John,

As you may have suspected. I am participating in this forum "only" for my interest in optical scaling. You, I am sure, are aware of that. A pissing contest with Hrant does not interest me. I am sure he can be most helpful.

Hearing of Herr Lege's work is very exciting. I will see if I can trace him down, or do you have any samples? In any event I wish to know more.

Gerald Giampa

I was with John, when Lege gived the custom Palatino, and I'm still septical.

Its somewhat a bit better than the actual Palatino, but when you know the ITC Bodoni story, the Palatino story is rather minor and quite similar than the non existant Minion MM who sound a first optical attempt by Adobe?

To what Sumner tell it to me, circa 1995, the 12pt was fisrt done from original Bodoni, then they done an interpolation between the 6 and 72, who have show very good result, very close to what Bodoni have done with its 12pt cut.
So they kept the interpolation. In ITC Bodoni, what is fabulous, is the incredible difference between the narrow delicate 72, and the 6 large "Clarendon style effect" Bodoni 6.

Interpolation is good for such task, but only if the two masters are really good in term of design. FB fonts generally propose good variations in this area, despite you can't completely agree with the style of the typefaces to which they applied the Optical scaling.

Jean,

Do you know of a good place to have a look at any of these? I live in a country with a language I do not understand either spoken or written and I am without my books and references. If you know anywhere to look on the web I would be most thankful.



Gerald Giampa

Go to www.myfonts.com
select ITC Bodoni 6, 12 and 72, and do the test drive at higher size?

For a term, I think "optical size compensation" is good, but it's too cumbersome, and "optical scaling" seems to strike a nice compromise.

> optical scaling means different fonts

Not necessarily. One could indeed argue that some features in some fonts cannot be adequately interpolated merely through Cartesian transformation*, but that's about it. When you consider the elements needed to deviate the smaller sizes from the "deliberative ideal", you realize that they can be interpolated just fine (although certainly not linearly wrt point size). Furthermore, it's a matter of quality, not absolutes. For example, you could say that the optical scaling in the Adobe fonts is weak, but it's still there.

* See my Founders Caslon elaboration (June 20) here:
http://www.typophile.com/forums/messages/30/12111.html

The Lege project seems interesting. However, interpolating [linearly] between 6 and 9 will not work very well, because 6 is below the immersive reading threshold and 9 is above. If you look at the 6 versus 8 versus 10 point of ATF Garamond for example, you'll see that in the 6 the descenders get abruptly much shorter. My theory is that they wanted to give more room to the caps, since the very small sizes are more often used for all-caps.

--

In my case, I'd love to formulate my own rules from scratch as to what needs to happen when sizes get small, but of course one can save a lot of effort by looking at existing/past practice. The ITC Bodoni stuff is nice, but in looking carefully at what different font houses have done from the very beginning, ATF stands out. There are many clues to this (such as their determination that 40 point -an odd size- is the upper limit for compensation), but their mastery is most obvious in ATF Garamond, as compared even to the glorious Imprimerie Nationale Garamont. My guess is they became experts motivated by their engineering background, but always using human vision (not geometry) as the guide.

To this effect, I'm working on printing a waterfall of 10 sizes of ATF Garamond, and will be doing some numerical analysis on the results.

hhp

Hrant said...

"because 6 is below the immersive reading threshold..."

I'm with you there.

"and 9 is above."

but not there. my observation has been that,
depending on the type design in question, the
immersive threshold ranges from 8 to 10. of course,
we must also take age and vision degradation into
consideration, which would lead to larger sizes yet for
some audiences (possibly 11-14, at the extreme end).

My wording was lousy.
I meant that 6 and 9 were on different sides of the lower limit of immersive reading. The range itself I would say is about 8 to 12 (depending on not just the reader, but also the size-on-body of the given font).

BTW, here's a very rough initial preview of what I'm doing with ATF Garamond:



The sizes are: 48, 36, 30, 24, 18, 14, 12, 10, 8 and 6.

hhp

Just to be clear: all those "a"s were scaled to the same body size.

BTW, you should see what happens to the "g"... :-)

hhp

A scale is not a spectrum.

