8 Macrocells

8.1 INTRODUCTION
This chapter introduces methods for predicting the path loss encountered in macrocells. Such
cells are commonly encountered in cellular telephony, where they are the main means of
providing initial network coverage over a wide area. However, the propagation models are
also applicable to broadcasting, private mobile radio and fixed wireless access applications
including WiMax. For such systems, in principle, the methods introduced in Chapter 6 could
be used to predict the loss over every path profile between the base station and every possible
user location. However, the data describing the terrain and clutter would be very large and the
computational effort involved would often be excessive. Even if such resources were avail-
able, the important parameter for the macrocell designer is the overall area covered, rather
than the specific field strength at particular locations, so models of a statistical nature are often
more appropriate.
The models presented in this chapter treat the path loss associated with a given macrocell
as dependent on distance, provided that the environment surrounding the base station is fairly
uniform. In consequence, the coverage area predicted by these models for an isolated base
station in an area of consistent environment type will be approximated as circular. Although
this is clearly inaccurate, it is useful for system dimensioning purposes. Methods will be
indicated at the end of this chapter and in Chapter 9 for improving the reality of this picture.

8.2 DEFINITION OF PARAMETERS

The following terms will be used in defining path loss models in this chapter and are
illustrated in Figure 8.1:
hm mobile station antenna height above local terrain height [m], often taken as
1.5 m
dm distance between the mobile and the nearest building [m]
h0 typical (usually the mean) height of buildings above local terrain height [m]
hb base station antenna height above local terrain height [m]
r great circle distance between base station and mobile [m]
R ¼ r ! 10"3 great circle distance between base station and mobile [km]
f carrier frequency [Hz]
fc ¼ f ! 10"6 carrier frequency [MHz]
l free space wavelength [m]
The basic definition of a macrocell is that hb > h0 Although buildings are not the only
obstructions in practice, they are usually by far the most significant at typical macrocellular
frequencies. In practice, base station heights are around 15–35 m if a mast is used, or around

Antennas and Propagation for Wireless Communication Systems Second Edition Simon R. Saunders and
Alejandro Aragón-Zavala
 2007 John Wiley & Sons, Ltd
164 Antennas and Propagation for Wireless Communication Systems

h0
hb

hm

dm
r

Figure 8.1: Definition of parameters for macrocell propagation models

20 m upwards if deployed on a building rooftop. The effective base station height may be
increased dramatically by locating it on a hill overlooking the region to be covered.

8.3 EMPIRICAL PATH LOSS MODELS

The two basic propagation models (free space loss and plane earth loss) examined in Chapter 5
plus the more detailed obstruction loss models described in Chapter 6 account, in principle, for
all of the major mechanisms which are encountered in macrocell prediction. However, to use
such models would require detailed knowledge of the location, dimension and constitutive
parameters of every tree, building and terrain feature in the area to be covered. This is far too
complex to be practical and would anyway yield an unnecessary amount of detail, as the system
designer is not usually interested in the particular locations being covered, but rather in the
overall extent of the coverage area. One appropriate way of accounting for these complex effects
is via an empirical model. To create such a model, an extensive set of actual path loss
measurements is made, and an appropriate function is fitted to the measurements, with para-
meters derived for the particular environment, frequency and antenna heights so as to minimise
the error between the model and the measurements. Note that each measurement represents an
average of a set of samples, the local mean, taken over a small area (around 10–50 m), in order to
remove the effects of fast fading (Chapter 10), as originally suggested by [Clarke, 68]. See
Chapter 19 for further details of measurement procedures. The model can then be used to design
systems operated in similar environments to the original measurements. A real example of an
empirical model fitted to measurements is shown in Figure 8.2. Methods of accounting for the
very large spread of the measurements at a given distance are the subject of Chapter 9.
The simplest useful form for an empirical path loss model is as follows:
PR 1 k
¼ ¼ n or in decibels L ¼ 10n log r þ K ð8:1Þ
PT L r
where PT and PR are the effective isotropic transmitted and predicted isotropic received
powers as defined in Chapter 5, L is the path loss, r is the distance between the base station and
the mobile and K ¼ "10 log10 k and n are constants of the model. Parameter k can be
considered as the reciprocal of the propagation loss that would be experienced at one metre
range ðr ¼ 1 mÞ. Models of this form will be referred to as power law models. A more
convenient form (in decibels) is
LðrÞ ¼ 10n logðr=rref Þ þ Lðrref Þ ð8:2Þ

where Lðrref Þ is the predicted loss at a reference distance rref .

Macrocells 165

−40

−50

−60
Received signal level [dBm]

−70

−80

−90

−100

−110

100 200 300 400 500600 700800 1000 2000 3000 5000
Distance from base station [m]

Figure 8.2: Empirical model of macrocell propagation: the dots are measurements taken in a
suburban area and the line represents a best-fit empirical model (reproduced by permission of
Red-M Services Ltd)

Both the free space loss and the plane earth loss can be expressed in this form. The
parameter n is known as the path loss exponent. It is found by measurement to depend on the
system parameters, such as antenna heights and the environment. Chapter 1 showed
that the path loss exponent is critical in establishing the coverage and capacity of a cellular
system.

8.3.1 Clutter Factor Models

Measurements taken in urban and suburban areas usually find a path loss exponent close to 4,
just as in the plane earth loss, but with a greater absolute loss value – i.e. larger K in Eq. (8.1).
This has led to some models being proposed which consist of the plane earth loss, plus (in
decibels) an extra loss component called the clutter factor, as shown in Figure 8.3. The various
models differ basically in the values which they assign to k and n for different frequencies and
environments.

Example 8.1
Calculate the range of a macrocell system with a maximum acceptable path loss of
138 dB, assuming hm ¼ 1:5 m; hb ¼ 30 m; fc ¼ 900 MHz and that path loss can be
modelled for this frequency and environment using the plane earth loss plus a clutter
factor of 20 dB.
166 Antennas and Propagation for Wireless Communication Systems

Figure 8.3: Clutter factor model. Note that the y-axis in this figure and in several to follow is the
negative of the propagation loss in decibels. This serves to make clear the way in which the received
power diminishes with distance.

