Discrete-Time Fourier Transform • Continuous-Time Fourier Transform • Discrete-Time Fourier Transform. • Discrete-Time Fourier Transform Theorems. • Band-Limited Discrete-Time Signals • The Frequency Response of an LTI Discrete-Time System • Phase and Group Delays Continuous-Time Fourier Transform Pair Equation Synthesis equation or Inverse FT . 1 +∞ jωt x(t ) = X ( j ) e d . ω ω ∫ 2π −∞ Analysis equation or Fourier Transform +∞ X ( jω ) = ∫ x(t )e −∞ − j ωt dt. Continuous-Time Fourier Transform 1 Inverse CTFT : − x(t ) = 2π CTFT : − X ( jω ) = ∫ t = +∞ t = −∞ ω = +∞ ∫ω = −∞ X ( jω )e jω t dω x(t )e − jω t dt CTFT → X ( jω ) x(t ) ← X ( jω ) = Re{ X ( jω )} + j Im{ X ( jω )} =| X ( jω ) | e j∠X ( jω ) | X ( jω ) | is the magnitude spectrum. ∠X ( jω ) is the phase spectrum −1 Im{ X ( jω )} = arg{ X ( jω )} = tan Re{ X ( jω )} Example 3.1 − at x(t ) = e u (t ) X ( jω ) = ∫ t = +∞ t = −∞ ∞ x(t )e =∫ e e 0 − at − jω t − jω t dt ∞ dt = ∫ e 0 − 1 − ( a + jω ) t ∞ = e |0 a + jω 1 CTFT − at e u (t ) ←→ a + jω − t ( a + jω ) Convergence of CTFT • CTFT of continuous-time signal will only exist if the Dirichlet conditions are satisfied:• (1) Signal must have finite number of discontinuities and finite number of maximum & minimum in any finite interval of time. • (2) Signal must be absolutely integrable:- ∫ ∞ −∞ | x(t )dt < ∞ Properties of Fourier Transform • • • • • • • • • • Linearity Time Shifting Conjugation Differentiation in the time-domain Integration in the time-domain Time and Frequency Scaling. Duality Parseval’s Relation Convolution Multiplication Summary of the approach to obtain the Fourier Transform for Discrete-time aperiodic signals. x[n] APERIODIC ~ - construct periodic signal x[n] for which one period is x[n]. ~ - x[n] has a Fourier series. ~ - as period of x[n] increases, ~ x[n] → x[ n], ~ and Fourier series of x[n] → Fourier Transform of x[n] Development of Discrete-time Fourier Transforms. DTFT. x[n} -N1 0 -N -N1 0 N2 N2 N Fourier Series of discrete - time periodic signal ~ x[n] = jk ( 2π / N ) n a e ......Eqn 5.1 ∑ k k =< N > Development of Discrete-time Fourier Transforms. Thus we have the Discrete - time Fourier Transform pair as : 1) Synthesis equation, Inverse Fourier Transform : 1 jω jω n x[n] = X ( e ) e dω . ∫ 2π 2π 2) Analysis equation, Fourier Transform or Fourier Spectrum : ∞ jω X(e ) = − jω n x [ n ] e . ∑ n = −∞ Discrete-time Fourier Transform DTFT jω x[n] ↔ X (e ) jω jω jω X (e ) = Re{ X (e )} + j Im{X (e )} jω = | X(e ) | e j∠X ( e jω ) jω | X(e ) | is known as magnitude spectrum. jω Im{X (e )} ∠X (e ) = arg{ X (e )} = tan jω Re{ X (e )} is known as the phase spectrum jω jω −1 Example 3.6 Discrete-time Fourier Transform. x[n] = a nu[n], 0 < a < 1 0 n Example 3.6 Discrete-time Fourier Transform. x[n] = a n u[n], 0 < a < 1 ∞ jω X (e ) = ∑ a u[n]e n ∞ − jω n = ∑ (ae 1 ) = 1 − ae − jω − jω n n =0 n = −∞ 1 (1 − a) |X(ω)| 1 (1 + a ) -π 0 tan −1 (a / 1 − a 2 ) Phase X(ω) -π − tan −1 ( a / 1 − a 2 ) 2π π π 2π ω Properties of Discrete-time Fourier Transforms. DTFT jω x[n] ↔ X (e ) DTFT x[-n] ↔ X (e Periodic : − jω ) X (ω ) = X (ω + 2πm) Properties of Discrete-time Fourier Transforms. Conjugation : DTFT x [ n ] ↔ X * ( e − jω ) * DTFT Pr oof From : − x[n] ↔ X (e jω ) DTFT xre [n] + jxim [n] ↔ X re (e jω ) + jX im (e jω ) Taking conjugation on both side of above equation : _ ie. replacing j = -j. DTFT xre [n] − jxim [n] ↔ X re (e − jω ) − jX im (e − jω ) DTFT ∴ x [n] ↔ X * (e − jω ). * Properties of Discrete-time Fourier Transforms. DTFT Conjugation : − x [n] ↔ X * (e − jω ) * Conjugate Symmetry : − DTFT If x[n] is real, then x[n] = x [n] ↔ X (e jω ) = X * (e − jω ) * By Defination : - X (e jω ) = X re (e jω ) + jX im (e jω ) and X* (e − jω ) = X re (e − jω ) − jX im (e − jω ) Since X (e jω ) = X * (e − jω ) because x[n] is real. ∴ X re (e jω ) + jX im (e jω ) = X re (e − jω ) − jX im (e − jω ) ∴ X re (e jω ) = X re (e − jω ), i.e. X re (e jω ) is an even function of ω . ∴ X im (e jω ) = − X im (e − jω ), i.e. X im (e jω ) is an odd function of ω . Properties of Discrete-time Fourier Transforms. Conjugate Symmetry F .T . If x[n] is real, then x[n] = x *[n] ↔ X (e jω ) = X * (e − jω ) By Defination : - X (e jω ) = X re (e jω ) + jX im (e jω ) replacing ω → -ω , X (e − jω ) = X re (e − jω ) + jX im (e − jω ) taking conjugation X* (e − jω ) = X re (e − jω ) − jX im (e − jω ) | X (e jω ) |=| X re (e jω ) + jX im (e jω ) |= X re2 (e jω ) + X im2 (e jω ) | X (e − jω ) |=| X re (e − jω ) + jX im (e − jω ) |= X re2 (e − jω ) + X im2 (e − jω ) | X * (e − jω ) |= | X re (e − jω ) − jX im (e − jω ) |= X re2 (e − jω ) + X im2 (e − jω ) ∴| X (e jω ) |=| X (e − jω ) |=| X * (e − jω ) | Since | X (e jω ) | =| X (e − jω ) |, the function | X (e jω | is an even function of ω . arg{ X(e jω )} = tan −1 X im (e jω ) X re (e jω ) arg{ X(e − jω )} = tan −1 X im (e − jω ) X re (e − jω ) arg{ X* (e − jω )} = tan −1 − jω ) − X im (e − jω ) −1 X im (e tan = − − jω − jω X re (e ) X re (e ) ∴ arg{ X* (e − jω )} = −arg{ X(e − jω )} and Since X (e jω ) = X * (e − jω ) because x[n] is real. arg{ X(e jω )} = arg{ X* (e − jω ) ∴ arg{ X(e jω )} = −arg{ X(e − jω )} i.e Phase angle X (e jω ) is an odd function of ω . Properties of Discrete-time Fourier Transforms. F .T . If x[n] is real, and x[n] ↔ X (e jω ) = Re{ X (e jω )} + jIM { X (e jω )} x[n] = Even part x[n] + Odd part of x[n] jω jω jω X (e ) = Even part X (e ) + Odd part X(e ). But from previous slide, Real { X (e jω )} is even function of ω . F .T . ∴ Ev{x[n]} ↔ Re{ X (e jω )} From previous slide, Im { X (e jω )} is odd function of ω . F .T . ∴ Odd{x[n]} ↔ j Im{X (e jω )} Discrete-time Fourier Transform Theorems. Time shifting : F .T x[n - n 0 ] ↔ e − jωn0 X (e jω ) Frequency shifting : e jω 0 n F .T x[n] ↔ X (e j (ω −ω 0 ) ) Linearity : F .T ax1[n] + bx 2 [n] ↔ aX 1 (e jω ) + bX 2 (e jω ) Parseval's relation : ∞ 1 | x[n] | = ∑ 2π n =-∞ 2 ∫π 2 | X (e jω ) |2 dω Convolution Theorem F .T h[n] * x[n] ↔ H (ω ) X (ω ) x[n] X(ω) h[n] H(ω) h[n]*x[n] H(ω)X(ω) Multiplication Theorem 1 h[n]x[n] ↔ H (ω ) * X (ω ) 2π DTFT DTFT x[n] h[n]x[n] x h[n] Band-Limited Discrete-Time Signals. (e.g. Bandpass) X ( jω ) = 0, for 0 ≤| ω |< ω a and , ω b ≤| ω |< Bandwidth = ω b - ω a − π − ωb − ωa ωa ωb π ω Lowpass Discrete-Time Signals X ( jω ) = 0, ω p ≤| ω |< π ≠ 0, 0 ≤| ω |< ω p Bandwidth = ω p −π −ω p ωp π ω Highpass Discrete-Time Signals X ( jω ) ≠ 0, ω p ≤| ω |< π = 0, 0 ≤| ω |< ω p Bandwidth = π - ω p . −π −ω p 0 ωp π ω e.g. Highpass Discrete-Time real signal Imaginary part 1 0.5 0.5 Amplitude Amplitude Real part 1 0 -0.5 -1 0 0.5 ω/π Magnitude Spectrum 0 -0.5 -1 1 0.5 0 0.5 ω/π Phase Spectrum 1 0 0.5 ω/π 1 4 Phase, radians Magnitude 1 0 0 0.5 ω/π 1 2 0 -2 -4 Frequency Response of LTI DTS x[n] = e jωn y[n] = H (e jω )e jωn h[n] y[n]=x[n]*h[n] In the time domain the output y[n] is calculated, by convolving input sequence with the system impulse response : y[n] = ∞ ∞ k = −∞ k = −∞ ∑ x[k ]h[n − k ] = ∑ h[k ]x[n − k ] Important property of LTI system, certain input signals known as the eigenfunction produces the