+ 
sin
(2pnft) + 
cos
(2pnft)
Introduction
Physical Layer
Baud Rate
Sending Bits
Telephone System
Effect of Noise
Transmission Media Types
Twisted Pair
Coaxial Cable
Radio Techniques
Microwave
Conclusion
Further Reading
This lecture seeks to explore the limitations placed upon signal transmission by transmission media and then looks at the transmission technologies used in WANs.
Limitations are placed on data transfer rates across our various transmission media.
To help look at these, let us look at the work of Fourier.
He said that any recurring signal or periodic function could be represented by a dc component plus a set of sine and cosine functions.
The general form appears below:
+ sin
(2pnft) + cos
(2pnft)
where:
g(t) is the periodic function
c is a function of the dc level
an and bn are the multipliers for sine and cosine functions of nth harmonic terms.
This can be used for periodic functions where the interval 0 to T and the interval from T to 2T are identical.
As an example, take the Fourier analysis of a square wave.
g(t) = 
sin
(2pnft)
Figure 2.1 above compares the voltages sent for each of the components of a square wave. This diagram only takes the series to the 9th harmonic which is 9 times the frequency of the fundamental frequency or first harmonic. It carries on to infinity.
This series, when fully expanded, gives us an infinite series of terms with the increasing numbers of harmonics for all odd values of n.
This is the number of signal changes per second in a channel. This is not always the same as the bit rate. Compare the bit rates for an 8 level and a 2 level signal.
In an 8 level system, each individual signal carries or conveys 3 bits ( 8 = 23) so the bps will be 3 times the baud rate.
In a 2 level system, each individual signal carries or conveys one bit ( 2 = 21) so the bps will be the same as the baud rate.
In a digital transmission system, let the bit rate be b bits per second.
Then the time to send one bit will be 
seconds.
To send a byte (= 8 bits) it will take 8 times this i.e. 
seconds
To illustrate this, if the bit rate was low, say 100 bits per second, then the time to transmit one bit would be:
second
or 0.01 seconds.
To transmit 8 bits would take 8 times this:
8 X = 
seconds
or 0.08 seconds.
The telephone system has an artificially engineered bandwidth with a cut-off frequency of around 3000 Hz.
Thus no signals above ~3000 Hz will be transmitted. Looking back at the Fourier series for a square wave, it can easily be seen that if the fundamental of the series is below 3000 Hz, the higher harmonics, being odd multiples of this will not be transmitted by the telephone system.
What effect will this have on the received signal?
Now look at this in a qualitative manner.
For example let us transmit the ASCII code for the letter b.
b = 0110001
We shall transmit 01100010, adding a parity bit. What parity are we using? Odd or even.
If we were to transmit this b over and over again it would be a repetitive signal and thus fall into the Fourier classification.
For this signal, the time to send would be seconds.
Thus the frequency of this is the reciprocal i.e. 
Hz.
If we now say that the cut-off frequency of the phone system is 3000 Hz, then the number of the highest harmonic passed is
= 
roughly.
Therefore at 9600 Baud (2 level transmission) = 9600 bps, the highest harmonic that will be transmitted is:
=
2.5 i.e. 2
We cannot use the 0.5 as the harmonics are always integer values.
The received signal will not have any frequencies above 3000 Hz within it so the received signal will not resemble a square wave as transmitted, rather a smeared out square wave with no sharp, well defined transitions.
Thermal noise will be present too and this is a function of the temperature of the medium, so can be measured. This noise further limits the transmission efficiency of our medium, reducing the bps value.
Harry Nyquist worked on transmission systems and proposed that the maximum data rate of a perfect, non-noisy channel would be
Maximum data rate = 2H log2V bits per second where:
H is the bandwidth of the medium
V is the number of signal levels
What data rate (bps) do we get for a standard telephone line having a bandwidth of 3000 Hz employing 2 level signaling?
Claude Shannon took this work further to give another expression for the maximum data rate of a medium
Maximum data rate = H log2 (1 + S/N) bits per second
The term S/N is to account for thermal noise. In the telephone system this is around 1000.