Solution
The plane earth model of Eq. (5.34) is now modified to the following empirical model:

Lemp ¼ 40 log r " 20 log hm " 20 log hb þ K

where K is the clutter factor [decibels]. Equating this to the maximum acceptable path
loss and rearranging for log r yields

Lemp þ 20 log hm þ 20 log hb " K

log r ¼
40
138 þ 20 log 1:5 þ 20 log 30 " 20
¼ & 3:78
40

Hence r ¼ 103:78 & 6 km. This shows how the results of Examples 5.4 and 5.5 can be
modified to predict more practical ranges.

A good example of a clutter factor model is the method by [Egli, 57], which is based upon a
large number of measurements taken around American cities. Egli’s overall results were
Macrocells 167

originally presented in nomograph form, but [Delisle, 85] has given an approximation to these
results for easier computation:

L ¼ 40 log R þ 20 log fc " 20 log hb þ Lm ð8:3Þ

where

76:3 " 10 log hm for hm < 10

Lm ¼ ð8:4Þ
76:3 " 20 log hm for hm ' 10

Note that this approximation involves a small discontinuity at hm ¼ 10 m. Although plane

earth loss is frequency independent, this model introduces an additional fc"2 received power
dependence, which is more representative of the results of real measurements. For very large
antenna heights, the loss predicted by (8.3) may be less than the free space value, in which
case the free space value should be used.
The mobile antenna characteristic is approximately linear for antennas which clear
the surrounding terrain features. Elsewhere there is a square law variation for heights
in the range 2–10 m, as for the plane earth loss model. The transition value 10 m presumably
corresponds to the mean building height, although no correction is made for other heights.
The average effect of polarisation is considered negligible.

8.3.2 The Okumura–Hata Model

This is a fully empirical prediction method [Okumura, 68], based entirely upon an
extensive series of measurements made in and around Tokyo city between 200 MHz
and 2 GHz. There is no attempt to base the predictions on a physical model such as the
plane earth loss. Predictions are made via a series of graphs, the most important of which
have since been approximated in a set of formulae by [Hata, 80]. The thoroughness of
these two works taken together has made them the most widely quoted macrocell
prediction model, often regarded as a standard against which to judge new approaches.
The urban values in the model presented below have been standardised for international
use in [ITU, 529]
The method involves dividing the prediction area into a series of clutter and terrain
categories, namely open, suburban and urban. These are summarised as follows:
( Open area: Open space, no tall trees or buildings in path, plot of land cleared for
300–400 m ahead, e.g. farmland, rice fields, open fields.
( Suburban area: Village or highway scattered with trees and houses, some obstacles near
the mobile but not very congested.
( Urban area: Built up city or large town with large buildings and houses with two or more
storeys, or larger villages with close houses and tall, thickly grown trees.
Okumura takes urban areas as a reference and applies correction factors for conversion to the
other classifications. This is a sensible choice, as such areas avoid the large variability present
in suburban areas (Chapter 9) and yet include the effects of obstructions better than could be
done with open areas. A series of terrain types is also defined for when such information is
available. Quasi-smooth terrain is taken as the reference, and correction factors are added for
the other types.
168 Antennas and Propagation for Wireless Communication Systems

Okumura’s predictions of median path loss are usually calculated using Hata’s approx-
imations as follows [Hata, 80]:

Urban areas LdB ¼ A þ B log R " E

Suburban areas LdB ¼ A þ B log R " C ð8:5Þ
Open areas LdB ¼ A þ B log R " D

where

A ¼ 69:55 þ 26:16 log fc " 13:82 log hb

B ¼ 44:9 " 6:55 log hb
C ¼ 2ðlogðfc =28ÞÞ2 þ5:4
D ¼ 4:78ðlog fc Þ2 " 18:33 log fc þ 40:94 ð8:6Þ
E ¼ 3:2ðlogð11:75hm ÞÞ2 " 4:97 for large cities; fc ' 300 MHz
E ¼ 8:29ðlogð1:54hm ÞÞ2 " 1:1 for large cities; fc < 300 MHz
E ¼ ð1:1 log fc " 0:7Þhm " ð1:56 log fc " 0:8Þ for medium to small cities

The model is valid only for 150 MHz ) fc ) 1500 MHz; 30 m ) hb ) 200 m;
1 m < hm < 10 m and R > 1 km. The path loss exponent is given by B=10, which is a little
less than 4, decreasing with increasing base station antenna height.
Base station antenna height hb is defined as the height above the average ground level in the
range 3–10 km from the base station; hb may therefore vary slightly with the direction of the
mobile from the base. The height gain factor varies between 6 dB per octave and 9 dB per
octave as the height increases from 30 m to 1 km. Measurements also suggest this factor
depends upon range.
Okumura found that mobile antenna height gain is 3 dB per octave up to hm ¼ 3 m and
8 dB per octave beyond. It depends partially upon urban density, apparently as a result of the
effect of building heights on the angle-of-arrival of wave energy at the mobile and the
consequent shadow loss variation (Chapter 9). Urban areas are therefore subdivided into
large cities and medium/small cities, where an area having an average building height in
excess of 15 m is defined as a large city.
Other correction factors are included in Okumura’s original work for the effects of street
orientation (if an area has a large proportion of streets which are either radial or tangential to
the propagation direction) and a fine correction for rolling hilly terrain (used if a large
proportion of streets are placed at either the peaks or valleys of the terrain undulations).
Application of the method involves first finding the basic median field strength in concentric
circles around the base station, then amending them according to the terrain and clutter
correction graphs.
Okumura’s predictions have been found useful in many cases [COST207, 89], particularly
in suburban areas. However, other measurements have been in disagreement with these
predictions; the reasons for error are often cited as the difference in the characteristics of
the area under test with Tokyo. Other authors such as [Kozono, 77] have attempted to modify
Okumura’s method to include a measure of building density, but such approaches have not
found common acceptance.
Macrocells 169

The Okumura–Hata model, together with related corrections, is probably the single most
common model used in designing real systems. Several commercial prediction tools essentially
rely on variations of this model, optimised for the particular environments they are catering for,
as the basis of their predictions; see www.simonsaunders.com/apbook for a list of such tools.

