ffiffi

Signal Transmission
and Filtering
2.T INTRODUCTION
Signal transmission is a process whereby a message (or information-bearing) signal is transmitted over
a communication channel. Signal filtering purposefully alters the spectral content of the signal so that a
better transmission and reception can be achieved. Many communication channels, as well as filters, can
be modeled as a linear time-invariant system. In this chapter we review the basics of a linear time-invariant
system in frequency domain.

2.2 IMPULSE RESPONSE AND FREQUENCY RESPONSE

A. Linear Time-invariant Systems
A system is a mathematical model of a physical process that relates the input signal (source or excitation
signal) to the output signal (response signal).
Let x(t) and y(t) be the input and output signals, respectively; of a system. Then the system is viewed
as a mapping of x(t) into y(r). Symbolically, this is expressed as

y(t) - cr-lx(t)) (2.r)

x(f) i-s*i..] y(tl
where e,7: is the operator that produces output y(r) from input j
x(r), as illustrated in Fig. 2.1.
If the system satisfies the following two conditions, then the Fig. 2.7 0perator representation
system is called a linear system. of a system

e,fi'lxr(t)*xr(t)l- c,T [x,(r)] t tF l*zf)l:ylt)+y2(t) (2.2)

for all input signals x,(t) and xz(t).

a-7* lax(t)l - aor'-lx(t)l: ay(t) (2.3)

for all input signals x(r) and scalar a.

[7,..t] Analog and Digital Communications

Condition (2.2) is the additivity property and condition (2.3) is the homogeneity property.
If the system satisfies the followirrg condition, then the system is called a time-invariant or fixed
system.
erv'lx(t - /o)l - y(t - ts) (2.4)

where /o is any real constant. Equation (2.4) indicates that the delayed input gives delayed output.
If the system is linear and time-invariant, then the system is called a linear time-invariqnt (LTI)
system.

B. Imputse Response
The impulse response h(t) of an LTI system is defined to be the response of the system when the input
is 5(r), that is,
h(t) cr'[6(t)) - (2.s)

The function h(t) is arbitrary, and it need not be zero for / < 0. If
h(t) :0 for r < 0 (2.6)
then the system is called causal.

C. Response to an Arbitrary Input

The response y(r) of an LTI system to an arbitrary input x(t) can be expressed as the convolution of x(r)
and the impulse response h(t) of the system, that is,

y(t) : x(t) * h(t) - f*.fl hQ - r) dr (2.7)

Since the convolution is commutative, we also can express the output as

y(r) - h(t) * x(t) - I]-rrrr-(r - r\ dr (2.8)

D. Frequency Response
Applying the time convolution theorem of the Fourier transform (1.28) to Eq. (2.7), we obtain
Y(w) : X(a)H(w) (2.e)
where X(r) - GT [x(r)], Y(w) : a7* [y(t)1, and H(r) - tv'lh(t)).
And l(u,,) is referred to asthefrequency response (or trans-
1 H(r)
fer function) of the system. Thus
Y(') I I
H(r) -- av- lh(t)1: x (a) (2.10) 6(f) h(0
x(r) Y(t\:x(0*h(0
The relationship represented by Eqs (2.5), (2.7), and (2.9) are
depicted inFig.2.2. I t
x(r) Y(''t\ : X(u)H(a)
By taking the inverse Fourier transform of Eq. (2.9), the
output becomes Fig. 2.2 Relationship between
inputs and outputs in
y(t) - IZTI!-N[* x(a)H(w)eiat da J
(2.1 t)
an LTI system
 Signal Tronsmission and Filteing

Thus, we see that either the impulse response h(t) or the Frequency response H(a) completely character-
izes the LTI system.

