Lecture 2 Signals and Systems (II) Principles of Communications Fall 2008 NCTU EE Tzu-Hsien Sang 1 Outlines • • • • • • • Signal Models & Classifications Signal Space & Orthogonal Basis Fourier Series &Transform Power Spectral Density & Correlation Signals & Linear Systems Sampling Theory DFT & FFT 2 Examples • Symmetry Properties of x(t) and Its Fourier Function * X X For real periodic x(t), n n For real aperiodic x(t), X ( f ) X * ( f ) 3 • Fourier Transform of Singular Functions (t ) is not an energy signal (hence doesn’t satisfy Dirichlet condition). However, its FT can be obtained by formal definition. FT FT (t ) 1, 1 ( f ), j 2f0 A (t t0 ) Ae FT • Example: The FT of , Ae jf0t FT A ( f f 0 ), (t nT ) ? n 0 4 • Fourier Transform of Periodic Signals— Periodic signals are not energy signals (don’t satisfy Dirichlet’s conditions). But we are doing it anyway (at least formally)… • Given a periodic signal jn 0t X ( f ) X n ( f nf0 ) x(t ) X n e n n 2f 0t • Example-1: cos • Example-2: (t nT0 ) n (A pulse train! What good are they for?) 5 Note: This table uses “” instead of “f”. But it doesn’t hurt the fundamental facts. 6 7 Transform Pairs (There is something nice to know in life…) 8 9 10 • Let FT of an aperiodic pulse signal p(t) be { p(t )} P( f ) • We can generate a periodic signal x(t) by duplicating p(t) at every interval Ts, then x(t ) [ (t nT )] * p(t ) p(t nT ) s s n n • From convolution theorem, X ( f ) {[ (t nT )]} P( f ) s n fs ( f nf ) P( f ) f P(nf ) ( f nf ) s n s n s s 11 Taking inverse FT of the eq. on previous page. 1 { X ( f )} x(t ) s n 1 f P ( nf ) s s { ( f nfs )} n p(t nTs ) n p(t nT ) { f P(nf ) ( f nf )} 1 n s s s j 2nf st f P ( nf ) e s s n j 2nf s t f P ( nf ) e s s n Poisson sum formula 12 Power Spectral Density & Correlation • Why should we care about the “frequency components” of a signal? • For energy signals: ( ) 1{G( f )} 1[ X ( f ) X * ( f )] 1[ X ( f )] 1[ X * ( f )] T x( ) x( ) x( ) x( )d lim x( ) x( )d T T • The time-averaged autocorrelation function • The squared magnitude of the FT represents the “energy” distributed on the frequency axis. (0) E signal energy. 13 • For power signals: R( ) x( ) x* (t ) 1 T * lim x(t ) x (t )dt, if aperiodicpowersignal T 2T T 1 * if periodicpowersignal x(t ) x (t )dt, T0 T0 R(0) S ( f )df S ( f ) {R( )} “Power spectral density function” • For periodic power signals: S ( f ) {R( )} | X n |2 ( f nf0 ) n 14 • The functions () and R() measure the similarity between the signal at time t and t+. • G(f) and S(f) represents the signal energy or power per unit frequency at freq. f. 2 R ( 0 ) power x (t ) R( ) , , max{R( )} R(0). • • R() is even for real x(t): R( ) x(t ) x* (t ) R( ). • If x(t) does not contain a periodic component: 2 lim R( ) x(t ) . | | • If x(t) is periodic with period T0, then R() is periodic in with the same period. • S(f) is non-negative. S ( f ) {R( )} 0, f 15 • Cross-correlation of two power signals: Rxy ( ) x(t ) y * (t ) x(t ) y * (t ) 1 lim T 2T T T x(t ) y * (t )dt • Cross-correlation of two energy signals: xy ( ) x(t ) y* (t )dt • Remarks: Rxy ( ) R ( ), xy ( ) * yx * yx ( ) 16 Signals & Linear Systems x (t ) Η y (t ) y (t ) H {x(t )} • The standard input/output black box model for linear systems. Q: Why does it work? • Linear: Satisfies superposition principle y(t ) H{1 x1 (t ) 2 x2 (t )} H{1 x1 (t )} H{ 2 x2 (t )} y1 (t ) y2 (t ) • Time-invariant: Delayed input produces an output with the same delay. H {x(t t0 )} y(t t0 ) 17 Describing LTI Systems with Impulse Responses • Let h(t) be the impulse response: h(t ) H { (t )}. x(t ) x( ) (t )d y(t ) H {x(t )} H { x( ) (t )d} x( )H { (t )}d If time-invariant, y(t ) H {x(t )} x( )h(t )d x(t ) * h(t ) H {x(t t0 )} x( )h(t t0 )d y(t t0 ) 18 Note: This example is a linear, but not time-invariant system. 19 • The convolution form holds iff LTI. • Duality of signal x(t) & system h(t): y(t ) x( )h(t )d h( )x(t )d • The Convolution Theorem: { y(t )} Y ( f ){ h( )x(t )d} H ( f ) X ( f ) Key application: generally H ( f ) X ( f ) is easier 20 than x(t ) h(t ) …