DESIGN
OCR-B
DESIGN OFTHE TYPEFACE
Signature
CONTENTS INTRODUCTION
5
MONOSPACE TYPEFACES
7
DESIGN PROCESS
8
DESIGN BY GRID
11
DEFINING A CENTRE-LINE
12
CHARACTER RECOGNITION
14
FROM TEST TO PUBLISHED
16
INFLUENCED BY UNIVERS
19
CHARACTERISTICS 20
<<DESIGN<OF<OCRB<<<<<<<<<<<<<<< D<AFRUTIGER<<ADRIAN<JOHANN<<<<<
DESIGN
04
INTRODUCTION In an era when computers were fairly underpowered, each character had to be as distinct as possible. Computers often struggled to clearly interpret indistinct typefaces. It was this fact that lead, in 1968, to the creation of the typefaces OCR-A and OCR-B. Each is designed to be easy for machines to interpret using the OCR process. In designing OCR-B, Adrian Frutiger created a typeface which still retains typographic flair while providing a solution to the problems surrounding OCR technology. It represents the best way technology should function â&#x20AC;&#x201C; human and machine both putting themselves to the fullest possible use. This booklet clearly defines the design process of Adrian Frutiger when designing his optical character recognition font OCR-B.
05
DESIGN
06
Monospace Typefaces A monospaced font is a font whose letters and characters each occupy the same amount of horizontal space. Monospace typefaces take their cue from typewriters, where all letters conform to a specific physical width, resulting in letterforms that must expand or condense to make the best use of the allotted space. Hence, the wide ‘i’s and tight ‘m’s.
They are also referred to as non-proportional typefaces, in contrast to typical proportional typefaces, where each character is a different width. Another feature of monospace typefaces – which can be seen as a pro or as a con – is that they are spaced perfectly evenly, creating nicely aligned columns of text. Optical character recognition has better accuracy with monospaced fonts, such as OCR-A and OCR-B.
07
Design Process
DESIGN
In designing the OCR-B typeface Adrian Frutiger and his colleagues, Robert Ranc and Gilbert Weill, would meet up every three months at one of the their offices. This series of meetings would continue over a span of five years. Their first task was to agree on a common grid. Then, at one of the following meetings, Frutiger was given a template and told that the typeface would be read according to these points. The characters had a consistent line weight and the most important thing was to determine the form-giving â&#x20AC;&#x2DC;centrelineâ&#x20AC;&#x2122;. It was needed for the milling of the typewriter face.
08
In the meetings Fruitger would always hand out photocopies to each participant, of the drawings he had produced in studio. Each of them would then go off and do their own maths in their respective companies and come up with a different result. If they came to the conclusion that part of a shape was too wide, too narrow, too high or too low, the changes would be noted with a pencil on the drawings right there and then in the meeting. The translation into coordinates was carried out by the participating firms.
09
DESIGN “ The
great stroke of luck in my life is to have been blessed first with an artistic feeling for the shapes, and second with an easy grasp of technical processes and of mathematics. “ – Adrian Frutiger
10
Design by Grid The most important prerequisite was the design of a suitable grid, fine enough to be acceptable for the reading methods of every computer manufacturer. The complete typeface was built up to this pattern. After numerous years of meetings, a common grid was decided upon and Adrian Frutiger received a template upon which to base his typeface design. The grid cells were only a few millimetres big and the system was considerably finer than the matrix of OCR-A with its 5 by 9 cells. Fruitger always drew the curves in his designs.
The engineers had agreed that adjusting them to the grid wasnâ&#x20AC;&#x2122;t the task of the designer, but the task of the computer. Hundreds of drawings were created in his studio, all of them then filled in with black. The grid was only superimposed later for copying purposes, so that the manufacturers could read the characterâ&#x20AC;&#x2122;s mass precisely. If a cell was more than half full it counted as a plus, if it was less than half full it counted as a minus. Initially only horizontal steps were possible but later the cells could also be divided diagonally. The resulting computations were done by the computer firms. They looked after legibility and the typewriter manufacturers looked after the execution of the typeface.
