Professional Documents
Culture Documents
333 PDF
333 PDF
B.P. Latbi
California State University - Sacramento
Roger A. Green
North Dakota State University
Printing number. 9 8 7 6 5 4 3 2 I
Chapter B Solutions 1
Chapter 1 Solutions 23
Chapter 2 Solutions 54
Chapter 3 Solutions 85
Chapter 4 Solutions . 138
Chapter 5 Solutions . 213
Chapter 6 Solutions . 298
Chapter 7 Solutions . 324
Chapter 8 Solutions . 355
Chapter 9 Solutions . 388
Chapter 10 Solutions . . 421
Chapter B Solutions
B.2. (a) For 1 + ;, r = J1 2 + 12 = ./2 and B = arctan (f) = 7r/4 = 0.584. Thus,
I + J = ./2e'"/4 = 1.414e'0.584.
e' + 1 = 1.755e'/2 •
(f) Using the results from B.2a and B.2b,
1 +J ./2eJ1r/4
--- =
-4 + ;3 5eJ2.498
= 0.283e- 3 1.1 i 3 ,
B.3. (a) Using Euler's identity,
3eirr/
4
= 3cos(7r/4) + ;3sin{7r/4) = 2.121 + ;2.121.
(b) Using Euler's identity,
~
eJ
= e- 1 = cos(-1) + JSin(-1) = 0.540 - J0.841.
(c) Expanding,
w-
- +-
w*= x + JY + x - JY =x= R ex+JY
( ) = Rew.
( )
2 2
(b)
w - w•
--- =
x + JY - x + .JY =y=mx+Jy=mw.
l ( ) I ( )
2J 2J
Re(e"') = excos(y).
(b) Im (JW1) = Im (J( x r + JY1)) = Im ( -y1 + JXi) = x1. Also, Re (w 1) = xi. Clearly,
2
(c) Re(w1)+Re(w2) = X1 +x2. Also, Re(w1 +w2) = Re(x1 +1Y1 +x2+1Y2) = X1 +x2.
Thus,
True. Re(w1) + Re(w2) = Re(w1 + w2)-
(d) Im(w1)+1m(w2) =Yi +y2. Also, Im(w1 +W2) = lm(x1 +JY1 +x2+JY2) = Y1 +y2.
Thus,
True. Im(w1) + Im(w2) Im(w1 + w2)- =
=
(e) Re(w1)Re(w2) = x 1x2. Also, Re(w1w2) Re((x1 + JY1)(x2 + JY2)) Re(x1x2 - =
Y1Y2+1(x1Y2 + x2y1)) = x1x2 - Y1Y2). Jn general x1x2-:/: X1x2 - Y1Y2, so
= Im ( :c2+JY2
"'1+w1)
Jn general
Similarly,
(d)
3
(g)
B.8. First, express w1 in both rectangular and polar coordinates. For rectangular form,
W1 ={3+14) 2 =9- 16+1{12+12) = -7+124. For polar form, r1=v(-7)2+24 2 = 25
and 81 = arctan ( :_~) = -1.287 + 7r = 1.855. Thus, w 1 = r 1 e181 = 25e'1.sss.
Second, express w 2 in both rectangular and polar coordinates. Since j = &"12 and
e - 140" = l , rectangular form is w2 = x 2 + JIJ2 = 12.5. For polar form, w 2 = r 2 e' 82 =
2.5e'"' 2 = 2.5eJ1·571 .
(a) From above,
(c)
(d)
W 1W2 = X + JY = - 60 + J - 17.5.
(g)
4
(b) From above,
W2 = X2 + JY2 = 1.543.
(c)
Similarly,
(d)
B.10. {a) Note, we can rewrite 2.5cos(3t) - 1.5sin(3t + rr/3) = ccos(3t + </>) as
Re (2.5e' 31 -i 11.5&<31 +"/ 3 l) = Re (ce1< 3 t+c1>l). Working with the left-hand side,
Re (2.5e'3 ' + Jl.5&C3t+"/ 3 l) =Re (e131 (2.5 + l.5&("/ 3 +"12 >)). The unknown con-
stants c and </>are determined by comparing the left- and right-hand sides.
and
·'
e
z(w+30)
- e
23
11:(w - 10)
e"'( w+10 ) -
1
(w _ 1Cl') e"'<w-30'). S ubstituting t he limits of integration yields
21
5
B.11 . Solutions to this problem are based on Euler's identity.
(a)
ex+1Y + e-x-JY
cosh(w) cosh(x + JY)
= 2
= 0.5 ((cos(y) + JSin(y))e"' + (cos(y) - JSin(y))e-x)
0.5 (cos(y)(e"' + e- "' ) + JSin(y)(e"' - e- "' ))
= cos(y) cosh(x) + Jsin(y) sinh{x)
Thus,
cosh(w) = cosh(x + JY) = cosh(x) cos(y) + Jsinh(x) sin(y) .
{b)
e"' +JY - e -X-JY
sinh(w) sinh{x + JIJ) =
2
0.5 ((cos(y) + JSin(y))e"' - (cos(y) - JSin(y))e-"')
0.5 (cos(y)(e"' - e- "' ) + J Sin(y)(ex + e-x))
cos(y) sinh(x) + JSin(y) cosh{x)
Thus,
sinh(w) = sinh(x + JY) = sinh(x) cos(y) + Jcosh(x) sin(y).
B.12. (a) (w) 4 = - 1 = e1i(n +2 ...i.> '* w = (e1<r.+2 ,,-i.>) 114 • Thus,
W = e1"(l/4+ k/ 2) for· k. = [O }1)2}3] .
o.• ..... . . . .. .
0 .2 .
I •. . ...... ..
~0.2
-0.6
x.
-o.•
-0.5
....
0 0.5
6
(b) Notice,
32
(w - (1 + ;2)) 5 = V2(1 + J) = 32eJ(1r/4+2 ""k)_
···· ....
' •i .•.
3.5 ' .
2.5 : . .. '• ~
I 2.
1.5 ' ..
·x .,
o.s
(c) The solution set of I w - 2; I= 3 describes a circle. To see this, note that
I w-2; 12 = (w-21)(w-2;)* = (x+;(y-2))(x+;(2-y)) = x 2 + (y-2) 2 = 32 = 9.
The circle has center (0, 2) and radius r = 3.
>>theta= linspace(0,2*pi,201); x = 3*cos(theta); y = 2+3*sin(theta);
>> plot(x,y,'k-'); axis equal ; grid; xlabel('Real'); ylabel('Imag');
(d) Graph w(t) =(I+ t)eJt for (-10 :St :S 10).
>> t = (- 10:.01:10); w = (1+t) . *exp(j*t);
>> im10 = find(t ==- 10); im = find(t<O);iO = find(t==O);
>> ip = f ind(t>O) ; iplO = find(t==10);
>> plot(real (w(im10)),imag(w(im10)),'vk',real(w(im)),imag(w(im)),'k-', . . .
>> real(w(iO)),imag(w(iO)),'ok',real(w(ip)),imag(w( i p)),'k:', ...
>> real(w(ip10)),imag(w(ip10)),'k- '); axis equal; xlabel('Real');
>> ylabel('Imag'); legend('t=-10',' t<O','t=O','t>O','t=10',0)
B.13. Since four distinct solutions are indicated, we know n = 4. The solutions to w" =
r;
w2 = r 2 eJ82 lie on a circle of radius 1n. The solutions to (w -wi)" = w2 lie on the
same circle shifted by w 1 . To find w 1 , drop perpendicular lines from the circle center to
7
4 .
_,.____._ _..___
-3 •2
_,
_,_;::::...._.._~=-----'----'-__J
0
Roal
the real and imaginary axes, respectively. As shown, two similar triangles are formed.
The circle center is w 1 = A + JA. F\Irthermore, we know that A + B = v'3 + 1 and
A- B = V'3- l. Clearly, A = v'3 and B = l. Thus, w 1 = v'3+ Jv'3. The value of w 2 is
now easily found by substitution: w 2 = ( v'3+1 - ( V3+Jv'3)) 4 = (l+JV3) 4 = 16e12"13 .
Thus,
n = 4, w 1 = v'3 + Jh, and w 2 = 16e12"1 3 .
B.14. We can write {J - w)l. 5 = (J - w) 312 = ,/8e1-:r/4 . Squaring both sides yields (J -
w) 3 = 8e1 <"'12+ 2 "k). Taking the third root of each side yields (J - w) = 2eJ("/6 + 2 rrk/3).
Rearranging yields three distinct solutions
8
A
2.5
2 ·
1.5
!"' 1
0.5
0 .......
-0.5
-1
-1.5'--~-~-~~-~-~-'--~--~--'
-2.5 -2 -1.5 _, -o s 0 0.5 1.5 2 .5
ReaJ
B.15. w = ,/) = (eJ(n/2 +2rrk)) 112 = e1<.,,!4+.,,k)_ Thus, there are two distinct solutions
9
MATLAB and other calculating devices generally give only the k = 0 solution 1 +pr.
B.17. log 10 (- l ) = log 10 e1<..-+ 2 "kl = J (11 + 211k) log 10(e). Since k can be any integer, there
arc an infinite number of solutions
MAT LAB and ot her calculating devices generally give only the k = 0 solution
J7r log 10 (e).
B.18. 1
(a) Jn (i~ ) = In ( J'2e~r.t•) =In (( J2)- 1 e1C-,../ 4 +2'r1.:J) = - In( J2) + J(-7r/4 + 27Tk).
Since k can be any integer, there are an infinite number of solut ions
1
In ( -- ) =- In( v'2) + J(- 7r /4 + 21T'k) for integer k.
l +J
MATLAB and other calculating devices genera lly give only t he k = 0 solution.
(b) cos(l+J) = 0.5 (e1<i+1J + e-1<i+1l) = 0.5 (e- (cos(l ) + 3sin(l))
1
+ e 1 (cos(l) - 3sin(l))) =
cos(l )cosh( l ) - 3sin(l) sinh( l ) . That is
B.21. (a) x 1 (t) = Re(2e<- 1 +12 r.lt) = 2e- 1 cos(27rt). This is ! Hz cosine wave that exponen-
tially decays by a factor of l - e- 1 = 0.632 every second. A signal peak is near
t = 0, where lhe signal has an amplitude of 2. See Figure SB.22a.
(b) x2(t) = Im(3-e< 1 - i2 ,.l1 ) = e' sin(27Tt). This is a llfa sine wave that exponentially
grows by a factor of e 1 = 2.718 every second. A signal peak is near t = 1/4, where
the signal has an amplitude of 1.284. See Figure Sl3.22b.
(c) x3(t) = 3 - Im(e< 1- i 2.-Jt) = 3 + e'sin(27rt). This is a !Hz sine wave that expo-
nentially grows by a factor of e 1 = 2.718 every second and has an offset of 3.
10
1.5
0 .5
i 0
-o.s
_,
-1.5
-2 '---~- -~-~-~--~-~--'----'
~ ~ ~ - u u
Figure SB.20: Plot of x(t) = e -at cos(wt) for w = 3(27r) and a= 0.5 ln(2).
A signal peak is near t = 1/4, where the signal has an amplit ude of 4.284. See
Figure SB.22c.
B.22. (a)» t = [0:.001:3); x_l = 2*exp(- t).*cos(2*pi*t);
>> plot(t,x_1,'k-',t,2*exp(-t),'k:',t,-2*exp(-t),'k:');
>> xlabel('t'); ylabel('x_l(t)');
• 1.5
_,.____o.s
0
..,___~~-~--~--~----'
1.5 2.5
I
B.23. Since cos(t) oscillates at 2~Hz, t should cover at least 2r. seconds t o span one period.
Since sin(20t) has a period of~~ = 0.314 seconds, the st ep size oft should be less t han
0.0314 to ensure at least ten samples per period of this fastest component .
11
20
15
10
s. 0 ::. :
-10
-15
-20
-»~--~--'---~'-----'~---'----'
0 u u u
I
2.5
20
,.
10
....······
;~
•S
•10
.,.
·20
·1
0 0.5 1.5 2.5
I
B.24. The highest frequency is lOHz, so the step size oft should be 0.01 or less to provide
ten samples per period of the fastest component. The lowest frequency is lHz, so t
should span at least one second to cover one period of the slowest component.
B.25. There are many approximations possible for the sound of a b ell. In the most simple
case, we can model a bell as a decaying exponential. A small, light bell will have a
high pitch and not sustain a sound for long. Thus, we might choose a base oscilla-
t ion of lkHz. A reasonably quick decay rate is obtained if the envelop decreases by
903 every second, or e 'n(O.J)t T hus, our bell model is x(t) = e 1n(O.l)t cos(27rl000t) .
The result, however, i!:i somewhat "flat". Adding harmonics, such as cos(2n2000t),
adds richness to the sound. Furthermore, some low frequency modulation, perhaps
as a result of a hand initially ringing the bell, improves the sound. For example,
12
...
...
•.
~ ~
0.2
-0.2
w
- 0.4 J
-o.•
~
-o•
-•'--~--'---'----'--_.__
1.6 18
If a large, heavy bell is desired, the frequency and decay rates need to be reduced. For
example, z(t) = e'n(O.S)t cos{27r3t) (cos{2nl00t) + 0.1 cos(2n200t)).
>> t = (0 :1/8000:5) ; a = log(0.5);
>> z = exp(a*t).*(cos(2*pi*3*t).*(cos(2*pi*200*t)+O.l*cos(2*pi*400*t)));
>> sound(z,8000);
2
B.26. (a) To express e-x as a Taylor series, recall that a Taylor series of eu about zero is
given by e" = 2::o *-.
Substituting -x 2 for u yields
00 ( 2) i
e-x• = '\'
0 2--.
.,
i=O t.
13
= "°'~
1 21
f e - :r dx = f "' L..,.=O i=.!t f
2
L.ri=O <- 1!>
1
(b) Integrating JvieJds
~
00
"' dx 1! x 2 idx or
_,.2 =
J e dx Loo (-l)'x2i+1
i=O
., 2t. + ] .
i.
(c) Since the lower limit of the definite integral is zero, it does not make any contribu-
. T.I
t1on . ms Joe
r1 -:r2d "'"°
c-1)' .,••+1 I
"'oo -1=-!l'._ F. ' ... ATLAB
x = L--i=O i ! 'li+i :=l = L-- i=O if(2i+l)· 1rs., '"'
is used to compute the first 10 terms of the sum.
>> i = 0:9; terms= ( - 1).-i./(gamma(i+1).*{2*i+1));
Next, one to ten term truncations are obtained using the MATLAB's cumulative
sum command.
>> cumsum(terms)
The results are
(l.0000, 0.6667, 0.7667, 0.7429, 0.7475, 0.7467, 0.7468, 0.7468, 0.7468, 0.7468).
3)i
e_"'3 = L.., ~..,
t.
'°'
00 (
i=O
(c) Since the lower limit of the definite integral is zero, it does not make any contribu-
. T
t1on. hus Joe
r1 -x3 _ " oo (-1)' x 3 '+' I
dx - L.,i=O - . , - 3i+ 1 x=l = L..,i=O ~· First, MATLAB "°'oo --1=_!L_ .
( 1.0000, o. 7500, 0.8214' 0.8048, 0.8080, 0.807 4' 0.8075, 0.8075, 0.8075, 0.8075).
+ e-ix')
2
B.28. (a) To express cos(x 2 ) = 0.5 ( e'"' as a Taylor series, recall that a Taylor
series of eu about zero is given by eu = L::o ~. Substituting ±Jx2 for u yields
cos(:c 2 ) = L
00
0.5
((JX 2);
-.- 1
-
t.
+
( -re"
:,
i.
")i)·
i= O
14
(b) Integrating J cos(x2 )dx
yields J L:i=O 0.5 (~
00
i! + ~)
i' dx
00 0 5( 3)'(1+(-l)') J (x 2 )i dx or
'1;""'
ui=O · 1!
(J);(l + (-l)i)
=L
x2;+1
J
oo
cos(x 2 )dx 0.5 .1 . .
i=O t. 2t + 1
(c) Since the lower limit of the definite integral is zero, it does not make
any contribution. Thus J;
cos(x2 )dx = L:: 0 0.5< 1 >'< ~!~-1)') ~:7;
1
= l:.=I
L::o 0.5 <1>:~~;t~l'). First, MATLAB is used to compute the first 10 terms
of the sum.
>> i = 0:9; terms = 0.5•(j).-i.*(1+(-1).-i)./(gamma(i+1).•(2•i+1));
Next, one to ten term truncations are obtained using the MATLAB's cumulative
sum command.
>> cumsum(terms)
The results are
(l.0000, 1.0000, 0.9000, 0.9000, 0.9046, 0.9046, 0.9045, 0.9045, 0.9045, 0.9045).
Thus,
fi(x) =L 2 00 (l)i+l x2i.
i=O
(b) Rewrite as h(x) = (0.5)"' = e- 1n< 2 >x. Recall that a Taylor series of eu about zero
is given by e« = L:~o ~: . Substituting - ln(2)x for u yields
B.30. (a) Begin by choosing a point on the unit circle, w = e'0 . Multiplying w by itself
yields ww = w 2 = e120 . Taking this result and again multiplying by w yields
ww 2 = w 3 = & 30 . At step n, the result is wn = e'nn. From Euler's identity,
we know wn = e1nn = cos(nf!) + JSin(nf!). The process does indeed provide
the desired quadrature sinusoids: the real part provides the cosine term and the
imaginary part yields the sine term.
(b) To produce a periodic signal, n needs to be a rational multiple of21l'. Most simply,
choose n = 27f IN) where N is the number of points computed per oscillation of
each sinusoid. For reasonable quality sinusoids, N should be some moderately
large integer, say 10 or 20. Although the quality of the sinusoids increases as
N is increased, the required processing speed also increases with N. Thus, N
represents a compromise between signal quality and processor speed. Taking
N = 20, for example, yields w = e1 "1 10 . In this case, only ioo~ooN = 500e - 9
seconds (500ns) are available to process each sample. This is feasible with current
processor technologies.
15
(c) Although not required by the procedure, a vector x!n) is maintained so that the
signal outputs can be plotted.
>> N = 20; w = exp(j*2*pi/N); w_n = w;
>>I= 40; x = zeros(1,I); x(l) = w_n;
>> for i = 1:!; x(i) = w_n; w_n = w_n*w; end
>> plot([1:I],real(x),'k- ',[1 : I],imag(x),'k:');
>> xlabel('n'); ylabel('Amplitude');
[~ ~ ~ l[~~ l
1 -1 0 X3
= Ax = [ ~
-3
l·
IAI = 0 + 3 - 1 - (2 - 3 - 0) = 3.
16
1 1 1
(a) 3 2 3 =0- 9 - 3 - (-6 - 3 - 0) = -3. Thus,
-3 -1 0
-3 -3
X1 =- = - = -1.
IAI 3
The same result is obtained in MATLAB by
>>A= [1 1 1;1 2 3;1 -1 O);
>> x_1 = det([[1;3;-3) ,A(:,2:3)])/det(A)
x_l = -1
1 l 1
(b) 1 3 3 = 0 + 3 - 3 - (3 - 9 - 0) = 6. Thus,
-3 0
6 6
X2 = TAT = 3 = 2.
0 0
X3 = TAT = 3 = o.
The same result is obtained in MATLAB by
>>A= [1 1 1;1 2 3;1 -1 O);
>> x_3 = det([A(:,1:2),[1;3;-3)])/det(A)
x_3 = 0
[ ad eb ] [ X1 ]
X2
= Ax = y
= [ fc ] .
17
increase the rank of A by one. For example,
[ ~ ~ ] [ ~~ ] = [ ~ ] .
The rank of A is one and the rank of (A, y] is two. Thus, there a.re no solutions.
MATLAB verifies the desired ranks a re obtained.
>>A= [1 1;2 2); y = [1;1);
>> [rank(A), rank([A,y) ))
ans =
1 2
(d) For an infinite number of solutions t,o exist, the matrix A must be rank deficient,
and [A, y] must not increase the rank of A . For example,
[ ~ ~ ] [ ~~ ] = [ ~ ] .
The rank of A is one and the rank of (A, y] is also one. Thus, there are an
infinite number of solutions. MATLAB verifies the desired ranks are obtained.
>>A = (1 1;2 2]; y = [1; 2);
>> [rank(A), rank ([A,y])]
ans
1 1
B.33. The system of equations is first written in matrix form.
[1
1
1
1
-1
l
1
- 1
-1
-! 1
-1
1 l[~~ l ~ l
X3
X4
=Ax = [ Q
-2
.
1 1
-2 3
0 -1
-2 3
18
x = -30.0000
8.0000
16.0000
7.0000
k2 = ki = 2.5J. Lastly, k3 = 52
"J;,5,;{6 Is = - 1 = l . Thus,
l - 2.5J 2.5J 1 5
H1(s) =- - + -- + -- = -- + -
s+1 s- J s+J s+ l
-.
s2 + 1
(b) }{· 2 ( S) _
-
)
H 1 (s) -
_ s3 +s 2
+s+l _
s 2 +5s+6 - S -
4 .1
-.-
15s+25 _
(s+2){s+3) - s
_ 4 + .!J...
.<± 2
+ As+J · U ·
smg
th
e
method of residues k = 15
•+ 25 1 = -5 and k2 = 15 25
st 1 = 20. Thus
' 1 •+3 s = -2 s+2 s=-3 '
-5 20
H2(s) = s-4+-- +--.
s+2 s+3
(c) H 3 (s) = 1
($+1)2(s2+1) --
1
(s+l)2(s+1)(s-1) -- .!L
s-; I S..
s+J + _ <lo_
(S+Ti' + ~
(s+l)' Using
the method of residues, k1 >L=1 = (1+ 12 ~ 1 )( 12) = -0.25. Since the
= (s+l)!(s+1
(d) H 4 ( s) -_ Js'>+
s2 +5s+6 _ l !Js/3+17/3 _ 13s/9+17/9 I h' r
2 s+ i - 3 + Js'+ 2 s+l - 5 2+ 2 s/J 1- l/J. n some cases, t 1s 1orm
is sufficient. A complete partial fraction expansion , however, requires the de-
nominator roots s =
-2/3±./4/9-4/3
2
= - l±r.12
3
= - 0.3333 ± 0.4714J. Thus,
H4(s) = t + •-(-l~J-./2)/J + s-(- 1';3 \1'2)/J' Using the method of residues,
1
k1 = ~•\9 +A:i 3 Is=(- l-;-./2)/3 = 0.7222 + 1.49283.
s- - +1
Since the system is real,
k2 = kj = 0.7222 - l.4928J. Thus,
B.36. The MATLAB residue command computes the partial fraction expansion of a rational
function by providing three quantities: the residues, the poles, and the direct terms.
(a)>> [r,p,k) = residue([l 5 6),(1 1 1 1))
r c 1.0000
0.0000 - 2.5000i
0.0000 + 2.5000i
p = -1.0000
-0.0000 + 1.0000i
19
- 0.0000 - 1.0000i
k = (]
Thus,
H ( ) 1 - 2.51 2.5J 1 5
1
s = s +1 + s- J + s +J = s+1 + s2 + 1·
(b) >> [r,p,k] residue([! 1 1 1) ,(1 5 6) )
r = 20 . 0000
-5.0000
p - 3 .0000
-2 . 0000
k = 1 -4
Thus,
20 -5
H2(s) = - -3 +--2 +s-4.
s+ s+
(c) >> [r,p,k] = r esidue(1 , poly([-1,-1,j, - j]) )
r = 0.5000
0.5000
-0.2500 - O.OOOOi
-0.2500 + O. OOOOi
p = -1. 0000
- 1 .0000
0.0000 + 1.0000i
0.0000 - 1 .0000i
k ,. []
Thus ,
H ( ) 05 0.5 - 0 .25 -0.25
3 s = (s + 1) + (s + 1) 2 + s- J + s +J ·
ao = -l,a1 = 1, and a2 = 0.
B.38. Many solutions are possible to this problem, but the procedure is the same in each
5 2 5 s ~ _,
cas e. Consider a fictitious phone number 555-5555. Then, HN(s) = s ~s,s:ss~-~s ! =
3 2
~= 3 !~!2!~:$g . T he partia l fraction expansion of H N(s) is obtained using the l'v1ATLAI3
residue command.
20
r -0.5000
c + 0.2887i
-0 . 5000 - 0.2887i
1.0000
p = -0 . 5000 + 0.8660i
- 0.5000 - 0.8660i
0
k = 1
Thus,
-0.5000 + 0.28871 - 0.5000 - 0.2887J 1
H N(S) =1+ +
s + 0.5000 - 0.8660] s + 0.5000 + 0.8660)
+ -.
s
..•
..•
.,_.
0.2
~ • (
.....
.....
.....
-o.a
...
... •'O.S
Ro(Q• o.s
(b} Multiplying w by e18 adds 8 to the angle of wand thereby rotates w by 8. Also,
we' 0 = (x+n)(cos(O)+Jsin(O)) = (xcos(B) - y sin(fJ)) +J(xsin(O)+ycos(B}). Fur-
tl.1ermore, R w -_ [ cos(B) - sin(B) ] [ x ] _ [ x cos(B) - y sin(B) J Th
. (B)
sm cos (B) y - xsm. (8) + ycos(B) . us,
8
Rw and we' are equivalent, and Rw rotates w by B.
(c) >>theta= lO*pi/180; R = [cos(theta) -sin(theta);sin(theta) cos(theta));
>> f = [fr;fi); f = R*f;
>> plot(f( 1,:),f(2,:),'k-'); xlabel('Re(Rf)'); ylabel('Im(Rf)');
>> axis((- 1 . 1 1.1 -1.1 1.1] ); axis equal;
(d) If Rf rotates f by 8, then RRJ = R(RJ) rotates f by 28. Similarly, RRRJ
rotates f by 38. In general, (RN )! rotates f by NO.
(e) As suggested in B.39b, multiplying f(w) by t he function eJ 8 simply rotates f by
0. For example, the previous plot is also obtained by
>> f = fr+ j*fi; f = f*exp(j*theta);
>> plot(real(f) , imag(f) ,'k- ');
>> xlabel('Re(fe-{j\theta})'); ylabel( 'Im(fe- {j\theta})');
>> axis( (-1. 1 1 .. 1 -1 .1 1.1]); axis equal;
21
0.0
0.1
O•
0.2
....
.....
....
-I
- 0 0.$
01
01
O•
O>
....
-o.•
_,
-o.s 0 05
""'")
Figure SB.39e: Lissajous figure rotated by alternate method.
22
Chapter 1 Solutions
2
1.1-1. (a) E = J0 (1) 2 dt + J;(-1) 2 dt = 3
(b) E = J;(-1) 2 dt + J;(1) 2 dt =3
(c) E = J:(2) 2 dt + J;(-2) 2 dt = 12
(d) E=J:(1)2 dt+J:(-1) 2dt=3
Comments: Changing the sign of a signal does not change its energy. Doubling a
signal quadruples its energy. Shifting a signal does not change its energy. Multiplying
a signal by a constant K increases its energy by a factor K 2 .
1.1-2.
j2(t - l) dt = 1 1
Ex 2 = 10
1
(-t) dt
2
= 31 t 31j0 = 3'
1
Ex 3 =
1
. 2
0
X
2
dx
1
= 3'
Ex = 1 2
(1) 2 dt = 2,
=1
1
2
Ex+v (2) dt = 4,
(" 2
Ev = lo (1) dt = 271"
2 2 2 2
Ex+y = fo " (sin t+ 1) 2 dt = fo 1f sin 2 (t)dt+2 fo " sin(t)dt+ fo 7r (1) 2 dt = 7r-l-0+27r = 311"
In both cases (a) and (b), Ex+y =Ex+ Ey· Similarly we can show t hat for both
cases Ex- y = E,. + Ey.
23
(c) As seen in part (a),
Ex = 11f sin 2
tdt = 7r/2
Furthermore,
By= 1tr (1) dt = n
2
Thus,
r r r r
37r
Ex+y =lo (sint+l )2 dt =lo sin 2 (t)dt+2 lo sin(t)dt+ lo (1) 2 dt = 7r/2+2(2)+7r 2 +4
Additionally,
r~ 37r
Ex-y= lo (sint-1}2dt=7r/2-4+7r=-z-4
In this case, Ex+y =/= Ex-y =/=Ex+ Ey. Hence, we cannot generalize the conclu-
sions observed in parts (a) and (b).
1.1-4. Px = t f~ 2 (t 3 ) 2 dt = 64/7
24
1.1-7. (a) i. By definition, E[Tx1(t)] = J1: _
2
00 (Tx 1 (t)) dt
T 2 J,:_ 00 xi(t)dt = T 2 E [x1 (t)].
E [x1(t-T)] = E[x1(t)].
2
iii. By definition, E [x1(t) + x2(t)] ft":-oo (x1(t) + x2(t)) dt
J1: _ 00 (xi(t) + 2x1(t)x2(t) + x~(t)) dt. However, x1(t) and x2(t) are
non-overlapping so their product x 1 (t)x 2 (t) must be zero. Thus,
E [x1(t) + x2(t)J = f 1":_ 00 (xi(t) + xHt)) dt = J1":_ 00 xi(t)dt +
ftO::-= x~(t)dt = E [x1 (t)] + E [x2(t)].
If (x 1 (t) =I= 0) => (xz(t) = 0) and (xz(t) f= 0) => (x1 (t) = 0) ,
Then, E [x 1(t) + x2(t)] = E [x 1(t)] + E [x2(t)J.
iv. By definition, E [xi (Tt)] = J1: _ 00 xf(Tt)dt.
First, consider the case T > 0. Substituting t' = Tt and dt' = Tdt yields
E [x (Tt)] = 1·00
I
x2(t' ) dt' = l. J.co
t'=-co l T T
x2(t')dt' = Efx1(t)J = ~.!ill .
t'=-= 1 T ITI
Next, consider the case T < 0. Substituting t' = Tt and dt' = Tdt yields
E[x1(Tt)J = J~7:°=xf (L')~J: = -y1 f 1':_=xf (t' )dt' = E[~~t)J _ For T < 0, we
know T = - ITI. Making this substitution yields E [x 1 (Tt)] = E[J'~ltJJ.
Since energy is the same whether T < 0 or T > 0, we know
l T/2 l T /2 n n .
Px = Jim -T
T-co 1
- T/2
x(t)x*(t)clt = Jim -
T-= T 1-T/2 LL D,,D·,.eJ(wk - w.-)t dt
k~mr= m
25
integrands are periodic signals (made up of sinusoids). These terms, when divided
by T-> oo, yield zero. The remaining terms (k = r) yield
(b) i.
Hence,
Px = 52 + l5e1" 13 12 + J5e- i,..t312 = 25 + 25 + 25 = 75.
Thought of another way, note that Do = 5, D ±i = 5 and Lhus Px = 52 + 52 +
52 = 75.
ii.
x(t) = 10 cos(100t+7r/3)+16sin(l50t+7r/5)
5e1,,./3e11001 + 5e-i"/3e-;1ooi _ j 8el"1sei1so1 _1 j 8e-i,../Se-;1sot
Hence,
l.2-3. (a) x 1 (t) can be formed by shifting x(t) to the left by l plus a time-inverted version
of x(t) shifted to left by l. Thus,
26
integrands are periodic signals (made up of sinusoids). These terms, when divided
by T-> oo, yield zero. The remaining terms (k = r) yield
(b) i.
Hence,
Px = 52 + l5e1" 13 12 + J5e- i,..t312 = 25 + 25 + 25 = 75.
Thought of another way, note that Do = 5, D ±i = 5 and Lhus Px = 52 + 52 +
52 = 75.
ii.
x(t) = 10 cos(100t+7r/3)+16sin(l50t+7r/5)
5e1,,./3e11001 + 5e-i"/3e-;1ooi _ j 8el"1sei1so1 _1 j 8e-i,../Se-;1sot
Hence,
l.2-3. (a) x 1 (t) can be formed by shifting x(t) to the left by l plus a time-inverted version
of x(t) shifted to left by l. Thus,
26
I
2: -!
" -.1t_,. 0 IS -.t...,.
-:1
:x~~)
:10 v24
i l:Z. ~
0 t. -> -t~
30 3tp ..... &
-1
Figure Sl.2-1
N d .: a-t)
-:z. 0 4 t~
1~
4
2
Figure Sl.2-2
(b) x2(t) can be formed by t ime-expanding x(t) by factor 2 to obtain x(t/2) Now,
left-shift x(t/2) by unity to obtain xe~ 1 ). We now add to this a time-inverted
version of x(1;-t) to obtain x2(t) . Thus,
t+l 1-t
x 2 (t) = x(-2- ) + x(-2-).
(c) Observe that x3(t) is composed of two parts:
First, a rectangular pulse to form the base is constructed by t ime-expanding x 2(t)
by a factor of 2. This is obtained by replacing t with t/2 in x 2 (t). Thus, we obtain
xz(t/2) = x(!-¥) + x( 2 ~t) .
Second, the two triangles on top of the rectangular base are constructed by time-
expanded (factor of 2) and shifted versions of x(t) according to x (t/2) + x(-t/2).
Thus,
t+2 2-t
x3(t) = x( 4-) + x(-- ) + x(t/2) + x(-t/2) .
4
(d) x 4 (t) can be obtained by t ime-expanding x 1 (t) by a factor 2 and then mu!-
27
tiplying it by 4/3 to obtain ~x 1 (t/2) = 1 [x(t~ 2 ) +x(2;-t)) . From this, we
subtract a rectangular pedestal of height 1/3 and width 4. This is obtained
by time-expanding x 2 (t) by 2 and multiplying it by 1/3 to yield !x 2 (t/2)
i [x(1!2 ) + x(2~t)). Hence,
x 4 (t) = 4[x(-
t+2
-) 2-t]
+ x(--) - l[x(-
t+2
-) 2-t]
+ x(--) .
3 2 2 3 4 4
(e) x 5 (t) is a sum of three components: (i) x 2 (t) time-compressed by a factor 2, (ii)
x(t) left-shifted by 1.5, and (iii) x(t) t ime-inverted and then right shifted by 1.5.
Hence,
E x(-t) = 1_: !
x(-t)] dt 1_: x (x) dx E:r
2 2
= =
oo 2 1
dt = ~
Joo x (x) dx =Ex/a
2
Ex(at) =
J -oo [x(at )] - oo
Ex(at-b) =
oo b)] 2
1 Joo
dt = - 2
x (x) dx =Ex/a,
J [x(at -
-oo a -oo
2 2
Ex(t/a) = 1_:[x(l/a)] dt = a1_: x (x) dt. = aEx
1.2-5. (a) Calling y(t) = 2x(-3t + 1) = t(u(-t-1)-u(-t + l)), MATLAB is used to sketch
y(t).
>> t = [-1.5 : . 001:1.5); y = inline('t.*((t<=-1)-(t<=1))');
>> plot(t,y(t),'k-'); axis([-1.5 1.5 -1.1 1.1));
>> xlabel( 't') ; ylabel('2x(-3t+1)');
1.2-6. MATLAB is used to compute each sketch. Notice that the unit step is in the exponent
of the function x(t) .
28
...
..•
...
0.2
~ ...,..•
-o.•
-06
-0.1
-·
~ ts ... -0.5 ...
Figure Sl.2-5a: P lot of 2x(- 3t + 1) .
...
•..
0.3
...
i'
..0
-0 •
... ,
-0.J
. ..
-05
..3 _,
Figure Sl.2-5b: Plot of x(t) = 0.5y(-t/3 + 1/3).
29
0.3
0.6
O.•
0.2
...
0.6
o.•
(c) True. If an energy signal x(l) has energy E, then the energy of :c(al) is ~ (a real
and positive).
(d) False. If a power signal x(l) has power P, then the power of x(at) is generally not
~· A counter-example provides a simple proof. Consider the case of x(t) = u(t),
which has P = 0.5. Letting a = 2, x(at) = x(2t) = u(2t) = u(t), which still has
power P = 0.5 and not P/a = P/2.
1.3-3. (a) For periodicity, x 1 (t) = cos{t) = cos{t + T1 ) = x 1 (t + T 1 ) . Since cosine is a
2n-periodic function, T1 = 211. Similarly, :c2 (t) = sin(nt) = sin(7rl + 7r7'2 ) =
sin(n(t + T 2 )) = x 2 (t + T2 ). Thus, 7rT2 = 2nk. The smallest possible value is
T2 = 2. Thus,
Ti = 271" and T2 = 2.
{b) Periodicity requires x3(t) = x3(t + T3) or cos( t) +sin( trt) = cos(t + T3) +sin(trt +
trT3 ) . This requires T3 = 2trk 1 and trT3 = 2trk2 for some integers k1 and k2.
Combining, periodicity thus requires T 3 = 2nk1 = 2k2 or 7f =kif k2 . However, 7f
is irrational. Thus, no suitable k1 and k2 exist, and x3(t) cannot be periodic.
(c)
30
1
J;2" cos2 ( t )dt
2~ (o.5(t + sin(t) cos(t))l~:0)
2 11
...1.. !27T
21f 22
= !2
~ f 0 sin2 ( ?Tt)dt
~ ( 2~(7rt-sin(7rt)cos(7rt))l:=o)
~ 2~ 27T = ~
limt-- 00 2~. J'!:r
(cos(t) + sin(7rt)) dt
2
1.3-4. No, f(t) =sin(wt) is not guaranteed to be a periodic function for an arbitrary constant
w. Specifically, if w is purely imaginary then f(t) is in the form of hyperbolic sine,
which is not a periodic function. For example, if w = J then f(t) = JSinh(t). Only
when w is constrained to be real will f(t) be periodic.
E = Ex ~ 10417 = 0. 0579 .
Yi 18 18
(b) Since y 2 (t) is just a (Ty 2 = 4)-periodic replication of x(t), the power is easily
obtained as
Ex
Py 2 = - = -Ex ~= 0.2604.
Ty, 4
Py, = .J.
1
f 2 2
1 (t ) dt = ;)
Thus,
P 11 , = 1/5.
A sketch of y 1 (t) over -3 :S: t :S: 3 is created using MATLAB.
>> t = (-3: .001:3]; mt= mod(t,2);
>> y_l = (mt<=1).•(mt.-2) + (mt>l).•((mt-2).-2);
>> plot(t,y_l,'k'); xlabel('t'); ylabel('y_l(t)'); axis tight;
31
...
_,
Figure Sl.3-6a: P lot of v1(t)
(b) Let
k 1 st< 1.5
t2 OS t < l
y,(t) = { -y2(-t) Vt
Y2(t + 3) Vt
With this form, y 2 (t) is odd and (T2 = 3)-periodic. The constant k is determined
by constraining the power to be unity, Pl/2 = 1=A (k 2
+ ~('l :=o)· Solving for
k yields k 2
=3- 2/ 5 = 13/5 or k = fi375. Thus,
1 st< 1.5
0$l<1
Vt
Vt
32
...
...
-0.5
. -·
- 1.S
_, _,
-> '---~-~-~--~--'--~
-1
2
:i
t~ 5 7 0 5 7 t~
(d )
Figure Sl.4-1
1.4-2.
33
(a) The impulse is located at T =t and x(r) at T =tis x(t). Therefore
1: x(r)o(l - r) dT = x(t).
1: O(T)x(t - T) dT = X(t).
Using similar arguments, we obtain
(c) 1
(d) 0
(e) e3
(f) 5
(g) x(-1)
(h) -e2
1.4-5. For sketches, refer to Figure Sl.4-5.
(a) Recall t hat the derivative of a funct ion at the jump discontinuity is equal to an
impulse of strength equal to t he amount of discontinuity. Hence, dx/dt contains
impulses 4o(t+4) and 2o(t-2). In addition, the derivative is -1 over the interval
(-4, 0), and is 1 over the interval (0, 2). The derivative is zero for t < - 4 and
t > 2. The result is sketched in Figure Sl.4-5(a).
{b) Using the procedure in part (a), Figure Sl.4-5(b) depicts <f2x/dt2 for the signal
in Figure Pl.4-2a.
'"@
i
-1 ...,-_4___o-t
Figure Sl .4-5
y(t) = fo
1
1 dx + /t (-1) dx =2- t 1 :=:; t < 3.
At t = 3, the impulse (of strength unity) yields an additional term of unity. Thus
(assuming t:--+ 0),
!3-e
1 1
1 t
y(t)= Idx+ (- l)dx+ o(x-3)dx=l+(-2) + 1 = 0 t>3
0 l 3-e
34
(b)
Figure Sl.4-6
j~
00
Therefore
o(t) = o(-t).
1.4-8. Letting at = x, we obtain (for a> 0)
J _
oo if>(t)o(at) dt If"'_ ° if>(-xa )o(x) dx = -1a 4>(0)
= -a
00 00
J_
oo
00
if>(t)o(at)dt 1
= lal</>(O) 1 _
= lal !""
00
</>(t)o(t)dt
Therefore
1
o(at) = j;ijo( t)
1.4-9.
0- J ¢(t)o(t) dt = - ¢(0)
35
(e) s = 2
(f) 5 = 5e0 t so thats= 0.
..J3 - •..-'!-
0
0
-J~ 0
Figure Sl.4-10
(b) Xe(t) = 0.5!tu(t) - tu(-t) ] = 0.5Jtl and x 0 (t) = 0.5!tu(t) + tu(-t)] = 0.5t.
(c) xe(t) = 0.5!sinwol+sin(-wot)J = 0 and x 0 (t) = 0.5[sinwot-sin(-wot)] =sin wot.
(d) xe(t) = 0.5[cos wot+ cos(-wot)J = cos wot and Xo(t) = 0.5!coswot - cos(-wot)J =
o.
(e) cos(w0 t+B) = cosw0 tcosO-sinw0 tsinB. Hence Xe(t) = cosw0 tcosB and x 0 (t) =
- sinw0 tsin0
(f) Xe(t) = 0.5[sinwotu(t) + sin(-w0 t)u.( - t)] = 0.5[sinwotu(t) - sin wotu(-t)] and
x 0 (t) = 0.5[sinw0 tu(t) - sin(-w0 t)u.( - t)] = 0.5[sinwotu(t) + sinwotu(-t)] =
0.5sinw0 t.
(g) Xe(t) = 0.5[cosw0 tu(l) + cos(-w0 t)u(-t)] = 0.5[cosw0 tu(t) + coswotu(-t)] =
0.5cosw0 t and x 0 (t) = 0.5[cosw0 tu(t) - cos(- wot)u(-t)] = 0.5[coswotu(t) -
cos wot u( - t)].
Figure Sl.5-1
1.5-2. (a)
1
x.(t) = 2
iu(t) + e21 u(-t)j
2re-
xo(t) = ~[e- 21 u(t) - e2 tu(-t)J
36
(b) Ex. = J~00 x~(t)dt. Because e- 2 tu(t) and e2 tu(-t) are disjoint in time, the
cross-product term in x~(t) is zero. Hence,
~ [ ('° e- 4tdt + f ~.
00 0
Ex.= f x;(t)dt = e4 tdt] =
Loo 4 Jo }_ 00 8
Also,
Hence,
(c) To generalize this result, we first consider causal x(t). In this case, x(t) and x(-t)
are disjoint. Moreover, energy or x(t) is identical to that of x(-t). Hence,
Identical arguments hold for anti-causal signals. Thus, for anti-causal signal x(t)
1.5-3. (a)
1
4fx(t) + x(-t))[x(t) - x(-t)]
1
4llx(t)l 2 - lx(-t)l 2 J
Since the areas under lx(t)l2 and lx(-t)l2 are identical, it follows that
1: x.(t)x 0 (t)dt =0
(b)
r '°
} _ Xc(t)dt
r=
= 2i }_= x(t)dt + 2
i
./_co x(-t)dt
r=
00
37
Because the areas under x (t) and x(-t) are identical, it follows that
It is worth pointing out that sometimes the unit step is defined as u(t) =
1 t>0
{
=
0.5 t 0 . Using this alternate definition, x 6 (t) is periodic.
0 t< O
1.5-6. (a) Using the figure, x(t} = (t+ l }(u(t+ 1)-u(t)) +(- t+ 1)(u(t)-u(t- 1)) . MATLAB
is used to plot v(t) = 3x ( - ~(t + 1))
>> x = inline ('(t+1). * ((t>=- l)&(t<0))+( -t+ l ) .* ((t>=O)&(t< l ))');
>> t = (- 5: .001:5); v = 3•x(-0.5*(t+l));
>> plot(t,v,'k-') ; xlabel('t ') ; ylabel('v(t)') ;
38
(c) Using l.5-6a, v(t) = (1.5t+4.5)( u(t+3)-u(t+ l)) + (- l.5t+l.5)(u(t+ l)-u(t - 1)).
MATLAB is used to determine and plot ve(t).
>> v_str = ['(l.5*t+4.5).*((t>=-3)&(t<- 1))+', .. .
'(-1.5*t+l.5) . *((t>=- 1)&(t<l))'];
>> v = inline(v_str); t = (- 4:.001:4); v_e = 0.5*(v(t)+v(-t)) ;
>> plot(t,v_e,'k'); xlabel('t'); ylabel('v_e(t)');
>> axis((-4 4 -.25 1 . 75));
..•
Thus,
3t/4 + 9/4 - 3 $t<- l
1.5 - 1$ t < 1
Ve(t ) = -3t/4 + 9/ 4 1~ t <3
{
0 otherwise
(d) Using l.5-6a, v(t) = (1.5t+4.5)(u(t+3) -u(t+l))+(-l.5t+l.5)(u(t+l)-u(t -1)).
MATLAB is used to create the four desired plots.
>> v_str = ['(1.5*t+4.5). * ((t>=- 3)&(t<-1))+' , ...
'(- 1.S*t+l.5).*((t>=-l)&(t<l))'];
>> v = inline(v_str);
>> t = (-6:.001:1); a= 2; b = 3; ax= (-6 1 - .5 9.5);
>> subplot(221); plot(t,v(a*t+b),'k-'); grid;
>> xlabel('t'); ylabel('v(at+b)'); axis(ax);
>> subplot(222); plot(t,v(a*t)+b,'k-'); grid;
>> xlabel('t'); ylabel( ' v(at) +b'); axis(ax);
>> subplot(223); plot(t,a*v(t+b),'k- '); grid;
>> xlabel('t'); ylabel('av(t+b)'); axis(ax);
>> subplot(224); plot(t,a*v(t)+b,'k-'); grid;
>> xlabel('t'); ylabel('av(t)+b'); axis(ax);
(e) Using l.5-6a, v( t) = (1.5t+4 .5)(u(t +3)-u(t+l)) +( - 1.5t+ 1.5 )(u(t+ l)-u(t - 1)).
MATLAB is used to create t he four desired plots.
>> v_str ={'(1.5*t+4.5).•((t>=- 3)&(t<-1))+', . . .
'(-1.5•t+1.5). * ((t>=-1)&(t<l))'];
>> v = inl i ne(v_str);
» t (-3: . 001:3); a= -3; b = -2; ax= (-3 3 -11.5 3.5);
39
" !
~·
.. ic
... .,
-· ~
. ..
& .•.•
..t • ................
t. . . . . ~
•2 ...
• .__
•• - -
·· ·- ··-· · - ··-· •_ _• .._
i ••••
__,
Figure Sl.5-6d: Plots of v(2t + 3), v(2t) + 3, 2v(t + 3), and 2v(t) + 3.
n -2
i' _,
T
~
-· .
-10 ·10
·•
-3
' --
-2
-
-1
-- ' - - - - - ' '--------'
-3 •2 .. ,
0 .. ' .. , . . . . .
- _,
{:
-·
-10 ·10
•ll ••
-3~----'---------'
-2 -1 -) -2 _,
Figure Sl.5-6e: Plots of v(-3t - 2), v(-3t) - 2, -3v(l - 2), and -3v(t) - 2.
1.5-7. (a) Using the figure, y(t) = t(u(t) - u(t - 1)) + (u(t - I) - u(t - 2)). MATLAB is
used to plot y 0 (t) = y(t) - 2v(-t) .
>> y = inline('t.*((t>=O)&(t<1))+((t>=l)k(t<2))');
>> t = (-5:.001:5); y_o = (y(t)-y(-t))/2;
>> plot(t,y_o,'k-'); xlabel('t'); ylabel('y_o(t)'); axis([-5 5 -.6 .6]);
40
...
•.
..,
..,
0.1
~· 0
·<U
-0.2
-··.....
....
-s ... -· -2 ·1
Thus,
-1/2 -2 $ t < -1
t/2 -1 $ t < 1
Yo(t) = 1/2 l $ t<2
{
0 otherwise
(b) Since y(t) = 0.2x(-2t-3), 5y(- 0.5t - l.5) = 5(0.2)x (- 2( - 0.5t- l.5) - 3) = x(t).
MATLAB is used to sketch x(t).
>> y = inline('t . *((t>=O)k(t<1))+((t>=1)k(t<2))');
>> t = (-8:.00 1: 0) ; x = 5•y(-0.5•t-1.5);
>> plot(t,x,'k-'); xlabel( 't'); ylabel('x(t)'); axis((- 8 0 - .5 5.5));
Thus ,
5 -7 $ t < - 5
x(t) = -5(t + 3)/2 - 5 $ t < - 3
{ 0 otherwise
41
(a) Since y(t) = -0.5x(-3t+2), -2y(-t/3+2/3) = -2(-0.5)x( - 3( - t/3+2/3)+2) =
x(t). MATLAB is used to sketch x(t).
>> y = inline('((t>=-1)&(t<O))+(-t+1).*((t>=O)&(t<1))');
>> t = [-2:.001:6); x = -2*y(-t/3+2/3);
>> plot(t,x,'k-'); xlabel('t'); ylabel('x(t)'); axis([-2 6 -2.5 0 . 5));
o.s.--~-~-~-~-~-~-~-~
-o.s
~ -1
-1.6
-2
-Z S '---'---'--~-~---'---'--~--'
-z -1
Thus,
- 2(t + 1)/ 3 -1 :5 t < 2
x(t) = -2 2 :5t<5
{ 0 ot herwise
(b) The even portion of x(t) is Xe(t ) = 0.5(x(t) + x(- t )).
>> x = inline('-2*(t+l)/3.*((t>=-1)&(t<2))-2*((t>=2)&(t<5))');
>> t = [-6: .001:6); x_ e = 0.5*(x(t)+x(-t));
>> plot(t,x_e,'k-'); xlabel( ' t ' ); ylabel('x_e (t)'); axis([-6 6 -1.5 0 . 5));
os .---~--~--~--~-~-~
o' - -
~- -0.5
- 1 ,$ '---~--~-~---'--~---'
-6 -2
42
Thus,
- 1 2 $ ltl < 5
(- ltl -1)/3 1 $ Jtl<2
x.(t) ={ -2/3 ltl < 1
0 otherwise
••
~- 0
.....
.,
•\,$'-----'---~-~~---'---~----'
-6 -2
Thus,
1 -5 $ t < - 2
(- t + 1)/3 -2$t<- l
-2t/3 - 1$ t < l
Xo(t) = -(t + 1)/3 1$l<2
-1 2$t<5
0 otherwise
1.5-9. Notice, w~8 ( -t) = 0.5(w(-t) + w•(t))' = 0.5(w*(-t) + w (t)) = Wcs(t). In Cartesian
form , this becomes w;. (- l) = x(-t) - JY( - t) = x (t) + JY(t) = Wcs(t). Equating
the real portions yields x(-t) = x(t), and equating the imaginary portions yields
- y( - t) = y(t). Thus, by definition, the real portion of Wcs(t) is even and the imaginary
portion of Wc 8 (t) is odd.
43
Consider, for example, point (2,1), which now corresponds to both t = 1 and
t = -1.
ii. If w(t) is odd, w(t) = -w( - t), it is odcl in both the real and imaginary
components. Thus, the graph reflects about both the real and imaginary
axes.
iii. If w(t) is conjugate symmetric, w(t) = w*( - t), it is even in the real compo-
nent and odd in the imaginary component. Thus, the graph reflects about
the real axis.
iv. If w(t) is conjugate antisymmetric, w(t) = -w*(- t), it is odd in the real
component and even in the imaginary component. Thus, the graph reflects
about the imaginary axis.
w(l}lt[v&(1 ""1:t)l$0dd
1.5 1.5
'
/,;.
fa
0.5 0.5
! 0 0 s 0 0
-0.5
~ -
-o.S
•1
"
-1.5
-2 -1
...
0
-1.5
-2 ., 0
r-
Ro
'·'
~
'
..•
.!i .,
0 s 0
-0.5
I
....
_,
-1.s
-2 _, 0
Ro
-1.!I
-i .,
..
0
(b) Since w(t) is only given over (0 :=:; t :=:; 1), w(3t) can be determined only for
(0 S t S 1/3). Since the function does not change, only the time at which it
occurs, the complex-plane graph of w(3t) looks identical to the original complex-
plane graph of w(t) with the exception that the points are assigned different
times. For example, point (2,1) occurs now at t = 1/3.
l.5-12. (a) We know x(t) = t2(1 + J) over (1 :=:; t :=:; 2). Since x(t) is skew-Hermitian,
x(t) = -x*(- t) and thus x(t) = - t2(1- J) over (- 2 :=:; t::; - 1). To minimize
energy, x(t) is set to zero everywhere else. Thus,
t 2 (1 + J) lStS2
:z:(t) = -t 2 (~ - J) -2St S -1
{
otherwise
44
-·
·2
-·
-2.s ..z -1.s -<>.5 0.5 ... 2.S
~ o l-------~
-·
-·
-·
•• ... ..
Figure Sl .5-12c: Plot of z{ t) = Real {Jx( -2t + 1)}.
(d) Since x(t) is finite duration, P,, = 0. Using symmetry, Ex= 2 f 12 (t2 (1 + J))(t2 (1-
J))dt = 2 f 1 2t 4 dt =
2 4
t 2
1
t= l
= ~(32 - 1) = 1; 4 . Thus,
124
E,, = 5 = 24 .8 and P,, = 0.
1.6-1. U x(t) and y(l) are t.he input and output, respectively, of an ideal integrator, then
y(t) = x(t)
45
and
0
y(t) = ft x (r)dr=f x(r)dr+ f' x(r)dr= y(O) + f'x(r)dr
-co -oo lo '-v-' lo
zero-input "-....-"
zero-state
and
1
v(t)=M ft -oo
1
x(r)dr=M f 0
-oo
1
x(r)dr+M 1t
0
1
x(r)dr=v(O)+M 1t0
x(r )dr
1.7-1. Only (b), (f), and (h) are linear. All the remaining are nonlinear. This can be verified
by using the procedure discussed in Example 1.9.
1.7-2. (a) The system is time-invariant because the input x(t) yields the output y(t) =
x(t - 2) . Hence, if the input is x(t - T), the out put is x(t - T - 2) = y(t - T),
which makes the system time-invariant.
(b) The system is time-varying. The input x(t) yields the output y(t) = x(-t).
Thus, the output is obtained by changing the sign oft in x(t). Therefore, when
the input is x(t - T) , the output is x(-t - T) = x(-[t + T]) = y(t + T) , which
represents the original output advanced by T (not delayed by T) .
(c) The system is t ime-varying. The input x(t) yields the output y(t) = x(at), which
is a scaled version of the input. Thus, the output is obtained by replacing t in
the input with at. Thus, if the input is x(t - T) (x(t) delayed by T), the output
is x(at - T) = x(a[t - ~J), which is x(at) delayed by T/a (not T). Hence the
system is time-vary ing.
(d) The system is time-varying. The input x(t) yields the output y(t) = tx(t). For
the input x(t-T), th e output is tx(t-T) , which is not tx(t) delayed by T. Hence
the system is time-varying.
(e) The system is time-varying. The output is a constant, given by the area under
x(t) over the interval !ti ~ 5. Now, if x(t) is delayed by T , the output, which is
the area under the delayed x(t), is another constant. But this output is not the
same as the original output delayed by T. Hence the system is time-varying.
(f) The system is time-invariant. The input x(t) y ields the output y(l), which is the
square of the second derivative of x(t). If the input is delayed by T, the output
is also delayed by T. Hence the system is time-invariant.
1.7-3. We construct the table below from the first three rows of data. Let rj denote t he jth
row.
46
In our case, the input x(t) = u(t + 5) - u(t - 5). From r 7 and the superposition and
time-invariance properties, we have
Figure Sl.7-5
l_
1.7-6. The solution is trivial. The input is a current source, which has infinite impedance.
Hence, as far as the output y(t) is concerned, the circuit behaves as shown in Figure
Sl.7-6. The nonlinear elements are irrelevant in computing the output y(t), and the
output y(t) satisfies the linearity condit ions. Yet, the circuit is nonlinear because it
contains nonlinear elements. o ·1 H
1
Figure Sl.7-6
1.7-7. (a) y(t) = x (t- 2). Thus, the output y(l) always starts after the input by 2 seconds
(see Figure Sl.7-7a). Clearly, the system is causal.
47
(b) y(t) = x(-t). The output y(t) is obtained by time inversion in the input. Thus,
if the input starts at t = 0, the output starts before t = 0 (see Figure Sl.7-7b).
Hence, the system is not causal.
(c) y(t) = x(at), a> 1. The output y(t) is obtained by time compression of the input
by factor a. Hence, the output can start before the input (see Figure Sl.7-7c),
and the system is not causal.
(d) y(t) = x(at), a< 1. The output y(t) is obtained by time expansion of the input
by factor 1/a. Hence, the output can start before the input (see Figure S l.7-7d),
and the system is not causal.
:
,
ylt)
\
yLt::)
)' .
0 0 t: -:. 0 -. ~ -"l: 0 t,.-:a
c.h) a. (cf)
Figure Sl.7-7
1.7-8. (a) Invertible because the input can be obtained by taking the derivative of the
output. Hence, the inverse system equation is y(t) = dx/dt.
(b) Not invertible for even values of n, because the sign information is lost. However,
the system is invertible for odd values of n. The inverse system equation is
y(t) = [x(t)]1fn.
(c) Not invertible because differentiation operation irretrievable loses the constant
pa.rt of x(t).
(d) The system y(t) = x(3t - 6) = x(3{t - 2]) represents an operation of signal
compression by factor 3, and then time delay by 2 seconds. Hence, the input
can be obtained from the output by first advancing t he output by 2 seconds,
and then time-expanding by factor 3. Hence, the inverse system equation is
y(t) = x(~ + 2). Although the system is invertible, it is not reali~able because it
involves the operation of signal compression and signal advancing (which makes
it noncausal). However, if we can accept time delay, we can realize a noncausal
system.
(e) Not invertible because cosine is a multiple valued function, and cos- 1 [x(t)J is not
unique.
(f) Invertible. x(t) =In y(t).
1.7-9. (a) Yes, the system is linear. Begin assuming y 1 (t) = r(t)x1(t) and Y2(t) = r(t)x2(l).
Applying ax 1 (t) + bx 2 (t) to the system yields y(t) = r(t) (ax 1 (l) + bx 2 (t)) =
ar(t)x1 (t) + br(t)x2(t) = ay1(t) + by2(t).
(b) Yes, the system is memoryless. By inspection, it is clear that the syst,em only
depends on the current input.
48
(c) Yes, the system is causal. Since the system is memoryless, the system cannot
depend on future values and must be causal.
(d) No, the system is not time-invariant. Since the system function depends on
the independent variable t, it is unlikely that the system is time-invariant. To
explicitly verify, let y(t) = r(t)x(t). Next., delay x(t) by ,,. to obtain a new
input x 2 = x(t - T). Applying x 2 (t) to the system yields y 2 (t) = r(t)x2(t) =
r(t)x(t - 'T) # r(t - 'T)x(t - T) = y(t - T). Since, the system operator and the
time-shift operator do not commute, the system is not time-invariant..
1.7-10. Using the sifting property, this system operation is rewritten as y(t)
0.5 (x(t) - x(-t)).
(a) This system extracts the odd portion of the input.
(b) Yes, the system is BIBO stable. If the input is bounded, then the output is neces-
sarily bounded. That is, if Jx(t)l :::; M:z: < oo, then ly(t)l = J0.5 (x(t) - x(-t)) I :::;
0.5 (lx(t) I + I - x(-t)I) ::=; Mx < oo.
(c) Yes, the system is linear. Let Y1(t) = 0.5(x1(t) - x1(-t)) and y2(l) =
0.5(x 2 (t) - x 2 (-t)). Applying ax 1 (t) + bx2(t) to the system yields y(t) =
0.5 (ax 1 (t) + bx2(t) - (a.x 1 (-t) + bx2(-t))) = 0.5a(x 1 (t) - X1(-t))+0.5b(x2(t) -
x2(-t)) = ay1 (t) + byi(t).
(d) No, the system is not memoryless. For example, at time t = 1 the output
y(l) = 0.5(x(l) - x(-1)) depends on a past value of the input, x(-1).
(e) No, the system is not causal. For example, at time t = -1 the out.put y(-1) =
0.5(x( - 1) - x( 1)) depends on a future value of the input, x(l).
(f) No, the system is not time-invariant. For example, let the input be x(t) =
t(u(t + l) -u(t - 1)). Since this input is already odd, the output is just the input,
y(t) = x(t). Shifting by a non-zero,,., x(t - T) is not odd, and the output is not
y(t - T) = x(t - T). Thus, the system cannot be time-invariant.
1.7-11. (a) No, the system is not BIBO stable. The system returns the time-delayed deriva-
tive, or slope, of the input signal. A square-wave is a bounded signal which, due
to point discontinuities, has infinite slope at certain instants in time. Thus, a
bounded input may not result in a bounded output, and the system cannot be
BIBO stable.
(b) Yes, the system is linear. Begin assuming y 1 (t) = fltx
1 (t - 1) and
Y2(t) = f,,x 2 (t - 1). Applying ax 1 (t) + bx 2(t) to the system yields y(t) =
ft
(ax1 (t - 1) + bx2(t - 1)) = a1tx1 (t - 1) + bf,,x2(t - 1) = ay 1 (t) + by2(l).
(c) No, the system is not memoryless. By inspection, it is clear that the system
depends on a past value of the input. For example, at t = 0, the output y(O)
depends on the time-derivative of x( - 1), a past value.
(d) Yes, the system is causal. By inspection, it is clear that the system does not
depend on future values.
(e) Yes, the system is time-invariant. To explicitly verify, let y(t) = f,,x(t - 1). Next,
delay x(t) by,,. to obtain a new input x 2 = x(t-r). Applying x 2 (t) to the system
yields Y2(t) = f,,x2(t) = -fix(t - 1 - 'T) = y(t - T). Since, t he system operator
and the time-shift operator commute, the system is time-invariant. In more loose
terms, the derivative operator returns the delayed slope of a. signal independent
of when that signal is applied.
49
1.7-12. (a) Yes, the system is BIBO stable. From the definition of the system, we know that
y(t) is either x(t) or 0. Correspondingly, if lx(t)I < oo then ly(t)i < oo, and the
system must be BIBO stable.
(b) No, the system is not linear. Consider two signals: x 1 (t) = 1 and x2(t) =
cos(t). The corresponding outputs of these individual signals is y 1 (t) = l and
cos(t) if cos(t) > 0 . . .
Y2(t) = { if cos(t) ~ . However, 1f we create a third input x 3(t) =
0 0
xi(t) + x2(t), the system output is y3(t) = 1 + cos(t) i= y1(t) + Y2(t). Since
superposition does not apply, the system cannot be linear.
(c) Yes, the system is memoryless. By inspection, it is clear that the system only
depends on the current input.
(d) Yes, the system is causal. Since the system is memoryless, the system cannot
depend on future values and must be causal.
(e) Yes, the system is time-invariant. Consider delaying x(t) by T to obtain a
new input x2 = x(t - 7). Applying x 2(t) to the system yields y2(t) =
x(t - 7) if x(t - 7) > 0 .
{ 0
if x(l _ T) ~
0
= y(t - 7). Smee the system operator and the
time-shift operator commute., the system is time-invariant.
1.7-13. (a) No, Bill is not correct. The x 1 (3t) term represents a compression rather than the
necessary dilation. One way to construct x 2 (t) is x 2 (t) = 2x1 (t/3) - x 1 (t - I).
However, this form is not unique; x 2 (t) = 2x 1(t) + x 1 (t - 1) + 2x 1 (t - 2) also
works and may be more useful.
