The Use of Digital Circuits in Data Transmission

,1969, No. 3 71

The use of digital circuits in data transmission

P. J. van Gerwen

Data transmission is the name given to the transmission of digital signals in such a form that they can be applied directly to a computer. Now that computers are being more and more widely used, it is becoming much more important to be able to transmit these signals reliably and at high speed. In this article it is shown that where data signals have to be transmitted over ordinary telephone lines, digital circuits have several advantages over conventional circuits containing inductors and capacitors. A survey is given of digital circuits designedfor this purpose. The use of such circuits has made it possible to design a data transmitter in the form of an integrated circuit on a chip a few millimetres square.

A significant development of the last few years is the transmission, but in order to use a telephone system growth of a new branch of computer technology: the for binary pulse trains an extra operation is required. combination of computer and telecommunications. To Telephone lines are, after all, designed for speech: only make the most economic use of the computer, it is signals at frequencies required for intelligibility are located at a central point and connected to the more or transmitted (ranging from 300 to 3400 Hz). Data sig- less distant places where the users are situated. The nals, however, have' a spectrum that also possesses a information to be processed then has to be supplied to d.c. component and a number of a.c. components out- the computer, often over a telephone line, and the side this band, including components at very low fre- resultant information is communicated to the user in quencies. Another difficulty is caused by the fact that the same way. The same situation is found where a carrier telephony systems with many channels are used computer memory store is used as an information pro- for long-distance communications. In this case there is cessing centre, e.g. for the book-keeping of bank trans- usually a small frequency difference between the carrier actions. Connection with the store allows any branch that is modulated in a channel at the transmitting end of the bank to have details of the state of customers' and the carrier that is .used for demodulation at the accounts without human intervention, and thus permits receiving end. Because ofthis there is a small frequency demands for payment to be accepted without delay. shift in the spectrum of the transmitted signals. This In this kind of procedure the information (the data) does not matter for speech, where only the spectral has to be transmitted in such a way that it can be distribution of the signals is of importance; for data directly processed by the computer; this is known as transmission, however, the shape of the signals is data transmission. Existing means of telecommunica- important, since it is determined by the time function. tion would seem to be the most convenient means for The shift in frequency upsets the harmonic relation this purpose, in particular telegraph and telephone between the components of the spectrum, and this may lines. Can these be used directly, without modification lead to more than the admissible level of distortion. for transmitting data signals, i.e. signals consisting of This distortion could also be aggravated by the fact series of binary pulses? Although telegraph systems are that the phase characteristic of telephone circuits is capable of handling binary signals, the bandwidth strongly curved, particularly at the edges of the trans- . available per telegraph channel is so small that the mitted spectrum. If one were to attempt to transmit speed at which the characters can be transmitted is far signals below 300 Hz in a telephone channel, the signals too low for modern data transmission. In telephone would be severely distorted. circuits the bandwidth is greater, giving a higher rate of However, it should be possible to avoid these difficulties by applying the extra operation mentioned above. P. J. van Genven is with Philips Research Laboratories, Eindhoven. In this operation the signals to be transmitted modulate 72 PHILIPS TECHNICAL REVIEW VOLUME 30

a separate carrier in such a way. that the principal com- phony, it is necessary to use filters in the transmitter. ponents of the modulated signallie within the frequen- As indicated in fig. 1, the data signal is usually put cy band available for telephony. The original signal is through a low-pass filter before being applied to the then obtained by demodulation with this "data car- modulator. After the filtering process, the higher- rier" in the receiver. Now if there is a frequency shift frequency components in the data signal are no longer of the spectrum because the signal has arrived via a present; the signal is then no longer binary. The output carrier telephony system, there will also be a shift in signal from the modulator is passed through a band- the data carrier. If the data carrier used for demodula- pass filter to suppress unwanted modulation products. tion in the receiver is given the same frequency shift, If both sidebands of the modulated signal are trans- which can be done by conventional methods, the mitted, the design of this filter presents no great prob- demodulated data signal again has the required shape. lems. In vestigial-sideband modulation however (see Amplitude, frequency or phase modulation may be below) this band-pass filter also has to nearly suppress used. Each of these methods has its advantages and one of the sidebands. The design of the filter is then disadvantages for data transmission, but we shall not much more difficult. go into this here. Fig. 2 shows the block diagram of a conventional receiver for data signals. A band-pass filter is used here The essentials of a data system at the input to suppress interference outside the re- Fig. 1 shows a block diagram of a conventional data quired frequency band. After demodulation the signal transmitter. The binary elements ofthe data signal, the again goes through a low-pass filter, and the original

