Behind the arduous process of making Chinese computer fonts

But there are tens of thousands of Chinese characters, and a 5×7 grid is too small to be legible. Chinese requires a grid of 16 x 16 or larger—that is, each character has at least 32 bytes of memory (256 bits). If you imagine a font containing 70,000 low-resolution Chinese characters, the total memory requirement will exceed 2 megabytes. Even a font containing only 8,000 of the most common Chinese characters requires approximately 256 KB to store bitmaps. This is four times the total memory capacity of most off-the-shelf personal computers in the early 1980s.

Although these memory challenges are serious, in the 1970s and 1980s, the most difficult problems faced by the production of low-resolution Chinese fonts were aesthetic and design issues. Long before anyone started using programs like Gridmaster, most of the work was done outside the computer, using pen, paper, and correction fluid.

The designer spent several years trying to design bitmaps that meet low memory requirements and retain some elegant calligraphy. Among those who created this character set, whether by hand-painting bitmap drafts of specific Chinese characters or digitizing them with Gridmaster, there are Ling Huanming (Ling Huanming) and Ellen Di Giovanni.

The core problem faced by designers is to switch between two completely different Chinese writing methods: hand-drawn characters, generated with a pen or brush, and bitmap fonts, generated with an array of pixels arranged on two axes. The designer must decide how (and whether) to try to recreate certain orthographic features of handwritten Chinese, such as entry strokes, thinning strokes, and exit strokes.

In the case of Sinotype III fonts, the process of designing and digitizing low-resolution Chinese bitmaps is thoroughly documented. One of the most fascinating archive sources of this period was a binder full of grids with hand-painted pound signs everywhere-the sketches were later digitized into bitmaps of thousands of Chinese characters. Each of these characters has been carefully laid out and in most cases edited by Louis Rosenblum and GARF, using correction fluid to erase any “bits” that the editor disagrees with. On top of the initial set of green hash marks, then a second set of red hash marks represents the “final” draft. Only then did the data entry work begin.

Louis Rosenblum Collection, Stanford University Library Special Collection

Considering the number of bitmaps that the team needs to design—if the machine has any hope of meeting the needs of consumers, at least 3,000 (preferably more)—people might assume that designers are looking for ways to simplify their work. For example, one way they can do this is to copy Chinese radicals—the basic components of a character—when they appear in roughly the same position, size, and orientation from one character to another. For example, when generating dozens of common Chinese characters containing “female radicals” (女), the GARF team could (and theoretically should) only create a standard bitmap, and then copy it to the radical in each character The molecule appears in it.

However, as the archival material shows, no such mechanical decision was made. Instead, Louis Rosenblum insisted that designers adjust each of these components—usually in almost imperceptible ways—to ensure that they are in harmony with the overall characteristics in which they appear.

In bitmap Juan (Juan, elegant) and mian (This way, delivery), for example—everyone contains radical women—radical changes are so small.In characters Juan, The middle part of the female radical occupies a horizontal span of six pixels, while the character has five pixels mianHowever, at the same time, the lower right curve of the female radical only extends one pixel outward in the character mianAnd in the characters Juan The stroke did not extend at all.

Louis Rosenblum Collection, Stanford University Library Special Collection

Throughout the font, this level of precision is the rule rather than the exception.

When we juxtapose the draft bitmap with its final form, we will see that more changes have been made.In the draft version round (Luo, collect, network), for example, the stroke in the lower left corner extends downward at a perfect 45° angle, and then gradually becomes a digital version of the outer stroke. However, in the final version, the curve has been “flattened”, starting at 45° and then flattening.

Louis Rosenblum Collection, Stanford University Library Special Collection

Although the space in which designers must work may seem small, they have to make a surprising number of choices. Each of these decisions affects all other decisions they make for a particular character, because even adding a single pixel often changes the overall horizontal and vertical balance.

The ruthless size of the grid affects the designer’s work in other unexpected ways. We see this most clearly in the demonic problem of achieving symmetry. Symmetrical layouts that appear in large numbers in Chinese characters are especially difficult to represent in low-resolution frames because, according to mathematical rules, creating symmetry requires an odd-sized space area. Bitmap grids with even sizes (such as 16×16 grids) make symmetry impossible. In many cases, GARF manages to achieve symmetry by using only part of the entire grid: only 15×15 areas are used in the entire 16×16 grid. This further reduces the available space.

Louis Rosenblum Collection, Stanford University Library Special Collection

When we started to compare bitmap fonts created by different companies or creators for different projects, the story became more complicated. Consider the water-based (氵) (bottom and right) that appears in the Sinotype III font instead of another early Chinese font created by HC Tien (left), a Chinese-American psychotherapist and entrepreneur. Chinese calculations in the 1970s and 1980s.

Louis Rosenblum Collection, Stanford University Library Special Collection

Although the above examples may seem small, each one represents another decision (among thousands) that the GARF design team must make, both in the drafting phase and the digitization phase.

Of course, low resolution did not remain “low” for a long time. Advances in computing have brought denser bitmaps, faster processing speeds, and lower and lower memory costs. In our current era of 4K resolution, retina displays, etc., it may be difficult to appreciate the artistry in the creation of early Chinese bitmap fonts-whether it is aesthetics or technology-although they are limited. But solving problems like this will eventually make computing, new media, and the Internet accessible to one-sixth of the world’s population.