Magnetic-core memory

The core memory, magnetic core memory, or even Ferritkernspeicher (English magnetic - core memory, or ferrite - core memory ) is an early form of non-volatile memory of the electronic calculating machines. It consists of threaded onto wires hard-magnetic ring cores ( engl. cores ) that can be re-magnetized and read by electric currents in the wires. The sign of the magnetic remanence of the individual ring cores represents the memory contents.

Core memory was used about 1950 to 1980 in the then usual calculators. You have now no more scope.

In a more general sense, the term is also used for memory in other, newer technologies, in terms of " memory in the core of the computer ".

History

The first time, she was born in 1949 in Shanghai physicist An Wang of Harvard University. In contrast to the MIT Harvard was not interested in patenting his own inventions. Wang received the patent itself under the name pulse transfer controlling device.

Jay Forrester's group, who had worked on the Whirlwind project at MIT, learned of Wang's work. Whirlwind needed a fast storage system for a real-time simulator. So far, run-time memory had to be used therefor. So-called storage tubes, based on cathode ray tubes such as the Williams tube or the Selectron, due to manufacturing difficulties and low reliability in operation obtained never a significant market position and the mid-1950s were replaced by core memory.

Two key inventions led to the development of core memory, which allowed only the development of the well-known in today's computer. The first, An Wang was the write -after -read cycle ( write - after - read cycle), which solved the problem, that the reading of information the same also destroyed: the magnetic polarity of the toroidal cores can only be determined by this be re-magnetized.

The second, Jay Forrester, was coincident -current system ( Coincident currents), which allowed the control of a large number of magnetic cores with a small number of wires ( see below, operation ). Core memory were prepared manually; the work was performed under the microscope and required fine tact.

In the late 1950s, factories were built in Asia, where low-wage workers established the core memory. The prices have been lowered, so that both the low-cost, but low in performance drum memory as well as the expensive high-performance systems could be replaced with electron tubes in the early 1960s.

Due to the mechanical complexity and thus bulky design, the capacity of the core memory was limited. It embodiments have been built up to a few megabytes. This, however, several cabinets were already required in less space and effort they came up with less than 100 kilobytes.

Although the production of core memory was briefly interrupted before its automation, the costs followed the then-unknown Moore 's law. The cost of technology from the beginning about one dollar per bit of data dropped to about $ 0.01 per data bit, until the core memory was replaced in the early 1970s by the silicon-based RAM.

The patent Wang was until 1955, when the technology was already in use, not approved. Several lawsuits prompted IBM, Wang patent for multi-million dollar abzukaufen. Wang used the money to expand the Wang Laboratories.

Core memory belonged to a family of technologies that the magnetic properties of materials made ​​it their own. In addition to the toroidal storing, for example, the design with magnetic pen store was used ( ICL computers in the 1970s ).

In the 1950s, the electron tubes were already mature, but still fragile and short-lived because of the heated filaments, unstable and the energy consumption too high. Core memory formed an energy-saving, miniaturized and reliable alternative. Essential, however, was that he, like the drum memory, the memory contents are not lost even when power is removed. According to a further miniaturization, the so-called Bubble Memories, he ( EEPROM / flash memory ) was then only really replaced by non-volatile semiconductor memory.

Description

Operation

A core memory essentially consists of a large number of magnetizable, magnetically hard ferrite cores, which are formed into rings, and are therefore referred to as a ring core. Magnetically hard in this context means that each core can store a bit of data in the sign of the residual flux density Br. For reading or writing of the cores run at least two electrical and mutually insulated wires through the annular openings, as illustrated in the adjacent figure on a core.

In the electric current in the write wire must be large enough for writing, that is exceeded by the current generated in the magnetic coercive force Hc in the magnetic circuit of the core. This results in the hard magnetic material of the ring, which has a nearly square hysteresis loop, to save the state in the sign of the residual flux density. The remanent flux density can assume two stable points, which are designated in the hysteresis loop with Br and Br.

Simultaneously with the write operation of the read line, a voltage pulse is in the second wire induced can be determined on the basis of which the orientation of the original stored in the remanent flux density of the core. The information may therefore be read only destructive. A possibly ummagnetisierte core must then be rewritten to produce the original data content again.

The selection process is illustrated in the adjacent figure the hysteresis curves: Was previously stored a positive residual flux density Br in the core, it comes with the writing of a "0" by the large change in the magnetic flux density in the core on the read line to a voltage pulse Ul in the order of some 100 mV. In the hysteresis curve on the left is thick drawn the curve of the flux density shows. After switching off the current in the write line, the residual flux density Br is left in the core, corresponding to the "0" state. Was previously stored in the core of a negative residual flux density Br, the flux passes through only a small part of the hysteresis curve and the rate of change is minimum. Consequently, the pulse voltage on the read line is also minimal. In both cases, the core is after reading out the state " 0", and the original memory content shall be able to be re- written by an inverse power - in.

Storing in addition to the core sense amplifiers are required for the operation, which convert the low -voltage pulses to the sense line in the appropriate logical voltage level. To describe current sources are necessary.