BP i/UV\r .----:::~::-:----,I____ - __ 1_ ~

Fig. 1. Block diagram of a conventional transmitter used for transmitting data via a telephone line. DATA data-signal source, Cl clock-signal source, Mod modulator, Carr carrier source, LP low-pass filter, BP band-pass filter, L telephone line. The waveforms ofthe signals are shown beside the appropriate connections. The modulation system used here is phase modulation.

"bits", are usually supplied to the transmitter in time form of the data signals is then reconstituted with the intervals controlled by a clock signal, which consists of aid of the clock signal. a series of equally spaced pulses, The frequency of this signal, the clockfrequency, is thus equal to the number Digital circuits of bits per second, the bit frequency. Transmission using The conventional filters widely used in telephony, a clock signal is known as synchronous data transmis- made up from inductors and capacitors, can in prin- sion. The clock signal can be transmitted in one form ciple also be used for. data transmission. Their use, or another to the receiver, where it is used for reconsti- however, involves particular problems. Besides the tution of the data signal already demodulated with the attenuation characteristic (attenuation of the signal appropriate carrier frequency.' as a function of frequency) the phase characteristic The data signal modulates the carrier at the trans- (phase shift as a function of frequency) now has to be mitter in a modulator. The carrier signal can be pro- taken into account as well. To transmit the bits duced by an oscillator. Ifthis is tied to the clock signal, correctly it is desirable that the phase shift should be so that a. fixed relation exists between the clock fre- proportional to the frequency, in other words that the quency and the carrier frequency, the data system is phase characteristic should be linear. (In telephony the said to be fully synchronized. Instead of using a separate phase characteristic is not usually important.) This oscillator, the carrier signal can also be derived from makes it difficult to build conventional filters for data the clock signal. transmission, especially if a sharp cut-off is required at Since the output signal from the transmitter has to the edge of the frequency band; i.e. if the attenuation remain within the frequency band laid down for tele- characteristic has to have steep sides. There is then no 1

1969, No. 3 DIGITAL CIRCUITS IN DATA TRANSMISSION 73

alternative but to apply a phase correction. Another feature of digital circuits that makes their A new principle of filter design has now been devel- use for data transmission attractive is their great ver- oped which presents new and attractive possibilities. As satility. We mean by this that an existing system can will be shown, signals in certain frequency bands can often be combined with other units to allow operation be suppressed by means of delay elements. This makes at other transmission rates, and that other features such it possible to obtain an attenuation characteristic which as the attenuation characteristic of a filter can easily be . has steep sides yet still has a linear phase characteristic. modified. In conventional techniques this usually re- Such filters therefore require no phase correction. These quires an entirely new circuit design. filters are particularly attractive for digital signals, since they can be designed entirely with digital tech- Transmission rate and bandwidth niques, using bistable circuits (flip-flops) as delay ele- The binary elements of a data signal can be trans- ments. The operation and application of such digital mitted at a rate which increases with the bandwidth of filters [1] is one of the subjects of this article. the transmission path. At a bandwidth of B hertz it is In conventional telecommunication systems induc- theoretically possible to transmit 2B binary elements tors or transformers are usually used in the modulators. per second. Many of the data systems in use today are In a data transmission system there are some advan- still a long way from this theoreticalIirnit, since t~e tages if these are also made up from digital circuits. amplitude and phase characteristics ofthe transmission Some circuits for digital modulators will be dealt with path also have an effect on the transmission rate. For in more detail shortly. example, a non-linear phase characteristic can cause

__ L_

Fig. 2. Block diagram of a conventional data receiver. L telephone line, BP band-pass filter; Dem demodulator, Carr carrier, LP low-pass filter, Reg circuit for reconstituting the data signals, Cl clock-signal source.

