Enforcer - Kerning tests

tourdeforce's picture

You know the moment when you're becoming pretty unsure if you did a good kerning and no matter how much times you look into the font, you always have some doubts whether if it's good or bad, near to finish - far away from being useful...

That's the moment that holds me last a couple of days with this new font.

I need some fresh (and refreshed) eyes to take look at this file and tell me if you noticed some irregularities.

I attached some PNGs that can be helpful:


- Dusan

Bendy's picture

I'm completely no expert, especially about spacing and kerning, but it looks to me like you might be using kerning as a remedy for spacing that's not quite there yet. Pairs like r_l, i_s and t_h shouldn't normally require kerning.

tourdeforce's picture

Thanks Bendy.

Things moved itself.

Frode Bo Helland's picture

Don’t even think about kerning until the spacing is absolutely perfect.

Arno Enslin's picture

Spacing and kerning are a disaster. I wonder, why you have doubts about the quality. It’s extremely uneven. Why is there kerning between i and j for example? Why is there positive kerning between w and o and negative between r and m? Another example: It’s and like. There is more space between the stems of I and t than between l and i. I and t is less critical in case of your font than t and h. While t and h may need more spacing, although the distance between the stems of t and h are the same as in i and h, the cross bar of the t is no reason for enlarging the distance between I and t.

By the way, it is better to kern in big point sizes with a big distance to the screen because of rounding imprecisions in FontLab. And kern words, but not whole sentences. You can use a free dictionary like the one, that OpenOffice uses, in FontLab. Try to create a list, in which you have an order of kerning/spacing, alphabetical like aa, ab, ac … or with regard to the forms ii/ij/hn, oo/oe/od, wo/we. (I prefer the alphabetical order.)

1. There are too many kerning pairs.
2. There are too many positive kerning pairs. In most well spaced fonts, that I have seen, positive kerning was the exception, but not the rule.

Check the following articles on "http://www.typeculture.com/academic_resource/articles_essays/": Basic Character Spacing in Type Design and Rhythm in Type Design

Don’t even think about kerning until the spacing is absolutely perfect.

Absolutely right. And not only with regard to the evenness, but also with regard to the general spacing.

tourdeforce's picture

This is how it looks now (in Regular weight only, there will be 3 more).

As I said in my previous message, I already saw some problematic situations and solve them (probably not all, but by the time I finish all weights, I'm hoping to see more and more irregularities).

@ Arno:

I can't agree with you completely, especially with 1) and 2).

There are total 1037 kerning pairs, partly cause there are 352 glyphs total in this font which is pretty OK number if you have in mind that some of well-know commercial fonts have more then 3000 pairs. I'm not saying that this font can't have less pairs, but that's what I was thinking in the first sentence of this post - as times goes by, everything will be more visible for me.

About positive pairs... 10-15% of all pairs in this font have positive kerning, basically cause in small sizes in FL I noticed some irregularities.

Of course I used spacing/kerning tests.

And of course I have a list with problematic glyphs.

But, sometimes, math is not equal to eye.

- - - -

I hope (and I know :)) that things will come on it's place at the end, the problem was maybe I expected that this will be more faster done then it's possible, but all good results require time.

- Dusan

Arno Enslin's picture

@ tourdeforce

It is not possible to decide about the quality of kerning in low sizes on screen. You even have rounding errors in 72 pt (Points? I am not sure at the moment about the unit. I mean the size in FontLab).

When I kern, I change the size frequently. And in the end I print the pairs (contained in words).

I cannot really value all pairs in your examples, but only the extreme ones and primary in the bigger size.

But there were so many red marks (positive kerning indicators) – it looked as more than 15%. Round about 50%.

1037 pairs are indeed not necessarily too many, but there seemed to be many wrong ones. And again, whole sentences irritate my eyes. Single pairs also don’t give a correct impression. The context is necessary, but words are enough for the kerning on screen (in my opinion). It’s partly dependent from the screen size. On a small TFT I would search for words, that don’t contain more than round about eight characters.

When you are ready, you could provide a PDF with a list of words, in which all pairs are contained. That’s better than screenshots.

By the way, I think, that your font is relatively hard to kern, because it does not forgive mistakes. This is probably caused by the regularity of the shapes and the squareness. I generally don’t get warm with the most sans serifs, especially not in big sizes. (I miss the serifs in big sizes more than in low sizes.)

And you should know, that I am a pure amateur. When I am talking about letters, it is primary feeling, but not a long year experience. But I think, that I have a good feeling for spaces. Probably more than for the features of the shapes. Maybe not, I don’t know. I often can’t comprehend the opinions of the professionals here about some fonts. In my eyes many of the lauded fonts are oversharp or overdetailed. I don’t like Maiola for example. But I extremely like Merlo. I like the Berthold Van Dijck more than the DTL Elzevir, because I can better read Van Dijck. This is partly caused by my printer, but probably not only. My eyes seem to draw lines in the reading process, if the shapes of glyphs are extremely sharp, especially when the irregularities are very limited in a font. Stempel Garamond is already a joy with my laser printer, but if it is printed more dirty and less sharp, I can better read it yet. What I want to say is, don’t give too much on my critique.

tourdeforce's picture

Thanks Arno.

