Professional Documents
Culture Documents
Antwoorden - Fourier and Laplace Transforms, Manual Solutions
Antwoorden - Fourier and Laplace Transforms, Manual Solutions
1.1
1.2
Put c1 = A1 ei1 and c2 = A2 ei2 , then f1 (t) + f2 (t) = (c1 + c2 )eit . Let
c = c1 + c2 , then f1 (t) + f2 (t) = ceit . The signal f1 (t) + f2 (t) is again a
time-harmonic signal with amplitude | c | and initial phase arg c.
1.5
A2 /
P =
A2 cos2 (t + 0 ) dt =
(1 + cos(2t + 20 )) dt
2 /
4 /
A2
.
=
2
1.6
The energy-content is E =
1.7
R
0
e2t dt = 12 .
3
1X
| cos(n/2) |2 = 12 .
4 n=0
1.8
The energy-content is E =
sum 1/(1 e2 ).
1.9
u( ) d | u( ) | d,
n=0
t
)
d
representing the response to u(t t0 ).
0
t1
Rt
d Calculating t1 cos( ) d gives the following response: (sin(t)
sin(t ))/ =
R t2 sin(/2) cos(t /2)/.
e Calculating t1 sin( ) d gives the following response: ( cos(t) +
cos(t ))/ = 2 sin(/2) sin(t /2)/.
f From the response to cos(t) in d it follows that the amplitude response
is | 2 sin(/2)/ |.
g From the response to cos(t) in d it follows that the phase response
is /2 if 2 sin(/2)/ 0 and /2 + if 2 sin(/2)/ < 0. From
1.12
a The frequency response of the cascade system is H1 ()H2 (), since the
reponse to eit is first H1 ()eit and then H1 ()H2 ()eit .
b The amplitude response is | H1 ()H2 () | = A1 ()A2 ().
c The phase response is arg(H1 ()H2 ()) = 1 () + 2 ().
a The amplitude response is | 1 + i | e2i = 2.
b The input u[n] = 1 has frequency = 0, initial phase 0 and amplitude
1. Since ein 7 H(ei )ein , the response is H(e0 )1 = 1 + i for all n.
c Since u[n] = (ein + ein )/2 we can use ein 7 H(ei )ein to obtain
that y[n] = (H(ei )ein + H(ei )ein )/2, so y[n] = (1 + i) cos((n 2)).
d Since u[n] = (1 + cos 4n)/2, we can use the same method as in b and
c to obtain y[n] = (1 + i)(1 + cos(4(n 2)))/2.
1.13
1.14
1.16
1.17
1.18
a If u(t) = 0 for t < 0, then the integral occurring in y(t) is equal to 0 for
t < 0. For t0 0 the expression u(t t0 ) is also causal. Hence, the system
is causal for t0 0.
b It follows from the boundedness of u(t), so | u( ) | K for some constant K, that y(t) is also bounded (use the triangle inequality and the
inequality from exercise 1.9b). Hence, the system is stable.
c If u(t) is real, then the integral is real and so y(t) is real. Hence, the
system is real.
d The response is
Z t
y(t) = sin((t t0 )) +
sin( ) d = sin((t t0 )) 2(cos t)/.
t1
1.19
a If u[n] = 0 for n < 0, then y[n] is also equal to 0 for n < 0 whenever
n0 0. Hence, the system is causal for n0 0.
b It follows from the boundedness of u[n], so | u[n] | K for some constant
K and all n, that y[n] is also bounded (use the triangle inequality):
n
n
n
X
X
X
u[l] K +
| u[l] | K +
K,
| y[n] | | u[n n0 ] | +
l=n2
l=n2
l=n2
n
X
l=n2
= (1)n (1 + (1)n0 ).
2.1
2.2
2.4
We have | z1 | = 4 2, | z
2 | = 4 and arg z1 = 7/4, arg z2 = 2/3. Hence,
| z1 /z2 | = |
z1 | / | z2 | = 2 and arg(z1 /z2 ) = arg(z1 ) arg(z2 ) = 13/12,
so z1 /z2 = 2e13i/12 . Similarly we obtain z12 z23 = 2048e3i/2 and z12 /z23 =
1 3i/2
e
.
2
2.5
2.6
2.7
4
4
4
Write z 5
z + z 1 as (z 1)(z + 1) and then solve z = 1 to find
the roots 2(1 i)/2. Combining linear factors
with complex
conjugate
roots we obtain z 5 z 4 + z 1 = (z 1)(z 2 + 2z + 1)(z 2 2z + 1).
2.8
2.9
2.11
2.12
Trying the first few integers we find the zero z = 1 of the denominator. A
long division gives as denominator (z 1)(z 2 2z + 5). We then split F (z)
as A/(z 1) + (Bz + C)/(z 2 2z + 5). Multiplying by the denominator
of F (z) and comparing the coefficients of z 0 = 1, z and z 2 we obtain that
A = 2, B = 0 and C = 1.
2.13
2.14
Use integration by parts twice and the fact that a primitive of ei0 t is
ei0 t /i0 . The given integral then equals 4(1 i)/03 , since e2i = 1.
Since 1/(2 eit ) = 1/ 2 eit and 2 eit 2 eit = 1, the result
R1
R1
follows from 0 u(t) dt 0 | u(t) | dt.
2.15
2.16
2.17
P
3
a Use that | an | = 1/ n6 + 1 1/n3 and the fact that
n=1 1/n converges (example 2.17).
P
2
b Use that | an | 1/n2 and the fact that
n=1 1/n converges.
n
ni
n
n
c
Use
that
|
a
|
=
1/
ne
e
=
1/(ne
)
1/e
and the fact that
n
P
n
1/e
converges
since
it
is
a
geometric
series
with
ratio
1/e.
n=1
a
n!
1
= lim
lim
= 0.
n (n + 1)!
n n + 1
b The series is convergent; proceed as in part a:
2n+1 + 1 3n + n
2 + 1/2n
1 + n/3n
2
= lim
= .
lim n+1
n
n
n 3 + (n + 1)/3 1 + 1/2n
n 3
+n+1 2 +1
3
2.19
2
1 + 1/n2
z
n + 1
lim
= lim 2 z 2
= 2 z2 .
2
n
2n
n (n + 1) + 1 2 z
n
1 + 2/n + 2/n2
This is less than 1 if z 2 < 1/2, that is, if | z | < 2/2. Hence, the radius
of convergence is 2/2.
2.20
2.23
2.25
a Using the ratio test we obtain as limit 5/3. This is less than 1 and so
the series converges.
P
b Since (n + in )/n2 = (1/n) + (in /n2 ) and the series
n=1 1/n diverges,
this series is divergent.
P
The series
cn (z 2 )n converges for all z with z 2 < R, so it has radius
n=0
of convergence R.
2.26
2.27
2.29
2.30
(1 + i)2n+2 z n+1
n+1
= lim | z | n + 1 (1 + i)2 = 2 | z | .
lim
n
n+2
(1 + i)2n z n n
n+2
This is less than 1 if | z | < 1/2, so the radius of convergence is 1/2.
b Calculate f 0 (z) by termwise differentiation of the series and multiply
this by z. It then follows that
zf 0 (z) + f (z) =
X
n=0
(1 + i)2n z n =
(2iz)n .
n=0
This is a geometric series with ratio 2iz and so it has sum 1/(1 2iz).
4 5
3
2
1 + 2i
2
2 12
3.2
k
X
a0
+
(am cos(m0 t) + bm sin(m0 t)).
2
m=1
Now substitute this for f (t) in the right-hand side of (3.4) and use the
fact that all the integrals in the resulting expression are zero, except for
R T /2
the integral T /2 sin(m0 t) sin(n0 t) dt with m = n, which equals T /2.
Hence, one obtains bn .
3.4
,
2
cn =
(1)n 1
n2
4 X cos((2k + 1)t)
.
2
(2k + 1)2
k=0
3.7
in 1 `
(1)n e1 1 .
2(n2 2 + 1)
The Fourier coefficients are calculated by splitting the integrals into a real
and an imaginary part. For c0 this becomes:
c0 =
1
2
t2 dt +
1
i
2
t dt =
1
1
.
3
(1)n (2 n)
.
n2 2
3.10
From the values of the coefficients cn calculated earlier in exercises 3.6, 3.7
and 3.9, one can immediately obtain the amplitude spectrum | cn | and the
phase spectrum arg cn (note e.g. that arg cn = if cn > 0, arg cn = if
cn < 0, arg cn = /2 if cn = iy with y > 0 and arg cn = /2 if cn = iy
with y < 0). This results in three figures that are given separately on the
website.
3.11
cn =
sin(n/4)
n
for n 6= 0,
c0 =
1
.
4
for n 6= 0.
cn = 0
Hence, the Fourier series is 1 (!). This is no surprise, since the function is
1 for all t.
3.12
1
,
2
cn = 0
for n 6= 0 even,
cn =
2
n2 2
for n odd.
3.14
We have that f (t) = 2p2,4 (t) q1,4 (t) and so the Fourier coefficients follow
by linearity from table 1:
c0 = 3/4, cn = (2n sin(n/2) 4 sin2 (n/4))/(n2 2 ) for n 6= 0.
3.15
Note that f (t) can be obtained from the sawtooth z(t) by multiplying the
shifted version z(t T /2) by the factor T /2 and then adding T /2, that
is, f (t) = T2 z(t T2 ) + T2 . Now use the Fourier coefficients of z(t) (table 1
e.g.) and the properties from table 2 to obtain that
c0 =
T
,
2
cn =
iT
2n
for all n 6= 0.
3.17
Shifts over a period T (use the shift property and the fact that e2in = 1
for all n).
3.19
1
4
2 sin(nt/4) dt.
2
an =
8
(cos(n/2) 1)
n2 2
for all n 6= 0.
32((1)n 1)
32((1)n + 1)
64(1)n
=
,
2
2
2
2
n
n
n2 2
which also gives the Fourier cosine series. One can write this series as
8
16 X 1
+ 2
cos(nx/2).
3
n=1 n2
For the Fourier sine series we extend the function to an odd function of
period 8. As above one can calculate the coefficients bn (the an are 0). The
result is
bn =
64((1)n 1)
n3 3
for all n 1.
3.24
If f is real and the cn are real, then it follows from (3.13) that bn =
0. A function whose Fourier coefficients bn are all 0 has a Fourier series
containing cosine functions only. Hence, the Fourier series will be even. If,
on the other hand, f is real and the cn are purely imaginary, then (3.13)
shows that an = 0. The Fourier series then contains sine functions only
and is thus odd.
3.25
b The even extension has period 2a, but it has period a as well. We can
thus calculate the coefficients an and a0 as follows (the bn are 0):
Z
Z
2 a/2
2 0
a0 =
2bt/a dt
2bt/a dt = b.
a 0
a a/2
while for n 0 we obtain from an integration by parts that
Z
Z
2 a/2
2 0
an =
(2bt/a) cos(2nt/a) dt
(2bt/a) cos(2nt/a) dt
a 0
a a/2
n
2b((1) 1)
=
,
n2 2
which gives the Fourier cosine series. It can also be determined using the
result of exercise 3.6 by applying a multiplication and a scaling.
The odd extension has period 2a and the coefficients bn are given by (the
an are 0):
Z
Z
1 a/2 2bt
1 a/2 2bt
2b) sin(nt/a) dt +
sin(nt/a) dt
bn =
(
a a
a
a a/2 a
Z a
1
2bt
+
(
+ 2b) sin(nt/a) dt
a a/2 a
8b
= 2 2 sin(n/2),
n
where we used integration by parts.
/2
b
1
2
b
1
2
4.1
| cn |2
n=
a2
8 X 1
+ 2 2
2
T
T 0 n=1 n2
P
P
2
1
since sin2 (n0 a/2) 1. The series
n= | cn |
n=1 n2 converges, so
converges. The periodic triangle function is treated analogously.
4.2
This follows immediately from (3.11) (for part a) and (3.8) (for part b).
4.4
Take t = T /2 in the Fourier series of the sawtooth from example 4.2 and
use that sin(n0 T /2) = sin(n) = 0 for all n. Since (f (t+) + f (t))/2 = 0,
this agrees with the fundamental theorem.
4.6
1
,
2
cn = 0
even n 6= 0,
cn =
i
n
odd n.
The Fourier series follows by substituting the cn . One can write the series
with sines only (split the sum in two pieces: one from n = 1 to and
another from n = 1 to ; change from n to n in the latter):
1
2 X sin(2k + 1)t
+
.
2
2k + 1
k=0
X
(1)k
= .
2k + 1
4
k=0
4.7
R
a We have that c0 = (2)1 0 t dt = /4, while the Fourier coefficients
for n 6= 0 follow from an integration by parts:
Z
(1)n i
(1)n 1
1
cn =
teint dt =
+
.
2 0
2n
2n2
The Fourier series follows by substituting these cn :
X
(1)n i
(1)n 1
1
+
+
eint .
4
2
n
n2
n=,n6=0
12
13
b From the fundamental theorem it follows that the series will converge
to 12 (f (+) + f ()) = /2 at t = (note that at there is a jump). If
we substitute t = into the Fourier series, take 4 to the other side of the
=-sign, then multiply by 2, and finally split the sum into a sum from n = 1
to and a sum from n = 1 to , then it follows that
X
(1)n 1
=2
(1)n
2
2
n
n=1
(the terms with (1)n i/n cancel each other). For even n we have (1)n
1 = 0 while for odd n this will equal 2, so (4.10) results:
X
2
1
=
.
8
(2k 1)2
k=1
4.9
8 X sin(2k + 1)t
.
3
(2k + 1)3
k=0
4.10
Now substitute t = 1/2 and use that f (1/2) = 1/4 and sin((2k + 1)/2) =
(1)n to obtain the required result.
P
2
2
Use (3.8) to write the right-hand side of (4.14) as a20 /4 + 21
n=1 (an + bn ).
