Komunikasi Data

1.1  The development of data communications 2

1.2  Types and sources of data 4

1.3  Communications models 6

1.4  Standards 7

1.5  Open Systems Interconnection 9

1.6  OSI Reference Model 13

1.7  Institute of Electrical and Electronic Engineers 802 standards 15

1.8  OSI Reference Model and other standards 16

This chapter commences with a review of the development of data communications which, although of interest in its own right, is in fact used as a vehicle to set the scene, and introduce a number of basic concepts, for later chapters in the book. Having established how data communications have developed we then turn our attention to the different forms of data and offer some examples of where such data occurs in practice. The general concept of communications models, which are a tool to enable design, implementation and understanding of communications networks and systems, is introduced. The case is then made for standards and the major standards organizations are discussed. A communications model which has greatly influenced current network philosophy is the Open Systems Interconnection (OSI) Reference Model for which standards have already have been produced and many others are under development. After introducing this model, later chapters explore how the model is used in practice, generally through a particular standard. The Institute of Electrical and Electronic Engineers, although not a standards organization in its own right, is discussed, as is its influence on the design of computer networks. A comparison then follows which relates standards of various organizations. The chapter concludes with a brief introduction to the ubiquitous TCP/IP standard which is the linchpin of today’s Internet.

1.1 The development of data communications

A brief overview of the development of data communications follows to enable the reader to appreciate the more complex concepts of later chapters. The appearance of digital electronics in the early 1960s led to the widespread use of digital computers.

Three main communities which rapidly deployed computers were large financial institutions such as banks and building societies in order to automate and reduce staffing, universities for the solution of complex problems and large commercial organizations to improve management and efficiency. Early computers, despite having relatively limited computing power compared with those of today, were expensive. Their use was therefore shared, often by many users. This raised the issue of how users gain access to computers. The issue of computer access which is both secure and fairly apportioned continues to be a topic of great importance and will be developed further in later chapters.

Another issue which appears in communications is that of a protocol. This is nothing more than a set of rules to govern an activity. Consider two people holding a conversation.

A simple protocol, often observed by some cultures, is that only one person speaks at a time. A person wishing to speak waits for a pause in the conversation before commencing. Some computer networks also use this protocol, in which case it is called stop-and-wait. In the case of a floppy disk, there clearly needs to be an agreed system for formatting and reading disks in order that they can be interpreted and processed by a computer. In other words, there needs to be a communication protocol.

One of the key features of a communications system is that of transmission speed. The principal measure of the speed of transmission of data is bit per second (bps). It is found by dividing the total number of bits transmitted by the time taken to complete such a transfer. Transmission speeds in some networks, which we shall explore

in Chapter 11, may be as high as one thousand million bits per second (Gbps), or more.

The emergence of modems in the early 1960s changed data communications dramatically. A modem is a device that converts digital data into analogue signals and therefore allows digital signals of computers to be converted into analogue-type signals suitable for transmission over the then almost universal analogue circuits of the telephone network or Public Switched Telephone Network (PSTN). This opened up the possibility of remote terminals being electrically connected to a distant or host computer, either permanently or temporarily, by means of a PSTN. This is an early example of on-line operation, which paved the way for interactive operation in which the stations at either end of a link can interact with each other. It would be an understatement to say this greatly speeded up data transmission and hence processing! Online operation also opened up new opportunities and applications, for example remote access of databases or resource sharing.

With the introduction of on-line operation, new protocols became necessary to format and control the flow of data. The possibility then existed of interconnecting a number of different computers, each of which may operate a different protocol. Networking, which is the technique of connecting two or more user systems together, was in its infancy and the emergence of the need for standards soon became evident.

Computer terminal equipment is generally called data terminal equipment (DTE).

Modems, as has been suggested above, may be used to interconnect a DTE to a The development of data communications 3 telephone line. A modem is an example of equipment that terminates the end of a communication network. Such equipment is referred to as data circuit terminating equipment (DCE). The connection between a DCE and a DTE is an example of an interface. Standard interfaces have been defined which govern the physical and procedural arrangements for connection. We shall consider the details of some standard interfaces in later chapters.

By the mid-1960s networks had advanced to the point that, instead of one DTE connecting with a single host computer, DTEs, or users, began communicating with each other. This enabled resource sharing and intercommunication to occur. To support such activity, users are assigned an address in the same way as a telephone line or house has a unique identity. Such addresses enable a network to successfully route messages from a send station to a receive station across a network. In order that communication networks are able to route connections they must contain either switches or routers to select the paths, or routes, a connection must follow. Hence many networks are called switched networks.

Simple DTE to host computer communication commonly uses a continuous connection for the duration of the message transfer. This is known as circuit switching, where a physical, or fixed path, is dedicated to a pair of devices for the duration of the communication. A dial-up modem arrangement to connect a DTE to a host computer is a good example of a circuit-switched connection.

New data networks began to be experimented with towards the end of the 1960s and operated on a packet basis whereby data messages, unless very short, are subdivided into a number of separate blocks of data, known as packets, each containing hundreds of bytes. Each packet may flow through the network independently of other packets. This now meant that each packet must contain some addressing information.

Connections may be connection oriented where a route through the network, known as a logical connection, is established (not necessarily a fixed end-to-end physical connection) for the duration of the message exchange and over which all subsequent packets flow. Connection-oriented operation is similar to circuit-switched operation. Once a  logical connection is established, data is then transferred, using as many packets as is necessary, until the data transfer is complete. The connection may then be terminated.

Connection-oriented operation is used extensively in data communications. An alternative method of operation is that of connectionless operation, which has become extremely popular in recent years. Here each packet is considered by the network as a unique entity and separate from any other packets, even though it may form part

of the same message transfer. This means that each packet must contain full address information so that it can be routed through a network independently. This contrasts with connection-oriented operation where, once a route is set up, subsequent packets follow the same route and do not have to be independently addressed. Packet switching will be explored in detail in Chapter 12. Public telephone and packet networks are generally operated on a national basis and cover a very large geographical area.

It is for this reason that they are called wide area networks (WANs).

In contrast to the 1960s, the 1970s saw computers become both available and affordable to society as a whole and appear in everyday industrial and commercial environments.

Local intercommunication between computers may, as has already been discussed, be provided by a WAN, be it a telephone or packet network. However, local

interconnection using a WAN is unwieldy and expensive. Local area networks (LANs) appeared in the mid-1970s. These networks today often comprise a number of personal computers (PCs) confined to a relatively small, or local, physical area and are suitable for a single site or complex. The appearance of LANs offered the facility of economically connecting machines together in, say, an office using a dedicated communication network. Where a connection is required between more distant users, for example users on separate LANs, connection may be made using an intermediate WAN.

The 1980s saw the development of metropolitan area networks (MANs) that connect a number of sites and typically span an area the size of a metropolitan region, or city. Optical fibre, rather than a metallic conducting medium, is generally used to support the relatively high data rate found in MANs over distances of the order of tens of kilometers. A MAN is often used as a backbone network to interconnect a number of physically disparate LANs. Both LANs and MANs are dealt with in later chapters.

The 1990s witnessed a rapid increase in popularity in internetworking. Just as packet switching grew out of a need to bring together a number of disparate systems into a network, a need has been perceived to bring together different networks. Although a user may not be aware of the existence of more than one network, the different networks that are interconnected in this way retain their own individuality and characteristics.

Such a grouping of networks is often referred to as an internet.

1.2 Types and sources of data

The previous section has already alluded in part to the nature of data and where it originates. In general, there are three main types of signal required to be transmitted: speech, video and data. Although speech and video signals are analogue in nature, the technology now exists to digitize virtually any signal at source. Once this process has occurred, any signal, irrespective of the type of information it represents, may, from then on, be regarded merely as data. Video signals are in the main found in videophones, video conferencing, multimedia applications, broadcast TV, cable TV, satellite and surveillance systems. Analogue operation continues today in some instances, even though the technology may be obsolescent, because it continues to provide adequate quality and the technology used continues to perform satisfactorily.

Although many different networks have evolved, there is increasing pressure for data transmission, through the introduction of standards, to operate at preferred speeds or rates. Speeds which are commonly used are 1200, 2400, 4800, 9600 bps, or multiples, and submultiples, of 64 kbps. A higher range of speeds of 1.5/2, 8, 34, 52, 155, 622 or 2488 Mbps are widely available in the long-haul communication links of WANs offered by Post, Telephone and Telecommunications operators (PTTs).

Figure 1.1 illustrates the major types of services currently found and their accompanying data rates. Additionally the services are also categorized into audio, visual and data. Services operating at under 2 Mbps are termed narrowband whereas higher rates are categorized as broadband. These themes will be developed further in Chapters 7 and 8.

Some of the services are self-explanatory. Video telephony is commercially available in the form of a ‘videophone’ where an image of the telephony user appears at

Types and sources of data 5

Figure 1.1 Types and sources of data.

the distant videophone. Video conferencing is a method of conferencing using TV

cameras and monitors between two or more sites. It attempts to provide a conferencing

facility and atmosphere without the need for participants to travel to the same

venue for a traditional conference or meeting. Speech and images are relayed and

close-ups may be included to view documents, photographs and drawings.

