Wireless communications — the fundamentals - Technion

illnurturedtownvilleΚινητά – Ασύρματες Τεχνολογίες

21 Νοε 2013 (πριν από 3 χρόνια και 8 μήνες)

108 εμφανίσεις

BT Technology Journal • Vol 25 No 2 • April 2007
11
Wireless communications — the
fundamentals
T G Hodgkinson
The primary aim of this paper is to provide an overview of wireless communication fundamentals, and the approach used
considers them within the context of the following four categories — radio propagation, wireless air interface, advanced
antenna systems and interference effects. In addition to this, a representative set of mobile systems are compared to show
that their differences are primarily due to them having different combinations of channel transport, modulation scheme and
regulatory constraints on transmit power, channel bandwidth, operating frequency and channel duplex.
1.Introduction
Wireless communication systems are not new, but they have
been continually evolving for many years, especially in the
area of mobile communications. First generation analogue
mobile systems began to emerge in the late 1970s and in the
early 1980s work began on what was to become the second
generation digital GSM system. In the early 1990s migration
to the second generation system started to gain momentum
and within two years all of the major European operators
had started to operate commercial GSM networks. During
the mid-1990s, ground work preparations started that
would eventually lead to the development of third
generation systems and the first commercial network was
launched in early 2000. More recently, mobile WiMAX has
begun to emerge and there is a growing interest in its
potential as an alternative to the other mobile networks and
their planned migration paths.
When third generation systems began to emerge this
coincided with the appearance of the first Wi-Fi systems,
which have the key difference of having been designed
primarily for indoor operation in licence-exempt spectrum
rather than outdoor operation in licensed spectrum.
However, it was not until the IEEE802.11a, b standards were
ratified in the late 1990s that Wi-Fi began to be widely
adopted. Since then these systems have continued to evolve
and grow in popularity.
Comparing the evolution paths of the various mobile
systems it will be found that the trend has largely been one
of migrating from analogue to digital, and from kbit/s
operating speeds for first generation systems through to the
low Mbit/s rates for third generation systems, with migration
to 100s of Mbit/s being on their fourth generation evolution
paths. For Wi-Fi, the trend has been a migration from low
Mbit/s rates to low 100s of Mbit/s. However, the advances
that have been made in both mobile and Wi-Fi systems to
date largely reflect technological advances that have
enabled increasingly advanced fundamental wireless
communication concepts to become practically viable.
Therefore, these advances are, in principle, generally
applicable for use in any type of wireless system.
For this reason, the aim of this paper is to provide an
overview of the wireless fundamentals, which will be
approached within the context of the block diagram shown
in Fig 1 using the following four categories:

radio propagation,

wireless air interface,

advanced antenna systems,

interference effects.
A high-level consideration of a representative set of
current mobile and Wi-Fi systems will be then used to argue
that their performance differences are largely a reflection of
different regulatory constraints and the fundamental limits
set by thermal noise, advanced modulation schemes and the
channel transport technique used.
2.Radio frequency propagation
All wireless systems, whether they use licensed or licence-
exempt spectrum, operate in exactly the same physical
environment and hence they are all susceptible to the same
fundamental propagation characteristics. All of these will
affect performance in some way, but the ones that have the
most significant impact are:
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
12

propagation path loss,

received signal fading,

uplink/downlink channel duplex,

wideband channel transport techniques.
Each of these aspects will now be considered in more
detail.
2.1 Propagation path loss
At the most basic level of consideration, propagation loss is
proportional to both the frequency used and the distance
travelled. Its dependence on frequency is shown by the plot
1
in Fig 2 for the 1 to 6 GHz frequency band expressed as an
increase on the loss at 1 GHz. The main point to note is that
over this 5 GHz frequency range the path loss difference is
almost 15 dB.
The loss with distance is equal to the distance travelled
raised to the power of the propagation coefficient, which for
free space propagation is equal to 2. In practice, however,
empirical path loss models are derived by curve fitting to
measurements and the propagation coefficients for these
models typically lie in the range 2.5 to 4 for urban macro-
cells and 2 to 8 for micro-cells. The extent to which this
increases relative path loss with respect to that of free-space
propagation can be appreciated from the plots in Fig 3.
These clearly show that, as the propagation coefficient
increases, the relative difference in path loss becomes
increasingly significant with distance. For example, doubling
the propagation coefficient gives a loss difference of 20 dB
at 10 m, but at 1 km this increases to 60 dB. Consequently,
practical operating ranges are significantly smaller than
would be achieved with ideal free-space propagation.
Fig 3 Difference between empirical model calculated propagation
path loss and ideal free-space propagation (α = 2) for a range of
propagation coefficients (α).
modulation
transmitter
coding
antenna array
demodulation
decoding
receiver
antenna array
transmit
power
received
signal power
RF interference
input
data stream
recovered
data stream
RF propagation
channel
wireless air interface
wireless air interface
thermal noise
Fig 1 Block diagram of a wireless communication system.
1
The results plotted were derived using the simple propagation loss model
L = 100αlog(d) + 20log(f
rf
) −147.6, where L is the path loss in dB at a
distance d from the transmitter, α is the propagation coefficient and f
rf
is
the radio frequency used. The final term, including its sign, is the numerical
value of 20log(4π/c), where c is the speed of light. Note that all range-
related calculations in this paper have been derived using this model.
Fig 2 Increase in propagation path loss with respect to the 1 GHz

value as a function of operating frequency.
16
path loss increase, dB
12
8
4
0
0 2 4 6
frequency, GHz
60
relative path loss, dB
40
20
0
1 10 100 1000
distance, m
α = 4
3.5
3
2.5
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
13
2.2 Received signal fading
Unfortunately, there is a drawback to the use of empirical
propagation models for calculating path loss, which is that
for a given distance from the transmitter the predicted loss is
the same for all directions of propagation. This is an issue
because in practice different propagation directions
experience different levels of clutter (buildings, trees, etc),
so their path losses differ over a given distance. This causes a
statistical variation in path loss, which is known as
shadowing or slow fading. Therefore, an operating margin is
needed to ensure that the probability of falling below the
minimum acceptable received signal-to-noise ratio is
acceptably small. The shadowing fade margin is plotted in
Fig 4 as a function of this probability, which is known as the
shadowing fade outage probability, for a range of location
variability
2

