1/41 Document Type Document Title

Author Date, Version

Unit/Dept. For internal use only

Guideline for Propagation Modelling with NetAct Planner

Version Date Author

1.0 28.9.2007 KSu
2/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Table of Contents
1. Overview............................................................................................ 3

2. Considerations................................................................................... 5

3. Digital Maps ....................................................................................... 5

4. Models ............................................................................................... 7
4.1 Standard Macrocell Model ............................................................................................... 7
4.2 Nokia propagation model ................................................................................................. 9
4.3 SUI Model for WiMAX .................................................................................................... 10

5. Model Tuning Process ..................................................................... 11

5.1 Model Tuning Strategy ................................................................................................. 13
5.1.1 Model tuning for coverage ............................................................................................. 13
5.1.2 Model tuning for interference ......................................................................................... 14

6. Preparation ...................................................................................... 16
6.1 Site Location and Route Planning ................................................................................. 16

7. Measurements ................................................................................. 22
7.1 CW signal measurement ............................................................................................... 22
7.2 Modulated signal measurement .................................................................................... 24

8. Model Tuning and Analysis ............................................................. 26

8.1 Parameter Selection ...................................................................................................... 30
8.2 Morpho correction auto tuning ....................................................................................... 38

9. Report preparations ......................................................................... 41

3/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

1. Overview
Terrestrial radio networks are being designed with tools using simulators or statistical
propagation models for predicting the radio coverage and interference over the target
areas. There are numerous different propagation models. Most of the ones used in cellular
telecom field coverage predictions are either semi-empirical models or their derivatives (like
the Nokia Model that is based on Okumura-Hata) or they are based on different Ray-tracing
algorithms. In this document Nokia Model is selected, but also standard model and SUI
model is presented..
Traditionally propagation modelling has been mostly done by using empirical propagation
models. Common for all models is the usage of environmental database. Typically, this
database contains information about land usage and terrain profile but sometimes a
separate street data and even detailed building data is used to gain better results in urban
areas. These models are simple as there typically is rather simple distance dependency
guided by certain factors and clutter offsets.
Most models, including the Nokia (Okumura-Hata) Model, need to be tuned before it can be
used for accurate predictions. The tuning takes the field measurements, a corresponding
transmitter configurations and tries to optimize the model parameters so that it estimates
propagations as close to measured ones as possible. The actual tuning is an iterative
process and makes use of computer algorithms for optimization of the parameters. With
careful propagation model tuning, these types of models can produce propagation
prediction with close to 0 dB average error and 6-8 dB standard deviation. If your prediction
vs. measured error’s standard deviation is considerably above 8 dB, you may consider
propagation model tuning as unsuccessful. This of course assuming that the area where
measurement results are gathered are reasonable, i.e. following normal guidelines of
propagation model tuning in terms of numbers and locations of measurement results. The
standard deviation of a measured signal depends upon the accuracy of the equipment,
which is used to measure the signal, the resolution of the terrain data and clutter. Also the
size of the area, which is used to average the binary samples (the minimum area depends
on the resolution of terrain and clutter data).
With propagation models we should predict the actual radio wave propagation as accurately
as possible. Radio wave propagation in reality is of course determined greatly on free
space loss added by shadowing effect due to environment. This means that path loss slope
has to be at least 20 dB/dec, typically in urban environment 30-40 dB/dec. (Note that 20
dB/dec path loss slope in logarithmic scale equals path loss exponent of 2 in linear scale,
i.e. free space loss). Figure 1 illustrates the effects of K parameters in Macrocell models in
NetAct Planner.
4/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 1 Basic propagation guidance parameters in Macrocell models in NetAct

Planner are shown above.

In model tuning process values of K factors should be found for the environment. K2 and
K6 parameters should have values such that the path loss slope result is according to
typical values given above. If K2 and K6 parameters have such values that resulted path
loss slope is under 30 dB/dec, it should be carefully considered if the result is too optimal
for the environment. Low path loss slope will result in optimistic coverage predictions but
pessimistic simulation results in terms of inter cell interference.
5/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