Interpolation creates any point on a spectrum (between big and small, thin and heavy, etc.) but a scale is a set of instances that work harmoniously - with each other, and also with other fonts, and also in relationship to layout proportions. Being non-scalar (incemental) for weight/condensing, Multiple Master typefaces lacked personality, a fault recognized by the addition, later in the game, of extra "master instances" between the ends of the spectrum.

Printing used to be highly conventional (materials and layout), and much of the aesthetic effect of the page was taken care of "in the font". But the situation has long since changed, there is no longer a dominant harmonic structure which type designs must conform to.

(Ironically, the scale of sizes in WP apps has put the masses in better touch with the principle of a harmony of type sizes in a page layout, than the majority of professional graphic designers, who get to create their own "atonal cacophony" out of peculiar size choices.)

Now size-specific harmony, aesthetic perfection, if you will, is not crucial in the font, because there are hundreds to choose from, and all kinds of environments to use them in. And 8.75 pt, horizontal scaling of 95%, tracked to minus 3, works quite well for newspaper text.

A number of foundries have realized that "size-specific" or "optical-size" is a rather one-dimensional term. For instance, HTF (and myself) have used the term "Fine" for what is ostensibly a large-size display font. Such would be suitable for newsprint headlines, but typographers can, and do, use these fonts for text work, when they have #1 coated stock.

Poynter Oldstyle is available in different "grades" or "tunings" for different amounts of press gain.

So I'm with Hrant, "human vision, not geometry". Automations like MM, optical scaling, and optical kerning, when used across the board as generic modifications of many typefaces, will always be sloppy approximations to the craft that individual type designers use to create remarkable, specific instances of a typeface.

Used with discretion, automations are marvelous tools for the type designer creating special typeface families with a limited set or scale of optical sizes that relate to the typeface design (you can deduce that from Hrant's example, where apparently the "g" varies differently than the a!), but beyond that they are nothing more than glorified versions of the faux-bold button.

"I meant that 6 and 9 were on different sides of the
lower limit of immersive reading."

okay, I'm with you again. I would also say that your
general range (8-12) is more realistic than mine.

it's interesting to see, in your samples, how the
aperture becomes relatively closed at the small end of
the scale. I wouldn't think this would be helpful at such
sizes.

in the course of your investigations, you might find
FyrisFonts' work interesting. they've created a minimal
set of optically scaled Jannon revival fonts.

This article on the Bentons is of relevance to your discussion: http://www.printinghistory.org/htm/journal/articles/31-32-Cost-Benton.pdf
Pages 14-16 of the PDF discuss how ATF adjusted the pantograph to cut different sizes; there is also mention of Stemple's practices.

> the "g" varies differently than the a!

Well, I didn't actually mean to imply that. My guess is that the whole falls into some logical method (even though the eye is indeed the gauge).

The dimension in the "g" missing from the "a" is the descender. Some of the most interesting things in optical scaling happen in the vertical proportions, and as you can see even the "a" falls into that. And the thing with compensations of proportion is that you can't fake it in the software...

> the aperture becomes relatively closed at the small end of the scale. I wouldn't think this would be helpful at such sizes.

I was wondering about that too, but then I realize that aperture is only one of the factors that has to be considered. Some others are: maintaining the character of the design at small sizes; increasing the weight; limiting the increase in x-height. If you look at the 6 point "a" again and try to picture just those four factors pushing against each other, you might agree that they did it pretty well.

BTW, thanks for that FyrisFonts link! I'd never even heard of them before - looks very interesting.

--

William, that's a great article. Note also that APHA's most recent issue is entirely devoted to ATF. I have a copy, but I'm almost too giddy to actually read it!

hhp

"If you look at the 6 point "a" again and try to picture
just those four factors pushing against each other, you
might agree that they did it pretty well."

I think it is a fantastic optical scale, no question there.
and I agree that the character is well maintained
throughout. I just wonder if there might be a
readability issue as a result of the reduced aperture.
however, the 6 point variation is the most drastic and
heavily affected in this regard and, as you implied
earlier, a 6 point font is intended more for legibility.

this scaling stuff is fuzzy, ain't it!?

Checking the APHA web site, I was touched to see that my late Uncle, J. Ben Lieberman was founder of APHA and that his friends endowed an annual memorial lecture in his name. His influence is what made has me a type nut,and to want to design the books I write, though I'm not in the design business.