(b) The output y 1 (t) is given for the input signal x 1 (t). The expression x 2 (t) =
2x 1 ( t)+x 1(t - 1) +2x 1 (t-2) forms x 2 (t) from a superposition of scaled and shifted
copies of x i (t). Since the system is linear and time invariant, the operations
of scaling, summing, and shifting commute with t.he system operator. Thus,
Y2(t) = 2y1(t) + y 1 (t - 1) + 2y1 (t - 2). Notice, it is not true that y 2(t) =
2y1 ( t/3) - !11 ( t - 1 ); linearity and time-invariance cannot help with the time-
scaling operation. MATLAB is used to plot Yz(l).
>> t = (-1:.001: 4); y_1 = inline('t.•((t>=O)&(t<1))+(t>=1)') ;
>> y_2 = 2*y_1(t) +y_1(t-1)+2•y_1(t-2);
>> plot(t,y_2,'k-'); xlabel('t'); ylabel('y_2(t)'); axis((-1 4 -.5 5.5));
'
Figure Sl.7-13b: Plot of y 2 (t) = 2y 1(t) + y 1 (l - 1) + 2y1 (t - 2).
50
1.8-1. The loop equation for the circuit is
Also
1
Dy1 (t) = Y2(t) ~ Y1(t) = DY2(t) (2)
3
(D; ) y 2 (t) = x(t) or (D + 3)y2(t) = Dx(t)
1.8-2. The currents in the resistor , capacitor and inductor are 2y2(t), Dy 2 (t) and (2/ D}y2 (t),
respectively. Therefore
2
(D + 2 + D)y2(t) = x(t)
or
(D 2 + 2D + 2)y2(t) = Dx(t) (1)
Also
1
y 1 (t) = Dy2(t) or Y2(t) = Dyi(t) (2)
JS c~-Y) KO<-y)
Figure Sl.8-3
(1)
51
Substituting of (I ) in (2) yields
I<1
------x(t) = (J D 2 + BD)B(t)
DL1 + R1
or
(JD 2 + BD)(DL1 + R1)8(t) = I<1x(t)
1.8-5.
But
qo(t) = Rh(t) (2)
Differentiation of (2) yields
. R
<io(t) = Rh(t) = A[q;(t) - <Jo(t)]
and
( D + ~) qo(t) = ~q;(t)
or
R
(D + a)qo(t) = aqi(t) a= - (3)
A
and
- 1 (1 a ) D
h(t) =A - D +a q;(t) = A(D +a) q;(t)
or
1
(D + a)h(t) = Aq;(t)
1.8-6. (a) The order of the system is zero; there are no energy storage components such as
capacitors or inductors.
(b) Using KVL on the !eh loop yields x(t) = R 1 y 1 (t) + Rz(Y1(t)-y2(t)) = 3y1(t) -
2y2(t). KVL on the middle loop yields 0 = R2 (y2 (t)-y1 (t))+R3 y 2(t)+R4 (y2(t)-
y3(t)) = -2y 1(t)+9y2(t)-4y3 (t) . Finally, KVL on the right loop yields R 4 (y 3 (t)-
Y2(t)) + (R5 + ~)y3(t) = -4y2(t) + 15y3(t). Combining together yields
52
(c) Cramer's rule suggests
3 -2 x(t)
-2 9 0
0 -4 0
y3(t) =
3 -2 0
-2 9 -4
0 -4 15
MATLAB computes the denominator determinant.
>> det([3 -2 0;-2 9 -4;0 - 4 15))
ans = 297
The numerator determinant is easy computed by hand as 0 + 0 + 8x(t) = 8x(t).
Thus,
8 8
y3(t) = x(t) = (2 - I cos(t)I) u(t - 1).
297 297
1.10-1. From Figure Pl.8-2, we obtain
x(t) = qif2 + Q1 + Q2
Moreover, the capacitor voltage q1 (t) equals the voltage across the inductor, which is
~<h· Hence, the state equations are
(1)
53
Chapter 2 Solutions
2.2-1. The characteristic polynomial is >.. 2 +5>..+6. The characteristic equation is >..2 +5>..+6 =
0. Also >. 2 + 5>. + 6 = (>.. + 2)(>.. + 3). Therefore the characteristic roots are >. 1 = - 2
and >.. 2 = -3. The characteristic modes are e- 2 t and e- 3 t. Therefore
and
lio(t) = -2c1e- 2 t - 3c2e- 3 t
Setting t = 0, and substituting initial conditions yo(O) = 2, lio(O) = -1 in this equation
yields
2.2-2. The characteristic polynomial is >.. 2 +4>..+4. The characteristic equation is >. 2 +4>.+4 =
0. Also >. 2 +4>.+4 = (>.+2) 2 , so that the characteristic roots are -2 and - 2 (repeated
twice). The characteristic modes are e- 21 and te- 2 c. Therefore
1 = C1 + C2 } ==:}
1 = -Cz
54
2.2-4. The characteristic polynomial is >. 2 + 9. The characteristic equation is >. 2 + 9 = O or
(>. + j3)(>. - j3) = 0. The characteristic roots are ±j3. The characteristic modes are
ei 3 t and e- i 3 t. Therefore
Yo(t) = ccos(3t + B)
and
Yo(t) = -3csin(3t + 8)
Setting t = 0, and substituting initial conditions yields
0 = ccosB } ccos8=0} c= 2
6 = -3csin B => csin8 = -2 ==> 8 = - 7r/2
Therefore 7r .
y0 (t) = 2cos(3l- 2') = 2sm3t
2.2-5. The characteristic polynomial is >. 2 +4>.+13. The characteristic equation is >. 2 +4>. +
13 = 0 or (>. + 2 - j3)(>. + 2 + j3) = 0. The characteristic roots are -2 ± j3. The
characteristic modes are c 1e<-2+i3)t and c2e<- 2 -i3)t. Therefore
5 = ccosB } ccos B = 5 c = 10
15.98 = -2ccosB- 3csinB => csin8 = -8.66 } => B=--rr/3
and
vo(t)
Yo(t)
Setting t = 0, and substituting initial conditions yields
4 =CJ+ C3 } C1 =5
3 = C2 - C3 ==} c2 =2
-1 = C3 C3 = -1
2.2-7. The characteristic polynomial is(>.+ l )(>. 2 + 5>. + 6) . The characteristic equation 1s
(>. + l)(>. 2 + 5>. + 6) = 0 or (>. + l)(>. + 2)(>. + 3) = 0. The characteristic roots are -1 ,
-2 and -3. The characteristic modes are e-t, e- 2t and e- 3 t. Therefore
Yo(t) = c1e-t + c2e- 2t + c3e- 3t
and
55
Set.t.ing t = 0, and substituting initial conditions yields
Ct = 6
~=-7
C3 =3
Therefore Yo(t) = 6e-t - 1e-2t + 3e-3t
2.2-8. The zero-input response for a LTIC system is given as y0 (t) = 2e-t + 3. Since two
modes are visible, the system must have, at least, the characteristic roots >.1 = 0 and
=
>-2 -1.
(a) No, it is not possible for the system's charact.eristic equation to be >. +1 = 0
since t.he required mode at >. = 0 is missing.
(b) Yes, it is possible for the system's characteristic equation to be J3(>. 2 + >.) = 0
=
since this equation has the two required roots >. 1 0 and >-2 = -1.
(c) Yes, it is possible for the system's characteristic equation to be>.(>. 1-1)2 = 0. This
equal.ion supports a general zero-input response of y 0(t) = c1 +c2e-t +c3te-t. By
letting c1 = 3, c2 = 2, and c3 = 0, the observed zero-input response is possible.
2.3-1. The characteristic equation is >. 2 + 4>. + 3 = (>. + l)(>. + 3) = 0. The characteristic
modes are e-t and e- 31 . Therefore
Therefore 1
Yn(l) = 2(e-t - e-3t)
C1 = 1
C2 = -1
Therefore
and
[P(D)yn(t)Ju(t) = !Yn(l) + 1y,..(l) + lly,.(t)]u(l) = (e- 2 t; e- 3t)u(t)
Hence
h(t) = bno(t) + [P(D)yn(t)]u(t) = o(t) + (e- 2t + e- 3 t)u(t)
56
2.3-3. The characteristic equation is .A + l = 0 and
Yn(t) = ce-t
In this case the initial condition is y~- 1 (0) = Yn (0) = 1. Setting t = 0, and using
Yn(O) = 1, we obtain c = 1, and
Yn(t)= e-t
P (D)yn(t) = 1-!in(t) + y 11 (t)Ju(t) = 2e-tu(t)
2.4-1.
Ac = l: c(t) dt
= A.,
AxA9
1: g( l - T) dt
This property can be readily verified from Examples 2.7 and 2.8. For Example 2.6,
we note that
oo 1
-oo J
e-ot dt = -
a
Use of this result yields A,, = 1, Ah = 0.5, and Ay = 1 - 0.5 = 0.5 = A xAh For
example 2.8, A:c = 2, A 9 = 1.5, and
Ac =
J I
- 1
- -l ( t+l) 2 dt+ /
6
-2 tdt+
1 3
2
1 2
4
- -l ( t 2 -2t-8) dt
6
4 14
= 9- + 1. + -9 = 3 = A"A9
57
2.4-2.
11""
-
a -oo
x(w)g(at-w)dw
l
-c(at) a~ 0
a
When a < 0, the limits of integration become from oo to -oo, which is equivalent to
the limits from -oo to oo with a negative sign. Hence, x(at) * g(at) = l~l c(at).
2.4-3. Let x(t) * g(l) = c(l). Using the time scaling property in Prob. 2.4-2 with a= -1,
we have x(- t) * g( - t) = c(- t). Now, if x(t) and g(t) are both even functions oft,
then x(t) = x( - t) and g(- t) = g(t). Clearly c(t) = c( - t). Using a parallel argument,
we can show that, if both functions are odd, c(t) = c(-t), indicating that c(t) is even.
But if one is odd and the other is even, c(t) = -c(- t), indicating that c(t) is odd.
2.4-4.
-bt
_e_e(b-a)T
I' = -bt
_e_ [e(b-a)t _ l} = e
-at
- e
-bt
b-a 0 b -a a- b
Because both functions are causal, their convolution is zero fort < 0. Therefore
2.4-5. (i)
Therefore
u(t) * u(t) = lu(t)
(ii) Because both functions are causal
e-at 1' dr
= le-at l ~0
and
e- 0 'u(t) * e- 0 'u(l) = te- 0
'u(t)
(iii) Because both functions are causal
58
u(T) = u(T - t) = 1 and
tu(t ) * u(t) = 1t 0
T dT = t
2
2
t 2::. 0
and 1
tu(t) * tt(t) = 2 t2u(t)
2.4-6. (i)
Because rand t-r are both nonnegative (when 0::; r::; t), u(T) =u(t-r) = 1,
and
sintu(t) * n(t) = (lot sinr dr) u(t) = (1 - cost)u(t)
(ii) Similarly
costu(t) * u(t) = (lot cosrdr) u(t) =sin tn(t)
2.4-7. In this problem, we use Table 2.1 to find the desired convolution.
(a) y(t) = h(t) * x(t) = e-tn(t) * u(t) = (1 - e- 1 )u(t)
(b) y(t) = h(t) * x(t) = e-tu(t) * e-tu(t) = te- tu(t)
(c) y(t) = e-tu(t) * e- 2 tu(t) = (e-t - e- 2 t)u(t)
(d) y(t) = sin3tu(t) * e-tu(t)
Here we use pair 12 (Table 2.1) with a = 0, f3 = 3, 8 = - 90° and >. = -1. This
yields
¢ = tan- 1 [ =~] = -108.4°
and
(cos l8.4°)e-t - cos(3t + 18.4°) ( )
sin3tu(t) * e-tu(t) JTii u t
0.9486e-t - cos(3t + 18.4°) ( )
- - -- - - -- - u ·t
JTii
2.4-8. (a)
59
(c)
2.4-9.
2.4-10. (a) For y(t) = 4e- 2tcos3lu(t) *U(t), We use pair 12 with a= 2, {3 = 3, 0 = 0,
.>.. = 0. Therefore
3
¢ = tan- 1 [ ~ ] = -56.31°
and
y(t) = 4 [ cos(56.31°) - ~s(3t + 56.31°)] u(t)
¢ = tan- 1 [ ~3 ] = -71.56°
and
[ cos(7L56°)e-t - ~ cos(3t + 71.56°)] u(t)
y(t) 4
4
hA. [0.316e-t - e- 2 t cos(3t + 71.56°)] u(t)
vlO
4 [e-t - Jioe- 2
tcos(3t+71.56°)] u(t)
60
The response is shown in Figure S2.4-lld.
- - - - - '!
~lf-)
t->
'
Y, (-t-1)
Figure S2.4- lld
-~-
----·--
2.4-12. (a)
0 -c ~ 0
Figure S2.4-l 2b
2.4-14. The output has ihe term e- 3 'u(t) that is not in ihe input. Hence, the input should
have the term e- 3 tu(t). There is also a possibility of an impulse term in the input
61
::x, L+)
l
--:s
~(-1-1)
-3 J
y(n))-:. I
.1
-I 3 -3
-3 3
y(lf- )-= I
-3 -3
-3
-6( A -
that will result in a term of the form e- 2 tu(t) in the output. Let us t ry
2a+ 2b = 4
-2b = 6 }~ a=5
b= - 3
Hence y(t) = SO(t) - 3e- 31u(t).
2.4-15.
- 2 -1
t +}
* u(t) = j"°_ T2
1
- -u(t
+ 1 - r) dr
00
Because u(t - r) = 1 for r < t and is 0 for r > t, we need integrate only up tor = l.
-2-
t
1 jt I
+ } *u(t) = _ -T 2-+-1 dr = tan- 1 rj
1
-oo
= tan- 1 t + -7r
00 2
Figure S2.4-15 shows ,2~ 1 and c(t) (the result of the convolution)
.1
()
0
Figure S2.4-15
62
j -=- .!2
c {t-)
0 · t~
0 1T . c~
Figure 82.4-16
Fort> 4tr (also for t < 0), x(t) * 9(t) = 0. Figure S2.4-17c shows c{t) .
~tt)
(b ) Figure S2.4-17
c(t) = J::t AB dr = A2
8 0 $ t $ 0.5
3
c(t) = j~+t AB dr = AB(l - t) 0.5 $ t $ 1
c(t) =0 t$ -4
63
(d)
c(t)
0 2: t 2: -3
(e)
c(t) = f_o 00
r2~ 1 dr = tan- 1 T lo-oo = ~
(f)
c(t) = .lt-3
r' . e-r dr = e-<t- 3> - e-t t 2: 3
c(t) = 0 t $ 0
(g) This problem is more conveniently solved by inverting x 1 (t) rather than x 2 (t)
c(t) = fct+J (r - t) dr = ~ t 2: 0
c(t) = J;+l(T - t) <fr= 4{1 - t2 ) 0 2: t 2: -1
c(t) =0 for t 2: 0
c(t)
c(t) 02:t2: - l
c(t) - 1 2: t 2: -2
c(t)
2.4-19.
By inspection, we find
Therefore
x(t) * W(t) [o(t) - o(t - 2)J * ~ (-t-1)
-
~ c; 1) -~C;3)
2
64
p I I
J r-1 8 r-1 I
I
A •
I
f
/\
I
I A I
I I
I I
I
. . I I
'C'
I
,
I
I
I I
2
I
I
't
i i t3 .).5+t
J.rt 2 . .5tf
8
A
r -
I
I
I
- 1
I
I
I
fJ
A
r--,
I
I
C(t4\ (. h)
"?;-"> ,, s i-~
0 1' tS.. 0 .}. -) O I -t ~
~ ~"j""
:- -, iI (_C)
-2 >ti D 1;~ -4 - I 0
~ }
)U c- ~o t:~ O
~~
I
\....:... -l.-l~~-==:::t::=::...;._~
~ 0 t 3-tt-
'J_ :;d
0
-11t
o I
(.; )
0 f+I
. 'I (l t
/ , -7;
t'.+ I
Cl.f-)
-I
2,._I__ ..J. '
o~t ~ -I
0 -2.. -1 tJ
Figure S2.4-19 shows x(l) * W(t) ;i .. .. ....
.1
-j
.. .. . , .
Figure S2.4-19
2.4-20. The unit impulse response of an ideal delay of T seconds is h(t) = b(t - T). Using Eq.
(2.48), we obtain
H(s) = f
-oo
00 b(t - T)e- 51 dt = e-sT
For an input x(t) =est, the output of the delay is y(t) = ss(t - T). Hence, according to
Eq. (2.49)
es(t - T)
H(s) = - est
- - = e-sT
65
2.4-21. y(t) = x(t) * h(t). Fort< -1, y(t) = O;
For -1 $ t < 0, y(t) = J~ 1 (T + l)dT = T
2
/2 + Tj:=-l = t 2 /2 + t + 1/2.
Fort 2 0, y(t) = J~ 1 (T + l)dT = 1/2.
Thus,
0 t < -1
y(t)= t2/2+t+1/2 -1$t<O
{
1/2 t20
2.4-22. Using the graph of the system response, h(t) = (-t/2 + l)(u(t) - u(t - 2)) . y(l) =
J.~00 htotal(T)x(l -T)dT. Since x(t) is causal , the upper limit of the integral is one.
F\1rthermore, since h(t) is causal, the total response htotai(t) = h(t) * h(t) is also
causal, which makes the lower limit of the integral zero. Over !O, 1], x(t) = u(t) = 1.
Thus, y(l) = f 0 htotal(T)dT. To compute y(l), it is only necessary to know htota1(t)
1
up tot= 1.
Over (0 $ t < 2), htotal (t) = J;(-T /2 + 1)(-(t - T)/2 + l)dr = J;(-r/2 + l)(T /2 +
1 - l/2)dr =
t3 t2
J;
( - r 2 /4 + r(I - 1 + t/2)/2 + (1 - t/2)) dr = -h
+ ~ + (1 - t/2)t =
24 - 2 + t.
Thus,
3 2 74 73 721 l 11 ~
y(l)=
1o
1
(r /24-7 /2+7)cfr= -
96
- -+-
6 2 r=O
1 1 1
= - - - + - = - = 0.343r5.
96 6 2 32
2.4-23. (a) Using KVL, x(t) = vL(t) + y(t). Also, ic(t) = C!f/t and vL(t) = L~ = L~ =
LC'!/-;J. Combining yields
2 1
>. +LC= 0.
•.
The characteristic roots are
±J
>-12=---.
. vr;c
(c) The form of the zero-input response is y 0 (t) = c1 e>"t + c2e.>. 2 t. Using >. 1 = ->. 2,
ic(O) = 0 = c!fltl
t=O
= C{c1>-1 + c2>-2) = C(c1>-1 - C2A1) = C>-1(C1 - C2). Thus,
c1 = c2 . Also, Vc(O) = 1 = y(O) = c 1 + c2 . Combining yields 2c1 = 2c2 = 1 or
c1 = c2 = 0.5. The zero-input response is thus y 0 (t) = 0.5(e1tf./LC + e-1tf./LC) .
Using Euler's identity yields
Yo(t) = cos(t/.JLC).
66
10 12 1-4 16 Ul
t(LCf,12
c1 = - J0.5. Thus, c2 = J0.5 and iio(t) = e''-;;-'' = sin(t). From this, the system
response is determined to be h(t) = L1c: sin(t)u(t) = sin(t)u(t).
Next, the zero-state response is computed as x(t) * h(t) = sin Te-< 1- r ld'T = J;
( e- t l; Imag (e 1,.. e.,.) dr) u(t) (Imag ( e-t l; er(l+J)dT)) u(t)
(rmag (e-t e'1 :~'l 1:=0 ) ) u(t)
1
(rmag (e-t e•C~·~:- 1)) u(t)
(Imag ( e''1+~-· )) u(t) (Imag (0.5e11 - J0.5e1 0.5(1 - 1)e- u(t) 1
-
1
))
(b) For a parallel connection, hp(t) = h 1 (t) + h2 (t). MATLAB is used to plot hp(t).
>> hl = inline('(1-t). * ((t>=O)-(t>=1))');
>> h2 = inline( ' t.*((t>=- 2) - (t>=2))');
>> t = linspace(-2.5,2.5,501); hp= h1(t)+h2(t);
67
~1 So 2 - l .$ _, -0 s ... " 2.$
>> plot(t,hp,'k');
» axis((-2.S 2.5 -2.5 2.5)); xlabel('t'); ylabel('h._p(t)');
' .
••
~· 0
_,
...
•2
M
"
Figure S2.4-24b: P lot of hp(t).
68
Combining all pieces yields
.....
••
•..
..,
!'·
-0.1
....
-o.•
.....,
Figure S2.4-24c: Plot. of h8 (t).
2.4-25. (a) Using KVL, :c(l) = RCy(t) + y(t) or y(t) + n1cy(t) = R'cx(t). The characteristic
. \ -1
roo t IS;'\ = RC'
The zero-input response has form y0 (t) = c 1e-tf(RC). Using !.he IC, Yo(O) = 2 =
c1 . T hus, y 0 (t) = 2e-tf(RC)_
The zero-state response is x(l) * h(t), where h(t) = b0 o(t) + !P(D)yo(t)Ju(t).
For this first-order system, Yo(t) = c1e- t/(RC) and !io(O) = 1 = C1. Using
y0 (t) = e-t/(RC ), bo = 0, and P(D) = R1c, the impulse response is h(t) =
i:fce - tf(RC)u(t). Thus, the zero-state response is (J; n1c e-rf(RC)dT) u(t) =
( -e-r/(RC)l:= o) u(t) = (1 - e-t/(RCl) u(t).
For t ;::::: 0, the total response is the sum of the zero-input response and the zero
state response,
y(l) = (1
+ e-t/(RC) ) u(t).
(b) From 2.4-25a, we know the zero-input response is y 0 (t) = y 0 (0)e-tf(RC). Since
the system is time-invariant, the unit step response from 2.4-25a is shifted by
one to provide the response to x (t) = u(t - l) . Thus, !.he zero-state response is
(1 - e - (t-l)/(RC)) u(t - 1). Summing the two parts together and evaluat ing at
69
t = 2 yields y(2) = 1/2 = y0 (0)e- 2 f<RC) + (1- e - I/(RC)). Solving for y 0 (0) yields
2.4-26. Notice, x(2t) is a compressed version of x(t). The convolution y(t) = x(t) * x(2t) has
several distinct regions.
For t < 0 and t ~ 3/2, y(t) = 0.
For 0 :S t < 1/2, y(t) = J~ 2r(t - r)dr = tr 2 - 2r3 /3 l:=o = t 3 /3.
For 1/2 :St< 1, y(t) = J; 12
2r(t - r)dr = tr 2 - 2r3 /31~:,20 = t/4 - 1/12.
For 1 :S t < 3/2, y(t) = J/!~ 2r(t - r)dr = tr 2 - 2r3 /3l~~t-l = t/4 - 1/12 - (t 3 -
2t 2 + t - 2t3 /3 + 2t2 - 2t + 2/3) = - l 3 /3 + 5t/4 - 3/4.
Thus,
t3/3 0 :St< 1/2
t/4 - 1/12 1/2 :St< 1
y(t) = -t 3 /3 + ~t/4 - 3/4 1 :St< 3/2
{
otherwise
>> t = linspace(-1/2,2,251);
>> y = (t.-3/3).*((t>=O)&(t<l/2));
>> y = y+(t/4-1/12).*((t>=1/2)&(t<1));
>> y = y+(- t.-3/3+5*t/4-3/4).*((t>=1)&(t<3/2));
>> plot(t,y,'k'); xl abel('t'); ylabel('y(t)');
0 .2
011
0.1&
0.14
..,,
~ O.l
0.011
0.00
0.04
0.02
0
-o.s 0.5 ,,.
Figure 82.4-26: Plot of y(t) = x(t) * x(2t).
2.4-27. Notice, VR(t) = VL, (t) = V£2(t) = v(t).
(a) KCL at the top node gives x(t) = y(t) +iL, (t) +in. Since v(t) = L 2y(t), we know
in(t) = v(t)/ R = ~y(t). Thus, x(t) = y(t) + ~y(t) + iL, (t). Differentiating
t his expression yields ±{t) = y(t) + I;{y(t) +i~\l(t). However, i~!(t) = v(t)/L1 =
70
i;!i(t). Thus, :t(t) = y(t) + 1ftii(t) + i;v(t) or
2.4-28. Since the system step response is s(t) = e-tu(t) -e- 2 tu(t), the system impulse response
is h(t) = fts(t) = -e-tu(t) + o(t) + 2e- 2tu(t) - o(t) = (2e- 21 - e- ')u(t). The input
x(t) = o(t - 7r) - cos(v'3)u(t) is just a sum of a shifted delta function and a scaled
step function. Since the system is LTI, the output is quickly computed using just h(t)
and s(t). That is,
For 0 $ t
t 3/2
< 1/2, y(t) =Io rdr + It+i rdr = T
·1t
T=O + "2
213/2
.,.=Hl
2 2
= t /2 + 9/8- (t /2 +
t+l/2) = -t+5/8.
For 1/2 $ l < 1, y(t) = I~ rdr = t 2 /2.
Combining,
-t +5/8 0 $ t < 1/2
t 2 /2 1/2 $ t < 1
y(t)= t- 1/ 2 1 $ t < 3/2
{ -t 2 /2 + t + 5/8 3/2 $ t < 2
y(t + 2) Vt
MATLAB is used to plot y(t) over (-3 $ t $ 3). This interval includes three p eriods
of the (T = 2)-periodic funct ion y(t).
>> t = linspace(-3,3,601); tm = mod(t , 2);
>> y = (-tm+5/8).*((tm>=O)k(tm<1/2));
>> y = y+(tm.-2/2).*((tm>=1/2)&(tm<1));
>> y = y+(tm- 1/2) . *((tm>=1)&(tm<3/2));
>> y = y+(-tm.-2/2+tm+5/8). * ((tm>;3/2)k(tm<2));
>> plot(t,y , 'k '); xlabel('t'); ylabel('y(t)');
71
Figure 82.4-29: Plot of y(t) = x(t) * h(t) over (-3 ~ t ~ 3).
2.4-30. (a) Using KVL, x(t) = i(t)R+vc, (t)+y(t) = RC2y(t)+vc, (t)+y(t). Differentiating
yields :i;(t) = RC2ii(t) + vc, (t) + y(t) = RC2y(t) + ~ 1 i(t) + y(t) = RC2jj(t) +
§;Y(t) + y(t). Thus,
The zero-state response is x(l) * h(t), where h(t) = bo8(t) + [P(D)yo(t)Ju(t). For
1
this second-order system, iio(t) = c1 +c2e- 31! 2, iio(O) = 0 = c1 +c2 and iib >(o) =
1 = -3c2 /2. Thus, c2 = -2/3 and c1 = 2/3. Using b0 = 0, and P(D) = O.SD,
the impulse response is h(t) = 0.5D(Y0 (t))u(t) = 0.50(2/3 - 2/3e-3112)u(t) =
0.5(-2/3(-3/2)e- 3112 )u(t) = O.se- 3112 u(t). Using x(t) = 4te-3112 u{t), the zero-
state response is (J;{4re-Jr/ 2 )(0.5e-J(t-r)/2 )dr) u(t) = (2e- 3112 J; rdr)
u(t) =
(2e- 1 t /2) u(l).
31 2 2
Thus,
Since the input is driving a natural mode, resonance is expected; thus, the t 2
term seems sensible.
For (t ~ 0), the total response is the sum of the zero-input response and the
zero-state response.
72
2.4-31. Since h(t) is only provided for over (0 $ t < 0.5), it is not possible to determine with
certainty whether or not the system is causal or stable. However, when looking at
h(t) the waveform appears to have a DC offset. This apparent DC offset can be very
troubling if h(l) is truly an impulse response function. If a DC offset is present, the
system is neither causal nor stable. Imagine, a non-causal, unstable heart ! Something
is probably wrong.
One simple explanation is that a blood-filled heart always has some ventricular pres-
sure. Unless removed, this relaxed-state pressure would likely appear as a DC offset
to any measurements. It would likely be most appropriate to subtract this offset when
trying to measure the impulse response function.
Another problem is that t he impulse response function is most approp riate in the
study of linear, time-invariant systems. It is quite unlikely that the heart is either
linear or time-invariant. Even if the impulse response could be reliably measured at a
particular time, it might not provide much useful information.
2.4-32. (a) x(t) * x(-t) = J~00 x(T)x(- (t - T))dT = J~00 x(T)x(T - t)dT = r.,.,(t).
(b) Since r,,.,(t) is an even function, we only need to compute Txx(l) for either t ~ 0
or t $ 0. In either case, the autocorrelation function is computed by convolving
the original signal with its reflection.
Fort < -2, rzx(t) = 0.
,.,. ( ) = Jo
ror -2$l<- l,r,,.,t rt+2 TdT=0.5T 21t+2
T=O =t2/2+2t+2.
t+ J
For - 1 $ t < 0, rxx(t) = J0
J
T(r-t)dT-1 ft+l TdT+ f 1t+2 dT = ( 3TJ - t-z ,t+l
.,.2
r=O) +
2 1
r 1 + 7 1T-1
t::.2 = t
3 2
t3t t3t+l _ t 3 +~t 2 tt + l _ ~ + (t + 2 - 1) = -t3 /6-
2 T:c tt J 3 2 2 2
2
t /2 + t/2 + 4/3.
Combining and using rut= r:u: -t yields
t 2 /2 + 2t + 2 -2 ~ t<- 1
-t3 / 6 - t 2 /2 + t/2 + 4/3 -1 ~t<0
r:u:(t) = t 3/6 - t 2 /2 - t/2 + 4/3 O~t<I
{ t 2 /2 - 2t + 2 1$t <2
0 otherwise
2.4-33. (a) KCL at the negative terminal of the op-amp yields *1- 0
+ Cy(t) = 0. Thus,
y(t) = - R~x(l).
(b) The zero-state response is y(t) = x(t)*h(t), where h(t) = b0o(t)+[P(D)y0(t)]1L(l).
This is a first order system with >. = 0, thus vo( t) = c1 e..\t = c1 . Since :Vo (O) =
1 = c1 , bo = 0, and P(D) = - n1c• the impulse response is h(t) = - n1cu(t).
73
~~
·-·
••
·-·
01
-~-·
_, -1.$
~· •• .. ...
Figure S2.4-32b: Plot of Txx(t).
Thus,
1
y(t) = ( [' - - dT) u(t)
) RC
= __t u(t)
RC .
0
Notice, ly(t)I ramps toward infinity as time increases. Intuitively, this makes
sense; a DC input to an integrator should output an unbounded ramp function.
2.4-34. The system response to u(t) is g(t) and t.he response t.o step u(t - T) is g(t - T). The
input x(t) is made up of step components. The step component at T has a height 6.f
which can be expressed as
The step component at n6.T has a height .i(n6.T)6.T and it can be expressed as
[x(n6T)6Tju(t - n6T). Its response 6y(t) is
1:
.6r - O
n=-oo
74
Where H(jw) is H(s)la=iw in Eq. (2.48). Hence
2.4-36. An element of length 6r at point n6r has a charge (Figure 82.4-36). A point xis at
a distance x - n6r from this charge. The electric field at point x due to the charge
Q(n6r)6r is
6.E = Q(n6r)6r
47re(x - n6r)2
The total field due to the charge along the entire length is
= j "° Q(r)
-oo 47re(x - r)2
dr
1
= Q(x) * -41fEX
-
Figure 82.4-36
2.4-37. (a) KCL at the negative terminal of the op-amp yields %(Rt)-o
on
+ y(Rt) - o
/
+ ic(t) = 0.
Also, ic(t) = Cy(t). Thus, ~R(
·t +
on
yR(t)
I
+ Cy(t) = 0 or
1 - 1
y(t) + C R/(t) = CRi" x(t) .
(b) The zero-state response is y(t) = x(l)*h(t), where h(t) = boo(l)+(P(D)y0 (t)Ju(t).
This is a first order system with >. = c~,, thus iio(t) = c1 eAt. Since
Yo(O) = 1 = c1, bo = 0, and P(D) = cR.'.,., the impulse response
is h(t) = -=-Le-tf(CR1>u(t)
en... . Thus, y(t) = -=-Le-... f(CR1)dr) u(t) =
Jo CR, ..
(rt
(
.!!t..e-T/(CR1) 1t
Rin .,.=O
) u(t) = .!!t..(e-tf(CR,)
Rin
- l )u(t).
75
(c) For this system, the characteristic root is only affected by C and R,. Using 103
resistors, the resistor R1 is generally expected to lie in the range (0.9R1 l.IR1). 1
(a) KCL at the negative terminal of the first op-amp yields =Jr?-- + Civ(t) = 0 or
Ri c-;x(l) = -v(t). KCL at the negative terminal of the second op-amp yields
1
1
y(t) + R:C2 y(l) = R1R2 C1 C2 x(t).
The characteristic equation is >. 2 + R 31c 2 >. = 0 and the characteristic roots are
1
>-1 = 0 and >-2 = - R C .
3 2
Since one root lies on thew-axis, the circuit is not BIBO stable. In particular, a
DC input results in an unbounded output.
(b) The zero-input response has form y 0 (t) = c1 +c2e- 2'. Each op-amp has an initial
output of one volt. Thus, y0 (0) = 1 = c1 + ez. KCL at the negative terminal of
the second op-amp yields _i2 + i,
+ C2!io(O) = 0 or ilo(O) = - n 21c 2 - n 31c 2 =
- 1- 2 = - 3. Thus, !io(O) = -3 = -2c2. Thus, c2 = 3/2and c1 = 1-3/2 = -1/2
and
Yo(t) = -1/2 + 3/2e- 2 '.
(c) The zero-state response is y(t) = x(t)*h(t), where h(t) = boo(t)+[P(Dhio(t)Ju.(t).
This is a second order system with >. 1 = 0 and >.2 = -2, so iio(l) = c1 + c2e- 2t.
Solving Yo(O) = 0 = c1 + c2 and y~ )(t) = 1 = - 2c2 yields c2 = - 1/2 and
1
76
expected to lie in the range (.25C2, l.25C2). Since .>.2 = - R 31c , the characteristic
root is expected to lie in the range (.A 2 /[(0.9)(0.75)J, .A2/[(l.1~(1.25)J) . Thus,
2.4-39.
2.4-40. (a) Yes, the system is causal since h(t) = 0 for (t < 0).
(b) To compute the zero-state response y 1 (t), the convolution of two rectangular
pulses is required: a pulse of amplitude J and width two and a pulse of amplitude
one and a width of one. The convolution involves several regions.
Fort < 0, Y1(t) = 0.
For 0 $ t < 1, Y1(t) = J; Jdt = Jt.
For 1 $ t < 2, y, (t) = J/- 1 Jdl = J(t - (t - 1)) = J.
For 2 $ t < 3, Y1(t) = ft2_ 1 Jdt =J(2- (t - 1)) = J(3-t).
Fort~ 0, Y1(t) = 0.
Thus,
0$t<l
1$t<2
2$t<3
otherwise
(c) To compute y2(t), first note that x2(t) = 2x 1 ( t - 1) - xi (t - 2). Using the system
properties of linearity and time-invariance, the output y 2 (t) is given by
2.5-l.
and 2 1 1
y(t) = --e-Jt + -e-4 t + _ t~0
3 2 6
(b) x(t) = e- 'u(t), Y<1>(t) = H(-1) = ~~=g =i
y(t) K1e-Jt + K2e-4t + ~e-t
. (t) = - 3K1e-3t - 4K2e-4t 1 -t
5e
Y -
77
Setting t = 0, and substituting initial conditions yields
and 1 2 1
y(t) = -e-Jt _ -e-4t + -e- t t~O
2 3 6
(c) x(t) = e- 2tu(t), yq,(t) = H (-2) = 0
y(t) = K1e-31 + l<2e- 4t
y(t) = -3K1 e- 31 - 4K2 e- 4 t
-3 3
y(t) Ke t cos(4t + 8) + 25
y(t) -3Ke- 3 t cos(4t + 0) - 4J<e- 3 t cos(4t + 0)
and
+ ~t)e- 2' - ~0
17
y(t) = ( 2e- 3 ' l
4 2
78
(b) x(t) = e-tu(t), Y¢(t) = H(-l)e- t = 0
y(t) (K1 + K 2 t)e- 2 t
y(t) = -2(K1 + K 2 t)e- 2 t + K 2 e- 2 t
and
y(t) = (-94 + -19
2
t)e-
2t
t~O
2.5-4. Because (>. 2 + 2>.) = >.(>. + 2), the characteristic roots are 0 and -2.
y,..(t) = Ki + K2e- 21
In this case :c(t) = u(t). The input itself is a characteristic mode. Therefore
y.p(t) = f3t
Dut y.p(t) satisfied the system equation
y(t)
and
y(t) = ~ - ~e- 2 t + ~t t ~0
2.5-5. The natural response Yn(t) is found in Prob. 2.5-1:
y9(t) = {Jte-3t
79
or jj,p(t) + 7y,p(t) + 12y<t>(t) = x(t) + 2x(t)
Substituting x(t) = e-Jt and y.p(t) = f3te- 31 in this equation yields
Therefore
y(t) K1e- 3 t + K2e- 4 t - te- 31
y(t) = -3K1 e- 31 - 4K2e- 41 + 3te- 3 t - e- 3 t
and
y(t) 2e-Jt - 2e- 41 - te- 31 t~0
(2 - t)e- 3 t - 2e- 41 t~0
80
(d) The integral of o(t) is u(t). The system response to o(t) is u(t). Clearly, the
system is an ideal integrator.
2.6-4. Assume that a system exists that violates Eq. (2.64) and yet produces a bounded
output for every bounded input. The response at t = t 1 is
1 if h(r) > 0
x(l1 - r) = {
- 1 if h(r) < 0
In this case
and
t 2 /2 + 2t + 2 -2 ~ t<0
O~t<2
y(t) =
{
-t 2 /~ + 2t 2~t< 4
otherwise
8l
,.,.---..--- ..---..---..---..--- ..---..---
••
yields
h(t) = (c(t) + o.5c(t - 1) + o.2sc(t - 2) + o. 12sc(t - 3) + .. ·)
2.7-1. (a) The time-constant {rise-time) of the system is Th = 10-s. The rate of pulse
communication < -A = 105 pulses/sec. T he channel cannot transmit million
pulses/second.
(b) The band width of t he channel is
B = _!__ = 105 Hz
Th
The channel can transmit audio signal of bandwidth 15 kHz read ily.
2.7-2.
1
Ti.= -
B
= -1014 = 10 - 4 = 0.l ms
The received pulse width = (0.5 + 0.1) = 0.6 ms. Each pulse takes up 0.6 ms interval.
The maximum pulse rate (to avoid interference between successive pulses) is
1
_ x _ :::::: 1667 pulses/sec
0 6 10 3
Tr = T,. = -;;1 = 10 - 4
82
2.7-4. (a) For a causal system with finite duration h(t}, the rise time is exactly equal to the
time when the signal is last non-zero. That is,
T,. = 4 seconds.
(b) The impulse response function h(t) is consistent with a channel that has the
following three characteristics: 1) a channel with delay from input to output {for
example, signal propagation delay), 2) a channel with low-pass character (pulse
dispersion that results in a o(t) input spreading into a square pulse), and 3) a
channel with two signal paths {for example, a primary signal path and an echo
path).
For systems with predominantly low-pass character, digital information can be
transmitted without significant interference at a rate of Fe= .J.. = 1/4. However,
this estimate is too conservative for the present system. Notice that h(t} = 0 for
0 :5 t < 1, corresponding to a transmission delay in the primary signal path.
The remaining portion of h(t) has a width of three, so it is therefore practical
to transmit at rates of Fe = 1/3. By clever interleaving of data, it is possible
to transmit at rates of Fe = 1/2. Consider transmitting the binary sequence
{bo, b1 , b2, b3, ... } using a (t = 1)-spaced delta train weighted by the pulse se-
quence {bo , bi, 0, 0, b2 , b3 , 0, 0, ... }. The output is the series of non-overlapping
unit-duration pulses given by {b0 , bi, b0 , b1 , b2 , b3 , b2 , b3 , .• •} . The effective trans-
mission rate is 0.5 bits per unit time.
(c) The resulting convolution y(t) = x(t) * h(t) has many regions.
Fort< 1, y(l) = 0.
For 1 :5 t < 2, y(t) = f 1t(-l) dt = 1- t.
For 2 $ t < 3, y(l) = f 12 {-l)dt = - 1.
For 3 $ t < 4, y(t) = f1~ 2 {-l)dt + J;( -l )dt = (t- 2) - 2 + 3- t = -1.
t < 5, y(t) = f 3 {-l)dt = -1.
4
For 4 $
For 5 $ t < 6, y(t) = J:1_2 {- l)dt = (t - 2} - 4 = t - 6.
For 6 $ t, y(t) = 0.
Thus,
1- t 1 $t<2
- 1 2$ t< 5
y(t) ={ t~6 5$ t < 6
otherwise
MATLAB is used to plot the result.
» t (0:.01:10); y = zeros (size(t));
>> y y + (1 - t).*((t>=1)&(t<2));
>> y y + (-1). * ((t>=2)&(t<S));
>> y y + (t-6). *((t>=5)& (t<6) );
>> plot(t,y,'k'); axis([O 10 -1 .2 .2));
>> xlabel( ' t' ); ylabel('y(t)');
83
••
Figure S2.7-4c: Plot of y(t) = x(t) * h(t).
84
Chapter 3 Solutions
1 ~ 1 . 19
P..r: = No+ 1 L lx[n]l2 = 6 [2(1}2 + 2(2}2 + 32] = 6
n= - 3
(b}
+ 2(3)2 ) = ~
1
Px = [2(1) 2 + 2(2) 2
12
(c)
3.1-3.
and
~. ~· 1~;· v,e''"-"i'
No - 1 No - 1 No- 1
~o L L Vre'rflon L v;,,e-jmOon
n = O r=O m=O
N 0 - I No - 1 [ No- I ]
Px - ~o ~ ~O VrV;,, ~ ei (r-m)Oon
85
The summation within square brackets is N 0 when r = m and 0 othenvise. Hence
No - 1 N0 -l
P" = L v,.v; = ,L l'Dr l2
r=O r= O
3.1-4. (a)
z.(nl
Ex = f)o.8) 2" =
n=O
f
n=O
0.64n = 1
1 - 0.64
= 2.78
(b) To find the energy of the even component Xe[n], we observe that both terms
0.8nu(nj and 0.8-"u(-nj are nonzero at n = 0. Hence, the two terms are not
disjoint, and the energy E"• is not the sum of the energies of the two terms. For
this reason, we rearrange xe(n] as
1
Xe[n] = o[n] + 2 {0.8nu(n - lj + 0.8- nu[n -1]}
All the above three terms are disjoint. Hence E"'• is the sum of energies of the
three terms. Thus
1 00 1 -oo
1 + -4"'0
L .64" + -4 "'
L 0.64- n
n= I n =- 1
= 1+ 2l ~ n 1 [ 0.64 ]
L 0.64 = 1 + 2 1 - 0.64 = 1.89
n=l
Hence
Ex,+ E;i; 0 = 1.89 + 0.89 = 2.78 =Ex
(c) Let us first consider a causal signal x(nJ, which can be expressed as
l 1
x[nJ = x[O]o[n] + 2 {x[n]u[n - l ] + x[-n]u(-n - 1]} + {x[n]u[n - l ] - x [-n[u [- n - I)}
2
1 oo I - oo 1 00
2
Ex.= x [0J + 4 L jx[n]l 2 + 4 L jx(-n] j2 = x 2 [0] 2 L 2
lx(n] l
n=l n = -1 n=I
86
Similarly
l 00
E:r =0 2 L lx[nJl2
n=l
Hence
00
Ex.+ Ex = x 2 (0) +
0 L lx[n) l2 =Ex
n= l
Similarly, we can show that this result applies to anticausal signals also.
A general signal is made up of causal and anticausal components, which are
disjoint. Hence the energy of a signal is the sum of energies of the causal and
anticausal components. Moreover the energy of each causal and anticausal com-
ponent is equal to the sum of the respective even and odd components. Also the
sum of the even components of the causal and anticausal signal equals the even
component of x[nJ . Same is true of odd components. Hence, it follows that the
energy of x[nJ is the sum of energies of even and odd components of x[nJ.
3.1-5. (a) As seen in Prob 3.1-4, the even and odd component of a causal x[nJ are
1 1
x[n) = x{0)6[n) + {x[n)u[n - 1) + x[-n)u[- n - 11} + {x[n]u[n - 1) - x[-nJu[-n - 1]}
2 2
:r,{n)
If x[O] = 0, then
l 00 l -oo l 00
(b) Observe that x[OJ6[n) is disjoint with x0 [nJ. Hence, the product x[0)6[n)x0 [nJ is
zero. Moreover, u[n-1) and u{-n - 1) is disjoint. Hence the cross-product terms
involving product u[n - l]u[- n - lJ are zero. Hence
()() 1 00
The two terms within the curly brackets are equal. Hence
00
L Xe[n]x~[n] =0
n= -oo
Using similar argument, we can show that this result holds for a nticausal signal.
A general signal is made up of causal and anticausal components, which are
disjoint. Hence it follows that
00
I: Xe[n]x~[nJ =0
n ::-oo
87
3.2-1. (a) The energy of xl-nJ is E 1 given by
n=-oo
E1 = L lx[mJl 2 = L lx[mJl2 = E .,
1n=oo m=-oo
3.2-2. Because I - x[nJl 2 = lx[nJl 2 , it follows that the power of -x[n] is P.,.
For parts b, c, d and e, using the arguments in Prob. 3.2-1, we show
Power of x [-n] = P.,
Power of x[n - mj = P:r
Power of cx[n] c2 Px=
Power of x[m - nj = P,,
This shows the time-shift or time-inversion operation does not affect the power of a
signal. Same is the case with sign change. Multiplication of a signal by a constant c
causes c2 -fold increase in power.
0.5
88
:x. r..- KJ • (' x. [ R.+'7 J
..,-rJl-l1
3
3
. . -
-·r.••-.i - ·•-.- •
{
12. ......
-3
'
- J.. - 1
1 :t_[-KJ
,._ 3 -s
-:£: [lf-t'-]
-3
(\ .. 0
't Cf :i:. t R· f,]
:;/b
.,,
/
- Cl
' (\l.;,.
• -q - 'l
0 3
~, "~ --
..:."t :;L' [_3 (\ 1. ~ [~] 9
- 1
(\, .. --
- 3
" ~ ~ /\ . ~
.. 'l
~-
89
(.b)
••
7
3 (_e.. )
(\ .--
Figure 83.3-3
-0·5
0
(J• ' " " 5 0 0·5 f•(.J.f~7 -, 0 0
Figure 83.3-5
(a)
1 ]
u!nJ = 2{u[nj + u[- nj} + 2 {u(nj - u[- nl}
90
(b)
... . ,r
I .., :··
.x~Lnj = D
Figure S3.3-7
Figure 83.4-1
3.4-2. The net growth rate of the native population is 3.3 -1.3 = 23 per year. Assuming the
immigrants enter at a uniform rate throughout. the year, their birth and death rate
will be (3.3/2)3 and (1.3/2)3, respectively of the immigrants at the end of the year.
The population p[nJ at the beginning of the kth year is p[n - 1] plus the net increase
in the native population plus i(n - l ], the immigrants entering during (n - l)st year
91
plus the net increase in the immigrant population for the year (n - 1).
[I
pn = pn- I 1I +in
I 1I + 3.3 - 1.3 pn- I -1I + 3.3x- 1.3 in
I - lI
100 2 100
1.02p[n - l J + l.Oli[n - 1]
or
p[nJ - 1.02p[n - lj = 1.0li{n - II
or
p[n + lJ - l.02plnJ = l.Oli[n]
3.4-3. (a)
X~:jl~1~£~ ~ J ~n ~
...D ref -ce !;~ fs v "\ t- d~ }<='\ ;t
Figure S3.4-3b
3.4-4. The input x[n] = u[nJ, which has a constant value of unity for all n ~ 0. Also
y[n] - y[n - 1] = Tu[nJ . Hence the difference between two successive output values
is always constant of value T. Clearly y(n] must be a ramp with a possible constant
component. Thus
y[n] = (nT + c)u[n)
To find the value of unknown constant c, we let n = 0 and obtain
y[O) = c
But from the input equation y[nJ - y[n - 1] = Tti[n], we find y[OJ =T !remember that
y[-lJ =OJ. Hence
d2 y dy
dl2 + a 1 dt + aoy(t) = x(t)
We use the notation y[nJ to represent y(nT), x[nJ t.o represent x(nT), · · · etc. and
assume that T is small enough so that the assumption T --. 0 may be made. We have
y(t) = y[nJ
dy y[nJ - y[n - lj
dl T
11lnl-y{n-l) _ yln-l) -y)n-21
<12y T T
dt 2 ~ T
92
y[n] - 2y[n - ll + y[n - 2]
r2
Hence, the given differential equation can be approximated by the following differential
equation
A1y(n] + A2y[n -1] + A3y(n - 2] = Bx[nJ
Substituting approximate difference expressions in the differential equation yields
y[n] - 2y[n - l] + y[n - 2J + a1T {y[n] - y[n - 1]} + a0T 2y[n] = T 2 x[n]
or
that is
v(n + 2] - ( 2 + ~) v[n + l] + v(nl = 0
3.4-7. (a) TI-ue; all finite power signals have infinite energy, and therefore cannot be energy
signals. Energy signals and power signals are mutually exclusive.
(b) False; a signal with infinite energy need not be a power signal. For example, the
signal x[n] = 2"u[nj has infinite energy and infinite power. Thus, it is neither an
energy signal nor a power signal.
(c) TI-ue; the system is causal. Even though the input is scaled by (n 1-1), the current
output only depends on the current input. Another way to see this is to rewrite
the expression as y(n] = nx(n] + x(n).
(d) False; the system is not causal. The current output depends on a future input
value. To help see this, substitute n' = n-1 to yield y[n'I = x[n' + 1); the output
at time n' requires the future input value at time n' + l.
(e) False; a signal x [n) with energy E does not guarantee that signal x [an] has
energy 1; 1. Although this statement is true for continuous-time signals, it is not
true for discrete-time signals. Remember, the discrete operat ion x[an] results
in a loss of information and thus a likely loss in energy. For example, consider
x(n] = o[n - l j, which has energy E = l. The signal y(nJ = x(2n] = 0 has zero
energy, not E/2 = 1/2.
3.4-8. Notice, yi[n) = -o[n) +o(n- 1] +2o[n-2J. Furthermore, x 2 [nl = xi(n-1] - 2xi[n - 2).
Since the system is LTl,
93
>> y1 = inline('-(n==O)+(n==1)+2*(n==2)'); n = (-2:8);
>> stem(n , yl(n-1)-2*y1(n-2),'k') ; axis((-2 8 -4.5 4.5));
>> xlabel('n'); ylabel('y_2[n]');
'
)
1 l
2
'
'
-2
3.4-9. Using the sifting property, this system operation is rewritten as y[n]
0.5 (x(n] + x[-n]).
(a) This system extracts the even portion of the input.
(b) Yes, the system is BIBO stable. If the input is bounded, then the output is neces-
sarily bounded. That is, if j:i;[nJI '.::: M:c < oo, t hen ly[n]I = 10.5 (x[n) + x[-nl) I '.: :
0.5 (lx[n]I + lx[-n]I) ~ Mx < oo.
(c) Yes, the system is linear. Let yi(n] = 0.5(xi[n) + xi[-n]) and Y2[nJ =
0.5(x2 [n] + x 2 [-n]). Applying axi(n] + bx 2 [n] to the system yields y[n] =
0.5 (ax 1 [nJ + bx2[n] + (axi[-nJ + bx2[-n])) = 0.5a(x1 [n] +xi [-nJ) +0.5b(x2(n] +
x2[-n]) = ay1 [n] + by2[nj.
(d) No, the system is not memoryless. For example, at t ime n = 1 the output
y[l] = 0.5(x[l] + x{-1]) depends on a past value of the input, x[-1) .
(e) No, the system is not causal. For example, at time n = -1 the output y[-1] =
0.5(x[-l] + x[l)) depends on a future value of the input, x[l].
(f) No, the system is not time-invariant. For example, let the input be xi[n) =
u[n + 1OJ - u.[n - llj. Since this input is already even, the output is just the input,
y 1 [n] = xi[n). Shifting by a non-zero integer N, the signal x 2 [n] = x 1 [n - NJ is
not even, and t.he output is y 2 [n] # yi(n - NJ = xi[n - NJ. Thus, the system
cannot be time-invariant.
3.4-10. It is convenient to substitute n' = n + 1 and rewrite the system expression as y[n'J =
x[n' - 11/x[n'J.
(a) No, the system is not BIBO stable. Input values of zero can result in unbounded
outputs. For example, at n' = 0 the bounded input x[n'J = o(n'] yields an
unbounded output y[l) = 1/0 = oo.
(b) No, the system is not memoryless. The current output relies on a past input. For
example, at n' = 0, the output y[n'J requires both the current input x{n'J and a
stored past input x [n' - l j.
94
(c) Yes, the system is causal. The current output y{n'J does not depend on any future
value of the input.
3.4-11. The operation y(l) = x(2t) is a one-to-one mapping, where no information is lost. Any
one-to-one mapping is invertible. In this case, x(t) is recovered by taking y(t/2).
Since every other sample of x[n] is removed in the operation y[n) = x [2n], one half of
x[n] is lost and the process is not. invertible. Thought of another way, the operation
y[n] = x[2n] is not a one-to-one mapping; many different signals x[n] map to t he same
signal y[n], which makes inversion impossible.
3.4-12. Notice that yi[n] is obtained by multiplying x [n) by the repeating sequence
{sin( ~n + l)} = {.. ., 0.5403, -0.8415, -0.5403, 0.8415, ... }. In this particular case,
x[n] is recovered by simply multiplying y[n] by the inverse repeating sequence
{ sin(tn+l)} = { .. ., 1.8508, - 1.1884,-1.8508, 1.1884, ... }.
In the case second, however, y2 [n] is obtained by multiplying x [n) by the repeating
sequence {sin(Hn+l))} = { ... , 0,-1,0,1,. .. }. Since the sequence includes zeros,
information is lost and the original sequence cannot be recovered.
3.4-13. Using the definition of the ramp function, the system expression is rewritten as y[n] =
nx{n)u[n].
(a) No, the system is not BIBO stable. For example, if the input is a unit step x{n] =
u[n] , then the output is a ramp function y[n) = r[n], which grows unbounded with
time.
(b) Yes, the system is linear. Let yi[n) = nx1 [n]u[n] and y 2{n) = nx 2[n]u[n] . Ap-
plying axdn] + bx2 [n] to the system yields y[n] = n(ax1 [n) +bx2[n])u{n] =
anx1[n ]u[n) + bnx2[n]u[n] = ayi[n) + by2[n].
(c) Yes, the system is memoryless. The current output only depends on the current
input. multiplied by a known (time-varying) scale factor.
(d) Yes, the system is causal. All memoryless systems are causal. The ou tput does
not depend on future values of the input or output.
(e) No, the system is not time-invariant. For example, applying X1 [n] = o[n] yields
the output yi(n] = no[n]u[n] = 0. Applying x2[n) = o[n - l ] yields the output
y2[n] = no[n - I ]u[n] = o[n - 1]. Note, x2[n) = x1[n - l] but Y2[nj :/= yijn - l].
Shifting the input does not produce a corresponding shift in the output.
3.4-14. (a) Position measurements x[n] are in met ers. The difference x[n] - x [n - l] is the
change in position, in meters, per frame of film. Since the camera operates at 60
fram es per second, dimensional analysis requires k = 60
d .
secon s
{b) Since v[n] = k(x[n) - x[n - 1]) is an estimate of the velocity v(t) = ftx(t), ii
is sensible to use a[n] = k(v[n] - v(n - lj) as an estimate of the acceleration
a(t) = -fiv(t). Combining estimates yields a[n] = k(k(x[n] - x[n - 1]) - k(x[n -
l] - x(n - 2])) or
95
the estimate a[n) depends only on current, and past values. A more balanced esti-
mate of a{t) is a shifted version a 2 [nJ = a[n+l] = k 2 (x[n + 1] - 2x[n] + x[n - l]).
While this may fix the problem of lag, the new system is no longer causal. Both
cases estimate derivatives using first-order differences; there are more sophisti-
cated (and more complex) methods to more accurately estimate derivatives.
By substituting o[n] for x[n], the impulse response is
3.4-15. There are many ways to solve this problem. For this solution, let d[nJ designate the
distance from the student's destination, which alternates between home and the exam
location, just before the student changes his mind. For even-valued n the destination
is home, and for odd-valued n the d estination is the exam location .
.Just before changing d irection, the student is a distance of d[nJ miles from his destina-
tion. Turning around, his next destination is t.herefore a distance of 2-d[nJ miles. The
2
student I.ravels one-half of this distance, which leaves -;lnl miles remaining. Thus,
a difference equation description of this problem is d[n + 1] = -;lnJ = 1 - 0.5d[n) =
2
(b)
y[n + lJ = - 2y[nJ + x[n + 1] {2)
Setting n = -1 , and substituting y[- IJ = 0, x[OJ = 1, yields
y[O] = 0+1=1
Setting n = 0, and substituting y(Oj = 1, x[l] = ~.yields
1 1 2 1
y(2] =-2(-2 +~} + e 2 = 4 - ~+ e 2 =3.399
96
3.5-2.
3.5-4.
y[n + 2) = -3y[n + 1) - 2y[n] + x[n + 2] + 3x[n + l] + 3x[nJ
97
Setting n = 0, and substituting y[- 1] = 2, y!- 2] = 3, x!OJ = 1, x[- 1] = 0, yields
y!O] = -2(2) - 3 + 2(1) - O = -5
1 1
yl2J = - 2(7.667) - (-5) + 2(9) - 3 = - 5.445
3.6-1.
(E2 + 3E + 2)ylnJ = 0
(E2 + 2E + l)y[nj = 0
The characteristic equation is~?+ 2'Y + 1 = ('Y + 1) 2 = 0.
(E2 - 2E + 2)ylnJ = 0
The characteristic equation is 'Y2 - 2'Y + 2 = ('Y - 1 - jl )('Y - 1 + jl) = 0. The roots
are 1 ± jl = ..f2e±i1r/4 .
y[nJ = c( v;;:.2)" cos( 7r4n + 8)
Setting n = - 1 and - 2 and substituting initial conditions yields
98
c 1f c .
0 = -
2
cos( - -
2
+ 0) = -2 sm 0
Solution of these two simultaneous equations yields
ccosO = 2 } c= 2
csin0=0 ~ 0=0
Hence, all the N characteristic roots are zero. Therefore, the zero-input component is
zero and the total response in given by the zero-state component.
3.6-5. (a) By definition, any element in the Fibonacci sequence is the sum of the previous
two. Thus, /[nJ = J[n - l j + J [n - 2J. Written in standard form, this yields
J[nJ - f [n - 1J - Jin - 2J = 0.
This is a somewhat unusual system in the fact that it has no input. In the lingo
of signa ls and systems, J[n] is a zero-input response that is completely driven by
the auxiliary conditions.
{b) The characteristic equation is 1 2 - / - 1 = 0. This yields two characteristic roots
11 - -
l+v'5
-
2
::::; 1.618 and 72 - -
1-v'5
-
2
::::; -0.618.
Since one characteristic root is in the right-half plane, the system is not stable.
(c) To determine a particular Fibonacci number, it is convenient to determine a
closed form expression for / [nJ. Since /[nJ is a zero-input response, it has form
J[n] = c111 + C2"f2 . The auxiliary equations yield /[l] = 0 = cn1 + c212 and
/[2] = 1 = c1 1[c2 1i. Solving yields c 1 = 1'l'1'2 (1' 2
1'2 - ")'1
JB-1 ~ 0.2764 and
) = v'S
2 5
c2 = 11
")'1 ")'2h2-")'il
= 275
~-! ~ 0.7236.
MATLAB is used to solve for the requested values of /[nj.
>> gammal • (l+sqrt(5))/2; gamma2 c (1-sqrt(S))/2;
>>cl = -gamma2/(gammal*gamma2•(ga.mma2-gammal));
>> c2 = gammal/(gamma1•gamma2•(gamma2-gammal) );
>> f50 = cl•gamma1 -so+c2•gamma2·so
f50 = 7.7787e+009
>> !1000 = cl•gamma1-1ooo+c2•gamma2· 1000
flOOO = 2.6864e+208
Thus,
99
and
1000 (
f!lOOOI = ,/5 - I l + ,/5 + ,/5 + l I - ,/5 ) 1000
: : : : 2.6864(10208 )
2,/5 ( 2 ) 2,/5 2
3.6-6 .
3.6-7. Since we are looking for the zero-input response, the term J3x[n - 81 is irrelevant
and the equation becomes y0 [nj + y 0 [n - 1] + 0.25yo[n - 21 = 0. The characteristic
equation for this second-order system is ')' 2 +1'+0.25 = 0. This yields a repeated root
at '/' = -0.5, and the zero-input response has form Yo[nl c 1(-0.5)" + c2 n(-0.5)n. =
The pair of equations y 0 ( - l] = 1 = - 2c1 + 2c2 and y 0 [lj = 1 = -ci/2 - c2 / 2 are
solved using MATLAB.
Thus,
Yo[n] = -1.25(0.5)" - 0.75n(0 .5)".
3.7- l. (a)
(E + 2)y(nJ = x[nl
100
Setting n = - 1, and substituting hl-1] = 6[-1) = 0, yields
h[OJ =0
Setting n = 0 in Eq. (1) and using h[O] = 0 yields
1 1
0= - +c ==> c= --
2 2
Therefore
h[n] = !6[n] - !(-2)"u[n]
2 2
(b) The characteristic root is -2, b1 = 0, a 1 = 2. Therefore
h [n + lJ + 2h[nJ = 6[n + lJ
Setting n = - 1, and substituting h [-1] = 0, 6[0] = 1, yields
h[O] =1
Setting n = 0 in Eq. (2) and using h[OJ = 0 yields
l=c
and
h[n] = (-2)nu[nJ
3.7-2. Characteristic equation is 1 2 - fry+ 9 = (I - 3) 2 = 0. Also a2 = 9, bz = 0. Therefore
(1)
We need two values of h[n] to determine c 1 and c2 . This is found from iterative solution
of
(E2 - 6E + 9)h[n] = E6[n]
or
h[n + 2] - 6h[n + lJ + 9h{n] = bin+ l] (2)
Also h[-1] = h(-2] = 6!-1] = 0 and 6[0J = 1. Setting n = -2 in (2) yields
0 = C1 }
1 = 3(c1 + c2) ==>
101
and
3.7-3.
We need two values of h[nJ to determine c and (}. This is done by solving iteratively
Setting n = 0 yields
h[OJ - 6(0) + 25(0) = 2(1) - 4(0) => h[O] =2
Setting n = 1 in (2) yields
= 2.061
ccos8 = 2
csin8 = -0.4931 }~ c
B = -0.244 rad
and
hlnJ = 2.061(5t cos{0.923n - 0.244)u[nJ
3.7-4. (a)
3.8-1.
00
= L (-2ru[m - l ]e-<n-m>u[n - m - 1]
Tn =- oo
102
However, u[m - l] = 0 for rn < 1 and u[n - m + 1] = 0 for rn > n + 1. Hence the
summation limits may be restricted for 1 :::; m :::; n + 1.
n+l 2
y[nl = e-n 2:)-2er e-n [ (-2e)"+ + 2e1
-2e- l
m=l
y[n+ 1]
103
8
and y[n] = [(-2t-(3}"]u[n- 1]
45
3.8-3. Here let us delay x[nJ by one unit and advance h[nJ by one unit. t.o obtain y[n).
3.8-1. Here let us delay the input x[n] by 3 units and advance h[n] by 4 units. The resulting
convolution yields y[n - 3 + 4J = y[n + 1J.
Hence
y[nJ -
2
3
°[3- n - 2"] u[n - lJ + 6 [3-(n-l) + (5n - 6}2(n- t)] u[n - l ]
25
{3-"u[n - l J - 2"u[n - l] } + ~n2nuln - 1]
446
-
75 5
3.8-5. Here, we advance x [n] by one unit and leave h[nJ unchanged. The resulting convolution
is y [n - l] . Hence
Hence
104
Hence
y[nJ = { ~(3rcos [7rn - 2.26] + l.273(2f} u[n -1]
3
3.8-6. #1
o[n - m - k] = 1 for m = n - k and is zero for all other values of m. Hence the
right-side sum is given by x[n - kJ.
