Information transmitted over a communication line is usually subjected to special coding, which improves the reliability of transmission. In this case, additional hardware costs for encoding and decoding are inevitable, and the cost of network adapters increases.

The coding of information transmitted over a network is related to the ratio of the maximum allowable transmission rate and the bandwidth of the transmission medium used. For example, with different codes, the maximum transmission rate over the same cable can differ by a factor of two. The complexity of the network equipment and the reliability of information transmission also directly depend on the chosen code.

To transmit discrete data over communication channels, two methods of physical encoding of initial discrete data are used - based on a sinusoidal carrier signal and based on a sequence of rectangular pulses. The first way is often called analog modulation, because coding is carried out by changing the parameters of the analog signal (amplitude, phase, frequency). The second way is called digital coding. At present, data having an analog form (speech, television image) is transmitted via communication channels in a discrete form. The process of representing analog information in discrete form is called discrete modulation.

5.1Analog modulation

The representation of discrete data as a sinusoidal signal is called analog modulation. Analog modulation allows you to represent information as a sinusoidal signal with different levels of amplitude, or phase, or frequency. You can also use combinations of changing parameters - amplitude and frequency, amplitude-phase. For example, if you form a sinusoidal signal with four amplitude levels and four frequency levels, this will give 16 states of the information parameter, which means 4 bits of information for one change.

There are three main types of analog modulation:

amplitude,

frequency,

Amplitude modulation.(AM) With amplitude modulation, for a logical one, one level of the amplitude of the carrier frequency sinusoid is selected, and for a logical zero, another (see Fig. 5.1). The frequency of the signal remains constant. This method is rarely used in its pure form in practice due to low noise immunity, but is often used in combination with another type of modulation - phase modulation.

Rice. 5.1 Different types of modulation

Frequency modulation. ( World Cup) With frequency modulation, the values ​​of logical 0 and logical 1 of the initial data are transmitted by sinusoids with different frequencies - f 1 and f 2 (see Fig. 5.1). The signal amplitude remains constant. This modulation method does not require complicated circuits in modems and is usually used in low speed modems.

Phase modulation. (FM) With phase modulation, the values ​​of logical 0 and 1 correspond to signals of the same frequency, but with a different phase (inverted), for example, 0 and 180 degrees or 0.90,180 and 270 degrees. The resulting signal looks like a sequence of inverted sine waves (see Figure 5.1). The amplitude and frequency of the signal remain constant.

Combined modulation methods are used to increase the transmission rate (increase the number of bits per one cycle of the information parameter). The most common methods quadrature amplitude modulation (Quadrature Amplitude Modulation, QAM). These methods use a combination of phase modulation with 8 phase shift values ​​and amplitude modulation with 4 amplitude levels. With this method, 32 signal combinations are possible. And although not all of them are used, the speed is still significantly increased, and due to redundancy, errors in data transmission can be controlled. For example, in some codes, only 6, 7 or 8 combinations are allowed to represent the original data, and the remaining combinations are prohibited. This coding redundancy is required for the modem to recognize erroneous signals, which are the result of distortion due to interference, which on telephone channels, especially switched ones, are very significant in amplitude and long in time.

Let's determine on which lines analog modulation can work, and to what extent this method satisfies the bandwidth of one or another used transmission line, for which we consider the spectrum of the resulting signals. For example, take the amplitude modulation method. The spectrum of the resulting signal with amplitude modulation will consist of a sinusoid of the carrier frequency f With and two side harmonics:

(f With -f m ) and (f With +f m ), where f m- modulation frequency (changes in the information parameter of the sinusoid), which will coincide with the data rate if two amplitude levels are used.

Rice. 5.2 Signal spectrum with amplitude modulation

Frequency f m determines the throughput of the line at this method coding. With a low modulation frequency, the width of the signal spectrum will also be small (equal to 2f m see Figure 5.2), so signals will not be distorted by the line if its bandwidth is greater than or equal to 2f m .

Thus, with amplitude modulation, the resulting signal has a narrow spectrum.

With phase and frequency modulation, the signal spectrum is more complex than with amplitude modulation, since more than two side harmonics are formed here, but they are also symmetrically located relative to the main carrier frequency, and their amplitudes decrease rapidly. Therefore, these types of modulation are also well suited for data transmission over lines with narrow bandwidths. A typical representative of such lines is the voice frequency channel, which is made available to users of public telephone networks.

From the typical frequency response of a voice-frequency channel, it can be seen that this channel transmits frequencies in the range from 300 to 3400 Hz, and thus its bandwidth is 3100 Hz (see Figure 5.3).

Rice. 5.3 frequency response of the voice frequency channel

Although the human voice has a much wider spectrum - from about 100 Hz to 10 kHz - for acceptable speech quality, a range of 3100 Hz is a good solution. The strict bandwidth limitation of the tone channel is associated with the use of multiplexing and circuit switching equipment in telephone networks.

Thus, for a voice frequency channel, amplitude modulation provides a data transfer rate of no more than 3100/2=1550 bit/s. If you use several levels of the information parameter (4 levels of amplitude), then the throughput of the voice frequency channel is doubled.

Most often, analog coding is used when transmitting information over a channel with a narrow bandwidth, for example, telephone lines in global networks. In local networks, it is rarely used due to the high complexity and cost of both encoding and decoding equipment.

Currently, almost all equipment that works with analog signals is being developed on the basis of expensive microcircuits. DSP (Digital Signal Processor). In this case, after modulation and signal transmission, it is necessary to carry out demodulation upon reception, and this is again expensive equipment. To perform the function of modulating the carrier sinusoid on the transmitting side and demodulating on the receiving side, a special device is used, which is called modem (modulator-demodulator). A 56,000 bps modem costs $100, and a 100 Mbps network card costs $10.

In conclusion, we present the advantages and disadvantages of analog modulation.

Analog modulation has many different information parameters: amplitude, phase, frequency. Each of these parameters can take on multiple states per carrier change. And, therefore, the resulting signal can transmit a large number of bits per second.

Analog modulation provides the resulting signal with a narrow spectrum, and therefore it is good where you need to work on poor lines (with a narrow bandwidth), it is able to provide high transmission speed there. Analog modulation can also work on good lines, here one more advantage of analog modulation is especially important - the ability to shift the spectrum in desired area, depending on the bandwidth of the line being used.

Analog modulation is difficult to implement and the equipment that does it is very expensive.

Analog modulation is used where it cannot be dispensed with, but in local networks other coding methods are used, for the implementation of which simple and cheap equipment is needed. Therefore, most often in local networks, when transmitting data in communication lines, the second method of physical coding is used - digital coding

5. 2.Digital coding

Digital coding- representation of information by rectangular pulses. For digital coding use potential and impulse codes.

Potential codes. In potential codes, only the value of the signal potential during the cycle period is used to represent logical ones and zeros, and its drops, which form complete pulses, are not taken into account. It is only important what value the resulting signal has during the cycle period.

impulse codes. Pulse codes represent a logical zero and a logical unit either by pulses of a certain polarity, or by part of the pulse - by a potential drop of a certain direction. The value of the pulse code includes the entire pulse along with its transitions.

Let's define the requirements for digital coding. For example, we need to transfer discrete data (a sequence of logical zeros and ones) from the output of one computer - the source - to the input of another computer - the receiver over the communication line.

1. For data transmission, we have communication lines that do not pass all frequencies, they have certain bandwidths depending on their type. Therefore, when encoding data, it must be taken into account that the encoded data is “passed through” by the communication line.

2. Sequences of discrete data must be encoded as digital pulses of a certain frequency. In this case, of course, it is best to achieve:

a) that the frequencies of the encoded signals be low to generally match the bandwidths of the communication links.

b) that the encoded signals provide a high transmission rate.

So good code should have less Hertz and more bits per second.

3. The data to be transmitted is an unpredictably changing sequence of logical zeros and ones.

Let's encode this data in a certain way with digital pulses, then how can we determine what frequency the resulting signal has? In order to determine the maximum frequency for us digital code it is enough to consider the resulting signal when encoding private sequences such as:

sequence of logical zeros

sequence of logical ones

alternating sequence of logical zeros and ones

Next, it is necessary to decompose the signal using the Fourier method, find the spectrum, determine the frequencies of each harmonic and find the total frequency of the signal, while it is important that the main spectrum of the signal falls within the bandwidth of the communication line. In order not to do all these calculations, it is enough to try to determine the fundamental harmonic of the signal spectrum, for this it is necessary to guess the first sinusoid from the signal shape, which repeats its contour of its shape, then find the period of this sinusoid. The period is the distance between two signal changes. Then you can also determine the frequency of the fundamental harmonic of the signal spectrum as F = 1/T, where F- frequency, T- signal period. For the convenience of further calculations, we assume that the bit rate of signal change is equal to N.

Such calculations can be made for each digital encoding method to determine the frequency of the resulting signal. The resulting signal in digital coding is a specific sequence of rectangular pulses. To represent a sequence of rectangular pulses as a sum of sinusoids to find the spectrum, a large number of such sinusoids is needed. The spectrum of a square wave sequence will generally be much wider than that of modulated signals.

If a digital code is used to transmit data on a voice frequency channel, then the upper limit for potential coding is achieved for a data rate of 971 bps, and the lower limit is unacceptable for any rates, since the channel bandwidth starts at 300 Hz.

So digital codes on voice-frequency channels are simply never used. But on the other hand, they work very well in local networks that do not use telephone lines for data transmission.

In this way, digital coding requires a wide bandwidth for high-quality transmission.

4. When transmitting information over communication lines from a source node to a receiver node, it is necessary to provide such a transmission mode in which the receiver will always know exactly at what point in time it receives data from the source, i.e. it is necessary to provide synchronization source and receiver. In networks, the synchronization problem is more difficult to solve than when exchanging data between blocks within a computer or between a computer and a printer. At short distances, a scheme based on a separate clocking communication line works well. In such a scheme, information is removed from the data line only at the moment the clock pulse arrives (see Fig. 5.4).

Rice. 5.4 Synchronization of receiver and transmitter over short distances

This synchronization option is absolutely not suitable for any network due to the heterogeneity of the characteristics of the conductors in the cables. Over long distances, signal velocity ripples can cause the clock to arrive so late or too early for the corresponding data signal that a data bit is skipped or reread. Another reason why networks refuse to use clock pulses is to save conductors in expensive cables. Therefore, networks use the so-called self-synchronizing codes.