High-definition television (HDTV) is an extension of existing traditional TV services,

the higher definition leading to greater picture detail and sharpness. To facilitate

improved information content a higher bit rate is necessary. Videotext is a service

which transmits alphanumeric and fairly simple graphics information. It is commonly

found as an additional service offered within TV broadcast services or for remote booking

services used, for instance, by travel agents for flight and holiday bookings. As

indicated in Figure 1.1, Videotex has a relatively low data rate of up to 100 kbps or so.

Remote computer-aided design (CAD) simply occurs when CAD information is

required to be transmitted. Such information occurs in a variety of CAD applications

often associated with engineering design and may be design drawings, instructions for

programs, etc. Bulk file transfer transmission occurs at a fairly high data rate and is

used to pass very large files of information between sites. An example is where large

amounts of stored computer information are passed to a second computer for backup

purposes in the event of the first computer failing. These back-ups are, of necessity,

performed quite frequently to maintain integrity and because of the large volume

of data involved they must be performed at high speed.

Data has already been discussed in terms of the need to transmit at a distance, or

more locally as in a LAN. Dependent upon the application, data may be transmitted at

very slow speeds of, say, the order of tens of bps typical of some telemetry applications.

LANs, however, have relatively high speeds of the order of tens, hundreds or even

thousands of Mbps.

1.3 Communications models

Data communications are predominantly associated with supporting communications

between two, or more, interconnected computer systems. Some of the main tasks necessary

for successful communication are listed below:

• Initialization and release of a communications link.
• Synchronization between sending and receiving stations in order to interpret signals correctly, for example at the start and end of a packet.
• Information exchange protocols to govern communication. For instance, a protocol needs to indicate whether a station may transmit and receive simultaneously, or alternately.
• Error control to determine if received signals, or messages, are free from error. Where errors are evident some sort of corrective action should be instigated.
• Addressing and routing. These functions ensure that appropriate routing occurs within a network so as to connect a sending station successfully to a receiving station with the correct address.
• Message formatting. This is concerned with the formatting or coding used to represent the information.

Although the above list is not exhaustive it does illustrate something of the range of

complexity involved in a data transfer between communicating parties. In order to build

a system, the communication task must be broken down into a number of manageable

subtasks and their interrelationships clearly defined. A common approach to such

analysis is to represent the tasks and their interrelationships in the form of a conceptual

communications model. If the model is sufficiently well defined then the system

may be developed successfully.

Although a given communications system may be modelled in a number of ways,

computer communications models lean heavily towards layered models. Each layer

of the model represents a related series of tasks which is a subset of the total number

of tasks involved in a communications system. Figure 1.2 illustrates a three-layer model

connecting two systems together.

A commonly used term in relation to communications models is that of an entity.

An entity is anything capable of sending, or receiving, information. Examples of

entities are user application programs, terminals and electronic mail facilities. Each

layer of the model communicates using a layer–layer protocol between the two systems.

This may be thought of as a communication at a particular layer between a pair

of different programs, each operating at the same layer. Within a layered communication

model, however, communication ultimately may only occur between the two

systems via the physical connection. In order to support a protocol at a particular layer

it must therefore operate via the lower layers of the model and then finally over the

Figure 1.2 A layered communications model.

physical connection. The boundary between the lowest layer and the physical connection

is a hardware interface. However, boundaries between the other layers are predominantly

software in nature.

A well-constructed layered model enables complex systems to be specified,

designed and implemented. This is achieved by the development of a model which

splits these tasks into manageable layers and in such a way that the function of each

layer is unique and can be implemented separately. Such a layered model then provides

a framework for the development of protocols and standards to support information

exchange between different systems. Although we have so far only considered

layered models conceptually, later on in this chapter we shall look in detail at the dominant

layered model which is shaping data communications today and in the future.

1.4 Standards

Standards are required to govern the physical, electrical and procedural characteristics

of communications equipment. Standards attempt to ensure that communications

equipment made by different manufacturers will satisfactorily interwork with each other.

The principal advantages of standards are that they:

1. Ensure a large market for a particular product.
2. Allow products from multiple vendors to communicate with each other. This gives purchasers more flexibility in equipment selection and use. It also limits monopoly and enables healthy competition.

Disadvantages are:

A standard tends to lock or freeze technology at that point in time.

1. Standards take several years to become established via numerous discussions and committee meetings. Often, by the time they are in place more efficient techniques have appeared.
2. Multiple standards for the same thing often exist. This means that standards conversion
3. is sometimes necessary. An example is the variety of TV standards existing throughout the world. In data communications an example is the USA’s use of 1.5 Mbps digital cable transmission systems operated by PTTs compared with Europe’s  use of 2 Mbps systems.

The world of data communications is heavily regulated, legally and de facto. There

exists a whole raft of standards bodies at international, regional and national level.

Internationally, the two principal standards bodies concerned with standards for data

communications are the International Telecommunications Union (ITU) and the

International Organization for Standardization (ISO).

International Telecommunications Union

The ITU is based in Geneva and is a specialized agency of the UN. It comprises

member countries, each with equal status and voting rights, as well as industrial

companies and international organizations. On 1 July 1994 it was restructured. The

Telecommunications Standardization Sector (ITU-T) is responsible for setting standards

for public voice and data services (formerly the remit of the Consultative Committee

on International Telegraphy and Telephony or CCITT). The Radio Communications

Sector (ITU-R) is responsible for radio-frequency spectrum management for both space

and terrestrial use. Both this and standards setting for radio used to be performed by

the International Radio Consultative Committee (CCIR). The third sector of the ITU

is the Development Sector (ITU-D), which is responsible for improving telecommunications

equipment and systems in developing countries.

Each sector also organizes conferences on a world and/or regional basis and operates

study groups. Standards eventually are produced as a result of such activity to

govern interworking and data transfer between equipment at an international level, rather

than within the confines of a single nation. ITU-T standards which have been produced

for use in data communication are:

G-series – Transmission systems and media, digital systems and networks

H-series – Audiovisual and multimedia systems

I-series – Integrated Services Digital Network (ISDN) transmission

Q-series – Switching and signalling

V-series – Data communications over the telephone network

X-series – Data networks and open system communication

Later chapters will be exploring systems and make reference to the relevant standards

as appropriate.

International Organization for Standardization

The ISO promotes the development of standards in the world with the view of facilitating

the international exchange of goods and services. Its sphere of interest is not merely confined to data communications, as seen, for instance, in ISO specifications

for photographic film. The organization is made up of members from most countries,

each representing the standards bodies of their parent country, for example BSI for

the UK and ANSI for the USA. The main implication of ISO in data communications

has been its development of the reference model for OSI which, as already mentioned,

is discussed later in this chapter. In the area of data communications, the ISO standards

are developed in cooperation with another body, the International Electrotechnical

Committee (IEC). Since the IEC is primarily interested in standards in electrical and

electronic engineering, it tends to concentrate on hardware issues, whereas ISO is more

concerned with software issues. In the area of data communications (and information

technology), in which their interests overlap, they have formed a joint technical committee

which is the prime mover in developing standards.

Other standards bodies

The European Telecommunications Standards Institute (ETSI) has 912 members

drawn from 54 countries inside and outside Europe. It represents PTT administrations,

network operators, manufacturers, service providers, research bodies and users.

ETSI plays a major role in developing a wide range of standards and other technical

documentation as Europe’s contribution to worldwide standardization in telecommunications,

broadcasting and information technology. Its prime objective is to support

global harmonization by providing a forum in which all the key players can

contribute actively and is officially recognized by the European Commission and EFTA

secretariat.

1.5 Open Systems Interconnection

As computer networks have proliferated, so the need to communicate between users

located on different networks has emerged. Such intercommunicating computer systems

may be termed distributed computer systems and are required to process information

and pass it between each other.

Historically, communication between groups of computers and DTEs was generally

restricted to equipment from a single manufacturer. Many systems used either IBM’s

Systems Network Architecture (SNA) or DEC’s Digital Network Architecture (DNA)

which are not directly compatible with each other. ISO formulated its Open Systems

Interconnection (OSI) reference model in the late 1970s specifically to address the problem

of interconnectivity between different user systems.

OSI gives users of data networks the freedom and flexibility to choose equipment,

software and systems from any vendor. It aims to sweep away proprietary systems

which oblige a user to build a system with kit from a single vendor. It is a concept

which relies upon the emergence of common standards to which components and systems

must conform. In this way full interconnectivity will occur between users, each

using equipment supplied by vendors of their choice.

N-layer service

The OSI reference model may be thought of as a series of conceptual layers. Such

layers operate on the principle shown in Figure 1.3. Layer N provides service N to

layer N+1. In order for layer N to fulfil its service to layer N+1, it in turn requests a

service from layer N−1. The interfaces or boundaries between adjacent layers or services

are known as N-layer service access points (SAPs).

Peer-to-peer protocols

The concept of an N-layer service is intended to break down the complex tasks of

networking into a series of logical and ordered subtasks, each of which becomes

relatively simple to design and implement. Another plank in this process of decomposition

of the networking task is the concept of peer-to-peer communication protocols

whereby any given layer ‘talks’ to its corresponding layer at the distant end.

Encapsulation

We have established that protocols are operating horizontally between layers in peerto-

peer fashion. However, in order to support such protocols, communication must in

reality occur up and down through the layers. The question then arises: how are protocols

implemented between layers and also across the network? The answer is illustrated

in Figure 1.4. When an application has a message to send, data is sent to the

application layer which appends an application layer header (H). The purpose of the

header is to include the additional information required for peer-to-peer communication.

