John Atanasoff was getting nowhere. It was a bitterly cold night in 1937, and he was stuck in the physics building at Iowa State University agonising over his stalled idea for a machine to solve algebraic equations. Exasperated, Atanasoff impulsively jumped into his car and drove aimlessly across the flat expanse of Iowa. He did not emerge from his reverie until 300 kilometres later, when he noticed he was crossing water. He’d reached the Mississippi River.
Stopping off at a half-empty tavern, Atanasoff warmed up with a couple of bourbons at a corner table and pondered his options. How could his machine work? Over the next few hours, he jotted down four ideas: fully electronic operations; binary calculations; regenerative memory; and a “black box” of logic circuits. With only his bar napkin to write on, Atanasoff had set down some of the fundamental concepts of modern computing.
YOU could say that John Atanasoff had been preparing for computing all his life. Growing up in Florida as the child of an electrical engineer and a maths teacher, at nine he was puzzling over the logarithms on his father’s slide rule, reading a college algebra text and secretly fixing his home’s newly installed electrical wiring. By 1925, he was living in the town of Ames, a 22-year-old graduate student of maths at Iowa State University. There he used mechanical gear-driven calculators, but was frustrated by their limitations.
Advertisement
Atanasoff focused on calculating large numbers of simultaneous linear equations. Variables shared by a large set of such equations can be solved by a process known as Gaussian elimination. These calculations can be rendered through many operations of the simple arithmetic that mechanical calculators already handled on a smaller scale. Tackling simultaneous linear equations mechanically might take weeks, however. A century earlier, a similarly frustrated Charles Babbage decided to harness steam power to do his calculations, so Atanasoff pondered on how to electrify this drudgery.
Mechanical calculators had been around since the 17th century, and Atanasoff himself had already invented a small one for geometrical equations. But the 1930s saw the dawn of a new electrical era: Charles Vannevar Bush built the thousand-geared Differential Analyser, Howard Aiken constructed the immense and mill-like Harvard Mark I, and Konrad Zuse in Germany invented his Z1 computer. Yet electromechanical computing was limited by innumerable moving parts demanding constant maintenance. Electricity could move faster and more smoothly than any gear or shaft. The future of computing, Atanasoff decided, lay in electronics.
By 1939, working in a dingy basement with student Clifford Berry, Atanasoff had built a table-top prototype, and began work on what is now called the ABC (Atanasoff-Berry Computer). Packed into a desk-shaped 2-metre steel frame, it used IBM punch cards to feed in equation coefficients, converting them to binary format. A revolving “memory drum” enabled vacuum-tube logic circuits to perform addition and subtraction; occasionally recharged capacitors kept the memory current. Intermediate results were burned electrically into “scratch” cards in binary form; a hole equalled 1 and no hole equalled zero. When an equation’s coefficients had been zeroed out and the variable’s value determined, the final result was displayed.
The machine was completed in 1941 for a mere $6000. A journalist on the Des Moines Register went to see it and found “an electrical computing machine said here to operate more like the human brain than any other such machine known to exist”.
But the ABC had a glitch. The scratch cards produced about 1 error in every 10,000 calculations, a serious problem given the immense number of calculations in the 29 simultaneous equations that Atanasoff set as his goal. When the US entered the second world war in late 1941, Atanasoff was called away to the Naval Ordnance Laboratory in Washington DC. Visiting Ames in 1948, he found that a colleague, coveting precious basement space, had dismantled the computer years before. The ABC was gone.
The question of who built the first computer provokes many heated arguments. In 1967, it set writs flying. Sperry Rand, the company that owned the patents on ENIAC (the Electronic Numerical Integrator and Computer), invented in 1942, was entangled in a lawsuit with rival manufacturer Honeywell. Sperry Rand had used its patents to demand licensing fees from anyone building electronic computers, creating a virtual monopoly. Honeywell went on the attack.
In one of the longest trials in the history of the US federal courts, Honeywell noted that in June 1941, John Mauchly, one of the inventors of ENIAC, had spent five days on a friendly visit to Atanasoff’s home in Iowa. While he was there, Honeywell contended, Mauchly had seen the ABC and he had used and patented its concepts in the ENIAC.
Mauchly maintained that he had been unimpressed by the ABC, that it did not work, and he had barely even seen the machine. But witnesses swore they had seen him by the machine with his shirtsleeves rolled up. His own letters and notes also contradicted his testimony. And despite Mauchly’s claim to have already long pondered computing, Atanasoff recalled that he had expressed “surprise that the base 2 number system was advantageous”.
ENIAC was both massive and massively impressive: it was not only programmable, which the ABC was not, it was also much faster. But Atanasoff was the first to use electronic logic and memory. In 1973, the judge ruled that ENIAC’s patent was invalid because its inventors had derived their ideas from Atanasoff. The ruling rocked the industry, smashing Sperry Rand’s monopoly. Now anyone could build computers. Many assumed that the desire to create a vital new industry was the real reason behind the ruling.
It was simpler than that. Although the ABC had never been patented, its “prior use” – being noted publicly in a newspaper – undercut anyone else trying to claim a patent. To be patentable or “usable” only requires a reasonable promise of functioning. The ABC’s crucial innovations did indeed work. Mauchly had seen these innovations, and they had been publicly demonstrated. So ENIAC’s priceless patents were unenforceable, and Sperry Rand wisely decided not to appeal. But being “the first computer” in the courthouse is one matter: in the public’s mind, ENIAC remained the original electronic computer.
In the mid-1990s, a student at Iowa State University taped a withering quote to the wall of Atanasoff’s old basement lab. ENIAC’s co-inventor John Presper Eckert Jr. had been outraged by Atanasoff’s claims to be making history: “It’s such an outlandish exaggeration to consider that he did it. He did some little thing which he never finished and which wouldn’t have worked if he ever did.” The quote served as a goad, for in that basement a team was rebuilding the ABC to settle the nagging question: did it work?
The team built a replica of Atanasoff’s invention as it stood in the summer of 1941. Clifford Berry had continued improving the ABC through 1942, but the replica would show exactly what Mauchly saw. “We resisted the temptation to improve any of the original electronics,” says project manager John Gustafson. The replica was unveiled in October 1997. And it worked. Even with the scratch-card problem, the ABC accurately handled five equations with five unknown values.
With the building of the replica, the magnitude of the ABC’s achievement has become clearer. “The amazing thing is how much is exactly like today’s computers,” says Gustafson. “Fifteen-point decimal precision; rotating discs for storage, dynamic memory… a system clock to synchronise all events, and an ordinary wall outlet for power.” Unlike the base-10 behemoths that followed, the ABC was an affordable, desk-sized binary machine. Your desktop computer has more than a little in common with the jottings on that bourbon-stained napkin.