Self-synchronizing codes- signals that indicate to the receiver at what point in time it is necessary to recognize the next bit (or several bits, if the code is oriented to more than two signal states). Any sharp drop in signal - the so-called front- can serve as a good indication for synchronization of the receiver with the transmitter. An example of a self-synchronizing code would be a sine wave. Since the change in the amplitude of the carrier frequency allows the receiver to determine the moment the input code appears. But this applies to analog modulation. In digital coding, there are also methods that create self-synchronizing codes, but more on that later.

In this way, a good digital code should provide synchronization

Having considered the requirements for a good digital code, let's move on to the consideration of the digital coding methods themselves.

5. 2.1Potential code without return to zero NRZ

This code got its name because when a sequence of 1s is transmitted, the signal does not return to zero during the cycle (as we will see below, in other coding methods, a return to zero occurs in this case).

Code NRZ (Non Return to Zero)- without returning to zero - this is the simplest two-level code. The resulting signal has two potential levels:

Zero corresponds to the lower level, unit - the upper. Information transitions occur at a bit boundary.

Let us consider three special cases of data transmission by the code NRZ: an alternating sequence of zeros and ones, a sequence of zeros and a sequence of ones (see Fig. 5.5, a).

Rice. 5.5 NRZ code

Let's try to determine whether this code satisfies the listed requirements. To do this, it is necessary to determine the fundamental harmonic of the spectrum with potential coding in each of the presented cases in order to more accurately determine which NRZ code has requirements for the communication line used.

The first case - information is transmitted, consisting of an infinite sequence of alternating ones and zeros (see Fig. 5.5, b).

This figure shows that when alternating ones and zeros, two bits 0 and 1 will be transmitted in one cycle. With the shape of the sinusoid shown in fig. 4.22b N- bit rate, the period of this sinusoid is equal to T=2N. The frequency of the fundamental harmonic in this case is equal to f 0 = N/2.

As you can see, with such a sequence of this code, the data transfer rate is twice the signal frequency.

When transmitting sequences of zeros and ones, the resulting signal is direct current, the frequency of the signal change is zero f 0 = 0 .

The spectrum of a real signal is constantly changing depending on what data is transmitted over the communication line and one should be wary of transmissions of long sequences of zeros or ones that shift the signal spectrum to the side low frequencies. Because NRZ code when transmitting long sequences of zeros or ones has a constant component.

It is known from signal theory that, in addition to the requirements for width, another very important requirement is put forward for the spectrum of the transmitted signal - no constant component(the presence of direct current between the receiver and transmitter), because the use of various transformer interchanges in the communication line does not pass direct current.

Therefore, some of the information will simply be ignored by this link. Therefore, in practice, they always try to get rid of the presence of a constant component in the spectrum of the carrier signal already at the coding stage.

Thus, we have identified one more requirement for a good digital code the digital code should not have a constant component.

Another disadvantage of NRZ is - lack of synchronization. In this case, only additional methods of synchronization will help, which we will talk about later.

One of the main advantages of the NRZ code is simplicity. In order to generate rectangular pulses, two transistors are needed, and complex microcircuits are needed to implement analog modulation. The potential signal does not need to be encoded and decoded, since the same method is used for data transmission inside the computer.

As a result of everything shown above, we will draw several conclusions that will help us when considering other digital coding methods:

NRZ is very easy to implement, has good error detection (due to two sharply different potentials).

NRZ has a DC component when transmitting zeros and ones, which makes it impossible to transmit on transformer isolated lines.

NRZ is not a self-synchronizing code and this complicates its transmission on any line.

The attractiveness of the NRZ code, because of which it makes sense to improve it, lies in the rather low frequency of the fundamental harmonic fo, which is equal to N/2 Hz, as shown above. Thus the code NRZ operates at low frequencies from 0 to N/2 Hz.

As a result, in its pure form, the NRZ code is not used in networks. Nevertheless, its various modifications are used, in which both the poor self-synchronization of the NRZ code and the presence of a constant component are successfully eliminated.

The following digital coding methods have been developed with the goal of somehow improving the capability of the NRZ code

5. 2.2. AMI Alternate Inversion Bipolar Coding Method

Method of bipolar coding with alternative inversion (Bipolar Alternate Mark Inversion, AMI) is a modification of the NRZ method.

This method uses three levels of potential - negative, zero and positive. Three signal levels is a disadvantage of the code, because in order to distinguish between three levels, a better signal-to-noise ratio is needed at the input to the receiver. The additional layer requires an increase in transmitter power of about 3 dB to provide the same bit fidelity on the line, which is a general disadvantage of multi-state codes compared to bilevel codes. In the AMI code, a zero potential is used to encode a logical zero, a logical one is encoded either by a positive potential or a negative one, while the potential of each new unit opposite of the previous one.

Rice. 5.6 AMI code

This coding technique partially eliminates the problems of the DC component and the lack of self-synchronization inherent in the NRZ code when transmitting long sequences of ones. But the problem of the constant component remains for him when transmitting sequences of zeros (see Fig. 5.6).

Let's consider particular cases of the code operation and determine the fundamental harmonic of the resulting signal spectrum for each of them. With a sequence of zeros - signal - direct current - fo \u003d 0 (Fig. 5.7, a)

Rice. 5.7 Determining the fundamental frequencies of the AMI spectrum

For this reason, the AMI code also needs further improvement. When transmitting a sequence of ones, the signal on the line is a sequence of bipolar pulses with the same spectrum as the NRZ code transmitting alternating zeros and ones, that is, without a constant component and with a fundamental harmonic fo = N/2 Hz.

When transmitting alternating ones and zeros, the fundamental harmonic fo = N/4 Hz, which is two times less than that of the NRZ code.

In general, for various combinations of bits on the line, the use of the AMI code leads to a narrower signal spectrum than for the NRZ code, and hence to a higher line throughput. The AMI code also provides some features for recognizing erroneous signals. Thus, a violation of the strict alternation of the polarity of the signals indicates a false impulse or the disappearance of a correct impulse from the line. A signal with incorrect polarity is called a forbidden signal. (signal violation).

The following conclusions can be drawn:

AMI cancels the DC component when transmitting a sequence of ones;

AMI has a narrow spectrum - from N/4 - N/2;

AMI partially eliminates synchronization problems

AMI uses not two, but three signal levels on the line and this is its drawback, but the following method managed to eliminate it.

5. 2.3 Potential code with inversion at unity NRZI

This code is completely similar to the AMI code, but only uses two signal levels. When zero is transmitted, it transmits the potential that was set in the previous cycle (that is, it does not change it), and when one is transmitted, the potential is inverted to the opposite.

This code is called potential code with inversion at one (Non Return to Zero with ones Inverted, NRZI).

It is convenient in cases where the use of the third signal level is highly undesirable, for example, in optical cables, where two signal states are reliably recognized - light and dark.

Rice. 5.8 NRZI code

The NRZI code differs in the shape of the resulting signal from the AMI code, but if you calculate the fundamental harmonics, for each case, it turns out that they are the same. For a sequence of alternating ones and zeros, the fundamental frequency of the signal is fo=N/4.(see Fig. 5.9, a). For with a sequence of units - fo=N/2. With a sequence of zeros, the same drawback remains fo=0- direct current in the line.

Rice. 5.9 Determining the fundamental frequencies of the spectrum for NRZI

The conclusions are as follows:

NRZI - provides the same capabilities as the AMI code, but uses only two signal levels for this and is therefore more suitable for further improvement. Disadvantages of NRZI are a DC component with a sequence of zeros, and lack of synchronization during transmission. The NRZI code became the basis for the development of more advanced coding methods at higher levels.

5. 2.4 Code MLT3

Code of three-level transmission MLT-3 (Multi Level Transmission - 3) has much in common with the NRZI code. Its most important difference is three signal levels.

One corresponds to the transition from one signal level to another. A change in the level of the linear signal occurs only if a unit is input, however, unlike the NRZI code, the generation algorithm is chosen in such a way that two adjacent changes always have opposite directions.

Rice. 5.10 Potential MLT-3 code

Consider special cases, as in all previous examples.

When transmitting zeros, the signal also has a constant component, the signal does not change - fo = 0 Hz. (See Figure 5.10). When all ones are transmitted, information transitions are fixed at the bit boundary, and one signal cycle can accommodate four bits. In this case fo=N/4 Hz - maximum code frequency MLT-3 when transferring all units (Fig. 5.11, a).

Rice. 5.11 Determining the fundamental frequencies of the spectrum for MLT-3

In the case of an alternating sequence, the code MLT-3 has a maximum frequency equal to fo=N/8, which is two times less than the NRZI code, therefore, this code has a narrower bandwidth.

As you noticed, the disadvantage of the MLT-3 code, like the NRZI code, is the lack of synchronization. This problem is solved with an additional data transformation that eliminates long sequences of zeros and the possibility of desynchronization. The general conclusion can be drawn as follows - the use of three-level coding MLT-3 allows you to reduce the clock frequency of the line signal and thereby increase the transmission rate.

5. 2.5 Bipolar pulse code

In addition to potential codes, impulse codes are also used when the data is represented by a full impulse or its part - a front.

The simplest case of this approach is bipolar pulse code, in which the unit is represented by a pulse of one polarity, and zero is the other. Each pulse lasts half a cycle (Fig. 5.12). Bipolar pulse code - three-level code. Let us consider the resulting signals during data transmission by bipolar coding in the same particular cases.

Rice. 5.12 Bipolar pulse code

A feature of the code is that there is always a transition (positive or negative) in the center of the bit. Therefore, each bit is labeled. The receiver can extract a sync pulse (strobe) having a pulse repetition rate from the signal itself. Binding is made to each bit, which ensures synchronization of the receiver with the transmitter. Such codes, which carry a strobe, are called self-synchronizing. Consider the spectrum of signals for each case (Fig. 5.13). When transmitting all zeros or ones, the frequency of the fundamental harmonic of the code fo=N Hz, which is twice the fundamental of the NRZ code and four times the fundamental of the AMI code. When transmitting alternating ones and zeros - fo=N/2

Rice. 5.13 Determination of the main frequencies of the spectrum for a bipolar pulse code.

This shortcoming of the code does not give a gain in data transfer rate and clearly indicates that the impulse codes are slower than potential ones.

For example, a 10 Mbps link requires a carrier frequency of 10 MHz. When transmitting a sequence of alternating zeros and ones, the speed increases, but not much, because the frequency of the fundamental harmonic of the code fо=N/2 Hz.

