Biblio5

190 LOCAL AND LONG-DISTANCE NETWORKS

Figure 8.11 Adaptive or state-dependent routing, network information (status) versus routing decisions.

8.4.2.2.2 Dynamic Routing Scheme. Routing schemes may also incorporate frequent automatic variations. Such changes may be time-dependent, state-dependent, and/ or event-dependent. The updating of routing patterns make take place periodically, aperiodically, predetermined, depending on the state of the network, or depending on whether calls succeed or fail in the setup of a route.

In time-dependent routing, routing patterns are altered at fixed times during the day or week to allow for changing traffic demands. It is important to note that these changes are preplanned and are implemented consistently over a long time period.

In state-dependent routing, routing patterns are varied automatically according to the state of the network. These are called adaptive routing schemes. To support such a routing scheme, information is collected about the status of the network. For example, each toll exchange may compile records of successful calls or outgoing trunk occupancies. This information may then be distributed through the network to other exchanges or passed to a centralized database. Based on this network status information, routing decisions are made either in each exchange or at a central processor serving all exchanges. The concept is shown in Figure 8.11.

In event-dependent routing, patterns are updated locally on the basis of whether calls succeed or fail on a given route choice. Each exchange has a list of choices, and the updating favors those choices that succeed and discourages those that suffer congestion.

8.4.2.3 Route Selection. Route selection is the action taken to actually select a definite route for a specific call. The selection may be sequential or nonsequential. In the case of sequential selection, routes in a set are always tested in sequence and the first available route is selected. For the nonsequential case, the routes are tested in no specific order.

The decision to select a route can be based on the state of the outgoing circuit group or the states of series of circuit groups in the route. In either case, it can also be based on the incoming path of entry, class of service, or type of call to be routed.

8.4.3 Call-Control Procedures

Call-control procedures define the entire set of interactive signals necessary to establish, maintain, and release a connection between exchanges. Two such call-control procedures are progressive call control and originating call control.

8.4.3.1 Progressive Call Control. This type of call control uses link-by-link signaling (see Chapter 7) to pass supervisory controls sequentially from one exchange to the next. Progressive call control can be either irreversible or reversible. In the irreversible

8.4 TRAFFIC ROUTING IN A NATIONAL NETWORK

191

case, call control is always passed downstream toward the destination exchange. Call control is reversible when it can be passed backwards (maximum of one node) using automatic rerouting or crankback actions.

8.4.3.2 Originating Call Control. In this case the originating exchange maintains control of the call set-up until a connection between the originating and terminating exchanges has been completed.

8.4.4 Applications

8.4.4.1 Automatic Alternative Routing. One type of progressive (irreversible) routing is automatic alternative routing (AAR). When an exchange has the option of using more than one route to the next exchange, an alternative routing scheme can be employed. The two principal types of AAR that are available are:

1. When there is a choice of direct-circuit groups between two exchanges; and

2. When there is a choice of direct and indirect routes between the two exchanges.

Alternative routing takes place when all appropriate circuits in a group are busy. Several circuit groups then may be tested sequentially. The test order is fixed or time-dependent.

8.4.4.2 Automatic Rerouting (Crankback). Automatic rerouting (ARR) is a routing facility enabling connection of call attempts encountering congestion during the initial call setup phase. Thus, if a signal indicating congestion is received from Exchange B, subsequent to the seizure of an outgoing trunk from Exchange A, the call may be rerouted at A. This concept is shown in Figure 8.12. ARR performance can be improved through the use of different signals to indicate congestion—S1 and S2 (see Figure 8.12).

•S1 indicates that congestion has occurred on outgoing trunks from exchange B.

•S2 indicates that congestion has occurred further downstream—for example, on outgoing trunks from D.

Figure 8.12 The automatic rerouting (ARR) or crankback concept. Note: Blocking from B to D activates signal S1 to A. Blocking from D to F activates signal S2 to A. (From Figure 4/ E.170 of Ref. 3.)

192 LOCAL AND LONG-DISTANCE NETWORKS

Figure 8.13 An example of preplanned distribution of load sharing. Note: Each outgoing routing pattern (A, B, C, D) may include alternative routing options.

