Bandpass Signaling

Chapter
BANDPASS SIGNALING
PRINCIPLES AND CIRCUITS
CHAPTER OBJECTIVES
Complex envelopes and modulated signals
Spectra of bandpass signals
Nonlinear distortion
Communication circuits (mixers, phase-locked loops, frequency
synthesizers, and detectors)
Transmitters and receivers
Software radios
This chapter is concerned with bandpass signaling techniques. As indicated in Chapter 1, the
bandpass communication signal is obtained by modulating a baseband analog or digital signal
onto a carrier. This is an exciting chapter because the basic principles of bandpass signaling are
revealed. The complex envelope is used, since it can represent any type of bandpass signal. This
is the basis for understanding digital and analog communication systems that are described in
more detail in Chapters 5 and 8. This chapter also describes practical aspects of the building
blocks used in communication systems. These building blocks are filters, linear and nonlinear
amplifiers, mixers, up and down converters, modulators, detectors, and phase-locked loops. The
chapter concludes with descriptions of transmitters, receivers, and software radios.
41 COMPLEX ENVELOPE REPRESENTATION OF BANDPASS WAVEFORMS
What is a general representation for bandpass digital and analog signals? How do we represent
a modulated signal? How do we represent bandpass noise? These are some of the questions
that are answered in this section.
237
238 Bandpass Signaling Principles and Circuits Chap. 4
Definitions: Baseband, Bandpass, and Modulation

DEFINITION. A baseband waveform has a spectral magnitude that is nonzero for fre-
quencies in the vicinity of the origin (i.e., f = 0) and negligible elsewhere.
DEFINITION. A bandpass waveform has a spectral magnitude that is nonzero for
frequencies in some band concentrated about a frequency f = fc, where fc 0. The
spectral magnitude is negligible elsewhere. fc is called the carrier frequency.
For bandpass waveforms, the value of fc may be arbitrarily assigned for mathematical
convenience in some problems. In others, namely, modulation problems, fc is the frequency of
an oscillatory signal in the transmitter circuit and is the assigned frequency of the transmitter,
such as, for example, 850 kHz for an AM broadcasting station.
In communication problems, the information source signal is usually a baseband
signalfor example, a transistortransistor logic (TTL) waveform from a digital circuit or
an audio (analog) signal from a microphone. The communication engineer has the job of
building a system that will transfer the information in the source signal m(t) to the desired
destination. As shown in Fig. 41, this usually requires the use of a bandpass signal, s(t),
which has a bandpass spectrum that is concentrated at fc, where fc is selected so that s(t) will
propagate across the communication channel (either a wire or a wireless channel).
DEFINITION. Modulation is the process of imparting the source information onto a
bandpass signal with a carrier frequency fc by the introduction of amplitude or phase
perturbations or both. This bandpass signal is called the modulated signal s(t), and the
baseband source signal is called the modulating signal m(t).
Examples of exactly how modulation is accomplished are given later in this chapter.
This definition indicates that modulation may be visualized as a mapping operation that maps
the source information onto the bandpass signal, s(t). The bandpass signal will be transmitted
over the channel.
As the modulated signal passes through the channel, noise corrupts it. The result is a
bandpass signal-plus-noise waveform that is available at the receiver input, r(t). (See Fig. 41.)
'
The receiver has the job of trying to recover the information that was sent from the source; m
denotes the corrupted version of m.
Complex Envelope Representation

All bandpass waveforms, whether they arise from a modulated signal, interfering signals, or
noise, may be represented in a convenient form given by the theorem that follows. v(t) will be
used to denote the bandpass waveform canonically; specifically, v(t) can represent the signal
Information
input ~ ~
m g(t) s(t) Transmission r(t) g(t) m
Signal Carrier medium Carrier Signal
processing circuits (channel) circuits processing
Transmitter Receiver
Figure 41 Communication system.
Sec. 41 Complex Envelope Representation of Bandpass Waveforms 239
when s(t) K v(t), the noise when n(t) K v(t), the filtered signal plus noise at the channel out-
put when r(t) K v(t), or any other type of bandpass waveform.
THEOREM. Any physical bandpass waveform can be represented by
v(t) = Re{g(t)e jvct} (41a)
Here, Re{} denotes the real part of {}, g(t) is called the complex envelope of v(t), and
fc is the associated carrier frequency (in hertz) where vc = 2p f c. Furthermore, two other
equivalent representations are
v(t) = R(t) cos[vc t + u(t)] (41b)
and
v(t) = x(t) cos vct - y(t) sin vct (41c)
where
g(t) = x(t) + jy(t) = g(t) ej
lg(t)
K R(t)eju(t) (42)
x(t) = Re{g(t)} K R(t) cos u(t) (43a)
y(t) = Im{g(t)} K R(t) sin u(t) (43b)
R(t) ! |g(t)| K 3x (t) + y (t)
2 2
(44a)
and
u 1t2 ! g1t2 = tan-1 a b

y(t)
(44b)
x(t)

Proof. Any physical waveform (it does not have to be periodic) may be represented
over all time, T0 : q , by the complex Fourier series:
n=q
jnv0t
v(t) = a cn e , v0 = 2p>T0 (45)
n= -q
1
Furthermore, because the physical waveform is real, c-n = c*n , and, using Re{# } = 2{
#}
+ 12{# }*, we obtain
v(t) = Ree c0 + 2 a cnejnv0t f

q
(46)
n=1
Furthermore, because v(t) is a bandpass waveform, the cn have negligible magnitudes for n in
the vicinity of 0 and, in particular, c0 = 0. Thus, with the introduction of an arbitrary parame-
ter fc, Eq. (46) becomes

The symbol K denotes an equivalence, and the symbol ! denotes a definition.

Because the frequencies involved in the argument of Re{} are all positive, it can be shown that the complex
function 2 g n = 1 cnejnv0 t is analytic in the upper-half complex t plane. Many interesting properties result because
q
this function is an analytic function of a complex variable.

n=q
v(t) = Re ea2 a cnej(nv0 - vc)t b ejvctf (47)
n=1
so that Eq. (41a) follows, where

q
g(t) K 2 a cnej(nv0 - vc)t (48)
n=1
Because v(t) is a bandpass waveform with nonzero spectrum concentrated near f = fc, the
Fourier coefficients cn are nonzero only for values of n in the range nf0 fc. Therefore,
from Eq. (48), g(t) has a spectrum that is concentrated near f = 0. That is, g(t) is a
baseband waveform.
The waveforms g(t), (and consequently) x(t), y(t), R(t), and u(t) are all baseband
waveforms, and, except for g(t), they are all real waveforms. R(t) is a nonnegative real
waveform. Equation (41) is a low-pass-to-bandpass transformation. The ejvct factor
in Eq. (41a) shifts (i.e., translates) the spectrum of the baseband signal g(t) from base-
band up to the carrier frequency fc . In communications terminology, the frequencies in
the baseband signal g(t) are said to be heterodyned up to fc . The complex envelope,
g(t), is usually a complex function of time, and it is the generalization of the phasor
concept. That is, if g(t) happens to be a complex constant, then v(t) is a pure sinusoidal
waveshape of frequency fc, and this complex constant is the phasor representing the
sinusoid. If g(t) is not a constant, then v(t) is not a pure sinusoid, because the ampli-
tude and phase of v(t) vary with time, caused by the variations in g(t).
Representing the complex envelope in terms of two real functions in Cartesian
coordinates, we have
g(x) K x(t) + jy(t)
where x(t) = Re{g(t)} and y(t) = Im{g(t)}. x(t) is said to be the in-phase modulation
associated with v(t), and y(t) is said to be the quadrature modulation associated with
v(t). Alternatively, the polar form of g(t), represented by R(t) and u(t), is given by
Eq. (42), where the identities between Cartesian and polar coordinates are given by
Eqs. (43) and (44). R(t) and u(t) are real waveforms, and in addition, R(t) is always
nonnegative. R(t) is said to be the amplitude modulation (AM) on v(t), u(t) is said to be
the phase modulation (PM) on v(t).
Example 41 IN-PHASE AND QUADRATURE MODULATED SIGNALING

Let x(t) = cos(2pt) and y(t) be a rectangular pulse described by
0, t 6 1
y(t) = c 1, 1 t 2
0, t 7 2
Using Eq. (41a), plot the resulting modulated signal over the time interval 0 6 t 6 4 sec.
Assume that the carrier frequency is 10 Hz. See Example4_01.m for the solution.
Sec. 43 Spectrum of Bandpass Signals 241
The usefulness of the complex envelope representation for bandpass waveforms cannot be
overemphasized. In modern communication systems, the bandpass signal is often partitioned into
two channels, one for x(t) called the I (in-phase) channel and one for y(t) called the Q (quadrature-
phase) channel. In digital computer simulations of bandpass signals, the sampling rate used in the
simulation can be minimized by working with the complex envelope g(t), instead of with the
bandpass signal v(t), because g(t) is the baseband equivalent of the bandpass signal.
42 REPRESENTATION OF MODULATED SIGNALS
Modulation is the process of encoding the source information m(t) (modulating signal) into a
bandpass signal s(t) (modulated signal). Consequently, the modulated signal is just a special
application of the bandpass representation. The modulated signal is given by
s(t) = Re{g(t)ejvct} (49)
where vc = 2pfc, in which fc is the carrier frequency. The complex envelope g(t) is a function
of the modulating signal m(t). That is,
g(t) = g[m(t)] (410)
Thus, g[ # ] performs a mapping operation on m(t). This was shown in Fig. 41.
Table 41 gives the big picture of the modulation problem. Examples of the mapping
function g[m] are given for amplitude modulation (AM), double-sideband suppressed carrier
(DSB-SC), phase modulation (PM), frequency modulation (FM), single-sideband AM sup-
pressed carrier (SSB-AM- SC), single-sideband PM (SSB-PM), single-sideband FM (SSB-FM),
single-sideband envelope detectable (SSB-EV), single-sideband square-law detectable (SSB-
SQ), and quadrature modulation (QM). Digital and analog modulated signals are discussed in
detail in Chapter 5. Digitally modulated bandpass signals are obtained when m(t) is a digital
baseband signalfor example, the output of a transistortransistor logic (TTL) circuit.
Obviously, it is possible to use other g[m] functions that are not listed in Table 41. The
question is; Are they useful? g[m] functions that are easy to implement and that will give
desirable spectral properties are sought. Furthermore, in the receiver, the inverse function
m[g] is required. The inverse should be single valued over the range used and should be easily
implemented. The mapping should suppress as much noise as possible so that m(t) can be
recovered with little corruption.
43 SPECTRUM OF BANDPASS SIGNALS
The spectrum of a bandpass signal is directly related to the spectrum of its complex envelope.
THEOREM. If a bandpass waveform is represented by
v(t) = Re{g(t)ejvct} (411)
then the spectrum of the bandpass waveform is
1
V(f) = 2 [G(f - fc) + G * (-f - fc)] (412)
242
TABLE 41 COMPLEX ENVELOPE FUNCTIONS FOR VARIOUS TYPES OF MODULATION a
Corresponding Quadrature Modulation

Type of Mapping Functions
Modulation g(m) x(t) y(t)
AM Ac [1+m(t2] Ac [1 + m(t2] 0
DSB-SC Ac m(t2 Ac m(t2 0
PM Ac e jDpm (t ) Ac cos[Dp m(t2] Ac sin [Dp m(t2]
t t
Ac cos cDf m(s) ds d Ac sin cDf m(s2ds d
t
FM Ac e jDf 1- q m(s) ds
L- q L- q
SSB-AM-SC b N (t2]
Ac [m(t2 j m Ac m(t2 ;Ac m (t2
N
SSB-PM b Ac e jDp[m (t) ; jmN (t)] Ac e < Dpmn (t) cos [Dpm(t2] Ac e<Dp m(t) sin [Dp m(t2]
t t
Ac e < Df1- q mn (s)ds cos cDf m(s) ds d sin cDf m(s) ds d
t
t
N t n (s)ds
m
SSB-FM b Ac e jDf 1- q [m(s) ; jm(s)] ds
Ac e < Df1- q
L- q L- q
SSB-EV b Ac e {ln[1m(t2] < j l nN | 1m(t2|} Ac [1 + m(t2] cos{lNn [1 + m(t2]} ; Ac [1 + m(t2] sin{lNn[1 + m(t2]}
SSB-SQ b Ac e (122 {ln[1m(t2] < j l nN | 1(t2|} Ac 21 + m(t) cos{12 lNn[1 + m(t2]} ; Ac 21 + m(t) sin{12 lNn [1 + m(t2]}
QM Ac [m1(t2+jm2(t2] Ac m1(t2 Ac m2(t2
e
0, m (t) 7 - 1
AM Ac |1+m(t2| Lc m(t2 -1 required for envelope
180, m(t) 6 -1
detection
e
0, m (t) 7 0
DSB-SC Ac |m(t2| L Coherent detection required
180, m (t) 6 0
PM Ac Dp m(t2 NL Dp is the phase deviation constant
(radvolt)
t
FM Ac Df m(s2 ds NL D f is the frequency deviation constant
L- s
(radvolt-sec)
SSB-AM-SC b N
Ac 2[m(t)]2 + [m(t)]2
tan 1[;m
N (t2m(t2] L Coherent detection required
b ^ (t)
; Dp m
SSB-PM Ac e Dp m(t2 NL
t
t
SSB-FM b Ac e ; Df1- q mN(s)ds Df m(s2 ds NL
L- q
SSB-EV b Ac |1 + m(t2| ;lNn [1 + m(t2] NL m(t2 -1 is required so that ln()

will have a real value
SSB-SQ b Ac 21 + m(t) ; 12lNn [1 + m(t2] NL m(t2 -1 is required so that ln()
will have a real value
QM Ac 2m21(t) + m22(t) tan1[m2(t2m1(t2] L Used in NTSC color television; requires
coherent detection
ba
Ac 0 is a constant that sets the power level of the signal as evaluated by the use of Eq. (417); L, linear; NL, nonlinear; [ ] is the Hilbert transform (i.e., the 90 phase-
shifted version) of []. (See Sec. 55 and Sec. A7, Appendix A.)
b
Use upper signs for upper sideband signals and lower signs for lower sideband signals.
c
In the strict sense, AM signals are not linear, because the carrier term does not satisfy the linearity (superposition) condition.
243
and the PSD of the waveform is

1
v( f) = 4 [g (f - fc) + g(-f -fc)] (413)
where G( f) = [g(t)] and g(f) is the PSD of g(t).
Proof.
v(t) = Re{g(t)ejvct} = 12 g(t)ejvct + 12 g*(t)e - jvct
Thus,
1 1
V(f) = [v(t)] = 2 [g(t)ejvc t] + 2 [g* (t)e-jvct] (414)
If we use [g*(t)] = G*(-f) from Table 21 and the frequency translation property of
Fourier transforms from Table 21, this equation becomes
1
V(f) = 2 {G(f- fc) + G *[-(f + fc)]} (415)
which reduces to Eq. (412).
Example 42 SPECTRUM FOR A QUADRATURE MODULATED SIGNAL

Using the FFT, calculate and plot the magnitude spectrum for the QM signal that is described in
Example 41.
Solution
See Example4_02.m for the solution. Is the plot of the FFT results for the spectrum correct?
Using Eq. (415), we expect the sinusoid to produce delta functions at 9 Hz and at 11 Hz. Using
Eq. (415) and Fig. 26a, the rectangular pulse should produce a |Sa(x)| type spectrum that is
centered at the carrier frequency, fc = 10 Hz with spectral nulls spaced at 1 Hz intervals. Note that
the FFT approximation of the Fourier transform does not give accurate values for the weight of
the delta functions in the spectrum (as discussed in Section 28). However, the FFT can be used
to obtain accurate values for the weights of the delta functions by evaluating the Fourier series
coefficients. For example, see study-aid problem SA41.
The PSD for v(t) is obtained by first evaluating the autocorrelation for v(t):
Rv(t) = 8v(t) v(t + t)9 = 8Re{g(t)ejvct} Re{g(t + t) ejvc(t + t)}9
Using the identity (see Prob. 274)

1
Re(c2) Re(c1) = 2 Re(c*2 c1) + 12 Re(c2c1)
where c2 = g(t)ejvct and c1 = g(t + t)ejvc(t + t), we get
Rv(t) = 812 Re{g *(t)g(t + t) e-jvct ejvc(t + t)}9 + 812 Re{g(t)g(t + t) ejvct ejvc(t + t)}9
Sec. 44 Evaluation of Power 245
Realizing that both 8 9 and Re{ } are linear operators, we may exchange the order of the oper-
ators without affecting the result, and the autocorrelation becomes
Rv(t) = 1
2 Re{8g* (t) g(t + t) ejvct9} + 1
2 Re{8g(t) g(t + t) ej2vctejvct9}
or
Rv(t) = 1
2 Re{8g*(t) g(t + t)9 ejvct} + 1
2 Re{8g(t) g(t + t)ej2vct9 ejvct}
But 8g * (t)g(t + t)9 = Rg(t). The second term on the right is negligible because ej2vct =
cos 2vct + j sin 2vct oscillates much faster than variations in g(t)g(t + t). In other words, fc
is much larger than the frequencies in g(t), so the integral is negligible. This is an application
of the RiemannLebesque lemma from integral calculus [Olmsted, 1961]. Thus, the autocor-
relation reduces to
1
Rv(t) = 2 Re{Rg(t)ejvct} (416)
The PSD is obtained by taking the Fourier transform of Eq. (416) (i.e., applying the
WienerKhintchine theorem). Note that Eq. (416) has the same mathematical form as Eq.
(411) when t is replaced by t, so the Fourier transform has the same form as Eq. (412). Thus,
v(f) = [Rv(t)] = 1
4 [g1f - fc2 + g* (-f - fc)]
But g*(f) = g(f), since the PSD is a real function. Hence, the PSD is given by Eq.
(413).
44 EVALUATION OF POWER
THEOREM. The total average normalized power of a bandpass waveform v(t) is

q
Pv = 8v2(t)9 = v(f) df = Rv(0) = 1
2 8|g(t)|29 (417)
L-q
where normalized implies that the load is equivalent to one ohm.
Proof. Substituting v(t) into Eq. (267), we get
q
Pv = 8v2(t)9 = v(f) df
L-q
q
But Rv(t) = -1[v(f)] = v(f)ej2pftdf, so
L-q
q
Rv(0) = v(f) df
L-q
Also, from Eq. (416),

1 1
Rv(0) = 2 Re{Rg(0)} = 2 Re{8g*(t)g(t + 0)9}
or
Rv(0) = 1
2 Re{8|g(t)|29}
But |g(t)| is always real, so
Rv(0) = 1
2 8|g(t)|29
Another type of power rating, called the peak envelope power (PEP), is useful for trans-
mitter specifications.
DEFINITION. The peak envelope power (PEP) is the average power that would be ob-
tained if |g(t)| were to be held constant at its peak value.
This is equivalent to evaluating the average power in an unmodulated RF sinusoid that

has a peak value of Ap = max [v(t)], as is readily seen from Fig. 51b.
THEOREM. The normalized PEP is given by

1
PPEP = 2 [max |g(t)|]2 (418)
A proof of this theorem follows by applying the definition to Eq. (417). As described
later in Chapters 5 and 8, the PEP is useful for specifying the power capability of AM, SSB,
and television transmitters.
Example 43 AMPLITUDE-MODULATED SIGNAL

Evaluate the magnitude spectrum for an amplitude-modulated (AM) signal. From Table 41, the
complex envelope of an AM signal is
g(t) = Ac[1 + m(t)]
so that the spectrum of the complex envelope is
G(f) = Acd(f) + AcM(f) (419)
Using Eq. (49), we obtain the AM signal waveform
s(t) = Ac[1 + m(t)] cos vct
See Example4_3.m for a plot of an AM signal that is modulated by a sinusoid.

