4. Fourier Series 4.1. Introduction In this section we will study periodic signals in terms of their frequency content. Remind that a signal f (t ) is said to be periodic if f (t + T ) = f (t ) (4.1) for every t, where T is a number. From this definition it follows that f (t + nT ) = f (t ) for every integer n, positive or negative. For example if n = 2 then f (t + 2T ) = f ((t + T ) + T ) = f (t + T ) = f (t ) . Similarly, for n = −1 the equation f (t − T ) = f (t' ) = f (t' +T ) = f (t ) holds. The smallest positive number T for which (4.1) holds is called a period. Let us consider the well known sinusoidal function Asin (ω0t + φ ) where A is the amplitude, ω0 is the angular frequency, and φ is the phase. Since Asin (ω0 (t + T ) + φ ) = Asin (ω0t + φ + ω0T ) (4.2) 116 and 2π is the smallest angle that satisfies the equation sinα = sin (α + 2π ) then T= 2π . ω0 (4.3) Let us consider two periodic signals f 1 (t ) and f 2 (t ) with periods T1 and T2 , respectively. The common period of f 1 (t ) and f 2 (t ) , if it exists, is the smallest positive number T satisfying f 1 (t + T ) = f 1 (t ) f 2 (t + T ) = f 2 (t ) and for every t. Observe that any T, such that f 1 (t + T ) = f 1 (t ) for all t, can be expressed in the form T = n1T1 where n1 is an integer. Likewise any T, such that f 2 (t + T ) = f 2 (t ) for all t, can be expressed in the form T = n 2T2 , where n2 is an integer. Hence, the condition under which the functions f 1 (t ) and f 2 (t ) have a common period is n1T1 = n2T2 . The above equation states that the ratio T1 n2 = T2 n1 must be a rational number. The common period is the lowest common multiple (over the positive integers field) of T1 and T2 . If the ratio T1 T 2 is an irrational number, then f 1 (t ) and f 2 (t ) do not have a common period. In such a case integers n1 and n2 such that n1T1 = n2T2 do not exist. Let h(x ) be an arbitrary function. If f (t ) is a periodic signal with period T then the function defined by the equation g (t ) = h( f (t )) 117 is periodic because g (t + T ) = h( f (t + T ) ) = h( f (t ) ) = g (t ) . Hence, the period of g (t ) is T or less than T (see Example 2). Example 4.1 Let us consider the sinusoidal signals f1 (t ) = 5cos(ω0t + 60°) f 2 (t ) = 10sin 3ω0t . The periods of these signals are: T1 = 2π ω0 T2 = , 2π . 3ω0 Consequently, we obtain T1 =3 T2 thus, the ratio T1 T 2 is a rational number. Therefore, the two signals have a common period given by T = T1 = 3T2 = 2π ω0 . Now we consider the signals: g1 (t ) = 5cos(ω0t + 45°) g 2 (t ) = 10sin 3 ω0t . The periods of these signals are: T1 = and the ratio 2π ω0 , T2 = 2π 3 ω0 118 T1 = 3 T2 is an irrational number. Hence, the signals do not have any common period. Example 4.2 Let us consider the periodic signal f (t ) = cost with period T1 = 2π and let h(x ) = x . The function g (t ) = h( f (t ) ) is g (t ) = cost . It is a periodic function with period T2 = T1 = π (see Fig.4.1) 2 f(t) (a) 0 (b) π 2π π 2π t g(t) 0 t Fig. 4.1. Plots of signals f (t ) = cos t and g (t ) = cos t Thus, the period of g (t ) is only half that of f (t ) . 