The action to be taken at exchange A upon receiving S1 or S2 may be either to block the call or to reroute it.

In the example illustrated in Figure 8.12, a call from A to D is routed via C because the circuit group B-D is congested (S1-indicator) and a call from A to F is routed via E because circuit group D-F is congested (S2 indicator).

One positive consequence of this alternative is to increase the signaling load and number of call set-up operations resulting from the use of these signals. If such an increase is unacceptable, it may be advisable to restrict the number of reroutings or limit the signaling capability to fewer exchanges. Of course, care must be taken to avoid circular routings (“ring-around-the-rosy”), which return the call to the point at which blocking previously occurred during call setup.

8.4.4.3 Load Sharing. All routing schemes should result in the sharing of traffic load between network elements. Routing schemes can, however, be developed to ensure that call attempts are offered to route choices according to a preplanned distribution. Figure 8.13 illustrates this application to load sharing, which can be made available as a software function of SPC exchanges. The system works by distributing the call attempts to a particular destination in a fixed ratio between the specified routing patterns.

8.4.4.4 Dynamic Routing. Let us look at an example of state-dependent routing. A centralized routing processor is employed to select optimum routing patterns on the basis of actual occupancy level of circuit groups and exchanges in the network which are monitored on a periodic basis (e.g., every 10 s). Figure 8.14 illustrates this concept.

Figure 8.14 State-dependent routing example with centralized processor.

8.4 TRAFFIC ROUTING IN A NATIONAL NETWORK

193

Figure 8.15 Example of time-dependent routing. (From Figure 7/ E.170 of Ref. 3.)

In addition, qualitative traffic parameters may also be taken into consideration in the determination of the optimal routing pattern.

This routing technique inherently incorporates fundamental principles of network management in determining routing patterns. These principles include:

•Avoiding occupied circuit groups;

•Not using overloaded exchanges for transit; and

•In overload circumstances, restriction of routing direct connections.

Now let’s examine an example of time-dependent routing. For each originating and terminating exchange pair, a particular route pattern is planned depending on the time of day and day of the week. This is illustrated in Figure 8.15. A weekday, for example, can be divided into different time periods, with each time period resulting in different route patterns being defined to route traffic streams between the same pair of exchanges.

This type of routing takes advantage of idle circuit capacity in other possible routes between originating and terminating exchanges which may exist due to noncoincident busy hours.11 Crankback may be utilized to identify downstream blocking on the second link of each two-link alternative path.

The following is an example of event-dependent routing. In a fully connected (mesh) network, calls between each originating and terminating exchange pair try the direct route with a two-link alternative path selected dynamically. While calls are successfully routed on a two-link path, that alternative is retained. Otherwise, a new two-link alternative path is selected. This updating, for example, could be random or weighted by the success of previous calls. This type of routing scheme routes traffic away from congested links by retaining routing choices where calls are successful. It is simple, adapts quickly to changing traffic patterns, and requires only local information. Such a scheme is illustrated in Figure 8.16 (Refs. 3, 4).

11Noncoincident busy hours: in large countries with two or more time zones.

194 LOCAL AND LONG-DISTANCE NETWORKS

Figure 8.16 Event-dependent routing in a mesh network.

8.5 TRANSMISSION FACTORS IN LONG-DISTANCE TELEPHONY

8.5.1 Introduction

Long-distance analog communication systems require some method to overcome losses. As a wire-pair telephone circuit is extended, there is some point where loss accumulates such as to attenuate signals to such a degree that the far-end subscriber is dissatisfied. The subscriber cannot hear the near-end talker sufficiently well. Extending the wire connections still further, the signal level can drop below the noise level. For a good received signal level, a 40-dB signal-to-noise ratio is desirable (see Sections 3.2.1 and 3.2.2.4). To overcome the loss, amplifiers are installed on many wire-pair trunks. Early North American transcontinental circuits were on open-wire lines using amplifiers quite widely spaced. However, as BH demand increased to thousand of circuits, the limited capacity of such an approach was not cost effective.

