Steer RF Chapter1

CHAPTER 1
Modulation, Transmitters
and Receivers
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 RF Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Analog Modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Digital Modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Ampliers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.6 Noise and Nonlinear Distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
1.7 Active Switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
1.8 Mixers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
1.9 Early Receiver Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
1.10 Modern Transmitter Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
1.11 Modern Receiver Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
1.12 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
1.1 Introduction
The frontend of a radio frequency (RF) communication receiver combines a
number of subsystems in cascade to achieve several objectives. Filters and
matching networks provide frequency selectivity to eliminate interfering signals.
Ampliers manage noise levels by boosting both received signals and signals to be
transmitted. Mixers coupled with oscillators translate the modulated information
from one frequency to another.
There are only a few types of receiver and transmitter architectures. In a receiver,
the central idea is to take information superimposed on an RF signal or carrier and
convert it to a lower frequency form which can be directly applied to a speaker
or digitized. In a cellular communication system, the low-frequency signal, often
called the baseband signal, could have a bandwidth of 30 kHz to 5 MHz and the
carrier frequency could be 500 MHz to 2 GHz. A transmitter takes the baseband
signal and superimposes it on an RF carrier which can be more easily radiated
2 MICROWAVE AND RF DESIGN: A SYSTEMS APPROACH
ANTENNA
BPF BPF
RF RF RF
LPF
MIXER MIXER
AMPLIFIER AMPLIFIER
ADC DSP
2
LO
1
LO
IF
1
IF
2
IF
1
IF
1
IF
2
(a)
IF
BPF
RF
MIXER
AMPLIFIER
ANTENNA
IF
LO
LPF ADC DSP
RF RF
(b)
IF
BPF
ANTENNA
LPF DAC DSP
RF RF RF IF
LO
MIXER
AMPLIFIER
(c)
Figure 1-1 Unilateral RF frontend: (a) a receiver with two mixing stages; (b) a receiver with
one heterodyne stage; and (c) a one-stage transmitter.
into space and propagates easily from one antenna to another. The essential
receiver and transmitter architectures are shown in Figure 1-1. In a receiver
mixers down-convert information superimposed on an RF carrier to a lower
frequency that can be directly connected to speakers or digitized by an analog-
to-digital converter (ADC). With a transmitter, the low-frequency information-
bearing signal is translated to a frequency that can be more easily radiated. The
most common receiver architecture is shown in Figure 1-1(a). First, an antenna
collects a broad portion of the electromagnetic spectrum. Antennas have relatively
lowfrequency selectivity (they have broad bandwidth) and unwanted signal levels
can be large, so additional ltering by a bandpass lter (BPF) is required to reduce
the range of voltages presented to the rst amplier. Eventually this signal is
digitized by an ADC but to do this the frequency of the information-carrying part
of the signal must be reduced. The stepping down of frequency is accomplished
by a mixer stage. With the mixer driven by a large local oscillator (LO) signal, the
output at the intermediate frequency (IF) is at the difference frequency of the RF
and LO(see Figure 1-2). Thus f
IF
= f
RF
f
LO
(although sometimes the LOis above
the RF so that f
IF
= f
LO
f
RF
). LOs generally have noise close to the operating
frequency so that there is a limit on how close the RF and LO can be in frequency
without oscillator noise appearing at the IF. If there is a single mixer, then the IF
MODULATION, TRANSMITTERS AND RECEIVERS 3
MIXER RF IF
LO
FREQUENCY
IF
RF
f f
LO
RF
LO
IF
DC
f
(a) (b)
Figure 1-2 Simple mixer circuit: (a) block diagram; and (b) spectrum.
may still be too high. A solution is to use two stages of mixing. A BPF between
the mixing (or heterodyning) stages further blocks unwanted signals. Eventually a
lowpass lter (LPF) allows only the nal IF (here IF
2
) to be presented to the ADC.
Once digitized, it is possible to further lter the intended signal which originally
appeared as modulation at the RF. Aone-stage receiver, see Figure 1-1(b), generally
requires a better ADC, but the elimination of a mixing stage reduces cost and size.
The architecture of a transmitter is similar, with a key difference being the digital-
to-analog converter (DAC) (see Figure 1-1(c)).
The major active elements in the RF frontend of both the transmitter and receiver
are the ampliers, mixers, and oscillators. These subsystems have much in
common using nonlinear devices to convert power at DC to power at RF. In the
case of mixers, power at the local oscillator(LO) is also converted to power at
RF. The frontend of a typical cellphone is shown in Figure 1-3. The components
here are generally implemented in a module and use different technologies for
the various elements, optimizing cost and performance. There are many variants
of the architecture shown here. At one extreme a module is used with all of
the components packed in a shielded structure perhaps 1 cm on a side and 2
3 mm thick. Another extreme is a single-chip implementation, usually in BiCMOS
(bipolar with complementary metal oxide semiconductor, CMOS), SiGe (silicon
germanium) technology, or high performance CMOS called RF CMOS. However,
it is necessary to use a gallium arsenide GaAs device to efciently achieve the
hundreds of milliwatts typically required.
Return now to the mixer-based transceiver (for receiver and transmitter)
architecture shown in a multichip form in Figure 1-3. Here, a single antenna is
used, and either a duplexer (a combined lowpass and highpass lter) or a switch
is used to separate the (frequency-spaced) transmit and receive paths. If the system
protocol requires transmit and receive at the same time, a duplexer is required to
separate the transmit and receive paths. This lter tends to be large, lossy, or costly
(depending on the technology used). Consequently a transistor switch is preferred
if the transmit and receive signals operate in different time slots. In the receive
path, a CMOS or BiCMOS chip initially amplies the low-level received signal,
and so low noise is important. This amplier is thus called a low-noise amplier
(LNA). The amplied receive signal is then bandpass ltered and frequency down-
converted by a mixer (indicated by a circle with a cross in it) to IF that can
be sampled by an ADC to produce a digital signal that is further processed by
/
RECEIVE
TRANSMIT
Bipolar
or
Diplexer
BPF
BPF
DMOS
GaAs
POWER
AMPLIFIER
CMOS
DSP
BPF
LO
T/R Switch
Figure 1-3 RF frontend organized as multiple chips.
digital signal processing (DSP). Variants of this architecture include having two
down-conversion stages, and a variant with no mixing that relies instead on direct
conversion of the receive signal using a subsampling ADC. In the transmit path
the architecture is reversed, with a DAC driven by the DSP chip that produces
an information-bearing signal at the IF which is then frequency up-converted by
a mixer, bandpass ltered, and amplied by what is called a power amplier to
generate the hundreds of milliwatts required. An alternative transmitter design
is direct digital synthesis (DDS), which bypasses the conversion stage. Direct
conversion and DDS are difcult to implement, but are essential for the highly
desired single or few chip solution.
This chapter describes the operation and design strategies for the RF frontend ar-
chitecture of Figure 1-3, looking at ampliers, mixers, switches, and oscillators.
This architecture is used in most high-performance RF and microwave communi-
cation and radar systems. While the subsystems are preferably linear at RF, this
can only be approximated, as the active devices used are intrinsically nonlinear.
Performance is limited fundamentally by distortion, which is related to the charac-
teristics of the RF signal, and this in turn is determined by the modulation scheme
that impresses information on an RF carrier.
1.2 RF Signals
Radio frequency communication signals are engineered to trade off efcient use
of the electromagnetic (EM) spectrum with the complexity and performance of
the RF hardware required to process them. The process of converting baseband
(or low-frequency) information to RF is called modulation of which there are two
types: analog and digital modulation. In analog modulation, the RF signal has a
continuous range of values; in digital modulation, the output has a number of
prescribed discrete states. There are just a few modulation schemes that achieve
the optimum trade-offs of spectral efciency and ease of use with hardware
complexity. The major modulation schemes are
Analog modulation
AM Amplitude modulation
FM Frequency modulation
PM Phase modulation
Digital modulation
FSK Frequency shift keying
PSK Phase shift keying
MSK Minimum shift keying (a form of FSK)
GMSK Minimum shift keying using Gaussian ltered data
BPSK Binary phase shift keying
QPSK Quadrature PSK (QPSK is also referred to as quar-
ternary PSK, quadriphase PSK, and quadra PSK )
/4-DQPSK /4 Differential encoded QPSK
OQPSK Offset QPSK
8PSK 8-state phase shift keying
3/8-8PSK 3 /8, 8-state phase shift keying
16PSK 16-state phase shift keying
QAM Quadrature amplitude modulation
Frequency modulation, and the similar PM modulation schemes, are used in
analog cellular radio. With the addition of legacy AM, the three schemes are the
bases of analog radio. The other schemes are used in digital radio including digital
cellular radio. GMSK is used in the GSM cellular system and is a form of FSK and
produces a constant amplitude modulated signal. The FM, FSK, GMSK, and PM
techniques produce constant RF envelopes, thus no information is contained in
the amplitude of the signal. Therefore errors introduced into the amplitude of the
system are of no signicance and so efcient saturating-mode ampliers such as
class C can be used. So there is a trade-off in the complexity of RF design, choice
of modulation format and battery life. In contrast, the MSK, /4-DQPSK, 3/8-
8PSK, and QAM techniques do not result in constant RF envelopes, so information
is contained in the amplitude of the RF signal. Thus more sophisticated RF
processing hardware is required.
1.3 Analog Modulation
Wireless modulation formats in conventional narrowband radio are based on
modifying the properties of a carrier by slowly varying the amplitude and phase
of the carrier. The waveforms and spectra of common analog modulation formats
are shown in Figure 1-4.
1.3.1 Amplitude Modulation, AM
Amplitude modulation (AM) is the simplest analog modulation scheme to
implement. Here a signal is used to slowly vary the amplitude of the carrier
according to the level of the modulating signal. The modulating signal is generally
referred to as the baseband signal and it contains all of the information to
be transmitted or interpreted. The waveforms in Figure 1-4 are stylized as the
U
(c)
(b)
(d)
(e)
(a)
BASEBAND
CARRIER
FREQUENCY
MODULATION
AMPLITUDE
MODULATION
PHASE
MODULATION
(i) (ii)
L
FREQUENCY
TIME
Figure 1-4 Analog modulation showing (i) waveform and (ii) spectrum for (a) baseband
signal; (b) carrier; (c) carrier modulated using amplitude modulation; (d) carrier modulated
using frequency modulation; and (e) carrier modulated using phase modulation.
variation in the carrier is relatively fast. They are presented this way so that the
effects of modulation can be more easily interpreted. The baseband signal (Figure
1-4(a)) is shown as having a period that is not too far away from the period
of the carrier (Figure 1-4(b)). In reality, there would be hundreds or thousands
of RF cycles for each cycle of the baseband signal so that the frequency of the
baseband signal would have frequency components which are a tiny fraction of
the frequency of the carrier.
With AM (Figure 1-4(c)) the amplitude of the carrier is modulated and this results
in a broadening of the spectrum of the carrier, as shown in Figure 1-4(c)(ii). This
spectrum contains the original carrier component and upper and lower sidebands
designated as U and L, respectively. In AM, the two sidebands contain identical
information, so all the information would be transmitted if the carrier and one
of the sidebands were suppressed. With the carrier present, it is easy to receive a
signal by bandpass ltering the incoming modulated signal, rectifying the result,
and then lowpass ltering the rectied signal to remove harmonics of the baseband
signal.
An AM signal x(t) has the form
x(t) = A
c
[1 +my(t)] cos
c
t, (1.1)
where m is called the modulation index and y(t) is the baseband information-
bearing signal that has frequency components which are below the carrier radian
V
PEAK
V
AVERAGE
V
PEAK
=1/2
Envelope
AM(100%)
Carrier
(b)
(c)
(a)
Figure 1-5 AMshowing the relationship between the carrier and modulation envelope: (a)
carrier; (b) 100% amplitude modulated carrier; and (c) modulating or baseband signal.
frequency
c
. Provided that y(t) varies slowly relative to the carrier, that is, the
frequency components of y(t) are signicantly below the carrier frequency, x(t)
looks like a carrier whose amplitude varies slowly. To get an idea of how slowly
the amplitude varies in actual systems, consider an AM radio that broadcasts at
1 MHz (which is in the middle of the AM broadcast band). The highest frequency
component of the modulating signal corresponding to voice is about 4 kHz. Thus
the amplitude of the carrier takes 250 carrier cycles to go through a complete
amplitude variation. At all times a cycle of the carrier appears to be periodic, but
in fact it is not quite. It is common to refer to the modulated carrier as being quasi-
periodic and to the apparent carrier as being the pseudo-carrier.
The concept of the envelope of a modulated RF signal is introduced in Figure 1-5.
Figure 1-5(a) is the carrier; the AM-modulated carrier is shown in Figure 1-5(b).
The outline of the modulated carrier is called the envelope, and for AM this is
identical to the modulating signal. Both the envelope and the modulating signal
are shown in Figure 1-5(c). At the peak of the envelope, the RF signal has maximum
power (considering the power of a single RF cycle). Since we are dealing with
100% AM modulation, m = 1 in Equation (1.1) and there is no RF power when the
envelope is at its minimum.
One of the characteristics of various modulation formats is the ratio of the power
of the signal when the carrier is at its peak (i.e., the power in one cycle of the
carrier when the envelope is at its maximum) relative to its average value (the
power averaged over all time). This is called the peak-to-average ratio (PAR)
and is a good indicator of how sensitive a modulation format is to the effects of
nonlinearity of the RF hardware.
It is complex to determine the PAR for a general signal, but a good estimate can
be obtained by considering that the modulating signal is a sinewave. Let y(t)
(= cos
m
t) be a cosinusoidal modulating signal with radian frequency
m
. Then
(for AM)
x(t) = A
c
[1 +mcos
m
t] cos
c
t . (1.2)
Thus if just one quasi-period of this signal is considered (i.e., one variation of the
modulated signal at the carrier frequency), then the signal has a power that varies
with time.
Consider a voltage v(t) across a resistor of conductance G. The power of the signal,
or the average power, must be determined by integrating over all time, which is
work, and dividing by the time period yields the average power:
P
avg
= lim
1
2
Gv
2
(t) dt . (1.3)
Now, if v(t) is a cosinusoid, v(t) = Acos t, then
P
avg
= lim
1
2
_

A
2
c
Gcos
2
(t) dt
= lim
1
2
_

A
2
c
G
1
2
[1 + cos (2t)] dt
=
1
2
A
2
c
G
_
lim
1
2
_

1 dt + lim
1
2
_

cos (2t) dt
_
=
1
2
A
2
c
G (1.4)
In the above equation, a useful equivalence has been employed by observing that
the innite integral of a cosinusoid can be simplied to just integrating over one
period, T = 2/:
lim
1
2
_

cos
n
(t) dt =
1
T
_
T/2
T/2
cos
n
(t) dt (1.5)
where n is a positive integer. In power calculations there are a number of other
useful simplifying techniques based on trigonometric identities. Some of the ones
that will be used are the following:
cos Acos B =
1
2
[cos(AB) + cos(A+B)]
cos
2
A =
1
2
[1 + cos(2A)] (1.6)
lim
1
2
_

cos t dt =
1
T
_
T/2
T/2
cos (t) dt = 0 (1.7)
1
T
_
T/2
T/2
cos
2
(t) dt =
1
T
_
T/2
T/2
1
2
[cos (2t) + cos(0)] (1.8)
=
1
2T
_
_
T/2
T/2
cos (2t) dt +
_
T/2
T/2
1 dt
_
=
1
2T
(0 +T) =
1
2
. (1.9)
More trigonometric identities are given in appendix A.3 on page 576. Also, when
cosinusoids cos At and cos Bt having different frequencies (A ,= B) are multiplied
together, then
_

cos At cos Bt dt =
_

[cos(A+B)t + cos(A+B)t] dt = 0, (1.10)

and, in general, if A ,= B ,= 0,
_

cos At cos
n
Bt dt = 0 . (1.11)
Now the discussion returns to characterizing an AM signal by considering long-
term average power and the short-term power of the signal. The maximum
amplitude of the pseudo-carrier at its peak amplitude is, from Equation (1.2),
x
p
(t) = A
c
[1 +m] cos
c
t . (1.12)
Then the power (P
peak
) contained in the peak pseudo-carrier is obtained by
integrating over one period:
P
peak
=
1
T
_
T/2
T/2
Gx
2
(t) dt =
1
T
_
T/2
T/2
A
2
c
G(1 +m)
2
cos
2
(
c
t) dt
= A
2
c
G(1 +m)
2
1
T
_
T/2
T/2
cos
2
(
c
t) dt =
1
2
A
2
c
G(1 +m)
2
. (1.13)
The average power (P
avg
) of the modulated signal is obtained by integrating over
all time, so
P
avg
= lim
1
2
_

Gx
2
(t) dt
= A
2
c
G lim
1
2
_

[1 +mcos (
m
t)] cos (
c
t)
2
dt
= A
2
c
G lim
1
2
_

__
1 + 2mcos (
m
t) +m
2
cos
2
(
m
t)
cos
2
(
c
t)
_
dt
= A
2
c
G lim
1
2
_

_
cos
2
(
c
t) + 2mcos (
m
t) cos
2
(
c
t)
+m
2
cos
2
(
m
t) cos
2
(
c
t)
dt
= A
2
c
G
_
lim
1
2
_

cos
2
(
c
t) dt
+ lim
1
2
_

2mcos (
m
t) cos
2
(
c
t) dt
+ lim
1
2
_

m
2
cos
2
(
m
t) cos
2
(
c
t) dt
_
= A
2
c
G
_
1
2
+ 0 + lim
1
2
_

m
2
cos
2
(
m
t) cos
2
(
c
t) dt
_
= A
2
c
G
_
1
2
+m
2
lim
1
2
_

1
4
[1 + cos (2
m
t)] [1 + cos (2
c
t)] dt
_
P
avg
= A
2
c
G
_
1
2
+
m
2
4
lim
1
2
_

[1 + cos (2
m
t) + cos (2
c
t) + cos (2
m
t) cos (2
c
t)] dt
_
= A
2
c
G
_
1
2
+
m
2
4
_
lim
1
2
_

1 dt + lim
1
2
_

cos (2
m
t) dt
+ lim
1
2
_

cos (2
c
t) dt + lim
1
2
_

cos (2
m
t) cos (2
c
t) dt
__
= A
2
c
G
_
1/2 +m
2
(1/4 + 0 + 0 + 0)
=
1
2
A
2
c
G(1 +m
2
/2) . (1.14)
So the PAR of an AM signal (i.e. PAR
AM
) is
PAR
AM
=
P
peak
P
avg
=
1
2
A
2
c
G(1 +m)
2
1
2
A
2
c
G(1 +m
2
/2)
=
(1 +m)
2
1 +m
2
/2
.
For 100% AM described by m = 1, the PAR is
PAR
100%AM
=
(1 + 1)
2
1 + 1
2
/2
=
4
1.5
= 2.667 = 4.26 dB. (1.15)
In expressing the PAR in decibels the formula PAR
dB
= 10 log
10
(PAR) was used,
as the PAR is a power ratio. As an example, for 50% AM described by m = 0.5, the
PAR is
PAR
50%AM
=
(1 + 0.5)
2
1 + 0.5
2
/2
=
2.25
1.125
= 2 = 3 dB. (1.16)
The PAR is an important attribute of a modulation format and impacts the types of
circuit designs that can be used. It is much more challenging to achieve low levels
of distortion when the PAR is high.
It is tempting to consider if the lengthy integrations can be circumvented. Powers
can be added if the signal components (the tones making up the signal) are
uncorrelated. If they are correlated, then the complete integrations are required.
1
Consider two uncorrelated sinusoids of (average) powers P
1
and P
2
then the
average power of the composite signal is P
avg
= P
1
+P
2
. However, in determining
peak power, the RF cycle where the two sinusoids align is considered, and here
the voltages add to produce a sinewave with a higher amplitude. So peak power
applies to just one RF pseudo-cycle. Generally the voltage amplitude of the two
sinewaves would be added and then the power calculated. If the uncorrelated
carriers are modulated and the modulating signals (the baseband signals) are
uncorrelated then the average power can be determined in the same way, but
the peak power calculation is much more complicated. The integrations are the
only calculations that can always be relied on. They can be used with all signals,
including digitally modulated signals.
1
For the purposes here, two signals are uncorrelated if the integral of their product over all time and
all offsets is zero. That is, x(t) and y(t) are uncorrelated if C =
x(t)y(t + ) dt = 0 for all ,