Yes, fuzzy it is! For example, some people would say that ATF went too far in the 6 point, and to a large extent it's a subjective issue. But not totally, because of the definite nature of human vision/reading.

BTW, I should mention that the sample I showed had too much impression pressure (by about 2 or 3 mils). Next week I'll do a better job of fighting gain, and the aperture will certainly increase.

hhp

Hrant

Just a note on the procedure. It will lessen the difficulty of "fighting gain," if you were to repro these in relative size groups. The reproduction press is going to ink smaller sizes heavier if it is "set" (impression, inking) to print the larger sizes correctly. If it is "set" to print the smaller sizes, it will underink the larger. Separate printings are really the only way around this on a reproduction press where makeready won't help much (and which could actually give you a false reading). Seriously.

Gerald

John Hudson, Hrant Papazian, William Berkson, Jean F Porchez, plain*clothes, Nick Shinn, and especially Tiffany.

Thank you. I think most of you know that optical scaling is the only reason why I had joined the forum. I am saddened that I will not be able to participate.

Hrant, I want you and Gerald to know that I have and never did have any ill will towards either of you.

I want that to be clear so there is no misunderstanding as to why I do not expect you will be hearing from me.

Good luck to you all.

Gerald Giampa

> .... Separate printings are really the only way around this ....

That dimension of the "physicality" of letterpress is still way beyond me... :-(

A couple of questions:
1) What about printing sideways? I remember you saying that the cylinder direction matters.
2) Is the change in settings from smaller to larger type a smooth progression? If it is, maybe that means I can just leave the whole form in there, and increase pressure and ink gradually, each run tailored to each size, and for each size scan up the run where it's optimal? This, to avoid changing the bed contents each time. Did I explain that OK?

hhp

Sideways is not best on the cylinder repro. Bit too much to get into right here and now.

Yeah, you explained it OK. You are either quite brilliant or one lazy SOB. You will need to mark your sheets quite carefully to avoid later confusion.

Gerald

Well, it's not necessity, it's actually laziness that's the mother of invention!

hhp

So I was looking through that FyrisFonts site - it's quite amazing.
Very broad and deep content, or at least it seems so - I don't read Swedish. But I tell you, I don't remember the last time I wanted so badly to be Swedish!

--

It looks like Lundhem takes optical scaling much more seriously than most, with some intelligent and novel analysis. Particularly interesting is the graph towards the bottom of:
http://www.fyrisfonts.com/artiklar/jenson/default.asp
It indicates that Adobe Jenson uses a "bi-linear" model, and it's notable that InDesign's "optical spacing" follows this model exactly as well.

But then in the Minion section, the scaling seems to be much more refined. Very interesting.

--

On the other hand, as for the work on Garajannon-MM (which seems not to be for sale except as distinct instances), if in fact he's made a linear interpolation between the 8 and 48 point, that's not optimal - although the effort itself is highly commendable.

hhp

Hrant

The physicality of letterpress is somewhat meaningless in terms of the end result is it not? Your idea of certain sizing ranges for optimization is not any different and, probably, based on the same notion. The physicality of letterpress is essentially, a mechanization of letterform. This is what gets in the way. Same with photofilm and digital. The abitrator is the eye. That is all you can trust. I hate to say it, but a course in calligraphy is heading your way.

Gerald

What I meant though was that there seem to be different dimensions to the physicality. I have a tentative handle on the relevance of some of them (like type height, and impression pressure), but I'm still in the dark concerning others (like ink behavior - especially in the shoulders, the direction of the cylinder, the dynamics of the paper, and others). And of course these all relate, which makes the whole largely mysterious (to me).

> a course in calligraphy is heading your way.

Totally. Know your enemy.
And after that, a course in cave painting.

The most useful thing about studying "history" is to learn form the mistakes.
Otherwise it's just cloying nostalgia.

hhp

Hrant said...

