This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Digital modulation techniques for mobile and personal
communication systems This paper describes in detail various digital modulation
techniques for mobile and personal communication
systems. Among others, these include 77." /4shift quadrature
phaseshift keying (QPSK), used in the second generation
digital cellular mobile systems in North America and
Japan, and Gaussian minimumshift keying (GMSK),
employed in the GSM system in Europe. It then briefly
discusses the current research activities in modulation
schemes for future systems. by A. H. Aghvami 1 Introduction 0 constant or near constant
The choice of modulation envelope technique has a direct impact on 0 low cost and ease of the capacity of a digital cellular implementation, mobile (DCM) communication
system. It determines the
bandwidth efficiency of a single To optimise all these features at the
same time is not possible as each physical channel in terms of the has its practical limitation and also number ofbits per second per is related to the others. For hertz (bil/s/Hz) and it is therefore example, to achieve high
important that this choice is bandw1dth efficiency one may
discussed in detail. choose to use hlghlevel modulation. However, if this is
done two consequent
disadvantages are introduced.
Firstly, the power efficiency of the
system is reduced. Secondly, the In selecting a suitable
modulation scheme for a DCM
system, consideration must be
given to achieving the following: 0 hggh bandWIdth éfﬁcwncy bandlimited highlevel modulated . hlgh power efflCIency signal has a large envelope 0 low carrierto»cochannel variation which, when the signal is
interference power ratio (CCI) passed through a power efficient 0 low outof—band radiation nonlinear amplifier, generates 0 low sensitivity to multipath large out—ofband radiation; this, in
fading turn, introduces interference to inphase branch U/B
converter incoming
bit stream U/B
converter quadrature branch 1 QPSK modulator ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL JUNE 1993 adjacent channels. The latter
disadvantage can be circumvented
by using linear power amplifiers
but these have poor power
efficiency. Hence, a tradeoff
among all the above features must
be adopted. In this paper, the power efficient
QPSK, OKQPSK, Ir/4shift QPSK
and GMSK modulation techniques
are first described and then the
modulation techniques for future
DCM systems are discussed. 2 Quadrature phase shift keying
(QPSK) systems
Quadrature PSK (also referred to
as 4ary PSK or 4phase PSK)
modems (modulator/
demodulators) are extensively
employed in operational satellite
communication systems. The block
diagram of a conventional QPSK
modulator is shown in Fig. l. The input unipolar binary
stream at a bit rate of fh is
converted into two bit streams
(inphase and quadrature streams),
each having a bit rate of j: =fb/2
(the symbol rate) by a serialto
parallel (S/P) converter. The two
unipolartobipolar (U/B)
converters convert these two
streams into two bipolar (i 1)
binary signals, which are passed
through two spectrally shaped
lowpass filters and then modulated
by the inphase and quadrature
carriers. The modulation operation
uses the doublesideband
suppressed—carrier amplitude bandlimited
QPSK signal 125 incoming
bit stream inphase
bit stream quadrature
bit stream 2 Operation of serialtoparallel converter modulation (DSSSC—AM)
technique. The two modulated
signals (each of which can be
considered as a binary PSK signal)
are combined to give a QPSK
signal. Finally, the QPSK signal is
filtered at the output of the
modulator to limit further its
power spectrum; this prevents
spillover into adjacent channels
and also removes out—ofband
spurious signals caused by the
modulation operations. It should
be noted that the spectrally shaped
filters can also be implemented as
a bandpass filter at the output of
the modulator. In this case, the
lowpass filters shown in the block
diagram are not required. quadrature T Fig. 2 illustrates the operation of
the S/P converter. In QPSK systems, the modulated
signal has four distinct phases
(Ir/4, 371/4, 57r/4 and 77r/4), each
representing one symbol. Each
symbol contains 2 bits of
information, The mapping of the
bits into symbols is frequently done
in accordance with the Gray code.