otherwise they are correlated (or partly correlated). For a more complete denition see reference [1].
Input Signal
Carrier
frequency
deviation
f
m
FREQUENCY
Voltage
Time
f
Sidebands
FREQUENCY
3
f
c
f
m
-2
f
c
f
m
- f
c
f
m
+
f
c
f
m
+2
f
c
f
m
+3
BW
V
O
L
T
A
G
E
(a) (b)
f
c
Channel
V
O
L
T
A
G
E
FREQUENCY
(c)
Figure 1-6 Frequency modulation by a sinewave: (a) signal varying the frequency of
carrier; (b) spectrum of the resulting waveform; and (c) spectrum when modulated by a
continuous baseband signal.
1.3.2 Phase and Frequency Modulation, PM and FM
The two other analog modulation schemes commonly used are phase modulation
(PM) (Figure 1-4(e)) and frequency modulation (FM) (Figure 1-4(d)). The signals
produced by the two schemes are identical; the difference is how the signals are
generated. In PM, the phase of the carrier depends on the instantaneous level of
the baseband signal. In FM, the amplitude of the baseband signal determines the
frequency of the carrier. The result in both cases is that the bandwidth of the time-
varying signal is spread out, as seen in Figure 1-6. A receiver must compress the
spread-out information to recreate the original narrowband signal, and this can be
thought of as processing gain, as the compression of correlated signals signicantly
increases the tolerance to noise. As will be seen, processing gain is essential in
digital radio, which uses digital modulation. The peak amplitude of the RF phasor
is equal to the average amplitude and so the PAR is 1 or 0 dB. A summary of the
PAR of the primary analog modulated signals is given in Figure 1-7.
Frequency modulation was invented by Edwin H. Armstrong and patented in
1933. FM is virtually static free and clearly superior to AM radio. However, it was
not immediately adopted largely because AM radio was established in the 1930s,
and the adoption of FM would have resulted in the scrapping of a large installed
Carrier
AM(100%)
FM
PAR=4.3dB
PAR=0dB
(b)
(a)
(c)
ENVELOPE
ENVELOPE
Figure 1-7 Comparison of 100%AMand FMhighlighting the envelopes of both: (a) carrier;
(b) AM signal with envelope; and (c) FM or PM signal with the envelope being a straight
line or constant.
infrastructure (seen as a commercial catastrophe) and so the introduction of FM
was delayed by decades. The best technology does not always win immediately!
Commercial interests and the interests of those heavily invested in an alternative
technology have a great deal to do with the success of a technology.
Carsons Rule
Frequency and phase modulated signals have unlimited bandwidth but the
information content of the sidebands drops off rapidly. The bandwidth required to
reliably transmit a PM or FM signal is subjective but the best accepted criterion is
called Carsons bandwidth rule or just Carsons Rule [2, 3]. It provides an estimate
of the bandwidth capturing approximately 98% of the energy when a carrier is
frequency or phase modulated by a continuous spectrum baseband signal. An FM
signal is shown in Figure 1-6. In particular, Figures 1-6(a) and 1-6(b) show the FM
function and then the spectrum that results when a single sinewave modulates
the frequency of a carrier. As time passes, the carrier moves up and down in
frequency synchronously with the level of the input baseband signal. The level
(typically voltage) of the baseband signal determines the frequency deviation of
the carrier from its unmodulated value. The frequency shift when the modulating
signal is at its maximum amplitude is called the peak frequency deviation,
f, and the maximum frequency of the modulating frequency is f
m
. Figure 1-
6(c) shows the spectrum that results when the modulating signal, or baseband
signal, is continuous. There are multiple sidebands, with the relative strength of
each dependent on a bessel function of the highest modulation frequency, f
m
,
and the maximum frequency deviation, f. Carsons Rule, derived from these
considerations, is
Bandwidth required = 2 (f
m
+ f) . (1.17)
Narrowband and Wideband FM
The FM signal, as used in FM broadcast radio, is also called wideband FM, as
the maximum frequency deviation is much greater than the highest frequency of
the modulating or baseband signal, that is, f f
m
. A more spectrally efcient
form of FM is called narrowband FM, where f f
m
. Narrowband FM was
developed as a more bandwidth efcient form of FM, but of course digital radio
has passed this now and narrowband FM is no longer an important modulation
type. The trade-off is that narrowband FM, as opposed to wideband FM, requires
more sophisticated demodulation and hence more complex circuits are required. It
should also be noted that FM as used in conventional FM broadcast radio is being
phased out so that spectrum can be used more efciently.
1.3.3 Two-Tone Signal
A two-tone signal is a signal which is the sum of two cosinusoids. Thus
y(t) = X
A
cos(
A
t) +X
B
cos(
B
t) (1.18)
is a two-tone signal. Generally the frequencies of the two tones are close with the
concept being that the two tones both t within the passband of a bandpass lter,
so it would be reasonable to assume that the individual tones have frequencies
that are within 1% of each other. A two-tone signal is not a form of modulation but
is commonly used to characterize the performance of RF systems. The composite
signal would then look like a slowly varying pseudo-carrier not unlike an AM
signal. The tones are uncorrelated so that the average power of the composite
signal, y(t), is the sum of the powers of each of the individual tones. The peak
power of the composite signal is the peak pseudo-carrier, so y(t) has a peak
amplitude of X
A
+X
B
. Similar concepts apply to three-tone and n-tone signals.
Example 1.1 PAR of a Two-Tone Signal
What is the PAR of a two-tone signal with both tones having equal amplitude?
SOLUTION: Let X
A
= X
B
= X, the peak pseudo-carrier has amplitude 2X, and so the power of the
peak RF carrier is (2X)
2
= 4X
2
. The average power is proportional to X
2
B
+X
2
B
= X
2
+X
2
= 2X
2
,
as each one is independent of the other, and so the powers can be added.
PAR =
4X
2
2X
2
= 2 = 3 dB. (1.19)
1.4 Digital Modulation
Digital modulation was rst employed in sending telegraph signals wirelessly in
which a carrier was switched, or keyed, on and off to create pulses of the carrier
signal. This modulation is now known as amplitude shift keying (ASK), but today
1 0 1 0
time
time
time
time
time
BASEBAND
CARRIER
FREQUENCY
SHIFT KEYING
AMPLITUDE
SHIFT KEYING
PHASE
SHIFT KEYING
(c)
(b)
(d)
(e)
(a)
Figure 1-8 Modes of digital modulation: (a) modulating bit stream; (b) carrier; (c) carrier
modulated using amplitude shift keying (ASK); (d) carrier modulated using frequency shift
keying (FSK); and (e) carrier modulated using binary phase shift keying (BPSK).
this scheme is little used. Several digital modulation formats are shown in Figure
1-8. The fundamental characteristic of digital modulation is that there are discrete
states, each of which denes a symbol, with a symbol representing one or more
bits. In Figure 1-8, there are only two states representing one of two values for
a bit (0 or 1). With multiple states, groups of bits can be represented. There are
many digital modulation formats that have proved successful and many of these
are considered below. In modern communication schemes it is important to be able
to recover the original carrier, so it is important that the amplitude of the carrier
not be small for an extended period of time as it is in the ASK scheme illustrated
in Figure 1-8(c)
1.4.1 Phase Shift Keying, PSK
The waveforms and spectrum of a PSK modulated signal are shown in Figure 1-
9. The incoming baseband bit stream (Figure 1-9(a)) is lowpass ltered and used
to modulate the phase of a carrier (Figure 1-9(b)). The spectrum of this signal is
shown in Figure 1-9(c). The PSK modulation scheme is similar to that represented
in Figure 1-10, with the FSK modulator replaced by a PSK modulator which shifts
the phase of the carrier rather than its frequency. There are many variants of PSK,
with the most fundamental characteristics being the number of phase states (e.g.,
with 2
n
phase states, n bits of information can be transmitted) and how the phasor
of the RF signal transitions fromone phase state to another. Generally PSKschemes
shape the spectrum of the modulated signal to t as much energy as possible
within a spectral mask. This results in a modulated carrier whose amplitude varies
(a)
INPUT
(b)
OUTPUT
(c)
SPECTRUM
TIME FREQUENCY
TIME
Figure 1-9 Characteristics of phase shift keying (PSK) modulation: (a) modulating bit
stream; (b) the waveform of the carrier modulated using PSK with the phase determined
by the 1s and 0s of the modulating bit stream; and (c) the spectrum of the modulated signal.
1
0
1
0
0
0
FSK Modulator
Filter
FSK Modulated Output
Carrier
Data
1 0 1
T
T
T/2
Figure 1-10 The frequency shift keying (FSK) modulation system.
(and thus a time-varying envelope). Such schemes require highly linear ampliers
to preserve the amplitude variations of the modulated RF signal. Other schemes
orchestrate the phase transitions to achieve a constant envelope modulated RF
signal but have lower spectral efciency. Two approaches to achieving this are rst
to slow the transitions down, and, second, to eliminate transitions from a phase
state to one which is rotated by 180
and so avoid the RF phasor traversing the

origin. The result of both approaches is that relatively simple hardware can be used
as amplitude distortion is not a problem. So system design affects RF hardware
complexity, and the sophistication of available and affordable hardware impacts
system design. There are only a few variants that achieve optimum properties and
many of these will be considered later in this chapter.
The communication limit of 1 symbol per hertz of bandwidth, the symbol rate,
comes from the Nyquist signaling theorem. Nyquist determined that the number
of independent pulses that could be put through a telegraph channel per unit of
time is limited to twice the bandwidth of the channel. With a modulated RF carrier,
this translates to a pulse of information on the I (or cosine) component, and a pulse
of information on the Q (or sine) component, in a unit of time equal 1/bandwidth.
I
2
3
Q
1
START
FINISH
I
Q
A B
C
D
(a) (b)
Figure 1-11 Constellation diagrams with possible transitions: (a) a binary modulation
scheme; and (b) QPSK, a four-state phase modulation scheme. Each state is a symbol.
Combining the I and Q channels, the phasor can move from one value to another
in a unit of time equal to 1/bandwidth. The phasor transition identies a symbol,
and hence one symbol can be sent per hertz of bandwidth.
1.4.2 Binary Phase Shift Keying, BPSK
Phase shift keying demodulation requires more sophisticated signal processing
than does FSK. PSK uses prescribed phase shifts to dene symbols, each of which
can represent one, two or more bits. Binary phase shift keying (BPSK), illustrated
in Figure 1-8(e), has one bit per symbol and is relatively a spectrally inefcient
scheme, with a maximum spectral efciency of 1 bit/second/hertz (1 bs
1
Hz
1
).
Although spectrally inefcient, it is ideally suited to low-power applications and
single-chip implementations, perhaps with an off-chip reference resonator. The
typical signal ow is from an antenna, through an RF-tuned amplier, with
quadrature mixing to produce I and Q channels which are then lowpass ltered.
The ltered I and Q channels are then commonly integrated over the duration
of a bit. In the most sensitive scheme the I and Q channels are oversampled (by
an ADC)at a multiple of the bit rate and the signal correlated with the expected
zero-crossing.
BPSKis commonly used in pagers and is used in Bluetooth. The operation of BPSK
modulation can be described using the constellation diagram shown in Figure 1-
11(a). The BPSK constellation diagram indicates that there are two states. These
states can be interpreted as the values of i(t) and q(t) at the sampling points
corresponding to the bit rate. The curves in Figure 1-11(a) indicate three transitions.
The states are at the ends of the transitions. If a 1, in Figure 1-11(a), is assigned to
the positive I value and 0 to a negative I value, then the bit sequence represented
in Figure 1-11(a) is 1001.
90
VCO
i(t)
q(t)
a(t)
b(t)
s(t)
WAVEFORM
SHAPING
WAVEFORM
SHAPING
SERIAL TO
PARALLEL
SERIAL BIT
STREAM
I
K
Q
K
I
Q
90 90
VCO
( )
( )
( )
( )
( )
WAVEFORM
SHAPING
WAVEFORM
SHAPING
WAVEFORM
SHAPING
WAVEFORM
SHAPING
SERIAL TO
PARALLEL
SERIAL TO
PARALLEL
SERIAL BIT
STREAM
Figure 1-12 Quadrature modulation block diagram indicating the role of pulse shaping.
1.4.3 Quadrature Phase Shift Keying, QPSK
QPSK modulation is usually referred to as quadrature PSK although it is also
referred to as quarternary PSK, quadriphase PSK, and quadraphase PSK. In QPSK
wireless systems, good spectral efciency is obtained by sending more than one
bit of information per hertz of bandwidth. Information is encoded in four phase
states. Thus referring to QPSK as quadraphase shift keying is more precise but
this is not the common usage. The higher order modulation schemes that achieve
more than two states require that the characteristics of the channel be taken
into account. The dominant characteristic of the wireless channel are deep fades
resulting fromdestructive interference of multiple reections. Fades can be viewed
as deep amplitude modulation and so it is difcult to transfer information in the
amplitude of a carrier. Consequently phase modulation schemes falling in the
class of M-ary phase shift keying (MPSK) are most appropriate in the mobile
context. In mobile environments there are just a few modulation formats that have
been found acceptable. These all fall in the class of either FSK-like schemes or
quadra phase shift keying (QPSK) (also called quadrature phase shift keying as
the modulation can be viewed as the superposition of two modulated quadrature
carriers ). The characteristic of QPSK modulation is that there are four allowable
phase states per symbol period, so two bits of information are transmitted per
change in the characteristic of the modulated signal. There are many other four-
state PSK schemes and there are schemes that have more than four phase states.
Quadrature phase shift keying modulation can be implemented using the
quadrature modulator shown in Figure 1-12. The constellation diagram of QPSK is
the result of plotting i(t) and q(t) on a rectangular graph (or equivalently A(t)
and (t) on a polar plot) in the generalized modulation circuit of Figure 1-12.
More commonly these quantities are referred to as I and Q. In Figure 1-12, the
input bit stream is rst converted into two parallel bit streams. Thus a two-bit
sequence in the serial bit stream becomes one I
K
bit and one Q
K
bit. The (I
K
,
Q
K
) pair constitutes the Kth symbol. A modulation scheme with four allowable
states A, B, C, Dis shown in Figure 1-11(b). In the absence of wave-shaping
circuits, i(t) and q(t) have very sharp transitions and the paths shown in Figure 1-
11 are almost instantaneous. This leads to large spectral spreads in the modulated
waveform, s(t). So to limit the spectrum of the RF signal s(t), the shape of i(t)
and q(t) is controlled; the waveform is shaped, usually by lowpass ltering. So a
pulse-shaping circuit changes binary information into a more smoothly varying
signal. Each transition or path in Figure 1-11 represents the transfer of a symbol
or minimum piece of information. The best efciency that can be obtained in
point-to-point communication is one symbol per hertz of bandwidth. In the QPSK
modulation scheme shown, there are three possible transitions from each point in
the constellation in addition to the possibility of no transition. Thus each symbol
contains two bits. So the maximum efciency of this type of modulation scheme
is 2 bs
1
Hz
1
(2 bits/second/hertz of bandwidth). What is actually achieved
depends on the pulse shaping circuits and on the criteria used to establish the
bandwidth of s(t). Various modulation schemes have relative merits in terms of
spectral efciency, tolerance to fading (due to destructive interference), carrier
recovery, spectral spreading in nonlinear circuitry, and many other issues that are
the realm of communication system theorists.
The waveforms corresponding to the state transitions shown in Figure 1-11(b)
most immediately affect the bandwidth of s(t) and the ability to demodulate the
signal. The constellation diagram is analogous to a phase diagram of the carrier
signal.
2
Thus, signal trajectories through the origin indicate that the amplitude of
the carrier is very small for many RF cycles, and this is particularly troublesome as
it is difcult to track the carrier in the presence of noise. The ability to demodulate
signals is equivalent to being able to reconstruct the original constellation diagram
of the modulation signal. Also, transitions through the origin indicate that there is
signicant amplitude variation of the RF signal and so this has high PAR.
Example 1.2 QPSK Modulation and Constellation
The bit sequence 110101001100 is to be transmitted using QPSK modulation. Show the transitions
on a constellation diagram.
SOLUTION: The bit sequence 110101001100 must be converted to a two-bit wide parallel stream
of symbols resulting in the sequence of symbols 11 01 01 00 11 00. The symbol 11 transitions to the
symbol 01 and then the symbol 01 and so on. The states (or symbols) and the transitions from one
symbol to the next required to send the bitstream 110101001100 are shown in Figure 1-13.
QPSK modulation results in the phasor of the carrier transitioning through the origin so that the
average power is lower and the PAR is high. A more signicant problem is that the phasor will fall
below the noise oor, making carrier recovery almost impossible.
2
This is true here but not in all cases. The constellation diagram is a way of representing symbols rst
and is not simply a phasor diagram.
01
I
Q
I
Q
0
10 00
01
I
Q
0
10 00
01
11
00 10 10 00 00 10
01 11
11 11
11 01
SYMBOL 5 SYMBOL 1 SYMBOL 3
I
Q
0
10 00
01
11
I
Q
I
Q
Figure 1-13 Constellation diagram states and transitions for the bit sequence
110101001100 sent as the set of symbols 11 01 01 00 11 00 using QPSK. Note that Symbols
2 and 3 are identical, so there is no transition and this is shown as a self-loop, whereas there
will be no transition in going from Symbol 2 to Symbol 3. The SYMBOL numbers indicated
reference the symbol at the end of the transition (the end of the arrowhead).
1.4.4 Frequency Shift Keying, FSK
Frequency Shift Keying is one of the simplest forms of digital modulation, with
the frequency of the transmitted signal indicating a symbol, usually either one or
two bits. It was the rst form of digital modulation employed. FSK is illustrated in
Figure 1-8(d). The schematic of an FSK modulation systemis shown in Figure 1-10.
Here, a binary bit stream is lowpass ltered and used to drive an FSK modulator,
one implementation of which shifts the frequency of an oscillator according to the
voltage of the baseband signal. This function can be achieved using a phase locked
loop (PLL) with considerably less sophistication than PSK-based schemes which
require digital signal processing to demodulate a modulated signal. With FSK, an
FM demodulator can be used to receive the signal. A characteristic feature of FSK
is that the amplitude of the modulated signal is constant so efcient saturating
(and hence nonlinear) ampliers can be used without introducing much distortion
of concern. Not surprisingly FSK was the rst form of digital modulation used
in mobile digital radio. Prior to digital radio FSK was used in analog radio to
transmit bits. A particular form of FSK is minimum shift keying (MSK), which
I
1
Q
0
I
01 00
10
Q
11
(a) (b)
Figure 1-14 Constellation diagrams of FSK modulation: (a) two-state FSK; and (b) four-
state FSK.
uses a baseband lowpass lter so that the transitions from one state to another are
smooth in time and limiting the bandwidth of the modulated signal.
In the preceding sections the constellation diagram was introduced as a phasor
diagram of the (modulated) carrier. However, the equivalence is only approximate
and the similarity is most distinct with FSKmodulation. Strictly speaking, a phasor
diagram describes a phasor that is xed in frequency. Still, if the phasor is very
slowly phase modulated then this approximation is good. That is, the frequency of
the modulated carrier is considered to be xed and the phase changes over time.
FSK modulation cannot be represented on a phasor diagram, as the information
is in the frequency transition rather than a phase transition. However, the discrete
states must be represented and the constellation diagram is used to graphically
represent them and the transitions. The departure from the phasor diagram can no
longer be ignored. In reality, with FSKmodulation, the frequency of the modulated
carrier changes slowly if the baseband signal is lowpass ltered. For example,
consider an FSK modulated signal with a bandwidth of 200 kHz and a carrier
at 1 GHz. This is a 0.02% bandwidth, so the phasor changes very slowly. So
going from one FSK state to another takes a very long time, about 5000 cycles.
In trying to represent FSK modulation on a pseudo-phasor diagram, the frequency
is approximated as being xed and the maximum real frequency shift is arbitrarily
taken as being a 180
shift of the phasor.

