Too much overshoot?

Ori Ben-Dor

While in 2017 1 & 7 look shorter than 2 & 0, I'm not so sure about in 010101 and 272727. What do you say? Should I overshoot a little less? (It's a display type, so the size here is within the designated range).

AbrahamLee

I think it looks fine. If you look at it long enough, then the eyes begin to distinguish the difference in height, but as a quick read-through (like most headers are), the difference is so small that the eye won't notice it. You could always try moving the top/bottom points inwards to see if that helps. General notion is 3%-8% overshoot and it looks like you're pretty close here.

John Hudson

It's pretty good. Looking at the 010101 sample, I think you might reduce the bottom overshoot of the zero just very slightly, as the 1 looks like it is floating upwards a bit.

Ori Ben-Dor

Interesting, I thought the top overshoot was more problematic. I'll play with both. Thank you both!

PabloImpallari

Do you have blue-zones defined and/or hintting already?
Is the sample rendered in mac or pc, photoshop, illustrator or web browsers?
All those factors affect how the overshoots are displayed, so you may want to test many (or ideally all) of them.

Ori Ben-Dor

None of those things, yet. It's just the raw glyphs.
The sample was rendered directly in the preview panel of FL5 for Windows.

PabloImpallari

BTW, in "2017" you zero looks a bit narrow to me, maybe because all the other glyphs have bigger white areas.

Ori Ben-Dor

Thanks, I'll play with the width too.

Georg Seifert

We had a similar discussion in another thread. And there we found that the size you are looking at changes the perception of the overshot quite a bit. 

Chris Lozos

The counter size of the zero is much affected by the sidebearings. A tighter fit would make the zero look less objectionable but then you compound the problem with the seven and the one.

Alex Kaczun

Visual perception is subjective. Match which square to round overshoot relationship looks correct to you within each row. Interesting test. Shape, weight and counters—all effect results.

AbrahamLee

That's a great exercise, @Alex Kaczun! A really helpful example to consider when considering how important things like this are for optically sized designs. I'm curious what everyone would say is the most optically-correct circle (1-10 from the left) in each row. Is there a way to do a poll in this forum?

Part of what I have learned from this is that it very much depends on how large they are being viewed at (or perhaps how far away I am from the image). For me, I'd say that 1 & 2 are too small at any size, 3 & 4 are marginal (perhaps would be okay at display sizes), 5 & 6 look great at a normal viewing distance, but 7-10 work progressively well at smaller sizes (or farther viewing distances) due to the higher frequencies of the overshoot not passing through to the eye.

Anyone else willing to share their thoughts on this experiment?

Alex Kaczun

Right you are Abraham—viewing size is key. By the way, you can double click on my image and it will expand to a separate window so that you can examine at larger point size. This is an extremely controlled test—ALL the squares are exactly the same height & width throughout, the first circle (starting from the left) is exactly the same height as the square. Each circle increases by 1% proportionately in height, from left to right. Each square and circle has the exact same stem weights—top and bottom. I could of included a condensed version or other design factors, but I wanted to keep the test simple. You can explore this sort of thing further while designing your fonts.

Of course, there are many other factors to consider as well in overall alignments. But this is a good start. It all really is somewhat subjective. Everyone will see it slightly differently. Every time I look at it with a fresh set of eyes—I go back and forth between 6-8. So, I guess, the sweet spot is somewhere at around 6-8% increase in overshoot for the circles. Remember, it's ok to increase the overshoot further if a glyph is very pointed at the top. Again, it's all visual. There are no hard and fast rules to any of this. Rely on your eyes—gets easier as time goes by.

Take a look at the Univers (Linotype) font family series as a guild. Adrian Frutiger is a master at alignments. As an example, his alignments for caps H and E are slightly different. Cap E slightly less in height than H. That sort of thing. Even the stems and hairlines vary top to bottom, and so forth. Slight alignment variations throughout—very interesting study, indeed. I hope this helps. Understand all the different methods and reasoning behind all of this, but always rely on your instincts and your own eyes when designing your fonts. If it looks right—go with it!

Sander Pedersen

I don't think the height is actually different in any adaptation of Univers. Frutiger writes in ‹The Complete Works› (p. 96) that it was a theory (illustrated in ‹Type Sign Symbol› and the issue devoted to Univers in Typographische Monatsblätter), but would have complicated the production, so it wasn't carried into practice.

Alex Kaczun

You are correct Sander. In the digital version's many of these subtle alignments and discrepancies were removed. Sorry, I should have been clearer on this. Thank you for pointing that out.

I worked at Mergenthaler (aka Linotype & Linotype-Hell) from 1976 to 1993, and as the principal type designer I had access to all the original Mergenthaler letter folio drawings and archives. I was involved in the digital transience from photo to digital and did a detailed study of the entire Univers family of fonts—letter folio archives based on Adrian Frutiger's hand and the approved letter drawings. Frutiger's theory was sound, it was just felt that all these variables would complicate digital PostScript production. Everything at this point became uniformity—to ensure better hinting, especially on low-resolution devices. I think today we can raise the bar again and consider all such theories and practices.

My only point was, to all type designers today, is to think about all this type of stuff and rely on their own instincts, experiment with other possibilities, and above all—visually rely on your own eyes. If it looks right—go with it!

Sander Pedersen

Right, I was just thinking that it's difficult to study something that you're hard pressed to find in praxis. I'm sorry for going off-topic.

Alex Kaczun

You are not going off-topic. I appreciate your input. I wish more people would just chime in more about this topic. It's all very interesting.

Simple visual test—which line (top or bottom) looks better to you—and why?

Sander Pedersen

It's subtle, and the viewing conditions are suboptimal, but I think the bottom line looks better; the overshoots seem more balanced with the baseline and cap height. The O looks perhaps a bit small in the top line, in comparison.

Ori Ben-Dor

Bottom. In the top line the O's could use some/more overshoot.

Alex Kaczun

You are both absolutely right. The lower line—Cap E & Z height was reduced -2/1000 shorter than the Cap H (hats off to Frutiger). And the Cap O overshoot was increased to that 7% overshoot as per test above. That confirms everything that we've talked about. Thanks.