The main requirements of semiconductor memories are that they occupy a small area, have a fast access time and operate with low power consumption. In addition they are non-volatile.
Semiconductor memories are available in integrated circuits (IC's). Basically there are two types of IC's bipolar and unipolar. In bipolar circuits charge carriers are both positive and negative (holes and electrons) are required for operation of the active elements. Whereas, in the unipolar case, either the holes or the electrons are required for the operation of the active elements.
Bipolar techniques offer fast access time at low packing density, whereas unipolar technology give a slower access time at much higher packing density. The unipolar memories are MOS or CMOS where as the bipolar memories are TTL, ECL or Schottky TTL. The slower speed of unipolar devices is due to parasitic capacitance associated with the bulk silicon substrate. Advance technologies like SOS (silicon on sapphire), NMOS and VMOS are gradually overcoming these difficulties.
In semiconductor memories, a static memory is one in which the stored information is maintained as long as the supply in ON whereas a dynamic memory is one in which the information is retained as a charge on a capacitor and i periodically subjected to a refresh cycle to compensate for the leakage of charge from the capacitor.
RAM is an array of storage cells. These cells are flip-flop. In some of these memories a word may be made up of say one row of flip-flops, which may be addressed simultaneously. Bipolar flip-flops have and access time of about 35 ns and one package may contain about 1 k cells whereas MOS devices have an access time of 400 ns but are large in size. The RAM's may be in static or dynamic mode.
Figure 1 shows bipolar static RAM cells respectively. Such cells are grouped together and are addressed by MAR. When particular X and Y select lines are at a low voltage, the current flows through transistor Q1 or Q2 (depending on the information stored) to the line which is at the low voltage. When a particular cell is selected the corresponding X and Y selected lines are raised to a high potential.
The current is thus divided from the selected lines to the data lines. The current flowing in the two data lines is unequal giving an indication of the cells state. To write in to a cell, the X and Y select lines must again be raised to a high potential, and the data lines held at a high or low potential to turn the desired transistor ON.
Dynamic RAM's are available in unipolar and bipolar configuration. These have large packing than static RAM's. These systems are faster and dissipate less power but have the disadvantage of the necessity of the refreshing cycle to overcome leakage problem. A simple MOS dynamic RAM is shown in Figure 2. The charge is stored in capacitor C1. If the transistor Q2 is turned ON, the information is refreshed (or changed). The cell is read by connecting the read data line to a negative potential (logic 1), then Q3 is turned ON. If a logic 1 is stored in C1 this turns transistor Q1 ON which causes the data line to be discharged. If a logic 0 was stored in C1, then the data line remains in logic 1. At the end of this read cycle, the data on the read line is the complement of the data stored in the cell. This is corrected by feeding the data back in through the write data line in order to refresh the cell after every read cycle.
A ROM (read only memory) is a memory device which is required to store information which is not likely to be changed as often as in a RAM. It is a low cost, high speed, non-volatile memory and is made up of similar arrays as in RAM's. A ROM can be used to realize an arbitrary truth tables, generate characters, convert codes or store editing or monitor programs. Several different forms of ROM's are available. The ROM's are also based on bipolar or unipolar devices. The classification of ROM's are shown in figure 3. The mask ROM's are factory programmed ROM's which are programmed to a standard specification. The PROM (programmable ROM) is supplied in blank form and the customer programs it according to his requirements.
There are two basic forms of PROM, the fusible link, which once programmed is permanent i.e. it becomes a ROM, and the ultra violet light erasable ROM (EPROM) in which the program can be completely wiped clean by exposing to strong UV light usually through a glass window in the package, then it is programmed. Finally there is a class known as RMM (read mostly memory) or electrically alterable ROM (EAROM), in which data is erased by applying a high voltage pulse to the programming pins. EAROM's has an advantage over EPROM in the sense that single word in EAROM can be erased and rewritten without affecting the rest of the content.
A read only memories (ROM's) are constructed with the combination of logic circuits. A simple decodes followed by an encoder makes a ROM as shown in figure 4.
The decoder is a logic circuit which accepts an n-bit word and establishes the state 1 on one and only one of 2n output lines. These 2n output lines are fed to the encoder which generats output at its K output lines.
The basic structure of ROM is an decoder followed by an encoder. THe arrangement of an 8 words x 4 bits ROM are shown in figure 5. There are three input or address lines. These address lines select one of the 8 output lines and depending upon the diode connections, the four output lines O0, O1, O2 and O3 will give the output. As an example, if the address bit are 011, the output will be 0011 because only two diodes are connected between line 011 and O2 and O3 lines. The diode means a 1 and no diode means a 0. The diodes may be replaced by transistor or MOSFET's.
Charged Coupled Devices (CCD)
Charge coupled devices store bit of information as packets of electric charge. These packets of charge are moved by an alternating electric field. This charge would eventually dissipate owing to small unavailable losses. The data can be recirculated by generating to fresh charge packets by the help of voltage to charge converts incorporated in the CCD's.