In FSK, the states are on a circle on the constellation diagram (see Figure 1-14).
Note that the constellation diagram indicates that the amplitude of the phasor
is constant, as FSK is a form of FM. In four-state FSK modulation (see Figure 1-
14(b)), transitions between states take twice as long for states that are on opposite
sides of the constellation diagram compared to states that are only separated by
90
. Filtering of the baseband modulating signal is required to minimizes the

bandwidth of the modulated four-state FSK signal. This reduces spectral efciency
to less than the theoretical maximumof 2 bits per hertz. In summary there are slight
inconsistencies and arbitrariness in using a phasor diagram for FSK but FSK does
have a dened constellation diagram which is closely related to a phasor diagram.
QPSK
FSK
Frequency
P
o
w
e
r
D
e
n
s
i
t
y
(
d
B
)
f + f
c
b
2
3 f + f
c
b
4 f + f
c
b
f
c
f + f
c
b
0
-10
-20
-30
-40
-50
(dB)
-4
E /N
b o
2 0 4 6 8 10 12 14 16 18 20
l
o
g
(
B
E
R
)
-2
-6
-8
-10
-12
-14
-16
-18
-20
FSK
QPSK
(a) (b)
Figure 1-15 Comparison of FSK and QPSK: (a) power spectral density as a function of
frequency deviation from the carrier; and (b) BER versus signal-to-noise ratio (SNR)as
E
b
/No (or energy per bit divided by noise per bit).
1.4.5 Comparison of FSK and QPSK Modulation
The modulation format used impacts the choice of circuitry, battery life, and
the tolerance of the system to noise. Figure 1-15 contrasts two types of digital
modulation: FSKas used in the Global Systemfor Mobile Communications (GSM)
cellular system, and QPSK used in the Digital Advanced Mobile Phone System
(DAMPS) cellular system. In Figure 1-15(a), f
b
is the bit frequency and it is seen
that FSK and QPSK have different spectral shapes. Most of the energy is contained
within the bandwidth dened by the bit frequency. At multiples of the bit
frequency, the power density with FSKis much lower than with QPSK, resulting in
less interference ( adjacent channel interference [ACI]) with neighboring radios
in adjacent channels. This is an important metric with radios that is captured by
the adjacent channel power ratio (ACPR), the ratio of the power in the adjacent
channel to the power in the main channel. Another important metric is the bit error
rate (BER). Different modulation formats differ in their susceptibility to noise. The
level of noise is captured by the ratio of the power in a bit, E
b
, to the noise power,
N
o
, in the time interval of a bit. This ratio, E
b
/N
o
(often referred to as E B N O),
is also the signal-to-noise ratio (SNR). In particular, consider the plot of the BER
against the SNR shown in Figure 1-15(b). QPSK is less susceptible to noise than is
FSK.
1.4.6 /4 Quadrature Phase Shift Keying, /4-QPSK
A major objective in digital modulation is to ensure that the RF trajectory from one
phase state to another does not go through the origin. The transition is slow so
that if the trajectory goes through the origin, the amplitude of the carrier will be
below the noise oor for a considerable time and it will not be possible to recover
its frequency. One of the solutions developed to address this problem is the /4
quadrature phase shift keying (/4-QPSK) modulation scheme. In this scheme the
n
/4
/2
/2
SYMBOL
I
Q
3/4
01
3/4
00 /4
10
11
1 /2
/2
SYMBOL
n+
Q
Q
11
00
01 0
10
(a) (b)
Figure 1-16 Constellation diagram of /4-QPSK modulation: (a) constellation diagram at
one symbol; and (b) the constellation diagram at the next symbol.
constellation at each symbol is rotated /4 radians from the previous symbol, as
shown in Figure 1-16.
Example 1.3 /4-QPSK Modulation and Constellation
The bit sequence 110101001000 is to be transmitted using /4-QPSK modulation. Show the
transitions on a constellation diagram.
SOLUTION: The bit sequence 110101001000 must be converted to a two-bit-wide parallel
stream of symbols resulting in the sequence of symbols 11 01 01 00 10 00. The symbol 11 transitions
to the symbol 01 and then the symbol 01 and so on. The constellation diagram of /4-QPSK modu-
lation really consists of two QPSK constellation diagrams that are shifted by /4 radians, as shown
in Figure 1-16. At one symbol (or time) the constellation diagram is that shown in Figure 1-16(a)
and at the next symbol it is that shown in Figure 1-16(b). The next symbol uses the constellation
diagram of Figure 1-16(a) and the process repeats. The states (or symbols) and the transitions from
one symbol to the next required to send the bitstream 110101001000 are shown in Figure 1-17.
One of the unique characteristics of /4-QPSKmodulation is that there is always a change, even if a
symbol is repeated. This helps with recovering the carrier frequency, which is an important function
in a demodulator. Also, the carrier phasor does not go through the origin and so the PAR is lower
than if QPSK modulation were used, as this would result in transitions through the origin. If the
binary bitstream itself (with sharp transitions in time) is the modulation signal, then the transition
from one symbol to the next occurs instantaneously and hence the modulated signal has a broad
spectrum around the carrier frequency. The transition, however, is slower if the bitstream is ltered,
and so the bandwidth of the modulated signal will be less. Ideally the transmission of one symbol
per hertz would be obtained. However, in /4-QPSK modulation the change from one symbol to the
next has a variable distance (and so takes different times) so that the ideal spectral efciency of one
symbol per hertz (or two bits/Hz) is not obtained. In practice, with realistic lters and allowing for
the longer transitions, /4-QPSK modulation achieves 1.62 bits/Hz.
01
I
Q
0
10 00
01
11
I
Q
0
11
10 00
01
I
Q
0
11
10 00
01
I
Q
11
00
10 01
I
Q
11
00
10 01
I
Q
11
00
10
Figure 1-17 Constellation diagram states and transitions for the bit sequence
110101001000 sent as the set of symbols 11 01 01 00 10 00 using /4QPSK modulation.
1.4.7 Differential Quadrature Phase Shift Keying,
DQPSK
Multiple transmission paths, or multipaths, result in constructive and destructive
interference and can result in rapid additional phase rotations. As a result, relying
on the phase of a phasor at the symbol sample time to determine the symbol
transmitted is prone to error. When an error results at one symbol, this error
accumulates when subsequent symbols are extracted. The solution is to use
encoding, and one of the simplest encoding schemes is differential phase encoding.
In this scheme the information of the modulated signal is contained in changes
(differences) in phase rather than in the absolute phase.
The /4-DQPSK modulation scheme is a differentially encoded form of /4-
QPSK. The /4-DQPSK scheme incorporates the /4-QPSK modulator and an
encoding scheme, as shown in Figure 1-18(a). The scheme is dened with respect
to its constellation diagram, shown in Figure 1-18(b) and repeated in Figure 1-18(c)
for clarity. The D indicates differential coding, while the /4 denotes the rotation
of the constellation by /4 radians or 45
from one interval to the next. This can

be explained by considering Figure 1-18(a). A four-bit stream is divided into two
quadrature nibbles of two bits each. These nibbles independently control the I and
DQPSK
QPSK
SERIAL TO PARALLEL
CONVERTER
BITSTREAM SPLIT INTO
AND CHANNELS I
SERIAL
BITSTREAM
I BITSTREAM
BITSTREAM
Q
Q
PHASE
ENCODER
DIFFERENTIAL
I BASEBAND
Q
BASEBAND
MODULATOR
RF
RF CIRCUITRY
/4 MODULATOR
/4
(a)
10
Q
11
/4
/4
00
01
3/4
3/4
I
/4
Q
00
/4
11
10
01
3/4
3/4
I
(b) (c)
Figure 1-18 A /4-DQPSK modulator consisting of (a) a differential phase encoder and a
/4-QPSK modulator; (b) constellation diagram of /4-DQPSK; and (c) a second example
clarifying the information is in the phase change rather than the phase state.
Table 1-1 Phase changes in a /4-DQPSK modulation scheme.
O
k
E
k

1 1 3/4
0 1 3/4
0 0 /4
1 0 /4
Q encoding, respectively, so that the allowable transitions rotate according to the
last transition. The information or data is in the phase transitions rather than the
constellation points themselves. The relationship between the symbol value and
the transition is given in Table 1-1. For example, the transitions shown in Figure 1-
19 for six successive time intervals describes the input bit sequence 000110110101.
Its waveform and spectrum are shown in Figure 1-20. More detail of the spectrum
is shown in Figure 1-21. In practice with realistic lters and allowing for the longer
transitions, /4-DQPSKmodulation achieves 1.62 bits/Hz, the same as /4-QPSK,
but of course with greater resilience to changes in the transmission path.
Sometimes a distinction is made between the transmitted symbols and the encoded
symbols. The encoded symbols already have the data represented as transitions
6
Q
START
FINISH
1
2
3
4
5
I
Figure 1-19 Constellation diagram of /4-DQPSK modulation showing six symbol
intervals coding the bit sequence 000110110101.
(c) (b) (a)
M
TIME FREQUENCY FREQUENCY
M
Figure 1-20 Details of digital modulation obtained using differential phase shift keying
(/4-DQPSK): (a) modulating waveform; (b) spectrum of the modulated carrier, with M
denoting the main channel; and (c) details of the spectrumof the modulated carrier focusing
on the main channel.
10
0
-10
-20
-30
-40
-50
-60
-70
-80
Lower
Adjacent
Channel
Main
Channel
Upper
Adjacent
Channel
S
P
E
C
T
R
U
M
(
d
B
)
FREQUENCY
Figure 1-21 Detailed spectrum of a /4-DQPSK signal showing the main channel and
lower and upper adjacent channels.
90
VCO
i(t)
q(t)
a(t)
b(t)
s(t)
WAVEFORM
SHAPING
WAVEFORM
SHAPING
SERIAL TO
PARALLEL
SERIAL BIT
STREAM
I
K
Q
K
Q
HALF BIT
DELAY
90 90
WAVEFORM
SHAPING
WAVEFORM
SHAPING
WAVEFORM
SHAPING
WAVEFORM
SHAPING
SERIAL TO
PARALLEL
SERIAL TO
PARALLEL
SERIAL BIT
STREAM
HALF BIT
DELAY
Figure 1-22 Block diagram of an OQPSK modulator.
from one transmitted symbol to the next. Similar reference is made to received
symbols and decoded symbols. The received symbols are the output of the /4-
QPSK demodulator, while the decoded symbols are the actual data extracted by
comparing one received symbol with the previous received symbol. The decoded
symbol is extracted in the DSP unit. In a differential scheme, the data transmitted
are determined by comparing a symbol with the previously received symbol, so
the data are determined from the change in phase of the carrier rather than the
actual phase of the carrier. This process of inferring the data actually sent from the
received symbols is called decoding. When /4-DQPSK encoding was introduced
in the early 1990s the DSP available for a mobile handset had only just reached
sufcient complexity. Today, encoding is used with all digital radio systems and
is more sophisticated than just the differential scheme of DQPSK. There are new
ways to handle carrier phase ambiguity. The sophistication of modern coding
schemes is beyond the hardware-centric theme of this book.
1.4.8 Offset Quadrature Shift Keying, OQPSK
The offset quadrature phase shift keying (OQPSK) modulation scheme avoids IQ
transitions passing through the origin on the constellation diagram (see Figure 1-
23(a)). As in all QPSK schemes, there are two bits per symbol, but now one bit
is used to directly modulate the RF signal, whereas the other bit is delayed by
half a symbol period as shown in Figure 1-22. The maximum phase change for a
bit transition is 90
, and as the I
K
and Q
K
are delayed, a total phase change of
approximately 180
is possible during one symbol. The constellation diagram is

shown in Figure 1-23(a).
1.4.9 Gaussian Minimum Shift Keying, GMSK
Gaussian minimum shift keying (GMSK) is the modulation scheme used in the
GSM cellular wireless system and is a variant of MSK with waveform shaping
coming from a Gaussian low pass lter. It is very similar to FSK modulation and
can be implemented with the same hardware, generally using a PLL.
11
0
10 00
01
I
Q
01
Q
11
10 00
I
1000
Q
1100
1101 0101 0001
0000 0100
0011 0111
0110 0010 1110
1111 1011
1010
1001
I
(a) (b) (c)
Figure 1-23 Constellation diagram for various formats: (a) OQPSK; (b) GMSK; and
16QAM.
GMSK modulation is also used in the Digital European Cordless Telephone
(DECT) standard. The spectral efciency of GMSK as implemented in the GSM
system (it depends slightly on the Gaussian lter parameters) is 1.35 bits/s/Hz
(1.35 bs
1
Hz
1
). Unltered MSK has a constant RF envelope and so the
linear amplication requirement is reduced. Filtering is required to limit spectral
spreadingin GMSK this results in amplitude variations of about 30%. However,
this is still very good, so one of the fundamental advantages of this modulation
scheme is that nonlinear, power-efcient amplication can be used. GMSK is
essentially a digital implementation of FM with a binary change in the frequency
of modulation. The switch from one modulation frequency to the other is timed
to occur at zero phase. Put another way, the input bitstream is shaped to form
half sinusoids for each bit of the input stream. The phase of the modulating
signal is always continuous, but at the zero crossings the half sinusoid continues
as a positive or negative half sinusoid depending on the next bit in the input
stream. The constellation diagram for GMSK (Figure 1-23(b)) is similar to that for
OQPSK, but on decoding, the information is not in the phase but the frequency.
So GMSK is an FSK scheme and can be implemented using traditional frequency
modulation and demodulation methods. While QPSK schemes can transmit more
data in a given channel bandwidth, GMSK (and other FSK techniques) have the
advantage that implementation of the baseband and RF hardware is simpler. A
GMSK transmitter can use conventional frequency modulation. On receive, an FM
discriminator can be used, avoiding the more complex I and Q demodulation.
In GMSK modulation, a data stream is passed through a Gaussian lter and the
ltered response drives an FM modulator with the FM deviation set to one-half of
the data rate. For example, an 8000bits/s GMSK data stream is modulated onto an
RF carrier with a peak deviation of 4 kHz or 2 kHz. One type of MSK and GMSK
modulator is shown in Figure 1-24.
Most GSM phones input the baseband signal to a PLL to implement frequency
modulation. The output of the PLL is input to a power amplier. This amplier
can be quite efcient, as amplitude distortion is not a concern.
WAVEFORM
SHAPING
SERIAL BIT
STREAM
sin( (t))
cos( (t))
90 90
VCO
i(t)
q(t)
a(t)
b(t)
s(t)
Q
I
(t)
Figure 1-24 Block diagram of an MSK modulator.

1.4.10 3/8-8PSK, Rotating Eight-State Phase Shift
Keying
The 3/8-8PSK modulation scheme is similar to /4-DQPSK in the sense that
rotation of the constellation occurs from one time interval to the next. This time
however the rotation of the constellation from one symbol to the next is 3/8. This
modulation scheme is used in the EDGE system, and provides 3 bits per symbol
(ideally) compared to GMSKused in GSMwhich has 2 bits per symbol. GSM/Edge
provides data transmission of up to 128 kbps, and faster than the 48 kbps possible
with GSM.
Quadrature modulation schemes with four states, such as QPSK, have two I
states and two Q states that can be established by lowpass ltering the I and
Q bitstreams. For higher order modulation schemes such as 8PSK this approach
will not work. Instead, I(t) and Q(t) are established in the DSP unit and then
converted using a DAC to generate the analog signals applied to the hardware
modulator. Alternatively the modulated signal is created directly in the DSP and
a DAC converts this to an IF and a hardware mixer up-converts this to RF. This
approach is required in multimode phones supporting multiple standards.
1.4.11 Quadrature Amplitude Modulation, QAM
The digital modulation schemes described so far modulate the phase or frequency
of a carrier to convey binary data and the constellation points lie on a circle of
constant amplitude. The effect of this is to provide some immunity to amplitude
changes to the signal. However, much more information can be transmitted if the
amplitude is varied as well as the phase. With sophisticated signal processing it is
possible to reliably use quadrature amplitude modulation (QAM). In particular, it
is necessary to characterize the channel. In wired and line-of-sight (LOS) systems,
the channel changes slowly. However, in non-LOS wireless systems it is necessary
to incorporate a pilot code or pilot signal with the data so that the characteristics
of the channel can be continually updated.
Table 1-2 Spectral efciencies of various modulation formats.
Modulation bits/s/Hz
BPSK 1
FSK 1
QPSK 2
GMSK 1.35
/4-DQPSK 1.63
3/8-8PSK 2.7
64-QAM 4
256-QAM 6
A sixteen-state rectangular QAM constellation is shown in Figure 1-23(c). This
constellation can be produced by separately amplitude modulating an I carrier
and a Qcarrier. Both carriers have the same frequency but are 90
out of phase. The

two carriers are then combined, with the result that the xed carrier is suppressed.
The most common form of QAM is square QAM, or rectangular QAM with an
equal number of I and Q states. The most common forms are 16-QAM, 64-QAM,
128-QAM, and 256-QAM. The constellation points are closer together with high-
order QAMand so are more susceptible to noise and other interference. Thus high-
order QAM can deliver more data, but less reliably, than can lower order QAM.
The constellation in QAM can be constructed many ways and while rectangular
QAM is the most common form, nonrectangular schemes exist; for example,
having two PSK schemes at two different amplitude levels. While there are minor
advantages to such schemes, square QAM is generally preferred as it requires
simpler modulation and demodulation. The rapid fading in a mobile environment
has a bigger impact on amplitude than on phase. As a result, PSK schemes have
fewer errors than QAM schemes in mobile use.
1.4.12 Digital Modulation Summary
The spectral efciencies of various digital modulation schemes are summarized in
Table 1-2. For example, in 1 kHz of bandwidth the 3/8-8PSK scheme (supported
in third generation cellular) transmits 2700 bits. Digital transmission requires
greater bandwidth than does analog modulation for transmission of the same
amount of information that was originally analog (e.g., voice). However, digital
modulation is essential for data, and digital modulation is also advantageous
for voice. Direct digitization of an audio waveform for high-quality reproduction
requires 8 bits of resolution captured at a sample rate of 8000 samples/s for
a total 64 kbps (64000 bits/s). The appeal of digital modulation for audio is
directly related to the reduction of bit rate accomplished by speech coding
algorithms. Acceptable speech is achieved with bit rates of 3.8 kbps and higher.
(The measure of speech quality is purely subjective.) The speech coding algorithms
achieve bit rate reduction by utilizing the characteristics of human hearing.
There is a lot of redundancy in speech, but this is not used specically. Human
hearing responds to time-varying spectral content, and the unique characteristic
is that the statistics of the signal, the autocorrelation functions and higher order
moments, can be captured at low resolution. The spectrum of the signal can
be adequately reconstructed from these statistics.
3
In a typical speech coding
algorithm implemented in a codec, units of 160 samples are characterized by just
a few autocorrelation and related parameters. These parameters generally do not
require many bits to enable fully intelligible speech to be synthesized so that a
reduction factor of 8 or even 16 in the bit rate is achieved.
The spectral efciencies shown in Table 1-2 are sometimes less than the ideal.
QPSK ideally conveys 2 bits per symbol, but in a communication system, lowpass
ltering at baseband is required to constrain the spectrum of the RF modulated
signal. It is clear that an ideal lter cannot be realized, and this in part limits
the achievable spectral efciency. A more signicant limitation can be understood
by considering the constellation diagram of actual schemes such as those shown
in Figure 1-23. Now consider that the constellation diagrams are equivalent to
phasor diagrams (which they are in rst- and second- generation radio). With the
same baseband bandwidth it will take different times for the phasor to make the
transition from one symbol to the next; longer transitions require more bandwidth
than do shorter transitions. As a result, the spectrumefciency will be less than the
ideal. So in a QPSK-like scheme 2 bits per symbol are achievable, but in practice
a symbol requires more than the minimum of one Hz of bandwidth. Also, with
QAM some of the outlying constellation points at the corners of the constellation
cube (high I and Q) are not used.
1.5 Ampliers
Ampliers can be optimized for low noise, moderate to high gain, or substantial
power output. Using GaAs metal epitaxy semiconductor eld effect transistors
(MESFETs), psuedomorphic high electron mobility transistors (pHEMTs) or
heterojunction bipolar transistors (HBTs), high power amplication can be
obtained at frequencies extending well into the millimeter-wave regions (above
30 GHz). silicon (Si) complementary metal oxide semiconductor (CMOS), and
Si BiCMOS (combining Bipolar Junction Transistors (BJTs), and CMOS), and
the higher-performance SiGe BiCMOS, provide gain above 30 GHz but with
limited power capabilities. At lower frequencies Si laterally diffused metal
oxide semiconductor transistors (LDMOS) dominate in basestation high-power
transistor ampliers at cellular frequencies (2 GHz and lower). A special high
frequency version of Si CMOS called RF CMOS, supporting microwave elements
and having a lower loss substrate, is beginning to dominate for low power
RF applications. SiGe BiCMOS has had an 8 GHz performance advantage over
RF CMOS for many years. However, RF CMOS is now capale of supporting
applications at 10 GHz covering most commercial applications. A late-2000s
comparison of these technologies is given in Table 1-3.
3
Moments themselves are not transmitted, however. Residuals are transmitted from which the time-
varying moments and spectra can be reconstructed. Note that the rst moment of a signal is its mean,
the second-order moment is the signals standard deviation, etc.
Table 1-3 Comparison of basic process technologies. Patterning uses optical lithography
except that e-beam lithography is required for lithography below 250 m with GaAs. Costs
do not include one-off, costs (i.e. non recurrent engineering [NRE] costs) such as mask and
design costs.
Substrate GaAs MESFET Si BIPOLAR Si/SiGe BiCMOS Si RF CMOS
HBT, pHEMT
150 mm 300 mm 300 mm 300 mm
semi-insulating semiconducting semiconducting semiconducting
Processing Implant/anneal epitaxial growth epitaxial growth epitaxial growth
Metal layers 23 2-4 2-4 411
Mask layers 13 2733 4050 3545
Cost (optical) 10 cents/mm
2
3 cents/mm
2
3 cents/mm
2
3 cents/mm
2
Cost (e-beam) 100 cents/mm
2
OR BASE
M
1
M
2
TRANSISTOR
INPUT
MATCHING
NETWORK
OUTPUT
NETWORK
MATCHING
RF
INPUT
RF
OUTPUT
CONTROL
CHIP
GATE
COLLECTOR OR DRAIN
FILTER
LOWPASS
FILTER
LOWPASS
AMPLIFIER
BIAS
Figure 1-25 Block diagram of an RF amplier including biasing networks.
An RF amplier requires the circuit arrangement indicated in the general block
form of Figure 1-25. The DC biasing circuit is fairly standard; it does not involve
any microwave constraints and it will not be discussed here. The lowpass lters
(the bias circuits) can have one of several forms and are often integrated into the
input and output network design sometimes through the choice of appropriate
alternative topologies. Synthesis of the input, output, and (occasionally) feedback
networks are the primary design objectives of any amplier. The essence of RF
amplier design is synthesis of the network, shown in Figure 1-26.
RF transistors used to amplify small signals should have high maximum available
gain and low noise characteristics. For transistors used in transmitters, where
the efcient generation of power is important, it is important that the transistor
characteristics be close to linear in the central region of the current-voltage
characteristics so that distortion is minimized when the RF voltage variations
range over a large portion of the current-voltage characteristics. The ultimate
limit on output power is determined by the breakdown voltage at high drain-
source voltages and also by the maximum current density. Finally, for efcient
amplication of large signals the knee voltage (where the current-voltage curves
bend over and start to atten at low drain-source voltages) should be low.
LOAD
M
1
M
2
TRANSISTOR
INPUT
NETWORK
SOURCE
OUTPUT
NETWORK
Z
21
S
12
L
S
22
S
11
Z
S
S
S L
Figure 1-26 The scattering parameter (Snm) and reection coefcients () associated with
a microwave transistor amplier.
1.5.1 Linear Amplier
The linear amplier is generally known as a Class A amplier and is dened by its
ability to amplify small to mediumand even large signals with minimal distortion.
This is achieved by biasing a transistor in the middle of its I-V (or current-voltage)
characteristics. Figure 1-27 shows the I-V characteristics of the FET and bipolar
transistors shown in Figures 1-28(a) and 1-28(b), together with the DC loadline.
The loadline is the locus of the output current and voltage, established by the
amplier congurations shown in Figures 1-28(c) and 1-28(d). For the Class A
ampliers in Figures 1-28(c) and 1-28(d) the loadlines are dened by
I
C
= (V
CC
V
CE
) /R
L
and I
D
= (V
DD
V
DS
) /R
L
. (1.20)
These are called single-ended ampliers as the input and output voltages are
referred to ground. The opposite type of amplier is the differential amplier
congurations to be considered shortly. An amplier using a bipolar transistor
(either a BJT, or an HBT) is shown in Figure 1-28(c), with the transistor terminals
labeled in Figure 1-28(a). Referring to Figure 1-27(a), the output voltage of the
bipolar amplier is V
CE
and this swings from a maximum value of V
CE,max
to
a minimum of V
CE,min
. For a bipolar transistor V
CE,min
is approximately 0.2 V,
while V
CE,max
for a resistively-biased circuit is just the supply voltage V
CC
. The
quiescent or bias point is shown with collector-emitter voltage V
Q
and quiescent
current I
Q
. For a Class A amplier, the quiescent point is just the bias point and
this is in the middle of the output voltage swing and the slope of the loadline is
established by the load resistor R
L
.
The I-V characteristics of a FET amplier are shown in Figure 1-27(b). The notable
difference between these characteristics and those of the bipolar transistor is that
the curves are less abrupt at low output voltage (V
C
or V
D
). This results in the
minimum output voltage (V
DS,min
) being larger than V
CE,min
. For a typical RF
FET amplier, as shown in Figure 1-28(d), the supply voltage (V
DD
) is 3 V, while
V
DS,min
is 0.5 V. So for the same supply voltage, the output voltage swing with a
FET amplier will be smaller than for a BJT amplier.
The bipolar and FET ampliers of Figure 1-28 use resistive biasing so that the
maximum output voltage swing is limited. As well, the bias resistor is also the
load resistor. Various alternative topologies have been developed yielding a range
of output voltage swings. The common variations are shown in Figure 1-29 for
an FET amplier. Figure 1-29(a) is a resistively biased Class A amplier with
,max
,min
BIPOLAR CHARACTERISTIC
= 1 mA
= 2 mA
= 3 mA
= 4 mA
= 5 mA
= 6 mA
= 7 mA
Q
C
CE
I
Q
V
CE
B
DC LOAD LINE
I
V
CE
V
Q
I
V
(a)
,max
= 0 V
,min
= 0.75 V
= 1.0 V
= 1.25 V
= 1.5 V
= 1.75 V
= 0.5 V
= 0.25 V
FET CHARACTERISTIC
Q
V
DS
GS
V
DC LOAD LINE
V
DS
V
DS
V
Q
I
Q
I
D
(b)
Figure 1-27 Current-voltage characteristics of transistor ampliers shown with a Class A
loadline: (a) FET amplier and bipolar amplier.
(d)
OUTPUT
INPUT
OUTPUT
INPUT
(a)
C
BASE B
COLLECTOR
EMITTER
E
(b)
G
D
GATE
DRAIN
S
SOURCE
(c)
V
CC
V
C
R
L
I
C
V
DD
V
D
R
L
I
C
I
B
I
D
V
GS
I
D
Figure 1-28 Class A single-ended ampliers: (a) BJT transistor with B for base terminal, C
for collector terminal, and E for emitter terminal; (b) MOSFET transistor with G for gate
terminal, D for drain terminal, and S for source terminal; (c) single-ended BJT Class A
amplier with resistive bias; and (d) single-ended MOSFET Class A amplier with resistive
bias.
the output voltage swing between V
DS,min
and V
DD
. The quiescent drain-source
voltage is halfway between these extremes. The load R
L
also provides correct
biasing. This amplier is also called a single-ended amplier to differentiate it from
a differential amplier. A more efcient Class A amplier uses inductive biasing
as shown in Figure 1-29(b). Bias current is now provided via the drain inductor
and the load R
L
is not part of the bias circuit. With the inductively loaded Class
A amplier, the quiescent voltage is V
DD
and the output voltage swing is between
V
DS,min
and 2V
DD
, slightly more than twice the voltage swing of the resistively
loaded amplier. Another topology that provides enhanced voltage swing is the
differential resistively biased amplier shown in Figure 1-29(c). This amplier
topology is also called a fully differential amplier (FDA). This is the topology
commonly found in silicon radio frequency integrated circuits (RFICs), where
the current source common to the sources of the FETs results in good differential-
mode gain (when the inputs to the two FET gates is 180
out of phase) and the

