Chapter 3 Discrete Fourier Transform Review The DTFT provides the frequency-domain () representation for absolutely summable sequences. The z-transform provides a generalized frequencydomain ( z ) representation for arbitrary sequences. Features in common Defined for infinite-length sequences. Functions of continuous variable ( or z ). They are not numerically computable transform. We need a numerically computable transform, that is Discrete Fourier Transform (DFT) Copyright © 2005. Shi Ping CUC Chapter 3 Discrete Fourier Transform Content The Family of Fourier Transform The Discrete Fourier Series (DFS) The Discrete Fourier Transform (DFT) The Properties of DFT The Sampling Theorem in Frequency Domain Approximating to FT (FS) with DFT (DFS) Summary Copyright © 2005. Shi Ping CUC The Family of Fourier Transform Introduction Fourier analysis is named after Jean Baptiste Joseph Fourier (1768-1830), a French mathematician and physicist. A signal can be either continuous or discrete, and it can be either periodic or aperiodic. The combination of these two features generates the four categories of Fourier Transform. Copyright © 2005. Shi Ping CUC The Family of Fourier Transform Aperiodic-Continuous-Fourier Transform X ( j ) x( t ) 1 2 x ( t )e j t dt X ( j )e j t d Copyright © 2005. Shi Ping CUC The Family of Fourier Transform Periodic-Continuous-Fourier Series X ( jk 0 ) 1 T0 T0 2 T0 2 x ( t )e x( t ) X ( jk 0 )e jk 0 t dt jk 0 t k 0 2F 2 T0 Copyright © 2005. Shi Ping CUC The Family of Fourier Transform Aperiodic-Discrete-DTFT X (e j ) x( n)e jn n x ( n) 1 2 X (e j )e jn d Copyright © 2005. Shi Ping CUC The Family of Fourier Transform Periodic-Discrete-DFS (DFT) N 1 X (k ) x( n)e j 2 nk N n0 x ( n) 1 N N 1 X ( k )e j 2 nk N k 0 Copyright © 2005. Shi Ping CUC The Family of Fourier Transform Summary Time function Frequency function Continuous and Aperiodic Aperiodic and Continuous Continuous and Periodic( T ) 0 Discrete ( T ) and Aperiodic Aperiodic and Discrete( 0 Periodic( s 2 T 2 T0 ) ) and Continuous Periodic( 2 ) s Discrete ( T ) and Periodic ( T ) 0 T and Discrete( 0 2 T0 ) Copyright © 2005. Shi Ping CUC return The Discrete Fourier Series (DFS) Definition Periodic time functions can be synthesized as a linear combination of complex exponentials whose frequencies are multiples (or harmonics) of the fundamental frequency Periodic continuous-time function x( t ) x( t rT ) fundamental frequency j 2 t T e x( t ) X (k )e j 2 kt T k Periodic discrete-time function x( n) x( n rN ) fundamental frequency j e 2 N n x ( n) 1 N N 1 X (k )e k 0 j 2 kn N Copyright © 2005. Shi Ping CUC The Discrete Fourier Series (DFS) 1 N N 1 N 1 e j 2 j rn N n0 x( n)e j 2 rN r mN 1, 1 1 e 2 j r N 0, N 1 e N rn N n0 1 X ( k ) k 0 N N 1 2 1 n0 N N 1 N 1 e n0 j 2 N N 1 X ( k )e elsewhere j 2 N k 0 ( k r )n kn j 2 rn N e X (r ) Copyright © 2005. Shi Ping CUC The Discrete Fourier Series (DFS) N 1 X (k ) x( n)e j 2 kn N n0 N 1 Because: X ( k mN ) x( n)e j 2 ( k mN ) n N n0 N 1 x( n)e j 2 N kn X (k ) n0 The X (k ) is a periodic sequence with fundamental period equal to N Copyright © 2005. Shi Ping CUC The Discrete Fourier Series (DFS) Let W N e j 2N ~ ~ X ( k ) DFS [ x ( n )] N 1 nk ~ x ( n )W N n0 1 ~ ~ x ( n ) IDFS [ X ( k )] N N 1 ~ nk X ( k )W N k 0 Copyright © 2005. Shi Ping CUC The Discrete Fourier Series (DFS) Relation to the z-transform ~ x ( n ), x(n) 0, 0 n N 1 elsewhere N 1 X (z) x(n)z n , n0 ~ X (k ) N 1 x ( n )( e j 2N k ) n n0 ~ X (k ) X ( z ) | ze j 2 k N ~ The DFS X ( k ) represents N evenly spaced samples of the z-transform X (z ) around the unit circle. Copyright © 2005. Shi Ping CUC The Discrete Fourier Series (DFS) Relation to the DTFT ~ x ( n ), x(n) 0, X (e j 0 n N 1 elsewhere N 1 ) x ( n )e j n , ~ X (k ) n0 ~ X (k ) X (e N 1 j x ( n )e 2 nk N n0 j ) | 2 k N The DFS is obtained by evenly sampling the DTFT at 2 N intervals. It is called frequency resolution and represents the sampling interval in the frequency domain. Copyright © 2005. Shi Ping CUC The Discrete Fourier Series (DFS) ~ X (k ) X (e j ) | 2 k N jIm[z] frequency resolution N=8 2 N k 0 Re[z] Copyright © 2005. Shi Ping CUC The Discrete Fourier Series (DFS) The properties of DFS Linearity ~ ~ ~ ~ DFS[ax1 ( n) bx2 ( n)] aX 1 ( k ) bX 2 ( k ) Shift of a sequence mk ~ ~ DFS[ x ( n+m )] W N X ( k ) e j 2 N mk ~ X (k ) Modulation ~ ~ DFS[W x ( n)] X ( k l ) ln N Copyright © 2005. Shi Ping CUC The Discrete Fourier Series (DFS) if then Periodic convolution ~ ~ ~ Y (k ) X 1 (k ) X 2 (k ) ~ ~ y ( n) IDFS[Y ( k )] N 1 ~ ~ x1 ( m ) x2 ( n m ) m 0 N 1 ~ (m ) x ~ (n m ) x 2 1 m 0 Copyright © 2005. Shi Ping CUC 1 ~ ~ ~ y ( n) IDFS[ X 1 ( k ) X 2 ( k )] N N 1 N 1 mk ~ x1 ( m )W N N k 0 m 0 1 ~ ~ nk X 1 ( k ) X 2 ( k )W N k 0 ~ nk X 2 ( k )W N N 1 ~ ( n m ) k X 2 ( k )W N k 0 N 1 N 1 1 ~ x1 ( m ) m 0 N N 1 N 1 ~ (m ) x ~ (n m ) x 1 2 m 0 ~ (m ) x ~ (n m ) x 2 1 m 0 Copyright © 2005. Shi Ping CUC return The Discrete Fourier Transform (DFT) Introduction The DFS provided us a mechanism for numerically computing the discrete-time Fourier transform. But most of the signals in practice are not periodic. They are likely to be of finite length. Theoretically, we can take care of this problem by defining a periodic signal whose primary shape is that of the finite length signal and then using the DFS on this periodic signal. Practically, we define a new transform called the Discrete Fourier Transform, which is the primary period of the DFS. This DFT is the ultimate numerically computable Fourier transform for arbitrary finite length sequences. Copyright © 2005. Shi Ping CUC The Discrete Fourier Transform (DFT) Finite-length sequence & periodic sequence x (n) Finite-length sequence that has N samples ~ ( n) periodic sequence with the period of N x Window operation ~ 0 n N 1 x ( n), x ( n) 0, elsewhere ~ ( n) R ( n) x ( n) x N Periodic extension ~ ( n) x x( n rN ) r ~ ( n) x (( n)) x N Copyright © 