119 4.2. Fourier series. Definition We consider a periodic signal f (t ) with the period T. If some conditions, which will be discussed later, are satisfied, we can express the function f (t ) as a series of sine and cosine functions f (t ) = a0 + a1cos ω0t + a2cos 2ω0t + + + b1sinω0t + b2sin 2ω0t + or f (t ) = a0 + ∞ ∑ (a cos nω t + b sin nω t ) n 0 n 0 (4.4) n =1 2π . Expression on the right hand side of (4.4) is called the T trigonometric Fourier series. The sum is called Fourier series and its terms are called the harmonics. The n-th harmonic is the term where ω0 = wn (t ) = an cos nω0t + bnsin nω0t = ⎛ ⎞ an bn = an2 + bn2 ⎜ cos nω0t + sin nω0t ⎟ . ⎜ a 2 + b2 ⎟ an2 + bn2 n ⎝ n ⎠ We label c n = a n2 + bn2 cos θ n = an an2 + bn2 sin θ n = (4.5) − bn an2 + bn2 (4.6) obtaining wn (t ) = cn (cos θ n cos ω0t − sin θ nsin nω0t ) = cn cos (nω0t + θ n ) . (4.7) Thus, the n-th harmonic is a sinusoidal function with the amplitude c n , the phase θ n and the angular frequency nω0 . The first harmonic 120 w1 (t ) = a1cos ω0t + b1sin ω0t = c1cos (ω0t + θ1 ) is called the fundamental and its frequency ω0 is called the fundamental frequency. The constant term a 0 is the 0-th harmonic. Hence, an equivalent form of the Fourier series f (t ) = a0 + c1cos (ω0t + θ1 ) + c2cos (2ω0t + θ 2 ) + = c0 + = ∞ ∑ c cos (nω t + θ ) n 0 (4.8) n n =1 with c0 = a0 follows. The plot of c n as a function of n is called the amplitude spectrum and the plot of θ n as a function of n is called the phase spectrum of the signal f (t ) . Together they are called the frequency spectra. Example 4.3 Let us consider a periodic signal f (t ) including the following harmonics: c0 = 2 w1 (t ) = 1.5cos (ω0t + 30°) w3 (t ) = cos (3ω0t − 45°) w5 (t ) = 0.5cos 5ω0t rad . s Figure 4.2 shows these harmonics and illustrates how the signal f (t ) is built up from its harmonics. where ω0 = 1000π (a) c0 2 t 121 (b) w1(t) t (c) w3(t) t w5(t) (d) t (e) f (t) t Fig. 4.2. Construction of the signal f (t ) consisting of four harmonics 122 4.3. Evaluation of Fourier series coefficients Our objective is to evaluate the a n and bn coefficients in the Fourier series (4.4). The coefficients are called the Fourier coefficients of the signal f (t ) . In order to determine a 0 we integrate both sides of equation (4.4) from 0 to T T ∫ 0 T ⎛ T ⎞ ⎜ f (t ) dt = a0dt + an cos nω0t dt + bn sin nω0t dt ⎟ . ⎜ ⎟ n =1 ⎝ 0 0 0 ⎠ ∞ T ∑ ∫ ∫ ∫ (4.9) Let us consider the integral T ∫ cos nω t dt . 0 0 The function cos nω0t has the period Tn = 2π nω0 whereas T= 2π ω0 . Hence, T = nTn holds. Since integral of cos nω0t over its period Tn is zero Tn ∫ cos nω t dt = 0 0 0 then Tn T ∫ ∫ (4.10) ∫ sin nω t dt = 0 . (4.11) cos nω0t dt = n cos nω0t dt = 0 . 