This code ensures that a single
symbol in error corresponds to a
single bit in error. The
constellation (signal—space
diagram) of a Gray—coded
unfiltered QPSK signal is shown in
Fig. 3. This Figure shows that the
carrier phase during any symbol
interval can be any one of the four 3 Constellation of Graycoded QPSK signal 126 phases 7r/4, 37r/4, 5717/4 and 77r/4
depending on the values of the
inphase and quadrature
components during that interval. Fig. 4 gives the block diagram of
a coherent QPSK demodulator.
The input bandpass filter removes
the outof—band noise and adjacent
channel interference. The signal at
the output of the filter is split into
two parts, each part being
coherently demodulated with the
inphase and quadrature carriers.
The two outputs are lowpass
filtered and applied to one—bit
analoguetodigital (A/D)
converters to regenerate the
inphase and quadrature baseband
signals. These two streams are
finally recombined in a parallel—to
serial (P/S) converter to give the
original bit stream. The carrier
recovery circuit provides inphase
and quadrature carriers which are
in synchronism with the received
unmodulated signal, i.e. which
have the same frequency and
phase as the unmodulated inphase
and quadrature carriers. The
symbol timing recovery circuit
provides the timing required to
operate the two A/D converters,
which must be in synchronism
with the received baseband
signals. Differential encoding/decoding for
QPSK Most practical carrier recovery
circuits introduce a phase
ambiguity into the recovered
carrier. In the case of QPSK, a
four—phase ambiguity may occur,
causing a considerable bit error
rate.I To remove this phase
ambiguity, a differential encoder
may be employed in the modulator
and a differential decoder in the
demodulator. Differential encoding, ensures
that the Changes in the phases of
the transmitted carrier represent
the information bits. Table 1 shows
the phase shifts (phase advances)
required for a QPSK signal to
transmit all possible pairs of the
information bits. The encoder can be
implemented using logic circuits
or lookup tables. The operation of
a QPSK differential encoder is
given in Table 2, where ak and bk Table 1 Phase shifts in
differential encoding Information bits Phase shift
0 O O
O 1 71/2
1 1 7c
1 O 371/2 ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL JUNE 1993 carrier
recovery
circuit received
QPSK signal 4 Coherent QPSK demodulator are the kth uncoded inphase and
quadrature symbols, respectively,
at the output of the S/P converter,
Ik and Qt represent the kth
differentially encoded symbols,
and [H and Q,“l are the kth
differentially encoded symbols
delayed by the symbol duration
(Ts = 2Tb)~ In the case of differentially
encoded QPSK (DEQPSK), the
decoding operation is the inverse
function of the encoding. For
differential QPSK (DQPSK), the
decoding operation is performed
on the received modulated signal,
thus avoiding the use of a carrier
recovery circuit. Fig. 5 shows a
differential QPSK demodulator
block diagram. Spectrum and spectral efficiency of
QPSK signals The power spectral density of an
unfiltered QPSK is of the form: _ sin 27r(f—fc)Th 2
3‘“ ‘ 4C“ i 276— ffirfi “) onebit A/D converter recovery symbol I timing P/S
converter recovered
bit stream oneebit circuit I A/ D converter where C is the average signal
power normalised across a 1 Q
resistor and Th = l/fb is the bit
duration. Table 2 Differential encoder operation for QPSK “k bk Ik—l Qk—I [k Qk
0 0 0 0 0 O
0 0 0 1 0 1
0 0 l l l 1
0 0 1 0 1 0
0 1 0 O l 0
0 1 0 1 0 0
0 I l l O l
0 1 1 0 1 l
l 1 0 O 1 l
l 1 0 l l 0
1 l l l O 0
1 1 l O O 1
1 O O 0 0 l
1 0 0 1 l 1
l 0 l 1 l 0
I 0 1 0 O 0 Assuming that the modulator
employs a spectrally shaped filter
having a square root of a raised
cosine function with a rolloff
factor of a (for optimum
performance), the filtered
spectrum ofa QPSK signal can
easily be obtained, and is as shown
in Fig. 6. The spectral efficiency of a
modulation scheme is defined as
the ratio of the bit rate to the
bandwidth, ﬁ/B, expressed in
bit/s/Hz. In the case onPSK, the
bandwidth of the signal is (l + 00f“
as in Fig. 6. Hence, its spectral
efficiency is: fr )2 Tiorsx=E= (1+ mfs = since ﬂ = ﬂ/Z. For the minimum bandwidth
case (a = 0), the theoretical
spectral efficiency of QPSK
systems is 2 bit/s/IIz. It is now
possible to realise practical filters
with a rolloff factor as low as 27
1+0: (2) a — 02 and hence a spectral onevbit A/D converter P/S
converter one—bit
A/D
converter 5 Block diagram of QPSK differential demodulator ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL JUNE 1993 127 frequency 6 Power spectral densities of unﬁltered and ﬁltered QPSK signal: (a) unﬁltered QPSK
spectrum; (1;) amplitude response of raisedcosine function with rolloff factor (x and
amplitude equaliser; (c) ﬁltered QPSK spectrum nonlinear m
‘D
2‘
(I)
C
O)
‘c
n:
..