The equation proposed by Shannon is independent of the sampling rate or how many levels of signaling are used.
What data rate (bps) do we get for a standard telephone line having a bandwidth of 3000 Hz having a S/N ratio of 1000?
If we equate these two equations and simplify, we get the result that:
V = 
For a perfect channel, what does this predicts for V ?
For a telephone channel having S/N of 1000, what does this predict for the number of signaling levels?
If this is inserted back into Nyquist's equation, what data rate does this give us?
Twisted pair lines are two insulated copper cables twisted together in a spiral. Before 1988 only Category 3 UTP (unshielded twisted pair) existed. This had a loose twist. After 1988, Category 5 cabling was introduced and this had a tighter twist giving rise to less crosstalk and a higher data rate and distance of transmission. This helps to minimise electromagnetic interference (EMI) between the two wires, sometimes known as coupling.
The TP lines were originally designed to carry voice communications only. The human voice occupies a frequency range about 20 Hz to 16 kHz. The telephone system uses 300 to 3400 Hz., i.e. a bandwidth of ~3100 kHz. Thus voices can be easily understood but there is a loss of timbre (or quality).
Pairs may be bundled together into cables containing hundreds of pairs. Typical thickness of the individual cables ranges from 0.04 to 0.9 mm. This is the most common transmission medium used for telephony and digital signaling. Telephone standard cable will support 64 Kbps. In a local area network twisted pair is also used (category 5) and data transmission rates of 100 Mbps are common but are limited to station number and length. Long distance dedicated twisted pair can support 4 Mbps.
Twisted pair is cheap and easy to work with, but limited in terms of data transmission rate and length. It couples well so picks up external signals. Different twisting reduces cross-coupling. Digital signals need repeaters every 2-3 km.
This consists of two conductors, an outer shielding cable within which is a hollow insulating dielectric shield around a single inner cable. This helps increase the range of frequencies over which it can operate. Diameter can range from 10 to 25 mm. Coaxial cable enjoys widespread use in LANs and short run computer links. It can support a large number of data types and equipment within a local site. Can also support high speed I/O channels on computer systems. Using FDM (frequency division multiplexing) it can carry many signals simultaneously.
Coaxial cable can be used up to several MHz so can support higher data transmission rates over long distances than twisted pair and due to the shielding it is much less susceptible to coupling. For digital signals repeaters are needed every km or less when supporting higher data rates.
For digital signals, the impedance is 50W, this is baseband cable.
For analogue signals, the impedance is 75W, this is broadband cable.
This was the cabling used for many telephone links, but is now confined to TV with many channels. Bandwidth is 300 to 450 MHz.
The use of radio techniques to carry data is now well established. Systems such as the Orange PCN carry data to and from mobile telephones using high frequency radio waves. Voice or text messages may be carried by the system. Some mobile sets use a technology that is known as WAP to access data from the World Wide Web.
The service area is divided into a set of hexagons, at the centre of which is a base station to transmit and receive from the mobile sets. Often the base station will make use of high bandwidth microwave links to carry the signals to and from the network’s mobile switching centre (MSC). The MSCs are themselves interconnected and also connected to the Public Switched Telephone Network (PSTN). More mobile telephony resources.
The picture below is of an Orange PCN aerial. The radio aerials for communicationg with mobile sets can be seen left and right. Top centre of the structureis a microwave transmitter to connect to the MSC.
Click for more photography of base
stations.
This is a cheaper solution than fibre optic. Towers can be ~80 km apart with 100 m poles but must be within line of sight of each other. The poles are generally situated on the top of hills in line of sight. MCI (Microwave Communications Inc.) started their network very quickly and cheaply with this technology, but it does suffer from attenuation when it rains. Otherwise this is a good high bandwidth solution.
Fibre optic consists of thin strands of transparent media that are capable of transmitting an optical ray. Fused silica is best but expensive. Multicomponent glass fibres are commonly used, each strand being clad with a material of different refractive index. Light is kept within the fibres by the effect of TIR, total internal reflection that happens at the interface of two materials of differing refractive indices. These fibres can be bundled together and covered with a protective sleeve. A recent trial in California has achieved a communication rate of 80 Gbit/ sec using a technique called Dense Wave Division Multiplexing (DWDM). Theory predicts a maximum data rate of ~50 Tbps but electrical to optical conversion is our current limitation.