0 0 Similarly, we have T 0 0 123 Substituting (4.10) and (4.11) into (4.9) we obtain T ∫ T f (t ) dt = a0dt = a0T ∫ 0 0 or 1 a0 = T T ∫ f (t ) dt . (4.12) 0 Formula (4.12) can be modified by adding an arbitrary value t0 to the lower and upper limit of integration 1 a0 = T t 0 +T ∫ f (t ) dt . (4.13) t0 Equation (4.13) shows that a0 is the average value of f (t ) over the period. To compute am we multiply both sides of the Fourier series (4.9) by cos mω0t and then integrate from 0 to T obtaining T ∫ T f (t )cos mω0t dt = a0cosmω0t dt + ∫ 0 + 0 (4.14) ∞ T ∑ ∫ (a cos nω tcos mω t + b sin nω tcos mω t )dt . n 0 0 n 0 0 n =1 0 As explained above, the first term on the right hand side of (4.14) is zero. To rearrange the other terms we use the following formulas: T ∫ sin nω tcos mω t dt = 0 0 (4.15) 0 0 ⎧⎪ 0 cos nω0tcos mω0t dt = ⎨ 1 T ⎪⎩ 2 0 T ∫ if m≠n if m=n . Hence, we have T f (t )cos mω0t dt = am cos 2 mω0t dt = 0 0 ∫ T ∫ 1 amT 2 (4.16) 124 or 2 am = T T ∫ 0 2 f (t ) cos mω0t dt = T t0 +T ∫ f (t )cos mω t dt m = 1, 2, 0 . (4.17) t0 In a similar manner we find 2 bm = T T ∫ 0 2 f (t )sin mω0tdt = T t 0 +T ∫ f (t )sin mω t dt m = 1, 2, 0 . (4.18) t0 Example 4.4 Let us consider the periodic signal with period T=2 having the waveform shown in Fig.4.3 f (t) 2 0 1 2 t -1 Fig. 4.3. Signal f (t ) for Example 4.4 We wish to express this signal by a Fourier series. The waveform of f (t ) is described as follows ⎧2 f (t ) = ⎨ ⎩− 1 0 < t <1 . 1< t < 2 for for To find a 0 , we apply (4.12) a0 = 1 2 ⎞ 1 1 ⎛⎜ 2dt − dt ⎟ = . ⎟ 2 2 ⎜⎝ 0 1 ⎠ ∫ ∫ Next, we calculate the Fourier coefficients a m given by (4.12): 125 1 2 1 2 ⎞ 2 1 2⎛ sin mω0t − sin mω0t . am = ⎜ 2cos mω0t dt − cos mω0t dt ⎟ = ⎟ mω0 2 ⎜⎝ 0 mω0 0 1 1 ⎠ ∫ Since ω0 = ∫ 2π = π we obtain T am = 0 m = 1, 2 , . To determine bm we use (4.18): bm = 1 2 1 2 ⎞ 2 ⎛⎜ 2sin mω0tdt − sin mω0t dt ⎟ = 2 sin mπt dt − sin mπt dt = ⎟ 2 ⎜⎝ 0 1 0 1 ⎠ ∫ ∫ ∫ ∫ = 1 2 −2 1 1 (− 2cos mπ + 2 + cos2mπ − cos mπ ) = cos mπt + cos mπt = mπ mπ mπ 0 1 = 3 (1 − cos mπ ) . mπ Hence, it follows ⎧ ⎪ 0 ⎪ bm = ⎨ ⎪ 6 ⎪ mπ ⎩ if m is even m = 1, 2 , if . m is odd Thus, the coefficients bm are b1 = 6 π , b2 = 0, b3 = 2 π , b4 = 0, b5 = 6 6 , b6 = 0, b7 = , 5π 7π and the Fourier series has the form f (t ) = 1 6 2 6 6 + sinω0t + sin 3ω0t + sin 5ω0t + sin 7ω0t + 2 π π 5π 7π . 126 Since cn = an2 + bn2 = bn and for n odd: an cosθ n = then θ n = − an2 + bn2 bn sinθ n = − =0 an2 + bn2 = −1 π for n odd. 2 The amplitude and phase spectra of the signal are shown in Fig.4.4 (a ) cn 6 2 .0 π 1 .5 1 .0 2 π 0 .5 0 (b) 2 1 6 5π 4 3 6 7π 6 5 n 7 θn 0 1 2 3 4 5 6 7 n π − 2 Fig. 4.4. Spectra of the signal for Example 4.4 4.4. Properties of Fourier series In this section we will discuss some waveform properties of periodic functions. 127 A function f (t ) is said to be even if, for every t, it satisfies the condition f (− t ) = f (t ) . Examples of the even functions are: cos ω0t , sinω0t , sin 2ω0t . Note that if a function is even, its graph is symmetric in the vertical axis (see Fig.4.5). f (t) − 0 T 2 t T 2 Fig. 4.5. An even function A function f (t ) is said to be odd if, for every t, it satisfies the condition f (− t ) = − f (t ) . Examples of odd functions: sin ω0t sin 3ω0t . Note that if a function is odd, its graph is symmetric