4.12
4.13
a The Fourier coefficients are (see table 1 or section 3.4.2 and use that
sin2 (n/2) = 1 for n odd and 0 for n even): cn = 2/n2 2 for n odd, 0 for
n 6= 0 even and c0 = 1/2. From Parseval for f = g, so from (4.14), it then
follows that (calculate the integral occurring in this formula):
1
1
8 X
1
= + 2
3
4
(2k 1)4
k=1
14
X
X
X
1
1
1
=
+
,
4
4
n
(2k)
(2k
+
1)4
n=1
k=1
k=0
1
4
4
1 X 1
=
S+
.
+
4
16
k
96
16
96
k=1
4.15
P
1
4
Solving for S we obtain
n=1 n4 = 90 .
Rb
Rb
Ra
Since a f (t) dt = T /2 f (t) dt T /2 f (t) dt, we can apply theorem 4.9
twice. Two of the infinite sums cancel out (the ones representing h0 in
theorem 4.9), the other two can be taken together and lead to the desired
result.
4.16
This follows from exercise 4.15 by using (3.8), so cn = (an ibn )/2 and
cn = (an + ibn )/2 (n N).
4.17
a
4
X
n=0
sin(2n + 1)t
.
2n + 1
b Since
Z t
cos(2n + 1)t
1
sin(2n + 1) d =
,
2n
+
1
2n
+1
4X
1
4 X cos(2n + 1)t
.
n=0 (2n + 1)2
n=0 (2n + 1)2
4.20
4.22
15
1
1
2X
1
sin t + +
cos 2kt.
2
1 4k2
k=1
1
4X
k
cos t
sin 2kt.
2
1 4k2
k=1
Write down the expression for Si(x) and change from the variable t to
t, then it follows that Si(x) = Si(x).
4.26
a From the definition of Si(x) it follows that Si0 (x) = sin x/x. So Si0 (x) =
0 if sin x/x = 0. For x > 0 we thus have Si0 (x) = 0 for x = k with k N.
A candidate for the first maximum is thus x = . Since sin x/x > 0 for
0 < x < and sin x/x < 0 for < x < 2, it follows that Si(x) indeed has
its first maximum at x = .
b The value at the first maximum is Si(). Since Si() = 1.852 . . . and
/2 = 1.570 . . ., the overshoot is 0.281 . . .. The jump of f at x = 0 is
= 3.141 . . ., so the overshoot is 8.95 . . .%, so about 9%.
4.28
4 X (1)n
2 X (1)n
sin nt + 2
sin(2n + 1)t.
n=1 n
n=0 (2n + 1)2
Use table 1 to obtain the Fourier coefficients and then apply Parseval, that
is, (4.13). Calculating the integral in Parsevals identity will then give the
first result; choosing a = /2 gives the second result.
4.30
a The Fourier series has been determined in the last example of section
3.6. Since f is continuous (and piecewise smooth), the Fourier series converges to f (t) for all t R:
16
f (t) =
2
4X
1
cos 2nt.
n=1 4n2 1
b First substitute t = 0 in the Fourier series; since f (0) = 0 and cos 2nt =
1 for all n, the first result follows. Next substitute t = /2 in the Fourier
series; since f (/2) = 1 and cos 2nt = (1)n for all n, the second result
follows.
c One should recognize the squares of the Fourier coefficients here. Hence
we have to apply Parsevals identity (4.14), or the alternative form given
in exercise 4.10. This leads to
Z
1
4
1 X
16
sin2 t dt = 2 +
.
2
2
Now change the variable from to and use that f2 is odd, then it
follows that (f1 f2 )(t) = (f1 f2 )(t).
b The convolution product equals
Z
1 1
(f f )(t) =
f (t ) d.
2 1
Since f is odd, part a implies that f f is even. It is also periodic with
period 2, so it is sufficient to calculate (f f )(t) for 0 t 1. First note
that f is given by f (t) = t 2 for 1 < t 2. Since 1 1 and
0 t 1 we see that t 1 t t + 1. From 0 t 1 it follows
that 1 t 1 0, and so close to = 1 the function f (t ) is given
by t . Since 1 t + 1 2, the function f (t ) is given by t 2
close to = 1. Hence, we have to split the integral precisely at the point
where t gets larger than 1, because precisely then the function changes
from t to t 2. But t 1 precisely when t 1, and so we
have to split the integral at t 1:
Z
Z
1 1
1 t1
(t 2) d +
(t ) d.
(f f )(t) =
2 1
2 t1
It is now straightforward to calculate the convolution product. The result
is (f f )(t) = t2 /2 + t 1/3.
c From section 3.4.3 or table 1 we obtain the Fourier coefficients cn of
the sawtooth f and applying the convolution theorem gives the Fourier
coefficients of (f f )(t), namely c20 = 0 and c2n = 1/ 2 n2 (n 6= 0).
d Take t = 0
R 1in part c; since f is odd and real-valued we can write
(f f )(0) = 12 1 | f ( ) |2 d , and so we indeed obtain (4.13).
e For 1 < t < 0 we have (f f )0 (t) = t 1, while for 0 < t < 1
we have (f f )0 (t) = t + 1. Since f f is given by t2 /2 + t 1/3 for
0 < t < 2, (f f )0 (t) is continuous at t = 1. Only at t = 0 we have that
f f is not differentiable. So theorem 4.10 implies that the differentiated
series represents the function (f f )0 (t) on [1, 1], except at t = 0. At
t = 0 the differentiated series converges to ((f f )0 (0+) + (f f )0 (0))/2 =
(1 1)/2 = 0.
f The zeroth Fourier coefficient of f f is given by
1
2
Z
(f f )(t) dt =
17
(t2 /2 + t 1/3) dt = 0.
This is in agreement with the result in part c since c20 = 0. Since this
coefficient is 0, we can apply theorem 4.9. The function
R t represented by the
integrated series is given by the (periodic) function 1 (f f )( ) d . It is
also odd, since f is even and for 0 t 1 it equals
Z 0
Z t
( 2 /2 1/3) d +
( 2 /2 + 1/3) d = t(t 1)(t 2)/6.
1
5.1
For a stable LTC-system the real parts of the zeroes of the characteristic
polynomial are negative. Fundamental solutions of the homogeneous equations are of the form x(t) = tl est , where s is such a zero and l 0
some integer. Since tl est = | t |l e(Re s)t and Re s < 0 we have that
limt x(t) = 0. Any homogeneous solution is a linear combination of
the fundamental solutions.
5.2
1
,
2
u2k = 0,
u2k+1 =
(1)k
(2k + 1)
(u = p,2 , so use table 1 and the fact that sin(n/2) = (1)k for n = 2k+1
odd and 0 for n even). Since H() = 1/(i + 1) and yn = H(n0 )un =
H(n)un it then follows that
y0 =
1
,
2
y2k = 0,
y2k+1 =
(1)k
.
(1 + (2k + 1)i)(2k + 1)
5.3
5.4
Note that u has period and that the integral to be calculated is thus the
zeroth Fourier coefficient of y. Since y0 = H(00 )u0 = H(0)u0 Rand H(0) =
5.5
H() =
2 + 1
.
2 + 4 + 2i
45
30
1
cos 4t +
sin 4t + .
104
104
8
We have that
1
H() =
.
2 + 02
18
19
,
2
u2k+1 =
2
.
(2k + 1)2 2
1
2 un ,
n2 +0
For the thin rod the heat equation (5.8) holds on (0, L), with initial condition (5.9). This leads to the fundamental solutions (5.15), from which the
superposition (5.16) is build. The initial condition leads to a Fourier series
with coefficients
Z
Z
2 L/2
2 L
An =
x sin(nx/L) dx +
(L x) sin(nx/L) dx,
L 0
L L/2
which can be calculated using an integration by parts. The result is: An =
(4L/n2 2 ) sin(n/2) (which is 0 for n even). We thus obtain the (formal)
solution
u(x, t) =
5.9
ut = kuxx
ux (0, t) = 0, u(L, t) = 0
u(x, 0) = 7 cos(5x/2L)
2 kt/4L2
cos((2n + 1)x/2L).
Superposition gives
u(x, t) =
An e(2n+1)
2 kt/4L2
cos((2n + 1)x/2L).
n=0
X
n=0
20
Since the right-hand side consists of one harmonic only, it follows that
A2 = 7 and An = 0 for all n 6= 2. The solution is thus u(x, t) =
2
2
7e25 kt/4L cos(5x/2L).
5.11
ut = kuxx
u(0, t) = 0, ux (L, t) = 0
u(x, 0) = f (x)
b Going through the steps one obtains the same fundamental solutions as
in exercise 5.9. The coefficients An cannot be determined explicitly here,
since f (x) is not given explicitly.
5.12
The equations are given by (5.17) - (5.20), where we only need to substitute
the given initial condition in (5.19), so u(x, 0) = 0.05 sin(4x/L) for 0
x L. All steps to be taken are the same as in section 5.2.2 of the textbook
and lead to the solution
u(x, t) =
An cos(nat/L) sin(nx/L).
n=1
n=1
Since the right-hand side consists of one harmonic only, it follows that
A4 = 0.05 and An = 0 for all n 6= 4. The solution is thus u(x, t) =
0.05 cos(4at/L) sin(4x/L).
5.15
An cos(nat) cos(nx).
n=0
X
n=0
An cos(nx) = kx
21
R
R
We have A0 = (2/) 0 kx dx = k and An = (2/) 0 kx cos(nx) dx for
n 6= 0, which can be calculated by an integration by parts: An = 0 for n
even (n 6= 0) and An = 4k/n2 for n odd. The solution is thus
u(x, t) =
k
4k X
1
5.16
a From H() = H() and yn = H(n0 )un follows that the response
y(t) to a real signal u(t) is real: since un = un we also have yn = yn .
b Since we can write sin 0 t = (ei0 t ei0 t )/2i, the response is equal
to (H(0 )ei0 t H(0 )ei0 t )/2i, which is ((1 e2i0 )2 ei0 t (1
e2i0 )2 ei0 t )/2i. This can be rewritten as sin 0 t 2 sin(0 (t 2)) +
sin(0 (t 4)).
P
2int
.
c A signal with period 1 has Fourier series of the form
n= un e
P
2int
, which is 0 since H(2n) = 0 for
The response is n= H(2n)un e
all n.
5.18
X
X
3
P =
| y(t) |2 dt =
| yn |2 =
| H(n0 )un |2 .
2 0
n=
n=
We have that
H() =
1 + i
.
4 2 + i(4 2 )
1
2
For the rod we have equations (5.8) - (5.10), where we have to take f (x) =
u0 in (5.10). The solution is thus given by (5.16), where now the An are the
Fourier coefficients of the function u0 on [0, L]. These are easy te determine
(either by hand or using tables 1 and 2): An = 0 for n even, An = 4u0 /n
for n odd. This gives
u(x, t) =
2 2
2
4u0 X
1
e(2n+1) kt/L sin((2n + 1)x/L).
2
n=0 (2n + 1)
Substituting x = L/2 in the x-derivative and using the fact that cos((2n +
1)/2) = 0 for all n leads to ux (L/2, t) = 0.
5.20
2 2
2
2a X 1
(1 cos(n/2))en kt/L sin(nx/L).
n=1 n
22
b The two rods together form one rod and so part a can be applied with
L = 40, k = 0.15 and a = 100. Substituting t = 600 in u(x, t) from part
a then gives the temperature distribution. On the boundary between the
rods we have x = 20, so we have to calculate u(20, 600); using only the
contibution from the terms n = 1, 2, 3, 4 we obtain u(20, 600) 36.4.
c Take k = 0.005, a = 100, L = 40, substitute x = 20 in u(x, t) from
part a, and now use only the first two terms of the series to obtain the
equation u(20, t) 63.662e0.0000308t = 36 (terms of the series tend to 0
very rapidly, so two terms suffice). We then obtain 18509 seconds, which
is approximately 5 hours.
6.1
6.2
6.4
R
We have to calculate (the improper integral) eit dt. Proceed as in
eaxample 6.1, but we now have to determine limB eiB . This limit
does not exist.
R
a We have to calculate G() = 0 e(a+i)t dt, which can be done precisely as in section 6.3.3 if we write a = + i and use that e(a+i)R =
eR ei(+)R . If we let R then this tends to 0 since > 0.
b The imaginary
of G() is /(a2 + 2 ) and applying the substituR part
2
tion rule gives /(a + 2 ) d = 21 ln(a2 + 2 ), so this improper integral,
which is the Fourier integral for t = 0, does not exits (limA ln(a2 + A2 )
does not exist e.g.).
c We have lima0 g(t) = lima0 (t)eat = (t), while for 6= 0 we have
that lima0 G() = i/.
To calculate the spectrum we split the integral at t = 0:
Z 1
Z 0
G() =
teit dt
teit dt.
1
2(cos 1)
2 sin
+
.
R1
For = 0 we have that G(0) = 2 0 t dt = 1. Since lim0 sin / = 1 and
lim0 (cos 1)/ 2 = 21 (use e.g. De lH
opitals rule), we obtain that
lim0 G() = G(0), so G is continuous.
6.5
F () = 2i
cos(a/2) 1
for 6= 0,
F (0) = 0.
b Using Taylor or De lH
opital it follows that lim0 F () = 0 = F (0),
so F is continuous.
6.7
6.8
6.9
sin2 (a/2)
12
+ 8i
.
2
4+
a 2
7
7
+
.
49 + ( )2
49 + ( + )2
a From the shift property in the frequency domain (and linearity) it follows that the spectrum of f (t) sin at is F ( a)/2i F ( + a)/2i.
b Write f (t) = p2 (t) sin t, obtain the spectrum of p2 (t) from table 3 and
apply part a (and use the fact that sin( ) = sin()), then
F () =
2i sin()
.