System designers turned to wideband radio and coaxial cable systems where each bearer or pipe carried hundreds or thousands of simultaneous telephone conversations.12 Carrier (frequency division) multiplex techniques made this possible (see Section 4.5). Frequency division multiplex (FDM) requires separation of transmit and receive voice paths. In other words, the circuit must convert from two-wire to four-wire transmission. Figure 8.17 is a simplified block diagram of a telephone circuit with transformation from two-wire to four-wire operation at one end and conversion back to two-wire operation at the other end. This concept was introduced in Section 4.4.

12On a pair of coaxial cables, a pair of fiber optic light guides, or a pair of radio-frequency carriers, one coming and one going.

8.5 TRANSMISSION FACTORS IN LONG-DISTANCE TELEPHONY

195

Figure 8.17 Simplified schematic of two-wire/ four-wire operation.

The two factors that must be considered that greatly affect transmission design in the long-distance network are echo and singing.

8.5.2 Echo

As the name implies, echo in telephone systems is the return of a talker’s voice. To be an impairment, the returned voice must suffer some noticeable delay. Thus we can say that echo is a reflection of the voice. Analogously, it may be considered as that part of the voice energy that bounces off obstacles in a telephone connection. These obstacles are impedance irregularities, more properly called impedance mismatches. Echo is a major annoyance to the telephone user. It affects the talker more than the listener. Two factors determine the degree of annoyance of echo: its loudness and the length of its delay.

8.5.3 Singing

Singing is the result of sustained oscillation due to positive feedback in telephone amplifiers or amplifying circuits. The feedback is the result of excessive receive signal feeding back through the hybrid to the transmit side, which is then amplified setting up oscillations. Circuits that sing are unusuable and promptly overload multichannel carrier (FDM) equipment.

Singing may be regarded as echo that is completely out of control. This can occur at the frequency at which the circuit is resonant. Under such conditions the circuit losses at the singing frequency are so low that oscillation will continue, even after cessation of its original impulse.

8.5.4 Causes of Echo and Singing

Echo and singing can generally be attributed to the impedance mismatch between the balancing network of a hybrid and its two-wire connection associated with the subscriber loop. It is at this point that we can expect the most likelihood of impedance mismatch which may set up an echo path. To understand the cause of echo, one of two possible conditions may be expected in the local area network:

196 LOCAL AND LONG-DISTANCE NETWORKS

1. There is a two-wire (analog) switch between the two-wire/ four-wire conversion point and the subscriber plant. Thus, a hybrid may look into any of (say) 10,000 deifferent subscriber loops. Some of these loops are short, other are of medium length, and still others are long. Some are in excellent condition, and some are in dreadful condition. Thus the possibility of mismatch at a hybrid can be quite high under these circumstances.

2. In the more modern network configuration, subscriber loops may terminate in an analog concentrator before two-wire/ four-wire conversion in a PCM channel bank. The concentration ratio may be anywhere from 2 : 1 to 10 : 1. For example, in the 10 : 1 case a hybrid may connect to any one of a group of ten subscriber loops. Of course, this is much better than selecting any one of a population of thousands of subscriber loops as in condition 1, above.

Turning back to the hybrid, we can keep excellent impedance matches on the fourwire side; it is the two-wire side that is troublesome. So our concern is the match (balance) between the two-wire subscriber loop and the balancing network (N in Figure 8.17). If we have a hybrid term set assigned to each subscriber loop, the telephone company (administration) could individually balance each loop, greatly improving impedance match. Such activity has high labor content. Secondly, in most situations there is a concentrator with from 4 : 1 to 10 : 1 concentration ratios (e.g., AT&T 5ESS).

With either condition 1 or condition 2 we can expect a fairly wide range of impedances of two-wire subscriber loops. Thus, a compromise balancing network is employed to cover this fairly wide range of two-wire impedances.

Impedance match can be quantified by return loss. The higher the return loss, the better the impedance match. Of course we are referring to the match between the balancing network (N) and the two-wire line (L) (see Figure 8.17).

Return LossdB c 20 log10(ZN + ZL)/ (ZN − ZL).

