INTRODUCTION TO YTANSMISSION FOR TELEPHONY
The basic building block for transmission is the telephone channel or voice channel.
“Voice channel” implies spectral occupancy, whether the voice path is over wire, radio, or coaxial cable or over a fiber-optic system. If a pair of wires of a simple subscriber loop is extended without loading, we can expect to see the spectral content from the average talker with frequencies as low as 20 Hz and as high as 20 kHz if the transducer of the telephone set was at all efficient across this band. Our ear, at least in younger people, is sensitive to frequencies from about 30 Hz to as high as 20 kHz. However, the primary content of a voice signal (energy plus emotion) will occupy a much narrower band of frequencies (approximately 100–4000 Hz). Considering these and other factors, we say that the nominal voice channel occupies the band from 0 to 4 kHz. CCITT defines the voice channel as the band of frequencies between 300 and 3400 Hz. Bell
Laboratories  states that “the optimum trade-off between economics and quality of transmission occurs when the telephone speech signal is band-limited to the range from about 200 to 3200 Hz.”
There are three basic impairments we must deal with regarding the voice channel.
• Attenuation distortion (frequency response)
• Phase distortion
Two additional impairments are echo and singing. We will deal with these two later.
Telecommunication System Engineering, by Roger L. Freeman
ISBN 0-471-45133-9 Copyright 2004 Roger L. Freeman
Level is another important parameter, especially in an analog network. Level must be controlled because it can surely impact quality of service (QoS).
2 THE THREE BASIC IMPAIRMENTS TO VOICE CHANNEL
2.1 Attenuation Distortion
A signal transmitted over a voice channel suffers various forms of distortion. That is, the output signal from the channel is distorted in some manner such that it is not an exact replica of the input. One form of distortion is called attenuation distortion and is the result of imperfect amplitude-frequency response. Attenuation distortion can be avoided if all frequencies within the pass band are subjected to exactly the same loss (or gain). Whatever the transmission medium, however, some frequencies are attenuated more than others. For example, on loaded wire-pair systems, higher frequencies are attenuated more than lower ones. On carrier equipment (see Section 4 of this chapter), band-pass filters are used on channel units, where, by definition, attenuation increases as the band edges are approached. Figure 5.1 is a good example of the attenuation characteristics of a voice channel operating over carrier multiplex equipment.
Attenuation distortion across the voice channel is measured against a reference frequency. The CCITT specifies 800 Hz as a reference, which is universally used in Europe, Africa, and parts of Hispanic America, whereas 1000 Hz is the common reference frequency in North America. Let us look at some ways attenuation distortion may be stated. For example, one European requirement may state that between 600 Hz and 2800 Hz the level will vary no more than −1 to +2 dB, where the plus sign means more loss and the minus sign means less loss. Thus if a signal at −10 dBm is placed at the input of the channel, we would expect −10 dBm at the output at 800 Hz (if there were no overall lossor gain), but at other frequencies we could expect a variation between −1 and
+2 dB. For instance, we might measure the level at the output at 2500 Hz at −11.9 dBm and at 1100 Hz at −9 dBm.
2.2 Phase Distortion
A voice channel may be regarded as a band-pass filter. A signal takes a finite time to pass through a telecommunication network. This time is a function of the velocity of propagation, which varies with the media involved.
The velocity of propagation also tends to vary with frequency because of the electrical characteristics associated with the network. Considering the voice channel, therefore, the velocity of propagation tends to increase toward band center and decrease toward band edge. This is illustrated in Figure 5.2, which shows relative delay across the voice channel.
The finite time it takes a signal to pass through the total extension of a voice channel or any network is called delay. Absolute delay is the delay a signal experiences while passing through the channel end-to-end at a reference frequency.
But we see that the propagation time is different for different frequencies, with the wave front of one frequency arriving before the wave front of another in the pass band. A modulated signal will not be distorted on passing through the channel if the phase shift changes uniformly with frequency, whereas if the phase shift is nonlinear with respect to frequency, the output signal is distorted compared to the input.