This is what I'll do

I'll make two font files:

1) one with existing spacing & kerning;
2) one with new spacing & kerning;

And let's see how things will work out.
My first aim is to have less kerning pairs and I think it's possible to go with less then 500 in this case. Second aim (first by importance) is to check spacing again and again and again. It simply require time to notice irregularities and like in book design, you see some typos after book is already printed, that's how it goes with font - it must sleep for a couple of days before I move on. It's pretty hard to see everything at once (at least for me).

Arno, no hard feelings, I really try to do good metrics for this font and reason why I started this topic is cause I wasn't so sure in what I done so far.

Hm.. I actually have an idea (which is probably already old, but whatever) about an additional way/help for doing metrics... I'm not good in scripting/coding, but idea is to do next:

To make software/script where you could define:

- number of pixels that will be between each glyph (for example, for all rounded glyphs volume between them would have e.g. 300px (300px as number of pixels that filled that space between them);

- should have an fixed size in points - for example it only works in 36pt, actually it's where's pixel's size 1:1;

Just an rough idea, there are questions what would happened with space between e.g. /C/ and /X/ when /C/ is open on the right side and /X/ on the left.

Just a thought....

Arno Enslin's picture

@ tourdeforce

You think about an autospacing or autokerning script? There is one autokerning service (iKern) and the DTL KernMaster. You could ask Igino Marini, what he wants to have for spacing your font. He doesn’t seem to have fixed prices. If it would be easy to write autospacing programs, that can compete with the human eye, there would be more of those programs.

I wish iKern would be a program, for which you can purchase a license, although it probably cannot beat the human eye.


There was an online kerning/spacing game. Unfortunately I have not bookmarked it, but it was very nice. Maybe somebody else remembers the link.

tourdeforce's picture

Thanks Arno.

I knew about iKern but the whole story is that I wanna do it by myself - spacing and kerning.
You gotta learn it one day, don't you ? :) (I mean me, not you :)).
Whatever time it takes > that's the price of knowledge :)

- Dusan

Frode Bo Helland's picture

Gornji: Although the screen is a bad place to judge, the text below has zero – 0 – kerning pairs.

Only at this point would I consider kerning. First: troublesome lc combinations, although most can be fixed with adjusted sidebearings and/or character shape. Second: combinations that leave a huge gap, like “T” followed by letters on the x-height, “L” followed by “T”, “A” followed by “V” and so on.

nina's picture

I've personally found that learning to see space takes serious time and practice, and is easily trickier than drawing glyphs. So while I don't think that (1) belittling the original poster or (2) showing how super well other people can do their spacing is necessarily the most helpful thing, and while I also don't agree that spacing and kerning can or should be completely separate processes, I'd also say: forget about kerning for the moment; scrap what you have for spacing; and start over with it.

Take your time with the spacing. Allow your eyes time to learn to see it. Use your eyes, but use a system for guidance. (I myself highly recommend the one Walter Tracy outlines in Letters of Credit as a starting point.) Spacing is very hard to critique from outside, so you'll probably largely need to trust your own eyes and your own system. Work on getting the spacing to a point where it works well in all but the most egregious exceptions, before you think of kerning again, which really should only be used to refine exceptions.

The reason BTW is not just that it's less work (in the long run) that way, but that you get better results in a variety of settings: Kerning is highly specific, whereas good spacing ensures a good overall «base» texture. If your font needs a lot of kerning to look decent, you'll easily be in trouble once someone tries to set a language you're unfamiliar with that uses character sequences you didn't test. And obviously you can never test all combinations that anybody might want to use, so it's advisable to have a stable basis and refine from there (through kerning). Also, there are still applications that don't support kerning.

Frode Bo Helland's picture

I second that, Nina. And my spacing is not perfect at all. I just wanted to illustrate my point.

tourdeforce's picture

Here's version completely without kerning - only spacing!

Enforcer Regular - Spacing Test (PDF ~450kb)

I used method of grouping all same/similar shapes (rounded with rounded, straight with straight...).

I noticed some irregularities (/e/ with /s/ is maybe too close...). Didn't checked so far how problematic combinations looks (/j/, /k/, /v/, /y/, /t/... /T/, /J/, /L/..), but at the this moment, this version is maybe the best so far.

- - -

Add-on: I just saw Nina's post. You're absolutely right about everything. It must take some time to figure it out by myself and all your comments were very useful in way making me suspicious if what I done in first place was good. If Arno didn't told me about too much kerning pairs, I'd never roll back to spacing as #1 in whole process.

Thanks for your kindly support and having a patience with me :)

Syndicate content Syndicate content