#2
n
n /n+l - l
/nu[n) * u[n] = L /m =
/
_
1
u [n] I -:/= 1
m=O
We multiply the result with u[n] because the convolution is zero for n < 0.
#3
n
tt.(n) * u[nJ = L u[m)u[n - m]
m=O
3.8-7. #4
12
n [(/1/r2)n+l -
1- Cr1h2)
l]
1 ln+l _ 1 2n+l] u [n)
[ /1 =!= /2
/1 -12
105
#6
1"u[n] * nu{n]
m= O
n n
= n L 'Ym - L m')'m
m=O m=O
'Yn+I - 1 'Y + (n('Y- 1) - lhn+l
n
'Y - 1 b -1)2
[
'Yb" - 1) + n(l - 'Y) u{nj 1
(1 - 'Y)2
3.8-8. #7
nl:m - l: m 2
n(n+1) n(n+I)(2n+l)
2 6
n(n2 - I) Il
= 6 un
#8
n
1"u(n] * 1nu{nl = L /mln -m
m =O
m=O
= (n + I)'Y"ulnJ
3.8-9. #9
n
L m7f''Y2'-m
m=O
nt = O
/1 -:/= 1'2
#11 Let
106
00
L 'Yi"'Y2-mu[m]u[-(n - m + l)]
m=-c:o
Consider c[nJ for n;::: 0. In this case -(n - m +I);::: 0 form;::: n + 1. Therefore
u[- (n - m + l)J = 0form<n+1 and equal to 1 form;::: n + l. Also u[m] = 1
form;::: n + 1 (for positive n). Hence
c[n] = 'Y2 f
m=O
('Y21
'Y
)m = 'Y2_'Y_2_-+'YI1_
Therefore
'Yn+l 'Yn+I
c[nJ = _I_ u[nJ + - 2-u[-(n + l)]
'Y2 - / J 'Y2 - 'Yl
= c(-2rYo[n]
10 = -~ :::} c == -20
2
T herefore
h[n] = (-2)"u[nJ
The zero-state response is
== - . e [~(e)-" + 2(-2)"Ju[nJ
2e + 1 e
[ 2e ~ 1 (e)- n + 2e2: 1 (- 2)"] u(nJ
107
{-20(-2)" +- 1
+ ~(-2)"]u[n]
- . (e)-n
2e + 1 2e + 1
1
- - [- (38e + 20)(- 2)" + (e)-nJu[n]
2e+l
3.8-11. (a)
(b)
(c)
y[n] -- 4~(2"+
3
1- 2 - (0 .sr+ 1- 2 Ju[n - 2] -- ~(2n-l
3 - (0 .5)"- 1 ]u(n - 2J
3.8-13. (a) Figure S3.8-13a represents the system in Figure P3.8-13 in a more convenient
fashion. A parallel connection requires the individual impulse responses to be
added, and a series connection requires the individual impulse responses to be
convolved. Thus, the overall impulse response is given by
(b) First, we shall simplify expressions for hi[n] and h 2 [nJ by using the facts that
=
u(n -1] = u[n] - o[n] and (0.9)"- 1 0 19 (0.9)" and (0.5)"- 1 = 0\(0.5)". Now
108
y[nJ
Similarly
09
h2 [nJ = 0.5nu[n) - · (0.5)"(u[n] - o[n))
0.5
= -0.8(0.5)"ulnJ + l.8o[n]
Recall that x[nJ * o[n] = x [n] and o[n] * o[n] = olnJ. Hence
4
h1 (nJ * h2[n] = 9(0.9)"u[n] * (-0.8)(0.5)nulnJ
5 4
-0.8(9)(0.5)nu[n] + l.8(9)(0.9)ntl[n] + o[n]
16 [ O gn+I - 0
- 45 . 0.4 .
5n+l] u(n] - 4 4
9(0.5)"u[nJ + 5(0.9)nu[nl + o(n]
= olnl
Thus the cascade of the two systems is an identity system.
3.8-14. (a) From Eq. (3.70b), for a causal system
n
g[nJ = I: h[kJ
k =O
109
3.8-15. The equation describing this situation is !see Eq. {3.9b)J
(E - a)h[n) = Eolnl
or
h[n + l) - ahlnJ = o[n + 11
Setting n = - 1 and substituting h[-1) = 0, 6[0) = 1, yields
h(O] = 1
Therefore
h(nJ = (aru[nJ = (l.Ol)"ulnl
The (zero-state) response is
3.8-16. This problem is identical to t he savings account problem with negative initial deposit
(loan). If M is t he initial loan, then y[O] = - M . Jfy[nJ is the loan balance, then [see
Eq. (3.9b)J
y[n + lJ - ay[n] = x[n + l ] a = l+r
or
(E - a)ylnJ = Ex[n]
The characteristic root is a, and the impulse response for I.his system is found in Prob.
3.8-15 to be
h(n) = a"u[n)
This problem can be solved in two ways.
First method: \Ve may consider the loan of M dollars as an a negative input -Mo[n) .
The monthly payment of P starting at n = 1 also is an input. Thus the tot.al input is
xlnl = -Mo[nJ + Pu[n - 1) with zero initial conditions. Because u[nJ = o[n] +u[n-1),
we can express the input in a more convenient. form as x[nJ = - (M + P)o[n) + Pu[n].
110
The loan balance (response) y[n] is
Also a = 1 + r and a - 1 = r where r is Lhe interest rate per dollar per month. At
n = N , the loan balance is zero. Therefore
Yo(n] = ca"u[n]
Setting n = 0, and substituting y 0 [0) = - M, yields c = -M and
Yo[nJ = - Ma"u[n)
The zero-state response y[n] is
an
Pa"u[n] * u[n - lJ = x[n - 1] = P [ - - - u[n - l]
a-1
1]
The total balance is y 0 [n] + y[n]
ar•
Yo[n] +y[n] = -Manu[nJ + P [----· u(n-1]
Ct - 1
I]
111
For n > 1, u[nJ = u[n - lJ = 1. Therefore
an
Loan balance= -Man+ P [- -
a-1
l] n>l
which confirms the result obtained by the first method. From here on the procedure
is identical to that of the first method.
3.8-17. We use the result in Prob. 3.8-16. In this problem r = 0.015, a= 1.015, P = 500,
M = 10000. Therefore
500 = 0000 (l.015)N (0.015)
l (J.015)N - 1
or
(l.015)N 1.42857
N Jn(l.015) Jn( 1.42857)
N
ln(l.42857) = 23 .956
Jn(l.015)
Hence N = 23 payments are needed. The residual balance (remainder) at the 23rd
payment is
2
1
y[23J = -10000( 1.015) 23 + 500[ (l.Ol 5) ; - J = - 471.2
0.01
3.8-18. (a) The two strips corresponding to u[nJ and u[nl (after inversion) are shown in the
Figure S3.8-18a for no shift (n = 0) and for one shift (n = 1). We see t hat if
c[nJ - u[nJ * u[nJ then
n y(nJ
0 0 + 1 + 2 + 3 + 4 + 5 = 15 y[n] =0 k~6
1 1+2 + 3 + 4 + 5 = 15
2 2 + 3 + 4 + 5 = 14 y[n] = 15 n < 0
3 3+4+5 =12
4 4+5=9
5 5
6 0
112
.,, .
.
c.. tn J . . .. .. .,, f. .
,. t • ' .,.;,,,.·
-.f,~. .,
I , ,
~ 2.. 3 ., . rn- 1
Figure S3.8- 18
3.8-20. From Figure S3.8· 20, we observe the following values for y[n]:
n ylnJ n yin]
0 5x5+5 x5 =50 ±11 Ox0+5x4=20
±1 5x4+ 0=20 ±12 0 x O+ 5 x 3 = 15
±2 5 x 3 + 0 = 15 ±13 Ox0+5x2 = 10
±3 5 x2+0=10 ±14 Ox0 + 5xl=5
±4 5xl +0=5 ±15 OxO +Ox O=O
±5 5x0 + 0 = 0 ±16 0
±17 0
±9 OxO+OxO=O ±18 0
±10 Ox0+5xl=5
Observe that
y[n] = 0 5 :5 lnl $ 9 and lnl ;::: 15
3.8·21. (a) From Figure 83.8-21, we observe the following values of y[n]:
n 0 ± 1 ±2 ± 3 ± 4 ±5 ±6 ± 7 lnl > 7
y[nj 7 6 5 4 3 2 I 0 0
(b) The answer is identical to that of (a). This is because when we lay the tapes
x[m] and g[- m] together , the situation is identical to that in (a).
113
11 = 0 1~1: 1~ 1m 1n ,. 1- .. , •s ~ [i\ ..J
';1 [.OJ:: /5 ., . I' '
'"'
I ~ '
'
, ' 2.
~ L •J::.15 15_
-
~ - r• I~[: l ~f~ jt/;I, I•,,.· ·
1 -2. c .1. 2-3~ 5 <07
f\ /I~
~ t .. 1].:.15
Figure 83.8-19
SD
(\ .::o I
~ Lf.\J::. 5 0
Figure 83.8-20
3.8-22. (a)
8 hiOJ
12 hllJ + hlOl => h[l] = 12 - 8 =4
14 h[2l +hill + hlOJ => hl2J = 2
15 hl3J + hl2J +hill+ hlOJ => h{3J = 1
15.5 h['1J + h[3J + hl2J + h[ll + h[OJ ==> h[4J = 0.5
15.75 hl5J + hl4J + hf3l + hf2J + hflj + hlOJ => hl5J = 0.25
(b)
H = r: r n= r~2 I2 n a-· =
f = W'y = [ ~2 i2 ~ ][ ;{/9
Hence the input sequence is: (I, 1/3, l /9, · · · )
l =[ :~:]
3.8-23. Using the sifting property, the zero-input response is rewritten as y 0 [n] = 2u[n] +
(l/3rulnJ .
(a) Since the system is second-order, it has two characteristic roots. Both roots
appear in the zero-input response, an<l are easily identified as /t = I and /2 = 1/3
114
0
,, . :
0 I ' I , ""'
f ' ,, I 0 0 I • Y\ =O
t I I 0 I
I I' I I It I 0 0 I' .
';J[o]::7 .T lj t/l J
0
0 0
I I
I
I
I
I
I
I
I
I 0 0 ,
I
I I I 0 ••
..
I f l DD . 1'1 :: -l
'
' I I
0 I I I " )
Figure S3.8-21
Let a 0 = 1 (standard form), so ry2 +a 11'+W2 = (J'- l ){J'-1/3) = 12- 4/3ry+ 1/3.
Thus,
ao = 1, a 1 = - 4/3, and a 2 = 1/3.
{b) To produce a strong response, the input should be close to a natural mode. Since
the system has two bounded modes, any linear combination of these modes will
produce a strong response. That is, a strong response is generated in response to
x[nJ = CJ ( - 1tu[n],
{b) For systems connected in parallel, the total impulse response is just t he sum of
the individual impulse responses
115
Figure S3.8-24a: Plot of hi[n] and h 2 [n].
~·
-z
_, l
-s
- 10 -$ ... -2 10
(c) For systems connected in series, the total impulse response is just the convolution
of the individual impulse responses
116
~- 0
l l
'
->
-J
-·
-5
l
.. ,o -· ;: -> 10
Notice, t.he coefficients of the polynomial expansion are the same as the terms
from the convolution.
(b) It is possible to expand the product. of two polynomial expressions by convolving
t he coefficients from each ordered polynomial expression. Repeated convolution
can be used to expand the product of more than two polynomia l expressions.
Care must be used to include all coefficient terms, including zeros, from the
highest power to the lowest, and coefficients need to be ordered in a consistent
manner.
(c) (z - 4 - 2z-3 + 3z- 2 )'1 = ((z- 4 - 2z- 3 + 3z- 2 ) 2 ) 2 • First, use MATLAB and
convolution to compute (z- 4 - 2z- 3 + 3z- 2 ) 2 .
>>temp= conv([l -2 3),(1 -2 3));
Convolution is used to square this intermed iate result and obtain (z- 4 - 2z- 3 +
3z- 2)4.
>> conv(temp,temp)
ans = 1 - 8 36 - 104 214 - 312 324 -216 81
Recognizing that the highest power of z - 1 in the result. must be 16, the expansion
is
117
>> conv (temp, (13 0 -5 0 1])
ans = 13 52 47 58 137 104 321 -44 257 10 204 O -95 O 25
Recognizing that the highest power of z in the result must be 10, the expansion
is
3.8-26. (a) Actually, only a first-order difference equation is necessary to describe this sys-
tem. On refill n, the amount of sugar y[n] is equal to the amount added x[n]
plus what.ever was still in the mug. Since Joe drinks 2/3 of his cup of coffee
before each refill, one-third of the sugar from the previous cup remains. Thus,
y[n] = y[n - l]/3 + x[n]. Thus,
a,= -l/3,a2 =O,bo = l,b1 = 0, and b2 = 0.
ln standard form, the difference equation is
(b) Since Joe adds two teaspoons of sugar each time he fills his cup,
x[n] = 2u[n].
(c) The total solution to a difference equation is the sum of the zero-input response
and the zero-state response. Since Joe starts with a clean mug, y[- 1] = 0 and
the zero-input response is necessarily zero, y 0 [n] = 0. Thus, the total solution is
just the zero-state solution.
To obtain the zero-state solution, the impulse response h[n] is needed. In this
case, h[n] = 1 ~ 3 +yo[n]u[nJ, where Yo[n] = cy". To determine c, input x[n] = o[n]
into the original difference equation to yield h[n) - h[n - l]/3 = o[n]. Since h[n]
is causal, h[O] - h[- 1]/3 = h[O] = o[O] = l = y0 [0]u[O) = c. Thus, h(n) = 3-nu(n].
The zero-state solution is x[n] * h[n] = (l:~=O 2(3)-n) u[n) = 2 1 -13~1 ;;» u[n] =
1
(3 - 3-n) u[n).
Thus,
y[n] = (3 - 3-n) u[n].
(d)
That is, after many cups of coffee, Joe's mug reaches a steady-state of three
teaspoons of sugar.
One way to make Joe's coffee remain a constant for all non-negative n is begin
at the steady-state value of three and then add two teaspoons of sugar at each
refill. That is,
x[n) = 2u[n) + o[n] .
The added o[n) "jump starts" the sugar content of the first cup to the steady-state
value.
If Joe desires a steady value of two teaspoons, which is two-thirds the original
steady-state value, the input is simply scaled by 2/3. That is, the steady level
118
ylnJ = 2u[nJ is achieved using the input
4 2
x[n] = 3u!n] + 3o[nJ.
3.8-27. (a.) The characteristic equation is rt+ 0.5 = 0. Thus, the characteristic root of the
system is 7 = j).5. The impulse response has form h!nJ = eo[nJ + C"fnulnl =
c(J0.5)"u[n]. To determine c, evaluate the original difference equation according
to Jh!O] + 0.5h{- lj = JhlOJ= - 5o[O] = -5. Thus, h[Oj = J5 = c. Taken together,
h[n] = J5(J0.5)nu[n].
(b) First, compute the zero-input response y 0 (t) = c(j).5)n. To find c, use the
initial condition y0 [- 1J = J = c(J0.5)- 1 Thus, c = J20.5 = - 0.5 and volnJ =
- 0.5(]0.5)" .
The zero-state response is h!nJ * x[n] = ( l:~:g J5(J0.5)k) u[n - 5J =
5 l - (,O.S)n - • ) u[n - 5J.
(J l - 30.5
Adding the zero-input and zero-state responses yields the total response for n ~ 0,
4
y[nj = - 0.5(Jo.sr + ( J5 1 - 1 (JO 5)n-
_ ;O.S
)
uln - SJ.
3.8-28. (a) MATLAB is used to sketch the function h[nJ = n(u[n - 2J - u[n + 2]).
>> n = (- 5 :5); h = n.* ((n>=2) - (n>=-2));
>> stem(n ,h,'k'); axi s((-5 5 -2.5 2.5));
>> xlabel('n '); ylabel (' h(n)') ;
..•
'
0.$
.,,...
-o.s
-·
-u
-2
- 25
-· .. -> -2 _,
(b) Using Figure 83.8-28a, write h(nJ = 2oln + 2J + o(n + lJ - o[n - lJ . A difference
equation representation immediately follows from this form ,
119
3.8-30. For this solution, consider the signals x[n] = o[nl - o[n - l] = [1, - lJ. y[nl = z[n] =
o[n] + o[n - ll = [l, 11. In this case, x[nl (y[nl * z[nJ) = [l, -2].
(a) Not equivalent. By counter-example, (x[nl * y[n])z[nl = [1] ::ft [1, -2] =
x[nJ (y[nJ * z[nj).
(b) Not equivalent. By counter-example, (x[nly[nJ) * (x[nJz[nJ) = [1, -2, lJ ::ft
[1, -2J = x[nJ (y[nJ * z[nJ).
(c) Not equivalent. I3y counter-example, (x[nJy[n]) * z[nJ = [1,0, - 11 ::ft [l, -21 =
x[n] (y[n) * z{nJ).
{d) True. None of the above expressions is equivalent to x[nl (y[nJ * z[nJ).
3.9-1.
(E + 2)y[n] = Ex[nl
The characteris~ic equation is 'Y + 2 = 0, and the characteristic root is -2. Therefore
1
= HIe-11e- n =
-1
y~[nJ -e1--e-n = - - e-n
e- +2 2e + l
1
y[n] B(-2)" I - - e- " n >_ 0
2e + l
Setting rt= 0, and substituting y[O] = 1 yields
1=B+--
1
B=~
2e + I 2e + l
and
1
y[nJ = --{2e{-2)" + e-"] n ~ 0
2e+ l
3.9-2.
y[nJ + 2y[n - 1] = x[n - lJ (1)
We solve this equation iteratively to obtain y[OJ. Setting n = 0, and substituting
y[-lJ = 0, x[- 11 = 0, we get
(E + 2)y[nJ = x[nJ
The characteristic root is - 2. Therefore
y,.[n] = B(-2)"
For x[nJ = e-"u[nJ = rnu[n] with r = e- 1 ,
120
Therefore
y[nl = B(- 2)n + _ e_ e-" n ;:: 0
2e+ 1
Setting n = 0 and substituting y[OI = 0 yields
e -e
O=B + - - =* B =--
2e + l 2e + 1
and
y[n] = -2e e+-[-
1
(- 2)n + e - nj n ;:: 0
3.9-3.
For x [nJ = 3n
[11J = H [3l3n = (3)2 + 3(3) + 3 3n = (21 )3n
Y¢ (3)2 + 3(3) + 2 20
The total response
- 4 =
16 =
121
and
y [n l :::: -11 ( -1 )n - -39 ( -2 )n + -21 (3)n
4 5 20
3.9-4.
Also the system equation is (£ 2 + 2£ + l)y[n] = (2£ 2 - E)x[n], and x[n] = ( ~ )n.
Therefore
2 33
I6
!)
4
v!nl = 'Lk
k: O
Hence
y [n + l] - y[n] = n +1
Hence, the required solution for y[n] is the solution of the above difference equa-
tion. To find the initial condition, we know that y[O] = 0.
We can rewrite the difference equation as
(E - l)y[n] = n +1
Hence the natural response is
Yc[n] = B(lf = B
From Table 3.2, we find the forced response form as
122
Substitution of this expression in the difference equation yields
or
c1 (n + 1) + c2 (n + 1)2 - c1n - c2n2 = n + 1
Equating coefficients of similar power yields
C1 + Cz = 1 }
~ C1 = C2 = -21
2c2 = l
Hence
y[n] = B + n(n2+ 1)
Setting n = 0 and using the initial condit ion y[OJ = 0 yields
O=B
Hence
y[n] = tk
k=O
= n(\+ 1)
(b)
"
y[nJ = 2:k3
k=O
y[n+ l] -y[nj = (n+ 1) 3
The natural response is
Yc[n] =B
The initial condition is y[O] = L?n=O k3 = 0. The forced
response , as seen from
Table 3.2, is
Yef>[nJ = c1n + c2n 2 + c3n3 + C4n4
Substitution Yef>{n] in the difference equation yields
CJ + C2 + C3 -f- C4 = l } C1 =Q
2c2 + 3c3 + 4c4 = 3 c2 = 1/4
3c3 + 6c4 = 3 ~ c3 = 1/2
4C4 = l C4 = 1/4
Hence
1 2 1 3 1 4
y[n] = B + -n + -n + -n
4 2 4
Also y[n] = 0. This yields B = 0, and
1 1 1 2( 2
y[n] = "\"'n "' k3 = -n2 + -n3 + -n4 = n n + 1)
L, 4 2 4 4
k=O
123
3.9-6.
n
y[nJ = Lkrk
k=O
Therefore
y[n + l J - y[nJ = (n + l)rn+l
The natural response (for characteristic root 1=1) is
Yc[nJ = B
The forced response, as seen from Table 3.2, is
CO + C1 - fll. = 1 } C1 = _r_
[ ==} r-_!,.
C1 - 7 =1 Co= {.'.-!)~
Hence
yin] = B + -r + -rn- ] rn
[ (r - 1)2 r - 1
Moreover y[O] ""' 0. Hence
y[nJ = r
+ [ -r
+ -rn]
- rn = r+[n(r - 1) - l]rn+l
--'---'----~,.---'---
(r - 1) 2 (r - 1) 2 r - 1 2
(r - 1)
3.9-7.
,
2
-1 + 0.16 = (I - 0.2)(1 - 0.8)
124
This yields
-0.12c(0.2)" = 0.2(0.2)"
Therefore
5
c=--
3
and
Y¢(n] = - ~n(0.2)"
y[n] = B 1 (0.2)" + fl2 (0.8)" - ~n(0.2)" n ~0
Setting n = 0, l, and substituting initial conditions y[O] = 1, y[l] = 2, yields
y(n] = - 2: (0.2)" +
32
9
(0.8)" - ~(0.2)" n ~0
3.9-8.
Therefore y(O] = 0 and y[l] = 0.5. For the input x(n] = cos(,.;'+~).
1Tn 7T
Y¢(n] = ccos( T + 3 + </>)
But y,p[n] satisfies the system equation, that is,
or
or
7Tn ;r ;rn ;r 1Tn 7T 7T 7T 7T
+ <f>) +csin( T +
-ccos(-z +
3 3 +</>) +0.16ccos(-z +
3 +</>)=cos( n+
2 3 + 2)
or 1Tn 7T 7Tn n 7T
2.27) =cos( 2 + 3 + 2)
L306ccos(
2 + 3 + </> -
Therefore
l.306c = 1 c = 0.765
125
</>- 2.27 = -7r </> = 3.84 = -2.44rad
2
Therefore
?TTL 7r
Y<t>[nj 0.765 cos( + -
2 3 2.44)
. 1Tn
0.765 cos( T - 1.393)
1Tn
y[nJ = B 1 (o.2r + B2(0.8r + 0.765 cos(
2 - 1.393)
0 = B1 + B 2 + 0.1354 } B1=0.241
0.5 = 0.2B 1 + 0.8B2 + 0. 753 => B 2 = -0.377
?rn
y[nJ = 0.241(0.2r - 0.377(o.8r + 0.765 cos( 2 - 1.393)
3.10-1. Assume that a system exists that violates (9.61) , and yet produces bounded output
for every bounded input. The system response at n = k 1 is
00
y[nd = E h[m]x[n 1 - m]
m=O
1 if h[m] > 0
x[n1 -m] ={ -1 if h[m] < 0
In this case
h[m]x[n 1 - m] = lh[mJI
and 00
y[n1] = L ih[rn] I = oo
m =O
This violates the assumption.
3.10-2. (a)
Roots are 0.2 and -0.8. Both are inside the unit circle. The system is BIBO
stable and asymptotically stable.
(b)
Roots are -1 and - 2. One root outside the unit circle and the other is on the
unit circle. The system is BIBO unstable and asymptotically unstable.
(c)
Roots are 1 (repeated twice) and - 0.5. Repeated root on unit circle. The system
126
is BIBO unstable and asymptotically unstable.
(d)
')'
2
+ ')' - 2 = ('Y + 0.5 + jl.5)(1' + 05 - jl.5)
Roots are -0.5 ± jl.5. Both roots outside the unit circle. The system is BIBO
unstable and asymptotically unstable.
(f)
Roots are ±1, ±jl. All the roots are simple and on unit circle. The system is
BIBO unstable and marginally stable.
3.10-3. The system S 1 is asymptotically (and BIBO) unstable. The system S2 is BII30 and
asymptotically stable. If we cascade the two systems, the impulse response of the
composite system is
(d) To have a mode that decays at a rate of 2-n, at least one characteristic root needs
to lie on the circle of radius one-half centered at the origin. By this criteria,
(e) For a second-order system with two finite roots to only have one mode, one
characteristic root needs to be located at the origin. By this criteria,
127
3.10-5. Notice, the system response can be written more simply as h[nJ = 8lnJ + ( k) '' u.[n-1J =
or 1L[nJ.
(a) Yes, the system is stable since the impulse response function is absolutely
summable. T hat is, I;~=-oo ih[n] I = I::=o (~)" = 1 ~~~3 = 3/2 < oo.
Yes, the system is causal since hln] = 0 for n < 0.
(b) MATLAB is used to plot x!nJ.
>> n = (-5:5); x = (n>=3)-(n>=-3);
>> stem(n,x,'k') ; axis((-5 5 - 1.2 1.2));
>> xlabel('n'); ylabel('x[n]');
.
.,•
~
.
-·••
, I
Figure S3.1G-5b: Plot of x[nJ = uln - 3) - uln + 3].
(c) The zero-state response is computed as vlnJ = xln] * hlnJ. This convolution
involves several regions.
For (n < - 3), yin}= 0.
For (- 3 :5 n < 2), y[n] = I:~= - 3 -{l/3r-k = -{1/3)" I:~= - 3 31<: =
- (1/3)" 3 -3;:_3~' H = r<n;3) _ 3.
For (n;:::: 2), y[n] = I;~=-l -(1/3)"- k = - {1/3)" I;~=- 3 31<: = - {1/3)" 3 -L~t =
- 75218(3) - ".
Combining yields
n <-3
-3 :5 n < 2
n;:::2
3.lG-6. (a) No, the system is not causal since h[n) i- 0 for (n < 0).
128
.s
! 6 v -
.5
· I .s
' _, 10
( b) '°'oo
l..Jn=-oo
lh[nJ ! ,~co
=
~n=-oo
I (!)lnl I =
2
"co
i...Jn=-oo 2
(!)lnl = '°'-1
L,,,n=-oo
O.s-n +
"co 0 5" = '°'-1 2n + '°'oo 0 · 5n = 0-20 + 0.5o-O = 1+2=3·
L-n=O · L-n=-oo L-n=O 1-2 1-0.S
2: !h!nJl=3.
n=-oo
9
Ex = oo and P" = 2.
(d) y[n) = x[n] * h[n] = L~-oo h[k)x[n - kJ. Thus, y{lO] = L~-oo h[kJx[lO -
kJ = L~-00 Q_5lkl3u[l0 - 5 - kJ = z;~-00 3(2k) + L~=O 3(0.5k) = 3 °;=:_;° +
3 o.so-o.s6 = 3 + 3fil - 285 ~ 8 91 Thus
1-0 .5 32 - 32 ~ . . '
285
y[lO] = 32·
3.M-1. To accommodate upsampling, the inline function needs to be modified so that it assigns
a zero for non-integer inputs.
The added term (n==fix(n)) is one if n is an integer and zero if n is not an integer.
The modified function is easy to test.
As hoped, f [n/2] inserts a zero between every sample off [n), which corresponds to
the desired upsample by two operation.
129
...
l
l
-2 10
3.M-2. There are many ways to solve this problem. For this solution , let din] designate the
distance from the student's dest ination, which alternates between home and the exam
location, just before the student changes his mind. For even-valued n the destination
is home, and for odd-valued n the destination is the exam location.
(a) Just before changing direction, the student is a distance of d[n] miles from his
destination. Turning around, his next. destination is therefore a distance of 2-d[n)
miles. The student travels one-half of this distance, which leaves 2 -;lnl miles
remaining. Thus, a difference equation description of this problem is d[n + 1] =
2
-;(nl = 1 - 0.5d[n] = ufn) - 0.5d[n). Rearranging and shifting by one yields
For this description, d[O] = 0. This auxiliary condition simply states that before
the student first decides to go to the exam, he is at home.
(b) MATLAB is used to iteratively simulate the difference equation.
>> n = [0:20); d = zeros (size(n));
>>for index= find(n>O), d(index) = 1- 0.5•d( index-1) ; end; d
d = 0 1 .0000 0.5000 0.7500 0.6250 0 . 6875 0.6563 0.6719
0.6641 0.6680 0.6660 0.6670 0.6665 0.6667 0.6666 0.6667
0.6667 0.6667 0 . 6667 0.6667 0.6667
As time increases, the remaining distance to the destination (alternating home
and exam) reaches a steady-state value of two-thirds of a mile. That is,
n-oo
Jim d[n] = 2/3.
Changing the problem so that the student travels two-thirds the remaining dis-
tance each time changes the result. In this case, the difference equation is
d[n] +~din - l ] = ~uln).
>> n = [0:20); d = zeros ( size (n));
>>for index = find(n>O), d(index) = (2-d(index- 1))/3; end; d
d 0 0 0.6667 0.4444 0.5185 0.4938 0.5021 0.4993 0.5002
0.4999 0.5000 0.5000 0.5000 0.5000 0.5000 0.5000 0.5000
0.5000 0.5000 0.5000 0.5000 0.5000
130
ln this case, the steady-state remaining distance is limn-co d[nJ = 1/2.
(c) The closed form, or total, solution is the sum of the zero-input and zero-state
responses. Since the auxiliary condition is d[OJ = 0, the zero-input response is
just zero. To compute the zero-state response, write the difference equation as
dfnl + 0.5d[n] = u[n - lJ = x[nJ, where the "input" x[nJ is just a shifted unit
step. In this way, h[nJ = (-1/2)°u[nJ. The final solution is d[nJ = h[n] * x[nJ =
(L:~;:~(- 0.5)k) u[n -11 = \-J(_00Wu[n -1]. Thus,
2
d[n] = J (1 - (-0.5)°) u[n - l].
131
>> xlabel('k'); ylabel('r_{xy}[k]');
•• ••
t
....
;;:
..
_,
• ,. " 20
.,
• 20
-·· 10
·•'---'---'--~~'---'---'---'-~--'---'
- 10 -6 10 15 20 25 )(I )$ 40
•
Figure S3.M-3(b)ii: Plots of x(n], y(nj, and r., 11 (k].
The largest magnitude of rz 11 [k] is five and occurs at k = 19. Signal x(n] has
a unit pulse of width five starting at n = 5. Signal y[n] has a similar feature:
a negative unit pulse of width five starting at n = - 14. These two similar
features are separated by a shift k = 5 - (-14) = 19. Since these are the
most similar features between the two signals, it seems very sensible that the
autocorrelation function has a large, negative value at shift k = 19.
3.M-4 . (a) function [y] = filtermax(x,N)
'l. function [y) filtermax(x,N)
132
10 20 30 40
->~------~
0 t0?03040
..
i
N
i
~
~
• .!! •
~
-·_, ·1
• 10 2<) :io 40
•1
0 10
.
2<) :io 40
not result in a sinusoidal output, the max filter cannot be a LTI system (the max
filter is TI but not linear).
3.M-5. (a) function (y] = filtermin(x,N)
'l. function (y] filtermin(x,N)
133
..
0
.__ _____....___,
102030-40
"' 0' - - ------"""'---'
102030"40
to 20 JO •O 10203040
......______.,,___,
0 10203040
..
0
______
.__
'°2030'1>
__,
..
11
.l
N
i
!
?: ... _,
" ., "'
·2
0 10 20 31) 40
·2
0 10 20 ,. 40
The plots are consistent with the expected behavior of a median filter. The
134
output magnitude tends to be smaller lhan the input magnitude. Unlike the
max or min filters, the median filter is not sensitive lo outliers in the input data.
Also notice that the median filter is an FIR filter. Thus, a sinusoidal input reaches
steady-state in after N - 1 samples. Furthermore, since a sinusoidal input does
not result in a sinusoidal output, the median filter cannot be a LTI system (the
median filter is TI but not linear).
3.M-7. (a) Replacing h[n] with y[n] and c5[n] with x[n] yields the desired difference equation,
y[nJ =
k=-(N-1)
~ (1 -1; I) x[n - k].
(b) To make the system causal, h[n] must be right-shifted by at least (N - 1).
Since the system is t ime-invariant, shifting h[n] causes the output to be delayed
(shifted) by the same (N - 1) amount.
(c) function [a,b] = i nterpfilter(N)
%function (a,b) = interpfilter(N)
a= 1; b = conv(ones(N,1),ones(N,1))/N;
(d) >> n = [0:9]; x = cos(n);
» tJ = 10; nup = (O:N*length(n)-1];
>> xup = (x;zeros(N-1,length(x))); xup xup(:);
>> (a,b) = interpfilter(N);
>> y = filter(b,a,xup);
>> subplot(311),stem(n,x,'k');
>> xlabel('n'); ylabel('x(n)');
>> subplot(312) ,stem(nup,xup,'k');
>> xlabel('n'); ylabel('x_{up}[n)');
>> subplot(313),stem(nup,y,'k');
>> xlabel('n'); ylabel('y(n] ');
<~ 0
'. ; l ; '. ~
1 z 3 ..
n
5 6
t ·=1
6'
'f~H2Y
.0 W N ~ ~ ~ " ~ ~ ~ ~
135
3.M-8. (a) First, rewrite the impulse response as h[n] (u[nJ - u[n - NJ) /N =
-}; L,~;01 o[n - kJ. Replacing h[n] with y[n] and o[n] with x[n] yields t he de-
sired difference equation,
N -1
y[n] =~ L x[n-k].
k=O
a= 1; b = ones(N,1)/N
(c) >> n = (0:44); x = cos(pi*n/5)+(n==30) - (n==35);
>> N = 4; [a,b) = filterma(N);
>> y1 = filter(b,a ,x);
>> N :: 8; (a,b) = filterma(N);
>> y2 = filter(b,a,x);
>> N = 12; (a,b) = filterma(N);
>> y3 = filter(b,a,x);
>> subplot(221); stem(n,x,'k'); axis((O 44 -2 .1 2.1));
>> xlabel('n'); ylabel('x[n] ');
>> subplot(222); stem(n,y1,'k'); axis((O 44 -2 . 1 2.1]);
>> xlabel('n'); ylabel('y(n) for N=4');
>> subplot(223); stem(n,y2,'k'); axis( [O 44 -2.1 2. 1));
>> xlabel('n'); ylabel('y(n) for N=8');
>> subplot(224); stem(n,y3,'k'); axis( (0 44 -2.1 2. 1));
>> xlabel('n'); ylabel('y[n) for N=12');
·· ~----~~~ ·· ~------~
0 t0 20JO .CO
~ z
~
-~
~
* ·1
·2
10 .,, ·2
10
0 20 30 0 20 30
"'
Figure S3.M-8c: P lots for JV-point moving average filters.
T he plots are consistent with the expected behavior of a moving average filter .
The output is a low-p ass version of the input. Larger values of N average over a
wider window; thus, larger N results in greater attenuation of the input sinusoid.
Large outliers have high frequency content and are significantly attenuated.
Also notice that t he moving average filter is an FIR filter. Thus, a sinusoidal
input reaches steady-state in a fter N - 1 samples. This particular filter is a LTI
filter, so a sinusoidal input will result in a steady-state sinusoidal output.
136
(d) The total impulse response of a cascade of two N -point moving average fi lters is
hcascadelnJ = ( u(nJ -~n-NJ) * ( uJn)-~n- NJ) = c~=~=O N-2) (u[nJ - u[n - NJ) +
(L:~=-,;_ (N-I) N- 2 ) (u[n - NJ - u[n - (2N - l)]) = ~ (u[n J - u[n - NJ) +
2
NN~- 1 (u [n - N J - u[n - (2N - l)]). That is, hcascade[nJ is a triangle-shaped
function with width 2N - 1 and maximum height l/N.
T he impulse response of a causal linear interpolation filter is hJininterplnJ =
L;~;~<N-i) (1 -1~1) o(n-(N - 1)-k). This function is identical to hcascade[nJ
except that it has a maximum height of 1. Thus, the cascade of moving average
fil ters is a factor 1/N d ifferent than the linear interpolation filter.
Nhcascade[nJ = hlininterpln].
,~1'. ; i ~ I
o , 2 3
1 '. '.
.. ~ s 1
'.
e a
s ~6 i r-'...~-~ ·:lj
0 • w ~ ~ •
n
~ ~ ~ ~ ~
·~~ 0 W W ~ ~ •
n
~ ro M W ~
137
Chapter 4 Solutions
4.1-1. (a)
- e-st 11
s 0
- ~re-$ -11
s
~(1 - e- s]
s
Note that the result is valid for all values of s; hence the region of convergence is
the entire s-plane. The abscissa of convergence is a 0 = -oo.
(b)
e-<s+l)t
00
- (s + 1)2 1-(s + l)t - llo
1
(s+1)2
x(t} = tcoswolu(t)
X(s) = 1 00
tcoswote- s' dt
1 [ 1 1 ] Re (s) > 0
2 (s - jc.,,·0 ) 2 + (s + jwo)2
138
{d)
100 dt - 2 fo°"
e-(•-2}t dt e -(•+l )t
l 2
= - ----
s-2 .s+l
We get the first term only if Re s > 2, and we get the second term only if Re
(s) > -1. Both conditions will be satisfied if Re (s) > 2 or ao > 2. Hence:
I 2
X(s) = s- 2- s+1 for ao > 2
(e)
00 (
cos w1 + w2 )l e- stdt + 1 lo
00
X( s ) 2I l {o {
2 cos ( w 1 - w2
)
te -stdl
= 1[
2
s
s 2 + (w1 + w2) 2
+ s
s 2 + (w1 - Wz)2
]
x(t) = cosh(at)u(t)
X(s)
s2 - a2
(g)
:i:(l) = sinh(at)u(t)
139
a
=
s2 -a2 Res> lal
(h}
Hence 1 -8
X s = -e' ( 1 ) + -1 e-3·o ( 1 )
( ) 2 s + 2 - j5 2 s + 2 + j5
This is valid if Re (s) > -2 for both terms; hence
X(s)= (s + 2)cos8-5sin0
s 2 + 4s + 29
4.1-2. (a)
X(s) =
1 ,0
te-st dt = -e-· t (-st - 1)
s
I'
0
=
1 (1- e-s - se-•)
2s
(b)
2s + 5 2s + 5 1 l
X(s) = = =-+-
s2+ss+6 (s+ 2}(s+3) s; 2 s+3
x(l) = (e - u + e- )u(t)
3 1
(b)
X( ) 3s + 5
= s 2 + 4s + 13
s
r= Jl17+25-60 = 3.018
13 - 4
140
(c)
X s - (s + 1)2 - (s + 1)2 .
( ) - s2 - s - 6 - (s + 2)(s - 3)
X 2s + 1 -1 As+ B
(s)= (s+l)(s2+2s+2) = s+l + s 2 +2s+2
s+2 2 k 1
X(s) = s(s+1) 2 =; + s+l - (s+l)2
0 = 2 +k +0 ==? k =- 2
141
and
2 2 1
X(s) -----
s s+1 (s + 1)2
x (t) (2 - (2 + t )e-tju(t)
(g)
x s = 1 = _1_ + ~ + k2 + k3 - 1
() (s+l)(s+2) 4 s+l s+2 (s +2) 2 (s+ 2)3 (s + 2) 4
Multiplying both sides by s and let s -+ oo. This yields
0 = 1 + k1 =? k1 = - 1
1 1 1 k2 k3 1
- - - -- = - - - - - +
(s+l)(s+2)4 s+l
+ - - --
s+2 (s + 2)2 (s+2)3 (s+2) 4
Set.ting s = 0 and -3 on both sides yields
1 1 k2 kJ 1
-= 1 - - + - +- - -
16 2 4 8 16
1 1
- 2 = - 2 + 1 + k2 - kJ - 1
1 1 1 1 1
X (s) ----- -
s+1 s+2 (s + 2)2 (s + 2)3 (s + 2) 4
t2 t3
x(t) [e-t - (1 + t + + )e-2tJu(t)
2 6
Comment: This problem could be tackled in many ways. We could have used
Eq. (B.47b), or after determining first two coefficienls by Heaviside method, we
could have cleared fractions. Also instead of lettings= 0 and - 3, we could have
selected any other set of values. However, in t.his case t.hese values appear most
(h) suitable for numerical work.
142
c = 5, b = v's=-4 = 1.
r= /5+1+4=Vi0
v
5-4
1 1 ../15
x(t) = !20 - 4(1 - 2t)e- 2 t + - -e- 21 cos(t + 71.56°)Ju(t)
5
(i)
s3 k 1/4 As+B
X(s) = (s+l)2(s2+2s+5) = s+l - (s+1) 2 + s 2 +2s+5
.Multiply both sides by s and let s ____, oo to obtain
I=k+A
4.2-1. (a)
and
X(s) L.:ju(t)] - L.:ju(l - l)J
-1 - e-s -1
s s
~(1 - e-•)
s
(b)
143
(c)
1 1
X(s) = e_.,. ( -- ) e_,.,. = ( -- ) e-C•+I)T
s+l s+l
(e)
Therefore
X( s) = ( 2Wo 2) e-s-r
s +w0
(g)
Therefore
X(s) = [coswor ( wo )
s 2 +w02
+ sinwor (A)]
s +
e-•T
w0
4.2-2. (a)
144
(b)
4.2-3. (a)
where
A = 1, B = 0, a = 1, c = 2, b = 1
X i ( s ) =s- - - {
s2 + 2s + 2 r = ,/2, 0 = tan- 1 (1) = 7r/4
(c)
(e)e-s 3
X(s) = +----
s2 - 2s + 5 s 2 - 2s + 5
1 3
e e- 8 +-- - -
s2 - 2s + 5 s 2 - 2s + 5
= eX1(s)e-s + X2(s)
145
where
1 1 .
X1(s) = s2 - 2s + 2
and x1(t) =
2
etsm2tu(t)
X2(s) =
s2 -
3
2s + 2
and x 2 (l) = ~etsin2tu(t)
Therefore
x(t) = ex1(t - 1) + x2(t)
= ~e<t- 1 >sin2(t - l)u(t- l) + ~etsin2tu(t)
2 2
(d)
= (e-• + e - 2s + 1) [ 1 ]
s2 + 3s + 2
(e- • + e- 2• + 1) [-1- -_1_]
s+ l s+2
and
G(s) X (s) + X (s)e-sTo + X(s)e- 2 $To + .. .
= X (s)p + e-•To + e-2aTo + e- 3sTo + .. .
X(s)
1 _ e-sTo
(b)
1
x(t) u(t) - u (t - 2) a nd X(s) = -(1 - e- 28 )
s
G(s)
=
X (s)
1 - e-Sj
1(1 -e-
=$ 1- e - 8s
2
•)
4.2-5. Pair 2 1 1
l
t
u(t) = o(r) dr <=> -(1) =-
o- s s
1 1 1
Pair 3
l
t
tu(t) = u(r) dr <=> -(-) = 2
o- s s s
Pair 4: Use successive integration of tu(t)
Pair 5: From frequency-shifting (4.23), we have
1 1
u(t) <=> - and e>.'u(t) <=> - -
s s - ,\
Pair 6: Because
146
l l
tu(t) <==> s2 and teA'u(t) {==> (s _ >.) 2
Pair 7: Apply the same argument to t 2 u(t), t 3 u (t), ... , and so on.
Pair 8a:
1
cosbtu(t) = .!.(e;bt + e-;bt)u(t) <==> ! (- -. +_ l_ _ ) = _ s_
2 2 s - Jb s + Jb s 2 + b2
Pair 8b: Same way as the pair 8a.
Pair 9a: Application of the frequency-shift property (4.23) to pair 8a cosbtu(t) <==>
s•-~1,'l yields
- at ( ) S +a.
cos bt u t <=> (
e
s+a.) 2 + b2
Pair 9b: Similar to the pair 9a.
Pairs lOa and lOb: Recognize that
Now use results in pairs 9a and 9b to obtain pair lOa. Pair lOb is equivalent to pair
lOa.
4.2-6. (a) (i)
dx
c5(t) - c5(t - 2)
dt
sX(s) I - e- 2 •
X(s) = !(l
s
- e-2•)
(ii)
dx
c5(t - 2) - c5(t - 4}
dt
sX(s) = e-2s _ e- 4•
dx
u(t) - 3u(t - 2) + 2u(t - 3)
dt
sX(s) ~ - ~e- 2 • + ~e- 3" [x(o- ) = 0)
s s s
X(s) 2_ (1 - 3e- 2• + 2e- 3")
s2
4·2-7· X( s } -- e - 3 e-• [ •
2
(s+l)(s+2)
] -
- e -3 e - s [1 + - 3s-2 ]
(s+l){s+2)
_
- e-3 e -s [1 + I
(s+I)
+ -4 ]
(s+2} ·
Thus,
4.2-8. First, note that the nth derivative of .~ 4 is (~~11~:~'1 . Thus, rewrite the transform as
_ 1 _ 1 12! _ 1 d' 2 . ( . s1gna
i · 1 ) s·mce CT > -1,
X ( s) - (s+l}'• - 12, ~ - 12 , the t1me-domam
dsT'1 •+I .
x(t) must be right sided. Repeated use of the differentiation in s property provides
147
the resulting inverse transform.
d
.C[tx (t)] = - dsX(s).
(b) y(t) = tx(t) = tiu(t) = u(t). Thus, Y(s) = J~00 u(t)e-stdt = J0 00
e-•tdt =
e~··s 1t=O For u > 0, this simplifies to Y(s)
00
• = l.
s
(c) Combining the previous two parts yields -f 3
X(s) = ~ - Thus,
4.3-1. (a)
1
(s2 + 3s + 2)Y(s) = s(-)
s
1 1 1
Y(s) = s 2 + 3s + 2 = s + l - s + 2
y(t) = (e - t - e- 21 )u(t)
(b)
1
(s2 Y(s) - 2s - l) + 4(sY(s) - 2) + 4Y(s) = (s + 1) -
s+
-
l
or
(s 2 + 4s + 4)Y(s) = 2s + 10
and
2s + 10 2s + 10 2 6
Y(s) - ---=
s2 + 4s + 4 (s + 2) 2
=--+
s I- 2
-(s +-2)-2
y(t) = (2 + 6t)e- 2 tu(t)
(c)
or 50 s + 32s + 50
2
(s2 +6s + 25)Y(s) = s+ 32+ -
s
=- -s--
and
s + 32s + 50
2
2 - s + 20
Y(s) - -- - -
s(s2+6s + 25)
= - + -
2 ---
s s +6s+25
y(t) · (2 + 5.836e- 3 t cos(4t - 99.86°))u(t)
4.3-2. (a) All init ial conditions are zero. The zero-input response is zero. T he entire re-
sponse found in P rob4 .3-2a is zero-state response, that is
148
Yzi (t) = 0
2s+9 1
Y(s) - - - - +-- --
s2 + 4s + 4 s 2 + 4s + 4
'----v---" '---.,.----"
zero-input zero-state
2 5 1
--+
$ +2 (s + 2)2
+(s-- -
+ 2)2
'--v--'
zero-input zero-state
y(t) = (2 + 5t)e-2 t + le-2t
..__,_.,
'----v----'
zero-input zero-state
s+7 25s + 50
Y(s) = - - - - + --::--- - ---,-
s2 + 6s + 25 s(s2 + 6s + 25)
'----v--' ..._____..,
zero-input zero-state
s+7 2 - 2s + 13
( s2 + 6s + 25) + ( + s2 + 6s + 25)s
y(t) [v'2e- 3
t cos(4t - ~ )] + [2 + 5.154e- 31 cos(4t -112.83°)]
zero-input zero-state
149
Y2(s) = 1 -:-- --,-1.,...-----,.. = _l/_4 __1_/3_ + _I/_1_2
s(s 2 + 5s + 4) s(s+ I)(s+4) s s+ l s+4
and
Y1 (t) ( ~ - ~ e-t - ~e- 4 t)u(t)
2 3 6
Y2(t) (1
= 4 - 3e- 1 t 1 4t ( )
+ 12e- )u t
If H1 (s) and H2(s) are the transfer functions relating y 1 (t) and y 2 (t), respectively
to the input x(t), thus
s+2 1
H I ()
S - --=---- - and H2{s) =
- s 2 + 5s + 4 s2 + 5s+ 4
(b) The Laplace transform of the equations are
s+ I s+ l l 0.724 0.276
Y1(s)
s(s2 +3s+ l ) s(s + 0.382)(s + 2.618)
- s s+ 0.382 s + 2.618
s+2 s+2 2 1.894 0.1056
Y2(s) = s(s 2 +3s+ l) = s(s -t 0.382)(s + 2.618) s s + 0.382 s + 2.618
H()
IS -
- s2
s+l
+ 3s -t 1
and H2 (s) = s 2 +s+2
3s + l
4.3-4. At t = 0, the inductor current y 1{0) = 4 and the capacitor voltage is 16 volts. After
t = 0, the loop equations are
If
150
Or
(2s + 5)Y1(s) - (2s + 4)Y2 (s) 8+ 40
s
1 - 8- 16
-(2s + 4)Y1 (s) + (2s + 4 + -)Y2(s)
s s
Cramer's rule yields
4(6s2 + 13s + 5) l6s + 28
Y1 (s) --'---------'- = 8- + - ---
s(s2 + 3s + 2.5) s s 2 + 3s + 2.5
t
YI (t) [8+17.89e-L5 t cos( - 26.56°)]u(t)
2
20(s + 2)
(s2 + 3s + 2.5)
= ~ t 7r
20v ;.:;e- i.st cos( - - - )u( t)
2 4
5s+3
4.3-5. (a) •'+11s+24
3s 2 +7s+5
(b) s3 +6s 2 -lls+6
3s+2
(c) s(s3+4)
I
(d) s+T
151
(b)
(iii) The input here is the input in (ii) delayed by 5 secs. Therefore X(s) =
1 -5s
s+4e
Y (s) = X(s)H(s)
Consider an input x 1 (t) = x(t). Then X 1 (s) = sX(s) . If the output is y 1 (t) and
152
its transform is Y 1 (s), then
H(s) = s +5
+ 2)
s2 (s
The characteristic roots are 0, 0, -2. There are repeated roots on imaginary
axis. Hence the system is BIBO (and asymptotically) unstable.
(iii)
H (s) = s(s + 2)
s+s
Although the characteristic root -5 is in LHP, because M > N, the system
is BIBO unstable.
(iv)
H( ) s+5
s = s(s + 2)
The roots are 0, and -2. One of the roots is on the imaginary axis which
makes the system BIBO unstable (but marginally stable).
(v)
s+5 s+5
H(s) = s 2 -2s-3 = (s -3)(s+l)
The roots are 3 and -1. One root in RHP makes system BIBO unstable (and
also asymptotically unstable).
(b) (i)
(D 2 + 3D + 2)y(t) = (D + 3)x(t)
or (D + I}(D + 2)y(t) = (D + 3)x(t)
The system transfer function is
H(s) - s +3
- (s + l)(s + 2)
The characteristic roots are -1 and -2 (both in LHP). Hence the system is
153
asymptotically a.nd BIBO stable.
(ii)
H( ) s+1 1
s = (s + l )(s + 2) = s +2
The characteristic roots are -1 and -2 (both in LIIP). The only pole of H(s)
is at -2. Hence the system is asymptotically and BIBO stable.
(iii)
(D 2 + D - 2)y(t) = (D - l)x(t)
(D 2 - 3D + 2)y(t) = (D - l)x(t)
The characteristic roots are 1 and 2. The only pole of H(s) is at 2. Hence
the system is asymptotically and BIBO unstable.
4.4-1. Figure 84.4-1 shows the transformed network. The loop equations are
1 1
(1 + - )Y1 (s) - -Y2(s)
s s (s+ 1)2
1 l
-- Yi(s) + (s + 1 + - )Y2(s) 0
s s
[:; 1
Y (s) ]
] [ Y2(s)
=[ ~ ]
0
Cramer's rule yields
1 1
~~-.,,...-.....,-~~~= ~~~
(s + 1)2 (s 2 + 2s + 2) (s + 1) 2 s2 + 2s + 2
154
vo (t)
1. s
=.T ' :
v(s)
0
Figure 84.4-1
4.4-2. Before the switch is opened, the inductor current is 5A, that is y(O) = 5. Figure
S4.4-2b shows t he transformed circuit for t ~ 0 with initial condition generator. The
current Y(s) is given by
5 2-
{b)
Awo A A
X(s) = s 2 +w0 2' Y (s) - Lw
oS
and y(l) = Lwo u(t)
155
4.4-4. At t = 0, the steady-state values of currents y 1 and y 2 is y 1 (0) = 2, y 2 (0) = 1. Figure
84.4-4 shows the transformed circuit for t :2: 0 with initial condition generators. The
loop equations are
6
+ 2)Yi(s) - Y2 (s)
(s 2+-
s
-Y1 (s) + (s + 2)Y2(s)
-
~
5
Figure S4.4-4
4.4-5. The current in the 2H inductor at t = 0 is lOA. The transformed circuit with initial
condition generators is shown in Figure S4.4-5 for t :2: 0.
= 20 /11
3(1.168)e-tfG cos(-- t - 31.1°)u(t)
6
7.787e- t/G cos(~ t - 31.1°)u(t)
156
(s+~)Y(s)=(s +l)( 20s+10 )= 20(s +1)(s+0.5)
2 2
s
.;~
V(s
) 1
,
- I
Figure 84.4-5
4.4-6. Figure 84.4-6 shows the transformed circuit with mutually coupled inductor replaced
by their equivalents (see Figure 4. 14b) . The loop equations are
100
s
-2sY1 (s) + (4s + l)Y2(s) 0
~ 25
Vfr")
0
Figure 84.4-6
4.4-7. Figure 84.4-7 shows the transformed circuit with parallel form of initial condition
generators. The admittance W(s) seen by the source is
13 s 2 +4s+ 13
s =-+s+ 4 = - - - -
W()
s s
The voltage across terminals ab is
V. (s) _ I(s) _ ~ +3 _ 3s + l
ab - W(s) - s2+4s+l3 - 82 + 4s + 13
Also
157
V.0 ( ) _ ~ V. ( ) _ 3s + 1
_____
s - 2 ob s - 2(s 2 + 4s + 13)
_ _..,..__
and vo(t) = l.716e- 2 t cos(3t + 29°)u(t)
\1/'l>
'\_ t/g
3 s
1.3
I
s I
-+3
.s
Figure S4 .4-7
4.4-8. The capacitor voltage at t = 0 is 10 volts. The inductor current is zero. The trans-
formed circuit with initial condition generators is shown for t > 0 in Figure 84.4-8. To
determine the current Y(s), we determine Za&(s), the impedance seen across terminals
ab:
1 3s+8
Zob(s) = ( ) = -
4s+ 11
1
1 + 2+•-t2
•-t3
l!Q
Also Y(s) = •3
~ + (4:ti81)
90(4s + 11)
= 3s2 + 28s + 55
30(4s + 11)
52 + ~s
3
+ 553
30(4s + 11)
(s + 2.8)(s + 6.53)
1.61 121.61
-s--+
+ 2.8 - --
s + 6.53
and y(t) = II21.61e- 6 · 53 t - l.61e- 2 · 8 tJu(t)
'fM z. 'l.
I\
~ -s s
~ ~
~
Figure 84.4-8
4.4-9. Figure 84.4-9 shows the transformed circuit (with noninverting op amp replaced by
its equivalent as shown in Figure 4.16) from Figure 84.4-9a
1 1 /{a 1
Vo(s) = J<Vi (s) = J<-CsR + Cs
-X(s) = - -
s+a
a=-
RC
158
Therefore Ka
H(s)=-
s+a
Similarly for the circuit in Figure P4.4-9b, we can show (see Figure S4.4-9)
H(s ) = Ks
s+a
X<s) rl t
v(~) + kVts.)
0 \J1ts>. I
c.s
Ch)
Figure 84.4-9
4.4-10. Figure S4.4-10 shows the transformed circuit. The op amp input voltage is V,,(s) ~ 0.
The loop equations are
6
I1(s) + (-$ + l)[Ii(s) - / 2 (s)J X(s)
3 6 3
--h(s)
s
+ (-$ + -2 )[Ii(s) - h(s)J 0
-s
Y(s) = - -2l [Ii{s) - I 2 (s)] = X(s)
s2 + 8 s+ 12
The transfer function -s
H(s)- - - --
- s2 + 8s + 12
3
5
-t .
.!..
2..
yls) Y(~)
'fr-~) )'.
Figure 84.4-10
6s 2 + 3s + 10
s(2s2 + 6s + 5)
lim sY(s) = 3
s- oo
159
y (oo) = Jim sY(s)
s-o
=2
(ii)
6s 2 + 3s+ 10
Y(s)
(s + 1)(2s 2 + 6s + 5)
y(O+) Jim sY(s) = 3
•-oo
y(oo) = lim sY(s) = 0
..-o
(b) (i)
2
Y(s) = s + 5s + 6
s2 + 3s + 2
This Y(s) is not strictly proper . We can express it as
2s +4
Y (s) = 1 + 2 3 2
s + s+
Hence y(o+) =
s(2s + 4) = 2
Jim
+ 3s + 2
..-oo s 2
and s3 + 5s2 + 6s
y(oo) = Jim sY(s) = lim 2 = 0
s-0 s-0 S + 3 S + 2
(ii)
s 3 + 4s 2 + lOs + 7
Y(s)- - - - - -- -
s2 + 2s + 3
y
(oo) = Jim s ( s3+ 4s2+10s +
.• -o s2 + 2s + 3
7) = 0
4.5-1. (a) At first glance, we are tempted to answer the question in affi rmative. Lei us
verify the reality.
(b) The loop equations are
160
_,_
~ .. '
- - -- 'fM
....
-J£0A-
- 2I1 + 412 = 0
411 - 2h X(s)
-2/1 + 4000212 0
In this case H(s) is very close to 1/ 4. This is because the second ladder section
=
causes a negligible load on the first. Let R 3 R.i = R. In this case, as R --+ oo,
we observe that H (s) --+ 1/ 4. The second ladder causes no loading in this case.
The Cascade rule applies only when the successive subsystems do not load the
preceding subsystems.
4.5-2. The transfer function of the two paths are e-st and ae-s(r+,.-). T he two paths a re in
parallel. Hence the transfer function of this communication channel is
H(s) e-sT + ae-•(T+r)
= e-sT(l + ae-sr)
®
or Figure 4.18d, it is immediately obvious that such an equalizer can be realized by
the following system
~ - /' >
LLD~'·>- rr J Figure S4.5-2
J
\Vhen this equalizer is placed in cascade with the communication channel, the effect ive
transfer function is given by
The effective system represents a pure delay ofT seconds, which makes it distortionless.
iv1oreovcr, the equalizer is realizable.
161
,.
~ · } -c:;;:eyies
. ~)·
- .
,_ _ ___._....,>--\ L.
-3/2...
~ 9/3
Figure S4.6-1
/6 f !r
4.5-3. (a) The system transfer function is
1
1
H(s) = s=12
l +s-1
- s+l
(i) We can verify that for K = 10, all the roots are in LHP and hence the system
is BIBO stable.
(ii) For K = 50, we can verify that two roots are in RHP and on LHP. Hence
the system is BIBO stable.
(iii) For K = 48, we verify that two roots are on imaginary axis at ±j J8 and one
is in the LHP. Hence the system is BIBO unstable (but marginally stable).
4.6-1.
H s
( )
= s +
2
2s
s3 + 8s 2 + 19s + 12
= (-s-)
s
(ss + 2) (-1-)
+ 1 s+
-1/6
s
+3
_s3/2+ + s +4
=
+ 1 3
8/3
4
4.6-2. The transposed version of the realizations for the transfer function in Prob. 4.6-1 are
shown in Figure S4.6-2.
4.6-3. (a)
3s(s + 2) 3s 2 + 6s
H(s) = + + 2s + 2)
(s l)(s 2 s + 3s 2 + 4s + 2
3
3s ) ( s+2 ) 3 6s +6
( s+ 1 s 2 + 2s + 2 = - s + 1 + s2 + 2s + 2
For the canonical form, we have a3 = 2, a 2 = 4, a1 = 3, and b3 =
0, b2 = 6, b1 = 3, bo = 0. Figure S4.6-3a shows a canonical, cascade and parallel
realizations.
(b)
2s - 4 2s - 4
H(s)
(s + 2)(s2 + 4) s3 + 2s2 + 4s + 8
2(s-2) (s-2) ( 2 ) 1 s
s3 + 2s2 + 4s + 8 = s + 2 s2 +4 = - s +2+ s2 +4
For a canonical forms, we have a 3 = 8, a 2 = 4, a 1 = 2, and b3 = - 4, b2 =
2, b1 = 0, bo = 0. Figure S4.6-3b shows a canonical, cascade and parallel realiza-
tions.
162
~~>;_- rf1J-"' '-'':~·-' .
L _,
'.l/2.
~Jj
u
'
.
i/?L..
)'.<•)
4.6-4. The transposed version of the realizations for the transfer function in Prob. 4.6-3 are
shown in Figure 84.6-4.
4.6-5.
J(_Ji
F
r·
...
;.·
i ··
i-'
r···
"
.:•
Y(s)
se.r1'e5
. .. . ..· .. ~ ·~--· .. ·- .
. . ·· .
.!
·j \=!~ s 4· 6-5
..~
I
;
·. .
;<t.5)
·;
... _ . ._.··_. ,.;_.. . .
ytt;:,)
o·"I
'
SP f P..S
o · C,
X(s) 1/:w
if
..·
Vs I/iv
- " ' " - ---+->--< 2-.J----'E:,----.
'' -3 - I
~:. t
·. 4
. .·
...
'
4.6-7.
H
5
_ s(s+l)(s+ .2) 3
s + 3s + 2s
2
=1_ ~+~ _~
( ) - (s + 5)(s + 6)(s + 8) s3 + 19s 2 + 118s + 240 s + 5 s+ 6 s+8
For a canonical form a 3 = 24, a 2 = 118, a 1 = 19, and b3 = 0, b-i = 2, b1 =
3, b0 = l. Figure 84.6-7 shows a canonical, cascade and parallel realizations.
4.6-8. The transposed version of the realizations for the transfer function in Prob. 4.6-7 are
shown in Figure 84.6-8.
4.6-9.
SJ s3
H(s) =
(s + l) 2 (s + 2)(s + 3) s 4 + 7s3 + 17s 2 + 17s + 6
27
= s ) ( s ) ( s ) ( 1 )
( s+1 s+ 1 s+2 s+ 3 = - s + 2 + s ~ 3 + s ~l! 1 ·-
8 l
(s: 1) 2
164
.;
J
. {'
.;
(
I
/ c.Cl.' V\ot'I k.
Jit-rtt i-
·; ·
.·...
·.•. ·~
.: .
xcs)
Yr.s)
r
..
:.:
..
..
.,
Yt.s ">
J
.,
"j.(s)
1
R'j · s 4-· ~-/ ·I
~:
·:~
·:
-I
.;<.(s)
. '
:· i
4.6-13. Application of Eq. (4.59) to Figure P4.6-13a yields
I l
H1(s) = ] ~bl, = -----
(s + a)2 + b2
+ (s+a) 2
Figure P4.6-13b is also a feedback loop with forward gain G(s) = s~a and the loop
• b?
gain (s+a)i . Therefore
-L s+a
H2(s) = s+~, = - - - 2 - -2
1+ ~ (s + a) +b
The output in Figure P4 .6-13c is the same of B - aA t..imes the output of Figure
P4.6-13a and A times the output of Figure P4.6-13b. Therefore its transfer function
is
4.6-14. These transfer functi ons are readily realized by using the arrangement in Figure 4.28
by a proper choice of Z1 (s) and Z(s). ..
(i) In Figure S4.6-14a
l
z,(s)
R1 +-L c,.
Z(s) = R
165
(.a)
(b '>
IOJf
·.
·.5K
4.6-15. One realization is given in Figure S4.6-11c. For the other realization, we express H(s)
as s +2 3
H (s) =- - =1 -
s + :>
-
s+5
166
IOK
Figure S4.6-/&
167
F (s)
(I>)
100 IUl
(b) The transfer function of this system is
We use the same argument as in part (a) to deduce that the de gain is 0.1 and
t.he 3-dB bandwidth is lOwc. Hence, the gain-bandwidth product is We.