LV
) values typical of a macro-cell.
Fig 4 Shadowing fade margin versus percentage outage probability
for a range of location variability (σ
LV
) values.
There is also another form of fading to take account of
which is known as multipath or fast fading. This type of
fading occurs when the antenna detecting the main radio
signal also receives reflected versions of itself from obstruc-
tions such as buildings, hills, etc (see Fig 5). Because these
multipath signals are time delayed with respect to each
other, and possibly also frequency spread due to the Doppler
effect [1], the resulting constructive/destructive interfer-
ence causes the received signal power to fluctuate with time.
The statistical nature of these fluctuations is characterised by
whether or not the propagation path is line-of-sight (Ricean
fading) or non-line-of-sight (Rayleigh fading) and also by
the value of the channel delay spread, which is a parameter
that depends upon the particular propagation environment.
Table 1 shows typical delay spread values for a range of
different propagation environments.
To counteract the performance variation caused by the
statistical nature of multipath fading, which depends upon
the number of reflections received and their relative powers
and phase relationships, an operating margin is needed to
ensure that the probability of falling below the minimum
acceptable signal-to-noise ratio threshold is acceptably
small. This fade margin is equal to the difference between
the threshold signal-to-noise ratio and the mean value
needed to achieve a particular multipath fade outage
probability. Figure 6 shows the relationship between
multipath fade margin and outage probability for Ricean
(solid lines) and Rayleigh (dashed line) flat
3
fading channels.
The factor associated with the solid lines, which is known as
the Ricean K factor
4
, represents the ratio of line-of-sight
power to total multipath signal power.
The plots in Fig 6 show that for non-line-of-sight
conditions and relatively low values of outage probability
there is significant benefit to be gained from using fade
mitigation techniques to reduce the amount of margin
needed, whereas for line-of-sight the benefit diminishes as
the main signal becomes increasingly dominant over the
multipath signal power. In principle, fading can be mitigated
by using time, frequency, space, code, angle-of-arrival and
polarisation diversity techniques — section 4 considers
some of these in more detail.
50
shadowing fade margin, dB
40
30
20
0
0.01 0.1 1 10
outage probability, %
10
σ
LV
= 12
8
5
Table 1 Typical delay spread values for a range of different
operating environments.
Environment Typical delay spread (μs)
Indoor micro-cell 0.001-0.05
Open area < 0.2
Suburban macro-cell < 1
Urban macro-cell 1-3
multipath
reflection
line-of-sight
non-line-of-sight
Fig 5 Line-of-sight and non-line-of-sight propagation
conditions.
2
Location variability is the name used for the standard deviation of the
random path loss variations caused by shadowing.
3
Flat fading occurs when the spectral components of the modulated carrier
are affected equally, whereas frequency selective fading occurs when they
are affected unequally.
4
If K is set to zero in the Ricean distribution the resulting expression is
identical to the Rayleigh distribution.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
14
2.3 Uplink/downlink channel duplex
In practice, the propagation characteristics are not
necessarily the same for both the uplink and the downlink
because the radio spectrum allocation process for licensed
spectrum specifies which one of two modes of operation is
to be used — these are known as frequency division duplex
(FDD) and time division duplex (TDD). The difference
between them is that the former uses different frequencies
for the uplink and the downlink, whereas the latter shares a
single frequency. The advantages of time division duplex are
that the same channel propagation characteristics apply for
both the uplink and the downlink and their bandwidth ratio
can be dynamically adjusted to match changing operating
conditions. Frequency division duplex, however, has the
advantages of providing better isolation between the uplink
and downlink and in some instances potentially has a greater
operating range due to using a narrower bandwidth receiver
than a TDD system (see section 6 for more details).
2.4 Wideband channel transport techniques
If channel bandwidth is increased to enable higher data-
rates to be used, a point is eventually reached where the
inter-symbol interference
5
caused by the time offsets
between the main signal and its multipath reflections cannot
be ignored, and the point at which this occurs depends upon
the channel symbol
6
rate and the delay spread
characteristics of the propagation environment. For symbol
rates that are much smaller than the reciprocal of the delay
spread, inter-symbol interference becomes negligible and
the fading is flat. For the converse situation, inter-symbol
interference cannot be ignored and the fading is frequency
selective. Consequently, increasing channel bandwidth
causes single carrier systems to become increasingly more
susceptible to inter-symbol interference and frequency
selective fading unless equalisation techniques are used to
counteract them.
Several channel transport techniques have emerged for
use in cases where the channel bandwidth exceeds the value
beyond which inter-symbol interference and frequency
selective multipath fading effects cannot be ignored.
Specific examples are direct sequence spread spectrum
(DSSS), orthogonal frequency division multiplexing (OFDM),
single carrier, frequency domain equalisation (SC/FDE) and
orthogonal frequency code division multiplexing (OFCDM).
2.4.1 Direct sequence spread spectrum
This is a single carrier technique that frequency spreads the
baseband modulation signal so that it fills the whole of the
channel bandwidth. This is achieved using frequency
spreading codes that are ideally designed to have a shorter
correlation time than the various multipath delay
differences. Provided this requirement is satisfied, the
individual multipath signals can be separated out if an
appropriate receiver is used (e.g. RAKE receiver [1]), thus
enabling the individual multipath signals to be extracted and
combined in a way that improves overall performance by
effectively removing the inter-symbol interference.
2.4.2 Orthogonal frequency division multiplexing
Unlike DSSS, this is a multi-carrier technique that transports
data over the channel by distributing it across an underlying
frequency multiplex. The basic principle involved is that the
larger the number of carriers (tones) used, the lower the
data rate they each have to carry. So, by using an
appropriate number of carriers the symbol duration can be
made to be significantly longer than the multipath delay
differences, and hence reduce the level of inter-symbol
interference and also create channels that have flat fade
characteristics. Furthermore, the inter-symbol interference
can be completely eliminated by the introduction of a cyclic
prefix [2] provided its duration is longer than the channel
delay spread. However, this is at the expense of increased
channel overhead.
2.4.3 Single carrier, frequency domain equalisation
Like DSSS, this is a single carrier transport technique that
uses frequency domain equalisation to overcome the
impracticality of using time domain equalisation to
counteract multipath fading in wide bandwidth channels.
Overall, SC/FDE, which also has a media access variant
known as SC-FDMA
7
, gives similar performance to OFDM for
essentially the same overall level of complexity, but because
it is a single carrier transport technique it has the advantage
of having a lower peak-to-average power ratio than OFDM,
which has certain cost advantages.
40
fade margin, dB
30
20
10
0
0.01 0.1 1 10
outage probability, %
k = 0 (Rayleigh)
2
5
10
30
5
Inter-symbol interference is when a data bit overlaps a time-slot
containing a different data bit.
6
Data bits transmitted over the RF channel are conventionally referred to as
symbols to distinguish them from input data bits. This is because a symbol
usually represents several input data bits (see section 3.2).
7
SC-FDMA is an extension of single carrier, frequency domain equalisation
(SC/FDE) that accommodates multiple-user access.
Fig 6 Multipath fade margin versus percentage outage probability
for a range of Ricean K factors.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
15
2.4.4 Orthogonal frequency code division multi-
plexing
This is a transport technique that adaptively exploits both
frequency and time domain spreading to maximise
performance according to the particular operating
environment. Essentially, it adaptively exploits OFDM to
provide frequency spreading and DSSS to provide time
spreading, and current claims are that it will improve on the
performance of current OFDM systems.
3.Wireless air interface
The performance of the wireless layer can be broken down
into five relatively independent component parts, which are:

receiver thermal noise,

channel modulation schemes,

channel coding techniques,

scheduling and adaptive modulation coding,

media access control.
The individual impact that these components have on
performance will now be described in more detail.
3.1 Receiver thermal noise
The performance of an ideal communication system is
limited only by the presence of thermal noise
8
at the
receiver, which is a natural phenomenon that cannot be
eliminated. Unfortunately, thermal noise power increases
with receiver bandwidth, so any increase in the latter will
require an increase in received signal power to maintain a
constant value of signal-to-noise ratio, and hence error
probability. Consequently, the maximum tolerable path loss
decreases with increasing receiver bandwidth. The increase
in noise power with receiver bandwidth is illustrated by the
plot in Fig 7 over a bandwidth range typical of wireless
systems that operate in the 1 to 6 GHz frequency band
9
. The
main point to note is that, for this range of receiver
bandwidths, the difference in maximum tolerable path loss is
30 dB.
The noise performance of practical receivers is usually
worse than the thermal noise limit due to receiver losses and
the additional noise introduced by electronic amplification.
This causes a further reduction in the maximum tolerable
path loss and by an amount equal to the ratio of actual noise
power to thermal noise power, which is known as the
receiver noise figure. Ignoring specialist receivers, such as
those that use cryogenic cooling, wireless system receivers
typically have noise figures in the range 1 to 10 dB.
3.2 Channel modulation schemes
The emphasis so far has been on the fundamental factors
affecting transmission distance. But the potential benefits of
using advanced modulation schemes as a means for
increasing the data rate transmitted over a fixed channel
bandwidth is also an important consideration. A point worth
noting is that although these modulation schemes enable
higher data-rate transmission to be achieved without any
increase in receiver bandwidth, and hence noise power, the
received signal power still has to be increased to maintain
the desired level of performance. The reason for this will
become clear later in this section.
For a given channel bandwidth, significant data-rate
increases can be achieved by using multi-level digital
amplitude and/or phase modulation schemes. These
schemes are known as M-ary amplitude shift keying (ASK),
M-ary phase shift keying (PSK) and M-ary quadrature
amplitude modulation (QAM), where M represents the
number of modulation levels, or symbols as they are also
known. However, QAM should not be thought of as being
distinct from the other two because it is effectively a
combination of M-ary ASK and 4-ary PSK; the latter is also
known as quadrature PSK (QPSK).
The fundamental principle behind multilevel modu-
lation is that a sequence comprising a predetermined
number of consecutive data bits is replaced by the
appropriate symbol from the symbol alphabet being used for
the coding process. This is illustrated in Fig 8 for a Grey code
mapping of two binary input bits into QPSK symbols; when
Grey code mapping is used, adjacent symbols are only
different by one binary bit.
The outcome of this is that the effective data rate
actually transmitted over the channel is equal to the symbol
rate multiplied by the number of bits-per-symbol.
Consequently, the improvement in effective data rate, or
spectral efficiency as it is also known, equals the number of
bits represented by a symbol.
40
increase in noise power, dB
30
20
10
0
0 10 20 30
increase in bandwidth, dB
Fig 7 Increase in thermal noise power as receiver
bandwidth is increased.
8
This is also referred to as additive white Gaussian noise (AWGN).
9
This includes some of the channel bandwidths being considered by the
WiMAX Forum.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
16
Fig 8 Grey code mapping of binary input pairs
into QPSK output symbols.
The black dots in Fig 9 show modulation constellations
for BPSK, 4-ary ASK, 8-ary PSK and 16-ary QAM and it is
clear that the M-ary schemes have a smaller symbol
separation distance than that of BPSK. This separation dis-
tance is known as the Euclidean distance, and the received
signal-to-noise ratio is proportional to this distance squared.
Therefore, it follows from Fig 9 that M-ary modulation
schemes will give worse performance than BPSK unless their
average received energy per bit is increased to the point where
their Euclidean distances become equal to that of BPSK.
The increase in energy per bit needed for the Euclidean
distance to become equal to that of BPSK is effectively the
performance penalty associated with using M-ary
modulation, and Fig 10 shows this penalty for the various
M-ary modulation schemes. Comparing the performance
penalty and spectral efficiency figures it will be seen that
with the exception of 4-ary PSK, any improvement in
spectral efficiency is always at the expense of having to
increase the received signal power, which effectively reduces
the maximum tolerable link loss. For example, increasing the
spectral efficiency by a factor of six reduces the maximum
tolerable link loss at worst by approximately 23 dB (64-ASK)
and at best by approximately 8 dB (64-QAM).
So far the emphasis has been on modulation schemes
that improve spectral efficiency, but if maximising the
transmission distance is more important M-ary frequency
shift keying (FSK) should be used. However, this is at the
expense of a significant reduction in spectral efficiency. For
example, 64-FSK improves on the BPSK signal-to-noise
ratio by approximately 4 dB, but this is at the expense of
requiring five times the channel bandwidth.
3.3 Channel coding techniques
With the exception of M-ary FSK, all of the aspects
considered so far are such that the desired error probability
can only be achieved by maintaining the received signal-to-
noise ratio constant. In the case of channel coding, however,
the fundamental aim is to enable the desired error
probability to be achieved at as low a received signal-to-
noise ratio as is practically possible. The difference between
the signal-to-noise ratios needed for uncoded and coded
transmission to operate with the same value of error
probability is known as the coding gain, and its actual value
depends upon the particular coding scheme used (e.g.
block, convolution, turbo, space-time) and its finer design
points (e.g. delay, number of states, constraint length). In
general, the nonlinear relationship between received signal-
to-noise ratio and probability of error results in coding gain
being a relatively complex function of the type of coding
used, especially when error correction is supported.
However, in general, coding gain is approximately equal to
binary
input pair
0, 0
0, 1
1, 1
1, 0
QPSK symbol
phase (rads)
π/4
3π/4
-3π/4
-π/4
Fig 9 Symbol constellation examples for BPSK and a range of M-ary modulation schemes showing their Euclidean distances in terms of the mean

energy per data bit (E
¯
b
). The date-rate improvement with respect to that of BPSK is also shown. The black dots represent the number of channel

modulation symbols used by each of the modulation schemes.
4E
b
1.6E
b
1.8E
b
1.6E
b
1× (ref)