The situation is different for the data receiver. The distortion which increases with the transmission rate. modulated signal entering the receiver should be con- Moreover, the permissible rate depends on the form in sidered as an analogue signal rather than a digital which the-modulated signal is transmitted. The theoret- signal. The advantages of digital circuits seem less ical-limit mentioned above of 2B elements per second obvious here. In fact, however, there are many ad- is based on the assumption that only one of the side- vantages in including digital circuits in data receivers, bands of the modulated signal is transmitted, Often, .and it seems likely that this will become standard prac- however, both sidebands are transmitted, and this tice in future. This point will be dealt with at the end reduces the permissible rate by a factor of 2. It is dif- of the article. ficult- to suppress one of the sidebands completely. The rapid increase in the use of digital circuits is Since the data signals contain components at very low partly due to the advances made in the microminiaturi- frequencies, signal components arise in the modulation zation of electronic circuits. Digital circuits require a process which have frequencies very close to the carrier fairly large number of elements, and it has now proved frequency.1t is therefore no easy matter to make a filter possible to design these circuits so as to use only tran- that can clearly separate the two-sidebands. A middle sistors and resistors - components that lend them- way can be taken by using vestigial-sideband modula- selves well to monolithic construction in an integrated circuit. This greatly reduces the disadvantage of using [1] See P. Leuthold, Filternetzwerke mit digitalen Schieberegi- a large number of elements, and later on in this article stern, Philips Res. Repts. Suppl. 1967, No. 5, also published in Mitteilungen aus dem Institut für Hochfrequenztechnik an we shall discuss a data transmitter incorporated in a der Eidgenössischen Technischen Hochschule in Zürich .crystal chip a few millimetres square. published by Prof. Dr. F. EiBorgnis, 1967, No. 1. 74 PHILIPS TECHNICAL REVIEW VOLUME 30

tion, in which one of the sidebands is only partly sup- +11 U2 = UI ~ Cic exp{-jw(n + k)'i}. (2) pressed [2]. k=-n In the foregoing we have been concerned with a carrier which If the delay network is symmetrical, in other words if is modulated by a binary signal, that is to say the carrier can be Cic C-Ic, then we can write (2) in the form: in two distinct "states". (This may relate to the amplitude, the = frequency or the phase, depending on the method of modulation used.) It has been found, however, that an even greater rate can U2 = UI [Co + 2 k~! c, cos kw'i] exp (-jwn'i). (3) be reached by modulating in such a way that more than two states ofthe carrier are used. With a "quaternary" system, for instance, The phase shift between UI and U2 is omx, which is using four different carrier states, the transmission rate can be thus again proportional to the frequency. The phase doubled. A method of modulating the carrier in such a way consists in starting from a data signal which is digital, but not characteristic is therefore linear. The amplitude charac- binary, and thus contains elements with different amplitudes. teristic is: (Quaternary modulation methods do also exist, however, which n start from a binary data signal.) A(w) = Co + 2 ~ Cic COS kort, ... (4) A disadvantage of this method of increasing the transmission k=! rate is that it makes the system more sensitive to noise and other interference. Moreover, it imposes even stricter requirements on The form of this characteristic can thus be influenced the amplitude and phase characteristics of the transmission path. by the choice ofthe weighting factors. The best approxi- . , mation to a specified shape for the amplitude charac- Filters based on delay elements

We have mentioned above that signal components at Del certain frequencies can be suppressed by using a delay network: This can be seen as follows. If an input signal UI of frequency w is given a delay 'i and this delayed signal is added to the original one (fig. 3), we may write the expression for the output signal as:

U2 = UI {I + exp (-jW'i)} = 2UI cos 1-W'i exp (-tjW'i). (1)

The amplitude of U2 thus varies with frequency as a Fig. 3. When a sinusoidal signalul is fed to a delay element De/, and the delayed signal is added to Ill, the result is an output signal cosine function. At the frequencies where cos 1-W'i= 0 U2 which is zero at certain frequencies. we have U2 = O. There is also a phase shift between U2 and UI, which is equal to 1-W'i,i.e. proportional to the frequency. teristic A(w) is obtained when the weighting factors . A filter may be obtained by using a large number of satisfy the equation: these delay elements, Fig. 4 shows a series arrangement niT of 2n elements, each of which gives a delay 'i. The out- c, = A(w) cos kw. dw. (5) put signals of the elements are added in a particular 2:I series of ratios, which are expressed by the introduetion -niT of weighting factors Cic (-n ~ k ~ n), which operate With the weighting factors chosen in this way the on the signals before addition. If the input signal is shape of the function A(w) is usually only matched again UI, then the output signal obtained in this way is: exactly for n = 00, i.e. for an infinitely large number