Arrangement in a matrix

So now not each core requires its own two wires and a separate sense amplifier, the following trick is used: The current in the write line is split into two wires which carry only one half of the required magnetic reversal current. These X -and Y- wires are arranged in a lattice structure ( matrix), and carry at each intersection of a core. If now a certain core be addressed, is the X - wire in question and the Y wire on each distinct half of the required flow contributed. Thus, other cores are either achieved only by half or no field strength and do not change their state.

For the realization of a 16 Kbit memory are thus 2 × 128 wires and as many controllable power sources needed.

Read / Write

The sample- wire (sense -line, S) or S- wire called, and in earlier core save even the lock wire ( - for reading and writing in a matrix two more wires are needed, which are looped through all the cores inhibit -line, Z).

Basically, a read and a write cycle is always run simultaneously. In the read cycle is re-magnetized with the X - and Y- wires of the corresponding core to logical "0" back. If the core is already a " 0" is stored in the read cycle happens not in "1 ", however, is induced due to the magnetic reversal of a pulse in the S- wire. In the write cycle, the core is magnetized again in the "1 " direction. In the case of a previously stored "0" sends a current in the opposite direction during the write cycle through the lock wire. This is sufficient to attenuate the field intensity of the X -and Y- wires so far that the core is not demagnetised in the "1 " direction.

Since the scanning wire and the lock wire will never be used at the same time, used later systems only one wire. Additional control switches between the two functions.

Computer systems with core memory often exploit the fact that not everyone read value must be preserved at all: For example, if a data word value is to be added, will initially only be read ( read cycle ). With the write cycle, the system waits until the addition is complete. Then, not the original value, but equal to the result of addition is written. The velocity of certain operations can be doubled.

The total time consumed for a Lese-/Schreibzyklus, called cycle time; It was a measure of the speed with which a memory core could be used. She was with computer systems of the 1960s often also a rough measure of the overall performance of the system, as later, the CPU clock rate.

Physical Properties

Early systems had cycle times (Read and Write Back ) of about 20 microseconds, they fell in the early 1960s to 2 microseconds, reaching the early 70s 0.3 microseconds. The possible clock rates between 50 kHz and 3 MHz were so in about the same magnitude as the home computer of the late 1970s and early 1980s, such as the Apple II and the Commodore 64

Data words with 32 bits of data are (depending an XY grid ) distributed in 32 layers, thus can access an entire data word in a read-write cycle.

Core memory are non-volatile memory - they get the information for an unlimited period without electricity. Also, core memories are robust against electromagnetic pulses, high temperatures and radiation. These are important advantages for military applications such as combat aircraft, but also in space vehicles. Several years of the commencement of the availability of semiconductor memories beyond this core storage were used.

Characteristic of core memory: they respond to the current, not the voltage - but the sense wire provides a voltage pulse. That was an important prerequisite for high clock rates with relatively large geometrical dimensions of the memory.

The Selektierstrom (half select current) Im / 2 was typically 400 mA for the late smaller and faster memory. Earlier memory needed larger currents.

The diameter of ring core moves in the order of 1 mm to 0.25 mm at the shortest access time.

A negative characteristic of the memory core is the dependence of the hysteresis on the temperature. The Selektierstrom is therefore adjusted by the control system - by a sensor, the temperature is measured. The Programmed Data Processor PDP-1 Digital Equipment Corporation is an example of this. Other systems worked around this problem by using the memory was housed in a temperature-controlled container. Examples here are the IBM 1620 ( it took up to 30 minutes to reach the operating temperature of 41 ° C) or heated in an oil bath housed called the IBM memory 709.

Other types

The destructive readout process and the forced rewriting of bits read in the classic ring core memory led in the 1960s and 1970s a number of developments which eliminate this disadvantage. A possibility is to construct the magnetic circuit so that the direction dependence of the magnetic flux density is used. Such cores are referred to as bi-axial cores.

Two holes are drilled in the block-shaped core, which are orthogonal to each other. Through a hole of the query wire is passed through the other hole, two wires. The writing and the reading wire wire Wherein the reading out of the core, a current pulse is sent only by the polling wire, which causes, depending on the residual flux density in the core of a positive or negative voltage pulse on the read wire. But it is characterized still no permanent change in the magnetic flux density in the core, the core does not lose its memory contents. Must be sent simultaneously to describe two sufficiently large current pulses in the appropriate direction by both the query wire and by the orthogonally related write wire. Only in this way the amount of coercive force is exceeded and it comes to storing the new state.

Other core designs which exploit different flux density distributions in magnetic circuits are referred to as Transfluxor. Two different sized holes are asymmetrically mounted and thereby led three lines on the toroidal core. By the different widths of the leg due to the asymmetry and size of the holes in the magnetic core material is obtained, also in this constructive version, the possibility of the memory contents can not be read destructive.

Miscellaneous

The usual technique in the slang term core dump, which stands as an expression of a dump in the form of a snapshot in particular in the hardware-oriented programming, is derived from the core store. For troubleshooting all cores ( cores ) can be read ( dump ) in order to make as program errors can locate.