BPSK
4-ASK
8-PSK
16-QAM
data-rate increase
increase in S/N, dB
30
20
10
0
4 8 16 64
M-ary
32
data-rate increase = log
2
(M-ary)
QAM
PSK
ASK
Fig 10 Signal-to-noise ratio increase relative to BPSK modulation

for a range of M-ary ASK, PSK and QAM modulations.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
17
the product of its code rate and the minimum Hamming
distance [3]; in its simplest form the latter is equal to the
smallest number of bit differences between the different
code words. Table 2 shows typical coding gains that can be
achieved when using convolution or turbo coding with
binary modulation schemes. Under some operating
conditions these coding schemes are also combined with
repetition coding, which simply involves sending the same
symbol multiple times, to give further performance
improvements but at the expense of reducing the effective
data rate. However, in practice, coding gain is bounded
because there is a minimum received signal-to-noise ratio,
known as the Shannon Limit [4], below which it is impossible
to design a coding scheme that can perform with an
arbitrarily small probability of error.
Table 2 Typical coding gains for convolution and turbo codes and
typical increase in bandwidth needed to achieve the gain.
The basic underlying principle for achieving coding gain,
irrespective of the particular coding technique used, is that a
predefined number of data bits are replaced by a code word
comprising a larger number of bits — the ratio of data bits to
code word bits is known as the coding rate. Unfortunately,
for binary modulated systems, coding rate increases the
channel bandwidth needed by an amount equal to the
reciprocal of the coding rate. However, with the appropriate
combination of coding and M-ary amplitude and/or phase
modulation, which is known as coded-modulation, the
improved spectral efficiency of multilevel modulation
compensates for the reduction caused by coding. For
example, using 8-ary PSK in conjunction with a 2/3 rate
code avoids the need for any increase in channel bandwidth.
However, M-ary modulation schemes require larger signal-
to-noise ratios than binary modulation, so for coded-
modulation the gain is smaller than can be achieved with
binary modulation, and typically by an amount similar to the
values given earlier in Fig 10.
3.4 Scheduling and adaptive modulation coding
Using dynamically adaptive control techniques with the
modulation and coding schemes previously outlined it is
possible to improve on the overall average link capacity that
would be achieved otherwise. This is known as adaptive
modulation coding and is achieved by using advanced link
adaptation control techniques to dynamically choose the
modulation scheme and coding rate that best matches the
instantaneous channel conditions. For example, higher order
modulation and coding rates, such as 64 QAM with a 5/6
rate code, are used when the signal quality is very good, but
as the signal deteriorates the modulation order and code
rate are reduced as necessary until the point is reached
where no further adjustments can be made. Adaptive
modulation coding combined with destination-based
scheduling, which selects the destination with the best
channel conditions at each moment in time, enables data to
be sent at rates that on average are higher than would have
normally been achieved. With the appropriate algorithm,
this type of scheduling can be used to avoid transmitting
over channels that are experiencing a significant level of
fade, and under certain operating conditions this can be
exploited to provide a form of diversity known as multiple
user diversity gain [5].
Ignoring any implementation and control overheads, the
ideal data rates for various combinations of the above
adaptive modulation and coding rates for channel
bandwidths in the range 1 to 20 MHz are given in Fig 11.
This shows that under ideal conditions, data rates in the
range of approximately 1 to 90 Mbit/s are achievable for a
bandwidth range of 1 to 20 MHz. In practice, due to link
layer overheads, these ideal values are reduced typically by
25% to 35%.
3.5 Media access control
Media access control is used to enable multiple users to
access the same channel, and in general there are three
distinct access categories — fixed assignment, random
access and demand assignment. For the fixed assignment
option user transmissions are controlled by the base-station.
Common examples of this option are time division multiple
access (TDMA), frequency division multiple access (FDMA),
code division multiple access (CDMA) and orthogonal
frequency domain multiple access (OFDMA). For the random
access option, users with relatively bursty traffic profiles
compete for access to the radio channel, so some form of
collision detection/avoidance control is necessary, which
introduces an additional overhead. A common example of
this option is carrier sense multiple access with collision
avoidance (CSMA/CA). For the demand assignment option,
users are assigned a channel on demand, which for bursty
traffic profiles can greatly improve the number of users that
a base-station can support. Dynamic TDMA and packet
reservation multiple access (PRMA) are examples of this
Example Coding gain (dB) Bandwidth increase (times)
Convolution code 4—9 2—3
Turbo coding 13 2
ideal data rate, Mbit/s
90
60
30
0
0 10 15 20
channel bandwidth, MHz
5
¾ 64-QAM
2
/
3
64-QAM
¾ 16-QAM
½ 64-QAM
½ 16-QAM
¾ QPSK
½ QPSK
Fig 11 Ideal data rates for various adaptive modulation and coding

combinations as a function of the available channel bandwidth.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
18
option. Typically, licensed spectrum systems use the fixed or
random assignment options, whereas licence-exempt
systems typically use the random access option, which has
the disadvantage of significantly reduced spectral efficiency
at higher traffic loads.
4.Advanced antenna systems
So far, although not explicitly stated, it has been assumed
that the wireless system comprises only a single antenna at
the transmitter and receiver. However, significant
performance benefit can be gained without any increase in
either transmit power or channel bandwidth by using
multiple antennas at the transmitter and/or receiver. There
are four possible antenna configurations — single-in single-
out (SISO), single-in multiple-out (SIMO), multiple-in
single-out (MISO) and multiple-in multiple-out (MIMO),
where ‘in’ refers to the transmit antennas and ‘out’ refers to
the receive antennas (see Fig 12).
In general, multiple antenna configurations offer a
means for increasing system coverage and/or capacity and
this improvement comes from one or more of the following
gains, which generally have to be traded off against each
other depending upon the particular operating conditions
and their different channel knowledge requirements:

array gain,

diversity gain,

spatial multiplexing gain,

interference suppression gain (this is considered later in
section 5) .
4.1 Array gain
Array gain is effectively a performance enhancement that an
antenna array can provide when propagation conditions are
such that there is very little multipath fading. In other words
there is a strong line-of-sight path and fading characteristics
are at worst Ricean with a high K value. Under such
conditions received signal powers can effectively be
considered to be deterministic rather than randomly varying.
This improved signal certainty significantly reduces the fade
margin needed (see Fig 6) and also enables an antenna array
to exploit constructive and destructive signal interference to
provide beam forming (see Fig 19). This requires that the
transmitted signals are identically modulated and that the
antennas are separated typically by half the wavelength of
the transmitted signal.
If the antenna array is used at the transmitter, as in the
MISO configuration, the direction of the transmission beam
is dynamically controlled by adjusting the relative
amplitudes and phases of the signals driving the antennas so
that the transmitted signals add constructively at the desired
destination antenna. When this is the case the resultant
received power will be greater than the sum of the individual
received signal powers. This increase in power is the array
gain provided by the transmit antenna array and for the
MISO configuration operated under conditions of constant
Fig 12 The four possible antenna configurations.
SISO
SIMO
MISO
MIMO
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
19
overall transmit power, the gain achieved is equal to the
number of transmit antennas used.
If the antenna array is used at the receiver instead of the
transmitter, as in the SIMO configuration, the direction of
the reception beam is dynamically controlled by adjusting
the relative amplitudes and phases of the different antenna
signals prior to combining them in the receiver. This
effectively focuses the reception beam on the source of the
transmission. With the appropriate delay settings these
signals combine coherently and, as in the case of transmit
array gain, the resultant received signal power is greater
than the sum of the individual signal powers from each
antenna. This increase in power is the gain of the receive
antenna array, and for the SIMO configuration, assuming the
received signal powers are equal, the gain is equal to the
number of receive antennas used.
It is clear from Fig 12 that the SIMO and MISO configur-
ations are effectively special cases of the MIMO config-
uration, and therefore, given that for most, if not all, practical
cases the beam width at the receiving antenna array will be
wider than the array width, the array gain for the MIMO
configuration is ideally the product of the transmit and
receive array gains. In other words it is equal to the product
of the number of transmit and receive antennas used.
4.2 Diversity gain
When operating conditions are such that there is no line-of-
sight path and the fading characteristics are predominantly
Rayleigh, which is typical of urban environments, the beam
forming needed for array gain deteriorates to the point
where most if not all of the array gain is lost. In addition to
this there is also a significant increase in the amount of fade
margin needed to ensure that the outage probability is
acceptably small (see Fig 6). Under these operating
conditions, the performance enhancement provided by the
various antenna configurations is that they reduce the
amount of fade margin needed rather than increasing the
effective received signal power as in the case of array gain.
This fade margin reduction is achieved using diversity, which
typically requires an antenna separation of at least the
coherence wavelength [4] of the channels to ensure that the
received signals have independent fading — this is a wider
spacing than used for array gain. Diversity gain can be
achieved using selection, switched equal-gain, or maximal
ratio combining techniques [1], and, although the last gives
the best performance, the spread of their respective
received signal-to-noise ratios needed for a given outage
probability is approximately only 2 dB.
Diversity reduces the amount of fade margin needed
because the diversity fade outage probability is equal to the
without-diversity value raised to the power of the diversity
gain, which is defined as being equal to the number of
independent signal paths between the transmit and receive
antennas. Figure 13 shows how the multipath fade margin
reduces with diversity gain for a range of outage
probabilities for the case of equal received signal powers and
Rayleigh fading channels. It is clear from these plots that
increasing diversity gain gives diminishing returns.
Consequently, the most significant reduction in fade margin
occurs at the smaller diversity gains and by an amount that
increases with decreasing outage probability. The reduction
in fade margin compared with that of a SISO antenna
configuration for identical operating conditions can be
considered to be the amount by which the received signal
power is effectively increased by diversity gain.
The SIMO, MISO and MIMO antenna configurations can
all provide diversity gain but they differ in how they achieve
it. The SIMO configuration exploits the space domain to
provide receive diversity and the gain achieved is equal to
the number of receive antennas used. The MISO
configuration, however, must exploit both space and time to
provide transmit diversity. This involves using space-time
coding techniques, but with appropriate coding it is possible
to achieve a transmit diversity gain that is equal to the
number of transmit antennas used. For the MIMO
configuration it is still necessary to use space-time coding at
the transmitter, but the overall diversity gain is now equal to
the product of the number of antennas used at the
transmitter and receiver.
4.3 Spatial multiplexing gain
A fundamental difference between the MIMO and SIMO/
MISO antenna configurations is that MIMO has several
independent pairings of transmit/receive antennas.
Consequently, multiple independent data streams can, in
principle, be simultaneously supported in the same
frequency channel, hence system capacity is increased.
Under ideal operating conditions this increase is equal to the
number of independent data paths, which has an upper limit
value that is always equal to the smaller of the number of
40
fade margin, dB
30
20
10
0
0 2 8 10
diversity gain, dB
4
6
P
outage
= 0.01%
0.1%
1%
10%
Fig 13 Fade margin as a function of diversity gain for a range of
outage probabilities (P
outage
).
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
20
transmit/receive antennas. This increase in capacity is known
as spatial multiplexing gain.
The key benefit of spatial multiplexing is that it is able to
increase the effective channel transmission rate by
exploiting random fading. Consequently, it requires a very
rich multipath environment, so it works best when there are
no direct line-of-sight paths between the transmit and
receive antennas — as is the case for diversity but the exact
opposite of the conditions for achieving array gain.
Unfortunately, spatial multiplexing gain has to be traded
against diversity gain [6]. This can be appreciated by taking
the view that spatial multiplexing and diversity have to share
the various paths available between the transmit and receive
antenna arrays, which means that as more paths are used for
spatial multiplexing there are less available for diversity use.
Therefore, the gain of one is increased at the expense of the
other, which means that a trade-off has to be made between
system capacity and the amount of fade margin. Their inter-
dependence is shown in Fig 14 for 2 × 2, 4 × 4 and 6 × 6
MIMO configurations.
Fig 14 Trade-off between spatial multiplexing and diversity gains
for a range of antenna array sizes.
5.Radio frequency interference
All wireless systems are susceptible to radio interference, but
if all extraneous sources are ignored it only leaves adjacent
channel and co-channel interference to consider, which are
interference sources that have to be traded against practical
design and spectral efficiency considerations. Both these
sources of interference become an issue when the coverage
provided by a transmitter is geographically close enough to
the coverage areas of other transmitters for their
transmissions to affect each other. The performance impact
of this inter-cell interference depends upon the relative
magnitudes of the mean received signal power of the
desired signal and the total power of the interfering signals
and whether or not their frequencies are different. In
addition, adjacent and co-channel intra-cell interference
can also occur depending upon which type of media access
control is used and whether or not the cell has been divided
into a number of different frequency sectors (see Fig 17).
The interference caused when using different
frequencies is known as adjacent channel interference. It is
caused by signal power ‘spilling-over’ into the adjacent
channels, and possibly ones beyond those, especially their
adjacent ones, which are known as the alternate channels.
This type of interference is caused by transmitter/receiver
imperfections associated with practical modulation
characteristics and the limitations of practical filters.
Generally, this type of interference reduces with increasing
channel separation and is usually managed by mandating
frequency guard bands and power spectral masks to which
practical transmitter/receiver designs must conform. The
performance margin needed to offset adjacent channel
interference is determined by the ratio of interfering power-
to-receiver thermal noise power
10
, as is shown in Fig 15.
The interference that is created when using the same
radio frequency is known as co-channel interference and its
impact on performance is a more complex process than for
interference caused by adjacent channels. This is because, in
addition to introducing the need for an interference margin,
which is derived in exactly the same way as for adjacent
channel interference, it also modifies the amount of fade
margin needed. This is because co-channel interference
causes the received signal power to fluctuate with time in
exactly the same way as multipath fading and for exactly the
same reason. Consequently, the combined impact of co-
channel interference and multipath fading on overall
performance is dependent upon their joint statistical
behaviour. The outcome of this dependence is a modified
relationship between outage probability and fade margin.
This can be appreciated by comparing the fade margin plots
in Fig 6 with the modified ones in Fig 16, which were derived
for Rayleigh (dashed line) and Ricean (solid lines) flat fading
channels in the presence of a single Rayleigh faded
interferer. Although both sets of plots still exhibit similar
trends, they show that the modified fade margin decreases
40
diversity gain, linear
30
20
10
0
0 2 4 6
spacial multiplexing gain, linear
array size 6 × 6
4 × 4
2 × 2
20
interference margin, dB
15
10
5
0
-10 0 10 20
interference to noise power ratio, dB
-5
5
15
Fig 15 Adjacent channel interference margin as a function of the