.s
Q
<1>
a.
(II
._
a:
3
o
a linear 7 Spectrum spreading of bandlimited QPSK signal at output of TWI‘A operating at
saturation4 efficiency of 17 bit/s/Hz can be
achieved in practice. Error probability performance of
QPSK systems The performance of a QPSK
system in the presence of system
impairments may be measured by
its symbol error probability
(symbol error rate) or bit error
probability (bit error rate). The
symbol error probability of a
system is defined as the average
probability of signalling elements
of the received signal being in
error (phase error in the case of
PSK systems). The bit error
probability of a system is defined
as the average probability of the
received bits being in error. In the presence of additive white
Gaussian noise (AWGN) only and
for transmission free of
intersymbol interference (181), the
expression for the symbol error
probability of a QPSK system is: P5 = erfc [‘5 (%) J % (3) where C/N is the carriertonoise
power ratio (CNR) and erfc() is
the complementary error function
given by: erfc(z) = 1 — erf(z) =fzje‘2dx ,
A. In practical measurements, it is
convenient to measure the CN R.
However, the bit energy (E b) to
noise power spectral density (N0)
ratio (E ,/N0) is a very useful
parameter when various
modulation schemes are
compared. Assuming that the
filtering strategy of the system
satisfies the Nyquist criterion, 3
general relation between these two
parameters for Mary modulation
schemes is given by: Q _ 1i
(N) w (M)
In the case onPSK, M = 4 and C _ a (r) 2 (Na) <5) By putting eqn.5 into eqn.3 the
symbol error rate of a QPSK
system as a function of E b/No is
obtained as: PE = erfc (7)15 (6) Relationship oflhe bit error rate, PB,
and symbol error rate, PE For Graycoded Mary
modulation schemes the
relationship between these two
figures of merit can be easily 128 ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL JUNE 1993 unipolar
to bipolar
converter S/P Converter unipolar
to bipolar
converter derived as: 1
p = _ P 7
e logzM E ( ) In the case of the QPSK system we
have: N (8) o a
PE = app = %erfc (Eh) The BER performance of QPSK
signals through nonlinear
channels depends on the type of
filtering and the amount of power
amplification. A detailed treatment
of the subject is given in Reference 2. 3 Offsetkeyed QPSK
(OKQPSK)
Most digital radio transmission
systems today operate their main
high power amplifiers (HPAs) near
to or at saturation for maximum
power efficiency. For this
operating mode, the HPA
introduces nonlinear amplitude
and phase distortions. One of the
damaging effects of these
nonlinearities is the spectral
spreading of the transmitted
signal, which increases
undesirable interference to the
adjacent channels (adjacent
channel interference). Because of their large phase
transitions (i 90°, i 180°) QPSK
signals, when bandlimited, have
large envelope variations (up to
100%). As a result, when a
bandlimited QPSK signal passes
through a nonlinear amplifier
operating at saturation. there is
almost a total regeneration of the
filtered sidelobes to their unfiltered
levels (nearly 100% spectral
spreading). Fig. 7 illustrates the
spreading of a bandlimited QPSK oscillator r  1
: sinusoidal 8 OKQPSK and MSK modulator ( for MSK only) quadrature inphase 9 Signal constellation of 7r/4QPSK
system signal by a nonlinear amplifier
operating at saturation. The
channel rolloff is assumed to be
40% (a = 04).3 For nonlinear systems,
alternative modulation schemes