The resultant header and data are termed a protocol data unit or PDU. Each

PDU is said to encapsulate data by adding such a header. This process is repeated a

Figure 1.3 N-layer service.

Open Systems Interconnection 11

number of times and generates a PDU at each layer. The data link layer also adds

a trailer to produce what is known as a frame. Finally the frame, with all of the

encapsulated PDUs, is transmitted over the physical medium. The receiving station

performs the reverse operation to encapsulation (known as decapsulation) as headers

are stripped off at each layer to separate the respective communications protocols

from the data units, the latter being successively passed up through the layers.

Primitives

The way in which a layer provides a service is by means of primitives. A primitive

specifies an operation or action to occur. It may be a request for a certain service, or

an indication that a certain action, or event, has happened. In general there are four

types of primitives available, Table 1.1.

Let us examine primitives in an everyday example such as making a telephone

call:

Caller lifts handset – Request a connection.

Dial tone, dialling and switching follow.

Ringing is applied to callee – Indication, ringing tone applied to callee to indicate

that a connection is being requested.

Callee answers – Response. Caller responds by accepting a connection.

Caller advised of callee’s response – Confirmation that call is accepted by callee.

Connection established – communication ensues.

Caller, or callee, hangs up to terminate connection.

In a similar way the communication, or in data communications terminology, the data

transfer phase and the disconnection phase, may make use of primitives. The above

example could therefore be further extended, and even shown diagrammatically, making

use of various service primitives as shown in Figure 1.5. This figure also illustrates

another point. The CONNECTION primitive uses a confirmed service whereas

the DATA and DISCONNECT primitives are unconfirmed. Confirmed services have

a response and confirmation, or acknowledgement, whereas unconfirmed services do

not use an acknowledgement. Connections must be confirmed so that both parties

are certain that a connection exists before transferring any information. Data may be

Figure 1.5 Connection-oriented data transfer sequence.

confirmed or unconfirmed, as in the above example, depending upon which service is

required. If voice is being conveyed over the connection confirmation is not usually

necessary. If some data is lost it may not even matter, or the receiving party could

ask for some repetition. Where a data transfer is more critical, for example in transferring

a webpage, any data loss could be of concern and therefore a confirmed service

could be used. Disconnections are invariably unconfirmed since once one party

has disconnected, the connection has ceased. In general either party may initiate a

disconnection.

1.6 OSI Reference Model

The ISO’s OSI seven-layer reference model is shown in Figure 1.6.

The reference model has been developed based upon some of the principles discussed

in general terms in the previous section. In addition the choice of seven layers

has sprung from the following:

           

1. Only sufficient layers have been agreed such that each layer represents a different and unique function within the model.
2. A layer may be thought of as the position of a related group of functions within the model which ranges from one specific to the user application at one end to one involving the transmission or data bits at the other. Layers between these two extremes offer functions which include interaction with an intermediate network to establish a connection, error control and data format representation.

3. A layer should be so organized that it may be modified at a later date to enable new functions to be added and yet not require any changes within any other layer.
4. The seven layers and their boundaries have attempted to build upon other models which have proved successful and in such a way as to optimize the transfer of information between layers.

Layer 1, the physical layer, defines the electrical, mechanical and functional interface

between a DCE and the transmission medium to enable bits to be transmitted successfully.

The layer is always implemented in hardware. A common example used

extensively in modems is the ITU-T’s V.24 serial interface which will be discussed

in detail in Chapter 6. No error control exists at layer 1 but line coding may be incorporated

in order to match data signals to certain properties of the communication channel.

An example might be to remove a dc component from the signal. Line coding is

introduced in Chapter 2.

Layer 2 is the data link layer, the function of which is to perform error-free, reliable

transmission of data over a link. Link management procedures allow for the setting

up and disconnection of links as required for communication. Having established

a connection, error detection, and optionally error correction, is implemented to

ensure that the data transfer is reliable. Flow control is also performed to provide for

the orderly flow of data (normally in the form of packets) and to ensure that it is not

lost or duplicated during transmission.

Layer 3 is the network layer, whose principal task is to establish, maintain and terminate

connections to support the transfer of information between end systems via

one, or more, intermediate communication networks. It is the only layer concerned

with routing, offering addressing schemes which allow users to refer unambiguously

to each other. Apart from the control of connections and routing, the layer, by engaging

in a dialogue with the network, offers other services such as a user requesting a

certain quality of service or reset and synchronization procedures.

The intermediate data communications network included in Figure 1.5 is not

strictly a part of the reference model. It is included to draw attention to the fact that

the network may be regarded as a black box. What happens within it does not necessarily

have to conform to any standards, although in practice this is rarely the case.

Rather, what is important is that interfaces and protocols are fully supported between

networks to ensure compatibility and interworking.

All of the lower three layers are heavily network dependent, for example ITU-T’s

X.25 recommendation for gaining access to packet-switching networks specifies operation

at layers 1, 2 and 3 only.

Layer 4 is the transport layer and separates the function of the higher layers, layers

5, 6 and 7, from the lower layers already discussed. It hides the complexities of data

communications from the higher layers which are predominantly concerned with supporting

applications. The layer provides a reliable end-to-end service for the transfer

of messages irrespective of the underlying network. To fulfil this role, the transport

layer selects a suitable communications network which provides the required quality

of service. Some of the factors which the layer would consider in such selection are

throughput, error rate and delay. Furthermore, the layer is responsible for splitting up messages into a series of packets of suitable size for onward transmission through the

selected communications network.

Layer 5, the session layer, is responsible for establishing and maintaining a logical

connection. This may include access controls such as log-on and password protection.

Secondly, the session layer performs a function known as dialogue management. This

is merely a protocol used to order communication between each party during a session.

It may be best explained by way of an example. Consider an enquiry/response

application such as is used for airline ticket booking systems. Although two-way communication

is necessary for such an interactive application it need not be simultaneous.

Suppose that the connection only provides communication in one direction at a

time. The protocol must therefore regulate the direction of communication at any one

instant. If, however, full simultaneous two-way communication is available then little

dialogue management is required save some negotiation at set-up time. The third, and

most important, function of the session layer is that of recovery (or synchronization).

Synchronizing points are marked periodically throughout the period of dialogue. In

the event of a failure, dialogue can return to a synchronizing point, restart and continue

from that point (using back-up facilities) as though no failure had occurred.

Layer 6 is the presentation layer and presents data to the application layer in a form

which it is able to understand. To that end, it performs any necessary code and/or data

format conversion. In this way there is no necessity for the application layer to be

aware of the code used in the peer-to-peer communication at the presentation layer.

This means that in practice, users may operate with entirely different codes at each

end and which may in turn be different again from the code used across the network

for intercommunication. Encryption may also be added at layer 6 for security of

messages. Encryption converts the original data into a form which ideally should be

unintelligible to any unauthorized third party. Such messages may usually only be

decrypted by knowledge of a key which of course must be kept secure.

Layer 7, the application layer, gives the end-user access to the OSI environment.

This means that the layer provides the necessary software to offer the user’s application

programs a set of network services, for example an e-mail service. It is effectively

the junction between the user’s operating system and the OSI network software.

In addition, layer 7 may include network management, diagnostics and statistics gathering,

and other monitoring facilities.

Most standards activity has centred on the lower layers to support communications

networks and their interfaces, for example ITU-T’s X.25 recommendation for packetswitched

network operation addresses layers 1, 2 and 3, only. ISO standards have more

recently addressed this imbalance with standards for some applications being available

at all seven layers to support a truly open system interconnection.

1.7 Institute of Electrical and Electronic Engineers 802 standards

The Institute of Electrical and Electronic Engineers (IEEE) is a US professional society

and is not a standards body in its own right. Its major influence on international

standards is through its IEEE 802 project which produced recommendations, initially

for use with LANs, in 1985 and which were adopted as standards by ISO in 1987.

Since then further standards have been developed for MANs. Work continues in the

development of new standards via technical advisory groups. The current activities of

the IEEE 802 Committee are shown below:

802. 1 Overview of 802 standards

802. 2 Logical Link Control

802. 3 Carrier Sense Multiple Access/Collision Detection CSMA/CD

802. 4 Token Bus

802. 5 Token Ring

802. 6 MANs

802. 7 Broadband LANs

802. 9 LAN Integration

802.10 LAN Security

802.11 Wireless LANs

802.12 100VG AnyLAN

802.17 Resilient Packet Ring

Logical Link Control (802.2) specifies the link control procedures for the correct flow

and interchange of frames for use with the three basic LAN types. These are distinguished

by the way in which a user gains access to the physical medium of the LAN,

a process known as medium access control (MAC), and are known as CSMA/CD,

Token Bus and Token Ring (802.3, 4 and 5 respectively). A number of differences

exist between these types of LAN which relate fundamentally to network topology

and access methods, signalling technique, transmission speed and media length. In Chapter

10 we shall examine MAC in some depth.

1.8 OSI Reference Model and other standards

Table 1.2 compares the OSI Reference Model and various other standards for a variety

of applications. In many instances the other standards are comparable with and

sometimes identical to those of ISO.

Some texts also include DNA and SNA in comparisons with ISO but as they are

not standards they are omitted here. ISO itself has standards to support applications

at the application layer such as:

FTAM – File Transfer & Access Management

VTs – Virtual Terminals

CCR – Commitment, Concurrency & Recovery

JTM – Job Transfer, Access & Management

ISO also has standards to implement the remaining protocol layer functions. The LAN

standards 802.2–802.5 have been adopted by ISO as 8802.2–8802.5 respectively. The

data link layer standard is 8802.2. Similarly at the physical layer, MAC procedures

and physical interconnection are offered as ISO 8802.3–8802.5.