in the origin (see Fig.4.6) f (t) − T 2 0 Fig. 4.6. An odd function T 2 t 128 A periodic signal f (t ) with period T is said to possess odd half-wave symmetry if, for every t, it satisfies the condition ⎛ T⎞ f (t ) = − f ⎜ t + ⎟ . ⎝ 2⎠ Function sinω0t has this property; another example is shown in Fig.4.7. f (t) 0 T 2 T t Fig. 4.7. A function possessing odd half-wave symmetry Even functions Let f (t ) be a periodic even function. Then for any t0 the equation t0 t0 −t0 0 ∫ f (t )dt = 2 ∫ f (t )dt holds. This equation implies a0 = 1 T T 2 T 2 ∫ f (t ) dt = T ∫ f (t ) dt . T − 2 2 0 If a function is even, then its Fourier series does not contain the terms (4.19) 129 bnsin nω0t n = 1, 2, because they violate the relation f (− t ) = f (t ) . As a matter of fact bnsin nω0 (− t ) = −bnsin nω0t . The expression for am can be reduced as follows: am = t0 +T T 2 ∫ f (t ) cos mω t dt = T ∫ f (t ) cos mω t dt . 2 T 2 0 t0 0 − T 2 Since both f (t ) and cos mω0t are even, then f (t ) cos mω0t = f (− t ) cos mω0 (− t ) . Consequently, we have T 2 am = 2 4 2 f (t ) cos mω0t dt = T 0 T ∫ T 2 ∫ f (t )cos mω t dt 0 m = 1, 2 , . (4.20) 0 Summarizing, we state that the Fourier series of an even periodic signal contains only a0 and cosine terms. Odd functions For an odd function f (t ) and any t0 the equation t0 ∫ f (t )dt = 0 −t0 holds; hence, we conclude that a0 = 0 . If a function is odd then its Fourier series does not contain the terms a0 and an cos nω0t because they violate the relation f (− t ) = − f (t ) . n = 1, 2, 130 Similarly as previously, it can be proved that the formula for bm can be simplified and has the form bm = 4 T T 2 ∫ f (t )sin mω t dt . (4.21) 0 0 Thus, the Fourier series of an odd periodic signal contains only sine terms. Odd half-wave symmetry functions In the case of odd half-wave symmetry the constant term a0 and all even ⎛ T⎞ harmonics cannot exist because they violate the condition f (t ) = − f ⎜ t + ⎟ , i.e. 2⎠ ⎝ a0 = 0 am = bm = 0 for m even . The coefficients am and bm for m odd are given by: am = bm = 4 T 4 T T 2 ∫ f (t )cos mω t dt 0 (4.22) 0 T 2 ∫ f (t )sin mω t dt . 0 (4.23) 0 The above formulas can be derived using the following relationships for m odd: ⎛ T⎞ ⎛ T⎞ f ⎜ t + ⎟ cos mω0 ⎜ t + ⎟ = − f (t ) cos(mω0t + mπ ) = f (t ) cos mω0t 2 ⎝ ⎠ ⎝ 2⎠ ⎛ T⎞ ⎛ T⎞ f ⎜ t + ⎟ sin mω0 ⎜ t + ⎟ = − f (t ) sin (mω0t + mπ ) = f (t ) sin mω0t . ⎝ 2⎠ ⎝ 2⎠ Using these expressions we can reduce the interval of integration twice and obtain formulas (4.22) and (4.23). 131 The advantages of the symmetry properties of periodic functions are as follows: (i) We know in advance what coefficients will be zero; consequently, we do not waste time to compute them. (ii) The other coefficients can be obtained using simplified expressions. 4.5. Exponential Fourier series We consider the trigonometric Fourier series f (t ) = a0 + ∞ ∑ (a cos nω t + b sin nω t ) n 0 n (4.24) 0 n =1 and substitute: cos nω0t = sin nω0t = ( ) ( ) 1 jnω 0 t e + e − jnω 0 t 2 1 jnω 0 t e − e − jnω 0 t . 