Del

. Fig. 4. Series arrangement of 211 delay elements Del, each with a delay time T. The output signals of the elements are added in certain ratios given by the weighting factors C-n ... C«. a input, b output. , ~ \1, JHJUfS' ilLufJL;;'nr(...... 1969, No. 3 DIGITAL CIRCUITS IN DATA TRANSMISSION • • rltAutflCÄbsl of delay elements. In practice sufficient elements are co A(w) = 2 :E Ck sin kWT:. . (6) chosen to give the required degree of approximation. k=l From equations (3) and (4) we see that the delay net- This is obtained exactly with weighting factors C« given by: works under consideration, with our symmetry condi- niT tion Ck = C-k, are able to provide all the amplitude c; = 2T: A(w) sin kWT: dw. . ... (7) characteristics that can be represented by a Fourier n.r series containing cosine terms only. This implies that -niT A is an even periodic function of os, An amplitude In this case it also follows from equation (;2) that besides a com- characteristic of this nature is shown in fig.5. Its ponent proportional to the frequency the phase shift between input and output contains a constant component equal to 90°. period Jr is equal to ï]», The value of -r: is usually This can be utilized for producing signals with single-sideband taken at the smallest that will give an adequate approxi- modulation [3]. mation to A(w) with the chosen number of elements. The frequency Jr then becomes so high that it is easy Filters made on this principle are known as "trans- to limit the passband to one of the bands indicated in versal filters" [4]. They have not been widely applied because it is difficult to make low-loss delay elements for analogue signals from inductors and capacitors. Another difficulty is that such filters are rather large JT\ lT\ since they require a large number of delay elements. o fa Fr -f 2fr Digital filters

Fig. 5. Amplitude characteristic of a filter with delay elements. Both of the above difficulties disappear if filters ex- This characteristic can be represented by a Fourier series con- clusively for digital signals are to be designed. The delay taining cosine terms only. The repetition period fr of the passbands is equal to lIT: (where T: is the delay time of the elements); elements can then be bistable circuits (flip-flops). These the cut-off frequency of the first passband is denoted by fa. If are connected in series and in effect form a shift register the first passband only is required, the others can be suppressed by means of a simple filter. (seefig. 6). The bistable circuits are switched from one

1----..---_--_------

Fig. 6. Digital filter made up from a shift register consisting of a series arrangement of bistable circuits (flip-flops) FF. R -n ... R« and r form a resistance network, which ensures that the output voltages of the bistable circuits are added in a particular series of ratios. Osc oscillator that pro- r duces the control pulses for the shift register, a input, boutput. the figure. For example, the signals at frequencies close state to the other by control pulses, causing the signal to Jr, 2Jr, etc., can be suppressed with a simple RC net- to travel through the shift register. The frequency of work, retaining only the lowest passband (cut-off fre- these pulses should be higher than the clock frequency quency Jo in fig. 5). When this is done the amplitude ofthe data signal, and is preferably anintegral multiple and phase of the signals in this passband are not of the clock frequency. The clock signals can for significantly affected, provided Jr is large. The result is thus a low-pass filter with a linear phase character- [2] One ofthe sidebands can be completely suppressed, to double istic. A band-pass filter can be made in much the same the transmission rate, if a special code is applied to the data signal, thus suppressing the components in the modulated way. . signal that are close to the carrier. This is described in: P. J. van Gerwen, On the generation and application of pseudo- A linear phase characteristic can also be obtained by choosing ternary codes in pulse transmission, Philips Res. Repts. 20, the weighting factors such that Ck = -C-k. It follows from (2) 469-484, 1965. [3] P. Leuthold and F. Tisi, Ein Einseitenbandsystem für Daten- that in this case the amplitude characteristic consists of a series übertragung, Archiv elektr. Ubertr. 21, 354-362, 1967. • with sine terms only. Assuming for simplicity that the number [4] See H. E. Kallmann, Transversal filters, Proc. LR.E. 28, of delay elements is infinitely large, the characteristic is given by: 302-310, 1940. 76 PHILIPS TECHNICAL REVIEW VOLUME 30

example be obtained by frequency division from the Since a digital filter with a shift register is far easier oscillator supplying the control pulses. Conversely, the to make than a filter containing a large number of delay control pulses can be derived by frequency multiplica- elements for analogue signals, digital filters have also tion from the clock signal, or the control-pulse oscilla- been used for analogue signals in some cases. In this tor can be synchronized with a higher harmonic of the case the analogue signal is first converted into a digital clock signal. signal in an analogue-to-digital converter. (This can be The weighting factors for the output signals of the done by using pulse-code modulation or delta modula- bistable circuits are obtained with a resistance network tion.) The filter is then followed by a decoder. Filters consisting of resistors R-n ... Rn and r, The voltage built in this way are known as analogue code filters. across r is then proportional to the sum of these signals, Since a bistable circuit can deliver two output Digital modulators signals in opposite phase, the weighting factor may be The design of a digital modulator is based on the use positive or negative as desired. The output signal does of a carrier in binary form, i.e. a' "square wave"; its of course contain components whose frequencies are repetition frequency should be an integral multiple of close to those of the control pulse and their higher half the clock frequency of the data signal. The data harmonics, These can be suppressed by connecting a signal also has to be supplied to the modulator in binary capacitor across r, form. (It is therefore not possible to use a low-pass filter Fig. 7 shows the attenuation characteristic of a filter as in fig. I.) With a carrier of this form, it is fairly easy of this kind which has 24 bistable circuits. The fre- to design digital modulators for amplitude modulation quency fr has been taken as 10fo. The dashed line and also for frequency and phase modulation, as we represents the characteristic that would be obtained shall now show. with an infinite number of elements. Amplitude modulation can be obtained by means of It is important to note that the weighting factors, and an AND gate with two inputs. If we apply to these hence the shape of the amplitude characteristic of the inputs two different binary signals A and B, whose filter, are determined by the relative values of the resis- instantaneous values can be referred to as 1 and 0 tors. This kind of filter (except for the capacitor) respectively, we obtain at the output a signal U which therefore lends itself very readily to integrated circuit has the value 1 if both A and B have this value. If techniques, since the relative values of the resistors in A = 0 or B = 0 (or A = B = 0), then U = O. In an integrated circuit can be fixed much more accu- Boolean algebra this function is expressed as A·B. rately than their absolute values. Fig. 8 shows the form of a binary data signal A with a "binary carrier" B whose frequency (the fundamental v frequency, in analogue terms) is twice the clock frequency of A. The carrier again occurs in the output signal U = A·B, but now modulated in amplitude by the data signal. Since the carrier is not sinusoidal, the modulation products of its higher harmonics are also obtained. As a rule these lie far outside the avail- 30~------~------~ able frequency band and can therefore be suppressed with à simple filter. The output signal U also contains the data signal itself, and since the waveforms of A and B never cross the zero line, ,U also contains a d.c. term. The data signallies in the frequency band of the modulated signal to be transmitted, and is therefore in effect a distortion of the modulated signal. It can easily be eliminated, however, by subtracting the data signal at half-amplitude from the output signalof the AND gate (fig. 9). This also has the result of removing the d.c. term from the output signal U. As can be seen from fig. 9, the signal thus obtained is no longer binary but ternary (with three levels). Such a signal can no longer be directly processed in a digital filter with Fig. 7. Attenuation characteristic of a digital filter containing bistable circuits, since these can only handle binary 24 bistable circuits, where fr = 10fo. With an infinite number of bistable circuits the characteristic would follow the dashed signals. line. . To obtainfi"equency modulation with a digital circuit t!" ..... 1969, No. 3 DIGITAL CIRCUITS IN DATA TRANSMISSION 77

