90 ANALYSIS AND TRANSMISSION OF SIGNALS To find the Fourier transform of this pulse, we differentia it successively, as shown in Fig. 3.23b and c. The second derivative consists of a sequence of impulses (Fig, 3.230). Recall thatthe derivative of signal at jump discontinuity san impulse of strength equal to the amount of jump. The function de(#)/dr has a positive jamp of 2/t at r= £12, EH 2 [5 (04.2) -289 44(0-3)] 50) From the time differentiation property [Eq. (3.49)|, ay 5 Be PPP GI) = ~2Af PG) (B5la) Also, from the time-shitting property [Eqs. (3.32) Bet = tp) a eH 351b) ‘Taking the Fourier transform of Eq, (3.50) and using the results in Eg. (3.51), we obtain (ang Pay) = 5 (oF 240") Steosafe—l) and 8 ga MEE\ _ efsinlar iP _ to (mit > 8 = ware (F) 3 [ SP Peyae(H) ex The spectrum Gif) is shown in Fig, 323d. This procedure of finding the Fourier transform can be applied to any function g(#) made up of straight-line segments with @(2) > O.as lil — 00. The second derivative of such a signal yields a sequence of impulses whose Fourier transform can be found by inspection. This example suggests a numerical method of finding the Fourier transform of an arbitrary signal g(s) by approximating the signal by slraight-line segments, ‘To provide easy reference, several important properties of Fourier transform are summa- rized in Table 3.2. 3.4 SIGNAL TRANSMISSION THROUGH A LINEAR SYSTEM ‘Alincar time-variant (LTD continuous time system can be characterized equally well in either the time domain or the frequency domain. The LTT system model, itlustrated in Fig. 3.24, ‘can often be used to characterize communication channels. In commmanication systems aad in signal processing, we are interested only in bounded-input-bounded-output (BIBO) stable linear systems, Detailed discussions on system stability ean be found inthe textbook by Lathi>Figure 2.25 Signa through inecr oyeon. ‘3.4 Signal Tronsmission Through a Linear System 9¥ TABLE 3.2 Properties of Fouriar Transform Operations Operation 0 en Superpoiion HOFRM GNF Seatac emtiplication ea) 160) Doality Gey ea) ‘Time seating wat) we (6) “Time shitting atte) Geen Frequency shitting — gine" GF fp) Tine convoiuion gy 210 GUNG Frequency convolution gia) Grif) Gag) Tinedimereniaion TEE yaepyreyy Time imteyration foogstarte $B + sour Topas upset wedoowin x9 FETT gsm | yuny= alent is Feqencydoman x) LM J vane nya xin A stable ITI system can be characterized inthe time domain by its impulse response h(t), which isthe system response to a unit impulse input, that is, MO = HO) — when x1) = 3G) ‘The system response to a bounded input signal x( follows the convolutional relationship YOO) = hfe) 0) es) ‘The frequency domain telationship between the input and the oulput is obtained by taking Fourier transform af both sides of Eq (3.53). We let 10 = XY) yey es FG) hn) exe HUN) ‘Then according o the convolution theorem, Eq. (3.53) becomes FH =HO-X) es Generally H (f), the Fourier transform of the impulse response h(), is referred to as the transfer function or the frequency response of the LTI system. Again, in general, Hf) is,92 ANALYSIS AND TRANSMISSION OF SIGNALS complex and can be written as AG) = Ae where [2/(f)is the amplitude response and é(f isthe phase response of the LTT system. 3.4.1 Signal Distortion during Transmission ‘The transmission of an input signal x(t) through a system changes it into the output signal y(). Equation (3.54) shows the nature of this change or modification, Here X (f) and Y«f}) are the spectra ofthe input and the output, respectively, Therefore, H (f) isthe spectral response of the system, The output spectrum is given by the input spectrum maltiplied by the spectral response of the system, Equation (3.54) clearly brings out the spectral shaping (or edification) of the signal by the system. Equation (3.54) can be expressed in polar form as WDM = KG AIM OA ‘Therefore, we have the amplitude and phase relationships bp an EO 3.55a) =A FOG) (3.55b) Daring the transmission, the input signal amplitude spectcum [X(f)| is changed to |X (>| HAG). Simiarly, the input signal phase spectrum 0, (/) is changed to Hf) + 6,(/). ‘An input signal spectral component of frequency fis modified in amplitude by a factor HAG)| and is shifted in phase by an angle 64(7). Clearly, |H(f)| is the amplitude response, ancl On(f) is the phase response of the system, The plots of [E(f)| and Ay(f) as functions of F show at a glance how the system modifies the amplitudes and phases of various sinusoidal inputs. This is why #(f) is called the frequency response of the system. During transmission through the system, some frequency components may be boosted in amplitude, while others ‘may be attenuated. The relative phases ofthe various components also change. In general, the ‘output waveform will be different from the input waveform. 