Computer components and logic circuits

Input, output, primary memory and secondary storage

Every computer system, from the controller in a washing machine to a data-centre server, has to do the same four jobs, and the syllabus names the four parts that do them. It must take data in (input), give results out (output), hold the program and data it is working on at this moment (primary memory), and keep programs and data when the power is off (secondary storage).

Primary memory is the memory the processor can address directly: it is fast, it is expensive per byte, and it is small compared with storage. It is where the operating system, the programs that are running and the data those programs are using are held while they are needed. It is usually volatile, so its contents disappear when the power goes, though the ROM part of it keeps its contents. Primary memory is what makes a stored program possible: instructions have to be in it before the processor can fetch and execute them.

Secondary storage is everything that keeps data without power: a magnetic hard disk, a solid state drive, a memory card, an optical disc, a USB flash drive. It is far larger, far slower and far cheaper per byte than primary memory, and it holds what is not being used at the moment - the operating system files, the applications, documents, photographs, backups. Removable secondary storage is the subset that can be taken out and carried: a USB flash drive, an SD card, an external hard disk, a CD, DVD or Blu-ray disc. It is what you use to move files between machines, to hand work in, to distribute software and films, and to keep a copy of data somewhere separate from the computer so that a fire, a theft or a ransomware attack does not destroy the only copy.

The four are not interchangeable, which is the point of the exam question. Without input there is nothing to process; without output the result is invisible and useless; without primary memory the processor has nowhere to keep the instruction it is executing; without secondary storage nothing survives the power being switched off, and there would be no operating system to load in the first place.

Embedded systems

An embedded system is a computer system built into a larger device, dedicated to one function, and normally invisible to the user. It has a processor, memory and input and output, but they are inside something else - a washing machine, a car, a pacemaker, a television, a traffic light controller, a microwave. Its program, usually called firmware, is normally held in ROM or flash memory, so the system starts doing its one job as soon as power is applied and cannot be given new applications by the user.

Most embedded systems are real-time: they must respond to an input within a guaranteed time. The controller that fires an airbag has to react in milliseconds, and the controller that keeps a car engine running has to read its sensors and adjust the injection many times a second. That is why an embedded processor is chosen for predictable response rather than for raw speed, and why the software is written to a fixed loop rather than to an operating system with a user interface.

Benefits. An embedded system is small and cheap enough to build into a product that sells in millions; it uses very little power, which matters for anything battery-powered; it does one job, so it can be tested exhaustively and is reliable; it responds immediately because there is no general-purpose operating system in the way; it cannot easily be misconfigured by the user, which reduces support costs; and because the program is in ROM it survives the power being cut and cannot be altered by accident.

Drawbacks. It is fixed in function, so upgrading it means replacing the firmware or the whole unit rather than installing new software; its memory and processing power are deliberately minimal, so it cannot be extended; it is hard to debug, because there is usually no screen, keyboard or log to inspect; faults can be difficult and expensive to fix in the field, especially when the device is sealed or in a remote place; security updates are often impossible or never issued, which leaves the device vulnerable for its whole life; and if the embedded system fails, the whole device stops working, because there is nothing else to take over.

• Dedicated: written for one task and normally cannot run other software.
• Firmware in ROM or flash: starts automatically, survives power loss, is not changed by the user.
• Real-time: must respond within a guaranteed time, which is why the processor and the program are kept simple.
• Cheap, small and low power: chosen so that the cost can be built into a mass-market product.
• Benefit to quote with a reason: low power and a fixed program mean long battery life and reliable, repeatable behaviour.
• Drawback to quote with a reason: fixed function and minimal memory mean it cannot be upgraded or extended, and faults are hard to diagnose.

Hardware devices: printers, sound and interactive devices

A laser printer starts by turning the page into a bitmap. The page arrives described in a page description language, an internal processor called the raster image processor converts that description into a grid of dots at the printer's resolution - one bit per dot - and the bitmap is held in memory and in the print buffer. The bitmap then switches the laser on and off, dot by dot, and everything after that is electrostatics.