ITU-T supports a number of application layer services including:

X.500 – Directory Services

X.400 – Message Handling Services

RTSE – Reliable Transfer

ROSE – Remote Operations

ACSE – Association Control

ITU-T also has standards defining operation at the lower three layers for ISDN, packet

switching and circuit switching.

The OSI Reference Model seeks to set out a method of developing intercommunication

or internetworking via one or more, often different, computer communication

networks. Additionally, having established the reference model, ISO is committed to

developing a full range of standards to enable such internetworking to be established

effectively and easily.

Finally, any discussion on OSI would be incomplete without reference to the

Internet and the TCP/IP suite. The US Department of Defense (DoD) had already

addressed and produced a solution to internetworking before ISO standardization

occurred. The DoD’s developments culminated in the Transmission Control Protocol/

Internet Protocol (TCP/IP), which is a suite of protocols used almost universally to

interconnect computer networks. As explained in Chapter 15, the Internet Architecture

Board (IAB) has taken over responsibility for Internet standards. The IAB publishes

documents called Request for Comments (RFCs), which are now in fact de facto

standards ratified by the ITU-T. Access is easily and inexpensively achieved by means

of subscription to any of the networks which form the Internet.

TCP/IP assumes that the Internet, or some of the underlying networks, is unreliable.

With reference to Figure 1.7, reliable operation on an end-to-end basis is achieved by

Figure 1.7 TCP/IP operation.

means of connection-oriented operation using TCP at layer 4. TCP is therefore only

provided at the two end-systems rather than in any intermediate network forming part

of the connection. The use of TCP means that lost packets can be retransmitted. IP is

implemented at the network layer and is used for routing each packet in all intermediate

networks and is a connectionless service.

Exercises

1.1  Suggest some alternative models to that of the OSI Reference Model. Compare and contrast them with that of the OSI Reference Model.

1.2  Are there any situations where a reduced version of the OSI Reference Model may be adequate? Draw such a model.

1.3  Explain the terms ‘peer-to-peer communication’ and ‘encapsulation’. What advantage do these concepts offer designers of open systems?

1.4  Explain what is meant by ‘service’ and ‘protocol’ for a given layer of the OSI Reference Model.

1.5  Summarize the structure of the OSI Reference Model using a sketch and indicate where the following services are provided:

a)      distributed information services

b)      code-independent message interchange service

c)      network-independent message interchange service.

1.6  A relatively long file transfer is to be made between two stations connected to an open system. Assuming that some form of switched communications network is utilized, outline all of the stages that occur to accomplish the file transfer. You should make appropriate reference to each layer of the OSI Reference Model.

1.7  The OSI Reference Model is often drawn with the inclusion of an intermediate communications network at the lower three layers. This network does not necessarily have to conform to any particular set of standards, but what is obligatory or inclusion of such networks within an OSI-based design?

1.8  Compare the TCP/IP stack with that of the OSI Reference Model. Explain what implications exist concerning the TCP/IP stack since not all of the equivalent layers compared with that of the OSI Reference Model are implemented.

2.1  Data transmission techniques 19

2.2  Network topology 29

2.3  Transmission media and characteristics 32

2.4  Baseband signalling 37

In Chapter 1 we discussed the nature of data and commonly found signal types that

are conveyed by it. In this chapter we shall look at the transmission of the data over

a transmission channel. We shall consider the techniques used to transmit the data,

the different configurations in which communications networks can be arranged and

the media over which the data may be transmitted; introduce one of the principal

ways in which data may be signalled over the media; and finally examine the effect

of practical channels upon signals.

2.1 Data transmission techniques

Probably the most fundamental aspect of a data communications system is the technique

used to transmit the data between two points. The transmission path between

the two points is known as a link and the portion of the link dedicated to a particular

transmission of data is a channel. Within a data network, the two ends of the link

are generally referred to as nodes. The term station has been used to mean node in

the past, but the term more aptly describes end-system or end-user nowadays. In this

way a clear distinction may be made between intermediate nodes within a network

and end-systems, which occur at the periphery. Most data networks multiplex a number

of different data sources over a link, in which case there would be a number of

channels within the link. Multiplexing is dealt with in detail in Chapter 8.

Communication modes

Communications systems may operate in either a one-way or a two-way mode. In

the context of data communications, an example of one-way communication is a

broadcast system in which data is transmitted from a central point to a number of

receive-only stations; there is no requirement for a return channel and therefore no

interaction exists. Teletex services are one-way transmission systems. Transmission

which is confined to one direction is known as simplex operation and is illustrated in

Figure 2.1(a).

Simplex operation is limited in terms of its operational capability but it is simple

to implement since little is required by way of a protocol. It has a major limitation in

that a receiver cannot directly indicate to a transmitter that it is experiencing any difficulty

in reception.

Most data applications require a channel that allows for communication in both

directions for either some form of dialogue or interaction, as is the case with Internet

shopping or travel agent booking services. Two-way communication is also required

if data is to be retransmitted when errors have been detected and some form of repeat

request has been sent back to the transmitter. Two possibilities exist for two-way

communication:

1.      Half-duplex operation, as shown in Figure 2.1(b), can support two-way communication, but only one direction at a time. This is typical of many radio systems which, for simplicity and cheapness, employ a common channel for both directions of Figure 2.1 Communication modes: (a) simplex; (b) half duplex; (c) full duplex. transmission. If a node is transmitting it cannot receive from a distant node at the same time. Some form of protocol is therefore necessary to ensure that one node is in transmit mode and the other is in receive mode at any one time as well as to determine when stations should change state.

2.      Full-duplex operation, as shown in Figure 2.1(c), can support simultaneous twoway communication by using two separate channels, one for each direction of transmission. This is clearly more costly but is simpler to operate. Although at first glance full-duplex operation appears to be an obvious choice, it must be remembered that two-way transmission is not always necessary. Furthermore, where transmission is predominantly in one direction, with little data traffic in the reverse direction of transmission, half-duplex operation may be quite adequate.

Parallel and serial transmission

Computer-based systems store and process data in the form of bits arranged in words

of fixed size. A computer memory typically consists of a series of stores, each of which

has a unique address. Computer systems may handle words of 8, 16, 32 or 64 bits.

Within many sections of a system data exists and is passed around in parallel form,

which means that each bit of a word has its own physical conducting path. Examples

of parallel transmission may be found on printed circuit boards and in printer interface

cables. Parallel interfaces, such as a computer interconnected to a printer, need

to signal in some way when data is available from the computer and also when the

printer is, or is not, ready to receive data. The main reason for such signalling is that

very often there are appreciable differences in operating speeds between two interfaced

devices. In the case of the computer–printer example, the computer is able to

send data at a much faster rate than the printer is able to print. This facility of matching

devices is achieved using an exchange of control signals known as handshaking.

Figure 2.2 shows an example of handshaking in a simple parallel interface. The computer

first places data upon the parallel data bus. It then signals to the printer that new

data is available by changing the state of the control signal DAV (Data AVailable)

from low to high. The printer then accepts the data currently on the parallel data bus.

When the printer is ready for the next character, it takes 1low to indicate to the

computer that the data was accepted. DAV and 1are then returned to their normal

states in readiness for signalling commencement of the next data transfer. Clearly

the computer must change data on the bus only when DAV is low. Thereafter the

sequence may be repeated to pass further characters to the printer.

Parallel transmission is a convenient method of interconnection or interfacing

between devices that are situated very close together. However, if signals are to be

transferred any distance, parallel transmission becomes impractical owing to the

increased cabling and connector requirements.

The alternative approach is to use serial transmission where only a single data line

exists and the data bits are sent one after the other along the line. Although serial transmission

is inherently slower than parallel transmission, the reduced cabling arrangements

enable longer connections and this is the usual method of sending data over

any distance. Figure 2.3(a) shows a basic serial interface arrangement. Figure 2.3(b)

Figure 2.2 (a) Parallel transmission and (b) handshaking.

Figure 2.3 Serial transmission.

shows data in the form of a serial bit stream. Each bit, or signal element, of this serial

data stream occupies the same interval of time. Signals may be represented by positive

or negative logic. In Figure 2.3(b) positive logic is employed and where the amplitude

of logic 1 is A Volts in this case.

Asynchronous and synchronous operation

Any data communications system comprises, as a minimum, a transmitter, a receiver

and some form of communication channel. The transmitter produces a data stream,

the timing of each bit being generated under the control of a local clock (alternate

logic ones and zeros).

Asynchronous transmission is where a transmitter may generate bits at any

moment in time. Transmission does not necessarily communicate any knowledge of

the transmitter clock, or bit, timing to the receiver. For the receiver to interpret incoming

signals it must produce a local clock of its own which is of the correct frequency

and is positioned at, or near, the centre of each received bit in order to interpret each

received bit correctly. Transmitter and receiver use separate independent clocks which

are of similar frequency to support a particular nominal bit transmission rate. To position

the receiver clock correctly a method of working over short point-to-point links

known as start–stop operation may be employed. An example of data transmitted using

such a system is shown in Figure 2.4.

