Semiconductor integrated memory

Introduction

Semiconductor Integrated Memory

SemiconductorMemory

There are two different characterizations of each storage unit. State "0" and "1" to store different information. Semiconductor memory is an important part that constitutes a computer. Compared with the magnetic memory, the semiconductor memory has the advantages of fast access speed, large storage capacity, small volume, and the memory cell array and the main peripheral logic circuitry, can be made on the same chip, so that the input and output interface is greatly simplified. Therefore, in the computer high speed storage, the semiconductor memory has replaced the past magnetic memory.

Main Advantages

This main advantage of this memory is:

1 storage unit array and major peripheral logic circuits are made on the same silicon chip, output and The input level can be compatible and matching the same surface. This makes the interface between the computer operation and control and the two parts of the storage of the two parts of the data; the deposit and reading speed of the data is more than three orders of magnitude than the magnetic memory, which can greatly improve the computer. The calculation speed;

3 utilizes a large-capacity semiconductor memory to make the volume and cost of the storage body sharply and decrease. Therefore, in the computer high speed storage, the semiconductor memory has all replaced the past magnetic memory. The semiconductor memory used as a large-scale integrated circuit is 1 k-bit dynamic random memory that starts production before and after 1970. With the improvement of process technology, in 1984, this type of product has reached 1 megabit storage capacity.

Classification

According to the functionality, the semiconductor memory can be divided into three categoriescent storage (RAM), read-only memory (ROM), and serial storage. With the development of the semiconductor integrated circuit process technology, the capacity of semiconductor memory is very fast, and the single-piece storage capacity has entered a mega level, such as 16 megaphoelectric random memory (DRAM) has been commercialized, 64 trillion, 256 mega DRAM in development. .

Random Memory

For any of the addresses, the memory (write speed, and readout speed) of the write data are randomly read and written at the same speed. The internal structure of the memory cell is generally in the form of a two-dimensional square matrix, that is, an address of an address (such as 64K × 1). But sometimes there is also a form (such as 8K × 8) to facilitate multi-bit output. The random memory is mainly used to form a system that requires a fast storage such as a computer main memory.

Different in operation, the random memory can be divided into static and dynamic. The unit circuit of the static random memory is a trigger. When one of the two transistors of A or B, "1" or "0" can be specified. The trigger is not changed as long as the power supply is high enough. Therefore, the information of each unit is stored, if not "forced" is changed, as long as there is sufficiently high supply voltage, there is no need to refresh (see metal-oxide-semiconductor static random memory). This memory is fast and easy to use. The unit of dynamic random memory is composed of a MOS capacitor and a MOS transistor. The data is stored in the capacitance in charge form, which is generally not charged "0", and there is a charge represent "1", and vice versa. The MOS transistor in the unit is a switch that controls the deposit and removal of the charge in the storage capacitor. Typically, the MOS capacitors and PNs coupled thereto have a weak leakage, and the charge becomes less, until the leak, the deposited data will be lost. Therefore, the dynamic random memory needs to rewrite the information stored in unit circuit every 2 to 4 milliseconds, which is referred to as refresh. Such a memory is characterized by a small number of units, high integration, and is the most widely used (see metal-oxide-semiconductor dynamic random memory).

read-only memory

is used to store long-term fixed data or information, such as various functions, characters, and fixed programs. Its unit has only one diode or triode. Generally, it is "1" when the device is turned on, it is "0" when the device is turned off, and vice versa. If you design a read-only memory mask version, the data is written in the mask graphic, and the lithography is transferred to the silicon chip. This is known as a mask read-only memory. After this memory is loaded into the whole machine, the user can only read the stored data without writing data. The advantage is suitable for mass production. However, the whole machine is in the commissioning phase, often needs to modify the content of the read-only memory, time-consuming, abundance, very flexible (see semiconductor read-only memory).

Serial Memory

It is arranged in a one-dimensional structure as a tape. The read time of the tail portion is very long, because the entire tape is passed in order. The cells in the semiconductor serial memory are also one-dimensional arrangement, and the data is read in each column, such as shift registers and charge coupling memory, and the like.

Different by manufacturing process technology, the semiconductor memory can be divided into two categories of MOS-type memory and bipolar memory. Since 1970s, NMOS circuits (see N-channel metal-oxide-semiconductor integrated circuits) and CMOS circuits (see complementary metallic-oxide-semiconductor integrated circuits) have the fastest development, and they can be used in extreme integration. The various semiconductor integrated memory of the degree. The read time of arsenide semiconductor memory such as 1024-bit static random memory has reached 2 milliseconds, and it is expected to develop in the super high speed.