What Are the Different Types of Sense Memory?

Magnetic core memory is the main form of random access computer memory and has existed for 20 years. This memory is often referred to as core memory, or informally as core memory.

The most common memory of early computers was made of various magnetic cores. This magnetic core memory has been

The magnetic core will be magnetized or change the direction of magnetization when a certain current flows on the wire. The minimum threshold of the current that can make the core magnetize can be obtained through experiments and material process control in advance. Each core has two XY wires running perpendicular to each other. There is also a readout line that runs through obliquely. These lines can be clearly seen in the photo above. These lines form an array. Addressing in different directions. The magnetic core can generate two opposite magnetizations according to the direction of the current during magnetization, which can record data as 0 and 1.

When writing, input a current slightly higher than the 50% magnetic ring magnetization threshold on the XY coordinate line corresponding to the core to be written, so that only the core corresponding to the XY coordinate will be in both lines at the same time. There are currents, and the current that exceeds the threshold value after superposition, the core is magnetized or the magnetization direction is changed to write a bit of data, and the current passed in all other cores is either 0 or 50% magnetization threshold, No magnetization current can be magnetized, so no data is written.

The reading is more complicated. The reading current is sent to XY. The reading current is slightly larger than the 50% magnetization threshold current when writing. The direction of the reading current is known in advance. There will be a current exceeding the threshold in the core corresponding to the XY addressing coordinates. If its original magnetic field direction is opposite to the magnetic field direction corresponding to the readout current, then due to the magnetic state of the magnetic core being reversed, there is a huge When the magnetic flux changes, there will be a large induced current on the oblique reading line, so we know that this magnetic core stores data opposite to the read signal. If its original magnetic field direction is the same as the magnetic field direction of the readout current, then since the magnetic state of the magnetic core has not changed, there will be no induced current on the oblique readout line, so we know this magnetic field. The core stores the same data as the read signal. The data in the magnetic core is thus read out, but it is not finished yet, because it is worth noting that after reading the data at this time, it is clear that no matter what data is stored on the original magnetic core, it will be written the same after reading it. The data has been read, that is, the read is destructive, so there must be a way to restore the stored data after the read. So after reading it, you need to write the original data again in order to recover the original data. The method is to write the data as described above, and use the data stored in the magnetic ring in the cache to write back. Therefore, the reading of the magnetic core memory is quite troublesome and relatively slow. The magnetic core that is not selected when reading is the same as that when writing, it will not change the magnetic state and generate an induced current, so it will not be read out and will not interfere with the read data of the selected core.

The core memory has a difference from our general storage concept. Generally, a memory write is always slower than a read, but the core memory is just the opposite. It is slower to read than write because it The readout is destructive, so the readout must include a write process to recover the data.

The term "core" comes from a traditional transformer with its windings surrounding the core. In core memory, wires pass through any given core-they are single-turn devices. The properties of the materials used for the memory core are significantly different from those used for power transformers. The magnetic materials used for core memory require a high degree of magnetic remanence, the ability to maintain a high degree of magnetization, and a low coercive force, thereby requiring less energy to change the direction of magnetization. The core can adopt two states, encoding one bit, which can be read when the "induction line" is "selected". The core memory contents are retained even if the memory system is powered off (non-volatile memory). However, when the kernel is read, it resets to a "zero" value. Circuits in the computer memory system then recover the information in an immediate rewrite cycle.

Core memory operation

The most common form of core memory, X / Y line coincidence current, is used in the computer's main memory. It consists of a large number of small toroidal ferrimagnetic ceramic ferrites (magnetic cores) in a grid structure (organized as a "stack" called a planar Layer ", the wire goes through the hole in the center of the core. In earlier systems there were four wires: X, Y, Sense and Inhibit, but later cores combined the last two wires into a Sense / Inhibit wire. Each toroidal coil stores one Bits (0 or 1). One bit in each plane can be accessed in one cycle, so each machine word in an array of words is distributed over a bunch of "planes". Each plane will operate on a word in parallel One bit, allowing a complete word to be read or written in one cycle.

The core relies on the "square ring" nature of the ferrite material used to make the ring. A current in a wire passing through the core creates a magnetic field. Only a magnetic field greater than a certain intensity ("selection") can cause the core to change its magnetic polarity. To select a memory location, one of the lines X and Y is driven by half the current ("half select") to cause this change. Only the combined magnetic field (logical AND function) generated at the intersection of the X and Y lines is enough to change the state; other cores can only see half of the required fields ("half-select"), or not at all. By driving a current through a wire in a specific direction, the induced field generated forces the magnetic flux of the selected core to circulate in one direction or the other (clockwise or counterclockwise). One direction is the stored 1 and the other is the stored 0.

The ring shape of the magnetic core is preferable because the magnetic circuit is closed and there are no magnetic poles, so there is little external magnetic flux. This allows the cores to be tightly packed together without their magnetic fields interacting. Alternating 45-degree positioning in the core array helps reduce any spurious coupling.

Core memory read and write

To read some core memory, the circuit attempts to flip the bit to the polarity designated as the 0 state by driving selected X and Y lines that intersect at that core.

1. If the bit is already 0, the physical state of the core is not affected.

2. If the bit was previously 1, the core changes magnetic polarity. After the delay, this change induces a voltage pulse to the Sense line.

The detection of such a pulse means that the bit contains 1. The absence of a pulse means that the bit contains 0. The delay in detecting a voltage pulse is called the core memory access time.

