Theory of spacing

raph's picture

I'm working on spacing my Century Catalogue (inspired in part by the recent digraph discussion), and have again come up against a fundamental question that's been bothering me for a while.

Tracy's theory of spacing in Letters of Credit makes a great deal of sense to me, especially for the metal type days when it was not practical to have kerning pairs for all possible digraphs. Essentially, Tracy recommends carefully determining the side bearings for 'n' and 'o' such that "nnonn", "nnonon", and "nnoonn" all look consistently spaced, then setting the side bearings for all other letters so that nxn and oxo look consistent.

However, to my eyes, no matter what side bearings I choose, "oooo" looks more tightly set than "nonon" and "nnnn". If I widen the "o" side bearings, then "ooo" looks good (just like "nnn"), but "nonon" is too loose relative to "nnnn".

I know there's a widespread sentiment that relying too heavily on kerning is just a symptom of bad side bearings, but it seems to me that kerning these standards might actually be the right way to solve the problem.

There are a few ways I could go with this:

1. Nice big side bearings on the "o" so that "oooo" looks great (just like "nnnn"), then kern no and on (and, of course, all other pairs from similar classes).

2. Nice narrow side bearings on the "o", then plus kern on "oo". This seems to be closest to the side bearings I see in the ATF catalogues, most likely because it's easier to add a thin space than subtract.

3. Halfway between, a bitty minus kern on "on" and "no", and a bitty plus kern on "oo". This seems to be the closest to the spirit of what Tracy is recommending, not least because it gives reasonably good results when kerning is turned off.

It's also possible that my sense of "oo" being too tight is just plain wrong; that the side bearings are the right way to do this and kerning is not needed.

Any insight would be much appreciated.

Nick Shinn's picture

You don't need any kerning between o's and n's. Once you start, you have opened pandora's box.

Certainly not for a metal-era face like Century, the forms of which were designed to work without kerning.

I have added kerning (positive) between lowercase round letters, in a sans serif. But it was for display use, very tight setting.

I put wider than usual sidebearings on round letters in Richler: it's good for reading text if you don't stop and think "that's not like Garamond". It worked for Richler as the rounds are quite square and the sides of the round letters were tending towards a vertical stem.

One thing to consider is that the relevant "sidebearing" for the n is not the distance to its serif extrema, but to the edge of its main vertical stem. (It could even be argued that the distance to consider is center-to-center between adjacent stems.) You may be able to remedy your problem by adjusting the width of the serifs on the n.

Nick Shinn's picture

Your tilte "Theory of spacing" got me thinking.

This diagram addresses the space between stems.
You will see that the internal distance is as important as "sidebearings".
Here, there are five possible "stem-to-stem" distances:
A, B, C, D, and E (the distance o-n is left out, as it is the same as n-o).

In ascending order: E,B,D,A,C.

I suspect that, for this traditional kind of typeface (Worldwide, my "Century") the ratio of these distances should follow a smooth mathematical progression, to create an aesthetically pleasing rhythm for the reader.

raph's picture

Thanks for your thoughtful and illuminating response, Nick. I particularly liked the idea of establishing a "rhythm", as opposed to the more commonly expressed goal of consistency (which, in the limit, becomes monotony). I wonder if you would have responded so nicely if you knew I was planning to release my Century Catalogue as a free font ;)

In any case, I feel a little foolish. The spacing I had in my font was based on the 18-pt metal, which I identified as the best source for the shapes. However, the 18-pt spacing is too tight for text use (although certainly not out of the range of what you see today). Trying to do tight spacing with non-kernable metal doesn't work that well, as you can clearly see in words like "sober" (line 6 of the 18-point in this scan). I think that, for this size at least, my point above stands: nn, no, and on all look good (the word "higher" is a good example of all these classes, with a bonus g thrown in), but oo is too tight.

By contrast, the text sizes, such as the 10-pt in particular, are beautifully spaced. You hardly miss the kern pairs; in fact, after loosening up my metrics a bit, many of the "v" and "y" pairs seemed over-kerned.

I propose the following tentative conclusions:

1. Spacing was a big part of the optical scaling done in the metal days, certainly for ATF fonts.

2. For the relatively loose spacing needed for text, good side bearings do almost all of what you need; kern pairs are only needed for true problem pairs such as oy.

3. Even so, if absolute consistency is your goal, you need slight plus kerning on oo pairs. But of course this wouldn't be at all authentic for a metal revival, aesthetic concerns aside.

4. For tighter spacing, such as display work, side bearings are not enough. ATF chose the compromise of relatively small side bearings for 'o' class, which at the very least left open the possibility of thin spacing 'oo' class pairs (although I haven't seen any evidence for this in the ATF specimen book).

bieler's picture


I'm not sure I quite follow the conclusions but...

in terms of metal type, the body, not the face, is where all the action is.

Practically all the well established metal type foundries practiced careful placement of the letterform on the body on a per size basis, long before the idea of optical scaling comes in to play. Any good old foundry type, especially from continental foundries, sets near perfectly, in harmony with whatever partner comes along.

Optical scaling itself was the result of the forced size range jumping made possible with the pantograph matrix engraver (and, ignoring the machine comp folks, not fully practiced by foundries until the 20th century). Optical sizing would be the more common term we now associate with the natural practice of the punch cutter.

Kerning was quite possible with metal type, far beyond the provision of alternate body widths. Moxon discusses the tools for it back in the late 17th century. In the twentieth century, the printer's saw was often used for kerning, and there was a machine, I had a couple, type mortisers, that were made specifically for the task.

And there were, of course, thin spaces, beginning with steel quarter point, copper half point, brass full point...

I would think the best way to authenticate the spacing of a metal foundry face would not be through specimen books but rather to proof it with type high thins between the bodies to provide information about placement. I believe Hrant did that with his investigation of Massive(?).

Nick Shinn's picture

>best way to authenticate the spacing of a metal foundry face

Reminded me of this sublime piece of detective work:
(One for book reviews, Tiffany?)
Type Spaces by Peter Burnhill

He does the measurements on pages from the Aldine Presss, deducing the systematic realtionship between page, line and letter placement which Aldus Manutius instituted.

Syndicate content Syndicate content