Elstob: a Fell-based font

Peter Baker

First a bit about me: I’m a professor of English medieval literature, fascinated since the 1980s by all things to do with type and lettering—both because paleography is a significant part of what I do and because of my struggles, as a young scholar, to get my KayPro II + dot-matrix printer to produce the characters needed for medieval texts. Since then I’ve produced several fonts based on medieval scripts, and also a general-purpose font called Junicode (loosely based on the Fell pica), which is pretty widely used—more because it’s free and has lots of goodies for medievalists than because of its design. I am self-taught and frankly a bit scared right now.

This is a variable font called “Elstob” after Elizabeth Elstob, a celebrated medieval scholar of the early 18th century. It takes as its starting point the Fell double pica used by the Oxford Press for dedications and the like in the decades around 1700. That face has its faults, but I think it has a kind of simplicity (for example, flat or flattish serifs, minimal or absent brackets) that make it a good model for a variable font. Elstob has three axes: weight, optical size, and grade. I have tried to keep the classic proportions and modernize the details, keeping the outlines as simple as possible to hold down file size. I hope to end up with a practical and readable typeface for websites having to do with medieval studies.

I won’t go into my own specific design concerns, but will just say that the farther I get from my models (heavier weights, smaller optical sizes) the more anxious I feel. Also, as I don’t think of this font as a revival, I am very willing to depart from the original for the sake of design improvements. I have been working on the roman and italic more or less simultaneously, so my samples have both (sorry). I welcome any advice this community can give.

Here are links to a PDF, a web page with the obligatory sliders and checkboxes, and a couple of images (here and here) from the 1925 specimen book, printed with type cast in the 1890s from the seventeenth-century matrices. And, of course, an image:

Craig Eliason

What you've identified as "regular" seems pretty light to me. 
The source pics show capital stem weights as much heavier than lowercase stems. Even if you don't want to go as far as that I still think your cap stems should be wider. Similarly, you may have lowered the tittles a little too much.
I'd ease up a bit on the lean of italic A and V.
Consider making the cap vertical serifs (on C/G/S, and maybe on the middle arms of E/F) more wedge-shaped instead of just line segments. 
b/d/p/q counters should capture about the same amount of white space.
Did you consider having the optical size axis change contrast, and not just proportions and sidebearings? I think a higher contrast master would make this face work at a broader range of sizes. 
The space glyph may be too wide.
Fun variable-font showoff page!

André G. Isaak

The thorn, B, and R runes (which are essentially recycled latin glyphs) don’t really fit with the others. Compare your Futhark thorn with your Futhark W; I’d redesign these so they're more structurally similar to the W rune.

André G. Isaak

Peter Baker said:

Since then I’ve produced several fonts based on medieval scripts, and also a general-purpose font called Junicode (loosely based on the Fell pica), which is pretty widely used—more because it’s free and has lots of goodies for medievalists than because of its design.

I think you may be selling yourself short here. Junicode isn't the only free font for mediaevalists out there, but it’s certainly one of the two most well-designed ones (the other being Andron Scriptor). I used to teach courses on the history of English and I used Junius (the predecessor of Junicode) for many of my handouts because I preferred it to all of the commercial fonts which I had looked at.

notdef

For runes with serifs, Johannes Bureus’ alphabet that he produced for Runa ABC is a very relevant reference: http://www.kb.se/F1700/ABC.htm AFAIK it was the first attempt at something like this. Runes were still in use well into the 1900s in rural Sweden.

Peter Baker

Thank you, Craig, André, and Frode. Lots to think about here.

Craig: It's going to take a while to work through all this great advice, but here's a start. I think you're right about the weight of the regular face. Moving my sliders from the current 380 to 420 produces something more like other regular weights:

A closer look at bdpq, which I spent some time last night trying to normalize:

I have been worrying about the weight of the capitals. I don't want to go all the way to the weight of my model—but I'll be experimenting with some compromises.

