‘stems pulse encat higher THe, es the aly de. Iowideh w ‘sually, vency outin filter: de re bie, it aF the volves ability fowith nitted as the fined retion form, ation: inded 2.10 2.10 WILUERT TRANSFORM 79 Note that the condition for stability involves only che potes of the transfer func ion H'(s); the zeros may indeed lie anywhere in the splane. Two types of systems may be distinguished here: + Minimumphase stens, characterized by a transfer function whose poles and zeros are all restricted to lie inside the left half of the s-plane, + Nonminimum-phase systems, whose transfer furictions are perinitted to have zeros fon the imaginary axis as well as the right half of the plane, In the case of low-pass filters where the principal requirement is to approx: imate the ideal amplitude response shown in Fig. 2.22, we may mention wo popular families of filters: Butterworth filters and Chebycheo fiters, both of which have all their zeros at s = ©. In a Butterworth filter, the poles of the transfer function H"(s) lie on a cirele with origin as the center and 2B as the radius, where Bis the 3dB bandwidth of the fitter. lo a Chebyshev filter, on the other hand, the poles lie on an ellipse, In both cases, of course, the polesare confined to the left half of the «plane. Turning next to the issue of physical realization of the filter, there are several options available to us, depending on the technology of choice: *+ Analog filters, built using (a) inductors and capacitors, or (b) capacitors, resis tors, and operational amplifiers, + Discretetime filter, for which the signals are sampled in eae buc their amplitude is continuous. These filters include switched-capacitor filters, and surface- acoustic wave (SAW) fil + Digital fters, for which the signats are sampled in time and their amplitude is also quantized. These filters are built using digital hardware; hence the name. An important feature ofa digital filter is that itis programmabls zhereby offering, a high degree of flexibifity in design. HILBERT TRANSFORM. ‘The Fourier transform, which has occupied so much of our attention thus far, is particularly useful for evaluating the frequency content of an energy signal or, in a limiting sense, that of a power signal, As such, it prosides the mathematic: basis for analyzing and designing frequency-selectve filters for the separation of signals on the basis of their frequency contents. Another method of separating. signals is based on phase seleivity, which uses phase shifis between the pertinent signals 0 achieve the desired separation. The simplest phase shift is that of 180 degrees, which is merely a polarity reversal in the case of a sinusoidal signal Shifting the phase angles of all components of a given signal by 180 degrees requires the use of an ideal transformer. Another phase shift of interest is that of 90 degrees. In particular, when the phase angles of all components ofa given signal are shifted by +90 degrees, the resulting function of time is known as the Hilbert wansform of the signal. To be specific, consider a signal g{0) with Fourier wansform G(f). The “Hilbert transform of g(4), which we shall denote by (0), is defined by"? Af 6 KOS ae (2.124)80 REPRESENTATION OF SIGNALS AND SYSTEMS Clearly, the Hilbert transformation of g(t) is Tinear operation. The inwerse Hit eit transform, by means of which the original signal g(1) is recovered from é(¢), is defined by / woe Af Be es The functions g0) ad g) ave sid 6 consite a Hil reson pais short table of Hibereransform pais given in Table 3 in Appendix 1] atthe end of the book We nate fern the definition of the ibe wransormn that gv) may be in terpreted as the convolution of g(t) with the time function 1/(7t). We also know from the comodution theorem (Pvopery 12 of tht Fourier tansform) tht the plication of their Fousier tanstorins i the frequency demain. For the ine function 1/4 we have ~ jgntS) 2.126) where sgn(f) is the signum function, defined in the frequency domain as 1 f>0 sn) =4 0 feo (2.127) “fo The Fouriersransform pair of Eq. (2.126) is obtained by applying the duality property of the Fourier transform to Eq. (2.74). In light of Eq. (2.126), 1 follows therefore that the Fourier transform G(J) of g(t) is given by AS) = mj sga NGO (2.128) Equation (2.128) states that given a signal g(2), we may obtain its Hithert transform (0) by passing g(?) through a linear wo-port device whose transfer function is equal 10 ~jsgn(S). This device may be considered as one that pro- ‘aces a phrase shift of ~ 90 degrees forall positive frequencies ofthe input signal and +90 degrees for all negative frequencies, as in Fig. 2.28, The amplitudes of all frequency components in the signal, however, are unaffected by transmission through the device. Such an ideal device is referred to as a Hilbert transformer. ‘The Hilbert transform has several important applications, which include the following: . 1. Itcan be used to realize phase selectivity in the generation of a special kind of modulation known as single sideband modulation. We shail have more 16 say about this application in Chapter 3. 2 It provides the mathematical basis for-the representation of band-pass sig. nals, This application is discussed im Sevtion 2.12 “The Hilbert transform, as defined aboxe, applies to any signal that is Fourier wansformable, Accordingly, it may be applied to energy signals as well as power signals.ews 2.10. HILBERT TRANSFORM st ity Hit wif W, | ») iH zp | i| sn lof | | ow i the | ait Figure 228 Phase “characteristic of ine {inear two-port device for obtaining the | Hilbert transform ofa real-valued signal. | “0 EXAMPLE 13 Sinusoidal Functions i | gO = cost2nf.t) i whose Fourier transform is i 2) 1 OS) = 5S 1) + 8S + FT 4 ly | Using this Fourier transform in Eq. (2.128), we get | os GS) = ~j en NGS) {i th i Las ~ fy + af + Ds } 28) : gOS ~ S) + 8S + Sd Isen(F) 1 Atal - fy - ar + 1 i fer a ij ve which represents che Poucier transform of the sine function sin 2uf.