2005. Shi Ping CUC The Discrete Fourier Transform (DFT) The definition of DFT N 1 X ( k ) DFT[ x ( n)] x( n)W N , 0 k N 1 nk n0 x ( n) IDFT[ X ( k )] 1 N N 1 X (k ) x( n)W nk N N 1 X ( k )W N nk , 0 n N 1 n0 ~ RN ( k ) X ( k ) RN ( k ) n0 x ( n) 1 N N 1 X ( k )W N nk ~ ( n) R ( n) RN ( n ) x N n0 Copyright © 2005. Shi Ping CUC return The Properties of DFT Linearity DFT[ax1 (n) bx2 (n)] aX 1 (k ) bX 2 (k ) N3-point DFT, N3=max(N1,N2) Circular shift of a sequence DFT[ x(( n m )) N RN ( n)] W km N X (k ) Circular shift in the frequency domain DFT[W nl N x( n)] X (( k l )) N RN ( k ) Copyright © 2005. Shi Ping CUC The Properties of DFT The sum of a sequence N 1 X (k ) k 0 N 1 x(n)W N nk n 0 k 0 x ( n) n 0 The first sample of sequence x ( 0) 1 N N 1 X (k ) k 0 DFT[ x( n)] X ( k ) DFT[ X ( n)] Nx(( N k )) N RN ( k ) Copyright © 2005. Shi Ping CUC The Properties of DFT Circular convolution x1 ( n) N N 1 x 2 ( n) x1 ( m ) x 2 (( n m )) N RN ( n) m 0 N 1 x 2 ( m ) x1 (( n m )) N RN ( n) x 2 ( n) m 0 DFT[ x1 (n) N N x1 ( n) x2 (n)] X 1 (k ) X 2 (k ) Multiplication DFT[ x1 ( n) x 2 ( n)] 1 N X 1 (k ) N X 2 (k ) Copyright © 2005. Shi Ping CUC The Properties of DFT Circular correlation Linear correlation n n rxy ( m ) x ( n) y * ( n m ) x ( n m ) y * ( n) Circular correlation N 1 rxy ( m ) x( n) y * (( n m )) N RN ( m ) n0 N 1 x(( n m )) N y * ( n)RN ( m ) n0 Copyright © 2005. Shi Ping CUC The Properties of DFT if R xy ( k ) X ( k ) Y ( k ) then rxy ( m ) IDFT[ R xy ( k )] * N 1 x( n) y * (( n m )) N RN ( m ) n0 N 1 x(( n m )) N y * ( n)RN ( m ) n0 Copyright © 2005. Shi Ping CUC The Properties of DFT Parseval’s theorem N 1 x ( n) y ( n) * n0 let 1 N N 1 X ( k )Y (k ) * k 0 x ( n) y( n) N 1 then x ( n) x * ( n) n0 N 1 x ( n) n0 2 1 N 1 X (k ) X N * (k ) k 0 1 N N 1 X (k ) 2 k 0 Copyright © 2005. Shi Ping CUC The Properties of DFT Conjugate symmetry properties of DFT xep (n) and xop (n) Let x (n) be a N-point sequence ~( n) x(( n)) x N 1 ~ 1 ~ ~ xe ( n) [ x ( n) x ( n)] [ x (( n)) N x (( N n)) N ] 2 2 1 ~ 1 ~ ~ xo ( n) [ x ( n) x ( n)] [ x (( n)) N x (( N n)) N ] 2 2 * ~ ~ It can be proved that x ( n) x ( n) e e * ~ ~ xo ( n) xo ( n) Copyright © 2005. Shi Ping CUC The Properties of DFT Circular conjugate symmetric component ~ ( n) R ( n) xep ( n) x e N 1 2 Circular conjugate antisymmetric component x(( n)) N x (( N n)) N RN ( n) ~ ( n) R ( n) xop ( n) x o N 1 2 x(( n)) N x (( N n)) N RN ( n) Copyright © 2005. Shi Ping CUC The Properties of DFT x( n) xep ( n) xop ( n) xep ( n) x (( N n)) N RN ( n) * ep xop ( n) x (( N n)) N RN ( n) * op Copyright © 2005. Shi Ping CUC The Properties of DFT X ep (k ) and X op (k ) X ( k ) X ep ( k ) X op ( k ) X ep ( k ) X (( N k )) N RN ( k ) * ep X op ( k ) X (( N k )) N RN ( k ) * op Copyright © 2005. Shi Ping CUC The Properties of DFT Re[X ep ( k )] Re[X ep (( N k )) N RN ( k )] Im[ X ep ( k )] Im[ X ep (( N k )) N RN ( k )] Re[X op ( k )] Re[X op (( N k )) N RN ( k )] Im[ X op ( k )] Im[ X