Fourier Analysis of Discrete Time Signals For a discrete time sequence we define two classes of Fourier Transforms: • the DTFT (Discrete Time FT) for sequences having infinite duration, • the DFT (Discrete FT) for sequences having finite duration. The Discrete Time Fourier Transform (DTFT) Given a sequence x(n) having infinite duration, we define the DTFT as follows: X ( ) DTFT x (n) j n x ( n ) e n 1 x (n) IDTFT X ( ) 2 X ( )e j n d X ( ) x ( n) ….. ….. N 1 n continuous frequency discrete time Observations: • The DTFT X ( ) is periodic with period 2 • The frequency ; is the digital frequency and therefore it is limited to the interval Recall that the digital frequency is a normalized frequency relative to the sampling frequency, defined as 2 F Fs one period of Fs 2 Fs / 2 0 0 X ( ) Fs / 2 Fs 2 F Example: DTFT x[n] 1 0 n N 1 since N 1 X ( ) e n 0 j n 1 e j N 1 e j e j ( N 1) / 2 sin N / 2 sin / 2 Example: x[n] A cos( 0n ) X ( ) A e j ( 0) A e j ( 0) Discrete Fourier Transform (DFT) Definition (Discrete Fourier Transform): Given a finite sequence x [ x(0), x(1),...,x( N 1)] its Discrete Fourier Transform (DFT) is a finite sequence X DFT( x) [ X (0), X (1),...,X ( N 1)] where N 1 X ( k ) x(n) wN kn , wN e j 2 / N n0 x DFT X Definition (Inverse Discrete Fourier Transform): Given a sequence X [ X (0), X (1),...,X ( N 1)] its Inverse Discrete Fourier Transform (IDFT) is a finite sequence x IDFT( X ) [ x(0), x(1),...,x( N 1)] where 1 x ( n) N N 1 X (k )w k 0 X kn N , wN e j 2 / N IDFT x Observations: • The DFT and the IDFT form a transform pair. back to the same signal ! x DFT x IDFT X X • The DFT is a numerical algorithm, and it can be computed by a digital computer. DFT as a Vector Operation Let x[0] x[1] , x x [ N 1 ] X [0] X [1] X DFT{x} X [ N 1 ] 1 wk ek N , k ( N 1) wN Then: x X [k ] ek*T x x 1 X [0]e0 X [1]e1 ... X [ N 1]eN 1 N ek 1 X [ k ]ek N 1 1 x[0] X [0] 1 x[1] X [1] 1 w N 1 w N N X DFT{x} ( N 1)( N 1) N 1 w w 1 ] 1 N [ x ] 1 N [ X N N WN X WN x 1 *T x WN X N WN1 Periodicity: From the IDFT expression, notice that the sequence x(n) can be interpreted as one period of a periodic sequence x p (n) : 1 x p (n) N N 1 X (k )w k 0 kn N 1 N N 1 X (k )w k 0 kn N wN kN 1 N x ( n) k 0 k ( n N ) N x p (n N ) n x p (n) N X (k )w original sequence N 1 2N N 1 periodic repetition N 2N n This has a consequence when we define a time shift of the sequence. For example see what we mean with x (n 1) . Start with the periodic extension x p (n) x p (n) A B D N N C n x p (n 1) A D B N N C n If we look at just one period we can define the circular shift x ( n) x(n 1) N A A B D B D C n C A B C D D A B C 0 1 2 3 0 1 2 3 D Properties of the DFT: • one to one • time shift where x(n) X ( k ) with no ambiguity; DFT x(n m) N wN km X ( k ) x(n m) N is a circular shift periodic repetition x ( n) x ( N m) x ( N 1) x(0) x(1) x( N m) x( N m 1) x( N 1) x(n m) N x ( N m) x ( N 1) x (0) x (1) x ( N m 1) X ( k ) X ( N k) • real sequences | X ( k )| | X ( N k )| • circular convolution y (n) x1 (n) x2 (n) N 1 x1 ( k ) x2 (n k ) N k 0 circular shift where both sequences x1 , x2 length N. Then: must have the same DFT x1 (n) x2 (n) X1 ( k ) X 2 ( k ), k 0,..., N 1 Extension to General Intervals of Definition Take the case of a sequence defined on a different interval: x[n] n0 n0 N 1 How do we compute the DFT, without reinventing a new formula? First see the periodic extension, which looks like this: x[n] n0 N 1 n0 Then look at the period n 0 n N 1 x[n] n0 n0 N 1 N 1 n Example: determine the DFT of the finite sequence x[n] 0.8|n| if 3 n 3 x[n] Then take the DFT of the vector 3 3 x x[0], x[1],...,x[3], x[3],...,x[1] n x[n] 3 n