A photoconductive drum is first given a uniform electrical charge by a charging roller. The laser beam sweeps across the spinning drum and is switched on and off by the bitmap, and each point it strikes loses its charge, so an invisible pattern of charged and discharged dots is written around the drum. Toner, a fine plastic powder carrying the opposite charge, is attracted only to the discharged dots, so the image appears on the drum in toner. The toner is transferred onto the paper by a transfer roller and then fused permanently by heat and pressure as the paper passes between the fuser rollers; the drum is cleaned and recharged, ready for the next page.

A 3D printer works in the opposite direction: instead of putting an image onto a sheet, it builds a solid object layer by layer. A model is created in CAD software, the printer's software slices it into thin horizontal layers, and each layer becomes a set of instructions - the path to follow, the speed, the temperature. In fused deposition modelling a plastic filament is fed into a heated nozzle, melted and extruded while stepper motors move the print head in the x and y directions; when a layer is finished, either the head or the build plate moves down by one layer thickness and the next layer is drawn on top of it. The layers bond as the plastic cools, and where the object overhangs, support material is printed underneath and broken away afterwards.

A microphone turns a pressure wave into an electrical signal. Sound pushes and pulls a thin diaphragm, and in a dynamic microphone that diaphragm moves a coil inside a magnet, so the changing magnetic flux induces a small alternating current that follows the shape of the sound wave; condenser microphones use a capacitor whose spacing changes with the diaphragm, and piezo types use a crystal that produces a voltage when it is squeezed. The output is analogue, so an ADC is needed before a computer can store or process it. A speaker is the same chain reversed: an alternating current from the amplifier flows through a coil in a permanent magnetic field, the varying force moves the coil and the cone attached to it, and the cone pushes the air to create the pressure waves you hear; a DAC is needed if the source is digital.

A touchscreen puts input and output in one panel. The commonest kind is capacitive: a transparent grid of conducting tracks lies over the display, and a finger touching the glass changes the capacitance at that point; a controller measures the change and passes the x and y coordinates to the computer. Resistive screens work by contact instead - two transparent layers are pressed together where you touch, closing a circuit - which is why they need a fingernail or a stylus but work through gloves. A virtual reality headset shows a separate image to each eye through its own lens, the two images being slightly different views of the same scene so that the brain reads the difference as depth. Motion sensors inside the headset - accelerometers and gyroscopes, sometimes helped by cameras or external beacons - track the position and rotation of the head many times a second, and the computer redraws both images to match, which is what makes the scene appear fixed to the world as you turn your head. A high refresh rate and low latency matter because a visible delay between your movement and the picture is what makes people feel unwell.

Storage devices: magnetic, solid state and optical

A magnetic hard disk stores data as magnetised regions on one or more rigid platters coated with a magnetic material. The platters spin constantly, at anything from about 5 400 to 15 000 revolutions per minute, and each surface has a read/write head on the end of an actuator arm that moves it across the platter and holds it a few millionths of a metre above the surface. Data is written along concentric tracks, each divided into sectors, so the head has to wait for the right sector to come round, which is why mechanical disks are slow at reading many small files. To write, the head changes the direction of magnetisation of the coating; to read, it detects the changes in magnetic flux as they pass under it and converts them back into an electrical signal.

Solid state (flash) memory has no moving parts. Each bit is held in a floating-gate transistor, where charge trapped on the floating gate decides whether the cell conducts, and so whether it reads as a 1 or a 0. Because there is nothing to spin or seek, access is almost immediate, start-up is instant, and the device keeps working when it is shaken or dropped. The controller has to manage the cells carefully: writes are slow compared with reads, data can only be erased in whole blocks, each block wears out after a limited number of erase cycles, and wear levelling spreads the writes so that no block is used up first. Flash is what a solid state drive, a USB stick and a memory card are made of, and it is also the usual home of firmware in embedded systems.

An optical disc reader/writer uses a laser focused onto a spiral track to read the pattern of pits and lands pressed into the disc. Light reflects differently from a pit and from a land because of interference, and a photodiode measures the changing amount of reflected light and turns it into a stream of 0s and 1s. A read-only disc such as a CD-ROM or DVD-ROM is stamped with its pits at the factory. A write-once disc such as a CD-R or DVD-R has a dye layer that the laser heats and darkens at the points that are to be read as pits, so those points can never be changed back. A rewritable disc such as a CD-RW uses a phase-change layer that the laser heats to one temperature to make it crystalline and another to make it amorphous, and the two states reflect differently, so the data can be written, erased and written again. Blu-ray uses a shorter-wavelength blue laser, which can be focused to a smaller spot and so read a much denser track.