"if in fact he's made a linear interpolation between the
8 and 48 point, that's not optimal - although the effort
itself is highly commendable."

unless I'm not understanding you, it sounds like you're
dodging the issue here. is the outcome a good scale or
not? no matter what their process was, it's the final
letterforms that matter.

unfortunately (for me), my insufficient knowledge
disallows me from judging the designs; but, the
differences throughout the scale seem to address
important issues. I really don't know what *optimal*
means here... but I'd like to.

since you've been browsing the site, you may have
already seen this, but the variations of their Spartacus
are quite interesting as well. I'm especially interested in
the possibility of using the display design (far left on
the graph, sharp endings) in concert with the normal
(far right on the graph, softened endings) for text.

I think that the bottom end of the Spartacus scale
might be a bit malformed in that the x-height seems
somewhat obese, in the spirit of ITC circa 1970
perhaps. since these sizes would be used for extended
text, I would like to see more adventurous extenders.
there seem to be other proportional issues with these
sizes as well, but I'm at a loss to define them. is this
some bias of mine from looking at typefaces too large
for too long, or am I on the right track? I certainly don't
know.

> is the outcome a good scale or not?

Yes, I agree that's the real question. But I don't have the font! I'd need to look closely at all the PDFs to be sure.

Like I said before, optical scaling comes in varying qualities - Garajannon certainly has optical scaling, and it seems much better than average too. Maybe "optimal" is confusing.

Spartacus is indeed interesting. And I think its vertical proportions are actually right on*: the x-height has to start getting "obese" at about 8 point. Look at the ATF Garamond stuff, and consider how little room is left for the ascenders in the smaller sizes.

Your "bias" there is probably an issue of deliberation versus immersion: you can't judge the merits of a text face by reading it large. You can only really do it to the extent that you understand readability. I don't mean to say that you don't understand readability, just that it's challenging to always work with this paradox all the time.

* Although the descenders seem slightly too long to be text-centric.

hhp

"The most useful thing about studying "history" is to learn form the mistakes. Otherwise it's just cloying nostalgia."

Might be the wrong path. I'd be looking at the techniques for the why, and how they contributed to later technologies.

I took a workshop in cave painting once a while back. Spit painting. Where you expell the colored "inks" from your mouth and guide and manipulate the airborne particles with your hands. Kind of like crude ink jet. But a bit more intuitive. Our inks were all organic and politically correct but some of the authentic inks used by the cave painters (at least the spit painters) were apparently quite dangerous.

Gerald

Hrant said...

"the x-height has to start getting "obese" at about 8
point. Look at the ATF Garamond stuff"

certainly there is a point (no pun intended) where the
body needs more room to breath than in larger sizes.
but how steep is the adjustment curve at the low end,
and where does it begin to flatten out? uhhhh, yeah...
I'm just not sure about that right now... (for all you
Office Space fans).

"Your "bias" there is probably an issue of deliberation
versus immersion: you can't judge the merits of a text
face by reading it large. You can only really do it to the
extent that you understand readability."

absolutely. it's a sloppy science trying to judge a bunch
of factors on a conscious level that are only fully
understood (at this point in history) in the
subconscious. reading studies provide guestimates, but
the contributing factors are so multifarious as to largely
discredit anything we can gather.

so far, it appears that modern scales are developed
from an automated system based on two subjective*
axes. since the only *solid* info we have to work with
at this point is theoretical, I'm not so much bothered by
that road block. I do think we might need to consider
added axes, however.

*I realize that there are a lot of supposed facts behind
these assumptions, but nothing is concrete yet.

what axes are we ignoring that might lead us to more
comprehensive scales? I imagine that the designers of
the past had intuitive systems (and physical
constraints) that lead to multi-faceted systems of
alteration. I think Hrant's (and others') investigations
into the construction of punches are the best way to
reveal these characteristics -- now we just need to
distill the info into usable guidelines. obviously, that's
why we're all here.

> how steep is the adjustment curve at the low end

Who knows...

The only solid thing I myself would state is that there are three ranges: immersive reading (generally about 8 to 12, but usually more like 9 to 11), below it, and above it. Below immersion you're basically making reading as painless as possible, not expecting high reading speeds, instead making sure there are no ambiguities. Above immersion you're appealing to deliberative, conscious aesthetic observation. But this doesn't mean aesthetics are irrelevant during immersion - who knows what the subconscious likes? It's only sure that the subconscious is a lot less worried about liking things, and more worried about getting the job done.