When there is no data to send the line remains in an idle state. The data is preceded

by a start signal which is normally of one signal element duration and must be

of opposite polarity to that of the idle state. This is followed by several data bits, typically

eight. Finally a stop signal is appended which has the same polarity as the idle

state. Stop elements have a time period equal to 1, 11/2 or 2 signal elements. A stop

bit is included to ensure that the start bit is clearly distinguishable from the last bit

of a preceding, contiguous, character. When the receiver detects the leading edge of

the start bit it starts the production of its receive clock. The first appearance of the

clock is so timed to fall at, or near, the centre of the first data bit and is used to strobe

the bit into a register. Subsequent clock pulses repeat the process for the remainder

of the data bits. Clearly, if the arrival rate of the received signal and the local clock

are identical, strobing will occur precisely at the centre of the bit durations. In practice,

this may not be the case for, as the number of data bits increases, the coincidence

of clock pulses and data bit centres progressively deteriorates. This limits

the length of asynchronous bit streams to a maximum of about 12 bits for reliable

detection, rendering asynchronous operation unsuitable for most data communications

networks.

Data is transmitted asynchronously using bits of 20 ms duration. Ignoring any constraints

on the number of data bits which may be accommodated between start and

stop bits, estimate the maximum number of bits which may be reliably received if the

receiver clock is operating at 48 Hz.

Assume that the start bit arranges that the receiver clock starts exactly in the centre

of the first received bit period:

Receiver bit duration = 20 ms

Receiver clock period = 1/48 Hz

=20.83 ms

It is clear from Figure 2.5(a) that after the first bit is received, the receiver clock slips

by approximately 0.83 ms on each successive bit. The problem therefore resolves into

finding how many such slips are necessary to slide beyond half of one received bit

period, or 10 ms:

10/0.83 = 12

Therefore the 12th clock pulse after the first one which was perfectly centred on bit 1,

that is the 13th actual clock pulse, occurs just on the trailing edge of the 13th bit as

may be seen in Figure 2.5(b). This means that bit 13 may or may not be successfully

received. The 14th receiver clock pulse occurs 20.83 ms after the 13th clock pulse

and is in fact 0.83 ms into the 15th bit. Hence the 14th bit is completely missed. As

a result the receiver outputs all of the bits up to and including the 12th correctly, may

or may not correctly interpret bit 13 and outputs the 15th bit received as the 14th bit.

Hence only 12 bits may be reliably received in this example.

A more efficient means of receiver clock synchronization is offered by synchronous

transmission. Data appears at a receiver as a continuous stream of regularly

timed data bits. Transmitter and receiver operate in exact synchronism. The

transmitter generates a clock which can be either transmitted to the receiver over a

separate channel or regenerated at the receiver directly from the transmitted data.

For the latter, sufficient timing information must be contained within the transmitted

signal to enable a component of the transmitter clock frequency to be recovered at

the receiver. This clock recovery process is shown in Figure 2.6.

Effective clock synchronization is an important factor in the design of data communications

networks, particularly if traffic which is sensitive to variations in time

delays, such as multimedia traffic, is to be transmitted over the network. An important

factor in ensuring that timing information is contained in transmitted data signals

is the choice of signalling technique. This issue is explored further in section 2.4. In

contrast to asynchronous operation, synchronous operation requires a continuous signal

to be transmitted, even if no data is present, to ensure that the receiver clock remains

in synchronism with that of the transmitter. Furthermore, for a receiver to interpret

the incoming bit stream correctly, some framing of bits into identifiable fixed-length

blocks or frames is necessary at the transmitter. Also, at start-up, it takes some time

Figure 2.5 Example 2.1

before a stable clock appears at the receiver. In the meantime, no data signals may be

reliably output by the receiver. Synchronism at the receiver is achieved by transmitting

a special bit sequence (or pattern) known as a preamble to enable the receiver

clock to be established. The additional circuitry at the transmitter and receiver to enable

clock recovery adds a degree of complexity to synchronous transmission which is not

required with asynchronous transmission. However, the framing bits in synchronous

transmission are relatively few compared with start and stop signalling elements in.

Figure 2.6 Synchronous receiver.

asynchronous operation. In consequence, the increased efficiency of synchronous transmission allows much higher data transmission speeds than asynchronous transmission.

Signalling rate

Transmission rate, also known as data rate, is the number of bits transmitted during

a period of time divided by that time and is measured in bits per second (bps). It is

important to distinguish between the transmission rate measured in bps and the signalling

rate or baud rate measured in baud. The signalling rate is the rate at which

an individual signalling element is transmitted and is probably best defined as the inverse

of the duration of the shortest signalling element in a transmission. The difference

between these two rates is not immediately obvious and is probably best explained by

an example.

Example 2.2

Asynchronous data is transmitted in the form of characters made up as follows: five

information bits each of duration 20 ms, a start bit of the same duration as the information

bits and a stop bit of duration 30 ms. Determine:

a)      The transmission rate in bps.

b)      The signalling rate in baud.

a)      The time taken to transmit a single character = (6 × 20) + 30 = 150 ms.

The number of bits transmitted during this time is 7.

The transmission rate = 7/(150 × 10−3) = 46.67 bps.

b)      The shortest signalling element has a duration of 20 ms, therefore the signaling rate = 1/(20 × 10−3) = 50 baud.

The difference between the bit rate and the baud rate arises because not all of the

bits are of the same duration. Another circumstance in which such a difference can

arise is if signals are modulated onto a carrier signal as outlined in Chapter 6.

ASCII code

So far, we have looked at the different techniques used to transmit data but not at

the way in which the data is represented. Text characters are normally represented

as fixed-length bit sequences. A number of characters are then grouped together in

what is, strictly speaking, an alphabet but which is more commonly called a code.

The alphabet most often used in data communications is the ITU-T alphabet number

5, or more commonly the US national version of this, known as the American

Standard Code for Information Interchange (ASCII). Each character in this alphabet

is represented by a 7-bit pattern, leading to the 27 different characters which are listed

in Table 2.1.

The alphabet consists of 96 print characters including both upper- and lower-case

letters and digits 0 to 9 (‘space’ and ‘delete’ are often included in this grouping) and

32 other characters which cannot be printed but which are associated with control functions.

This latter grouping includes ‘backspace’ and ‘carriage return’. A full list of the

ASCII control characters is provided in Table 2.2.

Table 2.1 ASCII code.

Figure 2.7 ASCII character ‘e’.

If ASCII characters are transmitted asynchronously, start and stop bits are added

along with an additional eighth bit which is known as a parity bit and is used as a

rudimentary check for errors. The parity bit is added so that the transmitted 8-bit sequences

contain only an even number of ones (even parity) or only an odd number of ones

(odd parity). Figure 2.7 shows the ASCII character ‘e’ as it would typically be transmitted

asynchronously using even parity. Note that the stop bit is twice as long as the

other bits.

At the receiver a check is made to determine whether the 8-bit sequences still have

the same parity. Any character which has not retained its parity is assumed to have

been received erroneously. If ASCII characters are transmitted synchronously, a number

of characters are normally made into frames.

Errors

In practice, transmission impairments result in data sometimes appearing in error

at the receiver. Thus a transmitted logic 0 may be output by a receiver as a logic

1 and vice versa. It is usual to express the number of errors that are likely to occur

in a system as a bit error rate (BER). A BER of 10−5 means that the probability

of a bit being received in error is 10−5. Alternatively we can say that, on average,

1 bit in every 100 000 (105) will be in error. One of the major causes of error

is noise, especially that introduced during transmission, which causes the receiver

to interpret bits wrongly on occasion. Noise and other transmission impairments,

errors and the techniques used to overcome them will be dealt with in detail in

Chapter 4.

Many systems employ some form of error control to attempt to improve the overall

BER. The simplest systems employ some form of error detection. In such a system,

the receiver may be aware of an error within a group of bits but does not know

which particular bit is in error. This system is therefore not able to correct errors. The

parity check mentioned above is a simple form of error detection which can detect

single errors. Error correction systems, however, attempt to identify the positions of

the bits which are in error and hence correct them. In a binary system, correction is

effected by a simple change of state of the offending bit (from 0 to 1 or from 1 to 0).

No error control systems can guarantee to detect or correct 100% of errors; they are

all probabilistic by nature and hence liable to miss some errors. The use of error control

only ever serves to improve the BER, although the resulting improvement is often

dramatic.

2.2 Network topology

Network topology is concerned with how communicating nodes are connected

together. The physical arrangement which is used to interconnect nodes is known

as the network topology and the process of determining a path between any two

nodes over which traffic can pass is called routing. Various topologies are shown in

Figure 2.8.

One such technology is a fully interconnected mesh, shown in Figure 2.8(a). Such

a mesh might seem an obvious first approach to interconnecting nodes. If there are n

nodes, each node requires n − 1 links to individually interconnect to each of the other

nodes. If one assumes that each such link is bidirectional, then the total number of

links required in the system may be halved and is given by:

Total number of links =

A little investigation reveals that even for modest values of n, the number of links

becomes excessive.

Figure 2.8 Network topologies: (a) fully interconnected mesh; (b) star; (c) bus; (d) ring.

Example 2.3

A network is to use a fully interconnected mesh topology to connect 10 nodes

together. How many links are required?

Number of nodes, n = 10

            Total number of links =

                                                                =             = 45

Even though the size of the network is small, an excessive number of links is necessary.