common-mode gain is low. This is important in RFICs, as noise in the substrate
will affect the inputs and outputs of both transistors equally. As can be seen in
Figure 1-29(c)(ii), the differential voltage swing will be approximately 4V
DD
less
the voltage drop across the current source. The supply voltage of RFICs is limited
so at the nal output stage the current source is sometimes sacriced so that larger
voltage swings can be obtained. The resulting amplier is the pseudo-differential
amplier (PDA), shown in Figure 1-29(d). The maximum output voltage swing
is now 4V
DD
2V
DS,min
almost four times the voltage swing, (or 16 times the
power into the same load) of a single-ended resistively biased Class A amplier.
1.5.2 Classes of Ampliers
The class A amplier has limited efciency mainly because there is always
substantial quiescent current owing whether or not RF current is owing. Higher
order classes of ampliers achieve higher efciency, but distort the RF signal.
The current and voltage locus of Class A, B, AB, and C ampliers have a similar
V
DD
min DS,
V
DD
R
L
L
D
V
DD
V
D
R
L
min DS,
_ 1
2
V
DD
+
_
min
_ 1
2
V
DD
min DS,
+
M1
M2
min DS,
min DS,
_
min
min
min
_
_
min
_
_
min
M1
M2
(i)
(iii)
(iii) (i)
(i)
(a)
SingleEnded
(b)
(c)
SingleEnded
Fully
Differential
(d)
Pseudo
Differential
Amplifier
Amplifier
Amplifier
Amplifier
Output Differential
Output
(ii)
(ii)
(i) (ii)
(ii)
( )
( )
DD
2
V
D2
V
D1
DD
V 2
DD
V 2
V
V
V
O
DD
V 2 V
DS,
D,
V
D
V
D
V
O
V
2
DD
V 2
V
D2
V
D,
V
V
D
V
DD
V
D
2
V
D,
V
D1
V
V
V
D,
V
DD
2V
DD
O
V
D
V
V
D
DD
V 2
O
V
V
DS,
V
Figure 1-29 Class A MOSFET ampliers with output voltage waveforms: (a) single-
ended amplier with resistive biasing; (b) single-ended amplier with inductive biasing;
(c) fully differential amplier with inductive biasing; and (d) pseudo-differential amplier.
Schematic is shown in (i), drain voltage waveforms in (ii); and differential output in (iii).
LOAD LINE
HIGH EFFICIENCY
I
D
DS
V
LOAD LINE
CONVENTIONAL
A
B
QUIESCENT POINT
AB CLASS AB AMPLIFIER (LOAD = R + L)
C
B CLASS B AMPLIFIER (LOAD = R + L)
A CLASS A AMPLIFIER (LOAD = R)
C CLASS C AMPLIFIER (LOAD = R + RESONANT)
LOAD
D
S
G
AB
A
A
C
AB
A
A
AB
AB
C
C
V
GS
I
D
I
G
Figure 1-30 Current-voltage characteristics of the transistor used in an amplier showing
the quiescent point and output waveforms of various amplier classes.
trajectory on the output current-voltage characteristics of a transistor. The output
characteristics of a transistor are shown in Figure 1-30, showing what is called the
linear loadline and the bias points for the various amplier classes. The loadline
is the locus of the DC current and voltage as the DC input voltage is varied.
With the Class A amplier, the transistor is biased in the middle of the transistor
characteristics where the response has the highest linearity. That is, if the gate
voltage varies from an applied signal the output voltage and current variations
are nearly linearly proportional to the applied input. The drawback is that there
is always considerable DC current owing, even when the input signal is vary
small. That is there is DC power consumption whether or not RF power is being
generated at the output of the transistor. This is not of concern if small RF signals
Table 1-4 Comparison of efciency metrics for an amplier producing 1 W RF output
power and consuming 2 W of DC power with various power gains. The industry standard
power-added efciency used by RF and microwave engineers is
PAE
.
Power gain (dB)
TOTAL

PAE

3 40% 25% 50%
6 44% 37% 50%
10 48% 45% 50%
15 49% 48% 50%
20 50% 50% 50%
40 50% 50% 50%
are to be amplied, as then a small transistor can be chosen so that the DC current
levels are small. It is a problem if an amplier must handle both large and small
signals.
This leads to a discussion of efciency. The efciency of a circuit is the useful
output power divided by the input power. Such a measure of efciency is called
power-added efciency (PAE). Two different denitions of PAE are used [4]. The
more universal denition that can be used with any two-port network is also called
the total power added efciency or transmit chain efciency and is denoted here
as
TOTAL
dened as
TOTAL
=
P
RF,out
P
DC
+P
RF,in
. (1.21)
At RF and microwave frequencies, the most common denition of PAE used with
power ampliers focuses on the additional RF power divided by the DC input
power. This is designated as
PAE
and is dened as
PAE
=
P
RF,out
P
RF,in
P
DC
. (1.22)
For high-gain ampliers, P
RF,in
P
DC
, and both
TOTAL
and
PAE
reduce to the
efciency of the amplier:
=
P
RF,out
P
DC

