likely line measure. But so what? Do we need training in what looks good when we read? Elite Reading 101?

With that in mind, you can argue that straight text -- as in book interiors -- isn't a matter for graphic design. You'd have the likes of Jan Tschichold on your side of that argument.

As soon as graphic design becomes an important element in the work, "proper spacing and fitting" belongs to how that design works.

Queneau's picture

Of course I can rely on instinct, and if have learnt a lot about using type. But there is a difference between something generally looking good, and the finetuning that comes with creating really good micro- and macrotypography. Well-functioning, well kerned typefaces help, but the job at hand might still require manual tweeking to make it work. My eye can still be improved a lot in this respect, and this is why I value the opinion of type professionals in this matter. Not to slavishly follow their views but to develop my own.

William Berkson's picture

Charles, I don't think that good kerning is obvious on the "micro" level, but does take training to recognize. It may well be that its end result of "smooth uninterrupted reading" is evident to a sensitive untrained reader, once it is printed out in a book.

But the art of kerning one letter to another so that you get the desired end result is not at all obvious. For starters, you have to have more than two letters present to do it, otherwise you can under- or over-kern easily. Even then, I think it takes time and training until one's eye-brain combination is "tuned" to see what will have an end result in text.

Once you have acquired the ability to "see" it, it may be generally crashingly obvious, but it is easy to forget that acquiring that ability to see takes time. Since I started doing this basically in old age, I can testify that in my case kerning decisions went from 3 minutes to 3 seconds, but that took many, many hours, working with similar words in Whitman, Adobe Garamond and Minion as comparators. As with everything involving type, experience makes a huge difference to one's perception, one's eye.

Syndicate content Syndicate content