Furthermore, with 10 nodes the maximum number of links which can be in use

simultaneously is only five. Thus, with 45 links, there is a nine-fold overprovision.

An attraction of a mesh-type configuration is that each node could have equal share

in any control strategy. A node is able to control the selection of the route to another

node directly, a process known as fully distributed control. Additionally, nodes may

be made to perform an intermediate switching function to interconnect two other nodes

should their direct link fail. This is known as adaptive routing. Example 2.3 highlights

one of the disadvantages of fully interconnected networks, that is excessive link

provision except in the case of very small networks.

Minimal connectivity, popular in LANs, can be provided by a star configuration

as shown in Figure 2.8(b). For a small overhead in switching provision at the central

node, a star network may make large savings in link provision, as in Example 2.3.

Clearly, only 10 links are required, which is more than a four-fold saving. Nevertheless

there is still an overprovision by a factor of 2 in that there are twice as many

links as are strictly required. This control arrangement is called centralized control,

a good example of which is a telephone exchange.

The bus topology of Figure 2.8(c) has been used extensively by LANs. A bus

topology may also be thought of as a tree-like structure. Points to consider are the

control of user access to the system and security, since every user potentially ‘sees’

every message transmitted. Multiple simultaneous accesses to a bus, known as collisions,

are impossible to avoid with a simple bus and form an important topic of discussion

within the section on LAN access control in Section 9.1.

The ring topology, as shown in Figure 2.8(d), has also been used by LANs and

appears as either a true physical ring or a physical bus which is logically ordered as

a ring (token bus). Logical rings simply order the passage of data around users in a

predetermined and consistent circular manner. A token is normally used to control

access to the medium in a ring arrangement, hopefully in a way that is fair to all users.

Apart from the differences in access control, a bus lends itself more easily to the

addition (or subtraction) of stations than does a ring. In some instances these alterations

may be done without interruption of service. This is clearly not the case with

a ring.

Network topology 31

In reality many networks are configured using a combination of both mesh and tree

configurations and where only the network core is arranged as a full mesh. Figure 2.9

shows the broad framework for the interconnection of telephone exchanges typical of

national telephone operators. Modern telephone exchanges, or switching centres, are

effectively real-time digital computer systems. They are virtually a network of computer

systems, or a data communications network in its own right. It should be noted

in passing that, at the time of writing, considerable effort is being put into the convergence

of telephone and data switching networks. The network shown contains a

number of fully interconnected, or meshed, Digital Main Switching Units (DMSUs).

A typical network may have 50 DMSUs and each serves a number of Digital Local

Exchanges (DLEs) to which customers are connected in a star fashion.

Clearly, the choice of topology for a network depends upon factors such as the number

of nodes, geographical location, traffic usage between node pairs and link capacities.

Choice focuses upon optimization of the number of switching points and overall

link provision in terms of both their number and capacity. A single star configuration,

for instance, does not suit a national network since many of the more distant nodes

Figure 2.9 Telephone network topology.

require long links to connect to the star point. Additionally, in a single-star configuration,

there may be an inordinate number of links to deal with at the hub as well as an

overwhelming amount of associated terminating and patching equipment. An example

is if the whole of the UK were to be served by a single telephone exchange!

There are also reliability and security issues to consider and, indeed, duplicate provision

may be deliberately built in to deal with these.

2.3 Transmission media and characteristics

In addition to the transmission techniques used in a data communications system the

different media over which the data is transmitted must also be considered. Media

available for transmission of signals are twisted-pair and coaxial-pair metallic conductors

and optical fibres. Additionally, signals may be transmitted using radio waves.

Twisted-pair cable is used for baseband communication over relatively short

distances, typically in modern LAN installations. A twisted pair possesses low inductance

but high capacitance which causes substantial attenuation of signals at higher

frequencies. Some environments where twisted-pair cables are deployed contain

excessive electrical interference which may easily ingress into cable pairs and lead to

excessive signal degradation. This typically occurs in the vicinity of high-voltage equipment.

In such situations shielded twisted-pair cable may be used. Shielding consists

of a tube of metallic braid, or winding a strip of foil, around each pair within the cable

to minimize the amount of interference. Twisted-pair cables are therefore described

as unshielded twisted pair (UTP) or shielded twisted pair (STP) and examples of their

construction are shown in Figure 2.10.

Transmission media and characteristics 33

Figure 2.11 Fibre attenuation characteristics.

Coaxial-pair cable, Figure 2.10(c), was developed to overcome the deficiencies of

twisted-pair conductors by enabling much higher frequencies of operation, typically

up to hundreds of MHz. In addition, its construction produces very little electromagnetic

radiation, thus limiting any unwanted emissions and reducing the possibility for

eavesdropping. Conversely, very little interference may enter coaxial pairs, affording

a high degree of immunity to external interference. Coaxial conductors may support

data rates in excess of 100 Mbps making them suitable for use in both LANs and

WANs. The velocity of propagation in metallic conductors, twisted-pair and coaxial,

is about two-thirds that of the free space velocity of propagation of electromagnetic

radiation and is about 200 000 km/s.

Optical fibre was specifically developed to handle even higher transmission rates

with lower losses than are possible with coaxial conductors. Digital transmission is

achieved by modulating data onto an optical carrier. There are two main wavelengths

where transmission loss is extremely low and which are used with optical fibre systems.

As Figure 2.11 shows, one occurs at a wavelength of 1.3 μm and the other at

1.5 μm, both of which are in the infrared region of the electromagnetic spectrum.

The longer wavelength is used in submarine systems where it is important for losses

to be as low as possible. In Figure 2.11, losses peak at about 1.4 μm owing to water

Figure 2.12 Light propagation in optical fibres: (a) step index; (b) graded index; (c) monomode.

contamination in the manufacturing process. At shorter wavelengths below 0.8 μm

loss rises rapidly as a result of metallic impurities within the glass.

Fibres are constructed of glass or plastic and contain an inner core and an outer

cladding. The refractive index of the core η1 is slightly higher than that of the core’s

surrounding cladding, η2, and, as a result, light is contained within the core due to

multiple reflections caused by total internal reflection at the core–cladding boundary.

Earlier optical communication systems predominantly used fibre with a relatively large

core diameter. This permits light rays to follow a range of different paths within the

core of the fibre. Such multipath propagation is known as multimode operation and

is illustrated in Figure 2.12(a) and (b). As a result of multimode operation, rays arrive

at the remote end of the fibre at slightly different times, leading to a spreading of the

received signal pulses, a phenomenon known as dispersion. The earlier systems used

a fibre with a step index profile in which the refractive index of the glass changed

abruptly at the boundary between the core and cladding. The use of graded index

fibre, in which the refractive index changes gradually, in multimode operation produces

an improvement over step index fibre. Here the gradual variation of the refractive

index of the core causes the rays of light to follow curved paths, as Figure 2.12(b)

shows. Consequently, the range of path lengths is less than in a step index fibre thus

reducing dispersion and allowing higher transmission rates to be used. Transmission

Although multimode operation is still used in short-haul systems and also in LANs,

monomode or single mode (Figure 2.12(c)) operation is now dominant for higher speed

operation such as high-speed trunk links operated by PTTs. Monomode operation uses

a much smaller core diameter and eliminates all but the direct ray of light that can be

thought of as passing down the centre of the core, thus greatly reducing dispersion

and enabling current systems to operate at speeds of 10 Gbps and beyond.

Figure 2.13 Radio propagation methods: (a) line of sight; (b) over the horizon.

Wireless operation, by means of radio propagation, is difficult to summarize simply

within an introductory topic. Propagation is dependent upon the frequency of operation,

geographical location and time, and is a complex subject for design engineers.

Radio may be used to good effect in point-to-point operation, especially over difficult

or hostile terrain, and is a natural choice in many broadcast applications, obviating

the need to install many separate physical circuits. Radio communication is of course

essential for most mobile networks.

Figure 2.13 illustrates the principal radio propagation methods and indicates

approximate ranges. Line of sight (LOS) operation, as the name suggests, requires a

clear unobstructed path between transmitter and receiver which can easily be arranged

in point-to-point operation. Many mobile communication systems, such as GSM, operate

on frequencies that operate in an LOS mode but cannot always ensure that this

condition is satisfied. In such circumstances radio waves which reflect from man-made

and natural obstacles provide alternative ray paths between transmitter and receiver

and which for much of the time enable satisfactory communication. Telecommunication

satellite communications is by means of LOS operation. At frequencies below

about 100 MHz radio communication over the horizon (OTH) operation is possible

using a variety of means. Radio waves at frequencies up to a few megahertz ‘bend’

and give rise to a ground wave. Such waves may travel tremendous distances if enough

power is transmitted.

Radio waves at frequencies around 3 to 30 MHz, that is the high-frequency (HF)

band, can be made to reflect from the ionosphere to a point some considerable distance

from the transmitter, in which case they are known as sky waves. Although not

apparent in Figure 2.13(b), sky wave propagation may traverse considerably greater

distances – many thousands of km – than is indicated, as a result of multiple

reflections of the wave from both the earth’s surface and the ionosphere.

Figure 2.14 relates these propagation methods to the corresponding frequencies, and

also compares those used for metallic and optical media. Frequencies in the ultra high- frequency (UHF) band can make use of scattering within the troposphere to achieve

similar range to that of HF communication.