Here's a really rough graph I once did:
http://www.themicrofoundry.com/other/optic_01.gif

> the only *solid* info we have to work
> with at this point is theoretical

Kinda. But the theory is derived in large part from empirical studies (from disparate fields). So you could say it's quasi-empirical. But where any empirical data at all is almost totally lacking is in fact in the numbers.

hhp

"It's only sure that the subconscious is a lot less
worried about liking things, and more worried about
getting the job done."

a fair guess, but we can't really say can we?

"Here's a really rough graph I once did:
http://www.themicrofoundry.com/other/optic_01.gif"

that's actually a nice visual reference -- I admire your
bravery in posting such rough thoughts! ;)

"where any empirical data at all is almost totally lacking
is in fact in the numbers."

I'm not sure I follow you. care to explain?

What I meant is that the empirical data points to the existence of three zones (or actually, two thresholds), but it doesn't say how steep the changes are on either side (with any usable accuracy).

But even what we know now is very usable. For example, we could say that the bi-linear model (as shown on the FyrisFonts site concerning Adobe Jenson-MM scaling, and as used by InDesign for spacing) which has one threshold (at 12 point) isn't good enough. And a tri-linear model might be the ideal compromise at this time, since we know there are three segments, but we don't really know their precise shapes.

--

BTW, could somebody please translate the "Viktig optisk brytpunkt" section on the Jenson-MM page, and the "Minons optiska 'masters'" section on the Minion-MM page**? That would be really great.

* http://www.fyrisfonts.com/artiklar/jenson/default.asp

** http://www.fyrisfonts.com/artiklar/minion/default.asp

hhp

"What I meant is [...] it doesn't say how steep the
changes are on either side (with any usable accuracy)."

okay, that's pretty much what my convoluted post was
getting at.

So where are our nice translating Swedes?! :-)

hhp

Hrant,

I am sure there will be lots of opinions as to what should be done in the way of optical scaling. I suspect each designer would like to have some control over his or her font product!

Do you have any ideas as to how the software should handle it? OpenType, page layout, font development software?

Point sizes for each font? That sort of opens up many complaints. I found one with Fletcher Challenge project. Tiffany had objections of her own.

Easily, I would think, there is something more elegant. Are you aware of what is, or is not going on in this area in the OpenType forum? I was led to believe there were objections to including optical scaling solutions other than "the awkward".

Frankly I think it should be somewhat like the kerning function. Either turn it on or turn it off in the page layout software. Changing fonts, I don't like that idea.

Do you think we should ask John Hudson?



Gerald Giampa

>Do you have any ideas as to how the software should handle it? OpenType, page layout...

John Hudson mentioned in another thread that there is the capacity in Open Type, not yet used, to tell the page layout program (probably InDesign) automaticaly to use different variants of the typeface for different sizes. I suppose this could be turned on and off also. This does sound like a very good idea to me.

> I am sure there will be lots of opinions as to what should be done in the way of optical scaling.

So what's your opinion?
What are some instances of superb optical compensation, and why?

> I suspect each designer would like to have some control over his or her font product!

Designer control is great. It doesn't preclude designers from approaching the problem from an analytical angle, not just one of preconceptions and personal preference. There are different ideologies out there, and each can have validity in certain contexts, but there are also the realities of human perception and cognition.

One central tricky question is this:
What's a good balance of character versus legibility at the smallest sizes?

For example, does the ATF Garamond go too far? Is the Imprimerie Nationale's version* better because it preserves more of the Garamond character (albeit at the cost of reducing legibility)? How much character can be conveyed at 6 point? And what about the greater proportion of all-caps setting at the smallest sizes (I think the reason for the sudden shrinking in the descenders of the ATF)?

* Thanks to Kent for helping me make these:
http://www.themicrofoundry.com/other/garas_12.gif
http://www.themicrofoundry.com/other/garas_10.gif
http://www.themicrofoundry.com/other/garas_06.gif

--

As for the mechanics of using optically scaled type, I think generally a continuous axis -like in MM- is just fine, although MM is no longer well supported. So the application would generate the right instance based on the point size. And if you have a series of sizes instead of an axis, it would be great to have application support for choosing the closest ideal size automatically. Of course, this is the type of stuff MM should have done from day one, but they never implemented any automation at all... probably because the average user (the target of Adobe, but hopefully not everybody) went "huh?"...

