Back

More on real-time operating systems

Introduction
According to the British Computing Society's 'A Glossary of Computing Terms', a real-time system is "one which can react fast enough to influence behaviour in the outside world". A key part of the above definition is the phrase 'fast enough'. Many real-time systems have outputs that change in fractions of a second as a result of changes to input information combined with the processing of that information. Examples of this include missile guidance systems and plane control systems. Some real-time systems, however, do not happen in fractions of a second. A plane ticket booking systems is an example of one such system. 

Another key part of the definition is the phrase 'influence behaviour'. It is important in real-time systems that the processing of data from the input sensors of a system environment actually influences the outputs before the environment in which the system operates has changed again of its own accord! For example, suppose you have a greenhouse with a ventilation window that can be opened and closed by various amounts, depending upon the temperature in the greenhouse. A temperature sensor is used to read the temperature in the greenhouse. Let's suppose it was very cold. This information is fed into a computer. The computer processes the fact that it is very cold and sends signals to the motor to close the window. It would be a very poor system if, by the time the window was actually closed, there was a sudden heat wave and it was really hot in the greenhouse! Outputs need to be changed quickly enough to affect the environment in which the system operates. 

Designing any real-time system - Input - Process - Output
When considering the design of any real-time system and the selection of sensors and actuators, it is important to try to picture the whole system and ask some questions: 

      • What data from the outside world do I need to monitor?
      • How will I get the data from the outside world to the processor? (What sensors do I need?)
      • Do I need to change the signal type of the sensor so it matches what the computer can handle? In other words, do I need an interface for the sensors?
      • What will I actually do with the data, what decisions need to be made?
      • What do I want to control and when? (This is the program you need to write!)
      • How will I achieve control? (What actuators do I need?)
      • Do I need to change the signal from the computer's type to what the actuators can handle? In other words, do I need an interface for any of the actuators?
      • How can the system be started and stopped? 

A plane booking system
Three different companies of travel agents sell plane tickets for a sightseeing flight around London. The plane is small, with only six seats available. The three companies use a central database to keep details of what seats are available.

real1

If the first customer comes in to Exotica and buys two seats then that information must be sent to the central database (the input) and the database must be updated (the processing) before anyone else can be allowed to buy tickets. This is so that all the agents can see exactly how many seats are left (the output). If this didn't happen, then a family of five could easily go into Sunshine Hols and buy five tickets. This would result in the plane being over-booked. This is an example of a real-time system. The inputs into the system are the booking details, the processing is the taking of the booking information and updating the database and the output is the updated database and tickets. Whenever a booking is made, it influences the database quickly enough so that over-booking is not possible. Booking systems are examples of real-time systems.
 
A missile guidance system
When a missile is fired at a moving target, the sensors will be feeding data into the missile's processor (the input). The software program will be processing the data it gets from the input and making decisions about adjustments to the direction of the missile (the processing). The processor then sends signals to the fins and motors of the missile that causes the direction of the missile to change (the output). This is another example of a real-time system, one that is happening very quickly indeed compared to the real-time ticket booking system! Note also in this example the use of feedback. Feedback is when the sensors feed back the current situation, in this example, some co-ordinates. The actual value is compared to the desired value and any adjustments in the output are made. Feedback is very common in control systems where a 'target' or 'desired value' is needed.

Features of real-time systems
Most control systems are time-critical. They need things to happen quickly. They also need things to happen reliably. So when, for example, a pilot steers a plane to the left, the plane should move to the left, quickly and every time! Applications such as these are written in purpose-designed software languages. ADA is one such example. It is a language that has its roots in Pascal. It has features that make it suitable for real-time applications. For example, it can deal with concurrent events and can be used in parallel processing architectures. Parallel processing might be needed if there is a lot of data and it cannot be processed within the required time by one processor. By using parallel processors, you can split the program up and run different parts of it at the same time. You can also use parallel processing to build into the system some backup systems (sometimes called ‘fail-safe systems’), in case the main one fails. It is possible to identify some features of real-time software as used in applications such as a plane's control systems.

    • The output times are predictable within specified limits. This is important. When you use your computer at school and send something to print or start multi-tasking, then the speed that your computer will work at slows and slows unpredictably. In many real-time systems, it is important to be able to specify that XXX will happen within YYY milliseconds, regardless of whatever else is happening in the system, regardless of any interrupts that might occur.
    • Backup systems will have been incorporated into both the software and hardware. If ever something fails in the system, there will be a backup system ready to take over immediately.
    • The software is very fast! It will have been designed and written so that it runs at the optimum speed.

Transducers
According to the British Computing Society "A Glossary of Computing Terms", a transducer is "an electronic component which converts one form of energy to another". Some examples of sensors are described in the next section.

Sensors
Sensors are devices that respond to a specific physical property in a system. They are transducers in that they take one form of energy and convert it into another. Here are some examples.

    • A thermistor is a temperature sensor (or temperature transducer). It converts a temperature into an analogue voltage. The voltage generated depends upon the temperature. There are an infinite number of temperatures and so there are an infinite number of voltages. Devices that generate an infinite range of values are known as 'analogue' devices. You could use a thermistor in a chemical plant to read the temperature of a reaction. The processor could convert the data from the transducer into a graph. Operators could then see the history of the temperature of a reaction.
    • A photocell is a light sensor (light transducer). It converts light energy into an analogue voltage. Light meters use photocells. A photographer can set up a camera according to the light available, as indicated on a light meter.
    • A pressure sensor (pressure transducer) in a chemical reaction vessel converts pressure into an analogue signal.
    • A pressure mat in a burglar alarm is a sensor. It senses whether someone is either on the mat or off it. It has two possible states: on and off so it is known as a digital device.
    • A switch or push-button is a digital sensor. It detects one of two positions: on and off. Switches can be used to send signals to a system. For example, an operator in a factory might start a system by pressing the 'start' button. This signal is fed into the computer, which processes the signal in its program.
    • A proximity sensor is a digital sensor. When something gets too close to it, it generates a voltage. When there is nothing next to it, it generates no voltage. Proximity sensors, for example, can be used to signal to a lift control system that a particular floor is coming up. The program can then order the motor to slow down.