2.3 FILTER CHARACTERISTICS OF LTI SYSTEMS

The frequency response H(r) is a characteristic property of an LTI system. It is, in general, a complex
quantiry that is,
H(a;) (2.12)
- lH(u)lsio'{")
In the case of an LTI system with a real-valued impulse response h(t), H(a) exhibits conjugate symmetry
[Eq.(1.65)], that is
H(-r) : H*(a) (2.13)

which means that lH(-u)l: lH(u)l ?otr)

-?n(w:) :
(2.14)
That is, the amplitudelH(a)l is an even function of frequency, whereas the phase 0o@) is an odd function
of frequency. Let
Y(w)
- lY(a)leie'(*'t x@) - lX(a)leiqt")
Rewriting Eq. (2.9), we have
i0 l') - i e,@ i 0 n@t
lx(a)llH (a)ls
i (0,(a) + 0 n@)) (2.ts)
ly(w)le W(r1
"
1H 1 w)ls -
Thus, we have
lr(r)l - lx(w)llH(a)l (2.r6)

0"(r):0,(a)*0p@) (2.17)

Note that the amplitude spectrum of the output signal is given by the product of the amplitude spec-
trum of the input signal and the amplitude of the frequency response. The phase spectrum of the output is
given by the sum of the phase spectrum of the input and the phase of the frequency response. Therefore,
and LTI system acts as a filter on the input signal. Here the wordy'lter is used to denote a system that
exhibits some sort of frequency-selective behavior.

2.4 TRANSMISSION OF SIGNALS THROUGH LTI SYSTEMS

A. Distortion less Transmission
For distortionless transmission through a system, we require that the exact input signal shape be repro-
duced at the output. Therefore, if x(t) is the input signal, the required output is
y(t)-Kx(t-tr) (2.18)
where r, is the time delay and K is a gain constant. This is illustrated in Fig. 2.3(a) and (b). Taking the
Fourier transform of both sides of Eq. (2.18), we get
Y(w)
- Ke't""Y(a) (2.1e)
From Eq. (2.9), we see that for distortionless transmission the system must have
H(a) - lH(w)leion@) - Ks-iut' (2.20)
That is, the amplitude of H(a) must be constant over the entire"frequency range, and the phase of H(u)
must be linear with frequency. This is illustrated in Fig. 2.3(c) and (d).
| . 2.4 | Analog and Digital Communicotions

(a) lnput signal (c) lH(o)l versus q,

0 fa t1+t6 t Slope = -td

(b) Output signal (d) d6r
06@) versus t..,

Fig.2.3 Distortion less tra nsmissio n

B. Amplitude Distortion and Phase Distortion

When the amplitude spectrum lH(r)l of the system is not constant within the frequency band of interest,
the frequency components of the input signal are transmitted with different amounts of gain or attenua-
tion. This effect is called amplitude distortion
When the phase spectrum 0o@) of the system is not linear with frequency, the output signal has a
different waveform from the input signal because of different delays in passing through the system for
different frequency components of the input signal. This form of distortion is calledphase distortion.

2.5 FILTERS
A. Ideal Fitters
By definition, an ideal filter has the characteristics of distortionless transmission over one or more speci-
fied frequency band and has zero response at all other frequencies.
An ideal bandpass filter (BPF) is defined by
1l
Fl"or(rl) : 'lIn_1r,., for ir,,
cr
-'
us l(.r,,-
' - c.2
(2.2r)
[0 otherwise

The amplitude and phase spectra of Hrpp(u) are shown in Fig. 2.4. The ideal BPF passes all input
signal components with frequencies betweefl u., and t,^.l,, rvithout distortion and all other signals
components are rejected. The parameters u., and uJ., are the lower and upper cutoff frequencies,
respectively.
An ideal low-pass filter (LPF) is defined by Eq. (2.21) with u., : 0.An ideal high-pass filter (HPF) is
defined by Eq. (2.21) with t^,1,,, > 0, Q,, : oo. An ideal bandstop filter (BSF) or notch filter is defined by

d llul(u.,
Huto(') : fo cl
-
l r
-
(.
(2.22)
- j'ttt
t, olherwise
 Signol Transmission and Filteing t-ti_-]