3.4.2 Distortionless Transmission In several applications, such as signal amplification or message signal transmission over a communication channel, we cequire the output waveform to be a replica of the input waveform, Insuch eases, we need to minimize the distortion caused by the amplifier or the communication cchanael. Iti therefore of practical interest to determine the characteristics of a system that allows a signal to pass without distortion (distortionless transmission). ‘Transmission is said to be distortionless ifthe input and the output have identical wave shapes within a multiplicative constant. A delayed output that retains the input waveform is alse considered distortionless, Thus, in distortionless transmission, the input x(@) and the output _»(0) satisfy the condition Mt) =k x(a) 8.56) ‘The Fourier transform of this equation yields YQ) = Axe PeFigure 3.25 near limeinwacion! spt hagoecy response fo dietoronlse fransmisson, 3.4 Signol Transmission Through a Linear System 93 Hen But because FP =XPHD), Wwe therefore have ees Hg This is the transfer function required for distortionless transmission, From this equation it {follows that [Hse B.5tay OM) = Daring @5Tb) ‘This shows that for distortionless transmission, the amplitude response |21(F}| must be a ‘constant, and the phase response 64(F) must be a linear function of f going through the origin f = 0, a shown in Fig. 3.25, The slope of (F) with respect to the angular frequency «1 = 2f is ~1g, where fg isthe delay of the output with respect to the input.” All-Pass vs. Distortionless System In circuit analysis and filter designs, we sometimes are mainly concerned with the gain of a system response. An al-pass system has a constant gain forall frequencies (ie. |H?)| = KI, without the linear phase requirement. Note from Eq. (2.57) that a distortionless aystem is always an all-pass system, Whereas the converse is not true. Because itis very common for beginners to be confused by the difference between allpass and distonionless systems, now is the best time to clarify ‘Tosee how anall-pass system may lead to distortion, letus consider an illustrative example, Imagine that we would like to transmit a recorded music signal from a violin-ello duet. The violin contributes to the high frequency par ofthis masic signal, while the cello contributes to the bases part. When tis music signal i transmitted through a particular all-pass system, both Parts have the same gain. However, suppose that this all-pass system would cause a l-second ‘area delay on the high-frequency content of the music (from the violin). As a sesul, the audience on the receiving end will hear a “rmusie” signal thats totally out of syne even though all signal components have the same gain and all are presen. The difference in transmission delay for components of different frequencies is contributed by the nonlinear phase of (7) in the al-pass fier. *Inadahion, we require hat (0) sth be (us sown ie Fig, 225) o¢ hae a constant wae (xan inter, nals, 3yF) =~ aig. The adion a the exces pase of mays mot change he sgn of he sg4 ANALYSIS AND TRANSMISSION OF SIGNALS ‘To be mote precise, the transfer function goin |# (f)| determines the gain of each input frequency component, whereas ZH} determines the delay of each component. Imagine a system input «(@) consisting of multiple sinusais (its spectral components}. For the output signal »(2) wo be distortionles, it should be the input signal multiplied by a gain snd delayed by ty. To synthesize such a signal, y() needs exactly the same components as those of x), with each component multiplied by & and delayed by tq. This means that the system transfer function H7() should be such that each sinusoidal component encounters the same gain {or Joss) & and eal component undergoes the same time delay of ty seconds. The frst condition requires that [HW =e We have seen eatlier (Sec. 33) that o achieve the same time delay ty for every frequency ‘component requires a linear phase delay 27g (Fis, 3.18) thrangh the origin Uf) = —2afg In practice, many systems have a phase characteristic that may be only approximately linear. A couvenient method of checking phase linearity is to plot the shope of ZH) as