ratio of interfering power-to-receiver thermal noise power.
10
This assumes the coarse approximation of treating interference power as
if it is an additional thermal noise source within the receiver.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
21
less rapidly for higher K values and that there is no
difference between the plots for the Rayleigh faded channel.
From an intra-cell interference perspective, the type of
media access control used to enable multiple users to
simultaneously use the transmission medium is another
potential source of either adjacent or co-channel
interference. For example, when using frequency division
multiple access, adjacent channel interference can occur
between the multiplexed sub-channels. On the other hand,
when using code division multiple access, co-channel
interference can occur due to the noise-rise effect [7], which
is caused by the loss of spreading code orthogonally and
requires the use of transmit power control to ensure its
impact on performance is kept within acceptable limits.
In principle, inter-cell co-channel interference could be
avoided completely by never reusing frequencies, but this is
impractical and very spectrally inefficient. For cellular
systems this problem has traditionally been addressed by
grouping a specific number of different frequency cells
together to form a cluster pattern, which is then duplicated
as necessary to achieve the desired coverage — see Fig 17.
However, because cluster size determines the shortest
possible distance that can be achieved between cells using
the same frequency, which is known as the co-channel reuse
distance, cluster size should, ideally, be made as large as
possible so that the co-channel interference is made to be as
small as possible. However, increasing the former reduces
spectral efficiency. This trade-off is illustrated in Fig 18. The
solid line plot shows the amount by which the co-channel
reuse distance increases with cluster size relative to that for a
single cell cluster; and the dashed line shows how the
spectral efficiency of a cluster comprising non-sectorised
cells decreases with its size. For example, a seven-cell cluster
increases the co-channel reuse distance by a factor of
approximately 2.5 times, but reduces the spectral efficiency
by approximately 86%. For sectorised cells the spectral
efficiency is worse than shown in Fig 18 because for non-
sectorised cells the number of frequencies per cluster is
equal to its size, whereas when they are sectorised this in-
creases by an amount equal to the number of sectors per cell.
Unfortunately, the approach of using cluster frequency
reuse patterns to control co-channel interference is only
really suited to systems where there is a high degree of
design control and supporting regulatory constraints, which
is more representative of systems that use licensed rather
than licence-exempt spectrum. For the latter it is anticipated
that the interference situation will generally be worse,
mainly because there are no constraints on technology,
channel selection or number of users. Consequently, new
wireless innovations in general, and increasing user/device
density in particular, will tend to create an interference en-
vironment that is increasingly more varied and less predict-
able than currently encountered in licensed spectrum.
However, systems that use antenna arrays should
experience some level of co-channel interference sup-
40
modified fade margin, dB
30
20
10
0
0.01 0.1 1 10
outage probability, %
K = 0 (Rayleigh)
2
30
5
10
Fig 16 Modified fade margin versus outage probability.
cell cluster
co-channel
reuse distance
three-sector cell
F
3
F
4
F
2
F
3
F
2
F
5
F
1
F
6
F
7
F
5
F
6
F
7
F
1
F
4
F
1b
F
1a
F
1c
Fig 17 A cluster pattern example showing the co-channel reuse
distance for a seven-cell cluster where F represents the cell
frequency. An example of a sectorised cell is also given.
0 4 8 10
cluster size, N
2
6
4
co-channel reuse distance
improvement, linear
3
2
1
0
100
80
60
40
20
0
spectral efficiency, %
Fig 18 Co-channel reuse distance and spectral efficiency as a
function of cluster size.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
22
pression, which in principle should be of benefit to both
licensed and licence-exempt systems, although it may be
more predictable for the latter. For example, if beam forming
is used at the receiver, co-channel interference is reduced
because any interference originating from outside the beam
coverage area is reduced by the array signal processing —
see Fig 19. This has the additional benefit of also
suppressing multipath signals produced by reflections from
objects that are located outside the beam coverage area —
hence channel delay spread is also reduced. Similarly, if
beam forming is used at the transmitter, any interference
originating from sources outside the beam coverage area will
be reduced. If a receive antenna array is used to provide
diversity gain rather than beam-forming array gain, co-
channel interference reduction is still possible, but this has
to be traded against diversity gain.
Overall, there is a continuing trend towards using as low
a frequency reuse factor as possible, with the ultimate aim
being to use only one frequency so that the maximum
possible spectral efficiency is achieved. When this is the case,
co-channel interference becomes more variable due to
increased interference levels and it becomes the limiting
factor for increasing network capacity. Co-channel
interference mitigation is an area of ongoing research and
potential areas for future improvements are interference
suppression offered by advanced antenna systems,
exploiting characteristics of OFDM to enable fractional
frequency reuse
11
[3] and advanced signal processing
techniques along the lines of the single antenna interference
cancellation method being considered for GSM [8]. Whether
or not advances in these areas will result in techniques
emerging that will be equally applicable in both licensed and
licence-exempt spectrum remains to be seen.
6.System performance insights
The overall performance of any wireless system is deter-
mined purely by the design options and operating conditions
considered in this paper. However, if propagation effects and
the benefits of coding and advanced antenna systems are
initially ignored, the resulting system performance is effective-
ly dependent only on channel transport technique and modu-
lation scheme used, and regulatory constrained parameters,
which are transmit power, channel bandwidth, operating
frequency and channel duplex. Table 3 contains information
relating to these aspects for a representative set of mobile
desired
user signal
interfering
user signal
multipath
reflection
beam-forming
antenna array
Fig 19 Example of beam forming being used to maximise the
desired user signal while at the same time minimising the unwanted

interfering user signal.
Table 3 Comparison of the features of a representative set of mobile and Wi-Fi networks.
Wireless
system
Uplink
power,
W
Frequency,
GHz
Duplex Channel
modulation
schemes
Channel
coding
Channel
bandwidth,
MHz
Sub-channel Bandwidth
concatenation
Channel
transport
GSM
GPRS
EDGE
1 1.8 FDD
GMSK12
8-PSK
Yes 0.2 Yes: fixed
Sub-channel
concatenation
TDM
CDMA2000
EV-DO Rev 0
Rev A
Rev B
Rev C
*
1 1.9 FDD
BPSK
QPSK
8-PSK
16-QAM
64-QAM
Yes
1.25
(Future 1.25—20)
Yes: variable
#
Channel
concatenation
TDM
DSSS
OFDM
UMTS
HSDPA
HSUPA
HSOPA
LTE
0.125 1.95 FDD
BPSK
QPSK
8-PSK
16-QAM
64-QAM
Yes
5
(Future 1.25—20)
Yes: variable
#

DSSS
OFDM
SC-FDMA
Mobile WiMAX
(IEEE802.16e)
0.2 2.5 TDD
QPSK
16-QAM
64-QAM
Yes 12.5—20
Yes: variable
#
— OFDM
Wi-Fi
IEEE802.11b
IEEE802.11g
0.1 2.4 TDD
BPSK
QPSK
16-QAM
64-QAM
Yes 20
Yes: fixed
(802.11b)
No (802.11g)

