Biblio5

In such an operation timing error only accumulates inside a character. Suppose the receiving device is 5% slower or faster than the transmitting device. Only 40% (5 × 8) timing error accumulates, well inside of the 50% maximum. Remember that the sampling takes place at mid-bit position of 50% of its timing interval.

Minimum lengths of stop elements vary, depending on the convention used on a particular network. In the commercial world, the stop element can be one or two unit intervals (bits) duration. With some older systems, the stop element was 1.42 unit intervals duration. In military operation the stop element is 1.5 unit intervals long. The start space is always 1 unit interval duration.

As we are aware, a primary objective in the design of data systems is to minimize errors in a received bit stream or to minimize the error rate. Two of the prime causes of errors are noise and improper timing relationships. With start–stop systems a character begins with a mark-to-space transition at the beginning of the start-space. Then 1.5-unit intervals later, the timing causes the receiving device to sample the first information element, which is simply a 1 or 0 decision. The receiver continues to sample at onebit intervals until the stop mark is received. It knows a priori where (when) it should occur. In start–stop systems the last information bit is the most susceptible to cumulative timing errors.

Synchronous data transmission is another story altogether. It is much more efficient because it does not have start and stop elements, which are really overhead. Synchronous data transmission consists of a continuous serial stream of information elements or bits, as illustrated in Figure 10.5. With start–stop systems, timing error could only accumulate inside a character, the eight bits of the example given previously. This is not so for synchronous systems. Timing error can accumulate for the entire length of a frame or over many frames.

With start–stop systems, the receiving device knows when a character starts by the mark-to-space transition at the start-space. In a synchronous transmission system, some marker must be provided to tell the receiver when a frame starts. This “marker” is the unique field. Every data frame starts with a unique field. A generic data frame is shown in Figure 10.7. View the frame from left to right. The first field is the unique field or flag, and it generally consists of the binary sequence 01111110. A frame always starts with this field and ends with the same field. If one frame follows another contiguously, the unique field ending frame No. 1 is the unique field starting frame No. 2, and so forth. Once a frame knows where it starts, it will know a priori where the following fields

Figure 10.7 A generic data frame. Flag pattern c unique field. Text is often called “info” field or information field.

272 DATA COMMUNICATIONS

begin and end by simple bit/ octet counting. However, with some data link protocols, the information field is of variable length. In this case, the information field length will appear as a subfield in the control field.

Synchronous data transmission systems not only require frame alignment, but also must be bit-aligned. This was an easy matter on start–stop systems because we could let the receive clock run freely inside the 5 bits or 8 bits of a character. There is no such freedom with synchronous systems. Suppose we assume a free-running receive clock. However, if there was a timing error of 1% between the transmit and receive clocks, not more than 100 bits could be transmitted until the synchronous receiving device would be off in timing by the duration of 1 bit from the transmitter, and all bits received thereafter would be in error. Even if the timing accuracy of one relative to the other were improved to 0.05%, the correct timing relationship between transmitter and receiver would exist for only the first 2000 bits transmitted. It follows, therefore, that no timing error whatsoever can be permitted to accumulate since anything but absolute accuracy in timing would cause eventual malfunctioning. In practice, the receiver is provided with an accurate clock that is corrected by small adjustments based on the transitions of the received bit stream, as explained in Section 10.7.3.

10.7.3 Timing

All currently used data-transmission systems are synchronized in phase and symbol rate in some manner. Start–stop synchronization has already been discussed. All fully synchronous transmission systems have timing generators or clocks to maintain stability. The transmitting device and its companion receiver at the far end of the circuit must maintain a timing system. In normal practice, the transmitter is the master clock of the system. The receiver also has a clock that in every case is corrected by some means to its transmitter’s master clock equivalent at the far end.

Another important timing factor is the time it takes a signal to travel from the transmitter to the receiver. This is called propagation time. With velocities of propagation as low as 20,000 mi/ s, consider a circuit 200 mi in length. The propagation time would then be 200/ 20,000 s or 10 ms. Ten milliseconds is the time duration of 1 bit at a data rate of 100 bps; thus the receiver in this case must delay its clock by 10 ms to be in step with its incoming signal. Temperature and other variations in the medium may also affect this delay, as well as variations in the transmitter master clock.