8.3.3 The COST 231–Hata Model

The Okumura–Hata model for medium to small cities has been extended to cover the band
1500 MHz < fc < 2000 MHz [COST231, 99].
LdB ¼ F þ B log R " E þ G ð8:7Þ

where

F ¼ 46:3 þ 33:9 log fc " 13:82 log hb ð8:8Þ

E is as defined in (8.6) for medium to small cities and

0 dB medium-sized cities and suburban areas

G¼ ð8:9Þ
3 dB metropolitan areas

8.3.4 The Lee Model

The Lee model is a power law model, with parameters taken from measurements in a number
of locations, together with a procedure for calculating an effective base station antenna height
which takes account of the variations in the terrain [Lee, 82; Lee, 93]. It can be expressed in
the simplified form
L ¼ 10n log R " 20 log hbðeffÞ " P0 " 10 log hm þ 29 ð8:10Þ

where n and P0 are given by measurements as shown in Table 8.1 and hbðeffÞ is the effective
base station antenna height. The measurements were all made at 900 MHz, and correction
factors must be applied for other frequencies, but these do not appear to have been specified in
the open literature.
The effective base station height is determined by projecting the slope of the terrain in the
near vicinity of the mobile to the base station location. Figure 8.4 shows how this effective
height varies for four mobile locations on gently sloping terrain.

Table 8.1: Parameters for the Lee model

Environment n P0
Free space 2 "45
Open area 4.35 "49
Suburban 3.84 "61.7
Urban Philadelphia 3.68 "70
Newark 4.31 "64
Tokyo 3.05 "84
New York city 4.8 "77
170 Antennas and Propagation for Wireless Communication Systems

hA
D
hB C
A B
hC
hD

Figure 8.4: Determination of effective base station antenna height for Lee’s model

8.3.5 The Ibrahim and Parsons Model

This method [Ibrahim, 83] is based upon a series of field trials around London. The method is
not intended as a fully general prediction model, but as a first step towards quantifying urban
propagation loss. It integrates well with a previous method [Edwards, 69] for predicting
terrain diffraction effects as the same 0.5 km square database is also used. Each square is
assigned three parameters, H, U and L, defined as follows.
Terrain height H is defined as the actual height of a peak, basin, plateau or valley found in
each square, or the arithmetic mean of the minimum and maximum heights found in the
square if it does not contain any such features.
The degree of urbanisation factor U is defined as the percentage of building site area
within the square which is occupied by buildings having four or more floors. For the 24 test
squares in inner London which were analysed, U varied between 2 and 95%, suggesting that
this parameter is sensitive enough for the purpose.
Land usage factor L is defined as the percentage of the test area actually occupied by any
buildings.
These parameters were selected empirically as having good correlation with the data. Two
models were proposed. The fully empirical method shows marginally lower prediction errors
but relies on a complex formulation which bears no direct relationship to propagation
principles.
The semi-empirical method, as with the Egli clutter factor method, is based upon the plane
earth loss together with a clutter factor b, expressed as a function of fc , L, H and U. Only the
semi-empirical method will be examined here, as it has been quoted in later work by the same
author [Parsons, 92], and as it forms a better basis for future development. The model is given
as

LT ¼ 40 log r " 20 logðhm hb Þ þ b

fc
where b ¼ 20 þ þ 0:18L " 0:34H þ K ð8:11Þ
40
and K ¼ 0:094U " 5:9
Macrocells 171

Data is extracted from the databases compiled by local authorities in the United Kingdom.
Although information on U is available only in highly urbanised city centres, K is set to zero
elsewhere. Considerably lower accuracy may be expected in such areas as K ¼ 0 would
correspond to U ¼ 63%. RMS errors calculated from the original data on which the model is
based vary from 2.0 to 5.8 dB as frequency is increased from 168 to 900 MHz. A comparison
is also shown for some independent data, but error statistics are not given.
The model is of limited use in suburban areas as U will normally be zero, giving no
measure of building height distribution.

8.3.6 Environment Categories

In an empirical model, it is crucial to correctly classify the environment in which the system is
operating. The models assume the characteristics of the environment to be predicted are
sufficiently similar to those where the original measurements were taken that the propagation
loss at a given distance will be similar. Good results will therefore be obtained only if the
correct classification is chosen. The categories of environment should also be sufficiently
numerous that the properties of different locations classed within the same category are not
too variable. The decision as to which category an environment fits into is usually purely
subjective and may vary between individuals and countries. For example, the Okumura–Hata
model uses four categories: large cities, medium-small cities, suburban areas and open areas.
Although the original measurements were made in Tokyo, the model relies on other parts of
the world having characteristics which are somehow similar to those in Tokyo. Although this
is an extremely questionable assumption, it is nevertheless true that the model has been
applied to many successful system designs.
Many more detailed schemes exist for qualitative classification of land usage; Table 8.2
shows one example. Schemes often correspond to sources of data, such as satellite remote-
sensing data which classifies land according to the degree of scattering experienced at various
wavelengths. This at least avoids the need for ambiguous judgements to be made. Similarly,
the Ibrahim and Parsons model uses a clear numerical approach to classification. Never-
theless, there is no guarantee that there is any one-to-one mapping between the propagation
characteristics and such measures of land usage. In order to find more appropriate parameters,

Table 8.2: British Telecom land usage categories [Huish, 88]

Category Description
0 Rivers, lakes and seas
1 Open rural areas, e.g. fields and heathlands with few trees
2 Rural areas, similar to the above, but with some wooded areas, e.g. parkland
3 Wooded or forested rural areas
4 Hilly or mountainous rural areas
5 Suburban areas, low-density dwellings and modern industrial estates
6 Suburban areas, higher density dwellings, e.g. council estates
7 Urban areas with buildings of up to four storeys, but with some open space between
8 Higher density urban areas in which some buildings have more than four storeys
9 Dense urban areas in which most of the buildings have more than four storeys
and some can be classed as skyscrapers; this category is restricted to the centres
of a few large cities
172 Antennas and Propagation for Wireless Communication Systems

the growing tendency in macrocellular propagation is towards models which have a physical
basis, and these are examined in the next section.