2j As a result, we obtain f (t ) = a0 + ∑ ⎜⎜⎝ 2 (e ∞ ⎛ an jnω 0 t ) + e − jnω 0 t + n =1 = a0 + ∞ ∑ n =1 ( bn jnω 0 t − e − jnω 0 t e 2j ⎛ an − jbn jnω 0 t an + jbn − jnω 0 t ⎞ + e e ⎜ ⎟. 2 2 ⎝ ⎠ )⎞⎟⎟ = ⎠ (4.25) Let c~0 = a0 (4.26) 1 c~n = (a n − jbn ) 2 (4.27) then we have 1 c~− n = (a − n − jb− n ) . 2 Since a− n = an and b− n = −bn (see (4.17) and (4.18)) 132 1 c~− n = (a n + jbn ) = c~n∗ 2 (4.28) Hence, we conclude that c~n and c~− n form a pair of complex conjugate numbers. Inserting (4.26), (4.27) and (4.28) into (4.25) yields f (t ) = ~ c0 + ∞ ∑ c~ e jnω 0 t n + n =1 ∞ ∑ c~ − ne − jnω 0 t . (4.29) n =1 The third term on the right hand side of (4.29) can be rearranged as follows ∞ ∑ −∞ c~− n e − jnω 0 t = n =1 ∑ ~c e jnω 0 t n . (4.30) n = −1 Using (4.30) we rewrite (4.29) in the compact form ∞ f (t ) = ∑ ~c e jnω 0 t (4.31) n n = −∞ called the exponential Fourier series. The coefficients c~n are given by (4.27) where an and bn are specified by (4.17) and (4.18) repeated below: 2 an = T bn = 2 T t0 +T ∫ f (t )cos nω t dt 0 t0 t0 +T ∫ f (t )sin nω t dt . 0 t0 Hence, it follows 1 2 ⎛⎜ c~n = 2T⎜ ⎝ 1 = T t0 +T ∫ ⎞ f (t )sin nω0t dt ⎟ = ⎟ ⎠ t0 +T f (t ) cos nω0t dt − j t0 ∫ t0 t 0 +T ∫ f (t )(cos nω t − jsin nω t ) dt . 0 t0 0 133 Using Euler’s formula e jα = cosα + jsin α , we obtain 1 c~n = T Usually we set t0 = − T t 0 +T ∫ f (t ) e − jnω 0 t n = 1, 2, dt . (4.32) t0 or t0 = 0 , then c~n is 2 1 c~n = T T 2 ∫ f (t ) e T − 2 − jn ω 0 t 1 dt = T T ∫ f (t ) e − jnω 0 t n = 1, 2, dt . (4.33) 0 The coefficients c~n are generally complex and can be expressed in the polar form c~n = c~n e jφ n n = 1, 2 , . (4.34) c− n = c~n and φ− n = −φn . Using (4.28) we obtain the relationships: ~ The plot of c~n versus n is called the amplitude spectrum and the plot of φn versus n is called the phase spectrum. Note that for n = 0 1 c~0 = T t 0 +T ∫ f (t ) dt = a 0 . (4.35) t0 Since − jtan a − jbn 1 2 c~n = n = an + bn2 e 2 2 −1 bn an then, using (4.5) and (4.6), we can express c~n in terms of cn and θ n 1 c~n = cn e jθ n . 2 (4.36) Assuming an arbitrary integer n = n̂ in the exponential Fourier series, we prove that the sum of two terms corresponding to n = n̂ and n = − n̂ gives the n-th harmonic: 134 ( c~n̂ e jn̂ω 0 t + c~− n̂ e − jn̂ω 0 t = ~ cn̂ e jn̂ω 0 t + ~ cn̂∗ e jn̂ω 0 t ) = 2Re(c~ e ) = ∗ jn̂ω 0 t n̂ ⎛1 ⎞ = 2Re⎜ cn̂ e j(n̂ω 0 t +θn̂ ) ⎟ = cn̂ cos (n̂ω0t + θ n̂ ) . ⎝2 ⎠ To derive the above relation, equations (4.28) and (4.36) have been applied. Example 4.5 Let us consider a periodic function f (t ) , shown in Fig.4.8, specified by the equation 1 f (t ) = Asin ω0t 2 where ω0 = for 0≤t ≤T 2π . For this function we find the coefficients c~n . T f(t) A 0 T 2T t Fig. 4.8. Periodic function for Example 4.5 To determine c~n we use (4.27). Since f (t ) is even then bn = 0 and formula (4.27) reduces to 1 c~n = an 2 (4.37) where an is specified by (4.20) repeated below for convenience an = 4 T T 2 ∫ f (t )cos nω t dt 0 0 n = 1, 2 , . 