The bipolar pulse code has a great advantage over previous codes - it is self-synchronizing.

The bipolar pulse code has a wide signal spectrum and is therefore slower.

The bipolar pulse code uses three levels.

5. 2.6 Manchester code

Manchester code was developed as an improved bipolar pulse code. The Manchester code also refers to self-synchronizing codes, but unlike the bipolar code, it has not three, but only two levels, which provides better noise immunity.

In the Manchester code, a potential drop, that is, the front of the pulse, is used to encode ones and zeros. In Manchester encoding, each clock is divided into two parts. Information is encoded by potential drops that occur in the middle of each cycle. It happens like this:

A unit is encoded by a low-to-high transition, and a zero is encoded by a reverse transition. At the beginning of each cycle, a service signal edge can occur if you need to represent several ones or zeros in a row.

Consider special cases of encoding (sequences of alternating zeros and ones, some zeros, some ones), and then we will determine the main harmonics for each of the sequences (see Fig. 5.14). In all cases, it can be seen that with Manchester coding, the signal change at the center of each bit makes it easy to isolate the clock signal. Therefore, the Manchester code has good self-synchronizing properties.

Rice. 5.14 Manchester code

Self-synchronization always makes it possible to transmit large packets of information without loss due to differences in the clock frequency of the transmitter and receiver.

So, let's determine the fundamental frequency when transmitting only ones or only zeros.

Rice. 5.15 Determination of the main frequencies of the spectrum for the Manchester code.

As can be seen when transmitting both zeros and ones, there is no constant component. Fundamental frequency fo=NHz, as in bipolar coding. Due to this, the galvanic isolation of signals in communication lines can be performed in the simplest ways, for example, using pulse transformers. When transmitting alternating ones and zeros, the frequency of the fundamental harmonic is equal to fo=N/2Hz.

Thus, the Manchester code is an improved bipolar code, improved by using only two signal levels for data transmission, and not three, as in bipolar. But this code is still slow compared to NRZI which is twice as fast.

Consider an example. Take for data transmission a communication line with a bandwidth 100 MHz and speed 100 Mbps. If earlier we determined the data rate at a given frequency, now we need to determine the frequency of the signal at a given line speed. Based on this, we determine that for data transmission by the NRZI code, the frequency range from N / 4-N / 2 is enough for us - these are frequencies from 25 -50 MHz, these frequencies are included in the bandwidth of our line - 100 MHz. For the Manchester code, we need a frequency range from N / 2 to N - these are frequencies from 50 to 100 MHz, in this range the main harmonics of the signal spectrum are located. For the Manchester code, it does not satisfy the bandwidth of our line, and, therefore, the line will transmit such a signal with large distortions (such a code cannot be used on this line).

5.2.7Differential Manchester code.

Differential Manchester code is a type of Manchester coding. It uses the middle of the clock interval of the line signal only for synchronization, and there is always a change in the signal level at it. Logic 0 and 1 are transmitted by the presence or absence of a signal level change at the beginning of the clock interval, respectively (Fig. 5.16)

Rice. 5.16 Differential Manchester code

This code has the same advantages and disadvantages as the Manchester one. But, in practice, it is the differential Manchester code that is used.

Thus, Manchester code used to be (when high-speed lines were a great luxury for a local area network) very actively used in local networks, due to its self-synchronization and lack of a constant component. It is still widely used in fiber optic and electrical networks. Recently, however, developers have come to the conclusion that it is still better to use potential coding, eliminating its shortcomings using the so-called logical coding.

5.2.8Potential code 2B1Q

Code 2B1Q- potential code with four signal levels for data encoding. Its name reflects its essence - every two bits (2B) are transmitted in one cycle by a signal that has four states (1Q).

Pare bit 00 corresponds potential (-2.5V), a couple of bits 01 corresponds potential (-0.833 V), couple 11 - potential (+0.833 V), and a couple 10 - potential ( +2.5 V).

Rice. 5.17 Potential code 2B1Q

As can be seen in Figure 5.17, this encoding method requires additional measures to deal with long sequences of identical bit pairs, since the signal is then converted to a DC component. Therefore, when transmitting both zeros and ones fo=0Hz. When alternating ones and zeros, the signal spectrum is twice as narrow as that of the code NRZ, since at the same bit rate the duration of the cycle is doubled - fo=N/4Hz.

Thus, using the 2B1Q code, you can transfer data over the same line twice as fast as using the AMI or NRZI code. However, for its implementation, the transmitter power must be higher so that the four potential levels (-2.5V, -0.833 V, +0.833 V, +2.5 V) are clearly distinguished by the receiver against the background of interference.

5. 2.9 Code PAM5

All the signal coding schemes we have considered above were bit-based. With bit coding, each bit corresponds to a signal value determined by the protocol logic.

With byte encoding, the signal level is set by two or more bits. In a five level code PAM5 5 voltage levels (amplitudes) and two-bit coding are used. Each combination has its own voltage level. With two-bit coding, four levels are required to transmit information (two to the second power - 00, 01, 10, 11 ). Transmitting two bits at the same time provides a halving of the signal change rate. The fifth level is added to create redundancy in the code used for error correction. This gives an additional margin of signal-to-noise ratio.

Rice. 5.18 Code PAM 5

5. 3. Logic coding

Logic coding runs until physical encoding.

At the stage of logical coding, the waveform is no longer formed, but the shortcomings of physical digital coding methods, such as the lack of synchronization, the presence of a constant component, are eliminated. Thus, first, corrected sequences of binary data are formed using logical coding tools, which are then transmitted over communication lines using physical coding methods.

Logical coding implies the replacement of bits of the original information with a new sequence of bits that carries the same information, but has, in addition, additional properties, such as the ability for the receiving side to detect errors in the received data. Accompanying each byte of the original information with one parity bit is an example of a very commonly used method of logical coding when transmitting data using modems.

Separate two methods of logical coding:

Redundant codes

Scrambling.

5. 3.1 Redundant codes

Redundant codes are based on splitting the original sequence of bits into portions, which are often called characters. Then each original character is replaced with a new one, which has large quantity bit than the original. A clear example of redundant code is the 4V/5V logic code.

Logic code 4V/5V replaces the original 4-bit characters with 5-bit characters. Since the resulting symbols contain redundant bits, the total number of bit combinations in them is greater than in the original ones. Thus, the five-bit scheme gives 32 (2 5) two-digit alphanumeric characters with a value in decimal code from 00 to 31. While the original data can contain only four bits or 16 (2 4) characters.

Therefore, in the resulting code, you can pick up 16 such combinations that do not contain a large number zeros and count the rest prohibited codes (code violation). In this case, the long strings of zeros are broken and the code becomes self-synchronizing for any transmitted data. The constant component also disappears, which means that the signal spectrum narrows even more. But this method reduces the useful capacity of the line, since the excess units user information do not carry, and only "occupy air time". The redundant codes allow the receiver to recognize corrupted bits. If the receiver receives a forbidden code, it means that the signal has been distorted on the line.

So let's look at work. logic code 4V/5V. The converted signal has 16 values ​​for information transfer and 16 redundant values. In the receiver decoder, five bits are decoded as information and service signals.

Nine symbols are allocated for service signals, seven symbols are excluded.

Combinations with more than three zeros are excluded (01 - 00001, 02 - 00010, 03 - 00011, 08 - 01000, 16 - 10000 ) . Such signals are interpreted by the symbol V and the receiver team VIOLATION- failure. The command means there is an error due to high level interference or transmitter failure. The only combination of five zeros (00 - 00000 ) refers to service signals, means the symbol Q and has the status QUIET- no signal on the line.

Such data encoding solves two problems - synchronization and noise immunity improvement. Synchronization occurs due to the elimination of a sequence of more than three zeros, and high noise immunity is achieved by the data receiver in a five-bit interval.

The price for these advantages with this method of data encoding is a decrease in the transmission rate. useful information. For example, As a result of adding one redundant bit to four information bits, the bandwidth efficiency in protocols with code MLT-3 and data encoding 4B/5B decreases respectively by 25%.

Encoding scheme 4V/5V presented in the table.

So, according to this table, the code is formed 4V/5V, then transmitted over the line using physical coding using one of the potential coding methods that is sensitive only to long sequences of zeros - for example, using the NRZI digital code.

The 4V/5V code symbols, 5 bits long, guarantee that no more than three zeros in a row can occur on the line for any combination of them.

Letter V in the code name means that the elementary signal has 2 states - from English binary- binary. There are also codes with three signal states, for example, in the code 8V/6T to encode 8 bits of the original information, a code of 6 signals is used, each of which has three states. Code Redundancy 8V/6T higher than code 4V/5V, since there are 3 6 = 729 resulting symbols for 256 source codes.

As we said, logical encoding occurs before physical, therefore, it is carried out by the network link-level equipment: network adapters and interface blocks of switches and routers. Since, as you yourself have seen, the use of a lookup table is a very simple operation, so the method of logical coding with redundant codes does not complicate the functional requirements for this equipment.

The only requirement is that the transmitter using the redundant code must operate at a higher clock rate to provide a given line capacity. Yes, to send codes 4V/5V with speed 100 Mb/s the transmitter must operate at a clock frequency 125 MHz. In this case, the spectrum of the signal on the line is expanded in comparison with the case when a pure, non-redundant code is transmitted over the line. However, the spectrum of the redundant potential code turns out to be narrower than the spectrum of the Manchester code, which justifies the additional stage of logical coding, as well as the operation of the receiver and transmitter at an increased clock frequency.

Thus, the following conclusion can be drawn:

Basically, for local networks it is easier, more reliable, better, faster - to use logical data encoding using redundant codes, which will eliminate long sequences of zeros and ensure signal synchronization, then use a fast digital code for transmission at the physical level NRZI, rather than using a slow but self-synchronizing Manchester code.

For example, to transmit data over a line with a bandwidth of 100M bit / s and a bandwidth of 100 MHz, the NRZI code requires frequencies of 25 - 50 MHz, this is without coding 4V / 5V. And if applied to NRZI also 4V / 5V encoding, now the frequency band will expand from 31.25 to 62.5 MHz. But nevertheless, this range still "fits" into the line bandwidth. And for the Manchester code, without the use of any additional coding, frequencies from 50 to 100 MHz are needed, and these are the frequencies of the main signal, but they will no longer be passed by the 100 MHz line.