Optical fibre is potentially much cheaper than coaxial conductors. Optical fibre’s

inherently low loss enables propagation over several tens of km without signals

requiring any regeneration, which is an appreciably greater spacing than is possible

using coaxial conductors. Coaxial-pair conductors (and twisted-pairs) suffer from varying

degrees of interference, but this is non-existent in fibre systems.

For secure transmission, coaxial or optical media are preferred since they radiate

very little, if at all. Radio is inherently insecure and twisted-pair is easily monitored.

Where these latter media are unavoidable data encryption may be used.

2.4 Baseband signalling

In this section we shall firstly consider how data is represented as a digital signal

and then outline why signals generally require to be encoded before being sent on a

transmission line. Finally the effect of restricting the range of frequency of a signal

transmitted over a channel and ways of mitigating the effect of any distortion which

may result is examined.

Data is almost universally represented in binary form using two elements: logic 1

and logic 0. These elements are very often mapped into two signal elements but such

mapping can occur in a number of ways:

·         l Unipolar, Figure 2.15(a):

–        All elements have the same polarity. Use of the word ‘all’ indicates that in some systems there may be more than simply two signal elements, or symbols, as we shall see.

–        Note that one element may be mapped to zero volts and therefore regarded as either positive or negative, and hence be consistent with the other signal element irrespective of its polarity.

·         Bipolar, Figure 2.15(b):

–        Two (or more) signal elements have different polarity.

Now consider some ways in which data may be represented as an electrical signal.

Only binary data is considered although all of the codes could be extended to signals

containing more than two symbols.

·         Non-return to zero level (NRZL), Figure 2.16(a):

–        High level represents logic 1.

–        Low level represents logic 0.

–        Note that NRZL does not of necessity define unipolar or bipolar operation; either may equally well be used.

·         l Non-return to zero-level inverted (NRZ-I), Figure 2.16(b):

–        Logic 0, no transition at commencement of signal. That is, the signal voltage remains the same as for the previous element.

–        Logic 1, transition at commencement of signal.

Figure 2.15 Signal representations: (a) unipolar; (b) bipolar.

Figure 2.16 NRZ signals.

It is evident from the figure that NRZ-I only has a transition for each logic 1, but

no transition in the case of a zero. This means that the receiver clock must be accurately

maintained in synchronism to ensure that strings of zeros are reliably received.

·         l Return to zero (RZ), Figure 2.17:

o   Each bit, or symbol, is encoded into three distinct signal elements. The three signal elements available are 0 V, a positive voltage which we shall call +V and a negative voltage, −V.

o   Any symbol to be transmitted starts and finishes with 0 V, which gives rise to the name of ‘return to zero’.

o   Binary 1 symbol is transmitted with a +V element in the centre of the symbol, and binary 0 a −V element.

RZ is an example of a binary code where one binary signal is encoded into three

signal elements. Each signal element is selected from one of three states and for

this reason RZ is an example of a ternary code. There is a generalized method of

describing codes. RZ is classified as 1B3T, meaning that one binary signal is encoded

into three signal elements, any one of which may have one of three (ternary) states.

Line coding

Most digital signals within electronic and computer systems are unipolar NRZL. Although

binary data signals found in computer systems may be transmitted directly onto a transmission

line, in practice it is usual for some form of encoding, called line coding, to

be applied first to the signals.

Transmitting signals either directly, or via line encoding, is known as baseband

transmission. The corollary is where signal elements are impressed upon a carrier, a

process known as modulation and discussed in detail in Chapter 6. Some form of

modulation must be used over radio and optical transmission media but baseband transmission

may be used over metallic cables.

Line coding is almost invariably performed in baseband transmission systems

longer than a few tens of metres. There are three principal reasons why line coding

is used immediately prior to transmitting signals to a line:

1.      Baseline wander: Systems interconnected by a transmission line each have their own power supplies. For reasons of safety and isolation of supplies, most lines are  ac coupled. This means that a line blocks dc and low frequencies due to either inductive, or capacitive, components at one, or both, ends of the line as shown in Figure 2.18. DC blocking means that low-frequency elements of a signal are rejected, or severely attenuated. In particular, and most importantly, ac coupled lines are unable to pass a steady, or dc, voltage. If NRZ signals are applied to such a line an effect known as baseline wander may occur as shown in Figure 2.19. Here we may see that transitions successfully pass through the line. However, any constant dc voltage representing several similar bits in succession appears at the receiver as a voltage that steadily decays towards zero volts, the baseline about which signals are referenced. The problem caused by baseline wander is that signal element voltages may become too small to determine accurately if they are above, or below, the baseline and hence become misinterpreted. This is compounded by such weaker signals having reduced immunity to noise. Both of these effects lead to increased BER.

Figure 2.17 RZ signals.

2.      Receiver clock synchronism: Synchronous operation is usually employed in lines of any appreciable length (i.e. a hundred metres, or more). The absence of transitions in NRZ signals conveying long strings of logic 0, or logic 1, means that the receiver has insufficient transitions with which to synchronize. In consequence its clock will drift over time leading to errors due to making decisions at the wrong moment, for similar reasons to those discussed in Section 2.1, earlier.

Figure 2.18 AC coupling.

3.      Bandwidth of the line: All transmission media have a limited bandwidth which
places a constraint upon the maximum signalling rate. Optical and radio channels exhibit a bandpass response positioned at relatively high frequency and are therefore incapable of passing baseband signals directly. Metallic conductors intrinsically offer a low-pass response. In practice, as indicated earlier, line terminating equipment places dc blocking elements in circuit, which severely attenuates low frequencies. As a result metallic conductors usually also exhibit a bandpass response, except for very short cables. Accepting that virtually all transmission links are in effect bandpass means that transmitted signals must have a spectrum that matches such a response. In particular signals ideally must not contain dc and lowfrequency components.

Over the years a variety of line codes have been devised to mitigate the bandlimiting

effect of practical lines. Their complexity and sophistication have increased

as advances in electronic circuits have occurred. Many such codes have been devised

and currently a variety of codes are used dependent upon the characteristics of a line

in question and the features desired. The principal design goals of a line code are:

·         negligible baseline wander to avoid any appreciable deviation of signal voltage from the ideal at the receiver. This is normally achieved by arranging that over time a signal has no long strings of the same polarity. In addition there should be an equal number of positive and negative voltage signals over time, in which case the signal is said to exhibit a zero balance of voltage. That is, the long-term line voltage averages to zero.

·         relatively frequent signal transitions sufficient to support reliable clock recovery at the receiver.

·         matching the transmitted signal’s spectrum to that of the line.

Example 2.4

How well does the RZ code satisfy the generalized design criteria of a line code? Contrast

RZ coding with that of NRZ.

RZ offers no protection against baseline wander since sequences of successive ones,

or zeros, are encoded into runs of same polarity. There is also no guarantee that sequences

of symbols have zero balance.

RZ offers good clock recovery prospects. It is said to be ‘clock rich’. Unlike NRZ,

for instance, RZ coding produces two transitions for every symbol, making clock recovery

at the receiver relatively trivial.

However, with two transitions per symbol, RZ results in double the spectrum that

might reasonably be expected. What this means in practice is that for a given channel

bandwidth, if RZ is to be used the effective data rate is halved.

In summary, RZ provides for reliable clock recovery at the receiver, cannot combat

baseline wander and is rather inefficient spectrally.

RZ produces at least twice as many transitions as NRZ and so provides good prospects

for clock recovery. This contrasts with NRZ coding which only produces one signal

transition for one binary symbol and no transition for the other, which does not easily

facilitate clock recovery. However, it does mean that RZ signals produce double the

spectrum of NRZ signals. Therefore, for a given bandwidth, NRZ may operate at a

higher data rate.

The above example serves to illustrate that NRZ and RZ codes are not suited as line codes

for transmission lines. Appendix 2 explores a number of line codes used in practice.

Band-limited channels

A metallic transmission line may be represented by an equivalent electrical circuit (Dunlop

and Smith 1994) which includes series inductance and shunt capacitance. This, in addition

to any ac coupling elements which may also be present, means that transmission

lines exhibit a low-pass filtering effect which gives rise to both attenuation and phase

distortion.

Theoretically baseband signals, which are examples of digital signals, have an infinite

spectrum which at face value implies that a band-limited channel is unable to handle

such signals. However, the low-pass filtering effect of a transmission line, or indeed any

practical channel, does not automatically mean that perfect recovery of a baseband

signal at a receiver is not possible. Signalling rate R has the unit baud and equals the

reciprocal of the duration of the shortest signalling element. Perfect recovery of digital

signals is possible providing that the signalling rate does not exceed twice the available

bandwidth, W. This rate is known as the Nyquist rate (Nyquist 1928, pp617–644):

R = 1/T ≤ 2W (2.1)

where T is the duration of one signalling element, or symbol. Alternatively the Nyquist

limit, for baseband signals, may be regarded as transmission of a maximum of two symbols

per hertz of bandwidth. (Note that for signalling over channels using modulation,

signalling is at best half the rate for baseband and, depending upon the type of modulation

employed, generally less than this.) If signals are transmitted at a rate in excess

of the Nyquist rate an effect known as aliasing (Read 1998, p48) occurs at the receiver.