PAE

TOTAL
(high gain) . (1.23)
These efciency metrics are compared in Table 1-4 for an amplier with 1 W RF
output power. The rst amplier has a power gain of 3 dB, which is commonly the
gain of the nal amplier stage producing the maximum output power available
from a particular transistor technology.
Returning now to a discussion of the efciency of the various classes of ampliers,
since the Class A amplier is always drawing DC current the efciency of Class
A ampliers is near zero when the input signal is very small. The maximum
efciency of Class A is 25% if resistive biasing is used and 50% when inductive
biasing is used. Efciency is improved by reducing the DC power and this is
achieved by moving the bias point further down the DC loadline, as in the Class
B, AB, and C ampliers shown in Figure 1-30. Reducing the bias results in signal
CLASS A
CLASS AB
INPUT
OUTPUT
CLASS C
CLASS B OUTPUT
INPUT
CURRENT
VOLTAGE
(a) (b)
Figure 1-31 Input and output waveforms for various classes of amplier: (a) Class A, B, C,
and AB ampliers; and (b) switching ampliers.
distortion for large RF signals. This can be seen in the various output waveforms
shown in Figure 1-31(a).
Class A ampliers have the highest linearity and Class B and C ampliers result
in considerable distortion. As a compromise, class AB ampliers are used in many
cellular applications, although Class C ampliers are used with constant envelope
modulation schemes, as in GSM. Nearly all small-signal ampliers are Class A.
This is also true for most broadband ampliers, as amplier stability is more
certain. The class A amplier presents impedances that are almost independent
of the level of the signal. However, a Class B, AB, or C amplier presents an
impedance whose value varies depending on the level of the RF signal. Thus
design requires more care, as the chances of instability are higher and it is
more likely that an oscillation condition will be met. Also, Class B, AB, and C
ampliers are generally not used in broadband applications or at high frequencies
mainly because of the problem of maintaining stability. Class A ampliers are
the preferred solution for ampliers at 10 GHz and above and for broadband
ampliers, again mainly because it is easy to ensure stability, and thus design is
much simpler and more tolerant to parasitic effects and variations.
The effect of parasitic capacitances and delay effects (such as those due to the time
it takes carriers to move across a base for a BJT or under the gate for an FET)
result in the current-voltage locus for RF signals differing from the DC situation.
This effect is captured by the dynamic or AC loadline which is shown in Figure
1-32(a). The Class A amplier is biased in the middle of the I-V characteristics
and the output from this amplier has the least distortion, as seen in Figure 1-31.
This seems very good, but the drawback is that the Class A amplier draws DC
current even when the input signal is negligible. This is a low efciency, but highly
linear class. When designers refer to a linear amplier they are referring to a
Class A amplier. The other amplier classes shown in Figure 1-30 have higher
efciencies but varying degrees of distortion. The outputs are shown in Figure 1-
DC LOAD LINE
I
AC LOAD LINE
D
V
DS
DC LOAD LINE
I
AC LOAD LINE
D
V
DS
(a) (b)
Figure 1-32 DC and RF loadlines: (a) loadlines of Class A, B and C ampliers; and (b)
loadlines of switching ampliers.
31. The output of the Class B amplier contains an amplied version of only half of
the input signal but draws just a small leakage current when no signal is applied.
With the Class C amplier there must be some positive RF input signal before
there is an output: there is more distortion but no current ows, not even leakage
current, when there is no RF input signal. The Class AB amplier is a compromise
between Class A and Class B ampliers. Less DC current ows than with Class
A when there is negligible input signal and the distortion is less than with Class
B. There are higher classes of amplier, Class D, E etc., and these rely on resonant
circuits to change the shape of the loadline to result in better trade-offs between
efciency and distortion than can be achieved with Class AB.
Class C ampliers are biased so that there is almost no drain-source (or collector-
emitter) current when no RF signal is applied, so the output waveform has
considerable distortion, as shown in Figure 1-31. This distortion is important
only if there is information in the amplitude of the signal. FM, GMSK, and PM
modulation schemes result in signals with constant RF envelopes, thus there is no
information contained in the amplitude of the signal. Therefore errors introduced
into the amplitude of a signal are of no signicance and efcient saturating mode
ampliers such as a Class C amplier can be used. In contrast, MSK, /4-DQPSK
and 3/8-8PSK modulation schemes do not result in signals with constant RF
envelopes and so information is contained in the amplitude of the RF signal. For
these modulation techniques reasonably linear ampliers are required.
Switching ampliers are a conceptual departure from Class A, AB, B, and C
ampliers, as can be seen in the typical AC loadline of a switching amplier
shown in Figure 1-32(b). This loadline is achieved by presenting the appropriate
harmonic impedances to the transistor amplier. The particular scheme of
harmonic termination (e.g., short or open circuits at the even and odd harmonics)
leads to the designation of a switching amplier as Class D, E, F, etc. The key
characteristic of a switching amplier is that when there is current through the
Table 1-5 Theoretical maximum efciencies of amplier classes.
Amplier Class Maximum Efciency
Class A (resistive bias) 25%
Class A (inductive bias) 50%
Class B 78.53%
Class C 100%
Class E 96%
Class F 88.36%
Table 1-6 Efciency reductions due to signal type. The class A amplier uses inductive
drain biasing.
Signal PAR Efciency Reduction Class A (L bias) Class E
(dB) Factor PAE PAE
FSK (MSK, GMSK) 0 1.0 50% 96%
QPSK 3.6 0.437 21.9% 42%
/4DQPSK 3.0 0.501 25.1% 48.1%
OQPSK 3.3 0.467 23.4% 44.8%
8PSK 3.3 0.467 23.4% 44.8%
64QAM 7.8 0.166 8.3% 15.9%
transistor, there is negligible voltage across the output. Also, when there is voltage
across the transistor, there is little current through it (see Figure 1-32(b)). The power
dissipated by the transistor is approximately the product of the current through it
and the voltage across the output. Thus the switching amplier consumes very
little DC power, transferring nearly all of the DC power to the RF signal. Bandpass
ltering of the output of the amplier results in a nal RF output with little
distortion. Switching ampliers are emerging as the preferred linear amplier in
both handsets and base stations of cellular systems.
The theoretical maximum power-added efciencies achieved by the various
amplier classes are given in Table 1-5. With modulated signals, the maximum
efciencies cannot be achieved, so that typically the average input power of the
amplier must be backed off by the PAR of the signal so that the peak carrier
portion of the signal has limited distortion. Generally the acceptable distortion
of the peak signal occurs at the 1 dB compression point of the amplier. This is
only an approximate guide, but useful. The PARs of several digitally modulated
signals are given in Table 1-6 together with their impact on efciency. If there are
two carriers, then the PAR of the combined signal will be higher, requiring greater
amplier back-off. In practice, the efciencies will differ from these theoretical
values because of loss in the amplier and the trade-off between efciency and
distortion. This is because the PAR does not fully capture the statistical nature of
signals, and because of coding and other technologies that can be used to reduce
the PAR.
Two microwave ampliers are shown in Figures 1-33 and 1-34. They are
called microwave monolithic integrated circuits (MMICs) and are fabricated
on compound semiconductor substrates. Both are broadband high-frequency
ampliers covering 8 to 12 GHz, which is known as X-band. Note that high-power
RF and microwave transistors are put in parallel yielding the required power.
1.5.3 Differential Amplier
Differential ampliers are the preferred amplier topology with silicon monolithi-
cally integrated circuits including RFICs. Figure 1-35(a) shows a fully differential
amplier (FDA) with resistive biasing in the drain legs. As well as providing bi-
asing current, the resistors are also the loads of the circuit. The supply voltage of
an RFIC can be quite low (a few volts or less), so choosing circuit topologies that
provide for large voltage swings is important, particularly for an output ampli-
er. Differential topologies lead to an almost doubling of the output voltage swing
compared to the output voltage swing of a single-ended amplier. An FDA in-
cludes a common current source (see Figures 1-35(a) and 1-35(b)). The circuit of
Figure 1-35(a) has a higher voltage swing, as previously described. Commonly the
schematic in Figure 1-35(c) is used. The current source at the common source point
of the FDA in Figure 1-35(a) limits the voltage swing, when larger output voltage
swings are required, the current source is eliminated and the resulting amplier
is called a pseudo-differential amplier (PDA), as shown in Figure 1-35(d). Again,
inductive biasing as shown in Figure 1-35(e) almost doubles the possible voltage
swing. The schematic representation of the PDA is shown in Figure 1-35(f).
1.5.4 Distortion
Distortion imposes a fundamental limit to the practical efciency that can be
realized by an amplier. Distortion originates when the output signal from an
amplier approaches the extremes of the loadline so that the output is not an exact
amplied replication of the input signal. For a one-tone signal, the amplitude gain
of the signal rollsoff as the input power increases, as shown in Figure 1-36(a). This
gure plots the amplitude of the output sinewave against the amplitude of the
input sinewave, putting both amplitudes in terms of power. The plot is called the
AM-to-AM (AM-AM) characteristic of the amplier. The ideal amplier would
follow the linear relationship between the output and input powers. The AM-AM
characteristic is linear at low input powers, but eventually the gain compresses
and the output power drops in proportion to the input power. At large powers, the
parasitic capacitances of the transistors in the amplier vary the signal phase, and
hence phase distortion results. Figure 1-36(b) shows what is called the AM-to-PM
(AM-PM) characteristic. The AM-AM distortion is generally more signicant, and
considerable departure from the linear response occurs before the output phase
varies appreciably. In Figure 1-36(a), the 1 dB gain compression point is at the
point where the difference between the extrapolated linear response is exceeds the
actual gain by 1 dB. P
1dB
is the output power at the 1 dB gain compression point
and is the single most important metric of distortion, and amplier designs use
P
1dB
as a point of reference. Phase distortion generally occurs at higher powers
(see Figure 1-36(b)). While Figure 1-36 presents the distortion characteristics for
INPUT
BIAS AND TEST POINTS
OUTPUT
G
G
G
G
G
G
G
AIRBRIDGE
RESISTOR
SPIRAL
INDUCTOR
TRANS-
ISTORS
1 2
STUB
(a)
2
AIRBRIDGE
MITRED BENDS
1
(b) (c)
Figure 1-33 A two-stage, two-transistor X-band (812 GHz) MMIC amplier producing
100 mW of power: (a) photomicrograph with key networks identied, G indicates ground;
(b) layout of the top spiral inductor; and (c) scanning electron microscope image of the
crosssection of the airbridge.
GROUND
GROUND
INPUT
BIAS
FIRST STAGE SECOND STAGE
BIAS
BIAS
GROUND
GROUND
OUTPUT
INPUT
MATCHING
INTERSTAGE
MATCHING
OUTPUT
MATCHING
Figure 1-34 An 812 GHz MMIC amplier producing approximately 1 Wof output power
with key networks identied. (Courtesy Filtronic, PLC, used with permission.)
a single sinewave, it has proved to be a good indicator of performance with
modulated signals. A two-tone signal consisting of two sinusoidal signals is a
better representation of modulated signals. A signal linearly combining (adding)
two sinusoidal signals of equal amplitude is shown in Figure 1-37(a). When the
input signal to a Class Aamplier is large the extremes of the signal on the loadline
(see Figure 1-27), are compressed when the signal reaches its extremities. This
results in the saturated output waveformshown in Figure 1-37(b). In the frequency
domain this distortion shows up as additional tones so that this distortion is said to
produce inter-modulation products (IMPs) as shown in Figure 1-38. Here f
1
and f
2
components have the frequencies of the two tones comprising the two-tone signal.
The extra ones in the output, f
3
and f
4
, are the intermodulation tones. In Figure
1-38, the tone at f
3
= 2f
1
f
2
is known as the lower IM3 (or lower intermod)
and f
4
= 2f
2
f
1
is known as the upper intermod. As well as the amplitude
distortion resulting in additional tones, there is phase distortion as captured by
the AM-PM characteristic. Ampliers, however, introduce less phase distortion
PSEUDO DIFFERENTIAL AMPLIFIER
V
_
V V
+
_
V
V
+
V
o
V
+
V
o
_
V
(a) (b) (c)
_
V
+
V
o
V
+
_
V
V
o
_
V
(f) (e) (d)
V
FULLY DIFFERENTIAL AMPLIFIER
o
V
+
V
o
FDA
PDA
Figure 1-35 Differential ampliers: (a) fully differential amplier (FDA); (b) FDA with
inductive biasing; (c) schematic representation; (d) pseudo-differential amplier (PDA); (e)
PDA with inductive biasing; (f) schematic representation.
NONLINEAR
in
(dB)
P
o
u
t
(
d
B
)
P
1dB
POINT
P
ACTUAL
EXTRAPOLATED
1 dB
LINEAR
GAIN 1 dB
COMPRESSION
P
in
(dB)
LINEAR
NONLINEAR
P
o
u
t
(
d
e
g
r
e
e
s
)
(a) (b)
Figure 1-36 Nonlinear effects introduced by RF hardware: (a) amplitude (AM-AM)
distortion; and (b) phase (AM-PM) distortion.
(a) (b)
Figure 1-37 A two-tone signal: (a) a two-tone input waveform; and (b) distorted output
showing compression (dashed waveform is undistorted).
SPECTRUM
A
M
P
L
I
T
U
D
E
f f
f
f
2
3
4
1
Figure 1-38 Spectrum at the output of a nonlinear amplier with a two-tone input signal.
than amplitude distortion, which is fortuitous since most communication systems
encode information in the phase or frequency rather than in the amplitude.
Amplier design consists of both design for good low-power linear operation
requiring maximum power transfer at the input and output of the amplier,
and a trade-off of acceptable distortion and efciency. In practice a certain level
of distortion must be tolerated, and what is acceptable is embedded in the
specications for the various wireless systems. Further discussion on the design
of ampliers is best undertaken after scattering parameters, or S-parameters,
are introduced. S-parameters describe power ow, and the design of RF and
microwave ampliers is intricately tied to power considerations, both noise power
and signal power. Linear amplier design is described in section 6.5.1 on page 316.
For low distortion, the peaks of the RF signal must be amplied linearly, however,
the DC power consumed depends on the amplier class. With Class A ampliers,
the DCpower must be sufcient to provide undistorted amplication of the largest
RF signal so that the DCpower is proportional to the peak ACpower. The situation
is similar for Class AB ampliers, with the difference being that the intent is to
live with some distortion of the peak signal so that the relationship between peak
power and DC power still exists, but the direct proportionality no longer holds.
For Class C and higher class ampliers, the DC power is mostly proportional
to the average RF power. So for Class A and Class AB ampliers, the average
operating point must be backedoff to allow for manageable distortion of the
peaks of a signal, with the level of back-off required being proportional to the
PAR. For Class C and higher classes, the back-off required comes from experience
and experimentation. The characteristics of the signal also determine how much
distortion can be tolerated.
The PAR of the signal is an indication of the type and amount of distortion that
can be tolerated. The PAR of the two-tone signal is 6 dB, and digitally modulated
signals can have PARs ranging from 0 dB to 10 dB or more. A signal with a
higher PAR results in lower efciency, as more back-off is required. Putting this
another way, the DC bias must be set so that there is minimum distortion when
the signal is at its peak, but the average RF power produced can be much less
than the peak RF power. (It is approximately an amount PAR below.) Thus for a
high-PAR signal, generally a higher DC power is required to produce the same RF
power. This is especially true for Class A ampliers. Efciency can be increased
by using switching ampliers. The PAR of a modulated signal is an indication
of how much information is being transferred in the amplitude of the signal. For
example, a GMSK signal has a PAR of 0 dB and there is no information in the
amplitude of a signal so that a highly efcient Class C amplier can be used as
any amplitude introduced does not matter. Signals that have higher PARs contain
increasing amounts of information in the amplitude of the signals.
AM-AM and AM-PM distortion, and two-tone distortion also provide indications
as to the distortion that occurs with digitally modulated signals. Distortion with
digitally modulated signals consists of in-band and out-of-band distortion. Out-
of-band distortion is represented in Figure 1-39, where the spectra at the input
and output of a nonlinear system are shown. The process that results in increased
levels in the adjacent sidebands is called spectral regrowth. This distortion is
approximately captured by the intermodulation distortion with a two-tone signal.
The generation of signal in the adjacent channel affects the function of other radios
and the level of signal is contained in system specications. Distortion generated
in-band affects the ability to interpret the constellation of the signal and hence
difculties in demodulating a signal. This type of distortion will be considered in
the next section.
1.5.5 Cochannel Interference
The minimum signal detectable in conventional wireless systems is determined by
the signal-to-interference ratio at the input. The noise is due to background noise
sources, including galactic noise and thermal noise. In cellular wireless systems,
the minimum signal detectable is also determined by the signal-to-interference
ratio (SIR), but now the dominant interference is due to other transmitters in the
cell and adjacent cells. The noise that is produced in the signal band from other
transmitters operating at the same frequency is called cochannel interference. The
level of cochannel interference is dependent on cell placement and frequency reuse
patterns. The degree to which cochannel interference can be controlled has a large
(a)
(b)
Figure 1-39 Input and output spectra of a digitally modulated signal: (a) a digitally
modulated signal at the input of a nonlinear amplier; and (b) at the output of the amplier.
LOWER
ADJACENT
CHANNEL OVERLAP OVERLAP
12.5 17.5
-30 0 30
ADJACENT
CHANNEL
UPPER
CHANNEL
A
M
P
L
I
T
U
D
E
-17.5 -12.5
Figure 1-40 Adjacent channels and overlap in the AMPS and DAMPS cellular systems.
effect on system capacity.
Control of cochannel interference is largely achieved by controlling the power
levels at the basestation and at the mobile units. Factors affecting interference are
The signal power falls off quickly with distance.
Transmitted power is reduced to the minimum acceptable signal-to-interference
ratio
Cochannel interference is not a nonlinear affect and is addressed using cell
placement.
1.5.6 Adjacent Channel Interference
Adjacent channel interference is the result of several factors. Since ideal ltering
cannot be achieved, there is inherent overlap of neighboring channels (Figure 1-
40). For this reason, adjacent channels are assigned to different cells. The nonlinear
behavior of transmitters also contributes to adjacent channel interference. Thus
characterization of nonlinear phenomena is important in RF design. Adjacent
channel interference occurs with both digitally modulated and analog modulated
RF signals. It turns out that conventional design approaches can be used to control
and predict adjacent channel interference for analog modulated signals but there
is as yet not a good design practice for digitally modulated signals.
The spectrum of a DAMPS signal is shown in Figure 1-41. The signal between
frequencies f
1
and f
2
is due to the digital modulation scheme and ltering. Most
of the signal outside this region is due to nonlinear effects which result in what is
called spectral regrowth, a process similar to third- and fth-order intermodulation
in two-tone systems. Using the frequency limits dened in Figure 1-41, the lower
channel ACPR is dened as
ACPR
ADJ,LOWER
=
Power in Lower Adjacent Channel
Power in Main Channel
=
_
f
4
f
3
X(f)df
_
f
2
f
1
X(f)df
, (1.24)
Figure 1-41 Denition of adjacent channel and main channel integration limits using a
typical DAMPS spectrum as an example.
where X(f) is the RF signal spectral power density.
1.5.7 Noise, Distortion, and Constellation Diagrams
Noise and nonlinear distortion affect the received constellation diagram and the
ability to demodulate signals. Noise is mostly introduced from the environment,
particularly from other radios, but the noise introduced by the RF hardware itself
is signicant when the signal received is close to the minimum detectable signal.
These distortion effects can be described in part by their effect on constellation
diagrams (see Figure 1-42). An additional impact is impairment introduced in
adjacent channels. It should be emphasized that the constellation diagram shows
the state of the system at the sampling instant that is determined by the recovered
clock. Errors in recovering the clock further distort the constellation diagram.
1.5.8 Error Vector Magnitude
The error vector magnitude (EVM) is a measure of the departure of a sampled
phasor from the ideal phasor located at the constellation point. Introducing an
error vector, X
error
, and a reference vector, X
reference
, which points to the ideal
constellation point, the EVM is dened as the ratio of the power of the error vector
IDEAL
I
Q
0
AMPLITUDE
DISTORTION
DISTORTED
IDEAL
I
Q
0
PHASE
DISTORTION
DISTORTED
WITH NOISE
I
Q
0
IDEAL
(a) (b) (c)
Figure 1-42 Impact of signal impairments on the constellation diagram of QPSK: (a)
amplitude distortion; (b) phase distortion; and (c) noise.
root mean square (RMS) power to the reference power. so that
EVM =
[X
error
[
2
[X
reference
[
2
. (1.25)
Expressing the error and reference vector in terms of the powers P
error
and
P
reference
respectively, enables EVM to expressed in terms of a power ratio:
EVM =
P
error
P
reference
. (1.26)
In decibels:
EVM(dB) = 10 log
10
P
error
P
reference
. (1.27)
or as a percentage,
EVM(%) =
P
error
P
reference
100% . (1.28)
If the modulation format results in constellation points having different powers,
the constellation point with the highest power is used as the reference.
1.6 Noise and Nonlinear Distortion
Noise and nonlinear distortion set the bounds on the signals that can be processed
in an RF circuit. Noise establishes the minimum detectable signal while nonlinear
distortion, by introducing distortion of the constellation diagram, sets the level of
the largest signal from which information can be reliably extracted. The range is
referred to as dynamic range and is one of the performance limits characterizing
analog circuits. In this section noise is considered rst and then expressions for
dynamic range developed.
s
v
L
R
v
n
G
G
o
N
e
S
i
N
i
S
i
N
o
S
o
S
o
i
S
i
N
G, F
N
(a) (b) (c)
Figure 1-43 Noise and two-ports: (a) amplier; (b) amplier with excess noise; and (c)
noisy two-port network.
1.6.1 Noise
Ampliers, lters and mixers in an RF frontend process (e.g., amplify, lter, and
mix) input noise the same way as an input signal. In addition, these components
can contribute what is called excess noise of their own. Without loss of generality,
the following discussion will consider noise with respect to the amplier shown in
Figure 1-43(a), where v
s
is the input signal. The noise signal with source designated
by v
n
is uncorrelated and random and must be described as an RMS voltage or by
its noise power. The most important noise-related metric is the signal-to-noise ratio
(SNR). With the noise power input to the amplier being N
i
and the signal power
input to the amplier S
i
, the input SNR is SNR
i
= S
i
/N
i
. If the amplier is noise
free then the input noise and signal powers are amplied by the power gain of
the amplier, G. Thus the output noise power is N
o
= GN
i
and the output signal
power is S
o
= GS
i
and the output SNR is SNR
o
= S
o
/N
o
= SNR
i
.
In practice, an amplier is noisy, with the addition of excess noise, N
e
, indicated in
Figure 1-43(b). The excess noise originates in different components in the amplier
and is either referenced to the input or the output of the amplier. Most commonly
it is referenced to the output so that the total output noise power is N
o
= GN
i
+N
e
.
In the absence of a qualier, the excess noise is referred to the output. N
e
is not
measured directly. Instead the ratio of the SNR at the input to that at the output is
called the noise factor, F:
F =
SNR
i
SNR
o
(1.29)
and this is the way it is normally measured. If the system circuit is nose free then
SNR
o
= SNR
i
and F = 1. If the circuit is not noise free, then SNR
o
< SNR
i
and
F > 1. F can be related to the excess noise produced in the circuit and from this
relationship the noise of a cascaded systemcan be calculated. With the excess noise,
N
e
, referred to the circuit output,
F =
SNR
i
SNR
o
=
SNR
i
1
1
SNR
o
=
S
i
N
i
N
o
S
o
=
S
i
N
i
GN
i
+N
e
GS
i
= 1 +
N
e
GN
i
. (1.30)
One of the conclusions that can be drawn from this is that the noise factor, F,
depends on the available noise power at the input of the circuit. As a standard
1
,2
2
G , F
1 1
G , F
2 2
N
i
S
i
N
o,1
o,1
S
o,2
S
N
o
Figure 1-44 Cascaded noisy two-ports.
reference the available noise power, N
R
, from a resistor at standard temperature
T
0
(290 K), and over a bandwidth B is used:
N
i
= N
R
= kT
0
B , (1.31)
where k is Boltzmanns constant. If the input of an amplier is connected to this
resistor and all of the noise power is delivered to the amplier, then
F = 1 +
N
e
GN
i
= 1 +
N
e
GkT
0
B
. (1.32)
Several random physical processes inside a circuit contribute to excess noise,
and not all of these process vary linearly with temperature. Consequently F is
a function of temperature although usually a weak one. It is also a function of
bandwidth and there is a problem in using F with cascaded systems in which
bandwidths vary for different subsystems. Even with all these problems, F is the
most important measure used to characterize noise. It can be used to determine the
noise performance of a cascade, when the noise factors and gains of the subsystem
constituents are known. F is the ratio of powers, and when expressed in decibels,
noise gure (NF) is used:
NF = 10 log
10
F = SNR
i
(dB) SNR
o
(dB) . (1.33)
Development of the noise factor of a cascade begins by considering the noise
contributions of the rst system, and then the next cascaded system, and so on.
The majority of RF and microwave systems are organized as cascades of two-port
networks with an input port and an output port, as shown in Figure 1-44.
If the excess noise contribution of an amplier is ignored, the output noise power
will be
N
o
= GkT
0
B . (1.34)
With excess noise, N
e
, from the amplier included, the output noise power is
N
o
= GkT
0
B +N
e
= GkT
0
B(1 +N
e
/(GkT
0
B)) = FGkT
0
B . (1.35)
Rearranging this equation the excess noise power can be written as
N
e
= (F 1)GkT
0
B . (1.36)
This result can be generalized for a system. Considering the second stage of the
cascade, the excess noise at the output of the second stage, due solely to the noise
generated internally in the second stage, is
N
e,2
= (F
2
1)kT
0
BG
2
. (1.37)
Then the total noise power at the output of a two-stage cascade is
N
o,2
= (F
2
1)kT
0
BG
2
+N
o,1
G
2
(1.38)
= (F
2
1)kT
0
BG
2
+F
1
kT
0
BG
1
G
2
. (1.39)
The second term above is the noise output from the rst stage amplied by the
second gain.
Generalizing the above result yields the total noise power at the output of the mth
stage:
N
o,m
=
m
n=2
_
(F
n
1) kT
0
B
n
i=2
G
i
_
+F
1
kT
0
B
m
n=1
G
n
. (1.40)
Thus an m-stage cascade has a total cascaded system noise factor F
T
=
N
o,m
__
G
T
N
i,1
_
, with G
T
being the total cascaded available gain and N
i,1
is the
noise power input to the rst stage. In terms of the parameters of individual stages
the total system noise factor is
F
T
= F
1
+
F
2
1
G
1
+
F
3
1
G
1
G
2
+
F
4
1
G
1
G
2
G
3
+ ; (1.41)
that is,
F
T
= F
1
+
m
n=2
F
n
1
m
i=n
G
i1
. (1.42)
This equation is known as Friis formula.
Example 1.4 Noise Figure of an Attenuator
What is the noise gure of a 20 dB attenuator in a 50 system?
SOLUTION: Denoting the attenuator as being in a 50 system indicates that an appropriate circuit
model to use in the analysis consists of the attenuator driven by a generator, with a 50 source
impedance, and the attenuator drives a 50 load. Also, the input impedances of the terminated
attenuator is 50 , as is the impedance looking into the output of the attenuator when it is connected
to the source. The key point is that the noise coming from the source is the noise thermally generated
in the 50 source impedance, and this noise is equal to the noise that is delivered to the load, as the
impedance presented to the load is also 50 . So the input noise, N
i
, is equal to the output noise:
N
o
= N
i
. (1.43)
o
IL = 3 dB
FILTER
S
o
N
o
S
o
G , F
2 2
N
i
S
i
G , F
1 1
N
Figure 1-45 Amplier and lter combination for which the total system noise gure is to
be calculated.
The input signal is attenuated by 20 dB (= 100). So
S
o
= S
i
/100 (1.44)
and thus the noise factor is
F =
SNR
i
SNR
o
=
S
i
N
i
N
o
S
o
=
S
i
N
i
N
i
S
i
/100
= 100 (1.45)
and the noise gure is
NF = 20 dB . (1.46)
So the noise gure of an attenuator (or lter) is just the loss of the component. This is not true for
ampliers of course, as there are other sources of noise, and the output impedance of a transistor is
not a thermal resistance.
Example 1.5 Noise Figure of Cascaded Stages
The cascaded two-port network in Figure 1-45 consists of a noisy amplier with a noise gure of 2 dB
and a gain of 20 dB followed by a lter with an insertion loss of 3 dB. Determine the total gain and
noise gure of the cascaded system.
SOLUTION: Friis formula can be used to calculate the total noise factor of the cascaded system,
from which the total noise gure can be determined. Since the insertion loss of the lter is 3 dB, the
gain of the lter is 3 dB and its noise gure is 3 dB, thus G
2
= 3 dB = 0.5 and NF
2
= 3 dB, so that
F
2
= 10
NF2/10
= 1.995. Also, for the rst stage, G
1
= 20 dB = 100 and F
1
= 10
NF1/10
= 1.585. The
total gain of the cascaded system is
G
T
(dB) = G
1
(dB) +G
2
(dB) = (20 dB) + (3 dB) = 17 dB . (1.47)
The total system noise factor is, from Equation (1.42),
F
T
= F
1
+ (F
2
1)/G
1
= 1.585 + (1.995 1)/100 = 1.595 . (1.48)
Thus the total noise gure of the system is NF
T
= 10 log
10
(1.595) = 2.03 dB. Note the importance
of the rst stage in determining the overall noise gure. If the gain of the rst stage is sufciently
large, subsequent stages have much less of an effect.
RESPONSE
FLOOR
N
o
1
1
S
N
R
M
I
N
P
i3
P
i3
P
o
3
P
o
1
3
P
o3
P
i
P
i
P
o
IP3
NOISE
OIP3
IP3
INTERCEPT
DR
SFDR
S
F
D
R
(dB)
IIP3
(
d
B
)
LINEAR
RESPONSE
Figure 1-46 Output power versus input power of a stage or system plotted as output
power in decibels versus input power in decibels. The IP3 response is a result of two-
tone intermodulation, and the input power is the combined power of the two signals that
have equal amplitude. Extrapolations of the 1:1 linear response and the 3:1 third-order
intermodulation response intersect at the IP3 point.
1.6.2 Dynamic Range
While modern communication and radar systems use digitally modulated signals,
two-tone signals are used to both approximately characterize nonlinearity, and
in manual calculations. At low powers before compression becomes a factor, the
fundamental response has a 1:1 slope with respect to the input, as shown in Figure
1-46. The IP3 response varies as the cube of the level of input tones when both
tones vary by the same amount, as is common in a two-tone test. Thus IP3 has
a 3:1 logarithmic slope with respect to the input. Since the relations are linear in a
log-log sense it is possible to describe the nonlinear performance of an amplier by
a single quantity called the dynamic range (DR) or by the spurious free dynamic
range (SFDR). SFDR and DR also capture noise properties.
In the following, an expression for SFDR is developed in terms of input-referenced
quantitiesthe input referred SFDR, (SFDR
i
). A similarly referenced dynamic
range DR (DR
i
) is also developed. SFDR describes the difference between a signal
and the noise oor, whereas DR incorporates the Minimum Detectable Signal
(MDS) which is the noise level plus a minimum acceptable SNR (SNR
MIN
),
expressed here in decibels.
Figure 1-46 illustrates the input-output relationship of a one-tone signal to the IP3
response of a subsystem and also graphically denes the dynamic ranges. The
point of intersection of the extrapolated linear output (of power P
o
) and third-
order (IP3) output (of power P
IP3
) is called the third-order intercept point (IP3
intercept). The point is identied by the output-referred intercept power (OIP3) or
by the input referred IP3 intercept power (IIP3), and these are key parameters in
describing the linearity of nonlinear subsystems.
In the linear gain region, P
o
versus P
i
has a slope of 1:1 so that
P
dBm,i
= P
dBm,o
G
dB
, (1.49)
where G
dB
is the power gain in decibels. P
o
is used here as the output power, with
P
dBm,o
indicating the output power in dBm. P
i
and P
dBm,i
are similarly dened.
In terms of input quantities
IIP3
dBm
= OIP3
dBm
G
dB
, (1.50)
where again the dBm subscript indicates that the quantity is expressed in decibels
referred to 1 milliwatt. The nonlinearity of RF active components results in
harmonics and intermodulation components. With the narrowband ampliers
of communication and radar systems, output lters conveniently lter out
harmonics. However, intermodulation distortion cannot be ltered out, as these
components are within the main passband. The intermodulation components
are therefore spurious tones. Generally just one of these denes the maximum
spurious tone and nearly always it is the third-order intermodulation tone with a
two-tone input. Consideration of the maximum spurious tone and the noise oor
denes the SFDR.
Examining Figure 1-46 leads to the following inequality describing the linear gain
of the amplier:
OIP3
dBm
P
dBm,o
IIP3
dBm
P
dBm,i
=
OIP3
dBm
P
dBm,o
(OIP3
dBm
G
dB
) P
dBm,i
= 1 . (1.51)
Here, P
o
, P
i
, OIP3, and IIP3 are the output power, the input power, the input-
referred IP3 intercept, and the output-referred IP3 intercept. The third-order
response is characterized by rst introducing an equivalent input power, P
dBm,i3
(P
i3
expressed in dBm), dened as the average power of the two-tone signal
that generates an IP3 of power P
dBm,o3
. Noting that P
dBm,o3
varies with a 3:1
logarithmic slope with respect to P
dBm,i3
, then
OIP3
dBm
P
dBm,o3
(OIP3
dBm
G
dB
) P
dBm,i3
= 3 (1.52)
or
P
dBm,i3
=
1
3
(2 OIP3
dBm
+P
dBm,o3
3G
dB
) . (1.53)
The SFDR can now be dened when the third-order intermodulation product
of two-tone excitation is the dominant spurious tone. The SFDR is dened as
the difference between P
i3
and P
i
when they produce IP3 and linear output
respectively that are both equal to the output noise power N
o
(see Figure 1-46);
that is, when P
o
= P
o3
= N
o
. Replacing P
dBm,o
in Equation 1.53 with N
o
gives
P
dBm,i3
=
1
3
(2 OIP3
dBm
+N
dBm,o
3G
dB
) (1.54)
and
P
dBm,i
= N
dBm,o
G
dB
. (1.55)
Note that the difference between the linear output and the third-order
intermodulation reduces as the input power increases above P
i3
. Thus the output-
referred SFDR is
SFDR
dB,o
= P
dBm,i3
P
dBm,i
(1.56)
=
2
3
OIP3
dBm
+
1
3
N
dBm,o
G
dB
N
dBm,o
+G
dB
=
2
3
(OIP3
dBm
N
dBm,o
) .
A similar development can be used to dene the input-referred SFDR:
SFDR
dB,i
=
2
3
(IIP3
dBm
N
dBm,i
) . (1.57)
Note that N
i
is the input-referred noise and includes noise applied to the
subsystem as well as the noise produced internally in the subsystem and
referred to the input. The SFDR provides a combined measure of distortion
and noise. However, for usable dynamic range the minimum acceptable SNR
must be considered. The minimum SNR (SNR
MIN
) required is determined by the
communication or radar modulation format, error coding, and acceptable BER. So
in dening DR the input power of the desired signal must increase sufciently
to produce an SNR of at least SNR
MIN
. Since the desired spurious level is still at
the noise oor, this implies a direct subtraction in decibels of the desired SNR.
Therefore the input-referred third-order dynamic range, preferred for use with
receivers, is
DR
i
=
2
3
(IIP3 N
dBm,i
) SNR
MIN
(1.58)
and the output-referred dynamic range, preferred for use with transmitters, is
DR
o
=
2
3
(OIP3 N
dBm,o
) SNR
MIN
. (1.59)
1.6.3 Probability Density Function and Distortion
Amplitude modulation is not inherently required for information transmission
with PSK modulation schemes. For instance, a QPSK signal consists of two digital
data streams, equal in amplitude, modulated in quadrature onto a carrier signal.
The resulting signal would have a constant envelope; however, the occupied
bandwidth is quite large, as the spectrum of a pulse train is sin(x)/x, the sinc
function. The rst sidelobe of the sinc spectrum is only 13 dB down from the
carrier level and is in the middle of the adjacent channel. To reduce the spectrum,
a low-pass lter is typically applied to each digital data stream to minimize the
out-of-band spectrum of the modulated signal. This comes with a drawback: the
lters cause a nite memory effect resulting in amplitude variations as the ringing
energy from a previous data pulse adds to the current ltered data pulse.
Amplitude variations of the modulated signal are characterized by measured
waveform statistics such as the PAR. A signal with a high PAR requires that
the RF system have high linearity to handle the average power requirements
-1
0
1
2
3
4
5
6
7
8
9
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
Envelope Magnitude (V)
CDMA OQPSK
CDMA QPSK
Complex Gaussian
Real Gaussian
Figure 1-47 Amplitude PDF for CDMA and Gaussian modulation signals. After [6].
and the peak amplitude excursions without generating excessive out-of-band
distortion. However, it is possible for a signal with a higher PAR to exhibit less
nonlinear distortion than a signal with lower PAR [5]. The reason for this apparent
inconsistency is because the signal peak is a singular point measurement with,
typically, a low probability of occurrence. Thus PAR is an incomplete statistic for
determining the linearity requirements for a transmitter to carry a signal.
The amplitude probability density function (APDF) is a more complete statistical
description of the amplitude variations of a modulated signal. The APDF denes
the maximum and minimum variation along with the relative probability of
occurrence of amplitudes within the variation. The APDF is typically estimated
from a histogram of amplitudes, with a uniform bin size, by
f(A) =
N
AN
c
, (1.60)
where N is the number of counts per bin, A is the bin amplitude width, and
N
c
is the total number of samples. The shape of the amplitude density between
the mean and peak amplitude inuences the sensitivity of a particular signal to
spectral regrowth due to nonlinear gain compression or expansion. For example,
Figure 1-47 shows the APDF for a CDMA mobile transmitter using OQPSK
modulation, the same signal using QPSK modulation, a real Gaussian signal, and
a complex Gaussian QPSKsignal (with I and Qeach having Gaussian distribution)
where the average power of each signal is set to 0 dBm. Gaussian signals are of
particular interest as their simple statistics lends them, and their interaction with
nonlinearities, to quasi-analytic treatment. The PAR for each signal is shown in
Table 1-7. The shape of the amplitude density after the mean differs for both signals
where it can be seen that it is difcult to determine, a priori, which signal will be
Table 1-7 Peak-to-average ratios in decibels for OQPSK and QPSK used in CDMA
compared to the PARs of Gaussian signals.
Signal Modulation PAR (dB)
OQPSK CDMA 5.4
QPSK CDMA 6.6
Real Gaussian 13.5
Complex Gaussian 11.8
more sensitive to nonlinear gain compression. For example, even though QPSK
has a higher PAR than OQPSK, the probability that the OQPSK signal is near the
peak is higher than for the QPSK signal. It is not surprising then that the measured
spectral regrowth of a OQPSK signal is higher than for a QPSK signal. So PAR is
only a rough guide to the distortion that is produced.
1.7 Switches
Microwave switches are commonly used to alternately connect an antenna to a
transmitter or a receiver. In some communication systems, such as GSM, a phone
does not transmit and receive simultaneously. Consequently a switch can be used
to separate the transmitted and received signals. In multi-band phones, a switch
is used to connect the right transmitter and receiver, which are band specic, to
the antenna. In radar systems, switches are used to steer an antenna beam by
changing the phase of the microwave signal delivered to each antenna in an array
of antennas. An ideal microwave switch is shown in Figure 1-48(a) where an input
port, designated RF
IN
, and an output port, identied as RF
OUT
, are shown. For
maximum power transfer between the ports, the switch should have little loss
and so small on resistance. At microwave frequencies, switches must be modeled
with parasitics and have nite on and off resistance. A realistic model applicable
to many switch types is shown in Figure 1-48(b). The capacitive parasitics, the
C
P
s, limit the frequency of operation of the switches, and the on resistance, R
ON
,
impacts the switch loss. Ideally the off resistance, R
OFF
, is very large however the
parasitic shunt capacitance, C
OFF
, is nearly always more signicant. The result is
that at high frequencies, there is an alternative capacitive connection between the
input and output through C
OFF
. The on resistance of the switch introduces voltage
division which can be seen by comparing the ideal connection shown in Figure 1-
48(c), and the more realistic connection shown in Figure 1-48(d). From the voltage
division ratio the loss of the lter can be calculated.
Switches are congured to provide connection from one or more inputs to one or
more outputs. The conguration of a switch is indicated by poles and throws and
several congurations are shown in Figure 1-49. Most commonly in microwave
applications single pole switches are used and this input is connected to an
antenna. The throws would be connected to different bands of a multi-band
phones for example.
There a four main types of microwave switches: mechanical, PIN diode, FET,
IN OUT
OUT
LOAD SOURCE LOAD
RF
IN
RF
IN
BIAS
GATE
CONTROL
RF
RF RF RF
IN
RF
OUT
RF
IN
RF
OUT
(e) (g) (f)
BIAS
FET MEMS BIAS
PIN DIODE
ACTUATOR
CONTROL
RF
(b) (a)
(c) (d)
SWITCH MODEL
IDEAL SWITCH
OUT
SOURCE
E
R
ON
R
OFF
C
OFF
C
P
C
P
R
S
R
L
R
ON
R
S
R
L
E
Figure 1-48 Microwave switches: (a) ideal switch connecting RF
IN
and RF
OUT
ports; (b)
model of a microwave switch; (c) ideal circuit model with switch on and with source and
load; (d) realistic low-frequency circuit model with switch on and with source and load; (e)
switch realized using a PIN diode; (f) switch realized using a FET; and (g) switch realized
using a MEMS switch; .
and MEMS switches. Mechanical switches are nearly ideal but tend to be large,
relatively expensive, and mostly used in laboratory settings. The other switches are
of most interest for use in systems. The PIN diode, FET, and MEMS switches are
shown in Figures 1-48(e), 1-48(f), and 1-48(g), respectively. With these technologies,
most higher order switches are based on interconnections of SPST switches. The
attributes of these switches is summarized in Table 1-8 for switches that are suitable
for cell phone applications. PIN diode switches are the most robust, handling the
most RF power, and operating at higher frequencies, than either FET- or MEMS-
based switches. However this comes at a price. The PIN diode used is similar to
a PN junction diode with the addition of an intrinsic layer between the p- and n-
type materials. With applied forward bias the diode has low series resistance. In
reverse bias the diode resistance is large. Forward bias requires DC current and
voltage, so control power is consumed when a PIN diode switch is on. The circuit
SPDT, Single Pole, Double Throw
SP4T, Single Pole, Four Throw
SPST, Single Pole, Single Throw
DPST, Double Pole, Single Throw
DPDT, Double Pole, Double Throw
Figure 1-49 Switch congurations.
Table 1-8 Typical properties of small microwave switches. (Sources:
1
Radant MEMS,
2
RF
Micro Devices, and
3
Tyco Electronics.)
Switch Conguration Power Loss Operating Actuation Response
Type Handling at 2 GHz Frequency Voltage Time
MEMS
1
SPST 1 W 0.15 dB to 12 GHz 40120 V 5 s
MEMS
1
SPST 0.5 W 0.27 dB to 40 GHz 40120 V 5 s
pHEMT
2
SPDT 2 W 0.5 dB to 2.5 GHz 6 V 0.5 s
pHEMT
2
SPDT 10 W 0.5 dB to 6 GHz 6 V 0.5 s
PIN
3
SPDT 13 W 0.35 dB to 2 GHz 12 V 0.5 s
PIN
3
SPDT 10 W 0.4 dB to 6 GHz 12 V 0.5 s
conguration for an SPST PIN diode switch is shown in Figure 1-48(e). Series bias
decoupling capacitors are required at the RF ports.
A eld effect transistors (FET) makes a good electronic switches; with the correct
bias applied to the gate, the drain-source connection looks like a small resistance.
Changing the bias to the other extreme removes free carriers from the channel
between the drain and source, and a large resistance is the result. Both Si and GaAs
switches are used at cellular frequencies with GaAs switches operating at extended
frequencies approaching 6 GHz. The operation of an FET can be described as
a variable drain-source resistance with the gate-source voltage controlling the
crosssection of the channel. The circuit for a a FET-based SPST switch is shown
in Figure 1-48(f). Series bias blocking capacitors are required at the RF ports.
Control power is only required to change the state of the switch; negligible power
is required to maintain the switch state.
A microelectromechanical system (MEMS) switch is fabricated using photolitho-
graphic techniques similar to those used in semiconductor manufacturing. They
are essentially miniature mechanical switches with a voltage used to control the
position of a shorting arm which is usually a cantilever or a membrane. As there is
no direct connection between the RF signal path and the control circuitry, MEMS
switches have inherently high operating frequencies. Power is required to change
the switch but once switching has been accomplished negligible DC power is re-
quired to maintain the connection.
1.8 Mixers
Frequency conversion, mixing or heterodyning, is the process of converting
information at one frequency (present in the form of a modulated carrier) to
another frequency. The second frequency is either higher, in the case of frequency
up-conversion, where it is more easily transmitted, or lower, when mixing is
called frequency down-conversion, where it is more easily captured. Capture of
the down-converted signal is nearly always by an Analog to Digital Converter
(ADC). Frequency conversion occurs with any nonlinear element. In Figure 1-
50(a) a nonlinear device is driven by two signals at
m
and
c
. The larger signal,
the LO, is called the pump and the other signal is called the RF. The spectrumof the
signals present in the circuit is shown in Figure 1-50(c). The aim here is to produce
a signal at the difference frequency (or intermediate frequency (IF)) with the same
modulation, and hence the same information, as the original RF signal. Two mixers
based on transistors are shown in Figure 1-51. The transistor mixer shown in Figure
1-51(a) uses ltering to separate the RF, LO, and IF components. Filters can be
large, so one of the particular advantages of the Gilbert mixer shown in Figure
1-51(b) is that it is a balanced mixer and ltering is not required to separate the
various signals. The symmetrical (or balanced) nature of this circuit means that
only differential mode signals at the input of the common source differential pairs
can appear at the output. Thus the largest signal present, the LO, is suppressed.
A balanced mixer can also be realized using diodes arranged in a ring to form
the diode ring double-balanced mixer shown in Figure 1-52. This mixer has the
advantage of being bidirectional, whereas the transistor mixer circuits of Figure
1-51 are unidirectional.
R
S
NONLINEAR
DEVICE
cos(w t) PUMP
cos(w t)
SIGNAL
LOAD
R
S
NONLINEAR
DEVICE
V
C
cos(w
C
t) PUMP
V
M
cos(w
M
t)
SIGNAL
LOAD
0.5
0.4
0.6
0.3
0.2
0.1
0.0
-0.1
-1.0 -0.6 -0.2 0.2 0.6 1.0
i
v (V)
(A)
(a) (b)
w
C
w
C
-
w
M
FREQUENCY
w
M
2w
C
w
C
+
w
M
3w
C
2
w
C
-
w
M
2
w
C
+
w
M
3
w
C
+
w
M
3
w
C
-
w
M
RF
or IF
SIGNAL
BASEBAND
VCO
PUMP
or LO
(c) (d)
Figure 1-50 Diode mixer: (a) circuit; (b) diode current-voltage characteristic; (c) spectrum
across the nonlinear device; and (d) schematic symbol for a mixer.
1.8.1 Mixer Analysis
A two-tone input
x(t) = [X
1
[ cos(
1
t +
1
) +[X
2
[ cos(
2
t +
2
)
can be written using complex notation as
x(t) =
1
2
_

