Microprocessors Lecture 7

Input and Output Interfaces

There is absolutely no point in having a computer at all unless there is some mechanism for entering data into the computer and outputting data from it. Obviously the most usual device is the VDU with its associated keyboard. Any source of information, however, can be connected to a computer, provided a suitable interface can be designed, from a simple two position switch to data from a satellite or the human voice. Similarly anything can be controlled by a computer from a simple LED to the position of a satellite or a robotic replacement arm or the national grid. All we need is a suitable interface.

What is an interface?

It is a system, usually electronic but may also contain mechanical parts which match the requirements and characteristics of the computer to those of the external device. What are these requirements and characteristics?
  1. Type of signals - Computers (at least Digital computers as opposed to Analogue computers) use digital signals throughout. If we wish to connect an analogue device then a D to A or A to D converter is needed.
  2. Electrical characteristics - The signals must correspond with the logic signals used in the computer system normally 0v for a logic zero and 5v for a logic one. Remember this is not necessarily true.
  3. Loading - Not only must the voltages be correct but the loading of an output device must be correct. It is perfectly possible for a device to require 5v to work but to take a current of hundreds of amps (Industrial Electrolysis equipment for example).
  4. Timing and Latching - The processor will normally input data by taking it from the data bus. As we have seen the timing of this operation is quite critical and it certainly no good for the device to put data onto the bus continually otherwise the processor would not be able to read instructions from memory. Similarly data is output by the processor on the data bus in one small part of the instruction cycle. Most devices will need the data to be present for rather longer. The signal must therefore be latched by the interface.
  5. Speed - Some devices require data at a precisely fixed speed. For example a disc drive has to write data onto the disc at the correct speed as the disc passes under the head and a sychronous communications link must supply data at exactly the speed of the link. The computer may not be able to read or write data at the required speed so an interface with a buffer memory would be needed. Also as we shall see when we deal with interrupts it is often not possible to guarantee that a processor will respond within a fixed time even if it can maintain the average speed required.
  6. Flagging - Some deviced generate data continuously (e.g. switches) but others generate data for a short time which is sometimes not even regular (A to D converter). It is the job of the interface to tell the processor when data is available by means of flags.

An example is given below of an interface using simple latches; normally we will not construct ad-hoc interfaces, but use a purpose-built interface chip such as a peripheral interface adapter (PIA or VIA) for parallel data. One such device, a VIA, is described at the end of these notes and again in the following lecture notes.

Simple example

Suppose we wish to connect a single switch to a computer so that its value (on or off) can be read at will. Obviously we can't just connect one side of the switch onto one line of the data bus and the other to 0v. If we did then when the switch is closed a continuous logic 0 would be put on the line and the processor would not work and when it is open then the processor works but any attempt to read the value on the switch results in a floating data bus line which might be one and might be zero.

We have two problems logic signals and timing. The logic problem is easily overcome by adding a resistor to +5v. When the switch is open the reistor pulls the output to logic 1 when closed the output is forced to logic 0.

The timing problem requires a little more thought. First we need an output which can be switched on and off - this means a tri-state logic gate or buffer. Then we need to determine when the data should be put on the bus. Presumably we want the switch to be at some memory location not used by other memory. Suppose we are not using the range $8000 to $BFFF (we might have RAM below $8000 and ROM from $C000 upwards). We could therefore choose one of these free locations $8010 for example. This means we want our interface to respond with the switch data when

  1. The correct address ($8010) is on the address bus.
  2. The processor is executing a read cycle (The programmer may make a mistake and try to write to this address - we don't want the interface and processor to output data at the same time).
  3. The E (Enable) signal is high to ensure correct timing.
To do this we need to AND all these events together.
and feed the output of this 18 input AND gate into the enable on a tri-state buffer. This would actually work if you can find an 18 input AND gate. Unfortunately most ICs of AND gates only have 14 pins and since we need 2 for power and 1 for an output we only have 11 left for inputs.

There are two solutions to this problem which are usually used in conjunction with each other.

