The best way to reproduce old printed specimens

Nick Shinn

As a collector of books about typography, nothing pisses me off more than halftone reproductions of type.

I recently paid $100 for a book, purchased online, which turned out to be full of halftone reproductions of 19th century specimens, reduced in size by an unspecified amount, in which the background of plain paper clocked in at 20%. I haven’t even bothered to read the damn thing, it is so painful to even open. This is not a new issue, Morison’s The Typographic Book (1963) also has halftone reproductions.

Obviously collotype and gravure are a bit expensive, but there is a simple method, for offset lithography, of reproducing printed specimens in a way that is both attractive and meaningful for typophiles: bitmaps.

I will first assume that typophiles are not so much interested in the document as they are in the type. Therefore we do not need to be reminded that the paper has yellowed, has foxing, and won’t easily lie flat—we would like to see the type as finely detailed, high contrast, black on white.

1. Make your scan or photo greyscale.

2. Move in the sliders at either end of Levels until the paper is 0% and the type 100%. Position the central slider so that the weight of the type appears correct. You now have a high contrast image.

3. Image > Mode > Bitmap. Set to output 2400 pixels/inch, 50% Diffusion dither.

4. Save as TIFF.

5. Now, when you place this file in your page layout document, you may even add a slight tone to the picture box, as a background to suggest the presence of the page, or as a graphic element of your document’s page layout. If you have the luxury of CMYK, make it something like 2/2/5/0 for a soft warm grey.

Jackson Showalter-Cavanaugh

Related, I'd love to hear someone's review of Stan Knight's Historical Types from Gutenberg to Ashendene, which is supposed to have very high quality photographic reproductions.

LeMo aka PatternMan aka Frank E Blokland

The problem with the conversion of letterpress images from halftone to line art is that the 50 Shades of Gray in the squash (you know, the halo effect around the edge of printed areas in letterpress printing) are flattened. Depending on how the contrast is interpreted, the letters will become lighter or bolder, but in any case the edges of the letters will become much harder than one sees in the original prints. One can soften this effect by giving the background a tone, but the outcome will by definition differ from the original.



In older books with reproductions the halftone dots are often crude and probably the photographic images were also not always perfect (or the original prints, like in case of the production of Monotype Poliphilus, but that is another story). In combination with size reduction this makes the reproductions often useless for studying the applied type. I believe that present-day printing techniques make it possible to do this much better though. And for forenamed reasons I prefer halftone then.

How to interpret original images and to explore what one sees and how one's perception is influenced by the Zeitgeist, is subject of a brief ‘Rosart project’ that is part of the Expert class Type Design course at the Plantin Institute of Typography. The students have to make decisions together as a team on the translation of eighteenth-century material into digital letterforms. This panel (although in Dutch) gives an idea of how original material is studied. And this panel (in English) on Rosart for the upcoming exhibition of EcTd-student’s work in the Museum Plantin-Moretus from 17 May till 31 August 2014, is made by Blondina Elms Pastel, who was a student of the 2012–2013 course.

Scott-Martin Kosofsky

There are a lot of variables that come under consideration when reproducing printed type. For one thing, there's the reproduction medium to consider: offset printing (and its line-screen resolution, whether linear or stochastic), high-quality inkjet, display screen. There's also the printing surface to consider, as there is the issue of dot-gain, especially on uncoated stock.

Nick, both gravure and collotype are very soft media. They can be very beautiful in their own right, especially to imply a wide tonal range in photography, but neither are very good for line art of any kind. They are also very rare, except for certain long run magazines that are printed by gravure.

I'm sorry, but I also disagree with your capture method. While there may be occasions when you need to use bitmaps, Photoshop does not have sufficient control over edge pixels for that particular conversion, especially with the slider settings you mention. You also need to keep the metadata in a format that offers the greatest longterm possibilities, a large color space such as LAB or Adobe RGB. Printed pages are never really line art--even the whitest paper has tone, so bitmaps are something to avoid, except in some special circumstances.

There's always a lot of personal interpretation involved, much of which depends on the factors mentioned above, and also the capture medium. Flatbed scanners, for instance, have a light source that approaches the type from only one direction. What you need, instead, is something more balanced. Well-balanced halide lighting is ideal, but it's an expensive setup. And you need the right kind of lens, too. If you want to see examples of really perfect page captures, take a look at the work of 42-Line, the company founded by E.M. Ginger, who edited the magazine Fine Print years ago: http://www.42-line.com

You need a range of pragmatic approaches depending on what you see in front of you. A few years ago, I repackaged Paul Shaw's Helvetica and the New York City Subway System for MIT Press, and reworked all the images. Many were poor, but others were ok. Some had serious color distortions. As an offset book, these were reproduced in CMYK. One of the techniques I used was the application of what's called UCR (under-color removal) to carefully bolster the black while minimizing the halo, which, in CMYK, contains very little black. The results were generally decent, though some images were only as good as they could be under the circumstances. If you compare the limited edition, which I had nothing to do with, to the MIT edition, I think you'll find that the MIT edition is far more faithful to the type specimens and has a lot more vitality.