(8.1)

If the balancing network (N) perfectly matches the impedance of the two-wire line (L), then ZN c ZL, and the return loss would be infinite.13

We use the term balance return loss (Ref. 5) and classify it as two types:

1. Balance return loss from the point of view of echo.14 This is the return loss across the band of frequencies from 300 to 3400 Hz.15

2. Balance return loss from the point of view of stability.16 This is the return loss between 0 and 4000 Hz.

“Stability” refers to the fact that loss in a four-wire circuit may depart from its nominal value for a number of reasons:

•Variation of line losses and amplifier gains with time and temperature;

•Gain at other frequencies being different from that measured at the test frequency. (This test frequency may be 800, 1000, or 1020 Hz.)

13Remember, for any number divided by zero, the result is infinity.

14Called echo return loss (ERL) in North America, but with a slightly different definition. 15Recognize this as the CCITT definition of the standard analog voice channel.

16From the point of view of stability—for this discussion, it may be called from the point of view of singing.

8.5 TRANSMISSION FACTORS IN LONG-DISTANCE TELEPHONY

197

• Errors in making measurements and lining up circuits.

The band of frequencies most important in terms of echo for the voice channel is that from 300 Hz to 3400 Hz. A good value for echo return loss for toll telephone plant is 11 dB, with values on some connections dropping to as low as 6 dB. For further information, the reader should consult CCITT Recs. G.122 and G.131 (Refs. 5, 6).

Echo and singing may be controlled by:

•Improved return loss at the term set (hybrid);

•Adding loss on the four-wire side (or on the two-wire side); and

•Reducing the gain of the individual four-wire amplifiers.

The annoyance of echo to a subscriber is also a function of its delay. Delay is a function of the velocity of propagation of the intervening transmission facility. A telephone signal requires considerably more time to traverse 100 km of a voice-pair cable facility, particularly if it has inductive loading, than it requires to traverse 100 km of radio facility (as low as 22,000 km/ s for a loaded cable facility and 240,000 km/ s for a carrier facility). Delay is measured in one-way or round-trip propagation time measured in milliseconds. The CCITT recommends that if the mean round-trip propagation time exceeds 50 ms for a particular circuit, an echo suppressor or echo canceler should be used. Practice in North America uses 45 ms as a dividing line. In other words, where echo delay is less than that stated previously here, echo can be controlled by adding loss.

An echo suppressor is an electronic device inserted in a four-wire circuit that effectively blocks passage of reflected signal energy. The device is voice operated with a sufficiently fast reaction time to “reverse” the direction of transmission, depending on which subscriber is talking at the moment. The block of reflected energy is carried out by simply inserting a high loss in the return four-wire path. Figure 8.18 shows the echo path on a four-wire circuit. An echo canceller generates an echo-canceling signal.17

Figure 8.18 Echo paths in a four-wire circuit.

17Echo canceller, as defined by CCITT, is a voice-operated device placed in the four-wire portion of a circuit and used for reducing near-end echo present on the send path by subtracting an estimation of that echo from the near-end echo (Ref. 7).

198 LOCAL AND LONG-DISTANCE NETWORKS

Figure 8.19 Talker echo tolerance for average telephone users.

8.5.5 Transmission Design to Control Echo and Singing

As stated previously, echo is an annoyance to the subscriber. Figure 8.19 relates echo path delay to echo path loss. The curve in Figure 8.19 traces a group of points at which the average subscriber will tolerate echo as a function of its delay. Remember that the longer the return signal is delayed, the more annoying it is to the telephone talker (i.e., the more the signal has to be attenuated). For example, if the echo delay on a particular circuit is 20 ms, an 11-dB loss must be inserted to make the echo tolerable to the talker. Be careful here. The reader should note that the 11 dB designed into the circuit to control echo will increase the end-to-end loudness loss (see Section 3.2.2.4) an equal amount, which is quite undesirable. The effect of loss design on loudness ratings and the trade-offs available are discussed in the paragraphs that follow.

If singing is to be controlled, all four-wire paths must have some amount of loss. Once they go into a gain condition, and we refer here to overall circuit gain, positive feedback will result and the amplifiers will begin to oscillate or “sing.” For an analog network, North American practice called for a minimum of 4-dB loss on all four-wire circuits to ensure against singing. CCITT recommends 10 dB for minimum loss on the national network. (Ref. 5, p. 3).

