Lecture on Communication Theory Chapter 2. Representation of signals and systems 2.1 Introduction Deterministic signals : A class of signals where waveforms are defined exactly as function of time. 2.2 F.T. 1) Let g(t) = non-periodic deterministic signal FT of g(t) G(f) = g (t ) exp( j 2ft)dt where f = frequency inverse F.T. g(t) = G( f ) exp( j 2ft)d f 2) To exist FT of g(t), sufficient condition (not necessary) Dirichlet’s conditions g(t) is single-valued, with a finite number of maxima and minima in any finite time interval g(t) has a finite # of discontinuities in any finite time interval g(t) is absolutely integrable, that is, g (t ) dt 3) Physical realizability existence of F.T. 4) All energy signal Fourier transformable 1 Lecture on Communication Theory 2.2 Fourier Transform 1. Notation t : time [sec] f : frequency [Hertz] w= 2f : angular frequency [radians/sec] 식 G(f) = F [g(t)] 식 g(t) = F-1 [G(f)] F, F-1 : linear operator g(t) G(f) 2. Continuous Spectrum 1) Continuous Spectrum G(f) = |G(f)| exp[j(f)] |G(f)| : Continuous amplitude spectrum of g(t) (f) : Continuous phase spectrum of g(t) 2) For real-valued g(t) G(f) = G*(f) = |G(-f)| = | G(f)| (-f) = - (f) Complex conjugation = conjugate symmetric Real : y 축 대칭, symmetric Imaginary : 원점 대칭, anti-symmetric 2 Lecture on Communication Theory Ex1) Rectangular pulse AT A -1/T T/2 -T/2 1/T 0 Def) rect(t) = 1 -1/2 < t < 1/2 0 |t| > 1/2 g(t) = A rect (t/T) sin( fT ) G ( f ) AT fT Def) sinc function sinc( λ) sin( ) G(f) = AT sinc (fT) A rect(t/T) AT sinc (fT) 3) Real Symmetric Real Symmetric FT Ex2) exp(-at) u(t) exp(at) u(-t) 1 a j 2f 1 a j 2f 3 Lecture on Communication Theory 2.3 Properties of the Fourier Transform 1. Linearity (Superposition) Let g1(t) G1(f) and g2(t) G2(f). Then for all constants C1 and C2, we have C1 g1(t) + C2 g2(t) C1 G1(f)+ C2 G2(f) 2. Time Scaling Let g(t) G(f). Then g(at) 1/|a| G(f/a) when a=-1, g(-t) G(-f) 3. Duality If g(t) G(f), then G(t) g(-f) 4. Time shifting If g(t) G(f), then g(t-t0) G(f)exp(-j2ft0) 5. Frequency shifting If g(t) G(f), then exp(j2fct)g(t) G(f-fc) where fc is a real constant modulation theorem 4 Lecture on Communication Theory Ex5) RF pulse g(t) =a rect(t/T)cos(2fct) G(f) = AT/2 {sinc[T(f-fc)] + sinc[T(f+fc)]} |G(f)| g(t) -fc T fc 2 T 6. Area under g(t) If g(t) G(f), then 7. Area under G(f) If g(t) G(f), then g (t )dt G (0) g (0) G( f )df 8. Differentiation in the Time Domain Let g(t) G(f), and assume that the 1st derivative of g(t) is Fourier transformable. Then d g (t ) j 2fG ( f ) dt n-th generalization dn g (t ) ( j 2f ) n G ( f ) n dt Ex6) Gaussian pulse g(t) = exp(-pt2) exp(-pf2) 5 Lecture on Communication Theory 9. Integration in the Time Domain Let g(t) G(f), then provided that G(0)=0, we have g ( )d t 1 G( f ) j 2f 10. Conjugate Functions If g(t) G(f), then for a complex-valued time function g(t), we have g*(t) G*(-f), when the asterisk