op (( N k )) N RN ( k )] Copyright © 2005. Shi Ping CUC The Properties of DFT Circular even sequences if x( n) x(( N n)) N RN ( n) then X ( k ) X (( N k )) N RN ( k ) Circular odd sequences if x( n) x(( N n)) N RN ( n) then X ( k ) X (( N k )) N RN ( k ) Copyright © 2005. Shi Ping CUC The Properties of DFT Conjugate sequences DFT[ x ( n)] X (( k )) N RN ( k ) * * X (( N k )) N RN ( k ) X ( N k ) * * DFT[ x (( n)) N RN ( n)] * DFT[ x (( N n)) N RN ( n)] X ( k ) * * Copyright © 2005. Shi Ping CUC The Properties of DFT Complex-value sequences DFTRe[x ( n)] X ep ( k ) 1 2 X (( k )) * N X (( N k )) N RN ( k ) DFT j Im[ x ( n)] X op ( k ) 1 2 X (( k )) N X (( N k )) N RN ( k ) * Copyright © 2005. Shi Ping CUC The Properties of DFT DFT[ xep ( n)] Re[X ( k )] 1 * DFT [ x (( n)) N x (( N n)) N ]RN ( n) 2 DFT[ xop ( n)] j Im[ X ( k )] 1 * DFT [ x (( n)) N x (( N n)) N ]RN ( n) 2 Copyright © 2005. Shi Ping CUC The Properties of DFT Real-value sequences if x( n) is real - value sequence then X ( k ) X (( N k )) N RN ( k ) * Imaginary-value sequences if x( n) only has imaginary part then X ( k ) X (( N k )) N RN ( k ) * Copyright © 2005. Shi Ping CUC The Properties of DFT Summary x ( n) Re[x ( n)] j Im[ x ( n)] X ( k ) X ep ( k ) X op ( k ) x ( n) xop ( n) xep ( n) X ( k ) Re[X ( k )] j Im[ X ( k )] example Copyright © 2005. Shi Ping CUC The Properties of DFT Linear convolution & circular convolution x1 ( n) N1 point sequence, 0≤n≤ N1-1 x2 ( n) N2 point sequence, 0≤n≤ N2-1 Linear convolution yl ( n) x1 ( n) x 2 ( n) N 1 1 x (m ) x 1 m 2 (n m ) x (m ) x 1 2 (n m ) m 0 yl (n) L point sequence, L= N1+N2-1 Copyright © 2005. Shi Ping CUC The Properties of DFT Circular convolution We have to make both x1 ( n) and x2 ( n) L-point sequences by padding an appropriate number of zeros in order to make L point circular convolution. x1 ( n), 0 n N 1 1 x1 ( n) N1 n L 1 0, x2 ( n), 0 n N 2 1 x 2 ( n) N2 n L 1 0, Copyright © 2005. Shi Ping CUC The Properties of DFT yc ( n) x1 ( n) L L 1 x 2 ( n) x1 ( m ) x 2 (( n m )) L RL ( n) m 0 L 1 x1 ( m ) x 2 ( n rL m ) RL ( n) r m 0 L 1 x1 ( m )x 2 ( n rL m ) RL ( n) r m 0 y l ( n rL ) RL ( n) r Copyright © 2005. Shi Ping CUC The Properties of DFT yc ( n) yl ( n rL) RL ( n) r if L N1 N 2 1 then that is yc ( n ) y l ( n ) x1 ( n) L x 2 ( n) x1 ( n) x 2 ( n) Copyright © 2005. Shi Ping CUC return The Sampling Theorem in Frequency Domain Sampling in frequency domain ~ X ( k ) X ( z ) |z W k N x(m )W N km m 1 ~ ~ x N ( n) IDFS[ X ( k )] N N 1 ~ kn X ( k ) W N k 0 km kn x ( m ) W W N N N k 0 m 1 N 1 1 x( m ) m N N 1 k ( mn) WN k 0 x( n rN ) r Copyright © 2005. Shi Ping CUC The Sampling Theorem in Frequency Domain ~ x N ( n) x( n rN ) r Frequency Sampling Theorem For M point finite duration sequence, if the frequency sampling number N satisfy: NM then ~ x N (n) x N (n) RN (n) x(n) Copyright © 2005. Shi Ping CUC The Sampling Theorem in Frequency Domain Interpolation formula of X (z ) N 1 X (z) n0 1 N 1 N N 