Setting Sidebearings

Eryk Kosinski

I would appreciate to know if there are any errors I have done while setting sidebearings for my glyphs. I would also like to know if left and right columns are balanced and not too dark on either side. The colums are test words from Emil Ruder and Niggli, Imprint of Braun Publishing AG, Switzerland.

Hrant Հրանդ Փափազեան Papazian

First of all: you need to settle on a grain size (because it helps you focus). I typically use 5.

BTW to quickly show spacing quality you can use a short phrase that has every letter in the middle of a word. I favor this thing I once concocted:

Incredibly, he makes a major life-change! For example: "I’ll require that the system have two sizes."

Source, with rationale:
https://github.com/davelab6/pangrams/blob/gh-pages/metagrams.md

Igor Petrovic

I would say that the color is pretty even in the left and right columns.

On the other side, I would say that spacing for capitals is significantly tighter than it's usual. Not sure what is the intended purpose of the font (text or display), because some display fonts intentionally go for obviously tight spacing. But given that the font is lighter than the standard "Regular" (the lighter weight the bigger spacing should be) and looking by its design I would expect bigger spacing for caps. 

The problem with too tight spacing is that it keeps unsolvable gaps like "TV" for example. You can make side bearings 0 and still have a very obvious gap if the rest is too tight, kerning can't help you unless to overlap letters.

 

Craig Eliason

Hrant H. Papazian said:

First of all: you need to settle on a grain size (because it helps you focus). I typically use 5.

Do you mean sidebearings should all be multiples of 5?! Hard disagree, I think with a UPM of 1000 (which I'd gather from the advance widths here) sidebearings should go unit by unit. 

John Savard

In general, the sidebearings of the text sample shown look excellent to me. But there is one exception. I think the capital letter "I" should have more space on both the right and the left.

Igor Petrovic

Craig Eliason said:

Hrant H. Papazian said:

First of all: you need to settle on a grain size (because it helps you focus). I typically use 5.

Do you mean sidebearings should all be multiples of 5?! Hard disagree, I think with a UPM of 1000 (which I'd gather from the advance widths here) sidebearings should go unit by unit. 

I use unit by unit for spacing (@1000 UPM), but kern in multiples of 5

Hrant Հրանդ Փափազեան Papazian

Craig Eliason said:
Do you mean sidebearings should all be multiples of 5?! Hard disagree, I think with a UPM of 1000 (which I'd gather from the advance widths here) sidebearings should go unit by unit. 

Not really *should* (note my "typically" above; I do go down to 2 on occasion) but I really believe going down to 1 is a case of over-assurance; even if a designer has the acuity and focus to –consistently– discern a difference of 1‰ it's pretty far-fetched to think non–type-designers (especially non-designer readers) can. And I don't mean [just] consciously, I mean their reading/experience will not be affected. Although I admit this is more true of text typography than display.

A –potentially– relevant parallel here: when fashion magazines horizontally squoosh photos of their models (to make them look thinner, which stinks of course) they can go down to 97% without people thinking "this is not a human". Although again, if this does apply it applies [more] to display than text.

@Igor Petrovic I've sometimes implemented such "asymmetry", because it saves effort without a marked loss in quality.

Johannes Neumeier

Try with a lot more different kinds of languages. Many languages have common letter combinations that are entirely absent from Latin-rooted languages, and they might be left looking odd in repetition.

It is also important to keep pushing as far as possible with metrics before doing kerning. At the same time try to not let the inevitably necessary kerns of some pairs drive their base metrics absurd by overcompensating too much. Not saying that is the case, just something I always find a tricky balance to strike myself.

As others have commented above, the uppercase seems quite tight. Often natural words in title case can benefit from this, at the expense of all caps looking too tight. You could also implement a capital spacing feature that spaces all caps out a little more. I agree with John about the I, but would say that extends to all the caps with straight vertical edges. They all seem a little dense (try some words with double N or H or followed by E/F/B/D/P). Also the capital S seems a bit loose to the right.

The lowercase harmonizes very well!

Matthijs Herzberg

To add onto what others are saying here, there appear to be some small inconsistencies in your sidebearings. For instance, /H has 65L and 65R, but /N has 62L and 62R, and M has 62L and 65R. /n is 67L 64R, /u is 64L 64R. Etcetera. I'm not going to inspect your entire table to look for these, but I'm sure there's plenty more.
These differences are minor and hard to notice, but if you're trying to produce a professional quality font, you'll need to fix them.

Hrant Հրանդ Փափազեան Papazian

@Matthijs Herzberg Since some of those things could actually make sense in a contrasted design (for example the stems of the "N" being thinner than the stems of the "H" might affect the space needed) I'm now wondering if the spacing was imported from somewhere...

Eryk Kosinski

Johannes Neumeier said:

Try with a lot more different kinds of languages. Many languages have common letter combinations that are entirely absent from Latin-rooted languages, and they might be left looking odd in repetition.

It is also important to keep pushing as far as possible with metrics before doing kerning. At the same time try to not let the inevitably necessary kerns of some pairs drive their base metrics absurd by overcompensating too much. Not saying that is the case, just something I always find a tricky balance to strike myself.

