Analyzing unitized glyph widths

[Jens Kutilek]

This topic was imported from the Typophile platform

Having been intrigued by the nice glyph width chart on the Neue Haas Grotesk page, I could think of no better way to spend half this holiday than to write a Python script that will plot a glyph width histogram.

The results are, not surprising, that different digitizations differ wildly in their glyph unit systems.

Here are some examples:

The Linotype Stempel Garamond is based on a rough 18 unit grid.

URW Garamond No. 8 has something like 3 times 18 units, 54 units.

The Berthold digitization of Stempel Garamond uses a completely incompatible grid, i.e. 48 units.

The same grids can be seen in the attached PDFs of different Univers digitizations.

If anyone is interested, I may clean up the Python script and make it available.

Cheers,
Jens

Attachment	Size
Univers-55_1800.otf_histogram.pdf	3.45 KB

[Jens Kutilek]

Can't get the second PDF to attach ... you can find it here: UniversB_2400_histogram.pdf.

[Indra Kupferschmid]

Cool, I’m intrigued, too. Especially that you found our page anchors :)

[Bendy]

Very nice. I've no idea about the 48-unit system, but seem to remember the 54-unit system indicates the URW Garamond was fitted for photo-typesetting. The 18 of course would match the Monotype unit system. (Was Linotype the same?) Interesting that nobody has re-fitted these typefaces for the digital world.

[oldnick]

Phototypesetting—first by film-strip or grid, later by digitization—on dedicated systems followed metal, and computerized front ends allowed for a finer grid. When the transition was made to the desktop, various foundries probably saw no need to reinvent the wheel.

It's interesting how all of the Garamonds shown have the highest frequency of characters almost dead-center in the spectrum of widths. If one synched this spectrum to the visible spectrum, one might say that Garamond is a decidedly green style.

[kentlew]

Ben — The original hot-metal Linotype did not require unitizing.

But when Mergenthaler developed its first phototypesetters, the stepping motor used for the character advance dictated some sort of unitization, and they adopted an 18-unit system, initially, because Monotype had already established that scheme so well.

Later when a continuous mechanism replaced the stepping motor, the unitizing approach was refined to 54 units (3×18), which could accommodate more nuance, while still being backward compatible with 18-unit designs.

[Edit: Oops, the foregoing contains several errors. See below.]

[hrant]

{To Follow}

[kentlew]

Sorry, my memory failed me. I have details wrong in my previous message there.

I found my correspondence from Matthew Carter on this subject.

I had the mechanisms backward. The original Linotype phototypesetters utilized a continuous-speed writing prism. Horizontal positioning was achieved by timing the shutter opening and closing. For this reason, discrete timings were needed, hence a unitized system.

MC said he thinks that the initial 18-unit system was adopted because of its precedence in the TTS newspaper system (which was probably influenced by Monotype’s 18-unit system, but that’s just speculation on my part). And newspapers were, of course, Linotype’s bread-and-butter clientele.

Later, the introduction of a stepping motor allowed discontinuous movement, which allowed a more refined system — even standing still, and thus zero-width characters, for complex scripts.

[eliason]

As I understand it, the reason the (metal) Linotype system didn't need to be unitized and the Monotype system did is that the latter essentially computed the spaces needed to justify the line, whereas the former used ingenious wedges to divvy up the space physically rather than mathematically.

[Indra Kupferschmid]

This is a complex topic and Nick Sherman and I are currently doing heavy research on it. Yes, Kent is right, a unit-system for Linotype fonts was first introduced with the TTS teletypesetting systems because those had to calculate the line length. A handful of Mergenthaler newspaper faces, spearheaded by the Legibility Group, were adjusted to the 18 units, mostly 15 different width with 4 units being the smallest. So unitized and normal versions of the same typefaces were available. All other Linotype mats did not need to adhere to a unit system but we speculate that this principle might have found its way into the production of normal typefaces, too, maybe for practicability reasons? (Currently trying to find evidence for that.)

Later in phototype, all manufacturers had their own unit system (the idea of incompatibility and dependence on a supplier was more or less the business model). Berthold used 48, the Lumitype 36, Monotype later 96 etc.

[hrant]

Indra, are you aware of the older ATF unitizing schemes? Assuming those are relevant to your effort.

hhp

[eliason]

Hrant, for widths?

[hrant]

Yes. Unlike ATF's height unitization however few fonts got that treatment. The point BTW was to save justification effort.

hhp

[Indra Kupferschmid]

Benton’s “self spacing” type, Hrant? Yea, I know about it.

[kentlew]

All other Linotype mats did not need to adhere to a unit system but we speculate that this principle might have found its way into the production of normal typefaces, too, maybe for practicability reasons? (Currently trying to find evidence for that.)

I’ve never encountered evidence for this (but that doesn’t mean it wasn’t so). What “practicability reasons” do you envision?

I would expect unitizing only for TTS faces, and possibly for any faces directly adapted from Monotype (like Times Roman, for instance, or the unique circumstance with Sabon). During Griffith’s tenure, anyway. Things might have changed under Burke — with more TTS and with nascent phototypesetting constraints.

[hrant]

Yes Indra, that's it (I had forgotten what they called it).

hhp

[kentlew]

Actually, Hrant, I think you’re probably thinking of ATF’s later incarnation of the concept, trademarked as “Quick-Set” (patented in 1918), which was an evolution of Linn Boyd Benton’s original “Self Spacing” scheme (patented in 1883).

Patricia Cost discusses Benton’s original invention in her book, The Bentons, p. 47 ff. (RIT Cary Graphic Arts Press, 2011). She also touches on some of its shortcomings, which the later Quick-Set system purports to overcome.

[hrant]

OK, thanks for fixing my memory. I don't have Cost's book, but I do remember reading her article about the Bentons in the APHA journal.

hhp

[dberlowgone]

"Interesting that nobody has re-fitted these typefaces for the digital world."

I did.

[Jens Kutilek]

"Interesting that nobody has re-fitted these typefaces for the digital world."

I did.

Another thing everybody should be grateful for to you, David :)

I have made a SourceForge project for my Python font tools, with the glyph width analyzer being the only available Python script for now. You can view and download it here: https://sourceforge.net/p/jkfonttools/code/2/tree/trunk/Macros/jkFontToo...

Cheers,
Jens