5. 3.2 Scrambling

Another method of logical coding is based on the preliminary "mixing" of the original information in such a way that the probability of occurrence of ones and zeros on the line becomes close.

Devices or blocks that perform this operation are called scramblers (scramble - dump, random assembly).

At scrambling the data is mixed according to a certain algorithm and the receiver, having received binary data, transmits it to descrambler, which restores the original bit sequence.

Excess bits are not transmitted over the line.

The essence of scrambling is simply a bit-by-bit change in the data stream passing through the system. Almost the only operation used in scramblers is XOR - "bitwise XOR", or else they say - addition by module 2. When two units are added by exclusive OR, the highest unit is discarded and the result is written - 0.

The scrambling method is very simple. First come up with a scrambler. In other words, they come up with what ratio to mix bits in the original sequence using "exclusive OR". Then, according to this ratio, the values ​​of certain bits are selected from the current sequence of bits and added up according to XOR between themselves. In this case, all bits are shifted by 1 bit, and the value just received ("0" or "1") is placed in the freed least significant bit. The value that was in the most significant bit before the shift is added to the coding sequence, becoming its next bit. Then this sequence is issued to the line, where, using physical encoding methods, it is transmitted to the recipient node, at the input of which this sequence is descrambled based on the inverse ratio.

For example, a scrambler might implement the following relationship:

where Bi- binary digit of the resulting code obtained on the i-th cycle of the scrambler, AI- binary digit of the source code, coming at the i-th cycle to the input of the scrambler, B i-3 and B i-5- binary digits of the resulting code obtained in the previous cycles of the scrambler, respectively, 3 and 5 cycles earlier than the current cycle,  - XOR operation (modulo 2 addition).

Now let's define the encoded sequence, for example, for such a source sequence 110110000001 .

The scrambler defined above will produce the following result code:

B 1 \u003d A 1 \u003d 1 (the first three digits of the resulting code will be the same as the original one, since there are no necessary previous digits yet)

Thus, the output of the scrambler will be the sequence 110001101111 . In which there is no sequence of six zeros that was present in source code.

After receiving the resulting sequence, the receiver passes it to the descrambler, which reconstructs the original sequence based on the inverse relationship.

There are other different scrambling algorithms, they differ in the number of terms that give the digit of the resulting code, and the shift between the terms.

The main problem of coding based scramblers - synchronization of the transmitting (encoding) and receiving (decoding) devices. If at least one bit is omitted or erroneously inserted, all transmitted information is irreversibly lost. Therefore, in scrambler-based coding systems, much attention is paid to synchronization methods. .

In practice, a combination of two methods is usually used for these purposes:

a) adding synchronization bits to the information stream, which are known in advance to the receiving side, which allows it, if such a bit is not found, to actively start searching for synchronization with the sender,

b) the use of high-precision time pulse generators, which makes it possible to decode the received bits of information "from memory" without synchronization at times of loss of synchronization.

There are more simple methods fight against sequences of units, also referred to the class of scrambling.

To improve the code Bipolar AMI two methods are used, based on the artificial distortion of the sequence of zeros by forbidden symbols.

Rice. 5.19 Codes B8ZS and HDB3

This figure shows the use of the method B8ZS (Bipolar with 8-Zeros Substitution) and method HDB3 (High-Density Bipolar 3-Zeros) to correct the AMI code. The source code consists of two long sequences of zeros (8- in the first case and 5 in the second).

Code B8ZS corrects only sequences consisting of 8 zeros. To do this, after the first three zeros, instead of the remaining five zeros, he inserts five digits: V-1*-0-V-1*.V here denotes a signal of a unit prohibited for a given cycle of polarity, that is, a signal that does not change the polarity of the previous unit, 1 * - a signal of the unit of correct polarity, and the asterisk sign marks the fact that in the source code in this cycle there was not a unit, but a zero. As a result, the receiver sees 2 distortions in 8 clock cycles - it is very unlikely that this happened due to noise on the line or other transmission failures. Therefore, the receiver considers such violations as coding of 8 consecutive zeros and, upon reception, replaces them with the original 8 zeros.

The B8ZS code is constructed in such a way that its constant component is zero for any sequence of binary digits.

Code HDB3 corrects any 4 consecutive zeros in the original sequence. The rules for generating the HDB3 code are more complex than the B8ZS code. Every four zeros are replaced by four signals that have one V signal. To suppress the DC component, the polarity of the signal V alternates with successive replacements.

In addition, two patterns of four-cycle codes are used for replacement. If the source code contained an odd number of ones before the replacement, then the sequence is used 000V, and if the number of units was even - the sequence 1*00V.

Thus, the use of logical coding in conjunction with potential coding provides the following advantages:

Enhanced candidate codes have a fairly narrow bandwidth for any sequences of 1s and 0s that occur in the transmitted data. As a result, codes derived from the potential by logical coding have a narrower spectrum than Manchester, even at an increased clock frequency.

Page 27 from 27 Physical basis of data transmission(Communication lines,)

Physical basis of data transmission

Any network technology must provide reliable and fast transmission of discrete data over communication lines. And although there are big differences between technologies, they are based on the general principles of discrete data transmission. These principles are embodied in methods for representing binary ones and zeros using pulsed or sinusoidal signals in communication lines of various physical nature, error detection and correction methods, compression methods, and switching methods.

linesconnections

Primary networks, lines and communication channels

When describing technical system, which transmits information between network nodes, several names can be found in the literature: communication line, composite channel, channel, link. Often these terms are used interchangeably and in many cases this does not cause problems. At the same time, there are specifics in their use.

Link(link) is a segment that provides data transfer between two neighboring network nodes. That is, the link does not contain intermediate switching and multiplexing devices.

channel(channel) most often denote the portion of the link bandwidth used independently in switching. For example, a primary network link may consist of 30 channels, each of which has a bandwidth of 64 Kbps.

Composite channel(circuit) is a path between two end nodes of a network. A composite link is formed by individual intermediate link links and internal connections in the switches. Often the epithet "composite" is omitted and the term "channel" is used to mean both a composite channel and a channel between adjacent nodes, that is, within a link.

Communication line can be used as a synonym for any of the other three terms.

On fig. two variants of the communication line are shown. In the first case ( a) the line consists of a cable segment with a length of several tens of meters and is a link. In the second case (b), the link is a composite link deployed in a circuit-switched network. Such a network could be primary network or telephone network.

However, for a computer network, this line is a link, since it connects two neighboring nodes, and all switching intermediate equipment is transparent to these nodes. The reason for mutual misunderstanding at the level of terms of computer specialists and specialists of primary networks is obvious here.

Primary networks are specially created in order to provide data transmission services for computer and telephone networks, which in such cases are said to work "on top" of primary networks and are overlay networks.

Classification of communication lines

Communication line generally consists of a physical medium through which electrical information signals are transmitted, data transmission equipment and intermediate equipment. The physical medium for data transmission (physical media) can be a cable, that is, a set of wires, insulating and protective sheaths and connectors, as well as the earth's atmosphere or outer space through which electromagnetic waves propagate.

In the first case, one speaks of wired environment, and in the second - wireless.

In modern telecommunications systems, information is transmitted using electric current or voltage, radio signals or light signals- all these physical processes are oscillations of the electromagnetic field of different frequencies.

Wired (overhead) lines ties are wires without any insulating or shielding braids, laid between poles and hanging in the air. Even in the recent past, such communication lines were the main ones for transmitting telephone or telegraph signals. Today, wired communication lines are rapidly being replaced by cable ones. But in some places they are still preserved and, in the absence of other possibilities, they continue to be used for the transmission of computer data. The speed qualities and noise immunity of these lines leave much to be desired.

cable lines have a rather complex structure. The cable consists of conductors enclosed in several layers of insulation: electrical, electromagnetic, mechanical, and possibly climatic. In addition, the cable can be equipped with connectors that allow you to quickly connect various equipment to it. Three main types of cable are used in computer (and telecommunications) networks: cables based on twisted pairs of copper wires - unshielded twisted pair(Unshielded Twisted Pair, UTP) and shielded twisted pair(Shielded Twisted Pair, STP), coaxial cables with a copper core, fiber optic cables. The first two types of cables are also called copper cables.

radio channels ground and satellite communications generated by a transmitter and receiver of radio waves. There is a wide variety of types of radio channels, differing both in the frequency range used and in the channel range. Broadcast radio bands(long, medium and short waves), also called AM bands, or ranges of amplitude modulation (Amplitude Modulation, AM), provide long-distance communication, but at a low data rate. Faster channels are those that use ranges are very high frequencies (Very High Frequency, VHF), which uses frequency modulation (Frequency Modulation, FM). Also used for data transfer. ultra-high frequency bands(Ultra High Frequency, UHF), also called microwave ranges(over 300 MHz). At frequencies above 30 MHz, the signals are no longer reflected by the Earth's ionosphere, and stable communication requires line-of-sight between transmitter and receiver. Therefore, such frequencies use either satellite channels, or microwave channels, or local or mobile networks where this condition is met.

Crosstalk at the near end of the line - determines the noise immunity of the cable to internal sources of interference. Usually they are evaluated in relation to a cable consisting of several twisted pair, when the mutual pickups of one pair on another can reach significant values ​​and create internal interference commensurate with the useful signal.

Reliability of data transmission(or bit error rate) characterizes the probability of distortion for each transmitted data bit. The reasons for the distortion of information signals are interference on the line, as well as the limited bandwidth of its pass. Therefore, an increase in the reliability of data transmission is achieved by increasing the degree of noise immunity of the line, reducing the level of crosstalk in the cable, and using more broadband communication lines.

For conventional cable communication lines without additional error protection, the reliability of data transmission is, as a rule, 10 -4 -10 -6 . This means that, on average, out of 10 4 or 10 6 transmitted bits, the value of one bit will be corrupted.

Communication line equipment(data transmission equipment - ATD) is the edge equipment that directly connects computers to the communication line. It is part of the communication line and usually operates at the physical level, providing the transmission and reception of a signal of the desired shape and power. Examples of ADFs are modems, adapters, analog-to-digital and digital-to-analog converters.

The DTE does not include the user's data terminal equipment (DTE), which generates data for transmission over the communication line and is connected directly to the DTE. A DTE includes, for example, a LAN router. Note that the division of equipment into APD and OOD classes is rather conditional.