1 k 0 N 1 k 0 x(n)z n N 1 1 n0 N N 1 X ( k )W nk N k 0 N 1 nk n 1 X (k ) W N z N n0 N 1 k 0 n z N 1 k 1 X (k ) W N z n0 1 W N Nk z N 1 z N X ( k ) k 1 N 1WN z N 1 X (k ) 1W k 0 n k N z 1 Copyright © 2005. Shi Ping CUC The Sampling Theorem in Frequency Domain X (z) 1 z N k (z) 1 N N 1 X (k ) 1W k 0 1 z N 1W N 1 k N z 1 X ( k ) k ( z ) k 0 N k N z 1 Interpolation function Copyright © 2005. Shi Ping CUC The Sampling Theorem in Frequency Domain Interpolation formula of X (e j ) N 1 X (e jw ) X ( k ) K ( e j N 1 ) k 0 ( ) k 0 sin 2N N sin 2 e j X ( k ) ( 2 k) N N2 1 Interpolation function Copyright © 2005. Shi Ping CUC return Approximating to FT (FS) with DFT (DFS) Approximating to FT of continuous-time aperiodic signal with DFT CTFT X ( j ) x( t ) 1 2 x ( t )e j t dt X ( j )e j t d Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Sampling in time domain t nT , X ( j ) x(t ) 2 x ( nt ) x ( t )e 1 1 2 dt T , j t dt T n dt T x ( nT ) e j nT n X ( j )e S 0 j t X ( j )e d j nT d s 2f s 2 T Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Truncation in time domain t : (0 ~ T0 ) , T0 NT , n : (0 ~ N 1) N 1 X ( j ) T x ( nT ) e j nT n0 x ( nT ) 1 2 S X ( j )e j nT d 0 Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Sampling in frequency domain k 0 , T0 1 F0 0T d 0 , N NT , fs 2 T T0 N 1 N 1 d 0 0 n0 0 2F0 2 T0 2 N X ( jk 0 ) T x ( nT ) e n0 s jk 0 nT N 1 T x ( n )e j 2 nk N n0 T DFT [ x ( n )] Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) 0 x ( nT ) 2 F0 N fs 1 N 1 N N 1 X ( jk 0 ) e k 0 N 1 X ( jk 0 ) e j 2 nk N k 0 N 1 jk 0 nT X ( jk 0 ) e j 2 nk N k 0 f s IDFT [ X ( jk 0 )] demo 1 T IDFT [ X ( jk 0 )] Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Approximating to FS of continuous-time periodic signal with DFS X ( jk 0 ) 1 T0 T0 x ( t )e dt 0 x( t ) jk 0 t X ( jk 0 )e jk 0 t k 0 2F0 2 T0 Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Sampling in time domain t nT , X ( jk 0 ) 1 dt T , T T0 N 1 n0 x ( nT ) e T0 NT jk 0 nT T0 N 1 dt 0 1 N T n0 N 1 j x ( n )e 2 nk N n0 DFS [ x ( n )] N Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Truncating in frequency domain T0 NT , f s NF0 , x(t ) X ( jk 0 ) e let k : (0, N 1) jk 0 t k N 1 x ( nT ) jk 0 nT X ( jk 0 ) e k 0 N 1 N X ( jk 0 ) e j 2 nk N k 0 N 1 N 1 X ( jk 0 ) e j 2 N nk N IDFS [ X ( jk 0 )] k 0 Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Some problems Aliasing Sampling in time domain: fs 2 fh , T 1 1 fs 2 fh Otherwise, the aliasing will occur in frequency domain 1 Sampling in frequency domain: T0 F0 T0 Period in time domain F0 Frequency resolution fs F0 T0 T N f h and F0 is contradictory Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Spectrum leakage x1 ( n), infinite - length sequence x2 ( n) x1 ( n) RN ( n), finite - length sequence X 2 (e j ) X 1 (e j ) W R (e j ) Spectrum extension (leakage) Spectrum aliasing demo Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Fence effect 0 2 N 0 2F0 fs fs , F0 fs N Frequency