DSSS
OFDM
# variable indicates sub-channel bandwidth is variable depending upon the spreading rate/level of sub-channelisation used
* this revision has been named Ultra Mobile Broadband
11
This is the technique of enabling different sub-channel groupings of
OFDM sub-carriers to be used by different cells at their edges to minimise
co-channel interference between them. This is particularly beneficial for
systems with a frequency reuse of one.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
23
systems, and it shows that all of them have evolved to the
point where they use some level of M-ary modulation and
that channel transport is tending to become predominantly
DSSS or OFDM. Also, although not shown in Table 3, they all
have plans to introduce advanced antenna technology in the
future. Overall, these trends suggest that these mobile
systems are all on a convergent evolution path and insights
into why can be gained by considering the fundamental
reasons for their current performance differences.
To enable the required high-level fundamental
performance insights to be gained, a relative rather than
absolute comparison will be used to compare and contrast
the various systems listed in Table 3. And the baseline
performance metric that will be used for this will be the
operating range they would achieve if they were binary
modulated at the maximum rate their channels can support.
The advantage of such an approach is that, with appropriate
assumptions about operating conditions, aspects such as
operating margins for mitigating propagation and interfer-
ence effects, dependence of minimum acceptable receive
signal-to-noise ratio on service type, system enhancements
from advanced modulation, coding, antenna gain/advanced
antenna systems can all effectively be normalised out and
hence be ignored as necessary. Although all of these factors
will affect absolute baseline performance values, they do not
affect the following features of this comparison method
(these are shown diagrammatically in Fig 20):

the baseline operating range is the range limit for the
maximum possible data rate that can be supported by
the channel bandwidth without resorting to M-ary
modulation schemes,

using M-ary modulation schemes enables operation at
data rates higher than the channel bandwidth, but
operation is restricted to region A in Fig 20,

operation beyond the baseline range, in other words
operation in region B in Fig 20, is only possible for
service rates that are less than the channel bandwidth
and then only if the receiver can effectively match its
bandwidth to that of the service rate.
Using the propagation model referred to earlier
(section 2.1) with the propagation coefficient set to 3.3, the
information in Table 3 was used to calculate the baseline
operating range for the listed wireless systems. The results
are given in Fig 21(a) referenced to that of GSM, and they
clearly show that the trend is for the baseline operating
range to fall with increasing channel bandwidth. This simply
reflects that receiver noise power is increasing with channel
bandwidth. The results in Fig 21(b) show the overall range
factor component parts associated with the difference in
channel bandwidth and the combined effect of the differ-
ence in operating frequency and transmit power. It is clear
from these results that the component representing the
combined effect of operating frequency and transmit power
has a very similar effect on all of the systems apart from GSM
and CDMA2000, which, although different from the rest, are
themselves very similar. The reason for this is that they have
a similar but higher regulatory power limit than the other
systems. This also explains why CDMA2000 has a better
baseline range than a 1.25 MHz channel WiMAX system.
opening margins, advanced
antennas and coding flex
these boundaries
range improvement region
available only to receivers
that can match their
bandwidth to that of the
service rate
region where data rate can
exceed channel bandwidth
baseline operating
range
T
x
A
B
Fig 20 Diagrammatic representation of the baseline comparison
metric used. T
x
represents the transmitting device.
(a)
(b)
Fig 21 (a) shows the overall range factor baseline results for the
representative set of mobile systems and (b) shows its two component