In essence we are dealing with phase linearity of a circuit. If the phase– frequency relationship over a pass band is not linear, distortion will occur in the transmitted signal. This phase distortion is often measured by a parameter called envelope delay distortion (EDD). Mathematically, envelope delay is the derivative of the phase shift with respect to frequency. The maximum difference in the derivative over any frequency interval is called envelope delay distortion. Therefore
EDD is always a difference between the envelope delay at one frequency and that at another frequency of interest in the pass band. Note that envelope delay is often defined the same as group delay—that is, the ratio of change, with angular frequency, of the phase shift between two points in a network .
2.2.1 Notes on Phase Distortion. Absolute delay is minimum around 1700 and 1800 Hz in the voice channel. This is shown in Figure 5.2. The figure also shows that around 1700 or 1800 Hz, envelope delay distortion is flattest. It is for this reason that so many data modems use 1700 or 1800 Hz for the characteristic tone frequency which is modulated by the data.
This brings up the next point. Phase distortion (or EDD) has little effect on speech communications over the telecommunication network. However, for data transmission, phase distortion is the greatest bottleneck for data rate (i.e., number of bits per second that the channel can support). It has probably more effect on limiting data rate than any other parameter .
2.3.1 General. Noise, in its broadest definition, consists of any undesired signal in a communication circuit. The subject of noise and noise reduction is probably the most important single consideration in analog transmission engineering.
It is the major limiting factor in system performance. For the discussion in this text, noise is broken down into four categories:
1. Thermal noise
2. Intermodulation noise
4. Impulse noise
2.3.2 Thermal Noise. Thermal noise occurs in all transmission media and all communication equipment, including passive devices. It arises from random electron motion and is characterized by a uniform distribution of energy over the frequency spectrum with a Gaussian distribution of levels.
Every equipment element and the transmission medium proper contribute thermal noise to a communication system if the temperature of that element or medium is above absolute zero. Thermal noise is the factor that sets the lower limit of sensitivity of a receiving system and is often expressed as a temperature, usually given in units referred to absolute zero. These units are kelvins.
Thermal noise is a general expression referring to noise based on thermal agitations.
The term “white noise” refers to the average uniform spectral distribution of noise energy with respect to frequency. Thermal noise is directly proportional to bandwidth and temperature. The amount of thermal noise to be found in 1 Hz of bandwidth in an actual device is Pn = kT (W/Hz) (5.1) where k is Boltzmann’s constant, equal to 1.3803 × 10−23 J/K, and T is the absolute temperature (K) of the circuit (device). At room temperature, T = 17◦C or 290 K; thus
Pn = 4.00 × 10−21 W/Hz of bandwidth = −204 dBW/Hz of bandwidth = −174 dBm/Hz of bandwidth
For a band-limited system (i.e., a system with a specific bandwidth), Pn = kTB (W), where B refers to the so-called noise bandwidth in hertz. Thus at 0 K we obtain Pn = −228.6 dBW/Hz of bandwidth; for a system with a noise bandwidth measured in hertz (B) and whose noise temperature is T we obtain Pn = −228.6 dBW + 10 log T + 10 logB (5.2)
2.3.3 Intermodulation Noise. Intermodulation (IM) noise is the result of the presence of intermodulation products. If two signals with frequencies F1 and F2 are passed through a nonlinear device or medium, the result will contain IM products that are spurious frequency energy components. These components may be present either inside and/or outside the band of interest for a particular device.
IM products may be produced from harmonics of the desired signals in question, either as products between harmonics or as one of the signals and the harmonic of the other(s) or between both signals themselves. The products result when two (or more) signals beat together or “mix.” Look at the mixing possibilities when passing F1 and F2 through a nonlinear device. The coefficients indicate the first, second, or third harmonics.
• Second-order products F1 ± F2
• Third-order products 2F1 ± F2; 2F2 ± F1
• Fourth-order products 2F1 ± 2F2; 3F1 ± F2 . . .