¢). The Hit nal bert transform of the cosine function is therefore equal to sin(2nf.t). { of . In a similar way, we find that the sine function sin(2nf,0) has a Hilbert ! on transform equal to ~ cos(2xf.t). Bi he Properties of the Hilbert Transform i | nd ‘The Hilbert transform differs from the Fourier transfarm in that it operates a ‘exclusively in the time domain, It has a number of useful properties, some of " which are listed below. To derive these properties, we make use of Eq. (2.128), which defines the relationship between the Fourier transform of a signal g(0) ‘s and that ofits Hilbert transform g(t). The signal g(t) is assumed to be real valued, which is the usual domain of application of the Hilbert transform, ver Property 1 A signal g(1) and its Hilbert transform &{t) have the same amplitude spectrum.82 REPRESENTATION OF SIGNALS AND SYSTEMS ‘Yo prove this property, we observe that the Fourier transtorm of @(0) is equal to —jsgn(f) times the Fourier transforin of g(4), and since the magnitude of ~jsgn(S) is equat to one for at f, then gl) and gl4) will both have the same amplitude specturm, ‘As a corollary to this property, we may state that if a signal g(0) is band limited, then its Hilbert transform g(2) will also be band limited. PROPERTY 2 If lt) is the Hilbert transform of gC), then the Hilbert transform of (0) is — gC). To prove this property, we note that the process of Hilbert transformation is equivalent to passing (0) through a linear ewo-port device with a transfer func. tion equal to ~jsgn{ J). A double Hilbert wansformation is therefore equivalent to passing g(¢) through a cascade of two such devices. The overall transfer func- tion of such a cascade is equal to [-jsgn(QP = ~ 1 forall f ‘The resulting output is thus — g(¢); that is, the Hilbert transform of @(1) is equal to ~ g(t). This result is subject to the requirement that G(O) = Ov where G4) is the Fourier transform of g(t) evaluated at f = 0. PROPERTY 3 A signal g(t) and its Hilbert transform @(1) are orthogonal, To prove this property, we use special case of the multiplication theorem described by Eq, (2.49). In particular, for a signal g(t) multiplied by its Hilbert ansform g(¢) we may write fF ccoae a= fener a (2129) Using Bq. (2.128) in (2.129), we get i BOM dt = if, sgn FGI GL-S) af = if, sem(L)GS)C*S) af (2.130) = i sen F) JEL at where, in the second line, we have used the fact that for a realvalued signal, G(-f) = GCS), The integrand in the righthand side of Bq, (2.130) is an odd function of f, being the product of the odd function sgn(f) and the even func- sion [GCP Hence, the imegrat is zer0, yielding the final resutems peal sof ind ent (oy ert 0) [ | | 2a 241 PRE-ENVELOPE 83 (2.381) J BD de This shows that an energy signal g(2) and its Hilbert transform g(¥) are orthiog- onal over che entire interval (~©, 2), Similarly, we may show that a power signal (0) and its Hilbért transform (4) are orthogonal over one period, as shown by itr fim ws HDR at = 0 (2.132) ‘The validity of Bq. (2.192) is readily demonstrated by the results presented in Example 13 on the Hilbert transforms of sinusoidal functions. PRE-ENVELOPE. Consider a realvalued signal g(), We define the freenaelope of the signal git) as the complex-valued function BO = git) + QO (2.138) where g(J) is the Hilbert cransform of g(¢). We note that the given signal g() is the real part of the pre-envelope g, (2), and the Hilbert transform of the signal is the imaginary part of the pre-envelope. just as the use of phasors simplifies manipulations of alternating currents and voltages, so we find that the pre- envelope is particularly useful in handling band-pass signals and systems. The reason for the name “pre-envelope” is explained later in Section 2.12, One of the important features of the preccnvelope g, (6) is the behavior of its Fourier transform. Let G, (f) denote the Fourier transform of g, (1). Then, wwe may write GS) = GA) + jl-jrgn MGS) Using the definition of the signum fanetion sgn(f) given in Eq. (2.127), we readily find that 2th), F>0 GAS) =4CM, fo (2.184) 0, seo where G(0) is the vale of G(s) at frequency f = 0, This means that the pre- envelope of a signal has no frequency content (ie. its Fourier transform van- ishes) for all negative frequencies, as illustrated in Fig, 2.990 for the case of a low-pass signal. Note that the use of tiangular spectrum for a low-pass signal in Fig. 2.29a is intended only for the purpose of illustration. From the foregoing analysis itis apparent that for a given signal g(t) we may determine its preenvelope g,. (2) in ane of two equivalent ways: 1. We determine the Hilbert transform g(4) of the signal g(Q), and then use Eq. (2.198) to compute the preenvelope £, ().REPRESENTATION OF SIGNALS AND SYSTEMS son 10) Ietny 2010) + w o Figure 2.29 (2) Amplitude spec- ‘trum of low-pass signal g(t). (b) ‘Amplitude spectrum of pre-enve- lope g (t. 2, We determine the Fourier wansfaren G(,') ofthe signal g(0), use Eq, (2-134) to determine G, (J), and then evaluate the inverse Fourier transforin of GCP) to obtain | Bi) = of CNexp enh) af (2.135) | For a particular signal g(1) of Fourier transform G(/), one-way may be better } than the other. For the purpose of illustration in Fig. 2.29, we have used a low-pass signal swith its spectrum limited to the band ~W-= = Wand centered at the origin. i Nevertheless, it should be