Large Scale Integration; Memories

There are a number of types of memory which can be used as ICs in digital electronics. These include

a. Random-Access Memory (RAM), wherein each memory word is accessed for reading or writing via a specific address, ac­cess time being approximately equal for any combination of suc­cessive locations.

b. Serial Access Memory (SAM), wherein the memory con­sists of a circular shift register (serial output connected to serial input). A counter keeps track of the "address" of the bit available for reading and writing (the serial output and input bits). To read or write a given address, the register is shifted until the counter matches the desired address: clearly a large change of address takes longer than a short one.

c. Read-Only Memory (ROM), in which the binary contents are wired in at the factory as a step in the 1C manufacturing pro­cedure. These act like PAMs, except it is not possible to change the contents. Although it is possible to obtain custom-designed ROMs, they are too expensive for production in small quantities.

d. Programmable ROMs (PROMs), which can be written us­ing special equipment. These hold their contents until erased with high-intensity ultraviolet light and re-programmed.

e. Programmed Logic Arrays (PLAs), some of which can also be programmed, do not have a full-scale memory complement, but are an expensive way of making a ROM-like device, in which not all inputs codes correspond to defined outputs and a given output can be specified by more than one input code.

RAMs are used for temporary data storage because they are volatile: that is their contents are lost if power is removed. ROMs, PROMs, and PLAs are non-volatile, but cannot be written on during normal operation.

A number of new memory types have recently appeared. We can expect corelike RAMs to become available in the near future.

RAMs, being used for temporary data storage, are good "scratch pads" for digital devices; they are used as computer mem­ories for the full range of computer sizes, often in a mixture of ROM, RAM, and core memory.

ROMs and PROMs are used for permanent storage, such as the programs in microcomputers, and start-up programs in larger machines. They are also used to sequence sequential machines from one state to the next, and they are very useful for data con­version, table lookup (trigonometric tables, for example), and generation of complex logical functions. A PROM is used to test a new memory content: if it is correct, a ROM is manufactured with the same content if the number of devices or speed requirements (ROMs are faster) justify the expense; PROMs are used for slower devices produced in smaller quantities.

Учитесь говорить.

6.17. Прочитайте текст. Используйте информацию тек­ста для беседы на тему «Надежность».

What mostly affects system performance is reliability: while increased speed may provide 5% more throughput (производи­тельность) increased reliability significantly affects the system output. System reliability can be quantified by MTBF (mean-time between failures), which is the reciprocal (обратная величина) of the product of the device failure rate and the number of com­ponents.

In semiconductor memories there are two types of failure mechanisms. The first is a hard error in which the device structure fails. The second, a soft error, is a random, non-recurring error caused by alpha particles.

As memories become denser, their storage area becomes smaller. As a result they can become more sensitive to soft errors.

6.18. Обсудите следующие темы:

1. Core memory. 2. Semiconductor memory. 3. Charge-cou­pled devices. 4. Magnetic bubble devices. 5. Electron beam-ad­dressed memories.

МАТЕРИАЛЫ ДЛЯ САМОСТОЯТЕЛЬНОЙ ВНЕАУДИТОРНОЙ РАБОТЫ

(ПОСЛЕ ВТОРОГО ЗАНЯТИЯ)

Учитесь читать и переводить.

Текст 6.8. Просмотрите текст. Озаглавьте его. Аргу­ментируйте свой выбор заголовка. Прочитайте текст еще раз. Подготовьте сообщение о разных типах памяти.

The electron beam is an addressing pointer of high definition and energy density that can easily be deflected. In storage tubes of the 1940's there were severe limitations to such addressing be­cause of the use of surface charge storage and inadequacies in fo­cusing and deflecting the beam. Two recent innovations, storage within a semiconductor and compounded deflection, may bring us closer to realizing the inherent potential of beam addressing.

The addressing is in two parts. First, the beam is deflected by a short conical structure of low aberration and strikes normally one of the appertures of a matrix of lenslets.

The matrix is made up of two metal plates that have an array of holes (an 18 by 18 array on 1.5-mm centres) and are maintained at different potentials. Second, the beam is deflected by bars run­ning along rows and columns between the holes of the matrix. No matter which lenslet is reached, the reduced beam will be sub­jected to the second deflection. In this compounded deflection the accuracy and stability at each step need only be a small fraction of what would be required with a single step.

Текст 6.9. Прочитайте текст и сделайте обобщение ин­формации об особенностях кэш-памяти.