Figure 6.10 Using a shift register
As a button is pressed the voltage output can be designed to change from logic 0 to logic 1 so to start with, we can assume no buttons are pressed and the response from each button is zero. Along comes a customer who, having read the instructions, inserted some money and re-read the instructions and stared at the buttons, eventually decides to press a button.
Pressing a button generates a burst of eight clock pulses and the value of each button is loaded into the shift register. Once the button has been pressed the zeros and ones corresponding to each of the buttons is loaded into the shift register. The output from each button is made available to external circuits and one such circuit will be activated and a ticket will drop down the chute.
For how long would the customer have to press the button?
The microprocessor is amazingly fast compared with us. If we feel the temperature of a piece of metal and it is too hot, we immediately take our hand off. But how long did this take? For most people the time to think and then respond would be about one-tenth of a second. In sport, it means that the person at the receiving end must use body movement or magic to predict what is going to happen. Waiting to respond to the flight of the ball will make them too late. Most people would therefore press a button for at least 0.1 s. So what can the microprocessor do in the same 0.1 s? A modern microprocessor can check a button in about 0.25 µs or 0.25 millionths of a second. In other words, it can check about 4 000 000 buttons in a second.
We have a best response time of 0.1 s. The microprocessor has a response time of about 0.25 µs. This means than the microprocessor lives at a speed of about 400 000 times faster than us. Can you imagine how we would feel faced with a creature called a ‘Waitabit’ that moves 400 000 times slower than us? It would take 11 hours to press the button. After all that effort, it may run off at 3 cm/h (1.2 in/h) to spend 11 years having a cup of coffee. By way of compensation, it may well live for 28 million years!
These are modified versions of the shift registers. There are only two simple changes necessary. The first is that the data is loaded in parallel.
This means that the data is loaded into each flip-flop in the register at the same time. This requires a separate connection to carry each bit but the good news is all the data is loaded under the control of a single clock pulse so it is very much faster. Once loaded, subsequent clock pulses cause the data to be moved along the register as before. The last bit of data is connected back to the other end of the register instead of dropping off the end into our bin. Have a look at Figures 6.11 and 6.12.
Figure 6.11 A rotate-left register
Figure 6.12 Data movement in a rotate-left register
As with shift registers, rotate registers can be made in rotate right as well as left versions. In microprocessors, the same register can be used to rotate or shift in either direction.
The benefit of using a rotate rather than a shift register is that the data is not destroyed. We have seen that a shift register is progressively emptied as bits fall into the bin at the end. With a rotate register, the data is not changed. If we rotate left say, six times, we only have to rotate right six times to recover all the original data.
The function of a memory is to store information – almost the same as we said for the register. Generally, a register lives within the microprocessor and stores small quantities of data for immediate use and it can do useful little tricks like shift and rotate. A memory is designed for bulk storage of data but that is all it can do – no tricks this time.
Well, almost no tricks – some types can remember the data even when the power is switched off. The ability to remember data after the power is switched off is the dividing line between the two main types of memory. If it loses its data when the power is switched off, then we call the memory RAM or volatile memory. If it can hold on to the data without power, we call it ROM or non-volatile memory (volatile means ‘able to evaporate’). This is seen in Figure 6.13.
Figure 6.13 The two classes of memory
The letters RAM stands for Random Access Memory which is a silly, out-of-date, name. It should be called read/write memory or RWM but it is so difficult to get something to change once it is established. Anyway, let’s leave the name for the moment and look at the memory.
The memory comes in an integrated circuit looking like a small microprocessor and is usually called a memory chip. Inside, there are a large number of registers, hundreds, thousands, millions depending on the size of the memory. Incidentally, when we are referring to memories, we use the word ‘cell’ instead of register even though they are the same thing.
So, each of the internal cells may have 4, 8, 16, 32, or 64 bits stored in flip-flops. Figure 6.14 shows the register layout in a very small memory containing only 16 cells or locations, each of which can hold 4 bits and is given a memory number or address.
Figure 6.14 The layout of cells in a memory
This RAM word
In prehistoric computing days, the memory would be loaded in order. The first group of bits would go into location 0, the next would go into location 2, then location 3 and so on rather like a shift register. This meant that the time to load or recover the information would increase as we started to fill the memory and have to move further down the memory. This was called sequential access memory (or serial access memory), abbreviated to SAM. This was OK when a large computer may hold 256 bits of information but would be impossibly slow if we tried this trick with a gigabyte.
To overcome this problem, we developed a way to access any memory location in the same amount of time regardless of where in the memory it happens to be stored. This system was called random access memory or RAM.
All memory, whether volatile or non-volatile is now designed as random access memory so it would be much better to divide the two types of memory into read/write and read only memory. But it won’t happen, RAM is too firmly entrenched.
Each location in a memory is given a number, called an address. In Figure 6.14, the 16 locations of memory would be numbered from 0 to 15, or in binary 0000–11112. The cells are formed into a rectangular layout, in this case a 4×4 square with four columns and four rows.
To use a cell, the row containing the cell must be selected and the column containing the cell must also be activated. The shaded cell in Figure 6.15 has the address 0110 which means that it is in row 01 and in column 10.
Figure 6.15 Selecting a memory location
To access this cell we need to apply the binary address to the row and column decoders. When the address 0110 is applied, the first half of the address, 01, is applied to the row decoder and the second half of the address is applied to the column decoder. A decoder circuit is a small logic circuit that, when fed with the address of the location, is able to switch on the appropriate row and column. The maximum number of locations that can be addressed will depend on the number of bits in the address. We have already seen that a 4-bit address can access 16 locations. This was because 24=16, so, generally 2n= number of locations where n is the number of bits in the address. To take a more realistic example, if we had 20 address lines we would have 220=1 048 576 or 1 Meg locations.