Professional Documents
Culture Documents
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
Contents
2.1 Introduction 2 2 The Fo 2.2 Fourier rier Transform 2.3 Properties of The Fourier Transform 2 4 The Inverse Relationship between Time and Frequency 2.4 2.5 Dirac Delta Function 2.6 Fourier Transform of Periodic Signals 2.7 Transmission of Signals Through Linear Systems 2.8 Filters 2.9 Low-Pass and Band-Pass Signals 2.10 Band-Pass Systems 2.11 Phase and Group Delay 2.12 Sources of Information 2.13 Numerical Computation of the Fourier Transform
2
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
We identify deterministic signals as a class of signals whose waveforms f are d defined fi d exactly tl as functions f ti of f time. ti In this chapter we study the mathematical description of such signals using the Fourier transform that provides the link between the time-domain and frequency-domain descriptions of signal. Another A h related l d issue i that h we study d in i this hi chapter h is i the h representation of linear time-invariant systems. Filters of different kinds and certain communication channels are important examples of this class of systems.
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
g(t) denote a nonperiodic deterministic signal. j = 1. variable f denotes frequency and t denotes time.
We have used a lowercase letter to denote the time function and a uppercase letter to denote the corresponding frequency function. function The functions g(t) and G( f ) are said to constitute a Fourier-transform pair. For the Fourier transform of a signal g(t) to exist, it is sufficient, but not necessary, that g(t) satisfies three sufficient conditions known collectively as Dirichlets conditions: The function g(t) is single-valued, with a finite number of maxima i and d minima i i in i any finite fi it time ti interval. i t l The function g(t) has finite number of discontinuities in any finite time interval. interval The function g(t) is absolutely integrable, i.e.
g ( t ) dt <
7
We may safely ignore the question of the existence of the Fourier transform of a time function when it is an accurately specified description of a physically realizable signal. Physical y realizability y is a sufficient condition for the existence of a Fourier transform.
2 1 T P = lim g ( t ) dt T T All energy signals i l are Fourier F i transformable. t f bl 2T Power Signal:0 < P < 2
Energy gy signals g :
g ( t ) dt < ; ( P = 0 )
( E = )
Plancherels theorem: if a time function g(t) is such that the value 2 of the energy g ( t ) dt < is defined and finite, finite then the Fourier transform G( f ) of the function g(t) exists and 2 A li g ( t ) G ( f ) exp ( j 2 ft lim f ) df dt d = 0. 0 A A
8
Notations time t measured in second (s) frequency f measured in Hertz (Hz) angular l frequency f second rad/s). rad/s) = 2 f (radians per second, A convenient shorthand notation for the transform relations: Fourier F i transformation t f ti
G( f ) = F g ( t )
Continuous Spectrum
By using the Fourier transform operation, operation a pulse signal g(t) of finite energy is expressed as a continuous sum of exponential q in the interval - ~ . The functions with frequencies amplitude of a component of frequency f is proportional to G( f ), where G( f ) is the Fourier transform of g(t). At any frequency f, the exponential function exp(j2 ft) is weighted by the factor G( f )df , which is the contribution of G( f ) i in an infinitesimal i fi it i l interval i t l df centered t d at t the th frequency f f. We may express the function g(t) in terms of the continuous sum of such infinitesimal components:
g ( t ) = G ( f ) exp ( j 2 ft ) df
10
If g(t) is a real-valued function of t, then G ( f ) = g (t ) e dt G(( f )=G*( f ) G * ( f ) = ( g (t ) e dt ) |G(-f )|=|G( f )|: an even function of f = g (t ) e dt = G ( f ) (( f )=)= ( f ): an odd function of f The spectrum of a real-valued signal exhibits conjugate symmetry.
j 2 ft j 2 ft f * j 2 ft
11
Define a rectangular function of unit amplitude and unit 1 1 duration: 1, < t < 2 2 rect ( t ) =
Real-Valued and Symmetric
0,
1 2
U AT sinc ( fT ) (2.10)
T /2
A ( cos(2 ft ) j sin(2 ft ) ) dt
T /2
= 2 A
cos(2 ft ) dt = 2 A
sin i ( 2 ft f ) 2 f
T 2
sin ( fT ) = AT ATsinc ( fT ) fT
12
sinc ( )
sin i ( )
sinc function
sinc ( )
sin ( )
As the pulse duration T is decreased, decreased the first zero-crossing zero crossing of the amplitude spectrum |G( f )| moves up in frequency. The relationship between the time time-domain domain and frequency frequency-domain domain is an inverse one. A pulse, narrow in time, has a significant frequency description over a wide range of frequencies, and vice versa.