On long communication lines, intermediate equipment is used, which solves two main tasks: improving the quality of information signals (their shape, power, duration) and creating a permanent composite channel (end-to-end channel) of communication between two network subscribers. In the LCN, intermediate equipment is not used if the length of the physical medium (cables, radio air) is not high, so that signals from one network adapter to another can be transmitted without intermediate restoration of their parameters.

V global networks high-quality signal transmission over hundreds and thousands of kilometers is ensured. Therefore, amplifiers are installed at certain distances. To create a through line between two subscribers, multiplexers, demultiplexers and switches are used.

The intermediate equipment of the communication channel is transparent to the user (he does not notice it), although in reality it forms a complex network called primary network and serving as the basis for building computer, telephone and other networks.

Distinguish analog and digital communication lines, which use different types intermediate equipment. In analog lines, intermediate equipment is designed to amplify analog signals that have a continuous range of values. In high-speed analog channels, a frequency multiplexing technique is implemented, when several low-speed analog subscriber channels are multiplexed into one high-speed channel. In digital communication channels, where rectangular information signals have a finite number of states, intermediate equipment improves the shape of the signals and restores their repetition period. It provides the formation of high-speed digital channels, working on the principle of time multiplexing of channels, when each low-speed channel is allocated a certain fraction of the time of the high-speed channel.

When transmitting discrete computer data over digital lines communication, the physical layer protocol is defined, since the parameters of the information signals transmitted by the line are standardized, and when transmitted over analog lines, it is not defined, since the information signals have an arbitrary shape and there are no requirements for the way the data transmission equipment represents ones and zeros.

The following are used in communication networks information transfer modes:

simplex, when the transmitter and receiver are connected by one communication channel, through which information is transmitted only in one direction (this is typical for television communication networks);

Half-duplex, when two communication nodes are also connected by one channel, through which information is transmitted alternately in one direction, then in the opposite direction (this is typical for information-reference, request-response systems);

duplex, when two communication nodes are connected by two channels (forward communication channel and reverse), through which information is simultaneously transmitted in opposite directions. Duplex channels are used in systems with decision and information feedback.

Switched and dedicated communication channels. In the TSS, there are dedicated (non-switched) communication channels and those with switching for the duration of information transmission over these channels.

When using dedicated communication channels, the transceiver equipment of communication nodes is constantly connected to each other. This ensures a high degree of readiness of the system for information transfer, higher quality of communication, and support for a large amount of traffic. Due to the relatively high costs of operating networks with dedicated communication channels, their profitability is achieved only if the channels are fully loaded.

Switched communication channels created only for the period of transmission of a fixed amount of information are characterized by high flexibility and relatively low cost (with a small amount of traffic). The disadvantages of such channels are: loss of time for switching (to establish communication between subscribers), the possibility of blocking due to the busyness of individual sections of the communication line, lower communication quality, high cost with a significant amount of traffic.

The initial information that needs to be transmitted over a communication line can be either discrete (computer output data) or analog (speech, television image).

Discrete data transmission is based on the use of two types of physical encoding:

a) analog modulation when encoding is performed by changing the parameters of the sinusoidal carrier signal;

b) digital coding by changing the levels of the sequence of rectangular information pulses.

Analog modulation leads to a much smaller spectrum of the resulting signal than with digital coding, at the same information transfer rate, but its implementation requires more complex and expensive equipment.

At present, the initial data, which has an analog form, is increasingly transmitted over communication channels in a discrete form (in the form of a sequence of ones and zeros), i.e. discrete modulation analog signals.

Analog modulation. It is used to transmit discrete data over channels with a narrow bandwidth, a typical representative of which is the voice frequency channel provided to users of telephone networks. Signals with a frequency of 300 to 3400 Hz are transmitted over this channel, i.e., its bandwidth is 3100 Hz. Such a band is quite sufficient for speech transmission with acceptable quality. The bandwidth limitation of the tone channel is associated with the use of multiplexing and circuit switching equipment in telephone networks.

Before the transmission of discrete data on the transmitting side using a modulator-demodulator (modem) modulation of the carrier sinusoid of the original sequence of binary digits is carried out. Reverse transformation(demodulation) is performed by the receiving modem.

There are three ways to convert digital data to analog form, or three methods of analog modulation:

Amplitude modulation, when only the amplitude of the carrier of sinusoidal oscillations changes in accordance with the sequence of transmitted information bits: for example, when transmitting one, the oscillation amplitude is set large, and when transmitting zero, it is small, or there is no carrier signal at all;

frequency modulation, when under the influence of modulating signals (transmitted information bits) only the frequency of the carrier of sinusoidal oscillations changes: for example, when zero is transmitted, it is low, and when one is transmitted, it is high;

phase modulation, when, in accordance with the sequence of transmitted information bits, only the phase of the carrier of sinusoidal oscillations changes: when switching from signal 1 to signal 0 or vice versa, the phase changes by 180 °.

In its pure form, amplitude modulation is rarely used in practice due to low noise immunity. Frequency modulation does not require complex schemes in modems and is typically used in low speed modems operating at 300 or 1200 bps. Increasing the data rate is provided by the use of combined modulation methods, more often amplitude modulation in combination with phase.

The analog method of discrete data transmission provides wideband transmission by using signals of different carrier frequencies in one channel. This guarantees the interaction of a large number of subscribers (each pair of subscribers operates at its own frequency).

Digital coding. When digitally encoding discrete information, two types of codes are used:

a) potential codes, when only the value of the signal potential is used to represent information units and zeros, and its drops are not taken into account;

b) pulse codes, when binary data is represented either by pulses of a certain polarity, or by potential drops of a certain direction.

The following requirements are imposed on the methods of digital coding of discrete information when using rectangular pulses to represent binary signals:

ensuring synchronization between transmitter and receiver;

Ensuring the smallest spectrum width of the resulting signal at the same bit rate (since a narrower spectrum of signals allows a higher data rate to be achieved on a line with the same bandwidth);

the ability to recognize errors in transmitted data;

Relatively low cost of implementation.

By means of the physical layer, only the recognition of corrupted data (error detection) is carried out, which saves time, since the receiver, without waiting for the received frame to be completely placed in the buffer, immediately rejects it when it recognizes erroneous bits in the frame. A more complex operation - the correction of corrupted data - is performed by higher-level protocols: channel, network, transport or application.

Synchronization of the transmitter and receiver is necessary so that the receiver knows exactly when to read the incoming data. Clock signals tune the receiver to the transmitted message and keep the receiver synchronized with the incoming data bits. The problem of synchronization is easily solved when transmitting information over short distances (between blocks inside a computer, between a computer and a printer) by using a separate timing communication line: information is read only at the moment the next clock pulse arrives. In computer networks, the use of clock pulses is abandoned for two reasons: for the sake of saving conductors in expensive cables and due to the heterogeneity of the characteristics of conductors in cables (over long distances, uneven signal propagation speed can lead to desynchronization of clock pulses in the clock line and information pulses in the main line , as a result of which the data bit will either be skipped or reread).

Currently, synchronization of the transmitter and receiver in networks is achieved by using self-synchronizing codes(SK). The coding of the transmitted data using the SC is to ensure regular and frequent changes (transitions) of the levels of the information signal in the channel. Each signal level transition from high to low or vice versa is used to trim the receiver. The best are those SCs that provide a signal level transition at least once during the time interval required to receive one information bit. The more frequent the signal level transitions, the more reliable the synchronization of the receiver is and the more confident the identification of the received data bits is.

These requirements for the methods of digital encoding of discrete information are, to a certain extent, mutually contradictory, therefore, each of the encoding methods considered below has its advantages and disadvantages compared to others.

Self-synchronizing codes. The most common are the following SCs:

potential code without return to zero (NRZ - Non Return to Zero);

bipolar pulse code (RZ code);

The Manchester code

· bipolar code with alternating level inversion.

On fig. 32 shows the coding schemes for message 0101100 using these CKs.

Rice. 32. Message encoding schemes using self-synchronizing codes

When transmitting discrete data over communication channels, two main types of physical coding are used - based on a sinusoidal carrier signal and based on a sequence of rectangular pulses. The first method is often also called modulation or analog modulation, emphasizing the fact that coding is carried out by changing the parameters of the analog signal. The second way is usually called digital coding. These methods differ in the width of the spectrum of the resulting signal and the complexity of the equipment required for their implementation.

When using rectangular pulses, the spectrum of the resulting signal is very wide. This is not surprising if we remember that the spectrum of an ideal momentum has an infinite width. The use of a sinusoid results in a much smaller spectrum at the same information rate. However, the implementation of sinusoidal modulation requires more complex and expensive equipment than the implementation of rectangular pulses.

Currently, more and more often, data that initially has an analog form - speech, a television image - are transmitted over communication channels in a discrete form, that is, in the form of a sequence of ones and zeros. Submission process analog information in discrete form is called discrete modulation. The terms "modulation" and "coding" are often used interchangeably.

2.2.1. Analog modulation

Analog modulation is used to transmit discrete data over narrow bandwidth channels, typified by tone frequency channel, made available to users of public telephone networks. A typical frequency response of a voice frequency channel is shown in fig. 2.12. This channel transmits frequencies in the range from 300 to 3400 Hz, so its bandwidth is 3100 Hz. Although the human voice has a much wider spectrum, from about 100 Hz to 10 kHz, for acceptable speech quality, a range of 3100 Hz is a good solution. The strict bandwidth limitation of the tone channel is associated with the use of multiplexing and circuit switching equipment in telephone networks.

2.2. Discrete data transfer methods at the physical layer 133

A device that performs the functions of modulating a carrier sinusoid on the transmitting side and demodulating on the receiving side is called modem(modulator-demodulator).

Analog modulation methods

Analog modulation is a physical coding method in which information is encoded by changing the amplitude, frequency, or phase of a sinusoidal carrier signal. The main methods of analog modulation are shown in fig. 2.13. On the diagram (Fig. 2.13, a) a sequence of bits of the original information is shown, represented by high level potentials for a logical one and a zero level potential for a logical zero. This encoding method is called a potential code, which is often used when transferring data between computer blocks.

At amplitude modulation(Fig. 2.13, 6) for a logical one, one level of the amplitude of the carrier frequency sinusoid is selected, and for a logical zero, another. This method is rarely used in its pure form in practice due to low noise immunity, but is often used in combination with another type of modulation - phase modulation.

At frequency modulation(Fig. 2.13, c) the values ​​0 and 1 of the initial data are transmitted by sinusoids with different frequencies - fo and fi. This modulation method does not require complex circuitry in modems and is typically used in low speed modems operating at 300 or 1200 bps.

