four, the outlier?

Craig Eliason

In figuring the thickness of the horizontal and vertical strokes of a four, are there any rules of thumb that could be helpful? Do/should/can they relate in any predictable way to the horizontals of 2/5/7, the straight stems of the uppercase or lowercase, each other, or any other stroke in the font? It seems like among the common glyphs these strokes uniquely have no clear analogues.

Ray Larabie

Not at all. Except in light weights, nothing in a 4 ends up being the same width as anything else. That's way I make the 4 last, then I test with a line line like 041424354647484940 and adjust until it feels right.

PabloImpallari

Simplepolator* is perfect for this kind of things.

1) Draw a glyphs having bolder stroke in one cell, and draw another glyph having a lighter stroke in another cell.
2) Select the 2 glyphs and use Simplepolator to generate the intermediate instances.
3) In the preview windows, type something like:
0@1@2@35@6@7@8@9@0
4) Use the arrow keys to move over the generated glyphs to see witch one works best.

*http://www.impallari.com/projects/overview/simplepolator

Gunnlaugur SE Briem's "How to make mistakes"
http://66.147.242.192/~operinan/2/2.3.3a/2.3.3.02.tests.htm

Craig Eliason

Thanks Pablo, I'm out of FontLab and into Glyphs these days--which makes it even easier, as Glyphs has built in a filter where I can set up an extreme in the main editing space, the other extreme in the background (mask), and use a sliding handle to interpolate on the fly between them--even more useful than simplepolator.
But one issue with the pesky 4 is that I'm simultaneously evaluating the right thickness for the horizontal and the vertical stroke, so rather than a string of possibilities ideally I'd have a two-dimensional matrix.
But anyway, tools aside, I appreciate the message: arrange your options in front of you and then trust your eye to figure it out.

James Puckett

But one issue with the pesky 4 is that I'm simultaneously evaluating the right thickness for the horizontal and the vertical stroke, so rather than a string of possibilities ideally I'd have a two-dimensional matrix.

I wonder if we can convince Georg to add anisotropic interpolation to that menu…

Rainer Erich Scheichelbauer

anisotropic interpolation

… gives you crappy results on diagonals. It cannot differentiate between upstroke and downstroke.

David Sudweeks

I've noticed I'm pretty arbitrary about stroke weights when it comes to four, particularly in heavier weights. I prioritize counter size, overall character width and color over any stroke-specific metrics.

Deleted Account

Here is a brief discussion supporting what's been said before. It is a clip from a non-upcoming retrospective full-length documentary film on the topic of figures in type.

Craig Eliason

★★☆☆ The pace is glacial and the screenplay is flimsy, but in his breakthrough rôle Frere-Jones already possesses the screen presence that will catapult him to superstardom.

Craig Eliason

On a somewhat related note, does anybody reduce stem width on stems with extenders? I notice in black weights that the extra length and thus extra surface area can look too heavy relative to stems that stay within the x-height (comparing d or q to a one-story a, for example).

But then I wonder if I'd be asking for trouble in even rasterization if I make stems mathematically inconsistent.

Deleted Account

"On a somewhat related note..."

Matthew C. has pictures of this too;) I was driving him and the two Zapfs among others, across southern Greece to Delphi, and as we all settled into the van I'd rented, Herr Zapf was just explaining to one of the lucky attendees who'd joined our van after the conference organizers tour bus we'd all signed up for was canceled, that "You always make the ascenders and descender a little lighter...". So I asked him later and this is a function of weight, as you say.

Also this dimension has a careful place between the cap and l.c. stem dependent on the class of the design.

I think the lowercase l, which has no interruption and must blend quietly with them all, is a good control character to add in extreme weight variation.

Alex Kaczun

"Also this dimension has a careful place between the cap and l.c. stem dependent on the class of the design."
I totally agree, David. And who could argue with Hermann Zapf. Thanks for sharing the story.

I usually tend to open up (space out) the lowercase "i", "j" and "l", proportionately based on how tightly fit the overall font is. You want these individual strokes to stand out visually from the rhythm and other vertical stems. Let these breath a little more so that the eye can pick them up more easily. Also, a little extra weight (proportionately more the heavier you go) to also further help these single strokes hold up visually. And, yes, I usually reduce the stem weights in busy descenders and ascenders glyphs, especially as pointed out by Zapf... "depending on the class of the design."

And Craig, I usually do not worry much as far as screen resolution goes. I design all my fonts based on visual cues. Always. Screen resolutions will always continue to change and improve.

Thomas Phinney

Careful choice of "standard stems" values in OTF can help maintain snap to a more limited number of values at lower resolutions. In TTF manual hinting might be necessary to get the same effect.

John Hudson

"You always make the ascenders and descender a little lighter..."

On the other hand, you make the longer verticals of the caps a little heavier. And long diagonals are more likely to need a bit more weight than a bit less. So there isn't a neat algorithmic relationship between stem length and stem weight, but a number of optical effects to manipulate.

Alex's point about the more open spacing of the narrow vertical letters is an excellent one: something that I'm only just beginning to appreciate after 20 odd years at this.

Deleted Account

"I usually tend to open up (space out) the lowercase "i", "j" and "l"..."

As we were taught. But Linotypocentric teaching of those three lowercase were, I think, for the needs of "12 pointing";)

The thinking that drove that teaching was limited by the task at hand, which rarely involved fonts under "stylistic stress". Most of the time we were designing 12 point masters of text faces. Fonts under stylistic stress are extra-compressed, extra-wide, Ultra, Hairline or masters for 1,000 or 6 pt. e.g.

So I didn't fully implement any typeface families under real stylistic stress until I had drawn some of the whopper families at FB and the RE versions of those families. Knowing that the thinking on RE is, "What would masters for very small sizes look like if the paper were free?", and thus needing to gobble up relatively massive amounts of horizontal space, I had to have a more complete idea of how i or l's needs meet m or w's across a spectrum of design classes for tiny type.

So, I ended up thinking i and l are a 1, gave 1.25 to jfrt, 1.5 to svxyz, 2 to abcdeghknopqu and 3 to mw. This helped.

I didn't class the figures yet, but the 4 seems like a 1.6667

Nick Shinn

When I first saw the title of this thread, I immediately thought of the descender in the lining figures of a certain style of 19th/20th century didone italic.

I dutifully replicated it in my Scotch Modern revival, but have absolutely no idea why it existed and persisted; sure, four can be difficult to design, but this solution does seem a bit eccentric.