James Puckett

Related, I'd love to hear someone's review of Stan Knight's Historical Types from Gutenberg to Ashendene, which is supposed to have very high quality photographic reproductions.

Knight did a smart thing almost nobody does—he printed photos instead of trying to make a facsimile that wouldn’t satisfy anyone he used high-quality grayscale photos on good coated stock. Full page images and enlargements are excellent, but the macro images are amazing. These light raking macro enlargements show tiny details like uneven ink coverage, the impression, and the texture of the paper. It’s like having a massive loupe that enlarges big chunks of pages instead of squinting at a few letters. The macro enlargements are much better than anything you can get with a macro lens or high-res scans.

Nick Shinn

ScottMartin:

I’m not suggesting that a bitmap file be used as a longterm resource, only that it is preferable to a halftone for printed reproduction. My method is a simple way to make the type sharp, which is surely what typophiles want to see. I’ve used it, and it works great, producing excellent facsimiles.

Also, I’m puzzled why you don’t think that Photoshop, at 2400 dpi, has sufficient pixel resolution! That’s a basic resolution that imagesetters burn digital type and halftone dots.

And I did cover the issue of whether the background page be left blank or given a slight tone. Of course, this is a postmodern question: whether the document reproduced is to be rendered as close to its initial state, or with the patina of age, or in a way that has been edited for the new document and contemporary eyes.

**

I recently supplied some images for an article in Codex 3 magazine. It was disappointing to see, for instance, on page 50, an old ad comprised of type and line art (an engraving), reproduced in CMYK—with no undercolor removal. It would have been so simple to use the method I described to produce a sharp reproduction.

Scott-Martin Kosofsky

Nick, while I agree with you about those Codex images, I'm afraid I cannot agree about halftones in general. They key to it is the judicious use of sharpening, using the techniques that are generally called "unsharp masking." That is how you control the edge pixels that give images of printed type their definition. How much you use, and at what settings (fade and such), depend on the sharpness of the image capture. Sharpening creates a sensation of contrast, and for that reason it's sometimes better to deal with contrast first, as a primary issue.

Why I prefer halftone for the representation of old specimens is that only halftone can give the viewer an idea of the printed context--impression, paper, ink. Without these, we're not really sure what we're looking at. For me, it's not a matter of reproducing the warts and all (no one intended the work to get dirty), but to make a knowledgable interpretation of what I think I'm seeing.

I think you misread what I said about resolution. The problem with converting to bitmap is not resolution, it's "threshold." You see, Photoshop arbitrarily assigns a 50% conversion threshold for bitmaps. That means that anything 50% and over will be black, and anything lower will be white. With letterpress images, especially old ones, part of the spread or halo may fall on either side of the 50% line, so even though what you're seeing may be sharp at high res, it's actually a distortion of the information to one degree or another, sometimes ok and sometimes not. An extreme adjustment of contrast (which is, in effect, what you're doing) will only exacerbate the distortion. With carefully made halftones, the viewers will be able to make their own interpretations--you are giving them all the information you can. Doing this well is a high craft in which experience counts. I have to say, too, that some very knowledgable people have different fast from my own.

Sharpness is also adjusted differently for different printing techniques and surfaces. Sometimes, a file that prints perfectly will look too sharp on the screen (obvious edge pixels), which means that it can't be used online successfully. You have to keep the metadata separate.

James, I think Stan Knight's book is great and we're all grateful for it, though if you don't mind my being picky, I'd say that it could be better. (The printing itself is mediocre.) Do take a look at the 42-line website. It's really good work.

I'm trying to think of how to post a good demo. Unfortunately, what I'm talking about requires high resolution for viewing. If you stop by my place, I can show you.

Nick Shinn

I’ve never been a fan of unsharp masking.

ScottMartin, if you’re converting a greyscale image to bitmap at 2400 dpi, the fact that the threshold is 50% doesn’t matter—you will still get a soft edge if there was a soft edge in the original image. Remember, this is how halftone dots are imageset.

I don’t think it’s possible, barring collotype or photographs, to get a really good idea of a letterpress document in a single image. So what I have done in the past is to show a halftone of a photograph of the document, seen as a 3D object, accompanied by a separate bitmapped image of the type at 100%, with the background knocked out. That way, the reader can see context and layout in one image, and in another also experience the type as solid black ink on paper. To me, this is preferable to looking at a picture of it through a fuzzy veil.

Scott-Martin Kosofsky

Chacun à son goût, Nick. That said, I think you're making generalizations that seem to me to be rather primitive, and your statement about the soft edge of black bitmap images is flat wrong. I don't think you understood what I said. That's not to say that using bitmaps in the way you suggest can't be applied successfully sometimes, but rather that it is, in my opinion, a last-ditch technique to be deployed under unfortunate circumstances.