(c) The transfer function of this system is
( ) G(s) _ ~ _ We
H s = 1 - G(s)H(s) - l - o.9w ,. - s + O.lw
s+wc c
We use the same argument as in part (a) to deduce that the de gain is 10 and
the 3-dl3 bandwidth is 0.lwc. Hence, the gain-bandwidth product is we.
( d) Included in previous parts.
4.8-1.
jw + 2 jw+2
H(jw)
(jw)2 + 5jw + 4 (4-w 2 ) + j5w
IH(jw)I
w2 +4 2
w +4V
(4-w 2 ) 2 + (5w) =w +17w 2 +16
2 4
w 5w
LH (jw) = tan - 1 ( )- tan - 1 ( _ w 2 )
2 4
(a) x(t) = 5 cos(2t + 30°). Here w = 2 and
IH(j2)1
LH(jw)
!is=v;
tan - 1 - tan- 1 (oo) = 45° - 90° = -45°
168
y(t) = 5
V2 cos(2t + 30° - 45°) = V2 cos(2t - 15°)
5
(b) :c(t) = 10 sin(2t + 45°)
Therefore
4.8-2.
jw+3
H(jw)
(jw + 2) 2
Jw2 +9 w w
LH(jw) = tan- 1 ( J) - 2)
1
IH(jw)I w2 + and tan- (
4
(a) x(t) = lOu(t) = l0ei 0 tu(t). Here w = 0 and H(jO) = 1. Therefore
169
Therefore
y(t)
Ji3 cos(2t + 60° -
=B 56.31°)u(t) = -Ji3
- cos(2t + 3.69°)u(t)
8
(c) x(t) = sin(3t - 45°)u(t) Here w = 3 and
IH(j3)1 = Vl8
13
and LH(j3) = 45° - 112.62° = -67.62°
Therefore
y(t) = -JIB
3
.
sm(3t - 45° - 67.62°)u(t) = -
Vl8 sin(3t - 112.62°)u(t)
1 13
(d) x(t) = e1 3tu(t)
= fw2 + 100 = 1
IH(jw)I
v;;?+ 100
LH(jw) tan - l (- ~) - tan - l( ~) = - 2 tan - I ( ~)
10 10 10
(a) x(t) = eiwt
y(t) = li(jw)ejwt = IH(jw)le1fwt+L H(jw)f = e' fwt-2tan- 1
(w/10)f
170
y(t) = cos[wt + B- 2tan- 1 ( ~)]
IH(jl) I 1
LH(jw) -2tan -lei)
- = -11.42 0
10
y(t) cos(t - 11.42°)
IH(j2)1 1
LH(j2) -2tan- 1 ( 2) = -22.62°
10
y(t) sin(2t - 22.62°)
IH(jlO) I 1
LH(jIO) -2tan- 1(10)
- = - 900
10
y(t) cos(lOt - 90°) = sin lOt
IH(jlOO)I = 1
171
LH(jlOO) -2tan- I (100)- = - 168.580
. 10
y(t) = cos( l OOt - 168.58°)
4.8-4. (a) From t he graph, the two system zeros are at s = ±11.5. Thus, s 2 +bis+ b2 =
(s+1L5)(s-11.5)=s 2 +2.25. Thetwo system polesareats = - l ±J0.5. Thus,
s 2 + a 1 s + a 2 = (s + 1+J0.5)(s+1 - J0.5) = s 2 + 2s + 1.25. At DC, the system
function is H(JO) = -1 = k~ = k~:~~ = k~ . Therefore,
(b) The DC gain is given as S(JO) = - 1. Titus, the input of 4 just becomes - 4.
To compute the output to cos(t/2 + 11 /3) , Ji (J0.5) is required. Graphically,
IH(J0.5) 1'-- lkl (gi~) = 9 ~ and LH(J0.5) = rr -n/2 + n/2 - (0 -Hr/4) = 3rr/4.
Thus, Lhe output to cos(t/2 + n/3 is just ~ cos(t/2 + n/3 + 3rr/4). Thus, the
172
output to x(t) = 4 + cos(t/2 + 7r/3) is
= 2 .;:, (~ + 1) ( 16 + 1)
H ) JOO s(ffii + l) ~ s(ffii+l)
(s =2x 20 ( ~ + 1) ( 16 + 1)
The amplitude response: The horizontal axis where the asymptotes begin is
2.5, which is 7.96 db. We draw the asymptotes at w = l (20 dB/dee.), 2 (-20
dB/dee.), 20 (-20 dB/dee.), and 100 (20 dB/dee.) as shown in Figure S4.9-la.
T he corrections are applied at various points as discussed in Examples 4.25 and
4.26. to obtain the Bode plot for amplitude response. We follow the similar
procedure for phase response.
(b) The transfer function can be expressed as
The amplitude response: The horizontal axis where the asymptotes begin is 2,
which is 6 db. Asymptotes start at w = 1 (-40 dB/dee.) , 10 (20 dB/dee.), 20 (20
dB/dee.), and 100 (-20 dB/dee.). The corrections are applied at various points
as discussed in Examples 4.25 and 4.26. to obtain the Bode plot.
(c) The transfer function can be expressed as
The amplitude response: The horizontal axis where the asymptotes begin is
173
-r~ . f<r.q-1
'''P-
;..,·_ . . :~-··
. ..·...:-~~. .- . · ,r ::: .:
··~--
. ...
. -J-·-
. . . .
-.
.
~ . ~ ~ ·~ ~
l~~ ~ ~
J'.~ ~
~ ~
ca..)
(b)
r~ •.
·1 1o'" , ••
1 s2
H(s) -
- 1 6(y + l)(~+~+l)
T he amplitude response: The horizontal axis where the asymptotes begin is 1/ 16,
which is -24 dB. Asymptotes start at w = 1 (40 dB/ dee.), 1 (-20 dB/dee.), 4 (-40
dB/ dee.). The corrections are applied at various points as discussed in Examples
4.25 and '1 .26. to obtain the Bode plot.
174
(b) The transfer function can be expressed as
1 s
H(s) - - - - - - - - - -
s;
- 100 (f + 1)( 1 0 + 0.1414s + 1)
The amplitude response: The horizontal axis where the asymptotes begin is
1/100, which is -40 dB . Asymptotes start at w = 1 (20 dB/dee.), 1 (-20 dB/dee.),
10 (-40 dB/dee.). The corrections are applied at various points as discussed in
Examples 4.25 and 4.26. to obtain the Bode plot.
(c) The 'transfer function can be expressed as
10 fo+l
s = -
H () --,2~~----
100 s( i"oo + 0.1414s + 1)
The amplitude response: The horizontal-axis where the asymptotes begin is 1/10,
which is -20 dB. Asymptotes start at w = 1 (-20 dB/dee.), 10 (20 dB/dee.), 10
(-40 dB/dee.) . The corrections are applied at various points as discussed in
Examples 4.25 and 4.26. to obtain the Bode plot.
4.9-3. To rise at 20dB per decade, a zero must be present before w = 0.L At w = 30, the
magnitude response begins to fall at -20dB per decade. This requires two poles at
that frequency: one pole to counteract the previous zero and another pole to cause
the -20dB per decade slope. The magnitude response level out at w = 500, which
requires the action of a zero. Thus, a second order system should be sufficient.
175
4.. ~.1.rHs ~ __ ~ _: _: : ...: ·:.
CD
"O -50
.
. . .
•
. . .:
~ · 74
-.
.
··· ·......,.
-:
. .
'
.
..
. .
-: -
..
.
.
.-: .
.
.
-. :-
.
.
c:
iii
. . . .
.. . . . . ' . .
<.:l ... j
-100
J
I
I
-1501
10·2 100
Frequency (rad/sec)
-180~
I
ci9l-)
10·2 100 10'
Frequency (rad/sec)
. . . ...
. ..:.. . ·-· ;.
; ~ - - .. - - .- • -.- • • ""'l'
-100 1 1
10· 10° 10
Fr9QUency (rad/sec)
180
t
~
.2
0.
0
(_ b ~
-180
10· 1 10° 10
1
Frequency(rad/sec)
50
CD
"O
c:
~ -2c
-50
-100 1 1
10· 10° 10
Frequency (rad/sec)
l II>
2"'
-90
0.
c~)
-180
(b) Notice, H( s) has two poles, one a t zero and another at a negative, real number.
It also has two zeros, one a t infinity and another at a negative, real number. Only
plots B and D show evidence of a finite zero as well as a finite pole. Of these,
only plot B can have the necessary pole at zero. Thus,
(c) At low frequencies, H(JW);:::; (C1+~2)JW. Thus, n doesn't affect IH (;w) at very
low frequencies .
(d) At high frequencies, H(Jw);:::; -Rng:_11t:,w2 = ~· Thus, R doesn't affect IH(3w)
177
- lf-J7 1-t J 7 I ;+{ .l i)\
- 1.
rt<!. 1.t.10- 1
4.10-1. V,Te plot the poles - l ±j7 and l±j7 in the s-plane. To find response at some frequency
w, we connect a ll the poles and zeros to the point jw as shown in Figure S4.10-l. Note
that the product of distances from the zeros is equal to the product of the distances
from the poles for all values of w. Therefore IH (jw)I = 1. Graphical argument shows
that LH (jw) (sum of the angles from the zeros- sum of the angles from poles) starts
at zero for w = 0 and then reduces continuously {becomes negative) as w increases.
As w-> oo, LH (w)-+ -27T.
4.10-2. (a) If rand dare the distances of the zero and pole, respectively from jw, t hen the
amplitude response IH (jw )I is t he ratio r/d corresponding to jw. This ratio is
0.5 for w = 0. Therefore, the d e gain is 0.5. Also the ratio r / d = 1 for w = oo.
Thus, the gain is unity at w = oo. Also the angles of the line segments connecting
the zero and pole to the point jw are both zero for w = 0, and are both 7r / 2 for
w = oo. Therefore LH (jw) = 0 at w = 0 and w - oo. In between the angle is
positive as shown in Figure S4 .10-2a.
(b) In this case the ratio r/d is 2 for w = 0. Therefore, the de gain is 2. Also the
ratio r/d = 1 for w = oo. Thus, t he gain is unit,y at w = oo. Also the angles
178
2 , tf <}A>) I
1 I tt (ju)) I
.i.
0 1
5 ..( >f( )u)) 0
. "'
<. v~
(b)
l~)
Figure $4.10-2.
of the line segments connecting the zero and pole to t.he point jw are both zero
for w = 0, and are both r. /2 for w = oo. Therefore LH (jw) = 0 at w = 0 and
w = oo. In between the angle is negative as shown in Figure S4.10-2b.
4.10-3. The poles are at -a ± jlO. Moreover zero gain at w = 0 and w = oo requires that
there be a single zero at s = 0. This clearly causes the gain to be zero at. w = 0. Also
because there is one excess pole over zero, the gain for large values of w is 1/w, which
approaches 0 as w --+ oo. therefore, the suitable transfer function is
H( s) = s = s
(s +a+ j lO)(s +a - jlO) s 2 + 2as + (100 + a2)
The amplitude response is high in the Yicinity of w = 10 provided a is small. Sma.ller
the a, more pronounced the gain in the vicinity of w = 10. For a = 0, the gain at
w = 10 is oo.
4.10-4. Cynthia is correct. Although the system is all-pass and has IH(JW)I = 1, the phase
response is not zero. Thus, the out.put generally has different. phase than the input.
Furthermore, the output can also include transient components that would not be
present in the original input.
4.10-5. Both Amy and Jeff are correct. By definition, a zero is any values thal forces H(s) = 0
179
and a pole is any value s that forces H( s) = oo. Thus, the system H(s ) = s = ;l:-r
has both a zero at s = 0 and a pole at s = oo. Remember, a rational system function
always has the same number of poles and zeros; if H(s) = s has an obvious zero at
s = 0 there must be a matching pole somewhere, even if it is not finite. By similar
argument, the system H(s) = ~ has a pole at s = 0 and a zero at s = oo.
4.10-6. At high frequencies, the highest powers of s dominate both the numerator and denom-
bns"
inator of H(s). That is, lims-oo H(s) = lim$-oo -;rv.
Lowpass and bandpass filters both require lims-oo H(s) = 0, which ensures a response
of zero at high frequencies. Only H( s) that are strictly proper (M < N) yield the
required lims-oo H(s) = 0.
Highpass and bandstop filters both require lim_,_ 00 H(s) = k, where k is some fi-
nite, non-zero constant. Only H( s) that arc proper (M = N) yield the required
lirns-oo H(s) = bo = k.
The case M > N is not considered, since such systems are not physically practical.
4.10-7. At high frequencies, the highest powers of s dominate both the numerator and de-
nominator of H(s). That is, lims-oo H (s) = lims-oo ~ 5~v . Thus, the log magnitude
response at high frequencies is given by lirn..,_ 00 log IH(JW)I = log(bo) + At log(w) -
N log(w). The fastest attenuation as a function of frequency requires M to be as
small as possible. Thus, for a given N, the attenuation rate of an all-pole lowpass
filter (M = 0) is faster than t he attenuation rate of any filter with a finite number of
zeros (M ;/; 0). .
4.10-8. No, it is not possible for such a system to function as a lowpass filter . For any choice
of (lk, b1 , ~, a 1 , a 2 ] E n), the system function H(s) = k ;::!:;!~; is proper. Thus, the
system function always has high-frequency gain of k. For k=/= 0, the system cannot be
lowpass. Furthermore, for k ::: 0 the system becomes a useless "nopa.ss filter" (again,
not lowpass) .
4.10-9. Nick is more correct than his professor. A cascade of two identical filters, each with
system response ll(Jw), gives a total response of H 2 (JW). Since realizable filters, such
as Butterworth filters, are not ideal, the cascade system will tend to have a faster
transition band and greater stopband attenuation. In a sense, the resulting fourth-
order system really does provide "twice the filtering" of the original second-order
system .
Unfortunately, there are also problems with Nick's approach. Simply cascading a
designed lowpass filter twice has negative consequences. For example, the cutoff fre-
quency shifts to a lower frequency than desired. As the cascaded RC example in
180
MATLAB Session 4 suggests, a cascade of low-order filters is inferior to a carefully de-
signed, equivalent-order filter. In general, a fourth-order Butterworth filter performs
better than a cascade of two second-order Butterworth filters .
4.10-10. (a) Using tables,
1 1 es/2 - e-s/2
H(s) = - - - e-s = e-s/ 2 - - - - -
s s s
Substituting s = jw yields the frequency response
. )
H(JW = e _.;w/2
,
ejw/2 _
e -jw/2
= e - jw/2 sm w
· ( /2)
= e
-jw/2 .
smc -
(w )
~ 02 2
The sine type of frequency response (with a linear phase shift of -w/2) represents
a lowpass system .
(b) Since h(t) is finite duration, the system has no finite poles. There are, however,
an infinite number of finite zeros for sinc(w/2) at w = 2-rrk ors = j27rk, where k
is any non-zero integer.
(c) In transform-domain, the inverse system is given by the reciprocal of Hc(s). Thus,
The inverse system has no finite zeros and an infinite number of finite poles.
Since poles lie on thew-axis, the inverse system cannot be asymptotically stable.
The same approach does not work in the time-domain. That is, h"; 1 ( t) -:/= hlt).
T he impulse response needs to be obtained from an inverse Laplace transform
of H; 1 (s) . Unfortunately, it is difficult to take the inverse Laplace transform of
Hc- 1 (s) ; no closed form solution for h~ 1 (t) is known to exist .
It is possible to approximate h; 1 (t). Consider the following idea. Replace the
denominator sin (fe) with a truncated Taylor series expansion. The result is
a rational approximation to H; 1 (s) that can be inverted using partial fraction
expansion techniques. Although not perfect, the result can perform reasonably
for many low-frequency inputs.
4.10-11. No, the suggested lowpass to highpass t ransformation HHp(s) = 1 - HLp(s) does
not work in general. Although it is possible to relate the ideal magnitude responses
according to IHHp(Jw) I = 1 - IHLp(Jw)I. the phase information contained in H(s)
generally makes HHp(s) -:/= 1 - HLp(s).
As an example, consider an ideal lowpass filter described by
4.10-12. (a) Yes, it is possible for the system to output y(t) = sin(lOOrrt)u(t) in response
to x(l) = cos(l007rt)u(t). Noting Y(s) = 5 2)(~~~"")2 and X( s) = 5 2 +(~oo,.-)Z •
181
one way to obtain y(t) from x(t) is using the system H(s) = Y(s)/X(s) =
10071' s 2 +{100?r) 2 _ lOOir
s'+{l001T)2 s - s
(b) Yes, it is possible for the system to output y(t) = sin(1007rt)u(t) in response
to x(t) = sin(507rt)u(t) . Noting Y(s) = .2)(~~~1Tl2 and X(s) = 82+(;01r)2,
one way to obtain y(t) from x(t) is using the system H(s) = Y(s)/ X(s) =
lOOrr s 2 +(507r) 2 _ 100?r s 2 +{50>r) 2 )
s2+(100n-)2 s - s(s +(lOO'lf) ·
(c) Yes, it is possible for the system to output y(t) = sin(l007rt) in response to
x(t) = cos(lOOr.t). To do this, the system must have H(;l007r) = e-J"/ 2 . That
is, the magnitude response at w = 100?T must be unity, and the phase response
at w = 1007T must be - ?T /2.
(d) No, it is not possible for the system to output y(t) = sin(l007rt) in response to
x(t) = sin(507rt). In an LTI system, an everlasting sinusoidal input of frequency
50r. cannot produce a different frequency output.
4.11-1. (a) Let x 1 (t) = x(t)u(t) = etu(t) and x 2 (t) = x (t) u(-t) = u(-t) . Then X 1 (s) has a
region of convergence a > l. And X 2 (s) has a region a < 0. Hence there is no
common region of convergence for X( s) = X 1 (s) + X 2 (s).
(b) x 1 (t) = e-tu(t), and X 1(s) = 1
.!
converges for a> -1. Also x2(l) = u(-t),
and X2(s) = -~ converges for a < 0. Therefore, the strip of convergence is
-1 <a< 0
Figure S4.ll-lb
(c)
1-e- st __, 0 }
as t _, oo if Re s 2: 0
i2+1 as t --> -oo if Re s ::;; 0
1
x(t) - - -
- 1 + et
_1_ e - •t __, 0 }
as t --> oo if Re s > -1
l+e' as t ....., -oo if Re s < 0
182
(e)
x(t) = e-kt,
as ·t -+ oo for any value of s
as t -+ -oo for any value of s
Hence the region of convergence is the entire s-plane.
4.11- 2. (a)
1
and X2( -s) = s +
1
1
and
X2(s) = - - <7 < 1
-s+ 1
1 1 -2
Hence: X(s) = X1 (s) + X2(s) = - - + - - = ~l - l <u< l
s + 1 -s + 1 s -
(b)
x(t) = e-ftl cost= e-t cost u(t) +et cost u(-t) = x 1 (t) + x2(t)
s+l s+ l
Hence X 1 (s) = (s + l )2 + 1 and X2(- s) = (s l )2 +1 <7 <l
1
s+l s- l 4 - 2s 2
X (s) = X1(s) + X2(s) = (s + l ) 2 +
1
-
(s - 1)2 + l
= - -
s4 - 4
- l < u<l
(c)
1 1
:c(t) = e1u(t) + e2 tu( - t); X 1 (s) = - -
s- 1
u >1 and X2( -s ) = --
s+2
l
X2(s) =- -·
-s+2
u < 2.
- 1
Hence X(s) = X1( s) + X2(s) = (s _ l)(s _ Z) l <u <2
(d)
e- t for t >0
x(t) = e-tu(t) ={ 1
for t<0
183
(c)
1
X 1 (s)=- a>O
s
1 1
X2(-s) = -s+ 1 X2(s) =- -
- s +1
<7 <1
l 1 -1
and hence: X(s) = s- s - 1 = s(s - l) O<a<l
(f)
O<a<l
4.11-3. (a)
2s + 5
X(s) = - 3 <a< - 2
(s + 2)(s + 3)
1 1
= --+-- - 3 <a< -2
s+2 s+3
The pole -2 lies to the right, and the pole -3 lies to the left of the region of
·convergence; hence the first term represents causal and the second term represents
anticausal signal:
(b)
2s - 5
X(s) 2<a<3
(s - 2)(s - 3)
1 1
- - + --
s-2 s-3
2<a<3
The pole at -2 lies to the left and t,ltat at 3 lies to the right of the region of
convergence; hence
x(l) = e2 tu(l) - e31 u( - t)
(c)
2s+3
X(s) <7 > -1
(s + l)(s + 2)
1 l
= - - +s+2
s-1 1
-- u > -1
184
Both poles lie to the left of the region of convergence, and
(d)
2s+3
X(s) = (s + l)(s + 2)
<1 <-2
1 1
- -
s+ l
+ --
s+2
a< -2
Both poles lie to the right of the region of convergence, and hence:
3s2 - 2s - 17
X(s) - l <a<5
(s + l)(s + 3)(s - 5)
1 1 1
= -
s+l
- +s+ --3 +-- s-5
The poles - 1 and -3 lie to the left of the region of convergence, whereas the pole
5 lies to the right:
4.11-4.
2
-(s +-2sl )(s
--- 2s - 6
- - - = -s +1-l - -
- l)(s + 2)
l
-+-
s- l
2
-
s +2
(a) Re s > 1: All poles to the left of the region of convergence. Therefore
x(t) = (e-t - e1 + 2e- 2 t)u(t)
(b) Re s < - 2: AII poles to the right of the region of convergence. Therefore
(c) - 1<Res<1: Poles - 1 and -2 to the left. and pole 1 to the right of the region
of convergence. Therefore
1•1 1
x(t) = e-2 , H(s) =- - a> - 1
s+l
1 1 1 l
And X (s)=-- - - - --<a <-
s + 0.5 s -·- 0.5 2 2
185
l - - -l - ]
1 l
hence: Ys -H sXs - -l- [ - - -- < u<-
( )- ( ) ( ) - s + 1 s + 0.5 s - 0.5 2 2
(b)
1 1
X(s) l <u<2
s-1 s-2
- 1
(s - l) (s - 2)
1
And H(s) = - - u > - 1
s+l
- 1
Hence: Y (s) = H ( s) X(s ) = (s + l)(s - l )(s - 2) 1 < (J < 2
1 1 1 1
X(s) = s +0.5 - s+0.25 --< a <-
(s + 0.S)(s + 0.25) 2 4
1
Also H(s)= - (j >- 1
s+l
186
(d)
1
X1(s) (1 >2
s-2
-1
X2(s) = (1 < 1
s-1
1
and H(s)= - (1 > -1
s+l
In this case, there is no region of convergence that is common to X 1 (s) and X 2 (s).
However, each of X 1 (s) and X 2 (s) have a region of convergence that is common
to H(s). Hence the output can be computed by finding the system response to
x 1 (t) and x 2 (t) separately, and then adding these two components. Thjs means
we need not worry about the common region of convergence for X 1 (s) and X2(s).
Thus:
where
1
Y1(s) = X1(s)H(s) = (1 >2
(s+ l )(s-2)
I 1
-3 + ___l_ >2
= s+l s-2
(1
Observe that both the poles ( -1 and 2) are to the left of the region of convergence,
hence both terms are causal, and:
- 1
- l<u<l
(s + l)(s - 1)
l !
= _..L._ _ _2_
-l<a< l
s+l s-1
The poles -1 and 1 are to the left and the right, respectively, of the strip of
convergence. Hence the first term yields causal signal and the second yields
anticausal signal. Hence
187
l 1
where X1(s) a> --
s + 0.25 4
-1 1
X2(s) = (! < --2
s+0.5
1
H (s) (! > -1
s+l
Here also, we have no common region of convergence, for X 1 (s) and X2(s) as in
part d. Let Y(s} = Y1 (s) + Y2 (s) where:
1 l
(! >--
(s + l}(s + 0.25) 4
4 4
- 3 + 3 l
(! > - -4
s+1 s + 0.25
-1 1
-1 <a< - -
(s + l)(s + 0.5) 2
2 2 1
-1 <a< - -
s+1 s + 0.5 2
l
where X1(s) a> - 3
s+3
- 1
X2(s) a<- 2
s+ 2
1
H(s) (! > -1
s+ l
In this case, there is a common region of convergence for X 1 (s) and H(s), but
there is no region of convergence common to X 2 (s) and H(s). Hence the output
y 1 (t) will b e fini te but Y2(t) will be oo.
4.11-6.
188
4.11-7. For u < 0, we know that c.- 1 [~] = -2u(-t). Additionally, 1:.- 1 (J/2] = 6/2. Using
properties, c.- 1 [s(I/2)] = ft (6(t)/2). Thus,
x(t) = -2u(-t) + :t (6(t)/2).
The function ft
(6(t)/2) is called the "unit doublet". Like 6(t), the unit doublet is
not a physically realizable signal. It is a mathematical constructions that is useful,
among other things, in finding function derivatives. Refer to the t,opic of generalized
derivatives.
4. 11-8. (a) Yes, x(t) can be left-sided. To be left-sided and absolutely integrable, the signal's
region of convergence must: 1) be left-sided, 2) include thew-axis, and 3) not
include any poles. W ith a pole at s = 7T, it is possible to achieve all three
necessary conditions. For example, x(t) = e" 1u(-t) has a pole at s = 7T, is
absolutely integrable, and is left-sided.
(b) No, x(t) cannot be right-sided. To be right-sided and absolutely integrable, the
signal's region of convergence must: 1) be right-sided, 2) include thew-axis, and
3) not include any poles. T hew-axis cannot be included in a region of convergence
that is to the right of the known pole at s = 7T.
(c) Yes, x(t) can be two-sided. To be two-sided and absolutely integrable, the signal
must: 1) have at least one pole in the right-half plane, 2) have at least one
pole in the left-half plane, and 3) have a region of convergence that includes
the w-axis. With a pole at s = ?T, these conditions are possible. For example,
x(t) = e"tu( - t) I- e- "tu(t) has a pole at s = 7T (and another at s = - 7T), is
absolutely integrable, and is two-sided.
(d) No, x(t) cannot be finite duration. To be finite duration, the signal's region of
convergence must include all finite values of s. However, since a pole is present
at s = 7T, this point cannot be included in the region of convergence. Thought,
of another way, a pole at s = 7T implies a signal component of either e"t u(t) or
e"t·u(- t) , both of which are infinite in duration.
4.11-9. (a) X 1 (s) = J::'00
x 1 (t)e-stdt = J::' 00
(1 + e1t)u(t)e-"'tdt = J0
00
+ e1<1-s)dt =
JC-st
( ::2se-st + e•;~-,;>) [ = ::2;(0 - e0) + 0;:.e_, 0
. For a > 0, this _simplifies to X 1(s) .
0
Thus,
X 1 (s) = -J + -1- for a> 0.
s s- J
-JS
X2(s) = -?-- for o < -1.
s- - 1
189
For
1
X 4 (s) = e-•" - for <7 < 0.
. S-J7r 12
4.11-10. {a) To be bounded amplitude, the region of convergence must include thew-axis.
The transfer function has two poles, at s = ±1, that musl be excluded from the
region of convergence. Thus, the region of convergence must be
-l<O"<L
(.M.. l!&.).
(b) Rewrite H(s) as 2• (s-1)(s ..+l) = e1n(s) s-1 + $+1 Using - 1 < a < 1 ' the time-
shifting property, and a table of Laplace transform pairs, the inverse transform
is found to be
h(t) = 0.5e-(t+lnC 2 »u(t + ln(2)} - 0.5et+ lnC 2 >u(- (t + ln(2)}).
» MS4P3(20)
ans = 524288 0 -2621440 0 5570560
0 -6553600 0 4659200 0
- 2050048 0 549120 0 -84480
0 6600 0 - 200 0
1
Thus,
4.M-2. (a) MATLAB is used for the design. To evaluate filter performance, the magnitude
response is plotted over the frequency range (0 $ f $ lOkHz).
>> N = 12; omega_c = 2•pi• 5000;
>>poles= roots([(j*omega_c)-(-2•N),zeros(l,2•N-1),1]);
>> B_poles = poles(find(real(poles)<O));
>> subplot(121), plot(real(B_poles), imag(B_pol es) , 'xk ');
>> xlabel('Real'); ylabel('Imag');
>> axis([-4e4 0 -4e4 4e4]); axis equal;
>>Ac poly(B_poles); A= A/A(end); B • 1;
>> f = linspace(0,10000,1001);
>> Hmag_B = abs(polyval(B,j•2•pi*f)./polyval(A,j*2*pi *f));
>> subplot(122), plot(f,Hmag_B,'k');
>> xlabel('f [Hz]'); ylabel('IH_{LP}(j2\pi f)I');
190
(!!'- 0.8
I o ~
i 0.6
•1
-2 ...
-3
-•"--~-~-~~~
-4 -3 -2 0
Roal 1110'
l(Hil
The resulting figures are consistent with a Butter.vorth design; the poles lie on
a semicircle in the left-half s-plane, and the magnitude response exhibits smooth
monotonic roll-off.
(b) Modifying program MS4P2, the Sallen-Key component values and magnitude re-
sponse plots are easily found.
>> omega_O = 5000*2*pi; f = linspace(0,10000,200);
>>psi= (7.5:15:90]*pi/180; Hmag_SK = zeros(6,200);
>> f or stage= 1:6,
>> Q = 1/(2*cos(psi(stage)));
>> disp(('Stage ',num2str(stage ), .. .
>> ' (Q = ', num2str(Q), .. .
>> '): Rl = R2 = ',num2str(100000)]);
>> disp([' Cl= ',num2str(2*Q/(omega_0*100000)), ...
>> ', C2 = ', num2str(l/(2*Q*omega_0*100000))));
>> B = omega_o-2; A [1 omega_O/Q omega_o-2);
>> Hmag_SK(stage,:) = abs(polyval(B,j*2*pi*f) ./polyval(A,j *2*pi*f));
>> end
» plot Cf ,Hmag_SK, 'k' ,f ,prod(Hmag_SK), 'k: ')
>> xlabel ('f [Hz)'); ylabel('Magnitude Responses')
St age 1 (Q 0.50431): R1 = R2 = 100000
Cl 3.2106e-010, C2 = 3.1559e- 010
Stage 2 (Q 0.5412): Rl = R2 = 100000
Ci 3.4454e-010, C2 = 2.9408e-010
Stage 3 (Q 0.63024): Rl = R2 = 100000
Cl 4.0122e-010, C2 = 2.5253e-010
Stage 4 (Q 0.82134): Rl = R2 = 100000
Ci 5.2288e-010, C2 = 1.9377e-010
Stage 5 (Q 1. 3066) : Rl = R2 = 100000
Cl 8.3178e-010, C2 = l.2181e-010
Stage 6 co 3.8306): R1 = R2 = 100000
Cl 2.4387e-009, C2 = 4.1548e-011
The resulting resistor and capacitor values are realistic. Each Sallen-Key stage
implements a complex-coujugate pair of poles. The flattest magnitude response
corresponds to the pair of poles that are furthest from the w-axis, or Stage l.
191
O0 1000 2000 3000 "OOCI 5000 6000 7000 IOOO IOOO 10000
l (H"tl
The most peaked magnitude response corresponds to the pair of poles that are
closest to the w-axis, or Stage 6. The remaining stages are ordered in between.
The dashed curve is the total magnitude response, and it is exactly the same as
t.he one shown in Figure S4.M-2a.
4.M-3. {a) MATLAB is used for the design. To evaluate filter performance, t.he magnitude
response is plotted over the frequency range (O $ f $ lOkHz).
>> omega_c = 2• pi•5000; R = 3; N = 12;
>>epsilon = sqrt(10"(R/10)-1);
>> k = [l:N); xi= 1/N•asinh(l/epsilon); phi= (k• 2-1)/(2•N)•pi;
>> C_poles = omega_c• (-sinh(xi)•sin(phi)+j•cosh(xi)•cos(phi));
>> subplot(121), plot(real(C_poles),imag(C_polcs),'xk');
>> xlabel('Real') ; ylabel ('Imag');
>> axis((-4e4 0 -4e4 4e4]) ; axis equal;
>> A = poly(C_pol es) ;
>> B = A(end)/sqrt(l+eps ilon-2) ;
>>omega= linspace(0,2• pi •10000, 2001);
>> Hmag_C = abs (MS4P1(B,A,omega) );
>> subpl ot (122); plot(omega/2/pi,abs(Hmag_C),' k'); grid
>> xlabel (' f [Hz)'); ylabel('IH_{LP} ( j2\ pi f )I ' );
The resulting figures are consistent with a Chebyshev design; the poles lie on
an ellipse in the left-half s-plane, passband ripples are equal in height and never
exceed R = 3dB, there are a total of N = 12 maxima and minima int.he passband,
and the gain rapidly and monotonically decreases after the cutoff frequency of
le= 5kHz.
{b) Modifying program MS4P2, the Sa.lien-Key component values and magnitude re-
sponse plots are easily found.
>> omega_c = 2•pi•5000; R = 3; N = 12;
>>epsilon= sqrt(10- (R/10)-1);
>> k = [l:N]; xi = 1/N•asinh(1/epsilon); phi• (k•2-1)/(2• N) • pi;
>> C_poles = omega_c•(-sinh(xi)•sin(phi)+j•cosh(xi) • cos(phi));
>> C_poles = C_poles(f ind (imag(C_poles)>O)); !. Quadrant 2 poles
>> f = linspace(0 , 10000 , 501); Hmag_SK = zeros(6,501);
192
: OJ
·~·
0.7 ·"'
••
I 0 "'~ 0$
..
·•
•..
03
-~
0.2
·3
.. 0.1 · · ·· ·I··• ·!·· i " •
.. ----~--~~
-2 _, 0 2
R• tll 11 10•
. .___.__._......_~~
o 2000 • 001>
'l"'l
t<>OO
......
eooo ,0000
193
Figure S4.M-3b: Order-12 Chebyshev LPF Sallen-Key Stage Responses.
error occurs since the Salley-Key stages are constrained to unity gain at de, yet
the Chebyshev filter requires gain ,; 1 ~,2 at de. This error is easily corrected by
adding a gain stage to the circuit.
4.M-4. (a) Using MATLAB, the Sallen-Key component values are easily found .
>> omega_O = 4000*2*pi; NS = 4;
>>psi= [90/(2*NS):90/NS:90]*pi/180;
>> Q = 1./(2*cos(psi));
>> Rl = 1e9/omega_O*ones(1,NS); R2 = Rl;
>>Ci= 2*Q./(omega_O*R1) ; C2 = 1./(2*omega_O*Q.*R2);
>> for stage = 1:NS,
>> disp(['Stage ',num2str(stage), ...
» ' (Q = ',num2str(Q(stage)), ...
>> '): Rl = R2 = ',num2str(R1(stage))));
>> disp([' Cl= ',num2str(C1(stage)), ...
>> ', C2 = ',num2str(C2(stage))));
>> end
Stage 1 (Q 0.5098): R1 = R2 = 39788.7358
Cl 1.0196e- 009, C2 = 9 :8079e-010
Stage 2 (Q 0.60134) : Rl = R2 = 39788.7358
Cl 1.2027e-009, C2 = 8.3147e- 010
St age 3 (Q 0.89998): Rl = R2 = 39788.7358
Cl 1 . 8e- 009, C2 = 5.5557e-010
Stage 4 (Q 2.5629): Rl = R2 = 39788.7358
Cl 5.1258e-009, C2 = 1 . 9509e-010
The resulting resistor and capacitor values are realistic.
(b) The transformed Sallen-Key circuit is shown in Figure S4.M-4b. Name the node
between capacitors v(t). In t ransform domain, KCL at the positive terminal of
the op-amp yields
Y(s) - V(s) Y(s) - 0
1
sC~ R2
Solving for V(s) yields
V( s ) = Y( s ) 1 + R2C~s
R· ' C'
·2 2S
.
194
+
y(t)
_J_
Rearranging yields
1
Y(s) ~'~~~s
3
IC~s + J/ R'1 + C~s] = c;sX(s) + Y(s) [C~s + 1/ R;J.
2 2
Rearranging yields
(
Y( s ) 1 +R;C~s
R'2 C'2S
1 + R',Cis + R;C~s
R'l
_ I+ R;C~s] = X( ) IC' ]
R't s 1 s .
Y(s) s2
JJ(s) = X(s) = ( ) 1
s2 + s ,,}c,4
"'? 2
+ ni1c
2
1
l
+ R'1 R'2 c'l C'2
195
Rl' = 33083.1247, R2' = 47853 .5056
Stage 3 (Q = 0.89998) : C1' = C2' = le-009
R1' = 22105.4372 , R2' = 71617.8323
Stage 4 (Q = 2.5629): Cl' = C2' = le-009
R1' = 7762.3973, R2' = 203950 .3311
The resulting resistor and capacitor values are realistic.
MATLAB also convenient ly computes magnitude responses and pole locations.
From H (s), it is clear that all zeros are at zero.
>> Hmag_SK = zeros(NS,200); Poles= zeros(NS,2);
>> f = linspace(O,omega_O/pi,200);
>> for stage= l:NS,
>> B ( 1 0 OJ;
>> A = [1,(1/(R2p(stage)•C2p(stage))+1/(R2p( stage )•Clp(stage))), ...
>> 1/(Rlp(stage) • R2p(stage)*C1p( stage)*C2p(stage)) ] ;
>> Poles(stage,:) = (roots(A)).';
>> Hmag_SK(stage,:) = abs(polyval(B ,j*2*pi*f)./polyval(A,j*2*pi*f));
>> end
>> subplot(121), plot(real(Poles(:)), imag(Poles(:)) , 'kx' ,0,0,'ko');
» axis(omega_0*(-1.1, . 1 - 1.1 1.1)); axis equal;
>> xlabel('Re(s) =\sigma'); ylabel( 'Im(s) - \omega') ;
>> subplot( 122) , plot(f,Hmag_SK,'k' ,f,prod(Hmag_SK),'k:');
>> xlabel('f [Hz] ') ; ylabel('Magnitude Responses');
... .
2S
,.
"
05
i
• l
0 0 ~ 1.5
i
··0$ it
_,
• 1,S
_, ·~
-2.$
_, ..... •
-2S -u
-·
Re(&>' o 1.10·
2000 • 000
IJ'<ll ""'° 8000
196
»Cl = 2• Q./(omega_O•Rl); C2 = 1 ./(2• omega_O•Q. • R2);
>> for stage = l:NS,
>> disp( [' Stage ' , num2s tr(stage) , . . .
>> ' (Q = ' , num2str ( Q(stage)) , ...
» '): Rl = R2 = ' , num2str(R1(stage)))) ;
>> disp( ( ' Cl= ',num2str (C1(stage)), ...
>> ' , C2 = ' ,num2str (C2 (stage))) );
>> end
Stage 1 CQ 0.50242): R1 = R2 = 106103. 2954
Ci 1.0048e-009, C2 = 9 .951 8e-010
St age 2 (Q 0.5225): R1 = R2 = 106103 .2954
Cl 1.045e-009, C2 = 9.5694e-010
St age 3 (Q 0.56694): R1 = R2 = 106103 . 2954
Ci 1.1339e-009, C2 = 8.8192e- 010
Stage 4 (Q 0.64682): Rl = R2 = 106103. 2954
C1 1 . 2936e-009, C2 = 7.7301e- 010
Stage 5 (Q 0 . 78815): Rl = R2 = 106103.2954
Cl 1.5763e-009, C2 = 6. 3439e- 010
Stage 6 (Q 1 . 0607): Rl = R2 = 106103 . 2954
Cl 2.1214e-009, C2 = 4 .714e- 010
St age 7 (Q 1. 7224 ) : Rl = R2 = 106103.2954
Cl 3 .4449e-009, C2 = 2 . 9028e-010
Stage 8 (Q 5 .1 01 1) : R1 = R2 = 106103.2954
Cl 1 . 0202e- 008, C2 = 9.8017e-011
The resulting resistor and capacitor values are realistic.
(b) See the solution to problem 4.M-4b.
(c) MAT LAB is used to transform the Butterworth LPF from 4.M-5a.
>> Rlp = 1./(Cl•omega_O) ; R2p = 1 ./(C2•omega_O);
>> Clp = 1. /(Rl•omega_O) ; C2p = 1./(R2•omega_O);
>> for s tage = l:NS,
» disp( [ ' St age ',num2s t r( stage) , . ..
» ' (Q = ',num2st r(Q (stage) ) , . ..
>> '): Cl''= C2'' = ',num2st r (C1p ( stage))));
>> disp((' Rl'' = ',num2str (Rlp(stago)), ...
>> ' , R2'' = ',num2str(R2p( st age ))));
» end
Stage 1 (Q = 0.50242): Cl' = C2' = le-009
Rl ' = 105592.379, R2' = 1066 16 . 6839
Stage 2 ( Q = 0 .5225): Cl'= C2' = l e-009
R1 ' = 101534.5231, R2' = 110877. 6498
Stage 3 ( Q = 0 . 56694) : Cl'= C2' = le-009
Rl ' = 93574 .7524, R2' = 120309.2608
Stage 4 (Q = 0.64682) : Cl ' = C2' = le-009
R1' = 82018.9565 , R2 ' = 137259.8455
Stage 5 (Q = 0.78815): Cl' = C2' = le-009
Rl' = 67311.218 , R2' = 167251.6057
Stage 6 (Q = 1 .0607): Cl' = C2' = le-009
R1' = 50016 .7472, R2' = 225082.7957
Stage 7 (Q = 1 .7224): Cl '= C2' = le-009
Rl' = 30800 . 1609, R2' = 365514.6265
197
Stage 8 (Q = 5.1011): Cl' = C2' = le-009
Rl' = 10399.9416, R2' = 1082497.3575
The resulting resistor and capacitor values are reasonably realistic.
MATLAB also conveniently computes magnitude responses and pole local.ions.
From H(s) , it is clear that all zeros are at zero.
>> Hmag_SK = zeros(NS,200); Poles= zeros(NS,2);
>> f = linspace(O,omega_O/pi,200);
>> for stage = l:NS,
>> B (1 0 O];
>> As [1,(1/(R2p(stage)*C2p(stage))+1/(R2p(stage)*Clp(stage))), ...
>> 1/(Rlp(stage)*R2p(stage)•Clp(stage)*C2p(stage)) ];
>> Poles(stage,:) = (roots(A)).';
>> Hmag_SK(stage,:) = abs(polyval(B,j*2*pi*f)./polyval(A,j*2*pi*f)) ;
» end
>> subplot(121), plot(r eal(Poles(:)),imag(Poles(:)),'kx',0,0,'ko');
>> axis(omega_0*(-1.1, .1 -1.1 1.1)); axis equal;
>> xlabel('Re(s) =\sigma'); ylabel('Im(s) u \omega');
>> subplot(122), plot(f,Hmag_SK,'k',f,prod(Hmag_SK),'k:');
>> xlabel('f [Hz]'); ylabel('Magnitude Responses');
...
••
•
-; .
0.2
0
i'
;.
r
-0.2
.... i
-··
-0,S
_,
-•0000
....---...,.~,.....,~--<000~-2000~-.~ ••
Figure S4.M-5c: Order-16 Butterworth HPF with we = 2rrl500.
The overall magnitude response plot looks like a highpass Butterworth filter;
the cutoff is correctly located at We and the response is smooth and monotonic.
Int.ercstingly, the Butterworth HPF poles look identical to the Butterworth LPF
poles. The only difference is seen in the zeros; all the zeros of the LPF are infinite,
and all the zeros of the HPF are located at. s = 0.
'1.M-6. (a) Using MATLAB, the Sallen-Key component values are easily found.
>> omega_c = 2•pi•4000; R = 3; N = 8;
>>epsilon= sqrt(lO-(R/10)-1);
>> k = [l:N]; xi= 1/N•asinh(l/epsilon); phi= (k•2-1)/(2*N)*pi;
>> C_poles = omega_c*(-sinh(xi)*sin(phi)+j•cosh(xi)•cos(phi));
>> C_poles = C_poles(find(imag(C_poles)>O)); I. Quadrant 2 poles
>> f = linspace(0,10000,501); Hmag_SK = zeros(6,501);
>> R1 = zeros(N/2,1); R2 = Rl; Ci= R1; C2 = Rl; Q = Rl; omega_O Rl;
198
>> for stage = 1 :N/2 ,
>> omega_O(stage) = abs(C_poles(stage));
>> psi= pi - angle(C_poles(stage));
>> Q(stage) = 1/(2*cos(psi));
>> R1(stage) 1e9/omega_c; R2(stage) = R1(stage);
>> Cl(stage) = 2*Q(stage) . /(omega_O(stage)*R1(stage));
>> C2(stage) = 1./(2*omega_O(stage)*Q(stage).*R2(stage));
>> disp(['Stage ',num2str(stage), . . .
>> ' (Q = ',num2str(Q(stage)), ...
>> '): R1 = R2 = ',num2str(Rl(stage))]);
>> disp([' C1 = ',num2str(C1(stage)), . ..
>> ', C2 = ',num2str(C2(stage))]);
>> B = omega_O(stage)-2; A= [1 omega_O(stage)/Q omega_O(stage)-2];
>> Hroag_SK(stage,:) = abs(polyval(B,j *2*pi*f)./polyval(A,j*2*pi*f));
>> end
Stage 1 (Q 22.8704): R1 = R2 = 39788.7358
Cl 4 . 6343e- 008, C2 = 2.215e-011
Stage 2 (Q 6.8251): R1 = R2 = 39788.7358
Cl 1.6274e-008, C2 = 8.7339e-011
Stage 3 (Q 3.0798): R1 = R2 = 39788.7358
Cl 1.0874e-008, C2 = 2.8659e-010
Stage 4 (Q 1.0337): R1 = R2 = 39788.7358
C1 9.2182e-009, C2 = 2.1569e-009
The resulting resistor and capacitor values are realistic.
(b) See the solution to problem 4.M-4b.
(c) MATLAB is used to transform the Chebyshev LPF from 4.M-6a.
>> Rip = 1./(C1*omega_c); R2p 1./(C2*omega_c);
>> Clp = 1./(Rl*omega_c); C2p = 1./(R2*omega_c);
>> for stage = l:N/2,
» disp( ('Stage ',num2str(stage), ...
>> ' (Q = ',num2str(Q(stage)), ...
>> '): Cl"= C2" = ',num2str(Clp(stage))]);
» disp([' R1'' = ',num2str(R1p(stage)), ...
>> ', R2'' = ',num2str(R2p(stage))));
>> end
Stage 1 (Q = 22 . 8704): Cl' = C2' = 1e-009
R1' = 858.5676, R2' = 1796313.2499
Stage 2 (Q = 6.8251): Cl' = C2' = 1e-009
Rl' = 2444.9937, R2' = 455567.9755
Stage 3 (Q = 3.0798): Cl'= C2' = le-009
Rl' = 3659.1916, R2' = 138833.4048
Stage 4 (Q = 1.0337): C1' = C2' = 1e-009
Rl' = 4316.3108 , R2 ' = 18446.8849
The resulting resistor and capacitor values possibly realistic; however , there is a
fairly la rge dynamic range between the largest and smallest resistors.
MATLAB also conveniently computes magnitude responses and pole locations .
By necessity, the transformation really stretches out the passband ; it is t herefore
important to plot the magnitude response over a broad range of frequ encies. To
facilitate a reasonable plot, the magnitude response is p lotted using both log-
199
magnitude and log-frequency scales. From H(s), it is clear that all zeros are at
zero.
>> Hmag_SK = zeros(N/2,5001); Poles= zeros(N/2,2);
>> f = logspace(2,5,5001);
>> for stage = l:N/2,
» B (1 0 0);
>> A= [1,(1/(R2p(stage)*C2p(stage)) +1/(R2p(stage)*C1p(stage))), ...
>> 1/(R1p(stage)*R2p(stage)*C1p(stage)*C2p(stage)));
>> Poles(stage,:) = (roots(A)).';
>> Hmag_SK(stage,:) = abs(polyval(B,j*2*pi*f)./polyval(A,j*2*pi*f));
>> end
>> subplot(121), plot(real(Poles(:)),imag(Poles(:)),'kx' ,0,0,'ko');
>>axis equal; ax = axis; axis([1.1*ax]);
>> xlabel('Re(s) = \sigma'); ylabel('Im(s) =\omega');
>> subplot(122),
>> semilogx(f,20*log10(Hmag_SK),'k',f,20*log10(prod(Hmag_SK)),'k:')
>> xlabel('f (Hz)'); ylabel('Magnitude Responses [dB] '); axis tight
>> axis([100 1e5 - 40 40));
II t0,
••
o.• 30
0.6
20
O•
0.2
•. !
i
10
• 0 0 i 0
I "'j
-<>.2
: §- - 10
~
-<>.•
-20
-<>.6
-0.4 ·JO
-· -<;
-· -?
Ro(s)=o
2
x 10·
-•O
10' 10'
l(HLJ
.•. 10'
4.M-7. Factored form is used to plot roots, and standard transfer function form is used to
compute magnitude response plots.
200
>> axis(omega_c*(-1.1 0 . 1 -1.1 1.1]); axis equal;
>> xlabel('Re(s) =\sigma'); ylabel('Im(s) =\omega');
>> f = linspace(0,7000,501);
>> (B ,A] = butter(6,omega_c,'s');
>> HLP = polyval(B,j•2• pi•f)./polyval(A,j*2*pi•f);
>> subplot(122) , plot(f,20*loglO(abs(HLP)),'k');
>> axis([O 7000 -40 2])
>> xlabel('f [Hz]'); ylabel(' IH_{LP}(j\omega)I');
1.5
.
.,
\ I
... -•5
-0.5
·ZS
_,
- 1.s
->S
·2
-•• '----~---~-'
-2 - \,S .,
A~(s}• o
*0.5 0
x 10•
•
Figure S4.M-7a: Order-6 Butterworth LPF with We= 27T3500.
(c) Order-6 Butterworth BPF with passband between 2kHz and 4kHz. Notice that
the command butter requires the parameter N = 3 to be used to obtain a
(2N = 6)-order bandpass fi lter.
>> omega_c = [2*pi*2000 , 2*pi*4000);
>> [z,p,k] = butter(3,omega_c , 's');
>> subplot(121),plot(real(p),imag(p),'kx' , ...
real(z),imag(z),'ko');
>> axis(omega_c(2)*[-1.1 0 . 1 -1.1 1.1]); axis equal;
>> xlabel ('Re(s) =\sigma'); ylabel('Im(s) =\omega');
>> f = linspace(0 ,7000, 501);
>> [B,A] = butter(3,omega_c,'s');
201
-$
1.S
-10
.. o.s
0
.,.
r
" 0
-0.S
-1
-t.S
-35
-•o'----'-~-~-~~
-2 - 1,S -1 -0.S 0 O 2000 4000 GOOO
Rt(•)•o x 10• 11m1
2.S
-s
1.S
·10
1l
o,s
_,,
• ~
f
0 0
.f-20
.....
·25
-1
- 1.S - 30
_,
- 3$
-2.s
-2.S
_, -1.S
-··
·1
R9(•) • <t
-<l.S 0
. 10·
0 :1000
"''"
r(tttJ
6000
Figure S4.M-7c: Order-6 Butterworth BPF with passband between 2kHz and 4kHz.
(d) Order-6 Butterworth BSF with stop band between 2kHz and 4kHz. Notice that
the command butter requires the parameter N = 3 to be used to obtain a
(2N = 6)-order bandstop filter.
>> omega_c = [2*pi*2000,2*pi*4000);
>> [z,p,k] = butter(3,omega_c,'stop','s');
>> subplot(121),plot(real(p),imag(p),'kx', ...
real(z) ,imag(z),'ko');
>> axis(omega_c(2)*[-1.1 0.1 -1.1 1.1)); axis equal;
>> xlabel('Re(s) =\sigma'); ylabel('Im(s) =\omega');
>> f ; linspace(0,7000,501);
>> [B,A) = butter(3,omega_c,'stop','s');
202
>> HBS = polyval(B ,j•2•pi•f )./polyval(A,j*2*pi•f);
>> subplot(122),plot(f,20•log10(abs (HBS)),'k');
>> axis( [O 7000 -40 2))
>> xlabel(' f [Hz]') ; ylabel(' IH_{BS}(j\omega) I ') ;
0
1.$
- 10
0$
_,.
•
-;;- 0
I
-o.s
-1
-· _,.
- ••'---'-'-'--'~~--'
-2'. S -2' -1.5 •l -0.5 0 O 2'000 cooo 1000
R.(~) c G • 1a • t(H.r)
Figure S4.M-7d: Order-6 Butterworth BSF with stopband between 2kHz and 4kHz.
4.M-8. Factored form is used to plot roots, and standard transfer function form is used to
compute magnitude response plots.
(a) Order-6 Chebyshev Type I LPF with We = 27r3500.
>> omega_c = 2*pi*3500;
>> [z,p,k) = cheby 1(6 ,3,omega_c,'s');
>> subplot(121),plot(real(p) , imag(p),'kx' , ...
real(z),imag(z) , 'ko');
>>axis equal; axis(1.1*axis);
>> xlabel('Re(s) =\sigma'); ylabel('Im( s ) = \omega' );
>> f = l inspace(0,7000,501);
>> [B,A] = cheby1(6,3,omega_c,' s');
>> HLP = polyval(B,j*2*pi•f)./polyval(A,j*2*pi*f);
>> subplot(122),plot(f,20*log10(abs(HLP)),'k');
>> axis([O 7000 -40 2))
>> xlabcl('f [Hz]'); ylabel('IH_{LP}(j\omcga) I');
(b) Order-6 Chebyshev Type I HPF with We = 27r3500.
>> omega_c = 2*pi*3500;
>> [z, p,k] = cheby 1(6,3,omega_c, 'high','s');
>> subplot( 121), plot(real(p),imag(p),'kx' , ...
real(z),imag(z),'ko');
>>axis equal; axis(l.l *axis);
>> xlabel('Re(s) = \ sigma' ); ylabel('Im(s) \omega');
>> f = linspace(0,7000 , 501);
>> [B,A] = cheby1(6,3 ,omega_c,'high','s');
>> HHP = polyval(B,j *2*pi*f)./polyval(A,j*2*pi*f);
>> subplot(122) ,plot(f,20*log10 (abs(HHP)),'k');
>> axis([O 7000 -40 2))
203
..
-··
o.s -IS
•
.. 0
r
-0.S
-z
... •05 0 0$ ,
kll{5)~CI I \0'
.10'
-$
-··
•
;
I
. 0
-z -ZS
• JO
(c) Order-6 Chebyshev Type I BPF with passband between 2kHz and 4kHz. Notice
that the command chebyl requires the parameter N = 3 to be used to obtain a
(2N = 6)-order bandpass fi lter.
>> omega_c = (2•pi• 2000,2•pi•4000);
>> (z,p,k] = cheby1(3,3,omega_c,'s');
>> subplot(121),plot(real(p),imag(p),'kx',-·.
real(z),imag(z), 'ko' );
>>axis equal; axis( 1.1•a.xis);
>> xlabel('Re(s) =\sigma'); ylabel('Im(s) \omega');
>> f = l inspace(0,7000,501);
>> (B,A] = cheby1(3,3,omega_c , 's');
>> HBP = polyval(B, j•2•pi • f)./polyval( A,j•2•pi• f);
>> subplot(122),plot(f ,20•log10(abs(HBP)),'k');
>> axis([O 7000 -40 2))
204
>> xlabel('f [Hz]'); ylabel('IH_{BP}(j\omega)I');
2.5
-<;
1.5
- ·10
0 .5 • 15
3 ';'
I
e 0
l -20
......
• 25
-u · 30
·2 ...
·ZS
-1.$ ·1 -O.$ 0
Re(a)•d
0.5 1
... 0 2000 4000
l[MzJ
6000
. 10·
Figure S4.M-8c: Order-6 Chebyshev Type I BPF with passband between 2kHz and 4kHz.
(d) Order-6 Chebyshev Type I BSF with stopband between 2kHz and 4kHz. Notice
that the command cheby1 requires the parameter N = 3 to be used to obtain a
(2N = 6)-order bandstop filter.
>> omega_c = (2*pi*2000,2*pi*4000];
>> (z,p,k] = cheby1(3,3,omega_c,'stop','s');
>> subplot(121),plot(real(p),imag(p),'kx', ...
real(z),imag(z),'ko');
>>axis equal; axis(1.1*axis);
>> xlabel('Re(s) =\sigma'); ylabel('Im(s) \omega');
>> f = linspace(0,7000,501);
>> [B,A] = cheby1(3,3,omega_c,'stop ','s') ;
>> HBS = polyval(B,j*2*pi*f)./polyval(A,j *2*pi*f);
>> subplot(122),plot(f,20*log10(abs(HBS)), 'k');
>> axis([O 7000 -40 2))
>> xlabel('f (Hz)'); ylabel('IH_{BS}(j\omega)I');
To demonstrate the effect of decreasing the passband ripple, consider magnitude
response plots for Chebyshev Type I LPFs with RP = {0.1, 1.0, 3.0}.
>> omega_c = 2*pi*3500; f = linspace(0,7000,501);
>> [B,A] = cheby1(6,.1,omega_c,'s');
>> HLP1 = polyval(B,j*2*pi*f)./polyval(A,j*2*pi*f);
>> [B,A] = cheby1(6,1,omega_c,'s');
>> HLP2 = polyval(B,j*2•pi*f)./polyval(A,j*2*pi*f);
>> [B,A] = cheby1(6,3,omega_c,'s');
>> HLP3 = polyval(B,j*2*pi*f)./polyval(A,j*2*pi*f);
» plot(f ,20*log10(abs(IILP1)), 'k-', .. .
f,20*log10(abs(HLP2)),'k--', .. .
f,20*log10(abs(HLP3)),'k:');
>> axis([O 7000 -40 2])
>> xlabel('f [Hz] '); ylabel('IH_{LP}(j\omega)I');
>> legend('R_p = 0.1','R_p = 1.0','R_p = 3.0',0);
205
-·
-10
_,.
_, - 25
-2
-3
-<OL--~•-'-~-~~
-3 -2 -t 0 2000 4000 6000
Re(s)~o :a to• r(HtJ
Figure S4 .M-8d: Order-6 Chebyshev Type I BSF with stopband between 2kHz and 4kHz.
- 10
,,
:. \
·. \
. '
'
'
.... \
f(ltlj
Thus, reducing the allowable passband ripple RP tends to broaden the transition
bands of the filter.
4.M-9. Factored form is used to plot roots, and standard transfer function form is used to
compute magnitude response plots.
(a) Order-6 Chebyshev Type II LPF with We= 2n3500.
>> omega_c = 2*pi*3500;
>> [z,p,k] = cheby2(6,20,omega_c,'s');
>> subplot(121),plot(real(p),imag(p),'kx', .. .
real(z),imag(z),'ko');
>>axis equal; axis(1.1*axis);
>> xlabel('Re(s) = \sigma'); ylabel('Im(s) \omega');
>> f = l inspace(0, 7000,501);
>> [B , A) = cheby2(6,20,omega_c,'s');
>> HLP = polyval(B,j*2*pi*f)./polyval(A,j*2*pi*f);
>> subplot(122),plot(f ,20*log10(abs(HLP)),'k');
>> axis((O 7000 - 40 2))
206
» xlabel( 'f [Hz]'); ylabel('IH_{LP}(j\omega)l');
a 10"
-s
- 10
0
• .P -15
•
; 0 ~ - 20
~
I
., • 'b _,.
-· -:io
-~
•,3.5
·•
... -· -2
A:.e(s) .s o
0
....
2
I 10•
"' 0
(
.. 0
·•
-··
. •• 0 . ..
r 0
r
·• s 0 ~
•I
•IS 0
...
-2
0
.... 0
Ae(s)o: o
207
(c) Order-6 Chebyshcv Type II BPF with passband between 2kHz and 4kHz. Notice
that the command cheby2 requires the parameter N = 3 to be used to obtain a
(2N = 6)-order bandpass filter.
>> omega_c - (2*pi*2000,2•pi•4000);
>> [z,p,k) = cheby2 (3,20, omega_c,' s');
>> subplot(121) ,plot(real(p),imag(p),'kx', .. .
real(z),imag(z),'ko');
>>axis equal; axis(1.1•axis);
>> xlabel('Re(s) =\sigma'); ylabel('Im(s) \omega');
>> f = linspace(0,7000,501);
>> [B,A) = cheby2(3 ,20, omega_c,'s');
>> HBP = polyval(B,j•2•pi•f)./polyval(A,j*2•pi•f);
>> subplot(122) ,plot(f,20•log10(abs(HBP)),'k');
>> axis((O 7000 -40 2))
>> xlabel('f [Hz)'); ylabel(' IH_{BP}(j\omega) I ');
J 10·
... 0
·•
... 0 ...
.••• ...
r .... 0
... 0
• 35
.. z,1
... -· ..... 0
•
R-<...).c o
.. .
,. to"
....•
"""' .....,"""" ....
Figure S4.M-9c: Order-6 Chebyshev Type II BPF with passband between 2kHz and 4kHz.
(d) Order-6 Chebyshev Type II BSF with stopband between 2kHz and 4kHz. Notice
that the command cheby2 requires the parameter N = 3 to be used lo obtain a
(2N = 6)-order bandstop filter.
>> omega_c ~ [2•pi•2000,2•pi•4000);
>> [z,p,k] = cheby2(3,20,omega_c,'stop','s') ;
>> subplot(121),plot(real(p), i mag(p),'kx', . . .
real(z),imag(z),' ko');
>>ax is equal; axis(l .l• axis);
>> xlabel('Re(s) =\sigma '); ylabel('Im(s) \omega');
>> f = linspace(0 ,7000,501 );
>> [B,A) = cheby2(3,20,omega_c,'stop' , 's');
>> HBS = polyval(B,j•2*pi•f)./polyval(A,j•2•pi•f);
>> subplot(122),plot(f,20• log10(abs(HBS)), ' k' ) ;
>> axis([O 7000 -40 2))
>> xlabel('f [Hz)'); ylabel(' IH_{BS}(j\omega) I ');
To demonstrate t he effect of increasing R. , consider magnitude response plots
for Chebyshev Type JI LPFs with R s = {10, 20, 30}.
208
• 10·
l
-5
0
0 - to ·
•IS
• i
I • .f -20
_, -M
-2
0
0
...
0 -35
-3
-2 0
fW(s)= o a 10·
_...,
• 2000 ....
•1114
6000
Figure S4.M-9d: Order-6 Chebyshev Type II BSF with slopband between 2kHz and 4kHz.
'I
' '
- 10
."
' '
'
-lO '
''
_,.
llH>I
209
(a) Order-6 Elliptic LPF with We = 27f3500.
>> oroega_c = 2*pi*3500;
>> (z,p,k] = ellip(6,3,20,omega_c,'s');
>> subplot(121),plot(real(p),imag(p),'kx', ...
real(z),iroag(z),'ko');
>>axis equal; axis(1.l*axis);
>> xlabel('Re(s) =\sigma'); ylabel('Im(s) \omega');
>> f = linspace(0,7000,2001);
>> (B,A] = ellip(6,3,20,omega_c,'s');
>> HLP = polyval(B,j*2*pi*f)./polyval(A,j*2*pi*f);
>> subplot(122),plot(f,20*log10(abs(HLP)),'k');
>> axis((O 7000 - 40 2))
>> xlabel('f [Hz]'); ylabel('IH_{LP}(j\omega)I');
(c) Order-6 Elliptic BPF with passband between 2kHz and 4kHz. Notice t hat the
command ellip requires the parameter N = 3 to be used to obtain a (2N = 6)-
order band pass fi lter.
>> omega_c [2•pi •2000,2•pi•4000);
>> [z,p,k) = ell i p(3,3,20,orocga_c,'s');
210
x 10·
'8
-·
-10
0
-15
3
r 0
-1
0
-2
p
-3
-2 -1
Ait<•> =• 11 10·
• 10·
0
2.S
·•
IS
_,.
0
•
05
-··
r ~
0
..
0
_,
!'
- 1.5
_,
_,. _,,
... ., ..,, 0
0
R•CO • o
05 ,
110·
.... 0
""'° •OOO
flHl)
IOOO
Figure S4.M-10c: Order-G Elliptic BPF with passband between 2kHz and 4kHz.
(d) Order-6 Elliptic DSF with stopband between 2kHz and 4kHz. Notice that the
command ellip requires the parameter N = 3 to be used to obtain a. (2N = 6)-
order bandstop filter.