13
14
It is i convenient i t to t mathematically th ti ll define d fi the th decaying d i exponential pulse using the unit step function. An unit step function is defined as:
1, t > 0 1 u (t ) = , t = 0 2 0 t < 0
= =
exp e p t ( a + j 2 f ) exp t a + j 2 f dt = ( ) ( a + j 2 f )
1 a + j 2 f
the Fourier-transform pair for the decaying exponential pulse of figure (a) is therefore
exp ( at ) u ( t ) U
16
1 a + j 2 f
G( f ) =
1 = exp = t a j 2 f dt ) ( a j 2 f The decaying and rising exponential pulses are both asymmetric functions of time t . Th i Fourier Their F i transforms f are therefore h f complex l valued. l d Truncated decaying and rising exponential pulses have the same amplitude spectrum, spectrum but the phase spectrum of the one is the negative of that of the other.
17
18
Chapter 2.3
1 f g ( at ) U G a a
where a is constant
20
g ( t ) dt = G ( 0 )
Mathematical Description g ( 0 ) = G ( f ) df
d g ( t ) U j 2 fG ( f ) dt
g ( ) d U
G ( 0) 1 G( f )+ (f) j 2 f 2
If g ( t ) U G ( f ) then g ( t ) U G ( f )
g1 ( t ) g 2 ( t ) U G1 ( )G ( f ) d
g1 ( )g 2 ( t ) d U G1 ( f ) G2 ( f )
g ( t ) dt = G ( f ) df
2
21
Proof: the proof of this property follows simply from the linearity of the integrals defining G( f ) and g(t).
22
This pulse Thi l may be b viewed i d as the th sum of f a truncated t t d decaying d i exponential pulse and a truncated rising exponential pulse. 1 1 2a G( f ) = + = 2 a + j 2 f a j 2 f a + ( 2 f )2 2a exp ( a t ) U 2 2 a + ( 2 f )
23
+1, sgn ( t ) = 0, 1,
t >0 t =0 t<0
g ( t ) = exp ( a t ) sgn ( t )
24
The Fourier transform is odd and purely imaginary. In general, a real odd-symmetric time function has an odd and purely l imaginary function f as its Fourier transform f .
25
Compression of a function in the time domain is equivalent to the expansion of its Fourier transform in the frequency domain, or vice versa.
ProofF g ( at ) =
g ( at ) exp ( j 2 ft ) dt
= at t =
Property 3Duality
If g ( t ) U G ( f ), then G ( t ) U g ( f )
Proof G ( f ) = g (t )e j 2 ft dt
g (t ) = G ( f )e j 2 ft df
7 t f
f t g ( f ) = G (t )e j 2 ft dt
g ( f ) = G (t )e j 2 ft dt = F{G (t )}
[Example 2.4]
t A rect T
Proof Let = ( t t0 )
F g ( t t0 ) = g ( t t0 ) exp ( j 2 ft ) dt
The amplitude of G( f ) is unaffected by the time shift, but its phase is changed by the linear factor -2ft0.
28
Proof
F g ( t ) exp exp ( j 2 f c t ) g ( t ) = j 2 t ( f f c ) dt = G ( f fc )
Multiplication of a function by the factor exp(-2fct) is equivalent q to shifting g its Fourier transform in the positive p direction by the amount fc.
29
Consider C id the th pulse l signal i l g ( t ) shown h in i figure fi (a) ( ) which hi h consists i t of f a sinusoidal wave of amplitude A and frequency f c , extending in duration from t = -T/2 to t = T/2. /2 This signal is sometimes referred to as an RF pulse when the frequency f c falls in the radio-frequency band. The signal g ( t ) of figure (a) may be expressed mathematically as follows
[Example 2.5]
t g ( t ) = A rect T
cos ( 2 f c t )
30
[Example 2.5]
we note that 1 cos ( 2 f c t ) = exp ( j 2 f c t ) + exp ( j 2 f c t ) 2 applying the frequency-shifting frequency shifting property to the Fourier-transform Fourier transform pair, we get the desired result AT + sinc T ( f + f c ) G( f ) = sinc T ( f fc ) 2
in the special case of fcT>>1, >>1 we may use the approximate result AT T ( f f c ) , 2 sinc G ( f ) 0, AT sinc T ( f + fc ) , 2
31
f >0 f =0 f <0
[Example 2.5]
The amplitude Th lit d spectrum t of f the th RF pulse l is i shown h in i figure fi (b). (b) This Thi diagram, in relation to figure in page 12, clearly illustrates the frequency-shifting frequency shifting property of the Fourier transform. transform
32
If g ( t ) U G ( f ), then g ( 0 ) = G ( f ) df That is, is the value of a function g(t) at t =0 0 is equal to the area under its Fourier-transform G( f ). This result can be obtained by putting t =0 in the formula of inverse Fourier transform. g ( t ) = G ( f ) exp ( j 2 ft ) df
33
Let L t g ( t ) U G ( f ) , and d assume that th t the th first fi t derivative d i ti of f g(t) i is Fourier transformable. Then
d (2.31) g ( t ) U j 2 f G ( f ) dt Th t is, That i differentiation diff ti ti of f a time ti function f ti g(t) h has th the effect ff t of f multiplying its Fourier transform G( f ) by the factor j2f. If we assume that the Fourier transform of the higher higher-order order derivative exists, then dn n U 2 g t j f () ( ) G( f ) n dt ProofThis result is obtained by taking the first derivative of both sides of the integral g defining g the inverse Fourier transform.
