Self-ref - a special purpose serif font

Hello,

I have created a special purpose font for my project. I do this for
boredom and curiosity. I'm new to font design. By profession, I'm a
programmer. I also do lot of photography and like visual arts. I live
in the Czech Republic.

I have created few small resolution bitmap fonts in past for some
games but this is my first outline font.

Purpose: few years back, I created this mathematical curiosity:
http://jtra.cz/stuff/essays/math-self-reference/index.html Almost ever
since that I wondered if I could do something similar that would not
be bitmap based, but with nice smooth font.

About a year ago I started learning math that would be needed for
this: Bézier curves.  I have the math part prepared. I made a program
to play with curves which I eventually turned into sort of glyph/font
editor (screenshot included) in which I designed following font. I
call the font 'Self-ref' based on purpose for creating it.

The font originally started as zero-contrast slab-serif but later I
tried increasing contrast and kept it that way.

Design objectives: To emphasize that used mathematics is flexible, it
should not feel like pure geometric font. It should not be drawn with
circular pen. Serifs. Contrast should not be extreme.

The font will be used for typesetting some mathematics and few texts
around.  I know TeX/LaTeX and I'm aware of its enormous use in
mathematics papers. TeX defaults to italics font for most of the
mathematics. It is nice that it stands out in wall of regular text,
but I don't feel like all the math needs to be typeset in italics. I
also feel creating italics font would be much harder for me. So I want
to use roman type for both mathematics and texts around. It will not
be much text anyway.

I have automatic kerning algorithm which I feel sort of works most of
the time except for symbols which are often kerned too tight (for now
I correct it manually after kerning).

I have not yet tried to use this to typeset much mathematics. I will
need few more glyphs and I will probably want to improve my automatic
kerning algorithm for symbols before that.

There is only a bitmap output, no PDF, so I include two bitmaps of the
same text, one to fit forum post and other high resolution
bitmap. There is no hinting so low resolution look is not great.

What problems and inconsistencies stand out in the font? Should I try
to simplify it (are three ways I do terminals too many)? Anything you
like? :-)

Link to higher resolution image (317KB)
Link to screenshot of my glyph designer (114KB)

Comments

  • I suggest to do more background reading about type design

    * DesignWithFontforge.com
    * GlyphsApp.com tutorials and handbook
    * fontlab stuff, especially their YouTube series
    * pick up a copy of http://www.tipo-e.com/publicaciones/how-to-create-typefaces.html


  • Ray Larabie
    Ray Larabie Posts: 1,438
    edited February 2018
    I feel like there should be more of a difference between the terminals such as on the y and some of the heavier vertical serifs such as on the capital EFLTZ. The terminals in some places serve to add presence where needed. But on the EFLTZ, they make the horizontal serifs seem too light.

    I like the diagonal chunk (chamfer?) action on the curves but it's too subtle to be seen except at very large sizes. They need to be exaggerated because at lower resolutions/smaller sizes, they'll merely read as noisy curves. If you look at the chunk on the top curve on the capital S, the way the curve hits the chunk is too subtle. It kind of half blends into it. The curve needs to bang into that chunk hard so it leaves a sharp corner. We are chunks and we want you to know it! You can cut the chunks deeper so what's left is something that as thin if not thinner than the thinnest part of your h.

    The lowercase v has a hearty chunk carved out of it while the lowercase u has a meek chip. Make all the chunks equally noticeable.

    I think it's fine that the chunk rule is used here and there. For example, there are S chunks but no O chunks. I like that you didn't over-apply the rule.

    The thing you're trying to do on bottom right of the a, and other similar letters is too weak. It's like a vestigial tail that doesn't seem to fit with anything else in the typeface. Same with the tail on the inside of the Q. 

    The left side of the b doesn't follow the rule you've established with the rest of the lowercase: no left facing serifs on the bottom. Also: compare how uh and ub fit together.