resolution F0 fs N 1 NT 1 T0 demo Copyright © 2005. Shi Ping CUC Approximating to FT (FS) with DFT (DFS) Comments Zero-padding is an operation in which more zeros are appended to the original sequence. It can provides closely spaced samples of the DFT of the original sequence. The zero-padding gives us a high-density spectrum and provides a better displayed version for plotting. But it does not give us a high-resolution spectrum because no new information is added. To get a high-resolution spectrum, one has to obtain more data from the experiment or observation. demo example Copyright © 2005. Shi Ping CUC return Summary z-transform of x(n) The frequency representations of x(n) Complex frequency domain X (z ) DTFT of x(n) j z e Frequency domain ZT interpolation 2 j k Time z e Nsequence x (n) DTFT X (e j ) 2 DFT of x(n) Discrete frequency domain k DFT N interpolation X (k ) Copyright © 2005. Shi Ping CUC return Illustration of the four Fourier transforms Fourier Transform Signals that are continuous and aperiodic Fourier Series Signals that are continuous and periodic DTFT Signals that are discrete and aperiodic Discrete Fourier Series Signals that are discrete and periodic Copyright © 2005. Shi Ping CUC ~ (m) x 1 n0 n1 n2 n3 n4 n5 n6 0 m ~ (n m) x 2 ~ y ( n) m 0 0 1 2 3 4 5 6 n return Copyright © 2005. Shi Ping CUC xep ( n) 1 2 x(n) x (( N n)) N RN ( n) x (n) 0 5 0 5 0 5 n x(( N n)) N 5 n xep (n) n return Copyright © 2005. Shi Ping CUC xep ( n) x (( N n)) N RN ( n) * ep xep (n) 0 5 0 5 n xep (( N n)) N n RN (n) return Copyright © 2005. Shi Ping CUC 10 (0.8) R11 ( n) n Original sequence x(n) 10 5 0 0 1 2 3 4 5 6 7 8 Circular conjugate symmetric component 9 10 n 0 1 2 3 4 5 6 7 8 Circular conjugate antisymmetric component 9 10 n 0 1 9 10 n xep(n) 10 5 0 xop(n) 4 2 0 -2 -4 2 3 4 5 6 7 8 return Copyright © 2005. Shi Ping CUC Circular even sequence x(n) 10 5 0 0 1 2 3 4 5 6 The DFT of x(n) 7 8 9 10 n X (k ) 40 20 0 0 1 2 3 4 5 6 7 8 9 10 k X (( N k )) N RN (n) 40 20 0 0 1 2 3 4 5 6 7 8 9 10 k return Copyright © 2005. Shi Ping CUC Circular odd sequence x(n) 4 2 0 -2 -4 0 1 2 3 4 5 6 7 The imaginary part of DFT[x(n)] 8 9 10 n 8 9 10 k 8 9 10 k X (k ) 10 0 -10 0 1 2 3 4 5 6 7 X (( N k )) N RN (n) 10 0 -10 0 1 2 3 4 5 6 7 return Copyright © 2005. Shi Ping CUC X ( k ) X (( N k )) N RN ( k ) * X (0) X (( N k )) N RN ( k ) * X ( 0) * k 0 X (0) is a real number if N is even X( N 2 ) X (( N k )) N RN ( k ) X( * k N 2 N X ( * N ) 2 ) is a real number 2 return Copyright © 2005. Shi Ping CUC X ( k ) X (( N k )) N RN ( k ) * X (0) X (( N k )) N RN ( k ) * X ( 0) * k 0 X (0) is an imaginary number if N is even X( N 2 ) X (( N k )) N RN ( k ) X( * k N 2 N X ( * N ) 2 ) is an imginary number 2 return Copyright © 2005. Shi Ping CUC x1 ( n) , x2 ( n) N-point real-value sequences X 1 (k ) DFT[ x1 (n)], X 2 (k ) DFT[ x2 (n)] y( n) x1 ( n) jx 2 ( n) Y ( k ) DFT[ y( n)] DFT[ x1 ( n) jx 2 ( n)] DFT[ x1 ( n)] jDFT[ x 2 ( n)] X 1 ( k ) jX 2 ( k ) X 1 ( k ) DFTRe[ y( n)] Yep ( k ) X 2 ( k ) DFTIm[ y( n)] 1 j Yop ( k ) 1 2 Y (k ) Y 1 2j Y (k ) Y (( N k )) N RN ( k ) (( N k )) N RN ( k ) return Copyright © 2005. Shi Ping CUC Linear convolution Circular convolution N = 6 12 12 10 10 8 8 6 6 4 4 2 2 0 0 0 1 2 3 4 5 6 7 8 9 n x (n) [1,2,2], 1 Circular convolution N=7 0 12 10 10 8 8 6 6 4 4 2 2 0 0 1 2 3 4 5 6 7 8 9 n 2 3 4 5 6 7 8 9 n x2 (n)Circular [1,2 ,3,2], convolution N = 5 12 0 1 0 1 2 3 4 5 6 7 8 9n return Copyright © 2005. Shi Ping CUC ( ) Magnitude Response, N = 8 1 0.8 2 4 N N 0.6 0.4 0.2 0 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 frequency in pi units 0.4 0.6 0.8 1 0.4 0.6 0.8 1 Phase Response 1 pi 0.5 0 -0.5 -1 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 frequency in pi units return Copyright © 2005. Shi Ping CUC X(k),N = 8 6 5 4 3 2 1 0 0 1 2 3 X (0)( ) 4 4 2 6 X (1) ( X ( 3))4( ) X ( 2)N( ) N N 5 6 7 k 2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 frequency in pi units 0.7 0.8 0.9 1 return Copyright © 2005. Shi Ping CUC xa ( t ) 10 (0.8) X a ( j) t 10 50 8 40 FT 6 30 4 20 2 10 0 0 5 10 t 15 20 25 0 -1 X (e x (n) 10 50 8 40 6 DTFT30 4 20 2 10 0 0 5 -0.5 10 15 n 20 25 0 -2 j 0 rad 0.5 1 ) -1 0 pi 1 2 Copyright © 2005. Shi Ping CUC X (e x(n) R11 ( n) j ) R(e j ) 50 10 40 8 30 DTFT 6 20 4 10 2 0 -10 -5 ~ ( n) x N 0 5 10 0 n 50 -2 -1 ~ X N (k ) 0 pi 1 2 10 40 8 DFS 30 6 20 4 10 2 0 -10 0 n 10 0 -10 0 k 10 Copyright © 2005. Shi Ping CUC x N (n) 50 X N (k ) 10 40 8 DFT 6 20 4 10 2 0 30 -10 0 n 10 0 -10 0 k 10 return Copyright © 2005. Shi Ping CUC x1 ( n) 0 X 1 (e n j X 2 (e n ) 0 x2 ( n) 0 R( e n ) 0 RN (n) 0 j 0 j ) return Copyright © 2005. Shi Ping CUC x( n) [1,1,1,1] 4 DTFT DFT 2 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 x( n) [1,1,1,1,0,0,0,0] 4 1.6 1.8 2 pi 1.8 2 pi 1.8 2 pi DTFT DFT 2 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 x( n) [1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0] 4 DTFT DFT 2 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 return Copyright © 2005. Shi Ping CUC signal x(n), x( n) cos( 0.48 n0<=n<=19 ) cos(0.52n) 2 1 0 -1 -2 0 2 4 6 8 10 n 12 14 16 18 20 20 X 20 ( k ) 15 10 5 0 0 0.1 0.2 0.3 0.4 0.5 pi 0.6 0.7 0.8 0.9 1 Copyright © 2005. Shi Ping CUC signal x(n), 0<=n<=19+80 zeros 2 1 0 -1 -2 0 10 20 30 40 50 n 60 70 80 90 100 20 X 100 ( k ) 15 10 5 0 0 0.1 0.2 0.3 0.4 0.5 pi 0.6 0.7 0.8 0.9 1 Copyright © 2005. Shi Ping CUC signal x(n), 0<=n<=99 2 1 0 -1 -2 0 10 20 30 40 50 n 60 70 80 90 100 50 X 100 ( k ) 40 30 20 10 0 0 0.1 0.2 0.3 0.4 0.5 pi 0.6 0.7 0.8 0.9 1 Copyright © 2005. Shi Ping CUC signal x(n), 0<=n<=99+300 zeros 2 1 0 -1 -2 0 50 100 150 200 n 250 300 350 400 60 X 400 ( k ) 40 20 0 0 0.1 0.2 0.3 0.4 0.5 pi 0.6 0.7 0.8 0.9 1 return Copyright © 2005. Shi Ping CUC Suppose F0 10 Hz, Determine T0 , Solution T 1 f h 4 kHz T, N T0 1 F0 1 1 0.1 s 10 1 fs 2 fh 2 4 10 N T0 0.1 T N 2 m 0.125 10 2 10 3 3 0.125 ms 800 1024 return Copyright © 2005. Shi Ping CUC