Under- and overshoot

mauro sacchetto

About under- and overshoot:

1.
for under- and overshoot is it better to have the same values (eg. baseline -16, x-height +16)?

2.
for uppercase and lowercase is it better to have the same undershoot values? Because in some fonts I have found for example that the uppercase is 0 / -16 and the lowercase 0 / -12. Could this result in an apparent imperfect horizontal alignment?

3.
For ascenders isn't a tighter overshoot enough?

4.
In general, is there a way to calculate the under- and overshoot in an absolute sense? I mean: how to decide if the most appropriate value for a given font is 12, or 14, or even 16, etc.?

Thank you!

Florian Pircher

Whatever looks best. 1. Outlines in my current project purposefully descend slightly more than they ascent. 2. Uppercase tends to need more overshoot to optically balance the larger outlines. 3. That depends on the design. 4. No, overshoots are used to counteract an optical phenomenon and are specific to the particular outlines, the font size, the tracking, and other variables.

John Hudson

There’s a tendency in digital type design to regularise alignment, which in practice means that shapes either overshoot or do not and, if they overshoot, all overshoot by the same amount (excepting, sometimes, uppercase letters may overshoot slightly more than lowercase). This is due to alignment rounding on pixel grids, where generally one wants to avoid some overshoots rounding one down and others rounding up at the same size: keeping the overshoots consistent avoids this.

In metal and phototype, one sees more fine tuned optical alignment, notably in some of Adrian Frutiger’s designs in which flat horizontal strokes as in Zz actually sit below the cap/x-height and above the baseline. I usually align the crossbar of f and t below the x-height for the same optical reason.

mauro sacchetto

Thanks for your helpful considerations.
But, from the image I am attaching, is it possible to understand if your expert eye does not find the overshoot of lowercase letters too narrow?

It is 14 pt with respect to the x-height (412), but it seems to me that there is also the problem of balancing with respect to the serif (I'm sorry, I dont't know the appropriate, technical term), whose highest point exceeds the x-height by 7pt.

PS

There is no kerning yet

Jasper de Waard

The ear on g is a bit shy, otherwise I really like it. Overshoots could maybe be a bit bigger, yes, but that is something you have to judge based on a high-quality printout at the intended size.

The placement of ` on è seems a bit off. I'd move it to the left a bit. And the tittles could go up a bit, I think. Okay I'll stop, please excuse my unsolicited advice

mauro sacchetto

No, indeed this is exactly the judgment, albeit impressive, that I was asking. Contribution of the changes, then I will publish a specimen in the appropriate section to have your opinion. Thank you

Cristóbal Henestrosa

Definitely I’d go for slightly bigger overshoots, above and below the x-height. Look at “rincomincia”: the first three letters (“rin”), as they have serifs, settle a precedent about where is the baseline, but then the next two letters (“co”), as they are rounded, looks like they are floating. After that, the next three letters (“min”) sit again in the baseline. And then it happens again in “nci”: the base of the “c” looks a little bit higher to me.

As for above the x-height, it is more difficult to appreciate, but look at “mezzo”: the upper part of the “o” is a bit lower than the “z”.

Probably the effect is more evident in your own image than in my enlargements.

Ori Ben-Dor

Having all glyphs either overshoot by the same amount or not at all also makes sense as a practical compromise. It's much easier to make it into a binary thing rather than start choosing for each glyph the best amount on its own.

Mark Simonson

The size of the curve (e.g., /o vs /period) makes a difference. Small curves usually need less overshoot than large curves. The target point size also makes a difference. Display faces need less overshoot than text faces. The final arbiter is your eye in any case.

Craig Eliason

If "Display faces need less overshoot than text faces" (with which I agree), does that suggest that enlarging the sample to better see the necessary overshoot is actually a problematic method? Should such judgments be made at the intended visual arc degrees?

Mark Simonson

Viewing much larger than the target size is okay as long as you're aware of the effect. You need to temper your judgement accordingly. The final decision should be from looking at a sample in the target size. The more experience you have, the more you get used to the way things like this look (or should look) when viewing zoomed in. Same thing with spacing.

mauro sacchetto

I increased the under- and overshoot from 12 to 18. Now I see whether to differentiate it by lower- and uppercase.
I recheck the alignment on the baseline (the case of | o | and | z | reported to me by @cristóbalhenestrosa.
This is intended to be a text face font, like serifs in general, intended to be printed as well, and not only displayed for example as .pdf.
For accents, @jasperdewaard is right, but the bad accent placement comes from the fact that I switched to the anchor method (it'clear I'm not a font designer, I learn progressively as I do things ...) and postponed the retouching until the base glyph has a definite shape.
One last consideration: Garamonds (Premiere, EB, ...) generally turn out quite small in 12pt print. This is already slightly larger. Does it make sense to increase the size a bit more, with printing as a destination?
Thank you so much for the very useful comments

PS
In a few days I will post a more complete specimen of the Roman in the appropriate section of the site

mauro sacchetto

Thanx to @CristóbalHenestrosa for his precise observations. I increased the under- and overshoot to 18pt, only to realize that it was too much compared to the previous one, which amounted to 12pt. Now I will try to take it to an intermediate size. However, I wonder if the optical disturbance is not also due to an excess of amplitude of the serif, which touches -6pt.

I read: «the first three letters ("rin"), as they have serifs, settle a precedent about where is the baseline». It could be that reducing it from -6pt to -4pt visually captures better that the actual baseline is at 0, and not below 0?

Jasper de Waard

Yes, your serifs are not visually on the baseline like this, but below it.