As others have commented above, the uppercase seems quite tight. Often natural words in title case can benefit from this, at the expense of all caps looking too tight. You could also implement a capital spacing feature that spaces all caps out a little more. I agree with John about the I, but would say that extends to all the caps with straight vertical edges. They all seem a little dense (try some words with double N or H or followed by E/F/B/D/P). Also the capital S seems a bit loose to the right.

The lowercase harmonizes very well!

I have made adjustment to uppercase glyphs and made S much less looser. Thanks for pointing out S glyph. I am still not sure if I glyph requires more spacing.

Eryk Kosinski

Matthijs Herzberg said:

To add onto what others are saying here, there appear to be some small inconsistencies in your sidebearings. For instance, /H has 65L and 65R, but /N has 62L and 62R, and M has 62L and 65R. /n is 67L 64R, /u is 64L 64R. Etcetera. I'm not going to inspect your entire table to look for these, but I'm sure there's plenty more.
These differences are minor and hard to notice, but if you're trying to produce a professional quality font, you'll need to fix them.

I am a begginer in type design and used book titled "Designing Type" to learn all principles. One of them is a way of setting sidebearings. 

Hrant Հրանդ Փափազեան Papazian

The guideline in that book is very coarse... which does not at all mesh with going down to single units.  :-/  BTW: the asymmetry in the "M" does not make general sense, especially in a contrastless design; the "i", "el" and "u" are also misleading, and the right sides of "c" and "e" should not be the same.

This particular design is actually relatively easy to space well, but it's a bit of a pain to untangle which glyph needs what and do a thorough job without a good test phrase... Could you please provide that?

Eryk Kosinski

Hrant H. Papazian said:

The guideline in that book is extremely coarse... which does not at all mesh with going down to single units.  :-/  (BTW the asymmetry in the "M" does not make general sense, especially in a contrastless design.)

This particular font is relatively easy to space well, but it's bit of a pain to untangle which glyph needs what and do a thorough job without a good test phrase... Could you please provide that? 

Here is a bigger M and O. For O it is 70 to 60 Units. For M, H, E and so on it is 68 and 62. Contrast is 90%.

Hrant Հրանդ Փափազեան Papazian

That's pretty mild contrast, but if you do decide to keep the spacing grain a single unit it can make sense... except I still think such a tiny grain itself doesn't.  :-)

Now compare the asymmetry in the "M' with the right sides of "c" and "e". Three units, versus one unit? That can't make sense, since the difference in the latter is much more pronounced... If you go with a coarser grain, the "M" would be symmetrical, and the "e" might have one unit more than the "c" (maybe two). Much easier to wrap one's head around, and only –some– type designers might complain.  :-)

Eryk Kosinski

Hrant H. Papazian said:

That's pretty mild contrast, but if you do decide to keep the spacing grain a single unit it can make sense... except I still think such a tiny grain itself doesn't.  :-)

Now compare the asymmetry in the "M' with the right sides of "c" and "e". Three units, versus one unit? That can't make sense, since the difference in the latter is much more pronounced... If you go with a coarser grain, the "M" would be symmetrical, and the "e" might have one unit more than the "c" (maybe two). Much easier to wrap one's head around, and only –some– type designers might complain.  :-)

Fun fact. As I edit in FontLab 7 in my metrics tab units of e and c on the right are the same. only in metrics panel it is different. Im curious if you experience the same thing as you work on your projects.

Hrant Հրանդ Փափազեան Papazian

Dunno, I always space by measuring myself, I never look at the panel.

Igor Petrovic

The guide in "Designing Type" is basically Walter Tracy's method, seen here and there in various forms. The main thing missing here is that spacing of capitals is usually somewhat bigger than for lowercase.

For example, if you pick 50% of the H counter as a value for side bearing for H then it makes sense to pick 50% of the n counter as the value for n sidebearing. Because the H counter is bigger than the n counter, hence sidebearing will be as well.

Somebody already mentioned the other problem, and that is: spacing of capitals will work well in "ALL CAPS" or in the "Title Case" setting. Can't work well for both (in theory). If it works for the "ALL CAPS" setting then spacing between the first (capital) letter and the next (lowercase) letter in "Title Case" looks too big. 

This can be solved by adding the OpenType feature which makes letter-spacing bigger when ALL CAPS is used. But more often you pick the side bearings for caps which are somewhat bigger than for lowercase but still look acceptable in "Title Case"

Chris Lozos

I always have a set of cap-to-cap kerns which opens up spacing.  It only requires 2 class kerns, H to H and O to H.

Eryk Kosinski

Igor Petrovic said:

The guide in "Designing Type" is basically Walter Tracy's method, seen here and there in various forms. The main thing missing here is that spacing of capitals is usually somewhat bigger than for lowercase.

For example, if you pick 50% of the H counter as a value for side bearing for H then it makes sense to pick 50% of the n counter as the value for n sidebearing. Because the H counter is bigger than the n counter, hence sidebearing will be as well.