135 Hence, we have T 2 an = 4A 1 sin ω0tcos nω0t dt T 0 2 ∫ n = 1, 2, and apply the trigonometric identity sinαcosβ = 1 (sin(α − β ) + sin(α + β ) ) 2 finding ⎛T ⎞ 2 ⎟ 2 A ⎜⎜ ⎛ 1 1 ⎞ ⎟ an = ⎜ sin( ω0 − nω0 )t + sin( ω0 + nω0 )t ⎟ dt = T ⎜0⎝ 2 2 ⎠ ⎟ ⎜ ⎟ ⎝ ⎠ ∫ T ⎡ ⎤2 −1 1 1 1 2A ⎢ ⎛ ⎞ ⎛ ⎞ ⎥ = cos ⎜ ω0 + nω0 ⎟ t ⎥ = cos ⎜ ω0 − nω0 ⎟ t − ⎢1 1 T ⎢ ω − nω ⎝2 ⎠ ⎝2 ⎠ ⎥ ω 0 + nω 0 0 0 2 ⎣2 ⎦0 ⎤ ⎡ ⎛1 ⎞ T ⎛1 ⎞ T cos ⎜ − n ⎟ω0 cos ⎜ + n ⎟ω0 ⎥ ⎢ − 2A 1 1 ⎝2 ⎠ 2 − ⎝2 ⎠ 2 − ⎥. ⎢ = + 1 1 1 T ⎢ ⎛1 ⎞ ⎛ ⎞ ω0 − nω0 ω0 + nω0 ⎥ ⎜ + n ⎟ω0 ⎥ ⎢ ⎜⎝ 2 − n ⎟⎠ω0 2 2 2 ⎝ ⎠ ⎦ ⎣ Since ω0 T = π and ω0T = 2π , then 2 ⎞ ⎛ ⎛1 ⎞ ⎛1 ⎞ ⎟ ⎜ cos ⎜ − n ⎟π cos ⎜ + n ⎟π − 2A ⎜ 1 ⎟ 1 2 2 ⎝ ⎠ ⎝ ⎠ an = = − + − ⎟ 1 1 1 1 2π ⎜ n n n n − − + + ⎟ ⎜ 2 2 2 2 ⎠ ⎝ ⎛ ⎞ π⎞ π⎞ ⎛ ⎛ ⎜ cos ⎜ nπ − ⎟ ⎟ cos ⎜ nπ + ⎟ A⎜ 1 ⎟ 1 2⎠ 2⎠ ⎝ ⎝ . = − − + 1 1⎟ 1 1 π⎜ n n n n − − + + ⎜ ⎟ 2 2⎠ 2 2 ⎝ 136 Next we use (4.37) and calculate: 1 2A ~ c1 = a1 = − 2 3π 1 2A c~2 = a2 = − 2 15π 1 2A c~3 = a3 = − 2 35π ………………… . To find c~0 = a 0 we apply equation (4.12) 1 c~0 = a0 = T T ∫ 0 1 cos ω0t A 1 A 2 f (t ) dt = sin ω0t dt = − T 0 2 T 1ω 0 2 T T ∫ = 2A π . 0 Example 4.6 Let us consider a rectangular-wave signal f (t ) indicated in Fig.4.9. f (t) 1 T 2 0 T -1 Fig. 4.9. Rectangular-wave signal f (t ) for Example 4.6 To find the coefficients c~n , we apply (4.33) repeated below 1 c~n = T where T ∫ f (t ) e 0 − jnω 0 t dt t 137 ⎧1 f (t ) = ⎨ ⎩− 1 and ω 0 = for for 0<t < T 2 T <t <T 2 2π . Hence, we have T T 2 T 1 − jnω 0 t 1 − jnω 0 t e dt − e dt = c~n = T 0 TT ∫ ∫ 2 T 1⎛ 1 ⎞ − jnω 0 t 2 1 ⎛ 1 ⎞ − jnω 0 t ⎟e ⎟⎟ e − ⎜⎜ − = ⎜⎜ − T ⎝ jnω0 ⎠ T ⎝ jnω0 ⎟⎠ 0 T . T 2 Using relationship ω0T = 2π we obtain after simple rearrangements ( ) j c~n = 2e − jnπ − 1 − e − jn 2π , 2πn n = ±1, ± 2, . (4.38) Note that the signal f (t ) is odd and has odd half-wave symmetry property; hence, c~0 = a 0 = 0 as well as a n = 0 and for every n bn = 0 for an even n. Consequently c~n = 0 for an even n. Taking into account the above statements and relationships (4.38) and (4.28) we obtain: π π 2 2 −j c~1 = − j = e 2 2 2 j c~−1 = c~1∗ = j = e 2 c~2 = 0 c~− 2 = c~2∗ = 0 π π π π π π 2 2 −j2 = e c~3 = − j 3π 3π 2 2 j2 = e c~− 3 = ~ c3∗ = j 3π 3π c~4 = 0 c~− 4 = c~4∗ = 0 π 2 2 −j2 = e c~5 = − j 5π 5π …………………… π 2 2 j2 = e c~− 5 = ~ c5∗ = j 5π 5π …………………… . 138 Thus, the exponential Fourier series expansion of f (t ) is 2⎛ f (t ) = j ⎜ π⎝ ⎞ ⎟. ⎠ 1 1 1 1 + e − j5ω 0 t + e − j3ω 0 t + e − jω 0 t − e jω 0 t − e j3ω 0 t − e j5ω 0 t − 5 3 3 5 The amplitude and phase spectra are indicated in Figs.4.10 and 4.11 ~ c n 2 2 π π 2 2 5π π -5 -4 -3 -2 0 -1 1 2 3 4 n 5 Fig. 4.10. Amplitude spectrum of the signal for Example 4.6 π φn 2 -5 -4 -3 -2 0 -1 − 1 2 3 4 5 n π 2 Fig. 4.11. Phase spectrum of the signal for Example 4.6 Since c~− n = c~n , the amplitude spectrum is symmetric in the vertical axis. Similarly, since φ− n = −φn , the phase spectrum is symmetric in the origin. 4.6. Convergence of Fourier series Not every periodic signal can be represented by a Fourier series and the question arises under what conditions the Fourier series exists. This question is answered by the Dirichlet conditions as follows: 139 (i) Over any period f (t ) must be absolutely integrable, i.e. T ∫ f (t ) dt < ∞ . 