2. Considerations
A special attention has to be paid when configuring the test equipment. The detailed
description of test transmitter and receiver setup needs to be followed for the path-loss
calculations. If the equipment is not setup correctly, the tuning may give a systematic error
for all the corresponding measurements thus compromising the success of the whole
process. This is also good time to start writing down notes.
The measurements are playing a very important role in model tuning. Bad quality at this
stage can lead to misleading data in actual tuning process and thus to a poor propagation
model. Minor short-time problems have no great impact on overall model, but a poor
equipment setup might ruin the chances to good quality models. In CW measurements, it
will take some time to build up all the settings and devices, but it will absolutely be worth it,
because results are more reliable. Minor problems have only little impact on results, but if
we have several little impacts, the total impact on the results can be huge. It is really
recommended, that only experts will do model tuning.
Next sections cover the following considerations
• Digital maps
• Propagation models
• Preparations
• Measurements
• Model tuning and Analysis

3. Digital Maps
The digital map provides information of the environment to the model. The quality of the
map is proportional to the quality of the resulting models. There are several issues, which
are affecting the map quality. A few critical factors are given below.

• The higher pixel resolution of the map does not necessarily mean a better map. For
example a lake that is correctly present in a 50m resolution map but it is missing
from the 5m map is making the 50m map a winner in that area regardless of its
lower resolution.
• In addition, it is good to know, how old the map is, when it was last time updated.
Infrastructure is changing all the time, new residential areas are born so the radio
network in changing all the time. Propagation model is a vital part of radio network,
so it is also as important to update existing data, maps etcFigure 2.
6/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

1995 2007

Figure 2 Digital maps has to be updated or the predictions will be wrong

 The clutter (land-usage) data should have categories those are differing from each
other while looking from the radio-signals point of view. The categories themselves
do not add or subtract extra accuracy, since the model is tuned with these particular
clutters Figure 3.

Figure 3 Even the resolution is same maps are totally different. Clutters have to
differ from each other from the radio signal point of view to get more accurate
results. Same model for these maps is not recommended.

 The resolution of topographic terrain model is typically much higher in vertical

direction. It is not useful to use of a map that has a 50m xy-resolution and 50m
vertical resolution. The vertical resolution should be typically between 1-4m. If
the resolution of the digital map is good enough and there is not enough of
different clutter, the meaning of the height database is not so important. If the
resolution of the map is not good enough, the right height information is more
important. For example over 5 m differences in antenna height affects on model
results really heavily. This 5 m difference can easily be generated in real life and
the effects on results are bigger if the antenna height is below rooftop.
 If it is known that there are some inaccuracies with used digital maps, it is still
really important to do measurements as accurate as possible, because these
same measurements can be reused later when more accurate digital map exists.
In addition, if some factors are inaccurate, it is important to try minimizing other
inaccurate factors.
7/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

4. Models
Propagation models are mathematical attempts to model the real radio environment as
closely as possible. Most propagation models need to be tuned by being compared to
measured propagation data, otherwise you will not be able to obtain accurate coverage
predictions. Few most important models are described here. There are also some other
models for some specific purposes, for example microcell model.

4.1 Standard Macrocell Model

There are three Standard Macrocell models available, 1, 2 and 3. The Standard Macrocell
models incorporate an optimal dual slope loss model with respect to distance from base
station. They also incorporate algorithms for effective base station heights, diffraction loss
and the effect of clutters.
The General formula for Macrocell models is as follows Equation 1.

Prx = Ptx – Ploss

Equation 1

,where
Prx = Received power(dBm)

Ptx = Transmit power(dBm)

Ploss = Path loss(dBm)

And Equation 2

Ploss = k1 + k2log(d) + k3(Hms) + k4log(Hms) + k5log(Heff) + k6(Heff)log(d) +

K7(diffn) + C_loss
Equation 2

Where: (see also Figure 1, above)

d Distance from base station to the mobile
Hms Height of the mobile station above ground(m). This figure may be specified
either globally or for individual clutter categories
Heff Effective base station antenna height(m)
diffn Diffraction loss calculated using either Epstein, Peterson, Deygot or Bullington
equivalent knife edge methods.
k1 and k2 Intercept and slope. These factors correspond to a constant pathloss offset
(dBm) and a multiplying factor (k2) for the log of the distance between the base
station and mobile Figure 4.
8/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

slope

intercept

Figure 4 Slope and intercept. These factors correspond to a constant pathloss

offset (dBm) and a multiplying factor (k2) for the log of the distance between the
base station and mobile.

k3 Mobile antenna height factor. Correction factor used to take into account the
effective mobile antenna height.
k4 Okumura-Hata multiplying factor for Hms. Compensates for gain due to mobile
height
k5 Effective antenna height gain. This is the multiplying factor for the log of the
effective antenna height.
k6 Log(Heff)Log(d). This is the Okumura-Hata type multiplying factor for
log(Heff)log(d)
k7 Diffraction. This is a multiplying factor for diffraction calculations. A choice of
diffraction methods is available.
C_loss Clutter specifications such as heights and separation are also taken into
account in the calculation process.