Choose between them by the job. A hard disk gives the most capacity for the money and is still used where large volumes are stored in one place, such as a server or a backup archive, but it is slow, noisy, power-hungry compared with flash and easily damaged by a knock while it is running. Flash is fast, silent, low power and shock-resistant, which is why it is in every phone, tablet and laptop, but it costs more per gigabyte, has a limited number of write cycles and is harder to recover data from when it fails. An optical disc is cheap, robust and long-lived, and is still used for distributing films and for archive copies that will not be rewritten, but it holds far less and is very slow.

Buffers

A buffer is a temporary store, usually a small area of RAM, used to hold data while it is being transferred between two parts of a system that work at different speeds. The faster side writes into the buffer at its own pace and the slower side empties it at its own pace, so neither has to sit idle waiting for the other. It is the shock absorber of a computer system.

The classic case is the printer buffer. A processor can hand a document over in a fraction of a second, while the printer may take a minute to put it on paper; without a buffer the processor would be held up for that whole minute. Instead the document is written into the printer's buffer as fast as the link allows and the processor carries on with other work, while the printer takes characters out of the buffer as it needs them. The same pattern appears in a keyboard buffer, which holds the codes for keys you pressed while the processor was busy so that no keystroke is lost, in a network buffer on a router, which holds packets while the outgoing link is busy, and in a streaming buffer, where a video player downloads several seconds ahead so that a brief drop in the connection does not interrupt playback.

Buffers are also used to even out a mismatch inside the machine. Double buffering keeps two buffers: while one is being filled, the other is being read, and then they swap. The graphics system uses this so that the screen is never updated while it is being drawn, which removes the flicker you would otherwise see on a partly redrawn image, and a disk or network interface uses it so that a transfer never has to pause while the buffer is emptied.

The point to hold on to is that a buffer is memory, not storage. It is small, it is volatile, and its contents are worthless once they have been passed on. It exists for a fraction of a second to keep two devices of different speeds working together, not to keep data for tomorrow.

• What it is: a temporary area of memory holding data during a transfer.
• Why it is needed: the sending and receiving devices work at different speeds, and the buffer lets each run at its own rate.
• Effect on the processor: it does not have to wait for a slow device, so it can be doing something else.
• Effect on the data: data is not lost when the receiving device is temporarily busy, and playback or printing runs smoothly.
• Double buffering: one buffer is filled while the other is emptied, so the transfer never pauses and the screen never shows a half-drawn image.
• Not storage: a buffer is volatile, small and temporary - do not confuse it with a disk cache or a backing store.

Memory: RAM, ROM, SRAM, DRAM and the programmable ROMs

RAM is the memory the computer works in. It is volatile, so it loses its contents when the power is removed, and it can be read from and written to freely and quickly. It holds the operating system once it has loaded, every program that is currently running and the data those programs are using. It is expensive per byte compared with storage, which is why a computer has a few gigabytes of it and a terabyte of disk. ROM is also primary memory, but it is non-volatile and, in normal use, read-only: its contents are fixed when the chip is programmed and cannot be altered by the running program. It is much smaller, and it holds the bootstrap - the first instructions that run when the power comes on, which test the hardware and load the operating system from storage - and the firmware of embedded devices such as a washing machine or a car's engine controller. ROM is used wherever the instructions must be present at every switch-on, must not be changed by the user, and must survive the power being cut.

Both SRAM and DRAM are RAM, and both are volatile; the difference is how a bit is held. SRAM (static RAM) stores each bit in a flip-flop made from four to six transistors, and the bit stays there as long as the power is on without anything having to refresh it. That makes it very fast, but each cell is large and the chip is expensive, so it is used in small amounts where speed is everything: cache memory close to the processor, and small buffers inside embedded systems. DRAM (dynamic RAM) stores each bit as charge on a tiny capacitor beside a single transistor. The capacitor leaks away in milliseconds, so the memory controller has to refresh every cell thousands of times a second, which takes time and makes DRAM slower than SRAM. In exchange a DRAM cell is about a quarter of the size and a fraction of the cost, so main memory is built from DRAM: that is how a machine can have gigabytes of working memory at a price anyone will pay.