Fibre optic displays very low thermal noise and has a very low error ratio in the order of 1:1013. Compare this with ARPANET which had an error ratio of 1:105at the lowly data rate of 56 kbps.
Sprint (Southern Pacific Railroad International), an American carrier, has always been fibre optic and just buried fibre optic besides its tracks.
Fibre optic is smaller and lighter than coax. It displays lower attenuation than coax or twisted pair. It is also unsusceptible to external electromagnetic interference. However, it is difficult to join and does not bend to small radii.
The cost of fibre optic cabling has dropped by more than tenfold in last 15 years and capacity has increased by around the same factor. It has good security characteristics, and needs to be bent to bleed information from itself. It is now used in most long distance land based telephony not being a lossy medium to transfer information. Repeaters can be up to ~100 km apart. In a few years fibre optics will be the dominant medium for fixed installations.
Light for slower links is supplied by LEDs (light emitting diodes) and for faster links ILDs (injection laser diodes) are used, the latter being faster, more expensive but less robust. In optical fibres, light travels best in three distinct "windows" centred on 850, 1300 and 1550 nm. Local applications tend to use 850nm. This is relatively cheap and limited to data rates of <100 Mbps and distances of a few km. For longer wavelengths (i.e. higher frequencies) laser sources are required and higher data rates and distances are achieved.
Data Transmission over Fibre Optic
According to Fourier, the square wave pulses of light have different frequency components. These different frequency components have differing transmission speeds and so will interfere with adjacent pulses. This is called dispersion.
To overcome this problem, we could either reduce the data rate, not a good idea, or alternatively shape the pulses to help overcome dispersion.
The shape used in practice is related to the reciprocal of the hyperbolic cosine of the signal and the rogue components cancel each other out.
This is currently being researched and will allow pulses to be sent 1000s of km without distortion. The pulses are called solitons.
The noise is purely thermal in nature so the solution here is to increase signal strength to make S/N very small.
Despite the efforts of long haul communications companies to provide high speed networks, the limitation has traditionally been the local loop. This is the copper cable from the local exchange to the customer's premises. An estimation of the copper in use estimates that world-wide, the local loops added end to end would stretch to the moon and back 1000 times, so replacing them is not a quick solution.
These copper links provide attenuation, delay distortion and noise and are the only analogue sections of communication in a modern WAN. Echo can be a problem, but can be solved using echo suppressers and echo cancelling.
Fibre To The House, (FTTH) is not a viable solution because of the cost. Fibre To The Curb (FTTC) is a more viable solution, bringing fibre optic to within a few hundred metres of every house , then utilising the local loops once more. The short length of UTP cable then could have a bandwidth of ~1 Mbps. This would have the ability to carry compressed video.
The telephone network was originally intended purely for speech communication. It is used because the parties in communication are remote from each other. There is a procedure to perform before a conversation can occur. The user of the telephone network knows only of the call progress signals (beeps, tones etc.) and the call commands necessary (i.e. the number to dial) to set up the path. There are however a number of intervening switching nodes in between caller and called parties, and signals have to be interchanged between these nodes in order to set up the path. The network signals are transparent to the user, they neither have to initiate or answer such signals. Machines have been designed which allow calls to be sent across the telephone network and allow the exchange of data between computers. These are the familiar modems and these were at the start of what is now termed networking. Modems merely carry out the same functions as human operators and then allow communication between computers and/or peripherals.
The possibilities offered by such connections is very limited but for many years it was deemed quite satisfactory. The control necessary is also quite limited because of the simplicity of the connections required. In practice these networks could operate totally as hardware configurations.
There are three basic components of which networks are constructed; their format is determined by the technology and the type of application. In computer networks these components need not be hardware they could indeed be software and protocols which are involved in data exchange.
As far as communication networks are concerned this a fairly new term but not a new concept. For the telephone network this is the method of access and the charging regime and could, as in electro-mechanical systems, be totally described by the hardware. In computer networks this is a complex software package developed, in the first instance, from ideas contained in the familiar disc operating systems of stand-alone computers.