g ( t ) = G ( f ) exp ( j 2 ft ) df
34
[Example 2.6]
Gaussian Pulse
We will W ill derive d i the th particular ti l form f of f a pulse l signal i l that th t has h the th same mathematical form as its own Fourier transform. By differentiating the formula for the Fourier transform G( f ) with respect to f, we have d j 2 tg ( t ) U G ( f ) G ( f ) = g ( t ) exp p ( j 2 ft f ) dt df
Add (2.31)
dg ( t ) dt
d g ( t ) U j 2 f G ( f ) dt
plus j times
j 2 tg ( t ) U
dG ( f ) + 2 tg ( t ) U j + 2 fG ( f ) df
dG ( f ) df
35
d G( f ) df f
If
dg ( t ) dt
= 2 tg ( t ) , then
= 2 fG ( f )
[Example 2.6]
Since the Si th pulse l signal i l g(t) and d it its Fourier F i transform t f G( f ) satisfy ti f the same differential equation, they are the same function, i.e. G( f ) )= g( f ), ) where g( f ) is obtained from g(t) by substituting f for t.
Since
dg ( t ) dt
2 g t = exp t ( ) ( ) = 2 tg ( t ), we can obtain
exp ( t 2 ) dt = 1
exp ( t 2 ) U exp ( f 2 )
36
(2 39) (2.39)
Proof:
d t g (t ) = g d ( ) dt
37
[Example 2.7]
Triangular Pulse
consider the doublet pulse g1(t) shown in Fig Fig. (a) (a). By integrating this pulse with respect to time, we obtain the triangular pulse g2(t).
The doublet pulse of figure (a) is real and odd-symmetric and its Fourier transform is therefore odd and purely imaginary. The h triangular i l pulse l of f figure fi (b) is i real l and d symmetric i and d its i Fourier transform is therefore symmetric and purely real.
38
[Example 2.7]
amplitude A, defined for the interval T t 0 Fourier transform: ATsinc( fT ) exp( j fT ) amplitude A, defined for the interval 0 t T Fourier transform: ATsinc( (f fT ) exp( p( j fT f )
Proof
g ( t ) = G ( f ) exp ( j 2 ft ) df
g ( t ) = G ( f ) exp ( j 2 ft ) df = G ( f ) exp ( j 2 ft ) df
Corollary: y
g ( t ) U G ( f )
40
[Example 2.8]
g ( t ) = Re g ( t ) j Im g ( t )
+ j Im g ( t ) g ( t ) = Re g ( t )
1 1 g t g t g t g t g t g Re Im = = + ( ) ( ) ( ) ( ) ( ) ( t ) 2 2j 1 U g t G f G Re + ( ) ( ) ( f ) 2 1 U g t G f G Im ( ) ( ) ( f ) 2j
(Im[g(t)]=0)
If g(t) is a real-valued time function, we have G( f )= G*(- f ). In other words, G( f ) exhibits conjugate symmetry.
41
G12 ( f ) =
Define:
= f f'
G12 ( f ) = G2 ( f ) g1 ( t ) exp ( j 2t ) dt d
42
The inner integral is recognized as G1() This integral is known as the convolution integral expressed in the h f frequency d domain, i and d the h function f i G12( f ) i is referred f d to as the convolution of G1( f ) and G2( f ). Th multiplication The lti li ti of f two t signals i l in i the th time ti domain d i is i transformed into the convolution of their individual Fourier transforms in the frequency domain domain. This property is known as the multiplication theorem. Notation: G12 ( f ) = G1 ( f ) G2 ( f )
G12 ( f ) = G1 ( ) G2 ( f ) d
Q.E.D.
g1 ( t ) g 2 ( t ) U G1 ( f ) G2 ( f ) G1 ( f ) G2 ( f ) = G2 ( f ) G1 ( f )
43
g1 ( )g 2 ( t ) d U G1 ( f ) G2 ( f )
Proof
= g1 ( ) g 2 ( t )d
Q.E.D.
44
We may thus state that the convolution of two signals in the time domain is transformed into the multiplication of their individual Fourier transforms in the frequency domain. This p property p y is known as the convolution theorem. Property 11 and property 12 are the dual of each other. Shorthand notation for convolution:
g1 ( t ) g 2 ( t ) U G1 ( f ) G2 ( f )
45
Let g(t) be defined over the entire interval -<t< and assume its F i transform Fourier f G( f ) exists. i If the h energy of f the h signal i l satisfies i fi 2 E = g ( t ) dt < then 2 2 g ( t ) dt = G ( f ) df
|G( f )|2 is defined as the energy spectral density (valid for energy signal). For power signal, signal we define power spectral density S( f ): 1 P S ( f ) df = lim T 2T
g ( t ) dt
( deterministic signal )
46
Proof:
E=
g ( t ) dt =
g * ( t ) g ( t )dt
= g * ( t ) G ( f ) e j 2 ft df dt = G ( f ) g * ( t ) e j 2 ft dt df
= G ( f ) g ( t ) e j 2 ft dt df
= G ( f ) G * ( f ) df
= G ( f ) df
2
47
[Example 2.9]
Consider C id the th sinc i pulse l A sinc( i (2Wt 2W ). ) Th The energy of f this thi pulse l equals 2 E = A sinc 2 ( 2Wt ) dt
The integral in the right-hand side of this equation is rather difficult to evaluate. evaluate From example 2.4, the Fourier transform of the sinc pulse A sinc(2Wt) is equal to (A/2W)rect( f /2W). Applying Rayleigh Rayleighs s energy theorem 2 A 2 f E = rect df 2W 2W A = 2W
2
A2 W df = 2W
W
48
Compression of a function in the time domain is equivalent to the expansion of its Fourier transform in the frequency domain, or vice versa.