Analysis of the practical aspects of the low-pass filtering effect of a transmission

line upon a baseband signal requires exact knowledge of a transmission line’s particular

frequency response or transfer function. Consider the effect of a perfect filter upon

a train of pulses where the line’s frequency response is assumed to be ideal as shown

in Figure 2.20(a). We shall assume a signalling rate equal to exactly twice the filter’s

bandwidth, W. That is:

Signalling rate = 1/T = 2W (2.2)

The effect of such filtering upon the pulse train may be found by multiplying the

signal, in the frequency domain, with that of the filter’s frequency response. In practice

it is easier to determine the effect of filtering by considering the time domain.

Here the effect of the filter may be found by convolving the time domain response

of the filter with that of the digital signal. It may be shown that the time domain response

of an ideal low-pass filter is given by the expression shown in Equation (2.3) and is

a so-called sinc function, Figure 2.20(b):

h(t) = sinc(2Wt) (2.3)

Figure 2.21 Transmission of a digital signal through ideal low-pass filter: (a) digital signal;

(b) output signal components.

Determination of the effect of convolving a digital signal consisting of pulses with

the above sinc pulse is not trivial. For simplicity we shall assume that each pulse is

an impulse, or delta, function and that the signal is bipolar. Such a signal is shown in

Figure 2.21(a). Convolution of a single impulse function with the sinc function response of the filter, Figure 2.20(b), produces another sinc function, but centred upon the instant in time

of the particular impulse function. Superposition may be brought to bear and hence

the output response of the filter is the sum of the channel response to each individual

impulse function. We may therefore conclude that the overall effect of the filter upon

the train of impulse functions shown in Figure 2.21(a) is the sum of a series of sinc

functions, each similar to that shown in Figure 2.20(b), each centred upon the time

interval of their respective impulse function. The individual sinc functions for each

impulse function are shown in Figure 2.21(b). This figure does not include the final

output voltage, that is the sum of the series of time-shifted sinc functions. However,

it is clear from the idealized channel output response shown that at the sampling intervals

(1/2W, or multiples thereof ) the amplitude is a peak and due entirely to the corresponding

impulse pulse to the line. Secondly, the amplitude of all other sinc pulses

at each sampling interval other than its own is zero. The receiver clock must be per-fectly synchronized so that the decision regarding the signal’s amplitude is made at

the correct interval in time. Under this condition the only voltage present is that purely

due to the desired received symbol, in which case no other symbols interfere. This

condition means that there is no interference between different symbols, that is there

is no intersymbol interference (ISI).

The above reveals that, even with a band-limited channel, each symbol may be correctly

received. In practice baseband signals are broader than impulse functions leading

to a ‘smearing’ of the received signal. Secondly the channel response is imperfect

and introduces distortion effects, of which phase distortion is the most detrimental.

Thirdly, clock timing, and hence decision timing of the received signal, is subject to

variation or jitter. All three effects introduce ISI which, in the extreme, may completely

change the meaning, or sense, of a received symbol at the moment of decision

which in turn will give rise to an error.

In practice a transmission line, although possessing a low-pass filtering effect,

does not have an abrupt cut-off frequency as considered earlier with the ideal lowpass

filter response. A typical baseband signal received in practice might be as shown

in Figure 2.22(a). In such cases the receiver must establish a decision threshold. A

voltage which, at the decision moment, is above that of the threshold level is output

as binary 1, and vice versa. In the case of two-level (binary) signalling the threshold

is simply mid-way between the maximum range of voltage excursion of the received

signal, or half the peak-to-peak amplitude. The receiver must also establish reliable

timing of the moment at which the signal should be sampled. This should be as closely

synchronized with that of the transmitter’s data clock as possible.

A useful and simple test performed in practice to monitor the effect of a transmission

link, and associated noise and interference, is that of an eye diagram. Consider

again the received digital signal shown in Figure 2.22(a). If an oscilloscope is synchronized

with the symbol rate such that each trace displays precisely one symbol,

successive traces become superimposed as shown in Figure 2.22(b). For ease of illustration

the symbols shown in Figure 2.22(a) have been numbered. Their corresponding

appearances within the eye diagram of Figure 2.22(b) are indicated. The exact

number of traces visible will depend upon the persistence capability of the cathode

ray tube of the oscilloscope.

Figure 2.22(b) appears similar in shape to an eye, hence its name. The receiver samples

the received signal periodically at the centre of each symbol, and hence eye. It

may be clearly seen from the eye diagram that the signal amplitudes at the moment

of decision vary. This is an indication that ISI is present and caused by distortion,

noise, interference and band limiting of the channel.

Figure 2.23 shows a generalized eye diagram. The fuzzy bands are stylized lines

and due to signal and ISI, as shown earlier in Figure 2.22, and accompanying variation

due to noise. Noise increases the overall width of the bands and its effect is to

reduce the open area of the eye. The eye diagram indicates the degree of immunity

to noise, or noise margin. This is the amount by which noise may be further increased

before complete closure of the eye occurs. If this occurs, no meaningful decision may

be made and reception becomes impossible. In practice, the decision level itself has

an associated tolerance which reduces the noise margin further. An eye which has a

wide opening horizontally, or sides of small slope, is tolerant to timing error variation,

or jitter, in regard to the moment of decision. Under this condition any jitter only has a marginal effect upon noise margin and the error rate is substantially unaltered.

Where the eye is narrow and the slope is steep, tolerance to jitter diminishes rapidly

as the moment of decision moves from the optimum position. Thus eye slope indicates

tolerance, or otherwise, to timing error or jitter.

In practice an equalizer may be used to compensate for a non-ideal channel

response found in practice. An equalizer may be positioned either between the transmitter

and line, or between the line and receiver. The aim of an equalizer is to reduce

Figure 2.22 Eye diagram.

the amplitude of oscillatory tails shown in Figure 2.20 and yet preserve the zerocrossing

feature of sinc pulses. This would result in a more open eye and therefore

reduced ISI and a commensurate reduction in BER. The combined response of

channel and equalizer can be made equivalent to multiplication in time an of ideal

low-pass filter response of bandwidth equivalent to the Nyquist minimum with a suitable

frequency response to satisfy the above requirements.

A popular response sought in practice is equivalent to that produced by convolving

the ideal rectangular response seen earlier with that of a truncated raised cosine

with double the bandwidth of the filter. Figure 2.24(a) illustrates the ideal and truncated

raised cosine frequency responses desired. The desired channel and equalizer response

found by their convolution may be represented as H( f ). It may be shown that the time

domain response (Read 1998, pp139–141) h(t) is given by the expression:

h(t)= sinc 2Wt ⋅sinc 4Wπt.(1-4²W²t²)^-1

(2.4)

Note that in practice the channel response is not rectangular and therefore the equalizer

response differs from the truncated raised cosine form. The equalizer response is,

as far as possible, such that when convolved with that of the channel it nevertheless

produces the time domain response indicated in Equation (2.4).

The response to an impulse produces a time response at the equalizer output as shown

in Figure 2.24(b) where we may see that, as with the sinc response of an ideal channel,

there is zero ISI. Raised cosine pulse shaping is slightly less tolerant to timing jitter

owing to the narrower pulse but the tails decay far more rapidly, which offers superior

overall ISI performance. This improvement does mean that a channel must have

a bandwidth at least twice that of the ideal value. A trade-off must therefore be made

between reduced ISI, and hence improved BER performance, and transmission rate.

In practical transmission systems pulses are broader than impulses in order to provide

greater energy and increase immunity to noise. The equalizer response is modified

Figure 2.23 Generalized eye diagram.

Figure 2.24 Raised cosine channel response.

to produce the same overall response as that required for an impulse. Hence the time

domain response to a finite width pulse remains the same as for an impulse described

above, but will result in improved BER performance.

In some cases baseband operation is not possible. This particularly applies to the

use of RF channels and optical fibre cables. In such cases signals must be modulated

to reposition them within a suitable spectral band to match that of the medium in use.

Modulation is discussed in Chapter 6.

Exercises

2.1  For each of the services shown below, state whether a simplex, half-duplex or full-duplex form of communication is most suitable. Give reasons for your choices.

(a)    Traditional terrestrial TV

(b)   Cable TV

(c)    Pay-TV

(d)    (d) Viewing webpages

(e)    (e) File transfers

(f)    (f ) PC banking

(g)   (g) Telephone call

2.2  Compare serial and parallel transmission. Why is parallel transmission unsuited for communication over appreciable distance?

2.3  With reference to Example 2.1, explain what happens if the first appearance of the clock pulse at the receiver is not exactly in the centre of the first incoming bit.

2.4  Data is transmitted asynchronously at 50 bps. Ignoring any constraints on the number of data bits which may be accommodated between start and stop bits, estimate the maximum number of bits which may be reliably received if the receiver clock is operating at 52 Hz.

2.5  Data is transmitted asynchronously in the form of short blocks consisting of a start bit, seven information bits, a parity bit and a stop bit. All bits are of 9.09 ms duration apart from the stop bit which is of double length. Determine:

(a)    the bit rate

(b)   the baud rate

2.6  Compare synchronous and asynchronous transmission.

2.7  Explain why synchronous transmission has the potential to operate at higher transmission rates than asynchronous transmission.

2.8  A computer network is to consist of 20 stations.

(a)    Determine the number of links required if: mesh topology is used star topology is used.

(b)   Suppose that four stations are designated as hubs, each fully interconnecte with the others. The remaining 16 stations are equally distributed to the hubs, any one station only being connected to one hub. Sketch the topology and determine the number of links.

(c) Compare all three topologies from (a) and (b).

2.9  Compare the relative merits of transmitting data over copper conductors, optical fibre and radio.