a function of frequency. This slope can be a function off in the general ease snd is given by aa) ui) (3.58) If the slope of is constant (that is, if 8 is linear with respect to f), all the components are delayed by the same time interval fy. But ifthe slope is not constant, then the time delay 4g varies with frequency. This means that different trequency components underge different amounts of time delay, and consequently the output waveform witl not be a replica of the input waveform (as in the example of the violin-cello duet). For a signal transmission ¢o be. distortionless, 2y(f} should be a constant r, over the frequency band af interest.” ‘Thus, thereisaclear distinction between all-pass and distortionless systems, tis acommon mistake to think that famness of amplitude response | (f)| alone can uarntee signal quality Assystem that has a flat amplitude respotse may yet distor a signal beyond recognition if the phase response is not linear (ty not constant), The Nature of Distortion in Audio and Video Signals Generally speaking, a human ear can readily perceive amplitude distortion, although it is relatively insensitive to phase distortion, For the phase distortion to become noticeable, the * Figne 3.25 shows tht fer distortions ansssion, the pha response nt only ine bt also pas ‘toni e origin This oer equzement cea be seme hat rele foe bapa signals The pha he eg h ray be ay corsan (54/9 = fy ~ 2a 2 (0) — ol Te reason forts cun be toons sa Ba. (3.57) whic shoves da ti ain of onsant phased fo spectuns fa Bandpass igh anon Oa pate ba he amar by oy. The mesangial (he envelope) ao aed. The orp envelope Ue ame a ee ape envelope delayed by called the group delay or envelope delay. and th np ane she sae a the ap cae delayed by ip = tut oop called the phase delay, wher fy ithe cee rogues of he pecan2.5 Idecl versus Proctical Filers 95 ‘variation in delay (variation in the slope of 9) should be comparable to the signal duration (or the physically perceptible duration, in case the signal itself is long). Inthe case of audio signals, ‘each spoken syllable can be considered to be an individual signal. The average duration of a spoken syllable is of a magnitude on the order of 0.01 10 0.1 second, The audio systems may have nonlinear phases. yet no noticeable signal distortion results because in practical audio systems, maximum variation in the slope of 5, is only a small fraction of a millisecond. This is the real reason behind the statement that “the human ear is relatively insensitive to: phase ddistortion.* Asa result, the manufacturers of audio equipment make available only 1H (f)|, the amplitude response charaeteristic of their systems. For video signals, on the other hand, the situation is exactly the opposite. The fuman eye is sensitive to phase distortion but is relatively insensitive (o amplitude distortion, The amplitude distortion in television signals manifests itself as a partial destruction ofthe relative hhalf-tone values of the resulting picture. which is not readily apparent to the human eye, The phase distortion (nonlinear phase), on the other hand, causes different time delays in different picture elements. This results ina sineared picture, which is readily apparent tothe human eye, Phase distortion i also very important in digital commonication systems because the nonlinear phase characteristic of a channel causes pulse dispersion (spreading ovt), which in tum causes, pulses fo interfere with neighboring pulses. This interference ean cause an error in the pulse amplitude atthe receiver: a binary 1 may read 3s 0, and vice versa. 3.5 IDEAL VERSUS PRACTICAL FILTERS Figure 3.26 (Geol low ose Boren response and ts inpuse fesponse, deal filters allow distortionless transmission of a certain band of frequencies and suppress all the remaining frequencies. The ideal low-pass filter (Fig, 3.26), for example, allows all ‘components below f = B Hz to pass without distortion and suppresses all components above f = B. Figure 3.27 shows idea! high-pass and bandpass filter characteristics. The ideal low-pass filter in Fig. 3.268 has a finear pase of slope —fg, which results in a time delay of ry seconds forall its input components of frequencies below B Hz. Therefore, if the input isa signal g(#) band-limited to B Hz, the output 7() is g(i) delayed by 1, that is, vO) = gt) ‘The signal g(¢) is transmitted by this system without distortion, but with time delay 2. For this filter |$'(F)| = TUF 2B), and 0 (F) = —2mfa. so that f apan( Seam (3.59) er) =n @ o