Using Eq. (412), we get the AM spectrum
1
S(f) = 2 Ac[d(f - fc) + M(f - fc) + d(f + fc) + M(f + fc)] (420a)
Sec. 44 Evaluation of Power 247
where, because m(t) is real, M*( f ) = M(-f ) and d ( f ) = d (-f ) (the delta function was defined to be
even) were used. Suppose that the magnitude spectrum of the modulation happens to be a triangular
function, as shown in Fig. 42a. This spectrum might arise from an analog audio source in which
the bass frequencies are emphasized. The resulting AM spectrum, using Eq. (420a), is shown in
Fig. 42b. Note that because G(f - fc) and G*(-f - fc) do not overlap, the magnitude spectrum is
1
Acd(f-fc) + 12 Ac |M(f-fc)|,
S(f) = e 21
f 7 0
1 (420b)
2 Acd(f+fc) + 2 Ac |M(-f-fc)|, f 6 0
The 1 in g(t) = Ac[1 + m(t)] causes delta functions to occur in the spectrum at f = ; fc, where fc
is the assigned carrier frequency. Using Eq. (417), we obtain the total average signal power
Ps = 1
2 A2c 8|1 + m(t)|29 = 1
2 A2c 81 + 2m(t) + m2(t)9
= 1
2 A2c [1 + 28m(t)9 + 8m2(t)9]
If we assume that the DC value of the modulation is zero, as shown in Fig. 42a, the average sig-
nal power becomes
1
Ps = 2 A2c [1 + Pm] (421)
where Pm = 8m2(t)9 is the power in the modulation m(t), 12 A2c is the carrier power, and 12 A2c Pm is
the power in the sidebands of s(t).
|M(f)|
1.0
B B
f
(a) Magnitude Spectrum of Modulation
|S(f)| Discrete carrier term

1
1 with weight = A c
Weight = A c 2
2
Lower Upper
Ac sideband sideband
2
f c -B fc f c+B f c -B fc f c +B
f
(b) Magnitude Spectrum of AM Signal
Figure 42 Spectrum of AM signal.
45 BANDPASS FILTERING AND LINEAR DISTORTION
Equivalent Low-Pass Filter

In Sec. 26, the general transfer function technique was described for the treatment of lin-
ear filter problems. Now a shortcut technique will be developed for modeling a bandpass
filter by using an equivalent low-pass filter that has a complex-valued impulse response.
(See Fig. 43a.) v1(t) and v2(t) are the input and output bandpass waveforms, with the cor-
responding complex envelopes g1(t) and g2(t). The impulse response of the bandpass filter,
h(t), can also be represented by its corresponding complex envelope k(t). In addition, as
shown in Fig. 43a, the frequency domain description, H( f ), can be expressed in terms of
v1(t)=Re [g1(t)ejct] Bandpass filter v2(t)=Re [g2(t )ejct]

h1(t) = Re [k1(t)e jct]
1 1
H(f)= K(f fc)+ K*(f fc)
2 2
(a) Bandpass Filter

|H(f )|
1 1
K*(f fc)| |K(f fc)|
2 2
fc fc f
(b) Typical Bandpass Filter Frequency Response
1 Equivalent low-pass filter 1

g1(t) g2(t)
2 1 2
1 G (f) 2 k(t)

1 1 1 G (f)
2
2 2 K(f) 2
(c) Equivalent (Complex Impulse Response) Low-pass Filter

1
|K(f)|
2
(d) Typical Equivalent Low-pass Filter Frequency Response

Figure 43 Bandpass filtering.
Sec. 45 Bandpass Filtering and Linear Distortion 249
K(f) with the help of Eqs. (411) and (412). Figure 43b shows a typical bandpass fre-
quency response characteristic |H(f)|.
THEOREM. The complex envelopes for the input, output, and impulse response of a
bandpass filter are related by
1 1
2 g2(t) = 2 g1(t) * 12 k(t) (422)
where g1(t) is the complex envelope of the input and k(t) is the complex envelope of the
impulse response. It also follows that
1 1
2 G2(f) = 2 G1(f) 12 K(f) (423)
Proof. We know that the spectrum of the output is
V2(f) = V1(f)H(f) (424)
Because v1(t), v2(t), and h(t) are all bandpass waveforms, the spectra of these waveforms are
related to the spectra of their complex envelopes by Eq. (415); thus, Eq. (424) becomes
1
2 [G2(f - fc) + G*2(-f - fc)]
1
= 2 [G1(f - fc) + G*1(- f - fc)] 12 [K(f - fc) + K*(- f - fc)] (425)
1
= 4 [G1(f - fc)K(f - fc) + G1(f - fc)K*(- f - fc)
+ G*1(-f - fc)K(f - fc) + G*1(-f - fc)K*(-f - fc)]

But G1(f - fc) K*(-f - fc) = 0, because the spectrum of G1(f - fc) is zero in the region of
frequencies around -fc, where K*(-f - fc) is nonzero. That is, there is no spectral overlap
of G1(f - fc) and K*(-f - fc), because G1(f) and K(f) have nonzero spectra around only
f = 0 (i.e., baseband, as illustrated in Fig. 43d). Similarly, G*1(-f-fc) K(f - fc) = 0.
Consequently, Eq. (425) becomes
312 G2(f-fc)4 + 312 G*2(-f - fc)4
= 312 G1(f-fc) 12 K(f-fc)4 + 312 G*1(-f-fc) 12 K* (-f-fc)4 (426)
Thus, 12 G2(f) = 12 G1(f) 12 K(f), which is identical to Eq. (423). Taking the inverse
Fourier transform of both sides of Eq. (423), Eq. (422) is obtained.
This theorem indicates that any bandpass filter system may be described and analyzed by
using an equivalent low-pass filter as shown in Fig. 43c. A typical equivalent low-pass frequency
response characteristic is shown in Fig. 43d. Equations for equivalent low-pass filters are usually
much less complicated than those for bandpass filters, so the equivalent low-pass filter system
model is very useful. Because the highest frequency is much smaller in the equivalent low-pass
filter, it is the basis for computer programs that use sampling to simulate bandpass communica-
tion systems (discussed in Sec. 46). Also, as shown in Prob. 417 and Fig. P417, the equivalent
low-pass filter with complex impulse response may be realized by using four low-pass filters with
real impulse response; however, if the frequency response of the bandpass filter is Hermitian sym-
metric about f = fc, only two low-pass filters with real impulse response are required.
A linear bandpass filter can cause variations in the phase modulation at the output,
u2(t) = l g2(t), as a function of the amplitude modulation on the input complex envelope,
R1(t) = |g1(t)|. This is called AM-to-PM conversion. Similarly, the filter can cause variations in
the amplitude modulation at the output, R2(t), because of the PM on the input, u1(t). This is
called PM-to-AM conversion.
Because h(t) represents a linear filter, g2(t) will be a linear filtered version of g1(t); how-
ever, u2(t) and R2(t)the PM and AM components, respectively, of g2(t)will be a nonlinear
filtered version of g1(t), since u2(t) and R2(t) are nonlinear functions of g2(t). The analysis of
the nonlinear distortion is very complicated. Although many analysis techniques have been
published in the literature, none has been entirely satisfactory. Panter [1965] gives a three-
chapter summary of some of these techniques, and a classical paper is also recommended
[Bedrosian and Rice, 1968]. Furthermore, nonlinearities that occur in a practical system will
also cause nonlinear distortion and AM-to-PM conversion effects. Nonlinear effects can be
analyzed by several techniques, including power-series analysis; this is discussed in the
section on amplifiers that follows later in this chapter. If a nonlinear effect in a bandpass sys-
tem is to be analyzed, a Fourier series technique that uses the Chebyshev transform has been
found to be useful [Spilker, 1977].
Linear Distortion
In Sec. 26, the general conditions were found for distortionless transmission. For linear
bandpass filters (channels), a less restrictive set of conditions will now be shown to be satis-
factory. For distortionless transmission of bandpass signals, the channel transfer function,
H( f ) = |H( f )|eju( f ), needs to satisfy the following requirements:
The amplitude response is constant. That is,

|H(f)| = A (427a)
where A is a positive (real) constant.
The derivative of the phase response is a constant. That is,
1 du(f)
- = Tg (427b)
2p df
where Tg is a constant called the complex envelope delay or, more concisely, the group
delay and u(f) = l H(f).
This is illustrated in Fig. 44. Note that Eq. (427a) is identical to the general requirement
of Eq. (2150a), but Eq. (427b) is less restrictive than Eq. (2150b). That is, if Eq. (2150b)
is satisfied, Eq. (427b) is satisfied, where Td = Tg; however, if Eq. (427b) is satisfied,
Eq. (2150b) is not necessarily satisfied, because the integral of Eq. (427b) is
u(f) = -2pfTg + u0 (428)
where u0 is a phase-shift constant, as shown in Fig. 44b. If u0 happens to be nonzero,

Eq. (2150b) is not satisfied.
Sec. 45 Bandpass Filtering and Linear Distortion 251
|H(f)|
Signal
bandwidth
fc
f
(a) Magnitude Response
u(f)
u0
fc
f
(b) Phase Response

Figure 44 Transfer characteristics of a distortionless bandpass channel.
Now it will be shown that Eqs. (427a) and (427b) are sufficient requirements for
distortionless transmission of bandpass signals. From Eqs. (427a) and (428), the channel
(or filter) transfer function is
H(f) = Aej(-2pfTg + u0) = (Aeju0)e-j2pfTg (429)
over the bandpass of the signal. If the input to the bandpass channel is represented by
v1(t) = x(t) cos vct - y(t) sin vct
then, using Eq. (429) and realizing that e-j2pfTg causes a delay of Tg, we find that the output
of the channel is
v2(t) = Ax(t - Tg) cos[vc(t - Tg) + u0] - Ay(t - Tg) sin[vc(t - Tg) + u0]
Using Eq. (428), we obtain
v2(t) = Ax(t - Tg) cos[vct + u(fc)] - Ay(t - Tg) sin[vct + u(fc)]
where, by the use of Eq. (2150b) evaluated at f = fc,
u(fc) = -vcTg + u0 = -2pfc Td
Thus, the output bandpass signal can be described by
v2(t) = Ax(t - Tg) cos[vc(t - Td)] - Ay(t - Tg) sin[vc(t - Td)] (430)
where the modulation on the carrier (i.e., the x and y components) has been delayed by the
group time delay, Tg, and the carrier has been delayed by the carrier time delay, Td. Because
u(fc) = -2p fc Td, where u(fc) is the carrier phase shift, Td is also called the phase delay.
Equation (430) demonstrates that the bandpass filter delays the input complex envelope
(i.e., the input information) by Tg , whereas the carrier is delayed by Td . This is distortionless
transmission, which is obtained when Eqs. (427a) and (427b) are satisfied. Note that Tg will
differ from Td, unless u0 happens to be zero.
In summary, the general requirements for distortionless transmission of either
baseband or bandpass signals are given by Eqs. (2150a) and (2150b). However, for the
bandpass case, Eq. (2150b) is overly restrictive and may be replaced by Eq. (427b). In
this case, T d Z T g unless u0 = 0 where T d is the carrier or phase delay and T g is the
envelope or group delay. For distortionless bandpass transmission, it is only necessary to
have a transfer function with a constant amplitude and a constant phase derivative over the
bandwidth of the signal.
Example 44 GROUP DELAY FOR A RC LOW-PASS FILTER

Using Eq. (427b), calculate and plot the group delay for a RC low-pass filter. Compare this result
for the group delay with that obtained in Example 218 for the time delay of a RC low-pass filter.
See Example4_04.m for the solution.
46 BANDPASS SAMPLING THEOREM
Sampling is used in software radios and for simulation of communication systems. If the
sampling is carried out at the Nyquist rate or larger ( fs 2B, where B is the highest frequency
involved in the spectrum of the RF signal), the sampling rate can be ridiculous. For example,
consider a satellite communication system with a carrier frequency of fc = 6 GHz. The
sampling rate required can be at least 12 GHz. Fortunately, for signals of this type (bandpass
signals), it can be shown that the sampling rate depends only on the bandwidth of the signal,
not on the absolute frequencies involved. This is equivalent to saying that we can reproduce
the signal from samples of the complex envelope.
Sec. 46 Bandpass Sampling Theorem 253
THEOREM. BANDPASS SAMPLING THEOREM: lf a (real) bandpass waveform has a

nonzero spectrum only over the frequency interval f1 6 | f | 6 f2, where the transmission
bandwidth BT is taken to be the absolute bandwidth BT = f2 - f1, then the waveform may
be reproduced from sample values if the sampling rate is
fs 2BT (431)
For example, Eq. (431) indicates that if the 6-GHz bandpass signal previously discussed had
a bandwidth of 10 MHz, a sampling frequency of only 20 MHz would be required instead of
12 GHz. This is a savings of three orders of magnitude.
The bandpass sampling theorem of Eq. (431) can be proved by using the Nyquist sam-
pling theorem of Eqs. (2158) and (2160) in the quadrature bandpass representation, which is
v(t) = x(t) cos vct - y(t) sin vct (432)

Let fc be the center of the bandpass, so that fc = (f2 + f1)2. Then, from Eq. (48), both x(t) and y(t)
are baseband signals and are absolutely bandlimited to B = BT2. From (2160), the sampling
rate required to represent the baseband signal is fb 2B = BT. Equation (432) becomes
n=q
a cxa f b cos vct - yaf b sin vct d c pf [t - (n/f )] d
n n sin{pfb[t - (n/fb)]}
v(t) = (433)
n= -q b b b b
For the general case, where the x(nfb) and y(nfb) samples are independent, two real sam-
ples are obtained for each value of n, so that the overall sampling rate for v(t) is fs = 2fb 2BT. This
is the bandpass sampling frequency requirement of Eq. (431). The x and y samples can be
obtained by sampling v(t) at t (nfb), but adjusting t slightly, so that cos vct = 1 and sin vct = -1
at the exact sampling time for x and y, respectively. That is, for t nfs, v(nfb) = x(nfb) when cos
vct = 1 (i.e., sin vct = 0), and v(nfb) = y(nfb) when sin vct = -1 (i.e., cos vct = 0). Alternatively
x(t) and y(t) can first be obtained by the use of two quadrature product detectors, as described by
Eq. (476). The x(t) and y(t) baseband signals can then be individually sampled at a rate of fb, and
the overall equivalent sampling rate is still fs = 2fb 2BT.
In the application of this theorem, it is assumed that the bandpass signal v(t) is recon-
structed by the use of Eq. (433). This implies that nonuniformly spaced synchronized samples
of v(t) are used, since the samples are taken in pairs (for the x and y components) instead of
being uniformly spaced Ts apart. Uniformly spaced samples of v(t) itself can be used with a min-
imum sampling frequency of 2BT, provided that either f1 or f2 is a harmonic of fs [Hsu, 1999;
Taub and Schilling, 1986]. Otherwise, a minimum sampling frequency larger than 2BT, but not
larger than 4BT is required [Hsu, 1999; Taub and Schilling, 1986]. This phenomenon occurs
with impulse sampling [Eq. (2173)] because fs needs to be selected so that there is no spectral
overlap in the f1 6 f 6 f2 band when the bandpass spectrum is translated to harmonics of fs.
THEOREM. BANDPASS DIMENSIONALITY THEOREM: Assume that a bandpass waveform

has a nonzero spectrum only over the frequency interval f1 6 | f| 6 f2, where the trans-
mission bandwidth BT is taken to be the absolute bandwidth given by BT = f2 - f1 and
BT f1. The waveform may be completely specified over a T0-second interval by
N = 2BTT0 (434)
independent pieces of information. N is said to be the number of dimensions required to

specify the waveform.
Computer simulation is often used to analyze communication systems. The bandpass

dimensionality theorem tells us that a bandpass signal BT Hz wide can be represented over a
T0-second interval, provided that at least N = 2BT T0 samples are used. More details about the
bandpass sampling theorem are discussed in study-aid Prob. SA45.
47 RECEIVED SIGNAL PLUS NOISE
Using the representation of bandpass signals and including the effects of channel filtering, we
can obtain a model for the received signal plus noise. Referring to Fig. 41, the signal out of
the transmitter is
s(t) = Re[g(t)ejvct]
where g(t) is the complex envelope for the particular type of modulation used. (See Table
41.) If the channel is linear and time invariant, the received signal plus noise is
r(t) = s(t) * h(t) + n(t) (435)
where h(t) is the impulse response of the channel and n(t) is the noise at the receiver input.
Furthermore, if the channel is distortionless, its transfer function is given by Eq. (429), and
consequently, the signal plus noise at the receiver input is
r(t) = Re[Ag(t - Tg)ej1vct + u(fc)2 + n(t)] (436)
where A is the gain of the channel (a positive number usually less than 1), Tg is the channel
group delay, and u(fc) is the carrier phase shift caused by the channel. In practice, the values for
Tg and u(fc) are often not known, so that if values for Tg and u(fc) are needed by the receiver to
detect the information that was transmitted, receiver circuits estimate the received carrier phase
u(fc) and the group delay (e.g., a bit synchronizer in the case of digital signaling). We will
assume that the receiver circuits are designed to make errors due to these effects negligible;
therefore, we can consider the signal plus noise at the receiver input to be
r(t) = Re[g(t)ejvct] + n(t) (437)
where the effects of channel filtering, if any, are included by some modification of the com-
plex envelope g(t) and the constant Ac that is implicit within g(t) (see Table 41) is adjusted to
reflect the effect of channel attenuation. Details of this approach are worked out in Sec. 86.
48 CLASSIFICATION OF FILTERS AND AMPLIFIERS
Filters
Filters are devices that take an input waveshape and modify the frequency spectrum to
produce the output waveshape. Filters may be classified in several ways. One is by the type of
construction used, such as LC elements or quartz crystal elements. Another is by the type of
Sec. 48 Classification of Filters and Amplifiers 255
transfer function that is realized, such as the Butterworth or Chebyshev response (defined
subsequently). These two classifications are discussed in this section.
Filters use energy storage elements to obtain frequency discrimination. In any physical
filter, the energy storage elements are imperfect. For example, a physical inductor has some
series resistance as well as inductance, and a physical capacitor has some shunt (leakage)
resistance as well as capacitance. A natural question, then, is, what is the quality Q of a circuit
element or filter? Unfortunately, two different measures of filter quality are used in the
technical literature. The first definition is concerned with the efficiency of energy storage in a
circuit [Ramo, Whinnery, and vanDuzer, 1967, 1984] and is
2p(maximum energy stored during one cycle)
Q = (438)
energy dissipated per cycle
A larger value for Q corresponds to a more perfect storage element. That is, a perfect L or C
element would have infinite Q. The second definition is concerned with the frequency selec-
tivity of a circuit and is
f0
Q = (439)
B
where f0 is the resonant frequency and B is the 3-dB bandwidth. Here, the larger the value of Q,
the better is the frequency selectivity, because, for a given f0, the bandwidth would be
smaller.
In general, the value of Q as evaluated using Eq. (438) is different from the value of Q
obtained from Eq. (439). However, these two definitions give identical values for an RLC
series resonant circuit driven by a voltage source or for an RLC parallel resonant circuit driven
by a current source [Nilsson, 1990]. For bandpass filtering applications, frequency selectivity
is the desired characteristic, so Eq. (439) is used. Also, Eq. (439) is easy to evaluate from
laboratory measurements. If we are designing a passive filter (not necessarily a single-tuned
circuit) of center frequency f0 and 3-dB bandwidth B, the individual circuit elements will each
need to have much larger Qs than f0B. Thus, for a practical filter design, we first need to
answer the question. What are the Qs needed for the filter elements, and what kind of
elements will give these values of Q? This question is answered in Table 42, which lists fil-
ters as classified by the type of energy storage elements used in their construction and gives
typical values for the Q of the elements. Filters that use lumped L and C elements become
impractical to build above 300 MHz, because the parasitic capacitance and inductance of the
leads significantly affect the frequency response at high frequencies. Active filters, which use
operational amplifiers with RC circuit elements, are practical only below 500 kHz, because
the operational amplifiers need to have a large open-loop gain over the operating band. For
very low-frequency filters, RC active filters are usually preferred to LC passive filters because
the size of the LC components becomes large and the Q of the inductors becomes small in this
frequency range. Active filters are difficult to implement within integrated circuits because
the resistors and capacitors take up a significant portion of the chip area. This difficulty is
reduced by using a switched-capacitor design for IC implementation. In that case, resistors

