This is the Kingsley stuff that Gerald Giampa is always hinting at, pardon the expression.

The really interesting part is the optical scaling module, since it probably incorporated the ATF wizardry from the Benton era. But my guess is that once the Kingsley boys realized that you can't get rich doing this and decided to call it quits, they also decided to take the knowledge to their graves, damn everybody else and the large benefit to the craft of type design.

But one of us will eventually reverse-engineer the stuff and pull the carpet from under them, hopefully before they all go six feet under, making them regret their grubbery. Even though I don't think that's a word.

hhp

Yes, it was released. I have disks and documentation for ATF Type Designer version 1.3 and 1.4 stashed away in storage somewhere. It was roughly equivalent to Fontographer 2.x at the time of its release and I think I recall it having some basic hinting tools. I'll see if I can dig it up this weekend and post some additional information.

Wow, Grant -- Image Club must have been more lucrative than I imagined!

I'm aware of the ATF Type Designer software, but am not really excited by it. For one, how do we know that they really got the optical scaling stuff right? It's not as if Morris Fuller Benton himself wrote any of the code.

Jens Alfke worked on the software. Maybe it's worth asking him about it. There's also a bit of an ad blurb by Henry Schneiker.

I'm also not really sure we need to understand exactly how ATF Type Designer accomplished its optical scaling. Adobe manages to crank out their share of optically scaled fonts, using Multiple Master technology at least internally, if not for release. Also, the landmark Metafont system (part of TeX) is capable of optical scaling, and the free Computer Modern fonts are optically scaled. Metafont is amazingly sophisticated, but unfortunately requires the equivalent of a computer science PhD to use effectively. I think a middle ground is possible - a highly interactive, visual tool that's powerful enough to make good optical scaling, but relatively simple to work with and not anywhere nearly as labor-intensive as current methods of optical scaling.

> Wow, Grant -- Image Club must have been more lucrative than I imagined!

Not really. We obtained ATF Type Designer on an evaluation basis and never actually used it for production. One of the reasons we were able to play with it was because we were licensing a ton of designs from ATF/Kingsley for the Image Club library. There were trying to get us hooked into supporting some of their attempts at 'diversifiction'. The multi-thousand dollar commercial license that Richard mentions above would have certainly been out of the question for Image Club during that time. We were a dyed-in-the-wool Fog shop from day one.

> how do we know that they really got the optical scaling stuff right?

We don't, but it's still our best bet. And maybe the Bentons documented everything (they were engineers) and the Kingsley boys got access to that.

> I think a middle ground is possible

I remeber discussing exactly this some years ago on the TD list. I think the consensus was that it would take too much effort (I'm talking about putting a GUI on Metafont) and there was no money in it.

Raph, I think you should contact Jens Alfke as soon as possible.
The Henry Schneiker link seems broken.

--

Grant, I actually have a copy of ATF Type Designer as well, but I've never tried it. It's Mac-only (at least my version), and I got it after having lost easy access to a Mac. What I'm wondering is this: does your version contain the optical scaling module? If it does, maybe mine does too, and that would make the effort to check it out worthwhile.

hhp

> What I'm wondering is this: does your version contain the optical scaling module?

Again, I'll have to check. I don't recall ever playing with that module, so perhaps it wasn't included. I'm pretty sure I have documentation for optical scaling though...

I had a brief email exchange with Jens. He didn't do the optical scaling; that was probably Henry Schneiker. He was able to tell me that it was based on stroking and filling, rather than interpolation.

I can come to some tentative conclusions, then. I think the characters were drawn in PostScript code by defining a path (moveto lineto curveto), then doing a gsave fill grestore stroke sequence. My strong guess is that the outlines were drawn at the largest optical scaling size (i.e. thinnest, most delicate features), then positive gain added for smaller sizes. Of course, in the Benton matrix engraver, you have a combination of positive and negative gain (erosion). It's possible, but rather difficult, to do erosion in PostScript code.

Because I am a curious sort, I hacked up some Type3 fonts to apply gain in this way to existing (Type1) fonts. The optical scaling results are pretty good, but I don't think the technique would work very well in production. Most importantly, optical scaling has profound effects on the metrics, and the Type3 approach has no way to communicate the adjustments to metrics to the layout application. Type3 fonts are notorious for flakiness and other problems; I'm not even going to post my file because it renders differently on the different PostScript interpreters I have available.