When I speak of unsharp masking, I am not referring only to the coarse Unsharp Mask filter in Photoshop, but more sophisticated tools that are available on the image-capture end, and within Photoshop as 3rd-party plugins or through the Smart Sharpen tool set.

I'm basing my opinion on decades of experience developing and producing much admired image-driven books, Some of them involve art and photography, while others are based heavily on documents. These are skills honed over many thousands of images. At the end of the day, you're left with your own taste, so if you prefer bitmaps, so be it.

It's hardly worth mentioning collotype, as it barely exists today. Despite its very seductive texture, it's still only as good as the image capture allows it to be.

Kent Lew

What I think ScottMartin may have missed in Nick’s original description is that he uses Diffusion Dither for his bitmap conversion, which is akin to a stochastic halftone screen on the midtones. So, thresholding does not produce the absolute edge cutoff that it does with a typical lineart conversion.

I think using an actual stochastic halftone (matched to press capabilities) with a properly prepared image along the lines of some of ScottMartin’s suggestions (unsharp masking and UCR/GCR if printing CMYK) can produce the best balance af tonal range/subtlety without some of the coarse, obscuring fuzziness of even a fine traditional halftone.

Nick Shinn

ScottMartin, I too am an old fart basing my opinion on decades of &c., &c., so I’m afraid that line of “reasoning” won’t convince me of anything!

I mentioned collotype and gravure because they produced the best reproductions I’ve seen, e.g. gravure in The Art & History of Books, by Norma Levarie, 1968, Meriden Gravure Co.

Scott-Martin Kosofsky

Nick, the Levarie book was printed by collotype, not gravure. Meriden Gravure never worked in true gravure, but rather in collotype, taking its name from "heliogravure," a reflection of how collotype plates were "cured" by sunlight.

Speaking of Meriden, take a look at one of their books from the next decade, Joe Blumenthal's Art of the Printed Book, which was printed in 1973 in 300-line screen tritone (offset) throughout. (I'm referring to the original printing, not the later reprints.) I think it was an exemplary way of reproducing typography, though a little bit soft for my taste. Several years later, the follow-up volume, The Printed Book In America, was produced as line art, the analog equivalent of your method. I think it's far less successful. Not only does it lack the "perfume" of its predecessor, but it conveys less useful information about the type. Which is not to say that it's bad, only that I don't think it's as good.

Kent was right, I did miss your reference to diffusion dither. That does help, but, all things being equal on the image capture front, I still don't think it's as good as fine-screen offset of well-made halftones, whether FM screening or 10-micron stochastic.

Again, I'll refer back to the work of 42-line. Tell me that's not better than what you're doing.

Nick Shinn

I haven’t seen that. It may well be better than what I’m suggesting. However, despite the title of this thread, my goal in making it was to ameliorate really dreadful reproductions made with coarse halftone screens, by explaining a simple method that will work with run-of-the-mill printing, not address the high end.

BTW, I’m pretty sure that my copy of the Levarie is gravure, because it has a mechanical screen pattern. It’s my understanding that the reticulation which occurs in collotype is random.

Blondina Elms Pastel

When determining the best method to reproduce old printed specimens, it seems the first
question to be answered is: What is the end purpose of the reproduction?
         Is it for a serious study? Is it for the pleasure of having a “copy”? When considering
reproduction for a serious study it would be most beneficial to use a method that retains
as much of the type specimen's original DNA as possible.
         Looking at the bitmap method, it appears to involve a lot of initial personal
interpretation (for example: How much contrast in enough contrast?) that will then
be interpreted by some printing method before being passed onto the enduser.
The aforementioned method, results in a document that will allow the study
of an interpretation of that specimen; rather than a study of the original type specimen.
The halftone method seems to retain more of the type specimen's DNA.
Data can be gathered from document (type of paper, pressure on paper, timeline)
that will certainly influence the way the printed type is perceived. For the enduser (not having
the original), it seems this document information is just as important as the printed type
if we want to make an in-depth study. Of course, the halftone method will also be an
interpretation of the original; however, it seems to be more faithful.
         Context, seems to be a key word here, without it, there is little to build on.
It appears that the halftone method brings context into play.

[Deleted User]

The user and all related content has been deleted.

LeMo aka PatternMan aka Frank E Blokland

It isn’t widely known that Iowa artist Grant Wood also made a black and white gothic version of his American Gothic by converting the painting to a bitmap.

Nick Shinn

So much for peer review.
Have any of you even seen the bitmap method, let alone tried it?!
This is not a theory.
Only Kent has demonstrated insight and understanding.

George Thomas

I tried bitmaps in the past at very high resolution with results that did not impress me. After re-reading your description of how you do it, I definitely will try again. Thanks for sharing it.

Dave Crossland

Frank, I do love your conspiracy theories of typographic and art history