A lumped element is a discrete R-, L-, or C-type element, compared with a continuously distributed RLC
element, such as that found in a transmission line.
TABLE 42 FILTER CONSTRUCTION TECHNIQUES
Description of Center Frequency Unloaded Q Filter

Type of Construction Elements or Filter Range (Typical) Applicationa
LC (passive) dc300 MHz 100 Audio, video,

IF, and RF

Active and dc500 kHz 200b Audio
+
Switched Capacitor
Crystal Quartz 1 kHz100 MHz 100,000 IF

crystal
Transducers
Mechanical 50500 kHz 1,000 IF

Rod Disk
Ceramic disk
Ceramic 10 kHz10.7 MHz 1,000 IF
Electrodes
One section
Fingers
Surface acoustic 10800 MHz c IF and RF

waves (SAW)
Piezoelectric Finger
substrate overlap region
/4
Transmission line UHF and 1,000 RF
microwave
Cavity Microwave 10,000 RF
a
IF, intermediate frequency; RF, radio frequency. (See Sec. 416.)
b
Bandpass Qs.
c
Depends on design: N = f0B, where N is the number of sections, f0 is the center frequency, and B is the bandwidth. Loaded
Qs of 18,000 have been achieved.
Sec. 48 Classification of Filters and Amplifiers 257
are replaced by an arrangement of electronic switches and capacitors that are controlled by a
digital clock signal [Schaumann et al., 1990].
Crystal filters are manufactured from quartz crystal elements, which act as a series resonant
circuit in parallel with a shunt capacitance caused by the holder (mounting). Thus, a parallel
resonant, as well as a series resonant, mode of operation is possible. Above 100 MHz the quartz
element becomes physically too small to manufacture, and below 1 kHz the size of the element
becomes prohibitively large. Crystal filters have excellent performance because of the inherently
high Q of the elements, but they are more expensive than LC and ceramic filters.
Mechanical filters use the vibrations of a resonant mechanical system to obtain the
filtering action. The mechanical system usually consists of a series of disks spaced along a
rod. Transducers are mounted on each end of the rod to convert the electrical signals to
mechanical vibrations at the input, and vice versa at the output. Each disk is the mechanical
equivalent of a high-Q electrical parallel resonant circuit. The mechanical filter usually has a
high insertion loss, resulting from the inefficiency of the input and output transducers.
Ceramic filters are constructed from piezoelectric ceramic disks with plated electrode
connections on opposite sides of the disk. The behavior of the ceramic element is similar to
that of the crystal filter element, as discussed earlier, except that the Q of the ceramic element
is much lower. The advantage of the ceramic filter is that it often provides adequate perfor-
mance at a cost that is low compared with that of crystal filters.
Surface acoustic wave (SAW) filters utilize acoustic waves that are launched and travel on
the surface of a piezoelectric substrate (slab). Metallic interleaved fingers have been deposited
on the substrate. The voltage signal on the fingers is converted to an acoustic signal (and vice
versa) as the result of the piezoelectric effect. The geometry of the fingers determines the
frequency response of the filter, as well as providing the input and output coupling [Dorf, 1993,
pp. 10731074]. The insertion loss is somewhat larger than that for crystal or ceramic filters.
However, the ease of shaping the transfer function and the wide bandwidth that can be obtained
with controlled attenuation characteristics make the SAW filters very attractive. This technology
is used to provide excellent IF amplifier characteristics in modern television sets.
SAW devices can also be tapped so that they are useful for transversal filter configura-
tions (Fig. 328) operating in the RF range. At lower frequencies, charge transfer devices
(CTDs) can be used to implement transversal filters [Gersho, 1975].
Transmission line filters utilize the resonant properties of open-circuited or short-
circuited transmission lines. These filters are useful at UHF and microwave frequencies, at
which wavelengths are small enough so that filters of reasonable size can be constructed.
Similarly, the resonant effect of cavities is useful in building filters for microwave frequencies
at which the cavity size becomes small enough to be practical.
Filters are also characterized by the type of transfer function that is realized. The trans-
fer function of a linear filter with lumped circuit elements may be written as the ratio of two
polynomials,
b0 + b1(jv) + b2(jv)2 + + bk(jv)k
H(f) = (440)
a0 + a1(jv) + a2(jv)2 + + an(jv)n
where the constants ai and bi are functions of the element values and v = 2pf. The parame-
ter n is said to be the order of the filter. By adjusting the constants to certain values,
desirable transfer function characteristics can be obtained. Table 43 lists three different
TABLE 43 SOME FILTER CHARACTERISTICS
Transfer Characteristic for the

Type Optimization Criterion Low-Pass Filtera
Butterworth Maximally flat: as many 1

derivatives of |H(f)| as |H(f)| =
21 + (f/fb)2n
possible go to zero
as f 0
Chebyshev For a given peak-to-peak 1
ripple in the passband of |H(f)| =
21 + e2C2n(f/fb)
the |H(f)| characteristic, the
|H(f)| attenuates the fastest e = a design constant; Cn (f) is the nth-order
for any filter of nth order Chebyshev polynomial defined by the recursion
relation Cn(x) = 2xCn-1(x) - Cn-2(x),
where C0(x) = 1 and C1(x) = x
Bessel Attempts to maintain Kn
linear phase in the H(f)
Bn(f/fb)
passband
Kn is a constant chosen to make H (0) = 1,
and the Bessel recursion relation is
Bn(x) = (2n - 1) Bn-1(x) - x2Bn-2(x),
where B0(x) = 1 and B1(x) = 1 + jx
a
fb is the cutoff frequency of the filter.
filter characteristics and the optimization criterion that defines each one. The Chebyshev
filter is used when a sharp attenuation characteristic is required for a minimum number of
circuit elements. The Bessel filter is often used in data transmission when the pulse shape is
to be preserved, since it attempts to maintain a linear phase response in the passband.
The Butterworth filter is often used as a compromise between the Chebyshev and Bessel
characteristics.
The topic of filters is immense, and not all aspects of filtering can be covered here. For
example, with the advent of inexpensive microprocessors, digital filtering and digital signal
processing are becoming very important [Oppenheim and Schafer, 1975, 1989].
For additional reading on analog filters with an emphasis on communication system
applications, see Bowron and Stephenson [1979].
Amplifiers
For analysis purposes, electronic circuits and, more specifically, amplifiers can be classified
into two main categories: Nonlinear and linear. Linearity was defined in Sec. 26. In practice,
all circuits are nonlinear to some degree, even at low (voltage and current) signal levels, and
become highly nonlinear for high signal levels. Linear circuit analysis is often used for the
low signal levels, since it greatly simplifies the mathematics and gives accurate answers if the
signal level is sufficiently small.
Sec. 49 Nonlinear Distortion 259
The main categories of nonlinear and linear amplifiers can be further classified into the
subcategories of circuits with memory and circuits with no memory. Circuits with memory
contain inductive and capacitive effects that cause the present output value to be a function of
previous input values as well as the present input value. If a circuit has no memory, its present
output value is a function only of its present input value.
In introductory electrical engineering courses, it is first assumed that circuits are linear
with no memory (resistive circuits) and, later, linear with memory (RLC circuits). It follows
that linear amplifiers with memory may be described by a transfer function that is the ratio of
the Fourier transform of the output signal to the Fourier transform of the input signal. As dis-
cussed in Sec. 26, the transfer function of a distortionless amplifier is given by Ke-jvcTd,
where K is the voltage gain of the amplifier and Td is the delay between the output and input
waveforms. If the transfer function of the linear amplifier is not of this form, the output signal
will be a linearly distorted version of the input signal.
49 NONLINEAR DISTORTION
In addition to linear distortion, practical amplifiers produce nonlinear distortion. To examine

the effects of nonlinearity and yet keep a mathematical model that is tractable, we will assume
no memory in the following analysis. Thus, we will look at the present output as a function of
the present input in the time domain. If the amplifier is linear, this relationship is
v0(t) = Kvi(t) (441)
where K is the voltage gain of the amplifier. In practice, the output of the amplifier becomes sat-
urated at some value as the amplitude of the input signal is increased. This is illustrated by the
nonlinear output-to-input characteristic shown in Fig. 45. The output-to-input characteristic may
be modeled by a Taylors expansion about vi = 0 (i.e., a Maclaurin series); that is,
q
v0 = K0 + K1vi + K2v2i + = a Knvni (442)
n=0
where
1 dnv0
Kn = ` a nb ` (443)
n! dvi vi = 0
There will be nonlinear distortion on the output signal if K2, K3, ... are not zero. K0 is the
output DC offset level, K1vi is the first-order (linear) term, K2v2i is the second-order (square-
law) term, and so on. Of course, K1 will be larger than K2, K3, ... if the amplifier is anywhere
near to being linear.
The harmonic distortion associated with the amplifier output is determined by
applying a single sinusoidal test tone to the amplifier input. Let the input test tone be
represented by
vi(t) = A0 sin v0t (444)

v0 output
Saturation
level
vi input
Figure 45 Nonlinear amplifier output-to-input characteristic.
Then the second-order output term is
K2A20
K2(A0 sin v0t)2 = (1 - cos 2v0t) (445)
2
This indicates that the second-order distortion creates a DC level K2A0>2 (in addition to any
2
DC bias) and second harmonic distortion with amplitude K2A20>2. In general, for a single-tone
input, the output will be
vout(t) = V0 + V1 cos(v0t + w1) + V2 cos(2v0t + w2)

+ V3 cos(3v0t + w3) + (446)
where Vn is the peak value of the output at the frequency nf0 hertz. Then, the percentage of
total harmonic distortion (THD) is defined by
4 a n = 2V2n
q
THD (%) = * 100 (447)

V1
The THD of an amplifier can be measured by using a distortion analyzer, or it can be evalu-
ated by Eq. (447), with the Vns obtained from a spectrum analyzer.
The intermodulation distortion (IMD) of the amplifier is obtained by using a two-tone
test. If the input (tone) signals are
vi(t) = A1 sin v1t + A2 sin v2t (448)

then the second-order output term is
K2(A1 sin v1t + A2 sin v2t)2 = K2(A21 sin2 v1t + 2A1A2 sin v1t sin v2t + A22 sin2 v2t)
The first and last terms on the right side of this equation produce harmonic distortion at
frequencies 2f1 and 2f2. The cross-product term produces IMD. This term is present only
when both input terms are presentthus, the name intermodulation distortion. Then the
second-order IMD is
2K2A1A2 sin v1t sin v2t = K2A1A2{cos[(v1 - v2)t] - cos [1v1 + v22t]}
It is clear that IMD generates sum and difference frequencies.

The third-order term is
K3v3i = K3(A1 sin v1t + A2 sin v2t)3

= K3(A31 sin3 v1t + 3A21A2 sin2 v1t sin v2t
+ 3A1A22 sin v1t sin2 v2t + A32 sin3 v2t) (449)
The first and last terms on the right side of this equation will produce harmonic distortion, and
the second term, a cross product, becomes
3
3K3A21A2 sin2 v1t sin v2t = 2
2 K3A1A2 sin v2t(1 - cos 2v1t)
3
= 2
2 K3A1A2{sin v2t - 12 [sin(2v1 + v2)t
- sin(2v1 - v2)t]} (450)
Similarly, the third term of Eq. (449) is
3K3A1A22 sin v1t sin2v2t

= 3
2 K3A1A225sin v1t - 12[ sin12v2 + v12t - sin(2v2 - v1)t]6 (451)
The last two terms in Eqs. (450) and (451) are intermodulation terms at nonharmonic fre-
quencies. For the case of bandpass amplifiers where f1 and f2 are within the bandpass with f1
close to f2 (i.e., f1 L f2 0), the distortion products at 2f1 + f2 and 2f2 + f1 will usually fall
outside the passband and, consequently, may not be a problem. However, the terms at 2f1 - f2
and 2f2 - f1 will fall inside the passband and will be close to the desired frequencies f1 and f2.
These will be the main distortion products for bandpass amplifiers, such as those used for RF
amplification in transmitters and receivers.
As Eqs. (450) and (451) show, if either A1 or A2 is increased sufficiently, the IMD
will become significant, since the desired output varies linearly with A1 or A2 and the IMD
output varies as A21A2 or A1A22. Of course, the exact input level required for the intermodula-
tion products to be a problem depends on the relative values of K3 and K1. The level may be
specified by the amplifier third-order intercept point, which is evaluated by applying two
equal amplitude test tones (i.e., A1 = A2 = A). The desired linearly amplified outputs will have
amplitudes of K1 A, and each of the third-order intermodulation products will have amplitudes
of 3K3 A34. The ratio of the desired output to the IMD output is then
a b
4 K1
RIMD = (452)
3 K3A2
The input intercept point, defined as the input level that causes RIMD to be unity, is shown in
Fig. 46. The solid curves are obtained by measurement, using two sinusoidal signal genera-
tors to generate the tones and measuring the level of the desired output (at f1 or f2) and the
IMD products (at 2f1 - f2 or 2f2 - f1) with a spectrum analyzer. The intercept point is a
fictitious point that is obtained by extrapolation of the linear portion (decibel plot) of the
desired output and IMD curves until they intersect. The desired output (the output at either f1
or f2) actually becomes saturated when measurements are made, since the higher-order terms
in the Taylor series have components at f1 and f2 that subtract from the linearly amplified
output. For example, with K3 being negative, the leading term in Eq. (451) occurs at f1 and
40
30
Intercept
20 point
3 dB
10
0 3-dB compression levels

RF output power (dBm)
10
20
45 dB
30
40 Desired output Third-order

intermodulation
product
50 (two-tone test)
60
70
90 80 70 60 50 40 30 20 10 0 10
RF input power (dBm)
Figure 46 Amplifier output characteristics.
will subtract from the linearly amplified component at f1, thus producing a saturated
characteristic for the sinusoidal component at f1. For an amplifier that happens to have the
particular nonlinear characteristic shown in Fig. 46, the intercept point occurs for an RF
input level of -10 dBm. Overload characteristics of receivers, such as those used in police
walkie-talkies, are characterized by the third-order intercept-point specification. This dBm
value is the RF signal level at the antenna input that corresponds to the intercept point. When
the receiver is deployed, input signal levels need to be much lower than that value in order to
keep the undesired interfering intermodulation signals generated by the receiver circuits to an
acceptable level. For transmitter applications, the intercept-point specification is the output
signal level corresponding to the intercept point.
Other properties of an amplifier are also illustrated by Fig. 46. The gain of the ampli-
fier is 25 dB in the linear region, because a -60- dBm input produces a -35- dBm output
level. The desired output is compressed by 3 dB for an input level of -15 dBm. Consequently,
the amplifier might be considered to be linear only if the input level is less than -15 dBm.
Furthermore, if the third-order IMD products are to be down by at least 45 dBm, the input
level will have to be kept lower than -32 dBm.
Another term in the distortion products at the output of a nonlinear amplifier is called
cross-modulation. Cross-modulation terms are obtained when one examines the third-order
products resulting from a two-tone test. As shown in Eqs. (450) and (451), the terms
3 2 3 2
2 K3A1A2sinv2t and 2 K3A1A2 sin v1t are cross-modulation terms. Let us examine the term
3 2 If we allow some amplitude variation in the input signal A1 sin v1t, so that it
2 K3A1A2sinv2t.
looks like an AM signal A1[1 + m1(t)] sin v1t, where m1(t) is the modulating signal, a third-
order distortion product becomes
3
2 K3A21A2[1 + m1(t)]2 sin v2t (453)
Thus, the AM on the signal at the carrier frequency f1 will produce a signal at frequency f2
with distorted modulation. That is, if two signals are passed through an amplifier having third-
order distortion products in the output, and if either input signal has some AM, the amplified
output of the other signal will be amplitude modulated to some extent by a distorted version
of the modulation. This phenomenon is cross-modulation.
Passive as well as active circuits may have nonlinear characteristics and, consequently,
will produce distortion products. For example, suppose that two AM broadcast stations have
strong signals in the vicinity of a barn or house that has a metal roof with rusted joints. The roof
may act as an antenna to receive and reradiate the RF energy, and the rusted joints may act as a
diode (a nonlinear passive circuit). Signals at harmonics and intermodulation frequencies may
be radiated and interfere with other communication signals. In addition, cross-modulation
products may be radiated. That is, a distorted modulation of one station is heard on radios
(located in the vicinity of the rusted roof) that are tuned to the other stations frequency.
When amplifiers are used to produce high-power signals, as in transmitters, it is desir-
able to have amplifiers with high efficiency in order to reduce the costs of power supplies,
cooling equipment, and energy consumed. The efficiency is the ratio of the output signal
power to the DC input power. Amplifiers may be grouped into several categories, depending
on the biasing levels and circuit configurations used. Some of these are Class A, B, C, D, E, F,
G, H, and S [Krauss, Bostian, and Raab, 1980; Smith, 1998]. For Class A operation, the bias
on the amplifier stage is adjusted so that current flows during the complete cycle of an applied
input test tone. For Class B operation, the amplifier is biased so that current flows for 180 of
the applied signal cycle. Therefore, if a Class B amplifier is to be used for a baseband linear
amplifier, such as an audio power amplifier in a hi-fi system, two devices are wired in push-
pull configuration so that each one alternately conducts current over half of the input signal
cycle. In bandpass Class B linear amplification, where the bandwidth is a small percentage of
the operating frequency, only one active device is needed, since tuned circuits may be used to
supply the output current over the other half of the signal cycle. For Class C operation, the
bias is set so that (collector or plate) current flows in pulses, each one having a pulse width
that is usually much less than half of the input cycle. Unfortunately, with Class C operation, it
is not possible to have linear amplification, even if the amplifier is a bandpass RF amplifier
with tuned circuits providing current over the nonconducting portion of the cycle. If one tries
to amplify an AM signal with a Class C amplifier or other types of nonlinear amplifiers, the
AM on the output will be distorted. However, RF signals with a constant real envelope, such
as FM signals, may be amplified without distortion, because a nonlinear amplifier preserves
the zero-crossings of the input signal.
The efficiency of a Class C amplifier is determined essentially by the conduction angle
of the active device, since poor efficiency is caused by signal power being wasted in the
device itself during the conduction time. The Class C amplifier is most efficient, having an
efficiency factor of 100% in the ideal case. Class B amplifiers have an efficiency of
p>4 * 100 = 78.5% or less, and Class A amplifiers have an efficiency of 50% or less
[Krauss, Bostian, and Raab, 1980]. Because Class C amplifiers are the most efficient, they are
generally used to amplify constant envelope signals, such as FM signals used in broadcasting.
Class D, E, F, G, H, and S amplifiers usually employ switching techniques in specialized
circuits to obtain high efficiency [Krauss, Bostian, and Raab, 1980; Smith, 1998].
Many types of microwave amplifiers, such as traveling-wave tubes (TWTs), operate on
the velocity modulation principle. The input microwave signal is fed into a slow-wave struc-
ture. Here, the velocity of propagation of the microwave signal is reduced so that it is slightly
below the velocity of the DC electron beam. This enables a transfer of kinetic energy from the
electron beam to the microwave signal, thereby amplifying the signal. In this type of amplifier,
the electron current is not turned on and off to provide the amplifying mechanism; thus, it is not
classified in terms of Class B or C operation. The TWT is a linear amplifier when operated at
the appropriate drive level. If the drive level is increased, the efficiency (RF outputDC input)
is improved, but the amplifier becomes nonlinear. In this case, constant envelope signals, such
as PSK or PM, need to be used so that the intermodulation distortion will not cause a problem.
This is often the mode of operation of satellite transponders (transmitters in communication
satellites), where solar cells are costly and have limited power output. The subject is discussed
in more detail in Chapter 8 in the section on satellite communications.
410 LIMITERS
A limiter is a nonlinear circuit with an output saturation characteristic. A soft saturating limiter
characteristic is shown in Fig. 45. Figure 47 shows a hard (ideal) limiter characteristic,
together with an illustration of the unfiltered output waveform obtained for an input waveform.
The ideal limiter transfer function is essentially identical to the output-to-input characteristic of
Sec. 410 Limiters 265
Ideal limiter vlim(t)

vout characteristic
VL VL
vin
VL
t
vin(t)
Figure 4 7 Ideal limiter characteristic with illustrative input and unfiltered output waveforms.
an ideal comparator with a zero reference level. The waveforms shown in Fig. 47 illustrate
how amplitude variations in the input signal are eliminated in the output signal. A bandpass
limiter is a nonlinear circuit with a saturating characteristic followed by a bandpass filter. In the
case of an ideal bandpass limiter, the filter output waveform would be sinusoidal, since the
harmonics of the square wave would be filtered out. In general, any bandpass input (even a
modulated signal plus noise) can be represented, using Eq. (41b), by
vin(t) = R(t) cos[vct + u(t)] (454)
where R(t) is the equivalent real envelope and u(t) is the equivalent phase function. The
corresponding output of an ideal bandpass limiter becomes
vout(t) = KVL cos[vct + u(t)] (455)
where K is the level of the fundamental component of the square wave, 4p, multiplied by the
gain of the output (bandpass) filter. This equation indicates that any AM that was present on
the limiter input does not appear on the limiter output, but that the phase function is preserved
(i.e., the zero-crossings of the input are preserved on the limiter output). Limiters are often
used in receiving systems designed for angle-modulated signalingsuch as PSK, FSK, and
analog FMto eliminate any variations in the real envelope of the receiver input signal that
are caused by channel noise or signal fading.
411 MIXERS, UP CONVERTERS, AND DOWN CONVERTERS
An ideal mixer is an electronic circuit that functions as a mathematical multiplier of two input
signals. Usually, one of these signals is a sinusoidal waveform produced by a local oscillator,
as illustrated in Fig. 48.
Mixers are used to obtain frequency translation of the input signal. Assume that the
input signal is a bandpass signal that has a nonzero spectrum in a band around or near f = fc.
Then the signal is represented by
vin(t) = Re{gin(t) ejvct} (456)
where gin(t) is the complex envelope of the input signal. The signal out of the ideal mixer is then
v1(t) = [A0 Re{gin(t)ejvct}] cos v0t
[g (t)ejvct + gin* (t)e-jvct]1ejv0t + e-jv0t2