Applying the ATF Bodoni scaling parameters to modern fonts yielded mixed results. The effects of optical scaling looked a bit heavy-handed to my eyes, especially in the 6pt size range. Some fonts looked great (most especially, my draft of the Centaur caps), while othersl, especially Helvetica, looked low-fi. The 'e' in Helvetica doesn't optically scale well using this technique, because it tends to close the aperture even more. I find it quite plausble that the fonts in the ATF catalog were explicitly designed to scale nicely using the Benton techniques.

In any case, my curiosity about the Kingsley/ATF software is now satisfied. I think I have a pretty good idea how to take these basic ideas and do them right, especially so they'll work well with modern software and processes. Even if the Kingsley software produced stunningly beautiful results (for which I have not yet seen any evidence), it is obviously not capable of grinding out OpenType's with optical scaling features, so that the Right Thing just happens in InDesign.

> The effects of optical scaling looked a bit
> heavy-handed to my eyes, especially in the
> 6pt size range.

It depends on how deliberatively you look at the results.
You can't expect functional 6pt type to also look pretty when magnified.

hhp

I dug out my copy of Type Designer this weekend. Following are a few shots of the packaging (binder) and disks for version 1.4 - the thumbnails link through to larger versions of the images on my server.



Fig. 1 - Binder with version 1.3 documentation



Fig. 2 - Cover letter for version 1.4 upgrade and floppies



Fig. 3 - Copyright and trademark information



Fig. 4 - Optical scaling chapter { This chapter has been intentionally left blank. }

Fig. 5 - ATF Type Designer disc contents

Apparently Type Designer does not like Mac OS 9...



Fig 6. - Boom!

Grant: thanks for this.

At this point, Occam's razor would suggest that the Optical Scaling feature was only implemented as a proof of concept, and never made it into production. The phrase "programs that set type must have the ability to compute correct character widths" echoes my above comment about communicating adjustments in metrics to the layout application. I'm unaware of any DTP app that shipped with support for Kingsley/ATF optical scaling, so it's most likely that it just never happened. InDesign does support the Adobe style of optical scaling, so I think we've come at least some of the way, baby.

Grant - thank you!

"Intentionally left blank" indeed. I wish I could've used that on college tests.

hhp

I hate to tease but I have the engineering for the Benton Optical scaling. Will post a sample sheet in several days, something particularly chosen not to be helpful however. But you wouldn't need any of it would you? It sounds like you already have it, or can make it up in your fantasy spin.

Yah, I know, I am a jerk!

Gerald: any original source material you have on the Benton Optical scaling would be greatly appreciated. It's fine with me if you try to choose it to not be helpful; I doubt you've got the analytic skills to be able to accurately identify what's helpful and what's not.

Original sources are always better than inferences and reconstructions, no matter how carefully worked out. Looking forward to seeing them.

Ralf,

Gerald: any original source material you have on the Benton Optical scaling would be greatly appreciated. It's fine with me if you try to choose it to not be helpful; I doubt you've got the analytic skills to be able to accurately identify what's helpful and what's not.

Do you see anything wrong with this comment? One can not

> One can not

Hrant,

But you haven't cut any!
And you have previously admitted -on Typophile- that you're unsure how to do it. Now you'll probably deny saying that, and when I dig up the archive you'll claim I'm ""mischaracterising"" what you said. The question will become: were you making things up then, or are you making things up now? It's really impossible to tell.

Grow up! Don't confuse software with hardware. Also I have never pretended to have photographic memory of software development especially since it was a group development.

Bentons are a very different matter. Photographic memory not required. I have the documentation. You will see by the posting.

Anyway Hrant. I am not willing to work with you on this project. Let me know when you have finished your class with Jim Rimmer?

By the way, those are not ink traps!

> I have the documentation.

So that's your claim to fame, that you have a manual?
Oh, and is it on a lifeboat somewhere in the Baltic?

> I am not willing to work with you on this project.

I don't mind, because -like I said- you have yet to show any actual insight whatsoever into optical scaling to begin with. It's all on other people's briny floppy disks that can't be found. Juvenile bravado stories. And you want to make money out of it - how tragicomic.