with smaller phase transitions
have been proposed. OKQPSK
signals have i 90° phase
transitions. As a result,
bandlimited OKQPSK signals
have a 33% envelope variation and
hence. when they are transmitted
through a nonlinear HPA, there is input bit
stream mapping S/P
converter 10 7r/4shift QPSK modulator ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL JUNE 1993 bandlimited
signal less spectral spreading than in the
case of a bandlimited QPSK
signal.4 A block diagram of an OKQPSK
modulator is shown in Fig. 8. The
modulator is identical to the QPSK
one except that the quadrature
stream is delayed with respect to
that in the inphase branch by one
bit interval. The OKQPSK
demodulator is also identical to the
QPSK one but with the inphase
branch shifted with respect to the
quadrature stream by T 1, seconds
just before the P/S converter (see
Fig. 4). For linear systems, the bit error
rate and power spectral density of
the OKQPSK system are the same
as those of QPSK systems. This modulation scheme has
been proposed for aeronautical
and land mobile satellite
communication systems. 4 7r/4shift QPSK ﬂ/4shift QPSK was first
introduced by Baker in 1962.5 It is
a compromise between QPSK and
offsetkeyed QPSK (OKQPSK) in
the sense that it has a maximum
phase change of 135° compared to bandlimited
Ir/4»shift
QPSK 129 11 Baseband differential detector 1800 for QPSK and 90° for OK
QPSK. Hence, bandlimited
7r/4—shift QPSK has less envelope
ﬂuctuation than bandlimited
QPSK but more than that of OK
QPSK. The main advantage of
7r/4shift QPSK is that it can be
noncoherently detected (by a
differential detector or by an FM
discriminator), which is not the
case for OK—QPSK. Also, in the
presence of multipath fading, it
outperforms OKQPSK. For these
reasons. it has been selected for the
Japanese and American second
generation cellular digital mobile
radio systems. In a n/4—QPSK modulator,
signalling elements of the
modulated signal are selected in
turn from two QPSK constellations
which are shifted by 71/4 with 12 IF differential detector 130 respect to each other, as shown in
Fig. 9. The modulator block diagram is
shown in Fig. 10. The input bit
stream is converted by a serialto
parallcl (S/P) converter into two
parallel streams (at, bk). each with
a symbol rate equal to half that of
the incoming bit rate. The kth
inphase and quadrature pulses (It
and Q) at the output of the signal mapping circuit are determined by their previous pulse levels, 1H and
QM, and the input symbols ak and
bk as given by [k : IHcos 0k — OH sin 9;.
Q = 1k_lsin 6k + QM cos 9k ak and bk, in turn, are related to the
phase shift changes of the
modulated signal according to converter converter P/S Table 3 Phase shifts as a
function of information bits for
7r/4shift QPSK Information bits Phase shift
0 0 It /4
0 1 3 7r/4
1 1 5 75/4
I 0 717/4 Table 3. The rest of the modulator
is identical to that of a QPSK
modulator. To demodulate the 7r/4shift
QPSK. we may use one of the
following differential detection
methods: (a) Basebanddifferential
detection: In this method, the
differential decoding is
performed on the recovered
inphase and quadrature
baseband signals as shown in
Fig. 11. It requires local
oscillators but, since the
phase error is removed by
differential detection at
baseband. phase coherence is
not needed. (17) IF differential detection: Fig.
12 shows the block diagram
of an 1F differential detector
for ﬂ/4shift QPSK.