Interfaces
Computers are digital devices. They can only use digital signals. A computer might use, for example the two signals 5 volts and 0 volts. Any signals going into the computer must be digital, use the same voltage the computer uses and be able to connect to the ports used by the computer. Suppose a proximity sensor, a digital device, uses 5 Volts but the wire carrying the signal is connected to a plug that can't be connected to the port at the back of the computer! Suppose a pressure mat uses 12 volts. How can this be connected up to a computer that can only take 5 Volts? How can you connect an analogue temperature sensor, with lots of possible voltages being generated, to a computer that can only accept two different voltages? You need an interface! 

real2

 

An interface converts analogue signals into digital (or vice versa). It converts signals from one voltage level to the computer's voltage level. It provides a means of ensuring that sensors and actuators can be physically connected to the computer. 

Analogue to Digital Converters (ADC)
One common requirement is for analogue signals to be converted into digital signals. This is done using an Analogue to Digital Converter, or ADC. It takes the analogue signal and converts it into digital information, which can be read by the computer.

real3

An example of the need for an interface with an ADC
Suppose we have a chemical reaction experiment in class and we want to measure the temperature during the reaction. We want to display the temperature at minute intervals on the computer. We will use a thermistor to read the temperature. A thermistor is a type of transducer. The voltage signal that comes out of the thermistor constantly changes as the temperature changes. There are an infinite number of temperatures possible so there are an infinite number of voltages possible. For this reason, we say that the thermistor is an analogue device. We will assume that the voltage varies anything from 0 volts to 1 volts for a range of temperatures from 0 deg C to 150 deg C. We cannot just plug the analogue device into the computer. Firstly, the computer is a digital device and only understands digital signals. Secondly, the voltages sent out of the thermistor may not be of the same level as the voltages used by the computer. Thirdly, it may physically not be possible to connect the wires from the thermistor into an I/O port at the back of the computer. For these reasons we need an interface. One part of the interface will be the ADC or Analogue to Digital Converter.

How does an Analogue to Digital Converter (ADC) help?
We need to be able to convert any analogue voltage into a digital one. There will also be times when we need a Digital to Analogue Converter (DAC), to take a digital signal and generate an analogue voltage. If we constantly recorded the analogue voltage over a 10-minute period, we may end up with a graph that looks like this:

real4

At the start of the experiment, the voltage from the thermistor is about 0.1 volts because the temperature is about 15 deg C. It rises steeply in the first minute and then steadily up to about 0.95 volts after 5 minutes, corresponding to about 140 deg C. It then declines steadily over the next 5 minutes. An ADC will read the voltage at set time intervals. To put it another way, it will sample the voltage regularly. How many samples we decide to take in any period of time is known as the sample rate and the only way to decide what is the best sample rate is to examine each problem on its own. In some applications, you may need to take a sample only once an hour, for example, measuring the temperature on top of a mountain. There may be other circumstances, such as in the above experiment where taking a sample once every hour would mean that important events are missed! A better sample time might be once a minute or once every 30 seconds. Of course, we could take 1000 readings per second if we wanted to but this would not give us any more information than taking a reading once a minute. All we would be doing is wasting computer storage. Choosing the right sample rate is a mixture of common sense and experience! Let us decide on a sample rate of one reading per minute. Each time a sample is taken, the voltage is converted into a digital value. The digital value will depend upon the type of ADC we have. For example, suppose we had an 8-bit ADC. That means that we have 256 different bit patterns (0 - 255).

    • If the ADC reads 0 volts this will be represented as 00000000
    • If the ADC reads 1/255 volt this will be represented as 00000001
    • If the ADC reads 2/255 volts this will be represented as 00000010
    • If the ADC reads 3/255 volts this will be represented as 00000011, and so on.

We can show this diagrammatically. The first two conversions have been shown on the graph.

real5

By using an 8-bit ADC, we can only divide up our 1 volt range into 1/256 divisions. We could get a much greater accuracy by using a 16-bit ADC. We could now divide up our 1 volt into about 65000 divisions and we could therefore measure even smaller changes to temperature. This is the same as saying we can measure the temperature more accurately! There is nothing stopping us using a 32-bit ADC for even better accuracy.

Actuators
According to the British Computing Society "A Glossary of Computing Terms", an actuator is "any device which can be operated by signals from a computer or control system causing physical movement". Sometimes, the processor in a system can change outputs directly (via an interface). E.g. lights could be switched on or off, an intruder alarm could be sounded or stopped, a graph could be plotted on the screen. However, when physical movement is required you need an actuator. You need a device that can cause movement when you send an electrical signal to it. E.g. if you needed to let water flow through a pipe, you might need to move a hydraulic arm that was connected to a valve. The hydraulic arm accepts electrical signals and moves. It is an example of an actuator. You might need to start and stop a lift by starting and stopping a motor. The motor is another example of a device that accepts electrical signals that causes physical movement. You might need to send a signal to a solenoid valve, so that it operates and opens a lock on a security door. The solenoid valve is an actuator. Just as you need interfaces to convert signals from various sensors into the signals used by the computer, you also need interfaces to convert signals from the computer into voltage types and levels that actuators (and other output devices) use. A motor might use 440 Volts ac. You would use an interface to allow 5 volts dc to switch on and off a device that uses a much bigger voltage. 

Back