And when I do receive Jim's instruction, even any potential insight you could offer will have totally evaporated, since once I have completed the foundation of my understanding I can extrapolate way beyond your own insular experience very quickly. As we say in Armenian: I put you in my pocket.

hhp

Hrant,

You are not the brightest blade in the drawer. Only five companies had Benton Pantographs.

Clue 1. Does this look like a Benton?
http://lanstontype.com/BentonPantograph.html

Clue 2. Does this look like a Benton?
http://www.atypi.org/news_tool/news_html?newsid=127&from=/

Clueless

Gerald wrote:

Finally your admitted "reverse engineering" error of 8% either way leaving your with a total of a 16% error factor is, simply put

Raph,

http://lanstontype.com/Burgess5.htmlhttp://lanstontype.com/Burgess5.html

This would be much easier for you without Hrant's mutant verse.

Hrant

> One can not

> the number of pattern plates used for the scaling of the metal Monotype Centaur

My semi-formal measurements done on the fascimile of Rogers's specimen booklet yields 7.

hhp

Hrant: yeah, I saw your earlier post on that, but I think you might be overcounting. In particular, I don't believe that variations in extender length should count. In the case of the ATF Bodonis, the descenders do get shorter as the font scales down, but I think that it's mostly done by adjustments in the pantograph, not by multiple pattern plates. For pqj, I think they just translated up the pattern plate while cutting the descender, and for gy I think they did nonuniform scaling.

Ha! Now I can make a point actually relevant to the thread. If what Jens told me is true about Kingsley/ATF optical scaling being based on adding a stroke rather than interpolation or any other Metafont-like manipulation of the coordinates, then it almost certainly lacked the ability to do scaling adjustments like pq in ATF Bodoni.

In any case, my analysis (from the same source as yours) yields four ranges: 6-9, 10-12, 14-24, and 30-72. The most obvious "tell" is the comma; the 14-24 design in particular is quite distinct. Other more subtle tells are the left serif of the u (angle varies) and the top curve of the f. For the smallest sizes, the clearest tell is the top bit of the t; it extends farther above the crossbar, which is the opposite of what you'd expect from simply raising the x-height.

I haven't fully figured out what's going on the 6 and 8pt sizes yet, except to note that the 8 looks considerably lighter than the 6 or the 9 (arguably a failure of optical scaling; it's very hard to find an example of this kind of lack of consistency anywhere in the ATF books), and that the 6pt has really short extenders.

I found some curious defects, too. Check out the nick in the e at 11:30 in the 48pt size, the too-small 36pt o, and the broken Q tail at 42 and 72. I found some other subtle curve problems as well. Are these flaws typical for Monotype faces? They would not be for ATF.

http://www.typemuseum.org/contact.htm
This museum holds a large number of plates, most of them from Monotype and they are keen to help.
Hope this helps,
Tim

Raph, your conclusion of 4 matches what I understand was the standard Monotype scheme. I was thinking that maybe they made an exception for Centaur - but maybe not.

If you're looking for the number of physical plates used, maybe extender variance should indeed not be counted (although things like the "f" and "g" make me note that size certainly affects structure, even ligation in the case of the "f"). But in terms of understanding optical scaling, for one thing by working around practical/physical constraints that are no longer relevant) what's going on in every part of the letterforms is important.

> it's very hard to find an example of this kind of
> lack of consistency anywhere in the ATF books

Interestingly, my waterfall of ATF Garamond from 6 to 48 reveals that the 7 is off in color. Not sure where that was coming from though, especially since there's no 7 in the 1923 specimen book. Also, the 14 that I had access to seemed to be a text 14, not a display 14, the latter being what's in the 1923 book*. This might be due to ATF simply changing their mind about where the text/display cutoff should be some time after casting the original version; and Garamond being so small on the body, making the 14 a text might make sense.

* Thanks to Kent for pointing this out.

Defects: well, Monotype was running a business, where quality is never the ultimate goal, it's a means; the process has to be cost-effective, first and foremost. And ATF were indeed better... although with their invention of the pantograph they certainly gave the best excuse to designers to start dumping optical scaling! :-/

> the 6pt has really short extenders.

This is also the case with ATF Garamond. I think it's because at really small sizes you're more likely to set all-caps, and doing so gives more room to the caps. The other factor (something else that Kent has pointed out here, more than once) is that the vertical proportions were integral: the absolute sizes above and below the baseline were based on typographic points, and 1/2 points were avoided*; so at Monotype as well the descenders might have been 2 points for the 8 but only 1 for the 6:
http://www.themicrofoundry.com/other/ATF_line.gif

* Most probably to make mixing different sizes/fonts on the same line a lot easier.

hhp

>he 6pt has really short extenders.