>> omega_c [2•pi•2000,2• pi•4000) ;
>> [z,p,k) = ellip(3,3,20,omega_c,'stop','s');
211
>> subplot(121),plot(real(p),imag(p),'kx', ...
real(z),imag(z),'ko');
>>axis equal; axis(l.l•a.xis);
>> xlabel('Re(s) =\sigma'); ylabel('Im(s) \omega');
>> f = linspace(0,7000,2001);
>> (B,A] = ellip(3,3,20,omega_c,'stop','s');
>>RBS= polyval(B, j•2•pi•f)./polyval(A, j•2•pi•f);
>> subplot(122),plot(f,20•log10(abs(HBS)),'k');
>> a.xis([O 7000 -40 2))
>> xlabel('f [Hz]'); ylabel('IH_{BS}(j\omega)I');
.,•. .
u
0 ON
,. 0
!'
-·
-··
.• ... i
., .
r .. 0
.j •20
r
., .,.
i,
' - ---·
• 1,5 • JO
., 0
• >S
.,, .
0
..•
-20000- 1$000-10000-SOCO
Ft.(!oJ•o
0 .... .
2000 .... ....
l (KtJ
Figure 84.M-lO<l: Order-6 Elliptic BSF with stopband between 2kHz and 4kHz.
4.M-11. First, the recursion relation CN(x) = 2xCN - i(x)-CN _2 (x) is rewritten as CN+i(x) =
ZxCN(x) - CN-1(x) or CN+i + CN-1 = ZxCN(x).
Letting 'Y cosh- 1 (x) and using Euler's formula, we know CN(x) =
1
cosh(Ncosh- (x)) cosh(N"f} = e"..,+2e_,.,., Thus, CN+1 + CN - 1 =
e<"'+'h+e-(N+th e<N- th+e-(N- th e"'"(e"+e--,)+e-N,.(e"'+e-")
2 + 2 2
2<e"+r"l Ce"..,+r""> = 2cosh('Y)cosh(N"f). Replacing "Y yields CN+1 + CN - 1 =
2cosh(cosh- 1 (x)) cosh(N cosh- 1 (x)) = 2xCN(x). Thus,
4.M-12. Note that Pk= <lk + JWk = wcsinh(~) sin(</>k) + JWccosh(~)cos(ef>k) · From the real
portion, we know <lk = We sinh({) sin(<i>k) or sin(¢>k) = w. sf.~h(~). From the imaginary
portion, we know Wk = We cosh(O cos(¢>k) or cos(<i>k) = "'• c~%h(O. From trigonometry,
we know 1 = cos 2 (¢>k) + sin 2(¢>k). Thus,
This is the equation of an ellipse. Since the Chebyshev poles Pk = ak + JWk satisfy the
equation, they must lie on the ellipse.
212
Chapter 5 Solutions
5.1-1. Given the fact that the t ime-domain signal is finite in duration, the region of con-
vergence should include the entire z-plane, except possibly z = 0 or z = oo.
X( Z ) = "\'oo I J - n = "\'7
L.Jn= - ooX n Z
( l)n -n
= °"'7 ( l/ )n
= l- (- l /z)
8
ThUS,
...__ - - L.J n=O - Z L,,n=O - Z
I-z-
1-(- 1/z).
8
X(z) =
1 + z- 1
; ROC lzl > 0.
In this form, X(z) appears to have eight finite zeros and one finite pole. The eight
zeros are the eight roots of unity, or z = e2 " k /S for k = (0, 1, ... , 7). The apparent
pole is at z = -1. However , there is also a zero z = - 1 (k = 4) that cancels this
pole. Thus, there are actually no finite poles and only seven finite zeros, z = e 2"'k/ S
fork = (0, 1, 2, 3, 5, 6, 7). MATLAB is used to plot the zeros in t he complex plane; the
unit circle is also plotted for reference.
...
..•
•.
O.l
I 0
-0.Z
-04
.....
....
•1
., -o.s
·~
0
A4'(t}
Figure S5. l-l: Pole-zero plot for xlnJ = (- 1r(ulnJ - uln - 81).
5.1-2. (a)
00
213
= z- m [ 1 + -1 + -+··
1 ·]
z z2
- m ( 1 ) Z
z 1- ~ = zm(z - 1)
X (z)
Hence
X[zJ
(f)
00
k=O
214
4 16 64
XlzJ = 1 + -z2 + -z 4 + -z6 + ...
Geometric progression with common ratio f, . Hence
2
X[z] = -
1-
1
-~4 = -z 2z- -4 lzl > 2
(g)
~ n - 1z - n
X[z] = L_,/ l
=-L_,
~ ('- )n
n= l / n=l z
(i)
x [nl = nu[n)
and
n=O
Using the results in Sec. B.7.4 with lzl > 1, we obtain
X z _ (1/z) _ z
I J- [(1/z) - 1]2 - (z - 1) 2 lzl > 1
(j)
1" l I "
=L =L I (-)
oo oo
X [z) 1z-"
n =O n. n=O n. z
Recall that
00 1
e"' = L: -xn
n =O n!
215
Therefore
X [z] = e'Yfz
(k)
XlzJ ~ [; ( ~) n+ ~ ( ~2) n]
= -1 [ -1- + -1- ]
2 1-~ l+~
= l [ z z ]
2 z-2+z+2
z2
=
z2 - 4
(1)
X
~n!
00
lzJ = "'"'-(1na)nz-n
I =
00
"'"' -l
~n!
(Ina)"
-
z
n= O n=O
5.1-3. Note that the signal x!nJ = nu!nJ has z-transform X(z) = I:::°=o nz- n = (z ~I)2 .
Thus, L:::'=o n(-3/2)-n is easily found by evaluating the X( z )lz=- 312 . That is,
"°"
L.m=O 1l
( - 3/2)-n -- (z-z1)2 I
z= - 3/2
-
- -3/-2 I )2 --
( -3/2 - ~
25[4 -- 50 -- -~
_g 25 · Th US,
5.1-4. (a)
-2
/n u(n - 2] = 21 {lnu{n] - o!nJ - 1o[n - l]}
I
Hence
1"-:.!u.[n - 2] ~ _!__
'Y2
[-z-
z- /
- 1- I]=
z
__ l_
z(z - 1)
(c)
1
x [nj = (2r+ 1u [n - 1] + (et- 1u [n] = 4(2)"- 1u [n - lj + ~(e)"u (n]
216
Therefore
4 l z
X lzJ = - + - -
z- 2 ez-e
(d)
Therefore
X z = z(z - 0.25) _ = 0.25(z - 1)
1
[ ) z2 - 0.5z + 0.25 z 2 - 0.5z + 0.25
(e)
x [nJ = n1"u [n - l] = n1"u [n) - 0 = n1"u [nJ
Therefore
-yz
X lzJ = (z -1)2
(f) Because n(n - l )(n - 2) = 0 for n = 0, 1, and 2
n = 0, 1, or 2. Therefore
x[n] = (-l)nnufn)
1 2 3 4 5
X [z] 0 - -+ -2 - - + - 4 - -+···
z z z3 z z5
00
1) n oo ( 1) 2n- l
~ 2n ~(2n + 1)
(
z2 - z2
00 00 00
= 2 L
n=O
n
( 1 )"
z2 -
2
; L
n =O
n
( 1 )"
z2 -
1
; L
n=O
( 1 )"
z2
2z 2 2 z2 z
X [z] (z2 - 1)2 - ; (z2 -1)2 - z 2 - 1
-z3 + 2z2 - z
(z2 1)2
- z(z - 1)2
(z + 1) 2 (z - 1) 2
-z
(z + 1) 2
217
(h)
00
x [n) = '_LkO[n - 2k + 1)
k=O
1 n = 2k - 1
o[n - 2k + 1J = { 0 n f: 2k - 1
We obtain
'.L kz-(2k-1)
00
=
k =O
= z~k (:2r
l/z2
Z ---'--~
2
[(l/z) 2 - 1)
z3
= (z2 - 1)2
5.1-5. (a)
X[zJ z- 4 2 1
z
=
(z - 2)(z - 3) z-2
= - - -z--
-3
_z_ _ _ z_
X[zJ = 2
z-2 z-3
x [n ] = [2(2)" - (3)"J u [n)
(b)
1 1
(z - e- 2 )(z - 2)
- ---..- -
z - e- 2
--
z - 2
z z
X[zJ
218
(d)
X!zJ = (z - 1)
2
= z2 - 2z + 1 = ! _ ~ + _!_
z3 z3 z z2 z3
Hence
x[nJ = o!n - lJ - 2o[n - 2J + o!n - 3]
(e)
0 = 3 + k + 0 ===:> k = -3
X[z] = 3_z_ -3-z- + 4 z
z+1 z- 2 (z - 2)2
x[nJ = [3(- 1r - 3(2r + 2n(2rJu[nJ
(g)
O=l+k =} k = -1
z z z
X[zJ = - - - - z-- +
z - 0.2 0.8
2- --
(z - 0.8)2
0.5 1
r = v'4 =2 .
e = cos - I
(-)
1
== -3
7f
8 = tan
- I
( M")
v3
7l'
= -3
1fn 7T trn 1T
x [n) = 2(1)" cos(
3 + '3 )u [n] = 2 cos( J + '3 )u [nJ
219
(i)
0 = -2 +A => A= 2
Set z = 0 on both sides to obtain
46 B
-25 = 2 + -25 => B =- 4
X [z] = _2 + _z_ +
z{2z -4)
z 2 - 6z + 25
z - l
For the second fraction on the right-hand side, we use pair 12c with A = 2,
B = -4, a = - 3, and hi= 5.
r=-
m
2
x [n] =
[-2 + -ffi
-(5)" cos(0.927n -
2
0.25) ] u[n]
(k)
0 =1 + A => A = -1
220
Setting z = 0 on both sides yields
11.34 1 B
- 50 = - 2 + 25 ~ B = 6 ·83
_ z z (- z + 6.83)
X [z J - - - + -...,.-- - ----
z - 2 z 2 - 5z + 25
For the second fraction on right-hand side, use pair 12c with A= - 1, B = 6.83,
a= - 2.5, and bl = 5.
- 1(-4.33)
/3 =cos - 1 (0.5) = 3
7r
8 = tan - - =- -3r.
-4.33 4
l 4
X [z] =2 - -
z
+ -z2 + · · ·
T herefore x [OJ= 2, x III = - 1, x [2J = 4.
(b) Long division of 2z 4
+ 16z 3 + 17z 2 + 3z by z 3 + 7z2 + 2z + 1 yields
1 4
X [zJ = 2z + 2- - + -2
z z
+ ···
X [z] = 'YZ
z2 - 2-yz + ,.2
z2 -
/Z
2-yz + 1·2
=2
z
+ 2 (1)
z
-f- 3 (2)
z
+ .. ·
221
Therefore x [OJ = 0, x [l] = /, :i; [2J = 212 , x [3] = 313 , · • ·, and
x[nJ = u [nJ - u [n - mj
z -m z 1 - z-m
X[z] = z- l - z z- 1 = l - z- 1
5.2-2.
x (nl = 6 [n - l] + 20 [n - 2J + 36 [n - 3j + 46 [n - 4] + 3o [n - SJ+ 2o [n - 6] + o [n - 7]
Therefore
1 2 3 4 3 2 1
X[zJ = ; + z2 + z3 + z4 + zs + z6 + z7
z6 + 2z5 + 3z4 + 4z3 + 3z2 + 2z + 1
= z7
Alternate Method:
222
1 [
= z 7 ( z - 1)Z z - 2z
8 4
+ 1]
Reader may verify that the two answers are identical.
5.2-3. (a)
/Z
(z - 1)2
1z(z + 1)
(z - 1)3
Let 'Y = 1
(b) Consider
z
/nuln] <==} - -
z - 'Y
Application of mult.iplication property to this pair yields
n1 ,,u [n l ~ - zdd-z ( - Z
z --1
)
= (z-"(Z1 )2
One more application of mu lt iplication property to this pair yields
2 n d [ /Z ] /z{z + 1)
11 'Y u[nj <==} -z dz (z - 1)2 = (z -1)3
2 2
n 3 i " u [nj = -z.:!:_ [1z(z + 1)] = 1z(z + 41z + 1 )
dz {z - 1 )3 (z- 7) 3
Now set t ing / = 1 in this result yields
3 [ J _ z(z
2
+ 4z + 1)
nun - (z - 1)3
(d}
X[z]
223
e ·· 2m [ 1 ]
= z ( z - e- 2 ) 2 2e (1 - m) + mz
(f)
x[n] = (n - 2) (0.5)"-3 u [n - 4J
= ~(n - 4 + 2)(o.s)"- 4 u[n - 4]
5.2-4. P air2:
u [n) = o[nJ + o[n - 1] + o[n - 2] + o[n - 3J + · · ·
1 1 1 1 z
u [nJ {:=}
1+ ~ + z2 + z3 + . · . = 1- l
%
=z- 1
Repeated application of Eq. (5.21) to pair 2 yields pair 3, 4, and 5. Application of Eq.
(5.20) to pair 2 yields pair 6, and application of time-delay (5.15b) to p air 6 yields
pair 7. Repeated application of Eq. (5.21) to pair 7 yields pair 8 and 9.
5.2-5. There is a t,y pographic error in the book problem: cos(7rn/A) sh ould be cos(7rn/4).
Given (Pair llb with bl= 1)
zsin fJ
sin ,Bnu[nJ <=> ( fJ)
z 2 - 2 cos z+ l
224
Therefore
7r
cos -rm[n] ¢:::>
-1
+ 1+ -
1
= - -
z(z - '7i)
--'----
4 ./2z(z2 - ./2z + 1) ./2z (z2 - ./2z + 1)
Moreover
Hence
1
13-nu[-n - lJ ¢:::> - - -1 = ....!!!_ lzl < 1/(3
1 - /3z 1 - /3z
Setting /3 = 1/'Y, we obtain
lzl <bl
5.2-7. (a)
(c) (i)
Hence
(ii) To find
Hence
z
7" COS7rnu[n] <==> - -
z +,.
225
5.2-8. (a)
k=O k=O
~ xlkJ ~ zX!z]
~ z-1
k =O
(b) Let x[n) = o[n]. which yields X!z] = L Application of the result in part (a} yields
n
~elk] = u!n] <=> _z_
~ z-1
k=<O
5.2-9. (a) In the time-domain, (z-~~75 >' is a convolution of two causal, decaying exponen-
tials. Thus, either plot 1 or plot 13 is possible. Using the IVT, the initial value
is lim,,_00 (z-~.•75 )2 = 1. Thus,
z2
Plot 13 corresponds to ( -r::) •
z - 0 .tv 2
(c) Note that L:!=oz - 2 k = 1 +z- 2 +z-4 +z- 6 + z - 8 • Thus, the time-domain signal
is o[n] + oln - 2] + o[n - 4] + oln - 6] + oln - 8] and
4
Plot 18 corresponds to L z- 2
k.
k=O
-$
(d) By inspection, 1
_:z 1 corresponds to a unit step that is shifted to the right by
five. Thus,
z-5
Plot 10 corresponds to .
1- z - 1
(e) Using synthetic division on zj~ 1 yields (z- 2 + z- 6 + z- 10 + ... ). In the time-
domain, the first non-zero term therefore occurs at n = 2. Thus,
z2
Plot 15 corresponds to - 4- - .
z - l
226
.IS 1·1m,_00 0 7Sz
{z-·o. 7 s)2 = 0. Tl lUS,
0.75z
Plot 1 corresponds to (z _ 0. ) 2 .
75
z+I
has the form of a sinusoid. Thus, either plot 3
or plot 4 is possible. Using the IVT, the initial value is limz_, 00 z~ - -A./2
2
z+
= 1. 1
Thus,
Plot 4 corresponds to
z2 -z/./2
rn .
z 2 - v2z + 1
grows linearly in the time-domain. Thus, plot 6 or plot 12 are possible. To dis-
tinguish between the two, first determine whether or not a root at z = 1 really
· , F. t 1· z - •-sz- 5 + 4z - 6 5 z 5 -sz+4 =
ex1s, s. •1rs , 1m,_ 1 s(i-.r '>2 = 1.1m,-1 sz(zS-Sz:t4 .
-.t )2 = 11m.. -1 S(z•- 2 ,•+z•)
- 1 s -•+4 - 6
0/0. This is indeterminant so
use L'Hospital's rule: limz-1 " 5( 1 = .:,-r/
rim, _ 1 -!; (S~G-1oz•+s1•)
f.<•"- 5 •+4l = rim,_ 1 3oz• -sz• - s
soz 4 +20z3
0/0 · Th·s
1 t 0 o is
. .. - 1_5z-S+4--6
indeterminant so use L'Hospital 's rule again: hm,_1 s(i z-l)2~ =
..i!_
J"
5
p(z -S • ·H)
IID z-I 5c's zG- J OzS+5z<)
= rIIDz-1 150r4-200z3
20z3
+60z1 = 20/10
= 2
' s·JnCC
.
11m,_1 , - •s(i
-s--•+4 z
.: .. - •J2
6
= 2 .J. •
r oo, no root exists at z =
1
an d
z- 1 - 5z-5 + 4z - 6
Plot 6 corresponds to S(l _ z- l )2
(i) In the time-domain, ,_z1. 1 is a growing exponential due to the root outside the
unit circle. Thus, plot 16 or plot 17 is possible. Using the IVT, the initial value
is lim,_ 00 , _ ..1. 1 = 1. Thus,
z
Plot 16 corresponds to - -- .
z - 1.1
0.25z- 1
Plot 7 corresponds to ( -l)( _ ).
1- z 1 - 0 .75z 1
5.3-1.
227
The z-transform of the system equation is
Pz
zY lz) + M z - -yY lz] =
z-1
Pz
(z - -y)Y lz] = -M + - -
z-1
and
-Mz Pz
YlzJ = -z --'")'+{z----y)(z
-- - -1)
Y[z] = -M P - M P [ 1 1 ]
z + (z - 7)(z - 1) = z - 'Y + "'( -
z - "'( 1 z - 'Y - z - 1
T he loan balance is zero for n = N, that is, y [NJ = 0. Setting n = N in the above
equation we obtain
This yields
p = T/N f\.f = rM
"'(N - l 1 - (1 + r)-N
5.3-2. Because part (b) requires us to separate the response into zero-input. and zero-state
components, we shall start with the delay operator forin of the equations, as
To determine the initial condition y[-lj, we set n = 0 in this equation and substitute
y!OJ = 1 to obtain
e-1]
Y[zj+2 [ -1 YlzJ + -- = ez
z 2 z - e- 1
The term (1 - e) on the right-hand side is due to the initial condition, and hence rep-
resents the zero-input component. the second term on t.he right-hand side represents
the zero-state component of the response. Thus
= 1- e +[ ez ]
z+2 (z-e- 1 )(z+2)
228
1- e 2e2 e
-z+2
-+ (2e+l)(z+2)
+- - - -- -
(2e+ l )(z-e-1)
Hence
z 2e2 z e z
Y[z] = (l - e) z + 2 + 2e + 1 z + 2 + 2e + 1 z - e- 1
The first term on t he right-hand side is the zero-input component and the remaining
two terms represent the zero-state component. Thus
2e2 e
y[nJ = (1 - e)( -2)"uln] + - - (- 2ru[n] + - - e- "u[nl
2e +l 2e+ l
zero-input comp. zero-state comp.
2y [nJ - 3y In - l] + y In - 2j = 4x [nJ - 3x In - l ]
Also
229
7: - Cl , 2;-
4y [n] + 4y [n - lj + y [n - 2) = x In - 1)
Also
1 1
y[n] Y [z] y[n-1] ~ -Y [z]
z
y [n - 2] <:::=}
2z Y [z] + 1
z 1
x[n]
z-1
x In - iJ -¢==> - -
z- 1
(x [- 1] = 0)
+ ~z Y!z] + z 2 Y [z] + 1 = ~ l
1
4Y [zj ( 1)
z-
2
4z + 4z + 1 y [z] = 2- z
(2)
z2 z- 1
and
Y[z] z(2 - z) z(2 - z)
z = 4(z - l)(z2 + z + 0 .25) 4(z - l )(z + 0 .5)2
1 [ 4/9 13/9 5/6 ]
4 z - 1 - z + 0.5 + (z + 0.5) 2
Y[zJ 14 [4 z
9z -
13 z
9 z + 0.5 + 6 (z + 0.5)2
1-
5 z ]
(b) To find the zero-input and the zero-state components, we observe that the only
t erm arising because of the initial conditions is 1 on the left-hand side of Eq. (1).
Hence, we can rewrite Eq. (2) with explicit zero-input and zero-state components
as
2
4z + 4z + Iy [z) = _ 1 + _ l _
z2 z- 1
Here the t erm -1 on the right-hand side represents the zero-input terms and
the second term on the right-hand side represents the zero-state component.
Rearranging the equation, we obtain
Y[z]
- z [ - 1+-1 -]
z - 4(z + 0.5) 2 z - 1
231
-z z
= 4(z + 0.5) 2 + 4(z - l)(z + 0.5) 2
'----.,----'
zero-input zero-state
= -1/4 + 1/8 + 1/9 - _!jJ_ + 1/12
z + 0.5 (z + 0.5) 2 z - 1 z + 0.5 (z + 0.5) 2
zero-input zero-state
Therefore
Y[z] = (-l/4)z + (1/8)z + (l/9)z _ (1/9)z + (1/12)z
z + 0.5 (z + 0.5)2 z- 1 z + 0.5 (z + 0.5)2
zero-input zero-state
(c) The terms which vanish as n -> oo correspond to the transient component and
the terms which do not vanish correspond to the steady-state component. Hence
~
y[n] = [ -
36 12
transient steady-state
y [nJ - 3y In - l] + 2y [n - 2] = x In - 1]
Also
1 1 2
y[n] Y[z] y[n - IJ-¢=> - Y[zJ+2 y [n - 2] <==> 2 Y [z] + - +3
z . z z
l
x[nJ X[zJ x[n-1] ~ - X{z]
z
z
X[zJ = -
z-3
The z-tra.nsform of the system equation is
1
z- 3
-3z + 12
z(z - 3)
232
9 z z 3 z
Y [z] ----6- - + - - -
2 z- l z -2 2 z- 3
y [n] - 2y [n - l] + 2y [n - 2] = x [n - 2J
1 l 1
Y [nJ ¢::::> Y (zJ y [n - l J ¢::::> -Y [z] + 1 y[n - 2] <==> -z 2 Y[zJ + -z
z
1 z
x(n - 2] <==> 2
z
x [z] and X [zJ =- -
z- 1
The z-transform of t he difference equation is
Y [zJ = 2z 2 - 4z + 3 1 z- 1
- -- - --- = --
(z - l )(z - 2z + 2)
2 z - l
+- -- -
z2 - 2z + 2
z
y In + 2J + 2y [n + 1J + 2y [n] = x [n + l] + 2x [n ]
z2 2z z(e +2)
z 2 Y [zJ - z + 2z Y [z] + 2Y [zJ -- - z + - - = -'-- -'-
z- e z-e z- e
z(e+2) z(z+2)
(z 2 + 2z + 2)Y (zJ = z + = --'---'-
z - e z-e
T herefore
Y [zj z +2 0.318 - 0.318z - 0.502
- - -2 - = - - + - -- --
(z - e)(z + 2z + 2) z- e z 2 + 2z + 2
233
y [zJ = 0.
318
_ z _ _ z{0.318z + 0 .502)
z - e z 2 +2z +2
For the second fraction on the right-hand side, we use pair 12c with A 0.318,
B = 0.502, a= 1, 11'12 = 2 and
- 1 -0.1 84
r = 0.367 fJ = cos- ( - )
l
= -3;r B = tan- 1 ( - - ) = -0.525
J2 4 0.318
3
y [n] = (o.318{e)n - 0.367{ht cos( ; n - 0.525)] u [n]
5.3-9. In transform domain, H (z) = z- 1 z~(/3 and Y (z) = z- 1 :;~~ - Since Y(z) = H (z)X(z),
we know X(z) = Y( z )/H(z) = z~~~ . Thus, X(z) = -3z;:~ . 3
Using tables, x[n] =
-3 (( 2)nu[n) - ~(-2)n- 1 u[n - ll) = -3 (-2(-2)n- lu[n] - ~(-2)" - 1 u[n - ll) or
x[n] = -3<5[n) + 7(- 2r- 1u[n - lj.
5.3-10. Taking the z-transform of b[m] = (1.0l)b[m - l] + p[m] and solving for B(z) yields
B(z) = P(z) 1 _i.~i z-I ·Thus, P (z) is required to solve for b[mj . One way to represent
Sally's deposit schedule is p[m] = 100 (u[m) - 2:~ 0 <5[m - (12k + ll)J). Defined t his
way, Sally deposits one hundred dollars on the first day of every month m except for
Decembers, (m= 12k+ll fork= {0,1,2, . . .}).
Taking the z-transform yields P(z) = 100 ( l-~- · - 2=:=- oo L~=O <5[m - {12k + ll)]z-m) =
100 ( l -~-1 - L~o 2=:=-oo <5[m - (12k + ll)]z-m)
100 C -~- • - 2:~0 z- < 12 1
k+i >). Substituting P (z) into the expres-
.
s10n 1r or B( z ) . Ids
y1e 8( z ) - 100 ( l- z-1
1 - ~oo
L..k=O z -(12k+ll) ) i-i.o1z-•
1 -
234
then
y{n] - y[n - 1] = n with y[O] =0
Setting n = 0 in this equation and y[O] = 0, yields
y[O] - y[-1] = 0 ==> y[-1] = 0
z -lY[zJ=z(z+ l)
z (z - 1)3
Hence
Y[zJ z(z + 1) 1 3 2
z
---'----'-
(z - 1)4
= (z - 1) 2
+ (z-1)3 + - --
(z-1) 4
z 3z 2z
Y[zJ - - + (z-1)3 +(z--1)4
-(z-1)2 --
3n(n - 1)
y[n] [n+ 2
+ n(n - l)(n - 2)] [ J
un
3
3
2n + 3n2 + n un
[ J = n(n + 1)(2n + 1) un
[ J
6 6
5.3-13. Let n
ylnJ = 2:k3
k=O
t.hen
y[nJ - y[n - 1J = n3 with y[O) =0
235
Setting n = 0 in this equation and using y[OJ = 0, we get
0 - y[-lJ = 0 = } y[-l J = 0
z(z2 + 4z + 1) 1 7 12 + - -6-
___;,__ _ __;.. = + +
{z-1)2 (z - 1)3 (z - 1)4
(z - 1)5 (z-1)5
z z z z
Y!z] - --+7 '+12 +6---
{z- 1)2 {z -1 )3 (z - 1)4 (z- 1)5
7 1
n+ n(n + l) + 2n(n - l)(n - 2) + l)(n - 2){n - 3)
y[n]
2 4n(n -
n 4 + 2n3 + n 2 n 2 (n + 1) 2
= 4 = 4
5.3-14. Let n
y!nJ =L kak a #1
k =O
then
y[nJ -y[n- lJ = nak with y[OJ =0
Setting n = 0 and y[OJ = 0 in this equation yields
0 - y[-lJ = 0 =} y[- lJ = 0
z - lY[z] = az
z (z - a) 2
or
a a a2
Y [z J az (a - I )2 ca.=iJ'f o:::I
-z- = (z - l)(z - a) 2 = ~ - ~ + (z - a) 2
and
236
Use of the result in Prob_ 5.2-8a yields
~ k {:::::::} z2 = z(z + 1) _ z
f;;;, (z -1)3 (z - 1)3 (z - 1)3
Hence
t
k =O
k= n2 _ n(n - 1) = n(n + 1)
2 2
(ii) Let
Then
X!z] = z(z + 1)
(z - 1) 3
Use of the result in Prob. 5.2-8a yields
Hence
Ln
k2 = n3 3n(n - 1) _ 2n(n - l)(n - 2) 2n3 + 3n2 + n n(n + 1)(2n + 1)
k=O 2 3 6 6
5.3-16. Lei
Then
X[z] = z(z2 + 4z + 1)
(z - 1) 4
Use of the result in Prob. 5.2-8a yields
~ k3 z 2 ( z 2 + 4z + 1) z 7z l 2z 6z
L..., - {:::::::} = + + + ---
k=O (z - 1) 5 (z - 1)2 (z - 1)3 (z - J)4 (z - l)S
Hence
4 3 2
~k3-
L...,
+7 ( l) ( l)(
- n -n n- +2n n- n-2) + n(n-l)(n - 2)(n - 3) =n +2n +n = -'----''-
n2(n+l)2
k =O 2 4 4 4
5.3-17. Let
Then
Xl z] = ( az )2
z-a
Use of the result. in Prob. 5.2-8a yields
2
az a [ z z z
Ln k
ka <==> = - - - - - + a(a - 1) -- -]
k =O (z - l){z - a)2 (a - 1)2 z - 1 z - a (z - a)2
237
Hence
5.3-18. (a)
z(2z + 3)
y [z] = X [z] H [zJ = (z - l ){z - 2)(z - 3)
Therefore
2z+ 3 5/2 7 9/2
- -- - - - - = - - - -- + - -
(z - l)(z - 2)(z - 3) z-1 z -2 z-3
5 z z 9 z
Y[z] - - - - 7--+ - - -
2z-l z-2 2z-3
y(n] = rn - 7(2r + ~(3t] u [n]
(b) From the given H[zJ, we can write
238
or
y[nl - 5y[n - 1) + 6y[n - 2] = 2x[n - I )+ 3x[n - 2)
5.3-20. All cases use the same transfer function. From the given H[z) (after dividing the
numerator and the denominator by 6), we can write
(z 2
- ~z + ~) Y[z) = (5z - l)X[z)
Therefore
Y[z] 5z - 1 12 48 36
z ~)(z - ~)(z - ~) = z - ~
(z - - z- S+ z - ~
Y[z] = 12-z- - 48-z- + 36-z-
z - l z-! z-l
4 3 2
This input is 16 times the input in part (a). Therefore the response will be 16
times the output in part (a) (linearity property). Therefore
This input is 16 times the input in part (b) . Therefore the 1·esponse will be ft
239
times the output in part (b) . Therefore
5.3-21. (a)
z(2z - 1)
y !zJ = X !zJ H [zJ = (z - l)( z 2 - l.6z + 0.8)
Y[z] 2z - l 5 5(z - 1)
z (z - l)(z2 - 1.6z + 0.8) z - 1 z2 - l.6z + 0.8
_ z_ _ z(z - 1)
Y[zj 5 5 2
z- 1 z - l.6z + 0.8
For the second fraction on the right-hand side, we use pair 12c with A = 1,
B = -1, a = -0.8, 'Y = -js, 111 2 = 0.8. Therefore
0.8J5
r = 1.118 f3 =cos- 1 ( --) = 0.464
2
H z _ z 2 + 3z + 3 _ z 2 + 3z + 3
[ J - z2 + 3z + 2 - (z + l)(z + 2)
Therefore
240
3 z 1 z
H[z] ---- +---
2 z+l 2z+2
h(n] rno(n) - (-1r + ~c-2r] u(nJ
(b)
2z2 - z z(2z - 1)
H [z] = z2 + 2z + 1 = -(z.._+___,.1)..,--
2
Therefore
2z - 1 2 3
= ---
(z + 1) 2 z + 1 (z + 1)2
z z
H[z] = 2( z+ 1) - 3 (z + 1)2
h(n] [2(-l)n + 3n(- 1rJu [n] = (2 + 3n)(-l)"u [n]
(c)
H [z] = z2 + 2z = z(z + 2)
z2 - z + 0.5 z2 - z + 0.5
Therefore
H(z] = z +2
z z 2 - z + 0.5
We use pair 12c with A= 1, B = 2, a= -0.5, hl 2 = 0.5, hi = 72, and
- 2.5
~
I
r = 5.099 {3 = cos- 1 (0.5v'5) = B =tan- ( - ) = -1.373
4 0.5
5.3-24. (a)
H[z] 1 - 1 1
z + 0.2)(z - 0.8) =z +
(z
-0.2
- +z - -
- 0.8
z z
H[z] -z--
+ 0.2 +z --0.8-
h[nJ 1-(-0.2)" + (0.8)"] u [n]
(b)
241
(c)
-I
5.4-1. For convenience, let / = 7s and rewrite h[nJ = (In+ (l•)n) u!nJ.
(a) By inspection, the structure is a parallel implementation of two modes, which
are easily identified in h[n]. The transfer function of the structure is H (z) =
l+},z - 1 + l+A~ z 1. Taking the transform of h!n], H (z) = l-~z - 1 + 1_ ./ z-1 ·
T hus,
l +J • I -;
A 1 = -1 = - v'8 and Az = -')' = - /8 .
(b) vo[n] = h[nJ * x[nJ = L::,_ 00 h[kJx[n - kl= {L:~!g 'Yn + (l•r} u[n + 3J . Thus,
1 - /n+4 1 - (1« )n+ 4 +J
Yo[n] = {
J - 'Y
+
l - 'Y*
}
u[n + 3J, where / = 1v8
ro .
242
5.4-2. (a)
2
H z _ 3z - l.8z _ 3z{z - 0.6) _ (~) (z -0.6)
[ J - z2 - z + 0.16 - (z - 0.2)(z - 0.8) - z - 0.2 z - 0.8
Parallel form : To realize parallel form, we could expand H [zJ or H [zJ/ z into
partial fractions. In our case:
1.2z - 0.48
[
H zI =
3
+ {z - 0.2){z - 0.8)
0.4 0.8
H [zI = 3 + (z - 0.2) + (z - 0.8)
Alternatively we could expand H [z] / z into partial fractions as:
and z z
H [z] = 2z - 0.2 + z - 0.8
The realizations are shown in Figure 85.4-2.
(b) Refer to Figure 85.4-2.
5.4-3. (a)
5z + 2.2 5z + 2.2
II [z]
(z + 0.2)(z + 0.8) z 2 + z + 0.16
1 ) ( 5z + 2.2) 2 3
( z + 0.2 z + 0.8 = ; + 0 .2 + z + 0.8
H [ J 3.8z - 1.1
z = z3 - 0.8z 2 + 0.37z - 0.05
II ( t ) ( 3.8z - u )
H z = z - 0.2 z2 - 0.6z + 0.25
243
;·:."·:.
1...
:.:
)<J_lf]u
-->- L:
.L
2. ..........,,,.-.--::.---1
. o"·'.L. o·&
... ·--..
........... ......
/(p
XbL
~J 3 .r.
¥:
-
CC\t'IOY\t~
Ji1 reci"
-o·IG
..3
o ·~
....
... .
xr~-~--.s__,
~~-1 :.:
•.·.
;
'•
3
i ..
I
........
. :-.
. .
·:.'.·
:::·.:::
. ··
1
11
... :
x [ ~]
3
1<1
Pt1vollef
..
:-
~)
( ..
.,
Cascade form:
H [z) = ( - z- ) ( l.6z - 1.8 )
z - 0.2 z 2 + z + 0.5
Parallel form:
H -2z 2z2 + 4z
[z] = z - 0.2 + z2 + z + 0.5
All the realizations are shown in Figure 85.4-5.
(b) Refer t o Figure 85.4-5.
5.4-6. (a) Cascade form:
2
H [z] = (-z- ) (2z + l.3z + 0.96)
z + 0.5 z2 -0.8z + 0.16
Parallel form:
z z 2z
H [z] = z + 0.5 +z- 0.4 + (z - 0.4) 2
244
.21
..... ·
' ..
i·-~.
'·:
:
. .··
..
. .·:
·· . .·.
1
CG!M m'c_
.,,.\Ired...
yqJ
- 1
-o ·5
·i
..
Cttt'\O Y\ le..
r.:\ \fp(' \-
i
• i
Xf'?>J
Pa'folle/
...
.. .
· ;._i
;·.
.::·.
trt~
. L.
5.4- 7.
The realization of thls transfer function is shown in Figure 85.4-7. It can be explained
in two ways. The realization has 5 paths in parallel, and each path represents one term
in the transfer function. The first path (which bypasses all the delays) has transfer
function 2. The second path (going through only one delay) has transfer function l /z,
and so on. Alternately we observe that this transfer functi on has a4 = a 3 = a 2 = a 1 =
0, and b4 = 8, b3 = 2, bi= 0.8, b1 = 1.J?o = 2. Therefore all the feedback coefficients
are zero, and there are no feedback paths. There are 4 feed forward paths with gains
8" 2, 0.8, 1, and 2 as shown in the reaiization.
'
Figure S5.4-7
5.4-8.
1 2 3 4 5 6
H [z) = -z + -z2 + -z3 + -z4 + -z5 + -z6
This transfer function is similar to that in Prob. 5.4-7. Its realization is shown in
Figure S5.4-8.
H (s) = (s + a)(s + b)
(s + c)(s + d)
Let us use abbreviated notation:
A lo denote s + a
B to denote s + b
C to denote s + c
D to denote s + d
250
Ar, Br, er, Dr, denote transposed realization of A, B, C, and D .
The realizat ion (~ ~) denote a direct form canonic realization by cascading (s +
a)/(s with (s + b)/(s + d}, and
+ c) (A;
~) represents a canonic realization using
the cascade of the transpose of (s + a)/(s + c) with direct form of (s + b)/(s + d}, etc
( ,i
C
3T) (~ BT) (fl. AT) ( fl. AT)
D ' D C ' D C ' C D '
This represent a total of 22 ways. We have not considered here the possibility of
change of variables, which can yield unlimited number of realizations.
5.5-1. (a)
JH [einJl2 = HH" = l = 1
· (ei 11 - 0.4)(e-in - 0.4) 1.16 - 0.8cosn
JH [ein) 1-- 1
Jr=l.=1=6=-=o=.8=c=o=sn=
and
LH [e1·n] sinn
= -tan- 1 - - --
cosn - 0.4
(b}
z 1
H lz] =z - 0 .4 = 1 - 0.4z-l
and H (ein] = 1 = 1
1- 0.4e-i 0 1 - 0.4cos n- jsin Q
Therefore
1 ~ V 1 0
IH [e ·nl I =vHH*= i = J l.16 - i 0.8cosn
1- 0.4e- i l - 0.4ei 0
251
and
LH[ejn] =-tan
· -i( 1 -0.4sinf1)
0.4cos n
(c)
2
H lz] = 3z - l.8z
z2 -z+0.16
LH[~-·irl] =tm
_1 (3sin2D-l.8sinD) -~n
_1 ( sin2f2-sinD) .
3cos2D - l.8cosn cos2D-cosD+0.16 ,.
5.5-2. (a)
0.5 2 2 0.5 1
[ l =l+-+-+-+-+-
Hz
z z2 z3 z4 z5
. ·50 [ 2 cos -D
2e-12 + -1 cos -3D +cos -5D]
2 2 2 2
Therefore · I
jH (eJ 0 ] j = 4cos n + cos 3D + 2cos T
2
5DI
2
and LH (eJ 0
] = -2.5D
(b) Using the same procedure as in ProbS.45a, we obtain:
252
and
H (z] = ~ [z4 + z3 + z2 + z
5 z4
+ l]
= ~ [ei40
5
+ ei30 + e;2n + e;n +
ei40
1]
~e-j2n [e;2n + e;n + 1 + e-;n + e-;20)
(ii). (i - ~ ) Y[z]
09
= X[z]
Ilence the transfer functions of these filters are
(i)
jl!
H [eill] = e
ei11 +0.9
= l+0.9e-ill
1 = 1
l+0.9cosf2-j0.9sin11
I
H [eill] I = Jl.81 + 1l.8cosn ' L H [eill] = - Lan - I [ - 0.9 sin n ]
1 + 0.9cosn
(ii)
·n
H [eill] = . e' = 1 . = l
eJ 11 - 0.9 1 - 0.9e-in 1 - 0.9cosn + j0.9sinfl
IH[ejll]J = 1
JL81- l.8cosn'
L H[-;n]
~ = - tan -i [ 0.9 sinfl]
1- 0.9cosn
Filter(i) has a zero at the origin and a pole at -0.9. Because the only pole is
near n == 1T(z = -1), this is a highpass filter, as verified from the frequency
response shown in Figure S5.5-4a.
Filter(ii) has a zero at the origin and a pole at 0.9. Because the only pole is near
n = O(z = 1), this is a lowpass filter, as verified from the frequency response
shown in Figure S5.5-4b.
253
For n = 0.997r
1
H [ei0·99.,,.] = = 9 58
./1.81 + 1.8 cos 0.997r '.
(ii) For n = 0.0h
1
H (ei0 ·0 I-rr] = = 9.58
,/1.81 - 1.8 cos 0.0h
For n = 0.997r
1
H (ei0·99"") = . = 0.5966
vl.81 - 1.8 cos o.997r
Filter(i) gain at Do is
IHI= 1
vl.81 - 1.8 cos no
Filter(ii) gain at 7r - n0 is
IHI= 1 = 1
.jl.81 - 1.8 cos( 7r - no) .jL81 + 1.8 cos( Do)
I HI
,o
0.51.
-'TI
Figure S5.5-4
5.5-5.
Hlz] = z+0.8
z -0.5
(a)
ein+o.8 _ (cosn+0.8)+jsinD
H [einJ
eifl-Q.5 - (cosn-0.5)+jsinD
(ein + 0.8)( e-jfl + 0.8) 1.64 + 1.6 cos n
IH (eiflJ 12 = H [eifl] H [e-Jn] (eifl - 0.5)(e-jll - 0.5) = 1.25 - cosn
LH [e1 n] t - 1 sin n - 1 ' sin n
an (cosn+o.8)-tan (cosn - 0.5)
(b) n=0.5
254
0.2784 - 0.9037 = -0.6253 rad
Therefore
1T
y (nj = 2.86 cos(0.5n -
3 - 0.6253) = 2.86 cos(0.5n - 1.6725)
5.5-6.
A = P[z)
(z - ')'1 )(z - ')'2) · · · (z - 'YN )(z - eiO)
I
z=e;n
and
y[nj = [tC;')';n + [ei0]ej!'ln] u[n]
H
The sum on the right-hand side consists of N characteristic modes the system. For
an asymptotically stable system b;I < 1 (i = 1, 2, · · ·, N) and the sum on the right-
hand side vanishes as n -+ oo. This sum is therefore the transient component of the
response. The last terms H [&0 ] ei 0 n, which does not vanish as n -+ oo, is the steady-
state component of t he response Yss [n]:
255
5.5-7. (a) n= 0.87T is in the fundamental range. Hence it is the apparent frequency.
{b) l.27T = 27T - 0.87T. Hence
cos(6.9n + 0) = cos{0.6168n + 8)
(e)
. (~) _ sin(7rn/2)
smc 2 - ( 1Tn/2)
(g)
5.5-8. Because 1.4r. = 2r. - 0.67T, cos{l.47rn + ~) = cos(- 0.61Tn + ~) = cos{0.67Tn - V· Also
cos ( 0.67rn +
7T) = cos 0.67Tn cos
7T - sm
. 0.67rn sin '6
1T
=
J3 cos 0.61Tn - Z
T I .
sm 0.67Tn
6 6
Similarly
cos ( 0.61Tn - i) =cos 0.67rn cos i +sin 0.61Tn sin i = ~cos 0.61Tn + v; sin 0.67rn
Therefore,
cos ( 0.61Tn + ~) +J3 cos ( l.47Tn + i) = J3 cos 0.611n+sin 0.67rn = 2 cos ( 0.67rn - i)
5.5-9. {a) fh = A= 5~~2 = 10 kHz.
256
{b) fs 2:: 2fh = 100
5.5-10. Taking t he z-transform of y[nJ
kHz, T $. * = lOµs .
II
~li.
OJ
..
o'--~..__~..__~'--~'--~"--~....._~_.___,
0
0
1
Figure S5.5-10a: jH(e'0 )j = v'S/4-cos(O) .
{b) Using H(z) and letting z = e30 , the phase response is LH (e30 ) = L 1 _ 0 _~., - ,u =
- L( l - 0.5 cos(fl) - 0.53sin(!1)). Thus,
n ( - 0.5 sin(n) )
LH (e' ) = -arctan l - O.Scos(fl) .
257
I•>
I
'fontname','symbol');
.
>•I
);' 0 .
'.' I•>
t '.'
' .
J.' p '], .. .
OJ .•••
r ...
".........
z
-OJ •
-OJ .
•
. 85 .5- lOb: L H (e 1n) = - arctan (
" -O.Ssin(fl)
F 1gure i- o.scos(fl) ) .
VL- 0 ,,rz_
lb )
-r
Figure SS.6-1
5.6-1. Figure 85.6-1 shows a rough sketch of the amplitude and phase response of this filter.
For the case (a), the poles are in the vicinity of n = % · Therefore, the gain IH[ei0 ]1
is high in the vicinity of n = 7r/4. In the case (b), the poles are in the vicinity of
n = 7r. therefore, the gain IH!ei0 JI is high in the vicinity of = lT . For case (a), the n
phases of the two poles are equal and opposite at n = 0. Hence LH!ei0 J st.arts at 0
258
(for n = 0). As n increases, the angle due to both poles increase. Hence, LH[ei 0 ]
increases in negative direction until it reaches the value - 21T at n = 1T. For case (b),
similar behavior is observed. Note that angle - 271" is the same as 0.
5.6-2. The two systems are very similar and have identical steady-state characteristics. There
is an important difference, however, between t he two systems. The system yin] -
y[n - l ] = xln] - x[n - l] is first-order and can support an initial condition; the
system y[n] = xln] is zero-order and cannot support an initial condition. If the initial
condition of the first system is non-zero, the output of the two systems can be quite
d ifferent.
5.6-3. (a) From the magnitude response plot, it is clear that this is a lowpass filter. Low
frequencies near n = 0 are attenuated, and high frequencies near n = ±7T are
passed with unity gain.
(b) From the magnitude and phase response plots, H(e1"1 2 ) = 72e13"1 2 . Thus, the
output to xi[n] = 2 sin( ~n + D is
(c) Notice, H(e'1 .,.14 ) = H(e-Pf 4 ). From the magnitude and phase response plots,
H (e-'"'1 4 ) ~ 0.07l e'2 .43 . Thus, the output to x2lnJ = cos( 74"n) is
71T
Y2[n) = 0.071cos( 4 n + 2.43).
HlzJ +1
= Kzz-a
Figure S5.6-6a shows the realization, Figure S5.6-6b shows the pole-zero configuration,
and Figure S5.6-6c shows the amplitude response of the filter. Observe that the pole
at a is close ton = 0. Hence, there is the highest gain at de. There is a zero at -1,
which represents n = 71". Hence, the gain is zero at S1 = 1T. This is a lowpass filter.
H [e;n] = K (e;n +
ein - a
1) 1
= K ( cos n + + ~s~nn)
cos n - a + J sm n
2 1
3.125K2 =K2 [ ( + cosn) ] ==} !1 = l.176
i.04 - o.4 cos n
259
S0L....V1'10N
s . tvJ- j_
Figure S5.6-4
Hence
5.6-7.
1 I
T '.S 2Bh = 40000 = 25µs
Select T = 25µs Frequency 5000 Hz gives
H[z] =
H [I) _J_<(_2_-_J_
22~) = 1
l + a2 - J2a
260
Figure 85.6-5
Figure SS.6-6
2
1
]{ + a -,:a= 1.707(1 + a 2 - ha)
2- 2
2 (ei2n - ..fi.ejn + l)(e-;2n - ..fi.e-;n + 1)
= K (ei2n _ ..fi.aeifl + a2)(e-;2n _ ..fi.ae-ifl + a2)
= K2 4 - 4../i.cosn + 2cos2n
(1 + a2)2 - 2../i.a(l + a2) cos n + 2a2 cos 2n
5.6-8.
I
This shows that the amplitude response is constant (j H [&11 ) = ~) for all values of
D. The filter is an allpass filter. This result can be generalized exactly the same way
for complex poles and zeros.
261
5.6-9. (a)
h2[n) = (- l )"hi[n)
e±j1rn h1[n]
Use of the property in Eq. (5.20) with/= e±i... , we obtain H 2 [zJ = Hi[ei±"zj.
Hence
(b) Figure S5.6-9 shows the frequency response of an ideal lowpass filter with cutoff
frequency De. Figure S5.6-9 also shows t he same frequency response shifted by
7r (with 27r-periodicity). It is clear that the shifted response corresponds to an
ideal high-pass filter with cutoff frequency 7r - .Ile.
5.6-10. (a) Let H(z) represent the original fi lter, either highpass or lowpass. The t ransformed
filter has system function HT(z) = H(z)lz=-z = H(-z) . The basic character
of the transformed filter can be assessed by its magnitude response, IHT(eJn)! =
IH( -e111 )1 = IH(&" e111 )1 = IH(e1 <n+rr)) I. That is, the magnitude response of a
transformed filter is just the magnitude response of the original filter shifted in
frequency by n. If the magnitude response of a digital LPF is shifted in frequency
by n, it becomes a highpass fi lter. Similarly, if the magnitude response of a digital
HPF is shifted in frequency by 7r, it becomes a lowpass filter. Put another way,
an original passband centered at .11 = 0 (LPF) is shifted by the transformation
to a passband centered at .11 = 7r (HPF), and vice-versa.
(b) If H(z) = L~=-ooh[nJz - n is the original fi lter, HT(z) = H(-z) =
L~=-oo hlnJ(-z) - n = L~= - 00 (-l)"h[nJz-n . Thus, the transformed filter has
impulse response
hT[nJ = (-1rh[nJ.
Put another way, the same transformation is accomplished by simply negating
the values of h(n) for every odd integer n.
r _
5. 6-11. Tlie b iIinear t rans1ormation states s -
2( 1-r -
T(l +z
1)
, )'
Rearranging yields z l + Ts/2
1-Ts/2"
l+;wT/2
Th us, s = JW maps to z = i-;wr; 2 ·
262
bilinear transformation maps (-oo :S w :S oo) to (-7r/2 ::; U7r/2) in a monotonic,
although non-linear, manner according ton= Lz = L ~~;:::~~; = arctan(wT/2)-
arctan( -wT/2) = 2 arctan(wT/2).
5.7-1. For the system in Eq. (3.15a), the transfer function H(s) is given by
1
H(s)=-
s+c
and
Using impulse invariance criterion, we obtain the equivalent digital filter transfer func-
tion from Table 5.3 corresponding to H(s) = as s!c
Hz _ Tz ,...., Tz assuming T --+ 0
[ ] - z - e-ct - z - (1 - cT)
- = -f3z-
H[z]
z+a
Substituting f3 = 1Jcr and a= 1 ~;1., we obtain
Tz
il[zJ= ~
z- l+cT
7s + 20 7s+20
H s _
a()- 2(s2+7s+10) ------ = --
1
+ -5/2
-
2(s + 2)(s + 5) s+2 s+5
Using Table 12.1, we get
H[zJ=T[ z
z - e- 2 r
+~-z~]
2 z - e-5T
First we select T:
7
Ha(O) =I and for s » 5 ==> Ha(s) ~
25
7
IH.. (jwo)I is 1% of IHa(O)j. Hence - = 0.01 and wo = 350, T = _!!_
2wo 350
263
Substituting this value of T in H[zJ yields
z 5 z ] [ z - 0.97474 ]
H(z] = 0.008976 [ z - 0.9822 + 2z - 0.9561 = 0.03 l'1lfiz z 2 - l.9383z + 0.9391
1) Canonical realization:
'
'f t
, n 3 'i)..3.
•
j
]~ I..._'. t. . ._r-_
~ "I V)Or\
'
IC-,
~J
~-------------4 O· 3q -I
Figure S5.7-2
H(zJ =[
z2 -
hTze-Tf-/2 sin ( r )
?2
2ze-T/-/2 cos ( ~) + e-hT
l
We now select T:
. 1
H 4 (0) = 1 and for highs, Ha (s) ~ 2
s
and JH0 (jw)I ~ w~ for high w
264
For negligible aliasing, we select the frequency w 0 to be that where
1
jH,.(jwo)I is 13 of IHa(O) j. Hence 2 = 0.01 . and w0 = 10
WO
1T
and T = Wo
- = 7r/l0
Substitution of this value of T in H[zj yields
0.0784z
H [z] = z 2 - l.5622z + 0.6413
Tz
H[z]= - and
z -1
Therefore,
T
--- .,IHo..\
- -- -
Figure 85.7-4
5.7-5. (a) Because an oscillator output is basically a system output with no input, a system
with zero-input response of the form sird1on (or cos(S1on + 0) with any value
of 0), where Do = woT will serve as the desired oscillator. A marginally stable
system with impulse response of the above form is a candidate. From Table 12.1,
265
pair llb, we see that a transfer function
H[z) = zsinwoT
z2 - 2z coswoT +I
has an impulse response (or zero-input response) of the form sin Don (Do = w0 T).
The period of the sinusoid is To = 27T / fl0 , and there are LO samples in each cycle.
Therefore, the sampling interval T = T 0 /10 = ?r/50.0 , and floT = rr/5. hence,
zsin {~)
H[z]
z 2 -2cos(i)z+ l
0.5878z
= z - l.618z + l
2
This is one possible solut ion. By varying the phase in the impulse response, we
could obtain variations of t his t ransfer function.
(b) Another approach is to consider an analog system wit h transfer function H 0 (s)
such that its impulse response ( or zero-input response) is of the form sin w0 t [or
cos(w0 t + 8) for any value of OJ. From Table 6.1 , pair 8b we find
wo
Ho(s) = -s 2- -
+w2
0
Now using Table 12.1 , we find the corresponding digital filter using impulse in-
variance method as
"
H [z] = TzsinwoT
z2 - 2zcoswoT + l
Earlier we found that w 0T = 1T/5. Because wo = 27T(l0,000) = 20,000rr, the
period T 0 = 10- 4 . There are 10 samples in each period. Hence the sampling
interval T = 10-5, and
-s 0.5878z
HIz I -- 10
z 2 - l.618z+ l
This is identical to t he answer in (a) except for an amplitude scaling by 10-s.
Because, we did not specify any amplitude requirement on the oscillator, different
answers will differ by a constant multiplier. This is a marginally stable system
and will oscillate without input with the response of the form
H0 (s) = -We-
s+wc
266
·1~
Figure S5.7-5
g[nj = z- 1 {z-
-z-H(zj }
1
Since gjnj = g,, {nT),
z
- H[zJ
z-1
z z
= - - - - ---
z - 1 z - e-w.T (z - l)(z - e-w.T)
Therefore
(b)
For Ha (s)= ~
S +we
z- 1 T
H[zJ = -Z[nTu[nlJ = -z-1
z
and
267
Hence,
7r
lwl::; T
IHo(jw) I = ~
w
Observe that this amplitude response is identical to that found by the impulse
invariance method in Prob. 5.7-4. He11ce, this amplitude response and the ideal
integrator amplitude response are the same as those in Figure S5.7-4. The only
difference between the answers obtained by these methods is that the phase re-
sponse of the step invariance response differs from that of the impulse invariance
method by a. constant ~.
5.7-7. (a) For a differentiator
Ha (s) =s
The unit ramp response r(t) is given by
1
r(t) = C- 1 F(s)H0 (s) = c- 1 2(s) = u(t)
s
Now we must design H[zJ such that its response to input nTu[nJ is u[nJ, that is
Now we design H[z] such that its response to nTu[nJ is !n2 T 2 u[nJ, that is
or
T 2 z(z + 1) Tz
2(z - 1)3 = (z - 1)2 H[zJ
Hence
H[z] = !_ ( z+ l)
2 z - 1
268
5.7-8.
If>.; = o; + j{3;, then e>-•T = e 0 •T efl3'T. When >.; is in LHP, a; < 0 and le>.'TI =
e0 •T < 1. Hence if >.; is in LHP, the corresponding pole of H[zj is within the unit
circle. Clearly if Ha(s) is stable, the corresponding H[zj is also stable.
5.7-9. {a) The w-axis is given by s = 3w. Rewriting the transformation s - 1 -.;.- · as
z = 1 _ 1sr and substituting s = JW yields z = i -~r· Thus, we need to show that
z = i-~r describes a circle centered at (1/2,0) with a radius of 1/2.
For a complex variable z, the equation lz - 1/212 = (1/2) 2 describes a circle
centered at (1/2,0) with a radius of 1/2. The transformation rule z = I-~wT is
substituted into this expression.
Since the equation is satisfied, the transformation rule z = l -.!w1 . maps thew-axis
to a circle centered at (1/2, 0) with a radius of 1/2.
Notice that <lifferent values of w can map to the same value of z (aliasing), which
makes an inverse transformation very problematic.
1
(b) First, rewrite the transformations= -;.- • as z = 1
_\r Next, notice
lzl 2 = zz•
l 1
I- a?' l-s·T
I
l-8T-s"T+ss •T2
I
= l-(0+1..,)T- (o-;w)T+(o'l+w2)r2
I
For a < 0, the denominator 1 - 2aT + (a 2 + w 2 )T2 > 1 and lzl 2 < 1. That is,
the left-half plane of s (u < 0) is guaranteed to map to the interior of the unit
circle in the z-plane.
5.9-1. (a)
z
z- 0.8
izl > 0.8
-z
X2 [nj ¢::::>
z- 2
lzl < 2
269
Hence
z z
X[zJ - ---
z-0.8
-
z-2
0.8 < lzl < 2
-l.2z
= 2 0.8 < lzl < 2
z - 2.8z + 1.6
(b)
z
X 1 [z) = - -
z-2
lzl > 2
z
X2[zJ - -
z-3
lzl <3
Hence
z z
X[z] = --+ --
z-2 z-3
2 < lzl < 3
z(2z - 5)
z 2 - 5z + 6
2 < lzl < 3
(c)
z
X i[zJ lzl > 0.8
z - 0.8
-z
lzl < 0.9
z - 0.9
z z
Hence XfzJ
z - 0.8 - z - 0.9
-z
0.8 < lzl < 0.9
lO(z 2 - l.7z + 0 .72)
(d)
(e)
z 3z
!(0.8)" + 3(0.4)") u. lnJ <=> - - + -- lzl > 0.8
z - 0.8 z - 0.4
4z(z - 0.7)
= (z - 0.4)(z - 0.8)
lzl > 0.8
(f)
270
(g)
(h)
and X[z]- z - _ z_
- z - e- 2 z - 2
(a) The region of convergence is lzl > 2. Both terms are causal. And
(b) The region of convergence is e- 2 < lzl < 2. In this case the 1st term is causal
and the second is anticausal.
(c) The region of convergence is lzl < e-2 . Both terms are anticausal in this case.
271
>> plot(real(zz),imag(zz),'ko',real(zp),imag(zp),'kx');
>> xlabel('Re(z)'); ylabel('Im(z)');
>> axis([- 1.5 1.5 -1 .5 1.5]); axis equal; grid;
u·~-~-~--.,.--.....---..--..----.--.
: 0
: 0
·ts•--~-__,:---~-....,.....-~--..,----7--~
·U -0..S O.S l.S
""<•>
z(z-!)
F igure S5.9-4a: Pole-zero plot for H ( z ) = (z3 - ~).
There are lwo possible regions of convergence, both of which exclude the thr~e
system poles: Jzl < 3/2 or lzl > 3/2.
(b) The poles and zeros of H- 1 (z) ~re just the zeros and poles, respectively, of fJ (z ).
Thus, lhe three zeros of H(z) satisfy z 3 = ¥,or z = 3/2eJ2 rrk/ 3 fork = (O, 1, 2).
There are two finite poles at z 0 and z =
1/2 as well as a pole at infinity. =
MATLAB is used to create the corresponding pole-zero plot.
>> k = [0:2]; zz = 3/2*exp(j•2•pi•k/3); zp = [0,1/2];
>> plot(real(zz),imag(zz),'ko',real(zp),imag(zp),'kx');
>> xlabel('Re(z)'); ylabel('Im(z)');
>> axis([- 1 .5 1.5 - 1.5 1.5)); axis equal; grid;
05 . . ~»
r o 0 . 0
.
..,
.,
· 1 ~ .~5-~.,~- ..,·~,...---~-~ .•
. ~.-~--.~
$ -~~
y;.
R.io
~"'( z-_
3
Figure S5.9-4b: Pole-zero plot for f:l- 1 ( z) =
There are two possible regions of convergence, both of which exclude the th ree
system poles: 0 < lzl < l / 2 or 1/2 < Jzl < oo.
272
5.9-5. It is known that x[nJ has a mode (1/2)n and that x 1 [n] = (1/3)nx[nJ is absolutely
summable. For this to be true, the mode at (1/2r must be right-sided. That is,
(1/2)"(1/:l)nu[nJ is absolutely summable but (1/2)"(1/3)nu[-n) is not. It is also
known that x2 [n] = (1/4)"x[n] is not absolutely summable. For this to be true, there
must be a pole somewhere in the annulus 3 < lzl < 4 that corresponds to a left-sided
signal; such a mode when multiplied by (1/3)" is still absolutely summable but when
multiplied by (1/4)" is not absolutely summable. Thus,
273
(b) X2(z) = L::=-oo x2[nJz- n L::=_ 00 (J)" cos(n + l)u[nJ =
L:::oJn0.5(&(n+l)+e- jmath(n+l))z-n = r::=o0.5e'(lff +
o.se-J
X2(z) = 1 - 0.5e' +
J&z- 1 1 - Je-iz- 1
; ROC lzl > 1.
(c) X3(z) = L::=-oo X3(n]z-n = L:: =-oo (J0.5(e" - e-n)uj- n + 1]) z-n =
L:~=-ooJ0.5 { (~)" - (~)" } . For lzl < e and izl < e- 1 , this becomes X3(z) =
2 2
2. { O- (e/z) _ O-(ez)- } Th S
2 1 - e/z. 1-(ez.)-l · U >
J Z-2 { e- 2 e2 }
X3(z) = -2- 1 - (ez)-1 + 1 - ez - l ; ROC lzl < e- 1.
4, this becomes
0 - (-4z- 2 ) 4z- 2
X4(z) = 1-
(-4z- 2 ) =
1 + 4 z- 2
; ROC lz l < 2.
Note: lz2 1 = lzl 2 , so the region of convergence is lzl 2 < 4 or lzl < 2.
5.9-10. (a) The signal X 1 (z) = 1+.!jz •+}z- •-kz 3 has three fu1ite poles, of which at least
one must be real. Using the region of convergence, we know that a real root
must be either 0.5, - 0.5, 2, or -2. Trying these values, we find that z = - 0.5
and z = -2 are both roots of the denominator. The remaining root can be
found by noting that the denominator 1 + .!jz- 1 + iz- 2 - ~z- 3 must be equal to
(l+O.sz- 1 )(1 +2z- 1 )(l+Az- 1 ). Equating the power of z -· 3 on each side yields
-1/3 =A. Thus, X1(z) = (l +o.sz- l)(i+~z-•)(i-z-1/J)" ExpandingyicldsX1(z) =
- I /5 + 1+8/7
i+o.sz-l z-1 + ~-2/35
2
usmg
. t h e region
. o f convergence (0.5 < Iz I < 2) an d
tables, the inverse is
-~ (-1/2)"u[n] - ~(-2)"u[-n -
2
xi[nJ = lJ + 35 (I/3)"u[n].
3
(b) X2(z) = , - il(2-z - 1)(1+2z 1) = z /2(1 - z- •;2)(1+2z- I) =
z 3 /2 ( t=%'=172
1/5
+ 4/ 5 ) .
i+'2?T Using the region of convergence (0.5 < lzl < 2)
and tables, the inverse is
274
(b) H 2 (Z ) -_ Z -3 ( (z-2)(z
z+1
tl/2)
) _
- z
-3 ~)
z-2 + z+l/2 - Z
(fil
- 4 (M -z/s )
z-2 + z+l/2 ·
s·mce _
h2[n] is stable, the region of convergence for H 2 (z) must be 1/2 < lzl < 2. Using
this ROC and z-transform tables, the inverse transform is
1
H(z) = - N; ROC lzl > l.
1 -z
Notice, H(z) has N poles that correspond to the N roots of unity. That is, the poles
of H(z) satisfy z = e1 2rrk/N fork = (0, 1, ... , N - 1).
5.9-13. For causal signals, the region of convergence may be ignored. We shall consider it only
for noncausal inputs
(a)
z2
y lzJ = X !zJ H lz] = (z - e)(z + 0.2)(z - 0.8)
-z2
y lzJ = (z + 0.2)(z - 0.8)(z - 2) 0.8 < lzl < 2
and
Y !zJ -z 1/11 2/3 0.758
-z- = (z + 0.2)(z - 0.8)(z - 2) = z + 0.2 + z - 0.8 - z - 2
1 z 2 z z
Therefore Y [z] = ii z + 0 _2 + 3 z _ 0 _8 - 0.758 z _
2
0.8 < lzl < 2
(c) The input in this case is the sum of the inputs in parts a and b hence the response
will be the sum of the responses in part a and b.