49
If the time-domain description of a signal is changed, then the frequency domain description of the signal is changed in an frequency-domain inverse manner, and vice versa. If a signal is strictly limited in frequency, frequency then the time-domain description of the signal will trail on indefinitely.
A signal g is strictly y limited in frequency q y or strictly y band limited if its Fourier transform is exactly zero outside a finite band of frequencies.
If a signal is strictly limited in time, then the spectrum of the signal i l is i infinite i fi i in i extent.
A signal is strictly limited in time if the signal is exactly zero outside a finite time interval.
51
Bandwidth
The bandwidth of a signal provides a measure of the extent of significant spectral content of the signal for positive frequencies.
When the signal is strictly band limited, the bandwidth is well defined. When the signal is not strictly band-limited, band limited there is no universally accepted definition of bandwidth.
A signal is said to be low-pass if its significant spectral content is centered around the origin. A signal is said to be band-pass if its significant spectral content is centered around fc, where fc is a nonzero frequency.
52
Bandwidth (cont.)
When th Wh the spectrum t of f a signal i l is i symmetric t i with ith a main i lobe l b bounded by well-defined nulls(i.e. frequencies at which the spectrum is zero), zero) we may use the main lobe as the basis for defining the bandwidth of the signal. When the signal is low-pass, the bandwidth is defined as one half the total width of the main spectral lobe, since only one h lf of half f this thi lobe l b lies li inside i id the th positive iti frequency f region. i When the signal is band band-pass pass with main spectral lobes centered around fc, where fc is large, the bandwidth is defined as the width of the main lobe for positive frequency. This definition of bandwidth is called the null-to-null bandwidth.
53
Bandwidth (cont.)
54
3-dB Bandwidth
When the signal is low-pass, the 3-dB bandwidth is defined as the separation between zero frequency, where the amplitude spectrum attains its peak value, value and the positive frequency at which the amplitude spectrum drops to 1/ 2 of its peak value. When the signal is band-pass, band-pass centered at fc, the 3-dB bandwidth is defined as the separation (along the positive frequency q y axis) ) between the two frequencies q at which the amplitude spectrum of the signal drops to 1/ 2 of the peak value at fc. Advantage : it can be read directly from a plot of the amplitude spectrum. Disadvantage: i d it i may be b misleading i l di if the h amplitude li d spectrum has slowly decreasing tails.
55
Ch a p t e r 2.4 Chapter 2 .4 The The I Inverse n ve r s e Re Relationship la t io n s h ip between b e t we e n Time and Frequency
3-dB Bandwidth
56
Root Mean Square (rms) bandwidth, defined as the square root of the second moment of a properly normalized form of the squared d amplitude lit d spectrum t of f the th signal i l about b t a suitably it bl chosen point. The rms bandwidth of a low-pass low pass signal is formally defined as:
f 2 G ( f ) 2 df Wrms = G ( f ) 2 df An attractive feature of the rms bandwidth is that it lends itself more readily to mathematical evaluation than the other two definitions of bandwidth, but it is not as easily measurable in the laboratory.
57
1 2
Time-Bandwidth product
For any family of pulse signals (e.g. the exponential pulse) that differ in time scale, the product of the signals duration and d it its bandwidth b d idth is i always l a constant, t t as shown h by b (duration) (bandwidth) = constant Th product The d t is i called ll d the th time-bandwidth ti b d idth product d t or bandwidth-duration product. If the duration of a pulse signal is decreased by reducing the time scale by a factor a, the frequency scale of the signal signals s spectrum, and therefore the bandwidth of the signal, is increased by the same factor a.
58
Consider C id the th rms bandwidth. b d idth Th The corresponding di definition d fi iti for f the th rms duration is
Trms
t 2 g ( t )2 dt = 2 g ( t ) dt
1 2
Th time-bandwidth The ti b d idth product d t has h the th following f ll i form: f 1 TrmsWrms 4 Gaussian pulse satisfies this condition with the equality sign.
59
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
The Dirac delta function, denoted by (t), is defined as having zero amplitude everywhere except at t = 0, 0 where it is infinitely large in such a way that it contains unit area under its curve; i.e.
