Where should diacritics lie within vertical metrics?

I am confused as to where to place my diacritics within the vertical metrics of a 1000 pt em square... for example, do the acute and grave accents, the circumflex, tilde, etc have to be included within the em square, or do they sit on top of the em square? If they must exist within the em square, to I have to increase the vertical space between the cap height and the ascender line to accommodate diacritics? Any direction here is much appreciated!


  • Chris LozosChris Lozos Posts: 1,290
    They do not have to be confined to the typical vertical metrics but you have to consider your users language.  Vietnamese needs more space because of stacking. Most users of frequent diacritics are accustomed to configuring their leading to go with the language in use.  If you make your xheight too small, users will increase size to accommodate.  A capable type user, familiar with the language, can make the needed adjustments.
  • Just to add to Chris's point, I always set my leading to a specific value and become very annoyed with any software which doesn't allow this and forces me to rely on the font's vertical metrics. The vertical metrics should be set to reasonable values, but I wouldn't worry too much about what approach is correct since I suspect your target audience won't rely on them.
  • In layout applications, the effect of the em square is primarily to determine how to scale glyphs. For example, if text is set to 12 points, then the em square is scaled to 12 points.

    The metrics used by the application for laying out a line are separate. When laying out text using an OpenType font, the sTypoAscender, sTypoDescender and sTypoLineGap are recommended for the application to use for placement of lines (as defaults: applications can always override).

    So, when considering how to size and position outlines within an em square, the first thing I think would be considered is how you want the size of text displayed with this font to appear in comparison to other fonts. Then, I think you'd want to place diacritics wherever you think is most appropriate; and then set the font's line metrics to be appropriate for your main, target use cases.

    Some languages may stack multiple diacritics above or below base letters, but you don't need to set line metrics to allow for that if it's not a main use case you're targeting. Math fonts, such as Cambria Math, have similar issues: they can have glyphs for brackets or integral signs, etc. that are very tall, but the default line metrics are not set very tall to accommodate them; otherwise non-math text using the font would by default probably get laid out with way-too-large line-to-line distance.

    Caveat: I'm not a font designer; I just have some familiarity with the font data and how it gets used in apps.

  • John SavardJohn Savard Posts: 705
    I would have thought that all accent marks must be within the em square, because otherwise if you don't specify additional leading, descenders will conflict with the accent marks. Although most text processing programs just seem to insert extra leading, but even this means that the font is badly behaved, and using it leads to unpredictable results.
    However, unlike most of the people here, I am not a professional font designer, so apparently there's some good reason why I'm wrong. I guess it's just not the convention to treat digital type like lead type.
  • John HudsonJohn Hudson Posts: 2,036
    It wasn’t uncommon for accents on capitals to extend beyond the body height in metal type, using vertical kerns (in the original sense of the term kern: a section of the face of type overhanging the body and resting on an adjacent sort or leading material). Casting moulds made this possible by being able to hold a matrix that is larger than the body size. I’ve been looking for pictures of vertical kerns, but have not found any yet, but this illustrates the concept in a classic horizontal kern:

  • Craig EliasonCraig Eliason Posts: 1,081
    There also seems to have been a healthy tradition of just leaving off problematic diacritics atop capitals in printing, at least in French. 
  • John SavardJohn Savard Posts: 705
    I'm now thinking that the reason that examples of vertical kerning are hard to find is tht when capital letters are accented, it may well be that piece accents are used.
  • John HudsonJohn Hudson Posts: 2,036
    Dan Rattigan reminded me that vertical kerns were also intrinsic to Monotype’s 4-line mathematical typesetting system. In this illustration, the grey rectangles indicate the body of the type and the white rectangles indicate the spacing materials supporting the kerns.

  • Craig EliasonCraig Eliason Posts: 1,081
    edited January 29
    Better not drop that b or X or Y on the shop floor!
  • Chris LozosChris Lozos Posts: 1,290
    @Craig Eliason  Metal type was a hard medium ;-)
  • But lead is a soft metal. Parts might not snap off, but they could bend.
  • John HudsonJohn Hudson Posts: 2,036
    Foundry type was hard; hot metal type was soft.

    Jean François shared this image from the Imprimerie national of a two-part, display size diacritic, tightly bound together:

    And Mathieu Triay provided this photograph of a vertically kerning Q:

Sign In or Register to comment.