The optical size axis does change the contrast, especially thickening narrow horizontal stems, where I was worrying about their getting too thin at small sizes. I'll experiment with a greater change in contrast, together with the other things you mention

André, Frode: I was looking at the Runic font in the 1704 Fell specimen book (it's also used in Hickes's Thesaurus), which looks like this:

I thought I could get away with the kind of thing they did with the Thorn and R runes, but I suppose not. It is jarring.

Thanks for the Bureus link, Frode! It seems likely that whoever was responsible for the Oxford runic font was looking at Bureus. I think the runes with serifs, harmonizing with a font's other capitals, are very much in the spirit of runes written in medieval manuscripts (as opposed to being carved in wood or stone), e.g. in the Exeter Book (scroll down for images).

Thanks for the kind words about Junicode, André. I'm glad a lot of people like it, but somehow I've never quite managed to love it myself. I am, however, a big fan of Andron.

John Savard

I certainly think this is a beautiful typeface.

Peter Baker

Thank you, John. Maybe it has a chance, anyway. Here's another attempt, increasing the weight of the regular, as above, and making the ratio of lc vertical stem width to uc vertical stem width match the model:

Better? Or have the caps gotten a little too thick now?

Craig Eliason

Maybe a touch too thick. But most also look like they're really crowding the ensuing lowercase letters. Give a little more sidebearing to them before making the final call on weight. I also like to set a paragraph in all caps and another in lowercase to compare overall color.

Peter Baker

I increased the spacing of the caps, but they still look heavy to me, and I get the same result from the experiment you mention with the para in caps. So I'm going to back up and try again, and will post when I have something I like.

Thanks again!

John Savard

Your sample text is in German. Monotype actually made a special version of Times Roman with lighter caps for German because that language capitalizes more words.

Peter Baker

Interesting! Did the font have both sets of caps, or just the lighter set? I chose German for the sample because of all the caps.

Peter Baker

I don't know if I'm there yet, but certainly getting closer:

There's a PDF with the longer passage and blocks set in caps and lc here. I can see some light spots (E) and dark spots (Z) in the caps, but overall (to my eye, anyway) it seems a fairly good match for the lc.

John Savard

Peter Baker said:

Interesting! Did the font have both sets of caps, or just the lighter set? I chose German for the sample because of all the caps.

Just the lighter set. Normally, when one thinks of Monotype Times Roman, one is thinking of Times Roman 327. But several other variants were made. One prominent example is Times Roman 569; this was developed specifically for use with Monotype's 4-line mathematics system, and its primary difference was that the italics were sloped at a more gentle angle. The version for German was Times Roman 727; there was also a version for French, Times Roman 827, which modified the shapes of a few letters to make them similar to forms popularized by the Romain du Roi.

Also, Times Roman 627 is a version with long descenders, and Times Wide 427 is a version which is not condensed in the way regular Times Roman is, these versions both intended to be more suitable to printing books.

Peter Baker

I guess it makes a lot of sense (and is affordable) when you have such a widely used product. It must be an immense money machine.

Peter Baker

I've adjusted the weight of the caps in the italic face along the same lines as in the roman, increasing the weight, but not as much as in the model:

Longer sample and pangram here. Also, the runic Thorn, B and R, which earlier duplicated roman letters, have been redone:

I'm not sure I like the R and B yet. Suggestions/reactions/vilifications solicited.

John Savard

Peter Baker said:

I guess it makes a lot of sense (and is affordable) when you have such a widely used product. It must be an immense money machine.

I'm sure it was an immense money machine, from 1935 on up to when it was licensed to Microsoft. I'm not so sure that it still is, at least to the same extent.

Peter Baker

I've been laboring all day on reworking my italic e and n, on the theory that if I can get those right, a whole lot of the italic will fall into place. First, here's a little sample from my model:

I like the spiky look of the i and the right side of n: I don't like the tucked-in look of the upper-left corner of the n, so I've changed that. Here's my attempt so far:

My problem is, the farther I get from the model (in weight and optical size), the less confident I am that I'm capturing the spirit of it. I'd love to have comments/suggestions.