275
5.9-14.
x [nJ ='--...r-'
2"u {n] + u [-(n + 1)]
,.___.,
:q [n] x2(nl
z
z-2
Jzl > 2
-z
z-1
lzl < 1
There is no region of convergence common to X 1 [zJ and X 2 [zJ
H[zJ- z
- (z + 0.2)(z - 0.8)
The region of convergence of H [z ] is lzl > 0.8 (assuming a causal system). We should
find the response to x 1 [n] and x2 [nJ separately.
z2
Yi [zJ = (z - 2)(z + 0.2)(z - 0.8) izl > 2
Ydz] = - .-
.I
-z - - -
2
-z - +0.758--
11 z + 0.2 3 z - 0.8
z
z - 2
and
Yi [n] = [-_!_(-0.2)n - ~(0.8)" + 0.758(2)n] u [n]
11 3
Similarly
-25 z 1 z z
Y2 [zJ = - - - + - - - +4- - 0.8 < lzl < 1
6 z- 1 6 z + 0.2 z - 0.8
and
and
5 10 ] 25
y [nJ = Y1 [nJ + Y2 [n] = [ 66 (-0.2)n + 3 (0.8)" + 0.758(2)n ti [nJ + 6 u (-(n + l)J
5.9-15.
-z
X[z]=--2
z-e
lzl < e2
and
H[z] = z lzl > 0.8
(z + 0.2)(z - 0.8)
No common region of convergence for X [z] and H [z] exists. Hence
y[nJ = oo
5.M-1. Taking the z-transforrn of 4y[n + 2] - y[n] = x[n + 2] + x[nJ y ields Y(z) ( 4z 2 - 1) =
276
X(z) (z2 + 1). Thus, the system function is H(z ) = ~~:~ = ,;;,-t:._11 = 0.25 1 ~~:~j4 •
(a) MATLAB is used to create the pole-zero diagram.
>> zz = roots ([! 0 1]); zp = roots((4 0 - 1));
>>theta= linspace(0 ,2•pi,201);
>> plot(real(zz),imag(zz),'ko',real(zp),imag(zp),'kx', .. .
cos(theta),sin(theta),'k');
>> xlabel('Re(z)'); ylabel('Im(z)'); grid;
>> axis((-1.1 1.1 -1.1 1.1) ); axis equal ;
04 ....... .....~.
02 . ' ' .
f 0 •. . .. ! . .. . ..... :............ •
-....
""°. ..................
··~·· · ··· ·· ·
-· _, -0.$ 0 o.s
n-ci>
2
Figure S5.M-la: Pole-zero diagram for H(z) = 2
42
/_\.
2
(b) H(e10 ) = 4', 2 -t;_\ lz.=e'n · MATLAB is used to plot the magnitude response
IH(e'n)I.
>>Omega = linspace(-pi,pi,501) ;
>> z = exp(j•Omega ) ; H = (z .-2+1)./(4•z. -2-1) ;
>> plot(Omega,abs(H),'k'); axis([- pi pi 0 1]); grid;
>> xlabel('\Omega'); ylabel( ' IH(e-{j\Omega})J ' );
>> set(gca,'xtick', [-pi:pi/4:pi],'xticklabel' ,('-p '; . . .
';'
'fontnamc','symbol ');
,;' ';, 0 ,. . , .,
) . ) .,
)
, ; ) p ') J •••
(c) The pole-zero plot of Figure S5.M-la and the magnitude response plot of Figure
S5.M-lb confirm that this is a band-stop system.
(d) Yes, the system is asymptotically stable. Referring to Figure S5.M-la, all t he
system poles are within the unit circle.
(e) Yes, the system is real. Since the system is expressed as a constant-coefficient
linear difference equation with real coefficient.s, the impulse response h.[nJ and
system are both real.
(f) For an input of the form x[n] = cos(51n), the greatest possible amplitude of the
output. corresponds to the greatest gain shown in the magnitude response plot of
Figure S5.M- l b. Thus, 2/3 is the greatest output amplitude given an input of
=
x[n] cos(51n). This output amplitude occurs when Q = ktr, for any integer k.
277
Of • • ... ·
... .
0.7 ... :· ..... ":·· ........: ..... ..
\o., .
~
.... ..
...
0.1
-
o'--~-'--~-"-~-'-~--L-~-'-~-"-~-'"~--'
Figure S5.M-lb: Magnitude response plot for 4y(n + 2} - y(n] = x[n + 2] + x(n].
1-----.--:t~ y[nJ
x[nJ -----~
1/4
1/4 - 1/4
(g) Inverting H(z) = ~~:~ = 0.25 1 :~:;; 4 provides y[n] - 0.25y[n - 2j = 0.25x[n] +
0.25x[n - 2j, which is a convenient form for implementation. Figure S5.M-lg
illustrates a TDFII implementation of the system.
(z-e'•l •)(: - e-'3•/•) Tl h d
5 .1M- 2 · ( a ) H( z ) = •- 0z2 -1
. 9 e,3• / • = z-O.!JeJ:S•t• . ms, t ere are zeros at z = e
3tr/4
an
z = e-13 .,,/4 , and there are poles al z = 0.9e'3"/4 and z = oo.
>> zz = roots((l 0 -j)); zp = 0.9*exp(j*3*pi/4);
>>theta= linspace(0,2*pi,201);
>> plot(real(zz),imag(zz),'ko', r eal(zp),imag(zp),'kx', ...
cos(theta), sin(theta) ,'k');
>> xlabel('Re(z)'); ylabel( 'Im(z)');
>> axis((-1.1 1.1 -1.1 1.1) ); axis equal; grid;
Since this is a complex system, poles and zeros need not occur in c~mplex con-
jugate pairs.
2
(b) Substituting z = &0 into H(z) = z-o~ 9 ;,t 7 ., IV1ATLAB is used to create the
magnitude response plot.
>>Omega= linspace(-2*pi,2*pi,1001); z = exp(j*Omega);
>> H = (z.-2-j)./(z-0.9*exp(j*3*pi/4));
>> plot(Omega,abs(H),'k'); axis([-2•pi 2*pi 0 21));
>> xlabel('\Omega'); ylabel('IH(e-{j\Omega})I'); grid;
278
...
••
j 0
.........:................ ·
...
-0.2 ·· ·: """""'".:
···-·· ....... .
;
~0.1 ······· ,, .......... .
_,
., ...a.5 0 0.5
~ti•)
l
Figure S5.M-2a: Pole-zero diagram for H{z) = .. _;, 9 ;,~ w/4.
II ,
.."
( "
l
..
10
"
5.M-3 . (a) H(z) = 2 (zf:o.~i,) has four finite zeros and two finite poles. The zeros' are the four
roots of unity, and the poles are at z = 0.9ePl 4 and z = 0.9e-1 3r./ 4 . MATLAB
is used to compute the pole-zero plot.
>> zz =roots([! 0 0 0 -1]); zp =roots([! 0 j•0.81]);
>>theta= linspace(0 ,2•pi,201);
>> plot(real(zz),imag(zz),'ko',real(zp),imag(zp),'kx', ...
cos(theta),sin(theta),'k');
279
>> xlabel('Re(z)'); ylabel('lm(z)');
>> axis( (-1.1 1.1 - 1.1 1.1) ) ; axis equal; grid;
1 .... ~-e--...:.'
· · ··· · · ·:""'
o.a ..
o.o
~ 0
-02
-0.4
-0.6
-0.0
-1
•1 -0.$ 0 0.$
Ao(>)
'fontname' , 'symbol');
(c) Since H(z) is an improper rational function and has poles at infinity, the system
is non-causal. By dividing the numerator by the denominator (in such a way as
to yield a right-sided impulse response), an improper rational function H(z ) will
y ield an impulse response h[nJ that is not wro for all negative n. Right-sided
280
functions with poles at infinity are sometimes called "not-quite causal" since they
are not causal only by a finite shift.
(d) Adding two poles at zero (a= b = 0) corresponds to a simple right-shift by two
and does not affect the magnitude response. That is, 1HcausaJ(e1°)l =ICe~1'~2) I=
1 ~e~T;~J 1
1 =1 )1. By adding two poles at zero to H( z ), Hcausa1(z) becomes
IH(e10
a proper rational function with only finite poles, and the corresponding impulse
response function becomes causal.
(e) Noting that Hcausa1 (z) = ~~:~ = J!~·;1~;, = ~;O.~i~~-;, the corresponding
difference equation is y[nJ + 0.8lJ1J[n - 2] = 0.5x[n] - 0.5x[n - 4]. Figure 85.M-3e
shows the corresponding TDFll implementation.
---~-- y[n]
112
0.8lj ..
-112
5.M-4. (a) From the block diagram, the corresponding difference equation is written directly.
(b) Taking the z-transform of y[n] - 0.5y[n-2] = x[n] yields Y(z){l-O.sz- 2 ) = X(z).
Thus, H(z) = ~~~~ = l-O.~z-•. Substituting z = eJ 0 into H(z), t he mag-
nitude response is IH(e1°)1 = Ii-o.s1e-,2n I = 1
y{l-0.5 cos(20))2+(-0.S sin(20)) 2
=
1
y 1-cos(20) +o. 25(cos• (2S1)+sin2 (211)) . Thus,
281
>> plot(Omega,Hmag,'k'); axis( [-pi pi 0 2.5)); grid;
>> xlabel('\Omega'); ylabel('fH(e~{j\Omega})I');
>> set(gca,'xtick',[-pi :pi/4:pi),'xticklabel', [' -p '; .. .
, ; , , ; > , ; 1 0 ';, ';' ';} , ; ' p '), .. .
'fontname','symbol');
u~----------------
"' . .... .
•
n
282
>> Hmag = 2*abs(cos(Omega));
>> plot(Omega,Hmag,'k'); axis((-pi pi 0 2.5)); grid;
>> xlabel('\Omega'); ylabel('IH(e-{j\Omega})I');
>> set(gca,'xtick',[-pi:pi/4:pi),'xticklabel' ,['-p '; ...
1 ;] I j) ) ; ) 0 ); ) J j J J; J ) ; J p >]I • ••
'fontname','symbol');
O.l ··• ,
-o'--~-'-~-l...~--''--~-'-~-'-~~~~-'-~~
0
0
Using Figure S5.M-6, this system is best described by a bandstop filter with gain.
That is, low and high frequencies have a boosted gain of two and middle frequencies
near n = 7r /2 are attenuated .
5.M-7. (a) » Omega = linspace(-pi,pi,501); z = exp(j*Omega);
>> H = cos(l./z);
>> plot(Omega,abs(H),'k'); axis((-pi pi 0 2]); grid;
>> xlabel('\Omega'); ylabel('IH(e-{j\Omega})I');
>> set(gca,'xtick',[-pi:pi/4:pi),'xticklabel',['-p '; .. .
,;, }; , ';, 0 ';, ';, ,;' ';, p ' ], .. .
'fontname','symbol ');
1.8 ....
...
0.2
-·
O '--~...L-~-'-~--'~~~~-'-~~~~~~~
0
0
283
From Figure S5.M-7a, it appears that this system behaves somewhat like a band-
pass filter. Frequencies near n = 7r /2 are boosted while frequencies near n = k1r
are somewhat attenuated.
(b) A Maclaurin series expansion of cos(x) is 2:;:':0 (( "':{.
2
Substituting n = z- 1
yields H (z) = cos(z- 1 ) = 2:;:':0 ~· Inverting yields
00
)k
htnJ = :L (-1 1
< k)' o[n - 2kJ = otnl - 1 atn -
2 . 2.
1 1
21+ 14. 01n - 4J - -61. a1n - 6J + ....
k =O
10
(c) Using Figure S5.M-7b, it is clear that the impulse response quickly decays to zero.
Thus, only the first few terms from h[nj are needed for a good approximation.
Note that h[n] ::::: o[n] - o[n - 2]/2 + o[n - 4]/24. An FIR difference equation is
found by letting h[n] = y[n] and o[n] = x[n]. That is,
This is a fourth-order FIR fi lter with only three non-zero coefficients. The mag-
nitude response is easily computed using MATLAB.
>> Omega = linspace(-pi,pi ,501); z = exp(j *Dmega);
>> H = 1 - z.-( - 2)/2 + z. -(-4)/24;
>> plot(Omega,abs(H),'k'); axis([-pi pi 0 2)); grid;
>> xlabel('\Omega'); ylabel('IH(e-{j\Omega})I');
>>
'.' .>•I
I
' f ontname','symbol');
, , .,
... .
set(gca,'xtick',[-pi :pi/4:pi),'xticklabel',('-p '; ...
1; .. p '] t •••
284
)JI .....
...
) I ..... .
O.• .• ,
o '--~-'-~-'-~-'-~-'-~-'-~-'-~--'-~--'
- 0
0
5.M-8 . Factored form is used to plot roots, and standard transfer function form is used to
compute magnitude response plots.
(a) Order-8 Butterworth LPF with De = Tr/3 .
>> Omega_c = pi/3; Omega= linspace(-pi,pi,1001); ~
>> [z,p,k) = butter(8,0mega_c/pi);
>> subplot(121),plot(real(p),imag(p),'kx', . . .
real(z),imag(z),'ko',cos(Omega),sin(Omega),'k');
>> axis([-1.1 1.1 -1.1 1 .1)); axis equal;
>> xlabel('Re(z)'); ylabel('Im(z)');
>> [B,A) = butter (8,0mega_c/pi);
>> HLP = polyval(B,exp(j *Omega))./polyval(A,exp(j*Omega));
>> subplot (122),plot(Omega,20*log10(abs(HLP)),'k');
>> axis([-pi pi -40 2)); grid;
>> xlabel('\Omega'); ylabel('IH_{LP}(e-{j\Omega})I');
>> set(gca,'xtick',[- pi:pi/3:pi),'xticklabel ', [' - p
';, ';, 0 ';' ';' , j, p '], .. .
'fontname','symbol');
(b) Order-8 Butterwort h HPF with nc= Tr/3 .
>> Omega_c = pi/3; Omega = linspace(-pi,pi,100 1);
>> [z,p,k) = butter(8,0mega_c/pi,'high');
>> subplot(121),plot(real(p) ,imag(p),'kx', . ..
r eal(z),imag(z),'ko',cos(Omega),sin(Omega),'k');
>> axis((-1.1 1.1 -1 .1 1 .1)); axis equal;
>> xlabel('Re(z)'); ylabel('Im(z)');
>> (B,A] = butter(8,0mega_c/pi,'high'); ,
>> HHP = polyval(B,exp(j*Omega))./polyval(A,exp(j*Omega));
>> subplot(122),plot(Omega,20*l og10(abs(HHP) ) ,' k');
>> axis((-pi pi - 40 2)); grid;
>> xlabel('\Omega' ); ylabel( ' IH_{HP}(e-{j\Omega})I');
>> set(gca ,'xtick ', [-pi:pi /3:pi],'xticklabel',['-p
';' ';' 0 , ; 1 1;, ';' p ']' . ..
' fontname','symbol');
285
...
....
-0.5 0 0.5 0
R1110:}
"
Figure S5.M-8a: Order-8 Butterworth LPF with flc = 7r/3.
1.5
_,, ..
{;._1 - lO
_,,
_,.,
- 1.5 _., ..
_, _..,
-0.5 0
A:eto
0 .5 ~
•
0
(c) Order-8 Butterworth BPF with passband between 5rr /24 and ll7r /24. Notice
that the command butter requires the parameter N = 3 to be used to obtain a
(2N = 8)-order bandpass filter .
>> Omega_c = [5*pi/24,ll*pi/24]; Omega; linspace(- pi,pi,1001);
>> [z,p,k] = butter(4,0mega_c/pi);
» subplot(121) ,plot (real(p) ,imag(p), 'kx', . . .
real(z),imag(z),'ko',cos(Omega),sin(Omega) , 'k');
>> axis([-1.1 1.1 -1. 1 1.1)); axis equal;
>> xlabel('Re(z)'); ylabel ('Im(z )');
>> (B,A] = butter(4,0mega_c/pi);
>> HBP = polyval(B,exp(j*Omega))./polyval(A,exp(j*Omega));
>> subplot(122),plot(Omega,20*1og10(abs(HBP)),'k');
>> axis( [-pi pi -40 2]); grid;
>> xlabel('\Omega'); ylabel('IH_{BP}(e~{j\Omega}) I ') ;
>> set(gca,'xtick' , [-pi:pi/3:pi] , 'xticklabel',[' - p '·
';' 1;' 0 '; ' ';' , ; 1 p '] ' . . .
'fontname','symbol');
286
1.$
-u
-1 --0.S 0 0 .5
R9(i)
Figure S5 ..M-8c: Order-8 Butterworth BPF with passband between 5rr/24 and 1l7r /24.
(d) Order-8 Butterworth BSF with stopband between 5r./24 and ll7r/24. Notice
that t he command butter requires the parameter N = 4 to be used to obtain a
(2N = 8)-order bandstop filter.
>> Omega_c = [5*pi/24,11*pi/24]; Omega= linspace(-pi,pi,1001);
>> [z,p,k] = butter(4,0mega_c/pi,'stop ' ); ~
>> subplot(121),plot(real(p),imag(p),'kx', . ..
real(z),imag(z) , 'ko',cos(Omega),sin(Omega),'k');
>> axi s((-1.1 1 . 1 -1 . 1 1 . 1)); axis equal;
>> xlabel('Re(z)'); ylabel('Im(z)');
>> [B,A) = butter(4,0mega_c/pi,'stop ' );
>> HBS = polyval(B,exp(j•Omega))./polyval(A,exp(j*Omega));
>> subplot(122),plot(Omega,20*log10(abs(HBS)),'k');
>> axis([-pi pi -40 2)); grid;
>> xlabel ('\Omega'); ylabel('IH_{BS}(e-{j\Omega})I');
>> set(gca,'xtick', [-pi:pi/3:pi] ,'xticklabel' ,[' - p
';, ';, 0 J;, , ;, ,; J p ']' ...
'fontname', ' symbol');
,.. _,
-10
- ll ...... . . . . .. .
- lO .
...
"'---~~-~-~
-o.s 0 05
Re(t) -·
-~~--~~~~~
0
0
Figure S5.M-8d: Order-8 Butterworth BSF with stopband between 5rr/24 and 1l1r /24.
287
5.M-9. Factored form is used to plot roots, and standard transfer function form is used to
compute magnit ude response plots.
(a) Order-8 Chebyshev Type I LPF with De= 7r/3.
>> Omega_c = pi/3; Omega= linspace(-pi,pi,1001);
>> [z,p,k) = cheby1(8,3,0mega_c/pi);
>> subplot(121),plot(real(p),imag(p),'kx', ...
real(z),imag(z),'ko',cos(Omega),sin(Omega),'k');
>> axis([-1.1 1.1 -1.1 1.1)); axis equal;
>> xlabel('Re(z)'); ylabel('Im(z)');
>> [B,A) = cheby1(8,3,0mega_c/pi);
>> HLP = polyval(B,exp(j*Omega))./polyval(A,exp(j*Omega));
>> subplot(122),plot(Omega,20*log10(abs(HLP)),'k');
>> axis([-pi pi -40 2)); grid;
>> xlabel('\Omega'); ylabel('IH_{LP}(e-{j\Omega})I');
>> set(gca,'xtick',[-pi:pi/3:pi),'xticklabel',['-p '; ...
,;' ';, 0 '; 1 ';, ';' p '], ...
'fontname','symbol');
...
.JO
-d .... . . . . . . .. . . .
_,, .
-JO
·15
....
•O.S 0
Rori:>
0.5
"
-· ~~-~~~~
0
288
-·
-10 ••
-u
..
.. -u
-· -JO
-u -» .....
_,
-· -05
•
Ro(z)
OS ~ .
-«>•'---'---"---'--'-' - ' - - - '
()
(c) Order-8 Chebyshev Type I BPF with passband between 57T/24 and l h/24. No-
tice that the command chebyl requires the parameter N = 4 to be used to obtain
a (2N = 8)-order bandpass filter.
>> Omega_c = [S*pi/24,11*pi/24]; Omega• linspace(-pi,pi,1001);
>> (z,p,k] = cheby1(4 ,3,0mega_c/pi); ~
>> subplot(121),plot(real(p),imag(p),'kx', . . .
real(z),imag(z),'ko',cos(Omega),sin(Omega),'k');
>> axis((-1.1 1.1 -1.1 1.1]); axis equal;
>> xlabel('Re(z)'); ylabel ('Im(z)');
>> (B,A] = cheby1(4,3,0mega_c/pi);
>> HBP = polyval(B,exp(j • Omega))./polyval(A,exp(j•Omega));
>> subplot(122),plot(Omega,20*log10(abs(HBP)),'k');
>> axis((-pi pi -40 2)) ; grid;
» xlabel ('\Omega'); ylabel (' IH_ {BP}(e-{j \Omega}) I ');
>> set(gca ,'xtick',[-pi:pi/3 :pi], 'xticklabel',( ' -p '; .. .
, ;, J;, 0 , ; ' '.' ';' p '), .. .
'fontna.me', 'symbol'); '
(d) Order-8 Chebyshev Type I BSF with stop band between 57T /24 and ll 7T/24. No-
t.ice t.hat the command cheby1 requires the parameter N = 4 to be used to obtain
a (2N = 8)-order bandstop filter.
>> Omega_c [S*pi/24,11*pi/24); Omega= linspace(- pi,pi,1001);
2
289
. .
_,
-··
- 1$ ..• •
:· ''; .. . ·:
.,. ...
. ....__~~-~--"'
-1 ~.s o o.
R.o(t)
.s
-
....,,.__........_._.._._.._._--'".....__.
Figure S5.M-9c: Order-8 Chebyshev Type I BPF with passband between 5n/24 and
0
n
l br /24.
'fontname' ,'symbol');
. _,
"
-··
·U
~z.•.. - 20
•U
_,.,
-1.$ _,, ..
_, _,
-1>.5
•
Re(>)
... ...- •n
Figure S5.M-9d: Order-8 Chebyshev Type I BSF with stopband between 57r/24 and
lln/24.
290
>> legend('R_p = 0.1','R_p = 1.0','R_p = 3.0',0);
>> set(gca,'xtick ' ,[-pi:pi/3:pi ] , 'xticklabel',['-p , ...
';, J;, 0 , ; ' ';, ';' p '], . ..
'fontname','symbol') ;
_, ................. .
S-,. '
r···
'
r :
_,,, 1::
i::
-).S •
..f•· ../ .. 'I . . ., ·•·· · •
. :•
. '
0
ll
Thus, reducing the allowable passband ripple Rp tends to broaden the transitioo
bands of the filter.
5.M-10. Factored form is used to plot roots, and standard transfer function form is used to
compute magnitude response plots.
(a) Order-8 Chebyshev Type II LPF with De = 7r / 3.
>> Omega_c = pi/3; Omega = linspace(-pi,pi,1001);
>> [z,p,k] = cheby2(8,20,0mega_c/pi);
>> subplot(121),plot(real(p),imag(p),'kx', ...
real (z),imag(z),'ko',cos(Omega),sin(Omega),'k');
>> axis([-1.1 1.1 -1.1 1.1)); axis equal;
>> xlabel('Re(z)'); ylabel('Im(z)');
>> [B,A] = cheby2(8,20,0mega_c/pi);
>> HLP = polyval(B,exp(j *Omega))./polyval(A,exp(j*Omega));
» subpl ot(122),plot(Omega,20*log10(abs(HLP)),'k') ;
>> axis((-pi pi -40 2)); grid;
>> xlabel('\Omega'); ylabel('IH_{LP}(e-{j\Omega})I');
>> set(gca,'xtick',(-pi:pi/3:pi),'xticklabel',('-p '·
, . , , ; ' 0 ';, ,. ' , ; , p ' ) ,. ..
' '
'fontname' ,'symbol');
291
• ~'
" _,
...
_.,
\;-111
- \ { ,. , .. I
.,, ...... ...
_,.
• 15 .,, ,..
.. ....__~~-:-:---~
-· ..o.s 0
A9(1)
o.s
....
1 0
n
.,.__
.,
...
_.,
_,,
...
-1.$ _,,
.. ...._~~~-~~
-0, S 0
Rt(:)
0$;
292
>> HBP = polyval(B,exp(j•Omega))./polyval(A,exp(j•Omega));
>> subplot(122),plot(Omega,20•log10(abs(HBP)),'k');
>> axis([-pi pi -40 2]); grid;
» xlabel( '\Omega'); ylabel(' IH_{BP}(e -{j \Omega}) I');
>> set(gca,'xtick', [-pi :pi/3:pi],•xticklabel', ['-p '; ...
,;, , ; , 0 ';, J; J >;, p '], ...
'fontname','symbol');
..···' ,.- ,-
_,
o.10 .•• .•:
}-~
[\~
0
·»
(
-JO
_,, ..
_..,
- 0
0
Figure S5.M-10c: Order-8 Chebyshev Type II BPF with passband between 57r/24 and
ll7r /24.
(d) Order-8 Chebyshev Type II BSF with stopband between 57r/24 and ll7r/24.
Notice that the command cheby2 requires the parameter N = 4 to be used to
obtain a (2N = 8)-order bandstop filler.
>> Omega_c = [5•pi/24,11•pi/24]; Omega = linspace(-pi,pi,1001);
>> [z,p,k) = cheby2(4,20,0mega_c/pi,'stop');
>> subplot(12 1),plot(real(p),imag(p),'kx', .. .
real(z),imag(z),'ko',cos(Omega),sin(Omega),'k');
>> axis((-1 .1 1.1 -1.1 1.1)); axis equal;
>> xlabel('Re(z)'); ylabel('lm(z)');
>> [B,A) = cheby2(4,20,0mega_c/pi,'stop');
>> HBS = polyval(B,exp(j•Omega))./polyval(A,exp(j•Omega));
>> subplot(122),plot(Omega,20•log10(abs(HBS)),'k');
>> axis([-pi pi -40 2)); grid;
>> xlabel('\Omega'); ylabel('IH_{BS}(e-{j\Omega})I');
>> set(gca,'xtick',[-pi:pi/3:pi],'xticklabel' ,['-p •·
.
,., ,;, 0 ,;' ,.
., ';, p '], ...
'fontname','symbol');
To demonstrate the effect of increasing R s. consider magnitude response plots
for Chebyshev Type II LPFs with Rs = {10, 20, 30}.
>> Omega_c = pi/3; Omega= linspace(-pi,pi,1001);
>> [B,A) = cheby2(8,10,0mega_c/pi);
>> HLP1 = polyval(B,exp(j•Omega))./polyval(A,exp(j•Omega));
>> [B,A) = cheby2(8,20,0mega_c/pi);
>> HLP2 = polyval(B,exp(j•Omega))./polyval(A,exp(j•Omega));
>> [B,A) = cheby2(8,30,0mega_c/pi);
293
1.S ·
-1.s
-·'-'--~~-----"'
-1 ~o.s o o.s 0
Ro(r} 0
Figure S5.M-10d: Order-8 Chebyshev Type II BSF with stopband between 57T /24 and
l h /24.
-10
o,
_, . m. 0
30
_,,
"' ' .
_,..
_,,,
.. '
·.
..
~ ,.
'
'
Ir •
. ::·:· · :·
.,'
1..
r,.
l ..
·.
.
I
'
I
1 1' :•
i f ..
' I
0
Q
294
>> [z,p,k) = ellip(S,3,20,0mega_c/pi);
>> subplot(121),plot(real(p),imag(p),'kx' , .. .
real(z),imag{z) ,'ko ',cos(Omega),sin(Omega) ,'k');
>> axis([-1.1 1.1 -1.1 1.1)); axis equal;
>> xlabel('Re(z)'); ylabel('Im(z)');
>> [B,A) = ellip(8,3,20,0mega_c/pi);
>> HLP = polyval(B,exp(j•Omega))./polyval(A,exp(j•Omega));
>> subplot(122),plot(Omega,20•log10(abs(HLP)),'k');
>> axis([-pi pi -40 2]); grid;
>> xlabel('\Omega'); ylabel('IH_{LP}(e-{j\Omega})I');
>> set(gca,'xtick',[-pi:pi/3:pi] ,'xticklabel',['-p '; ...
1;, J;, 0 , ; ' ,;, , ; ) p '], . ..
'fontna.me','symbol');
"
-··
-u
· t ...o.s o o_s
.......
~~~~~~
•
R.t(:) 0
'fontname','symbol');
0 ,;, .
'. J J;) p '], ...
(c) Ordcr-8 Elliptic BPF with passband between 5rr/24 and 1171"/24. Notice that the
command ellip requires the parameter N = 4 to be used to obtain a (2N = 8)-
order bandpass filter.
>> Omega_c ~ (5•pi/24,11•pi/24); Omega= linspace(-pi,pi,1001);
295
'·'
-J() •
-u
- JS · · · ·}· · ····
-2c_.__~~-~-~
-1 -05 0
Ro(>)
0.5
0
Q
nc = 7r/3.
"
·IS
- >S
"'-'---~~-~-~
-o ~ 0 0.5
Ro(z) -
-~~~~-~~~-
0
n
Figure 85.M-llc: Order-8 Ellipt ic BPF with passband between 57r/24 and ll7r/24.
(d) Order-8 Elliptic BSF with stopband between 57r/24 and ll1r/24. Notice that the
command ellip requires the parameter N = 4 to be used t.o obtain a (2N = 8)-
order bandstop filter.
>> Omega_c = (5*pi/24 ,11*pi/24); Omega = linspace(-pi,pi,1001);
296
>> [z,p,k) =. ellip(4,3,20,0mega_c/pi,'stop');
>> subplot(121),plot(real(p),imag(p),'kx' , . . .
real(z),imag(z),'ko';cos(Omega),sin(Omega),'k');
>> axis((-1 . 1 1 .1 -1 .1 1 . 1)); axis equal;
>> xlabel('Re(z)'); ylabel('Im(z)');
>> (B,A) = ellip(4,3,20,0mega_c/pi,'stop');
>> HBS = polyval(B,exp(j*Omega))./polyval(A,exp(j *Omega));
>> subplot(122),plot(Omega,20*log10(abs(HBS)),'k');
>> axis([- pi pi -40 2]); grid;
>> xlabel('\Omega'); ylabel('IH_{BS}(e-{j\Omega})I');
>> set(gca,'xtick' ,[- pi:pi/3:pi],'xticklabel',[' - p ' ·
,;) ';' 0 ';, '; > , ; , p ']' ...
'fontname','symbol');
t.5
-1.S
-JS .
....
-1 ..0,5 0
Ro{z)
o,s
-· ~~~~ ~~-~~~
0
0
Figure 85.M-lld: Order-8 Ellipt.ic BSF wit h stopband between 5n/24 and ll7r/24.
297
Chapter 6 Solutions
. ( b)
-f . ' -.. - ,_ - ,. w- ~· I
'.
J ri=,,;14>· l -D:-tir..i
· ·°'-
I 0 '- l
W' )11 II l. ~
I r
~ 5
, . ll...
6 ,.
( cl )
""f I>
<::>" "JO 2.IT
0 "!§
3
,,s
(f)
Figure S6.l- l
298
6.1-1. (a) To = 4, w 0 = ~~ = ~· Because of even symmetry, all sine terms are zero.
00
an = ~ [1
1
cos c~7T t) dt - /
2
cos c; t) dt] = n: n;
sin
4 ( 7rt 1 37rt
x(t) = - cos - - - cos - + -1 cos -57rt 1 7nt
- - cos - + · ··)
1r 2 3 2 5 2 7 2
Here bn = 0, and we allow Cn to take negative values. Figure 86.J -la shows the
plot of Cn.
(b) To = l07T, w0 = ~~ = ~· Because of even symmetry, all t he sine terms are zero.
00
an = - 112"
7T 0
- t cosntdt
21r
= 0, bn =-112rr -t sin ntdt = --
7r o 2 tr trn
1
and
299
0) because of odd symmetry.
bn = - 41"
1T 0
14
-4 tsin2ntdl
1T
= -2
1Tn
( -sm
2 . -1Tn - cos-
1Tn 2
1Tn)
2
4 1 . 8t
x(t) . 2t +
- sm -1 Sill
. 4t - -4, SIJI
. 6t - - Sill + ...
1T2 7r 97T2 27T
~
1T2
cos (2t - ~) + .!. cos (4t - ~) + ~2 cos (6t + ~) + .!. cos (st + ~) + · ..
2 7r 2 97T 2 7r 2
Figure S6.l-ld shows t he plot of Cn and Bn·
(e) To= 3, wo = 21T/3. 1
a0 =! 1 tdt=!
3 0 6
an= -
3
21 1
2n7T
tcos-tdt= -2-
3
3
21T n 2
27Tn 2rrn . 2?Tn
[cos-+-sm -
3 3 3
-lJ
21
b11 = -
3
0
0
1
2n7T
tsin - t d t
3
3
= --[sin
27T 2 n 2
2nn
-
3
-
2nn
-
3
2?rn
cos - ]
3
Therefore Co =~ and
and
~ cos 2>rn - sin 2rrn )
9n = tan - I 3 3 3
( cos 2 "n + '.hn sin 1!!! - 1
3 3 3
57T 7rr )
x(t) = 0.5 + 7T62 ( cos 3
n 2
t - 9cos1Tl +
l
25
cos 3t +
1
49
cos 3t + · · ·
Observe that even harmonics vanish. The reason is that if the de (0.5) is sub-
tracted from x(t), the resulting function has half-wave symmetry. (See Prob.
6.1-5). Figure S6.l-lf shows the plot of Cn.
6.1-2. (a) Here To = 1T, and w 0 = ~: = 2. T herefore
00
300
To compute the coefficients, we shall use the interval 1T to 0 for integration. Thus
a0 ..!..
11'
j 0
_,.
etf 2 dt = 0.504
~ 6n2 )
0
a,. ; j_ 1f
2
etf cos 2nt dt = 0.504 (
1+
b,. ; j_: 2
et/ sin 2ntdt = -0.504 (i + ~6n2 )
8
Therefore
Co = ao = 0.504
Ja~ + b~ = 0.504 (,;1+16n2 )
2
tan-
1
( ~:") = tan- 4n
1
2
x(t) = 0.504 + 0.504 L00
cos (2nt + tan- 14n)
n=l Vl + 16n 2
(b) This Fourier series is identical to that in Eq. (6.15a) with t replaced by -t.
(c) If x(t) =Co+ l:Cn cos(nw0 t + 0,.), then
Thus, time inversion of a signal merely changes the sign of the phase 0,.. Every-
thing else remains unchanged. Comparison of the above results in part (a) with
those in Example 6.1 confirms this conclusion.
6.1-3. (a) Here T 0 = 7r/2, and w0 = ~: = 4. Therefore
00
where
ao -
11'
21"/2 0
e- t dt = 0.504
and
b,. = ~n lo
f"
12
1
e- sin 4ntdt = 0.504 (
8
~ 2)
1+l n
301
(c) If x(t) =Co+ L: Cn cos(nwot +On), then
6.1-4. (a} Here To= 2, and w0 = ~ = 1T. Also x(t) is an even function oft. Therefore
00
Therefore
x(t) = BA
1T2
[ 1 ..
cos 1Tt + g cos 31Tt +
1
25
cos 51rt +
1
49
cos 7rrt + · · ·
]
(b) This Fourier series is identical to that in Eq. (6.16) with t replaced by t + 0.5.
(c) If x(t) =Co+ L: Cn cos(nwot +On), \.hen
Thus, time shifting by T merely increases the phase of the nth harmonic by nw0 T.
6.1-5. (a) For half wave symmetry
and
a,. = 2 [ (1'o/2
To Jo x(t)cosnwotdt+ Jo
(To/2 (
x T+
2
°
T, )
cosnw0
(
T,+
2
°
T, )
2 [ (1'o/2 (To/2 ]
To Jo x(t) cosnwotdt +Jo -x(T)[- cosnwoT] dT
4 [ ( To/2 ]
To Jo x(t)cosnw0 tdt
302
In a similar way we can show that
4 {To/2
b,. = To lo x(t)sinnw0 tdt
-4
8
[1 x(t)cos-tdt
0
4
nn
4
] =-
1
2
[1
0
2
t nn
-cos-tdt
2
]
4
nn n1T n1T
-24-2 ( cos-+-sin- -1 ) (n odd)
n 7T 2 2 2
4 (n1T n1T ) (n odd)
n2n2 T sin T - 1
Therefore
,..",,. (":i7t - 1) n= 1,5,9,13,···
- n2~> (n2" + 1) n = 3,7,ll,15,···
Similarly
bn 11 -
= -2
0
2
2
. -n1T t dt
t sm
4
=- 4- (.
n 2n 2
sm n1T
-
2
- -n-ir cos -n1T) = -4- sm
2 2 n27T2
. (n7T)
-
2
(n odd)
and
00
'i:""" n1T • nn
x(l) = L
n=l,3,5,-··
a,. cos
4 t + b,. sm 4 t
(ii) To = 27T, w 0 = 1, ao = 0 (by inspection). Half wave symmetry. Hence
00
-21"'
1T 0
e-tl10 cosntdt
2 [ e-t/10 ] ,..
- (-0.l cosnt+nsinnt) (n odd)
n n2 +00 .l 0
2 [ e- "110 1 ]
-7T n2 + 0.01 (01· ) - n2 + O.Ql ( - 01· )
2 ( -rr/10 l ) _ 0.0465
= 107r(n2 + 0.01) e - - n 2 + 0.01
and
2 [ e -t/10 ] "'
- (-0.lsinnt-ncosnt) (n odd)
r . n2 +00
. 1 0
2n _,,110 1.461n
(n2 + 0.01) (e - l) = n 2 + 0.01
303
6.1-6. {a) Here, we need only cosine terms and w0 = if· Hence, we must construct a pulse
such that it is an even function of l, has a value t over the interval 0 $ t $ 1,
and repeats every 4 seconds as shown in Fig. S6.l-6a. We selected the pulse
width W = 2 seconds. But it can be anywhere from 2 to 4, and still satisfy these
conditions. Each value of W results in different series. Yet all of them converge
to t over 0 to 1, and satisfy the other requirements. Clearly, there are infinite
number of Fourier series that will satisfy the given requirements. The present
choice yields
co
x(t) = ao + :z= a" cos
n=I
C;) t
By inspection, we find a0 = 1/4. Because of symmetry bn = 0 and
an = 4"41 0
1
mr t dt = n 47r (cos (rm)
t cos 2 2 2 2 n7r Sm
+2 . (n7r)
2 - 1]
{b) Here, we need only sine terms and w 0 = 2. Hence, we must construct a pulse
with odd symmetry, which has a value t over the interval 0 ~ t ~ 1, and repeats
every 7r seconds as shown in Fig. S6.l -6b. As in the case {a), the pulse width can
be anywhere from 1 to 7r. For the present case
co
x(t) = L bn sin 2nt
n=l
b11 = 7r !1 o
1
tsin2ntdl = ~(sin2n-2ncos2n)
7rn
(c) Here, we need both sine and cosine terms and w 0 = ~- Hence, we must. construct
a pulse such that. it has no symmetry of any kind, has a value t over the interval
0 $ t $ 1, and repeats every 4 seconds as shown in Fig. S6.l-6c. As usual, the
pulse width can be have any value in the range 1 to 4.
x(t) = a0 + ~
~an cos (n7r)
2 t + bn sin (n7r)
2 t
n=I
a,. = 21
-
4 0
1
t cos n7r
- t dt = -22-2 [cos (n7r)
2 n
-
11'
+ -nn sin (nn)
2
-
2
- 1]
2
21
-
4 0
1
nnt dl= -2- [sin ( -n7r ) - -cos
tsin -
2 n 2 7r2 2
n7r
2
(n7T)]
-
2
{d) Here, we need only cosine terms with w 0 = l and odd harmonics Qn)y. Hence,
we must. construct a pulse such that. it is an even function of t, has a value t over
the interva l 0 $ t ::::; 1, repeats every 27T seconds and has half-wave symmet.ry
as shown in Fig. S6.l-6d. Observe that the first half cycle (from 0 to 7r) and
the second half cycle (from 1T to 27r) are negatives of each other as required in
half-wave symmetry. This will cause even harmonics to vanish. The pulse has an
304
even and half-wave symmetry. This yields
00
a .. =-241T [1"
o
12
tcosntdt- 1" 1</2 7rn
]
(t-1T)cosntdt = - 2 2 (cosn7T-l)+ -sin-
2
n
mr
2
nodd
(e) Here, we need only sine terms with w 0 = 1T and odd harmonics only. Hence, we
must construct a pulse such that it is an odd function of t, has a value l over the
interval 0 :5 t :5 1, repeats every 4 seconds and has half-wave symmetry as shown
in Fig. S6.l-6c. Observe that the first half cycle (from 0 to 2) and the second half
cycle (from 2 to 4) are negatives of each other as required in half-wave symmetry.
This will cause even harmonics to vanish. The pulse has an odd and half-wave
symmetry. This yields
00
x(t) = L bn sin n 7T t
2
n• l
n odd
By inspection, a 0 = 0. Because of odd symmetry an = 0. Because of half-wave
""
symmetry (sec Prob. 6.1-5),
4 fI .
bn =;i Jo lsm
n1T
2 tdt+
f 2
(- t +2)sm
. rt1T
2
8 . Tt1T
tdt =n2 7T 2 sm
2 n odd
1
(f) Here, we need both sine and cosine terms with w0 = l and odd harmonics only.
Hence, we must. construct a pulse such that it has half-wave symmetry, but neither
odd nor even symmetry, has a value t over t he interval 0 $ t $ 1, and repeals
every 271" seconds as shown in F ig. S6.l-6f. Observe that the first half cycle from
0 to 7T) and the second half cycle (from 1T to 27T) are negatives of each other as
required in half-wave symmetry. By inspection, a0 = 0. This yields
00
x(t) = L a 11 cosnt+bnsinnt
r1 ::;;l
nodd
an 411
= -2
1T o
t cos nt dt = - 2 2 (cos n +
7Tn
n sin n - 1)
bn = - 411
2 7r 0
tsinntdt = - 2 2 (sinn -
7rTt
ncosn) n odd
305
• > •
I (
i
.. ' ,. , . " . .)
. .. 1t:;"
...
- :i-rr .2.tr -k-
Figure 86.1-6
6.1-7.
"
a b c d e f g h
periodic? yes yes no yes no yes yes yes yes
I 3 1 2
Wo 1 1 11' 70 4
8w
period 211" 2r. 2 14011' 3 211' 1T
Dn = 2_
211'
11-I
e - j(mr/2)t dt -f 1
3 e-j(mr/2)t dt = 2_ sin~
1Tn 2
lnl 2: 1
n=-oo
Dn
l
= -211' 12" -e
t - .,,,
21T
'
j
dt = - ,
21Tn
0
1 n>O
so that IDnl=- ,
21Tn n<O
306
(d) To= 7r, wo = 2 and Dn = 0
00
x(t) = L Dn~2nt,
n= -oo
4
4t · nt
where Dn = -
1
7r 1 1</
-tr/ 4 7r
-1. ( 2
-e-3 2 dt = - - 7rn 7rn
. 7rn
Sill -
2
-
7rn )
cos -
2
(e) 'T'
~Q -- - 2"
3 I w0-3.
00
x(t) = '°'
L.., Dn e;2w•t
-r ,
n=oo
where Dn = -13 1 0
1
le-'· h.!J.t
a dl =- 3- [ e- 3· 2..rul.
4rr 2n 2
a (j211'n
- - + 1) -1 ]
3
Therefore
IDnl=--3
47i2n2
[J 47r -
2+ -
9
n
- 2cos-
3
2
2rrn - 47rn
- sm
3
2
. 27rn]
3
-
x(t) = 0.5 +
n= -oo
:tt 0 $ t < A
1 A$l < 1r
x(t) =
{ 0
x(t + 27T)
7f ~t < 211'
otherwise
Do = ;, { x(t)dt
o lr0
307
, . )l"
(C\)
;
·( ' '··
.< ·'.' ·... •
-~ , .?; - J(
.. "P~
ID., l
_.. I i l
-:i.
l /'Tr..... w~
(d)
. L Dn 0
qo (e.)
o-.·a. U>,. I
·« -u- 'tE QJ •
3 3
-19 w.,
3 :J -<r! ..
, .
lf)
Figure 86.3-1
r"'
t +
1 ( lo Adt
27T j" dt)
A
1 ( t2 IA ,. )
271" 2A t=O t!t=A
~ (~2 +1T- A)
27T
21T - A
47T
Forni= 0,
308
=
~
2'1r
( te-jnt IA
-jAn t = O
-1A 0
e-jnt dt + e-int
jAn -jn t=A
I" )
1 (e-jnA e-int IA e-inw - e-inA)
-27T - -- -
- jn
-
-An2 t =O
+ --- -- -
- jn
1 (je-jmr e-jnA - 1)
= 27r --n- + An 2
=
_l_ (e-inA - 1 + je-jmr)
27rn An
Thus,
21'-A
4;f n=O
_l_ ( "-jnA_l + 1·e-jmr) otherwise
21rn nA
6.3-3. (a)
4
In the preceding expression, we could have expressed the term 2 cos {St - as
2 cos {St+ 2:n. Figure S6.3-3a shows amplitude and phase spectra.
;)
(b) By inspection of the trigonometric spectra in Fig. S6.3-3a, we plot the exponential
spectra as shown in Fig. S6.3-3b.
(c) By inspection of exponential spectra in Fig. S6.3-3a, we obtain
~ eit + G e-i'f ) e;st + ( e - i\") eiBt + ~e-it + ( ~ ei~ ) e-i5t + (ei¥) e-;st
(d) By inspection of the first line in part (c), we can immediately write x( t) in the
trigonometric form as
2 4
x(t) 3cost+cos(5t- ; ) +2cos ( st- ; )
309
w-
5
.,
LJ>n
5
-~ -5 5 0
Figure S6.3-3
~) + cos ( 3t - ~) + ~ cos ( 5t -
2
= 3 + 2 cos ( 2t - ;)
(d) By inspection of the first line in part (c) , we can immediately write x(t) in the
t rigonometric form as
= 3+2cos(2t-i)+cos(3t -~)+~cos(5t- ;)
2
x(t)
6.3-5. (a) The exponential Fourier series can be expressed with coefficients in Polar form
as
From this expression the exponential Spectra are sketched as shown in Figure
S6.3-5a.
(b) By inspection of t he exponential spectra in Figure S6.3-5a, we sketch the trigono-
metric spectra as shown in Figure S6.3-5b. From these spectra, we can write the
310
It'' 1 2 3 4 ':::>- w _,,
e1,
-111'-
-'11/2
I 2 3 4 5 W~
- 2....[
3
27rf~
3 I D,"'\ Ti!z. LDn
. !2.
.ii.
1.
f _,<. 1. ~ 4 fj
- I -s -:,}
-.[
w.-
'
' ~
-1V:z
-2'/r/~·
Figure 86.3-4
IDn I
'
3
,
' ' L..Dr"I
I :2
rr/z
":Tl~
1.
-3 -2 _, 1. 2 3 w~ -3 - 2. -1
-Tr/'2
~
-7r[:a. .
1' 5
e.,., 4 t1'
3 ~Y\ ~ ::t 3
-T/lf
- lr/'1
3 w ..
' '2.
Figure 86.3-5
311
6.3-6. (a)
7r
x(t) 2 + 2 cos(2t - r.) + cos(3t - 2)
2 - 2cos2t + sin3t
(d) Observe that the two expressions (trigonometric and exponential Fourier series)
are equivalent.
f :.L])n
2. ID() i
/
i
/
' I 2 3 \../-
1- '' '
r
·-if -'.> -2 -I
c•5
1-2-34 <i> ~
-iC
u.> -
Figure S6.3-6
6.3-7. (a) The exponential Fourier series, as found by inspection of Figure P6.3-6, is
x(t) 2r.)
= 2+4cos ( t+ 3 +2cos ( 2t+ 37r)
(d)
2
x(t) 2+4cos(t+ ; ) + 2cos(2t+i)
6.3-8. (a) The period is To= 8 and wo = -rr/4. Also Do= 0 (by inspection), and
00
x(t) = L D,,ein'it
n == -ex>
312
fl
f' b e~,
QI"'\ 4
z.Tl/3 4 .. .
.
--
2-
Tr/3
0 .1 1. .3 w~ 0 2- 3 cu ..-
Figure 86.3-7
D, = ~ [L: (~ 1)
1 + e-j2n(1f/4)tdl+ 1 (-~ i)
1
+ e-j2n(,,/4)tdt] =
This yields
n = ± 1, ±3, ±5, ± 7, · · ·
otherwise
Therefore
= '°'
00 4 . •
x(l) -e'"•'
~
. ..
n odd
7r2n2
(b) Observe that i(t) is the same as x(t) in Figure P6.3-8a delayed by 2 seconds.
Therefore "
00 00
Therefore
00
i(t) = L Dne'n".t
n - 1
n odd
where
(c) Observe that x(t) is the same as x(t) in Figure P6.2-8a time-compressed by a
factor 2. T herefore
00 00
Therefore
00
i:(t} =L DneJnftt
n-1
n odd
where
313
6.3-9. (a)
00
x(t) = L DneJnwot
n=-oo
00 00 00
(b)
00
x(t) I:
n=-co
Dnejnwot
00
x(t) x(at) = L:
n=-oo
DneJnwo(at)
l 4 N-l ( l)"
w(t) = - + ~
3 7r
L ~cosnnt
n
n=l
P,,
1
= -9 + 1T4
-
8
"°' - =
N-l
~
n=l
n4
1
0.198
314
The power of x(t) is
Px
111
=2 -l (At)
2
dt =T
A2
2A N l
w(t) = -(-1r+ 1 L:- sinmrt
7r
n=l
n
2
1 N 4A2
P.,,= -
2
2= 22
1 n
?0.3A
7T
2
N 4 N 1
Px = L 1Dnl2 = 7r2 L (1 - 4n2
)2 ? 0.49875
n=- N n;-N
Direct calculations using the above equation gives P: = 4/7r 2 = 0.4053 for N = 0
(only de), Px = 0.49535 for N = 1 (3 terms), and Px = 0.49895 for N = 2 (5 terms).
Thus, a 5-term Fourier series yields a signal whose power is 99.79% of the power of
the rectified sine wave. The power of the error in the approximation of x(t) by x(l) is
only 0.213 of the signal power Px.
6.4-1. Period T0 = 7r, and w 0 = 2, and
H (. ) jw and from Eq. (6.30b) Dn = 0.504
JW = (-w2 + 3) + j2w' 1 + j4n
00 00
6.4-2. (a)
315
= 4~ (ei(St- f) - e-i(St-f) - ei(2t+lf) + e-i(2t+t>]
'
. .
•• - y't .•
'.
.
Figure S6.4-2b -'7T
~
(c) Since none of the spectral components of x(t) appear in the pass-band of the
filter, the output is y(t) = 0.
6.4-3.
1
_,·nw tdt = --"'-'--""-----'-
Dn =
1 0
e
- t
e 0
(e - 1)( 1 - j27rn)
e( l + 47r2 n 2 )
The transfer function of the R-C drcuit is
1 jw
H (jw) -
- i + <L) -- - -
jw + i
x(t) = f
n=-oo
(e - 1)(1 - j2mi) ei2,,nt
e( I + 47r2n2)
2cly l2 = 2x.y
which yields the desired result.
6.5-2. (a)
316
Also
t2 x(t)[x(t) - cx(t)] dt = Jt2x(t)x(t) dt - c Jt2x 2 (t) dt
Jt1 t1 t,
But
ftt, x(t)x(t) dt
c= __,_l. , . - - - - -
ftt12 x2(t) dt
Substitution of c in the earlier equation yields
t2
J
t1
x(t) [x (t) - cx(t)] dt = 0
6.5-3. (a) If x(t) and y(t) are orthogonal, t hen we showed {see Eq. (6.67)] the energy of
x(t) + y(t) is Ex+ Ey. We now find the energy of x(t) - y(t):
1-: 2
lx(t) - y(t)J dt = l: l:lx(t)!2 dt + ly(t)l2 dt
1-: l: jx(t)J dt +
2 2
jy(t)l dt
The last result follows from the fact that because of orthogonality, the two in-
tegr als of the cross products x(t)y*(t) and x*(t)y(t) are zero [see Eq. (6.80)].
Thus the energy of x(t) + y(t) is equal to that of x(t) - y(t) if x(t) and y(t) are
orthogonal.
(b) Using similar argument, we can show that the energy of c1 x(t) + c2 y(t) is equal
to that of c 1 x(t) - c2 y(t) if x(t) and y(t) are orthogonal. This energy is given by
lcd 2 E:c + lc21 2 Ey . ·
(c) If z(t) = x(t) ± y(t), then
1-: 2
Jx(t) ± y(t)1 dt = l: 1-: lx(t)l2 clt + jy(t)l2 dt
317
6.5-4. (a) In this case Ex = f01 dt = 1, and
c=-E 111 % 0
x(t)y(t)dt=- 111
1 0
tdt=0.5
(b) Thus, x(t) ~ 0.5y{t), and the error e(t) = t - 0.5 over (0 $ t $ 1), and zero
outside this interval. Also Ex and E,, (the energy of the error) are
1 1 1
Ex = fo 2
x (t) dt = fo t
2
dt = 1/3 and Ee= fo (t - 0.5) dt = 1/ 12
2
c = 1
E:c 11
0
y(t)x(t)dt =3 110
tdt = 1.5
Thus, y(t) ::::: 1.5x(t.), and the error e(t) = y(t) - l.5x(t) = l- l.5t over (0 $ t $ 1), and
zero outside this interval. Also E e (the energy of the error) is Ee = (1 - 1.5t) 2 dt = J;
1/4.
6.5-6.
00
ao = 11 o
1
x(t) dt = [1 t dt = ~
lo 2
~l
1
a,.= 2 fo tcos27rntdt = 0 n (ninteger)
- 1
1
1
bn = 2 tsin27rnldt = -
o nn
Hence
00
l l 1
=-- - - - :L:- sin27rnt
2 n n=l n
318
From Eq. (6.77)
11 t2 dt - ~= ~- ~= 112
1 1 1
3 - 4- 2
rr 2 = 0.03267
-13 - -41 - -2rr1 2 - 8rr
- 1 2 = 0.02
~- ~- 1 1 1
- -2 - - -2 - - - 2 = 0.014378
3 4 2rr 8rr 18?T
6.5-7. (a) Figure S6.5-7a shows x 1{t) that is a periodic extension of x(t) to yield a series
with w 0 = 2rr and only sine terms. This requires T 0 = 27T /2w = 1 and odd
symmetry. From inspection, the de component. is 0.5. If we subtract. de (0.5)
from x 1 (t), the remaining signal x 1 (t) - 0.5 has odd symmetry (only sine terms).
Therefore
00
(b) w 0 = 7T and To = 2-;r /Tr = 2. For sine terms only, we need odd symmetry. Figure
S6.5-7b shows a suitable function x 2 (t). It has no de.
00
-411
2 0
2
tsinmrt<.lt = {- l)n+l_
nTr
(c) wo = ?T, To = 2Tr/wo = 2. For cosine terms only, we need an even function x3(t)
as shown in Figure 86.5-7c. By inspection de is 0.5. Therefore
1 00
2+ L n=l
ancosnTrt
=
411o
-
2
tcosmrt dt -4-
= -rr2n2 n = 1, 3, 5, ...
6.5-8. (a) The signal g(t) is the same as the signal x(t) in Example 6. 12 (Figure 6.23)
319
Figure 86.5-7
1 - 1
x 2 (t) dt = 27r and
J 2 1 2
x (t) dt - -c1
3
=2- 3
-
2
= 0.5
1 2
x (t) dt -
12 12
3c 1 - 7CJ = 0.28125
Since g(t) is the same as x(t) time-expanded by a factor 7r, all energies are in-
creased by the same factor (7r). Therefore
f1 = 0.5rr
€2 0.281257!"
6.5-9.
Hence
~
1
Co fo x(t)xo(t) dl =
320
c1 1 1
x(t)x1(t)dt = -~
C2 C4 = C5 = C6 = 0
C3 1 1
x(t)x3(t) dt = -~
= 1
1
1
c7 x(t)x1(t) dt = -
16
Hence
1 1 1 1
x(t) '.::= -xo(t) - - x 1 (t) - -x3(t) - -x1(t)
2 4 8 16
Also
1 1
2
x (t)dt= ~ and En= 1
If Ee(N) is the energy of the error signal in the approximation using first N terms,
then From Eq. (6.77)
~ - ~= 2- = 0.0833
3 12
1 2 2 1
- - C0 - C1
3
= -48 = 0.0204
1 2 2 2 1
- - C0 -
3
C1 - C3 = -192 = 0.0052
1 2 2 2 2 1 '
- - C0 -
3
C1 - C3 - C7 = -
768
= 0.001302
the corresponding trigonometric Fourier series found in Prob. 6.5-6 are
0.0833, 0.03267, 0.02, 0.014378. Clearly, the Walsh Fourier series gives smaller er-
ror than the corresponding trigonometric Fourier series for the same number of terms
in the approximation.
6.M-1. (a) To determine a suitable set of N = 10 frequencies wn, we first determine ten
points logarithmically spaced from I to 100.
>> f = logspace(0,2,10)
f = 1.0000 1.6681 2 . 7826 4 . 6416 7 . 7426
12.9155 21.5443 35.9381 59.9484 100.0000
The problem with these points is that they are not all rational, and the resulting
signal m(t) is thus aperiodic. Truncating to the four decimal places shown makes
the frequencies rational, but the resulting period T 0 is excessively long. An
approximately logarithmic sequence that results in smaller To is generated by
rounding the logarithmic frequencies to the nearest tenths of a hertz.
>> f = round(lO*logspace(0,2,10))/10
f = 1.0000 1.7000 2.8000 4.6000 7.7000
12.9000 21.5000 35.9000 59.9000 100.0000
With these frequencies, the signal m(t) = '2:~=l cos (wnl + Bn) has period To =
10. Thus, one reasonable choice of frequencies is
wn = 27rll, 1.7, 2.8, 4.6, 7.7, 12.9, 21.5, 35.9, 59.9, 100] for which m(t) has period To= 10.
321
>> m = inline('sum(cos(omega•t+theta•ones(size(t))))', ...
'theta','t','omega') ;
>>omega= 2•pi•f'; theta= zeros(size(omega)) ;
>> t = (-5:.01:5); plot(t,m(theta,t,omega),'k');
>> xlabel('t [sec]'); ylabel('m(t) [volts]');
I,
E{
-2
.. I _,
-·
-$
-· ·2
•
l{H<.1
322
>> mmag_opt = mmag; mmag = [mmag,zeros(l,9)];
>> for trial = 2:10;
>> theta_init = 2*pi*rand(N,1);
>> theta= fminsearch(maxmagm,theta_init,[],t,omega);
>> mmag(trial) = max(abs(m(theta,t,omega)))
>> if (mmag(trial)<mmag_opt),
>> theta_opt =theta'; mmag_opt = mmag(trial);
>> end
>> end
» mmag, theta_ opt
mmag = 6.6854 6.7069 6.6568 6.6421 6.7301
6.4846 6.5906 6.5294 6.5292 6.5888
theta_opt 2.7839 5.7162 5. 1883 4 . 2464 5.1741
4.0850 2.2366 1. 9871 1. 6831 3.7462
Thus, a good (but unlikely globally best) choice of phases is
On= 2n(2.7839, 5.7162, 5.1883, 4.2464, 5.1741, 4.0850, 2.2366, 1.9871, 1.6831, 3.7462).
In this case, the maximum value of m(t) is 6.4846, as shown in Figure 6.M-lb.
>> plot(t,m(theta_opt,t,omega),'k');
>> xlabel('t [sec]'); ylabel('m(t) [volts]');
(c) For environments with l/f noise, it is appropriate to have lower frequency com-
ponents have greater strength than higher frequency components. One simple
possibility is to adjust the magnitude of each sinusoidal component to match
the noise power at that frequ ency. Jn t his way, the signal-to-noise ratio is kept
constant for any frequency bin of the signal.
N
m(t) = L k
~cos (wnl +On).
n = l yWn
The constant k is selected to achieve the final desired signal power for the entire signal
m(t).
323
Chapter 7 Solutions
7.1-1.
If x(t) is an even function of t, x(t) sin wt is an odd function of t, and the second
integral vanishes. Moreover, x(t) cos wt is an even hmction oft, and the first integral
is twice the integral over t he interval 0 to oo. Thus when x(t) is even
= 2 fo
00
fo'' x(t)sinwtdt
0
X(w) = - 2j (2)
If x(t) is also real (in addition to being even), the int.egral (1) is real. Moreover from
(1)
X(-w) = 2 fo':io x(t)coswtdt = X(w)
Hence X(w) is real and even function of w. Similar arguments can be used to prove
the rest of the properties.
7.1-2.
x(t) ~
2r.
J
00
-oo
X(w)eiwt dw = ~
27r
J"°
-oo
IX(w)leiLX(w)ejwt dw
x(t) = -11"°
7f 0
!X(w)j cos[wt + LX (w)J dw
7.1-3. (a) Because x(t) = x 0 (t) + Xe(t ) and e-iwt =cos wt+ j sin wt
324
= 1_: [x0 (l) + Xe(t)J coswtdt - j 1_: [x0 (t) + Xe(t)] sin wtdt
Because Xe(t) cos wt and x 0 (t) sin wt are even functions and x 0 (t) cos wt and
Xe( t) sin wt are odd functions of t, these integrals [properties in Eqs. (B.43), p.
38J reduce to
00
X(w) = 21 Xe(t)coswtdt - 2j 1 00
x 0 (t)sinwtdt {1)
x 0 (t)sinwtdt (2)
l 1
u(t) - [u(t) + u(-t)J + -[u(t) - u(-t)J
2 2
1 1
- + -sgn(t)
2 2
'-v-' '-v--'
and
1
..-'\Aw)= 7r0(w) and X 0 (w) =-:-
JW
Clearly, X e(w) is the real part and X 0 (w) is the odd part of X(w).
We follow the same procedure for x(t) = e-atu(t) .
Xe (t) :z:.,(t)
Also
Xc(w) =~
2
[-1- __I_]= w2 +
jw + a jw - a
_ 2_a_
a2
and
Xo(w) = ~ [-1- + _1-] =
2 jw +a jw - a
2jw
w2 + a2
Clearly, Xe(w) is the real part and X 0 (w) is the odd part of X{w).
7.1-4. (a)
T 1T 1 -(jw+a)T
X(w) =
10
e-ate-iwt dt =
0
e - Ciw+a)t dt = - e
JW +a
,
(b)
T lT 1- -(jw-a)T
X(w) =
1 0
eat e -iwt dt =
. 0
e-Uw - a) dt = _ _e_ __
JW - a
325
7.1-5. (a)
1 !2 2e-jwt dt = 4 2 - jw 2 -j2w
X(w) =
1 0
4e- jwt dl +
J
- e . -
JW
e
(b)
o t . 1.,. -e-J"'tdt=
t . 2
X(w) =
J -T
- - e-J"' 1 dt+
T 0 T
-
TW
2
[coswr+wrsinwr-1 )
This result could also be derived by observing that x(t) is an even function.
Therefore from the result in Prob. 7.1-1
7.1-6. (a)
x(t) -
l j wo w e}wt dw = -l
2 jwt
_e_ [-w 2 t 2 - 2jwt + 2}
lwo
27r -wo 27r (jt)3 -wo
(w6t 2
- 2)sinw0 t + Zwotcoswot
=
7rt3
'lo
(b) The derivation can be simplified by observing that X(w) can be expressed as a
sum of two gate functions X 1 (w) and X 2 (w) as shown in Figure S7.l-6. Therefore
2rr _ 2
2
x(t) = 2 - j [X i (w) + X'.l(w)Je}wtdw = 2_
27r
{! 2
_2
e}wtdw+ j l e}wtdw } =sin 2t+sin t
_1 7rt
: ~ Xi rw~ ~1~/~)
- [ ( ·. + (·
-~ - 1 '. .2. -2 2- -' 0 ~
Figure S7 .1-6
7.1-7. (a)
x(t) -
1 j-rr/2 coswejwt dw
27r -r./2
e}wt . . Tr/2
{1tcosw+s111w} 12
2n ( 1 - l 2 ) _.,.