Fig.2.4 Frequency response of an ideal. BPF

B. Causat Filters
Notice that all ideal filters discussed in the preceding section are noncausal since h(t) * 0 for r < 0. It is
not possible to build ideal filters. As shown in Eq. (2.6), for a causal filter (or physically realizable filter)
its impulse response h(t) must satisfu the condition

h(t1 :0 for / < 0

C. Filter Bandwidth
The bandwidth WB of an ideal low-pass filter equals its cutoff frequency, that is, WB - u)c
lFig.2.12(a)l.Thebandwidthofanidealbandpassfilterisgivenby Wu: @,, - Ldc, (FiS.2.4).Themid-
.1
point ao:
;Z (, ,, I e,r) is the center frequency of the filter A bandpass filter is called narrowband
if WB ( ( c..ro. No bandwidth is defined for a high-pass or bandstop filter.
For nonideal or practical filters, a corlmon definition of filter (or system) bandwidth is the
3-dB bandwidth W, a". In the case of a low-pass filter, W, o, is defined as the positive frequency
at which the amplitude spectrum lH(w)l drops to a value equal to lfl(O)ll"{i, as illustrated in
Fig.2.5(a). In the case of a bandpass filter, Wro, is defined as the difference between the frequen-
cies at which lH(r)ldrops to a value equal to llli times the peak value lH(qi at the filter's middle
frequency r.,.,'o (called the midband frequency), as illustrated in Fig. 2.5(b). This definition is sornewhat
arbitrary and may become ambiguous and nonunique with multiple peak frequency responses, but it is a
widely accepted criterion of measuring a system's bandwidth. Note that each of the preceding bandwidth
definitions is defined along the positive frequency axis only and always defines positive-frequency, or
one-sided bandwidth only.

-u1 0 O1 d1 ag a2

lre F F'e *l
(a)

Filter bandwidth
J
I

I
 Anolog and Digital Communicotions

2.6 QUADRATURE FILTERS AND HILBERT TRANSFORMS

A. Quadrature Fitter
A quadrature filter [or - rl2 radian (-90") phase shifter] is an allpass system whose frequency response

is given by l1("r) : ln-j"''

.l
cu)o (2.23)
lel"'z c,,' ( 0

Since ,+irt2 : +j, H(u) can be rewritten as H(w): -7 sgn (c,r) (2.24)

The corresponding impulse response h(t) canbe obtained as (Solved Problem 2.19)

h(t)
l (2.2s)
- rt

B. Hitbert Transform
Let a signal x(r) $e the input to a quadrature filter (Fig. 2.6). Then by Eq. (2.6) the output y(t) - x(t) *
h(t) : x(/) * (llrt) will be defined as the Hilbert transform of x(r), denoted by i (r). Thus,
1-'[" l@-a, (2.26)
*(t1 -x(t)* ttt 7f 'r-e t-T
The Fourier transform of i (r) is given by (0
X (r) : H(a)X(a) : [-l sgn (u)]X(w) (2.27) Fig. 2.6 -7t/2 rad phase shifter

IMPULSE RESPONSE AND FREQUENCY RESPONSE

,,,.,
i:.

(a) "v- lxr(t) + xz!)): [x,(/) + xr(t)] cos a"t

- xrQ) cos u,t * xr(t) cos w,t : c*'[xr(r)] + @ [xz(t))

aF* lax(t)l
- [ax(/)] cos e,t : acF'tx(41
Hence, the system represented by (a) is linear.
(b) aF lxr(t) + xr(t)l: * x,(r) + x2(t)) cos u,t
lA
= r7= [r,(r)] * d- l*r(t)l
- lA + xr(01 cos act + lA + xr(t)l cos a.r"/
: l2A * x,(r) * xr(t)l cos u,t
 Signal Transmission and Filtering

Thus, the system represented by (b) is not linear. The system also does not satisfy the homogeneity
condition (2.3). Note that the system represented by (a) is called the balanced modulator for DSB
(double-sideband) signals in amplitude modulation (see Sec. 3.3), and the system represented by
(b) is the generator for ordinary AM (amplitude modulation) signals (see Sec. 3.4).

#* cffisidsf a *ffi ** *fiffiffiI;l

H:::,,
{lr) Is this system linear?
$l Is this system time-invariant?

(a) Let x(t)

- xr(/) * xr(t). Then
y(t) -- [r,(0 * xr(t)]6r(r) : xt(t)6r(t) * xr(t)6r(t) : y{t) * yzQ)
Similarly, let x(r) : oxr(/). Then y(t) : laxr(t))\rQ) : alx,(t)6r(t)l : ay,(t)
Thus, the system is linear.
(2n ',l

(b) Let x1(/) : .o, r,J

[.7
rhen y{t): i *,@r)6(t - nr): i .o,
l+ "r)6(t -
nr):,4 5(t - nr)

Next, let the input be x,(t) -,, [, - i) :.,, [?,)

rhen yzQ)
-,lrt, lT
*) 6(t - nry: o =r,
[,
_i)
Thus, the iystem is not time-invariant.
Note that this system is known as the ideal sampler (see Sec. 5.4)
,

e$iffiffi nu,o u ffiffi ih*

j:.,:::,:,:
:

[ : i, . hh((t), :, fl*xor{t
* r) dr

,,,.,,1:;,,,.,#.ffi.,ffiiiffiffi1ffi;ffi1i:rytput and input, respectively, of an LTI system whose impulse

fg$p0ilSe,ts ft(Ili ,,, :: ,,1,
:,:t: ,, a ,r::i l :

1, 1,,,r,
:

If the system is time-invaiant, then from Eq. Q.\ we have

a7* l6(t - ,)l : h(t-r)
Now, from definition (1.31) of 6(r - h), we can express x(r) as

x(t\:
\/ f'- x(z)5( t-r) dr
J-,o
Then from the linearity of the r,7'operator, we obtain
'lx(t)l: : *
y(t)
-. f**rr, a7'l6(t-r)l d,
f x(r)h(t-r) dr
 Anolog ond Digital Communications

2.4 The response of an LTI system to a unit step function u(t) is called the unit step response of tho
system and is denoted by a(t). Show that a(t) can be obtained as

a(t) - I_*ot"l o, (2.28)

and if the system is causal, then

o{t) * a" (2.2e)
linvt
From Eq. (2.8), a(t) : h(t) * u(t1 - [- ,Alu(t - r) dr
[t r{t
Since u(t - r): l r]t
[0
we have a(t): o"
I'_*r<r1
For a causal system, since h(;r)
- 0 for r 1.0,
a(t) - [' t{r1 a"

2.5 Let an LII system with impulse response ft(r) be represented by an operator i7'.lf
rr* lx(t)l .\r(r) (2.30)
-
then .\ is called lhe eigerwalue of ,F andr(r)r is called the associated eigenfunction of crT .
Show that the ftequency response H{*): (f tlr(f)l is the eigerwalue of the LfI system and ei*'
is the associated eigenfunction.

Using Eq. (2.8), we have

.r- leio,) : d,tt - 't dr: hlry s- t, , ar)r-,"' - H(w) ei'' (2.31)
[:h(r) lI ,.
Thus, we see that H(u) is the eigenvalue of the LTI system and ei'' is the associated eigenfunction.

2.6 AILTI system with impulse response h(t1

- {*'u(Ais driven by tfre inputx(f} * {P'uqtl,where
a, p > 0. Find the output of the system, in both frequency and time dornain.

Let the output be y(t), then by the convolution theorem,

Y(f \: x(f) H(.f ): [- I ) r 1-'l

ro+inf)l,"+Jhf)
llll
@ - a) (a * jz"f) @ - a) (0 + jz"f)
y(t): o7 -tlY(f)l: fdo'-
h "-P'1u1t1
 Signal Tronsmission and Filtering

'

2.? Determiui the impulse response of the circuit shown in Fig. 2.7(a).It is a typical example of a
'hoiding cilcuit, commonly used in pulse systems.
h(t)
x(0 At)
input output 1

Fig.2.7(a)
Let x(t) - 51v1

s(t)-6(t)-6(t-to)
y(t) [-
- h(r) - J-x OO) - 5(r- /0)] dr - u(t) - u(t - to)
where z(r) is the unit step function.

2.* afrc I-TI system to II(4 be A(r).

I-Et the response
Caa you ftnd t"he'reffin*e of this system to x(r) : cos (Zrt) from the information provided?
No, the input II(r) has a spectrum with zeros at frequencieslf- k (k = 0, k e Z) see Fig. 1.3(d),
r - 7, A : 1 and the information about the spectrum of the system at those frequencies will not
be present at the output. The spectrum of the signal x(t|: cos (2rt) consists of two impulses at
"f
: *.1, but we do not know the response of the system at these frequencies.
2.9 Consider the "RC network shown in Fig. 2.8(a). Find the frequency response H(a) and the impulse
,

reslonse ft(r) of the.RC network.

I I/^\
x(t) I
L /l
(0 ) +c
I

y{t)

! ot
{a) nCnstwork (b) lmpulse response of the RC network

Fig. 2.8

Using the principle of voltage divider, we can obtain the frequency response H(r,,,) by inspection:
U(iaC) : o o:
H(a\- 1
(2.32)
R+U(juC) jaRC*r ja*a RC
Next, from Solved Problem l.9 we obtain
o'u(t): I
h(t) (2.33)
- ae RC
e-,t(R€tu1)

which is sketched in Fig. 2.8(b). r

 Analog ond Digital Communications

2.1S,. eonsifu the,.sirytre;#C airclrit shown in Fig. 2.8(a). Find the unit step response a(f), From

h(t1: l' e-"(*')r(t)

RC
Thus, by Eq. (2.29) the unit step response is

a(t) - Jo \ / a, : !RCJo
f' lr(r) lt s-'rtnctdr - ft - e-"ac\u(t) (2.34)
.

i.tX Redo'Sdffi ffi 2.i0 *i* frequency response and Fourier imrersioa technique.

Now x(t)
-z(r). Thus, by Eq. (1.7g) X(w)
- r6(a)+ I
Ja
Next,byEq.(2.32) H(u): .o ,: t.
1a*a RC

Thus, by Eq. (2.9) we obtain

Y(w)-x(u\H(wl- [r..a1r)+ll t.O-l I - .+-

I ldl l.lu*a) -tr6(a\+Ja JQ+a
In the last step we have used the property (1.33) of 5 function and the partial fraction expansion
technique. Taking the inverse Fourier transform of I(r,,,), we obtain

y(t) - a(t) - u(t) -e "tu(r) - (7 - e "')u(t) a - ll(Rq

Fitter Characteristics of Linear Systems

2.12 Show that the .RC neturork of Solved Problem 2.9 [fig, 2.8(aX is a low-pass fi]ter. Also find its
3-dB bandwidth Wr*.

From Solved Problem 2.9,the frequency response H(r) is given by

H(,):i,u:#,*
where ao: lKRq. Writing H(r) -lH(w)lsi0n@)
a
we have and 0o@) - -tan-t ao

The amplitude spectrum lH(a)l and phase spectrum 0r(r) are plotted in Fig. 2.9. From Fig. 2.9
we see that the RC network of Fig. 2.8(a) is low-pass filter.
When e : ao
- r/(RC), lH(u)l
- rl Jr.Thus,
WNa:'o:
#
 Signal Transmission and Filteing

-us O uo:11(RC)
(a)

Fig. 2.9

Z.l3 The rise time t" of the low-pass RC filter of Fig. 2.8(a) is defined as the time required fot a unit
ry response
to
go from l0 to 90% of its final value. Shou'that
q.35
U.JJ
tr:T
-/3dB

wheref* a W36sl{2n} : 6,a

l/(ZrttC) is the 3-dB ba*dwidth (in hertz) of the fiIter.

Fat

From Eq. (2.34) of Solved Problem 2.10, the unit step response of the low-pass RC filter is found
to be
a(t)-(l-r-ti@c')u(t)
which is sketched in Fig. 2.10. By defini-
tion of the rise t1me,
tr: t2- t1
where a(tr) - | -e-ttt(Rc)
: 0.1 ---+ s-t1l(nc) : 0.9
a(tr) - l- dtzt(Rc)
: 0.9 -+ s-t2t(Rc) : 0.1
ir1
tl
ltz t
tl
Dividing the first equation by the second t-l
t'!
equation on the righrhand side, we obtain
Fig. 2.10
e(tr-tr)lRC):9
and t,: t2- tr: RC ln 9 :2.197RC : 2.t97 0.35
2nfzds fro,

2.14,' si@ r) *.. ,is apptied to a brick-wall low-pass filftr If{/)

. a: ::.:. t, . ::::: ,, a ,

Find the value of A zuch &at dre filter passes one-half the energy of x(r)
 ::_
-
--.+.-_

Anolog and Digitat Communications

x(t) ,4oont * X(f) _ I

- 400r + j2trf
Energy in x(t): E*: f ()dt: : I
I: ,[r* ,_8*rt 4, &oo;
Joules.