( t ) = 0, t 0
( t )dt = 1
The delta function (t) is an even function of time t. Sifting property of the delta function:
g ( t ) ( t t0 ) dt = g ( t0 )
Replication property of the delta function: the convolution of any function with the delta function leaves that function unchanged.
g ( t ) ( t ) = g ( ) ( t ) d = g ( t )
61
(t ) U 1
This relation states that the spectrum spectr m of the delta function f nction (t) extends uniformly over the entire frequency interval. We may view the delta function as the limiting form of a pulse of unit area as the duration of the pulse approaches zero.
62
DC Signal
By applying the duality property to the Fourier-transform pair of ( t ) U 1 and noting that the delta function is an even function, we obtain
If g ( t ) U G ( f ) then G ( t ) U g ( f )
1 U ( f )
1U ( f )
exp ( j 2 ft ) dt = ( f )
cos ( 2 ft ) dt = ( f )
63
exp ( j 2 f c t ) U ( f f c )
Sinusoidal Functions
64
Definition:
+1, sgn ( t ) = 0, 1,
t >0 t =0 t<0
The signum function does not satisfy the Dirichlet conditions and does not have a Fourier transform. Th signum The i f ti can be function b viewed i d as the th li limiting iti form f of f the th antisymmetric double-exponential pulse as the parameter a approaches 0.
exp ( at ) , g ( t ) = 0, exp ( at ) ,
t >0 t =0 t<0
65
j 4 f a 2 + ( 2 f )
1
2
F ( sgn ( t ) ) = lim
a 0
4 j f a + ( 2 f )
2 2
1 j f
sgn ( t ) U
j f
66
1 1, 1 u (t ) = , 2 0,
t >0 t =0
t<0
1 u (t ) = sgn ( t ) + 1 2
67
Let
y ( t ) = g ( ) d
The integrated signal y(t) can be viewed as the convolution of the original signal g(t) and the unit step function u(t) , as shown by
y ( t ) = g ( ) u ( t ) d = g ( t ) u ( t )
1, 1 u (t ) = , 2 0 0,
68
<t =t >t
Since
G ( f ) ( f ) = G ( 0) ( f )
Y(f )=
1 1 G ( f ) + G ( 0) ( f ) j 2 f 2
g ( ) d U
1 1 G ( f ) + G ( 0) ( f ) 2 j 2 f
69
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
A periodic signal can be represented in terms of a Fourier transform provided that this transform is permitted to include delta functions. functions Consider a periodic signal gT0(t) of period T0:
gT0 ( t ) =
n =
exp p ( j 2 nf f 0t )
T0 2
T0 2
gT0 ( t ) exp ( j 2 nf 0t ) dt
Let g(t) be a pulse like function, which equals gT0(t) over one period and is zero elsewhere; that is, T0 T0 gT0 ( t ) , t g (t ) = 2 2 elsewhere 0, gT0(t) may y now be expressed p in terms of the function g(t)
gT0 ( t ) =
m= =
g ( t mT )
0
g(t) is Fourier transformable and can be viewed as a generating function, which generates the periodic signal gT0(t).
1 cn = T0
T0 2
T0 2
The formula for the reconstruction of the periodic signal gT0(t) can be rewritten as: cn = f 0G ( nf 0 )
gT0 ( t ) =
n= =
exp ( j 2 nf 0t ) = f 0
n = =
G ( nf ) exp ( j 2 nf t )
0 0 0 0
gT0 ( t ) =
m =
g ( t mT ) = f G ( nf ) exp ( j 2 nf t )
0 0 n =
g ( t mT ) U f G ( nf ) ( f nf )
0 0 n = 0 0
(2.88)
The Fourier transform of a periodic signal consists of delta functions occurring at integer multiples of the fundamental f frequency f0=1/T 1/T0, including i l di the h origin, i i and d that h each h delta d l function f i is weighted by a factor equal to the corresponding value of G(nf0).
73
The function g(t), constituting one period of the periodic signal gT0(t), ) has a continuous spectrum defined by G( f ). ) The periodic signal gT0(t) has a discrete spectrum. Periodicity in the time domain has the effect of changing the frequency-domain frequency domain description or spectrum of the signal into a discrete form defined at integer multiples of the fundamental frequency.
74
An ideal sampling function, function or Dirac comb, consists of an infinite sequence of uniformly spaced delta functions.
T ( t ) =
0
m=
( t mT )
0
The generating function g(t) for the ideal sampling function T0(t) consists of the delta function (t). We therefore have G( f )=1 and G(nf0)=1 for all n. U i Eq. Using E (2.88) (2 88) g ( t mT0 ) U f 0 G ( nf 0 ) ( f nf 0 ) yields i ld
m = n =
m =
( t mT ) U f ( f nf )
0 0 n = 0
The Fourier transform of a periodic train of delta functions, spaced T0 seconds apart, consists of another set of delta functions weighted by the factor f0=1/ T0 and regularly spaced f0 Hz apart along the frequency axis.