(b)
326
Because X(w) is even function, the second integral on the right-hand side van-
ishes. Also the integrand of the first term is an even function. Therefore
x(t) = -11""
1T 0
12
w
-cos
WO
-'· . = -1- [costw
twuw
1TWo
+ twsintw]wo
t2 O
7.1-8.
X(w) = l: x(l)e-jwtdt
and x(t) -
1 !()()
X(w)dw
21T -()()
7.2-1. Figure S7.2- l shows the plots of various functions. The function in part {a) is a
gate function centered at the origin and of width 2. The function in part (b) can be
expressed as .6 ( 10~13 ). This is a triangle pulse centered at the origin and of width
100/3. The function in part (c) is a gate function rect(~) delayed by 10. In other
words it is a gate pulse centered at t = 10 and of width 8. The function in part (d) is
a sine pulse centered at the origin and the first zero occurring at ";' = 7r, that is at
w = 5. The function in part (e) is a sine pulse sine(~) delayed by 107r. For the sine
pulse sine(~), the first zero occurs at~= 1T, that is at w = 57r. Therefore the fun ction
is a sine pulse centered at w = l07T and its zeros spaced at intervals of 5tr as shown
in the figure S7.2-le. The function in part (f) is a product of a gate pulse (centered
at the origin) of width l07r and a sine pulse (also centered at the origin) with zeros
spaced at intervals of 57r. This results in the sine pulse truncated beyond the interval
±57r (ill ;:::: 5tr) as shown in Fig. f.
7.2-2.
X(w)
327
re.ci: ( 7)
1
t 1- . I
(f )
- 5 iT
Figure S7.2- l
7.2-3.
= _1 IO+n ;wt llO+>r
eiwt dw = _e__ = _ 1 _ [e;(lo+n)t _ e;(tO-w)t)
x(t) 2trJ
·1ot
10
_ ,.. 211(jw) 10 _ ,. j2trw
7.2-4. (a)
x(t)
2.. f "'O e-jwtoejwt d,w::: 2.. ! WO ejw(t-to) dw
27r - wo 2 r. - wo
1 e jw(t-to)\"'
0
=- -- -
sinwo(t wo . I (t -
to) = -s1ncwo to)]
(2tr)j(t - t 0 ) _..,
0
tr(t - to) 7r
(b)
x(t) = 2- [Jo
27r - wo
j ei"'t dw +
lo
ro -jeiwt dw]
= _ I_ eiw' \o _ _l_ei"''lwo = 1 - coswot
2r.t -wo 2trt 0 -rrt
7.2-5. (a) When a > 0, we cannot find the Fourier transform of e"tu(t) by setting s = jw
in the Laplace transform of e0 tu(t) because the ROC is Res> a, which does not
include the jw-axis.
(b) The Laplace transform of x(t) is
Interestingly, because x(t) has a finite width, the ROC of its X(s) is the entire
s-plane, which includes jw-axis. Hence, the Fourier transform
328
To verify this, we find the Fourier transform of x(t)
or
~ [ o(t) + j~t] ¢::::::} u(-w)
~2 [o( - t) - J7rt
~] -¢===> u(w)
(b)
7r[o(t + wo) + b(t - wo)J ¢::::::} 27r cos (-wow)= 21T cos (w0w)
'----v-'
X(t) 2rrx(-w)
Setting w 0 = T yields
o(t + T) + o(t - T) ¢::::::} 2cos Tw
(c)
329
Setting wo = T yields
7.3-2. Refer to the solution of P rob. 1.2-3 for description of t hese signals.
(a)
x 1 (t) = x(t + 1) + x(l -t)
x (t + 1) <==> X(w )ei"'
Time inversion of x(t + 1) yields x( - t + 1), Hence
x(l - t) <==> X( - w)e-i"'
and x C~ 1
) = x (~ + ~) ¢=> 2X(2w)ei"'
+2) <==>
X
(
-t
4
- 4X(4w)ei 2w
x (2 ~ t) <==> 4 X(-4w)e-i 2w
Hence x 3 (t) <==> 4 (X(4w)ei 2"' + X(-4w)e - i 2"' + 2X(2w) + 2X( - 2w)]
(d)
and x
(2-t) <==>
-
2
- 2X (-2w)e-3'2w
330
x C: 2
) {=>4X(4w)d 2w and x(
2
~ t) <==> 4X(- 4w)e-f w 2
X(w) = : 2
(ejw - jweiw - 1]
7.3-3. (a)
rect (
t ±T/2)
T <===> Tsinc ( w{) e±jwT/'l
= 2 ·r.
J
(wT) sm. wT
smc T T
~ sin2 (w;)
(b) From Figure S7.3-3b we verify that
Note that sin(t - 7r)u(t-7r) is sin t u(t) delayed by ;r. Now, sin tu(t) <===> f;[c5(w-
l) - 6(w + l)j + 1 _ 1w> and
Therefore
X(w) = {
7r
-:[c5(w - 1) - 6(w + l)] + - 1 - 2 } (1 + e _3..,..w)
~ 1-w
Recall that g(x)6(x - xo) = g(xo)o(x - Xo). Therefore o(w ± l)(l + e-j11'w) = 0,
and
X(w) = - -1(21 + e_,.,....,)
.
l -w
331
(c) From Figure S7.3-3c we verify that
Therefore
· - jnw/2 1
X(w) = ~2 + e = --(jw+e-i"wf
2
2
]
1- w 1 - w2 1- w
(d)
x(t) e-o.t[u(t) - u(t - T)] = e-a'u(t) - e-atu(t - T)
e-" 1u(t) - e- aT e-a(t- T}u(t - T)
1 -aT . 1
X(w) -._ _ _ ..!L_e-jwT = -.- - [ l _ e-(a+iw)TJ
JW + a. ;w + a JW +a
Figure S7.3-3
Therefore
x(t + T) + x(t - T) <===> X(w)eiw'I' + X(w)e-jwT = 2X(w) coswT
We can use this result to derive transforms of signals in Figure P7.3-4.
(a) Here x(t) is a gate pulse as shown in Figure S7.3-4a.
332
Also T = 3. T he signal in Figure S7.3-4a is x(t + 3) + x(t - 3), and
_, (o.)
Figure S7 .3-4
Therefore
1 . . 1
x(t) sin wot= j [x(t)e1"'0 t + x(t)e-1 "'0 t]
2
= 2j [X(w - wo) + X(w -wo)]
Therefore
and
2~[x(t+T}-x(t-T)] ~ X(w}sinTw
The signal in Figure P7.3-5 is x(t + 3) - x(t - 3) where
7.3-6. (a) The signal x(t) in this case is a triangle pulse 6( 2t,,) (Figure S7.3-6) multiplied
by coslOt.
x(t) = 6 c~) cos lOt
333
Also from Table 4.1(pair19) t:i.U.J <==} 7Tsinc2 {":n From the modulation prop-
erty {4.41), it follows that
x(t) = 6. ( t)
2
7T cos lOt <==}
7T { smc
2
. 2 [7T(w - 10)]
2
. 2
+ smc [7T(w+10) ]}
2
The Fourier transform in this case is a real function and we need only the
amplitude spectrum in this case as shown in Figure S7.3-6a.
(b) The signal x(t) here is the same as the signal in (a) delayed by 27T. From time
shifting property, its Fourier transform is the same as in part (a) multiplied by
e-iw(Z,,-). Therefore
The Fourier transform in this case is the same as that in part (a) multiplied by
e-i 21f"'.
This multiplying factor represents a linear phase spectrum -27Tw. Thus
we have an amplitude spectrum [same as in part (a)j as well as a linear phase
spectrum LX(w) = -27Tw as shown in Figure S7.3-6b. the amplitude spectrum
in this case as shown in Figure S7.3-6b.
Note: Jn the above solution, we first multiplied the triangle pulse 6.( 2~) by
cos lOt and then delayed the result by 27T. This means the signal in (b) ~s
expressed as 6.(t;;"')cos 10(t-27T).
We could have interchanged the operation in this particular case, that is, the
triangle pulse 6.( 2trr) is first delayed by 27T and then the result is multiplied
by cos lOt. In this alternate procedure, the signal in (b) is expressed as
6.( t;;") cos lOt.
This interchange of operation is permissible here only because the sinusoid
cos lOt executes integral number of cycles in the interval 211. Because of this
both the expressions are equivalent since cos lO(t - 21t) =cos lOt.
(c) Jn this case the signal is identical to that in (b), except that the basic pulse is
rect ( 2~) instead of a triangle pulse 6.( 2~) . Now
rect (
2~) <=:::} 2rr sine(1TW)
7.3-7. (a)
334
-'t - .l- 2. '+
, .,
Figure 87.3-6
(b)
Therefore
x(t) = ~sinc 2 (t)cos4t
'if
7.3-8. (a)
1 1
e>- 1u(t) <==? -.- - and u(t) <¢===> m5(w) + -:-
JW - A JW
-~o(w)
A
+ [~i + ~]
JW JW - A
becauseg(x)o(x) = g(O)o(x)
= ~,\ [JW-1- .,\ - (m5(w) + _;_)]
JW
335
Taking the inverse transform of this equation yields
1
x(l) = >:(e>.t - l)u(t)
(b)
1
e>.'tu(t) <==> -.--
JW - A1
Therefore
(c)
I l
X(w) = - l = ~ - ~
(jw - >-1)(jw - >-2) jw - ,\1 j w - ,\2
Therefore
I
x(t) = le>.•tu(t) + e>-21u(-t)J
,\2 - ,\I
Note that because ,\2 > 0, the inverse transform of is e>- 2 tu(-t) and not jw-_\,
-e>- tu(l) . The Fourier transform of the latter does not exist because >- 2 > 0.
2
(d)
1
e>.111L(- t)-¢===> - -. - - and
JW - A1
If x(t) = e>- 11
u( - t) * e>- 2 tu(-t), then
-1 -1
X(w) = l = ~ - r.::r,
(jw - ,\1)(jw - >-2) jw - >-1 jw - >-2
Therefore
l
x(t) = (e>.,t - e>- 21 )u(-t)
,\2 - ,\I
The remarks at the end of part (c) apply here also.
7.3-9. From the frequency convolution property, we obtain
1
x 2 (t) <=> - X(w)
2;r
* X(w)
Because of the width property of the convolution, the width of X(w) * X(w) is twice
t he width of X(w). Repeated application of this argument shows that the bandwidth
336
of xn(t) is nB Hz (n times the bandwidth of x(t)).
7.3- 10. (a)
X(w) =
J o e-iwt
-T
. dt - 1T . dt = -7"-ll -
O
e- 1"'t
JW
2 cos wTJ = ?__·4 sin2 (wT)
W
-
2
(b)
2J·rsmc
. (wT)
T sm. wT
2
j4 . (wT)
- sm 2-
w 2
(c)
X(w) =
J oo .
x(t)e-Jwt dt and -dF = - d J°" x(t)e- . dt 1 "'t
-oo dw dw -oo
-dF =
dw
J°" -
-oo
d (x(t)e-iwt)
dw
. = J°" [-jtx(t)Je-J"'t. dt
-00
T herefore
dF
- jtx(t)-¢::::? -
dw
337
(b)
l
jw+a
! Cw~a) = (jw-:a) 2
te- at u ( t) ~ (. 1 ) 2
JW +a
7.4-1.
l
H(w) = -
.-
JW + l
(a)
l
X(w) = --
jw + 2
I 1 l
Y (w) = = - - - --
(jw + l )(jw + 2) jw+ l jw + 2
y(t ) (e-t - e- 2 t)u(t)
(b)
l
X(w) = -jw+- l
1
Y(w) = (jw + 1) 2
y(t) = te- atu(t)
(c)
X (w ) = -jw-l--1
- 1 1/2 1/2
Y(w) - -- - -
(jw + 1) (jw - l) =- - ---
jw + 1 jw - 1
y(t) = 21 e- t
u(t) + 1 e t u(- t)
2
(d}
1
X(w) = 7TO(w) + -:--
JW
Y (w) =
1
-. - -
JW+ 1
[m5(w} + -l-]
JW
l
= 1TO(w) + ----
jw(jw + 1)
[becauseg(x)o(x) = g(O)o(x)J
1 1
= 1TO(w} +-:-- - -. - -
JW JW + l
y(t} = (I - e- t)u(t}
338
7.4-2. (a)
1 -1
X(w)=-.- and H(w)=-.-
JW +1 JW- 2
and
- 1 1 [ 1 1 ]
Y (w) = (jw - 2)(jw + 1) =3 jw + 1 - jw - 2
Therefore
(b)
-1 -1
X(w)=-. - and H(w)=-.-
JW- 1 JW- 2
and
1 - 1 -1
Y(w) = (jw - l)(jw - 2) = -j w---1 - J-.w-
- -2
Therefore
y(t) = [et - e2 tju(-t)]
The spectra Y1 (w) a nd Y2 (w) are also shown in Figure 87.4-3. Because y(t) =
y 1(t)y2(t), the frequency convolution property yields Y(w) = Yj(w) * Y2(w) . From
the width property of convolution, it follows that the bandwidth of Y(w) is the sum
of bandwidths of Yj(w) and Y2 (w). Because the bandwidths of Y1 (w) and Y2 (w) are
10 kHz, 5 kHz, respectively, the bandwidth of Y(w) is 15 kHz.
7.4-4 . H(w) = 10- 3 sinc ( 2 c);i 0 ) and P(w) = 0.5 x 10- 6 sinc2 ( 4 x~od
The two spectra are sketched in Figure 87.4-4. It is clear that H(w) is much narrower
than P(w), and we may consider P(w) to be nearly constant of value P(O) = 10- 6 /2
over t he entire band of H(w). Hence,
Y(w) = P(w)H(w) ~ P(O)H(w) = 0.5 x 10- 6 H(w) ==> y(t) = 0.5 x 10- 6 h(t)
Recall that h(t) is the unit impulse response of the system. Hence, the output y(t) is
equal to the system response to an input 0.5 x 10- 6 o(w) = Ao(w).
The two spectra are sketched in Figure S7.4-5. It is clear that P(w) is much narrower
than H(w) , and we may consider H(w) to be nearly constant of value H(O) = 10- 3
339
I XMU>)
.tt:'>- tcii:>
1 ·)t' {W)
Figure 87.4-3
Figure S7.4-4
P! w) I
I
I
015
:zrrt
tf I
Figure S7.4-5
7.4-6. Every signal can be expressed as a sum of even and odd components (see Sec. 1.5-2).
Hence, h(t) can be expressed as a sum of its even and odd components a
340
where h.,(t) = ~lh(t)u(t) + h(-t)u(-t)J and h0 (t) = !lh{t)u(t) - h(-t)u(-t)j. From
these equations, we make an important observation that
provided that h(t) has no impulse at the origin. This result applies only if h(t) is
causal. The graphical proof of this result may be seen in Figure 1.24.
Moreover, we have proved in Prob. 7.1-1 that the Fourier transform of a real and even
signal is a real and even function of w, and the Fourier transform of a real and odd
signal is an imaginary odd function of w. Therefore, if X(w) = R(w) + jX(w), then
R(w)
1
= -jX(w)
271"
2 1
*-:- = -
JW 7r
100
X(y)
- - dy
-oo W - Y
and
1 2
jX(w) = - R(w) *-;-
271" JW
=-;-
1 100
R(y)
- - dy
J?r _ 00 W - y
or
_.!. j
00
X(w) = R(y) dy
71 -oo w -y
7.5-1.
h{t) = _ l _ e-(t-to)2/4k
./47rk
This is noncausal. Hence the filter is unrealizable. Also
~ dw =
00 00
I ln IH{w)ll dw =
1 _
00
w2 + 1
{
} _ 00 w2 + 1
00
Hence the filter is noncausal and therefore unrealizable. Since h(t) is a Gaussian
function delayed by t 0 , it looks as shown in the adjacent figure. Choosing to= 3v'2/C,
h(O) = e- 4 · 5 = 0.011 or 1.13 of its peak value. Hence to = 3v'2k is a reasonable
choice to make the fi lter approximately realizable.
t0·
Figure 87.5-1
34 1
7.5-2.
hlt1
to
Figure 87.5-2
7.5-3. The uniL impulse response is the inverse Fourier transform of H(w}. Hence, we ha,'3
All the three systems are noncausal (and, therefore, unrealizable) because all the three
impulse responses start before t = 0.
For (a), the impulse response is a rectangular pulse starting at t = - 10- 6 . Hence,
delaying the h(t) by 1 µsecond will make it realizable. This will not change anything
in Lhe system behavior except the time delay of 1 µsecond in the system response.
For (b), the impulse response is a sine square pulse, which extends all the way to
- oo. Clearly, t his system cannot be made realizable with a fini te time delay. The
delay has to be infinite. However, because the sine square pulse decays rapidly (see
Figure 4.24d), we may truncate it at t = 10 - 4 , and then delay the resulting h(l) by
10 - 4 • This makes the fi lt er approximately realizable by allowing a time delay of 100
µseconds in the system response.
For ( c), the impulse response is 1, which never decays. Consequent ly, this filter cannot
be realized with any amount of delay.
7.6-J.
Ex=
1
00
- oo
x 2 (t)dt = - 1 2
2 na
1°"
-oo
e-
12
/"
2
dt
E - _1_ ..!:_
x - 27ra 2 ./2 _
loo
00
e-x212 dx -
-
,/2ii - _ 1_
2./2m1 - 2ufi
342
Ex = 2-j""
27r
IX(w)l2
-oo
dw = 2-j""
27r -oo
e- "
2
"'
2
dw
2-/
271" -oo
00
7r: [rect (~ )] dw
k 2k
2
=
lk
1T
2k2 -k dw =
11"
k
7.6-3. If x 2 (t) <==> A (w), then lhe output Y(w) = A(w)H(w), where H(w) is the lowpass
filter transfer funclion (Figure S7.6-3). Because this filter band 6./ _, 0, we may
express it as an impulse function of area 4tr6./. Thus,
Here we used the property g(x)c5(x) = g(O)c5(x) !Eq. (1.23a)J. This yields
y(t) = 2A(O)D.J
Next, because x 2 (t) <==> A(w), we have
}t{LD)
"
Figure 87 .6-3
and
Therefore
343
Interchanging the roles of x 1 (t) and x 2 (t) in the above development., we can show that
7.6-5. In the generalized Parseval's theorem in Prob. 7.6-4, if we identify g 1 (t) = sinc(Wt -
m7r) and g2 {t) = sinc(Wt - mr), then
7f
G 1 (w) = W rect (w)
W e~ 1• , and G2 (w) = -7r rcct ( - W ) e\V
j"ww
2 W 2W
Therefore
OO 1 ( 11' )2 Joo [rect ( W )]2 e W i!"-m)ww
J_
00
g1(t)g2(t)dt= 21T W -oo
2 w dw
n ;6 m
n=m
In evaluating the integqi.l, we used the fact that e±i2 "k = l when k is an integer.
7.6-6. Application of duality property jEq. (4.3l )J to pair 3 (Table 4.1) yields
'Poss - sc(t) m(t) cos 10, OOOt =cos lOOOtcos 10, OOOt
1
2 ...__,,.._,
- !cos 9000t +cos 11, OOOtJ
'--...---"'
LSB USB
(ii) For m(t) = 2 cos lOOOt +cos 2000t
'Posa sc(t) = m(t)coslO,OOOt = j2cosl000t +cos2000tJcosl0,000t
344
1
cos 9000t +cos 11, OOOt +
2[cos8000t +cos 12, OOOt]
[cos 9000t + 2'1 cos 8000t] +(cos 11, OOOt + 2'l cos 12, OOOt]
L8B U8B
-<CDC-
0
10 ' Mtw">
A)--""
w~
"
qcoc (I, co
11.tt t t· tt'2-K..,
01 CA.,) ..... B'I-\ q1<;
I
111<:.
-:. ..
'11
--;9~
t t-~K.
-1'.:ZK -8
Figure 87.7-1
J c...:~
tt
,.._ 81<\
.tt
'~"'
l 'fK
345
centered at ±3we as shown in Figure 87.7-2. The bandpass filter centered at
±we allows to pass the desired term ~m( t) cos wet, but suppresses the unwanted
term ~m(t) cos3wet· Hence, this system works as desired with the output
~m(t) cos wet.
(c) The minimwn usable value of we is 2TrB in order to avoid spectral folding at de.
(d)
The signal at point b consists of the baseband signal !m(t) and a modulated
signal !m( t) cos 2wet, which has a carrier frequency 2wet, not the desired value
wet· Both the components will be suppressed by the filter, whose center center
frequency is We· Hence, this system will not do t he desired job.
(e) The reader may verify that the identity for cosnwet contains a term cos wet whim
n is odd . This is not true when n is even. Hence, the system works for a carrier
cosn wet only when n is odd.
!
at-@
L;>...
- 3Wg..
~. -~ DI D w~ u; ...,,.
'~
3 4'~
:
.. ~ _ (DC..... I n w<!.-
'* ®
cZJ ->
Figure 87.7-2
7.7-3. This signal is identical to that in Figure 3.8a with period To (instead of 2n). We find
the Fourier series for this signal as
l + :;;:
x(t) = 2 2[COS Wet - 3
1 COS 3wct + S
l COS 5wet + · · ·]
y(t)::: m(t)x(t) = m(t) [~ + ~ (cos Wet - ~cos 3wct +~cos 5wct + ·· ·)]
The bandpass filter suppresses the signals m(t) and m(t) cos nwct for all n i= l. Hence,
346
the bandpass filter output is
2
km(t) cos Wet = -m(t) COS wet
1r
<
\L
a+@
C05 ( O l'(Jc.:f:.
. {
'- 4 ---t Lf
l--F";gu~1- F I+-->""'rri_2._l1 ) '•
·~
7.7-5. (a) Figure 87.7-5 shows the output signal spectrum Y(w).
(b) Observe that Y(w) is the same as M(w) with the frequency spectrum inverted,
that is, the high frequencies are shifted to lower frequencies and vice versa. Thus,
the scrambler in Figure P7.7-5 inverts the frequency spectrum. To get back the
original spectrum M(w), we need to invert the spectrum Y(w ) once again. This
can be done by passing the scrambled signal y(t) through the same scrambler.
7.7-6. Xa(t) =[A+ rn(t)] cos wet. Hence,
The first term is a lowpass signal because its spectrum is centered at w = 0. The
lowpass filter allows this term to pass, but suppresses the second term, whose spectrum
347
· y{cu)
-t5K 15 fc\
Figure 87.7-5
y(t) = A+ m(t)
When this signal is passed through a de block, the de term A is suppressed yielding
the output m(t) . This shows that the syst em cau demodulate AM signal regardless of
the value of A. This is a synchronous or coherent demodulation.
15 -
5 ~
0 --l\11H!lirH+ltttH-tf~~~lrlfHHil+
-E
10
· 0....++-+1+-+t+"~ltt+1~'itttttttt-ttt-tti
i
'
'-ID
I
l-.2Q
i--3~·' ~ ,
Figure 87.7-7
7.M-1. The signal x(t) = e- 0 tu(t) has Fourier Transform given by X(w) = JW~" and energy
E: = 2~. Using this information, MATLAB program MS7P2 is modified.
348
I. E_W = Energy contained in bandwidth W
W • 0; step = a; I. Initial guess and step values
X_squared = inline('l./(omega.-2+a.-2)','omega','a');
E = beta/(2•a); I. Desired energy in W
relerr = (E-0)/E; I. Initial relative error is 100 percent
while(abs(relerr) > tol) ,
if (relerr>O), I. W too small
W=W+step; I. Increase W by step
elseif (relerr<O), I. W too large
step = step/2; W = W-step; I. Decrease step size a.nd then W.
end
E_W c 1/(2•pi)*quad(X_squared,-W,W,[],(),a);
relerr = (E - E_W)/E;
end
7.M-2. To solve this problem, program MS7P2 is modified to solve for the pulse width to
achieve a desired essential bandwidth, rather than solving for the essential bandwidth
that corresponds to a desired pulse.
349
/. E_W = Energy contained in bandwi dt h W
tau = 1; st ep = 1 ; /. Initial guess and s t ep values
X_squared = inl ine ( ' (tau•MS7P1(omega•tau/ 2)) .-2' ,'omega','tau' ) ;
E_W = 1/(2•pi) • quad (X_squared, - W,W , (),(),tau);
E = beta•tau ; /. Desired energy i n W
rel err = (E - E_W)/E ;
while(abs(relerr) > tol) ,
i f (relerr>O), /.tau too small
tau=tau+step; /. Increase tau by step
elseif (relerr<O), /.tau too large
step = step/2;
tau = tau-step ; % Decrease step size and then tau.
end
E_W = 1/ (2*pi) *quad(X_squared , - W,W , [) ,(), tau);
E = beta• tau ; 'i. Desired energy in W
rel er r = (E - E_W)/E;
end
350
!. OUTPUTS: a = decay parameter
!. E_W = Energy contained in bandYidth W
a = 1; step = 1; !. Initial guess and step values
X_squared = inline('1 . /(omega.-2+a . -2)','omega' ,'a');
E_W = 1/(2•pi)•quad(X_squared,-W,W,[],[],a);
E = beta/(2•a); !. Desired energy in W
relerr = (E - E_W)/E;
while(abs(relerr) > tol),
if (relerr<O), !. a too small
a=a+step; !. Increase tau by step
elseif (relerr>O), !. a too large
step == step/2;
a = a-step; !. Decrease step size and then tau.
end
E W 1/(2*pi) *quad(X_squared, - W,W,[),[],a);
E = beta/(2*a); !. Desired energy in W
relerr = (E - E_W)/E;
end
351
function [W,E_W] = MS7P2mod4(beta,tol)
I. MS7P2mod4.m
I. Function M- file computes essential bandwidth W for a
I. unit-amplitude, unit duration triangle function.
% INPUTS : beta = fraction of signal energy desired in W
I. tol • tolerance of relative energy error
% OUTPUTS: wm essential bandwidth [rad/s]
I. E_W • Energy contained in bandwidth W
W = O; step • 1; I. Initial guess and step values
X_squared = inline('(sqrt(2)/2•MS7P1(omega/4)).-4','omega');
E = beta/3; % Desired energy in W
relerr = (E-0)/E; I. Initial relative error is 100 percent
while(abs(relerr) > tol),
if (relerr>O), I. W too small
W=W+step; I. Increase W by step
elseif (relerr<O), I. W too large
step= step/2; W = W-step; I. Decrease step size and then W.
end
E_W = 1/(2•pi)*quad(X_squared,-W,W);
relerr = (E - E_W)/E;
end
Dn
Tsmc
= To . (n1TT) .
70
(a) Setting 10 = 27r and r = 2r./3 yields
1 . (n7r)
Dn = 3smc 3 .
352
MATLAB is used to evaluate and plot the first ten coefficients.
>>tau= 2*pi/3; T_O = 2*pi; n = [0:10];
>> D_n = tau/T_O*MS7Pl(n*pi*tau/T_O);
>> stem(n,D_n,'k'); xlabel('n'); ylabel('D_n' );
>> axis((-0.5 10.5 -0.2 0.55]);
...
O.•
0)
0.2
o•
0. 1
- 0.1
X(w) = _e__
-w
2
/41oo e-(t'+3w2/v'2)/2dt'.
J2 -00
X(w) = ../1ie-w2/4.
MATLAB is used to plot x(t) and X(w).
353
>> subplot(211); plot(t,x,'k');
>> xlabel('t'); ylabel('x(t)');
>> subplot(212); plot(t,X, 'k');
>> xlabel ('\omega'); ylabel('X(\omega)');
...
0 .6
~
o,c
02
• ... ••
-10
-· "'
Figure $7.M-6 confirms that X(w) is just a scaled and stretched version of x(t). This"is
something remarkable; the Fourier Transform of a Gaussian pulse is itself a Gaussian
pulse!
354
Chapter 8 Solutions
8.1-1. The bandwidths of x 1 (t) and x 2 (t) are 100 kHz and 150 kHz, respectively. There-
fore the Nyquist sampling rates for x 1(t) is 200 kHz and for x2(t) is 300 kHz. Also
x 1 2 (t) ~ i,,F1 (w) * F1 (w), and from the width property of convolution the band-
width of x 12(t) is twice the bandwidth of x 1 (t) and that of x 23 (t) is three times the
bandwidth of x 2(t) (se also Prob. 4.3-10). Similarly the bandwidth of x1 (t)x2(t) is the
sum of the bandwidth of x 1(t) and x2(t). Therefore the Nyquist rate for x1 2(t) is 400
kHz, for x 23 (t) is 900 kHz, for x 1 (t)x 2(t) is 500 kHz.
8.1-2. (a)
w
sinc2(1007rt) ~ 0.016(- )
4 007r
The bandwidth of this signal is 200 r. rad/s or 100 Hz. The Nyquist rate is 200
Hz (samples/sec).
(b) The Nyquist rate is 200 Hz, the same as in (a), because multiplication of a signal
by a constant does not change its bandwidth.
(c)
. . 2 w 1 w
smc(l007rl) + 3 smc (607rt) <=> 0.01rect( 7T) + 6( 7T)
200 20 240
The bandwidth of rect( 2~,J is 50 Hz and that of 6( 2~,.) is 60 Hz. The band-
width of the sum is the higher of the two, that is, 60 Hz. The Nyquist sampling
rate is 120 Hz.
(d)
w
sine(50?Tt) 0.02re<:t(-)
20011
w
sinc{l007rt) 0.01rcct( 1T)
400
The two signals have bandwidths 25 Hz and 50 Hz respectively. The spectrum
of the product of two signals is l/27r times the convolution of their spectra.
From width property of the convolution, the width of the convoluted signal is
the sum of the widths of the signals convolved. Therefore, the bandwidth of
sinc(507rt)sinc(1007rt) is 25 + 50 = 75 Hz. The Nyquist rate is 150 Hz.
8. 1-3. (a)
355
Figure S8. l-3 shows the spectrum as a function off in Hz. The spectrum in the
range I/I < 14 Hz is the spectrum X(w).
(b) The Nyquist rate is 28 Hz. Hence 25% above this resu lts in the sampling rate
!. = 35 Hz. Figure S8.l-3 shows the sampled signal spectrum over the range
Ill < 50 Hz. To reconstruct. x(t), we pass the sampled signal through an ideal
lowpass filter of cutoff freq1,1ency anywhere between (14 + e) Hz to (21 - e) Hz
where € is a small positive number. The filter gain is T = 1/ /,, = 1/35.
.:t
_t ·- ·-
J ... .
.J_ _t-·
4,q
- 1-tt-\ ....-- -
Figure S8.l-3
br(t) ~ Ws L b (w - nw.. )
n=-OC>
Hence
00
= T L X(w - nw.)
n =-oo
1
T L 00
X(w - nw.,)e- 1 ""'•T
.
s n= - oo
356
8.1-5. If the corrupted spectrum is not filtered out, we need the minimum sampling rate
22 Hz. This is clarified by Figure S8.1-5a. For / ,. = 22, the uncorrupted spectrum
remains intact and can be recovered by a lowpass filter of cutoff frequency 10 Hz.
When the corrupted spectrum is suppressed, the resulting signal spectrum band is
only lOHz. Hence it is adequate to use /,. = 20 Hz, as shown in Figure S8.l-5b.
•1
...
.....____.......- -··- .......... .
._J _~-- - ·:--- --··-·-
JD ..2.d
-Fs
Figure S8.l-5
8.1-6.
The spectrum IX(w)I shows that most of the signal energy is concentrated within the
band of 1 Hz. It can be shown that 90.283 energy is contained within the band of 1
Hz. If we use 903 energy criterion for bandwidth, sampling rate of 2 Hz is adequate.
However, for a better approximation, (higher energy bandwidth criterion), we may go
to l s = 4 llz. Theoretically, of course, f 3 = oo.
8.1-7. (a)
X(w) = 6. (~)
207T
I- 1Tl<5(w + 207T) + <5(w - 207T)J
The bandwidth is 10 Hz. There is an impulse at lOHz, as seen from X(w) shown
in Figure S8. l-7a. The Nyquist rate is 20 Hz. Hence, ls = lOHz will not permit
reconstruction of x(t). This is verified from the sampled signal spectrum in Figure
S8. l-7a, shown as a function off in Hz.
(b) The Nyquist rate is 20 Hz. Hence the sampling rate /,. = 20 Hz is adequate
despite the fact that x(t) contains an impulse at the highest frequency 10 Hz.
This is because, the impulse component is cos 207Tt.
To reconstruct x(t) from the spectrum in Figure S8.1-7b, we need a n ideal lowpass
filter of cutoff frequency IO Hz and gain T = 1/20. Because the rect function
value is 0.5 at the edge (cutoff), t he lowpass filter gain at the cutoff frequency 10
Hz is 0.5 x 1/20 = 1/40. Hence for the input of an impulse of strength 407r at
357
X(fJ
f.s =- J0 f-1 15
-?ii 0 21f
which is in error.
+ e, it will miss the impulses at f = ±10 entirely;
If the fil ter bandwidth is Bo
and the reconstructed signal would be
358
iso
~~~-.. . _. , _ -~o ID 10
frequency, the spectrwn will shift by ±60n(n = 1, 2, 3, ··-)as shown in Figure S8. I-sp.
The negative frequency spectrum is labeled Po and the positive frequency spectrum
is labeled Q 0 . Both these segments repeat with period 60 Hz. Let us label Pk as the
segment P0 shifted by 60 kHz and P-k is Po shifted by -60 kHz. Similarly Qk and
Q_k represent Q 0 shifted by ±60 kHz. The sampled signal spectrum in Figure S8.l-8b
shows the labels of various repeated segments. It is clear from Figure S8.1-8b that
we can reconstruct. the signal x(t) from this spectrum by passing it through an ideal
bandpass filter (shown dotted) centered at 25 Hz and having bandwidth lOHz. The
filter gain is T = 1/60.
Figure S8.1-8c shows the sampled signal spectrum with ls =20 Hz (spectrum repeating
at every 20 Hz) . It is clear from this figure that we can reconstruct x(t) from this
spectrum also because the original segments Po and Qo are still intact without being
overlapped by any other repeating segments.
Figure S8.l-8d shows the spectrum of the signal y(t) sampled at a rate 20 Hz (spectrum
repeating at every 20 Hz). The figure shows that P0 overlaps with Q_ 2 and Q 0 overlaps
with P2 . Hence it is impossible to reconstruct y(t) from th.is spectrum.
8.1-9. This problem is trivial when worked out in the frequency-domain. The sampled signal
spectrum is given by
1 00
X(w) = T L X(w - 27rnfs)
n=:.-oo
We repeat the spectrum periodically with period (Ji + h) Hz, as shown in Figure
88.1-9. The amplitude at the origin is 1/T = 11 + h- From Figure S8.l-8a, it is
obvious that the resulting spectrum X(w) is constant for all wand has a value Ji+ /2.
Hence
X(w) = !i+h
and
x(t) = U1 + h)c(t)
Hence all the samples of x(t} at a rate l s = Ji+ hare zero except the sample at t = 0
359
:;q ,· ( X(u5) { .
~!fl-C6o... .
· 100
- ··· _ ! _,.,.-··.
-loD
:./___ J_~-i~'--'- ' M ·Y._ __
--.2.v j 20 30 /./ D GO gv qv 10.0
! (b) f ~
t
or n = 0.
I
.~
ji· i(
- ( f~ -rf1 J -·f-~.
,...,_ ..~
..._ f \
4-+-f
9---
' 1
X(lJ' I :· Cf..+ f'2...)
- -------· ····-·-· __
-· . --· ·- -· ·-··\-·- .
(.h ).
\
o\
Figure 88.1-9
8.1-10. The Maximum Information Rate: 2 Pieces of Information per Second per Hertz.
A knowledge of the maximum rate of information that can be transmitted over a
channel of bandwidth B Hz is of fundamental importance in digital communication.
\Ve can demonstrate a scheme which allows errorfree transmission of 2B independent
pieces of information per second over a channel of bandwidth B Hz. Recall that a
continuous-time signal x(t) of bandwidth B Hz can be constructed from its Nyquist
samples (which are at a rate of 28 Hz) using the interpolation formula (8.11). Letting
the the kth piece of information equal to x(kT), the kth Nyquist sample, and using
360
the interpolation formula (8.11), we can construct. a signal x(t) t.ha.t is bandlirnited to
B Hz. Clearly, this signal can be transmitted errorfree over the channel of bandwidth
B Hz. Moreover, the 28 pieces of information are readily obtained (errorfree) from
this signal by taking its Nyquist samples.
T his theoretical rate of communication assumes a noisefree channel. In practice, chan-
nel noise is unavoidable, and consequently, this rate will cause some detection errors.
We shall prove the converse of this result using the method of reduclio ad absur-
dum. Independent piece implies that each piece of information can be any one of the
(uncountably) infinite number of amplitudes. To prove the converse, let us assume
that a scheme exists that can transmit more than 2B independent pieces of infor-
mation/second. If this were the case, the interpolation formula implies that we can
t ransmit a signal of bandwidth higher than B Hz over a channel of bandwidth B Hz,
which is not possible.
8.1-11. (a)
This is a rect spectrum of band 4rr rad/s or 2 Hz. To find the sampled signal
spectrum X(w) , we multiply X(w) by ~ = 4 and repeat it periodically with
period / 3 = 4. the result is as shown in Figure 88.l-lla. Thus X(w) = 1 for i!;ll
w.
(b) To reconstruct x(t) from X(w) , we pass X(w) through an ideal lowpass filter of
gain T = ~and bandwidth f,/2 = 2Hz (Figure 88.l-llb) . The input is X(w) = l.
Hence the output is ~rect (~). The output is indeed x(t), as expected.
(c) If we sample x(t) at a rate 2 Hz (T = 1/2),
_
X (w)
00
= T1 L X(w - 21Tnfs) = 21 L
00
rect
(w-S7r41Tn)
n=-oo n = - oo
Thus X(w) is obtained by repeating 4rect (~) with period w = 41T (or J. = 2
Hz), as shown in Figure 88.l-llc. The two adjacent spectra, each of amplitude
1/2 overlap, yielding a constant value X(w) = 1. When this X(w) is applied to
t he fil ter in Figure S8.l-llb, the output be again x(t) = trect(w/87r).
(d) We get identical result using J. =l Hz, where X (w) repeats every 211", but the
amplitude is only 1/4. For any w, we get 4 overlapping spectra, yielding X (w) =
1.
(e) The signal x(t) = sinc(47Tt) is shown in Figure S8.l - lld. Observe t hat x(t) = 0
at t = ~ for a ll positive and negative integer value of n. This means sampling
x(t) at a rate 1f t
or T = yields all zero valued samples except at t = O(N = 0).
(f) T he signal x(l) = sinc(47rl) is bandlimited to B = 2 Hz. Hence, its Nyquist
rate is 4 Hz and the Nyquist interval is T = 1/4. As seen from Fig. 88.1-lld,
the samples of this signal a t t = nT are zero at l = M/4 for any positive or
negative integer M. In other words, if the sampling rate is 4M Hz, where M any
arbitrarily large inlegcr, the samples are given by
x(nT ) = { 1 n =0
0 n= ±J,±2,:.!:3, ...
361
Thus, we obtain the same set of samples if we sample the signal at a rate 4 Hz or
8 Hz, or 12 Hz, or 4M Hz, for any integer value of M. Therefore, regardless of
the value of M, we can reconst ruct the signal x(t) = sinc(4r.t) from these samples
using the interpolator in Fig. S8.l -llb.
(.b}
Figure S8. l- ll
8.2-1. The signal x(t) = sinc(200r.t) is sampled by a rectangular pulse sequence Pr(t) whose
period is 4 ms so that the fundamental frequency (which is also the sampling frequency)
is 250 Hz. Hence, Ws = 5007r. The Fourier series for pr(t) is given by
00
Use of Eqs. (3.66) yields Co=~' Cn = ;:,, sin( '~"'), that is,
362
the output y(t) is
y(t) = 0.2x(t)
Ilecause the spectrum X(w) has a zero value in the band from 100 to 150 Hz, we can
use an ideal lowpass filter of bandwidth B Hz where 100 < B < 150. But if B > 150
Hz, the filter will pick up the unwanted spectral components from the next cycle, and
the output will be distorted.
0 .\i7
to0 . ).50 . .
Figure S8.2-1
8.2-2. (a.) When the input is o(t), the input of the integrator is [o(t) - o(t -T)J. And, h(t),
the output of the integrator is:
H(w) = Tsinc(w;)e-;wT/2
IH(w)I = Tlsinc(w:)I
The amplitude response of t he filter is shown in Figure S8.2-2b. Observe that
the fi lter is a lowpass filter of bandwidth 271' /T rad/s or l/T Hz. The impulse
response of the circuit is a rectangular pulse. When a sampled signal is applied at
the input, each sample generates a rectangular pulse at the output, proportional
to the corresponding sample value. Hence the output is a staircase approximation
of the input as shown in Figure S8.2-2c.
1 ttt fc>) j
hlt)
T t:-!> -~ .. c '2.1T I 2T
-t:-
T T
Figure S8.2-2
363
8.2-3. (a) Figure S8.2-3a shows the signal reconstruction from its samples using the first-
order hold circuit. Each sample generates a triangle of width 2T and centered
at the sampling instant. The height of the triangle is equal to the sample value.
The resulting signal consists of straight line segments joining the sample tops.
(b) The frequency response of this circuit is:
h(t) = rect (
t - TT/2)
and
Hzoh(w) = Tsinc(wT/2)e-iwT/ 2
The frequency response of the cascade two sections of the ZOH circuits is given
by
Hcascade(w) = T 2 sinc2 (wT/2}e-iwT
ln part {b) , we found that the frequency response of an FOH circuit as
This shows that the frequency response of the cascading two ZOH circuits is T
times the frequency response of the FOH circuit with time delay T seconds. the
time delay is a desirable feature as it makes the FOH circuit causal, and therefore,
realizable. Thus, the cascade of two ZOH acts identical to an FOH circuit except
for the amplification by factor T and delay by T seconds.
8.2-4. Consider a sampling pulse p(t) of arbitrary width. The development in the text in
Sec. 8.1-1 applies to pulse of arbitrary width. Hence, we can use the result in Eq.
(8.7).
00
364
where C; are the coefficients of the compact trigonometric Fourier series for PT(t) with
p(t) = e- tu(t). Thus
0
00
This is a periodic signal of period T and we can find the Fourier series for PT(t)
00
Now, we use the arguments in Sec. 8.1-1 to show that x(t) can be constructed from
x(t) = x(t)P'.T·(t), provided the sampling rate is no less that 2B Hz or T < l/2B.
8.2-5. The signal x(t), when sampled by an impulse train, results in the sampled signal
x(t)o·r(t) (as shown in Figure 5.ld). If this signal is transmitted through a filter (Figure
88.2-5a) whose impulse response is h(t) = p(t) = rect( 0 .~25 ), then each impulse in the
input will generate a pulse p(t), resulting in the desired sampled signal shown in Figure
88.2-5. Moreover, the sampled signal spectrum (impulse sampling) is f, L:~=-oo X(w-
nw3). Hence, the output of the filter in Figure 88.2-5a is
... F(w)
T ::: 0·025
-Tp
l .b )
Figure 88.2-5
8.2-6.
fs = 20Hz
lfal = If - mfsl Ifal ~ fs/2 = 10
365
(a) f= 8 Hz is less that / 6 /2 = 10 Hz.
Hence, this frequency is not aliased and l! (ll = f = 8 Hz
(b) J = 12 Hz
l/(ll = 112 - 201= 8 Hz
(c) f = 20 Hz
If.. I = 120 - 201 = 0 Hz
(d) J = 22 Hz
I/al = 122 - 201=2 Hz
(e) f = 32 Hz
I/al = 132 - 401 = 8 Hz
8.2-7. (a) lo < 30 Hz and l! .. I = 20 Hz.
Hence, 20 = If o - 60ml ==> m = 0 and Jo = 20Hz
(b) 20 = l!o - 60ml ==> m = 1 and /o = 40Hz
(c) 20 = lfo - 60ml ==> m = 1 and /o = 80Hz
(d) 20 = l!o - 60ml ==> m == 2 and fo = lOOHz
8.2-8.
8.2-9. (a) The output in Eq. (8. lla) is the output of an ideal lowpass fi lter of bandwidth
B = l / 2T Hz. Clearly its bandwidt h must be~ l/2T llz.
(b) Suppose there is another signal x(t) that passes through the samples x(nT) and
has bandwidth smaller than that of x(t) obtained through Eq. (8- ll a). Clearly,
both x(t) and x(t) have bandwidth ~ l /2T and the signal x(t) - x(t) also has
bandwidth ~ l /2T Hz. Hence, we can reconstruct the signal x(t) - x(t) from
samples of x(t) -x(t) at a rate l /T Hz and using these samples in Eq. (8.lla). But
366
because both x(t) and x(t) pass through sample x(nT), the samples of x(l)-x(t)
at a rate l/T Hz are zero for all n. Clearly :r:(t) - i:(t) = 0 and x(t) = x(t).
8.2- 10. T = 1/ Rand the sample values are
We can use Eq. (8.lla) to reconstruct p(l) from these sample values. There is only
one nonzero-valued sample. Hence, we obtain
and
P(w) = ~rect ( ~)
R 21l'R
This is the only signal that has bandwidth 7rR rad/sec or R/2 Hz with samples p(O) =1
and p(nT) = 0 for all n '# 0.
8.2-11. Let us sample the pulse p(t) at a rate R Hz or T = l/R sec. The spectrum of the
sampled signal consists of TP(w) = RP(w) repeated periodically with perio~ R Hz,
as shown in Figure S8.2-ll. Because of the odd symmetry of P(w) about the dotted
axis, the overlapping spectra add to a constant value 1 for all w. Hence
,:4. 1·~u))
! ·.
Figure S8.2-ll
8.2-12. The Nyquist interval T = l/2B. The Nyquist samples are x(±n/2B) for (n =
0, 1, 2, 3, ... ). We are given that x(O) = x(l/2B) = 1 and x(n/2B) = 0 for all other n.
Hence, from Eq. (8. ll b)
367
8.2-13. (a) From Eq. (8.lla)
00
x(t) = L x(nT)sinc(Wt - mr) W =~
T
n= - oo
and
L: x(t)dt /_
00
-oo n=-oo
00
L sinc(Wt- mr)dt
L 00
x(nT) / _00 sinc(Wt - mr)dt
n=-<x:> -oo
00 /_00 sinc(t)dt
W
1
L x(nT)
n =-oo -oo
= L: x(t)x•(t)dt
1 -oo
lx(t}l2dt = l~ L
n=-oo
x(nT)x· (nT) =T L
n=-oo
lx(nT)J 2
8.2-14. Assume a signal x(t) that is simultaneously t imclimited and bandlimited. Let X(w) =
0 for Jwl > 27rD. Therefore X(w)rect( 4 : '8 ,) = X(w) for B' > B. Therefore from the
time-convolution property (4.42)
368
8.3-2. (a) The Nyquist rate is 2 x 4.5 x 106 = 9 MHz. The actual sampling rate = 1.2 x 9 =
10.8 MH7..
(b) 1024 = 2 10 , so that 10 bits or binary pulses are needed to encode each sample.
(c) 10.8 x 106 x 10 = 108 x 106 or 108 Mbits/s.
8.3-3. (a) For L = 16, we need a 4-bit binary code because 16 = 24 • The natural binary
code is given in the text. We give here two's complement code, which is used in
signal processing.
For a quaternary code, we use four symbols 0, 1, 2, and 3. For this code, we need
only a group of 2 symbols to form 16 combinations (4 x 4) = 16. One possible
quaternary code is given below.
and
bz log 2 = b,1 log 4 = 2b4 log 2
Hence
(0.2)(V) V
quantization error < = -
- 100 500
369
is encoded by 9 bits. Therefore, each signal uses 9 x 2400 = 21.6 kbits/second. Five
such signals are multiplexed. Hence, we need a total of 5 x 21.6 = 108 kBits/second
data bits.
8.4-1. In section 8.4, we have shown that when a timelimited signal x(t) is repeated periodi-
cally with a period T0 > r (the signal duration), the Fourier series coefficients for the
resulting periodic signal xr0 (t) are proportional to the samples of X(w), the Fourier
transform of x(t) at frequency interval of Jo= f.o Hz. This result is quite general and
applied even if x(t) is bandlimited, and therefore, nontimelimited. To sho,w this we
convolve x(t) with unit impulse train c5r(t). This will result in periodic repetition of
x(t) with period T.
Moreover
y(t)=x(t)*c5r(t) <==}
2~X(w)[~ow.(w)]
1 2n
TX(w)ow,(w) Ws = T
In the present case, T = L.iP and the fundamental frequency is w = 2n/T.
Therefore
Y(w) 1
= TX(w) ~
L- c5 ( w - T
27rn)
n = -oo
'it
Hence Y(w) represents the spectrum ~X(w) sampled at intervals of~ Hz. This means
y(t) is a periodic signal with fundamental frequency Jo = ~
00
y(t) = L Dnejnw, t
n= - oo
where
where
1
Co= Do= TX(O)
and
and
370
8.5-1.
1 1
To = - = - = 20ms
Jo 50
B = 10000 Hence f s 2:: 2B = 20000
1 1
T = f,, = 20000 = 50µs
To 20 x 10- 3
No = T = 50 x 10-6 = 400
20 x 10- 3
T = = 39.0625 µs
512
This gives To = 20 ms, f 0 = 50 Hz. And
1
fs =T= 25600Hz
There are also other possibilities of reducing T as well as increasing the freque~y
resolution. ··
Let us choose T = 1/8. Also To = 4. Therefore, N 0 = To/T = 32. The signal x(t)
repeats every 4 seconds with samples every 1/8 second. The samples are Tx(kT) =
(1/8)x(k/8). Thus, the first sample is (at k = 0) 1 x (1/8) = 1/8. The 32 samples are
(starting at k = 0)
17 3 5 13 1 1
8' 64' 32' 64' 16' 64' 32' 64' O, o, O, O, O, O, O, O,
113153 7
o, O, o, O, O, o, o, o, o, 64) 32, 64) 16' 64' 32) 64
The samples of x(t) and g(t) are shown in Figure 88.5-2.
8_~~
, I:
t '-
Figure 88.5-2
8.5-3.
1
x(l) = e-tu(t) X(w)=-.-
JW+ 1
(a) We take the folding frequency fs to be the frequency where IX(w)I is 13 of its
371
peak value, which happens to be 1 (at w = 0). Hence,
1
IX(w)I ~ - = 0.01=>w=27rB = 100
w
This yields B = 50/7r, and T :S: l/2B = 7r/l00. Let us round T to 0.03125,
resulting in 32 samples per second. The time constant of e-t is 1. For T 0 , a.
reasonable choice is 5 to 6 time constants or more. Value of T 0 = 5 or 6 results
in No = 160 or 192, neither of which is a power of 2. Hence, we choose To = 8,
resulting in N 0 = 32 x 8 = 256, which is a. power of 2.
(b)
1 1
IX(w)I = .Jw2 + 1 ~ ~ w » 1
0.997T
- - = tan
- 1 w
- => W = 63.66a = 63.66 rad/sec.
2 a
This yields B = ~ = 10.13 Hz. Also T :S: 1/2B = 0.04936. This results in the
sampling rate ~ = 20.26 Hz. Also To= 8 as explained in part (a). This yields
No= 20.26 x 8 = 162.08, which is not a power of 2. Hence, we choose the neh
higher value, that is N0 = 256, which yields T = 0.03125 and To = 8, the same
as in part {a).
8.5-4. (a)
2
x(t) = t2 + 1
Following the approach of Prob. 8.5-2, we find that the peak value of IX(w)I =
27Te- lwl is 27f (occurring at w = 0) . Also, 27fe-lwl becomes 0.01 x 27T (13 of
the peak value) at w = In 100 = 4.605. Hence, B = 4.605/27r = 0.733 Hz, and
T S l/2B = 0.682. Also,
2
x(O) = 2 and x(t) := t 2 t » 1
372
The energy within the band from w = 0 to W is given by
Ew = -81T2 1 W e- 2w dw = 27T(l - e- 2 W )
27T 0
But Ew = 0.99E1 = 0.99 x 27T. Hence,
5t
.L
·a
~~
~~-t-~~~--r1~~~~~~~--....~~~---~ ~-f>
i- ,. ~ 3l..
Figure 88.5-5
373
(c) Yes, X(r) is a valid DFT. Since X{r) is periodic with fundamental period equal
to 20, the DFT length can be any integer multiple of 20. That is, the DFT length
is N = 20, 40, 60, .... Since X (r) is not conjugate symmetric, the time-domain
signal x[nJ is complex.
(d) Yes, X(r) is a valid DFT. Since X (r) is periodic with fundamental period equal
to eight, the DFT length can be any integer multiple of 8. That is, the DFT
length is N = 8, 16, 24, . . .. Since X(r) is conjugate symmetric, the signal x[nJ is
real.
(e) Yes, X(r) is a valid DFT. Since X(r) is periodic with fundamental period equal
to ten, the DFT lengt h can be any integer multiple of 10. That is, the DFT
length is N = 10, 20, 30, ... . Since X(r) is not conjugate symmetric, the time
domain signal x[nJ is complex.
8.M-1. Given signal x[n] with DFT Xr = '2:~~ 0 1 xln]e-1rno 11 , the time-shifting property tells
us that y[nJ = x[n - n0 ] has DFT Yr = e-;rnono Xr. Therefore, if MATLAB computes
the DF'T Xr for a signal x[n] assuming that it starts at 0, then the DFT of the signal
shifted to start at no is found by scaling Xr by e-;rnono.
In MATLAB, this is easy to accomplish. Assuming constants n_O and Omega_O are
defined and DFT X is already computed, the corrected DFT X_shift is computed by
X_shift = exp(-j•([O:length(X)-1)')•0mega_O•n_O}.•X(:).
8.M-2. Ideally, x 1 [n] = & 2 ,,.n 3o/ioo + e32 'm33 /•00 is characterized by two spikes of equal heigiht
located at fr = 0.30 and fr = 0.33. For most cases, two OFT magnitude plots are
included: the first covers the entire range of digital frequencies and the second details
the range near the true frequency content of x 1 lnJ.
(a)>> n = (0:9); xl = exp(j•2•pi•n•30/100)+exp(j•2•pi•n•33/100);
>> X1 = fft(xl); f_r = (O :length(xl)-1)/length(xl);
>> stem(f_r-0.5 ,fftshift(abs(Xl) ),'k.');
>> xlabel('f_r'); ylabel(' IX_l(f_r)I');
>> axis((-0.5 0.5 0 20));
>&
..
"
For this DFT, only ten samples of xdnl are used. As a result, the DFT has
only 10 frequency bins uniformly spaced over the frequency interval l-0.5, 0.5).
As Figure S8.M-2a shows, there is insufficient frequency resolution to separately
identify the two closely spaced exponentials at fr = 0.30 and fr = 0.33.
374
(b) >> n = (0:9); x1 = exp(j•2*pi*n*30/100)+exp(j*2*pi•n•33/100);
>> xl = [x1,zeros(1,490)];
>> Xl = fft(xl); f _r = (O:length(xl)-1)/length(xl);
>> subplot(211),stem(f _r-0.5,fftshift(abs(X1)), 'k .');
>> xlabel('f_r'); ylabel('lX_l(f_r)I');
>> axis([-0.5 0.5 0 20));
>> subplot(212),stem(f_r-0.5,fftshift(abs(X1)),'k.');
>> xlabel('f_r'); ylabel('IX_l(f_r)I');
>> axis([0.2 0.4 0 20]);
For this DFT, only ten samples of x 1 [n] are used but the sequence is zero-padded
to a length of 500. Although the DFT has 500 frequency bins uniformly spaced
over the frequency interval l-0.5,0.5), there is insufficient information about
xi[n] (only 10 samples) to resolve the closely spaced exponentials at fr = 0.30
and fr = 0.33. Using t he picket fence analogy, zero-padding increases the number
of "pickets" in our DIT fence, but it docs not change what lies behind the fence.
Still, Figure S8.M-2b does show a concentration of signal energy centered at
fr = 0.315, the average of the lwo exponential frequencies.
(c) >> n ~ (0:99); x1 = exp(j*2*pi*n*30/100)+exp(j*2•pi*n*33/100);
>> Xl = fft(xl); f _r = (O:length(x1)-1)/length(x1);
>> subplot(211),stem(f_r-0.5,fftsbift(abs(X1)),'k.');
>> xlabel('f_r'); ylabel('IX_l(f_r)I');
>> axis([-0.5 0.5 0 110));
>> subplot(212),stem(f_r-0.5 ,fftshift(abs(X1)),'k.');
>> xlabel('f_r ') ; ylabel('IX_l(f_r)I');
>> axis((0.2 0.4 0 110));
For this DFT, 100 samples of x 1 [n] are used. As a result, the D.FT has 100
frequency bins uniformly spaced over the frequency interval [- 0.5, 0.5). T he set
of DFT bins also happens to include both exponential frequ encies fr = 0.30
and f,. = 0.33. As Figure S8.M-2c shows, the two exponentials are each easily
identified. It is rare, however, for the windowed data record x 1 [n] to contain
an integer number of periods, as occurs in this case, so Figure S8.M-2c paints a
somewhat optimistic picture of the performance of this 100-point DFT.
(d) >> n = (0:99); xl = exp(j*2•pi*n*30/ 100)+exp(j•2*pi*n*33/ 100) ;
375
..
l(IO
••
20
100
20
'·
Figure S8.M-2c: DFT of 100-point. x 1 [n j.
>> xl = [xl,zeros(l,400));
>> Xl • fft(xl); f _r = (O:length(xl) - 1)/length(xl);
>> subplot(211),stem(f_r-0.5,fftshift( abs(Xl)),'k.');
>> xlabel('f_r'); ylabel('IX_l(f_r)I');
>> axis( (-0.5 0.5 0 110));
>> subplot(212),stem(f _r - 0.5,fftsh ift(abs(X1)),'k.');
>> xlabel('f_r'); ylabel('(X_l(f_r)(');
>> axis([0.2 0.4 0 110));
..
20
.1, 11.
-i..) -<:i • «0.3 - 0.2 -0.1 Ot 01 0.3 O• OS
For this DFT, 100 samples of x 1 [nJ are used, and the sequence is zero-padded
to a length of 500. The DFT has 500 frequency bins uniformly spaced in t he
frequency interval {-0.5, 0.5), and t hese bins include both exponential frequencies
fr = 0.30 and fr = 0.33. As shown in Figure S8.M-2d, the two exponentials
can be separately identified, but there is some added clutter throughout the
frequency spectrum. This clutter is the result of applying a finit~length window
to the signal xi[n). Although Figure S8.M-2d may appear less accurate than
Figure S8.M-2d, both contain the same information about x 1 [n). Using the picket
fence analogy, Figure S8.M-2d uses more pickets than Figure S8.M-2c, but the
background behind both is the same. In many respects, Figure S8.M-2d paints a
376
more honest picture of the data than Figure S8.M-2c; recall that Figure S8.M-2c
looks uncommonly good since data record x 1 [nJ includes an integer number of
periods (a rare occurrence).
8.M-3. Ideally, x~dnl = e12rrn 3 o;ioo +e12"n3 i.s;ioo is characterized by two spikes of equal height
located at fr = 0.30 and fr = 0.315. For most cases, two DFT magnitude plots are
included: the first covers the entire range of digital frequencies and the second details
the range near the true frequency content of x 2 lnJ .
(a)>> n = (0:9); x2 = exp(j*2*pi*n*30/100)+exp(j*2*pi*n*31.5/100);
>> X2 = fft(x2); f_r = (O:length(x2) - 1)/length(x2);
>> stem(f_r-0.5,fftshift(abs(X2)),'k.');
>> xlabel('f_r'); ylabel('IX_2(f_r)I');
>> axis((-0.5 0.5 0 20));
1$
"
~ 10
..
•.___,,•~_,_~~·~_._~~'~-..__
-0.5 -0 A ..0.)
1 __,,I~_...._~~'~_,
"C,2 -0, 1 0 0, 1 02 0 ,J O.A 0.S
'·
Figure S8.M-3a: DFT of 10-point x 2 [nJ.
For t his DFT, only ten samples of x 2 [nJ are used. As a result, the DFT has
only 10 frequency bins uniformly spaced over the frequency interval [-0.5, 0.5).
As Figure S8.M-3a shows, there is insufficient frequency resolution to separately
identify the two closely spaced exponentials at fr = 0.30 and fr = 0.315.
(b) >> n = (0:9); x2 = exp(j*2*pi*n*30/100)+exp(j*2*pi*n*31.5/100);
>> x2 = [x2,zeros(1,490)];
>> X2 = fft(x2); f_r = (O:length(x2) - 1)/length(x2);
>> subplot(211),stem(f_r- 0.5,fftshift(abs(X2)),'k.');
>> xlabel('f_r'); ylabel('IX_2(f_r)I');
>> axis((- 0.5 0.5 0 20));
>> subplot(212),stem(f_r-0.5,fftshift(abs(X2)),'k.');
>> xlabel('f_r '); ylabel(' IX_2(f_r)I');
>> axis([0.2 0.4 0 20));
For this DFT, only ten samples of x 2 [nJ are used but the sequence is zero-padded
to a length of 500. Although the DFT has 500 frequency bins uniformly spaced
over the frequency interval [-0.5, 0.5), there is insufficient information about x 2 {n]
(only 10 samples) to resolve the closely spaced exponentials at f,. = 0.30 and
f,. = 0.315. Using the picket fence analogy, zero-padding increases the number
of "pickets" in our DFT fence, but it does not change what lies behind the fence.
Still, Figure S8.M-3b does show a concentration of signal energy centered around
fr = 0.308, the average of the two exponential frequencies.
377
...0,4 -0.3 - <l..l ~. ' 0. 1 O.l 0.3 0.4 0.5
'·
Figure S8.M-3b: DFT of 10-point x2 [n) zero-padded to 500-points.
100
••
~;: 60
••
20
•• i......~~~~~·~~~~~~Willru'o1~1J~IWllll1.1·uw.·~
~o.s -o.• -o., ....0.2 -o , o. 1 0.2 o.3 o.• o.5
..
100
••
20
For this DFT, 100 samples of x 2 [n] are used. As a result, the DFT has 100 fre-
quency bins uniformly spaced over the frequency interval [-0.5, 0.5). As Figure
S8.M-3c shows, the two exponentials are not easily distinguished; even the num-
ber of dominant frequency components is difficult to identify. There difficulties
partially occur because the exponentials are closely spaced and the data window
is insufficiently large. Features are also obscured since the frequency fr = 0.315
does not lie directly on a DFT frequency bin; therefore the effects of frequency
leakage and smearing are pronounced.
(d) >> n = (0:99); x2 = exp(j*2*pi*n*30/100)+exp(j•2•pi•n•31.5/100);
378
>> x2 = [x2,zeros(1,400)];
>> X2 = fft(x2); f_r = (O:length(x2)-1)/length(x2);
>> subplot(211),stem(f_r-0.5,fftshift(abs(X2)),'k.');
>> xlabel('f_r' ); ylabel('IX_2(f_r)I');
>> axis([- 0.5 0.5 0 110));
>> subplot(212),stem(f_r-0.5,fftshift(abs(X2)),'k.');
>> xlabel('f_r'); ylabel(' IX_2(f_r)I');
>> axis((0.2 0.4 0 110));
..
100
..
100
For this DFT, 100 samples of x 2 [n] are used, and the sequence is zero-padded to a
length of 500. The DFT has 500 frequency bins uniformly spaced in the frequency
interval [-0.5, 0.5). As shown in Figure S8.M-3d, two dominant frequency compo-
nents can be separately identified, but there is some added "clutter" throughout
the frequency spectrum. This "clutter" is t he result of applying a finite-length
window to the signal x 2 [n) . Comparing Figures S8.M-3c and S8.M-3d, it is clear
that zero-padding assists in separating and locating the two dominant frequen-
cies. That is, Figure S8.M-3d displays two discernable modes at the correct
frequencies fr = 0.30 and fr = 0.315 while Figure S8.M-3c cannot distinguish
these two features.