E.-L - I

Power transfer tunction of a

nerwort i, giul *n,.n is equar
^or'i!lri, to uniry for lfl< B.
' Eo":
"'
tr
-;
I
: [:, ng
x(-f) l' d-f :' -[o' ^P
I
ffi,*
r)
1rB IL
- 2", Jo ,ooL_lTf, I

['-[;no) )

:
*.#ran- t#l : #xtan- t#l
=+ #tan-r(*):#
tun, I B
)__ +oor, _ o
[2oo) 1600"- 4
B .rT

tan- _ I B 200 Hz.

200: -
2-rs how fta ru*,ffirr. droum in, Fig. 2.rr(a)is a high_pass
firter , , ,

In a manner similar to solved Problem 2.9,

wecan obtain the frequency response
r(,,,) by inspection:
H(a):rli!-- .iut(L/R)
- i(a/uo\
R+ial t+jw@R)'t+j@6,r:i, R
e.35)
is plotted in Fig. z.tt(b).rt
]llffigu:'$n]""' is seen that the RZ nerwork orFig.
2.u(a) is

(a) RL network
(b) Magnitude of frequen",
?"rronse of the nt nJwort
Fig.2.t!
 Signal Transmission ond Filteing

Fitters

2.16 Find the irnputse response }(r) of the ideal LPF with cutofffrequency c..r".

The frequency response of an ideal LPF with cutoff frequency c,.', is given by
( _;
for Ic"'lSa,
Hr.r, (a): f'-''" (2.36)
[0 otherwise

The amplitude and phase ofHrpr (w) are shown in Fig. 2.12(a). The impulse response of the ideal
LPF can be found by taking the inverse Fourier transform of Eq. (2.36), which yields

ilLPFG\ -
sin w'"(t - t') (2.37)
r(t _ to)
The impulse response ftr..(r) is shown in Fig. 2.12(b). Note that hrr, (t) = 0 for t < 0. Thus, the
ideal LPF is not a causal system.

td

2n
ac

(b)

Fi1.2.72 Frequency response and impulse response of an ideaL LPF

., .. :. , ,

iiy Consder eb E
ttt \
stnm1@ ih Fig. 2. 1 3(a). Find the impulse respons e hr(t) and its frequency reqponsg
ttlu)); i.,,,,,,.
,' .

From Fig.2.13(a) we can write the relation-

ship between input x(r) and output y(t) as
y(t)-x(t)-x(t-T)
Thus, by definition (2.5) we get

h(t)-5(t)-5Q-r)
Using Eq. (1.40) and property (1.18) we
(a)
obtain
Fig. 2.13 (a)
 Anolog and Digital Communications

H(w)-l-s-iur
- ,-iaTt27"jaTt2 _ e-iur/2)

:2 sin lJ,)r-i@r+tr\/z
l2) 0as
The magnitude of H(w) is shown in Fig. (b)
2.13(b). The system is known as comb Fig. 2.13(b)
filter.

i;,1$ U oOcurs when a transmitted signal arrives at,tl,re'r,oooffi, fu,f'wo.or,.msrs

, , ,, , ',,phfft$. .di . Ws;,A dmple model for a multipath sommunicati$t sh el is illtlsfreted
in Fig. 2.Ia(a).

(a) Find the,freqte cy systfln firnction H{u} for this channel andqlot f(r)l for a : I and 0.5.
," .r $)' ft .e$ ,ftfffi'ro,ffi4'indueed'distortioru an eqwdiiffiiam.,filter i,$.'often'uti'lized.

(a) Model for multipath transmission

(b) Tapped delay-liner filter

Fig.2.!4

H"r('):
H(w)
A tapped delayJine or transversalfilter, as shown in Fig. .2.14(b), is commonly used to approxi-
mate this equalization filter. Find the values for ay a2, ... ... oN, assuming r :
T and a << 1.
(a) y(t)-x(/)+ ax(t-r)
 Signal Transmission ond Filteing

Taking the Fourier transform of both sides, we have

Y(w)
- X(u) +ae-i"X(u) : (l + ae-i")X(r)
By Eq. (2.10),
(')
Y
\
H(u\ / -- x(r) - I -F e.s-t"tt

Using Euler'identity for e-i'' gives

H(r) : I -F 0 cos ur ja sin a,,r
Thus, lH(r)l- [(l * o cos o'r)2 + (a sin ,r)'lt'' - [(1 * a2 * 2crcoso'r)fi/z

a: l, :
I Tl
When lH(r)l [2(l * cos ,r)),,, - Z l.o, 2l
when o - L,
2
W@)l: (r.25 * cos ,r)r,,
1
Amplitude spectra lH(u)l for o
- I and a - are plotted in Fig. 2.L5.
2

-+ 0 + u

Fig. 2.15
(b) From Fig.2.14(b), we have
N
z(t)
- Dooylt-(k-t)Tl
k--l
Taking the Fourier transform of both sides gives
N
Z(w)
- \k:l a,g-iar-t)rY(a)
Thus, the frequency response Hn(w) of the transversal filter is given by

H,,(r):
' ?,\'! :fo,,"r"(k-Dr :ar*are-iur *aret'zr+... + are-i'(N-Dr
Y(w) -o=,
Now H"r(r): r1
fr t+M-*T
-l x+x2-r'+....1r1 <l
l*x -l- r
Using
 Analog ond Digital Communications

we can express r/"'(cu) as H"r(r) - I - ae-i" * a2e-i'z' +...

Thus, if r - T and lol < l, wohave or : l, a2: -(1, o3: s12,. . ., aN: (-r)'-'
Quadrature Filters and Hilbert Transforms

2.19 Veri$ Eq. (2.25J.

From Eq. (1.61) we have sgn(r) -- L

Ja
Applying duality property (1.22), we obtain

Z --- 2r sgn(-,,t) : -2r sgn(r,,,)

1t
from which we get I .,
-7sgn(c,r)
7rt

2t.20 Shorv *at a signal x($'aad its Hilbert transform i(r) have the same alnplitude speotrum.
By Eq. (2.27)

* @) : l-j sgn (r,,,)l X(o)

Since l-,1 sgn (cu)l
- l, we obtain

l* Q,))l - l-j sgn (u/)llx( a)l : lx(w)l

tf* , *- dffitransrormof i(rlisni(0r@#,

:,: . ... :, ..,: , :a,: : , j: tl i'j_

,,4*t ffi,nou*o,*'u
(2,38)
Let x(r) +-, X(a)
Then by Eq. (2.27)
i(r)
-l-j sgn(c,r)lx(0r) *-+ X(u)
and i(0 ., it l : l-j sgn(,r)l2x( a) : -X(u)
since [-7 sgn(o)]'-,1'[rgr(r)]' - -1. Therefore, we conclude that i(r) - -x(t).

i-tTi''I*li''x{r}ffi a'!eaX's1ffilffiffi,th*'i(r} and its l{ilbe#,trafisform i{r) are,oq hd$Onal, ttrat is,

{a'3eJ

Using Eq. (1.70), we have

f*.t>*(t) dt :* I*_*t"l k eo a,
If x(/) is real, then by Eq. (1.12) we have
* ed : lj sgn(--ur)lX(-:,r)
-,r sgn (r\fr * (ur)
 Signal Transmission and filteing

rhus,
f*.r,rr(t)dt
: *,/-]r*( ,)x(rv * (a)dw : *-[irr"r elx(a)1z du -0
since the integrand in the last integral is an odd function of a.,.

,.,:

From Eq. (1.76)

Then

- -jnl6(, - ro)- 6(, + ro)l

Thus, by the result of Supplementary Problem 1.10., we obtain iQ)
- sinu,,o/

Note that i(r) : cos lrf -il : siniro/.

l. " 2)

.,.: Lrlt x(t)

2,24 - 6(r).'