8.4 PHYSICAL MODELS

Although empirical models have been extensively applied with good results, they suffer from
a number of disadvantages:
( They can only be used over parameter ranges included in the original measurement set.
( Environments must be classified subjectively according to categories such as ‘urban’,
which have different meanings in different countries.
( They provide no physical insight into the mechanisms by which propagation occurs.
The last point is particularly significant, as empirical models are unable to account for factors
such as an unusually large building or hill which may greatly modify propagation in particular
locations. Although the plane earth model has a path loss exponent close to that observed in
actual measurements (i.e. 4), the simple physical situation it describes is rarely applicable in
practice. The mobile is almost always operated (at least in macrocells) in situations where it
does not have a line-of-sight path to either the base station or to the ground reflection point, so
the two-ray situation which the plane earth model relies upon is hardly ever applicable. In
order to find a more satisfactory physical propagation model, the remaining models in this
chapter examine diffraction as a potential mechanism.

8.4.1 The Allsebrook and Parsons Model

Although this model [Allsebrook, 77] is based upon a series of measurements, it may be
regarded as an early attempt to provide a physical basis for urban prediction models.
Measurements were made in three British cities (Bradford, Bath and Birmingham) at 86,
167 and 441 MHz. These cities cover a wide range of terrain and building classifications. A
40 dB/decade range dependence is again forced, as would be expected for plane earth loss.
This results in an Egli type model with a maximum RMS error of 8.3 dB at 441 MHz. (Note
that a least-squares curve fit at this frequency results in a range dependence of only 24 dB/
decade). A clutter factor b is introduced to account for excess loss relative to the plane earth
calculation.
The frequency dependence of the measured clutter factor is compared with an approxima-
tion to the excess loss expected from a 10 m absorbing knife-edge, placed 30 m away from a
2 m high mobile antenna. The predictions compare reasonably well with the mean values of b at
86 and 167 MHz, but considerably underestimate it at 441 MHz. The knife-edge calculation is
used as a generalised means of calculating diffraction from the rooftop of the building adjacent
to the mobile, with a UHF correction factor g included to force agreement with the measured
values above 200 MHz. It is suggested that this deviation is the result of building width being
more significant at the higher frequencies, but this is not confirmed by any analysis.
Allsebrook and Parsons’ ‘flat city’ model can be expressed as

LT ¼ LP þ LB þ g
!
h0 " hm ð8:12Þ
where LB ¼ 20 log pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
548 ðdm ! 10"3 Þ=fc
Macrocells 173

and LP is the plane earth loss. For ease of computation the prediction curve for g can be
replaced by the following quadratic approximation:

g ¼ "2:03 " 6:67fc þ 8:1 ! 10"5 fc2 ð8:13Þ

Note that, in the calculation of LP here, the effective antenna heights are those of the base
station and the building, giving an overall physical model which may be represented by
Figure 8.5.

Figure 8.5: Physical interpretation of Allsebrook and Parsons model

A published discussion of this model [Delisle, 85] finds that the correction factor g is
necessary in open areas as well as in (sub)urban areas, although the physical cause suggested
for g cannot apply in line-of-sight situations. Additionally, the quoted value of g is too large in
all situations, casting doubt upon the model’s generality. The model is physically valid only in
terms of the final building diffraction; the use of the plane earth calculation suggests the
existence of a specular ground reflection, which is highly unlikely in a built-up area. Despite
this, the model may be considered an improvement over empirical methods as it was the first
to make any allowance for the geometry of the specific path being considered.

8.4.2 The Ikegami Model

This model attempts to produce an entirely deterministic prediction of field strengths at
specified points [Ikegami, 91]. Using a detailed map of building heights, shapes and positions,
ray paths between the transmitter and receiver are traced, with the restriction that only single
reflections from walls are accounted for. Diffraction is calculated using a single edge
approximation at the building nearest the mobile, and wall reflection loss is assumed to be
fixed at a constant value. The two rays (reflected and diffracted) are power summed, resulting
in the following approximate model:

LE ¼ 10 log fc þ 10 logðsin %Þ þ 20 logðh0 " hm Þ

ð8:14Þ
% &
3
" 10 log w " 10 log 1 þ 2 " 5:8
Lr
174 Antennas and Propagation for Wireless Communication Systems

where % is the angle between the street and the direct line from base to mobile and Lr ¼ 0:25 is
the reflection loss. The analysis assumes that the mobile is in the centre of the street. The model
therefore represents the situation illustrated in Figure 8.6. It further assumes that the elevation
angle of the base station from the top of the knife-edge is negligible in comparison to the
diffraction angle down to the mobile level, so there is no dependence on base station height.

Base Single diffraction

station
Reflection

Mobile

Figure 8.6: Physical interpretation of Ikegami model

A comparison of the results of this model with measurements at 200, 400 and 600 MHz
shows that the general trend of variations along a street is accounted successfully. The predic-
tions suggest that field strength is broadly independent of a mobile’s position across the street.
This is confirmed by the mean values of a large number of measurements, although the spread of
values is rather high. Acceptable agreement is also obtained for variations with street angle and
width.
Although it accounts reasonably well for ‘close-in’ variations in field strength, it is a flawed
assumption that base station antenna height does not affect propagation. The same assump-
tion means that the free space path loss exponent is assumed, so the model tends to under-
estimate loss at large distances. Similarly, the variation with frequency is underestimated
compared with measurements.