The modern digital network with its A/ D (analog-to-digital) circuits in PCM channel banks provides signal isolation, analog-to-digital, and digital-to-analog. As a result, the entire loss scenario has changed. This new loss plan for digital networks is described in Section 8.5.7.

8.5.6 Introduction to Transmission-Loss Engineering

One major aspect of transmission system design for a telephone network is to establish a transmission-loss plan. Such a plan, when implemented, is formulated to accomplish three goals:

1. Control singing (stability);

2. Keep echo levels within limits tolerable to the subscriber; and

3. Provide an acceptable overall loudness rating to the subscriber.

8.5 TRANSMISSION FACTORS IN LONG-DISTANCE TELEPHONY

199

From preceding discussions we have much of the basic background necessary to develop a transmission-loss plan. We know the following:

•A certain minimum loss must be maintained in four-wire circuits to ensure against singing.

•Up to a certain limit of round-trip delay, echo may be controlled by adding loss (i.e., inserting attenuators, sometimes called pads).

•It is desirable to limit these losses as much as possible, to improve the loudness rating of a connection.

National transmission plans vary considerably. Obviously the length of a circuit is important, as well as the velocity of propagation of the transmission media involved.

Velocity of propagation. A signal takes a finite amount of time to traverse from point A to point B over a specific transmission medium. In free space, radio signals travel at 3 × 108 m/ sec or 186,000 mi/ sec; fiber-optic light guide, about 2 × 108 m/ s or about 125,000 mi/ sec; on heavily loaded wire-pair cable, about 0.22 × 108 m/ sec or 14,000 mi/ sec; and 19-gauge nonloaded wire-pair cable, about 0.8 × 108 m/ sec or 50,000 mi/ sec. So we see that the velocity of propagation is very dependent on the types of transmission media being employed to carry a signal.

Distances covered by network connectivities are in hundreds or thousands of miles (or kilometers). It is thus of interest to convert velocities of propagation to miles or kilometers per millisecond. Let’s use a typical value for carrier (multiplex) systems of 105,000 miles/ sec or 105 miles per millisecond (169 km/ ms).

First let’s consider a country of small geographic area such as Belgium, which could have a very simple transmission-loss plan. Assume that the 4-dB minimum loss for singing is inserted in all four-wire circuits. Based on Figure 8.19, a 4-dB loss will allow up to 4 ms of round-trip delay. By simple arithmetic, we see that a 4-dB loss on all fourwire circuits will make echo tolerable for all circuits extending 210 mi (338 km) (i.e., 2 × 105). This could be an application of a fixed-loss type transmission plan. In the case of small countries or telephone companies covering a rather small geographic expanse, the minimum loss to control singing controls echo as well for the entire system.

Let us try another example. Assume that all four-wire connections have a 7-dB loss. Figure 8.20 indicates that 7 dB permits an 11-ms round-trip delay. Again assume that the velocity of propagation is 105,00 mi/ sec. Remember that we are dealing with round-trip delay. The talker’s voices reach the far-end hybrid and some of the signal is reflected back to the talker. This means that the signal traverses the system twice, as shown in Figure 8.20. Thus 7 dB of loss for the given velocity of propagation allows about 578 mi (925 km) of extension or, for all intents and purposes, the distance between subscribers, and will satisfy the loss requirements with a country of maximum extension of 578 mi (925 km).

It is interesting to note that the talker’s signal is attenuated only 7 dB toward the distant-end listener; but the reflected signal is not only attenuated the initial 7 dB, but attenuated by 7 dB still again, on its return trip.

It has become evident by now that we cannot continue increasing losses indefinitely to compensate for echo on longer circuits. Most telephone companies and administrations have set a 45or 50-ms round-trip delay criterion, which sets a top figure above which echo suppressors are to be used. One major goal of the transmission-loss plan is to improve overall loudness rating or to apportion more loss to the subscriber plant so that subscriber loops can be longer or to allow the use of less copper (i.e., smaller-diameter