75
f0
n =
G ( nf ) exp ( j 2 nf t )
0 0
m=
76
m=
n =
A band-limited signal of finite energy, which only has frequency components less l than h fm Hertz, is i completely l l described d ib d by b specifying the values of the signal at instants of time separated by 1/2 fm seconds. seconds 1 TS or sampling rate f S 2 f m 2 fm A band-limited signal of finite energy, which only has frequency components less than fm Hertz, may be completely recovered f from ak knowledge l d of f its i samples l taken k at the h rate of f 2 fm samples l per second. The sampling rate of 2fm per second, for a signal bandwidth of fm Hertz, is called the Nyquist rate; its reciprocal 1/2 fm (measured in seconds) is called the Nyquist interval.
77
X S ( f ) = X ( f ) X ( f ) =
1 TS
n =
X ( f nf
78
79
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
System: any physical device that produces an output signal in response to an input signal. signal Excitation: input signal. Response: output signal. signal In a linear system, the principle of superposition holds, i.e., the response of a linear system to a number of excitations applied simultaneously is equal to the sum of the responses of the system when each excitation is applied individually.
Filter: a frequency-selective device that is used to limit the spectrum of f a signal i l to some band b d of f frequencies. f i Channel: transmission medium that connects the transmitter and receiver i of f a communication i ti system. t
81
Time Response
In th I the time ti domain, d i a linear li system t is i described d ib d in i terms t of f its it impulse response, which is defined as the response of the system (with zero initial conditions) to a unit impulse or delta function (t) applied to the input of the system. If t the e system syste is s time invariant inva iant, then t e the t e shape s ape of o the t e impulse pu se response is the same no matter when the unit impulse is applied to the system. Convolution Integral:
y ( t ) = x ( )h ( t ) d = x ( t ) h ( t )
= h ( )x ( t ) d = h ( t ) x ( t )
82
Causal: A system is said to be causal if it does not respond before the excitation is applied.
For a linear F li time-invariant i i i (LTI) system to be b causal, l the h impulse i l response h(t) must vanish for negative time, i.e. h(t)=0, t<0. A system y operating p g in real time to be p physically y y realizable, , it must be causal. The system can be noncausal and yet physically realizable.
Stable bl : A system is i said id to be b stable bl if the h output signal i l is i bounded for all bounded input signals.
Bounded input-bounded input bounded output (BIBO) stability criterion. For a LTI system to be stable, the impulse response must be absolutely integrable, i.e.
h ( t ) dt <
83
(2.100)
Consider a LTI system of impulse response h(t) driven by a complex exponential input of unit amplitude and frequency f
= exp ( j 2 ft ) h ( ) exp ( j 2 f ) d
Transfer function of the system is defined as the Fourier transform of its impulse p response p H ( f ) h ( t ) exp ( j 2 ft ) dt
H(f )
x (t )
y (t )
y ( t ) = H ( f ) exp ( j 2 ft ) = H ( f ) x (t )
x( t )=exp( j 2 ft )
84
or, equivalently, in the limiting form (a superposition of complex exponentials of incremental amplitude)
x ( t ) = lim
f 0 f = k f k =
X ( f ) exp ( j 2 ft ) f
y (t ) = H ( f ) x (t )
y ( t ) = Y ( f ) exp ( j 2 ft ) df
H ( f )X ( f ) exp ( j 2 ft ) df
Y( f )= H( f )X ( f )
= H ( f ) X ( f ) exp ( j 2 ft ) df
85
The transfer Th f function f i H( f ) is i a characteristic h i i property of f a LTI system. It is a complex quantity: H ( f ) = H ( f ) exp j ( f )
|H( f )|: amplitude response ( f ): phase or phase response If the impulse response h(t) is real-valued, the transfer function H( f ) exhibits conjugate symmetry:
H ( f ) = H ( f )
86
( f ) = ( f )
Chapter 2.7 Transmission of Signals Through Linear Systems Frequency Response (cont.)
87
A necessary and sufficient condition for a function ( f ) to be the gain of a causal filter is the convergence of the integral ( f ) 1 + f 2 df < this condition is known as the Paley-Wiener criterion. We may associate with this gain a suitable phase ( f ), such that the resulting filter has a causal impulse response that is zero for negative i time. i The Paley-Wiener criterion is the frequency-domain equivalent of the causality requirement. requirement A realizable gain characteristic may have infinite attenuation for a discrete set of frequencies, but it cannot have infinite attenuation over a band of frequencies.
88
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
2.8 Filters
A filter is a frequency-selective device that is used to limit the spectrum of a signal to some specified band of frequencies. frequencies Frequency response is characterized by a passband and a stopband. The frequencies q inside the passband p are transmitted with little or no distortion, whereas those in the stopband are rejected. There are low-pass, high-pass, band-pass, and band-stop filters.
low-pass | H ( f )| f band-stop | H ( f ) | f
90
band-pass | H ( f ) | f
0 high-pass | H ( f ) |
2.8 Filters
y ( t ) = x ( t t0 )
The ideal low-pass filter is noncausal because it violates the P l Wi Paley-Wiener criterion. it i This can be confirmed by examining the impulse response h(t)
h ( t ) = exp j 2f ( t t0 ) df B = sin 2 B ( t t0 ) ( t t0 )
91
B
= 2 Bsinc i 2 B ( t t0 )
(2 118) (2.118)
2.8 Filters
There is some response from the filter before the time t=0, so confirming that the ideal low-pass filter is noncausal. H However, we can make k th the delay d l t0 large l enough h such h that th t
sinc 2 B ( t t0 ) 1 for t < 0
By so doing, B d i we are able bl to build b ild a causal l filter fil that h closely l l approximates an ideal low-pass filter.