Devices passing multiple signals simultaneously, such as multichannel radio equipment, develop intermodulation products that are so varied that they resemble white noise. Intermodulation noise may result from a number of causes:
• Improper level setting. If the level of input to a device is too high, the device is driven into its nonlinear operating region (overdrive).
• Improper alignment causing a device to function nonlinearly.
• Nonlinear envelope delay. • Device malfunction.
To summarize, intermodulation noise results from either a nonlinearity or a malfunction that has the effect of nonlinearity. The cause of intermodulation noise is different from that of thermal noise. However, its detrimental effects and physical nature can be identical with those of thermal noise, particularly in multichannel systems carrying complex signals [12, 14].
2.3.4 Crosstalk. Crosstalk refers to unwanted coupling between signal paths.
There are essentially three causes of crosstalk: (1) electrical coupling between transmission media, such as between wire pairs on a voice-frequency (VF) cable system, (2) poor control of frequency response (i.e., defective filters or poor filter design), and (3) nonlinear performance in analog (FDM) multiplex systems.
Excessive level may exacerbate crosstalk.
There are two types of crosstalk:
1. Intelligible, where at least four words are intelligible to the listener from extraneous conversation(s) in a 7-s period.
2. Unintelligible: crosstalk resulting from any other form of disturbing effects of one channel on another.
Intelligible crosstalk presents the greatest impairment because of its distraction to the listener. Distraction is considered to be caused either by fear of loss of privacy or primarily by the user of the primary line consciously or unconsciously trying to understand what is being said on the secondary or interfering circuits; this would be true for any interference that is syllabic in nature.
Received crosstalk varies with the volume of the disturbing talker, the loss from the disturbing talker to the point of crosstalk, the coupling loss between the two circuits under consideration, and the loss from the point of crosstalk to the listener. The most important of these factors for this discussion is the coupling loss between the two circuits under consideration. Talker levels have been discussed elsewhere in this text. Also, we must not lose sight of the fact that the effects of crosstalk are subjective, and other factors have to be considered when crosstalk impairments are to be measured. Among these factors are the type of people who use the channel, the acuity of listeners, traffic patterns, and operating practice [3, 4].
2.3.5 Impulse Noise. Impulse noise is noncontinuous, consisting of irregular pulses or noise “spikes” of short duration, broad spectral density, and relatively high amplitude. In the language of the trade, these spikes are often called “hits.”
A technician may say that the circuit is getting “hit up.” Impulse noise degrades telephony ordinarily only marginally, if at all. However, it may seriously degrade data error performance on data or other digital waveforms. We discuss impulse noise further when covering data transmission in Chapter 10.
Level is a primary parameter in the analog network. By “primary” we mean very important. With the digital network level it is of secondary importance. In the context of this book, when we use the word level, we mean signal magnitude.
Level could be comparative. The output of an amplifier is 20 dB higher than the input. But more commonly, we mean absolute level, and in telephony it is measured in dBm (decibels referenced to 1 milliwatt) or in milliamperes. In radio (wireless) systems, we will more likely employ dBW (decibels referenced to 12 watts). When dealing with video systems (e.g., television), the unit of measure is voltage. The commonly derived unit is the dBmV, meaning decibels referenced to 1 millivolt.
In the telecommunication network, if levels are too high, amplifiers become overloaded, resulting in intermodulation and other types of distortion such as crosstalk. If levels are too low, customer satisfaction may suffer with a degraded loudness rating.
System levels are important parameters when engineering a telecommunication system. The values are usually taken from a level chart or a reference system drawing made by a planning group or as a part of an engineered job. On the chart a 0 TLP (zero test level point, refer to Chapter 3, Section 15) is established. A test-level point is a location in a circuit or system at which a specified test-tone level is expected during alignment. A 0 TLP is a point at which the test-tone level should be 0 dBm. Here the decibel unit, the dBr,∗ may enter the discussion.
As we briefly discussed in Chapter 3, the dBm can be related to dBr and dBm0 by the following formula: dBm = dBm0 + dBr (5.3)
For instance, a value of −32 dBm at a −22-dBr point corresponds to a reference level of −10 dBm0. A −10-dBm0 signal introduced at the 0 dBr point (0 TLP) has an absolute signal level of −10 dBm [4–6].