The three programmable ROMs differ only in how they can be changed. PROM is supplied blank and programmed once, by the user, in a device called a programmer that burns out fuse links inside the chip; after that it behaves exactly like ROM and cannot be altered. EPROM can be erased and reprogrammed, but only by removing the chip, exposing the quartz window on its top to strong ultraviolet light for several minutes, and then writing it again at a higher voltage - so it is a workshop job, not something the machine can do to itself. EEPROM is erased and written electrically, in circuit, and can be changed one byte at a time, so a device can update its own stored settings or firmware without being opened. That convenience costs money and limits the number of times a cell can be rewritten. Flash memory, the technology inside a solid state drive, is the same idea arranged to erase in whole blocks rather than byte by byte, which makes it much cheaper and faster per byte.

Monitoring and control systems: sensors, actuators and feedback

Monitoring means using sensors to measure conditions and then recording, displaying or transmitting the readings. The system changes nothing: a person looks at the data and decides what to do. An automatic weather station, a data logger recording the temperature of a greenhouse every ten minutes, a car dashboard gauge and a hospital heart monitor are all monitoring systems. Control goes one step further: the system uses the readings to operate actuators and change the conditions itself, with no person in the loop. A greenhouse that switches heaters and vents to hold 22 °C, a washing machine holding the water at 40 °C, an air-conditioning unit and a car's engine management system are control systems. The same sensors can serve both; what separates monitoring from control is whether anything is altered automatically.

A sensor converts a physical quantity into an electrical signal, which an ADC converts into a digital value that the processor can compare with a stored number. The four the syllabus names are worth knowing by name, by quantity and by one example. A temperature sensor is usually a thermistor, whose resistance falls as temperature rises, so the voltage across it in a potential divider changes with the temperature; it is used in ovens, fridges, air conditioning and greenhouses. A pressure sensor uses a diaphragm or a strain gauge whose resistance changes as it is deformed, or a piezo crystal that produces a voltage when squeezed; it is used in a car's tyre-pressure monitor, in a boiler and in a barometer. An infra-red sensor either detects the heat radiated by a warm body (passive infra-red, used in intruder alarms and automatic doors) or uses an infra-red emitter and detector pair to sense a reflection or a break in the beam (used in remote controls, line-following robots and obstacle detection). A sound sensor is a microphone, which converts pressure waves into a voltage; it is used in a baby monitor, a clap-activated switch and noise-level logging.

An actuator does the opposite of a sensor: it converts an electrical signal into a physical action. A motor or servo turns or moves something, a solenoid or a valve opens or closes a pipe, a relay or transistor switches a heater, a pump or a lamp on and off, and a loudspeaker turns a signal back into sound. The processor's output is a digital signal, so a DAC is needed where the actuator requires a varying analogue voltage rather than simply on or off.

Feedback is what makes control work, and it is the part students leave out. The processor does not switch the heater on and then walk away: it reads the sensor, compares the reading with the set value stored in memory, and acts on the difference - if the temperature is below the set value, switch the heater on; if it is above, switch it off. It then reads the sensor again, and repeats. That loop, in which the effect of the action is measured and used to adjust the next action, is feedback. Without it a system would overshoot - the heater would stay on past the target and the room would get too hot - and it could not compensate for a disturbance such as a cold night or an open window, because it would have no idea that anything had changed.

Logic gates: six functions and their truth tables

A logic gate is a circuit with one or two inputs and one output, whose output depends only on the combination of inputs present at that moment. The syllabus uses six, and in this syllabus every gate except NOT has two inputs only, which matters when you draw a circuit for an expression with three variables. NOT has one input and inverts it. AND gives 1 only when both inputs are 1. OR gives 1 when either input is 1, or both. NAND is AND followed by NOT, so it gives 0 only when both inputs are 1. NOR is OR followed by NOT, so it gives 1 only when both inputs are 0. XOR (also written EOR) gives 1 when the inputs are different and 0 when they are the same. The 'X' stands for exclusive: OR allows both inputs to be 1, XOR does not.