This is also the case with ATF Garamond. I think it's because at really small sizes you're more likely to set all-caps, and doing so gives more room to the caps.

6pt type would be most likely used for notes and other secondary material. Since this tended to space-sensitive, with as much text as possible crammed into as little space as possible, extenders would be made short in order to minimise necessary leading.

As a related aside, when I began researching traditional Byzantine cursive ligatures for Greek, I discovered that the smaller sizes of Granjon's Greek fonts contained more ligatures than the larger sizes. My guess is that this too was a space saving technique, since ligated text is more compact than unligated. The smaller the type, the more likely that the typesetter would be trying to cram as much text as possible into a small area or small number of pages, so type makers would facilitate this in various ways.

> extenders would be made short in order to minimise necessary leading.

But the point is that the descenders are disproportionately short.

Six point type would indeed be used mostly for notes, etc. in a book, but in a newspaper or advertisement all-caps (like the small legalese that mentions that the 0% commission only applies if your mother is from another planet) benefits a lot from a shorter descender depth.

BTW, not really "leading" (since type that's larger on the body generally needs more leading), but maybe "line height".

> The smaller the type, the more likely that the typesetter
> would be trying to cram as much text as possible

Certainly true, but that seems to be overplaying economy: there's also the issue of apparent size (legibility), since smaller type (I mean below like 9 point) needs to be larger on the body. I think this is a better way to look at it, since type with a smaller x-height actually sets in less space!

hhp

Hrant

We've probably discussed this before. I have the ATF/Kingsley disks. But as I may have mentioned, I tossed the manuals a while back (maybe this is a telling attribute to the power of print over that of digital media?).

They never did get to the optimization feature as far as I know. I did talk to the Adobe type manager when it first came out and all he would say was he had problems loading it. So did I. It never seemed to have actually worked for anyone that I know.

Gotta say though, as I recall, the manual had some good introductory info in it. Must have been a day of lunacy when I tossed it. Maybe I was thinking better things were coming down along the old type road. Hah!

Gerald

Hey all

This is my first post here - and possibly my only post, given the tenor of some of the comments I have been reading tonight - but as I think I am probably unique on this list in that I own both a Benton Engraving Machine AND the Optical Scaling module for ATF Type Designer, I thought I might weigh in with a snippet of clarification and a couple of other comments.

1. ATF released the OpScale module integrated into Release 1.5 of their software. I have a feeling this was not a general release, and I further believe that any potential plans for its release were shelved when the company began to flounder so spectacularly. I have tried unsuccessfully to make use of it several times, but - as Henry Schneiker pointed out to me when I bought him a meal here in about 1993 - the information the module expects is directly related to ATF's original cutting slips. These slips are slippery beasts; I have never seen one, and I suspect most of them went into the many and copious bins apparently lined up outside ATF on the day of the infamous auction. Henry further pointed out that these cutting slips were in no way standardised, and that interpreting them required a personal and particular familiarity with the typeface being cut, the slips used to record that typefaces's cutting schedule, and the actual machine itself used to cut the matrices or punches. That is, one workman's notes were another workman's nonsense.

I recall Henry mentioning cutter/follower sizes and ratios, sliding scales, expansions and contractions, etc; all the things you'd expect if you are familiar with the processes involved. I also recall him saying that without the cutting slips the OpScale module was essentially useless, that it was designed to allow ATF's staff to enter existing information when digitising their extant designs rather than develop tables for new faces. And that has been my experience.

I am unfortunately completely unqualified to comment on the underlying mathematics, although Henry described it minute detail. He also showed me some tricky (read: baffling) semi-interactive postscript where a character outline was modified disproportionately depending on some variables I didn't fully grasp. I remember wondering if it was similar to the way Metafont did its magic, but didn't get a chance to ask.

