Chapter 5 FOURIER TRANSFORMS 5.1 INTRODUCTION Fourier series are powerful tools for problems involving functions that are periodic or are of interest on a finite interval only. Since many problems involve functions that are non periodic and are of interest on either semi-infinite interval or the whole x -axis, we ask what can be done to extend the method of Fourier series to such functions. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS Chapter 5 FOURIER TRANSFORMS 5.1 INTRODUCTION The Fourier integral can be regarded as the limiting case of a Fourier series representation of a function f (x ) defined over an interval −L < x < L as L → ∞. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 Fourier Integral We know that the expansion of a function f (x ) defined in an interval [−L, L] into Fourier series has the form ∞ a0 X + (an cos ωn x + bn sin ωn x ) f (x ) = 2 n=1 where ωn = nπ L , Z 1 L an = f (t) cos(ωn t)dt and L −L Z 1 L bn = f (t) sin(ωn t)dt. L −L Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL The substitution of the expressions of an and bn into Fourier series leads to 1 f (x ) = 2L + Z L f (t)dt 1 L −L ∞ Z L X n=1 f (t) cos ωn t cos ωn x + sin ωn t sin ωn x dt. −L Thus, 1 f (x ) = 2L Z L ∞ 1X f (t)dt+ L n=1 −L Z L f (t) cos ωn (t−x )dt −L (1) Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL Now we set ∆ω = ωn+1 − ωn = Then 1 L = π (n + 1)π nπ − = . L L L ∆ω π and we may write (1) in the form Z L 1 f (x ) = f (t)dt 2L −L Z L (2) ∞ 1X f (t) cos ωn (t − x )dt. + ∆ω π n=1 −L Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL Suppose f (x ) is defined throughout the x -axis and piecewise smooth in any finite interval [−L, L] and thus can be expanded into Fourier series in every such interval. Then the representation is valid for any fixed L, arbitrarily large, but finite. We shall assume that f is absolutely integrable on (−∞, ∞), that is the improper integral Z ∞ |f (t)|dt = Q −∞ is convergent (and so Q is finite). Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL Letting L → ∞ we find that Z L Z L 1 1 f (t)dt ≤ |f (t)|dt 2L −L 2L −L Z ∞ Q 1 → 0. |f (t)|dt = ≤ 2L −∞ 2L Hence 1 lim L→∞ 2L Z L f (t)dt = 0. −L Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL Note that the improper integral Z ∞ φ(ω) := f (t) cos ω(t − x )dt −∞ is absolutely convergent since Z Z ∞ f (t) cos ω(t−x ) dt ≤ −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai ∞ |f (t)|dt = Q < ∞. −∞ CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL Moreover, the improper integral φ(ω) is absolutely convergent, and hence, if f is a continuous function, the function φ is continuous as well. For sufficiently large values of L the integral Z L f (t) cos ωn (t − x )dt −L can be regarded as being almost equal to the value of the function φ(ω) assumed for ω = ωn , that is, Z L f (t) cos ωn (t − x )dt ≈ φ(ωn ). −L Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL The second term on the right side of (2) therefore can be rewritten as Z L ∞ ∞ 1X 1X ∆ω φ(ωn )∆ω. f (t) cos ωn (t−x )dt ≈ π n=1 π n=1 −L P π In the sum ∞ n=1 φ(ωn )∆ω the quantity ∆ω = L tends to zero as L → ∞ while, in the limit, the variable ωn run through all the values of ω from 0 to ∞. Therefore it is natural to expect that this sum tends to the integral Z ∞ Z ∞ Z ∞ φ(ω)dω = dω f (t) cos ω(t − x )dt. 0 0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai −∞ CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL On substituting the expression obtained into the formula of f (x ) we obtain from (2), Z ∞ Z 1 ∞ dω f (t) cos ω(t − x )dt (3) f (x ) = π 0 −∞ The integral on the right is called Fourier integral and the formula itself is Fourier integral formula. It is clear that our naive approach merely suggest the representation (3). Sufficient conditions for validity of (3) are as follows. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL Theorem 5.1 (Fourier integral theorem) Let f be absolutely integrable over the interval −∞ < x < ∞ and suppose that f (x ) satisfies Dirichlet conditions: In any finite interval, f is piecewise continuous and has only a finite number of maxima and minima. Then the Fourier integral representation of f (x ), given in (3), converges to f (x ) at all points where f is continuous and to 12 [f (x −) + f (x +)] where f is discontinuous. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL Since cos ω(t − x ) = cos ωt cos ωx + sin ωt sin ωx , formula (3) can be rewritten in the form Z ∞ Z 1 ∞ f (t) cos ωtdt cos ωxdω f (x ) = π 0 −∞ Z ∞ Z 1 ∞ f (t) sin ωtdt. + sin ωxdω π 0 −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL By defining the functions A(ω) and B(ω) as Z 1 ∞ f (t) cos ωtdt and A(ω) = π −∞ Z 1 ∞ B(ω) = f (t) sin ωtdt π −∞ (4) the Fourier integral representation in (3) can be written in the simpler form Z ∞ f (x ) = A(ω) cos ωx + B(ω) sin ωx dω 0 (5) Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL Relation (5) is analogous to an expansion of a function into a Fourier series, and expressions (4) are similar to the formulae for the Fourier coefficients. Example 1.1 Find the Fourier integral representation of f (x ) = e −|x | . 2 , B(ω) = 0, π(1 + ω 2 ) Z 2 ∞ cos ωx dω. = π 0 1 + ω2 A(ω) = e −|x | Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.2 FOURIER INTEGRAL FOR EVEN AND ODD FUNCTIONS For an even or odd function the Fourier integral becomes simpler. Just as in the case of Fourier series, this is of practical interest in saving work and avoiding errors. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.1 FOURIER INTEGRAL If f is an even function, then f (t) sin ωt is odd, so B(ω) = 0 in (4) and Z 2 ∞ f (t) cos ωtdt. A(ω) = π 0 The Fourier integral (5) then reduces to the Fourier cosine integral Z ∞ f (x ) = A(ω) cos ωxdω (6) 0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.2 FOURIER INTEGRAL FOR EVEN AND ODD FUNCTIONS Similarly, if f is an odd function, then f (t) cos ωt is odd, so A(ω) = 0 in (4) and Z 2 ∞ f (t) sin ωtdt. B(ω) = π 0 The Fourier integral (5) then reduces to the Fourier sine integral Z ∞ f (x ) = B(ω) sin ωxdω (7) 0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.2 FOURIER INTEGRAL FOR EVEN AND ODD FUNCTIONS Example 1.2 Find Fourier cosine and Fourier sine integrals of f (x ) = e −βx where x > 0 and β > 0. ANS. f (x ) = e −βx 2β = π Z 0 ∞ cos ωx 2 dω = ω2 + β 2 π Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 Z 0 ∞ ω sin ωx dω ω2 + β 2 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.2 FOURIER INTEGRAL FOR EVEN AND ODD FUNCTIONS Summary of Fourier integral representations Let f (x ) satisfy the conditions of Theorem 5.1. (a) f (x ) has the general Fourier integral representation Z ∞ 1 f (x +)+f (x −) = [A(ω) cos ωx+B(ω) sin ωx ]dω, 2 0 where Z 1 ∞ f (t) cos ωtdt and A(ω) = π −∞ Z 1 ∞ B(ω) = f (t) sin ωtdt. π −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.2 FOURIER INTEGRAL 5.2.2 FOURIER INTEGRAL FOR EVEN AND ODD FUNCTIONS (b) If f (x ) is even, it has the Fourier cosine integral representation Z ∞ 1 f (x +) + f (x −) = A(ω) cos ωxdω. 2 0 (c) If f (x ) is odd, it has the Fourier sine integral representation Z ∞ 1 B(ω) sin ωxdω. f (x +) + f (x −) = 2 0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM An integral transform is a transformation in the form of an integral that produces from given functions new functions depending on a different variable. The Laplace transform is of this kind and is by far the most important integral transform in engineering. The next in order of importance are Fourier transforms. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.1 FOURIER INTEGRAL IN COMPLEX FORMS The starting point for the development of the Fourier transform is the complex form of the Fourier integral representation of a function f (x ). To derive this representation in which f (x ) is defined over the interval (−∞, ∞), we write Z ∞ Z ∞ (3) 1 dω f (t) cos ω(t − x )dt f (x ) = π 0 −∞ Z Z ∞ 1 ∞ = dω f (t) cos ω(x − t)dt π 0 −∞ where we have used the result cos ω(t − x ) = cos ω(x − t). Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.1 FOURIER INTEGRAL IN COMPLEX FORMS As the integrand in the last integral is an even function of ω, Z ∞ Z ∞ 1 dω f (t) cos ω(x − t)dt f (x ) = 2π −∞ −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 (8) FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.1 FOURIER INTEGRAL IN COMPLEX FORMS The function sin ω(x − t) is an odd function of ω, so Z ∞ Z ∞ 1 dω f (t) sin ω(x − t)dt (9) 0= 2π −∞ −∞ Multiplying equation (9) by j, adding the result to equation (8), and using the Euler formula e jθ = cos θ + j sin θ, we arrive at the complex Fourier integral representation Z ∞ Z ∞ 1 dω f (t)e jω(x −t) dt f (x ) = 2π −∞ −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 (10) FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR We write the factor e jω(x −t) in (10) as the product e jω(x −t) = e jωx · e −jωt and have 1 f (x ) = 2π Z ∞ e jωx Z −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai ∞ f (t)e −jωt dt dω (11) −∞ CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR Definition 3.1 For a given function f : R → C the function F (ω) (for ω ∈ R) is defined by Z ∞ (12) F (ω) = f (t)e −jωt dt −∞ provided the integral exists as an improper Riemann integral. The function F (ω) is called the Fourier transform or spectrum of f (t). Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR F (ω) exists when f (t) is absolutely integrable on the interval (−∞, ∞). The transformation from f (t) to F (ω) is called the Fourier transformation. We will also write F{f }(ω) instead of F (ω). The symbol F denotes the Fourier transformation. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR • f (t) will represent a function depending on time, while F (ω) usually depends on frequency. • Thus we say that f (t) is defined in the time domain and F (ω) is defined in the frequency domain. Note that F (ω) will in general be a complex-valued function, so F : R → C. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR It follows from (11) that Z ∞ 1 F (ω)e jωt dω f (t) = 2π −∞ (13) Since formula (13) recovers f (t) from its Fourier transform F (ω), f (t) is termed the Fourier inverse transform of F (ω). As with the Laplace transform, the pair {f (t), F (ω)} is called a Fourier transform pair. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR There are a number of variations of definition of F (ω) and you need to be aware that different authors may use slightly different formulae. This can cause confusion when the Fourier transform is first met. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR Although the convention that we have adopted here is fairly standard, some authors associate the factor 1 with (12) rather than (13), while others 2π 1 associate a factor of √ with each of (12) and 2π (13). We stress that this is a choice that we have made, but you should have no difficulty in using any form. All that is required of the factors is that their 1 product be and so, the pair combine to give the 2π Fourier integral (10). Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR Example 3.1 Find the Fourier transform of the one-sided exponential function f (t) = H(t)e −at , (a > 0) where H(t) is the Heaviside unit step function. ANS. F{f (t)} = 1 . a + jω Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR Example 3.2 Calculate the Fourier transform of the rectangular pulse ( A if |t| ≤ T , f (t) = 0 if |t| > T . Solution F{f (t)} = Z T Ae −jωt dt = 2AT sinc ωT . −T where sinc x is defined by ( sinc x = sin x x 1 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai if x ̸= 0, if x = 0. CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR Example 3.3 Show that the Fourier transform of the unit step function H(t) does not exist. Solution Z ∞ Z H(t)e −jωt dt = −∞ ∞ e 0 −jωt ′ Since e Z ∞ −jωt Z dt = lim b→∞ b e −jωt dt. 0 = −jωe −jωt , it follows that b 1 lim e −jωt 0 −jω b→∞ −∞ j = lim e −jωb − 1 . ω b→∞ However, this limit does not exists. H(t)e −jωt dt = Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.2 FOURIER TRANSFORM PAIR Table of Fourier transforms f (t) e −at H(t) (a > 0) F{f } 1 a + jω 1 (a + jω)2 te −at H(t) (a> 0) ( A (|t| ≤ T ) 2AT sinc ωT 0 (|t| > T ) 2a e −a|t| (a > 0) a2 + ω 2 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.3 CONTINUOUS FOURIER SPECTRA Fourier transforms are generally complex-valued functions of the real frequency variable ω. Writing F (ω) in the exponential form F (ω) = |F (ω)|e j arg F (ω) , plots of |F (ω)| and arg F (ω), which are both real-valued functions of ω, are called the amplitude and phase spectra respectively of the signal f (t). The two spectra represent the frequency-domain portrait of the signal f (t). Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.3 CONTINUOUS FOURIER SPECTRA Example 3.4 Determine the amplitude and phase spectra of the causal signal f (t) = e −at H(t) (a > 0) and plot their graphs. ANS. |F (ω)| = √ 1 a2 + ω 2 −1 arg F (ω) = tan (1) − tan Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai −1 ω CALCULUS 3 Chapter 5 a FOURIER TRANSFORMS 5.3 FOURIER TRANSFORM 5.3.3 CONTINUOUS FOURIER SPECTRA The Fourier transform discussed so far have two properties in common. First, the amplitude spectra are even functions of the frequency variable ω. This is always the case when the time signal f (t) is real. The second common feature is that all the amplitude spectra decrease rapidly as ω increases. This means that most of information concerning the “shape” of the signal f (t) is contained in a fairly small interval of frequency axis around ω = 0. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.4 PROPERTIES OF FOURIER TRANSFORM 5.4.1 LINEARLY PROPERTY Theorem 5.2 (Linearity of the Fourier Transform) If f (t) and g(t) are functions having Fourier transforms and if α and β are constants, then F{αf (t) + βg(t)} = αF{f (t)} + βF{g(t)} As a consequence of this, we say that the Fourier transform operator F is a linear operator. Clearly the linearity property also applies to the inverse transform operator F−1 . Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.4 PROPERTIES OF FOURIER TRANSFORM 5.4.2 TIME-DIFFERENTIATION PROPERTY Theorem 5.3 (Fourier Transform of Derivatives) Let f (t) be continuous on the t-axis and f (t) → 0 as |t| → ∞. Furthermore, let f ′ (t) be absolutely integrable on the t-axis. Then F{f ′ (t)} = jωF{f (t)} (14) Two successive applications of (14) give the transform of the second derivative of f F{f ′′ (t)} = −ω 2 F{f (t)} Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.4 PROPERTIES OF FOURIER TRANSFORM 5.4.2 TIME-DIFFERENTIATION PROPERTY Repeating the argument n times, it follows that n o (n) (15) F f (t) = (jω)n F{f (t)} The result (15) is referred to as the time-differentiation property, and may be used to obtained frequency-domain representation of differential equations. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.4 PROPERTIES OF FOURIER TRANSFORM 5.4.2 TIME-DIFFERENTIATION PROPERTY Example 4.1 Given 2 √ 2 F e −t (ω) = π e −ω /4 , 2 find F{te −t }(ω). ANS. F{te −t 2 r π 2 }(ω) = − jωe −ω /4 . 2 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.4 PROPERTIES OF FOURIER TRANSFORM 5.4.3 TIME-SHIFT PROPERTY Theorem 5.4 (Time-shift Property) If f (t) has a Fourier transform F (ω), then F{f (t − t0 )} = e −jωt0 F (ω) (16) The result (16) is known as the time-shift property, and implies that delaying a signal by a time t0 causes its Fourier transform to be multiplied by e −jωt0 . (16) implies F−1 e −jωt0 F (ω) = f (t − t0 ) = f (t) t→t−t0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.4 PROPERTIES OF FOURIER TRANSFORM 5.4.3 TIME-SHIFT PROPERTY Example 4.2 Determine the Fourier transform of the rectangular pulse ( A, 0 ≤ t ≤ 2T f (t) = 0, elswhere. ANS. F (ω) = 2ATe −jωT sinc ωT . Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.4 PROPERTIES OF FOURIER TRANSFORM 5.4.4 FREQUENCY-SHIFT PROPERTY Theorem 5.5 (Frequency-shift Property) If f (t) has a Fourier transform F (ω), then F e jω0 t f (t) = F (ω − ω0 ) (17) The result (17) is known as the frequency-shift property, and indicates that multiplication by e jω0 t simply shifts the spectrum of f (t) so that it is centered on the point ω = ω0 in the frequency domain. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.4 PROPERTIES OF FOURIER TRANSFORM 5.4.4 FREQUENCY-SHIFT PROPERTY Example 4.3 of the signal Determine the frequency spectrum f (t) cos ωc t if F (ω) = F{f (t)} is known. ANS. 1 1 F{f (t) cos ωc t} = F (ω − ωc ) + F (ω + ωc ). 2 2 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.1 ENERGY AND POWER In this section we define two quantities associated with time signals, namely signal energy and signal power. These quantities play an important role in characterizing signal types. The energy associated with the signal f (t) is defined as Z ∞ 2 E= f (t) dt −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.1 ENERGY AND POWER If f has a Fourier transform F (ω), Z ∞ 1 f (t) = F (ω)e jωt dω, 2π −∞ then Z ∞ 1 jωt E= f (t)f (t)dt = f (t) F (ω)e dω dt 2π −∞ −∞ −∞ Z ∞ Z ∞ Z ∞ 1 1 jωt F (ω) f (t)e dt dω = F (ω)F (−ω)dω = 2π −∞ 2π −∞ −∞ Z ∞ Z ∞ 1 1 2 = F (ω)F (ω)dω = F (ω) dω. 2π −∞ 2π −∞ Z ∞ Z ∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.1 ENERGY AND POWER Theorem 5.6 (The Parseval relation for the Fourier transform) If f (t) has the Fourier transform F (ω), then Z ∞ Z ∞ 2 1 2 f (t) dt = F (ω) dω 2π −∞ −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 (18) FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.1 ENERGY AND POWER Equation (18) relates the total energy of the signal 2 f (t) to the integral over all frequencies of F (ω) . 2 For this reason, F (ω) is called the energy 2 spectral density, and a plot of F (ω) versus ω is called the energy spectrum of the signal f (t). The result (18) is called Parserval’s theorem. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.1 ENERGY AND POWER Example 5.1 densities of Determine the energy spectral (a) the one-sided exponential function f (t) = e −at H(t), (a > 0); ( A if |t| ≤ T (b) the rectangular pulse f (t) = 0 if |t| > T . Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.1 ENERGY AND POWER Solution (a) By Example 3.1, F (ω) = 1 . a + jω The energy spectral density of the function f (t) is therefore 1 1 2 F (ω) = = . |a + jω|2 a2 + ω 2 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.1 ENERGY AND POWER (b) By Example 3.2, F (ω) = 2AT sinc ωT . Thus the energy spectral density of the pulse is 2 F (ω) = 4A2 T 2 (sinc ωT )2 . Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.1 ENERGY AND POWER There are important signals f (t), defined in general for <2t < ∞, for which the integral R ∞ −∞ −∞ f (t) dt is unbounded. For such signals, instead of considering energy, we consider the average power P, frequently referred to as the power of the signal: Z 2 1 T P = lim f (t) dt T →∞ T −T Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.2 TRANSFORMS OF IMPULSE FUNCTIONS We now consider the Fourier transform of the Dirac delta function δ(t). Recall that ( Z b g(c) if a < c < b g(t)δ(t − c)dt = 0 otherwise a Thus we readily obtain the following two Fourier transforms: Z ∞ F{δ(t)} = δ(t)e −jωt dt = 1 Z−∞ ∞ F{δ(t)} = δ(t − t0 )e −jωt dt = e −jωt0 −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.3 FOURIER TRANSFORMS OF A CONVOLUTION Now we examine how the Fourier transform behaves under convolutions. 2 2 R∞ R∞ Suppose that −∞ f (t) dt and −∞ g(t) dt are finite. The convolution of f and g, denoted f ∗ g, is defined by Z ∞ f (t − x )g(x )dx , f ∗ g (t) = −∞ or, equivalently, Z f ∗ g (t) = ∞ f (x )g(t − x )dx −∞ Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.3 FOURIER TRANSFORMS OF A CONVOLUTION Theorem 5.7 (Convolution theorem) Let F (ω) and G(ω) be the Fourier transforms of f and g. Then F{f ∗ g}(ω) = F (ω)G(ω) F {F ∗ G}(t) = 2πf (t)g(t) −1 The Fourier transform of the convolution of two functions equals the product of the Fourier transforms of each of the functions. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.5 TRANSFORMS OF THE STEP AND IMPULSE FUNCTIONS 5.5.3 FOURIER TRANSFORMS OF A CONVOLUTION Example 5.2 Use the convolution theorem to evaluate F f (t) ∗ g(t) , where ( 1, −1 ≤ t ≤ 1 f (t) = and g(t) = e −t H(t). 0, otherwise Solution According to Examples 3.1 and 3.2, sin ω 1 F (ω) = 2 and G(ω) = . ω 1 + jω Thus, by Theorem 5.7, sin ω F f (t) ∗ g(t) = 2 . ω(1 + jω) Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 DISCRETE FOURIER TRANSFORM 5.6.1 DEFINITION OF THE DISCRETE FOURIER TRANSFORM In this section we show how a function f (t), with Fourier transform F (ω), can be sampled at intervals of T to give a sequence of values f (0), f (T ), f (2T ) . . .. The discrete Fourier transform takes such a sequence and processes it to produce a new sequence which can be thought of as a sampled version of F (ω). The discrete Fourier transform is important as it is the basis of most signal and image processing methods. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 DISCRETE FOURIER TRANSFORM 5.6.1 DEFINITION OF THE DISCRETE FOURIER TRANSFORM Definition 6.1 Let N be a positive integer and let {fn }N−1 n=0 be a sequence of N complex numbers. The discrete Fourier transform (DFT) of {fn }N−1 n=0 is another sequence Fk , also having N terms, defined by Fk = N−1 X fn e −j2nkπ/N , k = 0, 1, . . . , N − 1. n=0 We also write {Fk }N−1 k=0 by F{fn } to denote the DFT of the sequence {fn }N−1 n=0 . Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 DISCRETE FOURIER TRANSFORM 5.6.1 DEFINITION OF THE DISCRETE FOURIER TRANSFORM Example 6.1 Find the DFT of the sequence {fn }30 = {1, 2, −5, 3}. Solution Here N = 4, so 3 3 X X −j2nkπ/4 Fk = fn e = fn e −jnkπ/2 , n=0 k = 0, 1, 2, 3. n=0 So, F0 = F1 = 3 X n=0 3 X fn e 0 = 1 + 2 + (−5) + 3 = 1, fn e −jnπ/2 = · · · = 6 + j, n=0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 DISCRETE FOURIER TRANSFORM 5.6.1 DEFINITION OF THE DISCRETE FOURIER TRANSFORM F2 = F3 = 3 X n=0 3 X fn e −j2nπ/2 = 3 X fn e −jnπ · · · = −9 n=0 fn e −j3nπ/2 = · · · = 6 − j. n=0 So, the DFT of the sequence 1, 2, −5, 3 is the sequence 1, 6 + j, −9, 6 − j. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 DISCRETE FOURIER TRANSFORM 5.6.2 THE INVERSE DISCRETE FOURIER TRANSFORM The Inverse Discrete Fourier Transform Suppose we know the numbers Fk , the N-point DFT of some sequence {fn }N−1 n=0 . We would like to retrieve {fn }. We can prove that N−1 1X Fk e j2nkπ/N , fn = N n = 0, 1, . . . , N − 1. k=0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 DISCRETE FOURIER TRANSFORM 5.6.2 THE INVERSE DISCRETE FOURIER TRANSFORM Definition 6.2 The inverse DFT of the sequence {Fk }N−1 k=0 is the sequence {fn }, also having N terms, given by N−1 1X Fk e j2nkπ/N , n = 0, 1, . . . , N−1. F {Fk } = fn = N −1 k=0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 DISCRETE FOURIER TRANSFORM 5.6.2 THE INVERSE DISCRETE FOURIER TRANSFORM Example 6.2 Using the definition, find the inverse DFT of the sequence {−4, 1, 0, 1}. Solution In this sequence N = 4 and F0 = −4, F1 = 1, F2 = 0, F3 = 1. From the definition, for n = 0, 1, 2, 3, 3 3 k=0 k=0 1X 1X Fk e j2nkπ/4 = Fk e jnkπ/2 fn = 4 4 1 − 4 + 1 · e jnπ/2 + 0 · e jnπ + 1 · e j3nπ . 4 Then, taking n = 0, 1, 2, 3 gives the sequence {fn }3n=0 = {−1/2, −1, −3/2, −1}. = Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.3 MATRIX REPRESENTATION OF THE DFT The DFT of a sequence {fn }N−1 n=0 is given by Fk = N−1 X fn e −j2nkπ/N n=0 Define w = e −j2π/N so that Fk = N−1 X fn w nk n=0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.3 MATRIX REPRESENTATION OF THE DFT Writing out this sum explicitly for each k we find F0 = f0 w 0 + f1 w 0 + f2 w 0 + · · · + fN−1 w 0 F1 = f0 w 0 + f1 w 1 + f2 w 2 + · · · + fN−1 w N−1 F2 = f0 w 0 + f1 w 2 + f2 w 4 + · · · + fN−1 w 2(N−1) .. . FN−1 = f0 w 0 + f1 w N−1 + f2 w 2(N−1) + · · · + fN−1 w (N−1)(N−1) . These equations can be written in matrix form as follows: Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.3 F0 F1 F2 .. . MATRIX REPRESENTATION OF THE DFT = FN−1 Likewise, f0 f1 f2 = 1 N .. . fN−1 w0 w0 w0 .. . w0 w1 w2 .. . w0 w2 w4 .. . ··· ··· ··· .. . w 0 w N−1 w 2(N−1) · · · w0 w0 w0 .. . w0 w1 w2 .. . w0 w2 w4 .. . ··· ··· ··· .. . w 0 w N−1 w 2(N−1) · · · Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai w0 w N−1 w 2(N−1) .. . w (N−1)(N−1) w0 w N−1 w 2(N−1) .. . w (N−1)(N−1) CALCULUS 3 Chapter 5 f0 f1 f2 .. . . fN−1 (19) F0 F1 F2 .. . FN−1 FOURIER TRANSFORMS . 5.6 FAST FOURIER TRANSFORM 5.6.3 MATRIX REPRESENTATION OF THE DFT Example 6.3 (a) Find the matrix representing a three-point DFT. (b) Use the matrix to find the DFT of the sequence {fn }2n=0 = {4, −7, 11}. Solution (a) Here N = 3 and so √ 3 −1 − j . w = e −j2π/3 = 2 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.3 MATRIX REPRESENTATION OF THE DFT The matrix representing a three-point DFT is 0 1 1 1 w w0 w0 w 0 w 1 w 2 = 1 e j2π/3 e j4π/3 w0 w2 w4 1 e j4π/3 e j8π/3 1 1√ 1√ = 1 −1−j2 √3 −1+j2 √3 . 1 −1+j2 3 −1−j2 3 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.3 MATRIX REPRESENTATION OF THE DFT (b) 1 F0 F1 = 1 F2 1 1√ −1−j 3 2√ −1+j 3 2 1√ −1+j 3 2√ −1−j 3 2 4 −7 11 8 = 2 + j15.5885 . 2 − j15.5885 So, the required DFT is the sequence 8, 2 + j15.5885, 2 − j15.5885. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM The fast Fourier transform (FFT) is an algorithm that has been developed to compute the DFT in an extremely economical fashion. This algorithm exploits the periodicity and symmetry of trigonometric functions to compute the transform. It makes the DFT a practical tool for large N. Here one chooses N = 2p (p integer) and uses the special form of the Fourier matrix to break down the given problem into smaller problems. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM We consider the case where N = 4 = 22 measurements. Then w = e −j2π/4 = e −jπ/2 = −j and thus, w nk = (−j)nk . Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM Thus the matrix representing a four-point DFT is w0 w0 w0 w0 w0 w1 w2 w3 w0 w2 w4 w6 w0 1 1 1 1 3 j w 1 −j −1 6 = 1 −1 1 −1 w 9 w 1 j −1 −j Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 . FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM Let the sample values {gn }3n=0 be given. We find F{gk }. Step 1. Determine the matrix representing the DFT. By (19), Gk = F{gk } representation as 0 G0 w G1 (19) w 0 G2 = w 0 G3 w0 is given in the matrix w0 w1 w2 w3 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai w0 w2 w4 w6 w0 g0 3 w g1 w 6 g2 w9 g3 CALCULUS 3 Chapter 5 . FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM Since w 4 = 1, w 5 = w , w 6 = w 2 , w 7 = w 3 , w 8 = w 4 = 1, w 9 = w 5 = w , we have G0 G1 G2 G3 1 1 1 1 g0 1 w 1 w 2 w 3 g1 = 1 w 2 w 0 w 2 g2 1 w3 w2 w1 g3 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 . (20) FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM Step 2. Factorize the matrix: 1 1 1 1 1 w0 1 w2 w0 w2 1 w2 1 w1 w2 w3 = 0 0 1 w3 w2 w1 0 0 0 0 1 0 w0 0 0 1 0 w0 0 0 . 1 w1 1 0 w2 0 1 w3 0 1 0 w2 (21) The matrix on the left-hand side of (21) is the coefficient matrix of (20), but with rows 2 and 3 interchanged. Thus we can write (20) as Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM G0 1 w0 G2 1 w 2 G1 = 0 0 G3 0 0 0 0 1 0 w0 0 g0 0 0 0 0 1 0 w g1 . 1 w 1 1 0 w 2 0 g2 1 w3 0 1 0 w2 g3 (22) We now define a vector ′ 1 0 g0 g1′ 0 1 g2′ = 1 0 0 1 g3′ g0 w0 0 0 0 w g1 w 2 0 g2 g3 0 w2 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 . (23) FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM It then follows from (23) that g0′ = g0 + w 0 g2 g1′ = g1 + w 0 g3 and g2′ = g0 + w 2 g2 = g0 − w 0 g2 g3′ = g1 + w 2 g3 = g1 − w 0 g3 . Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM (22) can be rewritten in the 1 w0 0 G0 G2 1 w 2 0 G1 = 0 0 1 0 0 1 G3 Hence, G0 G2 G1 G3 form 0 0 w1 w3 g0′ g1′ . g2′ g3′ (24) = g0′ + w 0 g1′ = g0′ + w 2 g1′ = g2′ + w 1 g3′ = g2′ + w 3 g3′ . Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM Example 6.4 Use the method of the FFT algorithm to compute the Fourier transform of the sequence {gn }3n=0 = {1, 2, 1, 0}. In this case N = 4 = 22 , so n n w n = e −j2π/4 = e −jπ/2 = (−j)n . Solution. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM We computing the vector g0′ 1 0 w0 0 g0 g1′ (23) 0 1 0 w 0 g1 g3′ = 1 0 w 2 0 g2 0 1 0 w2 g3 g3′ 1 0 1 0 1 0 1 0 1 2 = 1 0 −1 0 1 = 0 1 0 −1 0 Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 2 2 . 0 2 FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM Set [G0′ G1′ G2′ G3′ ]T = [G0 G2 G1 G3 ]T . From (24) we find that this vector is given by ′ ′ G0 1 w0 0 0 g0 G1′ 1 w 2 0 0 g1′ G2′ = 0 0 1 w 1 g2′ G3′ 0 0 1 w3 g3′ 2 4 1 1 0 0 1 −1 0 0 2 0 = 0 0 1 −j 0 = −2j 2 2j 0 0 1 j Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 . FOURIER TRANSFORMS 5.6 FAST FOURIER TRANSFORM 5.6.4 THE FAST FOURIER TRANSFORM Finally, since G0′ = G0 , G1′ = G2 , G2′ = G1 , and G3′ = G3 , we recover the transform vector [G0 G1 G2 G3 ]T as ′ G0 G0 4 G1 G2′ −2j G2 = G1′ = 0 . G3 G3′ 2j Therefore the Fourier transform of the sequence 1, 2, 1, 0 is the sequence 4, −2j, 0, 2j. Assoc. Prof. Dr. Nguyen Dinh & Dr. Nguyen Ngoc Hai CALCULUS 3 Chapter 5 FOURIER TRANSFORMS