4 Fourier Transform A periodic function is obtained by repeating a signal after some fixed period of time T’ (refer to Fig. 10). We know that such a periodic signal has a Fourier series representation. For a given time function f(t), we use the notation f(t) to show the periodic function obtained by repeating f(t) for the time period T. This means that, frit . f(t-nT) (109) fay a a : bie ot 7 - Figure 10: Generation of a periodic signal and the corresponding Fourier series coefficients. ‘We saw in the case of discrete Fourier series that as the period 1” is made larger, [(i)] the fundamental frequency (2x/T) becomes smaller, [(ii) ] the spacing between the frequency lines (2/7) decreases, [(ii)] the amplitude of each frequency component decreases, the general shape of the spectrum remains unchanged. We now consider the limiting case of T + 00. Tims fr(t) = f(t) (110) felt) = So Rye (ui) 27where 2 1 pte inwat, Pag] fete mtd wy (112) Define: Wy = Rwy (113) Pn) = TP (14) This results in ow ad frlt)= 0 Fet= 5) Pilon (115) ma “unt Foon) = [fle dt (116) Spacing between adjacent frequency lines is equal to: Aw = 2x/T. Substituting in (115), we obtain, fel) = 3 Plesent 52 (uz) As tends to infinity, Aw = 2n/T, which is the spacing between the frequency lines, tends to zero and the frequency spectrum becomes continuous. In this e, the relationships (116), (117) change into, P(w) = Lf f(t)e dt = F{f(t)} (118) tee oe L0) = 5 [ Plwjedo =F P(w)} (119) We also have, F,= FF (nu) (120) ‘The function F(w) is known as the [spectral density] function of f(t). One can think of the Fourier transform as the representation of a give signal in terms of an infinite sum of complex exponentials cach weighted by F(w)dw. This means that we have infinite number of terms where the contribution of each term tends to zero. In contrast to this, a period waveform has all its amplitude components at discrete frequencies where each component has a definite contribution. Note that relationship (120) expresses the Fourier series coefficients of a periodic signal in terms of the samples of the Fourier transform of one period of the same signal.It is usual to express the Fourier transform in terms of f or w = 2nf. In changing from f tow, or vice versa, the following relationship is useful, Af) = 2nd(w) (121) Example:] itera rect(t/r (222) 0, Itl>7P ~ a Pw) =f sect(t/rye*tar = oat of [emt evierl? _ giur/ty (—juy) — p8i0(@T /2 ( Ni -fa) = ae F {rect(t/r)} = r8a(wr/2) (124) 4.1 Parseval’s theorem [@ra= [ soroa [10 [x [% Peereman) a = [10 ef. Fr(wje a] a= (125) x £ Fw) ft Feat] Coe xe [Pe Ployde = 5 |” |PwyPde ‘This means that, [Linord = 2 [~ reypae (026) The quantity |F(w)|? is called the [energy spectral density] (unit of joule per Hz) of f(t) The quantity |F(w)|?du is the energy in a small frequency band of dw around w. Note that in the case of a periodic signal, the parseval relationship computes the average power of the signal and is of the form, oa aL ecard = Se (27) In this case, |Fa|? is the power in the frequency nuy = n2x/T. 294.2. Some Fourier transforms involving impulse functions F{S(t)} = Lf. 8(t)etdt = 1 F{8(t— to} = Lf 5(t = to)etde = eve Fao Fund} =z [sw Funletde = Loti Fletint} = 2x6 Fy) wy =0 + Fil} = 2nd(w) Fos wot} = Fie" + feiaty = a[8(w — uo) + dw + w0)] F{sinat} = Figo = yo} 5 Flite — wp) — 6(w + wo)] f any =) 9, ~1, t<0 F{ogu(t)} = F {lime sole*sem(t)]} = time wo { J [erWagn(eyje-*ae} = ° oo fingaa {= f° oe ears [Perera = taf 2 Fsgn(t)} = limo { 5 30 (128) (129) (130) (131) (132) (133) (134) (135) (136) (137) (138) (139) (140)u(t) = ; + Soentt) (141) Fult)} = SFO) + SF seul) (asa) Ffu(t)} = 15(w) + Ajo (143) ‘The spectral density function of the unit step contains an impulse at w = 0 corresponding to the average value of 1/2 in the step function, It also has all the high frequency components of the sgn() function reduced by one-half. ‘Table 1 contains a list of some important Fourier transform pairs. Figure 11 shows some of these Fourier transform pairs. 