2 1
23
24
6.10
Use section 6.3.3 (or exercise 6.2) and the modulation theorem 6.17, and
write the result as one fraction, then
(F (t)eat cos bt)() =
a + i
.
(a + i)2 + b2
Similarly it follows from section 6.3.3 (or exercise 6.2) and exercise 6.9a
that
(F (t)eat sin bt)() =
6.12
b
.
(a + i)2 + b2
Write
Z
F () =
f (t)eit dt +
f (t)eit dt
6.14
Calculate the spectrum in a direct way using exactly the same techniques
as in example 6.3.3 (or use (6.20) and twice an integration by parts):
F () =
2i
.
1 + 2
6.16
R a/2
The spectrum is given by a/2 teit dt, which can be calculated using an
integration by parts. The result is indeed equal to the formula given in
example 6.3.
6.17
2
of the Gauss function, of course) it follows that i e /4a /(2a a) is
the spectrum of tf (t).
b If we divide the Fourier transform of f 0 (t) by 2a, then we indeed
obtain the same result as in part a.
6.18
Two examples are the constant function f (t) = 0 (k arbitrary), and the
2
Gauss function et /2 with k = 2. Using exercise 6.17a we obtain the
2
function tet /2 with k = i 2.
6.19
Use table 3 for (t)eat and then apply the differentiation rule in the frequency domain, then the result follows: (a+i)2 . (Differentiate (a+i)1
just as one would differentiate a real function.)
6.20
6.21
Use the fact that limx xa ex = 0 for all a R and change to the variable
2
x = at2 in tk /eat (separate the cases t 0 and t < 0). Then part a follows
and, hence, part b also follows since we have a finite sum of these terms.
6.22
Apply the product rule repeatedly to get an expression in terms of the derivatives of f and g (this involves the binomial coefficients and is sometimes
called Leibniz rule). Since f and g belong to S, tn (f (t)g(t))(m) will be a
sum of terms belonging to S, and so the result follows.
6.23
6.25
6.26
25
R
We have that ( )(t) = 0 (t ) d . Now treat the cases t > 0 and
t 0 separately, then it follows that ( )(t) = (t)t. (If t 0, then
t < 0 for > 0 and
R t so (t ) = 0; if t > 0 then (t ) = 0 for
> t and the integral 0 1 d = t remains.) Since (t)t is not absolutely
integrable, the function ( )(t) is not absolutely integrable.
R
From the causality of f it follows that (f g)(t) = 0 f ( )g(t ) d . For
Rt
t < 0 this is 0. For t 0 it equals 0 f ( )g(t ) d .
a
2
the spectrum of tg(t) is iG0 () = i 2e /2 . Since (F tg(t))(0) = 0,
2
we may apply the integration rule to obtain that F1 () = 2e /2 .
b Apply the differentiation rule in the frequency domain with n = 2, then
2
F2 () = 2(1 2 )e /2 .
c Since f3 (t) = f2 (t 1), it follows from the shift property that F3 () =
ei F2 ().
2
d From part a and exercise 6.9 it follows that F4 () = ( 2e(4) /2 +
2
2e(+4) /2 )/2i.
e Use the scaling property from table 4 with c = 4, then F5 () =
G(/4)/4.
6.29
26
7.1
7.2
7.3
Take C > 0, then it follows by first changing from the variable Au to v and
then applying (7.3) that
Z C
Z AC
sin Au
sin v
lim
du = lim
dv = .
A 0
A 0
u
v
2
7.4
Split 1/(a+i) into the real part 1/(1+ 2 ) and the imaginary part /(1+
2 ). The limit of A of the integrals over [A, A] of these parts gives
limA 2 arctan A = for the real part and limA (ln(1 + A2 ) ln(1 +
(A)2 )) = 0 for the imaginary part.
7.6
2
a In exercise 6.9b it was shown that F ()
R = 2i sin()/(
R 1). The
function f (t) is absolutely integrable since | f (t) | dt = | sin t | dt <
. Moreover, f (t) is piecewise smooth, so all conditions of the fundamental
theorem are satisfied. We now show that the improper integral of F ()
exists. First, F () is continuous on R according to theorem
R 6.10, so it
is integrable over e.g. [2, 2]. Secondly, the integrals 2 F () d and
R 2
F () d both exist. For the former integral this can be shown as
F
()
d
d
2 1
2
2
since | 2i sin() | 2 (and 2 1 > 0 for > 2). The integral in the
right-hand side is convergent.
b Apply the fundamental theorem, then
Z
2i sin() it
1
f (t) =
e d
2 2 1
for all t R (f is continuous). Now use that F () is an odd function and
that 2 sin sin t = cos( t) cos( + t), then
Z
1 cos( t) cos( + t)
d.
f (t) =
0
1 2
7.8
2
Since f (t) is continuous for t > 0 and t 6= a we have for these values that
Z
2 1 cos a
f (t) =
sin t d.
0
27
28
7.11
7.12
The spectrum of qa (t) is F () = 4 sin2 (a/2)/(a 2 ) (see table 3). From duality it then follows that the spectrum of sin2 (at/2)/t2 is (a/2)qa (). (We
can apply duality since qa is continuous, piecewise smooth, and absolutely
integrable and since its Fourier integral exists as improper integral; this
latter fact follows immediately if we use that F () is even and continuous
and that e.g. F () 1/ 2 for 1).
7.14
7.15
These results follow immediately from duality (and calculating the right
p
2
2
constants). For example: /aet /4a 2ea , now divide by 2.
7.16
7.17
Again, this is an important exercise: it teaches to recognize useful properties for the inverse transform.
a We immediately use table 3 to obtain that 1/(4 + 2 ) is the spectrum
of f (t) = e2| t | /4.
b Apply a shift in the frequency domain to the spectrum p2a () of
sin(at)/t, then it follows that f (t) = (ei0 t + ei0 t ) sin(at)/(t), so f (t) =
2 cos(0 t) sin(at)/(t).
c As in part b it follows that f (t) = 3e9it /((t2 + 9)).
7.19
29
(F Pb )() = e(a+b)| | , where we also used table 3. But also (F Pa+b )() =
e(a+b)| | , and since F is one-to-one (theorem 7.4) it then follows that
Pa+b = Pa Pb .
7.21
a Use the result of exercise 6.14 (G() = 2i/(1+ 2 )), the fundamental
R
R
theorem and the fact that the spectrum is odd to change from to 0 .
It then follows that (use x instead of )
Z
x sin xt
dt = et .
2
1
+
x
2
0
Since g is not continuous at t = 0, this result is not correct at t = 0. Here
one should take the average of the jump, which is 0.
R
b We apply Parseval (formula (7.19)) and calculate | g(t) |2 dt =
R0
R
R
fact that the integrand is even. Writing x instead of , the result follows.
7.22
Parseval (7.18)R with f (t) = ea| t | and g(t) = eb| t | and calculate
RUse
2
2
2a/(a + ) and 2b/(b2 + 2 ) (table 3).
7.23
a Since sin4 t/t4 is the square of sin2 t/t2 and (F sin2 t/t2 )() = q2 ()
(table 3), it follows from the convolution theorem in the frequency domain
that (F sin4 t/t4 )()
R = (/2)(q2 q2 )().
b The integral sin4 t/t4 dt is the Fourier transform of sin4 x/x4 calR
culated at = 0, hence sin4 x/x4 dx = (/2)(q2 q2 )(0). Using that
q2 is an even function we obtain that
Z
Z 2
(q2 q2 )(0) =
q2 (t)q2 (t) dt = 2
(1 t/2)2 dt.
7.26
The Gauss function f (t) = eat belongs to S and so we can apply Poissons
p
2
summation formula. Since F () = /ae /4a (see table 3), it follows
from (7.23) with T = 1 that
X
n=
ean =
X
p
2 2
/a
e n /a .
n=
Take f (t) = a/(a2 +t2 ), then F () = ea| | (see table 3); we can then use
(7.22) with T = 1 (in example 7.8 the conditions were verified) to obtain
30
X
n=
a
=
a2 + (t + n)2
1+
2n(a+it)
n=1
!
2n(ait)
n=1
Here we have also split a sum in terms with n = 0, n > 1 and n < 1, and
then changed from n to n in the sum with n < 1. The sums in the righthand side are geometric series with ratio r = e2(a+it) and r = e2(ait)
respectively. Note that | r | < 1 since a > 0. Using the formula for the
sum of an infinite geometric series (example 2.16), then writing the result
with a common denominator, and finally multiplying everything out and
simplifying, it follows that
a X
1
1 e4a
.
=
2
2
4a
n= a + (t + n)
1+e
e2a (e2it + e2it )
a To determine the spectrum we write sin t = (eit eit )/2 and calculate
the integral defining F () in a direct way:
Z
Z
1
i(1)t
i(1+)t
F () =
e
dt
e
dt .
2i
0
0
Writing the result with a common denominator and using the fact that
ei = ei = 1 gives F () = (1 + ei )/(1 2 ). From theorem 6.10
we know that F () is continuous, so we do not have to calculate F () at
the exceptional points = 1.
b Apply the fundamental theorem, so (7.9), noting that f (t) is continuous
on R. We then obtain
Z
1
1 + ei it
f (t) =
e d.
2 1 2
Split the integral at t = 0 and change from to in the integral over
(, 0]. Then
Z it
1
e + eit + ei(t) + ei(t)
f (t) =
d,
2 0
1 2
which leads to the required result.
c Take t = /2 in part b and use that f (/2) = 1, thenRthe result follows.
| F () |2 d = .
2
2
i/2
Since F () can
cos(/2)/(1 2 ) and we have
be rewritten as 2e
i/2
2
that e
= 1, it follows that | F () | = 4 cos2 (/2)/(1 2 )2 . This
integrand being even, the result follows.
7.29
a We know from table 3 that p2a (t) 2 sin a/ and e| t | 2/( 2 + 1).
From the convolution theorem it then follows that p2a (v)e| v | 4f () =
G().
b We now determine g explicitly by calculating the convolution product
(use the definition of p2a ):
Z t+a
Z a
(p2a (v) e| v | )(t) =
e| t | d =
e| u | du
a
ta
31
8.1
8.2
Z
1
2 sin a
lim
f (t ) d = f (t)
2 a
for any absolutely integrable and piecewise smooth function f (t) on R that
is continuous at t. Using (8.3) this can symbolically be written as (take
t = a)
Z
()f (a ) d = f (a)
8.4
1 (t) dt +
2 (t) dt,
For S there exists a constant M > 0 such that (e.g.) (1+t2 ) | (t) | M
for all t R. Hence,
Z
Z
Z
|
(t)
|
dt
M
dt <
(t)
dt
1 + t2
0
0
0
32
8.7
33
R
(the latter integral equals [arctan]
0 = /2). The integral 0 (t) dt thus
exists and one can now show that is indeed a distribution precisely as in
exercise 8.4 (linearity of integration).
R
In example 8.4 it was already motivated why the integral | t | (t) dt
exists: there exists a constant M > 0 such that (e.g.) (1 + t2 ) | t(t) | M
for all t R. Hence,
Z
Z
Z
|
t(t)
|
dt
M
|
t
|
(t)
dt
dt < .
2
1 + t
8.9
a
Z
R
hence, | t |
is integrable over [1, 1]. Since 0 t1/2 dt = 2 limR R
does not exist, | t |1/2 is not integrable over R.
R
b We first show that | t |1/2 (t) dt exists for S. To do so, we
split the integral in an integral over [1, 1] and over | t | 1. For the first
integral we note that | (t) | M1 for some constant M1 > 0. From part a
we then get
Z 1
Z 1
Z 1
1/2
1/2
|
t
|
(t)
dt
|
t
|
|
(t)
|
dt
M
| t |1/2 dt < .
1
1/2
1/2
| t |1/2 (t) dt
| (t) | dt
| (t) | dt.
| t |1
| t |1
In example
8.1 itEhas been shown that the latter integral exists. This shows
D
1/2
that | t |
, exists and one can now show that | t |1/2 is indeed a
distribution (linearity of integration; see e.g. exercise 8.4).
8.10
dt < .
t(t) dt
| t(t) | dt M
1
+
t2
(1 +
8.14
a From (8.17) it follows that the complex number (3) (0) is assigned.
b This number (3) (0) is meaningfull for all functions that are 3 times
continuously differentiable.
8.15
a First apply example 8.3, then definition 8.4, and finally integration by
parts, then
34
(sgn t)0 , =
Since the function | t | from example 8.9 is continuously differentiable outside t = 0, it follows from the jump formula that | t |0 = sgn t (at t = 0
there is no jump and outside t = 0 this equality holds for the ordinary
derivatives). The function from example 8.10 has a jump of magnitude 1
at t = 0, while for t < 0 the derivative is 0 and for t > 0 the derivative is
sin t. Hence, the jump formula implies that ((t) cos t)0 = (t) (t) sin t.
8.17
a The function pa has a jump of magnitude 1 at t = a/2 and of magnitude 1 at t = a/2. Outside t = 0 the ordinary derivative is 0, so it
follows from the jump formula that pa (t)0 = (t + a/2) (t a/2).
b The function (t) sin t has no jump at t = 0, for t < 0 the ordinary
derivative is 0 and for t > 0 the ordinary derivative is cos t, so it follows
from the jump formula that ((t) sin t)0 = (t) cos t.
8.18
8.19
From the jump formula it follows that the derivative as distribution is given
by a(t)eat + (t), so f 0 (t) af (t) = (t) as distributions.
8.22
8.23
8.25
First apply definition 8.6 and then the definition of pv(1/t) from example
8.5 to obtain
Z
Z
t(t)
ht pv(1/t), i = lim
dt = lim
(t) dt.
0 | t |
0 | t |
t
Since R S is certainly integrable over R, the limit exists and it will be
8.26
35
hT (t), (t)i = hT (t), (t)i for all S, where we used definition 8.7.
Similarly for odd T .
8.27
a From the definition of sgn t in example 8.3 it follows that hsgn t, (t)i =
hsgn t, (t)i for all S. This shows that sgn t is odd according to
exercise 8.26. Similarly for pv(1/t) (change from t to t in the integrals
defining pv(1/t)).
b From the definition of | t | in example 8.4 it follows that h| t | , (t)i =
h| t | , (t)i for all S (change from t to t in the integral defining | t |).
This shows that | t | is even according to exercise 8.26.
8.29
and in e.g. exercises 8.10, 8.12b and 8.18a we have seen that such integrals
are well-defined for S. This gives a mapping from S to C and the
linearity of this mapping follows precisely as in e.g. exercise 8.3 or 8.4.
Hence, f indeed defines a distribution Tf .
b Apply the jump formula (8.21): outside t = 0 the function f is continuously differentiable with derivative f 0 (t) = 2t for t > 0 and f 0 (t) = 2 for
t < 0. Note that f 0 again defines a distribution Tf 0 . At t = 0 the function
has no jump, hence (8.21) implies that Tf0 = Tf 0 .
c Again we have that the function f 0 is continuously differentiable outside t = 0 and f 00 (t) = 2 for t > 0 and f 00 (t) = 0 for t < 0. Let Tf 00
be the distribution defined by f 00 . At t = 0 the function f 0 has a jump
f 0 (0+) f 0 (0) = 0 2 = 2, and according to (8.21) (applied to Tf 0 and
using that Tf0 = Tf 0 and so Tf00 = Tf0 0 ) we have that
Tf00 = Tf0 0 = Tf 00 + (f 0 (0+) f 0 (0))(t) = Tf 00 2(t).
The second derivative of f considered as distribution is the same as the
second derivative of f outside t = 0, minus the distribution 2(t) at t = 0.
8.30
a Since 00 (t) can be defined for all twice continuously differentiable functions, the product f (t) 00 (t) can also be defined for all twice continuously
differentiable functions f (t) by hf (t) 00 (t), (t)i = h 00 (t), f (t)(t)i. This is
because it follows from the product rule that the product f (t)(t) is again
twice continuously differentiable.
b From part b and the definition of the second derivative of a distribution
(formula (8.17) for k = 2) we obtain hf (t) 00 (t), (t)i = h(t), (f (t)(t))00 i.
Since (f (t)(t))00 = f 00 (t)(t) + 2f 0 (t)0 (t) + f (t)00 (t) we thus obtain
that hf (t) 00 (t), (t)i = f 00 (0)(0) + 2f 0 (0)0 (0) + f (0)00 (0), which equals
hf 00 (0)(t) 2f 0 (0) 0 (t) + f (0) 00 (t), (t)i ( S). This proves the identity.
c Apply part b to the function f (t) = t2 and use that f (0) = f 0 (0) = 0
and f 00 (0) = 2, then t2 00 (t) = 2(t). Next apply b to f (t) = t3 and use
that f (0) = f 0 (0) = f 00 (0) = 0, then it follows that t3 00 (t) = 0.
d According to definition 8.7 we have that
00
9.1
9.3
9.4
a From example 9.1 (or table 5) we obtain the result e5it /2.
b See example 9.2: 2 cos 2t.
c The spectrum of pv(1/t) is isgn (table 5). Note that 2 cos =
ei + ei and that the spectrum of (t a) is eia (table 5). Hence the
answer is i 1 pv(1/t) + (t 1) + (t + 1).
9.5
9.7
It is obvious that (t) = (1 + sgn t)/2 by looking at the cases t > 0 and
t < 0. Since 2() is the spectrum of 1 and 2ipv(1/) is the spectrum
of (t), it follows that (t) has spectrum () ipv(1/).
9.8
00
a Subsequently
apply
definitions 9.1 and 8.7: hFT (at), i = hT (at), i =
36
37
From table 5 it follows that 0 (t) i. Using (9.12) we then obtain that
it 0 (t) (i)0 = i, so t 0 (t) 1. Exercise 8.23c gives: t 0 (t) = (t)
and since (t) 1 we indeed get t 0 (t) 1 again. Similarly we get
t 00 (t) 2i using (9.12) or using exercise 8.30b: t 00 (t) = 2 0 (t).
9.12
9.13
The linearity follows as in definition 8.6. The main point is that one has
to show that eiat (t) S whenever S. So we have to show that
for
that tn eiat (k) (t) < M for some M > 0 and all k, n Z+ . But since
iat
e = 1 this means that we have to show that tn (k) (t) < M for some
M > 0 and all k, n Z+ , which indeed holds precisely because S.
9.15
9.16
a Use table 5 for (t) and apply a shift in the time domain, then it follows
that (t 1) ei (() ipv(1/)).
b Use table 5 for (t) and apply a shift in the frequency domain, then it
follows that eiat (t) ( a) ipv(1/( a)).
c We have (t) () ipv(1/) and if we now write the cosine as
a combination of exponentials, then we can use a shift in the frequency
domain (as in part b) to obtain that (t) cos at 12 (( a) ipv(1/(
a)) + ( + a) ipv(1/( + a))).
d Use that 1 2() and 0 (t) i (table 5), so 3i 6i() and
(apply a shift) 0 (t 4) e4i i; the sum of these gives the answer.
e First note that (t)sgn t = (t) and the spectrum of this is known;
furthermore we have that t3 2i3 (3) () (table 5), so the result is
2 2 (3) () + () ipv(1/).
9.17
a Use table 5 for the sign function and apply a shift: 21 ieit sgn t.
b Write sin t as a combination of exponentials and apply a shift to 21 isgn t,
then we obtain the result 14 (sgn(t + 3) sgn(t 3)).
c Apply reciprocity to (t), then we obtain ((t) ipv(1/t))/2
(). Now (t) = (t) and pv(1/t) = pv(1/t), hence, the result is:
1
i 1 pv(1/t) + 12 (t).
2
d Apply the scaling property (table 6) to 1 2() to obtain 1
6(3). Next we apply a shift in the frequency domain (table 6), which
38
results in e2it/3 6(3 2). Using the differentiation rule in the frequency domain (table 6) we obtain from 1/2 () that (it)2 /2
00 (). From linearity it then follows that (31 e2it/3 t2 )/2 (3
2) + 00 ().
9.19
9.20
Use exercise 9.20 with T (t) = (t b). The convolution theorem leads to
the obvious eib eia = ei(a+b) .
9.24
9.21
2 () 4
k=
9.25
9.26
39
10.1
a When the system is causal, then the response to the causal signal (t)
is again causal, so h(t) is causal. On
R t the other hand, if h(t) is causal, then
it follows that y(t) = (u h)(t) = h(t )u( ) d and if we now have
a causal input u, then the integral will be 0 for t < 0 and so y(t) is causal
as well, proving that the system is causal.
b When the system is real, then the response to the real signal (t) is
again real, so h(t) is real. On the other hand, if h(t) is real, then it follows
from the integral for y(t) = (u h)(t) that if u(t) is real, then y(t) is also
real, proving that the system is real.
10.2
10.3
We can express p2 (t) as p2 (t) = (t + 1) (t 1). Since a(t) is (by
definition) the response to (t) and we have a linear system, the response
to p2 (t) = (t + 1) (t 1) is a(t + 1) a(t 1).
10.5
(e.g., both are smaller than 0 e3t dt), this is indeed the case and hence
the system is stable.
10.6
10.7
10.8
10.9
41
b The function u has a Fourier series with terms cn eint (note that 0 = 1).
But the response to eint is H(n)eint and H(n) = 0 for n > 1 and n < 1.
So we only have to determine c0 , c1 and c1 . These can easily be calculated
from the defining integrals: c0 = 21 and c1 = c1 = 1/. Hence, the
2
response follows: y(t) = H(0)c0 +H(1)c1 eit +H(1)c1 eit = 12 3
cos t.
10.11
=1
.
(s 1)(s + 2)
3s1
3s+2
3 i 1
3 i + 2
Now (t) 1 and (t)e2t 1/(i + 2) (table 3, no. 7) and from time
1
=
reversal (scaling with a = 1 from table 4, no. 5) it follows for i1
t
t
2t
1
1
2
4
that
(t)e
.
Hence,
h(t)
=
(t)
+
(t)e
(t)e
.
i()+1
i1
3
3
b The impulse reponse h(t) is not causal, so the system is not causal.
c The modulus of H() is 1, so it is an all-pass system and from Parseval
it then follows that the energy-content of the input is equal to the energycontent of the output (if necessary, see the textbook, just above example
10.7).
10.13
2 02
.
2 i 20 02
1 + /2
1
i + 1 +
= +
,
2i +
2
2i +
where we also used a long division. Using the tables it then follows that
h(t) = 12 ((t) + (1 + /2)(t)et/2 ).
b We have to interpret the differential equation the other way around, so
with input and output interchanged. This means that the system function
is now 1/H(), so
2i +
2+
=2
i + 1 +
i + 1 +
42
where we also used a long division. Using the tables it then follows that
h1 (t) = 2(t) (2 + )(t)et(1+) .
c Note that the spectrum of (h h1 )(t) is the function H() (1/H()),
which is 1. Since (t) 1, it follows that (h h1 )(t) = (t).
10.16
i cx i cx
X(x)
=
1
if
c
=
0
and
e
,
e
if
c > 0. Similarly Y (y) = 1 if c =0 and
where s R.
1 | t |
e
2
Since
10.18
t1
t
so h(t) = ((t) (t 1))e . Applying table 3, no. 7 and a shift in the
time domain gives H() = (1 e(1+i) )/(1 + i).
b The impulse response is causal, so the system is causal.
c It is straightforward to verify that the impulse response is absolutely
integrable, so the system is stable.
d The response y(t) to the block function p2 (t) is equal to the convolution
of h(t) with p2 (t), which equals
Z 1
Z t+1
y(t) =
h(t ) d =
h( ) d.
1
t1
This is 0 for t < 1 or for t > 2. For 1 < t < 0 it equals 1 e(t+1) , for
0 t < 1 it equals 1 e1 , and for 1 t < 2 it equals e(t1) e1 .
10.19
43
b The frequency response is H() = i/(1 + i). (Apply e.g. the differentiation rule to h(t) = a0 (t).)
c We have Y () = H()U () (using obvious notations), so Y () =
i/(1 + i)2 = 1/(1 + i) 1/(1 + i)2 . The response is the inverse
Fourier transform of Y (): y(t) = (1 t)et (t).
10.21
a The frequency response H() is the triangle function qc (). The inverse Fourier transform follows from table 3: h(t) = 2 sin2 (c t/2)/(c t2 ).
b Since a0 (t) = h(t) and h(t) 0, the function a(t) is a monotone increasing function.
10.22
1 + i
.
2 + 2i + 2
1
1
+
.
2(i + 1 i)
2(i + 1 + i)
The inverse Fourier transform is then h(t) = (et cos t)(t). Integrating
this over (, t] gives the step response a(t) = 12 (1 + et (sin t cos t))(t).
b We have Y () = H()U () (using obvious notations), so Y () =
1/( 2 + 2i + 2). Applying partial fraction expansion we obtain
Y () =
10.23
1
1
.
2i(i + 1 i)
2i(i + 1 + i)
10.24
y/2
where s R.
44
1
Since te (t) 1/(1+i)2 this means that F (s) = 2(1+is)
2 . The (formal)
solution is thus given by
Z
2
1
1
u(x, y) =
es y/2 eisx ds
2 (1 + is)2
Z
(1 s2 ) cos sx + 2s sin sx s2 y/2
1
=
e
ds.
2
(1 + s2 )2
11.1
11.2
3y 2x 6 + i(2y + 3x + 9)
(3i 2)(x + 3 iy)
=
.
(x + 3 + iy)(x + 3 iy)
(x + 3)2 + y 2
11.3
Apply definition 11.3 and expand the squares; several of the exponentials
cancel and only 1/2+1/2 remains, so sin2 z +cos2 z = 1. Similarly it follows
by substitution that 2 sin z cos z = sin 2z.
11.5
11.6
The proofs can be copied from the real case; this is a straightforward matter. The same applies to exercise 11.7.
11.8
This rational function is continuous for all z C for which the denominator
is unequal to 0. But the denominator is 0 for z = 1, z = i or z = 2i. So
g(z) is continuous on G = C \ {1, i, 2i}.
11.9
The proof can be copied from the real case; this is a straightforward matter:
lim
wz
11.11
f (w) f (z)
w2 z 2
= lim
= lim (w + z) = 2z.
wz w z
wz
wz
Using definition 11.3 and the chain rule it follows that (cos z)0 = (ieiz
ieiz )/2 = (eiz + eiz )/2i = sin z.
45
46
a This is not true; take e.g. z = 2i, then cos 2i = (e2 + e2 )/2 > 3.
b Use definition 11.3; expand the exponentials in the resulting expression
for cos z cos w sin z sin w.
c Write z = x + iy and use part b, then cos(x + iy) = cos x cos iy
sin x sin iy. But cos iy = cosh y and sin iy = i sinh y (see exercise 11.5a), so
cos(x + iy) = cos x cosh y i sin x sinh y, which gives the real and imaginary
parts.
d Since ez is analytic on C, it follows from theorem 11.5 that cos z =
(eiz + eiz )/2 is also analytic on C.
11.17
c This function is analytic when z 2 6= 3, so on C \ {i 3}. The derivative is given by ez (z 2 2z + 3)/(z 2 + 3)2 (use the quotient rule).
d This function is analytic on C since both ez and sin w are analytic on
C. The derivative is given by ez cos ez (use the chain rule).
12.1
a The complex number eiR lies on the unit circle for all R and so the
limit R does not exist.
12.2
The integrand equals e(as)t and a primitive of this is given by e(as)t /(a
s). The lower limit 0 leads to 1/(s a), while the upper limit results in the
limit limR e(as)R (note that 1/(as) does not influence the outcome of
this limit). Write s = + i, a = + i and use that limR e()R = 0
only if < 0, so if > . Hence, the Laplace transform equals 1/(sa)
for Re s > Re a.
12.3
12.4
One can use the method of example 12.9 and apply an integration by parts
for s 6= 0. It then follows that
Z
1
2 st
F (s) = lim R2 esR +
te
dt.
s R
s 0
The remaining integral is the Laplace transform of t for Re s > 0, which
is 1/s2 (example 12.9). Since limR R2 eR = 0 for > 0, it follows as
before that limR R2 esR = 0 for Re s > 0. This shows that F (s) = s23 .
12.5
a From examples 12.2 and 12.8 it follows that 1/(s + 2) is the Laplace
transform and that a = c = 2.
b From example 12.7 it follows that e4s /s is the Laplace transform and
that a = c = 0.
c From exercise 12.2 it follows that 1/(s 2 3i) is the Laplace transform
and that a = c = 2.
12.6
a
b
c
d
12.8
1
1
1
+
2 si
s+i
From
From
From
From
+ 12 eit ), we
1
1
s
1
+
= 2
.
2 sa
s+a
s a2
12.9
Follow the hint, e.g. write cos(at + b) = cos at cos b sin at sin b, then apply
47
48
linearity and the fact that we know the transforms of cos at and sin at. In
table 7, lines 8 and 9 the answers are given.
12.10
In
a
b
c
d
e
f
g
h
12.11
12.12
The function equals cos t for 0 t < 2 and is 0 elsewhere since cos(t
2) = cos t. The Laplace transform of cos t is s/(s2 + 1). A shift in
the time-domain shows that se2s /(s2 + 1) is the Laplace transform of
(t 2) cos(t 2). Adding these results leads to s(1 e2s )/(s2 + 1).
12.15
12.16
For all these exercises one first needs to recognize the basic form of the
function, then apply table 7, in combination with a shift in the time- or
s-domain.
a 1/(s 2)2 ; use a shift in the s-domain.
b 2es /s3 ; use a shift in the time-domain.
c 5/((s + 3)2 + 25); use a shift in the s-domain.
d (s b)/((s b)2 + a2 ); use a shift in the s-domain.
e se3s /(s2 1); use a shift in the time-domain.
f 2e3 /(s 1)3 ; use a shift in the s-domain.
12.17
12.18
Write f (t) = (t)t (t 1)t, then it follows from exercise 12.17c and table
7 that the Laplace transform is given by (1 (1 + s)es )/s2 .
12.19
For this exercise one again first has to recognize the basic form of the
Laplace transform, e.g. using table 7, and then, if necessary, combine it
with properties like a shift in the time- or s-domain.
a 2e3t
b 3 sin t
c 4 cos 2t
d (sinh 2t)/2
e (t 2)(t 2); use a shift in the time-domain.
f (t 3) cos(t 3); use a shift in the time-domain.
49
s+1
1
,
(s + 1)2 + 1
(s + 1)2 + 1
Apply De lH
opitals rule repeatedly (n times) to the limit t of tn /et
and use that limt et = 0 for any > 0. Theorem 12.3 implies that
the Laplace transform of tn exists for Re s > 0 (it is easy to show that
tn = (t)tn is of exponential order for > 0 arbitrary).
12.23
12.25
a We know (e.g. from table 7) that (Ltn )(s) = n!/sn+1 and using the
shift rule in the s-domain we then obtain that F (s) = n!/(s a)n+1 .
b We know (e.g. from table 7) that (Leat )(s) = 1/(s a) and using the
differentiation rule in the s-domain we then obtain that
F (s) = (1)n
dn 1
n!
=
.
dsn s a
(s a)n+1
12.27
The causal function sinh at is continuous on R, so it follows from the integration rule (table 8) that
Z t
1
L
sinh a d (s) = F (s)
s
0
Rt
with F (s) = (L sinh at)(s) = a/(s2 a2 ). Hence f (t) = 0 sinh a d =
(cosh at 1)/a.
12.28
a Use the differentiation rule in the s-domain and table 7 for the Laplace
transform of cos at. Then f has Laplace transform
2s(s2 3a2 )
d2
s
=
.
ds2 s2 + a2
(s2 + a2 )3
b Using the differentiation rule in the s-domain and table 7 for the Laplace
transform of sinh 3t one obtains that
(Lt sinh 3t)(s) =
3
6s
d
= 2
ds s2 9
(s 9)2
and
(Lt2 sinh 3t)(s) =
18(s2 + 3)
3
d2
.
=
ds2 s2 9
(s2 9)3
50
18(s2 + 3)
18s
6
2
+ 2
.
(s2 9)3
(s 9)2
s 9
a The function equals 2(t)t for t < 1 and since t = 2t t it follows that
f (t) = 2(t)t (t 1)t = 2(t)t (t 1)(t 1) (t 1).
b From part a, table 7 and the shift rule in the time domain it follows
that
2
es
es
.
2
2
s
s
s
c The function is not differentiable at t = 0 and t = 1. Apart from these
two points the derivative equals 0 for t < 0, 2 for 0 < t < 1 and 1 for
t > 1. Hence, f 0 (t) = 2(t) (t 1) for t 6= 0, 1. Since the Laplace
transform does not depend on the value at these points, we have that
(Lf 0 )(s) = (2/s) (es /s).
d According to the differentiation rule in the time domain one should have
(Lf 0 )(s) = s(Lf )(s), so
F (s) =
2 es (s + 1)
2 es
=s
.
s
s2
This is not correct since we cannot apply the differentiation rule in the
present situation: f has a jump at t = 1 and so it isnt differentiable on R.
12.31
a From the shift rule in the s-domain it follows that (Leibt f (t))(s) = F (s
ib) and since sin at = (eiat eiat )/2i it follows that (Lf (t) sin at)(s) =
(F (s ia) F (s + ia))/2i.
b First apply the scaling property and then the shift rule in the s-domain,
then (Le2t f (3t))(s) = e(s+2)/3 /(s + 2).
c We can apply the integration
here (t3 f (t) is continuous on R).
R t rule
3
Hence the Laplace transform of 0 f ( ) d is given by G(s)/s with G(s) =
(Lt3 f (t))(s). Now apply the differentiation
rule in the s-domain, then it
Rt
follows that the Laplace transform of 0 3 f ( ) d is given by
12.32
1 d3
F (s).
s ds3
a Since 3et2 = 3e2 et it follows that (L3et2 )(s) = 3e2 /(s 1) (see
table 7). We also have that (L(t 2))(s) = e2s /s so (for Re s > 1)
F (s) =
3e2 s + e2s (s 1)
.
s(s 1)
s2 2s + 2
.
s3
c From (L(t 4))(s) = e4s /s and the shift property in the s-domain
(table 8) it follows that F (s) = e4(s2) /(s 2).
d Note that f (t) = e2it et = e(1+2i)t and applying table 7 we thus obtain
that F (s) = 1/(s + 1 2i). (One can also use the Laplace transforms of
sin 2t and cos 2t and a shift in the s-domain.)
e The Laplace transform of sin t is 1/(s2 +1). Applying a shift in the timedomain we obtain that (L(t 2) sin(t 2))(s) = e2s /(s2 + 1). Finally we
note that et+3 = e3 et and so we apply a shift in the s-domain:
F (s) =
51
e3 e2(s1)
.
(s 1)2 + 1
s ln 3
.
(s ln 3)2 + 4
g One can write f (t) as the following combination of shifted unit step
functions: f (t) = (t) (t 1) + (t 2). From table 7 we then obtain
that
F (s) =
12.33
1 es + e2s e3s
.
s
s2
4
+
.
(s 2)2 + 16
(s 2)2 + 16
From table 7 and a shift in the s-domain we then obtain that f (t) =
3e2t cos 4t + e2t sin 4t.
e The denominator equals (s+4)2 and if we now use that s+3 = (s+4)1,
then it follows that
1
1
F (s) =
.
s+4
(s + 4)2
From table 7 and a shift in the s-domain we then obtain that f (t) =
e4t (1 t).
f Apply a shift in the time domain to (Lt2 e2t )(s) = 2/(s 2)3 (table 7),
then we obtain that f (t) = 21 (t 4)e2t8 (t 4)2 .
g Applying the integration rule to (L sin 3t)(s) = 3/(sR2 + 9) (the causal
t
function sin t is continuous on R) we obtain that (L 0 sin 3 d )(s) =
2
3/(s(s + 9)). But the integral equals (1 cos 3t)/3, so (L(1 cos 3t))(s) =
9/(s(s2 + 9)). From a shift in the time domain it then follows that f (t) =
(t 1)(1 cos 3(t 1))/9.
13.1
The integral defining the convolution can be calculated by using the formula
for the product of two cosines. The convolution then equals 12 t cos t +
1
sin t. Using the convolution theorem we obtain s2 /(s2 +1)2 as the Laplace
2
transform. On the other hand we obtain from the Laplace transforms of
cos t and sin t and the differentiation rule in the s-domain the Laplace
transform (s2 1)/(2(s2 + 1)2 ) + (1/(2(s2 + 1)), which agrees with the
result obtained from the convolution theorem.
13.2
13.4
13.5
This is not possible, since lims sn does not exist, contradicting theorem
theorem 13.2.
13.6
13.8
13.9
For the functions cos t and sinh t the value f () does not exist and so the
final value theorem cannot be applied.
13.11
a For a periodic function f () will in general not exist and so the final
value theorem cannot be applied.
b Theorem 13.5 implies that
52
sF (s) =
s
1 esT
53
f (t)est dt.
Taking the limit s 0 gives (note that the integral is over a bounded interRT
RT
val) lims0 0 f (t)est dt = 0 f (t) dt. From the definition of derivative
(definition 11.7) it follows that (ezT )0 (0) = lims0 (esT 1)/s and since
(ezT )0 = T ezT we thus obtain that
lim
s0
s
1
= ,
1 esT
T
RT
which shows that lims0 sF (s) = T1 0 f (t) dt.
c In example 13.4 we have sF (s) = 1/(1 + es ) and so lims0 sF (s) = 12 .
R2
The function has period 2 and so the integral is 12 0 f (t) dt = 12 , which
verifies the result of part b for f .
13.12
a For t < a we have (t) = (t); for t < 2a we then have (t) = (t)
2(t a); finally, for all t we have (t) = (t) 2(t a) + (t 2a). If (s)
is the Laplace transform of (t), then it follows from table 7 that
(s) =
1
2eas
e2as
+
.
s
s
s
Since f has period 2a we then obtain from theorem 13.5 (or table 8) that
F (s) =
1 2eas + e2as
.
s(1 e2as )
b Let (t) denote the restriction of f (t) to one period, then (t) = t(t)
(t 2)(t 2) 2(t 1). Since (Lt)(s) = 1/s2 , we obtain from a shift in
the time-domain that (L(t 2)(t 2))(s) = e2s /s2 . Also (L(t 1))(s) =
es /s and so
(s) =
e2s
2es
1
2
.
2
s
s
s
1
2es
1
1
= 2
.
s2
s(1 e2s )
s
s sinh s
Apply (13.7) (or definition 13.2) and the definition of distribution derivative, then it follows that
D
E
D
E
(L (n) (t a))(s) = (n) (t a), est = (1)n (t a), (est )(n) .
Since (est )(n) = (s)n est , we then obtain from the definition of the
shifted delta function that (L (n) (t a))(s) = sn eas .
13.18
This follows immediately from table 9, no 2 and (9.21) together with (13.9).
13.19
Consult tables 7 and 9 for this exercise and use linearity, and for part d
the convolution theorem.
a 1 + (1/(s2 + 1)) = (s2 + 2)/(s2 + 1),
b s + 3s2 ,
54
Consult tables 7 and 9 for this exercise, but now in the opposite direction,
and use mainly linearity.
a 0 (t) + 3(t) (t 2),
b 00 (t) 4 0 (t) + 4(t) + e2t (first write (s 2)2 as s2 4s + 4),
c (t 2) sin(t 2) + (3) (t 2) (apply a shift in the time-domain to the
Laplace transform of sin t),
d (t) sin t (first write s2 /(s2 + 1) as 1 1/(s2 + 1)).
13.22
.
(s2 1)2
2 ds s2 1
2 s2 1
We have (L sinh t)(s) = 1/(s2 1) and (L cosh t)(s) = s/(s2 1) and the
differentiation rule in the s-domain then implies that
(Lt cosh t)(s) =
d
s
.
ds s2 1
55
To the rational part we apply partial fraction expansion (again we first put
y = s2 ), which leads to (1/s2 ) + (1/(s2 1)). From table 7, a shift in the
time-domain and table 9 it then follows that f (t) = 00 (t 2) + (t 2) +
(t 2)(t 2 + sinh(t 2)).
13.24
F (s) =
1 s
2
.
2 s2 + 4 s2 + 4
From table 7 and the convolution theorem (or table 8) it then follows that
(g h)(t) = (cos 2v 21 sin 2v)(t).
Rt
b The definition of convolution gives f (t) = 12 0 cos 2 sin(2t2 ) d and
using the trigonometricRformula 2 cos a sin b = sin(a+b)sin(ab) this can
t
be written as f (t) = 14 0 (sin 2t sin(4 2t)) d . Caculating this integral
1
gives f (t) = 4 t sin 2t.
c Applying the differentiation rule in the time-domain to (L sin 2t)(s) =
2/(s2 + 4), we obtain that (Lt sin 2t)(s) = 4s/(s2 + 4)2 , so f (t) = 14 t sin 2t.
13.25
13.26
s+1
1
s
(s + 1)2 + 4
and from table 7 and a shift in the s-domain it then follows that f (t) =
1 et cos 2t. We see that f () = 1 = lims0 sF (s), which verifies the
final value theorem.
b Note that lims0 sF (s) = 0. We now determine the inverse Laplace
transform of F (s). Since F (s) is a function of y = s2 , we use partial
fraction expansion for y/(y 1)(y + 4), which gives
F (s) =
2 2
1 1
+
.
5 s2 1
5 s2 + 4
From table 7 we then obtain that f (t) = (sinh t + 2 sinh 2t)/5. Since f ()
does not exits, the final value theorem cannot be applied.
13.27
2es
2
2es
es
2 + 3 3 ,
s
s
s
s
56
2
1
2
2
+ 2
+
.
s
s
s1
(s 1)2
1
s2 (s2 + 4)
and
H(s) = es
s5 4s4 8s + 64
.
s2 (s2 + 4)
We first determine the inverse Laplace transform g(t) of G(s). Since G(s)
is a function of y = s2 , we apply a partial fraction expansion to 1/y(y + 4),
which gives G(s) = (1/4s2 ) (1/4(s2 + 4)) and so (by table 7) g(t) =
(2t sin 2t)/8. Next we determine the inverse Laplace transform h(t) of
H(s). Since the degree of the numerator is larger than the degree of the
denominator, we first perform a long division:
s5 4s4 8s + 64
4s3 16s2 + 8s 64
=s4
.
s2 (s2 + 4)
s2 (s2 + 4)
Partial fraction expansion of the rational function gives
4s3 16s2 + 8s 64
2
16
2s
= 2 + 2
.
s2 (s2 + 4)
s
s
s +4
Hence,
H(s) = ses 4es es
16
2s
2
2 + 2
s
s
s +4
From table 7 and the shift property in the time-domain it then follows that
h(t) = 0 (t ) 4(t ) (t ) (2 16(t ) + 2 cos 2(t )) and
then f (t) = g(t) + h(t).
14.1
Here e and e cancel each other and the integral that remains can be
calculated by an integration by parts. This gives tet sin tet and so
y(t) = u(t) + tet sin tet = tet .
Next we determine y(t) using the Laplace transform: Y (s) = H(s)U (s) and
since U (s) = 1/((s+1)2 +1) (table 7 and a shift in the s-domain), it follows
that Y (s) = 1/(s + 1)2 and so (inverse Laplace transform) y(t) = tet .
14.2
Again use the important formula Y (s) = H(s)U (s), so H(s) = Y (s)/U (s).
In this case U (s) = 1/s2 and Y (s) = 1/s2 s/(s2 + 4), so H(s) = 1
s3 /(s2 + 4) = 1 s + 4s/(s2 + 4) (divide s3 by s2 + 4). The inverse Laplace
transform gives h(t) = (t) 0 (t) + 4 cos 2t.
14.3
14.4
Taking Laplace transforms of the left- and right-hand side gives (note the
condition of initial rest) Y (s) = 48/((s2 + 4)(s2 + 16)). A partial fraction
expansion (in the variable y = s2 ) results in 4/(s2 + 4) 4/(s2 + 16) and
from table 7 it then follows that y(t) = 2 sin 2t sin 4t.
14.5
58
Y (s) = U (s)H(s) = 1/((s 1)(s 4)(s 2)) we use partial fraction expansion:
1
1
1
Y (s) =
+
.
3(s 1)
2(s 2)
6(s 4)
From table 7 we then obtain that y(t) = (e4t + 2et 3e2t )/6.
e Use time-invariance (and linearity): 3(t 1) has response 3h(t 1) =
(t 1)(e4t4 et1 ).
14.7
14.8
14.10
Let Y (s) be the Laplace transform of y(t), then we obtain from the differential equation (and tables 7 and 8) that (s2 Y (s)sy(0)y 0 (0))+Y (s) = 1/s2 .
Since y(0) = 0 and y 0 (0) = 1 it follows that s2 Y (s) 1 + Y (s) = 1/s2 , so
(s2 + 1)Y (s) = 1 + 1/s2 = (s2 + 1)/s2 and thus Y (s) = 1/s2 . The inverse
Laplace transform of Y (s) is y(t) = t.
14.12
Let Y (s) be the Laplace transform of y(t), then we obtain from the differential equation and the initial conditions y(0) = 3 and y 0 (0) = 1 (and
tables 7 and 8) that (s2 Y 3s 1) 4(sY 3) 5Y = 3/(s 1). Solving
for Y gives (s2 4s 5)Y (s) + 11 3s = 3/(s 1), hence,
Y (s) =
(s
3
3s 11
3s2 14s + 14
+ 2
=
.
4s 5)
s 4s 5
(s 1)(s2 4s 5)
1)(s2
31
3
19
+
.
12(s + 1)
8(s 1)
24(s 5)
From table 7 we obtain the inverse Laplace transform y(t) = (62et 9et +
19e5t )/24.
14.13
59
1 e2s
1
s
1
s
Y (s) = 2 2
+ 2
= (1 e2s )
.
+ 2
s (s + 1)
s +1
s2
s2 + 1
s +1
From tables 7 and 8 we obtain the inverse Laplace transform y(t) = t
sin t + cos t (t 2)(t 2 sin(t 2)).
14.14
2(s + 1)
2e2s
1
+
+
.
(s + 1)2 + 4
(s + 1)2 + 4
s(s2 + 2s + 5)
Using table 7 and a shift in the s-domain it is easy to get the inverse Laplace
transform of the first two terms since (Let cos 2t)(s) = (s+1)/((s+1)2 +4)
and (Let sin 2t)(s) = 2/((s + 1)2 + 4). For the third term we use partial
fraction expansion:
(s + 2)
1
1
=
.
s(s2 + 2s + 5)
5s
5(s2 + 2s + 5)
We write the second term as (s + 1)/5(s2 + 2s + 5) + 1/5(s2 + 2s + 5) and
taking everything together now we get
Y (s) =
9(s + 1)
2
1
1
+
+ e2s
.
5s
5((s + 1)2 + 4)
5((s + 1)2 + 4)
(s + 1)2 + 4
From our previous remarks and a shift in the time-domain it then follows
that y(t) = (2 + 18et cos 2t et sin 2t)/10 + (t 2)e(t2) sin 2(t 2).
14.16
Let X(s) and Y (s) be the Laplace transforms of x(t) and y(t). From table
7 we know that (L cos 2t)(s) = s/(s2 + 4) and (L sin 2t)(s) = 2/(s2 + 4).
Applying the Laplace transform to the system and substituting the initial
conditions x(0) = 1 and y(0) = 0 we obtain the algebraic system
sX + Y = 1 + 2s/(s2 + 4),
X + sY = 2/(s2 + 4).
Next we solve this system of two linear equations in the unknowns X =
X(s) and Y = Y (s). We can find Y (s) by multiplying the second equation
by s and adding it to the first equation; we then obtain
Y s2 Y =
2s
2s
1 2
= 1
s2 + 4
s +4
2s2
2
+s+ 2
s2 + 4
s +4
and so X(s) = 2/(s2 + 4) s/(s2 1). The inverse Laplace transform gives
the solution x(t) = sin 2t cosh t, y(t) = sinh t.
14.17
Let X(s) and Y (s) be the Laplace transforms of x(t) and y(t). Applying
the Laplace transform to the system and substituting the initial conditions
60
1
1
2
1
1
s
X =
= 2 2
,
s
s
s(s2 + 1)
s
s
s +1
2s
2s + 1
1
+ 2
+
.
(s 1)2 (s + 1)
(s + 1)(s 1)2
s1
2 4s
5s 2
1
2
.
(s + 1)(s 1)2
(s + 1)(s 1)2
s1
1
5
s
1
1
+
+
+
2
,
2(s + 1)
s1
2(s 1)2
2(s2 + 1)
s +1
3
7
5
s
5
+ 2
+
.
2(s + 1)
2(s 1)
2(s 1)2
s +1
2(s2 + 1)
The inverse Laplace transform gives the solution x(t) = (et +2et +5tet +
cos t 2 sin t)/2, y(t) = (3et 7et 5tet + 2 cos t + 5 sin t)/2.
14.20
Apply Laplace transform with respect to t to the partial differential equation and substitute the initial conditions u(x, 0) = 2 sin 2x and ut (x, 0) =
0, then it follows as in example 14.16 that
s2 U (x, s) (2 sin 2x)s = 4Uxx ,
where U (x, s) is the Laplace transform of u(x, t). Hence, we have obtained
the ordinary differential equation
U 00
s2
1
U = s sin 2x.
4
2
61
2s
sin 2x.
s2 + (4)2
2s
sin 2x.
s2 + (4)2
The inverse Laplace transform gives the solution u(x, t) = 2 cos 4t sin 2x.
14.22
As in exercise 4.20, for example, one obtains the ordinary differential equation
U 00 (s + 6)U = cos(x/2).
The general solution of the homogeneous equation is
s+6
U (x, s) = Aex
+ Bex
s+6
U (x, s) = Aex
s+6
+ Bex
s+6
4
cos(x/2).
4s + 25
1
cos(x/2).
s + 25/4
The inverse Laplace transform gives the solution u(x, t) = e25t/4 cos(x/2).
14.23
a The impulse response is the derivative (in distribution sense if necessary) of the step response. Since the step response has no jump at t = 0,
it follows that h(t) = a0 (t) = 2 sinh 2t + 2 sin t et . The transfer function
is the Laplace transform of h(t) and from table 7 we obtain that
H(s) =
2
1
4
+ 2
.
s2 4
s +1
s+1
4
2
1
+ 2 2
2
.
s2 (s2 4)
s (s + 1)
s (s + 1)
2
1
1
1
2
+
.
s2 4
s +1
s
s+1
1
2
sinh 2t 2 sin t + 1 et .
62
H(s) =
1
1
,
L s2 + 02
1
1
.
L (s + a)(s2 + 02 )
1
1
1
s
a
1
2
.
+ 2
a2 + 02 s + a
a + 02 s2 + 02
a + 02 s2 + 02
1
a
at
q(t) =
e
+
sin
cos
t
.
0
0
L(a2 + 02 )
0
d Note that 0 6= a. Since V (s) = s/(s2 + a2 ) it follows as in part c that
s
1
s
s
1
=
Q(s) =
,
L (s2 + a2 )(s2 + 02 )
L(02 a2 ) s2 + a2
s2 + 02
where we also applied a partial fraction expansion. The inverse Laplace
transform gives
q(t) =
e
1
(cos at cos 0 t) .
L(02 a2 )
Q(s) =
1
2s
.
L (s2 + 02 )2
d
0
2s0
= 2
.
ds s2 + 02
(s + 02 )2
a One can write u(t) = cos t + (t ) cos(t ). Taking the Laplace
transform gives U (s) = (Lu)(s) = (1 + es )s/(s2 + 1). Applying the
Laplace transform to the differential equation and substituting the initial
conditions gives (s2 +s2)Y s2 = U (s). Since s2 +s2 = (s1)(s+2)
we thus obtain, after a little simplifying, that
Y (s) =
s
1
+ 1 + es
.
s1
(s + 2)(s 1)(s2 + 1)
For the second term we use a partial fraction expansion, which eventually
leads to
`
1
Y (s) =
+ 1 + es
s1
2
1
3s
1
+
.
15(s + 2)
6(s 1)
10(s2 + 1)
10(s2 + 1)
63
From table 7 and a shift in the time domain it follows by the inverse
Laplace transform that y(t) = et + g(t)/30 + (t )g(t )/30 with
g(t) = 4e2t + 5et 9 cos t + 3 sin t.
b We have (L(t2))(s) = e2s and (L 0 (t3))(s) = se3s (table 9). Applying the Laplace transform to the differential equation and substituting
the initial conditions gives as in part a:
Y (s) =
1
3e2s + 6se3s
+
.
s1
(s + 2)(s 1)
1
1
1
1
2
2s
3s
Y (s) =
+e
+ 2e
+
.
s1
s1
s+2
s1
s+2
From a shift in the time domain it follows by the inverse Laplace transform
that y(t) = et + (t 2)(et2 e2t+4 ) + 2(t 3)(et3 + 2e2t+6 ).
14.26
E
1
.
2L s(s + 4R/L)
After a partial fraction expansion we obtain from the inverse Laplace transform the solution
E
1 e4Rt/L .
i1 (t) = 2i2 (t) =
4R
b Since (L sin 2t)(s) = 2/(s2 + 4) we obtain as in part a that
I2 =
1
1
.
L (s2 + 4)(s + 4R/L)
After a partial fraction expansion we obtain from the inverse Laplace transform the solution
L
2R
4Rt/L
i1 (t) = i2 (t) =
e
cos
2t
+
sin
2t
.
8R2 + 2L2
L
14.27
1
s2
U = sin x.
4
2
64
2
sin x.
s2 + 4 2
15.1
15.2
15.3
15.4
Since [k] = 0 for k < 0 and [k] = 1 for k 0 we can write the sum in the
given right-hand side as
[n k] =
k=0
[k][n k].
k=
From theorem 15.1 it then follows that the sum equals [n].
15.7
15.9
n=
We have to determine the Nyquist frequency of the convolution. By definition of convolution we have
Z
Z
Z +
F () =
p ( u)p2 (u) du =
p ( u) du =
p (u) du.
F ( ks )ps ().
k=
Hence,
Fr () = ei0
( 0 ks )ps ()
k=
+ ei0
( + 0 ks )ps ().
k=
65
66
( 0 ks )ps (0 + ks )
k=
+ ei0
( + 0 ks )ps (ks 0 ).
k=
15.13
2 sin(3t/8)
sin(t/8)
.
t
t
h(t) =
2 sin2 (t/2)
.
2 t2
1
1 X sin((t nT ))
.
n=1
t nT
To determine y(t) it suffices to know the response to the signal (sin t)/t
since the system is linear and time-invariant. The spectrum of (sin t)/(t)
is p2 () and the spectrum of the corresponding output is then given by
p2 ()q () = q (). The response to (sin t)/(t) is thus the inverse
Fourier transform of q (), which is 2 sin2 (t/2)/( 2 t2 ) (table 3). Hence,
2 sin2 ((t 1)/2)
2 sin2 ((t + 1)/2)
2 sin2 (t/2)
+
+
.
2
2
2
2
t
(t 1)
2 (t + 1)2
P
a We know that f [n] = 3k=0 f [k]4 [n k] holds for all periodic discretetime signals with period 4. Hence, we only have to show that the sampling
has period 4, which is easy because
n
f [n] = f (
+ 2) = f ((n + 4) ) = f [n + 4].
2
2
y(t) =
15.15
67
ck ( k0 ).
k=
sin (t + 1)
sin (t 1)
t sin t
+
=
.
2(t + 1)
2(t 1)
(1 t2 )
X
1 sin (t nT )/T
f [n]
dt.
f (t) dt = T
t nT
n=
R
P
The integral in the right-hand side is , so f (t) dt = T
n= f [n].
d According to Parsevals identity we have
Z
Z
Z
1
1
1
E=
| f (t) |2 dt =
| F () |2 d =
cos2 d = .
2
2
2
16.1
F [k] =
f [n]e2ink/2 .
n=0
i
Since e
F [k] =
1
X
= 1 we have
f [n](1)nk = f [0] + (1)k f [1].
n=0
16.2
The signal f [n] = (1)n has period 2 and, hence, period 4 as well. The
2-point DFT F2 [k] is given by 1 (1)k (see 16.1). The 4-point DFT F4 [k]
is by definition given by
F4 [k] =
3
X
n=0
N
1
X
n=0
f [n] =
N
1
X
f (nT /N ).
n=0
It is better to define the function value at the jumps as the average value
of the left-hand and right-hand limit. Using the DFT we then find a better
approximation of the Fourier coefficients. (In exercise 16.4 we do have
c0 = F [0]/N for all N .)
16.7
16.8
Since F [k] = | F [k] | ei arg(F [k]) we obtain from the given amplitude spectrum and phase spectrum that F [k] = 2eik/2 and so F [0] = 2, F [1] = 2i,
F [2] = 2, F [3] = 2i. Next apply the inverse DFT:
68
69
and hence, f [0] = 0, f [1] = 0, f [2] = 0, f [3] = 2, that is, f [n] = 24 [n 3].
16.10
|t |
f (t) = 1
=
2 t for t 2.
The signal is real, so F [k] = F [k], which gives F [3] = F [1] = F [1] = i.
Applying the inverse DFT leads to f [n] = (1 + in+1 + (i)n+1 )/4.
16.13
Apply the definition of the cyclical convolution and use theorem 15.2, then
one obtains that (f f )[n] = f [n]+f [n1] = N [n]+2N [n1]+N [n2].
16.14
Use the convolution theorem: first determine the functions f1 and f2 with
f1 cos(2k/N ) = (e2ik/N + e2ik/N )/2 and f2 sin(4k/N ) =
(e4ik/N e4ik/N )/2i and then calculate f [n] = (f1 f2 )[n]. Since
N [n m] e2imk/N (table 11 and table 12, shift in the n-domain),
we have f1 [n] = (N [n 1] + N [n + 1])/2 and f2 [n] = (N [n + 2]
N [n 2])/2i. From the convolution theorem and theorem 15.2 it then
follows that f [n] = (f1 f2 )[n] = (f2 [n + 1] + f2 [n 1])/2, which equals
(N [n + 3] + N [n + 1] N [n 1] N [n 3])/4i.
16.16
From table 11 we have that N [n] 1 and so (table 12, shift in the ndomain) N [n l] e2ilk/N . Since cos2 (k/N ) = (1 + cos(2k/N ))/2 =
(2 + eik/N + eik/N )/4 it follows that f [n] = (2N [n] + N [n 1] + N [n +
1])/4. The power equals
N 1
1 X
| f [n] |2 .
N n=0
1
1
3
1 1
+
+
=
.
P =
N 4
16
16
8N
16.18
a We calculate G[k] from the expression for the 5-point DFT. Note that
g[3] = g[2] = c2 = 1 and g[4] = g[1] = c1 = 2. Hence,
G[k] = g[0] + g[1]e2ik/5 + g[2]e4ik/5 + g[3]e6ik/5 + g[4]e8ik/5
= 1 + 2e2ik/5 + e4ik/5 + e6ik/5 + 2e8ik/5
= 1 + 4 cos(2k/5) + 2 cos(4k/5).
b Since the Fourier coefficients of f are known, we can express f as a
Fourier series: f (t) = c0 + c1 ei0 t + c1 ei0 t + c2 e2i0 t + c2 e2i0 t .
Hence,
f (2m/50 )
70
X
1
| f (t) |2 dt =
| cn |2 .
T 0
n=
P2
P
2
2
Since ck = 0 for | k | 3 we obtain that
n=2 | g[n] | =
n= | cn | =
P4
2
n=0 | g[n] | . Applying Parseval for the DFT one obtains the result.
16.19
a From table 11 we have that N [n] 1 and so (table 12, shift in the
n-domain) f [n] e2ik/N 1 + e2ik/N = 2 cos(2k/N ) 1.
b Calculate the convolution using (16.14) and theorem 15.2, then
(f g)[n] =
N
1
X
l=0
c First write g[n] as complex exponentials and then determine the N point DFT using table 11 and the shift rule in the k-domain:
G[k] =
N
(N [k 2] + N [k + 2]).
2
16.20
a Since f (t) is real and even, the sampling f [n] is real and even since
f [n] = f (nT /5) = f (nT /5) = f [n].
b Apply the inverse DFT, where the values of F [3] and F [4] are calculated using the fact that F has period 5 and is even. Hence, f [n] = (1 +
2e2in/5 + 2e8in/5 + e4in/5 + e6in/5 )/5, which equals (1 + 4 cos(2n/5) +
2 cos(4n/5))/5.
c The function f is band-limited with band-width 10/T . The Fourier
coefficients ck of f contribute to the frequencies k0 = 2k/T . Hence,
these are 0 for | k | 3. This means that f (t) is equal to the Fourier series
c0 + c1 ei0 t + c1 ei0 t + c2 e2i0 t + c2 e2i0 t . Substituting t = nT /5
(and rearranging) we obtain that f [n] = c0 + c1 e2in/5 + c2 e4in/5 +
c2 e6in/5 + c1 e8in/5 . But this is precisely the expression for the inverse DFT, which implies that c0 = F [0]/5, c1 = F [1]/5, c2 = F [2]/5,
c2 = F [3]/5 = F [2]/5, c1 = F [4]/5 = F [1]/5. Hence ck = F [k]/5 for
| k | 2. Since F [k] has period 5 and ck = 0 for | k | > 2 we do not have
ck = F [k]/5 for | k | > 2.
17.1
17.2
As in exercise 17.1 we can write the formula for the inverse DFT in matrix
form:
01 1
1
1
1 1 0 F [0] 1 0 f [0] 1
2
3
w
w4 C B F [1] C B f [1] C
B1 w w
1B
CB
C B
C
B 1 w2 w4 w w3 C B F [2] C = B f [2] C .
5@
A @
A
3
4
2 A@
F [3]
f [3]
1 w
w w
w
4
3
2
1 w
w
w
w
F [4]
f [4]
The matrix in the left-hand side is thus the inverse of the matrix in exercise
17.1.
17.4
f [0] f [2]
2 2
Mf =
=
.
f [1] f [3]
0 1
The 2-point DFT of the rows of this matrix gives
4 0
C=
.
1 1
Multiplying this by the twiddle factors w4 with w4 = ei/2 = i gives
4 0
Ct =
.
1 i
Now calculate the 2-point DFT of the columns of this matrix to get the
4-point DFT:
5 i
MF =
.
3 i
Hence, F [0] = 5, F [1] = i, F [2] = 3, F [3] = i.
17.6
f [0] f [2] . . . f [N 2]
Mf =
.
1
1
...
1
The N/2-point DFT of the first row of this matrix is A[k], while the N/2point DFT of the second row follows from table 11. This gives the matrix
C:
71
72
C=
A[0]
N/2
A[1]
0
...
...
A[N/2 1]
0
Multiplying this by the twiddle factors will not change this matrix because
of the zeroes in the second row of C and hence Ct = C. Now calculate the
2-point DFT of the columns of Ct = C to get the matrix MF and, hence,
the required N -point DFT of f [n]:
F [0]
F [1]
. . . F [N/2 1]
=
.
F [N/2] F [N/2 + 1] . . . F [N 1]
17.7
Let A1 [k] be the 2N -point DFT of f [2n] and B1 [k] the 2N -point DFT of
f [2n + 1]. Then we have for the 4N -point DFT of f [n] (see (17.14)):
F [] = A1 [] + w4N
B1 [],
F [2N + ] = A1 [] w4N
B1 [],
A1 [] = A[] + w2N
C[],
C[],
A1 [N + ] = A[] w2N
D[],
B1 [] = B[] + w2N
D[],
B1 [N + ] = B[] w2N
D[],
B[] + w4N
C[] + w4N
F [] = A[] + w2N
3
D[],
B[] w4N
C[] + w4N
F [N + ] = A[] w2N
3
F [2N + ] = A[] + w2N
C[] w4N
B[] w4N
D[],
3
D[],
B[] + w4N
F [3N + ] = A[] w2N
C[] w4N
where = 0, 1, . . . , N 1.
17.8
RT
Since f (t) is causal we have FT () = 0 f (t)eit dt. Apply the trapezium
rule to the integral and substitute = (2k + 1)/T , then it follows that
FT ((2k + 1)/T )
N 1
T X in/N
e
f [n]e2in/N .
N n=0
This shows that the spectrum at the frequencies = (2k + 1)/T can be
approximated by the N -point DFT of ein/N f [n].
17.9
T
2k
) (F [k] + F [k])
T
N
73
First we determine the DFTs of the signals using table 11. Since f1 [n] =
N [n] + N [n 1] F1 [k] = 1 + e2ik/N and f2 [n] = N [n] + N [n + 1]
F2 [k] = 1 + e2ik/N we obtain the DFT of the cross-correlation as follows:
12 F1 [k]F2 [k] = (1 + e2ik/N )2 = 1 + 2e2ik/N + e4ik/N .
17.13
17.14
Since the 3-point DFT of g[n] is given by G[k] = g[0] + g[1]w3k + g[2]w32k
we conclude that
(+N )
F [N + ] = A[] + w3N
17.15
2(+N )
B[] + w3N
C[].
m1
Take N1 = 3 and N2 = 3
and consider the N1 N2 -matrix
0
1
f [0] f [3] . . . f [N 3]
Mf = @ f [1] f [4] . . . f [N 2] A .
f [2] f [5] . . . f [N 1]
Let the N2 -point DFT of the rows f [3n], f [3n + 1] and f [3n + 2] be given
by A[k], B[k] and C[k], then the matrix C is given by
0
1
A[0] A[1] . . . A[N2 1]
C = @ B[0] B[1] . . . B[N2 1] A .
C[0] C[1] . . . C[N2 1]
Multiplying this by the twiddle factors and then applying the 3-point DFT
of the columns gives the matrix MF containing the N -point DFT of f [n]:
0
1
F [0]
F [1]
. . . F [N2 1]
MF = @ F [N2 ]
F [N2 + 1] . . . F [2N2 1] A .
F [2N2 ] F [2N2 + 1] . . . F [N 1]
74
17.16
2N
X
for n = 0, 1, . . . , 2N .
l=0
2N
X
1
Fp [k]Gp [k]e2ink/(2N +1)
2N + 1
k=0
18.1
F (z) =
f [n]z n =
n=
1
X
f [n]z n +
n=N
N
X
f [n]z n .
n=0
The anti-causal part converges for all z, while the causal part converges
for all z 6= 0. The region of convergence is thus given by 0 < | z | < . If
f [n] = 0 for n 1 then the z-transform converges for all z C.
18.2
The anti-causal part converges for all z. The causal part can be written as
n n
X
1
1
+
.
2z
3z
n=0
Hence, the z-transform converges for | 2z | > 1 and | 3z | > 1, that is, for
| z | > 1/2.
18.3
X
z n
n=0
z n
3
This series converges for | z/2 | < 1 and | z/3 | < 1, hence, for | z | < 2.
b The z-transform is given by
F (z) =
cos(n/2)z n .
n=0
P
n
Since | cos(n/2) | 1 for all n and since
converges
n=0 | z |
P for | z | > 1,
the z-transform also converges for | z | > 1. Moreover,
n=0 cos(n/2)
diverges since limn cos(n/2) 6= 0. We conclude that the z-transform
converges for | z | > 1.
c From parts a and b it follows immediately that the region of convergence
is the ring 1 < | z | < 2.
18.4
4z 1
4z 2
2z 1
+
=
(z 1)2
z1
z(z 1)2
for | z | > 1.
75
76
F (z) =
c
2z 3
4z 2
2z 3 4z 2
=
.
2
(1 z)
1z
(1 z)2
(1)n z n =
n=0
1
1+z
for | z | < 1.
zn =
n=4
z 4
1z
for | z | < 1.
One can also use [n] 1/(1 z) and apply a shift rule.
e From example 18.6 it follows that (n2 n)[n] 2z/(z 1)3 and
n[n] z/(z 1)2 . Adding these results gives
n2 [n]
z(z + 1)
(z 1)3
for | z | > 1.
From example 18.2 we obtain that 4n [n] z/(z 4) for | z | > 4. The
differentiation rule implies that n4n [n] 4z/(z4)2 for | z | > 4. Together
these results give
F (z) =
18.5
z(z + 1)
4z
+
(z 1)3
(z 4)2
for | z | > 4.
F (z) =
cos(n/2)z n = 1 z 2 + z 4 =
n=0
z2
1
=
.