same function at the output of the system multiplied with complex constant known as eigenvalue. Frequency Response of LTI DTS x[n] = e jωn y[n] = H (e jω )e jωn h[n] y[n] = ∞ ∞ k = −∞ k = −∞ ∑ x[k ]h[n − k ] = ∑ h[k ]x[n − k ] ∞ y[n] = ∑ h[k ]e ∞ jω ( n − k ) =e j ωn k = −∞ ∑ h[k ]e − j ωk k = −∞ y[n] = H (e jω )e jωn ∞ jω where H (e ) = ∑ h[k ]e − j ωk k = −∞ This is the DTFT of h[n] by changing the variable k to n. H (e jω ) is the frequence response of the LTI DTS and is related to its impulse response h[n] through DTFT. DTFT i.e. h[n] ⇔ H (e jω ) Frequency Response of LTI DTS x[n] h[ n] H ( e jωn ) jω X (e ) y[n] jω Y (e ) Time Domain y[n] = x[n] * h[n] But by DTFT theorem : convolution in time domain ≡ multiplication in frequency domain DTFT x[n] * h[n] ⇔ X (e jω ) H (e jω ). DTFT y[n] ⇔ Y (e jω ). jω Y ( e ) jω i.e. H (e ) = X ( e jω ) Frequency Response LTI DTS H (e jω ) is a complex function and is known as Frequency Response. H (e jω ) = H re (e jω ) + jH im (e jω ) =| H (e jω ) | e jθ (ω ) | H (e jω ) |= H re2 (e jω ) + H im2 (e jω ) ...............Magnitude Response θ (ω ) = arg{H (e jω )}.............................Phase Response. G (ω ) = 20 log 10 | H (e jω ) | decibels(dB)....Gain Function. A(ω ) = -G(ω )..........Attenuation or Loss Function Frequency Response of LTI FIR Discrete-Time System Input - Output Relationship given by : N2 y[n] = ∑ h[k]x[n - k], N1 < N 2 . k = N1 Taking DTFT with linearity & time - shifting properties : Y(e jω ) = N2 - j ωk jω h[k]e X(e ), ∑ k = N1 jω N2 Y(e ) - jωk h[k]e ∴ Frequency Response H (e jω ) = = ∑ jω X(e ) k = N1 jω jω H (e ) is a polynomial in e . Frequency Response of LTI IIR Discrete-Time System Input - Output Relationship given by : M N ∑a k y[n - k] = ∑ b k x[n - k]. k =0 k =0 Taking DTFT with linearity & time - shifting properties : M N ∑ a ke k =0 - jωk Y(e ) = ∑ b k e - jωk X(e jω ). jω k =0 M jω Y(e ) jω ∴ Frequency Response H (e ) = = jω X(e ) - jωk b e ∑ k k =0 N - jωk a e ∑ k k =0 H (e jω ) is a rational function in e jω . Concept of Filtering • Discrete Signals are sequences of numbers in time-domain. • Through DTFT, Discrete Signals are transformed into frequency domian which are represented as magnitude and phase spectrum of frequencies. • Shape the Spectrums to the desired forms by multiplying them with frequency response of LTI Discrete-Time Systems. – Frequency shaping filters – Frequency selective filters. Frequency Response of LTI DTS x[n] h[ n] H ( e jωn ) jω X (e ) y[n] jω Y (e ) Time Domain y[n] = x[n] * h[n] But by DTFT theorem : convolution in time domain ≡ multiplication in frequency domain DTFT x[n] * h[n] ⇔ X (e jω ) H (e jω ). DTFT y[n] ⇔ Y (e jω ). jω Y ( e ) jω i.e. H (e ) = X ( e jω ) In Frequency Domain X ( jω ) = 0, ω p ≤| ω |< π ≠ 0, 0 ≤| ω |< ω p Bandwidth = ω p Signal Magnitude Spectrum −π −ω p ωp π Magnitude Response Of LTI DTS (Filter) ωc ω What happen to Phase? • Phase and Group Delays need to be considered since both DTFT of Input and impulse response are complex functions. • Multiplying these two complex functions will result in addition of their phase components (argument of complex function) • Group delay is defined as the negative derivative of phase function with respect to angular frequency. Phase & Group Delay jω jω jω Y(e ) = X (e ) H (e ) ∠Y(e jω ) = ∠X (e jω ) + ∠H (e jω ) IF ∠H (e jω ) is denoted as θ (ω ) IF θ (ω ) < 0, the output will lag the input. IF θ (ω ) > 0, the output will lead the input. θ (ω ) dθ (ω ) Group Delay = dω θ (ω 0 ) Phase Delay = - ω0 ω0