8.4.3 Rooftop Diffraction

When a macrocell system is operated in a built-up area with reasonably flat terrain, the
dominant mode of propagation is multiple diffraction over the building rooftops. Diffraction
can occur around the sides of individual buildings, but this tends to become highly attenuated
over reasonable distances as many interactions with individual buildings are involved.
The diffraction angle over most of the rooftops is small for typical base station heights and
distances, usually less than 1* . In these cases the diffraction is largely unaffected by the
particular shape of the obstacles, so it is appropriate to represent the buildings by equivalent
knife-edges. The one exception to this is diffraction from the ‘final building’ at which the
wave is diffracted from rooftop level down to the street-level antenna of the mobile
(Figure 8.7). It is usual to separate these processes into multiple diffraction across the first
ðn " 1Þ buildings, treated as knife-edges, and a final building which can be treated either as a
knife-edge or as some more complex shape for which the diffraction coefficient is known.
The small diffraction angles encountered have two negative consequences for prediction
of these effects. Firstly, a large number of building rooftops may appear within the first
Fresnel zone, all contributing to the propagation loss. Secondly, the near-grazing incidence
Macrocells 175

Figure 8.7: Multiple diffraction over building rooftops

implies that the approximate models described in Chapter 6 will fail, leading to very
inaccurate predictions. The full multiple edge integral must instead be applied (Section
6.6), which could lead to very long computation times, particularly as it is desired to predict
the base station coverage over a wide area, which would require a large number of individual
path profiles. Special methods have been developed to enable reasonably rapid calculation of
the multiple diffraction integral for cases where accurate results are required and where the
necessary data on the building positions and heights is available [Saunders, 94]. Such data is
usually too expensive for general use in macrocells, although satellite imagery is reducing the
data cost considerably nowadays. Two simplified solutions with reduced data and computa-
tional requirements are therefore examined here.

8.4.4 The Flat Edge Model

In this model [Saunders, 91], the situation is simplified by assuming all of the buildings to be
of equal height and spacing. The values used can be average values for the area under
consideration, or can be calculated individually for each direction from the base station if the
degree of urbanisation varies significantly. The geometry is shown in Figure 8.8, illustrating
the following parameters additional to the definitions in Section 8.2: distance r1 from the base
station to the first building [m] and elevation angle a of the base station antenna from the top
of the final building [rad]. In Figure 8.8, buildings are arranged normal to the great circle path.

Figure 8.8: Geometry for the flat edge model

176 Antennas and Propagation for Wireless Communication Systems

As this will not normally be the case in practice, the value of w used should be an effective one
to account for the longer paths between the buildings for oblique incidence.
The excess path loss is then expressed as

Lex ¼ Ln"1 ðtÞLke ð8:15Þ

where Lke accounts for single-edge diffraction over the final building and Ln"1 accounts for
multiple diffraction over the remaining ðn " 1Þ buildings. It turns out, provided r1 + nw (i.e.
the base station is relatively distant from the first building), that the multiple diffraction
integral (6.24) can be completely solved in this special case. The result is that Ln"1 is a
function of a parameter t only, where t is given by
rffiffiffiffiffiffi
&w
t ¼ "a ð8:16Þ
l
It is given by the following formula

n"1
1X pffiffiffiffiffiffiffiffiffiffiffiffi
Ln ðtÞ ¼ Lm ðtÞFs ð"jt n " mÞ for n ' 1 L0 ðtÞ ¼ 1 ð8:17Þ
n m"0

where
2
(" rffiffiffi! # " rffiffiffi! #)
e"jx 2 1 2 1
Fs ðjxÞ ¼ pffiffiffiffi S x þ þj C x þ ð8:18Þ
2j & 2 & 2

and S(.) and C(.) are the standard Fresnel sine and cosine integrals defined in Appendix B.
This formulation is extremely quick and simple to compute and it applies for any values of a,
even when the base station antenna height is below the rooftop level. The number of buildings
can be increased to extremely high values with no difficulties.
The flat edge model may be calculated either directly from (8.17), or the results may be
estimated from the prediction curves in Figure 8.9, which show the cases where hb ' h0 . An
alternative approach is to use the approximate formula

Ln ðtÞ ¼ "20 log An ðtÞ

ð8:19Þ
¼ "ðc1 þ c2 log nÞ logð"tÞ " ðc3 þ c4 log nÞ ½dB.

where c1 ¼ 3:29; c2 ¼ 9:90; c3 ¼ 0:77; c4 ¼ 0:26. This approximates the value of (8.17) with
an accuracy better than /1:5 dB for 1 ) n ) 100 and "1 ) t < 0. It also enables us to
investigate the behaviour of the effective path loss exponent for the flat edge model, as for
fixed n, we can rewrite (8.19) with L being the path loss as a power ratio as:
% rffiffiffiffiffiffi&"ðc2 =10Þ log n % rffiffiffiffiffiffi&"ðc2 =10Þ log n
&w hb " h0 &w
L / ð"tÞ"ðc2 =10Þ log n ¼ a & ð8:20Þ
l r l
where the approximation holds if ðhb " h0 Þ 1 r. This is the excess field strength, so the
overall path loss exponent, including an extra 2 from the free space part of the loss, is as
follows:
Macrocells 177

Excess diffraction loss to top of final building [−dB]

t=−1.0
0

−5

t=0
−10

−15

−20

−25
100 101 102
Number of edges

Figure 8.9: Flat edge model prediction curves for elevated base antennas: curves relate to t varying
from 0 to "1 in steps of 0.1. The crosses indicate the number of edges required for a settled field
according to (8.24)

Path loss exponent ¼ 2 þ ðc2 =10Þ log n ð8:21Þ

This expression is shown in Figure 8.10, where it is apparent that, for reasonably large
numbers of buildings, the path loss exponent for the flat edge model is close to 4, just as
observed in practical measurements. More generally, we can state
Multiple building diffraction accounts for the variation of path loss with range which is
observed in macrocell measurements.
Figure 8.9 shows that, for ðhb > h0 Þ (i.e. t < 0), the field at the top of the final
building eventually settles to a constant value as the number of edges increases. This number,
ns , corresponds to the number required to fill the first Fresnel zone around the ray
from the base station to the final building. The first Fresnel zone radius r1 is given approxi-
mately by
pffiffiffiffiffi
r1 & ls ð8:22Þ

where s is the distance along the ray from the field point, provided s 1 r. Hence, for small a,
pffiffiffiffiffiffiffiffiffiffi
r 1 lns w
a ¼ tan"1 & ð8:23Þ
ns w ns w
178 Antennas and Propagation for Wireless Communication Systems

3.8

3.6

3.4
Path loss exponent

3.2

2.8

2.6

2.4

2.2

2
0 10 20 30 40 50 60 70 80 90 100
Number of buildings

Figure 8.10: Path loss exponent for the flat edge model

So
l &
ns & ¼ ð8:24Þ
a2 w t2

This is marked in Figure 8.9. Note that the number of edges required for settling rises very
rapidly with decreasing a. Whenever a ) 0 the field does not settle at all, but decreases
monotonically for all n.
The flat edge model is completed by modelling the final building diffraction loss and the
reflections from the buildings across the street using the Ikegami model from Section 8.4.2.
Thus the total path loss is given by
LT ¼ Ln ðtÞ þ LF þ LE ð8:25Þ

where Ln ðtÞ can be found from (8.17), Figure 8.9 or (8.19), LF is the free space loss (5.20) and
LE is given in (8.14).