emphasized that the pre-envelope can be defined for any signal, be it low-pass or band-pass, so long as it possesses a spectrum. Equation (2.438) defines the preenvelope g, (1) for positive frequencies ‘Symmeutically, we may define the pre-envelope for negative frepuenties A = eh) ~ 58Ct) (2.136) “The evo pre-envelopes g, (4) and g_ (1) are simply the complex conjugate of each other, as shown by c= gO (2.137) ‘The spectsum of the pre-envelope g, (0) is norizero only for positive frequencies, as emphasized in Eq. (2.134); hence, the use of & plus sign as the subscript. In contrast, the spectrum of the other pre-envelope g_ () is nonzero only for negaws 22 7) of 15) ter nal for ies. 36) of s7) cs, In gan CANONICAL REPRESENTATIONS OF BANO-PASS SIGNALS 385 tive frequencies, as shown by the Fourier transform , sro GAN) = 460), f=0 (2.138) 2G), F<0 ‘Thus, the pre-envelopes g. (t) and g_(#) constitute a complementary pair of complex-valued signals. Note also that the sum of g, (0) and g. (J) is exactly twice the original signal g(1). CANONICAL REPRESENTATIONS OF BAND-PASS SIGNALS We say chat a signal ¢(¢) is 2 band pars signalif its Fourier transform G(S) is non- negligible only in a band of frequencies of total extent 2W, say, centered about some frequency +/,. This is illustrated in Fig. 2.302. We refer to fas the carrier ‘fiequency. In the majority of communication signals, we find that the bandwidth ‘2Wis small compared with J,, and so we refer to such a signal as a narew-band signal. However, a precise statement about how small the bandwidth must be in order for the signal to be considered narrow-band is not necessary for Our present discussion, Let the presenvelope of a narrow-band signal g(t), with its Fourier transform G(S) centered about some frequency =f, be expressed in the form BO = KOexp(Pahy (2.199) We refer to g(¢) as the complee envelope of the signal. Equation (2.139) may be viewed as the basis of a definition for the complex envelope g(t) in terms of the preenvelope g, (1). We note that the spectrum of g, (t) is limited to the fre- quency band f. - W= f = f, + W, as illustrated in Fig, 2.306. Therefore, applying the frequeneyshifting property of the Fourier transform to Eq, (2.199), we find that the spectrum of the complex envelope (1) is limited to the band - Wf = Wand centered at the origin as illustrated in Fig, 2.30¢. That is, the complex envelope g(t) of a band-pass signal g(1) is a low-pass signal, which is an important result, By definition, the given signal g(t) is the real part of the pre-envelope g, (0). We may thus express the original band-pass signal g(¢) in terms of the complex. envelope gi) as folfows: elt) = Rel glsexpl j2afs)) (2.340) Jn general, GU) isa complex-valued quantity; to emphasize this property, we may express it in the form BY) = a) + Feats) (24) where g(t) and go(2) are both real-valued low-pass functions; their low-pass prop- enty is inherited from the complex envelope g(#). We may therefore use Eqs.86 REPRESENTATION OF SIGNALS AND SYSTEMS rey 16,91 preven | © Figure 2.30 (a) Amplitude spectrum of band- pass signal g(t. (b) Amplitude spectrum of pre-envelope g , (t). () Amplitude spectrum of complex envelope Gt), (2.140) and (2.141) to express the original band-pass signal g(0) in the canonical Jorm: LO = Blt)cos(2nfd) ~ golt)sin 2h) (2.142) We refer to gy(0) as the in-phase component of the bandspass signal g(#) and to Belt) 8 the quadrature component of whe sigual, his womenclature recognizes that2.12 CANONICAL REPRESENTATIONS OF BAND-PASS SIGNALS 87 \ eet iystems, \ @ ” | ttn 7 _ | ° { Figure 2.31 Illustrating on interpretation of the complex envelope g(t) and its mul- tiplication by exp (j2f.t). i | spect to cos(2nf,) (ie, the multiplying factor of g,(!)}. i According to Eq, (2.141), the complex envelope g(t) may be pictured as a i time-varying phasor positioned at the origin of the ggeplane, as indicated in Fig. 2.314. With time varying, the end of the phasor moves about in the plane. Figure 2.315 shows the phasor representation of ‘the complex. exponential exp(j2n/.1). In the definition given in Eq. (2.140), the complex envelope ¢(") is multiplied by the complex exponential exp(2n/,t). The angles of these two phasors therefore add and their lengths multiply, as shown in Fig. 2.314 More- over, in this latter figure, we show the ggg plane rotating with an angular velocity equal to 2f, radians per second. Thus, in the picture portrayed here, the phasor representing the complex envelope g(") moves in the gjgerplane and at the same time the plane itself rotates about the origin. The original band-pass signal g(t) | is the projection of this time-varying phasor on a fixed fine representing the real axis, as indicated in Fig, 2316 Both g(t) and go(t) are low-pass signals limited to the band ~W== f= Hence, except for scaling factors, they may be derived from the band-pass signal (0) using the scheme shown in Fig. 2.924, where both low-pass filters are iden- tical, each having a bandwidth equal to W (see Problem 2.29). To reconstruct (0) from its in-phase and quadrature components, we may use the scheme shown in Fig. 2.920 wnical ( 142) sthat | |REPRESENTATION OF SIGNALS AND SYSTEMS ‘ogame fie join seinstor somp p= fel? a0 -80" phase Figure 2.32 (a) Scheme for deriving the in-phase and quadra- ture components of a band-pass signal. (b) Scheme for recon- structing the band-pass signal from its in-phase and quadrature components, ‘The two schemes shown in Fig. 2.92 are basic to the study of linear modulation systens. The multiplication of the: low-pass in-phase component g(t) by cos(2af.t) and the multiplication of the low-pass quadrature component