1. Incomplete decoding. The process of selecting the appropriate address is known as address decoding and what we have done here is to completely decode the address. I.e. if any other address appears on the address bus then the interface doesn't respond. But if we look at our system the entire range of addresses from $8000 to $CFFF is unused so we could in theory allow the switch' interface to respond to any or in fact all of these addresses. All we have to do is to make sure that it doesn't respond to addresses outside that range. If we look at how the address space is divided up we see that it is controlled by A15 and A14. 

                   A15 A14    Address range        Devices addressed
                    0    0    $0000 - $3FFF               RAM
                    0    1    $4000 - $7FFF               RAM
                    1    0    $8000 - $BFFF            Interface
                    1    1    $C000 - $FFFF               ROM
All we have to do, therefore is to detect when A15 is 1 AND A14 is zero. The inputs to our AND gate are now only four.
The penalty we have paid for this simplification is that every address in the range now addresses our switch. This is not a problem but it does mean we have no address space left for any other interface and we are going to need at least one output interface for the system to perform any useful task at all.

The solution is to add a few more address inputs to our AND gate but not to decode all of them. For example we might add A0,A1 and A4 so that the interface responds to $8010 as originally intended.

It will also respond to $8030 and to $8050 and $8014 and many others but it will not respond to $8011 nor $8012 nor $8013 nor $8000 so we have created plenty of room for other interfaces.

2. Address Paging - The other approach, which as I have already said is often used in conjunction with incomplete decoding is address paging. Here the decoding is done in stages. We might, for example, first create four pages of $4000 locations by decoding A15 and A14

We might then break down each of these large pages into smaller pages of $1000 locations
Each of these sub-pages could be broken down into smaller pages and so on until we end up with individual locations. The snag with this approach if taken to extremes is that we have a large number of decoders in series and therefore more time will be required for address changes to propagate through.

In practice we use a combination and in this particular example we might take Interface Enable with addresses A4, A1 and A0 and of course R into a 5 input AND gate which is a fairly normal size.

We have looked at a very simple example of an input interface and seen how in this case we can solve the problems of timing and electrical compatibility. Obviously it we wish to extend the interface so that up to eight switches are used then we simply add more tri-state buffers and use the remaining data bus lines.

A simple output interface

If we wish to output data on a set of lights or LEDs we have similar problems of timing and compatibility but also the additional problem of latching the data. The processor will only put the data on the data bus for a small part of one cycle, say 400 microsecs. If we actually want to see the data it needs to be present for rather longer, say seconds. In other words the interface needs to store the data and output it continuously to the displays. For this we will need a register of some sort which we can load from the data bus at the correct instant. To follow on from our previous example we might want this device to be at memory location $8011. We can generate a suitable'timing signal by ANDing Interface Enable with A4, A1 and A0. Since we only want the device to respond when data is being output by the processor we will have to use the inverse of the R/W signal. The timing signal now loads data into the register where it remains until new data is written, and of course it appears on the output lines. The register will only have a normal TTL drive capability and so, depending on the size of the LEDs or lamps we will need additional buffers or drivers to supply the current. These might be simple transistor circuits as shown above or, for mains driven lights, thyristor controllers.

General purpose interfaces

The problem of designing interfaces is one which faces every computer system designer and frequently the problems are identical and fall into general categories. The interfaces we have looked at are parallel interfaces, that is, the data to be input or output is continously available on a number of parallel connections. Such an interface can drive lamps and switches but can also drive more complex devices such as a printer, grahics input device, analogue devices via D to A or A to D convertors, stepper motors, pressure transducers, valve actuators, 7-segment displays etc. These are all devices which can be positioned near the computer and so the existence of many connections is not a serious problem. If a device is to be located at a distance then serial data is preferable as the cost of wiring is much reduced.

Parallel Interface

The obvious answer when many applications require a similar solution is to design an integrated circuit to perfom the task. The costs involved in designing such a crcuit are of course very high and so it should be as versatile as possible so that the potential market is large and the returns on the design investment as great as possible. When you come to look at the problem you find that all applications are different requiring different numbers of inputs and outputs, some using 8 bit data, some 12 and some 16-bit. An almost universal approach has been to provide two 8-bit parallel interfaces on a single IC.