Differential decoding is
performed on the received IF
signal using a delay line and
two mixers and hence no
local oscillator is needed in
the receiver. (c) Limiter FM discriminator
detection: The block diagram ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL JUNE 1993 l3 Limiter FM discrimator detector of an FM discriminator
detector is shown in Fig. 13.
The FM discriminator
extracts the instantaneous
frequency deviation of the
received signal. The integrate
and—dump circuit integrates
the frequency deviation over
each symbol duration, the
integral being the phase
difference between two
sampling instants. Finally, the
4level threshold comparator
detects the output phase
difference. It has been shown that these
schemes have practically the same
performance in the presence of
cochannel interference and
Gaussian noise if the circuit
elements are properly designed" The modulation adopted for the
American digital cellular mobile
(DCM) communication system
(1354) is 7r/4shift QPSK with a
roll—off factor of (1:035. The burst
transmit signal with a bit rate of
486 kbit/s through a 30 kHz
channel bandwidth gives a spectral
efficiency of 162 bit/s/HZ for this
system. For the Japanese DCM
system a 7r/4shift QPSK
modulation with a roll—off factor of
(1:05 is used. The burst
transmission bit rate and the
physical channel bandwidth are
42 kbit/s and 25 kHz, respectively.
This gives a bandwidth efficiency
of 168 bit/S/Hz. 5 Gaussian minimum shift
keying (GMSK) Minimum shift keying (MSK) is a special case of binary continuous phase frequency shift keying (FSK) with a modulation index of 05 or with a frequency deviation of f2 — f1 = 1
2 4T,,
where f2 and f1 are two discrete frequencies of the MSK signal
representing logic states 1 and 0, Af= integrate
and
dump FM discriminator respectively, and Tb is the bit
duration. The MSK modulator can be
implemented using a voltage
controlled oscillator (VCO)l or a
quadrature form similar to that of
OK—QPSK. In the latter case, the
rectangular shaped pulses before
the two multipliers are converted to
halfsinusoidal waveforms by two
sinusoidal pulse shaped filters as
shown in Fig. 8. The demodulator
may also use coherent detection
similar to that of OK—QPSK. For linear systems, the bit error
rates of QPSK, OKQPSK and
coherent MSK are identical. An
unfiltered MSK signal has a
constant envelope and also a faster
spectral rolloff than QPSK or
OKQPSK but its mainlobe OPSK/OKOPSK I P/S  converter bandwidth is 15 times wider (see
Fig. 14). For higher bandwidth efficiency
and sharper cutoff, an MSK signal
should be filtered before
transmission. This introduces
envelope variations of the
bandlimited signal and as a result,
when it is passed through a
nonlinear highpower amplifier, its
filtered power spectral density
spreads. This may cause
undesirable interference to the
adjacent channels (adjacent
channel interference). However, if
a Gaussianshaped filter is used for
filtering the transmitted signal, the
constant envelope property of the
signal can be kept, since a
premodulation lowpass filter (LPF)
with Gaussianshaped transfer fc+2f5 3
fc+2f5 frequency 14 Unﬁltered power spectral densities of QPSK, OKQPSK and MSK signals ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL JUNE 1993 131 15 GMSK modulator 16 I’LLTYPE GMSK modulator function is easily realised. A simple
method for the modulator is to use
a VCO directly with a
premodulation Gaussianshaped
LPF as shown in Fig. 15. The
impulse response of the ﬁlter is
given by: h(t) = e‘lZ/ZUZTZ/O'TWZW) where 0' = \l(ln 2)/27rBT andB is
the 3 dB bandwidth of the filter
and T the bit duration. However, it is difficult to
maintain the centre frequency
accurately in an acceptable range.
To overcome this drawback, a
PLLtype GMSK modulator can be
used. The block diagram of this
modulator is shown in Fig. 16.7
The coherent demodulation
t...
View
Full Document
 Fall '08
 Goldsmith
 Communications

Click to edit the document details