8.4.5 The Walfisch–Bertoni Model

This model can be considered as the limiting case of the flat edge model when the number of
buildings is sufficient for the field to settle, i.e. n ' ns . The multiple diffraction process was
investigated in [Walfisch, 88] using a numerical evaluation of the Kirchhoff–Huygens
Macrocells 179

integral and a power law formula is fitted to the results for the settled field. The Walfisch–
Bertoni model was the first to actually demonstrate that multiple building diffraction accounts
for the variation of distance with range which is observed in measurements.
The settled field approximation is as follows:
% rffiffiffiffi&0:9 - "t .0:9
a w
Asettled ðtÞ & 0:1 ¼ 0:1 ð8:26Þ
0:03 l 0:03
This is valid only for 0:03 ) t ) 0:4. For large ranges, we can again put
rffiffiffiffiffiffi rffiffiffiffiffiffi
&w hb " hm &w
t & "a &" ð8:27Þ
l r l

Hence Lsettled / r "1:8 . The free space loss is proportional to r "2 , so this model predicts that
total propagation loss is proportional to r "3:8 , which is close to the r "4 law which is
commonly assumed in empirical models and found in measurements. A single knife-edge
approximation with a reflection from the building opposite is again used, just as in the
Ikegami model, to account for the diffraction from the final building. The complete model is
expressed as
Lex ¼ 57:1 þ LA þ log fc þ 18 log R " 18 logðhb " h0 Þ
R2 ð8:28Þ
/ 0
" 18 log 1 "
17ðhb " h0 Þ

where
/- . 0
w 2 2
LA ¼ 5 log þðh0 " hm Þ " 9 log w
2
/ 01 ð8:29Þ
"1 2ðh0 " hm Þ
þ 20 log tan
w
The use of the settled field approximation requires that large numbers of buildings are present,
particularly when a is small. Despite this limitation, the Walfisch–Bertoni model is the first to
have accounted for observed path loss variation using realistic physical assumptions rather
than relying upon forcing agreement using propagation models of entirely different situa-
tions, such as the use of the plane earth model in clutter factor models.
A later paper by some of the same authors [Maciel, 93] provides an alternate means of
computing the settled rooftop field which is valid over a wider range of the parameter t.
Ln ðtÞ ¼ "10 logðG2 Q2 ðtÞÞ ð8:30Þ
where
QðtÞ ¼ "3:502t " 3:327t2 " 0:962t3 ð8:31Þ
The antenna gain G2 is that in the direction to the highest building edge visible from the base
station antenna when it is below the surrounding rooftops, or the gain in the horizontal plane when
the base station antenna is above the surrounding rooftops.
The authors state that this expression predicts Ln with an accuracy greater than 0.5 dB over
the range "1 ) t ) "0:01, again provided that the number of buildings in the path is large
180 Antennas and Propagation for Wireless Communication Systems

enough for the field to settle. This expression was found to describe measurements made in
Denmark for small r with good accuracy [Eggers, 90].

8.4.6 COST 231/Walfisch–Ikegami Model

The Walfisch–Bertoni model for the settled field has been combined with the Ikegami model
for diffraction down to street level plus some empirical correction factors to improve
agreement with measurements in a single integrated model by the COST 231 project
[COST 231, 99].
For non-line-of-sight conditions the total loss is given by

L ¼ LF þ Lmsd þ Lsd ð8:32Þ

where LF is the free space loss, Lmsd accounts for multiple knife-edge diffraction to the top of
the final building and Lsd accounts for the single diffraction and scattering process down to
street level. L is given a minimum value of LF in case the other terms become negative. The
individual terms are
ðh0 " hm Þ2
Lsd ¼ "16:9 þ 10 log fc þ 10 log þ Lð%Þ ð8:33Þ
wm
where wm is the distance between the building faces on either side of the street containing the
mobile (typically wm ¼ w=2), and the final term accounts for street orientation at an angle % to
the great circle path:
for 0* < %< 35*
8
< "10 þ 0:354%
Lð%Þ ¼ 2:5 þ 0:075 ð% " 35* Þ for 35* ) %< 55* ð8:34Þ
4:0 " 0:114 ð% " 55* Þ for 55* ) % ) 90*
:

Finally, the rooftop diffraction term is given by

Lmsd ¼ Lbsh þ ka þ kd log R þ kf log fc " 9 log w ð8:35Þ

where
"18 log½1 þ ðhb " h0 Þ. for hb > h0
Lbsh ¼ ð8:36Þ
0 for hb ) h0

54 for hb > h0
8
>
54 " 0:8 ðhb " h0 Þ for R ' 0:5 km and hb ) h0
<
ka ¼ ð8:37Þ
: 54 " 0:8 ðhb " h0 ÞR
>
for R < 0:5 km and hb ) h0
0:5

8
< 18 for hb > h0
kd ¼ ð hb " h0 Þ ð8:38Þ
: 18 " 15 for hb ) h0
h0
Macrocells 181
% &
fc
kf ¼ "4 þ 0:7 "1 for medium-sized city and
925
suburban areas with medium tree density ð8:39Þ
% &
fc
kf ¼ "4 þ 1:5 "1 for metropolitan centres
925

For approximate work, the following parameter values can be used:

3nfloors for flat roofs

h0 ¼
3nfloors þ 3 for pitched roofs ð8:40Þ
*
w ¼ 20 " 50 m; dm ¼ w=2; % ¼ 90

where nfloors is the number of floors in the building. The model is applicable for
800 MHz ) fc ) 2000 MHz, 4 m ) hb ) 50 m, 1 m ) hm ) 3 m and 0:02 km ) R ) 5 km.
An alternative approach is to replace the Lmsd term by Ln ðtÞ from the flat edge model. This
would enable the path loss exponent to vary according to the number of buildings and to be
uniformly valid for hb ) h0. Note, however, that for very low base station antennas other
propagation mechanisms, such as diffraction around vertical building edges and multiple
reflections from building walls, are likely to be significant. Models intended for microcells are
then more likely to be appropriate, as described in Chapter 12.