gol!) by sin(2af.t) represent linear forms of modulation. Given that the Carrier fre- quency J, is sufficiently large, the resulting band-pass function g(t) defined in Eq, (2.142) is referred to as a passband signaling waveform. Correspondingly, the mapping from g,(1) and go(t) into g(0) is known as passband modulation. Equation (2.141) is the “Cartesian” form of expressing the complex enve- lope g(). Alternatively, we may express it in the “polar” form &() = alspexpl jo(O1 (2.143) where a(¢) and $(4) are both real-valued low-pass func . Based on this polar> systems ‘modulation © go) by ment go(t) carrier fre- defined in dingly, the »plex enve- (2.148) 2.12 CANONICAL REPRESENTATIONS OF BANDPASS SIGNALS epresentation, the original band-pass signal g(t) is defined by Bt) > altheosl2aft + HO] (2189) We refer ta a(t) as the natural envelope or simply the envelope of the band-pass signal g(¢) and to (0) as the phase of the signal. Equation (2.144) represents a ‘ybrid form of amplitude modulation and angle modulation; indeed, ic includes am- plitude modulation, frequency modulation, and phase modulation as special From this discussion it is apparent that whether we represent a band-pass (roodutated) signal g(i) in terms ofits insphase and quadrature components as in Eg, (2.148) or ia cerms of its envelope and phase as in Eq, (2.144), the infor- mation content of the signal g(t) is completely represented by the complex envelope g(e). The particular virtue of using the complex envelope g(2) co rep- resent the band-pass signal is an analytical one, and will become evident in Sec tion 2.23 Terminology ‘The distinctions among the three different envelopes that We have introduced to describe a band-pass signal g(¢) should be carefully noted. We may summarize their definitions as follows: 1. The preeavelope g, (#) for positive frequencies is defined by Bo(O = at) 45D where g(t) is the Hilbere eransform of the signal g(i). According to this representation, £4) may be viewed as the quadrature function of g(0). Cor respondingly, in the frequency domain we have 2a), foo GS) = 460), f=0 OF f0 G_(f) = 4 G0), f=0 26), F< 2 The complex envelope g(t) equals a frequencyshified version of the pre- envelope gy (1), as shown by RO > gy Mexp(— 72h) where f, is the carrier frequency of the band-pass signal g(¢) 3. The envelope a(1) equals the magnitude of the complex envelope (1) and also that of the preenvelope g. (1), as shawn by a= 120) = le col90 REPRESENTATION OF SIGNALS AND SYSTEMS Note that for a band-pass signal g(4), the pre-envelope g(t) sa complex band- pass signal whose value depends on the carrier frequency f. On the other hand, the envelope a(t) is always a real low-pass signal and, in general, the compiex envelope g(2) is a complex low-pass signal; the values ofthe latter «wo envelopes are independent of the choice of the carrier frequency f. Jn Chapter 3 itis shown that the signal a(t) results from envelope detection (ces rectification and low-pass filtering) of the band-pass signal g(t). For this, reason we call a(t) the erivelope of g(t), and call the complex signals g(!) and &+(0) the complex envelope and pre-envelope of g(), respectively. The envelope a(¢) and phase (0) of g(t) are related to its quadrature com Ponents g(t) and go() as follows (see the time-varying phasor representation ofFig. 21a): nse . ay = VEO + EO b() = tan Conversely, we may write EX) = ale) costawy) gol!) = att) sin 60] Thus, each of the quadrature components of a band-pass signal contains both amplitude and phase information. Both components are required for a unique definition of the phase $(2), modulo 2a, EXAMPLE 14 RF Pulse (continued) Suppose we wish to determine the different envelopes of the RF pulse defined N e() = Avec) costes We assume that J. >> 1, so that the RF pulse g(#) may be considered narrove sand. From Example 5, we recall that the Fourier transform of g(t) is given by sine (Ff ~ f)], f>0 s=0 0, at 2 sine TF + J, F< 0 (v sine( TUF - f)], £0 0. sso 213 243rier local oscillator phase drifts from its proper value by a small angle ¢ radians. The | The Zchannel output will remain essentially unchanged, but there will now be some i g the signal appearing at the Qchannel output, which is proportional to sing = ¢ for small 4. This Q.channe! output will have the same polarity as the Fchannel out- put for one direction of local oscillator phase drift and opposite polarity for the ‘opposite direction of local oscillator phase drift, Thus, by combining the # and channel outputs in a phase discriminator (which consists of a multiplier followed, i ATION 3.4 DOUBLE SIDEBANO-SUPPRESSED CARRIER MODULATION 139 | | © for by a low-pass filter), as shown in Fig, 3.15, a de control signal is obtained that This automatically corrects for local phase errors in the voltage-contvallad oscillator. j an Tis apparent that phase control in the Costas receiver ceases with modula local tion and chat phase-lock has to be reestablished with the reappearance of mod- \ The ulation. This is not a serious problem when receiving voice transmission, because i ue the lockup process normally occurs so rapidly that no distortion is pereepubte (Fi path ver Multiplex if ide. ‘Quadrature-Carrier Multiplexing (| | such “The quadrature nul effect ofthe coherent detector may also be put to good use {6 in the consiruetion of the so-called quadratarecanter muliplexing OF quadrature ia lator { Amplitude madalaion (QAM). This scheme enables two DSB-SC modulated waves | the | (resulting from the application of two physically independent message signals) to f out ‘occupy the same channel bandwidth, and yeti allows for the separaon of the i tthe Shome ‘A