Another difference is that the analogue plan allows a small number of 2-wire links to be used in the connection .sx 13.6 THE CONTROL OF SIDETONE .sx 'Sidetone' is the name given to an effect on 2-wire systems where the speaker hears his own voice in his own earphone while he is speaking .sx ( We first introduced it in Chapter 2 .sx ) Too little sidetone can make speakers think their telephone is dead , while too much leads them to lower their voices .sx CCITT recommends 'sidetone reference equivalents' of at least 17 dB .sx 13.7 THE CONTROL OF ECHO .sx By ensuring that the CCITT recommended loss of 6 to 7 dB at least is encountered between the two reference points at either end of the 4-wire part of a connection , we not only safeguard the circuit against the effects of instability ; we also make it unlikely that any signal echo is generated by the 2-to-4-wire converter ( the so-called 'hybrid' converter) .sx These echoes are caused by reflection of the speaker's voice back from the distant receiving end , and Figure 13.8 shows a 'hybrid' converter causing a signal echo of this kind .sx The problem of echo arises whenever a hybrid converter is used , and it results from the non-ideal electrical performance ( i.e. the 'mismatch' ) of the device .sx No matter whether the 4-wire line is digital or analogue , the result is an echo .sx To understand the cause of the echo we have first to consider composition of the hybrid itself .sx It consists of a 'bridge' of four pairs of wires ; one pair correspond to the 2-wire circuit , two more pairs correspond to the 'receiver' and 'transmit' pairs of the 4-wire circuit , and the final pair is a balance circuit , the function of which is explained below .sx The wires are connected to the 'bridge' in such a way as to create a separation of the receive and transmit signals on the 2-wire circuit from or on to the corresponding receive and transmit pairs of the 4-wire circuit .sx It is easiest to understand the type of hybrid which uses a pair of 'cross-coupled' transformers as a bridge .sx This is shown in Figure 13.9. Any signal generated at the telephone in Figure 13.9 appears on the transmit pair ( but not on the receive pair ) , while any incoming signal on the receive pair appears on the telephone ( but not on the transmit pair) .sx It works as follows .sx figures&captions .sx Electrical signal output from the telephone produces equal magnetic fields round windings W 1 and W 2 .sx Now the resistance in the balance circuit is set up to be equal to that of the telephone so as to induce equal fields around windings W 3 and W 4 , but the cross-coupling gives them opposite polarity .sx The fields of windings W 1 and W 3 tend to act together and to induce an output in winding W 5 .sx Conversely , the fields of windings W 2 and W 4 cancel one another ( due to the cross-coupling of W 4 ) , with the result that no output is induced in winding W 6 .sx This gives an output on the transmit pair as desired , but not on the receive pair .sx In the receive direction , the field around winding W 6 induces fields in W 2 and W 4 .sx This produces cancelling fields in windings W 1 and W 3 because of the cross-coupling of windings W 3 and W 4 .sx An output signal is induced in the 2-wire telephone circuit but not in the 'transmit' pair , winding W 5 .sx Unfortunately , the balance resistance of practical networks is rarely matched to the resistance of the telephone .sx For one thing , this is because the tolerance on workmanship in real networks is much greater .sx Also , the use of exchanges in the 2-wire part of the circuit ( if relevant ) means that it is impossible to match the balance resistance to the resistance of all the individual telephones to which the hybrid may be connected .sx The fields in the windings therefore do not always cancel out entirely as intended .sx So for example , when receiving a signal via the receive pair and winding W 6 , the fields produced in windings W 1 and W 3 may not quite cancel , and a small electric current may be induced in winding W 5 .sx This manifests itself to the speaker as an echo .sx The strength of the echo is usually denoted in terms of its decibel rating relative to the incoming signal .sx This is a value called the 'balance return loss' , or sometimes , the 'echo return loss' .sx The more efficient the hybrid , the greater the balance return loss ( the isolation between receive and transmit circuits) .sx A variety of other problems can be caused by echo , the two most important of which are :sx electrical circuit instability ( and possible 'feedback' ) .sx talker distraction .sx If the returned echo is nearly equal in volume to that of the original signal , and if a rebounding 'echo' effect is taking place at both ends of the connection , then the volume of the signal can increase with each successive echo , leading to distortion and circuit overload .sx This is circuit instability , and as we already know , the chance of it occurring is considerably reduced by adjusting the 4-wire circuit to include more signal attenuation .sx The total round-loop loss in the UK digital network shown in Figure 13.7 is at least 14 dB , probably inflicting at least 30 dB attenuation on echoes even if the hybrid has only a modest isolating performance .sx Talker distraction ( or data corruption ) is another effect of echo , but if the time delay of the echo is not too long , then distraction is unlikely , because all talkers hear their own voices anyway while they are talking .sx The echo delay time is equal to the time taken for propagation over the transmission link and back again , and is thus related to the length of the line itself .sx The longer the line , the greater the delay .sx Should the one-way signal propagation time exceed around 8 ms , giving an echo delay of 15 ms or more , then corrective action is necessary to eliminate the echo which most telephone users find obtrusive .sx A one-way propagation time of 8 ms is inevitable on all long lines over 2500 km , so that undersea cables of this length and all satellite circuits usually require echo suppression .sx Further propagation delay can also be caused by certain types of switching and transmission