2.4.1 Typical Levels. Earlier measurement of speech level used the unit of measure VU, standing for volume unit. For 1000-Hz sinusoid signal, 0 VU = 0 dBm. When a VU meter is used to measure the level of a voice signal, it is difficult to equate VU and dBm. One of the problems, of course, is that speech transmission is characterized by spurts of signal. The damping on a VU meter tends to smooth out the spurts of voice signal. We can relate VU to dBm in the following formula:
Average power of a telephone talker ≈ VU − 1.4 (dBm) (5.4) Today, when discussing speech transmission, we often talk about equivalent peak level (EPL) and long-term conversational level. The unit of measurement for both is the dBm0 (dBm referenced to the zero test level point or 0 TLP).
EPL can be described approximately as the 95% point on the cumulative probability distribution of instantaneous talker power. Periods of silence are excluded from the distributio
In the long-distance network, EPLs can be characterized by a mean of −11.1 dBm0 with a standard deviation of 4.7 dB. Average power for calls in the long-distance network can be characterized by a mean of −25.5 dBm0 and a standard deviation of 5.3 dB. For the local area, the corresponding quantities are a mean EPL of −11.8 dBm0 with a standard deviation of 4.7 dB and a mean average power of −26.5 dBm0 with a standard deviation of 5.4 dB. The distributions of EPL and average power are approximately log-normal (i.e., normal when the independent variable is expressed in dB units) [4, 9, 12, and 13].
2.5 Signal-to-Noise Ratio
When dealing with transmission engineering, signal-to-noise (S/N) ratio is perhaps more frequently used than any other criterion when designing a telecommunication system. S/N ratio expresses in decibels the amount by which a signal level exceeds the noise within a specified bandwidth.
As we review several types of material to be transmitted, each will require a minimum S/N ratio to satisfy the customer or to make the receiving instrument function within certain specified criteria. We might require the following S/N ratios with the corresponding end instruments:
• Voice: 40 dB
_ based on customer satisfaction.
• Voice: 45 dB
• Data: ∼15 dB, based on a specified error rate and modulation type.
In Figure 5.3 a 1000-Hz signal has an S/N ratio of 10 dB, assuming a nominal 4-kHz bandwidth for his example. The level of the noise is +5 dBm, and the signal level is +15 dBm.
TWO-WIRE AND FOUR-WIRE TRANSMISSION
3.1 Two-Wire Transmission
A telephone conversation inherently requires transmission in both directions.
When both directions are carried on the same pair of wires, it is called two-wire transmission. The telephones in our homes and offices are connected to a local switching center (exchange) by means of two-wire circuits. A more proper definition for transmitting and switching purposes is that when oppositely directed portions of a single telephone conversation occur over the same electrical transmission channel or path, we call this two-wire operation.
3.2 Four-Wire Transmission
Carrier and radio systems require that oppositely directed portions of a single conversation occur over separate transmission channels or paths (or use mutually exclusive time periods). Thus we have two wires for the transmit path and two wires for the receive path, or a total of four wires for a full-duplex (two-way) telephone conversation. For almost all telephone systems, the end instrument (i.e., the telephone subset) is connected to its local serving exchange on a two-wire basis. In other words, the subscriber loop is two-wire.
In fairly well developed nations, the output of the local serving exchange, looking toward the toll network, is four-wire. In many less developed nations the two-wire to four-wire conversion does not take place until the output of the toll-connecting exchange. This is the same point, of course, where the A/D conversion (conversion to PCM, Chapter 8) takes place.
To simplify the explanation, Figure 5.4 illustrates a typical PSTN network showing the two-wire-to-four-wire conversion from the calling-subscriber end, and the converse, conversion from four-wire to two-wire at the called subscriber end. Schematically, the four-wire interconnection is shown as if it were a single channel wire-line system with amplifiers. However, it would be more likely be a multichannel digital carrier system on cable, fiber-optic light guide, and/or multiplex over radio. The amplifiers in Figure 5.4 serve to convey the ideas this section considers. As we detail in the figure, conversion from two-wire operation to four-wire is carried out by a terminating set, more commonly referred to in the industry as a term set, which contains a four-port balanced transformer (ahybrid) or, less commonly, a resistive network.