At phase modulation(Fig. 2.13, d) data values ​​0 and 1 correspond to signals of the same frequency, but with a different phase, for example, 0 and 180 degrees or 0.90,180 and 270 degrees.

In high-speed modems, combined modulation methods are often used, as a rule, amplitude in combination with phase.

Chapter 2. Fundamentals of Discrete Data Communication

Spectrum of the modulated signal

The spectrum of the resulting modulated signal depends on the type of modulation and the modulation rate, that is, the desired bit rate of the original information.

Let us first consider the spectrum of the signal with potential coding. Let a logical unit be encoded by a positive potential, and a logical zero by a negative potential of the same magnitude. To simplify the calculations, we assume that information is transmitted consisting of an infinite sequence of alternating ones and zeros, as shown in Fig. 2.13, a. Note that in this case, the baud and bits per second values ​​are the same.

For potential encoding, the spectrum is directly obtained from the Fourier formulas for the periodic function. If discrete data is transmitted at a bit rate N bit/s, then the spectrum consists of a constant component of zero frequency and an infinite series of harmonics with frequencies fo, 3fo, 5fo, 7fo,..., where fo = N/2. The amplitudes of these harmonics decrease rather slowly - with coefficients 1/3, 1/5,1/7,... of the harmonic amplitude fo (Fig. 2.14, a). As a result, the potential code spectrum requires a wide bandwidth for high-quality transmission. In addition, it must be taken into account that in reality the spectrum of the signal is constantly changing depending on what data is transmitted over the communication line. For example, the transmission of a long sequence of zeros or ones shifts the spectrum towards low frequencies, and in the extreme case, when the transmitted data consists of only ones (or only zeros), the spectrum consists of the zero frequency harmonic. When transmitting alternating ones and zeros, there is no DC component. Therefore, the spectrum of the resulting potential code signal during the transmission of arbitrary data occupies a band from some value close to 0 Hz to about 7fo (harmonics with frequencies above 7fo can be neglected due to their small contribution to the resulting signal). For a voice frequency channel, the upper limit for potential coding is achieved for a data rate of 971 bps, and the lower limit is unacceptable for any rates, since the channel bandwidth starts at 300 Hz. As a result, potential codes on voice frequency channels are never used.

2.2. Discrete data transfer methods at the physical layer 135

With amplitude modulation, the spectrum consists of a sinusoid of the carrier frequency fc and two side harmonics: (fc + fm) and (fc - fm), where fm is the frequency of the change in the information parameter of the sinusoid, which coincides with the data rate when using two amplitude levels (Fig. 2.14, 6). The frequency f m determines the line capacity for a given coding method. With a small modulation frequency, the signal spectrum width will also be small (equal to 2f m), so the signals will not be distorted by the line if its bandwidth is greater than or equal to 2f m . For a voice frequency channel, this modulation method is acceptable at a data rate of no more than 3100/2=1550 bps. If 4 amplitude levels are used to represent data, then the channel capacity increases to 3100 bps.

With phase and frequency modulation, the signal spectrum is more complex than with amplitude modulation, since more than two side harmonics are formed here, but they are also symmetrically located relative to the main carrier frequency, and their amplitudes decrease rapidly. Therefore, these modulations are also well suited for data transmission over a voice-frequency channel.

Combined modulation methods are used to increase the data rate. The most common methods are quadrature amplitude modulation (Quadrature Amplitude Modulation, QAM). These methods are based on a combination of phase modulation with 8 phase shift values ​​and amplitude modulation with 4 amplitude levels. However, not all of the possible 32 signal combinations are used. For example, in codes Trellis only 6, 7 or 8 combinations are allowed to represent the original data, and the remaining combinations are prohibited. Such coding redundancy is required for the modem to recognize erroneous signals resulting from distortion due to interference, which on telephone channels, especially switched ones, are very significant in amplitude and long in time.

2.2.2. Digital coding

When digitally encoding discrete information, potential and impulse codes are used.

In potential codes, only the value of the signal potential is used to represent logical ones and zeros, and its drops, which form complete pulses, are not taken into account. Pulse codes allow binary data to be represented either by pulses of a certain polarity, or by a part of the pulse - by a potential drop of a certain direction.

Requirements for digital coding methods

When using rectangular pulses to transmit discrete information, it is necessary to choose a coding method that would simultaneously achieve several goals:

Had at the same bit rate the smallest width of the spectrum of the resulting signal;

Provided synchronization between transmitter and receiver;

Had the ability to recognize mistakes;

Has a low cost of implementation.

136 Chapter 2 Discrete Data Transfer Basics

A narrower spectrum of signals allows you to achieve a higher data transfer rate on the same line (with the same bandwidth). In addition, the requirement is often imposed on the signal spectrum that there is no constant component, that is, the presence of direct current between transmitter and receiver. In particular, the use of various transformer circuits galvanic isolation prevents the flow of direct current.

Synchronization of the transmitter and receiver is needed so that the receiver knows exactly at what point in time it is necessary to read new information from the communication line. This problem is more difficult to solve in networks than in the exchange of data between closely spaced devices, for example, between units within a computer or between a computer and a printer. At short distances, a circuit based on a separate clocking communication line (Fig. 2.15) works well, so that information is removed from the data line only at the moment the clock pulse arrives. In networks, the use of this scheme causes difficulties due to the heterogeneity of the characteristics of the conductors in the cables. Over long distances, signal velocity ripples can cause the clock to arrive so late or too early for the corresponding data signal that a data bit is skipped or reread. Another reason why networks refuse to use clock pulses is to save conductors in expensive cables.

Therefore, networks use the so-called self-synchronizing codes, the signals of which carry indications for the transmitter at what point in time it is necessary to recognize the next bit (or several bits, if the code is oriented to more than two signal states). Any sharp edge in the signal - the so-called front - can be a good indication for synchronization of the receiver with the transmitter.

When using sinusoids as a carrier signal, the resulting code has the property of self-synchronization, since a change in the amplitude of the carrier frequency allows the receiver to determine the moment the input code appears.

Recognition and correction of distorted data is difficult to implement by means of the physical layer, therefore, most often this work is undertaken by the protocols that lie above: channel, network, transport or application. On the other hand, error recognition at the physical layer saves time, since the receiver does not wait for the frame to be completely placed in the buffer, but rejects it immediately upon recognition of erroneous bits within the frame.

The requirements for coding methods are mutually contradictory, so each of the popular digital coding methods discussed below has its own advantages and disadvantages compared to others.

______________________________2.2. Discrete data transfer methods at the physical layer _______137

Potential code without return to zero

On fig. 2.16, and shows the previously mentioned method of potential encoding, also called encoding without returning to zero (Non Return to Zero, NRZ). The last name reflects the fact that when transmitting a sequence of ones, the signal does not return to zero during the cycle (as we will see below, in other coding methods, a return to zero occurs in this case). The NRZ method is easy to implement, has good error recognition (due to two sharply different potentials), but does not have the self-synchronization property. When transmitting a long sequence of ones or zeros, the signal on the line does not change, so the receiver is unable to determine from the input signal the times when it is necessary to read the data again. Even with a highly accurate clock generator, the receiver can make a mistake with the moment of data acquisition, since the frequencies of the two generators are never completely identical. Therefore, at high data rates and long sequences of ones or zeros, a small mismatch of clock frequencies can lead to an error in a whole clock cycle and, accordingly, reading an incorrect bit value.

Another serious disadvantage of the NRZ method is the presence of a low frequency component that approaches zero when transmitting long sequences of ones or zeros. Because of this, many communication channels do not provide

138 Chapter 2 Discrete Communication Basics

those with a direct galvanic connection between the receiver and the source do not support this type of encoding. As a result, in its pure form, the NRZ code is not used in networks. Nevertheless, its various modifications are used, in which both the poor self-synchronization of the NRZ code and the presence of a constant component are eliminated. The attractiveness of the NRZ code, which makes it worthwhile to improve it, is the rather low fundamental frequency fo, which is equal to N/2 Hz, as shown in the previous section. Other coding methods, such as Manchester, have a higher fundamental frequency.

Bipolar coding method with alternative inversion

One of the modifications of the NRZ method is the method bipolar coding with alternative inversion (Bipolar Alternate Mark Inversion, AMI). In this method (Fig. 2.16, 6) three levels of potential are used - negative, zero and positive. To encode a logical zero, a zero potential is used, and a logical unit is encoded either by a positive potential or a negative one, while the potential of each new unit is opposite to the potential of the previous one.

The AMI code partially eliminates the DC and lack of self-timing problems inherent in the NRZ code. This happens when sending long sequences of ones. In these cases, the signal on the line is a sequence of bipolar pulses with the same spectrum as the NRZ code transmitting alternating zeros and ones, that is, without a constant component and with a fundamental harmonic of N/2 Hz (where N is the data bit rate) . Long sequences of zeros are also dangerous for the AMI code, as well as for the NRZ code - the signal degenerates into a constant potential of zero amplitude. Therefore, the AMI code needs further improvement, although the task is simplified - only sequences of zeros remain to be dealt with.

In general, for various combinations of bits on the line, the use of the AMI code leads to a narrower signal spectrum than for the NRZ code, and hence to a higher line throughput. For example, when transmitting alternating ones and zeros, the fundamental harmonic fo has a frequency of N/4 Hz. The AMI code also provides some features for recognizing erroneous signals. Thus, a violation of the strict alternation of the polarity of the signals indicates a false impulse or the disappearance of a correct impulse from the line. A signal with incorrect polarity is called forbidden signal (signal violation).

The AMI code uses not two, but three signal levels per line. The additional layer requires an increase in transmitter power of about 3 dB to provide the same bit fidelity on the line, which is a general disadvantage of codes with multiple signal states compared to codes that only distinguish between two states.