There are basically three methods of overcoming these problems. One is to provide a separate synchronizing circuit to slave the receiver to the transmitter’s master clock. However, this wastes bandwidth by expending a voice channel or subcarrier just for timing. A second method, which was quite widely used until twenty years ago, was to add a special synchronizing pulse for groupings of information pulses, usually for each character. This method was similar to start–stop synchronization, and lost its appeal largely because of the wasted information capacity for synchronizing. The most prevalent system in use today is one that uses transition timing, where the receiving device is automatically adjusted to the signaling rate of the transmitter by sampling the transitions of the incoming pulses. This type of timing offers many advantages, particularly automatic compensation for variations in propagation time. With this type of synchronization the receiver determines the average repetition rate and phase of the incoming signal transition and adjusts its own clock accordingly by means of a phase-locked loop.

In digital transmission the concept of a transition is very important. The transition is what really carries the information. In binary systems the space-to-mark and mark- to-space transitions (or lack of transitions) placed in a time reference contain the infor-

mation. In sophisticated systems, decision circuits regenerate and retime the pulses on the occurrence of a transition. Unlike decision circuits, timing circuits that reshape a pulse when a transition takes place must have a memory in case a long series of marks or spaces is received. Although such periods have no transitions, they carry meaningful information. Likewise, the memory must maintain timing for reasonable periods in case of circuit outage. Note that synchronism pertains to both frequency and phase and that the usual error in high-stability systems is a phase error (i.e., the leading edges of the received pulses are slightly advanced or retarded from the equivalent clock pulses of the receiving device). Once synchronized, high-stability systems need only a small amount of correction in timing (phase). Modem internal timing systems may have a long-term stability of 1 × 10−8 or better at both the transmitter and receiver. At 2400 bps, before a significant timing error can build up, the accumulated time difference between transmitter and receiver must exceed approximately 2 × 10−4 s. Whenever the circuit of a synchronized transmitter and receiver is shut down, their clocks must differ by at least 2 × 10−4 s before significant errors take place once the clocks start back up again. This means that the leading edge of the receiver–clock equivalent timing pulse is 2 × 10−4 in advance or retarded from the leading edge of the pulse received from the distant end. Often an idling signal is sent on synchronous data circuits during periods of no traffic to maintain the timing. Some high-stability systems need resynchronization only once a day.

Note that thus far in our discussion we have considered dedicated data circuits only. With switched (dial-up) synchronous circuits, the following problems exist:

•No two master clocks are in perfect phase synchronization.

•The propagation time on any two paths may not be the same.

Thus such circuits will need a time interval for synchronization for each call set-up before traffic can be passed.

To summarize, synchronous data systems use high-stability clocks, and the clock at the receiving device is undergoing constant but minuscule corrections to maintain an in-step condition with the received pulse train from the distant transmitter, which is accomplished by responding to mark-to-space and space-to-mark transitions. The important considerations of digital network timing were also discussed in Chapter 6.

10.7.4 Bits, Bauds, and Symbols

There is much confusion among professionals in the telecommunication industry over terminology, especially in differentiating, bits, bauds, and symbols. The bit, a binary digit, has been defined previously.

The baud is a unit of transmission rate or modulation rate. It is a measure of transitions per second. A transition is a change of state. In binary systems, bauds and bits per second (bps) are synonymous. In higher-level systems, typically m-ary systems, bits and bauds have different meanings. For example, we will be talking about a type of modulation called QPSK. In this case, every transition carries two bits. Thus the modulation rate in bauds is half the bit rate.

The industry often uses symbols per second and bauds interchangeably. It would be preferable, in our opinion, to use “symbols” for the output of a coder or other conditioning device. For the case of a channel coder (or encoder), bits go in and symbols come out. There are more symbols per second in the output than bits per second in the

274 DATA COMMUNICATIONS

input. They differ by the coding rate. For example, a 1/ 2 rate coder (used in FEC) may have 4800 bps at the input and then would have 9600 symbols per second at the output.

10.7.4.1 Period of a Bit, Symbol, or Baud. The period of a bit is the time duration of a bit pulse. When we use NRZ (nonreturn-to-zero) coding (discussed in Section 10.7.5), the period of a bit, baud, or symbol is simply 1/ (bit rate), 1/ (baud rate) or 1/ (symbol rate). For example, if we were transmitting 9600 bits per second, what is the period of a bit? It is 1/ 9600 c 104.16 msec. For 2400 baud c 1/ 2400 c 416.6 msec; for 33.6 kbps c 1/ 33,600 c 0.0297 msec, or 29.7 ns.