Operation of a Hybrid
A hybrid, in terms of telephony (at voice frequency), is a transformer. For a simplified description, a hybrid may be viewed as a power splitter with four sets of wire-pair connections. A functional block diagram of a hybrid device is shown in Figure 5.5. Two of the wire-pair connections belong to the four wire path, which consists of a transmit pair and a receive pair. The third pair is the connection to the two-wire link that is eventually connected to the subscriber subset via one or more switches. The last wire pair of the four connects the hybrid to a resistance–capacitance balancing network, which electrically balances the hybrid with the two-wire connection to the subscriber subset over the frequency range of the balancing network. An artificial line may be used for this purpose.
Signal energy entering from the two-wire subset connection divides equally, half of it dissipating in the impedance of the four-wire side receive path and the other half going to the four-wire side transmit path, as shown in Figure 5.5. Here the ideal situation is that no energy is to be dissipated by the balancing network
(i.e., there is a perfect balance). The balancing network is supposed to display the characteristic impedance of the two-wire line (subscriber connection) to the hybrid. Signal energy entering from the four-wire side receive path is also split in half in the ideal situation where there is perfect balance. Half of the energy is dissipated by the balancing network (N) and half at the two-wire port (L)
The reader notes that in the description of the hybrid, in every case, ideally half of the signal energy entering the hybrid is used to advantage and half is dissipated or wasted. Also keep in mind that any passive device inserted in a circuit, such as a hybrid, has an insertion loss. As a rule of thumb, we say that the insertion loss of a hybrid is 0.5 dB. Thus there are two losses here that the reader must not lose sight of:
Hybrid insertion loss 0.5 dB
Hybrid dissipation loss 3.0 dB (half of the power) 3.5 dB (total)
As far as this section is concerned, any signal passing through a hybrid suffers a 3.5-dB loss. This is a good design number for gross engineering practice.
Hybrids used on short loops, however, may have higher losses, as do special resistance-type hybrids.
In Figure 5.5, consider the balancing network (N) and the two-wire side of the hybrid (L). For the older, conventional analog network, the two-wire side of the hybrid connected the subscriber through at least one two-wire switch. Because of the switch, the two-wire side of the hybrid could look into at least 10,000 possible subscriber connections—some short loops, some long loops, and other loops in poor condition. Because of the fixed conditions on the four-wire side, we can pretty much depend on holding a good impedance match. Our concern under these conditions is impedance match on the two-wire side. That is the match between the compromise network (N) and the two-wire side (L).
We measure the capability of impedance match by return loss. In this particular case we call it balance return loss:
Balance return lossdB = 20 log10
ZL + ZN
ZL − ZN
Let’s say, for argument’s sake, that we have a perfect match. In other words the impedance of the two-wire subscriber loop side on this particular call was exactly
900 _ and the balancing network (N) was 900 _. Substitute these numbers in the formula above and we get
Balance return lossdB = 20 log
900 + 900
900 − 900
Examine the denominator. It is zero. Any number divided by zero is infinity.
Thus we have an infinitely high return loss. And this happens when we have a perfect match, an ideal condition. Of course it is seldom realized in real life. In real life we find that the balance return loss for a large population of hybrids connected in service and serving a large population of two-wire users has a median more of the order of 11 dB with a standard deviation of 3 dB .
This is valid for North America. For some other areas of the world, balance return loss median may be lower with a larger standard deviation.
When return loss becomes low (i.e., there is a poor impedance match), there is a reflection of the speech signal. This is echo. We define the cause of echo as any impedance mismatch in the network. Most commonly this mismatch occurs at the hybrid. Echo that is excessive becomes singing. This is caused by high positive feedback on intervening amplifiers. Singing is a highly undesirable impairment