A0
=
4 in
A0
= [gin(t) ej(vc + v0)t + gin* (t) e-j(vc + v0)t + gin(t) ej(vc-v0)t + gin* (t) e-j(vc - v0)t]
4
or
A0 A0
v1(t) = Re{gin(t)ej(vc + v0)t} + Re{gin(t)ej(vc - v0)t} (457)
2 2
Equation (457) illustrates that the input bandpass signal with a spectrum near f = fc
has been converted (i.e., frequency translated) into two output bandpass signals, one at the
Mixer
vin(t) v1(t) v2(t)
Filter
vLO(t)=A0 cos (0 t)
Local
oscillator
Figure 48 Mixer followed by a filter for either up or down conversion.

Sec. 411 Mixers, Up Converters, and Down Converters 267
up-conversion frequency band, where fu = fc + f0, and one at the down-conversion band,
where fd = fc - f0. A filter, as illustrated in Fig. 48, may be used to select either the up-
conversion component or the down-conversion component. This combination of a mixer
plus a filter to remove one of the mixer output components is often called a single-side-
band mixer. A bandpass filter is used to select the up-conversion component, but the
down-conversion component is selected by either a baseband filter or a bandpass filter,
depending on the location of fc - f0. For example, if fc - f0 = 0, a low-pass filter would be
needed, and the resulting output spectrum would be a baseband spectrum. If fc - f0 7 0,
where fc - f0 was larger than the bandwidth of gin(t), a bandpass filter would be used, and
the filter output would be
A0
v2(t) = Re{g2(t)ej(vc - v0)t} = Re{gin(t)ej(vc - v0)t} (458)
2
For this case of fc 7 f0, it is seen that the modulation on the mixer input signal vin(t) is pre-
served on the mixer up- or down-converted signals.
If fc 6 f0, we rewrite Eq. (457), obtaining
A0 A0
v1(t) = Re{gin(t)ej(vc + v0)t} + Re{gin* (t)ej(v0 - vc)t} (459)
2 2
because the frequency in the exponent of the bandpass signal representation needs to be
positive for easy physical interpretation of the location of spectral components. For this case
of fc 6 f0, the complex envelope of the down-converted signal has been conjugated compared
to the complex envelope of the input signal. This is equivalent to saying that the sidebands
have been exchanged; that is, the upper sideband of the input signal spectrum becomes the
lower sideband of the down-converted output signal, and so on. This is demonstrated mathe-
matically by looking at the spectrum of g*(t), which is
q q *
[gin* (t)] = gin* (t)e-jvt dt = c gin(t)e-j(-v)t dt d
L-q L-q
= Gin* (-f) (460)
The -f indicates that the upper and lower sidebands have been exchanged, and the conjugate
indicates that the phase spectrum has been inverted.
In summary, the complex envelope for the signal out of an up converter is
A0
g2(t) = gin(t) (461a)
2
where fu = fc + f0 7 0. Thus, the same modulation is on the output signal as was on the input
signal, but the amplitude has been changed by the A02 scale factor.
For the case of down conversion, there are two possibilities. For fd = fc - f0 7 0, where
f0 6 fc,
A0
g2(t) = g (t) (461b)
2 in
This is called down conversion with low-side injection, because the LO frequency is below
that of the incoming signal (i.e., f0 6 fc). Here the output modulation is the same as that of the
input, except for the A02 scale factor. The other possibility is fd = f0 - fc 7 0, where f0 7 fc ,
which produces the output complex envelope
A0
g2 = g * (t) (461c)
2 in
This is down conversion with high-side injection, because f0 7 fc. Here the sidebands on the
down-converted output signal are reversed from those on the input (e.g., an LSSB input signal
becomes a USSB output signal).
Ideal mixers act as linear time-varying circuit elements, since
v1(t) = (A cos v0t)vin(t)
where A cos v0t is the time-varying gain of the linear circuit. It should also be recognized that
mixers used in communication circuits are essentially mathematical multipliers. They should
not be confused with the audio mixers that are used in radio and TV broadcasting studios. An
audio mixer is a summing amplifier with multiple inputs so that several inputs from several
sourcessuch as microphones, tape decks, and CD deckscan be mixed (added) to
produce one output signal. Unfortunately, the term mixer means entirely different things,
depending on the context used. As used in transmitters and receivers, it means a multiplying
operation that produces a frequency translation of the input signal. In audio systems, it means
a summing operation to combine several inputs into one output signal.
In practice, the multiplying operation needed for mixers may be realized by using one
of the following:
1. A continuously variable transconductance device, such as a dual-gate FET.

2. A nonlinear device.
3. A linear device with a time-varying discrete gain.
In the first method, when a dual-gate FET is used to obtain multiplication, vin(t) is usually
connected to gate 1 and the local oscillator is connected to gate 2. The resulting output is
v1(t) = Kyin(t)vLO(t) (462)
over the operative region, where vLO(t) is the local oscillator voltage. The multiplier is
said to be of a single-quadrant type if the multiplier action of Eq. (462) is obtained only
when both input waveforms, vin and vLO(t), have either nonnegative or nonpositive values
[i.e., a plot of the values of vin(t) versus vLO(t) falls within a single quadrant]. The multi-
plier is of the two-quadrant type if multiplier action is obtained when either vin (t) or
vLO(t) is nonnegative or nonpositive and the other is arbitrary. The multiplier is said to be
of the four-quadrant type when multiplier action is obtained regardless of the signs of
vin(t) and vLO(t).
In the second technique, a nonlinear device can be used to obtain multiplication by
summing the two inputs as illustrated in Fig. 49. Looking at the square-law component at the
output, we have
vin(t)
+
v1(t) vout(t)
Nonlinear
device Filter
+
vLO(t)=A0 cos (0 t)
Local
oscillator
Figure 49 Nonlinear device used as a mixer.
v1(t) = K2(vin + vLO)2 + other terms

= K21v2in + 2vinvLO + v2LO2 + other terms (463)
The cross-product term gives the desired multiplier action:
2K2vinvLO = 2K2A0vin(t) cos v0t (464)
If we assume that vin(t) is a bandpass signal, the filter can be used to pass either the up- or
down-conversion terms. However, some distortion products may also fall within the output
passband if vc and v0 are not chosen carefully.
In the third method, a linear device with time-varying gain is used to obtain multiplier
action. This is demonstrated in Fig. 410, in which the time-varying device is an analog
switch (such as a CMOS 4016 integrated circuit) that is activated by a square-wave oscillator
signal v0(t). The gain of the switch is either unity or zero. The waveform at the output of the
analog switch is
v1(t) = vin(t)s(t) (465)
Analog switch
(a linear time-varying device)
vin(t) v1(t) vout(t)

Filter
s(t)
Multivibrator
(a square-wave oscillator)
Figure 410 Linear time-varying device used as a mixer.

where s(t) is a unipolar switching square wave that has unity peak amplitude. (This is analogous
to the PAM with natural sampling that was studied in Chapter 3.) Using the Fourier series for a
rectangular wave, we find that Eq. (465) becomes
cos nv0t d
q 2 sin (np/2)
1
v1(t) = vin(t)c + a (466)
2 n=1 np
The multiplying action is obtained from the n = 1 term, which is
2
v (t) cos v0t (467)
p in
This term would generate up- and down-conversion signals at fc + f0 and fc - f0 if vin(t) were
a bandpass signal with a nonzero spectrum in the vicinity of f = fc. However, Eq. (466) shows
that other frequency bands are also present in the output signal, namely, at frequencies f =
| fc nf0|, n = 3, 5, 7, ... and, in addition, there is the feed-through term 12vin(t) appearing at the
output. Of course, a filter may be used to pass either the up- or down-conversion component
appearing in Eq. (466).
Mixers are often classified as being unbalanced, single balanced, or double balanced.
That is, in general, we obtain
v1(t) = C1vin(t) + C2v0(t) + C3vin(t)v0(t) + other terms (468)
at the output of mixer circuits. When C1 and C2 are not zero, the mixer is said to be
unbalanced, since vin(t) and v0(t) feed through to the output. An unbalanced mixer was illus-
trated in Fig. 49, in which a nonlinear device was used to obtain mixing action. In the
Taylors expansion of the nonlinear device output-to-input characteristics, the linear term
would provide feed-through of both vin(t) and v0(t). A single-balanced mixer has feed-
through for only one of the inputs; that is, either C1 or C2 of Eq. (468) is zero. An example
of a single-balanced mixer is given in Fig. 410, which uses sampling to obtain mixer action.
In this example, Eq. (466) demonstrates that v0(t) is balanced out (i.e., C2 = 0) and vin(t)
feeds through with a gain of C1 = 12 . A double-balanced mixer has no feed-through from
either input; that is, both C1 and C2 of Eq. (468) are zero. One kind of double-balanced mixer
is discussed in the next paragraph.
Figure 411a shows the circuit for a double-balanced mixer. This circuit is popular
because it is relatively inexpensive and has excellent performance. The third-order IMD is
typically down at least 50 dB compared with the desired output components. This mixer is
usually designed for source and load impedances of 50 and has broadband input and output
ports. The RF [i.e., vin(t)] port and the LO (local oscillator) port are often usable over a
frequency range of 1,000:1, say, 1 to 1,000 MHz; and the IF (intermediate frequency) output
port, v1(t), is typically usable from DC to 600 MHz. The transformers are made by using
small toroidal cores, and the diodes are matched hot carrier diodes. The input signal level at
the RF port is relatively small, usually less than -5 dBm, and the local oscillator level at the
LO port is relatively large, say, +5 dBm. The LO signal is large, and, in effect, turns the diodes
on and off so that the diodes will act as switches. The LO provides the switching control
+ +
RF LO Local
vin vLO oscillator
port port
IF port
+
v1(t)
Load
(a) A Double-Balanced Mixer Circuit
vin(t) vLO(t) vin(t) vLO(t)
v1(t) v1(t)
(b) Equivalent Circuit When vLO(t) Is Positive (c) Equivalent Circuit When vLO(t) Is Negative
s(t)
+1
1 t
T0
(d) Switching Waveform Due to the Local Oscillator Signal
Figure 411 Analysis of a double-balanced mixer circuit.
signal. This circuit thus acts as a time-varying linear circuit (with respect to the RF input
port), and its analysis is very similar to that used for the analog-switch mixer of Fig. 410.
During the portion of the cycle when vLO(t) has a positive voltage, the output voltage is pro-
portional to +vin(t), as seen from the equivalent circuit shown in Fig. 411b. When vLO(t) is
negative, the output voltage is proportional to -vin(t), as seen from the equivalent circuit
shown in Fig. 411c. Thus, the output of this double-balanced mixer is
v1(t) = Kvin(t) s(t) (469)

where s(t) is a bipolar switching waveform, as shown in Fig. 411d. Since the switching waveform
arises from the LO signal, its period is T0 = 1f0. The switching waveform is described by
q sin(np/2)
s(t) = 4 a cos nv0t (470)
n=1 np
so that the mixer output is

q sin(np/2)
v1(t) = [vin(t)]c4K a cos nv0td (471)
n=1 np
This equation shows that if the input is a bandpass signal with nonzero spectrum in the vicin-
ity of fc, the spectrum of the input will be translated to the frequencies |fc nf0|, where n = 1,
3, 5, . . . . In practice, the value K is such that the conversion gain (which is defined as the
desired output level divided by the input level) at the frequency |fc f0| is about -6 dB.
Of course, an output filter may be used to select the up-converted or down-converted
frequency band.
In addition to up- or down-conversion applications, mixers (i.e., multipliers) may be
used for amplitude modulators to translate a baseband signal to an RF frequency band, and
mixers may be used as product detectors to translate RF signals to baseband. These applica-
tions will be discussed in later sections that deal with transmitters and receivers.
412 FREQUENCY MULTIPLIERS
Frequency multipliers consist of a nonlinear circuit followed by a tuned circuit, as illustrated

in Fig. 412. If a bandpass signal is fed into a frequency multiplier, the output will appear in
a frequency band at the nth harmonic of the input carrier frequency. Because the device is
nonlinear, the bandwidth of the nth harmonic output is larger than that of the input signal. In
general, the bandpass input signal is represented by
vin(t) = R(t) cos[vct + u(t)] (472)
The transfer function of the nonlinear device may be expanded in a Taylors series, so that the
nth-order output term is
v1(t) = Knvnin(t) = KnRn(t) cosn[vct + u(t)]
or
v1(t) = CRn(t) cos[nvct + nu(t)] + other terms

mth-order output terms, where m 7 n, may also contribute to the nth harmonic output, provided that Km
is sufficiently large with respect to Kn. This condition is illustrated by the trigonometric identity 8cos4x = 3 + 4
cos 2x + cos 4x, in which m = 4 and n = 2.
Sec. 412 Frequency Multipliers 273
Frequency multiplier
vin(t) v1(t) vout(t)

Nonlinear Bandpass
device filter
(a) Block Diagram of a Frequency Multiplier
Vcc
Bandpass filter
(a circuit tuned to nfc)
Biased in
nonlinear region
Frequency out=n fc
Frequency in=fc
(b) Circuit Diagram of a Frequency Multiplier
Figure 412 Frequency multiplier.
Because the bandpass filter is designed to pass frequencies in the vicinity of nfc, the output is
vout(t) = CRn(t) cos [nvct + nu (t)] (473)
This illustrates that the input amplitude variation R(t) appears distorted on the output signal
because the real envelope on the output is Rn(t). The waveshape of the angle variation, u(t), is not
distorted by the frequency multiplier, but the frequency multiplier does increase the magnitude of
the angle variation by a factor of n. Thus, frequency multiplier circuits are not used on signals if
AM is to be preserved; but as we will see, the frequency multiplier is very useful in PM and FM
problems, since it effectively amplifies the angle variation waveform u(t). The n = 2 multiplier
is called a doubler stage, and the n = 3 frequency multiplier is said to be a tripler stage.
The frequency multiplier should not be confused with a mixer. The frequency multiplier
acts as a nonlinear device. The mixer circuit (which uses a mathematical multiplier operation)
acts as a linear circuit with time-varying gain (caused by the LO signal). The bandwidth of the
signal at the output of a frequency multiplier is larger than that of the input signal, and it ap-
pears in a frequency band located at the nth harmonic of the input. The bandwidth of a signal
at the output of a mixer is the same as that of the input, but the input spectrum has been trans-
lated either up or down, depending on the LO frequency and the bandpass of the output filter.
A frequency multiplier is essentially a nonlinear amplifier followed by a bandpass filter that is
designed to pass the nth harmonic.
413 DETECTOR CIRCUITS
As indicated in Fig. 41, the receiver contains carrier circuits that convert the input bandpass
waveform into an output baseband waveform. These carrier circuits are called detector cir-
cuits. The sections that follow will show how detector circuits can be designed to produce
R(t), u(t), x(t), or y(t) at their output for the corresponding bandpass signal that is fed into the
detector input.
Envelope Detector
An ideal envelope detector is a circuit that produces a waveform at its output that is propor-
tional to the real envelope R(t) of its input. From Eq. (41b), the bandpass input may be
represented by R(t) cos[vct + u(t)], where R(t) 0; then the output of the ideal envelope
detector is
vout(t) = KR(t) (474)
where K is the proportionality constant.
A simple diode detector circuit that approximates an ideal envelope detector is shown in
Fig. 413a. The diode current occurs in pulses that are proportional to the positive part of the
input waveform. The current pulses charge the capacitor to produce the output voltage wave-
form, as illustrated in Fig. 413b. The RC time constant is chosen so that the output signal
will follow the real envelope R(t) of the input signal. Consequently, the cutoff frequency of
vin(t) C R vout(t)
(a) A Diode Envelope Detector
Input signal, vin(t)

Output signal, vout(t)
(b) Waveforms Associated with the Diode Envelope Detector
Figure 413 Envelope detector.