The propagation model can be tuned by modifying the k-factors. For improved near and
far performance, dual slope attenuation can be introduced by specifying both near and
far values for k1 and k2 and the crossover point. Typical parameter values can be seen
below Figure 5.

Note:If you are using clutter heights, you must specify both separation and a clutter
height per clutter.
9/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Parameter 450 MHz 900 MHz 1800 MHz 2000 MHz

Mobile Rx Height 1.5 1.5 1.5 1.5

Earth Radius 8493 8493 8493 8493

k1 142.3 150.6 160.9 162.5

k2 44.9 44.9 44.9 44.9

k1 (near) 129.00 0.00 0.00 0.00

k2 (near) 31.00 0.00 0.00 0.00

k3 -2.60 -2.55 -2.55 -2.55

d< 0.00 0.00 0.00 0.00

k4 -0.8 0.00 0.00 0.00

k5 -11.70 -13.82 -13.82 -13.82

k6 -4.30 -6.55 -6.55 -6.55

k7 0.4 0.7 0.8 0.8

Effective Antenna Relative Relative Relative Relative

Height (Heff)

algorithm

Diffraction Loss Epstein Peterson Epstein Epstein Epstein

algorithm Peterson Peterson Peterson

Merge knife edges Leave unaltered Leave Leave Leave unaltered

closer than unaltered unaltered (0.00)
(0.00) (0.00) (0.00)

Dual slope 0.0m effectively No No No

making it single
slope

Clutter parameters N/A N/A N/A N/A

Use clutter heights No No No No

Use clutter separations No No No No

Use mobile heights No No No No

Figure 5 In this table suggested default parameter could be seen. These values
represent sensible starting values for the Standard Macrocell models based on an
urban environment.

Note that these factors define the basic propagation; the values given here are to be
considered as guiding values only! User should be careful when adjusting these
parameters. These values should not change very much and model tuning should
mostly be affecting on clutter offsets.

4.2 Nokia propagation model

The Nokia propagation model comprises five parts:

• The basic path loss model

• Line-of-sight checking
10/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

• Calculation of the base station effective antenna height

• Topography correction
• Morphography corrections

Most of these parts have a set of selectable correction functions with user definable
parameters. This, and the fact that each cell can have a unique model, enables you to
specify a suitable model for each propagation environment.

In this document Nokia model has been selected for detailed analysis.

4.3 SUI Model for WiMAX

The SUI (Stanford University Interim) propagation model is one of the recommended
models for use in analyzing fixed wireless broadband communication networks such as
WiMAX at frequencies below 11GHz. The SUI models are divided into three types of
terrains1, namely A, B and C. Type A is associated with maximum path loss and is
appropriate for hilly terrain with moderate to heavy foliage densities. Type C is
associated with minimum path loss and applies to flat terrain with light tree densities.
Type B is characterised with either mostly flat terrains with moderate to heavy tree
densities or hilly terrains with light tree densities.
Multiradio Planner enables you to set up WiMAX propagation models based on the SUI
standard.

The SUI model is an empirical model governed by the basic path loss equation as
follows Equation 3, Equation 4:

For d > d0

Equation 3

For d < d0

Equation 4

Where
11/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

The SUI model in MultiRadio Planner can also account for diffraction loss and clutter
loss. Therefore the above equations can be modified as follows Equation 5, Equation 6:

For d > d0

Equation 5

For d < d0

Equation 6

Where Ld is the Diffraction Loss Scale Factor.

The Mobile RX Height and standard deviation of interference values are not used in the SUI
model calculation. There is a value you must specify for Customer Premises Equipment
(CPE) height.
Note: In NetAct Planner It is not possible to use the CW measurement functionality with the
SUI model.

5. Model Tuning Process

The diagram below Figure 6 shows the process flow of model tuning. This process is
suitable for all frequency measurements. Models and model parameters by themselves are
different for different frequencies.
12/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Start

Preparation

Measurements

Model tuning

Analysis

Report writing

End

Figure 6 The process flow of the whole model tuning process is shown above. This
process can be used for example for 2G and 3G model tuning measurements.