2
1+z
1 + z2
sin(n/2)z n = z 1 z 3 + =
n=0
z 1
z
=
.
1 + z 2
1 + z2
A direct calculation of the z-transform of ein [n] gives (for | z | > 1):
ein z n = 1 + ei z 1 + e2i z 2 + =
n=0
1
z
=
.
1 ei /z
z ei
A direct calculation of the z-transform of 2n ein [n] gives (for | z | < 2):
0
X
2n ein z n =
n=
2n ein z n =
n=0
1
.
1 ei z/2
z
1
+
.
z ei
1 ei z/2
One could apply a partial fraction expansion here (see e.g. exercise 18.10).
However, in this case it is easy to obtain the z-transform in a direct way
by developing F (z) in a series expansion. Since the z-transform has to
converge for | z | > 2 (there are poles at z = 2i and f [n] has a finite
switch-on time) we develop F (z) as follows:
1
1
4
16
F (z) = 2
= 2 1 2 + 4 + .
z +4
z
z
z
77
From the series we now obtain that f [n] = 0 for n 0 and that f [2n] =
(1)n1 22n2 , f [2n + 1] = 0, for n > 0.
18.9
1
1
z4
z2
F (z) =
=
+
+
.
1
4(1 + z 2 /4)
4
4
16
From the series we obtain that f [n] = 0 for n 1 and f [2n] = (1)n 22n2 ,
f [2n 1] = 0, for n 0.
18.10
The poles are at z = 1/2 and z = 3; the signal f [n] must have a finite
switch-on time, hence, the z-transform has to converge for | z | > 3. A
partial fraction expansion of F (z)/z gives
F (z)
1/10
18/5
=1+
.
z
z + 1/2
z+3
Applying (18.14) to the expansion of F (z) gives
(1/2)n [n]
z
z
for | z | > 1/2, (3)n [n]
for | z | > 3.
z + 1/2
z+3
z
for | z | < 3.
z+3
M1
X
f [l]g[n l]
l=N1
78
F (z) =
z
1
=
z2 + 1
z
1
1
1 2 + 4 + .
z
z
From the series we obtain that f [n] = 0 for n < 0 and f [2n + 1] = (1)n ,
f [2n] = 0, for n 0.
b The convolution theorem gives (f f )[n] F 2 (z). Hence,
!
n
X
X
h[n] = (f f )[n] =
f [l]f [n l] =
f [l]f [n l] [n].
l=
l=0
From this we obtain that h[n] = 0 for n < 0 while for m 0 we have:
h[2m + 1] = f [0]f [2m + 1] + f [1]f [2m] + + f [2m + 1]f [0] = 0,
h[2m] = f [0]f [2m] + f [1]f [2m 1] + + f [2m]f [0]
= 0 + (1)0 (1)m1 + 0 + + (1)m1 (1)0 + 0 = m(1)m1 .
18.14
Apply the definition of the convolution product and use that [n l] = 0
for l > n and [n l] = 1 for l n.
18.15
Use that f [n] F (z) and [n] 1 and apply the convolution theorem
(assuming that the intersection of the regions of convergence is non-empty)
then (f )[n] F (z) 1 = F (z). Hence, f [n] = (f )[n].
P
ln
Define a discrete-time signal h[n] by h[n] = n
f [l], which is the
l= 2
n
n
convolution product of f [n] with 2 [n]. Now 2 [n] z/(z 1/2) for
| z | > 1/2 and f [n] F (z). Hence, h[n] zF (z)/(z 1/2). Assuming
that | z | = 1 belongs to the region of convergence of the z-transform of f [n]
we get
18.17
X
ei F (ei )
=
h[n]ein .
ei 1/2
n=
2
X
1
1
1
i
| f [n] |2 =
F
(e
)
d
=
cos2 d = .
2
2
2
n=
(One can also determine f [n] first and then calculate
ectly.)
n=
| f [n] |2 dir-
18.20
We have that [n] = (g f )[n] with g[l] = f [l]. The convolution theorem
and property (18.25) (or table 15, entry 2) imply that the spectrum of [n]
is given by G(ei )F (ei ). But G(ei ) = F (ei ) (combine table 15, entries
2
2 and 5) and hence [n] F (ei )F (ei ) = F (ei ) .
18.22
79
a The signal f [n] is absolutely summable. The z-transform has one pole
at z = 2 and therefore the region of convergence is | z | < 2, since it
must contain | z | = 1. Since F (z)/z = 1 2/(z + 2) we have F (z) =
z 2z/(z + 2) for | z | < 2. The inverse transform of this gives f [n] =
[n + 1] (2)n+1 [n 1].
b The Fourier transform of f [n] equals F (ei ) = e2i /(ei + 2).
c We have f [n] F (z) for | z | < 2. The scaling property (table 14, entry
5) gives 2n f [n] F (z/2) for | z | < 4. The spectrum of 2n f [n] is thus equal
to F (ei /2) = e2i /(2ei + 8).
18.24
a The signal f [n] is causal. The poles of F (z) are at i/2 and at 0. The
region of convergence is the exterior of the circle | z | = 1/2. This contains
the unit circle and so the signal is absolutely summable. The spectrum of
f [n] equals F (ei ) = 1/(ei (4e2i + 1)).
b First apply a partial fraction expansion to F (z)/z (the denominator
equals z 2 (2z + i)(2z i)):
F (z)
1
i
i
= 2 +
.
z
z
z i/2
z + i/2
Applying (18.14) to F (z) gives f [n] = [n 1] + (i(i/2)n i(i/2)n )[n].
c If F (ei ) = F (ei ), then the signal is real. Since
F (ei ) = 1/(ei (4e2i + 1)) = F (ei ),
18.25
the signal is indeed real. (One can also write (i(i/2)n i(i/2)n ) =
in+1 2n (1 (1)n ), which equals 0 for n = 2k and 21n (1)k+1 for
n = 2k + 1, showing clearly that it is real.)
P
For n < 0 we have that n
l=0 g[l]g[n l] = 0 since g[n] is causal. Thus f [n]
is also causal. Since g[l] = 0 for l < 0 and g[n l] = 0 for l > n we can
write
f [n] =
g[l]g[n l] = (g g)[n].
l=
19.2
h[n] =
n1
X
l=
n1
X
2ln u[l] =
l=
l=
h[n l]u[l].
l=
1 X
u[l] ([n l + 1] + [n l 1])
2
l=
l=
1
= (u[n + 1] + u[n 1]) = y[n].
2
P
ln
c We now have h[n] =
[l] = 2n [n] and
l=n 2
h[n l]u[l] =
X
l=
19.3
h[n l]u[l] =
u[l]2ln [l n] =
l=
u[l]2ln = y[n].
l=n
| h[n] | =
n=
n=
c
PThisn system is not causal, and it is not stable since
n=0 2 = .
n=
| h[n] | =
19.4
19.5
Since [n] = [n] [n 1], it follows from linearity and time-invariance
that h[n] = a[n] a[n 1]. Now use (19.3) to calculate the response y[n]
to u[n] = 4n [n]:
y[n] = (h u)[n] =
a[l]u[n l]
l=
X
l=
80
a[l]u[n l]
a[l 1]u[n l]
l=
X
l=
a[l]u[n 1 l].
81
We now calculate the first sum; the second one then follows by replacing n
by n 1.
a[l]u[n l] =
l=
X
2l [l] 3l [l 1] u[n l]
l=
X
l ln
2 4
l=0
n
X
[n l]
!
l=1
n
[n] 4
l=0
3l 4ln [n l]
n
X
(4/3)l [n 1]
l=1
The transfer function follows from an [n] z/(z a) for | z | > | a |; this
is because we can write cos n = (ein + ein )/2, and hence
z
z
1
+
h[n]
2 z ei /2
z ei /2
i
for | z | > e /2 = 1/2. We thus obtain:
H(z) =
for | z | > 1/2. The poles are at ei /2 and ei /2, which is inside the unit
circle. Therefore the system is stable.
19.9
a The impulse response h[n] can be determined by a partial fraction expansion of H(z)/z. Since
1/3
H(z)
1
1
=
z
9 z + 1/3
(z + 1/3)2
we have
H(z) =
1
9
z/3
z
z + 1/3
(z + 1/3)2
1
8 + 6i
Im
ein/2 = Im
(cos(n/2) + i sin(n/2)) ,
9 + 6i + 1
100
which is (3 cos(n/2) + 4 sin(n/2))/50.
19.11
a The frequency
be written as H(ei ) = 1 + e2i + e2i .
Presponse can
i
in
Since H(e ) = n= h[n]e
it follows that h[0] = 1, h[2] = h[2] = 1
and h[n] = 0 for all other n. Hence, the impulse response is h[n] = [n] +
[n 2] + [n + 2]. The input is u[n] = [n 2]. Since [n] 7 h[n], we have
82
From (19.15) and the fact that H(ei ) is even we obtain that
Z
Z
1
1 b
h[n] =
cos(n) d
H(ei ) cos(n) d =
2
a
1
=
(sin nb sin na ),
n
which can be written as 2(sin 21 n(b a ) cos 12 n(b + a ))/(n) for n 6= 0
and
Z
1 b
b a
h[0] =
d =
.
a
19.14
19.15
a Apply the z-transform to the difference equation, using the shift rule
in the n-domain. We then obtain that (1 + 21 z 1 )Y (z) = U (z). Since
H(z) = Y (z)/U (z) it follows that
H(z) =
z
1
=
.
1 + z 1 /2
z + 1/2
l=
P
l
Now if n < 0 then this is 0, while if n 0 then it equals (1/2)n n
l=0 (1) =
n+1
n
(1/2)
(1 + (1) ).
c The transfer function H(z) has one pole at z = 1/2, which is inside
the unit circle, so the system is stable.
19.17
z(z + 1)
1 + z 1
= 2
.
1 z 2 /4
z 1/4
83
A(z)
8/3
3/2
1/6
=
.
z
z1
z 1/2
z + 1/2
Multiply this by z and use table 13 to obtain a[n] = ((8/3) 3(1/2)n+1
(1/3)(1/2)n+1 )[n].
d Since u[n] = [n] + [n 2] and [n] 7 a[n] we get u[n] 7 a[n] + a[n 2].
19.18
.
z
z
z + 1/2
(z + 1/2)2
Multiply this by z and use table 13 to obtain that h[n] = [n] ((1/2)n
(1/2)n(1/2)n )[n].
b According to (19.9) we have that z n 7 H(z)z n . Substituting z = 1
we get the response to the input (1)n . Since H(1) = 0, the response is
the null-signal.
c The (rational) transfer function H(z) has one pole at z = 1/2, which
is inside the unit circle, so the system is stable.
d The impulse response is real, so the system is real.
e Let u[n] 7P
y[n], then Y (ei ) = H(ei )U (ei ). Furthermore we have
in
that U (ei ) =
. Comparing this with U (ei ) = cos 2 =
n= u[n]e
2i
2i
(e
+e
)/2 we may conclude that u[2] = u[2] = 1/2 and u[n] = 0
for n 6= 2, hence, u[n] = ([n 2] + [n + 2])/2. By superposition it then
follows from [n] 7 h[n] that y[n] = (h[n 2] + h[n + 2])/2.
19.19
N 1
1 X
U [k]H(e2ik/N )e2ink/N .
N
k=0
On the other hand we have from the inverse DFT for y[n] that
y[n] =
N 1
1 X
Y [k]e2ink/N ,
N
k=0
a Since H(ei ) = cos 2 = (e2i + e2i )/2 we see from definition (19.11)
of the frequency response that h[2] = h[2] = 1/2 and h[n] = 0 for n 6= 2.
Hence, h[n] = ([n + 2] + [n 2])/2.
b Using the inverse DFT we can recover u[n] from the DFT F [k] of u[n]:
u[n] =
3
1X
F [k]e2ink/4 ,
4
k=0
a Since [n] = [n] [n 1] we have for the responses that h[n] =
a[n] a[n 1] and hence h[n] = n2 (1/2)n [n] (n 1)2 (1/2)n1 [n
84
1]. For the z-transforms we have (use the shift rule in the n-domain)
H(z) = A(z) A(z)/z = A(z)(z 1)/z. From table 13` it follows that
n
2
n
(1/2)n [n] z/(z 1/2), n(1/2)n [n]
` (z/2)/(z 1/2) , 2 (1/2) [n]
(z/4)/(z 1/2)3 and since n2 = 2 n2 + n it then follows that
z/2
z/2
1 z(z + 1/2)
+
=
.
(z 1/2)3
(z 1/2)2
2 (z 1/2)3
A(z) =
1 (z 1)(z + 1/2)
.
2
(z 1/2)3
z+1
z 1 + z 2
= 2
.
1 z 1 /2
z z/2
z+1
.
(z 1/2)(z 1)
.
z
z
z1
z 1/2
Multiply this by z and use table 13 to obtain that a[n] = 2[n] + (4
6(1/2)n )[n].
b The (rational) transfer function H(z) has poles at z = 0 and z = 1/2,
which lie inside the unit circle, so the system is stable.
19.23
6(z 2 1)
6 6z 2
= 2
.
1
2
6 5z + z
6z 5z + 1
.
z
z
z 1/3
z 1/2
Multiply this by z and use table 13 to obtain that h[n] = 6[n] +
(16(1/3)n 9(1/2)n )[n].
b The frequency response H(ei ) is obtained from the transfer function
H(z) by substituting z = ei , so H(ei ) = 6(e2i 1)/(6e2i 5ei + 1).
c From (19.10) we know that Y (ei ) = H(ei )U (ei ). Since y[n] is
identically 0, we know that Y (ei ) = 0 for all frequencies . Since H(ei ) 6=
0 for e2i 6= 1 we have U (ei ) = 0 for e2i 6= 1. The frequencies = 0
or = may still occur in the input. These frequencies correspond to
85