92
2.8 Filters
Consider C id a rectangular t l pulse l x(t) of f unit it amplitude lit d and dd duration ti T, which is applied to an ideal low-pass filter of bandwidth B. The problem is to determine the response y(t) of the filter. filter Using Eq. (2.118), and setting t0=0 for simplification h ( t ) = 2 Bsinc ( 2 Bt ) Gibbs phenomenon the resulting filter response
y ( t ) = x ( )h ( t ) d
= 2B
T 2
sin 2 B ( t ) 2 B ( t )
T 2
2.8 Filters
Design of Filters
Design of filters is usually carried out in the frequency domain. domain There are two basic steps:
The approximation pp of a p prescribed frequency q y response(i.e. p ( amplitude p response, phase response, or both) by a realizable transfer function. The realization of the approximating transfer function by a physical device. device
For an approximating transfer function H( f ) to be physically realizable it must represent a stable system. realizable, system Stability is defined here on the basis of the bounded input bounded output p criterion described in Eq. q (2.100). ( ) In the following, we specify the corresponding condition for stability y in terms of the transfer function. The traditional approach is to replace j2 f with s.
94
2.8 Filters
Design of Filters
Ordinarily, the approximating transfer function H( Ordinarily (s) is a rational function, which may be expressed in a factored form as: H ' ( s ) = H ( f ) j 2 f = s
s z1 )( s z2 )" ( s zm ) ( =K ( s p1 )( s p2 )" ( s pn )
where K is scaling factor; z1, z2, , zm are the zeros of the transfer function; ; p1, p2, ,, , pn are its p poles. For low-pass and band-pass filters: m<n. If the system y is causal, , all the poles p of the transfer function H(s) should be inside the left half of the s-plane, i.e. Re[pi]<0.
95
2.8 Filters
Two popular T l families f ili of f low-pass l filt filters: B Butterworth h filters fil and Chebyshev filters. All their zero are at s= and the poles are confined to the left half of the s-plane plane. Butterworth filter
The poles of the transfer function lie on a circle with origin as the center and 2B as the radius, where B is the 3-dB bandwidth of the filter. Is said to have a maximally flat passband response. The poles lie on an ellipse. P id faster Provide f t roll-off ll ff than th Butterworth B tt th filter filt by b allowing ll i ripple i l in i the th frequency response. Type yp 1 filters have ripple pp only y in the p passband. Type 2 filters have ripple only in the stopband and are seldom used.
96
Chebyshev filter
2.8 Filters
Comparison of the amplitude response of 6th order Butterworth low-pass low pass filter with that of 6th order Chebyshev filter. filter
97
2.8 Filters
A common alternative to both the Butterworth and Chebyshev filters is the elliptic filter, which has ripple in both the passband and the stopband. Elliptic p filter p provide even faster roll-off for a g given number of poles p but at the expense of ripple in both the passband and stopband. Butterworth filters are the simplest p and elliptic p filters are the more complicated to design in mathematical terms. The finite-duration impulse response (FIR) filter is often used in digital signal processing. The FIR filter is the equivalent of the tapped delay-line filter d described ib d in i the h previous i section. i The FIR filter has only zeros; it is thus inherently stable.
98
2.8 Filters
99
2.8 Filters
100
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
Communication using low-pass signals is referred to as baseband communication. In some transmission media, there is insufficient spectrum at baseband ( (e.g., g , radio waves) ) or the p properties p of media are not conductive to conducting signal at baseband (e.g., optical fibers). In these cases, we employ band-pass communications.