Craig Eliason

Your changes from lighter to heavier weights seem to only be due to the strokes getting thicker. The spaces between the strokes--the counters and interletter spacing--seem more similar, e.g. similar areas of white space surrounding both /n/s. But conventionally (and wisely I think) a type design will get bolder by both increasing the black and decreasing the white. (Without so doing, you get overexaggerated changes in width between weights, as you have here.)
I would add space in and around the letters of your extralight master, and take away space in and around the letters of your extrabold master. 

Peter Baker

Heavens! I knew that, but it wasn't in my head when I needed it! I'm not sure this gets me all the way to where I want to go. But closer?

Craig Eliason

I have to chuckle because now what becomes apparent is the reasoning behind that original cramped lead-in of the /n/ that you disliked in the original. Your more generous instroke pushes off the e and thus opens too much space in the interletter space below. (I really like these moments of coming to understand the parts of classic designs that at first seemed puzzling.)
Keep working on even spacing (your top line reads like en en en en en and bottom reads like e ne ne ne ne ne ne), and where the shapes hinder even spacing they may need to be altered.

Peter Baker

Hmm. You're right that the spacing is now seriously off. But I think it's simply (or mostly) that my sidebearings are bad. When you measure the amount of horizontal space taken by the incoming and outgoing strokes of the original n, they are the same:

It may be that the designer of this Double Pica italic wanted the heavier incoming stroke to compensate for the light vertical stroke that it leads into, but others around this time (e.g. the designer of the Fell English) are producing a structurally similar n with the incoming and outgoing strokes looking much more balanced—a different stylistic choice. (BTW, the u in the Fell Double Pica appears to be nothing other than the n turned upside down. I've been scratching my head over that for quite a while.)

Craig Eliason

Maybe you're right, and it's not so much smaller space the punchcutter was after as a stroke that would point more downward towards the preceding outstroke. 

As for the n from the u... it's a pain in the ass to cut a punch!

André G. Isaak

I don't know how determined you are to be faithful to the original, but have you thought about lengthening the outgoing stroke on the e? To me, it seems overly short which adds whitespace where it isn't wanted.

Peter Baker

Craig Eliason said:

As for the n from the u... it's a pain in the ass to cut a punch!

I love this explanation. Honestly, it's the only one that makes any sense!

Peter Baker

André G. Isaak said:

I don't know how determined you are to be faithful to the original, but have you thought about lengthening the outgoing stroke on the e? To me, it seems overly short which adds whitespace where it isn't wanted.

Like this?

(The outgoing stroke now projects a tiny bit beyond a guideline drawn at the italic angle and tangent to the rightmost curve of the e.)

As to faithfulness to the original, I think of this model as a convenient starting point, not a thing to be reproduced. In fact, one nice thing about working with an old but non-classic face like this one is that one has a good bit of freedom. I'm pretty confident that no one is ever going to complain that I took too many liberties with this typeface.

André G. Isaak

You might even go longer than that, but I’d want to see it next to letters without an incoming stroke. Can you provide a 'eneenn' sample?

Peter Baker

Always eager to please . . .

And huge thanks to both André and Craig for looking at this!

Claudio Piccinini

(In general). I wouldn’t do a "display extra light". These typefaces are made for, (albeit minimal in text or in your case large sizes), modulation.

Peter Baker

Display ExtraLight is an odd duck. I'm showing it here because it's at a crucial corner of the design space, and flaws there are going to show up everywhere. And it will be theoretically available in the variable font. But when I generate desktop instances I don't bother with it.

Thomas Phinney

The thins seem a bit heavy for “display”—I see more more modulation in the original typeface, at 24 pt (double pica). If you are doing an optical size axis, I would expect your display master to be perhaps optimized for an even larger size, with more, rather than less, contrast.

(Of course, you are doing a reinterpretation, so there is room to change things.)