--

I can even see software generating optical scaling automatically. Although it can never be as good as the man-made stuff, if the software could make the font wider, looser and darker for the smaller sizes, that would already help a lot. And if it could fiddle with the vertical proportions too, you'd be 90% there.

As for turning it off, it should be like an airbag: it should be on by default, and not trivial to turn off. People should only turn it off if they really think they really know what they're doing.

hhp

Hrant,

"What are some instances of superb optical compensation, and why?" I am not sure I have arrived at any conclusions as what is superb. What is not superb is what we have now, nothing. I suppose we could build something that is worse than nothing but let's not do that.

But when I said that the designer may wish for input you have found the right illustrations, which are very good. ATF seems to have moved strongly in favour of legibility. (Also from an analytical angle) But I am not to sure if they have gone too far? My own leanings favour Imprimerie Nationale's. However, I admit Lanston went the way of ATF. Both enhance legibility, the Imprimerie Nationale's preserves more characteristics of the face.

My prediction is that there will always be a designer who favours preserving characteristics of the design but seeking "legibility gain". And there will be the strongly "functional personalities" which will lean towards "maximizing legibility" at any cost.

Maybe an on and off function? Oooh, that sounds like three stations on the dial, an "off", an "artistic", and "maximizing legibility" station. What do you think of calling it "radio scaling"? After all optical scaling is somewhat misleading.

Also I am in agreement that tackling this with an analytical angle is imperative.

Speaking of on and off functions. I am also in agreement with you that by default the best solution should be activated. But you can appreciate sometimes it nice to be able to turn things off such as kerning. That Birds & Tree
http://www.lanstontype.com/SvartSkepp.html
colour ornament would not work as it does if that were not the case.

With the "Birds & Tree" you turn the kerning off, colour the elements in the unit, turn it on and they register. How much do you want for your dot. com? Write me.

But by default, yes keep the kerning on. Make the default the most useful function. That would also apply to "radio scaling". Actually when I think about "radio scaling" it is not a bad word for it. It conjures up "other solutions". "Optical scaling" as a name can indicated the opposite of what we are trying to achieve. After all, optical scaling is what we have now, type as though shrunk down by a lens.

What we want is type "scaled for human optics". So it is receptive optics that we seek, not projective optics.

Optical scaling was illustrated by inherent mechanical weakness.

What I am saying is that Garamond did not use the word "optical scaling". ATF did?

I am thinking out loud. Sorry. Anyway I know names can be silly. I think "radio scaling" is dynamic, may lead to several solutions. Certainly it is better than calling it "optical scaling" which is misleading or, a variation such as "receptive optical scaling" which is not elegant naming.




Gerald Giampa
Lanston Type Company

OK, here goes; translations of the selected pieces from the articles in question by Stefan Lundhem/Fyrisfonts.

I apologize, most of all to Stefan, for any misinterpretations.


---------------------------------------------------------------------------

The Jenson article:

Important optical crossing point

The optical axis gives an even shape adjustment between 12 and 72 points,but from 12 down to 6 points there's a dramatic change. Here the typeface has a crossing point, which makes text in small point sizes really work text-wise.
This is excellent and a step forward compared to the cautious optical adjustment in Minion MM.


The Minion article:

Minions optical "masters"

Minion MM:s optical masters for 6 pt and 72 pt generates point size adjusted fonts when you choose a digit from the optical axis.
This choice won't happen automatically when you choose size, instead a certain category must be created in ATM or a program that handles MultipleMaster-typefaces, like Illustrator.
The crossing point for interpolation is 11 pt. Here Minion MM is identical with old Minion. From 11 to 6 pt the letters get gradually wider and sturdier, from 11 points to 72 pt they get gradually thinner and more narrow.
This interpolation isn't made linear. The smaller point sizes are in a larger proportion based on the 6 point master than they should've been, had it been linearly interpolated.
The optical adjustment works in captions, but be careful in the smaller point sizes.

---------------------------------------------------------------------------

Thank you thank you thank you!

hhp