block diagram of the quadraturecarrier multiplexing system is shown in Fig. 3.16. The transmitter part of the system, shown in Fig, 9.16a, involves the use of two separate product mpdulators that are supplied with two cartier waves { of the same frequency but differing in phase by ~90 degrees. The transmitted i] signal s(t) consist ofthe sum of these two product modulator outputs, as shown | ” } s(t) = Agmy(‘eos(2afe} + A.any(t)sin(2af,t) (3.20) where m,(t) and m(2) denote the two different message signals applied to the | the carrier frequency J, where Wis the message bandwidth of m,(t) oF m(t) According to Eg, (3.20), we may view A,n,(t) as the in-phase component of the I multiplexed band-pass signal s(¢) and ~A,mg(t) as its quadrature component. | ‘The receiver part ofthe system is shown in Fig. 3.165, The multiplexed signal s(t) is applied simultaneously to two separate coherent detectors that are sup- i plied with two local carriers of the same frequency, but differing in phase by j | ~90 legrees. The output of the top detector is 41m), whereas the output of the botiom detector is $4,ma(t). For the system to operate satisfactorily, itis important to maintain the correct phase and frequency relationships between | the local oscillators used in the transmitter and receiver parts of the system. ‘To maintain this synchronization, we may use a Costas receiver described above. Another commonly used method is to send a pilot signal outside the pass- . band of the modulated signal, In the latter method, the pilot signal typically | 1 consists of a low-power sinusoidal tone whose frequency and phase are related | to the carrier wave e(); at the receiver, the pilot signal is extracted by means of \ | a suitably tuned circuit and then translated to the correct frequency for use in | the coherent detector.CONTINUOUS-WAVE MODULATION Durtpened we Product OAc coat) ange Le Niconi2afen Mutiplexes ‘ig ot 90" a Figure 3.16 Quadrature-carrier multiplexing system. (a) Trans- miter. (0) Receiver. FILTERING OF SIDEBANDS ‘The next iasme we wish 10 discuss is how to process a DSBSC modulated signal 50 a5 to generate a vestigial sideband (VSB) or single sideband (SSB) modulated signal. An intuitively satisfying method of achieving this requirement is the fre ‘quency discrimination mathod that involves the use af an appropriate filter following, 4 product modulator responsible for the generation of the DSB-SC modulated signal. Natorally, the exact specification of the filter depends on the desired type’ of modulation. Consider then the circuit described in Fig. 8.17¢, where u(t) denotes the product modulator ourput, as shown by at) = Aam(ayeos2af,t)signal ulated he fre lowing ulated d ype" es the a5 FILTERING OF SIDEBANDS sy wean [nf aga | __ Moma mt) + Hg st) o Figure 3.17. (a) Filtering scheme for processing sidebands. (6) Coherent detector for recovering the message signal, Let Hf) denote the transfer function of the filter following the product mod- ulator. The spectrum of the modulated signal s(t) produced by passing u(t) through the filter is given by SU) = USA) way fear ~ fy + m+ ven where M(f) is the Fourier transform of the message signal m(t). The problem we wish to address is t6 determine the particular Af) required (0 produce a ‘modulated signal s(t) with desired spectral characteristics, such that the original message signal m(t) may be recovered from s(t) by coherent detection, The first step in the coherent detection process involves multiplying the modulated signal s(2) by a locally generated sinusoidal wave A!cos(2n/t), which is synchronous with the carrier wave 4,cos(2a/,1), in both frequency and phase as in Fig. 9.176. We may thus write v(t) = Arcos(2mrft)s(¢) ‘Transforming this relation into the frequency domain gives the Fourier trans- form of u(t) as A VS) = GIST — Sf) + SUF + LN (3.22) Therefore, substitution of Eq, (8:21) in (8.22) yields WS) = SALMA US - AFH +N (3.28) + Adin s ~ aH ~ fy + MU + ME + 11an CONTINUOUS-WAVE MODULATION The high eequency components of u(t) represented by the second term sn bq. (8.28) are removed by the low-pass filter in Fig. 8.17610 produce an output v(t), the spectrum of which is given by the remaining components: BAacsycens = f) + HS + 21 0.24) vat) For a distortionless reproduction of the original baseband sighal m0) at the coherent detector output, we require ¥,(f) to be a scaled version of M(/).(This, means, therefore, that the transfer function H(S) must satily the condition ACL ~ 4) + HS © Sy = 2S (9.25) where H(J), the value of H(J) at J = J,, is 4 constant. When the mestage (baseband) spectrum M(f) is zero outside the frequency Fange ~ Wf = W, we need only satisfy Eq. (8.25) for values of J in this interval, Also, ea simplify the exposition, we set HU) = 1/2. We thus require that H(f) satisfies the condition: HU - f+ HS +Qel -Wsfsw (3.26) ‘Thereisa great deal of flexibility mn the selection of HL) to satisy this condition, as discussed later in Sections 3.6 and 8.7. In any event, under the condition described in Eq, (3:28), we find from Eq, (324) that the coherent detector output in Fig. 3.17bis given by 20 = Aen 2 Equation (3:21) defines the spectrum of the modulated signal s(t). Recog- nizing that s(¢) is a band-pass signal, we may formulate its time-domain descrip- tion in terms of in-phase and quadrature components, using the passband sig- naling method described in Section 2.12. In particular, s(t) may be expressed in the canonical form May = sAt)costem.t) — solt)sin(2n,0) (3.28) where s,(t) is the in-phase component of s(2), and so(¢) is its quadrature com- ponent, To determine 50), we note that its Fourier transform is related to the