8tO------<--I~

A :,-~1L___~ ~L_ ~L_ ___

8 :no 0 D 0 0 0 0 0 0 0 0 A ~_-~I -L ~ ~ ___

,8t :--00 D n 0 0 n 0 0 n n 0000 _I

Fig. 8. AND gate with two inputs. When two binary signals A and B are applied to these inputs, the resultant output signal is U = A . B. (Some of the symbols in this figure and figs. 9, 10 and 11 follow recent international recommendations; these are indicated below fig. 10.) 0000 DD

u ~--O0 D DOOD D C 0000 _t Fig, 10, Digital frequency modulator made up from two AND gates, one OR gate and one NOT gate, which supplies the inverted signal A. A data signal, BI and B2 binary carriers àt different frequencies, The output signal U is now a binary signal with frequency modulation. ooon uuuu_I I mOO2 Fig. 9. When the data signal A at half-amplitude is subtracted 1 0 from the output signalof an AND gate a ternary output signal U is obtained. This does not contain the data signalor a d.c. term, AND gate OR gate exclusive NOT gate but only the carrier, modulated by the data signal. OR gate (inversion)

it is necessary to have two digital carriers at different quency changes whenever the data signal switches from frequencies. (These can be obtained by frequency multi- the value 0 to 1 or vice versa. plication or by frequency division using a single oscilla- A digital phase modulator can be obtained by using tor.) The circuit must now work in such a way that a circuit known as a "modulo-Z adder". This circuit, either the one or the other carrier is transmitted, de- also referred to as an exèlusive OR gate, deli~ers an pending on whether the data signal has the value I or O. output signal at the value 1 when the input signals are An example of such a circuit is shown in fig. 10. It is unequal. The output signal is 0 when the input signals built up from two'AND gates, an OR gate and a NOT are equal (0 or 1). This relation between-the input sig- gate. The data signal is applied to the input A and the nals A and B and the output signal U is represented two 'carriers to the inputs BI and B2.' The frequency by the equation: of BI here is twice that of B2. The NOT gate delivers U = A . jj + A· B. the inverted data signal Ä. The output signal is now written as A· BI + ;r. B2 in Boolean algebra.' This In fig. 11 this signal is shown together with its .com- signal, together with its various component parts, is ponent parts. As long as the data signal A is' zero, U.is also shown in fig. 10. It is' a digital signal whose fre- equal to the carrier; if A has the value 1, then U is equal 78 PHILIPS TECHNICAL REVIEW VOLUME 30

of the modulating signal. The result is a signal whose spectrum is symmetrical with respect to the carrier wave. In data transmission, however, a carrier of say 1800 Hz is modulated with a data signal with a clock frequency of for example 1200 Hz. (The clock fre- A :- ....I ..L.._ __ ---'- __j__ _ quency and the carrier frequency may even be iden- tical.) Since the data signal is digital, it contains higher harmonics, and in this case their frequencies are higher B :un 0 ODD 0 D D 0 0 than the carrier frequency. If the carrier frequency is an integral multiple ofhalfthe bit frequency, the spectrum of the modulated signal will be asymmetrical. This is AB :----0 0 nooo illustrated in jig. 13, which shows the spectrum of a

: ____ hum 0 0 Ä.a 00 u :----0 n n 00 n no -t

Fig. 11. Exclusive OR gate. A data signal, B carrier. In the output signal U the carrier is modulated in phase with the data a signal. to the inverted carrier. This is in fact phase modulation. With binary signals a simple relation exists between phase modulation and amplitude modulation. If the inverted carrier is added to a phase-modulated digital signal the result is an amplitude-modulated signal. Fig. J 2 shows the signal U from fig. 11, together with the inverted carrier Jj and their sum U'. We see that the latter (ternary) signal is in fact amplitude-modulated by the data signal A of fig. 11. b The modulators discussed above differ from the ones normally used in radio engineering and in other branches of electronics in another respect besides the digital techniques. In ordinary modulators the carrier frequency is generally much higher than the frequency Fig. 13. a) Spectrum of a data signal consisting of periodic pulses. b) Spectrum of a phase-modulated signal, where the carrier frequency is equal to the clock frequency of the data signal.