0 (ii) In any period f (t ) is of bounded variation, i.e. f (t ) has at most a finite number of maxima and minima. (iii) In any period f (t ) has at most a finite number of discontinuities and each of these discontinuities is finite. Under these conditions the Fourier series of f (t ) converges to f (t ) at all points where f (t ) is continuous and the series converges to [ ( ) ( )] 1 f t i+ + f t i− 2 at any point ti of discontinuity. Note that the condition (i ) guarantees that each coefficient c~n is finite, because 1 c~n = T T ∫ f (t ) e 0 − jnω 0 t 1 dt ≤ T T ∫ f (t ) e 0 − jnω 0 t 1 dt = T T ∫ f (t ) dt < ∞ . 0 A periodic function that violates this condition is depicted in Fig.4.12 f (t) − 3 T 2 -T − T 2 T 2 T 3 T 2 Fig. 4.12. Function that violates the first Dirichlet condition An example of a function that does not satisfy condition (ii) is t 140 ⎛ 2π ⎞ f (t ) = e t −1sin ⎜ ⎟ ⎝1− t ⎠ 1 0 ≤ t <1 with T = 1 as depicted in Fig.4.13. f (t) 0 2 1 t Fig. 4.13. Function that violates the second Dirichlet condition An example of a function that violates condition (iii) is indicated in Fig.4.14. f (t) 0 t T T 2 2 Fig. 4.14. Function that violates the third Dirichlet condition − Dirichlet conditions do not require that a function f (t ) be continuous in order to possess a Fourier expansion. Thus, its waveform may consists of a number of disjointed arcs as depicted in Fig.4.15. 141 f (t) c b a t1 0 T 2T Fig. 4.15. Signal with discontinuity points t The signal shown in Fig.4.15 has two points of discontinuity at t1 and T over the period. Setting t = t1 into the Fourier series of this signal we obtain 1 (a + c ) . 2 Similarly, setting t = T we obtain 1 b. 2 It can be shown that Fourier coefficients tend to zero as n approaches infinity. Therefore the Fourier series can be truncated after a finite number of terms. The truncated series including N terms will be called the partial sum and labeled s N (t ) . Let us consider a periodic signal f (t ) having the waveform depicted in Fig.4.16. This signal satisfies the Dirichlet conditions. f(t) 1 − T 2 − T 4 0 T 4 T 2 Fig. 4.16. Rectangular pulse train signal t 142 T the points of discontinuity occur. Figure 4.17 shows the partial sums 4 s N (t ) of this signal for N = 7 , N = 11 , N = 15 and N = 25 . Note that the partial sums exhibit ripples and oscillations. Furthermore, the peak value of these ripples At t = ± (a) s7(t) T − 2 T − 4 (b) 0 T 4 T 2 T 4 T 2 t s11(t) 0 T − 2 T − 4 t 143 (c) s15(t) 0 T − 2 (d) T 4 T − 4 T 2 t s25(t) 0 T T T T − − 2 4 2 4 Fig. 4.17. An illustration of the Gibbs phenomenon t in the neighborhood of the discontinuity points does not decrease as N increases and remains approximately 9% of the height of the signal for any finite N. Thus, the deviations in the vicinity of the discontinuities are rather large. This behavior of s N (t ) is known as the Gibbs phenomenon.