    Capital B feels out of place. You've got a low waistline on the EFHR and high on the B with a little bowl on top.
  • Dave,
    thank you. A "How to create typefaces" looks like good book. I have looked at books locally, but I did not find anything useful beyond few books that focused more on lettering rather than typography. I have read a lot of online texts, including parts of the DesignWithFontforge.com. I watched a lot of episodes of Fontribute by Thomas Jockin on Youtube. But there is always more to read and see.

    Ray,
    thank you very much on specific feedback. I will have to try how your suggestions work for me. I will report back in a week or so.
  • Hi Jakub,

    I agree with the comments above. "How to create Typefaces is a good place to start".
    I'd also take a look at https://youtu.be/LR-CG5eB3nQ by Thomas Phinney. He has some other good videos on YouTube well worth a look.


    I'll add the following quick observations. Some of the serifs are very heavy (E) compared with those on ABC. The crossbar on the H is probably mathematically centered, but should be higher — you want optical centering. Similarly with the /s

    The way the bowl strokes intersect with the verticals on /a /b and /d are too heavy, it should taper much like you've tapered the /m.

    The shoulder on the /r is too wide.
    I hope this is a help. Keep at it!
  • Hi,
     
    Thank you all for feedback. I have made some changes the to font, mostly based on feedback.

    Here is how it looks now:

    Here is a comparison image:


    Here are links to higher resolution variants of above images:
    http://jtra.cz/typeface-review/selfref-r2-sample-big.png (310KB)
    http://jtra.cz/typeface-review/selfref-r2-sample-comparison-big.png (1.2MB)
     
    List of changes:
    /a and /d have larger bottom serif and more similar to others. Thanks for advice here - I was looking at them for too long (they were among earliest letter I draw) so I accepted the original look without questioning. I like the new look much better.
     
    /b no longer has bottom serif. I have experimented with several variants and this was look best to me.
     
    /g has slightly fatter ear.
     
    /b /d /p /q now have more unified bowls. I liked some aspects of the old look, but new one is more consistent.
     
    /r has slightly shorter shoulder. I feel like making it even more shorter would require different approach. I see most fonts have very short shoulder compared to this one though.
     
    /s has spine slightly higher.
     
    /v /w /x now have smaller diagonal chamfer to fit more in others.
     
    /y has new tail more consistent with other glyphs and it is slightly wider.
     
    /B has waistline lower.
     
    /E and /H but not /F have waistline higher.
     
    /Q has a new tail.
     
    /S has just slightly more prominent chunk at the top.
     
     
    I have made digits wider. /1 now has curved arm (I'm not sure what is proper terminology for this), but I have not decided whether to use curved or go back to straight. I have adjusted waistline of digits /6 /3 /9. I have made dots in math symbols bigger and changed comma.
     
     
    I tried to resolve difference between fat horizontal-to-vertical serifs and thin vertical-to-horizontal serifs (in both upper and lower case) in a experiment where I made thin ones 50% fatter, but I did not liked the end result. I tried to make fat ones thinner, but I did not liked the end result either. I don't see a way how to approach this without making font very different.
     
    I will try more exaggerated chunks next, but I don't mind them being less noticeable at lower resolutions/smaller sizes. Intended output is digital and zoomable without ability to adapt for actual zoom unlike regular fonts. I'm ok with glyphs looking more normal when small but somewhat more eccentric when enlarged.




  • Ray Larabie
    Ray Larabie Posts: 1,438
    I like it. I think the lowercase qu combination produces too much of a gap. Similar to the way ub fit together in the old version. Maybe try going a little higher with the lowercase accents and ij dots. Mathematical symbols: type 1+2+3+4 etc. Move + vertically until it doesn't look too low. Adjust the rest of the mathematical symbols to the same height. They're usually seen with numerals so they don't need to match the hyphens.
  • Thomas Phinney
    Thomas Phinney Posts: 2,901
    I think you need more overshoot, at least on the caps (but probably the lowercase as well).

  • In terms of typography it is quite imperfect, but I prefer some imperfection over not completing it. Thank you for typography advice.