The PIA (Peripheral Interface Adapter) from Motorola and the VIA (Versatile Interface Adapter) from Rockwell are typical and popular examples sharing many common features. Each has these two 8-bit parallel interfaces which occupy consecutive locations in memory so that they can if necessary be used in conjunction to provide up to sixteen bits. (Thus a LDD instruction loads the data from both interfaces into D = A:B). Additionally each of the 16 interface connections can be programmed to be either an input or an output connection thus catering for control applications where each line performs a different function (controlling an open or shut valve or detecting the presence or absence of a component on a conveyor belt). This programming is performed as we have already seen using two further registers in the interface. In the case of the VIA these are organised as follows:- 

                    Address n       Interface Register B
                            n+1     Interface Register A
                            n+2     Data Direction Register B
                            n+3     Data Direction Register A
Each bit of the Data Direction register controls the direction of one of the interface connections. Thus DDRA0 controls the direction of data on line PRA0 and DDRA2 on PRA2 etc.

The address at which the whole interface appears is determined by address decoding as before and also by two Chip Select inputs CS1 and CS2. The first is active high and the second active low which means that "10" on these is required to enable the chip. The advantage of this is that by suitably juggling the address connections to these pins it may be possible in small systems to avoid any additional address decoding. For example connecting A15 to CS1 and A14 to CS2 puts the interface at memory location $8000 (but note that the decoding is incomplete, which means that the interface will also appear at many other addresses. 

           A15 A14 A13 A12 A11 A10 A9  A8  A7  A6  A5  A4  A3  A2  A1  A0
            1   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
           CS1 CS2                                            RS3 RS2 RS1 RS0
The versatility of these devices does not end there. If a sequence of data is to be transfered to or from an external device then some sort of handshake mechanism is required to ensure that data is received correctly. For example if a page of text is to be sent to a printer then the processor is capable of sending a new data byte every few instructions by stepping through memory using the index register. Thus a new byte can be sent every 10 microsecs. A printer can typically print at 100 characters per sec or one character every 10 millisecs. Thus there is a difference in speed of 1000 or if the processor is sending data continuously then 1 character in 1000 will be printed correctly. To overcome this we use two handshaking signals which have a variety of names the most common of which are 'Data Available' and 'Data Accepted'. 
          Data Available      This signal is used to indicate that a new
                              data byte is available on the interface lines
                              which requires printing.

          Data Accepted       This indicates to the processor that the
                              printer has printed the current data byte (or
                              at least read the byte into a buffer ready
                              for printing) and is ready for the next one.
The sequence of signals is as follows:-
1. The processor sets up the first data byte on the interface lines.
2. The processor asserts Data Available.
3. The printer reads the data byte.
4. The printer asserts Data Accepted.
5. The processor removes Data Available.
6. The printer removes Data Accepted.
7. The process repeats with the next data byte.

The handshaking ensures that the processor does not change the data byte before the printer has read it and that the printer does not read data which is incorrect. Although there are many variations on this it can always be carried out using two signals, one from the sending to the receiving device and the other going in the opposite direction.

To enable handshaking to take place the VIA (or PIA) has two Control Signals for each of the two interfaces - these are labelled CAI and CA2 for A interface and CB1 and CB2 for B interface. CA1 and CB1 can only be used as inputs but CA2 and CB2 can be used as outputs (for handshaking) or as inputs (for applications not requiring handshaking).

The operation of the control lines is controlled by a further register in the interface - the Peripheral Control Register, which resides in memory location n + C. 

                     Address n + C     Peripheral Control Register
The processor is not normally concerned with the actual logic level on the input control line. What it needs to know is when a transition has occurred on the line which then initiates some action (outputting the next data byte for example). The interface therefore detects transistions and when such a transition occurs a flag is set which the processor can detect. In some applications the transition being looked for is the 1 -> 0 transition and in some the 0 -> 1. Sometimes both may be important, The control register has a bit which determines which transition is detected. Another bit does the same thing for CA2 (CB2) when it is used as an input. Obviously there must also be a bit which determines whether CA2 (CB2) is used as an input or output and when programmed as an output another bit must determine what value that output takes up.
Back