11
Defining the Centre-Line
DESIGN 12
The first designs were made in the form planned for letterpress printing. The characters had a consistent line weight and the most important thing was to determine the form-giving â&#x20AC;&#x2DC;centrelineâ&#x20AC;&#x2122;. This line became the basis of reference for all modifications, and the guideline for the engraving of the typewriter typeface.
“ The
skeleton of a letter is like a keyhole engraved on the reader’s memory. The letter that is read is the key that seeks and finds its lock. When the designer strays too far from the base form, then you get friction, frustration or unreadability. “ – Adrian Frutiger
13
Character Recognition The formal principle of OCR-B was based on the premise that each character must differ from another by at least 7 per cent in the worst possible case. To check this, two characters were superimposed in such a way that they covered each other optimally.
DESIGN 14
Additionally, this test was carried out using two different printing weights: a fine weight caused by weak pressure on the keys of the typewriter keyboard or through a lack of ink on the typewriter ribbon, as well as a fat, squashed weight caused by strong pressure on the keys or by bleeding ink. Even if the fine and fat weights were superimposed â&#x20AC;&#x201C; the original weight could be fattened by a factor of up to 1.5 â&#x20AC;&#x201C; the difference of 7 per cent still had to be guaranteed. Generous character spacing was needed to guarantee correct processing; serifs, on the other hand, were rather detrimental to performance since they increased the coverage ratio of the characters. Furthermore, the paper should not be reflective and the type should not bear any stains.
“ We
worked for five years to achieve sufficient visual difference between character-pairs that look quite similar to a machine. “ – Adrian Frutiger
15
From Test to Published 1963
DESIGN 16
The first test version of OCR-B dates from 1963 and contained 109 characters. The bowl shapes of the majuscules were static, while there are two types of bowl shapes in the minuscules: a round, static bowl, for instance in c d p, and a flat, dynamic one in b g q. All numerals had dynamic shapes but the curves varied: flat curves in 2 3, leaning towards the diagonal in 5 and clearly diagonal in 6 9. Initially Frutiger designed the majuscules so they were of the same height as the numerals, but for the first test version the former were scaled down to differentiate them clearly.
1965 The version published as Standard ECMA-11 in 1965 contained 112 characters including three additional letters with diacritics. Some characters had undergone considerable correction. This is obvious with the W, whose outer diagonals became curved; with the numeral 0, which received a more oval shape to differentiate itself better from the capital O; with j, which now had a normally placed dot: as well as with the aforementioned b g p, which now featured static bowl shapes. The same was true for the $ sign. The @ had obviously changed, whereas the slight incline in the numeral 1 was hardly visible. Altogether the typeface now had a more consistent shape compared to the test version.
1969 A further extension and correction phase took place from 1969 onwards. Five more characters were added: the section mark §, the two Dutch ligatures IJij, the German esszett ß and the Japanese currency symbol ¥. The British pound sign £ was changed considerably. There was still problems in differentiating D O 0 and B 8 &. With the D, the curved stroke now started directly at the stem, the O was more oval: the zero, on the other hand, had become more angular. The Q was adapted in shape to the O and the tail of the Q was altered. The B now featured a markedly wider shape that resolved the conflict with the 8, whereas the ampersand & got a smaller lower bowl at the expense of the upper one. The j was changed yet again and, similar to the i, it now had a horizontal bar while the y received a curved descender. Eventually all corrections were not beneficial in terms of shape; instead Frutiger had acknowledged the overarching goal of character recognition.