8.M-4. Ideally, y 1 [n] = 1 +e1 2 rrn3 o/ioo +0.5*eJ2rrn 43 /ioo is characterized by three spikes located
at fr = 0, fr = 0.3, and fr = 0.43. The spikes at fr = 0 and fr = 0.3 should have
equal height and the spike at fr = 0.43 should have have a height that is half as high
as the other two. For most cases, two DFT magnitude plots are included: the first
covers the entire range of digital frequencies and the second details the range near the
true frequency content of x 2 [n].
(a)>> n = (0:19); yl = 1+exp(j*2*pi*n*30/100)+0.5*exp(j*2*pi*n*43/100);
>> Yl = fft(yl); f_r = (O:length(yl)-1)/length(yl); '
>> stem(f_r-0.5,fftshift(abs(Yl)),'k.');
>> xlabel('f_r'); ylabel(' IY_l(f_r)I');
>> axis([-0.5 0.5 0 25]);
For this DFT, only 20 samples of yi[n] are used. As a result, the DFT has only
20 frequency bins uniformly spaced over the frequency interval [-0.5, 0.5) . As
Figure S8..t-.il-4a shows, strong content is seen at f,. = 0 and f,. = 0.3, but there
379
,.
20
1S
~..
~
..
J
~ 10
>•.---·-~---~---~----,
380
for this distortion is frequency leakage that results from applying a rectangular
window to the signal yi[n].
(c) >> n = (0: 19); yl = l+exp(j *2•pi• n•30/ 100)+0 . 5•exp(j•2• pi•n•43/100);
>> yl = yl. • window(©hanning,length(yl))';
>> Yl = fft(yl) ; f _r = (O: l ength(yl)-1)/length(yl);
>> stem(f_r-0.5,fftshift(abs(Yl)),'k.');
>> xlabel('f_r') ; ylabel('IY_ l (f_r)I');
>> axis((-0.5 0.5 0 12));
••
381
1',-----.,..-----.--- - - . - - - - - - .
..
1•~~l~1111 1 1 1 1 1m 1 ~l l ~l
0.3 O.M
'•
0 .'4
computed in 8.M-4b; the reason for this improvement is lhat the Hanning window
has lower side lobes than the rectangular window, and is thus less susceptible to
leakage from distantly spaced components.
In this particular case, the Hanning window improves the analysis of the signal.
The lower side lobes of the Hanning window result in reduced leakage. Alth ou~h
the Hanning's broad main lobe results in increased smearing, signal components
are spaced sufficiently far apart that each component can still be distinguished.
8.M-5. Ideally, y 2 !nJ = 1 + e12'rn3 o/ rno +0.5*e12 ""38/ lOO is characterized by three spikes localed
at fr = 0, f r = 0.3, and fr = 0.38. T he spikes at fr = 0 and fr = 0.3 should have
equal height and the spike at fr = 0.38 should have have a height that is half as high
as t he other two. For most cases, two OFT magnitude plots are included: the first
covers the entire range of digital frequencies and the second details the range near the
true frequency content of y 2 [nJ.
(a)>> n = (0:19); y2 = l+exp(j*2•pi•n•30/ 100)+0 .5•exp(j •2*pi•n•38/100);
>> Y2 = fft(y2); f_r = (O:length(y2) - 1)/length(y2);
>> stem(f_r- 0 .5 ,fftshift(abs(Y2)),'k.');
>> xlabel('f _r'); ylabel('IY_2(f_r)I');
>> axis((-0.5 0.5 0 25));
For this DFT, only 20 samples of y 2 [nJ are used. As a result, the DFT has only
20 frequency bins uniformly spaced over t he frequency interval [-0.5, 0.5). As
Figure S8.M-5a shows, strong content is seen at f r = 0 and fr = 0.3, but there
is insufficient detail to identify the component at fr = 0.38. As a result, it is not
really possible to determine the relative strength of the two non-DC components.
(b) >> n = (0:19); y2 = 1+exp(j*2•pi•n•30/100)+0. 5•exp (j•2•pi•n•38/100);
>> y2 = [y2,zeros(1,480));
>> Y2 = fft(y2); f_r = (O:length(y2)-1)/length(y2);
>> subplot(211),stem(f_r-0.5,fftshift(abs(Y2)),'k.');
>> xlabel('f_r'); ylabel('IY_2(f_r)I');
>> axis((-0.5 0.5 0 25));
>> subplot(212) ,stem(f_r-0.5,ffts hift(abs(Y2)),'k.');
>> xlabel('f _r') ; ylabel('IY_2 (f_r)I');
>> axis([0.25 0.45 0 25));
382
5
0
-0.S
I I
-.0...
I I
-0.3
'
-0.2 -0.1
I '
0. 1
I
02
I I
0.3 0 .-4
I O.S
0 .1 0 .2 o.3 o .• o.s
"~---~---~--------~
383
2
•
•
•
2
0
0., 0.2 0 .3 0 .4
I 0.$
-0.5 -0.4 ..0 .3 -0.2 -0.,
Figure S8.M-5a; the DFT of the Hanning-windowed signal has broader features
than the DFT of the rectangular-windowed signal. As with Figure S8.M-5a, the
component at fr = 0.38 is not discern able in Figure S8.M-5c. Finally, notice that
the peak amplitudes in Figure S8.M-5c are lower than the peak amplitudes in
Figure S8.M-5a. This is primarily because the Hanning-window attenuates tr\e
edges of the original signal, resulting in a loss of signal energy (and thus smaller
DFT coefficients).
>> n = (0:19); y2 = 1+exp(j*2*pi*n*30/100)+0.5*exp(j*2*pi*n*38/100);
>> y2 = [y2.*window(©hanning,length(y2))' ,zeros(l,480)];
>> Y2 = fft(y2); f_r = (O:length(y2)-1)/length(y2);
>> subplot(211),stem(f_r-0.5,fftshift(abs(Y2)),'k.');
>> xlabel('f_r'); ylabel('IY_2(f_r)I');
>> axis((-0.5 0.5 0 12));
>> subplot(212),stem(f_r-0.5,fftshift(abs(Y2)),'k.');
>> xlabel('f_r'); ylabel('IY_2(f_r)I');
>> axis([0 . 25 0.45 0 12));
10
.,L-~-~-~-
-0.5 -0.4 -0,3 -0.2
....
-0.1 0 ,1 oz 0.3 0,4 0.5
384
As shown in Figure S8.M-5c, zero-padding does not improve the picture of the
Hanning-windowed signal y2[n] . While the components near fr= 0 and fr= 0.3
ca.n be identified, the component at fr = 0.38 appears mostly lost! As such,
the relative strengths of the components at fr = 0.3 and fr = 0.38 cannot be
determined.
In this particular case, the Hanning window does not improve the analysis of the
signal. While the lower side lobes of the Hanning window may reduce leakage,
the Hanning's broad main lobe smears the components fr = 0.3 and fr = 0.38
to the point that the component fr = 0.38 is completely obscured.
8.M-6. (a) MATLAB is used to plot the four samples corresponding to one period of the
periodic sign a.I x[n] = cos(mr /2).
>> N = 4; n = (O:N-1); x = cos(n*pi/2);
>> stem(n,x,'k.'); xlabel('n'); ylabel('x[n]');
>> axis([-.5 3.5 -1.1 1.1]);
0 .8
o.•
•.
..,
*•-0~
-OA
-OA
~·
-I
385
....
..
1.2
:;;- I
~
••
...
••
,
•
· <>.> • 0, 4 ·<>.3 -0.2 ..... 0. 1 0.2 0.3 O.• 05
386
Figure S8.M-6e: sjn] from zero-padded DFT.
387
Chapter 9 Solutions
9.1-1.
Therefore
x[n] = 2ej0.4rrn + 2ejl6"" + ei(l.2•rn-ir/2 ) + ei(0.8"n+1T/2)
We have first, second, third and fourth harmonics with coefficients
D1 = D2 = 2 D3 = - j D4 = j
ID1I = ID2I = 2 jD3j = ID41 = l
and
lT/ 2-
3
-TT/ :J
Figure S9.1 -1
9.1-2.
1
x[nJ = cos 2.271'n cos 3.3nn = 2[cos5.57rn +cos l.lnn]
1
2lcos 1.5rrn +cos l.lnnJ
388
= ![ejl.57rn + e - jl.5rrn + ejl.1'm + e-jl.l;ornJ
2
! [ejl.5orn + ej0.5orn + eJ1. l1rn + eJ0.9ornl
2
The fundamental frequency no = 0.1 and No = ~,,. = 20. There are only 5th, 9th,
11th and 15th harmonies with coefficients .o
Figure 89.1-2
9.1-3.
The hmdamental frequency no = 0.47r and N 0 = ~: = 5. Only 2nd, and 3rd harmonics
are present.
L lJ,,...
I tJ~ ~ 1• I I •
l 2. 3 }f 1{' ~
0. f.f /r
3 It _, V'~
-o•lflT
Figure S9.l-3
9.1-4. To compute coefficients Dn we use Eq. (9.13) where summation is performed over any
interval N 0 . We choose this interval to be -N0 /2, (N0 /2) - 1 (for even No). Therefore
(No/2)- 1
D,. = ~o L x[n]e - jrflon
n = -No/2
389
In the present case No = 6, no = ~: = ~, and
2
l "'""" ( J -
D r=-0xne jrZ!n
3
6 n=-3
6
1 ~ 2~
'6 [3 + 4cos( r) + 2cos( r)J
3 3
3 2 l 2
Do = Z D1 = 3 Dz = O D3 = 6 D4 = 0 Ds = 3
9.1-5. In this case No= 12 and no= i·
x[OJ = 0 x[l] = 1 x[-lJ = -1 x[2) = 2 x[-2] = -2
x [3] = 3 x[-3J = -3 x[±4] = x{±5] = x[±6] = 0
Therefore
9.1-6. Here, the period is N 0 and no= 2~/No . Using Eq. (9.9), we obtain
Therefore
390
Interchanging the order of summation yields
N 0 - 1 No-1 [N0 -1
Px = ~o L L D,.D;n L ej(r-m)Oon
]
From Eq. (5.43), in Appendix 5.1, the sum inside the parenthesis is N 0 when r = m,
and is zero otherwise. Hence
l No-1 No-1
Px = No L lx[nJl2 = L IDrl
2
n=O r=O
9.1-8. (a) Yes, the sum of aperiodic discrete-time sequences can be periodic. For example,
consider two signals xi(n] = sin(n)u[n] and x 2 [n] sin(n)u[-nJ. The sum of these
two aperiodic signals is the periodic function x 1 [n] + x2[n] = sin(n).
(b) No, it is not possible for a sum of periodic discrete-time sequences to be aperiodic.
Consider arbitrary periodic signals xi[n] and x 2[n) with periods N 1 and N2 ,
respectively. Let y[n] ::::: x 1 [n] + x2[nJ. Notice that y[n + N1N2] = x1 [n + N1N2] +
x2ln + N1N2J- By periodicity, xdn + kNi} = x1[n] and x2(n + kN2] = x2(n] for
any k. Thus, y[n+N1 N 2 ] = x 1 (n)+x2 [n) = y[n]. That is, the sum of two periodic
signals must also be periodic.
9.2-1.
Since IX(D)I is an even function and LX(n) is an odd function of n, the integrand
in the second integral is an odd function of S1, and therefore vanishes. Moreover the
integrand in the first integral is an even function of n, and therefore
x[n] =- 11
7r 0
00
9.2-2. (a) Because x[nJ = Xo[n] + Xe[nJ and e-jOn =cos nn - j sin nn
co
n=-oo
00 00
Because xe[n]cosDn and x 0 [nJsinf2n are even functions and x 0 [n]cosDn and
xe[n] sin Dn are odd functions of n, these sums reduce to
00 00
Also, using the parallel development, we obtain the results similar to those of
391
Prob. 7.1-1 ,_ by which we have
00 00
x n 1 1 - acosn . -a sinn
( )= 1 - ae-i!l = 1 + a2 - 2a cos n + J 1 + a2 -2a cos n
Hence
-n [ J 1 1
a u -n {::=} ( 1 ) "
- e-J
·n - 1 +1 = 1 - aeJ
·n
a
and
and
1 I )
DTFT{x 0 (n]} = 0.5 (
1 - ae-J
·n -
1 - ae
n = l +a- 2jasinn
- 2a cos
n = J.lm X (f2)
9.2-3. (a)
00
X(D) = L o[n]e-jOn = 1
n ---oo
(b)
00
(c)
X(f2)
392
IX(U)I = LX(w) =- tan- 1 ( sinn
n )
cos - -y
Observe that
X(n) = _'Y_ = __;_-y_e-_;_n~
eiO - 'Y 1 - -ye-JO
Comparison of this equation with Eq. (9.34a) shows that. X(n) in the present
case is ')'e-jO times the X(D) for -y"u[n]. Clearly, the amplitude spectrum in this
case is -y times that in Figure 9.4b. Moreover, the angle spectrum in the present
case is equal to -n plus that in Figure 9.4c. This is shown in Figure S9.2-3a.
(d)
sinn )
IX(D)I = LX(D) = 2!l - tan- 1 ( n
-y/l + 1'2 - 2-ycosn cos - 'Y
Observe that
e;2n e;nh
X(U) = 'Y(eif'I - 1·) = -l---l'-e--1=·0
Comparison of this equation with Eq. (9.34a) shows that X (U) in the present caie
is ~e-;n times the X(U) for ')'"u[nJ . Clearly, the amplitude spectrum in this case
is lh times that in Figure 9.4b. l'\forcover, the angle spectrum in the present
case is equal to n plus that in Figure 9.4c. This is shown in Figure S9.2-3b.
(e)
x[n] (--ytu[n]
00
x(n) .Lc-1rc-jo"
n=O
n=O
1
= 1 +1e-iO
1
(f)
DTFT of both those components are found in the text Eq. (9.34) and Eq. (9.36).
Hence
1 1 _ e-iO(l _ -y2)
X(f2) = ,. , + -i e - 10 - l - ----=,..,-~,;,...--
1 - -ye-1" .., (1 - ')'e-i")(e - if'I - -y)
393
.•I. 1 f\
hii
a. u [ A- t]
LL
~ 0
ft c- 0 1.. ii-- - l ~
i\.-.
:~JL)I
1)<.cvv)\
~~)I XC"'-)\
Figure 89.2-3
9.2-4. (a)
X(D) e1kfl
sincl(n + k)'JT)
= o[n + kl
This follows from the fact that both n and k are integers and sin[(n + k)7r) = 0
for all n /; -k. For n = - k, sinc[(n + k)7rJ = 1. Hence, the result.
(b)
l
x[nJ = 2(o[n + kJ + cln - kJ)
(c)
394
The DTFT if x 2 [nJ = ~ cosn is found in part (b). Hence
1 1
x[nJ = 2o[n] + 4 (o[n + l j + o[n - IJ)
(cl)
X(D)
x[n]
(e)
x[nJ = einon
(f)
x (O)
x[nJ =
7r
2
7r7r 1:
[o(n - n o)+ o(n + no)I
[o(D - Do)+ o(D:+ D0 )] eiOndn
9.2-5.
3rr/4 einn 13,,/4
x[nJ _2._
27r J-n/4
einn dD = -.-
27rJn _,, 14
= -._I_ (ei(31</4)n _ e-j{.,./4)n]
J21Tn
ei"n/ 4 . .
= -.- - [2jsin11n/2J
J 27rn
= 0.5sinc(mi/2)e' r>•n/ 4
No No No+l -j{No+I)n 1
X(f2) = ~ a"e-i11n = ~ (ae- iO)n = a e . -
£__; £__; ae- ,n - 1
n ""O n=O
The result applies Lo both parts (a) and (b). The only difference being in part (a),
395
a< 1 and in part (b) , a> 1.
9.2-7. (a)
6 12
X(fi) 2 L e-;nn + L e- jnn
n=O n=>7
e- j7n - 1 e- jl3n - e-j7n
2
e-1n.
- 1 + - -e-;n
-=----
- 1
e-;1n + e-;13n _ 2
e-jn -1
(b)
-(No-1) N0 - l
X(fi) " -=!!_e-;nn + " __n_e-;nn
~
n=:-1
No-1 ~ No-1
n=O
No-1 No-1
" _!!!:__e-jrnn +" __n_ e-jnn
~ No-1 ~ No-1
m=l n=O
No - 1 No- 1
" _!!!:___e-jmll
~ No-1
+ L __
n_e-jnn
No-1
m=O n=O .· . 'i)
9.2-8. (a)
(b) The derivation can be simplified by observing that X(D) can be expressed as a
sum of two gate functions X 1 (D) and X 2 (D) as shown in Figure S9.2-8. Therefore
x[n]
2~ 1: [X1(D) + X 2(11)jejnn dfl
2~ { 1:
sin 2n +sin n
dnn dfl + 1 1
1
e;nn dfl}
nn
396
-. .
~~ - 1
;l..
:l
Y:(fl)
\ .2.
- -2
l : 1.
-x,rD-J.
2..
I
( +
- I
.i l
Xz{Y2-)-
0
I
~
Figure S9.2-8b
9.2-9. (a)
xlnJ -
27r
1 1"/2
-Tr/2
COS S1 eiOn dSl
eiOn
27r(l _ n 2 ) {jncosn + sinn} :rr/
12
2
1 (7rn)
7r(l - n2) cos 2
(b)
l
x[nJ=-
27r
j" X(Sl)e1°ndn=
-'If
. 1 [111' X(S1)cosSlndn+j 1" X(S1)sinS1nd~ ]
27r ->r _,,
Because X(Sl) is even function, X(S1) sin On is an odd function of n. Hence, the
second integral on the right-hand side vanishes. Also the integrand of the first
term is an even function. Letting i10 = 7r/4, we obtain
9.2-JO. (a)
3
X(Sl) = L xlnJe-iOn = 3 + 2(e-;n + ei0 ) + (e-i2!'l + ei 211 )
n=-3
3 + 4 cos n + 2 cos 2n
(b)
X(Sl) =
6
L x [nJ e-jnO
n=O
= e-iO + 2e-;2n + 3e-i3 f! + 2e-i4 f! +e-jr;n
= e-i311 [(ei 211 +e- j 211 )+2(ei11 + e- ;n) + 3j
e- 1311 [3 + 4 cos n + 2 cos 20]
397
(c)
3
X(f2) = L x[n]e- fnfl = 3e-i 0 - 3ei 0 + 6e-12n - 6e-i20 + 9e-i30 - 9ei 3 l1
n=-3
6j(sin n + 2 sin 2n + 3 sin 3f2)
(d)
2
X(n) = L x(nje-in!l = 2e-i11 + 2ei 0 + 4e-i211 + 4e-i2 fl
n=-2
= 4 cos n + B cos 2n
9.2-11. (a)
(b)
x[nj
= { x ~];]
00
n=0,±1,±2,···
L x[k]o[n - Lkj
otherwise
k=-oo
x,,[n]
(b)
398
Interchanging the order of the summation
X e(D) = ,,f;, 00
x[k] c~oo o[n - Lk]e-Jnn)
00
{c) z[n] is the signal x!nJ = 1 expanded by factor L = 3. Hence, from the above
result and pair 11, we obtain
2k
L L
00 00
9.2-13. (a) We shall consider spectra within the band ID! :S 7r only.
Pair 8: ~sine (Den) *=* rect C~J
This is identical to pair 18 in Table 7.1 with We replaced by De and t replaced by
n.
Pair 9: In the same way, we see that pair 9 is ident ical to pair 20 in Table 7.1.
Pair 11: is identical to pair 7 in Table 7 .1.
Pair 12: is identical to pair 8 in Table 7.1.
Pair 13: is identical to pair 9 in Table 7.1.
Pair 14: is identical to pair 10 in Table 7.1.
(b) This method cam1ot be used for pairs 2, 3, 4, 5, 6, 7, 10, 15 and 16 because in
all these cases X(D) is not bandlimited.
9.2-14. (a) Not valid because not 27r-periodic.
(b ) Valid because it is a constant.
(c) Valid because it is i~-periodic, hence a lso 27r-periodic.
( d) Not valid because it is 207r-periodic and not 27r-periodic.
(e) Not valid because it is not 2ri-periodic.
9.3-1. (a)
x[n] u[n] - u [n - 9]
0
X(D) [
-.-
,. . -
ei + 7ro(D)J - [-.-ein
0- + 1fo(D)J e-;9n
e1" - 1 e' - 1
399
because
eiO
a"u[n] <==> - .n- -
e' -1
we obtain
n- m eiHe-jmO ei(l - m)O
a u[n - m] {:=} . = -.0- -
e' 0 - / e1 -1
(c)
Hence
eiO eiO .
X(D) = a-3 - ._ _ _ -e- 1 1on
.-
a1-
ei 0 - a e'n - a
eifl (a - 3 _ a1e- 1on)
eifl - a
(d)
Hence
..
(e)
Hence
(f)
x[n] = (n - m)an-mu[n - m]
(g)
Hence
1eifl meifl
X(st) = ( eifl - i)2 (eifl - i)2
ein(1- m eifl + m1)
= (eifl - i)2
400
(h)
Hence
X(n)
9.3-2.
Note that at n = 0, both x[n - 4J and x [-n - 4] have value 4. This duplication is
corrected by the term -4o[nJ. Thus
sin(5S1/2) ·n
y(n] = z-l sin(Sl/2) e-J2 = o[n] + o[n - lJ + o[n - 2] + o[n - 3J + o[n - 4J
Both x[nJ and y[nJ are nonzero over 0 S n S 4 and are zero outside this range clearly.
401
9.3-4. (a) Apply modulation property [Eq. {9.55)] to pair 2 to obtain
1 [ ei(0-!10) ei(0+!1o) ]
ancosDonu[n] <==} -
2 ei(!1-0o) - a
+ -,.,,,,,..--=-:--
ei(n+no) - a
(b)
x[n] = n 2anu[n)
Apply 'multiplication by n' property [Eq. (9.50)] to pair 2 to obtain
d [ eifl ] aeiO
na"u[n] <=> j dr. - .,...-- = ( ' n )2
H eJ" - a eJ - a
d ( aeJn ] aeJn(eJn + a)
n2anu[nJ <=> j d!l (eJn - a)2 = (eJn - a)3
(c)
X(D) =
9.3-5. We shall consider spectra within the band IDI ::; 1T only
Pair#ll
1 = u[nJ + ti[- (n + l)J
But
Hence
and
e-in
u[-(n + l)] = u[-n] - o[nl <=> e-jn - 1 + m5(S1) - 1
and
ejn e-Jn
1 <=> -.-n-
eJ - 1
+ r.o(D) + e _1.0 - l
+ 7r<5(D) - l = 2m5(D)
402
Pair#l2
Apply frequency-shift property [Eq. (9.54)] to t he result for pair 11 to obtain
Pair#l3
Apply modulation property [Eq. (9.55)] to the result for pair 11 to obtain
Pair#l4
Apply modulation property [Eq. (9.57)] with 8 = -7r/2, to the result for pair 11 to
obtain
sin Don ~ j7r[o(f2 +Do) - o(st - Do)]
Pair#15
Apply modulation property [Eq. (9.55)] to the result for pair 10 to obtain
1 [ ei(O-Oo) eJCO+Oo) ] 7r
cosnonu[n] {=}
2 eJ<n-no) _ 1 + eJ<n+no) _ 1 + 2 [o(n - Do) + o(D + !10 )J
ei20 - eill cos no 7f
eJ
·2n ·n n
- 2eJ cos 0 + 1
+ - [o(D - Do) + o(n + Do)]
2
,.
'
Pair# 16
Apply modulation property [Eq. (9.57)] with 8 = -1r/2 to pair 10 to obtain
j [ ei (!Hllo) ei(ll-flo) l j7r
sinf2on·u[nJ <==> 2 eiCn+rio) - 1 - eiCO-no) - l j + 2!0(!1 +Do)+ o(D - f2o) ]
ei 211 sin Do r.
·2n
e'
n D
- 2eJ cos o + 1
+ -:{o(D
2;
- Do) - o(D + D 0 )]
9.3-6.
Hence
x[n + k] + x[n - k] <==> X(!t) [eikn + e-Jkn] = 2X(D) cos kD
403
(b) Figure S9.3-6b shows the signal ll(i)· Hence,
x[n] = 6 (-n-8)
8
- + 6 (n+8)
- -
8
The DTFT of 6 ( ~) can be found in several way. Here we shall use the method
of convolution. The reader can verify that the convolution w[n) shown in Figure
S9.3-6c with wl-n) yields 6 (i).
From pair 7, we obtain
sin 2n -;1.sn
q [n I ¢==> . 0 5r.. e
Sin . H
Therefore
sin(- 2n) _;isn sin2n jisn
[ I~
q-n
sin( -0.5S1)
e- · = sin o.sn e ·
Because
6 (i) = q[n] * q[-n]
6 (n) sin (2n)
2
- :i. i ,_ n~
lo..) I
~
· ~W[n]
·-··· ~
. -...2. .
r - -·-
' Y\ ::31'
l(c )
Figure S9.3-6
x [n ± k] ¢==> X (n)e±;kn
Therefore
404
From pair 7 (Table 9.1)
sin 2.5n 1 20
w[nJ = u[n] - u(n - 5] <==> .
sm 0.511
e-
and
x[n] =- {w[n - 21- w[n + 21} <=> - sin 2.5D .
sm .5
.
. 0 n sm2ne-J 20
9.3-8. If
ein
w (n) = n ---
-ei-=- -r
then
X(D) = W(n)W(O)
and
n
2: -rm-rn-m n~ o
m=O
'Yn t 1 ={ nr n~O
n<O
= n-ynu[nJ
9.3-9. Pair 2:
')'nu[n] = b[nJ + -y§[n - 1] + -y2b[n - 2] + .. . + ...
Therefore
00
m=O
1 &0
1_ e-iO
1
= eiO _ 'Y
Pair 3:
Use time inversion property to obtain
p.1> 1
and
Hence
!>.I > I
Letting >. = 1/-y, we obtain
n -lh ei!l
--y u [- (n + l )] <==> ·n
eJ - l 7
/ = eJ·n - -y "' < 1
405
Pair 4:
llnl = ')'"u[n ] - ( ~) n u!-(n + l)J
Hence
l -12
Pair 5:
Apply 'multiplication by n' property [Eq. (9.50)J to pair 2 to obtain
Pair 6:
Hence
Pair 7:
9.3-10. We shall consider the spectrum only within the band lfll ~ 1r.
x[n] = e3·nn -~
= eJ l x e3-~
2
2~ (27ro ( n - ~ ) * 2m5 ( n - ~ )]
0 0
X(n) = 1n1 ~ 7f
27ro (n- ~0 ) *O (n- ~0 ) 1n1~7f
27f 1: o( x - ~0 ) o(n - n + ~0 ) dx 1n1 ~ 7f
406
From the sampling property [Eq. (1.24)] of the impulse, we obtain
X(11) = L x [nje-ifln
n=-oo
Hence
00 00
L x[n]e-j"n ej!ln
n=-oo
00
n=-oo
= X(D - 1f)
Hence, we have
Moreover
rect (
n -c
211
n) = 0 at if
Hence
00
X(O) = L (-1rsinc(ncn) =o
n=-oo
(c)
407
Using the argument in part (a), we obtain the desired result.
(d) Using the argument in part (b) and (c), we obtain the desired result.
(e)
Hence
x[OJ = 2_ (" X (0)dn
211' } _.,,.
(1)
L 2
!sinc (0cn)j
2
n=-oo
= -
1T 12nc (1 -2n- + -n2 ) dD
n~ 0 De st~
9.3-12.
408
9.4- L
1 ei 0
X(H)
} + 0.5e-jO eifl + 0.5
ei11 (ei 0 + 0.32)
Y(ft) X(O)H(n) = (eJn + 0.5) (ei 11 + 0.8) (ein + 0.2)
Y(ft) ei11
+0.32
eiO (eifl + 0.5) (ein + 0.8) (eiO + 0.2)
2 8/3 2/3
eiO + 0 .5 - eiO + 0.8 + eifl + 0.2
eifl 8 eifl 2 elf!
Y(n) 2
ein + 0.5
-- +--~-
3 eiO + 0.8 3 ei 0 + 0.2
y[n] [ 2 (-0.5t - ~ (-0.8)" + ~ (-0.2)"] u[nJ
9.4-2.
I.32r. ei 11 (ei 11
+ 0.32)
Y(Sl) = 2.16 o(D) + (ein - l){ei 0 + 0.2)(ei 11 + 0.8)
Using partial fraction expansion , we obtain
·n ·n ·n
Y(D) = l.32r. o(n) + 1.32 _ eJ_ _ ! e' 4 e'
· 2.16 2.16 ei11 - 1 6 ein + 0.2 - 9eitl + 0.8
and
yin]= 0.6llu[nJ - [~(-o.2r + ~(-0.8)n] u[n]
9.4-3.
ejn zeifl
X(D) ---
eifl-0.8 eifl-2
e;2n 2ei2fl
Y(D)
X(D)H(O) = (ein - 0.5) (eiO - 0.8) (eiO - 0.5) (ei 11 - 2)
-5/3 8/3 2/3 8/3
ei0 - 0.5 + eifl - 0.8 + ei 11 - 0 .5 - eifl - 2
-1 8/3 8/3
eifl - 0.5 + eifl - 0.8 - ei 0 - 2
eifl 8 eifl 8 eifl
Y(D)
= - ein - 0.5 + 3ein - 0.8 - 3eiil - 2
409
9.4-4. (a) When x[n] = o[n], the output is h[n], given by (assuming causal accumulator)
n
h[nJ = I:: o!kl = u!nJ
k=O
and
e jfl
H(D) = ejn _
1
+ ?To(il)
(b) If this accumulator is used as a digital processor for the digital integrator, dis-
cussed in example 3.7, and the input is x(t) = u(t), then the sampled x(t) yields
the digital input x[n] = x(nT) = u[n]
9.4-5. (a)
Therefore
y[n] =sine [ 7r(n
2- Z)]
(b)
. n= O
smc(7rn) ={ 1
0
h .
ot erw1se
Hence
sinc(7rn) = o[n] <==> l
The output corresponding to this input is
and
y[n] = smc1. [7r(n - 2)]
2 2
(c)
410
and
(b) Figure S9.4-6a. shows 7"u[n} and (-l)Jnu[n}. The spectra. for (-1)n7nu[nJ a.re
the same as those for 7"u[n} (Figure S9.4-6 band c) but shifted by rr, as shown
in Figure 89.4-6 b over the fundamental band lfil ::; rr.
(c)
/
{
¥
\f.l.J2.J~; [_'fl.SL)
I
/~
·t t
:".
l ',,.
.·\
. .:..:-" l .· \ ~- -
I
-iC .·-=-- o ·---=t[,
- - -
SL-
---'<--
-IL .
(..b)
Figure 89.4-6
9.4-7.
and because
411
Y(D.) = Q(D. - 7r) X(D. - 27r)H(D. - 7r)
X(D.)H(D. - rr)
X(D.)H1(D.)
Therefore
Figure S9.4-7
Now consider the system S 2 , whose ai and b; coefficients are ( -1 )k times the
corresponding coefficients of S 1 . Hence
'°'N b e-jk(fl-")
L-k=O k = H1(D--rr)
"N
1 + Lk=l ake-ik(!t- ")
(b) As shown in Prob. 9.4-6, if H 1 (D.) is a lowpass filter, then H 2 (0.) = H 1(0. --rr) is
a highpass fi lter.
(c) The DTFT of the system equation yin] - 0.8y[n - l] = x[n] is
H2
(n) -_ Y2(if2) _ i
X(j!1) - 1 + 0.8e-;n
= 1
1 - 0.8e-iCO-w) = Hi
(n
- 1T
)
Hence the frequency response spectra for H 2 (f2) are same as those for H 1 (f2)
frequency-shifted by 7r. Hence, H2 (D.) is a highpass system.
9.4-9. (a) Let us compute the response h[nJ to the unit. impulse input o[nJ . Because the
system contains time varying multipliers, however, we must also test whether it is
a time varying or a time-invariant system. It is therefore appropriate to consider
the system response to an input o[n - kJ. This is an impulse at n = k. Using the
fact that x[n) o[n - kJ = x[njo[n - kJ, we can express the signals at various points
as follows:
at a1 2 cos (f2ck) o[n - kl
a2 2sin (f2ck) o[n - kl
bi 2cos (D.ck)h0 [n - k]
~ 2sin (f2ck)h0 [n - kl
C1 2 cos (D.ck) cos (f2cn)h0 [n - kJ
c2 2sin (f2ck) sin (f2cn)h 0 [n - kJ
d 2h0 [n - k] [cos (.Qc k) cos (f2cn) + sin(f2ck) sin (f2cn)J
= 2h0 [n - kl cos (f2c[n - kl)
Thus, the system response to the input o[n - kl is 2h0 [n - kJ cos[f2cln-k]J. Clearly,
the system is linear time-invariant, with impulse response
H(w ) n -
= rect ( 2W nc)
+ rect 2W (
n + n c)
The transfer function H(w) [Figure S9.4-9bj represents an ideal bandpass fi lter
when nc + w $ 71.
9.M-1. (a) The inverse DTFS is given by x[nJ = 2:~~; 1 Vre;r!lon. Like the DTFS, the
IDTFS can be computed using a matrix based approach. First, define W No =
e300 , which is a constant for a given No. Substituting W No into the IDTFS
equatio11 yields x[n] = 2:~:':~ VrW,V~. An inner product of two vectors computes
1
x[nj.
Vo
Vi
x[n) = [ 1 W" w2n ... w<No
' JVo> No> ' No
t)r]
V2
413
H(w)
0
(c)
Figure S9.4-9b
x!O] 1, I, 1, 1 Do
W(No-1)
x!l] 1, w~o' w~o' D1
wft'J...ro - 1)
xl2] 1, w~o' w~o' No
V2
2
x!No - l] I, W(No-1 ) W2(No - l) W(No-1 ) VNo-1
No ' No ' No
(b) MATLAI3 code, similar to that presented in MATLAB session 9, is used to test
the execution speed of the matrix IDTFS approach to the inverse FFT approach.
First, test vectors and IDTFS matrices are created.
>> XlO = fft(randn(10, 1));
>> X100 = fft(randn(l00,1));
>> X1000 = fft(randn( 1000,1));
>> W10 = (exp(j•2•pi/ 10)) . -((0:10-1)' • (0: 10-1));
>> W100 = (exp(j • 2•pi/100)).-((0:100-1)'•(0:100-1));
>> W1000 = (exp(j •2•pi/ 1000)).-((0:1000-1)'•(0 :1000-1));
\end{ verbatim)
Next, execution speeds are measured by repeating calculations within a loop .
Notice , output from MATLAB's {\tt ifft} command must be scaled by
$1/N_O$ to compute the IDTFS .
\begin{ verbatim}
>> tic; for t=1:50000, ifft(Xl0)/10; end; T10ifft =toe;
>> tic; for t=1:50000 , W10•X10; end; T10mat =toe;
>> tic; for t=1:5000, ifft(Xl00)/100; end; T100ifft =toe;
414
>>tic; for t=1:5000, W100•X100; end; T100mat = toe;
>>tic; for t=1:500, ifft(X1000)/1000; end; T1000ifft =toe;
>> tic; for t=1:500, W1000•X1000; end; T1000mat =toe;
>> [T10mat/T10ifft, T100mat/T100ifft, T1000mat/T1000ifft)
ans= 1.0323 3.5000 101.4754
For these trials, these results indicate that the IFFT approach is about as fast as
the matrix approach for N 0 = 10, about an order of magnitude faster than the
matrix approach for No = 100, and about two orders of magnitude faster than
the matrix approach for N 0 = 1000. While actual times will vary considerably
from computer to computer and from trial to t rial, the general trend is clear: the
matrix based approach is less efficient than the inverse FFT approach, and this
difference grows rapidly as N 0 increases.
(c) Substituting 1) = ~ow NoX into x = WNo1) yields x = w;.,o ~ow NoX =
~o WNo W N 0 X . For equality, ~o WNo W No = IN0 , where !No is the No-by-No
identity matrix. Thus, w;:.,,o w No = NolNo = w No WNo.
Thus, multiplying the DFT matrix W No by the inverse DTFS matrix W;:.,,0 , or
vice versa, yields the scaled identity matrix Nol No:
1 - sin(stc)
Q- - ---
- cos(Dc) ·
415
Thus, the difference equation is
This first order system has one pole at z =a= 0.1584. Since this pole is inside
the unit circle, the system is stable. Frequency response is computed using the
signal processing toolbox function freqz:
>>Omega= linspace(O,pi,1001); H = freqz(B,A,Omega);
>> H3dB = freqz(B,A,(-Omega_c,Omega_c]);
>> plot(Omega,(abs(H)),'k', .. .
(O,pi],[1/sqrt(2),1/sqrt(2)],'k:', ...
[Omega_c,Omega_c),(0,1),'k:');
>> axis((O,pi,0,1));
>> xlabel('\Omega'); ylabel('IH(e-{j\Omega})I');
>> set(gca,'xtick' ,(O:pi/5:pi],'xticklabel',(' 0 '·
, p/5';'2p/5';'3p/5';'4p/5';' p '], ...
'fontname','symbol');
..•
...
...
••
0,)
••
0
As seen if Figure S9.M-2b, the filter is high pass with a cutoff frequency De = 27r/5,
as desired.
(c) Since a, and therefore H(z), is held constant, the cutoff frequency nc remains
constant as well. That is, changing the sampling frequency does not affect the
digital cutoff frequency De of the filter. However, the cut-off frequency expressed
in hertz scales directly with the sampling frequency. That is, as Fs is increased
to Fs = 50kHz, Jc = lkHz is increased to Jc = lOkHz.
-2- ) (~).
1
(d) The inverse to H(z) = (1.±2)
2 ( l-az-
1
-z - 1 ) is H- 1 (z) = ( l+a 1-z - , Since the
inverse has a root on the unit circle, it is not BIBO stable and therefore not well
behaved.
(e) For nc = 7r/2, a = ~!~~~c = 1 01 , which is indeterminant. Using L'Hospital's
1
416
filter! Notice that t.he impulse response is h[nJ = 0.50(n] - 0.50[n - lJ. which is a
finite duration signal.
9.M-3. (a) It would be unlikely if not impossible to achieve this exact magnit.ude response
with a practical FIR filter. Since the magnitude response has points of derivative
discontinuities, an infinite length filter would be required, which is not practical.
(b) MATLAB is used to design a length-31 FIR fill.er that reasonably approximates
the desired response. Different approximations are easily accomplished by chang-
ing N. To perform computations, program MS9P3 and MS5P1 are utilized. Notice,
it is important to define n over (0, 2?r) not (-rr, 11-).
>> H_d = inline(('((mod(Omega,2•pi)>s O)k(mod(Omega,2•pi)<pi/4))' , ...
'.•(4•mod(Omega,2•pi)/pi)+', ...
'((mod(Omega,2•pi)>=pi/4)&(mod(Omega,2•pi)<pi/2))', . . .
'.•(2-4•mod(Omega,2• pi)/pi)+', ...
'((mod(Omega ,2•pi)>7•pi/4)&(mod(Omega,2•pi)<=2•pi))', ...
'.•(-4•(mod(Omega,2•pi) -2•pi)/pi)+', ...
'((mod(Omega,2•pi)>3• pi/2)&(mod(Omega,2•pi)<=7•pi/4))', ...
'.•(2+4•(mod(Omega,2•pi)-2•pi)/pi)']);
>> N = 31; h = MS9P3(N,H_d);
>>Omega= linspace(0,2•pi,1001);
>> Omega_samples = linspace(0,2•pi•(1- 1/N),N)';
» H = MS5P1(h,1,0mega);
>> subplot(2,1,1); stem([O:N-1),h,'k');
>> xlabel('n'); ylabel('h[n] ');
» subplot(2,1,2); plot(Omega_samples,H_d (Omega_saroples) ,'ko', ...
Omega,H_d(Omega),'k:',Omega,abs(H) ,'k');
>> axis([O 2•pi -0.1 1.3)); xlabel('\Omega'); ylabel(' IH(\Omega) I');
>> legend('Samples','Desired','Actual',0);
>> set(gca,'xtick',[O:pi/2:2• pi),'xticklabel',[' 0
I p/2 I p; 3p/2) ; 2p ) )
) ) ; I I ; ) I I • • •
'fontname','symbol');
••
02
•.
f
~
-0 '
....~. ,. ••. ,. 25 JO
~
l.l
......
••
i·'
.
n
Repeating the above code using N = 101 yields a much closer approximation.
9.M-4. A simple first-order highpass filter is given by HHp(z) = k(l - z- 1 ). To achieve a gain
of 3, solve 3 = klJ - e- 1"1=2k . Thus, k = 3/2. To realize the desired comb filler, the
417
...
0.2
•..
z
....
....
,,
••
418
9.M-5. {a) Reversing the order of the elements of column vector x can be accomplished using
a N 0-by-N0 permutation matrix RNo that is simply a 90-degree rotated N 0-by-N0
identity matrix. For example, R 5 is
(b) Let integer i E {O, 1, ... , No - l} be used to designate row or column of W No·
Row i of W No is represented as r ; = e' 2"IO,l,. .. ,No-1Ji/No. Column i of W No is
represented as c; = e' 2" [o,i , ... ,No-l)Ti/No. For i ~ 1, notice that column No - i
is CNo-i = eJ2TT[0,l, ... ,No-l)T{No-i)/No = eJ21f(0, l, ... ,No -l lT e - J2"(0, l ,. .. ,No-ll°ri/No =:
e-12,,(o, 1,. .. ,No - 11T i/No = r f 1 . That is, for i ~ 1, column (No - i) is the complex-
conjugate transpose of row i. Also notice that W No is composed of orthogonal
rows,
0 i :f; k
r; r f: = {
No i = k
Combining these facts yields
~No
...
l
1 0 0
0 "
wi, [ RNo-1
0
1 0 0 0 0
0 0 0 0 1
W~ = 5 0 0 0 1 0
0 0 1 0 0
0 1 0 0 0
By inspect.ion, it is clear that W 'i.0 is a scaled permutation matrix. The operation
W 'J.,0 x scales and reorders the vector x: t he first element of x is not moved, but
the order of the remaining No - 1 elements are reversed.
MATLAB is used to confirm these conclusions.
>> x = (1 2 3 4 5]';
>> W_5 • dftmtx(S);
>> real(W_5•W_5•x)'
ans = 5.0000 25.0000 20.0000 15.0000 10.0000
The last. line includes the real command to remove minute imaginary·components
that result due to computer round-off. As expected, vector x is scaled by No = 5
and t he order of the last four elements is reversed.
(c) Using the previous result, W 'Jv0 x = (W 'J.,0 )(wi0 )x = N6 x. The first multiplica-
tion by (wi 0 ) scales x by N 0 and reverses t.he order of the last No - 1 elements.
The second multiplication again scales x by N 0 for a total of NJ and reverses
the previously reversed last N 0 - 1 elements, effectively leaving the order of x
unchanged. MATLAB is used to confirm these conclusions.
419
>> x = [1 2 3 4 5]';
>> W_5: dftmtx(5);
>> real(W_5*W_5*W_5*W_5*X)'
ans = 25.0000 50.0000 75 . 0000 100.0000 125 . 0000
The result is just x scaled by NJ = 25.
420
Chapter 10 Solutions
10.1-1. (a)
ii+ lOy + 2y = x
Choose: Qi =y and Q2 = ii = cii ==:} ci2 = Y
<ii= Q2
hence:
1fa = -2q1 - lOQ2 +x
In matrix form we get:
[ qi ] [ 0 1 ] [ QI ] +[ 0 ]
ci2 ~ -2 - 10 Q2 1 x
(b)
jj + 2eYiJ +logy= x
Choose Qi =y and Q2 = iJ = ci1
Qi = Q2
hence:
</2 = - 2e"'Q2 - JogQ1 +x
It is easy to see that this set is nonlinear.
(c)
where
and
. x - Q1 - ch
t= 2
1. x - Qi - <12
Also we have: 2 Q1 = 2 - Q1
421
We can also write:
. .
q2 = 3i2 = 3 [x-qi-<i2
2
- q2
]
= 32x - 3 3.
2q1 - 2q2 - 3 q2
5. 3 3 3
Hence 2q2 = -2q1 - q2 + 2x
3 6 3
or <i2 = -5qi - 5q2 + 53; (2)
3 6 3 ] 12 6 2
Qi= -3q1 +x - [ --qi - -q2 + -x = - -qi+ -Q2 + - x
5 5 5 5 5 5
Hence the state equations are:
[ ~·
q2
]=[ -~
-~
_; ] [ : ] + [ : ] x(t)
Figure Sl0.2-1
~Q1
2
= - 2q1 - q2 - 3q1 + 3x = - 5qi - Q2 + 3x
(2)
Hence from (1) and (2) the state equations are found as:
-10
1 =~ ][:~ ] +[ ~ ]x
422
The output equation is: y = <h = Qi - Q2
or y = [ 1 -1 ) [ ~~ ]
Figure SI0.2-2
10.2-3. Let's choose the voltage across the capacitor and the current through the inductor as
state variables q 1 and Q2 , respectively. Writing the loop equations we get:
And thus:
Qi = -5q1 + Q2 + 5xi
ch = - 2q1 -2q2 + 2x1 - 2x2
Hence the state equations are
·!II
Figure SI0.2-3
with and
X = 2i1 +Qi +Qi = 2Q1 + 2Q2 +Qi +QI = 3Q1 +Qi + 2<i2 (1)
423
Substituting <h in the equation (1) we get:
. 5 2 5
q1 = - -qi +-q2+-x (4)
3 3 3
From (3) and (4) the state equations are obtained as:
_o! ] [ ~~ ] + [ t]x(t)
-
2. L:,L
i, -='
~ "-•
vek
f'. L. - )(
:t I
Figure Sl0.2-4
Qi+ ci1
<i1
= - 2- +:i:--2 -
x - Qi
424
Figure Sl0.2-5
I s Yes)
!--+..-- . - -- -
s +z
S +-1
Figure Sl0.2-6
From (l), (2) and (3) the state equations can be written as:
[
~I
Q2
<h
l [-}Q ~2 ~1 l[:~ l ~ l
= 0
1 0 - 1 q3
+ [
0
x
Q1 )qq1
Q2 = A2Q2 + X1
q3 A3q3 + X2
Q4 A4q4 + X2
][ l pq[: l
or
[~]=[1 ~ ~
0 0
.>..2 0
0 .>..3 +
0 0 .>..4 q4 0 1
425
The output equation is:
Yi == Q1 + qz 1 0
Y2 == Q2 + q3 1 1
10.2-8.
H s _ 3s + 10
( ) - s 2 +7s+12
[
<ii ]
42
=[ 0
- 12
l ] [
-7
Qi ]
Q2
+[ 0 ]x
1
y == [ 10 3 ] [ :: ]
Transposed direct form II: In this case the block diagram can be drawn as shown in
Figure Sl0.2-8a.
(5)
;t_,
1__..~--- Yes)
hence: 4i = -7qi + Q2 + 3x
<i2 - 12q1 + lOx
or
[ ci1 ] [ -7 1 ] [
Q2 = - 12 0
Qi ]
Q2 +[ 3 ]
10 x
The output equation is:
426
The cascade form:
and
Parallel form:
2 1
H(s ) = - + -
s+4 s+ 3
QI = -4q1 -!-X
Q2 = - 3q1 + x
And the output equation is:
10.2-9. (a)
H( s ) _ 4s = 4s
- (s+l)(s +2)2 s3+5s2 + 8s+4
Direct form II:
And
427
;:t,
In this case:
<i1 = -5qi + Q2
<i2 = -8qi + Q3 + 4x
q3 =-qi
And:
[~~ l
Q3
= [ =! ~ ~ ~~
-4 0 0
l[ l ~ l
Q3
+[
0
x
Cascade form:
H(s)= C~1)C~2)C~2)
From the block diagram we have:
41 = -2ql + q2
+ 2q2 = 41fa ==>
Q2
Qi = - 2qi
+ Q2
~2 = - 4q3 - 2q2 + 4x
<h = - q3 + x {
q3 = - q3 +x
~ ~ ~ ~2 ;~
2
[ ] [ ] [ :: ] +[ : ] x
Aud the output:
P arallel fonn:
H(s) = _-_4_ + _4_ + _ _8_
s +I s+2 (s + 2)2
We have:
428
Figure Sl0.~9a: parallel
ci1 = - Q1 +x
<fa - 2q2 + q3
<.b -2q3 +x
[: l~ [-~I ~2 12 ][:: l ~ l
And the output is:
+ [ x
[ ::. ] = [~
1 0
0 1
0 0
q4 -2 -7 -9
l!~ l
And the output is:
y ~ ( 0 12 7 I j
429
And
Cascade form:
H s _ s(s + 3)(s + 4) _
( ) - (s + 2)(s + 1)3 -
(-1
+s
)(_!!_)
+ 2
(~) (s + 4)
+ s + 1 s l s l
And
y = Q1 = [ l
6 11 7 6
H(s)=-- + - + - -
s+2 s+l (s+1)2 (s + 1) 3
<ii = -2q1 + x
q2 = -qz + q3
<h = -q3 + q4
<i1 = -q4 + x
From the block diagram, we have
430
Yrs)
Xt s)
or
10.3-1.
q=Aq + B x
Q(s)
= <I>(s)q(O)
=
[
2(s+3) +2
(s+l) (s+2)
-2+s
(s+l}(s+2)
l[=
2s+8
(s +l) (s+2)
s- 2
(s+l )(s+2}
]=[ ·!31- •:2]
s+ I + .t+2
And finally:
2
(t ) = [ Q1 (t) ] = c-t [ Q (5 ) ] = [ (6e-t - 4e- t)u(t) ]
q Q2(t) (- 3e-t + 4e- 21 )u(t)
10.3-2.
431
-6
(s+3)(s+2) ]
<I>(s) =[ .,...(s_+.,,,;3):o,....(s-+2=)
s+5
(s+3)(s+2) (s+3)(s+2)
And hence:
(s+3}(s+2}
~(s_+"'" :~'"'-"~-+2~}
(s+3}(s+2}
] [ 5+ •;~~o• ]
-34.02
s+2
+ 39.03
s+3
- 10- s ]
2
s 2 +l0 4
[ 17.01 10.01 0
s+2 - s+3 - s2+10•
10.3-3.
s ~ ] and <I>(s)
2 1 [ s+1
(sI - A)= [ s_:- = (sl - A)- 1 = tt 0 ]
1 1 (s + l)(s + 2) 1 s+2
..!2 0 ]
[ (s+ 1)1(s+2} .!1
Also x(t) = u(t) ==> X(s) = ~
And thus:
l
s+2
Q(s) = [
t
(s + l)(s+2)
Hence:
10.3-4.
(S I - A) = [ s+o 1 s -1 ] d'.... ()
+ 2 an '*' s = ( sI - A) - 1 1
= {s + l)(s + 2)
[ s+2
0 s+1
1 ]
432
and x(t) = [ u(t)
c5(t)
] ==> X(s) = [ ~1 ]
BX(s) = 1
[ 0
1 ] [
1 i1 ]
= [
T
s+ 1 ]
1 2
(2s+l}(s+2}+3s
s(s+l;(•+2) = sl + s+~
4
-
3
s+2
]
[
s+2 s+2
And hence:
2
(t) = .c-1(Q( )) = [ q1(t) ] = [ (1+4e-t - 3e- t)u{t) ]
q s q2 (t) 3e- 2 tu(t)
10.3-5.
so q(o)+BX(s) = [ 2 ~~] = [
2
·r]
and <P(s)[ (O)+BX (s)J
q
=[ (s+lJ(s+2)
-2
(s+ I )(•+2)
(s+l)l(s+2) ]
s+3
(s+ !}(s+2)
[ 2s:1 ] =
O
[ (s+2i9)t1+2)
-2{2s+ l )
s(s+i}Cs+2)
l
2
{s+ 1){}+2)
Y (s) = C <P(s)[q(O) + BX(s)J = [ 0 1 ]
[
_ 2 ( 2 •+1)
s(s+l)(s+2)
-4s-2 - 1 1 3
Y(s) = s(s + l)(s + 2) = s -2· s+ 1 + s+2
y(t) = .c- 1[y(s)] = (-1 - 2e- t + 3e- 2t)u(t)
10.3- 6.
433
C{<I>(s)!q(O) + BX(s)J} + DX (s)
s+l
(sl - A) = -l ] and<I>(s)=(sl -A)- 1 = 1 [s + l 1 ]
[ 1 s+l s2 + 2s + 2 - l s+1
= [ a•~!f+2 s2 +~s+2 ]
a•+2s+2 a•~"i:+2
2 2
C<I>(s)!q(O) + BX(s)J =(1 1 ) ~(s)(q(O) + BX(s)J
+ 3s + 1+s +1 ]
= [ 2s s{(s + l) 2 + l} ~
Also: DX(s) = ~
= sfts f1~;!~J
2
Hence Y (s) = C<I>(s)!q(O) + BX(s)j + DX(s) + ~ = aft:f1~;;.~J
2
Y(s) = 4s + 5s + 4 = C + As + B
s(s2 + 2s +2) s s 2 +2s+2
Using partial fractions and clearing fractions we get:
Y( ) 2 2s + 1 2 (s + 1) 1
s =5+(s+l) 2 +1 2 =5+ 2 (s+1)2 + 12 (s+l)2+1 2
H s
( )-
- (-1-)
+ s
(3s+l0) -
+ 3 s 4 -
3s+l0
s 2 + 7 s + 12
This is the same transfer function as in Prob. 10.2-8, where the cascade form state
equations were found to be
[ :~ ] = [ ~4 ~3 ] [ :~ ] + [ ~ ] x
And y = q1 = [ 1 0 ] [ :~ J
In lhis case
s+4 -1 ] 1 [ s+3 1 ]
sl - A = [ O and <I>(s) = (sl - A)- 1
s+ 3 (s + 3)(s + 4) O s+4
434
1
= [ s!4 (s+3}(s-H)
1
s+3
]
Also in our case:
3s+ 10
Hence: C<I>(s)B = = H(s)
s2 + 7s+ 12
10.3-8.
H(s) = C <I>(s)B + D
in Prob. 10.3-5 we have found cI>(s). And
5
][~]=[
8
Hence
-2
C <I>(s)B =[1 0 ] <J?(s)B = (s + l)(s + Z) and since
-2
H(s) = C <I>(s)B = s2 + 3s + 2
10.3-9. From Prob. 10.3-6,
s+l
][~]=[
(s+1)2+1 (s+l)'+l
<J?(s)B = _
1
[ s+l
(s+ l)2 +l (s+1) 2 + 1
And:
And
2
+ 1 = 2 + 38 + 4
82
H(s) = C<I>(s)B + D = s +2
(s + 1) + 1 s + 2s + 2
10.3-10. In this case:
sl - A= [ :
8
~\ ] and <I>(s ) = (sl - A)-
1
(s+1)2
1 [s + 2 l]
-1 s
435
~ l
[ (s+l)
-1
(s+J)'2
(s+1)2
s
(s+l)'
]
And:
.. +2
~] [ ~ ~] [ (ss!~)> ]
1
<I>(s) B =
[
<•~; (•+!}"
s
(s+J)2 (s+l}2 (s+1)2 (s+J)2
[4 2: ] [~
s+2 ]
(s+JJ2
C <I>(s)B = 11 -I
(s;'.'1)2 (s+l) 2
2s+l
~]
(•+1)2
4s+7
and tifu
(•+l) (•+1)2
1 1
•+l •+l
l
And cA t * Bx(t) [ e- 2 t *sin lOOt - e- 3 t *sin lOOt
_2e- '
100
2
+ 2cosl00t
100
+ 3e-100 3
' _ 3cosl00t
100
[ + e- 2• _ cos lOOt _ e- 3 • +~
100 100 100 100
-o.02e- 2
' + 0.03e- 3
t - 0.01 cos lOOt ]
[ o.01e- 2 t - 0.01e- 3t
Hence:
At A [ -34.02e- 21 + 39e-3 t + 0.01 cos lOOt ]
q(t) =e [q(O)J +e t * Bx(t) = 17.0le-2t - 10.oe- Jt
Hence
A A [ - 34.02e- 2 t + 39.03e- 3 t + 0.01 cos lOOt ]
q(t) =e tq (O) +e t *Bx= 17.0le-2t - 10.0le-3t
2
Hence: eAt = Lr-1(;,...(
'!' S
)) = [ -te- t -2t
e -e e~t ]
2
And: eA t q(O) = [ e-te-- e-2t
t 0 ] [ -1
e-t O ] = [ -e-t
O ]
s!l~s!2
l
s+l (s+l)(3+2) = [ s! l ]
4>(s) = [
0 l ]
s+2 0 s+2
And y(t) is given by: y(t) = C[eA 1 q(O) + eAtB * x(t)I + D x(t)
438
Since D = 0 = } y(t) = C[eA'q(O) + eA t * B x(t)]
And:
And hence:
10.3-16.
Hence:
And:
And
10.3-17.
H s _ 3s + 10
( ) - s 2 + 7s + 12
439
l<rom Eq. (10.65) we have:
</>(t)B = ¢(t) [ O]
1
= [ e=tsint]
e cost 1
l [.;. -
10.3-19. From Prob. 10.3-10,
[ ,,., s
~
1 l
l
(s+l)2 (•+1)2 s+ l - (s+J)'l
(s~t1'>2 ~
4 3
¢(s) = (s+l) = .!1 + c:]1p S-tJ + (s+l)2
s+2 l
1 + S-t1
1
•+l s+l s+l
10.4-1.
[ Q1 ] = [ 0 1 ] [ Q1 ] +[ 2 ] x
Q2 - 1 -1 Q2 1
440
The new state equation of the system is given by:
p-l = [ ~ ~l ] =>PA= [ ~l ~ ] [ ~l ~l ] = [ =~ -1
-2
J
p AP-1 = [ - 1 - 1 ] [ 1 - 1 J= [ - 2 1 ]
-1 -2 1 0 -3 1
And: PB= [ ~l ~ ] [ ~ ] = [ ~l ] x
Hence [ ~~ ] = [ =; ~ ] [ :~ ]+ [ ~l ] x
· I[ s+2
Isl - A l= 3 s -1
_1 JI = (s + 2)(s - 1) + 3 = s 2
- s + 2s - 2 + 3 = s2 + s + 1
-1 ±jJ3
S1,2 =
2
which are the same as in the original system.
10.4-2.
<i1 ]
[ 9.2
=[ -2
0 1 ] [
-3
Q1 ]
Q2
+[ 0 ] x t)
2 (
A = [ - 1 0 ]
0 -2
441
-pu = 2P12 ==>Pu = 2p12 } P12 = 3p12 - 2P12
P 12 = 3p12 - Pi 1 ==> If we choose Pn = 2 then P12 =1
P21 = Pn
P22 = 3p22 - P21
And if P21 =1 then P22 =1
Therefore P = [ ~ ~]
and hence
y=[ - 3 0
10.4-3.
s - 1 0
Isl - Al = 0 s - l
0 2 s+3
s{(s)(s + 3) + 2} = s(s2 + 3s + 2) = s(s + I)(s + 2) = 0
Hence the eigenvalues are: >. 1 = 0, >.2 = - 1 and >. 3 = -2.
And
[~ 0
-1
0
00
-2
l [
Pu
P21
P31
P12
P22
P32
P13
P23
p33
l =[
Pn
P21
P31
P12
P22
p32
P13
P23
p33
l[ 0
0
0
1
0
-2
442
10.4-4.
where:
(¢(s)) - 1 = [sl - A] =
[
s+ l
0
0
0
.s + 3
0 s+2
0
0
l
},. l
Hence:
I s -t-a.
'--~~~
s-r a ::t.. S' +h
'--~~~
.::t.
[ Q1
Q2
]
=[ -a
0
1 ] [
-b
Qi ]
Q2 + [ 1
(a - b)
Jx
t he output is:
443
The characteristic equation is
Isl - Al = 0 =
l s+ a
0
-1 1=
s+b (s + a)(s + b) = 0
=
we also have: w Pq and w = P AP- 1 w + PBx.
We are looking for P such that: P AP- 1 = A or AP = P A
P12
P22
= -apu
={
-ap11
lfp11 = (b-a) then P12 = l,p21 = 0
-bpz1 = -ap21 ==> P21 = 0 and p2 2 can be anything;
- ap12 = Pu - bp12 = 0
Jet's take P22 = 1
-bpz2 = P21 - /Jpz2 ==> P21 = 0
And thus:
hence:
\Ve notice that in C, there is no column with all elements zeros, hence we conclude
that the system is observable.
Controllability: In the new system (diagonalized form):
B = P B = [ b -0 a 11 ] [ a-1 b ] = [ a-b
O ]
the lst row in B is zero. We affirm that this system is not controllable.
(b) State equations:
and: y = q1 = [ 1 0 ] [ :~ ]
444
Observability:
B = PB = [ ~b ~a ] [ 1 1 ) = [ -b -a ]
in B, there is no row with all elements zeros; hence this system is controllable.
445
1 - 2z- 1 0 ]-1 [ 1 - ;2
= 0 ]-l = [ z~2 0 ]-l
[ -z-1 1 - z-1 _l
z
1_ l
z
_l
z
z-1
z
z2 [ z-1
z 0 ] [ .. .:2 0 ]
(z-l)(z-2) ~ z.-2 - % z
-z- (z-l){z-2) z-1
Also:
(zl -A)- 1 [
z - 2
-1 z-1
0 ]-l (z-l)(z-2)
1 [ z - 1
1
0 ]
[ _z-2
l
1
{z-J )(z-2)
1
z-1
2
1 1
C(I - z- Ar q (O) = [ (z-1)(z-2) + z.:1 ] = (z - l;(z _ 2)
Also
10.6-2.
E+0.32
y[nJ = £2 + E + 0.16x[nJ
(a) In this case:
z+ 0.32
H( -) = Y(z)
"" X(z) z 2
+ z + 0. 16
z + 0.32 0.2 0.8
------ = - - +- -
(z + 0.2)(z + 0.8) z + 0.2 z + 0.8
446
(b) State and output equations for the direct form 11: using the output of each delay
as a state variable we get:
q2[n)
-0.16qi(nJ - q2[n) + x[n]
Direct. fonn II
ra rat/el
Figure Sl0.6-2:
10.6-3.
nJ = E(2E + 1) x[n]
y l E 2 +E-6
(a)
z(2z + 1) _ 2z 2 + z
Y(z) = H(z)
X(z) z 2 + z - 6 - z2 + z - 6
448
(b) State and output equations for the direct form II:
Q1 [n+ lj q2lnJ
Q2ln + l] = 6q1 [nj - Q2[n) + x [n)
and
[
Q1 [n + 1J ] = [ 0
q2[n + l]
1 ] [ Q1 [n] ] + [ 0
6 - 1 Q2 [nJ 1
Jx[n)
The output equation is:
[:~[~: il ] = [ ~
3
; ] [ ~~!~l ]+ [ i ]x[n]
The output equation is:
qi[n
[ 92 [n
+ l] ] = [ 2 0 ] [ Q1 [n] ] + [ 1 ] x[nJ
+ lJ 0 -3 (f2[n] 1
The output equation is:
'149
yin]= [ 2 - 3 ] [ qi[n] ] + 2:c[nJ
q2lnJ
X[ l'=J
::t,
x['J'_J_--1
Pam He\
Figure Sl0.6-3
10.M-l. Figure 10.M-1 is used to help determine the state and output equations.
x[n] y[n]
-116
Figure 810.M-1: Direct form II realization of
y[n] +h fn - l] + ~y[n - 2] = xlnJ + ~xln - l J.
Directly from the d iagram, note that qdn+ lJ = q2lnJ+OxlnJ and q2ln + l J = - i q2lnJ-
hi [n] + xlnJ. Taken together, these yield t he state equation
y(nJ ]
= [ - ~ -! ] [ ~~!~l + l x(n] = CQlnJ + D x(n] .
10.l\ll-2. Figure 10.M-2 is used to help determine the state and out put equations.
450
x{nJ - ----- >--- -....-- y[n]
112 -516
-116
Direct.ly from the diagram, note that y[nj = v2 !n] +x[n]. In standard form, the out.put.
equation is thus
Also using the diagram, not.e t.hat v2ln + l] = vi[n] + - ~y[n] + 4x!n] and vi(n + l] "=
- hfn]. Substituting y[n] = v2[n] + x!n] into each yields v2 [n + lj = vi(n] - ~(v2!nJ +
x[n])+ ~x[n] and vi[n+ l] = - t(vz!n] +x[nJ). Simplifying to standard form, the state
equations are represented in matrix form by
451