4.3. Periodic signals For a periodic signal fr(t), we have, felt) = 35 Pre™*, wy = 20/7 (144) ripen =7{ > Raemot (145) FUrO}= SS RFlO™) (46) FUfslt)} <2" So Fys(w— muy) (a7) iodic signal of unit impulses: p(t) = > (tn) (148) S(t) = > Reinet (149) where ear —jnwot, 5 Bland dt (150) in)=% SS ome (151) 31fi) Fla) = Afi} 1. e*ule) 1a + je) 2 te “ult) Ifa + jw)? 3. ew 2a/(a? + w*) 4 ertae oVine 2 5. sgn (1) 2/(Je) 6. i/(nt) sgn (w) 7. ult) 78a) + 1/(jw) 8. al) 1 94 2n8(w) iS Gane 2n8(w F wo) 11. cos wot a{8(w — a) + 3(w + w)] 12. sin wot —ja[ Bw ~ wo) ~ B(w + w)) 13, reet (t/1) 7 Sa (wr/2) 14, “sa cw) rect (0/W) an 1s. Z sawn rect (w/(2W)) 16. A(t/r) [Sa (wr/2)P v7, fsa one Aw/W) an 18. cos (at/7) rect (t/r) Botan 2W __cos (Wt) 9. Se cos [w/(2W)] rect [o/(2W)] 20. 8(0) 8.(t), Where wy = 2r/T ‘Table 1: Some selected Fourier transform pairs.> 3 3 x * F a zg 4 = _= Figure 11; Some selected Fourier transform pairs. 33 > & > 2 gf. 7 xq + ov = € i é 8 3 £ ; we 2 3 2 g 2 é é eof £ £ € e €FASr(t)} = 2 x 5(w ~ nwo) = wo > 5(w — nw) (152) F{Sr(t)} = wodug(w) (153) Example:] Consider the rectangular pulse, 1 lel < 7/2 rect(t/r) = (154) 0, itl> 7/2 We already saw that, Ff{rect(t/r)} = rSa(wr/2) (155) Now consider a periodic signal obtained by repeating rect(t/r) over a period of T. Using (120), the corresponding Fourier series coefficients are equal to, 1 Fa = pF (neo) = (156) which is the same as the result obtained earlier The Fourier transform of this period rectangular signal is equal to, Ffrectr(t)} = 2" > Fd(w — nwo) = (157) n(navor /2) SOT 5(w — ny (nwt /2) ( 0) ‘These are shown in Fig. 12. 4.4 Properties of Fourier Transform « [Linearity (superposition) Flas falt) + a2folt)} = arF{ filt)} + a2 falt)} (158) « [Complex conjugate FPO} = [~ piyear 7 iz seyemat]” (159) = F(-w) 34Flw)=Sa*/,) y Sore ) FUG) o «@ Figure 12: (a) The rectangular pulse, (b) its Fourier transform. (c) The Fourier series transform of the periodic coefficients of the periodic rectangular signal, (d) The Four rectangular signal. = Flw). Real f(t), then f*(t) = f(t) and F*(— Using this result, it is easy to show that for real f(#), the real/imaginary parts of F(w) should have even/odd symmetry.Any signal can be written as the sum of an even part and an odd part, f(t) = Felt) + folt). We have, FA felt)} = Fel) and real FAfo(t)} = Fw) and imaginary (160) ‘To prove, consider an even function f,(t). We have, FUL} =f faltyertat sy =a S(t) cos(wt)at — 7 f~ felt) sin (wt at The first integral involves an even function and is equal to two times the value of the integral over [0,00]. The second integral involves an odd function and is equal to zero. This results in, Ff} = 2 [ felt) cos(wt)at (162) This is a real, even| function of w. Similarly, for an odd function, we have, FURY = [~ faltyerat aa (163) = ja folt) cos(wt)dt ~ if folt)sin(we)at ‘The first integral involves an odd function and is equal to zero. The second integral involves an even function and is equal to two times the value of the integral over (0, co]. This results in, F{folt)} = —25 | folt) sin(wt)dt (164) imaginary, odd] function of w. Using these results together, we conclude 36