Potential code with inversion at unity

There is code similar to AMI, but with only two signal levels. When zero is transmitted, it transmits the potential that was set in the previous cycle (that is, it does not change it), and when one is transmitted, the potential is inverted to the opposite. This code is called potential code with inversion at unity

2.2. Discrete data transfer methods at the physical layer 139

(Non Return to Zero with ones Inverted, NRZI). This code is useful in cases where the use of a third signal level is highly undesirable, for example, in optical cables, where two signal states are reliably recognized - light and dark. Two methods are used to improve potential codes like AMI and NRZI. The first method is based on adding redundant bits containing logical ones to the source code. Obviously, in this case, long sequences of zeros are interrupted and the code becomes self-synchronizing for any transmitted data. The constant component also disappears, which means that the signal spectrum narrows even more. But this method reduces the useful bandwidth of the line, since redundant units of user information are not carried. Another method is based on preliminary "mixing" of the initial information in such a way that the probability of the appearance of ones and zeros on the line becomes close. Devices or blocks that perform this operation are called scramblers(scramble - dump, disorderly assembly). When scrambling, a known algorithm is used, so the receiver, having received binary data, transmits them to descrambler, which restores the original bit sequence. Excess bits are not transmitted over the line. Both methods refer to logical rather than physical coding, since they do not determine the shape of the signals on the line. They are studied in more detail in the next section.

Bipolar pulse code

In addition to potential codes, networks also use pulse codes, when the data is represented by a full pulse or its part - a front. The simplest case of this approach is bipolar pulse code, in which the unit is represented by an impulse of one polarity, and zero is the other (Fig. 2.16, v). Each pulse lasts half a cycle. Such a code has excellent self-clocking properties, but a DC component may be present, for example, when transmitting a long sequence of ones or zeros. In addition, its spectrum is wider than that of potential codes. So, when transmitting all zeros or ones, the frequency of the fundamental harmonic of the code will be equal to N Hz, which is two times higher than the fundamental harmonic of the NRZ code and four times higher than the fundamental harmonic of the AMI code when transmitting alternating ones and zeros. Due to the too wide spectrum, the bipolar pulse code is rarely used.

Manchester code

In local networks, until recently, the most common coding method was the so-called Manchester code(Fig. 2.16, d). It is used in Ethernet and Token Ring technologies.

In the Manchester code, a potential drop, that is, the front of the pulse, is used to encode ones and zeros. In Manchester encoding, each clock is divided into two parts. Information is encoded by potential drops that occur in the middle of each cycle. A unit is encoded by a low-to-high transition, and a zero is encoded by a reverse transition. At the beginning of each cycle, a service signal edge can occur if you need to represent several ones or zeros in a row. Since the signal changes at least once per transmission cycle of one data bit, the Manchester code has good

140 Chapter 2 Discrete Communication Basics _____________________________________________

self-synchronizing properties. The bandwidth of the Manchester code is narrower than that of the bipolar pulse. It also does not have a constant component, and the fundamental harmonic in the worst case (when transmitting a sequence of ones or zeros) has a frequency of N Hz, and in the best case (when transmitting alternating ones and zeros) it is equal to N / 2 Hz, like in AMI codes or NRZ. On average, the bandwidth of the Manchester code is one and a half times narrower than that of the bipolar pulse code, and the fundamental harmonic oscillates around 3N/4. The Manchester code has another advantage over the bipolar pulse code. The latter uses three signal levels for data transmission, while Manchester uses two.

Potential code 2B1Q

On fig. 2.16 d shows a potential code with four signal levels for encoding data. This is the code 2B1Q the name of which reflects its essence - every two bits (2B) are transmitted in one cycle by a signal that has four states (1Q). Bit 00 is -2.5V, bit 01 is -0.833V, AND is +0.833V, and 10 is +2.5V. sequences of identical pairs of bits, since in this case the signal is converted into a constant component. With random bit interleaving, the spectrum of the signal is twice as narrow as that of the NRZ code, since at the same bit rate the clock duration is doubled. Thus, using the 2B1Q code, you can transfer data over the same line twice as fast as using the AMI or NRZI code. However, for its implementation, the transmitter power must be higher so that the four levels are clearly distinguished by the receiver against the background of interference.

2.2.3. Logic coding

Logic coding is used to improve potential AMI, NRZI or 2Q1B type codes. Logic coding should replace long sequences of bits leading to a constant potential with interspersed ones. As noted above, two methods are characteristic of logical coding - redundant codes and scrambling.

Redundant codes

Redundant codes are based on splitting the original sequence of bits into portions, which are often called characters. Then each original character is replaced with a new one that has more bits than the original. For example, the 4V/5V logic code used in FDDI and Fast Ethernet technologies replaces original 4-bit symbols with 5-bit symbols. Since the resulting symbols contain redundant bits, the total number of bit combinations in them is greater than in the original ones. So, in the 4B / 5B code, the resulting symbols can contain 32 bit combinations, while the original symbols - only 16. Therefore, in the resulting code, you can select 16 such combinations that do not contain a large number of zeros, and count the rest prohibited codes (code violation). In addition to eliminating the DC component and giving the code the property of self-synchronization, redundant codes allow

2.2. Discrete data transfer methods at the physical layer 141

receiver to recognize garbled bits. If the receiver receives a forbidden code, it means that the signal has been distorted on the line.

Correspondence of source and resulting codes 4V/5V is presented below.

The 4B/5B code is then transmitted over the line using physical coding using one of the potential coding methods that is only sensitive to long sequences of zeros. The 4V/5V code symbols, 5 bits long, guarantee that no more than three zeros in a row can occur on the line for any combination of them.

The letter B in the code name means that the elementary signal has 2 states - from English binary - binary. There are also codes with three signal states, for example, in the 8B / 6T code, to encode 8 bits of initial information, a code of 6 signals is used, each of which has three states. The redundancy of the 8B/6T code is higher than that of the 4B/5B code, since there are 3 6 =729 resulting symbols per 256 source codes.

Using the lookup table is a very simple operation, so this approach does not complicate network adapters and interface blocks of switches and routers.

To provide a given line capacity, a transmitter using a redundant code must operate at an increased clock frequency. So, to transmit 4V / 5V codes at a rate of 100 Mb / s, the transmitter must operate at a clock frequency of 125 MHz. In this case, the spectrum of the signal on the line is expanded in comparison with the case when a pure, non-redundant code is transmitted over the line. Nevertheless, the spectrum of the redundant potential code turns out to be narrower than the spectrum of the Manchester code, which justifies the additional stage of logical coding, as well as the operation of the receiver and transmitter at an increased clock frequency.

Scrambling

Shuffling the data with a scrambler before putting it on the line with a candid code is another way of logical coding.

Scrambling methods consist in bit-by-bit calculation of the resulting code based on the bits of the source code and the bits of the resulting code obtained in previous cycles. For example, a scrambler might implement the following relationship:

Bi - Ai 8 Bi-s f Bi. 5 ,

where bi is the binary digit of the resulting code received at the i-th cycle of the scrambler, ai is the binary digit of the source code received at the i-th cycle of the

142 Chapter 2 Discrete Data Transfer Basics

scrambler input, В^з and B t .5 - binary digits of the resulting code obtained in the previous cycles of the scrambler, respectively, 3 and 5 cycles earlier than the current cycle, 0 - XOR operation (modulo 2 addition).

For example, for the source sequence 110110000001, the scrambler will give the following result code:

bi = ai - 1 (the first three digits of the resulting code will be the same as the original one, since there are no necessary previous digits yet)

Thus, the output of the scrambler will be the sequence 110001101111, which does not contain the sequence of six zeros that was present in the source code.

After receiving the resulting sequence, the receiver passes it to the descrambler, which reconstructs the original sequence based on the inverse relationship:

Various scrambling algorithms differ in the number of terms that give the digit of the resulting code, and the shift between the terms. So, in ISDN networks, when transferring data from a network to a subscriber, a transformation is used with shifts of 5 and 23 positions, and when transferring data from a subscriber to a network, with shifts of 18 and 23 positions.

There are also simpler methods of dealing with sequences of ones, also classified as scrambling.

To improve the Bipolar AMI code, two methods are used based on the artificial distortion of the sequence of zeros by forbidden characters.

On fig. Figure 2.17 shows the use of the B8ZS (Bipolar with 8-Zeros Substitution) method and the HDB3 (High-Density Bipolar 3-Zeros) method to correct the AMI code. The source code consists of two long sequences of zeros: in the first case - from 8, and in the second - from 5.

The B8ZS code corrects only sequences consisting of 8 zeros. To do this, after the first three zeros, instead of the remaining five zeros, he inserts five digits: V-1*-0-V-1*. V here denotes a signal of one forbidden for a given cycle of polarity, that is, a signal that does not change the polarity of the previous one, 1* is a signal of a unit of correct polarity, and an asterisk marks that

2.2. Discrete data transfer methods at the physical layer 143

the fact that in the source code in this cycle there was not a unit, but a zero. As a result, the receiver sees 2 distortions in 8 clock cycles - it is very unlikely that this happened due to noise on the line or other transmission failures. Therefore, the receiver considers such violations as coding of 8 consecutive zeros and, upon reception, replaces them with the original 8 zeros. The B8ZS code is constructed in such a way that its constant component is zero for any sequence of binary digits.

The HDB3 code corrects any four consecutive zeros in the original sequence. The rules for generating the HDB3 code are more complex than the B8ZS code. Every four zeros are replaced by four signals that have one V signal. To suppress the DC component, the polarity of the V signal is reversed in successive changes. In addition, two patterns of four-cycle codes are used for replacement. If the original code contained an odd number of 1s before the replacement, then the OOOV sequence is used, and if the number of 1s was even, the 1*OOV sequence is used.

Enhanced candidate codes have a fairly narrow bandwidth for any sequences of 1s and 0s that occur in the transmitted data. On fig. 2.18 shows the spectra of signals of different codes obtained by transmitting arbitrary data, in which various combinations zeros and ones in the source code are equally probable. When constructing graphs, the spectrum was averaged over all possible sets of initial sequences. Naturally, the resulting codes may have a different distribution of zeros and ones. From fig. 2.18 shows that the potential NRZ code has a good spectrum with one drawback - it has a constant component. The codes obtained from the potential by logical coding have a narrower spectrum than the Manchester one, even at an increased clock frequency (in the figure, the spectrum of the 4V / 5V code should approximately coincide with the B8ZS code, but it is shifted

144 Glovo2 Fundamentals of discrete data transmission

to the region of higher frequencies, since clock frequency increased by 1/4 compared to other codes). This explains the use of potential redundant and scrambled codes in modern technologies such as FDDI, Fast Ethernet, Gigabit Ethernet, ISDN, etc. instead of Manchester and Bipolar Pulse Coding.

2.2.4. Discrete modulation of analog signals

One of the main trends in the development of network technologies is the transmission of both discrete and analog data in the same network. Discrete data sources are computers and other computing devices, and analog data sources are devices such as telephones, video cameras, sound and video equipment. In the early stages of solving this problem in territorial networks, all types of data were transmitted in analog form, while computer data that was discrete in nature was converted to analog form using modems.