equipment .sx Indeed , a significant problem encountered with digital transmission media is that the time required for intermediate signal regeneration ( detection and waveform reshaping ) means that the overall speed of propagation is actually reduced to only 0.6 times the speed of light .sx This means that even quite short digital lines require echo suppression .sx Two methods of controlling echoes on long-distance transmission links are common .sx These are termed 'echo suppression' and 'echo cancellation' .sx An echo suppressor is a device inserted into the transmit path of a circuit .sx It acts to suppress retransmission of incoming 'receive path' signals by inserting a very large attenuation into the transmit path whenever a signal is detected in the receive path .sx Figure 13.10 illustrates the principle .sx The device in Figure 13.10 is called a 'half-echo suppressor' since it acts to suppress only the transmit path .sx A full echo suppressor would suppress echoes in both transmit and receive paths .sx It is normal for a long connection to be equipped with two half-echo suppressors , one at each end , the actual position being specified by the formal transmission plan .sx Ideally half-echo suppressors should be located as near to the source of the echo as possible ( i.e. as near to the 2-to-4-wire conversion point as possible ) , and work best when near the ends of the 4-wire part of the connection .sx In practice it may not be economic to provide echo suppressors at all exchanges in the lower levels of the hierarchy , and so they are most commonly provided on the long lines which terminate at regional ( class 1 of Figure 13.6 ) and international exchanges .sx Sophisticated inner-exchange signalling is used to control the use of half-echo suppressors .sx Such signalling ensures that on tandem connections of long-haul links intermediate echo suppressors are 'turned off' in the manner illustrated by Figure 13.11. This ensures that a maximum of two half-echo suppressors ( one at each end of the 4-wire part of the connection ) are active at any one time .sx figures&captions .sx The amount of suppression ( i.e. attenuation ) required to reduce the subjective disturbance of echo depends upon the number of echo paths available , the echo path propagation time , and upon the tolerance of the telephone users ( or data terminal devices) .sx CCITT recommends that echo suppression should exceed ( 15 + n ) dB , where n is the number of links on the connection .sx Unfortunately echo suppressors cannot be used in circuits carrying data , because the switching time between attenuation-on and attenuation-off states is too slow and can itself cause loss or corruption of data .sx Most data modems designed for use on telephone circuits are therefore programmed to send an initiating 2100 Hz tone over the circuit , in order to disable the echo suppressors .sx Another form of echo control device , called an 'echo canceller' , can be used on either voice or data circuits .sx Like an echo suppressor , an echo canceller has a signal detector unit in the receive path .sx However , instead of using it to switch on a large attenuator , it predicts the likely echo signal and literally subtracts this prediction from the transmit signal , thereby largely 'cancelling' out the real echo signal .sx Other signals in the transmit path should be unaffected .sx Listeners rate the performance of echo cancellers to be better than that of echo suppressors .sx This , coupled with the fact that they do not corrupt data signals , is making them a popular alternative to suppressors , common on digital line systems and exchanges , and standard equipment in some countries ( e.g. USA) .sx 13.8 SIGNAL ( OR 'PROPAGATION' ) DELAY .sx An important consideration of the network transmission plan is the overall signal delay or 'propagation time' .sx Excessive delay brings with it not only the risk of echo , but also a number of other impairments .sx In conversation , for example , long propagation times between talker and listener can lead to confusion .sx In the course of a conversation , when we have said what we want to say , we expect a fairly prompt response .sx If we are met with a silent pause , caused by a propagation delay , then we may well be tempted to speak again , to check that we have been heard .sx Inevitably , as soon as we do that , the other party starts speaking , and everyone is talking at once .sx On video the effect of signal delay is even more revealing .sx For example , on live satellite television broadcasts , whoever is at the far end always gives the impression of pausing unduly before answering any question .sx Nothing can be done to reduce the delay incurred on a physical cable or satellite transmission link .sx Thus intercontinental telephone conversations via satellite are bound to experience a one-way propagation delay of about 1/4 second , giving a pause between talking and response of 1/2 second .sx Furthermore , the extremely rapid bit speeds and response times that computer and data circuitry is capable of can be affected by line lengths of only a few centimetres or metres .sx Line lengths should therefore be minimized and circuitous routings avoided as far as possible .sx It is common for maximum physical line lengths to be quoted for data networks .sx Similarly , in telephone networks , rigid guidelines demand that double or treble satellite hops or other excessive delay paths ( i.e. those of 400 ms one-way propagation time or longer ) are avoided whenever possible .sx Excessive delays can be kept in check by appropriate network routing algorithms , as we shall see in Chapter 14 .sx 13.9 NOISE AND CROSSTALK .sx Noise and crosstalk are unwanted signals induced on to the transmission system by adjacent power lines , electrostatic interference , or other telecommunication lines .sx The only reliable way of controlling them is by careful initial planning and design of the transmission system and the route .sx One source of noise results from the induction of signals on to telecommunications cables which pass too close to high-power lines .sx Another source of noise is poorly soldered connections or component failures .sx