X
1
e
1
t
+

X
1
e
1
t
+

X
2
e
2
t
+

X
2
e
2
t
_
.
Note that the coefcient of the positive exponential frequency component is one-
half that of the phasor. Thus the phasor of the
1
component is

X
1
= [X
1
[e
()
=
2

X
1
, and the phasor of the
2
component is

X
2
= [X
2
[e
()
= 2

X
2
. So the rst three
powers of x can be easily expanded manually, for example, expanding x
2
gives
x
2
(t) =
_
1
2
_
2 _
X
2
1
e
21t
+ 2X
1
X
1
+ 2X
1
X
2
e
(1+2)t
+ 2X
1
X
2
e
(12)t
+ (X
1
)
2
e
21t
+ 2X
1
X
2
e
(21)t
+ 2X
1
X
2
e
(1+2)t
+X
2
2
e
22t
+ 2X
2
X
2
+ (X
2
)
2
e
22t
, (1.61)
LO
FILTER
RF
FILTER
LO
RF
IF
FILTER
IF
GATE
BIAS
LO
FILTER
LO
FILTER
RF
FILTER
RF
FILTER
LO LO
RF RF
IF
FILTER
IF
IF
FILTER
IF
FILTER
IF
GATE
BIAS
(a)
LO
BALUN
RF
BALUN
LO
RF
IF
BALUN
IF
LO
BALUN
LO
BALUN
RF
BALUN
RF
BALUN
LO LO
RF
IF
BALUN
IF
BALUN
IF
(b)
Figure 1-51 Transistor-based mixer circuits: (a) single-ended FET mixer with LO, RF and
IF bandpass lters; (b) Gilbert mixer with baluns producing differential LO and RF signals.
RF
IF
LO
Figure 1-52 Diode ring double-balanced mixer.
and similarly expanding x
3
yields
x
3
(t) =
_
1
2
_
3 _
X
3
1
e
31t
+ 3X
2
1
X
1
e
1t
+ 3X
2
1
X
2
e
(21+2)t
+ 3X
2
1
X
2
e
(2
1
2
)t
+ 3X
1
(X
1
)
2
e
1
t
+ 6X
1
X
1
X
2
e
2
t
+ 6X
1
X
1
X
2
e
2t
+ 3X
1
X
2
2
e
(1+22)t
+ 6X
1
X
2
X
2
e
1t
+ 3X
1
(X
2
)
2
e
(122)t
+ (X
1
)
3
e
31t
+ 3(X
1
)
2
X
2
e
(221)t
+ 3(X
1
)
2
X
2
e
(21+2)t
+ 3X
1
X
2
2
e
(221)t
+ 3X
1
(X
2
)
2
e
(1+22)t
+X
3
2
e
32t
+ 6X
1
X
2
X
2
e
1t
+ 3X
2
2
X
2
e
2t
+ 3X
2
(X
2
)
2
e
2t
+ (X
2
)
3
e
32t
, (1.62)
Table 1-9 The intermodulation products resulting from x, x
2
, and x
3
, where x is a two-
tone signal, showing only the positive frequencies. The rst column gives the complex
amplitudes of the frequency components.
Intermodulation Product Frequency Order
1
2
X
1
X
1
0 2
1
2
X
2
X
2
0 2
2
1
2
X
1