000000 data signal, consisting of periodic pulses (fig. 13a), and the spectrum that results when phase modulation is applied with a carrier frequency equal to the clock 8:---0 D D D 0 0 D D DOl frequency of the data signal (fig. 13b). This latter spectrum consists of groups of components near to the carrier and near to its odd harmonics. The asymmetry is particularly evident near the carrier. A data system is generally designed in such a way that only the spectral region surrounding the carrier is transmitted, while the other components are sup- Fig. 12. When the inverted carrier ij is added to the phase- modulated signal U (see fig. IJ) an amplitude-modulated ternary pressed by a filter. The original objective that we men- signal U' is obtained. tioned at the beginning of the article, bringing the 1969, No. 3 DIGITAL CIRCUITS IN DATA TRANSMISSION 79 spectrum inside the telephony band, has thus been achieved. The asymmetry in the spectrum of the transmitted signal can cause interference effects. It may give rise to crosstalk between the bits of the data signal, result- ingin a smaller permissible interference margin during the reconstitution of the signal in the receiver. We shall just note here that when the carrier frequency is an integral multiple of half the clock frequency the asymmetry in the spectrum can be removed by means of a band-pass filter tbat passes the modulated signal (see r fig. I). If this filter is given an asymmetrical attenuation characteristic, a symmetrical spectrum can again Fig. 14. Block diagram of a data transmitter using phase modula- be obtained. The shape of the attenuation character- tion. Mod modulator, FD frequency dividers, Del shift register istic of a digital filter can be infl uenced, as we have seen, consisting of 12 bistable circuits, Tterminal for the control pulses Cl 0 by the choice of the resistors that determine the weight- for the shift register, clock signal, output. ing factors (fig. 6).

A data transmitter as an integrated circuit transmitter almost entirely in the form of an integrated Since digital circuits, which contain no inductors and circuit. Fig. 14 shows the block diagram of a vestigial- only a single capacitor, can now I be used both for sideband transmitter which, as an integrated circuit, modulators and for filters, it is possible to build a data measures only 2.7 x 2.1 mm. Fig. 15 is a much enlarged

Fig. IS. The data transmitter of fig. 14 as an integrated circuit, measuring 2.7 x 2.1 mm and including 203 transistors and 172 resistors. The upper three rows each contain four bistable circuits, the lower row contains the modulator and the two frequency dividers. 80 PHILlPS TECHNICAL REVIEW VOLUME jo picture of this monolithic circuit, whi~h includes '203 further ways of increasing the versatility of a trans- transistors and 172resistors, The transmitter.comprises mitter (at least, in so far as no limitations are set by a modulator for phase modulation, a filter, èontaining the unit to be integrated). The clock pulses and the twelve bistable circuits, and two frequency dividers carrier in this case are derived from the oscillator that (step-down ratio 1 : 2) which derive the clock signal delivers the control pulses for the shift register, by and the éarrier from the control pulses of the fiIt~r, means of three frequency dividers, one of which has a Bistábie circuits with an exceptionally low power con- fixed step-down ratio of 2. (The carrier frequency is sumption are used [51, giving a total power .consump- thus always an integral multiple of half the clock fre- , " tien of only 120 mW. The control pulses come from quency.) By suitably choosing the other step-down an oscillator which is not included in the circuitIf.the ratios, m and k, the transmitter can be made to , -sÓr rÓ: ,-" " frequency of this oscillator is 9.6 kHz, a datasignàl operate at different transmission rates with the oscilla- can be transmitted at a rate of 2400 bits per second. tor frequency kept fixed. For example, with the same The carrier frequency is then equal to the bit frequency, oscillator frequency, clock signals can be obtained at arid the 'Spectrum of the modulated output signal lies frequencies of 600, 1200 or 2400 bits/s, and there are within the range fro~ 600 to 3000 Hz. This transmitter various possibilities for the carrier frequency. With can-however also be used for a much higher transmis- an oscillator frequency of say 28.8 kHz, and with sion rate (a channel with a larger bandwidth is then m = 8 and k = 3, we obtain a clock signalof required). If the oscillator frequency is increased by a 1200 bitsis and a carrier frequency of 1800 Hz. factor of 20, the transmission rate and the carrier fre- The transmitter of fig. 16 also allows a signal to be quency are increased by, the same factor. This means transmitted with vestigial-sideband modulation pr with. that the transmission rate is then 48 000 bits per double-sideband modulation. For this purpose the second; ,J:h~ spectrum of, the modulated, signal then shift register is provided with two resistance networks, " J:. I ranges from 12 to 60 kHz. " which can be connected in by two switches SI and S2. Th~ersatilitY'ofthe transmitter-is alsodemonstrated In 'One of these networks the resistors have values by the relative ease with which it can be altered from which allow the lower sideband to be transmitted, and amplitude modûlation to phase modulation. We have in the other they are chosen to allow the upper side- already shown.how thls-can be done by simply adding band to be transmitted. If both switches are closed, the inverted ca?rier.....to th;'pbase-modulated signal both sidebands are transmitted. Fig. 17 shows the cor- , . -.... _, ~ (fig. 12). The resultis a ternary signal, aiïd-we have seen responding attenuation characteristics, measured at a thatthe spectrum of this sign~nnot be l~fèd",,~ith carrier frequency of 1800 Hz and a clock signalof a digital filter made up from bistable ci;C'Uifs:What ~ 1200 bits/s. Finally, fig, 18 shows some of the corre- "', . , ~ be done, however, IS to pass the phase-modulated sponding signals: the data signal, the modulated signal ~bjn~ry) sign~l thr?ugh the filter bero~e, a~ding the (with phase modulation), the output signal from the inverted cam er to It. . . c-, filter and the same signal after suppression of fre- e-, Another" block diagram (fig. 16) demonsffätes quency-components in the higher passbands. "

