Early Digital Computing

By 1952, the basic concept for the air defense project was approaching a degree of maturity. A radar network had been assembled, and Lincoln Laboratory was ready to begin testing. MIT’s Whirlwind computer showed promise for providing the real-time computational capability needed for SAGE.  However, the reliability of the computer still posed a significant problem. Before plans for a nationwide air defense system could be taken seriously, the computer would have to become much more reliable.

Magnetic-Core Memory

One of the most significant limitations of computers at the time was the storage tubes that were used for internal memory. These tubes were large and slow, and worst of all, they were unreliable. One of the greatest breakthroughs in the development of Whirlwind was the invention of magnetic-core memory. That invention was the key development leading to the widespread adoption of computers for industrial applications because, unlike computers with storage-tube memories, computers with magnetic-core memories were reliable.

Forrester with memory arrayJay Forrester inspects a 64 × 64 memory array. Photograph courtesy of the MIT Museum.

In 1947, while working on Whirlwind in the MIT Servomechanisms Laboratory, Jay Forrester began to think about developing a new type of memory. He conceived of a new way of configuring memory units—in a three-dimensional structure. Although Forrester initially thought of using glow-discharge tubes, preliminary tests indicated that these tubes were too unreliable. Lacking a good way to implement a three-dimensional memory, Forrester dropped work on his concept for a couple of years. Then, in the spring of 1949, he saw an advertisement from the Arnold Engineering Company for a reversibly magnetizable material called Deltamax. Forrester immediately recognized that this was the material he needed for the three-dimensional memory structure.

Forrester directed one of his students, William Papian, to study combinations of small toroidal-shaped cores made of ferromagnetic materials possessing suitable hysteresis loop characteristics. Papian's master's thesis, "A Coincident-Current Magnetic Memory Unit," completed in August 1950, described the concept of magnetic-core memories and showed how the cores could be combined in planar arrays, which could in turn be connected into three-dimensional assemblies. Papian fabricated the first magnetic-core memory, a 2 × 2 array, in October 1950. The early results were encouraging, and by the end of 1951 a 16 × 16 array of metallic cores was completed.

Whirlwind computer


The Whirlwind computer, seen here in the Barta Building on the MIT campus, was the precursor to the computer developed to accommodate the demands of the SAGE system.

The organization and direction of Project Whirlwind now went through a major change. The task of developing a flight simulator was abandoned, and the focus of the program shifted to air defense. In September 1951, all members of the Servomechanisms Laboratory who were working on Whirlwind were assigned to a new laboratory—the MIT Digital Computer Laboratory, headed by Jay Forrester. Six months later, the Digital Computer Laboratory was absorbed by Lincoln Laboratory as the Digital Computer Division. Lincoln Laboratory took over the development of magnetic-core memories.

Operation of the early metallic magnetic-core memories was still unsatisfactory—switching times were 30 µsec or longer. Therefore, in cooperation with the Solid State and Transistor Group, Forrester began an investigation of ferrites. These nonconducting magnetic materials had weaker output signals than did the metallic cores, but their switching times were at least ten times faster. In May 1952, a 16 × 16 array of ferrite cores was operated as a memory, with an adequate signal and a switching time of less than a microsecond. So promising was the performance of the new array that the Digital Computer Division began construction of a 32 × 32 × 6 memory, the first three-dimensional memory.

Close-up of memory arrayThis close-up photograph of a magnetic-core memory array shows the 64 × 64 arrangement of magnetic elements on the surface plane.

Whirlwind was by this time in considerable demand, so a new machine called the Memory Test Computer was built to evaluate the 16,384-bit core memory. When the Memory Test Computer went into operation in May 1953, the magnetic-core memory, in sharp contrast to the electrostatic-storage-tube memory in Whirlwind, was highly reliable. Forrester promptly removed the core memory from the Memory Test Computer and installed it in Whirlwind.

The first bank of core storage was wired into Whirlwind on August 8, 1953. A month later, a second bank went in. A different memory was subsequently installed in the Memory Test Computer, enabling that machine to be used in other applications.

Whirlwind megnetic-core memory banksWhirlwind magnetic-core memory banks.

The improvement in Whirlwind's performance was dramatic. Operating speed doubled; the input data rate quadrupled. Maintenance time on the memory dropped from four hours per day to two hours per week, and the mean time between memory failures jumped from two hours to two weeks.

The invention of core memory was a watershed in the development of commercial computers. The technology was quickly adopted by International Business Machines (IBM), and the first nonmilitary system to use magnetic-core memories, the IBM 704, went on the market in 1955. Magnetic cores were used in virtually all computers until 1974, when they were superseded by semiconductor integrated-circuit memories.

Part 2: MIT Lincoln Laboratory extends computer capability—the AN/FSQ-7 computer and software development

Adapted from E.C. Freeman, ed., Technology in the National Interest, Lexington, Mass.: MIT Lincoln Laboratory, 1995.

top of page