1
1
2(
1
2
)
3
3X
2
1
X
1

1
3
2(
1
2
)
3
6X
1
X
2
X
2

1
3
2
1
2
X
2

2
1
2(
1
2
)
3
3X
2
2
X
2

2
3
2(
1
2
)
3
6X
1
X
1
X
2

2
3
2(
1
2
)
2
X
2
1
2
1
2
2(
1
2
)
2
X
2
2
2
2
2
2(
1
2
)
3
X
3
1
3
1
3
2(
1
2
)
3
X
3
2
3
2
3
2
1
2
X
1
X
2

1
+
2
2
2
1
2
X
1
X
2

1

2
2
2(
1
2
)
3
3X
2
1
X
2
2
1
+
2
3
2(
1
2
)
3
3X
2
1
X
2
2
1

2
3
2(
1
2
)
3
3X
1
X
2
2

1
+ 2
2
3
2(
1
2
)
3
3X
1
X
2
2
2
2

1
3
so that the output of the cubic equation
y(t) = a
0
+a
1
x(t) +a
2
x
2
(t) +a
3
x
3
(t)
can be calculated for a two-tone input. Table 1-9 lists these phasors and groups
themaccording to frequency. The phasors of the various intermodulation products
resulting from x, x
2
, and x
3
can be taken as the coefcients of the positive
exponential frequency components after the factor of two correction for terms
other than DC. Terms of the same frequency are summed to obtain the output
at a particular frequency. For example, the phasor output at
1
is given by the sum
of three intermodulation products:
Y
1
= a
1
_
1
2
_
X
1
+ 3a
3
_
1
2
_
3
X
2
1
X
1
+ 6a
3
_
1
2
_
3
X
1
X
2
X
2
. (1.63)
1.9 Early Receiver Technology
In this section, historical receivers are considered rst, in part because the terms
associated with the early receivers are still used, but also because the early trade-
offs inuence the architectures used today. Today receivers use DSP technology,
very stable LOs and sophisticated clock recovery schemes. This was not always
so. One of the early problems was using an LO to demodulate a signal when
transmitter oscillators drifted by many kilohertz. Radio at rst used AM and
the carrier was sent with the information-carrying sidebands. With this signal, a
simple rectier circuit connected to a bandpass lter could be used to rectify, but
the reception was poor. A crystal rectier consists of a single diode with lters. To
improve performance it was necessary to lock an oscillator to the carrier and then
amplify the received signal. Here some of the early schemes that addressed some
of the problems are discussed. There were many more variants, but the discussion
covers the essential ideas.
1.9.1 Heterodyne Receiver
The heterodyning principle mixes a single-tone signal, the LO, with a nite
bandwidth signal to produce a lower frequency version of the information-bearing
signal. With the LO frequency set appropriately, the low-frequency signal would
be in the audio range. If the information-bearing signal is an AM signal, then
the low-frequency version of the signal is the original audio signal which is the
envelope of the AM signal. This type of receiver is called a tuned radio frequency
(TRF) receiver, and performance is critically dependent on the stability of the
LO and the selectivity of the receive lter. The TRF receiver required the user
to adjust a tunable capacitor so that, with a xed inductor, a tunable bandpass
lter was created. Such a lter has limited Q
4
and a bandwidth that is wider than
the bandwidth of the radio channel. Even worse, a user had to adjust both the
frequency of the bandpass lter and the frequency of the LO. The initial radios
based on this principle were called audions, used a triode vacuum tube, and have
been in use since 1906. They were an improvement on the crystal detectors, but
there was a need for something better.
1.9.2 Homodyne Receiver
The homodyne [7], syncrodyne (for synchronous heterodyne) [8], and autodyne
(for automatic heterodyne) circuits were the needed improvements on the audion
and are based on the regenerative circuit invented by Edwin Armstrong in 1912
while he was an electrical engineering student at New York Citys Columbia
University [9]. Armstrongs circuit fed the input signal into an amplifying circuit
and a portion of this signal was coupled back into the input circuit so that the
signal was amplied over and over again. This is a positive feedback amplier. A
4
Q is the quality factor and is the ratio of the energy stored to the energy resistively lost each cycle.
Good frequency selectively in a lter requires high Q components. Tunable components have lower
Q than xed components.
OUTPUT
ANODE
+
GRID
CATHODE
(a) (b)
Figure 1-53 The Colebrooks original homodyne receiver: (a) circuit with an antenna,
tunable bandpass lter and triode amplier; and (b) triode vacuum tube.
small input RF signal was amplied to such a large extent that it resulted in the
amplifying circuit becoming nonlinear and consequently it rectied the amplitude
modulated RF signal. Colebrook used this principle and developed the original
homodyne receiver shown in Figure 1-53(a). This serves to illustrate the operation
of the family of regenerative receivers. The antenna shown on the left-hand side is
part of a resonant circuit that is in the feedback path of a triode oscillator. The triode
vacuum tube is annotated in Figure 1-53(b). Here the grid coils (which control the
ow of carriers between the bottom cathode
5
and top anode) are weakly coupled
to the anode circuit. When an AC signal appears at the top anode, the part within
the passband of the tuned circuit is fed back to the grid and the signal reinforced.
The radio signals of the day were AM and had a relatively large carrier and so the
oscillator tended to lock on to the carrier. The AMsidebands were then successfully
heterodyned down to the desired audio frequencies.
The autodyne worked on a slightly different principle in that the oscillation
frequency was tuned to a slightly different frequency from the carrier. Still, the
autodyne combined the functions of an oscillator and detector in the same circuit.
1.9.3 Superheterodyne Receiver
The superheterodyne receiver was invented by Edwin Armstrong in 1918 [10].
The key concept was to heterodyne down in two stages and to use xed lters
and use a tunable LO. The receiving antenna was connected to a bandpass lter
that allowed several channels to pass. This relaxed the demands on the receive
lter, but also lters with higher selectivity could be constructed if they did not
need to be tuned. Today we use high-order lters that are manually or machine
tuned, as manufacturing tolerances do not allow high-Q high-order lters to
be manufactured unmodied. The ltered received signal is then mixed with
an offset LO to produce what was called a supersonic signala signal above
the audio rangeand hence the name of this architecture. The performance of
the superheterodyne (or super HET) receive architecture has only recently been
achieved at cellular frequencies using direct conversion architecture requiring
large-scale integrated (and hence silicon) circuits. However, the superheterodyne
5
The cathode is heated (the heater circuit is not shown) and electrons are spontaneously emitted in a
process called thermionic emission.
90 90
VCO
i(t)
q(t)
a(t)
b(t)
s(t)
w
c
w
c
+ w
m
w
c
-w
m
FREQUENCY
w
c
w
c
+ w
m
w
c
-w
m
FREQUENCY
w
c
w
c
+ w
m
w
c
-w
m
FREQUENCY
w
c
w
c
+ w
m
FREQUENCY
Figure 1-54 Quadrature modulator showing intermediate spectra.
architecture is still superior above about 6 GHz.
1.10 Modern Transmitter Architectures
Modern transmitters maximize both spectral efciency and electrical efciency.
Spectral efciency is achieved by suppressing the carrier on transmit and
transmitting a single sideband. The classic technique for achieving this is
quadrature modulation, described in the next section. Electrical efciency must
be achieved with tight specications on allowable distortion and designs
must achieve this with minimum manual adjustments. Electrical efciency has
resulted in compound semiconductor transistors, including GaAs HBTs and
pHEMTS, mostly preferred for cellular handsets. For basestation and point-to-
point applications Si LDMOS is the dominant technology below a few gigahertz,
with high-breakdown galliumnitride (GaN) FETs being introduced. Another trend
is the development of universal amplier concepts so that the same RF frontend
can be used for a number of different applications. Multifunctional capability is
a cost driving transmitter architectures to minimize the RF analog hardware. The
discussion here focuses on narrowband communications when the modulated RF
carrier can be considered as a slowly varying RF phasor.
1.10.1 Quadrature Modulation
Quadrature modulation describes the frequency conversion process in that the
real and imaginary parts of the RF phasor are varied separately. A subsystem
which implements quadrature modulation is shown in Figure 1-54. This is quite an
ingenious circuit. The operation of this subsystem is described by what is known
as the generalized quadrature modulation equation:
s(t) = i(t) cos [
c
t +
i
(t)] +q(t) sin[
c
t +
q
(t)] . (1.64)
Here, i(t) and q(t) embody the particular modulation rule for amplitude,
i
(t)
and
q
(t) embody the particular modulation rule for phase, and
c
is the carrier
radian frequency. In terms of the signals identied in Figure 1-54, the quadrature
modulation equation can be written as
s(t) = a(t) +b(t) (1.65)
a(t) = i(t) cos [
c
t +
i
(t)] (1.66)
b(t) = q(t) sin[
c
t +
q
(t)] , (1.67)
where a(t) describes the output of the mixer at the top and b(t) describes the output
of the mixer on the bottom. The spectrum of a(t) as shown in Figure 1-54 has two
bands above and below the frequency of the carrier,
c
. Similarly the spectrum of
b(t) has two bands above and below the frequency of the carrier. However, there is
a difference. The LO(here designated as the voltage-controlled oscillator (VCO) is
shifted 90
(perhaps using an RC delay line) so that the frequency components of

b(t) have a different phase relationship to the carrier than those of a(t). When a(t)
and b(t) are combined the carrier content is canceled, as is one of the sidebands.
This is exactly what is desired: the carrier should not be transmitted, as it contains
no information. Also it is desirable to transmit only one sideband, as it contains
all of the information in the modulating signal. This type of modulation is called
suppressed carrier single-sideband (SCSS) modulation. The actual characteristics
depend on the particular forms of i(t), q(t),
i
(t) and
q
(t), and these dene the
modulation schemes. Only a few have the optimum properties of a well-dened
spectrum with steep sidewalls so that adjacent channels can be closely packed. In
the next section frequency modulation is used to demonstrate the SCSS operation.
In digital modulation, i(t) and q(t) are each derived from a bitstream, perhaps
simply by ltering a binary waveform.
1.10.2 Frequency Modulation
Frequency modulation is considered here to demonstrate suppressed carrier
single-sideband operation. Let i(t) and q(t) be nite bandwidth signals centered
at radian frequency
m
with (
q
(t)
i
(t)) being 90
on average. This is shown in

Figure 1-54, where
m
represents the frequency components of i(t) and q(t). With
reference to Figure 1-54, and setting
i
(t) = 0 =
q
(t),
i(t) = cos(
m
t) and q(t) = sin(
m
t) (1.68)
and the general quadrature modulation equation, Equation (1.64), becomes
s(t) = i(t) cos (
c
t) +q(t) sin(
c
t) = a(t) +b(t)
where
a(t) =
1
2
cos[(
c

m
)t] + cos[(
c
+
m
)t] (1.69)
and
b(t) =
1
2
cos[(
c
+
m
)t] cos[(
c

m
)t] . (1.70)
So the combined frequency modulated signal at the output is
s(t) = a(t) +b(t) = cos[(
c
+
m
)t] , (1.71)
and as well as the carrier being suppressed, so is the lower sideband. This lower
sideband is also referred to as the image. In modulators it is important to suppress
this image and in demodulators it is important that undesired signals at the image
frequency not be converted along with desired signals .
1.10.3 Polar Modulation
Polar modulation is a relatively new modulation scheme for impressing
information on a carrier. In polar modulation, the i(t) and q(t) quadrature signals
are converted to polar form as amplitude A(t) and phase (t) components.
This is either done in the DSP unit or if a modulated RF carrier is all that is
provided, using an envelope detector to extract A(t) and a limiter to extract the
phase information corresponding to (t). Two polar modulator architectures are
shown in Figure 1-55. In the rst architecture, Figure 1-55(a), A(t) and (t) are
available and A(t) is used to amplitude modulate the RF carrier which is then
amplied by a Power Amplier (PA). The phase signal, (t) is the input to a phase
modulator implemented as a PLL. The output of the PLL is fed to an efcient (i.e.,
nonlinear) amplier operating near saturation (also called a saturating amplier).
The outputs of the two ampliers are combined to obtain the large modulated RF
signal to be transmitted.
In the second polar modulation architecture, Figure 1-55(b), a low-power
modulated RF signal is decomposed into its amplitude and phase modulated
components. The phase component, (t), is extracted using a limiter which
produces a pulse-like waveformwith the same zero crossings as the modulated RF
signal. Thus the phase of the RF signal is captured. This is then fed to a saturating
amplier whose gain is controlled by the carrier envelope or A(t). Specically
A(t) is extracted using an envelope detector, with a simple implementation being
a rectier followed by a lowpass lter with a corner frequency equal to the
bandwidth of the modulation. A(t) then drives a switching (and hence efcient)
power supply that drives the saturating power amplier. Polar modulation is
nding application in the Universal Mobile Telecommunications System (UMTS),
as the 8PSK modulation used in UMTS results in low efciency if direct power
amplication of a modulated carrier is used.
1.10.4 Direct Conversion Modulation
The transmitted signal has a controlled spectral content, being just a single
channel. Issues such as image rejection and interferers are not problems. The
overwhelming majority of wireless transmitters are based on the heterodyne
principle where baseband I and Q signals are used in a vector modulator to
produce a modulated IF that is upconverted in a mixer to an RF signal. The nal
( ) t A
( ) t
COMBINER
PHASE
MODULATOR
AMPLITUDE
MODULATOR
RF
CARRIER
MODULATED RF
OUTPUT
PA
PA
(a)
ENVELOPE
DETECTOR
MODULATED
RF INPUT
( ) t A
SWITCHING
POWER SUPPLY
MODULATED RF
OUTPUT
( ) t
LIMITER
PA
(b)
Figure 1-55 Polar modulator architectures: (a) amplitude and phase modulated compo-
nents amplied separately and combined; and (b) the amplitude used to modulate a power
supply driving a saturating amplier with phase modulated input.
TRANSMIT
MODEM
QUADRATURE
MODULATOR
PA
PREDISTORTER
QUADRATURE
MODULATION
CORRECTION
DATA
I
Q
TRANSMIT
MODEM
QUADRATURE
MODULATOR
PA PA
PREDISTORTER
QUADRATURE
MODULATION
CORRECTION
DATA
I
Q
Figure 1-56 Architecture of a direct conversion transmitter.
RF signal is then amplied by a power amplier. A direct conversion transmitter
generates the RF signal directly without an IF stage using the architecture shown
in Figure 1-56. Here, the transmit modem rst produces I and Q baseband
signals from the data. This is then translated directly to RF via a quadrature
modulator and then amplied by a PA. Practically, the nonlinearities of the PA
must be linearized using predistortion, and quadrature modulation errors must
be accounted for in a quadrature modulation corrector. The transmit modem, the
predistorter and the quadrature modulation corrector can be combined in a DSP,
so considerable complexity is shifted to the DSP chip. One of the major problems
in this architecture is the noise introduced by errors in step-size mismatch. Noise-
shaping techniques implemented in a DSP have been developed to shift this noise
outside the bandwidth of the generated signal. Similar errors are associated with
mismatches of the ADC and of the analog circuit paths of the separate I and Q
paths. This distortion is also pushed out of band by the noise-shaping algorithm.
1.11 Modern Receiver Architectures
Communication receivers most commonly use mixing of the RF signal with a xed
signal called an LO to produce a lower frequency replica of the modulated RF
signal. Some receiver architectures use one stage of mixing, while others use two
stages of mixing. In cellular systems, the receiver must be sensitive enough to
detect signals of 1 pW or less. Some of the architectures used in modern receivers
are shown in Figure 1-57. Figure 1-57(a) is the superheterodyne architecture in
much the same form it has been used for almost a century. Key features of this
architecture are that there are two stages of mixing, and ltering is required to
suppress spurious mixing products.. Each mixing stage has its own VCO. The
receiver progressively reduces the frequency of the information bearing signal. The
image rejection mixer in the dashed box achieves rejection of the image frequency
to produce an IF (or baseband frequency) that can be directly sampled. It is,
however, difcult to achieve the amplitude and phase balance, especially when
the image reject lter is realized on an integrated circuit. Instead the architecture
shown in Figure 1-57(b) is used. The lter between the two mixers can be quite
large. For example, if the incoming signal is 1 GHz, the frequency of the signal after
the rst mixer could be 100 MHz. Filters become smaller at higher frequencies for
the same performance, so the dual-conversion receiver shown in Figure 1-57(c).
This is very similar to the traditional superheterodyne architecture except that the
intermediate frequency between the two mixers is high. In the previous example
it could be 3 GHz. The required bandpass lter can be realized as an integrated
circuit. The Low-IF or Zero-IF receiver shown in Figure 1-57(d) uses less hardware
and is used in less demanding communication applications.
1.11.1 Homodyne Frequency Conversion
Homodyne mixing and detection is one of the earliest wireless receiver
technologies and is used in AM radio. Homodyne mixing can be used for
detecting modulation formats other than AM, including digitally modulated
signals, and is a particularly attractive architecture for monolithically integrated
circuits. In homodyne mixing, the carrier of a modulated signal is regenerated
and synchronized in phase with the incoming carrier frequency. Mixing the carrier
with the RF signal results in an IF signal centered around zero frequency. The only
simple way to ensure that the pump is in phase is to transmit the carrier with
the RF signal. AM transmission does just this but at the cost of transmitting large
carrier power, as well as the additional prospect of interference that goes along
with this. Homodyne mixing can be used with digitally modulated signals.
Signal spectra that result in homodyne mixing are shown in Figure 1-58. In
Figure 1-58(a), the RF signals are shown on the right-hand side and the baseband
signals are shown on the left hand side. It is usual to show both positive and
negative frequencies at the lower frequencies so that the conversion process is
more easily illustrated. The characteristic of homodyne mixing is that the pump
corresponds to the carrier and is in the middle of the desired RF channel. RF
BPF
HIGH
f
MED
f
LOW
f
HIGH
BPF
BPF
VCO1
VCO2
VCO
IF
I
Q
I
Q
I
Q
I
Q
VCO1
VCO2
SUPERHETERODYNE, HARTLEY IMAGE REJECTION
SUPERHETERODYNE RECEIVER
VCO1
VCO2
DUAL CONVERSION RECEIVER
LOWIF OR ZEROIF RECEIVER
(b)
(c)
(d)
(a)
BPF
90
f
LOW
LPF
LPF
ADC
ADC
90
90
LPF ADC
f
BPF
LPF
LPF
ADC
ADC
90
f
HIGH
f
LOW
f
HIGH
BPF BPF
LPF
LPF
ADC
ADC
90
f
HIGH
f
MED
f
LOW
+
Figure 1-57 Architecture of modern receivers: (a) superheterodyne receiver using the
Hartley architecture for image rejection; (b) superheterodyne receiver; (c) dual-conversion
receiver; low-IF or zero-IF receiver. PBF, bandpass lter; LBF, LowPass Filter; ADC, analog
to digital converter; VCO, voltage controlled oscillator; 90, 90
phase shifter; I, in-phase

component, Q, Quadrature component; fHIGH, fMED and fLOW indicate relatively high,
medium and low frequencies in the corresponding section of the receiver.
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
FREQUENCY
DC
FILTER
FREQUENCY
DC
FILTER
DC
FILTER
(a) (b)
Figure 1-58 Frequency conversion using homodyne mixing: (a) the spectrum with a large
local oscillator or pump; and (b) the baseband spectrum showing only positive frequencies.
signal components mix with the pump and it appears that the entire RF spectrum
is down-shifted around DC. Of course, the actual baseband spectrum (the term
for the lowest frequency signals) is only dened for positive frequencies, so the
negative frequency baseband signals and the positive frequency baseband signals
add to yield the baseband spectrum shown in Figure 1-58. With other modulation
schemes, this loss of information is avoided using quadrature demodulation
discussed later in the discussion of zero IF conversion. An amplitude modulated
signal has identical modulation sidebands and so the collapsing of positive and
negative frequencies at baseband results in no loss of information. Then simple
amplitude detection circuitry, such as a rectier, is used and the rectied signal is
passed directly to a speaker.
1.11.2 Heterodyne Frequency Conversion
In heterodyne mixing, the pump and the main RF channel are separated in
frequency, as shown in Figure 1-59. In this gure the RF signals (shown as three
discrete channels on the right-hand side of the spectrum) mix with the pump
signal to produce signals at a lower frequency. This lower frequency is usually
not the nal baseband frequency desired and so is called the IF. The intermediate
frequency of the main channel is at the difference frequency of the RF signal and
the pump. The pump must be locally generated and so is called the LO signal.
There are several important renements to this. The rst of these is concerned with
limiting the number of signals that can mix with the pump signal. This is done
using an RF preselect lter, resulting in the spectrum shown in Figure 1-60(b).
The important characteristic is that the signals at frequencies below the pump
have been suppressed, however, the image channel is still of concern. To see the
difculties introduced by the image channel consider the frequency conversion-
to-baseband process described in Figure 1-61. The RF spectrum after RF preselect
ltering is shown in Figure 1-61(a) and the baseband (or IF) spectrum is shown in
Figure 1-61(b). Again, positive and negative frequencies are used to better illustrate
the down-conversion process. Note the down-converted image and the main
channel are equidistant from DC. So referring the signals to a positive frequency-
only spectrum, Figure 1-61(c), it can be seen that the image channel interferes with
the main channel. In the worst-case scenario, the IF image could be larger than that
of the desired channel. Fortunately there is a circuit x that compensates for this.
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
INTERMEDIATE
FREQUENCY
FILTER
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
INTERMEDIATE
FREQUENCY
FILTER
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
INTERMEDIATE
FREQUENCY
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
INTERMEDIATE
FREQUENCY
FILTER FILTER
(a)
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
FILTER
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
FILTER
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
FILTER FILTER
INTERMEDIATE
FREQUENCY
INTERMEDIATE
FREQUENCY
INTERMEDIATE
FREQUENCY
BASEBAND
2
(b)
Figure 1-59 Frequency conversion using superheterodyne mixing: (a) the pump is offset
from the main channel producing a down-converted channel at the intermediate frequency;
and (b) a second pump down-converts the main channel, nowat the intermediate frequency,
to the baseband frequency.
FREQUENCY
DC
PUMP
MAIN
CHANNEL
RF PRESELECT
FILTER
IMAGE
(a)
FREQUENCY
DC
PUMP
MAIN
CHANNEL
RF PRESELECT
FILTER
RF PRESELECT
FILTER
IMAGE
(b)
Figure 1-60 Frequency conversion using heterodyne mixing showing the use of an RF
preselect lter to reduce the image signal.
FREQUENCY
DC
PUMP
MAIN
CHANNEL
RF PRESELECT
FILTER
RF PRESELECT
FILTER
IMAGE
(a)
FREQUENCY DC
DOWN-CONVERTED
MAIN CHANNEL
IF
FILTER
IF
FILTER
IMAGE
IF
+IF
(b)
FREQUENCY DC
MAIN
CHANNEL
WITH IMAGE
+IF
(c)
Figure 1-61 Frequency conversion using heterodyne mixing showing the effect of image
distortion: (a) the RF spectrum following ltering using an RF preselect lter; (b) the
baseband down-converted signal showing positive and negative frequencies; and (c) the
single-sided baseband spectrum following IF ltering showing the contamination of the
nal signal by the image signal.
The block diagrams of the circuits that correct this image problem are shown in 1-
62(a) and Figure 1-62(b) for both transmit and receive functions. For now, consider
the quadrature receiver shown in Figure 1-62(a). Shown here is an antenna that
takes in the RF signal, and the RF preselect function is performed by a bandpass
lter (BPF). This signal is initially amplied (usually), split and applied equally to
two mixers. In an ideal mixer, the two signals are multiplied together. The mixers
have pump (LO) signals that are 90
out of phase (i.e., in quadrature) with each

other and contain different information. The phase relationships at the outputs of
the two mixers have very particular relationships with each other such that when I
and Qare added the image is eliminated, as shown in Figure 1-62(e). With digitally
modulated signals, however, the I and Q waveforms are individually sampled by
ADCs.
Figure 1-62(c) shows the baseband spectrum at the I output of the heterodyne
receiver. Figure 1-62(d) shows the baseband spectrum at the Q output of the
heterodyne receiver. It also shows the negative frequency components inverted.
This is a shorthand way of saying that the negative (image) frequency components
Q
LO
90
o
RF
I
BPF
LPF
LPF
RF
LO
90
o
BPF
I
Q
(a) (b)
FREQUENCY DC
DOWN-CONVERTED
MAIN CHANNEL
IF
FILTER
IMAGE
IF
+IF
FREQUENCY DC
DOWN-CONVERTED
MAIN CHANNEL
IF
FILTER
IF
FILTER
IMAGE
IF
+IF
(c)
FREQUENCY DC
DOWN-CONVERTED
MAIN CHANNEL
IMAGE
IF
+IF
(d)
FREQUENCY DC
MAIN CHANNEL WITH
IMAGE CANCELLATION
+IF
FREQUENCY DC
MAIN CHANNEL WITH
IMAGE CANCELLATION
+IF
(e)
Figure 1-62 Quadrature mixing: (a) receive modulator; and (b) transmit modulator.
Frequency conversion using heterodyne mixing and quadrature mixing; (c) the baseband
spectrum at the I output of the heterodyne receiver; (d) the baseband spectrum at the Q
output; and (e) the positive spectrum following the summation of the I and Q channels at
the output of the heterodyne receiver.
are 180
out of phase with the down-converted image components of the in-phase