8.5 ITU-R MODELS

It is often difficult to select the best model for a given application from the many described in
this chapter. Recommendations produced by the International Telecommunications Union are
a good reference source in this situation, as they summarise in simple form some recom-
mended procedures. They may not always represent the most accurate model for a given case,
but they have the benefit of being widely accepted and used for coordination and comparison
purposes. Two recommendations in particular are relevant to this chapter and are briefly
summarised below. The full text of the recommendations may be obtained online from
http://www.itu.int/ITU-R/

8.5.1 ITU-R Recommendation P.1411

This recommendation [ITU, 1411] contains a model relevant to macrocell applications which
is applicable to non-line-of-sight systems operating from 20 m to 5 km. The model is
essentially a version of the COST 231 Walfisch–Ikegami model with simplified and general-
ised calculation procedures. The validity range is quoted as follows:

hb : 4"50 m;
hm : 1"3 m;
fc : 800"2000 MHz for hb ) ho
fc : 800"5000 MHz for hb > ho :
182 Antennas and Propagation for Wireless Communication Systems

Other models from this recommendation are relevant to microcell situations and will be
examined in Chapter 12.

8.5.2 ITU-R Recommendation P.1546

For longer ranges up to 1000 km this recommendation [ITU, 1546] contains useful models.
The model is based on series of curves (or tables), originating from measurements, allowing
predictions for wide area macrocells and for broadcasting and fixed wireless access applica-
tions. As well as the basic prediction curves there are a set of correction methods to account
for factors including terrain variations, effective base station antenna height, paths which
cross mixtures of land and sea and predictions for different time percentage availability. The
method gives similar results to the Okumura–Hata model for distances up to 10 km. Overall
the method is valid for fc ¼ 30 MHz to 3 GHz and distances from 1 to 1000 km. As such it is
applicable to a wide range of system types, including wide area macrocells, Private Mobile
Radio systems and terrestrial video and audio broadcasting networks.

8.6 COMPARISON OF MODELS

The path loss predictions of most of the models described in this chapter are compared in
Table 8.3. It shows the exponents of path loss variation predicted by each model. Thus a –2 in
the hb column means the model predicts that path loss is inversely proportional to the square
of the base station antenna height. In some cases it is difficult to express the variation in this
form, but otherwise it is useful as a means of comparison.

Table 8.3 Comparison of macrocell propagation models

Path loss
Model exponent hb hm fc
Free space (Chapter 5) 2 0 0 2
Plane earth (Chapter 5) 4 "2 "2 0
Egli 8.3.1 4 "2 "1ðhm < 10Þ 2
"2ðhm > 10Þ
Okumura–Hata
(Section 8.3.2)
COST231–Hata
(Section 8.3.3) a b See (8.5) & 2.6
Lee (Section 8.3.4) 2–4.3 "2 "1 NS
Ibrahim (Section 8.3.5) 4 "2 "2 10fc =400
Allsebrook (Section 8.4.1) 4 "2 ðh0 " hm Þ2 &2
Ikegami (Section 8.4.2) 1 0 ðh0 " hm Þ2 2
Flat edge (Section 8.4.4) 2–4 &"2 ðh0 " hm Þ2 &2.1
Walfisch–Bertoni (Section 8.4.5) 3.8 & "1.8 &"1 2.1
COST-231 Walfisch–Ikegami
(Section 8.4.6) and [ITU, 1411]
NLOS model (Section 8.5.1) 3.8 &"1.8 &"1 See (8.32)
NS ¼ not specified, a ¼ 4:5 " 0:66 log hb , b ¼ "1:38 " 0:66 log r
Macrocells 183

8.7 COMPUTERISED PLANNING TOOLS

The methods described in this chapter are most often implemented for practical planning
within computer software. The development of such software has been motivated and enabled
by a number of factors:
( The enormous increase in the need to plan cellular systems accurately and quickly.
( The development of fast, affordable computing resources.
( The development of geographical information systems, which index data on terrain,
clutter and land usage in an easily accessible and manipulable form.
Such techniques have been implemented in a wide range of commercially available and company-
specific planning tools. Some of them are listed at http://www.simonsaunders.com/apbook.
Although most are based on combined empirical and simple physical models, it is anticipated
there will be progressive evolution in the future towards more physical or physical-statistical
methods as computing resources continue to cheapen, clutter data improves in resolution and
cost and as research develops into numerically efficient path loss prediction algorithms.

8.8 CONCLUSION
Propagation path loss modelling is the fundamental method of predicting the range of a
mobile radio system. The accuracy of the path loss predictions is crucial in determining
whether a particular system design will be viable. In macrocells, empirical models have been
used with great success, but deterministic physical models are being increasingly applied as a
means of improving accuracy, based on the use of multiple rooftop diffraction as the key
propagation mechanism. This accuracy comes at the expense of increased input data require-
ments and computational complexity. Another generation of models is expected to appear
which combine sound physical principles with statistical parameters, which can economically
be obtained in order to provide the optimum balance between accuracy and complexity.
The path loss may be taken, very roughly, to be given by

PR 1 h m h2
¼ ¼ k 4 2b ð8:41Þ
PT L r fc

where k is some constant appropriate to the environment. It should be emphasised that this
expression, and all of the models described in this chapter, account only for the effects of
typical clutter on flat or gently rolling terrain. When the terrain variations are sufficient to
cause extra obstruction loss, then the models must be supplemented by calculations of terrain
loss in a manner similar to Chapter 6.

