Elemaints - A Serif Family with Optical Sizes

Linus Romer

@Christian Thalmann: So here is the return to the /zeta. Additionally, I have again worked on /vartheta and /sigmafinal:

Christian Thalmann

Looks good to me overall!

Is the bottom of /theta a bit light though? And the top a bit too dark? The overall shape also still strikes me as a bit unusual; a bit pointier on the bottom than I'd expect, and bulging out to the right.

Also some very tight spacing here and there, such as in Ωκ and άκ.

Linus Romer

Yes, the botoom of the /theta was too light and the top too dark. I do not find it too pointy at the bottom, but I relaxed it a bit. After you mentioned it, I could see the bulge to the right, too. I tried to fix this all.

The letters /zeta, /xi, /chi and /a have also been repolished. The spacing was redone (/Omega and /alpha had to low right side bearings amongst others). The kerning is not done yet, hence some pairs may still look odd spaced.

Adam Jagosz

/chi consistently looks tilted to the right to me

Christian Thalmann

Greek looks very good to me now.

The top of /delta still contains a lot of tension. Maybe it would profit from widening the curve radius of the top left turn, which would require lowering the diagonal and maybe adding some compensatory weight to the flag.

The ano teleia should reside at the x-height, not at half the x-height, as far as I know.

I don't mind the /chi's forward drive, it just looks a tad too wide to me.

Is that /a for the Tiny version? Not sure what to make of it... the terminal is better now, and the symmetry breaking of the arch is a valid design choice... but the current implementation of that arch looks fractured in two places to me. (Then again, I feel like I'm meddling in your /a way too much already. I should probably just wait and see where this is headed.)

Linus Romer

I tried to follow the advice of Adam and Christian for /chi, /delta, ano teleia and /a:

The /a that I have shown in the last post was for the Caption face (as well as the Greek letters). There should be less abrupt turns now...

@Christian Thalmann Actually, I am glad about your meddling. I appreciate your advice much!

Christian Thalmann

I agree with all changes. That’s a very pretty display /a now!

Dave Crossland

Looking forward to seeing this on CTAN

Will you use OFL?

Linus Romer

@Dave Crossland Yes, the outline fonts will be quite probably published under OFL with reserved font name.

I have meanwhile worked on the tabular oldstyle figures (still for the caption face):

I am not sure, whether the /zero should be a bit smaller ( = more side bearings).

I have also begun with double-struck capitals:

Again, any comments on the tabular oldstyle figures or the double-struck letters are welcome.

Belleve Invis

@Linus Romer
Love it! Will you do Italics and more weights? I want to use it in my paper.

Steve Gardner

I'm seeing a bit of a bulge in the curves of the /U and similar glyphs (e.g /⊆, /∈).

Linus Romer

@Belleve Invis Yes, I intend to create italics and bold and semibold for each optical size as well. So I will end up with 5 (optical sizes ) × 3 (weights) × 2 (upright/italic) = 30 faces for the whole family. The design of the upright faces is already defined (but the faces are not done yet) and I will have to struggle a while with italics. Anyway, the completion of the family will take quite a while, so you (and me as well) have to wait some time before we can use it in our papers.

@Steve Gardner Where do you see these bulges? At the joint of straight lines with the curved lines? In the following picture, the change of curvature between straight and curved lines has been decreased. Does it look better now?

Steve Gardner

@Linus Romer I've shown where I'm seeing bulges in the attached PDF using broken red lines.  Maybe the nodes at these points could be lower, reducing the gap between the unbroken horizontal lines and making the curve appear a little less 'pointy'?

Linus Romer

@Steve Gardner I tried to follow your advice (old version left, new version right):

Do you think, the new version looks better? I had a look at the /U's of professional serif fonts and have the impression, that there is always a bulge at the joint between a straight horizontal/vertical line and a curve (as a matter of rasterization).

Steve Gardner

Personally, I think that's a huge improvement! 

AbrahamLee

The curve continuity is much better now, I agree, but I like the old general shape better. Doesn’t feel so squashed on the bottom. So, I’d take what you have and, with the interpolated move feature on, select the bottom tangent points and move them back up a little until the bowl has a similar overall curve as before.

My opinion.

AbrahamLee

You also might want to make the curves that go from stem to serif a bit more consistent. At small sizes, I doubt it would get noticed much, but at larger sizes it might.

Thomas Phinney

You can get the shape of the bottom of the bowl more round and less flattened, while still maintaining the gradual join on the sides. I was glad to see the elimination of the sudden join you had originally—it just looks like a novice error.

Linus Romer

I tried to follow the advice of @Thomas Phinney and @AbrahamLee (bowl more round, join of the serifs softer):

Comparison to the old (round) shape (old left, new right):

Thomas Phinney

That's better. Not only because of the join, but also in the old version, the thin stroke thickened a bit too suddenly at the start of its curving.

Linus Romer

In my original sources, many arcs relied on superarcs which were constructed with an intermediate points (hence two bezier paths). FontForge removed these intermediate points in a suboptimal way (see picture below) and therefore I did not remove those intermediate points. But the remarks of @AbrahamLee and @Thomas Phinney made me try Fontlab. Fontlab did a much better "remove point" job than FontForge. Up to this moment I did not realize that the approximation with a single bezier path could be as good...

Thereupon, I rewrote my sources based on the following mathematical approximation (s stands for the superness of the superarc, a stands for the relative position of the control point between a borderpoint and the intersection of the border tangents):

The own approximation is approximately the same as Fontlab's approximation. So, because of Abraham Lee and Thomas Phinney the font file will be much more elegant. Thank you!

Belleve Invis

@Linus Romer

Cubic Bezier curves are, well, a modified form of cubic Hermite spline.
So if you have the parametric equation of a derivable curve, you can easily get the control points.

There are multiple approximation strategies, and you can try the best:

• Derivative: $b(0)=f(0)$, $b(1)=f(1)$, $b'(0)=f'(0)$, $b'(1)=f'(1)$

• Tangent-middle: $b(0)=f(0)$, $b(\tfrac{1}{2})=f(\tfrac{1}{2})$, $b(1)=f(1)$, $b'(0)=k_1 f'(0)$, $b'(1)=k_2 f'(1)$

Linus Romer

@Belleve Invis Is the tangent-middle strategy the one that is used by FontLab?

Linus Romer

Meanwhile, I have completed the math encoding blocks for the caption face (in LaTeX: OT1, OML, OMS, OMX, AMSa, AMSb, TS1):

(The other optical sizes and the italics will be added later.) The blackbord bold and the calligraphic capitals are displayed below. The calligraphic capitals are based on TIX2/Euler.

The attached PDF shows the math usage of Elemaints in detail. Any comments on shapes, stroke widths and glyph positioning are welcome.

Linus Romer

I have added polytonic Greek and small caps to Elemaints Caption. Furthermore, I have refined the spacing and completed the kerning. Any comments regarding spacing, kerning, small caps or polytonic Greek are welcome.

Christian Thalmann

Shouldn't Greek small caps follow the all-caps rules for accents?

Linus Romer

@Christian Thalmann I did not think about it until your comment. After some research I would say: Yes, your suggestion would be the modern convention. And that is what the Greek Font Society did for GFS-Didot and GFS-Bodoni (actually they just removed any accents, which is wrong again as John Hudson explains in http://www.typophile.com/node/83725).

An extract of the link above:

In Arno Pro the accents are placed above the small caps. Is it a rule? John, do you place the accents above or in the left of small caps?

Above. Then I use OpenType Contextual Alternates substitutions to suppress marks in both all-caps and small-caps settings, following the modern convention, which allows users to turn that feature off and display marks if they desire.

As far as I understand, modern Greek does actually not use polytonic glyphs anyway, but some still use it (https://en.wikipedia.org/wiki/Greek_diacritics#Modern_use_of_polytonic_system ).

Maybe, I will ask Dimitrios Filippou from The Eutypon what he would prefer (because he seems to use still polytonic).

Linus Romer

@Christian Thalmann Meanwhile, I have asked Dimitrios Filippou: He would prefer the variant with accents above the small caps.

Of course, an additional OpenType feature that enables the other rule would be cool. I will consider adding it, as soon as someone requests it.

Dimitrios also found some errors (e.g. that some omegas and upsilons were not small caps) that I have corrected now:

Linus Romer

... (sigh) ... the errors are hopefully corrected now (omicron with varia and omicron with oxia had omega as base):

Linus Romer

I have refined the display face of Elemaints:

Latin:

Greek:

Any suggestions and comments are welcome.