Narrow-band signal: the bandwidth 2W is small compared to the carrier frequency f0. A real-valued band-pass signal g(t) with non-zero spectrum G( f ) in the vicinity y of fc may y be expressed p in the form:
g ( t ) = a ( t ) cos 2f c t + ( t )
a(t): envelope (non-negative) (t ) : phase Using the relationship cos(A+B)=cos(A)cos(B)-sin(A)sin(B)
(2 123) (2.123)
2 a ( t ) = g I2 ( t ) + gQ (t )
gQ ( t ) ( t ) = tan g t ( ) I
1
Eq (2.123) Eq. (2 123) may be written as (2.126) ( t ) exp ( j 2f c t ) g ( t ) = Re g where we define g (t ) = g I (t ) + jgQ (t ) (t ) as the complex The g I (t ) and gQ (t ) are real, we refer to g envelope of the band band-pass pass signal. 1 exp 2 g (t ) = g t j f t g + ( ) ( ) ( t ) exp ( j 2fct ) c 2 g (t ) U G ( f ) F ( f f )+G * ( f f ) G ( f ) = 1 G c c 1 2 Re ( A) = ( A + A* )
( f )| |G
fc
W 0 W
|G( f )|
fc
f
104
2W
2W
T determine To d t i the th complex l envelope l of f the th RF pulse l t g ( t ) = A rect cos ( 2 f c t ) T Assume fcT>>1, so that g(t) is narrow-band
t g ( t ) = Re A rect exp ( j 2f c t ) T the complex envelope is t ( t ) = A rect g T and the envelope equals
t ( t ) = A rect a (t ) = g t T
105
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
x(t) represents the message signal signal, y(t) is the received or output signal, and h(t) is the impulse response of the channel or filter. X( f )=F[x(t)], )] H( f )= F[h(t)], )] Y( f )= F[y(t)]. )] Time domain y ( t ) = x ( )h ( t ) d
Frequency domain Y ( f ) = H ( f ) X ( f ) These equations are valid for linear systems. systems Time i domain d i
Band-pass system
y (t ) =
x ( )h ( t ) d
Frequency domain Y ( f ) = H ( f ) X ( f )
107
Band-pass systems
( t ) exp ( j 2f c t ) g ( t ) = Re g (2.126)
When h(t) is the impulse response of a bandpass filter, filter by analogy with g(t) of Eq. 2.126, it may be represented as ( t ) exp ( j 2 f t ) h ( t ) = Re h c ( t ) is the complex impulse response of the bandpass where h
( t ) exp ( j 2 f c t ) y ( t ) = Re y
where y ( t ) is the complex envelope of y(t).
109
The analysis of a band-pass system is complicated due to the multiplying p y g factors cos(2 ( fct) and sin(2 ( fct) ). The significance of Eq. (2.140) is that, we need only concern the (t ) , h ( t ) , and y (t ). low-pass p functions, x In other words, the analysis of a band-pass system is replaced by an equivalent but much simpler low-pass analysis that completely retains the essence of the filtering process.
1 3
1 (t ) = h (t ) x (t ) y 2
2
111
Target: compute the response of an ideal band-pass filter H( f ) to an RF pulse of duration T and carrier frequency fc ( fc T>>1) t x ( t ) = A rect cos ( 2 f c t ) T
low-pass equivalent
2, H(f )= 0,
113
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
Whenever a signal is transmitted through a dispersive (frequencyselective) device such as a filter or communication channel, channel some delay is introduced into the output signal in relation to the input signal. g In an ideal filter, the phase response varies linearly with frequency inside the passband of the filter, in which case the filter introduces a constant delay. Question: what if the phase response of the filter is nonlinear? Signal Models: assume that a steady sinusoidal signal at frequency fc is
transmitted through a dispersive channel that has a total phase-shift of ( fc). Phase delay of the channel: ( fc)/2 fc [sec] is the time taken by the received signal phasor to sweep out this phase lag.
Phase delay y is not necessarily y the true signal g delay. y The true signal delay is represented by the envelope or group delay.
115
Assume that the dispersive channel is described by the transfer function: H ( f ) = K exp j ( f ) where K is a constant the phase ( f ) is a nonlinear function of frequency. The input signal x(t) consists of a narrow-band signal: x ( t ) = m ( t ) cos ( 2 f c t ) where m(t) is a low-pass low pass (information (information-bearing) bearing) signal with its spectrum limited to the frequency interval | f | W. Assume fc >> W. By using the Taylor series about the point f f=f fc and retaining only the first two terms: ( f ) ( fc ) + ( f fc ) ( f ) f
116
f = fc
Taylor series at f = f c
n =0
n)
( fc )
n!
(f
fc )
H ( f ) K exp j 2 f c p j 2 ( f f c ) g
2 H ( f ) , 0,
f >0 f <0
The Fourier transform of the complex envelope of the received signal: 1 (f) Y( f )= H( f )X (2.139) 2 K exp ( j 2 f c p ) exp ( j 2 f g ) M ( f )
The term exp(p( j2fg)M( f ) represents p the Fourier transform of the delayed signal m(t-g). Complex envelope of the received signal:
( t ) K exp ( j 2 f c p ) m ( t g ) y
= Km ( t g ) cos 2 f t ( ) c p
118
The sinusoidal carrier wave cos(2fct) is delay by p seconds, hence p represents the phase delay. Sometimes, Sometimes p is also referred to as the carrier delay. The envelope p m(t) is delayed y by y g seconds; ; hence, , g represents p the envelope or group delay. g is related to the slope of the phase ( f ), measured at f=fc. When the p phase response p ( f ) varies linearly y with frequency, q y, the signal is delayed but undistorted. When this linear condition is violated, we get group delay distortion.
119
Wireless Information Transmission System Lab. Institute of Communications Engineering g g National Sun YatYat-sen University
121
Some source are digital in the sense that the information can be naturally represented as a sequence of zeros and ones ones. The digital waveform can be represented as:
g ( t ) = bk p ( t kT )
k =0
122