REFERENCES
[Allsebrook, 77] K. Allsebrook and J. D. Parsons, Mobile radio propagation in British cities at
frequencies in the VHF and UHF bands, IEEE Transactions on Vehicular
Technology, 26 (4), 95–102, 1977.
[Clarke, 68] R. H. Clarke, A statistical theory of mobile radio reception, Bell System
Technical Journal, 47 (6), 957–1000, 1968.
184 Antennas and Propagation for Wireless Communication Systems

[COST-207, 89] Final report of the COST-207 management committee, Digital land mobile
radio communications, Commission of the European Communities, L-2920,
Luxembourg, 1989.
[COST 231, 99] COST 231 Final report, Digital Mobile Radio: COST 231 View on the
Evolution Towards 3rd Generation Systems, Commission of the European
Communities and COST Telecommunications, Brussels, 1999.
[Delisle, 85] G. Y. Delisle, J. Lefevre, M. Lecours and J. Chouinard, Propagation loss
prediction: a comparative study with application to the mobile radio channel,
IEEE Transactions on Vehicular Technology, 26 (4), 295–308, 1985.
[Edwards, 69] R. Edwards and J. Durkin, Computer prediction of service areas for VHF
mobile radio networks, Proceedings of the IEE, 116 (9), 1493–1500, 1969.
[Eggers, 90] P. Eggers and P. Barry, Comparison of a diffraction-based radiowave propaga-
tion model with measurements, Electron Letters, 26 (8), 530–531, 1990.
[Egli, 57] J. J. Egli, Radio Propagation above 40MC over irregular terrain, Proceedings of the
IRE, 1383–91, 1957.
[Hata, 80] M. Hata, Empirical formula for propagation loss in land mobile radio services,
IEEE Transactions on Vehicular Technology, 29, 317–25, 1980.
[Huish, 88] P. W. Huish and E. Gurdenli, Radio channel measurements and prediction for
future mobile radio systems, British Telecom Technology Journal, 6 (1), 43–53,
1988.
[Ibrahim, 83] M. F. Ibrahim and J. D. Parsons, Signal strength prediction in built-up areas,
Proceedings of the IEE, 130F (5), 377–84, 1983.
[Ikegami, 91] F. Ikegami, T. Takeuchi and S. Yoshida, Theoretical prediction of mean field
strength for urban mobile radio, IEEE Transactions on Antennas and Propaga-
tion, 39 (3), 299–302, 1991.
[ITU, 529] International Telecommunication Union, ITU-R Recommendation P.529-2:
Prediction methods for the terrestrial land mobile service in the VHF and
UHF bands, Geneva, 1997.
[ITU, 1411] International Telecommunication Union, ITU-R Recommendation P.1411-3:
Propagation data and prediction methods for the planning of short-range
outdoor radiocommunication systems and radio local area networks in the
frequency range 300 MHz to 100 GHz, Geneva, 2005a.
[ITU, 1546] International Telecommunication Union, ITU-R Recommendation P.1546-2:
Method for point-to-area predictions for terrestrial services in the frequency
range 30 MHz to 3 000 MHz, Geneva, 2005b.
[Kozono, 77] S. Kozono and K. Watanabe, Influence of environmental buildings on UHF land
mobile radio propagation, IEEE Transactions on Commununications, 25 (10),
1133–43, 1977.
[Lee, 82] W. C. Y. Lee, Mobile Communications Engineering, McGraw-Hill, New York,
1982, ISBN 0-070-37039-7.
[Lee, 93] W. C. Y. Lee, Mobile Design Fundamentals, John Wiley, New York, 1993, ISBN
0-471-57446-5.
[Maciel, 93] L. R. Maciel, H. L. Bertoni and H. Xia, Propagation over buildings for ranges of
base station antenna height, IEEE Transactions on Vehicular Technology, 42 (1),
41–45, 1993.
Macrocells 185

[Okumura, 68] Y. Okumura, E. Ohmori, T. Kawano and K. Fukuda, Field strength and its
variability in VHF and UHF land mobile radio service, Review of the Electrical
Communications Laboratories, 16, 825–73, 1968.
[Parsons, 92] J. D. Parsons, The Mobile Radio Propagation Channel, John Wiley & Sons, Ltd,
Chichester, ISBN 0-471-96415-8 1992.
[Saunders, 91] S. R. Saunders and F. R. Bonar, Explicit multiple building diffraction
attenuation function for mobile radio wave propagation, Electronics Letters,
27 (14), 1276–77, 1991.
[Saunders, 94] S. R. Saunders and F. R. Bonar, Prediction of mobile radio wave propagation
over buildings of irregular heights and spacings, IEEE Transactions on
Antennas and Propagation, 42 (2), 137–44, 1994.
[Walfisch, 88] J. Walfisch, and H. L. Bertoni, A theoretical model of UHF propagation in
urban environments, IEEE Transactions on Antennas and Propagation, 36
(12), 1788–96, 1988.

PROBLEMS
8.1 Use the Okumura–Hata model to compare the maximum cell radius available from
systems having maximum acceptable path loss of 130 dB in urban and suburban areas at
900 MHz and 1800 MHz.
8.2 Use the flat edge model to predict the range corresponding to a path loss of 110 dB
at 900 MHz. Assume hb ¼ 15 m; h0 ¼ 10 m; LE ¼ 15 dB; w ¼ 40 m and hm ¼ 1:5 m.
8.3 Repeat the previous example using the Okumura–Hata suburban model.
8.4 If two co-channel base stations with omnidirectional antennas are separated by 5 km,
calculate the minimum value of signal-to-interference ratio encountered, assuming a cell
radius of 1 km and negligible directivity in the elevation plane.
8.5 Repeat Problem 8.4, assuming that the antennas are 15 m above level ground, with the
elevation pattern of a vertical half-wave dipole.
8.6 Discuss the limitations of the Okumura-Hata model, assuming that the frequency range at
which has been validated is appropriate.
8.7 Estimate the range of a communication system operating in the GSM 900 MHz band
using the Lee macrocell model for a suburban area, if the base station is assumed to have
an effective height of 25 m. Make appropriate assumptions for any parameters not
provided.
8.8 Prove Eq. (8.22)
8.9 Repeat Problem 8.7 using the flat edge model, making use of the approximate formula-
tion (8.19).