00,

,----+----0-.,.-----+--..,------:-;,~- --_ ...... 1-0

Fig. 16, Block diagram of a more versatile data transmitter, The switches SI and S2 allow either the upper sideband, the lower sideband or both sidebands to be transmitted, The appropriate choice of the step-down ratios III and k gives a fairly wide margin of freedom in the choice of the clock frequency and the carrier frequency. 1969, No. 3 DIGITAL CIRCUITS IN DATA TRANSMISSION 81

\ J / I, ,I'._ " , , a , I v 5 , I 7~ I 20 '>S2-""""_ \ I \ I b \ I \ I 10 \ I \ I \ I \I

3kHz

Fig. 17. Attenuation characteristics that can be obtained with the circuit of fig. 16 with a carrier frequency of 1800 Hz and a clock frequency of 1200 bits/so When switch SI is closed, the c lower sideband is transmitted and the upper sideband is strongly attenuated; when S2 is closed the filter transmits the upper sideband. When both switches are closed, both sidebands are transmitted.

Digital circuits in data receivers As we noted earlier, the signals reaching a data receiver via a telephone line are no longer binary (or ternary, etc.) because of the bandwidth limitation and the possibility of a freq uency shift. For this reason conventional receivers generally use analogue circuits. However, the rapid advances in integrated-circuit techniques present new possibilities here as well. This d applies particularly to the filters transmitting the signal before and after the demodulator (see fig. 2). A difficulty arises with the analogue code filters mentioned earlier, in which the analogue signal is fed to a digital filter through an analogue-digital converter and the output signalof this filter passes through a decoder. In order to obtain a sufficient accuracy in encoding the analogue Fig. 18. Signals occurring in the circuit shown in fig. 16. a data signal, b phase-modulated signal before the filter, c signal after signal digitally a fairly high frequency has to be used the filter with components in the higher passbands, d signal for the control pulses of the shift registers. This high arising when the latter components are suppressed by means of an extra filter. frequency necessitates the use of a large number of shift register elements. However, when integrated circuits are used, this is no longer a real difficulty, and it therefore seems likely that the use of analogue code Summary. The circuits used in conventional data transmission systems are very similar to those used in other branches of elec- filters in data receivers will in due course become an tronies: filters, modulators, etc., which are mainly made up from economic proposition. Since digital circuits are already inductors and capacitors. Since data signals consist of series of binary pulses, however, many of the circuit functions can be being used for other receiver functions, in particular performed by digital circuits, which are easily made as integrated for reconstituting the data signal, both data receivers circuits. Digital filters can be made by using shift registers, and logic circuits such as AND, OR and NOT gates can be used as and data transmitters will then consist almost 'entirely digital modulators. Tn this way it was possible to make a com- of digital circuits. plete data transmitter in integrated form on a crystal chip measuring 2.7 x 2.1 mm. A particular feature of the circuits is their great versatility, since various characteristics of the data trans- [5] See A. Slob, Fast logic circuits with low energy consumption, mitter can be easily modified. Wider use of digital circuits in data Philips tech. Rev. 29, 363-367, 1968 (No. 12). receivers is also to be expected.