denotes the complex conjugate operation <Corollary> g*(-t) G*(f) 11. Multiplication in the Time Domain (Multiplication Theorem) Let g1(t) G1(f) and g2(t) G2(f), Then g1(t)g2(t) G ( )G ( f )d 1 2 G1 ( f ) G2 ( f ) 12. Convolution in the Time Domain (Convolution Theorem) Let g1(t) G1(f) and g2(t) G2(f), Then g1(t) g2(t) G1(f)G2(f) 6 Lecture on Communication Theory 2.4 Rayleigh’s Energy Theorem 1. Rayleigh’s Energy Theorem E = g (t ) dt G( f ) df Ex9) E = A 2 2 2 linear transform A 2 f 2 sinc ( 2 Wt ) dt ( ) rect ( )df 2W 2W A 2 W ( ) df W 2W A2 2W 7 Lecture on Communication Theory 2.5 The inverse relationship btw time and frequency 1. Time과 frequency 관계 1) linear transform of FT 2) Time Frequency wide low narrow high cannot specify arbitrary function of time and frequency. 3) Strictly limited in time infinite in freq. infinite in time strictly band-limited in freq. cannot be strictly limited in both time and freq. 2. Bandwidth 1) def) extent of significant spectral content of the signal for signal positive frequencies 2) strictly band-limited 경우가 아닐 경우의 definitions BW = Main lobe bounded by well-defined nulls Base-band Pass-band modulation -1/T 1/T 0 0 BW = 2/T BW = 1/T 2 배 차이 8 fc fc + 1/T Lecture on Communication Theory 3 dB Bandwidth : 1/ 2 of its peak value rms(root mean square) BW Wrms f 2 G ( f ) 2 df G( f ) 2 df 1/ 2 (장점) mathematical evaluation (단점) not easily measurable in lab 3. Time-Bandwidth Product 1) for base-band sync function = (duration T) • (BW of main lobe = 1/T) =1 2) Trms Wrms 1/4 “ = “ when Gaussian pulse 2.6 Dirac Delta Function 1. Definition (t) = 0, t 0 (t )dt 1 (t) is even function of t 2. Shifting property g (t ) (t t )dt g (t ) 0 0 9 Lecture on Communication Theory 3. Replication property g(t) (t) = g(t) g ( ) (t )d g (t ) 4. Fourier Transform F [(t)] = 1 g(t) G(f) 1.0 t f 5. Applications of the Delta Function 1) dc signal 1 (f) exp( j 2ft )dt cos( 2ft ) dt 2) Complex Exponential Function exp(j2fct) (f-fc) 3) Sinusoidal Functions cos (2fct) = 1/2 [exp (j2fct) + exp (-j2fct)] 1/2 [(f-fc) + (f+fc)] f Sin(2fct) 1/2j [(f-fc) - (f+fc)] f 10 Lecture on Communication Theory 4) signum Function sgn[t] = g(t) +1, t > 0 0, t = 0 -1, t < 0 + 1.0 t 0 - 1.0 j 4f a ( 2f ) 2 exp(-|a|t)sgn(t) F[sgn(t)] = lim a 0 |G(f)| 2 j 4f 1 a 2 (2f ) 2 jf 5) Unit step function u(t) = f 0 g(t) 1, t>0 1/2, t = 0 0, t < 0 t |G(f)| u(t) = 1/2 [sgn(t) + 1] u(t) 1/2 1 1 (f ) j 2f 2 f 11 Lecture on Communication Theory 6) Integration in the Time Domain (Revisited) Let y (t ) g ( ) d g ( )u (t ) d g (t ) u (t ) 1 1 Y ( f ) G ( f )[ ( f )] j 2f 2 1 1 Y ( f ) G ( f ) G (0) ( f ) j 2f 2 t 12 Lecture on Communication Theory 2.7 Fourier Transforms of periodic signals Let a periodic signal be gTo(t) with period T0 complex exponential Fourier series g T (t ) cn exp( j 2nf 0t ) 0 1 n cn 1 T /2 g To (t ) exp( j 2nf0 t )dt T0 T / 2 2 0 0 where f0 = 1/T0 ; fundamental