However, as the technology for receiving and transmitting analog data has developed, it has become clear that transmitting them in analog form does not improve the quality of the data received at the other end of the line if they are significantly distorted during transmission. The analog signal itself does not give any indication either that distortion has occurred or how to correct it, since the waveform can be anything, including the one that was recorded by the receiver. Improving the quality of lines, especially territorial ones, requires huge efforts and investments. Therefore, analog technology for recording and transmitting sound and images has been replaced by digital technology. This technique uses the so-called discrete modulation of the original time-continuous analog processes.

Discrete modulation methods are based on discretization of continuous processes both in amplitude and in time (Fig. 2.19). Consider the principles of spark modulation using the example pulse code modulation, PCM (Pulse Amplitude Modulation, PAM), which is widely used in digital telephony.

The amplitude of the original continuous function is measured with a given period - due to this, time discretization occurs. Then each measurement is represented as a binary number of a certain capacity, which means discretization by function values ​​- a continuous set of possible amplitude values ​​is replaced by a discrete set of its values. A device that performs this function is called analog-to-digital converter (ADC). After that, measurements are transmitted over communication channels in the form of a sequence of ones and zeros. In this case, the same coding methods are used as in the case of transmission of initially discrete information, that is, for example, methods based on the B8ZS or 2B1Q code.

On the receiving side of the line, the codes are converted into the original bit sequence, and special equipment called digital-to-analog converter (DAC), performs demodulation of the digitized amplitudes of a continuous signal, restoring the original continuous function of time.

Discrete modulation is based on theory of the Nyquist-Kotelnikov mapping. According to this theory, an analog continuous function transmitted as a sequence of its time-discrete values ​​can be accurately reconstructed if the sampling frequency was two or more times higher than the frequency of the highest harmonic of the spectrum of the original function.

If this condition is not met, then the restored function will differ significantly from the original one.

The advantage of digital methods for recording, reproducing and transmitting analog information is the ability to control the reliability of data read from a carrier or received via a communication line. To do this, you can apply the same methods that are used for computer data (and are discussed in more detail below), - the calculation of the checksum, the retransmission of corrupted frames, the use of self-correcting codes.

For high-quality voice transmission in the PCM method, a quantization frequency of the amplitude of sound vibrations of 8000 Hz is used. This is due to the fact that in analog telephony, the range from 300 to 3400 Hz was chosen for voice transmission, which transmits all the main harmonics of the interlocutors with sufficient quality. According to the Nyquist-Koteltkov theorem for quality voice transmission

146 Chapter 2 Discrete Communication Basics

it is sufficient to select a sampling frequency that is twice the highest harmonic of the continuous signal, i.e. 2 x 3400 = 6800 Hz. The sample rate of 8000 Hz actually chosen provides some margin of quality. The PCM method typically uses 7 or 8 code bits to represent the amplitude of a single sample. Accordingly, this gives 127 or 256 gradations of the audio signal, which is quite sufficient for high-quality voice transmission. When using the PCM method, a bandwidth of 56 or 64 Kbps is required to transmit one voice channel, depending on how many bits each sample is represented. If used for this purpose

7 bits, then with a measurement transmission frequency of 8000 Hz we get:

8000 x 7 = 56000 bps or 56 kbps; and for the case of 8 bits:

8000 x 8 - 64000 bps or 64 Kbps.

Standard is digital channel 64 kbps, which is also called elementary channel of digital telephone networks.

The transmission of a continuous signal in a discrete form requires networks to strictly adhere to a time interval of 125 μs (corresponding to a sampling rate of 8000 Hz) between adjacent measurements, that is, it requires synchronous data transmission between network nodes. If the synchronism of incoming measurements is not observed, the original signal is not restored correctly, which leads to distortion of voice, image or other multimedia information transmitted over digital networks. For example, timing distortion of 10 ms can lead to an "echo" effect, and shifts between samples of 200 ms lead to loss of recognition of spoken words. At the same time, the loss of one measurement, while maintaining synchronism between the remaining measurements, has practically no effect on the reproduced sound. This is due to smoothing devices in digital-to-analog converters, which are based on the inertia property of any physical signal- the amplitude of sound vibrations cannot instantly change by a large amount.

The quality of the signal after the DAC is affected not only by the synchronism of the measurements received at its input, but also by the discretization error of the amplitudes of these measurements.

8 the Nyquist-Kotelnikov theorem assumes that the amplitudes of the function are measured accurately, while at the same time using for their storage binary numbers with a limited bit depth somewhat distorts these amplitudes. Accordingly, the restored continuous signal is distorted, which is called sampling noise (in amplitude).

There are other discrete modulation methods that allow you to represent voice measurements in a more compact form, for example, as a sequence of 4-bit or 2-bit numbers. At the same time, one voice channel requires less bandwidth, for example, 32 Kbps, 16 Kbps or even less. Since 1985, the CCITT voice coding standard has been used, called Adaptive Differential Pulse Code Modulation (ADPCM). ADPCM codes are based on finding differences between successive voice samples, which are then transmitted over the network. The ADPCM code uses 4 bits to store one difference and the voice is transmitted at 32 Kbps. More modern method, Linear Predictive Coding (LPC), makes sampling of the original function more infrequent, but uses methods to predict the direction of change in signal amplitude. Using this method, you can lower the voice rate to 9600 bps.

2.2. Discrete data transfer methods at the physical layer 147

Digitally presented continuous data can be easily transferred through a computer network. To do this, it is enough to place several measurements in the frame of some standard network technology, provide the frame with the correct destination address, and send it to the destination. The recipient must extract the measurements from the frame and submit them with a quantization frequency (for voice - at a frequency of 8000 Hz) to a digital-to-analog converter. As the next frames with voice measurements arrive, the operation should be repeated. If the frames arrive synchronously enough, then the voice quality can be quite high. However, as we already know, frames in computer networks can be delayed both in end nodes (while waiting for access to a shared medium) and in intermediate communication devices - bridges, switches and routers. Therefore, the voice quality when transmitted digitally via computer networks usually low. For high-quality transmission of digitized continuous signals - voices, images - today special digital networks are used, such as ISDN, ATM, and networks digital television. However, for the transfer of intracorporate telephone conversations Today, frame relay networks are typical, the frame transmission delays of which are within acceptable limits.

2.2.5. Asynchronous and synchronous transmission

When communicating at the physical layer, the unit of information is a bit, so the means of the physical layer always maintain bit synchronization between the receiver and the transmitter.

The link layer operates on data frames and provides synchronization between the receiver and transmitter at the frame level. It is the responsibility of the receiver to recognize the start of the first byte of the frame, recognize the boundaries of the frame fields, and recognize the end of the frame flag.

It is usually sufficient to ensure synchronization at these two levels - bit and frame - so that the transmitter and receiver can ensure a stable exchange of information. However, if the quality of the communication line is poor (usually this applies to telephone switched channels), to reduce the cost of equipment and increase the reliability of data transmission, additional funds byte-level synchronization.

This mode of operation is called asynchronous or start-stop. Another reason for using this mode of operation is the presence of devices that generate data bytes at random times. This is how the keyboard of a display or other terminal device works, from which a person enters data for processing by a computer.

In asynchronous mode, each byte of data is accompanied by special signals "start" and "stop" (Fig. 2.20, a). The purpose of these signals is, firstly, to notify the receiver of the arrival of data and, secondly, to give the receiver enough time to perform some timing-related functions before the next byte arrives. The start signal has a duration of one clock interval, and the stop signal can last one, one and a half, or two clocks, so one, one and a half, or two bits are said to be used as a stop signal, although these signals do not represent user bits.

The described mode is called asynchronous because each byte can be slightly shifted in time relative to the bitwise cycles of the previous one.

148 Chapter 2 Discrete Data Transfer Basics

bytes. Such asynchronous transmission of bytes does not affect the correctness of the received data, since at the beginning of each byte there is an additional synchronization of the receiver with the source due to the "start" bits. More "free" time tolerances determine the low cost of the equipment of the asynchronous system.

In synchronous transfer mode, there are no start-stop bits between each pair of bytes. User data is collected in a frame, which is preceded by synchronization bytes (Fig. 2.20, b). The sync byte is a byte containing a pre-known code, such as 0111110, which notifies the receiver that a data frame has arrived. Upon receiving it, the receiver must enter into byte synchronization with the transmitter, that is, correctly understand the beginning of the next byte of the frame. Sometimes several sync bytes are used to provide more reliable synchronization between the receiver and transmitter. Since the receiver may have problems with bit synchronization when transmitting a long frame, self-synchronizing codes are used in this case.

» When transmitting discrete data over a narrowband voice-frequency channel used in telephony, analog modulation methods are most suitable, in which the carrier sinusoid is modulated by the original sequence of binary digits. This operation is carried out by special devices - modems.

* For low-speed data transmission, a change in the frequency of the sinusoid carrier is applied. Higher speed modems operate on combined Quadrature Amplitude Modulation (QAM) techniques, which are characterized by 4 levels of carrier sinusoid amplitude and 8 levels of phase. Not all of the possible 32 combinations of the QAM method are used for data transmission, forbidden combinations allow you to recognize distorted data at the physical level.

* On broadband communication channels, potential and pulse coding methods are used, in which data is represented by different levels of a constant signal potential or by the polarities of a pulse or its front.

* When using potential codes, the task of synchronizing the receiver with the transmitter is of particular importance, since when transmitting long sequences of zeros or ones, the signal at the input of the receiver does not change and it is difficult for the receiver to determine the moment of picking up the next data bit.

___________________________________________2.3. Data link layer transmission methods _______149

* The simplest potential code is the non-return-to-zero (NRZ) code, however it is not self-clocking and creates a DC component.

» The most popular pulse code is the Manchester code, in which the information is carried by the direction of the signal edge in the middle of each cycle. Manchester code is used in Ethernet and Token Ring technologies.

» To improve the properties of a potential NRZ code, logical coding methods are used that exclude long sequences of zeros. These methods are based on:

On the introduction of redundant bits into the original data (4V/5V type codes);

Scrambling of the original data (codes like 2B1Q).

» Enhanced potential codes have a narrower spectrum than pulse codes, so they are used in high-speed technologies such as FDDI, Fast Ethernet, Gigabit Ethernet.