2. Someone said: 'the 6pt has really short extenders'...

I believe that this was done for purely practical reasons, and without any consideration at all for the intended use of the type. This extreme shortening of extenders, which is generally to be seen most prominently below about 8-point, is required to maximise x-height. This is supported further by the fact that usually it is the descenders which suffer the most, an explicit acknowledgement of and concession to the dominant role played by the upper half of a character image in maintaining readability. It is entirely feasible (and has been done, see Cancellaresca Bastarda for one example) to cast a size of type on a larger body as standard release. (And I don't mean, '6-point on a 7-point body'; that's a typographer's specification rather than a founder's.) But when foundries sell 6-point type, they're selling 6-point type, not 7-point, so they make the design fit a 6-point body; and they sacrifice the least important part, the descenders. I think that considerations of use, such as solid setting for subsidiary matter, are of consequential rather than primary significance.

3. Regarding 'discolouration' in size-series.

It's pertinent to remember that foundries occasionally released versions and sizes of types without regard to the tonal contiguity of their neighbours. These apparent discontiguities were simply the result of variants which had been cut for a specific reason and were thus out of the normal series or progression, but which were sometimes shown in the specimens regardless. I had a conversation with David Saunders of Monotype a decade ago and brought up this very issue, and that was his response.

4. 'Most probably to make mixing different sizes/fonts on the same line a lot easier.'

This is never easy, ever. You should try it on your adana!

5. And lastly, it's plain silly to say that Monotype's drawing or production quality was inferior to ATF's; you could argue the toss either way and no-one would win. And to blame this imagined inferiority on the fact that Monotype was running a business...no, sorry, that's right, ATF was a charity; I'd forgotten. Both companies produced solid designs and superb types. You can't take seriously any criticism of 'types' based on their printed images alone; or not these kinds of criticisms, anyway. "And ATF were indeed better..." Uh-huh.

All the best.

Nick

ps. I have chosen 'mr_benton' for my username, but after reading through some of the forums, it really seems that Mr Giampa is the real mr_benton, and that I am merely a pale imposter. If you get around to reading this, Mr Giampa, I would appreciate it very much if you would contact me off-list at your convenience. I feel we may have a few things to discuss. Or, rather, that I desperately desire someone to talk BEM with. Thanks

> I own both a Benton Engraving Machine AND the Optical Scaling module for ATF Type Designer

Wow. We're lucky - please don't leave! :-)

Reading your most revealing post, one thing I realize is that tracking down the cutting slips is probably a waste of time. Reverse-engineering from the actual fonts seems better.

But let me ask:
What are the chances we can get Henry Schneiker on board?

> it's plain silly to say that Monotype's drawing or production quality was inferior to ATF's

Just last night I was setting and printing ATF and Monotype cuts of Baskerville. The difference in quality (of the alloy, the face depth, and certainly the fitting*) is staggering. Sorry. The one aspect where the Monotype is better however is that it has trapping - which I think is telling, considering the [probable] greater reliance on the pantograph at ATF.

* For example, the Monotype's "j" has an elephantine left sidebearing, predictably.

hhp

Nick,

Thanks very, very much for your comments, which add much light to the questions I'm struggling with. I certainly hope this is not your last post.

I'm definitely not trying to make a blanket statement about the relative quality of ATF and Monotype fonts; while the particular data points I've seen seem to favor ATF, I'm sure the real story is a lot more complicated. Of course, since my main sources are the printed specimen books, there are a great many variables other than the quality of the type itself. Indeed, I believe that the 1912 and 1923 ATF books are notable largely for the insane attention to detail in printing quality. It's rare to find an instance of an objective technical flaw such as a broken hairline, misaligned baseline, etc.

I have, by looking at these printed samples, inferred quite a bit about the optical scaling "algorithms" used by ATF. I am excited about the possibility of adapting these algorithms to the digital world. One of the things I'd like to do soon is write up my findings so far, especially so that people who know more about metal type will be able to tell me what I'm missing. I would be honored to have you look over a draft of this paper, if you're so willing.

Hello chaps,

I recall Henry mentioning cutter/follower sizes and ratios, sliding scales, expansions and contractions, etc; all the things you'd expect if you are familiar with the processes involved. I also recall him saying that without the cutting slips the OpScale module was essentially useless, that it was designed to allow ATF's staff to enter existing information when digitising their extant designs rather than develop tables for new faces.

Read this and weep!

Attention Raph

Just letting you know I have sent you an email or two at your acm address. Wasn't sure which address to use. Thanks.

Nick

you can use my raphaelfreeman@gmail.com

Actually, I meant Raph Levien, unless that's you in another guise.

Nick