freq. Generating function g(t) gTo(t) , - T0/2 t T0/2 0, elsewhere g(t) = from cn f 0 g (t ) exp( j 2nf 0t )dt f 0G (n f 0 ) 2 m n 3 g T 0 (t ) g (t mT0 ) f 0 G (nf 0 ) exp( j 2nf 0t ) Use 1 3 Poisson’s sum formula g (t mT ) f G (nf ) ( f nf ) m 0 0 n 0 0 where G(f) g(t) (Observations) periodicity in time domain discrete spectrum defined at nf0 13 Lecture on Communication Theory ex11) (t mT ) f ( f nf ) m 0 0 n 14 0 Lecture on Communication Theory g(t) 0 sampling 1 T (t ) T 0 T/2 T 2 G(f)* 2 ( f ) T 0 2/T 0 2/T 2(f) T G(f) sampling 2 T 0 T/8 2 RECT 2 ( f ) T 0 2/T 0 2/T G(f)* RECT 2 ( f ) T <HW> 2.1, 2.4, 2.8, 2.17, 2.18 15 t t f f f 0 rectT (t ) 2T t f f Lecture on Communication Theory 2.8 Transmission of signals through linear system Def) linear system : a system which holds the principle of superposition x1(t) h(t) y1(t) x2(t) h(t) y2(t) c1x1(t) + c2x2(t) h(t) c1y1(t) + c2y2(t) 1. Time Response 1) Impulse Response : the response of system to (t) (t) h(t) h(t) h(t) y(t) 2) Convolution x(t) y(t) = x(t) h(t) h(t) x(t) x( )h(t )d h( ) x(t )d excitation time response time t system - memory time t- ; a weighted integral over the past history of the input signal, weighted according to the impulse response of the system. 16 Lecture on Communication Theory ex) x(KT) 0 1 2 3 KT 1 h(KT ) h(KT) -1 y (2T ) x( KT )h(2T KT )dk 0 1 2 KT 2 h( KT ) -2 -1 0 1 KT 3 KT 4 KT 5 KT 6 KT 7 h(2T KT ) -4 -3 -2 -1 y (2T ) 1 4 0 0 h(T KT ) y (T ) y ( 0) y (T ) 1 1 1 5 6 7 1 3 4 y (2T ) 5 y (3T ) 4 KT y (4T ) 2 y (5T ) 1 KT y(kt) y (6T ) 0 -1 0 17 1 2 34 5 KT Lecture on Communication Theory t T T * * t T -2T 2T t t T x(t ) y (t )dt x(t) 1 y(t) T t * * correlation t 남자 여자 0 -1 18 1 0 -1 2T Lecture on Communication Theory ex12) Tapped-Delay-line Filter assumptions) 1 h(t) = 0 for t < 0 2 h(t) = 0 for t Tf y (t ) 0T h( ) x(t )d f sampling with , t=n, =k N 1 y(n ) h(k ) x(n k ) k 0 where N = Tf let wk = h(k ) N 1 y(n ) = wk x(n k ) k 0 = w0x[n ] + w1 x[n - ] + ..... + wN-1 x[n -(N-1) ] 19 Lecture on Communication Theory 2. Causality & stability 1) Causal if a system does not respond before the excitation is applied h(t) = 0, t < 0 causality Real time으로 동작 하는 system must be causal Memory 기능이 있는 것 can be non-causal 2) Stable if the output signal is bounded for all bounded input signals (BIBO) cf, BIBO: bounded input-bounded output BIBO stability criterion BIBO stability h(t ) dt 3. Frequency Response 1) x(t) y(t) h(t) y(t) = x(t) h(t) Y(f) = X(f)H(f) 20 Lecture on Communication Theory 2) H(f)의 (freq. domain에서의) 표현 H(f) = |H(f)| exp(j(f)) |H(f)| : Amplitude response (f) : phase response h(t)가 real H(f) : conjugate symmetry |H(f)| = |H(-f)| even (f) = - (-f) odd polar form ln H(f) = ln |H(f)| + j (f) = a(f) + j (f) a(f):gain[nepers] (f):[radians] ) a’(f) = 20 log10 |H(f)| [dB] decibels a’(f) = 8.69 a(f) 1 neper = 8.69 dB 3) BW - 3 dB BW B -B -fc 0 fc-B fc+B 4. Paley-Wiener