1976 Even with the international standard ISO 1073/II in 1976 the OCR-B project had not yet come to a close. Instead the number of characters was successively extended. From 121 characters in 1976 the font grew to 147 by 1994. The additional characters were mainly due to the inclusion of special characters for different languages. Some character shapes of Linotype’s digitized version of OCR-B are not identical to the ECMA original – there is more similarity in Berthold’s, but only in the typewriter version.
17
DESIGN “ OCR-B
would have been better still, had I not used the Univers terminals. But at the time I was still completely faithful to Univers. Today I would prefer oblique terminals. But the criteria were very difficult to develop. “ – Adrian Frutiger
18
Influenced by Univers The font family Univers is one of the greatest typographic achievements of the second half of the 20th century. The family has the advantage of having a variety of weights and styles, which, even when combined, give an impression of steadiness and homogeneity. The clear, objective forms of Univers make this a legible font suitable for almost any typographic need. In 1954 the French type foundry Deberny & Peignot wanted to add a linear sans serif type in several weights to the range of the Lumitype fonts. Adrian Frutiger, the foundryâ&#x20AC;&#x2122;s art director, suggested refraining from adapting an existing alphabet. He wanted to instead make a new font that would, above all, be suitable for the typesetting of longer texts quite an exciting challenge for
a sans-serif font at that time. Starting with his old sketches from his student days at the School for the Applied Arts in Zurich, he created the Univers type family. In 1957, the family was released by Deberny & Piegnot, and afterwards, it was produced by Linotype. Frutiger remained extremely close to the style of Univers when designing OCR-B, and many of its characteristics can be seen in his optical character recognition font. For example, the majuscules are wider and higher than the numerals in Univers, whereas the opposite is true for OCR-B. The characters that are very similar in shape get a serif, horizontal bar or curved stroke in OCR-B. And in contrast to Univers, and to the letters of OCR-B, the numerals feature dynamic curves.
19
Typeface Characteristics
DESIGN
NUMERAL 5 The numeral 5 features a diagonal incline in the upper section of the character.
Adrian CHARACTER SPACING For the typewriters all characters had to be of even width, these were monospaced faces. Therefore we had to draw a narrow m.
20
CAP HEIGHT Of utmost importance was the difference between capitals and numerals. Identical heights were experimented with for a long time, but there would always be pairs that didnâ&#x20AC;&#x2122;t work. The B â&#x20AC;&#x201C; 8 combination caused some major headache: specifically, the machines that read the counter would never recognise the difference correctly. Eventually Adrian Frutiger came up with the idea to keep the numerals higher than the capitals. Since the numerals were of correct proportions right from the start and thus formed the basis for the standard, all the capitals of the typeface were eventually scaled down.
UPPERCASE W The uppercase W has outer diagonals that are slightly curved, wheras in the inner diagonals remain straight.
21
SERIFS
DESIGN 22
Adrian Frutiger staunchly refused to introduce any serifs if it wasnâ&#x20AC;&#x2122;t absolutely necessary. Serifs increae the similarity between characters, and are therefore less suitable for machine readers. But because of the danger of them being confused there was no other choice with certain characters, such as with i j and I.
SIMILAR CHARACTERS Adrian Frutiger made use of the shapes of similar characters to create others. This can be seen clearly in the lowercase q, p, d and b. It almost seems like he designed one single character and then rotated this to create the rest.
The major differences between each letter can be viewed in the ascender/descender lenght. The d and b have a significantly longer ascender than the p and qâ&#x20AC;&#x2122;s descender.
VARIATIONS With the K, the arm and tail donâ&#x20AC;&#x2122;t come to an acute point on the stem. This has never been done before in any of Frutigerâ&#x20AC;&#x2122;s many typefaces, but there was no other option technically. If there had been a gap in the centerline, the reader would possibly have read the K as a stroke and a chevron. The same goes for the Q, in which the tail enters inside the counter in OCR-B, unlike anywhere else in his work.
UPPERCASE D The curve of the uppercase D almost perfectly traces the curve of a semi-circle.
23
DESIGN
24
OCR-B