After any such read, this bit contains 0. This explains why core memory accesses are called destructive reads: any operation that reads the core contents erases them and must be recreated immediately.

To write to some core memory, the circuit assumes that there is a read operation and that the bit is in the 0 state.

To write 1 bit, the selected X and Y lines are driven with the current direction opposite to the read operation direction. As with reading, the magnetic core at the intersection of the X and Y lines changes magnetic polarity.

To write 0 bits (in other words, write 1 bit is prohibited), the same amount of current is also sent through the Inhibit line. This reduces the net current through the corresponding core to half the selection current, thereby suppressing changes in polarity.

The access time plus the rewrite time is the memory cycle time.

Other forms of core memory

Word line core memory is typically used to provide register memory. Other names for this type are linear selection and 2-D. This form of core memory typically weaves three lines on a plane through each core, word read, word write, and bit sense / write. To read or clear a word, full current is applied to one or more word read lines; this clears the selected core and any trigger will generate a voltage pulse in its bit read / write line. For reading, usually only one word read line is selected; however, for clarity, multiple word read lines can be selected while ignoring the bit sense / write lines. To write a word, a half current is applied to one or more word write lines, and a half current is applied to each bit sense / write line for a bit to be set. In some designs, read words and word write lines are combined into a single wire, resulting in a memory array with only two wires per bit. For writing, multiple word writing lines can be selected. This provides a performance advantage over the X / Y line coincidence current, as multiple words with the same value can be cleared or written in a single cycle. A typical machine register set typically uses only a small facet of this form of core memory. Some very large memories are built using this technology, such as the Extended Core Storage (ECS) auxiliary memory in the CDC 6600, which has up to 2 million 60-bit words.

Another type of core memory, called core rope memory, provides read-only storage. In this case, a magnetic core with more linear magnetic material is used only as a transformer; no information is actually stored within each core. Every part of the word has a core. Reading the contents of a given memory address generates a current pulse in the wire corresponding to that address. Each address line passes through a core to represent binary, or outside the core to represent binary. As expected, the kernel is much larger than the kernel that reads and writes the kernel. This type of memory is very reliable. One example is the Apollo-guided computer used for moon landings.

Core memory features

The performance of early core memory can be characterized in today's terminology as very similar to a clock rate of 1 MHz (equivalent to home computers in the early 1980s, such as Apple II and Commodore 64). The early core memory system had a cycle time of about 6 s, which had fallen to 1.2 s by the early 1970s and 600 ns (0.6 s) by the mid-1970s. Some designs have higher performance: The memory cycle time of the CDC 6600 in 1964 was 1.0 s, and the core used required a half-select current of 200 mA. Do as much work as possible to reduce access time and increase data rates (bandwidth), including using multiple core grids at the same time, each grid storing one data word. For example, a machine may use a network of 32 cores, each core has a single bit of a 32-bit word, and the controller can access the entire 32-bit word in a single read / write cycle.

Core memory is non-volatile memory-it can retain its contents indefinitely without power. It is also relatively immune to EMP and radiation. These are important advantages of first-generation industrial programmable controllers, military devices and fighter aircraft and other spacecraft applications, and have led to the use of cores for many years after semiconductor MOS memory is available (see also MOSFET). For example, the space shuttle IBM AP-101B flight computer originally used core memory and retained its memory even when the challenger disintegrated in 1986 and subsequently fell into the sea. [16] Another feature of the early core is that the coercive force is very sensitive to temperature; a proper semi-selective current at one temperature is not a proper semi-selective current at another temperature. Therefore, the memory controller will include a temperature sensor (usually a thermistor) to properly adjust the current level for temperature changes. An example of this is the core memory used by Digital Equipment Corporation for its PDP-1 computers; this strategy continues with all subsequent core storage systems built by DEC for its PDP air-cooled computer series. Another way to deal with temperature sensitivity is to "stack" the magnetic core in a temperature controlled oven. Examples of this are the IBM 1620's heated air core memory (which may take 30 minutes to reach operating temperature, approximately 106 ° F (41 ° C)) and the IBM 7090's heated oil bath core memory, early IBM 7094s and IBM 7030.

The core is heated rather than cooled because the main requirement is a consistent temperature and it is easier (and cheaper) to keep the temperature well above room temperature rather than equal to or below room temperature.

In 1980, the price of a 16 kW (kw, 32 kB equivalent) core memory board installed in a DEC Q-bus computer was about $ 3,000. At that time, the core array and supporting electronic components were mounted on a single printed circuit board with a size of about 25 × 20 cm. The core array was mounted a few millimeters above the PCB and protected with a metal or plastic board ^[1] .

Diagnosing hardware problems in core memory requires running time-consuming diagnostic programs. Although a quick test checks whether each bit contains 1s and 0s, these diagnostics test the core memory using worst-case mode and must run for several hours. Since most computers have only one core memory board, these diagnostic programs also move in memory, so you can test every bit. The advanced test is called the "Schmoo test", in which the half-select current is modified with the test time ("gating") of the test sense line. The data chart for this test looks like a cartoon character named "Schmoo" with a stuck name. In many cases, errors can be resolved by gently tapping a printed circuit board with a core array on the table. This slightly changes the position of the cores along the wires running through them and can solve the problem. This process was rarely needed because core memory proved to be very reliable compared to other computer components of the time.