Fourier transform of s(1) as follows (see Problem 2.29): SL =f) + 9S + -we sun = (0 Dr ss + hy WSTSW 5.0) elsewhere Hence, substituting Eq. (8.22) in (3.29), we find that the Fourier ansform of (0) is given by SK = AMS LHS ~ f) + HU + LY) (3.80) PAM, Ws fs Ww35 1a FILTERING OF SIDEBANDS where, in the second fine, we have made use of the condition ia Bq. (8.26) a oa imposed op H(J). From Eq, (3:30) we readily see thatthe in-phase component of the modulated signal si) is defined by t B24) sft) = ghante) (3.31) w the which, except for a scaling factor, is the same as the original message signal m() ‘This ‘To determine the quadrature component s9() of the modulated signal s(1), om wwe recognize that its Fourier transform is defined in terms of the Fourier want- form of s(t) as follows (sce Problem 2,29) 3.25) (Sf -f) - f+ fl, -wsf=sw sah) = {? (3.82) , elsewhere | oat “Therefore, substi, (8.21) in (3.32), we get if SoS) = SAMA UALS = J) — HS + S91 (3.33) 7 | i This equation suggests that we may generate o(t), except for a scaling factor, by passing the message signal m(1) through a new filter whose transfer function 'y is related to that of the filter in Fig, 3.172 as follows: 4 ulated wave s(2). There are two important points to note here: Hef) = {HS =f) - WS + Dl -watsw (334) | 527) Let m'(2) denote the output of this filler produced in response to the input m(¢) ‘ Hence, we may express the quadrature component of the modulated signal s(t) ' cog. as 1 { crip. ada te 4 salt) = 5am) (3.35) \ din i 5 Accordingly, substituting Eqs. (3.31) and (3.35) in (3,28), we find that s(t) | 28) may be written in the canonical form. j 4 om. 1 / 1, | the m(t) = FAan(theost@mhe) ~ 3.Aon' (sin mio (8.36) | Equation (8.36), except for the scaling factor 1/2, suggests the circuit of Fig. 5) 18 as the basis ofe phase dlscrimination mutbd fs the generation of the mod \ | 1. The inphase component 540) is completely independent of the transfer of function H1(f) of the band-pass filter involved in the generation of the mad- ulated wave s(¢) in Fig. 3.174, so long as it satisfies the condition of Eq. (3.26). 2. The spectra) modification attributed to the transfer function H(f) is con- fined solely to the quadrature component s9() 0) ‘The-role of the quadrature component is merely to interfere with the in-phase } component, s0 as to reduce or eliminate power in one of the sidebands of the i modulated signal s(1), depending on the application of interest3.6 CONTINUOUS-WAVE MODULATION osuating row vee odalooe 0) Avo (nee te iter ~. ep Shier in 20h) mig] roast >| moscltor Figure 3.18 Block diagram of phase discrimination method for processing sidebands. ‘The discussion presented thus far has been of a fairly general nature, In the next two sections, we consider the generation of VSB and SSB modulated signals 1s special cases of the modulated signal s(¢) defined in Eq. (3.36) VESTIGIAL SIDEBAND MODULATION’ Assuming that the requirement is to generate a vestigial sideband (VSB) modu- lated signal containing a vestige of tho lower sideband, we find that Eq, (3.26) is satisfied by using a band-pass filter whose transfer function H(f) is as shown in Fig, 3.1; to simplify matters, only the response for positive frequencies is mn ° ih hth hw Figure 3.19 Amplitude response of VSB filter; only Positive-frequency portion is shown.MATION 36 VESTIGIAL SIDEBAND MODULATION 145 Figure3.20_ Frequency response of filter for producing Rd the quadrature component of the VSB wave. | shown here. This frequency response is normalized, so that |H()|is one half at the carrier frequency J, The important feature to note, however, is that the cutoff portion of the frequency response around the carrier frequency J, exhibits odd symmetry That is to say, inside the transition interval f, ~ = [fis J, + f,the sum of the values of |#I(f)] at any two frequencies equally displaced above and below fis unity; fis the width of che vestigial sideband. Note also that outside ‘nthe the frequency band of interest (ie. |S] >, + W), the transfer function H(S) ignals may have an arbitrary specification, “The corresponding frequency response of the filter producing the quadra- ture component of the VSB modulated signal in accordance with Eq, (3.34) is as shown in Fig. 3.20. Figures 8.19 and 3.20 apply to a VSB modulated signal containing a vestige of the lower sideband. For a VSB modulated signal containing a vestige of the vodu- ‘upper sideband, we have similar eesults except for the following differences: The 3.26) "upper cutoff portion of H(f) is controlled to exhibit odd symmetry around the i town carrier frequency f, whereas the lower cutoff portion is arbitrary, This has the | ies is effect of replacing the minus sign at the summing junction at the output of Fig. | i 8.18 with a plus sign. «Television Signals A discussion of vestigial sideband modulation would be incomplete without a mention of its role in commercial television (TV) broadcasting.’ The exact de- sails of the modulation format used to transmit the video signal characterizing a TV system are influenced by two factors: 1. The video signal exhibits a large bandwidth and significant low-frequency content, which suggest the use of vestigial sideband modulation, 2. The circuiuy used for demodulation in the receiver should be simple and therefore cheap; this suggests the use of envelope detection, which requires the addition of a cartier to the VSB modulated wave. ‘With regard to point 1, however, it should be stressed that although there is indeed a basic desire to conserve bandwidth, in commercial TV broadcasting theCONTINUOUS-WAVE MODULATION fine) | ° tee oe te re srg Normalized response | 1 { t t Hl i ne) ue 36 Co) |< ~ channet banawiath —>] MH a) Figure 3.21 (a) idealized amplitude spectrum of a transmit- ted TV signal. (b) Amplitude response of VSB shaping filter in the receiver. transmitted signal is not quite VSB modulated, The reason is that at the trans suitter the power levels are high, with the result that it would be expensive 10 rigidly control the filtering of sidebands, Instead, a VSB filter is inserted in each receiver where the power levels are low. The overall performance is the same as conventional vestigial-sideband modulation, except for some wasted power and bandwidth. These remarks are ilhistrated in Fig. 8.2). In particular, Fig. 321¢ shows the idealized spectrum of a transmitted TV signal. The upper sideband, 25 percent of the lower sideband, and the picture cartier are transmitted. The Frequency response of the VSB filter used ta do the required spectrum shaping in the receiver is shown in Fig. 3.21 ‘The channel bandwidth used for TV broadcasting in North America is 6 MHz, as indicated in Fig. 3.216. This channel bandwidth not only accommodates the bandwidth requirement of the VSB modulated video signal but also provides for the accompanying sound signal that modulates 4 eatrier ofits own, The values presented on the frequency axis in Figs. 3.214 and 3.214 pertain to a specific TV channel. According to this figure, the picture carrier frequency is at 55.25 MH2,>BULATION he trans- ensive t0 fin each ideband, ted. The shaping rica is 6 modates provides ne values 37 37. SINGLE SiDEBANO MODULATION 107 and the sound carrier frequency is at 59.75 MHz, Note, however, that the infor- mation content of the TV signal lies in a baseband spectrum extending from 1.25 MHz below the picture cartier to 4,5 MHz above it With regard to point 2, the use of envelope detection (applied to a VSB modulated wave plis carrier) produces waveform distortion in the video signal recovered at the detector output. The distortion is produced by the quadrature component of the VSB modulated wave; this issue is discussed! next, Waveform Distortion ‘The use of the imedomain description given in Eq, (8.36) enables the deter- mination of the waveform distortion caused by the envelope detector. Specifi- cally, adding the carrier component 4, cos(27/,¢ © Eq, (8.36), the tater being scaled by a factor h,, modifies the modulated signal applied to the envelope detector input as «eo alr + Seam | cos(2af,2) -} AyAcnt (sin (Qad.t) (9.97) where the constant &, determines the percentage modulation. The envelope detector output, denoted by a(¢), is therefore 1 jz i / j2) 2 (3.38) Ly dante 14 sham) 1+|— 1+ Zhe) Equation (3.38) indicates that the distortion is contributed by m’(e), which is responsible for the quadrature component of the incoming VSB modulated sig- nal. This distortion can be reduced by using two methods: + By reducing the percentage modulation to reduce the amplitude sensitivity hy + By increasing the width of the vestigial sideband to reduce m' (0) . Both methods are in fact used in practice. In commercial TV broadcasting, the width of the vestigial sideband (which is about 0.75 MHz or one-sixth of a fall sideband) is determined to keep the distortion due to m'(1) within .olerable limits when the percentage modulation is nearly 100. SINGLE SIDEBAND MODULATION Consider next the generation of a SSB modulated signal containing the upper sideband only. From a practical viewpoint, the most severe requirement of SSB generation usually arises from the unwanted sideband, the nearest frequency component of which is separated from the desired sideband by twice the lowest148 CONTINUQUS-WAVE MOOULATION Th Oh h 7 lo Eoerey we Figure 3.22. Spectrum of a message signal mit) with an energy gap centered around the origin frequency component of the message (tmodulating) signal. The implication here is that for the generation of an SSB modulated signal to be possible, the niessage spectrum must have an energy gap centered at the origin, as illustrated in Fig. 8.22, This requirement is naturally satisfied by voice signals, whose energy gap is about 600 Hz wide (ie., it extends from — 300 10 +800 Ha). We may then, generate an SSB signal containing the upper sideband, say, by using a band-pass filter, the frequency response of which is ideally related to the carrier frequency fas shown in Fig, 3.234, Thus, given the message spectrum defined in Fig. 3.22, Joh 0 Theth teh m1 o Figure 3.23 (a) Wdeallzed frequency response of band-pass fier. (b) Spectrum of SSB signal containing the upper sideband.“ATION Phere ossage pig. gapis then ‘-pass uency 3.22, n of 3.7 SINGLE SIDEBAND MODULATION 149 Prodet Ay 08 nfl A, co nf) Figure 3.24 Block diagram of a two-stage $SB modulator. we Bnd that the corresponding spectrum of the SSB signal is as shown in Fig. 3.236, In designing the band-pass filter in Fig. $.17a for the generation of an SSB modulated wave, we must satisfy three basic requirements: + The desired sideband lies inside the passband of the filter. + Tp unwanted sideband lies inside the stopband of the filter. + The filter's transition band, separating the passband from eke scopband, is twige the lowest frequency component of the message signal ‘This'kind of frequency discrimination usually requires the use of highly selective filters, which can only be realized in practice by means of crystal resonators.