parts. For convenience the wireless systems are given in order of in-
creasing bandwidth and the two WiMAX entries represent its two
extreme values of channel bandwidth.
1.0
overall range factor
0.8
0.6
0.4
0
GSM
2000
WiMAX
1
WiMAX
2
Wi-Fi(g)
CDMA
UMTS
Wi-Fi(b)
0.2
1.0
component range factors
0.8
0.6
0.4
0
GSM
2000
WiMAX
1
WiMAX
2
Wi-Fi(g)
CDMA
UMTS
Wi-Fi(b)
0.2
channel bandwidth component
frequency and power component
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
24
Although receiver noise power increases with channel
bandwidth it only degrades the effective received signal-to-
noise ratio for systems that use TDM or simple forms of
OFDM (sub-channelisation not supported). This is not,
however, the case for systems that use DSSS or the more
advanced forms of OFDM (those that support sub-
channelisation). The reason for this is that in the case of
DSSS systems increasing the spreading rate is effectively
equivalent to reducing the receiver bandwidth, and hence
noise power, which improves the received signal-to-noise
ratio; and in the case of OFDM systems it is because using a
subset of the available OFDM tones increases the power per
tone, assuming constant transmit power, which increases the
received signal-to-noise ratio. If the spreading rate and level
of sub-channelisation are variable, the receiver bandwidth
can be considered as being matched to the service rate and
hence maximises the received signal-to-noise ratio. This
capability is supported by CDMA2000, UMTS and WiMAX,
but not by Wi-Fi because the OFDM version (IEEE802.11g)
does not support sub-channelisation and the DSSS version
(IEEE802.11b) uses fixed rate spreading at a rate that
provides no noise power reduction benefit for service rates
less than 1 Mbit/s.
Assuming the limiting cases of using only one WiMAX
OFDM tone or in the case of CDMA2000 and UMTS using
one of their higher value spreading rates, which in all three
cases equates to a service rate of approximately 10 kbit/s,
the overall range factor improvement is as shown by the dark
grey columns in Fig 22(a). To aid comparison with the
baseline results in Fig 21(a), which are plotted against a
different vertical scale, they have been reproduced in
Fig 22(a) using dashed line columns to represent them.
Comparing these two sets of results clearly shows that
variable rate spreading/sub-channelisation can provide
significant improvements, and provided the ratio of channel
bandwidth to spreading rate equals the service rate the
improvement is both maximised and independent of
channel bandwidth. The Wi-Fi results, however, show that
there is no improvement for IEEE802.11g and that for
IEEE802.11b the improvement is less significant than for the
other systems. This is because the IEEE802.11b spreading
factor effectively equates to the spreading rate that would
be used for a 1 Mbit/s service rate whereas the other systems
have effectively used the optimal value of spreading rate/
sub-channelisation for a 10 kbit/s service rate.
The associated improvement in the channel bandwidth
component is shown by the grey columns in Fig 22(b), and
for the same reason as before the relevant baseline results in
Fig 21(b) have been reproduced using dashed line columns
to represent them. Comparing these two sets of results
clearly shows that being able to match the receiver
bandwidth to that of the service rate improves the channel
bandwidth range factor significantly. Also, comparing the
dark grey and lighter grey column results in Fig 22(a) and
(b), respectively, clearly shows that the overall range factor
improvements largely reflect those of the channel
bandwidth component (although the improvement appears
to be larger for CDMA2000, this simply reflects that is
allowed to use a higher transmit power).
In general, for service rates significantly smaller than the
channel bandwidth, the ability to use variable rate
spreading/sub-channelisation improves significantly on the
baseline operating range. For example, GSM channel
bandwidth is eight times that of a typical voice service rate,
so its operating range is approximately half of what would be
achieved if it were able to match receiver bandwidth to that
of the service rate (see Fig 23). By way of another voice
service example, CDMA2000, UMTS and WiMAX will always
exceed the operating range that can be achieved by an
IEEE802.11b voice service by at least a factor of four (this is
derived by dividing the appropriate overall range factor
values in Fig 22(a) by the IEEE802.11b value).
Receiver noise considerations also show that for overall
equivalent bandwidth and very similar operating conditions,
2.5
overall range factor
2.0
1.5
1.0
0
GSM
2000
WiMAX
1
WiMAX
2
Wi-Fi(g)
CDMA
UMTS
Wi-Fi(b)
0.5
(a)
2.5
channel bandwidth range factor
2.0
1.5
1.0
0
GSM
2000
WiMAX
1
WiMAX
2
Wi-Fi(g)
CDMA
UMTS
Wi-Fi(b)
0.5
(b)
Fig 22 (a) shows the overall range factor improvement (dark grey
columns) achieved by matching the receiver bandwidth to the service
rate compared to the baseline results (dashed line columns. (b) shows
the channel bandwidth component improvement (light grey
columns) compared to the baseline results (dashed line columns). For
convenience the wireless systems are given in order of increasing
bandwidth and the two WiMAX entries represent its two extreme
values of channel bandwidth.
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
25
an FDD system could in some cases have a performance
advantage over a TDD system. This is because the bandwidth
of a TDD receiver has to be twice that of an FDD receiver. For
example, for constant transmit power, doubling the
bandwidth of an OFDM receiver would halve the power per
OFDM tone, whereas for a TDM system it would double the
receiver noise power. Under such operating conditions the
operating range for an FDD system would ideally be 1.2
times that of the equivalent TDD system. However, TDD has
the advantages that the uplink to downlink bandwidth ratio
can be dynamically adjusted and its channel characteristics
are generally identical for both directions.
So far the discussion has essentially concentrated on the
impact of receiver noise on operating range, but the baseline
comparison metric can also be used to gain insights into
system capacity. In principle, capacity can be increased by
concatenating channels and/or sub-channels and also by
using M-ary modulation schemes, all of which have been used
either individually or in combination to improve the perfor-
mance of the systems listed in Table 3. However, as will now
be explained, these three approaches are not all capable of
giving the same increase in data rate and in all but one case
the increase is achieved at the expense of operating range.
For systems that use TDM sub-channels, the data rate
can be increased by concatenating them and ideally this
would not cause any reduction in operating range. The
reason for this is that the receiver noise bandwidth for such
systems is usually determined by the channel bandwidth, as
is the case for GSM, and therefore concatenating sub-
channels does not change receiver noise power. However,
the increase in data rate that can be achieved is limited
because this can never exceed the channel bandwidth. If the
data rate is increased by concatenating channels, rather than
sub-channels, receiver noise power increases in proportion
to the number of channels concatenated, so there is a loss of
operating range if total transmit power is assumed constant;
effectively this is no different than increasing the channel
bandwidth. An example of using channel concatenation is
3xEV-DO which concatenates three 1.25 MHz channels, and
Fig 23 showed that increasing the bandwidth by a factor of
three reduces the operating range by 30%. Overall, because
the data-rate increase that can be achieved using either
sub-channel or channel concatenation is limited by the
number of channels available for concatenation, to operate
at data rates beyond this limit M-ary modulation has to be
used. However, this increases the minimum acceptable value
of received signal-to-noise ratio (see Fig 10), which reduces
the operating range. Figure 24 shows the range reduction
for M-ary PSK and M-ary QAM modulation expressed as a
fraction of the range that would have been achieved with
BPSK modulation. Using Figs 23 and 24, it will be found that
for a given increase in data rate, the range reduction caused
by using M-ary QAM is never worse than that caused by
increasing the channel bandwidth to support the equivalent
data-rate BPSK system.
Fig 24 Range reduction factor as a function of M-ary PSK and QAM
for a range of M-ary values.
7.Conclusions
This paper has provided an overview of the design choices
and fundamental communication concepts that determine
the performance of wireless systems. In addition to this, it
has also considered the fundamental baseline performance
differences between a representative set of mobile systems
and argued that these are primarily due to them having
different combinations of channel transport, modulation
scheme and regulatory constraints on transmit power,
channel bandwidth, operating frequency and channel
duplex. Consequently, it is anticipated that as these systems
evolve their technological differences are likely to disappear
which suggests that they are on a convergent evolution path
that will eventually result in their baseline differences being
dictated purely by regulatory factors.
References
1 Saunders S R: ‘Antennas and Propagation for Wireless Communication
Systems’, Wiley (1999).
2 ‘Mobile WiMAX Part I: A Technical Overview and Performance
Evaluation’, WiMAX Forum (February 2006).
3 Proakis J G and Salehi M: ‘Communication Systems Engineering’,
Second Edition, Prentice-Hall (2001).
0.9
range factor, times
0.8
0.7
0.6
0.5
0 4 6 10
2 8
1.0
channel bandwidth increase, times
Fig 23 Reduction in operating range as a function of the increase in

channel bandwidth.
range reduction factor
1.0
0.8
0.6
0
4 8 16 64
M-ary
32
QAM
PSK
0.4
0.2
Wireless communications — the fundamentals
BT Technology Journal • Vol 25 No 2 • April 2007
26
4 Sklar B: ‘Digital Communications: Fundamentals and Applications’,
Prentice-Hall (2001).
5 Eskafi F H: ‘Multiuser Diversity and Opportunistic Beamforming in
Wireless Networks’, Berkeley — http://caleng.berkeley.edu/archive/
2006spring/
6 Barbarossa S: ‘Multiantenna wireless communication systems’, Artech
House (2004).
7 Holma H and Toskala A: ‘WCDMA for UMTS: Radio Access for Third
Generation Mobile Communication,’ Third Edition, Wiley (2002).
8 Nickel P and Gerstacker W: ‘Single Antenna Interference Cancellation
using Prefiltering and Multiuser Joint Detection based on the M-
Algorithm’, Conference International Symposium on Telecommuni-
cations (2005) — http://www.lnt.de/lmk/publikationen/IST05_Nickel.
pdf
Terry Hodgkinson’s research interests have
covered many aspects of telecommunication
s
and associated technologies over the past 30
years. This has included researching into
optical, ATM and IP communications con-
cepts, but more recently he has been
concentrating on researching the potential
opportunities/threats that could be created
by future Wi-Fi/WiMAX wireless systems and
pervasive communication environments. Hi
s
research has been recognised by several
external awards, the most notable being a DS
c
Higher Doctorate and the 1984 Opto-
Electronic Rank Prize. He has published
extensively over the years (over 90 papers, 3 book chapters and 18 patents)
and has held several advisory roles associated with European collaborative
projects and international conference technical committees. He is a member
of the IET and IEEE, and is also a Director of the Mobile VCE.
Unless otherwise stated, copyright of the papers appearing in the BT Technology Journal is reserved by British Telecommunications plc. The views of the contributors are not necessarily those
of the Editorial Board nor of the BT Group. Mention of products and services available from suppliers outside the BT Group does not imply an endorsement.
The papers in this Journal describe processes, products and services that may be the subject of patents or patent applications. The Journal is indexed/abstracted in ABI Inform.