Start
At the start point the model tuning strategy is defined. This is also the point where big lines
are drawn, is it possible to measure out own site or do we have to use existing site. How
much time can be used with this model tuning project etc. Preparations are based on that
strategy which is defined here.
Preparation
This part consist the analysis of the digital map and defining the test transmitter locations
and the routes to be driven. Configuring the measurement equipment and verifying the
functionality. Preparation is playing very important role and special attention must be paid
on this part. It is assumed that before preparation all the cable loss attenuations are
measured and power parameters / antenna gains should be known. If these are not known,
these should be done before preparations.

Measurements
13/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Measurements will be done according to the agreed drive test plan. Notes will be made
during the measurements for later use during the tuning process. LOS and NLOS samples
must be known during the measurement. The purpose is that additional measurements
are not needed. However, if more measurements are needed, it recommended to
start measurements from the start and preparation points. After that, these phases are
even more important.

Model Tuning & Analysis

The tuning phase is the most time consuming part of the process. Deep understanding of
all variables related to the desired model and errors in the source material is required. If
analysis with existing samples is not accurate enough, it is recommended to investigate
what is causing problems, meaning the whole process will start again from the start point,
as mentioned before.

Reporting
Information about the tuning and each new model must be extensively documented. This is
the output of the project and it is delivered to the customer. The model will be used as it is
documented. Using the wrong model in wrong purpose might lead to great errors in
predictions thus lowering the quality (reliability) of the models tuned.

5.1 Model Tuning Strategy

Before the model tuning measurement, it is good to define, what is this model for, is it for
coverage analysis or is it for interference analysis? Please notice that even the model is
tuned well, standard Equation 7 deviation is 6-8dB. If only one model is purpose to use,
total deviation calculation should be taken into account more seriously, when for example
coverage thresholds are defined.

1 2

Equation 7 For statistically independent parameters the total standard deviation is

calculated as above.

Use of one model requires a great knowledge of model tuning and network planning in
order to avoid negative surprises, for example bad coverage dip areas and interference
areas. These problems can be very expensive to repair later, lots of new sites or other
changes will be needed and it will cost a lot of money.

5.1.1 Model tuning for coverage

If the main purpose for propagation model is to predict coverage and particularly coverage
dips, it is recommended to weight bad signal level samples while doing model tuning. The
signal level can vary even over 30dB within the one pixel, which is much for example
compared to standard deviation. In figure below Figure 7, one example is shown. These
14/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

points (point1, 2, 3) mentioned in the figure does not mean that there is only one sample,
which differs from predicted value. This means more like area, amount of averaged
samples, which are weaker than predicted model values.

Signal levels Point2: Signal level is over 12dB w eaker

point1:-89dBm compared to predicted model. This coverage problem area
Point2:-95dBm can ’t be found, because model is not tuned for the
Point3:-72dBm “ dip ” values.
Drive test route

point1

-70dBm
point2
point3
-100dBm
Point3 -85dBm
-110dBm

Figure 7 If the coverage and coverage dips are main concerns in radio network
planning model for coverage is recommended to use. Real life example can also
be seen from figure7.

Coverage weighted model is also recommended to use if the radio network should work
also in some special area, for example in the road which is located in the bottom of the
canyon. If the model is not tuned for the coverage dips, here for example for canyon
coverage, it might be that after site places are selected, almost the whole canyon road is
out of coverage. After that, it is expensive to repair the coverage. If the site places had
selected for this canyon coverage (based on coverage-weighted model) the radio network
would have been better and many sites would have been saved. Many times, it is really
worth thinking what the main purpose for the model is.

5.1.2 Model tuning for interference

If the main work for network planning is to do optimising and interference analysis, it is
recommended to weight more peak values while doing model tuning. In C/I analysis I is as
important as C. Coverage is not the main issue any more. Long distance measurements
are strongly recommended to make sure that slope, parameter C, for the model is also
working in long distances Figure 8.
15/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Signal levels Point1: Signal level is over 15dB better

point1:-95dBm compared to predicted model. This interference
Point2:-107dBm can ’t be found, because model is not tuned for the
Point3:-101dBm “ peak” values.
Drive test route

point1

-70dBm
point3 -100dBm
point2
-85dBm

point3

-110dBm

Figure 8 If the interference is the main concern in radio network planning model
for interference analysis is recommended to use. Peak values are more weighted.
Real life example can also be seen from figure7.

Below is one example how the propagation model is working and what are the critical parts
when coverage analysis and interference analysis are made. