First Chinese Personal Computer: A Story of Engineering Daring

China's Digital Rise and the Challenge of Early Computing

Currently, China stands as a leading global digital economy, possessing an unparalleled hardware supply chain. Numerous highly successful companies, including Alibaba, Tencent, and ByteDance, are at the forefront of worldwide innovation. However, this advanced technological landscape is built upon a solution to a significant computing problem that originated four decades ago: the creation of Chinese word processing capabilities.

The Expansion of Computing in China During the 1980s

Starting in the early 1980s, China significantly increased its acquisition of computers from the United States and Western nations. The number of imported microcomputers rose dramatically from 600 in 1980 to 130,000 by 1985. Companies based in the United States, Japan, and Europe actively sought to capitalize on this surge in demand.

A Critical Obstacle to Adoption

A substantial hurdle existed for both prospective Chinese computer users and Western manufacturers. Western-made personal computers, printers, monitors, operating systems, and software were initially unable to process Chinese character input or output, particularly in the early and mid-1980s, and not without significant modification. Without substantial adjustments, personal computers were largely ineffective for those needing to work with the Chinese language.

The Memory Challenge and Chinese Fonts

A primary issue revolved around memory – specifically, the amount of memory needed for Chinese fonts. Early Western computing engineers established that an English font could be constructed using a 5-by-7 bitmap grid, requiring only 5 bytes per character. While not visually refined, this grid provided adequate resolution for displaying Latin alphabet letters on terminals or printouts.

Storing the 95 printable characters of U.S. ASCII demanded just 475 bytes of memory, a small portion of the 48 KB of memory available on an Apple II at the time.

The Increased Demands of Chinese Characters

Achieving comparable legibility for Chinese characters necessitated a much larger grid size. Engineers were compelled to increase the grid dimensions geometrically, from 5-by-7 pixels to 16-by-16 pixels or larger, requiring at least 32 bytes of memory per Chinese character. Storing the bitmaps for the 8,000 most frequently used Chinese characters – either in simplified or traditional form, but not both, and excluding any metadata – would consume approximately 256 KB of memory.

This was four times the total memory capacity of most commercially available personal computers in the early 1980s, and that was before considering the memory requirements of the operating system and application software.

A Story of Innovation and Ingenuity

These circumstances set the stage for a remarkable engineering story, showcasing entrepreneurial spirit and inventive problem-solving within the global digital revolution.

Introducing the Sinotype III

This is the first of two articles exploring the Sinotype III, an early experimental machine capable of handling Chinese-language input and output. Constructed using a commercially available Apple II, but enhanced with a custom-programmed word processor and operating system, the Sinotype III demonstrated the feasibility of adapting Western computers for Chinese language use, thus unlocking a substantial new market.

Focus of This Article

In this initial installment, we will delve into the significant technical challenges related to computer memory, fonts, and operating systems faced by the creators of the Sinotype III. We will also examine the innovative solutions they developed to overcome these obstacles.

The Genesis of Chinese Computing: A Story of Innovation

The narrative unfolds with the Graphic Arts Research Foundation (GARF), the organization widely considered the birthplace of Chinese computing. The Ideographic Composing Machine, also known as the Sinotype, was initially conceived in the late 1950s by MIT electrical engineer Samuel Hawks Caldwell, supported by GARF funding. Unfortunately, Caldwell’s passing in 1960 brought the project to a temporary halt.

Throughout the 1960s and 1970s, the Sinotype project persevered, benefiting from the involvement of various entities, including Itek Corporation, RCA, and ultimately, GARF once more.

Louis Rosenblum: Reviving a Pioneering Project

The revival of Sinotype can largely be attributed to Louis Rosenblum. Born in New York City in 1921, he was another alumnus of MIT, earning a bachelor's degree in Applied Math in 1942.

Rosenblum studied under Harold Edgerton, the celebrated electrical engineering professor known for his iconic “milk drop coronet” photographs. Immediately after graduating, he joined Polaroid, collaborating with Edwin Land on projects like the development of instant photography.

In 1954, Rosenblum transitioned to Photon, where he focused on photocomposition for writing systems beyond the Latin alphabet. Having a thorough understanding of Samuel Caldwell’s groundbreaking work on Sinotype, he embraced the project and actively revitalized it when he became a consultant at GARF in the mid-1970s.

GARF's Advisory Board and the Path to Microcomputers

GARF’s continued development of the Sinotype project extended into the early 1980s. During this period, they assembled an advisory board comprised of distinguished scholars and experts in China-related fields.

Harvard linguist Susumo Kuno joined the board, alongside Richard Solomon, who played a crucial role in arranging Richard Nixon’s historic visit to the PRC in 1972 and then led the Social Science Department at the RAND Corporation.

Despite the expertise of this group, the significant advancement of the Sinotype project – the shift from a minicomputer-based system (Sinotype II) to a microcomputer-based one (Sinotype III) – was sparked by a student with limited prior involvement at GARF.

Bruce Rosenblum: An Unexpected Catalyst

This catalyst was Bruce Rosenblum, Louis Rosenblum’s son. In 1979, he completed a short, two-week assignment at GARF, focusing on data management for the Sinotype II project.

At the University of Pennsylvania, Bruce balanced his studies with his role as photo editor for the Daily Pennsylvanian, a student-run newspaper. The newspaper was remarkably well-equipped and staffed by students with considerable technical skill.

By the fall of his junior year, the newspaper’s aging typesetting equipment (two Compugraphic typesetters) required replacement. Bruce, along with three colleagues, researched alternatives and ultimately secured a $125,000 contract with Mycro-Tek in Wichita, Kansas, and Compugraphic in Wilmington, Massachusetts.

The Inspiration from a TRS-80 and a Bold Proposal

A pivotal moment for the Sinotype project occurred in early May 1981. Bruce, having finished his final exams, visited the newspaper’s offices and observed his colleague, Eric Jacobs, working on a TRS-80 Model II personal computer from Radio Shack.

Jacobs was exploring the potential of this microcomputer to manage the newspaper’s business operations. Bruce watched for approximately 30 minutes, an experience that would profoundly influence his future.

“It was the first time I’d ever seen anyone work on a microcomputer,” Bruce later recalled. “Those few minutes inspired the entire Sinotype III project and ultimately my career in computers.”

Later that week, Bruce casually suggested to his father that a microcomputer could potentially replicate the functionality of GARF’s expensive Data General hardware, used for Sinotype II, at a significantly lower cost – perhaps $10,000 compared to the existing $100,000+ investment.

His father was intrigued and asked Bruce if he would be willing to undertake the programming task. Despite lacking formal computer science training, Bruce had experience with PDP-8 assembly language and BASIC. He responded with “the chutzpah of a newly minted graduate who had no immediate job prospects,” agreeing to the challenge.

Sinotype III: A Proposal and Successful Implementation

In June 1981, Bruce formally presented his Sinotype III proposal to Bill Garth, Prescott Low, and his father, Louis, in New York. He dressed professionally for the occasion, wearing a three-piece suit.

His proposal outlined hardware costs of $7,500 and programming fees of $5,000, promising a Chinese word processor running on an Apple II, deliverable within four months. Success would represent a substantial reduction in cost.

Bruce secured the assignment and programmed Sinotype III from June to November 1981, concurrently working as a tour guide at Independence Hall in Philadelphia. He would write assembly code by hand during breaks and transcribe it at night. After his tour guide position ended on Labor Day, he dedicated two months to completing the code and delivering it to GARF.

Memory Enhancement in Early Chinese Computers

One of the initial challenges encountered by GARF and the Rosenblums revolved around the limitations of computer memory. Early personal computer developers in China actively sought methods to maximize the available memory within their systems. We will examine two primary approaches, frequently used independently or in combination: Adaptive Memory and the utilization of Chinese Character Cards.

The Sinotype III system was composed of several key elements. These included a Sanyo DM5012CM 12-inch monitor, an Epson MX-70 printer, a Corvus 10 MB “Rigid Disk Storage” unit for storing the Chinese character bitmap database and associated “descriptor codes,” an Apple Disk Drive for text file storage, and the core Apple II computer.

Initially, the Apple II was equipped with 32 KB of RAM, expandable to 48 KB directly on the motherboard. Bruce Rosenblum noted, via email, that they maximized this capacity immediately upon purchase. However, 48 KB proved insufficient for their needs. Consequently, Bruce implemented a common modification among “power users” of the time: installing a 16 KB memory card in slot 0, increasing the total available memory to 64 KB.

Even 64 KB was not enough, however. “A complete encoding system required more memory,” he explained, “as did the 16-by-16 bitmaps for the 100 most frequently used ideographs.”

Bruce then began investigating a unique Apple II modification that was largely unexplored. “I discovered a way to install a second 16 KB board in slot 2 of the Apple II,” he stated, “resulting in a total of 80 KB of memory.” This was a completely unconventional approach, yet it functioned using readily available components.

This modification, however, exceeded the inherent limitations of the machine. The 6502 microprocessor in the Apple II could directly address only 64 KB of memory. Therefore, despite adding an extra 16 KB with the second memory board, there was no native method for the Apple II to simultaneously access these additional memory addresses. The unconventional nature of this mod surprised an Apple engineer during a conversation with Bruce, who had never considered such a possibility.

To enable access to the full 80 KB of memory, Bruce bypassed the standard operating system and developed his own in assembly language. A crucial aspect of this custom program was the ability to “switch between two overlapping 16 KB banks.” Essentially, while only 64 KB of memory locations were accessible at any given moment, rapid switching between the two expansion cards created the illusion of accessing both, at a speed imperceptible to the user. This yielded a 25% increase in usable memory, potentially allowing for the inclusion of up to 400 additional Chinese characters in on-board memory.

Bruce delivered the completed code to GARF just before Thanksgiving, then embarked on an extended backpacking trip across Europe and Asia. From this point forward, the development of Sinotype III was primarily managed by Louis Rosenblum and GARF, though Bruce continued to provide consultation, maintaining regular communication with his father from his various locations throughout Europe, China, India, and beyond.

Achieving Real-Time Chinese Character Input

Despite the innovative design of the system, Louis and Bruce calculated that only between 600 and 1,000 Chinese characters could be accommodated within the available on-board memory. Considering the operating system size, application programs, and the memory footprint of each individual character, the majority of the machine’s Chinese character lexicon would necessitate storage on external media like floppy disks or a hard drive.

Initially, Bruce briefly explored the use of PROM (programmable read-only memory) chips. However, this approach quickly proved impractical. In the early 1980s, the largest available PROM chips offered a maximum capacity of 2 KB, which equated to only 28 to 51 Chinese characters. Storing a 7,000-character lexicon using this method would have required either 138 or 250 PROM chips, a substantial quantity.

The possibility of utilizing floppy disks for character storage was also investigated. This option was deemed unsuitable due to the sheer number of disks required and the slow access times associated with retrieving character bitmaps from floppy disk storage. Consequently, GARF decided to equip Sinotype III with an external hard drive, a relatively uncommon microcomputer peripheral at the time.

This decision, however, introduced performance limitations. Compared to the rapid speeds of most computing operations, hard drives were considerably slower. These drives relied on rotating magnetic disks – “platters” – similar to record players. Data was read by a head, analogous to a record player’s needle, and retrieval speed depended on the head’s position and the disk’s rotational alignment.

Specifically, accessing characters stored on the hard drive took 10 times longer than accessing those in RAM. Characters in RAM could be retrieved in approximately 100 milliseconds, a time imperceptible to human users. Conversely, retrieving characters from external storage required up to a full second – a duration easily noticed by the user.

A one-second input delay would have been unacceptable in the context of 1980s personal computing, where users were becoming accustomed to real-time typing experiences. This delay, being ten times longer than the 100-millisecond RAM access time, would have been consistently noticeable when inputting less frequently used characters.

To address this issue, Louis Rosenblum conceived of a system he termed “adaptive temporary storage.” Sinotype III would dynamically adjust the characters stored in RAM based on recent user input. Upon startup, the on-board RAM would initially contain only a core set of frequently used characters.

When a less common character residing on the hard drive was inputted, it would take up to one second to access. However, as Rosenblum explained, “its code and dot matrix pattern will be noted in the random access memory.” Essentially, these characters would be temporarily copied from the hard drive into the RAM cache, accelerating subsequent retrieval times.

Chinese-on-a-Chip

Despite utilizing techniques like toggling and adaptive memory, a substantial number of Chinese characters remained outside the scope of these methods. While frequently used Chinese characters comprised a significant portion of typical text, creating technical or specialized content would inevitably require accessing the “off-site” character repository. Increasing the number of these “low-frequency” characters available “on-site” was crucial for achieving a Chinese computing experience comparable to that enjoyed by users of English-language systems.

In the late 1970s and early 1980s, engineers began investigating a distinct hardware solution, known as “Chinese Character Cards” (Hanka), “Chinese Cards” (Zhongwenka), “Chinese Character Generators,” “Chinese Font Generators” (Hanzi zimo fashengqi), or, as one source noted, “Chinese-on-a-Chip.” Similar to memory and graphics cards, these “Chinese character cards” were intended for installation into motherboard expansion slots.

These cards contained thousands of pre-programmed Chinese bitmaps and input encodings. Essentially, they functioned like an external hard drive, but with significantly faster speeds and improved reliability. This dedicated hardware offered a substantial performance boost.

The development of “Chinese-on-a-chip” cards wasn’t a primary focus at GARF. Instead, they evolved from earlier custom-designed Chinese systems predating the personal computing era.

Examples include the Ideographix IPX, created by Chan Yeh, and the Olympia 1011. These systems utilized microprocessors specifically for generating character bitmaps and storing input information. The Olympia 1011 Chinese word processor, essentially a specialized electric typewriter, dedicated one of its three Intel 8085 processors solely to Chinese character generation.

By the early 1980s, these character generators were mass-produced and sold as standalone products. Users no longer needed to purchase a complete word processor like the Olympia 1011 to access on-board character generation capabilities.

Instead, a “Chinese Character Card” could be purchased and installed into a personal computer.

Tsinghua University was among the first institutions to concentrate on Chinese Character Cards. Researchers there developed a card capable of storing approximately 6,000 Chinese bitmap patterns in a 32-by-32 dot matrix format.

Throughout the mid- and late-1980s, dozens of different “Hanka” cards became available, manufactured and sold by companies in Japan, China, Taiwan, Hong Kong, the United States, and other regions.

By the mid- and late-1980s, the “Chinese-on-a-chip” method had become so prevalent that nearly all computers supporting Chinese or Japanese languages included a character generator card.

From Caldwell’s Sinotype in the 1950s to the Rosenblum team and GARF’s work on Sinotype III in the 1980s, resolving the memory challenges associated with Chinese characters was fundamental to opening the Chinese market to computing. Innovative solutions, including increasing computer memory, creating adaptive memory algorithms, and developing dedicated hardware, overcame these obstacles and initiated the computer revolution in China.

The subsequent challenge involved extending this capability beyond the computer itself to encompass connected devices. The next article in this series, forthcoming on TechCrunch, will explore the difficulties of designing and programming early computer monitors, printers, and other peripherals capable of handling Chinese text output.