Somebody already mentioned the other problem, and that is: spacing of capitals will work well in "ALL CAPS" or in the "Title Case" setting. Can't work well for both (in theory). If it works for the "ALL CAPS" setting then spacing between the first (capital) letter and the next (lowercase) letter in "Title Case" looks too big. 

This can be solved by adding the OpenType feature which makes letter-spacing bigger when ALL CAPS is used. But more often you pick the side bearings for caps which are somewhat bigger than for lowercase but still look acceptable in "Title Case"

Later on in the project I plan to make true low capitals. In other words custom glyphs and apply lowercase letter measurements for width of stems as well as sidebearings. Is it a good way of approaching this problem? 

Hrant Հրանդ Փափազեան Papazian

Smallcaps should be spaced to work together versus with lowercase, but regular caps (which is sadly what virtually everybody will use even if you have smallcaps) still have to work great with lc. As Chris suggests, kerning will generally take care of regular all-caps setting (and discerning users will track all-caps tighter).

Chris Lozos

(and discerning users will

That is the key. The skill and eye of the user was not questioned when they were all typographers.  Today, anyone can use type but that "anyone" is not trained or discerning.  I prefer to design type for the capable user who can and will use type in a discerning way, rather than focus on the naive user who may not see or use type any better than the simplest mechanical typewriter. I truly wish them well but assume the only people who actually buy type are those who should know type.

Hrant Հրանդ Փափազեան Papazian

@Chris Lozos I generally agree... while keeping in mind that we might still be able to help (or at least reduce harm to) the end-users (readers) in spite of the "typesetters"...  :-)

Thomas Phinney

Lots of good advice here. Spacing seemed tight for the size being displayed, especially in caps. Strange that the caps are slightly tighter than the lowercase; should be slightly looser, as discussed.

One broad piece of advice: how tight or loose the spacing should be overall depends somewhat on the intended usage size. Pick a specific size that you want the spacing to be ideal for, and accept that users may need to track it a bit in other situations. If you space it well for 10 pt text, it will look noticeably too loose at 48+ pt—and that is OK and even a good sign. 

Laurensius Adi

A bit of tips from me, how I learned as a beginner here. I download open source typefaces, such from Google Fonts, that looks about the same with the typeface I'm working on. Then open the font on your software, and see the font resolve spacing. How one sidebearing relate to other glyphs, look for proportion or ratio. I hope this helps.

John Savard

I approve of increasing the space between the capital letters. But I noticed what seemed to me to be a glaring omission in your set of sample words for testing. There were words in all capitals, and words all in lower-case. But what about lower-case words that start with a capital letter? The transition from capital to lower-case is very important.

However, that may be something you have to control with kerning. Not being a type designer myself, I made the suggestion that to avoid the need to use kerning for the most common case, one could get the proper balance by having a narrow right side bearing, and a wide left side bearing, for all the capital letters, so that words in all capitals would have a looser spacing than that between a capital letter and a following lower-case letter.

I remember that created many responses explaining why it was a silly idea, however, so I'm not really recommending it.

Vasil Stanev

Don't bother with numbers. The negative space between letters should look as big as the one inside them until you get a regular gray impression when you squint your eyes at the type. You start with an H,O, n,o and work up from there. Your first attempts will be clumsy until you develop an eye for it, later it gets precise.
Your eyegauge is the arbiter... simple as that.

I say look because there are many little ways to twist that. Mathematically correct spacing might look worse than what your heart tells you. You are an artist, not a scientist doing calculations.

Erwin Denissen

Matthijs Herzberg said:

To add onto what others are saying here, there appear to be some small inconsistencies in your sidebearings. For instance, /H has 65L and 65R, but /N has 62L and 62R, and M has 62L and 65R. /n is 67L 64R, /u is 64L 64R. Etcetera. I'm not going to inspect your entire table to look for these, but I'm sure there's plenty more.
These differences are minor and hard to notice, but if you're trying to produce a professional quality font, you'll need to fix them.

Personally I think there is nothing wrong with unequal side-bearings for symmetrical glyphs, as long as it results in better overall color.

Disclaimer: I am a software developer, and by no means a professional font designer.

Hrant Հրանդ Փափազեան Papazian

@Vasil Stanev Numbers do provide a foundation that saves time... because even one's eye isn't fully unreliable.

Hrant Հրանդ Փափազեան Papazian

John Savard said:

I remember that created many responses explaining why it was a silly idea, however

Could you please point me to that?

John Savard

Unfortunately, I don't know if this was on Typophile or Type Drawers, and I doubt I'll be able to find the thread with a Google search.

One attempt I have made, though, indicates that the Typophile archive is no longer online.

With lead type, there is an obvious issue: space can only be added, not taken away, so this would be incompatible with the rare case of a capital letter in the middle of a lower-case word.

Also, spaces between all upper-case words and all lower-case words would be affected.

With computer typography, though, I would think that it's more debatable. Shifting the sidebearings would allow the most common cases to work well with renderers that don't support kerning, and kerning instructions for the less common case of a lower-case character followed by an upper-case character, and for an upper-case character preceded by a space, could be included in the font.