Sec. 413 Detector Circuits 275
the low-pass filter needs to be much smaller than the carrier frequency fc and much larger than
the bandwidth of the (detected) modulation waveform B. That is,
1
B fc (475)
2pRC
where RC is the time constant of the filter.
The envelope detector is typically used to detect the modulation on AM signals. In this
case, vin(t) has the complex envelope g(t) = Ac[1 + m(t)], where Ac 7 0 represents the strength
of the received AM signal and m(t) is the modulation. If |m(t)| 6 1, then
vout = KR(t) = K|g(t)| = KAc[1 + m(t)] = KAc + KAcm(t)
KAc is a DC voltage that is used to provide automatic gain control (AGC) for the AM
receiver. That is, for KAc relatively small (a weak AM signal received), the receiver gain is
increased and vice versa. KAcm(t) is the detected modulation. For the case of audio (not
video) modulation, typical values for the components of the envelope detector are R = 10 k
and C = 0.001 fd. This combination of values provides a low-pass filter cutoff frequency
(3 dB down) of fco = 1(2pRC) = 15.9 kHz, much less than fc and larger than the highest
audio frequency, B, used in typical AM applications.
Product Detector
A product detector (Fig. 414) is a mixer circuit that down-converts the input (bandpass
signal plus noise) to baseband. The output of the multiplier is
v1(t) = R(t) cos[vct + u(t)]A0 cos(vct + u0)

1 1
= 2 A 0R(t) cos[u(t) - u0] + 2 A 0R(t) cos[2vct + u(t) + u0]
where the frequency of the oscillator is fc and the phase is u0. The low-pass filter passes only
the down-conversion term, so that the output is
1 1 -ju0
vout(t) = 2 A0R(t) cos[u(t) - u0] = 2 A0Re{g(t)e } (476)
1
vin(t)=R(t) cos[c t +(t)] vout(t)= A0 Re[g(t) ej0]
v1(t) 2
Low-pass
or filter
vin(t)=Re[g(t) ejc t]
where g(t)=R(t) ej(t)
v0(t)=A0 cos[c t +0]
Oscillator
Figure 414 Product detector.

where the complex envelope of the input is denoted by
g(t) = R(t)eju(t) = x(t) + jy(t)
and x(t) and y(t) are the quadrature components. [See Eq. (42).] Because the frequency of the
oscillator is the same as the carrier frequency of the incoming signal, the oscillator has been
frequency synchronized with the input signal. Furthermore, if, in addition, u0 = 0, the oscilla-
tor is said to be phase synchronized with the in-phase component, and the output becomes
1
vout(t) = 2 A0x(t) (477a)
If u0 = 90,
1
vout = 2 A0y(t) (477b)
Equation (476) also indicates that a product detector is sensitive to AM and PM. For
example, if the input contains no angle modulation, so that u(t) = 0, and if the reference phase
is set to zero (i.e., u0 = 0), then
1
vout(t) = 2 A0R(t) (478a)
which implies that x(t) 0, and the real envelope is obtained on the product detector
output, just as in the case of the envelope detector discussed previously. However, if an
angle-modulated signal Accos [vct + u(t)] is present at the input and u0 = 90, the product
detector output is
1 j[u(t) - 90]
vout(t) = 2 A0 Re{Ace }
or
1
vout(t) = 2 A0Ac sin u(t) (478b)
In this case, the product detector acts like a phase detector with a sinusoidal characteristic,
because the output voltage is proportional to the sine of the phase difference between the
input signal and the oscillator signal. Phase detector circuits are also available that yield
triangle and sawtooth characteristics [Krauss, Bostian, and Raab, 1980]. Referring to
Eq. (478b) for the phase detector with a sinusoidal characteristic, and assuming that the
phase difference is small [i.e., |u(t)| p>2], we see that sin u(t) u(t) and
1
vout(t) L 2 A0Acu(t) (479)
which is a linear characteristic (for small angles). Thus, the output of this phase detector
is directly proportional to the phase differences when the difference angle is small. (See
Fig. 420a.)
The product detector acts as a linear time-varying device with respect to the input vin(t),
in contrast to the envelope detector, which is a nonlinear device. The property of being either
linear or nonlinear significantly affects the results when two or more components, such as a
signal plus noise, are applied to the input. This topic will be studied in Chapter 7.
Detectors may also be classified as being either coherent or noncoherent. A coherent

detector has two inputsone for a reference signal, such as the synchronized oscillator
signal, and one for the modulated signal that is to be demodulated. The product detector is an
example of a coherent detector. A noncoherent detector has only one input, namely, the
modulated signal port. The envelope detector is an example of a noncoherent detector.
Frequency Modulation Detector

An ideal frequency modulation (FM) detector is a device that produces an output that is
proportional to the instantaneous frequency of the input. That is, if the bandpass input is
represented by R(t) cos [vct + u(t)], the output of the ideal FM detector is
d
Kd[vct + u(t)] du(t)
vout(t) = = Kcvc + (480)
dt dt
when the input is not zero (i.e. R(t) Z 0).

Usually, the FM detector is balanced. This means that the DC voltage Kvc does not
appear on the output if the detector is tuned to (or designed for) the carrier frequency fc. In this
case, the output is
du(t)
vout(t) = K (481)
dt
There are many ways to build FM detectors, but almost all of them are based on one of
three principles:
FM-to-AM conversion.
Phase-shift or quadrature detection.
Zero-crossing detection.
A slope detector is one example of the FM-to-AM conversion principle. A block

diagram is shown in Fig. 415. A bandpass limiter is needed to suppress any amplitude varia-
tions on the input signal, since these would distort the desired output signal.
The slope detector may be analyzed as follows. Suppose that the input is a fading signal
with frequency modulation. From Table 41, this FM signal may be represented by
vin(t) = A(t) cos[vct + u(t)] (482)
where
t
u(t) = Kf m(t1) dt1 (483)
L-q
vin(t) v1(t) v2(t) vout(t)

Bandpass Differentiator Envelope
limiter detector
Figure 415 Frequency demodulation using slope detection.

A(t) represents the envelope that is fading, and m(t) is the modulation (e.g., audio) signal. It
follows that the limiter output is proportional to
v1(t) = VL cos[vct + u(t)] (484)
and the output of the differentiator becomes
v2(t) = - VL cvc + d sin[vct + u(t)]

du(t)
(485)
dt
The output of the envelope detector is the magnitude of the complex envelope for v2(t):
vout(t) = ` - VL cvc + d`
du(t)
dt
Because wc du>dt in practice, this becomes
vout(t) = VL cvc + d
du(t)
dt
Using Eq. (483), we obtain
vout(t) = VLvc + VLKfm(t) (486)
which indicates that the output consists of a DC voltage V L v c , plus the ac voltage
VLKf m(t), which is proportional to the modulation on the FM signal. Of course, a capaci-
tor could be placed in series with the output so that only the ac voltage would be passed to
the load.
The differentiation operation can be obtained by any circuit that acts like a frequency-
to-amplitude converter. For example, a single-tuned resonant circuit can be used as illustrated
in Fig. 416, where the magnitude transfer function is |H(f)| = K1 f + K2 over the linear
(useful) portion of the characteristic. A balanced FM detector, which is also called a balanced
discriminator, is shown in Fig. 417. Two tuned circuits are used to balance out the DC when
the input has a carrier frequency of fc and to provide an extended linear frequency-to-voltage
conversion characteristic.
Balanced discriminators can also be built that function because of the phase-shift
properties of a double-tuned RF transformer circuit with primary and secondary windings
[Stark, Tuteur, and Anderson, 1988]. In practice, discriminator circuits have been replaced by
integrated circuits that operate on the quadrature principle.
The quadrature detector is described as follows: A quadrature signal is first obtained
from the FM signal; then, through the use of a product detector, the quadrature signal is
multiplied with the FM signal to produce the demodulated signal vout(t). The quadrature sig-
nal can be produced by passing the FM signal through a capacitor (large) reactance that is
connected in series with a parallel resonant circuit tuned to fc. The quadrature signal voltage
appears across the parallel resonant circuit. The series capacitance provides a 90 phase
shift, and the resonant circuit provides an additional phase shift that is proportional to the
Frequency-to-
amplitude converter Envelope detector
C
vin(t) L C0 R0 vout(t)
(a) Circuit Diagram of a Slope Detector

Frequency band for linear
frequency-to-amplitude conversion
|H(f)|
0 f0
f
fc
(b) Magnitude of Filter Transfer Function
Figure 416 Slope detection using a single-tuned circuit for frequency-to-amplitude conversion.
instantaneous frequency deviation (from fc) of the FM signal. From Eqs. (484) and (483),
the FM signal is
vin(t) = VL cos[vct + u(t)] (487)
and the quadrature signal is
d
du(t)
vquad(t) = K1VL sincvct + u(t) + K2 (488)
dt
where K1 and K2 are constants that depend on component values used for the series capacitor
and in the parallel resonant circuit. These two signals, Eqs. (487) and (488), are multiplied
together by a product detector (e.g., see Fig. 414) to produce the output signal
sin cK2 d
1 2 du(t)
vout(t) = 2 K1VL (489)
dt
Tuned circuit Envelope

#1 @ f1 detector
Tuned circuit # 2 Tuned circuit #1

characteristic characteristic
+
v1(t) 0 f2 fc f1 vout(t)
Overall frequency-to-
voltage characteristic

0
f
Tuned circuit Envelope

#2 @ f2 detector
(a) Block Diagram
Tuned circuit #1
L1 C1
v1(t) vout(t)
L2 C2
Tuned circuit #2
(b) Circuit Diagram
Figure 417 Balanced discriminator.
where the sum-frequency term is eliminated by the low-pass filter. For K2 sufficiently small,
sin x x, and by the use of Eq. (483), the output becomes
1 2
vout(t) = 2 K1K2VLKfm(t) (490)
This demonstrates that the quadrature detector detects the modulation on the input FM signal.
The quadrature detector principle is also used by phase-locked loops that are configured to
detect FM. [See Eq. (4110).]
As indicated by Eq. (480), the output of an ideal FM detector is directly proportional
to the instantaneous frequency of the input. This linear frequency-to-voltage characteristic
may be obtained directly by counting the zero-crossings of the input waveform. An FM
detector utilizing this technique is called a zero-crossing detector. A hybrid circuit (i.e., a
circuit consisting of both digital and analog devices) that is a balanced FM zero-crossing
detector is shown in Fig. 418. The limited (square-wave) FM signal, denoted by v1(t), is
LPF
Q v2(t)
Monostable +
R C
vin(t) v1(t) multivibrator vout(t)
(Q pulse width Differential
Limiter
Tc 1 amplifier
is = ) v3(t)
2 2fc Q
-
R C
LPF
(a) Circuit
1
Waveform for the case of the instantaneous frequency fi>fc where fi=
Ti
Ti
v1(t)
Limiter output
t
Q
v2(t)=DC level of Q
Q monostable output
Tc/2
t
Q v3(t)=DC level of Q

Q monostable output
(b) Waveforms ( fi>fc )
Figure 418 Balanced zero-crossing FM detector.

shown in Fig. 418b. For purposes of illustration, it is assumed that v1(t) is observed over that
portion of the modulation cycle when the instantaneous frequency
1 du(t)
fi(t) = fc + (491)
2p dt
is larger than the carrier frequency fc. That is, fi 7 fc in the illustration. Since the modulation volt-
age varies slowly with respect to the input FM signal oscillation, v1(t) appears (in the figure) to
have a constant frequency, although it is actually varying in frequency according to fi(t). The
monostable multivibrator (one-shot) is triggered on the positive slope zero-crossings of v1(t).
For balanced FM detection, the pulse width of the Q output is set to Tc 2 = 12fc, where fc is the
carrier frequency of the FM signal at the input. Thus, the differential amplifier output voltage is
zero if fi = fc. For fi 7 fc [as illustrated by the v1(t) waveform in the figure], the output voltage is
positive, and for fi 6 fc, the output voltage will be negative. Hence, a linear frequency-to-voltage
characteristic, C [ fi(t) - fc], is obtained where, for an FM signal at the input, fi(t) = fc +
(12p)Kf m(t).
Another circuit that can be used for FM demodulation, as well as for other purposes, is
the phase-locked loop.
414 PHASE-LOCKED LOOPS AND FREQUENCY

SYNTHESIZERS
A phase-locked loop (PLL) consists of three basic components: (1) a phase detector, (2) a
low-pass filter, and (3) a voltage-controlled oscillator (VCO), as shown in Fig. 419. The
VCO is an oscillator that produces a periodic waveform with a frequency that may be varied
about some free-running frequency f0, according to the value of the applied voltage v2(t). The
free-running frequency f0, is the frequency of the VCO output when the applied voltage v2(t)
is zero. The phase detector produces an output signal v1(t) that is a function of the phase
difference between the incoming signal vin(t) and the oscillator signal v0(t). The filtered sig-
nal v2(t) is the control signal that is used to change the frequency of the VCO output. The PLL
configuration may be designed so that it acts as a narrowband tracking filter when the
vin(t) v1(t) Low-pass v2(t)

Phase detector filter (LPF)
(PD) F(f )
v0(t)
Voltage-controlled
oscillator (VCO)
v0(t)
Figure 419 Basic PLL.

Sec. 414 Phase-Locked Loops and Frequency Synthesizers 283
low-pass filter (LPF) is a narrowband filter. In this operating mode, the frequency of the VCO
will become that of one of the line components of the input signal spectrum, so that, in effect,
the VCO output signal is a periodic signal with a frequency equal to the average frequency of
this input signal component. Once the VCO has acquired the frequency component, the fre-
quency of the VCO will track the input signal component if it changes slightly in frequency.
In another mode of operation, the bandwidth of the LPF is wider so that the VCO can track
the instantaneous frequency of the whole input signal. When the PLL tracks the input signal
in either of these ways, the PLL is said to be locked.
If the applied signal has an initial frequency of f0, the PLL will acquire a lock and the
VCO will track the input signal frequency over some range, provided that the input fre-
quency changes slowly. However, the loop will remain locked only over some finite range of
frequency shift. This range is called the hold-in (or lock) range. The hold-in range depends
on the overall DC gain of the loop, which includes the DC gain of the LPF. On the other
hand, if the applied signal has an initial frequency different from f0, the loop may not acquire
lock even though the input frequency is within the hold-in range. The frequency range over
which the applied input will cause the loop to lock is called the pull-in (or capture) range.
This range is determined primarily by the loop filter characteristics, and it is never greater
than the hold-in range. (See Fig. 423.) Another important PLL specification is the
maximum locked sweep rate, which is defined as the maximum rate of change of the input
frequency for which the loop will remain locked. If the input frequency changes faster than
this rate, the loop will drop out of lock.
If the PLL is built using analog circuits, it is said to be an analog phase-locked loop
(APLL). Conversely, if digital circuits and signals are used, the PLL is said to be a digital
phase-locked loop (DPLL). For example, the phase detection (PD) characteristic depends on
the exact implementation used. Some PD characteristics are shown in Fig. 420. The sinu-
soidal characteristic is obtained if an (analog circuit) multiplier is used and the periodic
signals are sinusoids. The multiplier may be implemented by using a double-balanced mixer.
The triangle and sawtooth PD characteristics are obtained by using digital circuits. In addition
to using digital VCO and PD circuits, the DPLL may incorporate a digital loop filter and
signal-processing techniques that use microprocessors. Gupta [1975] published a fine tutorial
paper on analog phase-locked loops in the IEEE Proceedings, and Lindsey and Chie [1981]
followed with a survey paper on digital PLL techniques. In addition, there are excellent books
available [Blanchard, 1976; Gardner, 1979; Best, 1999].
The PLL may be studied by examining the APLL, as shown in Fig. 421. In this figure,
a multiplier (sinusoidal PD characteristic) is used. Assume that the input signal is
vin(t) = Ai sin[v0t + ui(t)] (492)

and that the VCO output signal is
v0(t) = A0 cos[v0t + u0(t)] (493)
where
t
u0(t) = Kv v2(t) dt (494)
L-q
Vp
v1
(a) Sinusoidal Characteristics
Vp
v1

e
(b) Triangle Characteristics
Vp
v1

e
(c) Sawtooth Characteristics
Figure 420 Some phase detector characteristics.
vin(t) v1(t)= Km vin (t)v0(t) LPF v2(t)

F(f)
v0(t)
VCO
Figure 421 Analog PLL.
and Kv is the VCO gain constant (radV-s). Then PD output is

v1(t) = KmAiA0 sin[v0t + ui(t)] cos[v0t + u0(t)]
KmAiA0 KmAiA0
= sin[ui(t) - u0(t)] + sin[2v0t + ui(t) + u0(t)] (495)
2 2
where Km is the gain of the multiplier circuit. The sum frequency term does not pass through
the LPF, so the LPF output is
v2(t) = Kd[sin ue(t)] * f(t) (496)
where
ue(t) ! ui(t) - u0(t) (497)
KmA iA 0
Kd = (498)
2
and f(t) is the impulse response of the LPF. ue(t) is called the phase error; Kd is the equivalent
PD constant, which, for the multiplier-type PD, depends on the levels of the input signal Ai
and the level of the VCO signal A0.
The overall equation describing the operation of the PLL may be obtained by taking the
derivative of Eqs. (494) and (497) and combining the result by the use of Eq. (496). The
resulting nonlinear equation that describes the PLL becomes
t
due(t) dui(t)
= - KdKv [sin ue(l)]f(t - l) dl (499)
dt dt L0
where ue(t) is the unknown and ui(t) is the forcing function.
In general, this PLL equation is difficult to solve. However, it may be reduced to a linear
equation if the gain Kd is large, so that the loop is locked and the error ue(t) is small. In this
case, sin ue(t) ue(t), and the resulting linear equation is
due(t) dui(t)
= - KdKvue(t) * f(t) (4100)
dt dt
A block diagram based on this linear equation is shown in Fig. 422. In this linear PLL model
(Fig. 422), the phase of the input signal and the phase of the VCO output signal are used
instead of the actual signals themselves (Fig. 421). The closed-loop transfer function
0(f)i(f) is
0(f) KdKvF(f)
H(f) = = (4101)
i(f) j2pf + KdKvF(f)
i(t) e(t) LPF v2(t)

Kd F1(f)=F(f)

0(t)
VCO
Kv
F2(f)=
j2f
0(t)
Figure 422 Linear model of the analog PLL.

where 0( f) = [u0(t)] and i( f ) = [ui(t)]. Of course, the design and analysis techniques
used to evaluate linear feedback control systems, such as Bode plots, which will indicate
phase gain and phase margins, are applicable. In fact, they are extremely useful in describing
the performance of locked PLLs.
The equation for the hold-in range may be obtained by examining the nonlinear behav-
ior of the PLL. From Eqs. (494) and (496), the instantaneous frequency deviation of the
VCO from v0 is
du0(t)
= Kvv2(t) = KvKd[sin ue(t)] * f(t) (4102)
dt
To obtain the hold-in range, the input frequency is changed very slowly from f0. Here the DC
gain of the filter is the controlling parameter, and Eq. (4102) becomes
v = KvKdF(0) sin ue (4103)
The maximum and minimum values of v give the hold-in range, and these are obtained
when sin ue = 1. Thus, the maximum hold-in range (the case with no noise) is
1
fh = K K F(0) (4104)
2p v d
A typical lock-in characteristic is illustrated in Fig. 423. The solid curve shows the
VCO control signal v2(t) as the sinusoidal testing signal is swept from a low frequency to a
high frequency (with the free-running frequency of the VCO, f0, being within the swept
band). The dashed curve shows the result when sweeping from high to low. The hold-in range
fh is related to the DC gain of the PLL as described by Eq. (4104).
The pull-in range fp is determined primarily by the loop-filter characteristics.
For example, assume that the loop has not acquired lock and that the testing signal is swept slowly
toward f0. Then, the PD output, there will be a beat (oscillatory) signal, and its frequency | fin - f0|
will vary from a large value to a small value as the test signal frequency sweeps toward f0.
Pull-in Hold-in
range range
v2(t)
Direction fp fn
of sweep
Direction
of sweep
0 f0 fin
Hold-in Hold-in
range range
fh fp
Figure 423 PLL VCO control voltage for a swept sinusoidal input signal.
As the testing signal frequency comes closer to f0, the beat-frequency waveform will become
nonsymmetrical, in which case it will have a nonzero DC value. This DC value tends to change the
frequency of the VCO to that of the input signal frequency, so that the loop will tend to lock.
The pull-in range, fp, where the loop acquires lock will depend on exactly how the loop filter F(f)
processes the PD output to produce the VCO control signal. Furthermore, even if the input signal
is within the pull-in range, it may take a fair amount of time for the loop to acquire lock, since the
LPF acts as an integrator and it takes some time for the control voltage (filter output) to build up to
a value large enough for locking to occur. The analysis of the pull-in phenomenon is complicated.
It is actually statistical in nature, because it depends on the initial phase relationship of the input
and VCO signals and on noise that is present in the circuit. Consequently, in the measurement of
fp, several repeated trials may be needed to obtain a typical value.
The locking phenomenon is not peculiar to PLL circuits, but occurs in other types of
circuits as well. For example, if an external signal is injected into the output port of an oscil-
lator (i.e., a plain oscillator, not a VCO), the oscillator signal will tend to change frequency
and will eventually lock onto the frequency of the external signal if the latter is within the
pull-in range of the oscillator. This phenomenon is called injection locking or synchronization
of an oscillator and may be modeled by a PLL model [Couch, 1971].
The PLL has numerous applications in communication systems, including (1) FM
detection, (2) the generation of highly stable FM signals, (3) coherent AM detection, (4)
frequency multiplication, (5) frequency synthesis, and (6) use as a building block within
complicated digital systems to provide bit synchronization and data detection.
Let us now find what conditions are required for the PLL to become an FM detector.
Referring to Fig. 421, let the PLL input signal be an FM signal. That is,
t
vin(t) = Ai sin cvct + Df m(l) dl d (4105a)
L-q
where
t
ui(t) = Df m(l) dl (4105b)
L-q
or
Df
i(f) = M(f) (4105c)
j2pf
and m(t) is the baseband (e.g., audio) modulation that is to be detected. We would like to find
the conditions such that the PLL output, v2(t), is proportional to m(t). Assume that fc is within
the capture (pull-in) range of the PLL; thus, for simplicity, let f0 = fc. Then the linearized PLL
model, as shown in Fig. 422, can be used for analysis. Working in the frequency domain, we
obtain the output
aj b F1(f)
2pf
Kv
V2(f) = i(f)
b
2pf
F1(f) + ja
KvKd
which, by the use of Eq. (4105c), becomes
Df
F1(f)
Kv
V2(f) = M(f) (4106)
b
2pf
F1(f) + ja
KvKd
Now we find the conditions such that V2( f) is proportional to M( f). Assume that the band-
width of the modulation is B hertz, and let F1( f) be a low-pass filter. Thus,
F(f) = F1(f) = 1, |f| 6 B (4107)
Also, let
KvKd
B (4108)
2p
Then Eq. (4106) becomes
Df
V2(f) = M(f) (4109)
Kv
or
v2(t) = Cm(t) (4110)
where the constant of proportionality is C = DfKv. Hence, the PLL circuit of Fig. 421 will
become an FM detector circuit, where v2(t) is the detected FM output when the conditions of
Eqs. (4107) and (4108) are satisfied.
In another application, the PLL may be used to supply the coherent oscillator signal for
product detection of an AM signal (Fig. 424). Recall from Eqs. (492) and (493) that the
VCO of a PLL locks 90 out of phase with respect to the incoming signal. Then v0(t) needs
to be shifted by -90 so that it will be in phase with the carrier of the input AM signal, the
requirement for coherent detection of AM, as given by Eq. (477). In this application, the
bandwidth of the LPF needs to be just wide enough to provide the necessary pull-in range in
order for the VCO to be able to lock onto the carrier frequency fc .
Figure 425 illustrates the use of a PLL in a frequency synthesizer. The synthesizer
generates a periodic signal of frequency
fout = a bf
N
(4111)
M x

This results from the characteristic of the phase detector circuit. The statement is correct for a PD that
produces a zero DC output voltage when the two PD input signals are 90 out of phase (i.e., a multiplier-type PD).
However, if the PD circuit produced a zero DC output when the two PD inputs were in phase, the VCO of the PLL
would lock in phase with the incoming PLL signal.
Quadrature
detector
AM signal
LPF
v1=Ac [1+m (t)] sin(c t)
v1(t) v0(t)
In phase 90 phase
detector VCO
shift
vout(t)
LPF
Figure 424 PLL used for coherent detection of AM.
where fx is the frequency of the stable oscillator and N and M are the frequency-divider para-
meters. This result is verified by recalling that when the loop is locked, the DC control signal
v3(t) shifts the frequency of the VCO so that v2(t) will have the same frequency as vin(t). Thus,
fx fout
= (4112)
M N
which is equivalent to Eq. (4111).
Classical frequency dividers use integer values for M and N. Furthermore, if program-
mable dividers are used, the synthesizer output frequency may be changed by using software
that programs a microprocessor to select the appropriate values of M and N, according to Eq.
(4111). This technique is used in frequency synthesizers that are built into modern receivers
with digital tuning. (See study aid Prob. SA46, for example, of frequency synthesizer design.)
For the case of M = 1, the frequency synthesizer acts as a frequency multiplier.
Equivalent noninteger values for N can be obtained by periodically changing the divider
count over a set of similar integer values. This produces an average N value that is noninteger
and is called the fractional-N technique. With fractional-N synthesizers, the instantaneous
value of N changes with time, and this can modulate the VCO output signal to produce
unwanted (spurious) sidebands in the spectrum. By careful design the sideband noise can be
Oscillator vx(t) Frequency vin(t) v1(t) v3(t)

(frequency standard) divider LPF
f=fx M
v2(t)
Frequency v0(t)
divider VCO
N
vout(t)
Figure 425 PLL used in a frequency synthesizer.

Microprocessor PCM Synthesized

(with digitized samples signal waveform
Digital-to-analog
of waveform stored
converter
in random-access memory
or read-only memory)
Figure 426 Direct digital synthesis (DDS).
reduced to a low level [Conkling, 1998]. More complicated PLL synthesizer configurations
can be built that incorporate mixers and additional oscillators.
415 DIRECT DIGITAL SYNTHESIS
Direct digital synthesis (DDS) is a method for generating a desired waveform (such as a sine
wave) by using the computer technique described in Fig. 426. To configure the DDS system
to generate a waveform, samples of the desired waveform are converted into PCM words and
stored in the memory (random-access memory [RAM] or read-only memory [ROM]) of the
microprocessor system. The DDS system can then generate the desired waveform by playing
back the stored words into the digital-to-analog converter.
This DDS technique has many attributes. For example, if the waveform is periodic, such as
a sine wave, only one cycle of samples needs to be stored in memory. The continuous sine wave
can be generated by repeatedly cycling through the memory. The frequency of the generated sine
wave is determined by the rate at which the memory is read out. If desired, the microprocessor
can be programmed to generate a certain frequency during a certain time interval and then switch
to a different frequency (or another waveshape) during a different time interval. Also, simultane-
ous sine and cosine (two-phase) outputs can be generated by adding another DAC. The signal-to-
quantizing noise can be designed to be as large as desired by selecting the appropriate number of
bits that are stored for each PCM word, as described by Eq. (318).
The DDS technique is replacing analog circuits in many applications. For example, in
higher-priced communications receivers, the DDS technique is used as a frequency synthesizer
to generate local oscillator signals that tune the radio. (See Sec. 416.) In electronic pipe
organs and music synthesizers, DDS can be used to generate authentic as well as weird sounds.
Instrument manufacturers are using DDS to generate the output waveforms for function gener-
ators and arbitrary waveform generators (AWG). Telephone companies are using DDS to
generate dial tones and busy signals. (See Chapter 8.)
416 TRANSMITTERS AND RECEIVERS
Generalized Transmitters
Transmitters generate the modulated signal at the carrier frequency fc from the modulating
signal m(t). In Secs. 41 and 42, it was demonstrated that any type of modulated signal could
be represented by
v(t) = Re{g(t)ejvct} (4113)
Sec. 416 Transmitters and Receivers 291
or, equivalently,
v(t) = R(t) cos[vct + u(t)] (4114)
and
v(t) = x(t) cos vct - y(t) sin vct (4115)
where the complex envelope
g(t) = R(t)eju(t) = x(t) + jy(t) (4116)
is a function of the modulating signal m(t). The particular relationship that is chosen for g(t) in terms
of m(t) defines the type of modulation that is used, such as AM, SSB, or FM. (See Table 41.) A
generalized approach may be taken to obtain universal transmitter models that may be reduced to
those used for a particular type of modulation. We will also see that there are equivalent models that
correspond to different circuit configurations, yet they may be used to produce the same type of
modulated signal at their outputs. It is up to the designer to select an implementation method that
will maximize performance, yet minimize cost, based on the state of the art in circuit development.
There are two canonical forms for the generalized transmitter, as indicated by Eqs.
(4114) and (4115). Equation (4114) describes an AMPM type of circuit, as shown in
Fig. 427. The baseband signal-processing circuit generates R(t) and u (t) from m(t). The R
and u are functions of the modulating signal m(t), as given in Table 41 for the particular
type of modulation desired. The signal processing may be implemented by using either non-
linear analog circuits or a digital computer that incorporates the R and u algorithms under
software program control. In the implementation using a digital computer, one ADC will be
needed at the input and two DACs will be needed at the output. The remainder of the
AMPM canonical form requires RF circuits, as indicated in the figure.
Figure 428 illustrates the second canonical form for the generalized transmitter. This
uses in-phase and quadrature-phase (IQ) processing. Similarly, the formulas relating x(t) and
y(t) to m(t) are shown in Table 41, and the baseband signal processing may be implemented
by using either analog hardware or digital hardware with software. The remainder of the
canonical form uses RF circuits as indicated.
RF circuits
Basebands circuits
R(t) v(t)=R(t) cos[vc t+(t)]
Modulated signal out

Baseband signal- Carrier
m(t) processing (Type I) oscillator fc
circuit may
Modulation
be nonlinear
in
cos[vc t+(t)]
(t)
Phase
modulator
Figure 427 Generalized transmitter using the AMPM generation technique.

RF circuits
Basebands circuits
x(t) I channel
Baseband signal- +
m(t) processing (Type II) v(t)=x(t) cos(vc t)-y(t) sin(vct)
Modulation
circuit may cos(vc t)

be nonlinear
in
y(t) Q channel
Carrier 90 sin(vc t )
oscillator fc phase
shift
cos(w c t)
Figure 428 Generalized transmitter using the quadrature generation technique.
Example 45 GENERATION OF A QM SIGNAL BY IQ PROCESSING

Rework Example 41, but generate the QM signal by using IQ processing as shown in
Fig. 428. See Example4_05.m for the solution. Note that the QM signal generated by IQ
processing is identical to that obtained by complex envelope processing used for Example 41.
Once again, it is stressed that any type of signal modulation (AM, FM, SSB, QPSK,
etc.) may be generated by using either of these two canonical forms. Both of these forms
conveniently separate baseband processing from RF processing. Digital techniques are espe-
cially useful to realize the baseband-processing portion. Furthermore, if digital computing
circuits are used, any desired type of modulation can be obtained by selecting the appropriate
software algorithm.
Most of the practical transmitters in use today are special variations on these canonical
forms. Practical transmitters may perform the RF operations at some convenient lower RF
frequency and then up-convert to the desired operating frequency. In the case of RF signals
that contain no AM, frequency multipliers may be used to arrive at the operating frequency.
Of course, power amplifiers are usually required to bring the output power level up to the
specified value. If the RF signal contains no amplitude variations, Class C amplifiers (which
have relatively high efficiency) may be used; otherwise, Class B amplifiers are used.
Generalized Receiver: The Superheterodyne Receiver

The receiver has the job of extracting the source information from the received modulated
signal that may be corrupted by noise. Often, it is desired that the receiver output be a replica
of the modulating signal that was present at the transmitter input. There are two main classes
of receivers: the tuned radio-frequency (TRF) receiver and the superheterodyne receiver.
The TRF receiver consists of a number of cascaded high-gain RF bandpass stages that are
tuned to the carrier frequency fc , followed by an appropriate detector circuit (an envelope detec-
tor, a product detector, an FM detector, etc.). The TRF is not very popular, because it is difficult
to design tunable RF stages so that the desired station can be selected and yet have a narrow
bandwidth so that adjacent channel stations are rejected. In addition, it is difficult to obtain high
gain at radio frequencies and to have sufficiently small stray coupling between the output
and input of the RF amplifying chain so that the chain will not become an oscillator at fc. The
crystal set that is built by the Cub Scouts is an example of a singleRF-stage TRF receiver that
has no gain in the RF stage. TRF receivers are often used to measure time-dispersive (multipath)
characteristics of radio channels [Rappaport, 1989].
Most receivers employ the superheterodyne receiving technique as shown in
Fig. 429. The technique consists of either down-converting or up-converting the input sig-
nal to some convenient frequency band, called the intermediate frequency (IF) band, and
then extracting the information (or modulation) by using the appropriate detector. This
basic receiver structure is used for the reception of all types of bandpass signals, such as
television, FM, AM, satellite, cellular, and radar signals. The RF amplifier has a bandpass
characteristic that passes the desired signal and provides amplification to override addi-
tional noise that is generated in the mixer stage. The RF filter characteristic also provides
some rejection of adjacent channel signals and noise, but the main adjacent channel
rejection is accomplished by the IF filter.
The IF filter is a bandpass filter that selects either the up-conversion or down-conversion
component (whichever is chosen by the receivers designer). When up conversion is selected,
the complex envelope of the IF (bandpass) filter output is the same as the complex envelope for
the RF input, except for RF filtering, H1( f), and IF filtering, H2( f). However, if down conver-
sion is used with fLO 7 fc , the complex envelope at the IF output will be the conjugate of that
for the RF input. [See Eq. (461c).] This means that the sidebands of the IF output will be
RF inputs Mixer
vin(t) RF
(radio-frequency)
amplifer, H1(f)
LO (local oscillator)
Baseband
output
IF (to speaker
IF (intermediate- out CRT, etc.)
Baseband
frequency) Detector
amplifer
amplifer, H2(f)
Figure 429 Superheterodyne receiver.

Dual-conversion superheterodyne receivers can also be built, in which a second mixer and a second IF stage
follow the first IF stage shown in Fig. 429.
inverted (i.e., the upper sideband on the RF input will become the lower sideband, etc., on the
IF output). If fLO 6 fc, the sidebands are not inverted.
The center frequency selected for the IF amplifier is chosen on the basis of three
considerations:
The IF frequency should be such that a stable high-gain IF amplifier can be economi-
cally attained.
The IF frequency needs to be low enough so that, with practical circuit elements in the
IF filters, values of Q can be attained that will provide a steep attenuation characteristic
outside the bandwidth of the IF signal. This decreases the noise and minimizes the
interference from adjacent channels.
The IF frequency needs to be high enough so that the receiver image response can be
made acceptably small.
The image response is the reception of an unwanted signal located at the image frequency due
to insufficient attenuation of the image signal by the RF amplifier filter. The image response
is best illustrated by an example.
Example 46 AM BROADCAST SUPERHETERODYNE RECEIVER

Assume that an AM broadcast band radio is tuned to receive a station at 850 kHz and that the LO
frequency is on the high side of the carrier frequency. If the IF frequency is 455 kHz, the LO
frequency will be 850 + 455 = 1,305 kHz. (See Fig. 430.) Furthermore, assume that other
signals are present at the RF input of the radio and, particularly, that there is a signal at 1,760 kHz;
this signal will be down-converted by the mixer to 1,760 - 1,305 = 455 kHz. That is, the unde-
sired (1,760-kHz) signal will be translated to 455 kHz and will be added at the mixer output to the
desired (850-kHz) signal, which was also down-converted to 455 kHz. This undesired signal that
has been converted to the IF band is called the image signal. If the gain of the RF amplifier is
down by, say, 25 dB at 1,760 kHz compared to the gain at 850 kHz, and if the undesired signal is
25 dB stronger at the receiver input than the desired signal, both signals will have the same level
when translated to the IF. In this case, the undesired signal will definitely interfere with the
desired signal in the detection process.
For down converters (i.e., fIF = | fc - fLO|), the image frequency is
fimage = e
fc + 2fIF, if fLO 7 fc (high-side injection)
(4117a)
fc - 2fIF, if fLO 6 fc (low-side injection)
where fc is the desired RF frequency, fIF is the IF frequency, and fLO is the local oscillator
frequency. For up converters (i.e., fIF = fc + fLO), the image frequency is
fimage = fc + 2fLO (4117b)

|V LO (f)|
|V in (f)|
|H 1(f)|
Image attenuation
1,760 1,305 850 f

fc = 850 fLO = 1,305 fimage = 1,760 kHz
455 kHz 455 kHz
910 kHz
Figure 430 Spectra of signals and transfer function of an RF amplifier in a superheterodyne receiver.
From Fig. 430, it is seen that the image response will usually be reduced if the IF frequency
is increased, since fimage will occur farther away from the main peak (or lobe) of the RF filter
characteristic, |H1( f)|.
Recalling our earlier discussion on mixers, we also realize that other spurious responses
(in addition to the image response) will occur in practical mixer circuits. These must also be
taken into account in good receiver design.
Table 44 illustrates some typical IF frequencies that have become de facto standards.
For the intended application, the IF frequency is low enough that the IF filter will provide
good adjacent channel signal rejection when circuit elements with a realizable Q are used;
yet the IF frequency is large enough to provide adequate image-signal rejection by the RF
amplifier filter.
The type of detector selected for use in the superheterodyne receiver depends on the
intended application. For example, a product detector may be used in a PSK (digital) system,
and an envelope detector is used in AM broadcast receivers. If the complex envelope g(t) is de-
sired for generalized signal detection or for optimum reception in digital systems, the x(t) and
y(t) quadrature components, where x(t) + jy(t) = g(t), may be obtained by using quadrature
TABLE 44 SOME POPULAR IF FREQUENCIES IN THE UNITED STATES.
IF Frequency Application
262.5 kHz AM broadcast radios (in automobiles)

455 kHz AM broadcast radios
10.7 MHz FM broadcast radios
21.4 MHz FM two-way radios
30 MHz Radar receivers
43.75 MHz TV sets
60 MHz Radar receivers
70 MHz Satellite receivers
I channel x(t)
LPF
IF signal
vIF(t)=Re[g(t)evIFt] 2 cos(vIFt)
Q channel y(t)
LPF
2 cos(vIFt)
Oscillator 90
f=fIF phase shift
Figure 431 IQ (in-phase and quadrature-phase) detector.
product detectors, as illustrated in Fig. 431. x(t) and y(t) could be fed into a signal processor
to extract the modulation information. Disregarding the effects of noise, the signal processor
could recover m(t) from x(t) and y(t) (and, consequently, demodulate the IF signal) by using the
inverse of the complex envelope generation functions given in Table 41.
The superheterodyne receiver has many advantages and some disadvantages. The main
advantage is that extraordinarily high gain can be obtained without instability (self-oscillation).
The stray coupling between the output of the receiver and the input does not cause oscillation
because the gain is obtained in disjoint frequency bandsRF, IF, and baseband. The receiver is
easily tunable to another frequency by changing the frequency of the LO signal (which may be
supplied by a frequency synthesizer) and by tuning the bandpass of the RF amplifier to the
desired frequency. Furthermore, high-Q elementswhich are needed (to produce steep filter
skirts) for adjacent channel rejectionare needed only in the fixed tuned IF amplifier. The
main disadvantage of the superheterodyne receiver is the response to spurious signals that will
occur if one is not careful with the design.
Zero-IF Receivers
When the LO frequency of a superheterodyne receiver is selected to be the carrier
frequency (fLO = fc) then fIF = 0, and the superheterodyne receiver becomes a zero-IF
or direct conversion receiver. In this case, the IF filter becomes a low-pass filter (LPF).
This mixerLPF combination functions as a product detector (and the detector stage of
Figure 429 is not needed). A quadrature down converter can also be added so that the x(t)
and y(t) components of the complex envelope can be recovered. In this case the zero-IF
receiver has a block diagram as shown in Fig. 431, where the input signal is at fc and vc
replaces vIF in the figure. The components x(t) and y(t) may be sampled and digitized with
ADC so that the complex envelope, g(t) = x(t) + jy(t), may be processed digitally with DSP
hardware, which is discussed in Sec. 417. The analog LPF acts as an antialiasing filter for
the sampler and the DSP hardware. The zero-IF receiver is also similar to a TRF receiver
with product detection.

A direct-conversion receiver is also called a homodyne or synchrodyne receiver.
Sec. 417 Software Radios 297
Zero-IF receivers have several advantages. They have no image response. The same
zero-IF receiver hardware can be used in many different applications for manufacturing econ-
omy. Since DSP hardware is used, the effective RF bandpass characteristics and the detector
characteristics are determined by DSP software. (See next section.) The software can
be changed easily to match the desired application. The same zero-IF hardware can be used
for receivers in different VHF and UHF bands by selecting the appropriate LO frequency
(FLO = fc) and tuning the front-end filter (usually a single-tuned circuit) to fc.
The zero-IF receiver has the disadvantage of possibly leaking LO radiation out of the
antenna input port due to feed-through from the mixer. Also, there will be a DC offset on the
mixer output, if there is LO leakage into the antenna input since a sine wave (LO signal)
multiplied by itself produces a DC term (plus an out-of-band second harmonic). The use of a
high-quality balance mixer and LO shielding will minimize these problems. The receiver can
also have a poor noise figure, since the front end usually is not a high-gain, low-noise stage.
As in any receiver, the hardware has to be carefully designed so that there is sufficient
dynamic range to prevent strong signals from overloading the receiver (producing spurious
signals due to nonlinearities) and yet sufficient gain for detecting weak signals. In spite of
these difficulties, the zero-IF receiver provides an economical, high-performance solution for
many applications. A practical zero-IF receiver with excellent selectivity provided by DSP
filtering is described in QST [Frohne, 1998].
Interference
A discussion of receivers would not be complete without considering some of the causes of
interference. Often the owner of the receiver thinks that a certain signal, such as an amateur
radio signal, is causing the difficulty. This may or may not be the case. The origin of the
interference may be at any of three locations:
At the interfering signal source, a transmitter may generate out-of-band signal compo-
nents (such as harmonics) that fall in the band of the desired signal.
At the receiver itself, the front end may overload or produce spurious responses. Front-
end overload occurs when the RF or mixer stage of the receiver is driven into the
nonlinear range by the interfering signal and the nonlinearity causes cross-modulation
on the desired signal at the output of the receiver RF amplifier.
In the channel, a nonlinearity in the transmission medium may cause undesired signal
components in the band of the desired signal.
For more discussion of receiver design and examples of practical receiver circuits, the reader
is referred to the ARRL Handbook [ARRL, 2010].
417 SOFTWARE RADIOS
Software radios use DSP hardware, microprocessors, specialized digital ICs, and software to
produce modulated signals for transmission (see Table 41 and Fig. 428) and to demodulate
signals at the receiver. Ultimately, the ideal software receiver would sample and digitize
received signals at the antenna with analog-to-digital conversion (ADC) and process the
signal with digital signal-processing (DSP) hardware. Software would be used to compute the
receiver output. The difficulty with this approach is that it is almost impossible to build
ADCDSP hardware that operates fast enough to directly process wideband modulated signals
with gigahertz carrier frequencies [Baines, 1995]. However, the complex envelope of these
signals may be obtained by using a superheterodyne receiver with quadrature detectors
(Fig. 431). For sufficiently modest bandpass bandwidth (say, 25 MHz), the I and Q compo-
nents, x(t) and y(t), of the complex envelope can be sampled and processed with practical DSP
hardware so that software programming can be used.
In another approach, a high-speed ADC can be used to provide samples of the IF signal
that are passed to a digital down-converter (DDC) integrated circuit (e.g., Intersil, HSP50016)
[Chester, 1999]. The DDC multiplies the IF samples with samples of cosine and sine LO
signals. This down-converts the IF samples to baseband I and Q samples. The DDC uses
ROM lookup tables to obtain the LO cosine and sine samples, a method similar to the direct
digital synthesis (DDS) technique discussed in Sec. 415. To simultaneously receive multiple
adjacent channel signals, multiple DDC ICs can be used in parallel with the LO of each DDC
tuned to the appropriate frequency to down convert the signal to baseband I and Q samples for
that signal. (For more details, see the Intersil Web site at http: www.intersil.com.)
The I and Q samples of the complex envelope, g(t) = x(t) + jy(t), can be filtered to
provide equivalent bandpass IF filtering (as described in Sec. 45). The filtering can provide
excellent equivalent IF filter characteristics with tight skirts for superb adjacent channel inter-
ference rejection. The filter characteristic may be changed easily by changing the software.
Raised cosine-rolloff filtering is often used to reduce the transmission bandwidth of digital
signals without introducing ISI. For minimization of bit errors due to channel noise, as well as
elimination of ISI, a square-root raised-cosine filter is used at both the transmitter and the
receiver [as shown by Eq. (378) of Sec. 36].
AM and PM detection is accomplished by using the filtered I and Q components to
compute the magnitude and phase of the complex envelope, as shown by Eqs. (44a) and
(44b), respectively. FM detection is obtained by computing the derivative of the phase, as
shown by Eq. (48).
The Fourier transform can also be used in software radios, since the FFT can be com-
puted efficiently with DSP ICs. For example, the FFT spectrum can be used to determine the
presence or absence of adjacent channel signals. Then, appropriate software processing can
either enhance or reject a particular signal (as desired for a particular application). The FFT
can also be used to simultaneously detect the data on a large number of modulated carriers
that are closely spaced together. (For details, see Sec. 512 on OFDM.)
The software radio concept has many advantages. Two of these are that the same hard-
ware may be used for many different types of radios, since the software distinguishes one type
from another, and that, after software radios are sold, they can be updated in the field to include
the latest protocols and features by downloading revised software. The software radio concept
is becoming more economical and practical each day. It is the way of the future.
For additional reading about practical software radio design and designed circuits, see
the 2011 ARRL Handbook [ARRL, 2010]. To explore hands-on design of a software radio, go
to http:gnuradio.org. GNU Radio is a free software toolkit for learning about, building, and
deploying Software Defined Radio systems. A description of GNU Radio is also available on
Wikipedia.
Sec. 419 Study-Aid Examples 299
418 SUMMARY
The basic techniques used for bandpass signaling have been studied in this chapter. The
complex-envelope technique for representing bandpass signals and filters was found to be
very useful. A description of communication circuits with output analysis was presented for
filters, amplifiers, limiters, mixers, frequency multipliers, phase-locked loops, and detector
circuits. Nonlinear as well as linear circuit analysis techniques were used. The superhetero-
dyne receiving circuit was found to be fundamental in communication receiver design.
Generalized transmitters, receivers, and software radios were studied. Practical aspects of
their design, such as techniques for evaluating spurious signals, were examined.
419 STUDY-AID EXAMPLES
SA41 Voltage Spectrum for an AM Signal An AM voltage signal s(t) with a carrier
frequency of 1,150 kHz has a complex envelope g(t) = Ac[1 + m(t)]. Ac = 500 V, and the modula-
tion is a 1-kHz sinusoidal test tone described by m(t) = 0.8 sin (2p1,000t). Evaluate the voltage
spectrum for this AM signal.
Solution. Using the definition of a sine wave from Sec. A1,
0.8 j2p1000t
m(t) = [e - e-j2p1000t] (4118)
j2
Using Eq. (226) with the help of Sec. A5, we find that the Fourier transform of m(t) is
M(f) = - j 0.4 d(f-1,000) + j 0.4 d(f + 1,000) (4119)
Substituting this into Eq. (420a) yields the voltage spectrum of the AM signal:
S(f) = 250 d(f-fc) - j100 d(f-fc - 1,000) + j100 d (f-fc + 1,000)

+ 250 d(f + fc) - j100 d(f + fc - 1,000) + j100 d(f + fc + 1,000) (4120)
See SA4_1.m for a plot of the AM signal waveform, and a plot of its spectrum which
was calculated by using the FFT. Compare this plot of the spectrum with that given by
Eq. (4120).
SA42 PSD for an AM Signal Compute the PSD for the AM signal that is described in
SA41.
Solution. Using Eq. (271), we obtain the autocorrelation for the sinusoidal modulation m(t)
namely,
A2 A2 jv0t
Rm(t) = cos v0t = [e + e-jv0t] (4121)
2 4

Because m(t) is periodic, an alternative method for evaluating M(f) is given by Eq. (2109), where c-1 = j0.4,
c1 = -j0.4, and the other cns are zero.
where A = 0.8 and v0 = 2p1,000. Taking the Fourier transform by the use of Eq. (226) we obtain
the PSD of m(t):
A2
m(f) = [d(f-fo) + d(f + f0)]
4
or
m(f) = 0.16 [d(f-1,000) + d(f + 1,000)] (4122)
The autocorrelation for the complex envelope of the AM signal is
Rg(t) = 8g *(t)g(t + t)9 = A2c 8[1 + m(t)][1 + m(t + t)]9
= A2c [819 + 8m(t)9 + 8m(t + t)9 + 8m(t)m(t + t)9]
But 819 = 1, 8m(t)9 = 0, 8m(t + t)9 = 0, and 8m(t) m(t + t)9 = Rm(t). Thus,
Rg(t) = A2c + A2c Rm(t) (4123)
Taking the Fourier transform of both sides of Eq. (4123), we get
g(f) = A2c d(f) + A2c m(f) (4124)
Substituting Eq. (4124) into Eq. (413), with the aid of Eq. (4122), we obtain the PSD for the
AM signal:
s(f) = 62,500 d(f-fc) + 10,000 d(f-fc - 1,000)

+ 10,000 d(f-fc + 1,000) + 62,500 d (f + fc)
+ 10,000 d(f + fc - 1,000) + 10,000 d(f + fc + 1,000) (4125)
(Note: We realize that this bandpass PSD for s(t) is found by translating (i.e., moving) the baseband
PSD of g(t) up to fc and down to -fc . Furthermore, for the case of AM, the PSD of g(t) consists of
the PSD for m(t) plus the superposition of a delta function at f = 0).
SA43 Average Power for an AM Signal Assume that the AM voltage signal s(t), as
described in SA41, appears across a 50- resistive load. Compute the actual average power
dissipated in the load.
Solution. From Eq. (421), the normalized average power is
1 2
(Ps)norm = (Vs)2rms = 2 Ac [1 + (Vm)2rms]
0.8 2
2 (500) c1 + a b d = 165 kW
1 2
= (4126a)
12

Because m(t) is periodic, Eq. (2126) can be used as an alternative method of evaluating m(f). That is, by
* = A>(2j) = -j0.8>2 = -j0.4 (and the other c s are zero), Eq. (4122) is obtained.
using Eq. (2126) with c1 = c-1 n
Note: An alternative method of computing (Ps)norm is to calculate the area under the PDF for s(t).
That is, by using Eq. (4125),
q
(Ps)norm = (Vs)2rms = Ps(f) df = 165 kW (4126b)
L-q
Using Eq. (4126a) or Eq. (4126b), we obtain the actual average power dissipated in the
50- load:
(Vs)2rms 1.65 * 105

(Ps)actual = = = 3.3 kW (4127)
RL 50
SA44 PEP for an AM Signal If the AM voltage signal of SA41 appears across a
50- resistive load, compute the actual peak envelope power (PEP).
Solution. Using Eq. (418), we get the normalized PEP:
1 1 2
(PPEP)norm = 2 [ max |g(t)|]2 = 2 Ac [1 + max m(t)]2
1 2
= 2 (500) [1 + 0.8]2 = 405 kW (4128)
Then the actual PEP for this AM voltage signal with a 50- load is
(PPEP)norm 4.50 * 105

(PPEP)actual = = = 8.1 kW (4129)
RL 50
SA45 Sampling Methods for Bandpass Signals Suppose that a bandpass signal s(t) is
to be sampled and that the samples are to be stored for processing at a later time. As shown
in Fig. 432a, this bandpass signal has a bandwidth of BT centered about fc, where fc BT and
BT 7 0. The signal s(t) is to be sampled by using any one of three methods shown in Fig. 432.
For each of these sampling methods, determine the minimum sampling frequency (i.e., minimum
clock frequency) required, and discuss the advantages and disadvantages of each method.
Solution.
Method I Referring to Fig. 432a, we see that Method I uses direct sampling as described in
Chapter 2. From Eq. (2168), the minimum sampling frequency is (fs)min = 2B, where B is the
highest frequency in the signal. For this bandpass signal, the highest frequency is B = fc + BT2.
Thus, for Method I, the minimum sampling frequency is
(fs)min = 2fc + BT Method I (4130)
For example, if fc = 100 MHz and BT = 1 MHz, a minimum sampling frequency of (fs)min =
201 MHz would be required.

If s(t) is a current signal (instead of a voltage signal), then (Ps)actual = (Is)2rmsRL.

Thanks to Professor Christopher S. Anderson, Department of Electrical and Computer Engineering,
University of Florida, for suggesting Method II.
|S(f)|
BT
fc fc
f
Sampler
s(t)
Bandpass signal input Sampled output
Clock
(a) Method IDirect sampling
Down converter
Sampler
s(t) Bandpass
filter
Bandpass Sampled output
signal input
2cos v0 t
Clock
Local
oscillator
(b) Method IIDown conversion and sampling
Sampler
Low-pass In-phase sampled output
filter
2cos vc t
s(t)
Oscillator Clock
Bandpass
signal input 2sin vc t Sampler
Quad-phase sampled output
Low-pass
filter
(c) Method IIIIQ (in-phase and quad-phase) sampling
Figure 432 Three methods for sampling bandpass signals.
Method II Referring to Fig. 432b, we see that Method II down converts the bandpass signal
to an IF, so that the highest frequency that is to be sampled is drastically reduced. For maximum
reduction of the highest frequency, we choose the local oscillator frequency to be f0 = fc - BT2.

Low-side LO injection is used so that any asymmetry in the two sidebands of s(t) will be preserved in the
same way in the down-converted signal.
The highest frequency in the down-converted signal (at the sampler input) is B = (fc + BT2) - f0 =
fc + BT2 - fc + BT2 = BT, and the lowest frequency (in the positive-frequency part of the down-
converted signal) is (fc - BT2) - f0 = fc - BT2 - fc + BT2 = 0. Using Eq. (2168), we find that
the minimum sampling frequency is
(fs)min = 2BT Method II (4131)
when the frequency of the LO is chosen to be f0 = fc - BT2. For this choice of LO frequency,
the bandpass filter becomes a low-pass filter with a cutoff frequency of BT. Note that Method
II gives a drastic reduction in the sampling frequency (an advantage) compared with Method I.
For example, if fc = 100 MHz and BT = 1 MHz, then the minimum sampling frequency is now
(fs)min = 2 MHz, instead of the 201 MHz required in Method I. However, Method II requires
the use of a down converter (a disadvantage). Note also that (fs)min of Method II, as specified
by Eq. (4131), satisfies the (fs)min given by the bandpass sampling theorem, as described by
Eq. (431).
Method II is one of the most efficient ways to obtain samples for a bandpass signal.
When the bandpass signal is reconstructed from the sample values with the use of Eq. (2158)
and (2160), the down-converted bandpass signal is obtained. To obtain the original bandpass
signal s(t), an up-converter is needed to convert the down-converted signal back to the original
bandpass region of the spectrum.
Method II can also be used to obtain samples of the quadrature (i.e., I and Q) components
of the complex envelope. From Fig. 432b, the IF signal at the input to the sampler is
vIF(t) = x(t) cos vIF t - y(t) sin vIF t
where fIF(t) = BT2. Samples of x(t) can be obtained if v IF(t) is sampled at the times correspond-
ing to cos vIF t = 1 (and sin vIF t = 0). This produces BT samples of x(t) per second. Likewise,
samples of y(t) are obtained at the times when sin vIF t = 1 (and cos vIF t = 0). This produces BT
samples of y(t) per second. The composite sampling rate for the clock is fs = 2BT. Thus, the sam-
pler output contains the following sequence of I and Q values: x, -y, -x, y, x, -y, ... The sampling
clock can be synchronized to the IF phase by using carrier synchronization circuits. Method III
uses a similar approach.
Method III From Fig. 432c, Method III uses in-phase (I) and quadrature-phase (Q) product
detectors to produce the x(t) and y(t) quadrature components of s(t). (This was discussed in
Sec. 416 and illustrated in Fig. 431.) The highest frequencies in x(t) and y(t) are B = BT2.
Thus, using Eq. (2168), we find that the minimum sampling frequency for the clock of the I
and Q samplers is
(fs) min = BT (each sampler) Method III (4132)
Because there are two samplers, the combined sampling rate is (fs)min overall = 2BT. This also
satisfies the minimum sampling rate allowed for bandpass signals as described by Eq. (431).
Thus, Method III (like Method II) gives one of the most efficient ways to obtain samples of
bandpass signals. For the case of fc = 100 MHz and BT = 1 MHz, an overall sampling rate of
2 MHz is required for Method III, which is the same as that obtained by Method II. Because IQ
samples have been obtained, they may be processed by using DSP algorithms to perform
equivalent bandpass filtering, as described by Sec. 45, or equivalent modulation of another

type, as described by Sec. 42. If desired, the original bandpass signal may be reconstructed
with the use of Eq. (432).
SA46 Frequency Synthesizer Design for a Receiver LO. Design a frequency synthesizer
for use as the local oscillator in an AM superheterodyne radio. The radio has a 455-kHz IF and can
be tuned across the AM band from 530 kHz to 1,710 kHz in 10-kHz steps. The synthesizer uses a
1-MHz reference oscillator and generates a high-side LO injection signal.
Solution. Referring to Eq. (459) and Fig. 429 for the case of down conversion and high-
side injection, we find that the required frequency for the LO is f0 = fc + fIF. If fc = 530 kHz
and fIF = 455 kHz, the desired synthesizer output frequency is f0 = 985 kHz. Referring to the
block diagram for the frequency synthesizer (Fig. 425), we select the frequency of the
mixer input signal, v in (t), to be 5 kHz, which is one-half the desired 10-kHz step. Then
M = fxfin = 1,000 kHz5 kHz = 200, and an integer value can be found for N to give the
needed LO frequency. Using Eq. (4112), we obtain N = f0fin. For f0 = 985 kHz and fin = 5
kHz, we get N = 197. Thus, to tune the radio to fc = 530 kHz, the required values of M and N
are M = 200 and N = 197. In a similar way, other values for N can be obtained to tune the
radio to 540, 550, ..., 1,710 kHz. (M remains at 200.) Table 46 lists the results. The selected
values for M and N are kept small in order to minimize the spurious sideband noise on the
synthesized LO signal. M and N are minimized by making the size of the step frequency, fin,
as large as possible.
The spectral sideband noise on the synthesizer output signal is minimized by using a
low-noise reference oscillator and a low-noise VCO and by choosing a small value of N for
the reduction in the number of intermodulation noise components on the synthesized signal.
The bandwidth of the loop filter is also minimized, but if it is too small, the pull-in range will
not be sufficient for reliable locking of the synthesizer PLL when it is turned on. In this
example, N can be reduced by a factor of about 12 the IF frequency is chosen to be 450 kHz
instead of 455 kHz. For example, for fIF = 450 kHz and fc = 530 kHz, we need f0 = 980 kHz.
This LO frequency is attained if M = 100 (for a step size of f in = 10 kHz) and N = 98,
compared with (see Table 46) M = 200 and N = 197, which was needed for the case when
fIF = 455 kHz.
TABLE 45 SOLUTION TO SA46 WITH DIVIDER RATIOS M AND N FOR AN AM RADIO-

FREQUENCY SYNTHESIZER
Reference frequency = 1,000 Hz; IF frequency = 455 kHz
Received frequency, fc Local oscillator frequency,

(kHz) f0 (kHz) M N
530 985 200 197

540 995 200 199
550 1,005 200 201
o o o o
1,700 2,155 200 431
1,710 2,165 200 433
Problems 305
PROBLEMS
jv t
41 Show that if v(t) = Re{g(t)e c }, Eqs. (41b) and (41c) are correct, where g(t) = x(t) + jy(t) =
ju(t)
R(t)e .
42 An AM signal is modulated by a waveform such that the complex envelope is
g(t) = Ac{1 + a[0.2 cos(p250t) + 0.5 sin(p2500t)]}
where Ac = 10. Find the value of a such that the AM signal has a positive modulation percentage
of 90%. Hint: Look at Ex. 43 and Eq. (55a).
43 A double-sideband suppressed carrier (DSB-SC) signal s(t) with a carrier frequency of 3.8 MHz
has a complex envelope g(t) = Acm(t). Ac = 50 V, and the modulation is a 1-kHz sinusoidal test
tone described by m(t) = 2sin (2p 1,000t). Evaluate the voltage spectrum for this DSB-SC signal.
44 A DSB-SC signal has a carrier frequency of 900 kHz and Ac = 10. If this signal is modulated by
a waveform that has a spectrum given by Fig. P33. Find the magnitude spectrum for this DSB-
SC signal.
45 Assume that the DSB-SC voltage signal s(t), as described in Prob. 43 appears across a 50-
resistive load.
(a) Compute the actual average power dissipated in the load.
(b) Compute the actual PEP.
46 For the AM signal described in Prob. 42 with a = 0.5, calculate the total average normalized
power.
47 For the AM signal described in Prob. 42 with a = 0.5, calculate the normalized PEP.
48 A bandpass filter is shown in Fig. P48.
L C
v1 (t) R v2 (t)
Figure P48
(a) Find the mathematical expression for the transfer function of this filter, H( f ) = V2( f )V1( f ),
as a function of R, L, and C. Sketch the magnitude transfer function |H( f )|.
(b) Find the expression for the equivalent low-pass filter transfer function, and sketch the corre-
sponding low-pass magnitude transfer function.
49 Let the transfer function of an ideal bandpass filter be given by
1, f + fc 6 BT>2
H(f) =1, f - fc 6 BT>2
L 0, f elsewhere
where BT is the absolute bandwidth of the filter.
(a) Sketch the magnitude transfer function |H(f)|.

(b) Find an expression for the waveform at the output, v2(t), if the input consists of the pulsed
carrier
v1(t) = A(t>T) cos (vct)
(c) Sketch the output waveform v2(t) for the case when BT = 4T and fc BT.
(Hint: Use the complex-envelope technique, and express the answer as a function of the sine
integral, defined by
u
sin l
Si(u) = dl
L0 l
The sketch can be obtained by looking up values for the sine integral from published tables
[Abramowitz and Stegun, 1964] or by numerically evaluating Si (u).
410 Examine the distortion properties of an RC low-pass filter (shown in Fig. 215). Assume that the
filter input consists of a bandpass signal that has a bandwidth of 1 kHz and a carrier frequency of
15 kHz. Let the time constant of the filter be t0 = RC = 105 s.
(a) Find the phase delay for the output carrier.
(b) Determine the group delay at the carrier frequency.
(c) Evaluate the group delay for frequencies around and within the frequency band of the signal.
Plot this delay as a function of frequency.
(d) Using the results of (a) through (c), explain why the filter does or does not distort the band-
pass signal.
411 A bandpass filter as shown in Fig. P411 has the transfer function
Ks
H(s) =
s + (v0/Q)s + v20
2
where Q = R3C>L, the resonant frequency is f0 = 1>(2p3LC), v0 = 2pf0, K is a constant,

and values for R, L, and C are given in the figure. Assume that a bandpass signal with fc = 4 kHz
and a bandwidth of 200 Hz passes through the filter, where f0 = fc .
R=400
L=1.583 mH C=1 mF
Figure P411
(a) Using Eq. (439), find the bandwidth of the filter.
(b) Plot the carrier delay as a function of f about f0.
(c) Plot the group delay as a function of f about f0.
(d) Explain why the filter does or does not distort the signal.
412 An FM signal is of the form
t
s(t) = Ac cos cvct + Df m(s) ds d
L-q
Problems 307
where m(t) is the modulating signal and vc = 2pf c, in which fc is the carrier frequency. Show that
the functions g(t), x(t), y(t), R(t), and u(t), as given for FM in Table 41, are correct.
413 The output of a FM transmitter at 96.9 MHz delivers 25kw average power into an antenna system
which presents a 50- resistive load. Find the value for the peak voltage at the input to the
antenna system.
414 Let a modulated signal,
s(t) = 100 sin(vc + va)t + 500 cos vct - 100 sin(vc - va)t
where the unmodulated carrier is 500 cos vct.

(a) Find the complex envelope for the modulated signal. What type of modulation is involved?
What is the modulating signal?
(b) Find the quadrature modulation components x(t) and y(t) for this modulated signal.
(c) Find the magnitude and PM components R(t) and u (t) for this modulated signal.
(d) Find the total average power, where s(t) is a voltage waveform that is applied across a 50-
load.
415 Find the spectrum of the modulated signal given in Prob. 414 by two methods:
(a) By direct evaluation using the Fourier transform of s(t).
(b) By the use of Eq. (412).
416 Given a pulse-modulated signal of the form
s(t) = e-at cos [1vc + v2t]u(t)
where a, vc, and v are positive constants and the carrier frequency, wc v,
(a) Find the complex envelope.
(b) Find the spectrum S( f ).
(c) Sketch the magnitude and phase spectra |S( f)| and u(f) = l S(f).
417 In a digital computer simulation of a bandpass filter, the complex envelope of the impulse
response is used, where h(t) = Re[k(t) ejvct], as shown in Fig. 43. The complex impulse
response can be expressed in terms of quadrature components as k(t) = 2hx(t) + j2hy(t),
1
where hx(t) = 12 Re[k(t)] and hy(t) = 2 Im [k(t)]. The complex envelopes of the input and
output are denoted, respectively, by g1(t) = x1(t) + jy1(t) and g2(t) = x2(t) + jy2(t). The bandpass
filter simulation can be carried out by using four real baseband filters (i.e., filters having real
impulse responses), as shown in Fig. P417. Note that although there are four filters, there are
only two different impulse responses: hx(t) and hy(t).
(a) Using Eq. (422), show that Fig. P417 is correct.
(b) Show that hy(t) K 0 (i.e., no filter is needed) if the bandpass filter has a transfer function with
Hermitian symmetry about fcthat is, if H(-f + fc) = H(f + fc), where | f | 6 BT 2 and BT
is the bounded spectral bandwidth of the bandpass filter. This Hermitian symmetry implies
that the magnitude frequency response of the bandpass filter is even about fc and the phase
response is odd about fc.
418 Evaluate and sketch the magnitude transfer function for (a) Butterworth, (b) Chebyshev, and
(c) Bessel low-pass filters. Assume that fb = 10 Hz and P = 1.
419 Plot the amplitude response, the phase response, and the phase delay as a function of frequency
for the following low-pass filters, where B = 100 Hz:
Baseband filter
hx(t)

x1(t) x2(t)

Baseband filter
hy(t)
g1(t) g2(t)
Baseband filter
hy(t)

y1(t) y2(t)

Baseband filter
hx(t)
Figure P417
(a) Butterworth filter, second order:

1
H( f) =
1 + 32(jf>B) + (jf>B)2
(b) Butterworth filter, fourth order:
1
H(f) =
[1 + 0.765( jf>B) + (jf>B)2 ][1 + 1.848(jf>B) + (jf>B)2 ]
Compare your results for the two filters.

420 Assume that the output-to-input characteristic of a bandpass amplifier is described by Eq. (442)
and that the linearity of the amplifier is being evaluated by using a two-tone test.
(a) Find the frequencies of the fifth-order intermodulation products that fall within the amplifier
bandpass.
(b) Evaluate the levels for the fifth-order intermodulation products in terms of A1, A2, and
the Ks.
421 An amplifier is tested for total harmonic distortion (THD) by using a single-tone test. The output
is observed on a spectrum analyzer. It is found that the peak values of the three measured har-
monics decrease according to an exponential recursion relation Vn + 1 = Vn e-n, where n = 1, 2, 3.
What is the THD?
422 The nonlinear outputinput characteristic of an amplifier is
vout(t) = 5vin(t) + 1.5v2in(t) + 1.5v3in(t)
Problems 309
Assume that the input signal consists of seven components:

1 4 6 1
vin(t) = + 2 a cos[(2k - 1)pt]
2 p k = 1 (2k - 1)2
(a) Plot the output signal and compare it with the linear output component 5vin(t).
(b) Take the FFT of the output vout(t), and compare it with the spectrum for the linear output
component.
423 For a bandpass limiter circuit, show that the bandpass output is given by Eq. (455), where
K = (4p)A0. A0 denotes the voltage gain of the bandpass filter, and it is assumed that the gain is
constant over the frequency range of the bandpass signal.
424 Discuss whether the Taylor series nonlinear model is applicable to the analysis of (a) soft limiters
and (b) hard limiters.
425 Assume that an audio sine-wave testing signal is passed through an audio hard limiter circuit.
Evaluate the total harmonic distortion (THD) on the signal at the limiter output.
426 Using the mathematical definition of linearity given in Chapter 2, show that the analog switch
multiplier of Fig. 410 is a linear device.
427 An audio signal with a bandwidth of 10 kHz is transmitted over an AM transmitter with a carrier
frequency of 1.0 MHz. The AM signal is received on a superheterodyne receiver with an envelope
detector. What is the constraint on the RC time constant for the envelope detector?
428 Assume that an AM receiver with an envelope detector is tuned to an SSB-AM signal that has a
modulation waveform given by m(t). Find the mathematical expression for the audio signal that
appears at the receiver output in terms of m(t). Is the audio output distorted?
429 Referring to Table 41, find the equation that describes the output of an envelope detector as a
function of m(t), if the input is a
(a) DSB-SC signal.
(b) FM signal.
430 Evaluate the sensitivity of the zero-crossing FM detector shown in Fig. 418. Assume that the
differential amplifier is described by vout(t) = A [v2(t) - v3(t)], where A is the voltage gain of the
amplifier. In particular, show that vout = Kfd, where fd = fi - fc, and find the value of the sensitiv-
ity constant K in terms of A, R, and C. Assume that the peak levels of the monostable outputs Q

and Q are 4 V (TTL circuit levels).
431 Using Eq. (4100), show that the linearized block diagram model for a PLL is given by
Fig. 422.
432 Show that Eq. (4101) describes the linear PLL model as given in Fig. 422.
433 Using the Laplace transform and the final value theorem, find an expression for the steady-state
phase error, limt S q ue(t), for a PLL as described by Eq. (4100). [Hint: The final value theorem
is limt S q f(t) = lims S0 sF(s).]
434 Assume that the loop filter of a PLL is a low-pass filter, as shown in Fig. P434.
0(f)
(a) Evaluate the closed-loop transfer function H(f) = for a linearized PLL.
i(f)
(b) Sketch the Bode plot [|H(f )|]dB 20 log |H(f )| for this PLL.
=
F(f)
Figure P434
435 Assume that the phase noise characteristic of a PLL is being examined. The internal phase noise
of the VCO is modeled by the input un(t), as shown in Fig. P435.
i(t) v2(t)
Kd F1(f)

0(t) VCO
v
F2(f)=
K
j2f

n(t)
0(t)
Figure P435
(a) Find an expression for the closed-loop transfer function 0(f)n(f), where ui(t) = 0.
(b) If F1(f) is the low-pass filter given in Fig. P434, sketch the Bode plot [|0 (f)n(f)|]dB for
the phase noise transfer function.
436 The input to a PLL is vin(t) = A sin(v0t + ui). The LPF has a transfer function F(s) = (s + a)s.
(a) What is the steady-state phase error?
(b) What is the maximum hold-in range for the noiseless case?
437 (a) Refer to Fig. 425 for a PLL frequency synthesizer. Design a synthesizer that will cover a
range of 144 to 148 MHz in 5-kHz steps, starting at 144.000 MHz. Assume that the frequency
standard operates at 5 MHz, that the M divider is fixed at some value, and that the N divider is
programmable so that the synthesizer will cover the desired range. Sketch a diagram of your
design, indicating the frequencies present at various points of the diagram.
(b) Modify your design so that the output signal can be frequency modulated with an audio-
frequency input such that the peak deviation of the RF output is 5 kHz.
438 Assume that an SSB-AM transmitter is to be realized using the AMPM generation technique, as
shown in Fig. 427.
(a) Sketch a block diagram for the baseband signal-processing circuit.
(b) Find expressions for R(t), u(t), and v(t) when the modulation is m(t) = A1 cos v1t +
A2 cos v2t.
Problems 311
439 Rework Prob. 438 for the case of generating an FM signal.

440 Assume that an SSB-AM transmitter is to be realized using the quadrature generation technique,
as shown in Fig. 428.
(a) Sketch a block diagram for the baseband signal-processing circuit.
(b) Find expressions for x(t), y(t), and v(t) when the modulation is m(t) = A1 cos v1t + A2 cos v2t.
441 Rework Prob. 4 40 for the case of generating an FM signal.
442 An FM radio is tuned to receive an FM broadcasting station of frequency 96.9 MHz. The radio is
of the superheterodyne type with the LO operating on the high side of the 96.9-MHz input and
using a 10.7-MHz IF amplifier.
(a) Determine the LO frequency.
(b) If the FM signal has a bandwidth of 180 kHz, give the requirements for the RF and IF filters.
(c) Calculate the frequency of the image response.
443 A dual-mode cellular phone is designed to operate with either cellular phone service in the
900-MHz band or PCS in the 1,900-MHz band. The phone uses a superheterodyne receiver with
a 500-MHz IF for both modes.
(a) Calculate the LO frequency and the image frequency for high-side injection when the phone
receives an 880-MHz signal.
(b) Calculate the LO frequency and the image frequency for low-side injection when the phone
receives a 1,960-MHz PCS signal.
(c) Discuss the advantage of using a 500-MHz IF for this dual-mode phone.
(Note: Cellular and PCS systems are described in Chapter 8.)
4 44 A superheterodyne receiver is tuned to a station at 20 MHz. The local oscillator frequency is
80 MHz and the IF is 100 MHz.
(a) What is the image frequency?
(b) If the LO has appreciable second-harmonic content, what two additional frequencies are
received?
(c) If the RF amplifier contains a single-tuned parallel resonant circuit with Q = 50 tuned to
20 MHz, what will be the image attenuation in dB?
445 An SSB-AM receiver is tuned to receive a 7.225-MHz lower SSB (LSSB) signal. The LSSB sig-
nal is modulated by an audio signal that has a 3-kHz bandwidth. Assume that the receiver uses a
superheterodyne circuit with an SSB IF filter. The IF amplifier is centered on 3.395 MHz. The LO
frequency is on the high (frequency) side of the input LSSB signal.
(a) Draw a block diagram of the single-conversion superheterodyne receiver, indicating frequen-
cies present and typical spectra of the signals at various points within the receiver.
(b) Determine the required RF and IF filter specifications, assuming that the image frequency is
to be attenuated by 40 dB.
4 46 (a) Draw a block diagram of a superheterodyne FM receiver that is designed to receive FM sig-
nals over a band from 144 to 148 MHz. Assume that the receiver is of the dual-conversion
type (i.e., a mixer and an IF amplifier, followed by another mixer and a second IF amplifier),
where the first IF is 10.7 MHz and the second is 455 kHz. Indicate the frequencies of the sig-
nals at different points on the diagram, and, in particular, show the frequencies involved when
a signal at 146.82 MHz is being received.
(b) Replace the first oscillator by a frequency synthesizer such that the receiver can be tuned in
5-kHz steps from 144.000 to 148.000 MHz. Show the diagram of your synthesizer design and
the frequencies involved.
447 An AM broadcast-band radio is tuned to receive a 1,080-kHz AM signal and uses high-side LO
injection. The IF is 455 kHz.
(a) Sketch the frequency response for the RF and IF filters.
(b) What is the image frequency?
448 Commercial AM broadcast stations operate in the 540- to 1,700-kHz band, with a transmission
bandwidth limited to 10 kHz.
(a) What are the maximum number of stations that can be accommodated?
(b) If stations are not assigned to adjacent channels (in order to reduce interference on receivers
that have poor IF characteristics), how many stations can be accommodated?
(c) For 455-kHz IF receivers, what is the band of image frequencies for the AM receiver that uses
a down-converter with high-side injection?

Bandpass Signaling

Uploaded by

Copyright:

Available Formats

Bandpass Signaling

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bandpass Signaling

Uploaded by

Copyright:

Available Formats

Chapter

41 COMPLEX ENVELOPE REPRESENTATION OF BANDPASS WAVEFORMS

Definitions: Baseband, Bandpass, and Modulation

Complex Envelope Representation

u 1t2 ! g1t2 = tan-1 a b

v(t) = Ree c0 + 2 a cnejnv0t f

this function is an analytic function of a complex variable.

so that Eq. (41a) follows, where

Example 41 IN-PHASE AND QUADRATURE MODULATED SIGNALING

42 REPRESENTATION OF MODULATED SIGNALS

43 SPECTRUM OF BANDPASS SIGNALS

TABLE 41 COMPLEX ENVELOPE FUNCTIONS FOR VARIOUS TYPES OF MODULATION a

Corresponding Quadrature Modulation

SSB-EV b Ac |1 + m(t2| ;lNn [1 + m(t2] NL m(t2  -1 is required so that ln()

and the PSD of the waveform is

v(t) = Re{g(t)ejvct} = 12 g(t)ejvct + 12 g*(t)e - jvct

Example 42 SPECTRUM FOR A QUADRATURE MODULATED SIGNAL

Rv(t) = 8v(t) v(t + t)9 = 8Re{g(t)ejvct} Re{g(t + t) ejvc(t + t)}9

Using the identity (see Prob. 274)

where c2 = g(t)ejvct and c1 = g(t + t)ejvc(t + t), we get

THEOREM. The total average normalized power of a bandpass waveform v(t) is

Also, from Eq. (416),

But |g(t)| is always real, so

This is equivalent to evaluating the average power in an unmodulated RF sinusoid that

THEOREM. The normalized PEP is given by

Example 43 AMPLITUDE-MODULATED SIGNAL

g(t) = Ac[1 + m(t)]

so that the spectrum of the complex envelope is

G(f) = Acd(f) + AcM(f) (419)

Using Eq. (49), we obtain the AM signal waveform

s(t) = Ac[1 + m(t)] cos vct

See Example4_3.m for a plot of an AM signal that is modulated by a sinusoid.

(a) Magnitude Spectrum of Modulation

|S(f)| Discrete carrier term

45 BANDPASS FILTERING AND LINEAR DISTORTION

Equivalent Low-Pass Filter

v1(t)=Re [g1(t)ejct] Bandpass filter v2(t)=Re [g2(t )ejct]

(a) Bandpass Filter

(b) Typical Bandpass Filter Frequency Response

1 Equivalent low-pass filter 1

(c) Equivalent (Complex Impulse Response) Low-pass Filter

(d) Typical Equivalent Low-pass Filter Frequency Response

+ G*1(-f - fc)K(f - fc) + G*1(-f - fc)K*(-f - fc)]

312 G2(f-fc)4 + 312 G*2(-f - fc)4

= 312 G1(f-fc) 12 K(f-fc)4 + 312 G*1(-f-fc) 12 K* (-f-fc)4 (426)

The amplitude response is constant. That is,

u(f) = -2pfTg + u0 (428)

where u0 is a phase-shift constant, as shown in Fig. 44b. If u0 happens to be nonzero,

(a) Magnitude Response

(b) Phase Response

H(f) = Aej(-2pfTg + u0) = (Aeju0)e-j2pfTg (429)

v1(t) = x(t) cos vct - y(t) sin vct

Using Eq. (428), we obtain

v2(t) = Ax(t - Tg) cos[vct + u(fc)] - Ay(t - Tg) sin[vct + u(fc)]

where, by the use of Eq. (2150b) evaluated at f = fc,

u(fc) = -vcTg + u0 = -2pfc Td

Thus, the output bandpass signal can be described by

Example 44 GROUP DELAY FOR A RC LOW-PASS FILTER

46 BANDPASS SAMPLING THEOREM

SSB-EV b Ac |1 + m(t2| ;lNn [1 + m(t2] NL m(t2 -1 is required so that ln()

v1(t)=Re [g1(t)ejct] Bandpass filter v2(t)=Re [g2(t )ejct]

+ G1(-f - fc)K(f - fc) + G1(-f - fc)K*(-f - fc)]

= 312 G1(f-fc) 12 K(f-fc)4 + 312 G1(-f-fc) 12 K (-f-fc)4 (426)

vLO(t)=A0 cos (0 t)

vLO(t)=A0 cos (0 t)