Computer Memory Devices - Part 2
August 1960 Electronics World

August 1960 Electronics World   Table of Contents Wax nostalgic about and learn from the history of early electronics. See articles from Electronics World, published May 1959 - December 1971. All copyrights hereby acknowledged.

The first installment of this two-part "Computer Memory Devices" series discussed the use of magnetic data storage in the form of drums and tapes. Both types provide long-term, non-volatile storage, but both suffer from a relatively slow execution of writing and reading to and from, respectively, the media. In 1960 when Electronics World magazine printed the articles, drums and tape were used during execution of programs because electronic storage in the form of vacuum tube circuits was extremely costly in terms of power, cost, and physical space. As recently as the early 1980's, magnetic tape storage still dominated the data storage field, especially where huge amounts on information needed to be stored and retrieved. Semiconductor memory, while less voluminous and less power hungry, still added a lot to the cost of computers. If you were around at the time and used a PC, you remember that 64 kilobytes or RAM was considered high end and many (maybe most) off-the-shelf PCs came with just 16 kB or 32 kB of RAM. As reported in Part 2, the advent of magnetic core memories greatly improved the speed of I/O operations, but it was also expensive because building the arrays was very labor intensive, requiring assemblers to thread hair-thin wires through the tiny cores.

Computer Memory Devices - Part 2

By Ed Bukstein, Northwestern TV & Electronics Inst.

Not limited by mechanical delays, magnetic cores provide speedier access time than tape or drums.

Of the various means for magnetically recording and storing data in the "memory" sections of electronic computers, two were described in the first portion of this article. These two, magnetic tape and magnetic drums, were found to complement each other. While the former provides the greater storage capacity, the latter permits quicker access to any desired portion of the information being held for use.

Because of these characteristics, many computers use these two in combination, with the drum serving as the main memory and the tape providing back-up storage. Blocks of information are transferred from the tape to the drum at some time prior to the need for that data in the computer. After the computer has made use of this data, other blocks of information can be transferred to the drum. However, there are some applications where it is desirable to have even shorter access time than that provided by drums. In these, magnetic cores are used for the main memory. In fact, the drum may then become the back-up storage device.

The magnetic core is a ring-shaped piece of magnetic material. Since the core can be magnetized in either of two directions (clockwise or counterclockwise), it can be used to store a bit of binary information. Magnetization of the core in one direction can be used as a representation of binary 1, and magnetization in the opposite direction can represent binary 0. The binary number 10101, for example, can be stored in five magnetic cores as shown in Fig. 11. Here the arrows indicate direction of magnetization.

Physically, the magnetic cores are small - 0.08 inch outside diameter is representative-and are placed on an array of perpendicular sets of wires, as shown in Fig. 12. For simplicity, a 3-by-3 array of cores is shown here, but 32-by-32 and 64-by-64 arrays are commonly used in practice.

Information is written into a core by passing currents through the wires on which the core is mounted. Assume, for example, that all of the cores in Fig. 12 are initially magnetized in the 0 direction, and that it is desired to store a 1 in the core located at the intersection of wires X₁ and Y₂. This would be accomplished by passing currents through wires X₁ and Y₂ simultaneously. The values chosen for these two currents are such that either one individually is not sufficient to reverse the magnetization of the core, but the combined effect of both at the intersection X₁Y₂ is sufficient to reverse the core at this location.

The individual currents are frequently referred to as half-currents because each has a value equal to half of the total current required to reverse the magnetization of the core. This technique, known as coincident-current switching, makes it possible to write a 1 in any selected core by passing half-currents through the appropriate X and Y wires.

The process of reading a core is somewhat similar to the writing process, except that the currents are passed through the X and Y wires in a direction opposite that used for writing. Assume, for example, that the core at X₁Y₂ is to be read. This would be accomplished by passing half-value "read" currents through wires X₁ and Y₂ simultaneously. These read currents are al-ways in such direction as to switch the selected core to the 0 direction of magnetization. The read currents will therefore switch core X₁Y₂ to the 0 direction, and the reversing magnetic field will induce voltage in the sense wire. This output, which is a few millivolts in amplitude, is amplified (by the sense amplifier in Fig. 12) and then used to trigger a flip-flop stage. As a result, the flip-flop stage is now switched to the 1 condition, and the 1 which was previously stored in core X₁Y₂ has now been transferred to the flip-flop.

If core X₁Y₂ had been in the 0 rather than the 1 condition, the read currents would not have reversed this core's magnetization. Under these conditions, there would have been no output from the sense wire and the flip-flop would have remained in the 0 condition. The read currents therefore cause the flip-flop, in either case, to assume the condition (0 or 1) of the core being read.

Since reading a core which has a 1 stored in it causes this core to switch to the 0 condition, the read-out process destroys the information in the core (although the information is still available in the flip-flop). For this reason, the reading operation is followed by a writing operation to switch the core back to the 1 condition, so that it may retain stored data.

In addition to the wires shown in Fig. 12, one other wire, not shown, is threaded through all of the cores in the array. This is known as an inhibit wire and is used during the process of writing. As explained previously, a 1 can be written into a selected core by passing currents through the associated X and Y wires. If however, a 0 is to be written into the core, something must be done to prevent the write currents from switching this core to 1. This is accomplished by passing a half-current through the inhibit wire at the same time the two half-currents are passed through the X and Y wires. The inhibit current is in such direction that it opposes the write currents and therefore prevents the selected core from being switched to 1. At first consideration, it may seem more reasonable to write a 0 into a core simply by preventing the write currents from passing through the X and Y wires. It happens that the associated circuitry, however, is much simpler if the write currents are used but nullified by an inhibit current.

Since the cores in an array are selected for reading or writing on a one-at-a-time basis, this type of storage would be relatively slow if all of the bits of a given number were stored in the same array. Under these conditions, a number would have to be written (or read) one bit at a time. For this reason, each bit of a number is stored in a separate array of cores. Each of these arrays is known as a plane. Each of the cores shown in Fig. 11 would therefore be placed at corresponding locations in five different planes and could be read or written simultaneously. The total storage capacity of a core memory is determined by (1) the number of cores in each plane and (2) the number of planes. A twelve-plane 64-by-64 memory would, for example, be capable of storing 4096 twelve-bit numbers.

Magnetic drum and tapes are cyclic storage devices, involving mechanical motion of the recording surface with respect to the heads. Recorded data passes the reading heads in the order in which the data is located on the surface of the drum or tape. By contrast, magnetic cores constitute a random-access type of storage. This means that the cores can be read in any order at once simply by pulsing the appropriate X and Y wires. It is for this reason that the access time is very short as compared to tape or drum storage, and magnetic cores are frequently used in the main memory of high-speed computers.

Posted January 2, 2023