10.7.5 Digital Data Waveforms

Digital symbols may be represented in many different ways by electrical signals to facilitate data transmission. All these methods for representing (or coding) digital symbols assign electrical parameter values to the digital symbols. In binary coding, of course, these digital symbols are restricted to two states, space (0) and mark (1). The electrical parameters used to code digital signals are levels (or amplitudes), transitions between different levels, phases (normally 08 and 1808 for binary coding), pulse duration, and frequencies or a combination of these parameters. There is a variety of coding techniques for different areas of application, and no particular technique has been found to be optimum for all applications, considering such factors as implementing the coding technique in hardware, type of transmission technique employed, decoding methods at the data sink or receiver, and timing and synchronization requirements.

In this section we discuss several basic concepts of electrical coding of binary signals. In the discussion reference is made to Figure 10.8, which graphically illustrates several line coding techniques. Figure 10.8a shows what is still called by many today neutral transmission (also see Figure 10.4). This was the principal method of transmitting telegraph signals until about 1960. In many parts of the world, neutral transmission is still widely employed. First, the waveform is a nonreturn-to-zero (NRZ) format in its most elementary form. Nonreturn-to-zero simply means that if a string of 1s (marks) is transmitted, the signal remains in the mark state with no transitions. Likewise, if a string of 0s is transmitted, there is no transition and the signal remains in the 0 state until a 1 is transmitted. A “0” (zero) infers the 0 voltage line as shown in the figure. Figure 10.8c is the conventional NRZ waveform most encountered in the data industry.

Figures 10.8b and 10.8d show the typical return-to-zero (RZ) waveform, where, when a continuous string of marks (1s) or spaces (0s) is transmitted, the signal level (amplitude) returns to the zero voltage condition at each element or bit. Obviously RZ transmission is much richer in transitions that NRZ.

In Section 10.6.2 we discussed neutral and polar transmission systems. Figure 10.8a shows a typical neutral waveform where the two state conditions are 0 V for the mark or binary 1 condition and some positive voltage for the space or binary 0 condition. On the other hand, in polar transmission, as illustrated in Figures 10.8c and 10.8d, a positive voltage represents a space and a negative voltage a mark. With NRZ transmission, the pulse width is the same as the duration of a unit interval or bit. Not so with RZ transmission, where the pulse width is less than the duration of a unit interval. This is because we have to allow time for the pulse to return to the zero (voltage) condition.

Bi-phase-L or Manchester coding (Figure 10.8e) is a code format that is being used ever more widely on digital systems such as on certain local area networks (LANs) (see Chapter 11). Here binary information is carried in the transition. By convention, a logic 0 is defined as a positive-going transition and a logic 1 as a negative-going transition.

Figure 10.8 Digital data transmission waveforms.

Note that Manchester coding has a signal transition in the middle of each unit interval (bit). Manchester coding is a form of phase coding.

10.8 DATA INTERFACE: THE PHYSICAL LAYER

When we wish to transmit data over a conventional analog network, the electrical representation of the data signal is essentially direct current (i.e., has a frequency of 0 Hz). As such it is incompatible with that network that accepts information channels in the band 300 –3400 Hz.6 A data modem is a device that brings about this compatibility. It translates the electrical data signal into a modulated frequency tone in the range of 300 –3400 Hz. In most cases with modern modems that tone is 1800 Hz. Also, more often than not, the digital network (as described in Chapter 6) has extensions that are analog and require the same type(s) of modem. If the digital network is extended to a user’s premise, a digital conditioning device (CSU/ DSU) is required for bit rate and waveform compatibility.

For this discussion of data interface, we will call the modem or digital conditioning

6This is the conventional CCITT voice channel.

276 DATA COMMUNICATIONS

Figure 10.9 Data circuit interfaces, physical layer.

device data communication equipment (DCE). This equipment has two interfaces, one on each side, as shown in Figure 10.9. The first, which is discussed in this section, is on the user side, which is called data terminating equipment (DTE), and the applicable interface is the DTE–DCE interface. The second interface is on the line side, which is covered in Section 10.9. It should be noted that the DTE-DCE interface is well defined.

The most well-known DTE–DCE standard was developed by the (U.S.) Electronics Industries Association (EIA) and is called EIA-232E (Ref. 5). It is essentially equivalent to international standards covered by CCITT Recs. V.24 and V.28, and ISO IS2110.