Criterion : freq. domain equivalent of causality requirement a(f) : gain of causal filter a( f ) 1 f 2 df 21 Lecture on Communication Theory 2.9 Filters 1. Ideal Low Pass Filter 1, -B f B 0, |f| > B h(t) = 2B sinc[2B(t-t0)] (f) = -2f t0 : linear phase |H(f)| = t0 =0 1/B t0 >0 t0 To make a causal filter, | sinc[2B(t-t0)] | << 1 for t < 0 If making digital filter, non-causal is O.K. 2. computer experiment I pulse response of ideal LPF X(t) -T/2 T/2 H(f) B -B f B=5/T, BT=5 BT=10 1/T 10/T 5/T 22 Lecture on Communication Theory (observations) overshoot = 9 % overshoot is independent of B frequency of ripple is B BT 5 10 20 100 Gibb’s phenomenon Ringing oscillation freq. 5 Hz 10 Hz 20 Hz 100 Hz overshoot (%) 9.11 8.98 8.99 9.63 3. Fig 27. B = 1 Hz 일 때 f0 = 0.1(T=5), 0.25(T=2), 0.5(t=1), 1(T=0.5) Hz 입력이 들어갔을 때의 output conclusion) BT 1 이 되어야 recognizable output이 나온다 4. Design of Filters 1) Basic Design steps approximation of a prescribed frequency response by a realizable transfer function Realization of the approximating transfer function by a physical device 23 Lecture on Communication Theory 2) Stable system : BIBO complex frequency s = j2f plane 상에서 표시 H’(s) is a rational function H’(s) = H(f) | j2f=s = k ( s z1 )( s z 2 )...( s z m ) ( s p1 )( s p2 )...( s pn ) SI z1 z2, ... zm : zeros p1 p2, ... pn : poles SR stability Re[pi] < 0 for all i Minimum-phase systems Re[pi] < 0, Re[zi] < 0 Non-minimum phase systems Re[pi] < 0, - Re[zi] 3) Butterworth filters Chebyshev filters 4) Implementation of filters Analog filters : (a) L, C (b) C, R, OP-amp Discrete-time filters : switched-capacitor filters (SAW) surface accoustic wave filters Digital filters : FIR, IIR (장점) programmable, flexibility 24 Lecture on Communication Theory 2.10 Hilbert Transform 1. Frequency selective filters Phase selective filters 180º : ideal transformer ± 90 º : Hilbert Transform 2. Def. of Hilbert Transform 1 g ( ) gˆ (t ) d t 1 gˆ ( ) g (t ) d t sgn(f) 1 And 1/t -jsgn(f) f -1 Gˆ ( f ) j sgn( f )G( f ) 3. Applications 1) SSB Modulation 2) Mathematical basis for the representation of band-pass signals 25 Lecture on Communication Theory ex13) g (t ) cos( 2f c t ) 1 G ( f ) ( f f c ) ( f f c ) 2 ˆ G ( f ) j sgn( f )G ( f ) j ( f f c ) ( f f c )sgn( f ) 2 1 ( f f c ) ( f f c ) 2j gˆ (t ) sin( 2f c t ) sin( 2f c t ) Hillbert cos( 2f c t ) 4. Properties of Hilbert transform Assumption) g(t) is real-valued property 1) g (t ) & gˆ (t ) have the same amplitude spectrum 2) if g (t ) Hillbert gˆ (t ) then gˆ (t ) Hillbert - g (t ) pf ) j sgn( f ) 2 1 3) g(t) and gˆ(t) are orthogonal pf) G(f) f x0 jG(f) f 26 Lecture on Communication Theory 2.11 Pre - envelope 1. Pre-envelope for positive frequency. For a read-valued signal g(t) pre-envelope of g(t) g (t) g(t) jĝ(t ) G ( f ) g( f ) j j sgn( f )G( f ) FT. , f 0 , f 0 , f 0 2G( f ) G( 0 ) 0 G(f) -W W G+(f) f W -W 2. Pre-enveloped for negative frequency. g (t) g(t) jĝ(t ) FT. g (t) g (t) G ( f ) g( f ) j j sgn( f )G( f ) 0 G( 0 ) 2G( f ) 27 , f 0 , f 0 , f 0 f Lecture on Communication Theory G(f) -W W G-(f) f -W W f Pre-envelope for negative freq. 3. 용도: Useful in handling band-pass signals and systems 2.12 Canonical Representations of Band-Pass signals 1. Base-band 신호와 Pass-band신호의 관계 Base-band signal ( 복소수 ) : complex envelope g I (t) : In - phase component g~ (t ) g I (t ) jg Q (t ) g Q(t) : Quadrature component ; low - pass signal Pass-band signal ( 실수 ) : band-pass signal, narrow-band signal ~(t ) g Re( • ) exp( j 2f c t ) ~ (t ) exp( j 2f t ) g (t ) Reg c g I (t ) cos( 2f c t ) g Q (t ) sin( 2f c t ) ; canonical form 28 g (t ) Lecture on Communication Theory Pre-envelope of g(t) ~ (t ) exp( j 2f t ) g (t ) g c ~ G( f ) w w f G( f ) fc fc BW 2 W BW W physical line 2 line(복소수) G ( f ) 1 line(실수) spectral efficiency is same 29 f fc BW 2 W Lecture on Communication Theory 2. Physical Implementation of pass-band signal 30 Lecture on Communication Theory 3. Expression in the polar form g~(t ) a(t ) exp( j (t )) : base - band a(t), (t) : real Pass-Band g (t ) a(t ) cos[ 2f c t (t )] a(t ) : natural envelope amplitude mod (t ) : phase angle mod . 4. Envelope 용어 정리 a(t) : envelope of g(t) ~ g (t) : complex envelope g (t) : pre - envelope a(t) g~ (t ) g (t) 31 Lecture on Communication Theory t ~ ex14) g (t ) A rect( ) T t g (t ) A rect( ) cos( j 2f c t ) T t g (t ) A rect( ) exp( j 2f c t ) T ~ (t ) A rect( t ) a (t ) g T ~(t ) g g+(t)real t -T/2 T/2 g(t) g +(t)image ~ G( f ) G ( f ) G( f ) f 0 -fc 0 0 fc 32 fc Lecture on Communication Theory 2.13 band-pass systems 1.Pass-band 신호를 Base-band에서 표현 Base - band Pass - band ~ Input x (t ) X ( f ) x(t ) Re ~ x (t ) exp( j 2f ct ) ~ ~ Filter h (t ) H ( f ) h(t ) Re h (t ) exp( j 2f ct ) Output ~ y (t ) Y ( f ) y(t ) Re ~y (t ) exp( j 2f ct ) Pass - band Base - band y (t ) x(t ) h(t ) ~ 2~ y (t ) ~ x (t ) h (t ) 33 Lecture on Communication Theory 2. Complex Filter의 구현 ~ x ( t ) xI ( t ) j xQ ( t ) ~ h ( t ) hI ( t ) j hQ ( t ) ~y ( t ) y ( t ) j y ( t ) I Q ~ x( t ) h ( t ) 2 ~y ( t ) ~ xI ( t ) j xQ ( t )hI ( t ) hQ ( t ) xI ( t ) hI ( t ) xQ ( t ) hQ ( t ) jxI ( t ) hQ ( t ) xQ ( t ) hI ( t ) 34 Lecture on Communication Theory 3. Band-pass system의 response를 구하는 과정 summary 1) x(t ) ~ x (t ) ~ 2) h(t ) h (t ) ~ 3) y (t ) ~ x (t ) h (t ) 4) y (t ) Re~ y (t ) exp( j 2f t ) c 결론 Pass - band filtering 은 (Base - band filtering Modulation ) 4. Computer Experiment II. Response of Ideal Base-pass Filter to a pulsed RF Wave BT=5일 경우의 RF파형 Base-band 파형 + Modulation 35 Lecture on Communication Theory 2.14 Phase and Group Delay A dispersive channel in a polar form H ( f ) K exp j ( f ) ( f ) : nonlinear function of frequency input signal x(t ) x(t ) m(t ) cos( 2f c t ) By expanding ( f ) in a Taylor series ( f ) ( fc ) ( f fc ) Define p ( f ) f ff c ( fc ) 2f c 1 ( f ) g 2 f f fc ( f ) 2f c p 2 ( f f c ) g H ( f ) K exp j 2f c p j 2 ( f f c ) g Equivalent low - pass filter ~ H ( f ) K exp j 2f c p 2f g 36 Lecture on Communication Theory Recevied signal ~ ~ y (t ) Y ( f ) 1 ~ ~ ~ Y( f ) H ( f ) X ( f ) 2 K exp j 2f c p exp j 2f g M ( f ) ~ y (t ) K exp j 2f m(t ) c Finally p g y (t ) Re~ y (t ) exp j 2f c t Km(t g ) cos2f c (t p ) Conclusion 1) p 2) g ( fc ) ; phase delay, carrier delay 2f c 즉, carrier is delayed by p . 1 ( f ) 2 f ; envelope delay, group delay f fc 즉, m(t ) is delayed by g (해석) if g constant no ploblem if g function of f group delay variation 을 규정해야 한다. 37 Lecture on Communication Theory 2.15 Numerical Computation of the Fourier Transform 1. DFT & IDFT 1) Nyquist Sampling Theorem Sampling Rate should be greater than twice the highest frequency component of the input signal to avoid aliasing. t fs 2 T fs 2 T=NTs fs fs fs=Nf 2) DFT & IDFT {g0,g1,•••,gN-1} Given finite data sequence ex) analog gn=g(nTs) DFT of gn N 1 Gk g n exp( n 0 IDFT of Gk j 2 kn) where k=0,1,•••,N-1 N 1 N 1 j 2 g n Gk exp( kn) where n=0,1,•••,N-1 k 0 N N {G0,G1,•••,GN-1} : transform sequence k=0,•••, N-1 : frequency index 38 Lecture on Communication Theory 3) Physical meaning of DFT N 1 G0 g n DC n 0 2n ) n 0 N N 1 2 2n G0 g n exp( ) n 0 N N 1 G0 g n exp( 1cycle/N 2cycle/N ••• ••• 2. Interpretation of the DFT and the IDFT. DFT & IDFT: Gk and gn must be periodic. DFT & IDFT: are linear. 39 Lecture on Communication Theory 3.FFT Algorithms. DFT of gn N 1 Gk g nW nk n 0 where (1) k = 0, 1, 2, , , , , , ,N-1 j 2 W exp( ) N W n 1, W N / 2 1 W ( k lN )( n mN ) W kn m,l 0,1,2, i.e. Wkn : periodic with period N. Assume N=2L, where L is integer N 1 2 (1) N 1 Gk g nW g nW nk nk n 0 n N 1 2 N 1 2 n 0 n 0 N 2 g nW nk g N W N 1 2 n k k n 2 2 kN 2 ( g n g N W )W kn n 0 n 2 k 0,1, , N 1 N 1 2 ( g n (1) k g n 0 N n 2 40 )W kn Lecture on Communication Theory 1 N k even , i.e, k 2l , l 0,1, 1 2 N 2 1 G2 l ( g n g n 0 n 2 ln )( W ) N 2 N 1 2 xn (W 2 ) ln ; N po int 2 FFT 2 n 0 k odd , i.e, k 2l 1, l 0,1, N 1 2 N 1 2 G2 l 1 ( g n g n 0 n ( 2 l 1) n ) W N 2 N 1 2 ( y nW n )(W 2 ) ln n 0 N point FFT = 2 [ ; N -point 2 41 N/2-point FFT FFT] Lecture on Communication Theory ex) Fig 2.38 ex) Fig 2.38 Butterfly. 42 Lecture on Communication Theory # of computation of DFT # of computation of FFT N2complex(x)+N(N-1)complex(+) N log 2 N if decimation in freq. 2 one complex(x) + 2complex(+) ex) N=1024=1k DFT 1M 200배 차이 FFT 5K 단 FFT는 2N개로 해야 함 Other algorithm : Decimation-in-time algorithm N log2N computations 4. Computation of IDFT 1 N 1 g n GkW kn N k 0 N 1 Ng n Gk W kn n 0,1, , N 1 k 0 GK Ng n complex Ng n complex GK FFT conjugate conjugate where : complex conjugate 43 N qn Lecture on Communication Theory <HW1> 2.30, 2.32 문제) N=16(g0~g15)일 때의 Decimation-in-freq. Algorithm을 그림으로 그려라. <Computer HW> 2.1 (see book) Bit reversed index. 44