® When itis necessary to generate an SSB signal occupying x frequency band. that is much higher than that of the baseband signal (e.g., tanslating a voice signal to the high-frequency region of the radio spectrum}, i¢ becomes very dif ficult to design an appropriate filter that will pass the desired sideband and reject the other, using che simple arrangement of Fig. 9.17. tn such a skeration itis necessary to resort to a multiplemodulation process so as to ease the filtering requirement. This approach, involving two stages of modulation, is iDhustrated in Fig. 3.24. The SSB signal 5,(1) atthe first banchpass filter output is used as the ating wave for the second product modulator, which produces a DSB-SG signal with a spectrum that is symmetrically spaced about the second carrier frequency J. The frequency separation between the sidebands of this DSB-SC signal is effectively twice the first carrier frequency f,, thereby making it relatively easy to remove the unwanted sideband by the second band-pass filter, and thereby generate the SSB signal (1). ime-Domain Description of SSB Modulated Signal The issue to be considered next isthe time-domain description of an SSB mod- ulated signal s(4). Given the idealized frequency response of Fig. 3.284 as the dlescription of H(J), we find from Eq, (3:34) thatthe corresponding description of Half) responsible for the generation of the-quadrature component sg(t) is as shown in Fig, 8.25. Using the definition of the signam function sga(/), we readily see from Fig. 3.25 that ih) = ~jsan(f (3.39) which is recognized as the transfer function of the Hilbert transformer (see Section 2,10). In other words, the m'(2) in Eg, (3.36) is exactly the Hilbert rans. form of the original message signal m(1), Thus, using s(t) to denote this Hilbert transform, we may forgully describe an SSB modulated signal comaining onlyCONTINUOUS-WAVE MODULATION Figure 3.25 Frequency response of quadrature filter for single sideband modulation containing the upper sideband, the upper sideband as follows: . 0 Shami) cos(2ah.t) — RA) sin( 22) (340) ‘This means that the filter responsible for the generation of the quadrature com ponent of s(¢) in the modulator of Fig. 3.18 consists of a Hilbert transformer. Also, if we want to generate an SSR ruadulated signal containing only the lower sideband, we simply replace the minus sign at the summing junction at the our put of Fig. 3.18 with a plus sign, An SSB modulator based om the scheme shown in Fig. 3.18 in the manner described here is called the Hartley modulator. Demodulation of SSB Signals ‘The demodulated signal v,(¢) defined in Bq. (3:27) assumes perfect synchronism, beqween the ascillator in the coherent detector of Fig. 3.17b and the oscillator in the modulator of Fig. 3.174, both in phase and in frequency. This ean be provided by one of two methods « Transmitting a low-power pilot carrier in addition to the selected sideband, or + Using, in the receiver, a highly stable oscillator « as the carrier frequency. fed to the same frequency In the latter method, it is inevitable that there would be some phase error‘ he local oscillator signal with respect to the carrier wave used to generate the incoming SSB modulated wave s(t). Denoting the local oscillator signal by Alcos(2mf.t + $), we find that (after some straightforward steps) the resulting demodulated signal is given by (for the case when the upper sideband only ig, transmitted) v(é) = Faaitnco) cosh + s(t) sind} (3.41)Ton 40) be 3.8 FREQUENCY TRANSLATION 151 3.8 Unlike the idealized coherent detection process described in Eq. (3.27), the demodulated signal v,(2) of Eq. (3:41) contains an unwanted component pro- portional to 1(@) sing, which cannot be removed by filtering. This unwanted component appears 28 a phase distortion. To show this, we take the Fourier transform of v4) in Eq. (8.41) co obtain vif) = AALS) cosh + Mf) sind) 3.42) ‘But fcom the definition of the Hilbert transform si(), we know that the Fourier transform of si) is related to the Fourier transform of the original signal m(¢) by MS) = =j sgn PMCS) (3.43) ‘Therefore, substituting Eq. (3.43) in (3.42) and then simplifying, we get AIM Aexp(—i0)s fro Vf) = (3.44) PAM Dexp(jo), F< 0 Thus the phase error @ in the local oscillator output results in pase distortion, where each frequeney component of the original message signal m(1) undergoes a constant phase shift 6 in the course of demodulation at the receiver. This phase distortion may be tolerated with voice communications, because the human ear is relatively insensitive to phase distortion; the presence of phase distortion gives rise to a Donald Duck voice effect. In the transmission of music and video signals, ‘on the other hand, the presence of phase distortion in the form of a constant phase difference is utterly unacceptable. FREQUENCY TRANSLATION ‘The basic operation involved in single sideband modulation is in fact a form of frequtncy translation, which is why single sideband modulation is sometimes re- ferred to as frequency changing, mixing, or helerodyning, This operation is clearly illustrated in the spcetrum of the signal shown in Fig. 3.23 compared to'that of the original message signal in Fig. 3.22. Specifically, we see that a message spe trum occupying the band from f, to J, for positive frequencies in Fig. 3.22 is shifted upward by an amount equal to the carrier frequency fin Fig. 8.28% the message spectrum for negative frequencies is translated downward in a symme- tic fashion, The idea of frequency translation described herein may be generalized as follows. Suppose that we have a modulated wave s(t) whose spectrum is centered on a carrier frequency f,, and the requirement is to translate it upward in fre- quency such that its carrier frequency is changed from f, to a new value f. This requirement may be accomplished using the mixer shown in Fig. 3.26, which is similar to the scheme of Fig. 3.174, Specifically, the mixeris a device that consists