EIA-232E and most of the other standards discussed here are applicable to the DTE–DCE interface employing serial binary data interchange. It defines signal characteristics, mechanical interface characteristics, and functional descriptions of the interchange circuits. EIA-232E is applicable for data transmission rates up to 20,000 bps and for synchronous/ asynchronous serial binary data systems.

Section 2.1.3 is quoted from EIA-232E (Ref. 5). It is crucial to the understanding of signal state convention and level:

For data interchange circuits, the signal shall be considered in the marking condition when the voltage (V1) on the interchange circuit, measured at the interface point, is more negative than minus three volts with respect to circuit AB (signal ground). The signal shall be considered in the spacing condition when the voltage V1 is more positive than plus three volts with respect to circuit AB. . . . The region between plus three volts and minus three volts is defined as the transition region. The signal state is not uniquely defined when the voltage (V1) is in this transition region.

During the transmission of data, the marking condition is used to denote the binary state ONE and the spacing condition is used to denote the binary state ZERO.

Besides EIA-232 there are many other interface standards issued by EIA, CCITT, U.S. federal standards, U.S. military standards, and ISO. Each defines the DTE–DCE interface. Several of the more current standards are briefly described in the following.

EIA-530 (Ref. 7) is a comparatively recent standard developed by the EIA. It provides for all data rates below 2.1 Mbps and it is intended for all applications requiring a balanced electrical interface.7 It can also be used for unbalanced operation.

Let us digress for a moment. An unbalanced electrical interface is where one of the signal leads is grounded; for a balanced electrical interface, no ground is used.

EIA-530 applies for both synchronous and nonsynchronous (i.e., start–stop) operation. It uses a standard 25-pin connector; alternatively it can use a 26-pin connector. A list of interchange circuits showing circuit mnemonic, circuit name, circuit direction (meaning toward DCE or toward DTE), and circuit type is presented in Table 10.2.

7EIA-530 is more properly called ANSI/ EIA/ TIA-530-A.

278 DATA COMMUNICATIONS

In amplitude shift keying, the two binary conditions are represented by a tone on and a tone off. ASK, used alone, is not really a viable alternative in today’s digital network. It is employed in many advanced hybrid modulation schemes.

Frequency shift keying is still used for lower data rate circuits (≤ 1200 bps). The two binary states are each represented by a different frequency. At the demodulator the frequency is detected by a tuned circuit. So, of course, there are two tuned circuits, one for each frequency. Binary 0 is represented by the higher frequency and binary 1 by the lower frequency.

Phase shift keying is widely used with data modems, both alone and with hybrid modulation schemes. The most elementary form is binary PSK. In this case there are just two phases, which are separated one from the other by 1808 phase difference. The demodulator is based on a synchronous detector using a reference signal of known phase. This known signal operates at the same frequency of the incoming signal carrier and is arranged to be in phase with one of the binary signals. In the relative-phase system a binary 1 is represented by sending a signal burst of the same phase as that of the previous signal burst sent. A binary 0 is represented by a signal burst of a phase opposite to that of the previous signal transmitted. The signals are demodulated at the receiver by integrating and storing each signal burst of 1-bit period for comparison in phase with the next signal burst.

10.9.3 Critical Impairments to the Transmission of Data

The effect of various telephone circuit parameters on the capability of a circuit to transmit data is a most important consideration. The following discussion is intended to familiarize the reader with problems most likely to be encountered in the transmission of data over analog or mixed analog/ digital circuits. We make certain generalizations in some cases, which can be used to facilitate planning the implementation of data systems.

10.9.3.1 Phase Distortion. Phase distortion “constitutes the most limiting impairment to data transmission, particularly over telephone voice channels” (Ref. 8). When specifying phase distortion, the terms envelope delay distortion (EDD) and group delay are often used. The IEEE (Ref. 1) states that envelope delay is often defined the same as group delay; that is, the rate of change, with angular frequency, of the phase shift between two points in a network.8

Let us try to put this another way that will be easier to understand. We are dealing with the voice channel which will be transported in an analog network or in a mixed analog/ digital network. It is a band-limited system. Phase distortion arises from the fact that not all frequency components of the input signal will propagate to the receiving end in exactly the same elapsed time. This is true when the signal passes through filters or their equivalent. Particularly troublesome are loaded cable circuits and FDM carrier circuits. Even PCM channel banks have filters.