This comprises the hardware that enables, firstly, speech to be changed to electrical current variations and then later for data to be encoded as current variations too. Transponders such as the telephone handset or the VDU (visual display unit) provide the signals in their necessary form. In addition to single pairs of wires for individual communication paths there are methods of using cables for multiple paths or channels known as multiplexing. Early computer networks relied heavily on 'Statmuxes' or 'Concentrators' for an economic network structure.
The early type of communication involved just speech in the case of the telephone network or just text in the case of the Telex network. It was deemed satisfactory communication if the information content was communicated. Early broadcast systems were only concerned with slightly more stringent requirements that of the fidelity of the broadcast. More sophisticated systems of television and multi-media computer systems demand that more of the atmosphere of the original is present in the received transmission.
The telephone exchanges and interconnecting cables provide the network over which speech and/or data communication occurs in the public switched telephone network (PSTN). The topology or interconnection pattern defines the type of network. In the telephone network the topology is so complex and diverse that a model of the complete network is impossible however a diagram showing the typical interconnection patterns of nodes is possible. Figure 2.2 shows the hierarchical nature of this network. This is true for all large networks and is the case for the Internet too.
Note: Line thickness indicates data capacity of cables
Figure 2.2 Hierarchical Switching Structure of POTS
The idea behind such a structure is that links between local exchanges should be used if the call is between subscribers on exchanges connected by such links, provided that the links are available. A call can be routed via group switching centres when the traffic intensity means that the local route is not available and so on up through the hierarchy.
The network is organised like a tree or group of trees whose roots have grown together. The actual interconnections between the exchanges depend on the patterns of traffic entering and leaving each exchange. Each exchange is optimised for a particular function. A call requiring a service that cannot be performed by a lower class exchange is usually forwarded to the next higher exchange in the network for further processing. The network makes connections by attempting to find the shortest path from the local exchange serving the caller to the local exchange serving the called party. The high usage inter-office trunk groups which provide direct connection between exchanges of equal or lower level are used first. If they are busy, trunk groups at the next higher level are used. Digital logic circuits in the common control of each exchange make decisions on rules stored in memory that specify which trunk groups are to be tried and in which order.
The local network directly serves residential and business telephones and these are all connected to the local exchange by wire pairs (generally) which fan out like a tree from the exchange throughout the serving area. These serving areas can vary from around 10 square miles in an urban area to over 100 square miles in rural areas.
Exchange area networks are the intermediate links between the local network and the long haul network. Exchanges are interconnected with exchange area transmission systems. These systems may consist of open wire pairs on poles, wire pairs in cables, microwave radio links and fibre optic cable links. The majority of links today in UK and USA are fibre optic and the system used is SDH/ SONET (Synchronous Digital Hierarchy/ Synchronous Optical NETwork). The exchange area network normally interconnects local exchanges and tandem exchanges. Tandem exchanges are those that make connections between central offices when an interoffice trunk is unavailable. A tandem exchange is to local exchanges as a local exchange is to customer telephone equipment.
The long haul network is a set of interconnections between long distance exchanges. Each local exchange has at least one connection to a long haul exchange. These links are of high capacity and use fibre optic links (SDH/ SONET) although cable and microwave are still used in older installations. Today all backbone routes are fibre optic owing to the extremely high bandwidth offered by the medium. Each link can carry thousands of calls simultaneously.
We have seen the limitations placed upon data transmission rates by the characteristics of both the digital signals and the media upon which they are carried.
We have seen that fibre optic is the best and fastest data transmission medium available today. We have also seen the impossibility of replacing all the world's local loops due to the cost of laying the new cables.
We have seen how the telephone network is organised and the redundancy it employs for reliability.
Tannenbaum, Computer Networks, Prentice Hall:
Limitations, bandwidth limited signals, channel capacity p77
Transmission media p82
Fibre optics vs. copper cable p92
Telephone system architecture p102
Local loop limitations p108
(c) MMClements 2001 Back to Top of Page