A truth table lists every possible combination of inputs once, in a fixed order, with the output for each. A gate with n inputs needs 2 to the power n rows: two inputs give four rows, three inputs give eight. Write the inputs in binary counting order - 00, 01, 10, 11 - so that a row is never missed or repeated, and give each gate in a circuit its own output column, working from the inputs towards the final output. Teachers and examiners read a table by its rows, so a column that is right in three rows and wrong in one loses a mark and, worse, tells the examiner that the function was not understood.

You also have to recognise and draw the symbols. AND has a flat back and a rounded, D-shaped front. OR has a curved back and a pointed front, like a shield. NOT is a triangle with a small circle, called a bubble, on its output. NAND and NOR are the AND and OR symbols with that same bubble on the output, and the bubble is the only thing that makes them different - which is exactly why they are the opposite functions. XOR is the OR symbol with an extra curved line across its input side. When you draw them, keep the input lines on the left, the output on the right, and put the bubble at the output where there is one.

Boolean notation is worth recognising even though the exam asks for circuits and truth tables. A AND B is written A.B or simply AB, A OR B is written A + B, and NOT A is written with a bar over the A, which you show as NOT A on paper. When you evaluate an expression, brackets come first, then NOT, then AND, then OR - so A OR B AND C means A OR (B AND C), not (A OR B) AND C. If you are unsure, put the brackets in yourself before you start drawing.

• NOT: one input; output is the opposite, so 0 -> 1 and 1 -> 0.
• AND: output 1 only when both inputs are 1; otherwise 0.
• OR: output 1 when at least one input is 1; output 0 only when both are 0.
• NAND: the opposite of AND, so output 0 only when both inputs are 1.
• NOR: the opposite of OR, so output 1 only when both inputs are 0.
• XOR (EOR): output 1 when the inputs are different, 0 when they are the same.

Moving between a problem statement, an expression, a circuit and a truth table

From a problem statement. Name the inputs and the output first, and say what 1 means for each. In a car seat-belt warning, let I mean the ignition is on, S mean the driver's seat is occupied and B mean the belt is fastened; the lamp L is to come on when the ignition is on and the seat is occupied and the belt is not fastened, which gives L = I AND S AND NOT B. Writing that sentence down as an expression before you draw anything is what keeps the circuit correct, and it also tells you how many gates you need.

From an expression to a circuit. Work from the inside out: brackets first, then NOT, then AND, then OR. Give every intermediate result its own label so you can connect it onwards without confusion - for L above, NOT B is one gate, I AND S is a second, and the third AND gate combines the two intermediate outputs. Remember that AND, OR, NAND, NOR and XOR have two inputs each in this syllabus, so a three-input AND is built from two two-input AND gates, as here. The output of the last gate is L, and the inputs on the left are I, S and B.

From a circuit to a truth table. Count the inputs and write 2^n rows in binary order, then add a column for each gate output in the order the signal reaches them, and work along the rows. From a circuit to an expression, do the same thing in symbols: label the output of each gate with the expression for its inputs, substitute the labels into the next gate, and keep going until you reach the final output.

From a truth table to an expression. Take every row in which the output is 1. For each one, write the inputs joined by AND, putting a variable in as it is when its value in that row is 1 and as NOT when its value is 0, and then OR all those terms together. For XOR: the output is 1 in the rows 01 and 10, so the expression is (NOT A AND B) OR (A AND NOT B). Check it by reading the expression back into the table - if every row matches, the expression is right, and if one does not, you have misread a 0.

• Problem statement -> name the inputs, name the output, state what 1 means, then write the expression in words and in symbols.
• Expression -> circuit: brackets, then NOT, then AND, then OR; label every intermediate output; build three-input gates from two-input gates.
• Circuit -> expression: write the expression at each gate output and substitute it into the next gate.
• Circuit or expression -> truth table: 2^n rows in binary order, one column per gate, filled from the inputs towards the output.
• Truth table -> expression: one AND term for each row where the output is 1, with NOT for each input that is 0, then OR the terms.
• Always check the finished answer against the thing you started with - a table with four rows for a three-input circuit is wrong before you read a single value.