Figure 10.10 shows a typical frequency-delay response curve in milliseconds of a voice channel due to FDM equipment only. For the voice channel (or any symmetrical passband, for that matter), delay increases toward band edge and is minimum about band center (around 1800 Hz). As the bit rate is increased on the channel, there is more and more opportunity for the delayed energy from bit 1 to spill into the principal energy component of bit 2. When it does, we call it ISI (intersymbol interference).

Phase or delay distortion is the major limitation of modulation rate on the voice

8Angular frequency is 2p × frequency in Hz. It is sometimes called the radian frequency.

Figure 10.10 Typical differential delay across a voice channel, FDM equipment back-to-back. “Back- to-back” means that the transmit and receive portions of the equipment are placed on a test bench and appropriately interconnected.

channel. The shorter the pulse width (the width or period of 1 bit in binary systems), the more critical will be the EDD parameters. As we will discuss in Section 10.9.5, it is desirable to keep delay distortion in the band of interest below the period of 1 baud.

10.9.3.2 Attenuation Distortion (Frequency Response). Another parameter that seriously affects the transmission of data and can place definite limits on the modulation rate is attenuation distortion. Ideally, all frequencies across the passband of a channel of

real working system.

In Rec. G.132 (Ref. 9), the CCITT recommends no more than 9 dB of attenuation distortion relative to 800 Hz between 400 Hz and 3000 Hz. This figure, 9 dB, describes the maximum variation that may be expected from the reference level at 800 Hz. This variation of amplitude response is often called attenuation distortion. A conditioned channel, such as a Bell System C-4 channel, will maintain a response of 2 dB to +3 dB from 500 Hz to 3000 Hz and 2 dB to +6 dB from 300 Hz to 3200 Hz− .

Considering tandem operation, −the deterioration of amplitude response is arithmetically cumulative when sections are added. This is particularly true at band edge in view of channel unit transformers and filters that account for the upper and lower cutoff characteristics. Figure 10.11 illustrates a typical example of attenuation distortion (amplitude response) across carrier equipment back-to-back. Attenuation distortion, phase distortion, and noise were introduced in Section 3.3.

10.9.3.3 Noise. Another important consideration in the transmission of data is noise. All extraneous elements appearing at the voice channel output that were not due to the input signal are considered to be noise. For convenience, noise is broken down into four categories: (1) thermal, (2) crosstalk, (3) intermodulation, and (4) impulse (Ref. 27). Thermal noise, often called “resistance noise,” “white noise,” or “Johnson noise,”

280 DATA COMMUNICATIONS

Figure 10.11 Typical attenuation distortion across a voice channel, carrier equipment back-to-back.

is of a Gaussian nature or completely random. Any system or circuit operating at a temperature above absolute zero inherently will display thermal noise. The noise is caused by the random noise of discrete electrons in the conduction path. Crosstalk is a form of noise caused by unwanted coupling from one signal path into another. It may be caused by direct inductive or capacitive coupling between conductors or between radio antennas. Intermodulation noise is another form of unwanted coupling, usually caused by signals mixing in nonlinear elements of a system. Carrier and radio systems are highly susceptible to intermodulation noise, particularly when overloaded. Impulse noise can be a primary source of errors in the transmission of data over telephone networks. It is sporadic and may occur in bursts or discrete impulses called “hits.” Some types of impulse noise are natural, such as that from lightning. However, man-made impulse noise is ever increasing, such as that from automobile ignition systems and power lines. Impulse noise may be of a high level in conventional analog telephone switching centers as a result of dialing, supervision, and switching impulses that may be induced or otherwise coupled into the data-transmission channel. The worst offender in the switching area is the step-by-step exchange, some of which are still around. Impulse noise in digital exchanges is almost nonexistent.

For our discussion of data transmission, two types of noise are considered, (1) random (or Gaussian) noise and (2) impulse noise. Random noise measured with a typical transmission measuring set appears to have a relatively constant value. However, the instantaneous value of the noise fluctuates over a wide range of amplitude levels. If the instantaneous noise voltage is of the same magnitude as the received signal, the receiving detection equipment may yield an improper interpretation of the received signal and an error or errors will occur. Thus we need some way of predicting the behavior of data transmission in the presence of noise. Random noise or white noise has a Gaussian distribution and is considered representative of the noise encountered on a communications channel, which, for this discussion is the voice channel.9 From the probability distribution curve of Gaussian noise shown in Figure 10.12, we can make some statistical

9The definition of the term Gaussian distribution is beyond the scope of this book. It essentially means to us that this type of noise is well characterized and we will use the results of this characterization such as found in Figure 10.12.