,,](bJUsetheresuItof(o)toconfirmthat(Solvedr,muterrr2.|9}

-
I r* 7 sgn (a,,)
rt
(a) By definition (2.26) and Eq. (1.37) we obtain
rl
-i(r)-6(r)* I -
rt -lrt
(b) From Eq. (1.a0) 6(r)*X(u)-1
Then, by Eq. (2.27) we have * @)
- - j sgn (cuX( a) : -f sgn(cu)

Thus, we conclude that

I .,
-,1sgn (o)
nt

2.1 Consider the system whose input-output 2.3 A system is called BIBO stable if every
relation is given by the linear equation bounded input produces a bounded output.
Y(t): ax(t) + b Show that the system is stable if its impulse
where x(r) and y(t) are input and output of response is absolutely integrable, that is,
the system, respectively, and a and b are
[* w<'lldr < x
constants. Is this system linear? pns. No]
2.2 Consider the system that is represented by lHint: Take the absolute value of both sides of
Eq. (2.8) and use the fact that lx (t - r)l < K.l
cf
[x(t)] - x*(t) 2.4 Consider the simple RC circuit shown in
where x*'(t) is the complex conjugate of Fig. 2.8(a). Find the output y(t) when the
x(r). Is this system linear? [,,4ns. No] . input x(t) - p"(t) [see Eq. (1.56)].
 Analog and Digitat Communicotions

y(t) p,(t) o(t * ,(! + (b) Find n so that lH(r)l' is constant to

fAns. - - e ",' a) +
e-a(t-r)u(t - all within I dB over the frequency range
2.5 Find the frequency response
the H(u) of of lc''''l
: 0'8c''ro'
network shown in Fig. 2.16, and show that
the network is a high-pass filter.
lurr.(a) Note that lim lO)" _
lAns. H(u) - -LCuzl(r - LCG + jwRQl I '--luo J

HC [m for c.., I

2.10 If
l; il;ll''' "-')
the unit impulse response of a causal LTI
system contains no impulse at the origin,
Fig. 2.16 then show that with

2.6 Find the frequency response H(r) of the H('):A(w)+iB(a)

network obtained by interchanging C and L A(a) and B(a) satisfy the following equa-
in Fig. 2.16, and show that the network is a tions:
low-pass'filter.
fAns. H(r) - 1/(1
2.7 Determine the impulse response and the 3-
- LCuz + jaRQl
[- r(rl
A(,)- 17lr-x,D-A ar
dB bandwidth of the filter whose frequency
response is f(c,,')
- l0l(a2 + 100). r- e(o),a\
B(*,):-lirJ--;-A
frAns. h(t) - ! a'ot'l' W, o, - 6.44radians
per second (radls)l These equations are known as the Hilbert
transform pair.
A gaussian filter is a linear system whose lHint: Let h(t) : h"(t) + h,(t) and use the
frequency response is given by causality of h(t) to show that h"(t) : h,(t)
H('): s-auz s-iuto lsgn(r)1, ho(t) - ft"(r)[sgn(r)].1

calculate (a) the 3-dB bandwidth W, o, and 2.ll Show that

(b) the equivalent bandwidth IZ.o defined by
/- t olf'dt - [* t*{Df a,
rr/"q:
; ,, f*tru<r>la, lHint: Use Eq. (2.27) and apply Parsevalb
theorem (1. l6) for the Fourier transform.]
I osq(b) W"c- 0'886 I

lAns.(a)ryas:
r, ry
,la J;
|

)
2.12 Show that

2.9 A Butterworthlow-pass filter has x(t)-*(t)*

I
t-*)
f,Hint: Use Eq. (2.27)l
lH(w)l -
2.13 Let (a) x(t) - sin wof; @) x(r) -- m(t) cos uct;
where n is the number of reactive compo- (c) x(t) - m(t) sin u"t. Find i(r).
nents (i.e., inductors or capacitors). lHint: Use Eq. (2.27).)
(a) Show that as n ---+ x,lH(w)l approaches lAnq.(a) *(t): - cos r,,,,0/; @) i1t1 - m(t) sin
the characteristics of the ideal low-pass w"t; (c) i(t1 -m(t)cosr.,,r"/l
-
filt% shown in Fig. 2.12(a) with cue - (rc.