signal. However, the down-converted main channel and neighbors are in phase in
both spectra. Thus the combination of the I and Q outputs suppresses the image
signals from baseband. Figure 1-62(e) illustrates the positive spectrum following
the summation of the I and Q channels at the output of the heterodyne receiver.
Image rejection is key to all commonly used wireless communications systems
today.
The quadrature transmitter shown in Figure 1-62(b) works almost identically, but
in reverse. The I and Q waveforms are each applied to mixers, with one pumped
by an LOand the other by the LOshifted by 90
. The outputs of each mixer contain

upper and lower sidebands on either side of the LO (or carrier) frequency. When
these are combined at the summing node, one of the sidebands (called the image)
is canceled and only a single carrier is presented to the bandpass lter (BPF) and
then radiated by the antenna. In QPSK digitally modulated systems, the I and Q
waveforms are lowpass ltered digital bitstreams.
Heterodyne systems implement signal processing such as ltering, modulation
and demodulation, and image rejection at RF and IF using hard to integrate
discrete components leading to expense and limitations on size reductions.
Heterodyne architectures are regarded as approaching their limit in size,
integration, and fabrication cost. The primary issue in mixer design is limited
image rejection resulting from gain and phase mismatches of the I (in-phase) and
Q (quadrature) paths.
1.11.3 Direct Conversion Receiver
Zero-IF direct conversion receivers are very similar to quadrature homodyne
receivers with an LO signal placed at the center of the RF channel. The difference
is that in homodyne receivers the phase of the carrier (i.e., is the phase of the LO)
is precisely known as the carrier is transmitted with the signal. In virtually all RF
transmission schemes (above a few megahertz) the carrier is not transmitted. Thus
in zero-IF direct conversion schemes the LO signal has inherent phase error with
the original carrier. The important characteristic is that there is only one level of
mixing. The conversion process is described in Figure 1-63. Aparticular advantage
of direct conversion is that the relatively large IF lters are eliminated. Thus the I
and Q mixer outputs are necessary, as the two sides of the RF spectrum contain
different information, and there would be irreversible corruption if a scheme was
not available to extract the information in each of the sidebands.
The main nonideality of this design is the DC offset in the down-converted
spectrum. DC offset results mostly from self-mixing, or rectication, of the LO.
This DC offset can be much larger than the down-converted signal itself, and
because of the nonlinearities of baseband amplication stages, either severely
limits the dynamic range of the receiver or places limitations on the modulation
format that can be used. One way of coping with the DC offset is to highpass
lter the down-converted signal, but highpass ltering requires a large passive
component (e.g., a series capacitor) at least to avoid dynamic range problems with
active lters. Highpass ltering the down-converted signal necessarily throws
away information in the signal spectrum and it is only satisfactory to do this if
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
(a)
(b)
FREQUENCY
DC
FILTER
FREQUENCY
DC
FILTER
(c)
FREQUENCY
DC
FILTER
FREQUENCY
DC
FILTER
I Q
Q I
FREQUENCY
DC
PUMP
MAIN
CHANNEL
DOWN
CONVERTED
CHANNEL
(a)
(b)
FREQUENCY
DC
FILTER
FREQUENCY
DC
FILTER
(b)
FREQUENCY
DC
FILTER
FREQUENCY
DC
FILTER
(c)
FREQUENCY
DC
FILTER
FREQUENCY
DC
FILTER
I Q
Q I
Figure 1-63 Frequency conversion using homodyne mixing and quadrature mixing: (a)
the baseband spectrum at the I output of the homodyne receiver; (b) the baseband spectrum
at the Q output of the homodyne receiver; and (c) the positive spectrum following the
summation of the I and Q channels at the output of the homodyne receiver.
there is very little information around DC to begin with.
The primary effort in zero-IF converters is overcoming the DC offset problem, and
to a lesser extent coping with jitter of the LO. The primary LO noise of concern is
close-in phase noise, which can be at appreciable levels 100 kHz from the carrier.
This noise is commonly referred to as icker noise and increases rapidly as the
offset from the carrier reduces. This is of concern in all conversion processes.
However, one of the properties of heterodyne mixing is that the RF signal is
considerably offset from the large phase noise region. Consequently the LO phase
noise at the frequency of the RF reduces the impact on the resulting offset IF signal.
For these reasons, heterodyne mixing provides higher performance than direct
conversion. Also, direct conversion receivers are difcult to implement for 8-state
(8PSK etc.) and higher order modulation.
In cellular wireless, the radio signals are spectrally efcient and the spectrum is
fairly constant across the channel. So the near-DC signals that result from direct
conversion have appreciable information content and cannot be discarded so easily
without signicant distortion.
The main problems of zero-IF conversion in cellular radio applications are
1. Spurious LO leakage. Retransmission of the LO is possible because the LO is
tuned precisely to the RF signal frequency and reverse leakage through the
RF path will radiate from the antenna. Spurious LO transmission is severely
regulated. The limit on this in-band LO radiation is between 50 and 80 dBm.
The problem is reduced by using differential LOs and using multiple RF
amplier stages to increase the reverse isolation between the mixer and the
antenna.
2. Interferer leakage. A large RF interferer can leak through the RF amplier and
enter the mixer through both the LO port as well as the RF port. Mixing of these
components results in DC offset.
3. Distortion. Direct conversion receivers are more sensitive to undesired signals
than are heterodyne receivers. The nonlinearities of the input mixers will rectify
strong spurious RF signals to produce output components around DC. This is
the result of second-order nonlinearities and so this effect can be suppressed
through the use of balanced RF circuits which will have only odd-order
nonlinearities. This reduces the baseband signals that can result from large RF
interferers, but it is still possible to produce baseband distortions if the RF
interferer is large enough to produce third harmonics in the mixer. The effect
of this distortion can be largely eliminated through the use of highpass ltering
at the baseband, but this is not acceptable for cellular radio signals. Heterodyne
conversion is susceptible to distortions resulting from odd-order nonlinearities,
but in zero-IF converters second order distortion is also a problem.
4. LO self-mixing. Mixing of the LO with itself will produce a DC signal in the
mixer output. This DC level may be many orders of magnitude larger than the
baseband signal itself, so it can desensitize or saturate the baseband amplier.
DC offsets can also result from circuit mismatch problems. The DC offset that
results primarily from LO self-mixing is the most signicant problem in the
use of zero-IF architectures in cellular wireless. The DC offset can be reduced
through the use of balanced designs, but circuit mismatch errors still result in
very large DC offsets.
5. LO frequency error. A difference between the LO and the carrier will cause the
RF signal to be asymmetrically converted around DC.
6. Second-order distortion. Because of second-order distortion, second harmonics
of the signal can appear in the baseband. This is a problem if the RF signal is
large to begin with. This problem can be circumvented by using designs that
utilize differential signals.
7. I/Q mismatches. Mismatches of the I and Q paths also results in DC offsets.
These offsets, however, vary negligibly with time, and analog or digital
calibration techniques can be used to remove their effect.
The problem of DC offset is made worse because the DC level can vary with time
as the amplitude of the interferer varies, or the LO that leaks from the antenna
reects off moving objects and is received as a time-varying interferer itself.
1.11.4 Low-IF Receiver
In a low-IF receiver, single-stage heterodyne mixing is used to down-convert the
modulated RF carrier to a frequency just above DC, perhaps to a few hundred
kilohertz or a few megahertz depending on the bandwidth of the RF channel.
In doing this, the DC offset problem of a direct conversion receiver is avoided.
The particular advantage of a low-IF receiver is that it can be used with higher
order modulation formats (8PSK and higher). It does, however, require a higher
performance ADC than does a direct conversion receiver.
1.11.5 Subsampling Analog-to-Digital Conversion
Subsampling receivers overcome the DC offset problem typical of other direct
conversion receivers. The idea is to sample the modulated RF signal using an
exact subharmonic of the carrier of the RF channel to be converted. The sampling
rate must be at least twice the bandwidth of the RF signal and the track-
mode bandwidth must be greater than the carrier frequency. Thus the sampling
aperture is the critical parameter and must be several times smaller than the
period of the carrier. Fortunately the aperture times of CMOS tracking circuits are
adequate. It is critical that an RF preselect lter be used to eliminate unwanted
interferers and noise outside the communication band. Aliasing of signals outside
the Nyquist bandwidth onto the baseband signal is a consequence of subsampling.
Adjacent channel signals are converted without aliasing, but these will lie outside
the bandwidth of the baseband signal. Flicker noise on the sampling clock is
multiplied by the subsampling ratio and appears as additional noise in baseband.
1.11.6 First IF-to-Baseband Conversion
In a superheterodyne conversion architecture there are two heterodyne stages,
with the IF of the rst stage in the range of 20 to 200 MHz. The assignment of
frequencies is known as frequency planning and this is treated as proprietary by
the major radio vendors. This IF is then converted to a much lower IF, typically
around 100 kHz or higher. This frequency is generally called baseband, but strictly
it is not because the signal is still offset in frequency from DC. Some direct
conversion architectures leave the rst heterodyne mixing stage in place and use
direct conversion of the rst IF to baseband (true basebandaround DC).
1.11.7 Bilateral Double-Conversion Receiver
The receivers considered so far are suitable for narrowband communications
typical of point-to-point and consumer mobile radio. There are many situations
were the received or transmitted RF signal covers a very wide bandwidth, such as
with emergency radios, television, and military communications. If narrowband
RF frontend architectures are used, a switchable lter bank would be required
and this would result in an impractically large radio. Tunable bandpass lters
are one option being explored. The current preferred solution to covering wide
bandwidths is the double-conversion transceiver architecture shown in Figure 1-
64. The frequency plan of a typical radio using 0.1 MHz channels between 20 MHz
and 500 MHz is shown. The key feature of this radio is that bidirectional mixers
are used, such as the diode ring mixer of Figure 1-52 on page 64. Following the
FILTER
1262 MHz
ADC
DAC
BANDPASS
FILTER
BANDPASS
782 MHz to
f
LO1
f
LO2
f
RF
20 MHz to
500 MHz
f
IF
TUNABLE TUNABLE
1282 MHz
FIXED
10 MHz WIDE
68 MHz
FIXED
1 MHz WIDE
NOTCH
FILTER
FILTER OR
LOWPASS
BLOCKS
IMAGE
TUNABLE
68 MHz
0.1 MHz Channel
SWITCH
Rx
Tx
Figure 1-64 Bilateral double conversion transceiver for wideband operation.
RF chain from left to right, the RF is rst mixed up in frequency, bandpass ltered
using a high-Q distributed lter, and then down-converted to a lower frequency
which can be sampled directly by an ADC. A much higher performance passive
(and hence bidirectional) lter can be realized at gigahertz frequencies than at a
fewtens of megahertz. On transmit, the function is similar, with the mixers and LO
source reused. As a receiver, the notch lter or lowpass lter is used to block the
image frequency of the rst mixer so that only the upper sideband IF is presented
to the rst bandpass lter. The lowpass or notch lter may be xed, although with
the plan shown there must be at least two states of the lters. On transmit, the
lowpass or notch lters prevent the image frequency from being radiated.
1.12 Summary
This chapter presented the RF frontend architectures used from the beginnings
of wireless communications up to those used in modern systems. Similar
architectures are used in the frontends of radar and sensor systems. Wireless
systems proliferate, and even in established domains such as cellphones,
architectures are evolving to achieve greater efciency, greater multifunctionality,
and lower cost primarily by monolithically integrating and digitizing as much
as possible of the RF frontend. Size drives the replacement of superheterodyne
architecture by eliminating large intermediate lters.
1.13 Exercises
1. An amplier has a power gain of 1200. What is the power gain in decibels?
2. The PAR of a signal is an important parameter in determining the efciency
that can be achieved by an amplier with an allowable amount of distortion.
The following questions are about determining the PAR of various signals.
Note that the average power level is not necessarily the average of the
minimum and maximum power levels. A full power calculation and integral
should be performed. [WS3]
(a) Write down a formula for the power of a signal x(t). You can consider x(t)
to be a voltage across a 1 resistor.
(b) What is the PAR of an FM signal at 1 GHz with a maximum modulated
frequency deviation of 10 kHz?
(c) What is the PAR of a two-tone signal (consisting of two sinewaves
at different frequencies that are say 1% apart)? First, use a symbolic
expression and then consider the special case when the two amplitudes
are equal. Consider that the two tones are close in frequency.
(d) What is the PAR of a three-tone signal (consisting of three sinewaves say
1% apart in frequency) when the amplitude of each sinewave is the same?
(e) What is the PAR of an AM signal with 75% amplitude modulation?
3. Short answer questions on gain calculations.
(a) An amplier with 50 input impedance and 50 load impedance has a
voltage gain of 100. What is the gain in decibels?
(b) An attenuator reduces the power level of a signal by 75%. What is the gain
of the attenuator in decibels?
(c) What is the wavelength in free space of a signal at 4.5 GHz?
4. An amplier consists of three cascaded stages with the following characteris-
tics
Characteristic Stage 1 Stage 2 Stage 3
Gain 10 dB 15 dB 30 dB
NF 0.8 dB 2 dB 2 dB
What is the noise gure (NF) and gain of the cascaded amplier?
5. The frontend of a receiver for a cellular phone has a bandpass lter with a
25 MHz passband and loss of 2 dB and is followed by two amplier stages.
The rst stage has a gain of 20 dB and a noise gure of 0.5 dB and the second
stage has a gain of 60 dB and a noise gure of 2 dB.
(a) Sketch the block diagram of the system as described.
(b) What is the gain of the system?
(c) What is the noise gure of the bandpass lter?
(d) What is the noise gure of the system?
(e) The system is now connected to an antenna with an effective noise
temperature of 30 K and which delivers a signal of 10 pW to the bandpass
lter. Determine the noise temperature at the output of the system and
hence the noise power in the 25 MHz bandwidth. Determine the signal-
to-noise ratio at the output of the frontend system.
6. An FM signal has a maximum frequency deviation of 20 kHz and a modu-
lating signal between 300 Hz and 5 kHz. What is the bandwidth required to
transmit the modulated RF signal when the carrier is 200 MHz? Is this consid-
ered to be narrowband FM or wideband FM?
7. Consider two analog signals combined together. One signal is denoted x(t)
and the other y(t) where x(t) = 0.1 sin(10
9
t) and y(t) = 0.05 sin(1.01 10
9
t).
What is the PAR of this combined signal? Express PAR in decibels.
8. A high-delity stereo audio signal has frequency content ranging from 50 Hz
to 20 kHz. If the signal is to be modulated on an FM carrier at 100 MHz, what
is the bandwidth required for the modulated RF signal? The maximum fre-
quency deviation is 5 kHz when the modulating signal is at its peak value.
9. The following sequence of bits 0100110111 is to be transmitted using QPSK
modulation. Take these data in pairs, that is, as 01 00 11 01 11. These pairs, one
bit at a time, drive the I and Q channels. Show the transitions on a constella-
tion diagram.
10. The following sequence of bits 0100110111 is to be transmitted using OQPSK
modulation. Take these data in pairs, that is as 01 00 11 01 11. These pairs, one
bit at a time, drive the I and Q channels. Show the transitions on a constella-
tion diagram. [WS24]
11. The following sequence of bits 0100110111 is to be transmitted using /4-
DQPSK modulation. Take these data in pairs, that is, as 01 00 11 01 11. These
pairs, one bit at a time, drive the I and Q channels. Use four constellation dia-
grams, with each diagram showing one transition. On each diagram, indicate
the initial symbols and the next symbols (rotated of course by 45
). [WS25]
12. The cascaded two-port network in Figure 1-45 consists of a lter with an in-
sertion loss of 3 dB followed by an amplier with a noise gure of 2 dB and a
gain of 20 dB. Calculate the total gain and noise gure. This problem parallels
example 1.4 on page 237.
13. A Class A HBT amplier is biased with a collector emitter quiescent voltage of
5 V and a quiescent collector-emitter current of 100 mA. When operated at the
1 dB compression point, the input RF power is 10 mW and the output power
is 100 mW.
(a) What is the quiescent DC power consumed? Express your answer in
milliwatts.
(b) What is the output power in dBm?
(c) What is the efciency of the amplier? Note that the efciency of a class A
amplier can be more than 25% if distortion can be tolerated.
(d) What is the power-added efciency of the amplier?
(e) If the input power is reduced by 10 dB so that the amplier is no longer in
compression, will the DC quiescent point change? Explain your answer.
(f) If the input power is reduced 10 dB so that the amplier is no longer in
compression, what is the output power in dBm? Ignore any change in the
quiescent point.
(g) With 1 mW input power, what is the power-added efciency of the
amplier if the quiescent point does not change?
14. The BJT amplier in Figure 1-65(a) has a load, R
L
, and a maximumundistorted
efciency of 25%. Derive this efciency.
V
V
O
OUTPUT
INPUT
V
I
CC
L
R
INPUT
V
CC
V
O
V
I
C
OUTPUT
L
L
R
(a) (b)
Figure 1-65 Class A HBT amplier with load RL: (a) with the collector bias also provided
through R
L
; and (b) collector supplied through an inductor L which is an open circuit at
AC frequencies.
15. The BJT amplier in Figure 1-65(b) has an RF choke providing collector current
and acting as an open circuit at RF. The load, R
L
, is driven through a capacitor
C which is effectively a short circuit at RF. The maximum undistorted
efciency of this circuit is 50%. Derive this efciency. Ignore the base-emitter
voltage drop, V
CE,min
, and note that the maximum of V
O
is V
CC
, allowing a
voltage swing of V
CC
around the collector quiescent operating voltage.
16. A 75 attenuator has a loss of 16 dB and is between a source with a Thevenin
impedance of 75 and a load of 75 .
(a) What is the noise power, N
i
, available from the 75 source resistor at
standard temperature (270 K) in a 1 MHz bandwidth?
(b) Now consider that the attenuator is connected to the attenuator which is
also connected to the load. If the source generates a modulated signal that
is 1 MHz wide and has an available power, S
i
, of 1 fW, what is SNR
i
at the
input to the attenuator at standard temperature?
(c) With the attenuator connected to the source, what is the Thevenin
equivalent impedance looking into the output of the attenuator?
(d) Calculate the noise power N
o
available fromthe attenuator with the source
attached at standard temperature (270 K) in a 1 MHz bandwidth?
(e) What is the signal power, S
o
, delivered to the load?
(f) What is the SNR at the load, SNR
o
?
(g) What is the noise factor, F, of the attenuator?
(h) What is the noise gure, NF, of the attenuator?
17. A superheterodyne receiver has in order an antenna, a low-noise amplier, a
bandpass lter, a mixer, a second bandpass lter, a second mixer, a lowpass
lter, an ADC, and a DSP which will implement quadrature demodulation.
Develop the frequency plan of the receiver if the input RF signal is at 2 GHz
and has a 200 kHz single-channel bandwidth. The nal signal applied to the
ADCmust be between DCand 400 kHz so that I/Qdemodulation can be done
in the DSP unit. Noise considerations mandate that the LO of the rst mixer
must be at least 10 MHz fromthe input RF. Also, for a minimumsize bandpass
lter between the two mixers, the lter should be at as high a frequency as
possible and it has been determined that a 100 MHz lter is available at an
acceptable cost, so it has been decided that it will be used.
(a) Draw a block diagram of the receiver and annotate it with symbols for the
frequencies of the LOs and the RF and IF signals.
(b) What is the LO frequency f
LO1
of the rst mixer?
(c) What is the LO frequency f
LO2
of the second mixer?
(d) Specify the cutoff frequency of the lowpass lter following the second
mixer.
(e) Briey discuss in less than half-page other design considerations as they
relate to the frequency plan, lter size, and lter specication.
18. Short answer questions. Each part requires a short paragraph of about ve
lines and a gure where appropriate to illustrate your understanding.
(a) Consider a two-tone signal and describe intermodulation distortion.
(b) Describe the effect of a lossy lter on the SNR. Consider signals at the
input and output of the lter?
(c) What is meant by 1 dB gain compression?
(d) Consider a digitally modulated signal and describe the impact of a
nonlinear amplier on the signal. You must include several (at least)
negative effects.
19. Short answer questions. Each part requires a short paragraph of about ve
lines and a gure where appropriate to illustrate your understanding.
(a) Explain the operation of a superheterodyne receiver?
(b) Compare zero-IF and low-IF receivers.

Steer RF Chapter1

Uploaded by

Copyright:

Available Formats

Steer RF Chapter1

Uploaded by

Document Information

Original Title

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

Steer RF Chapter1

Uploaded by

Copyright:

Available Formats

CHAPTER 1

[cos(A+B)t + cos(A+B)t] dt = 0, (1.10)

x(t)y(t + ) dt = 0 for all ,

and so avoid the RF phasor traversing the

shift of the phasor.

. Filtering of the baseband modulating signal is required to minimizes the

from one interval to the next. This can

is possible during one symbol. The constellation diagram is

Figure 1-24 Block diagram of an MSK modulator.

out of phase. The

out of phase) and the

(perhaps using an RC delay line) so that the frequency components of

on average. This is shown in

phase shifter; I, in-phase

out of phase (i.e., in quadrature) with each

out of phase with the down-converted image components of the in-phase

. The outputs of each mixer contain

You might also like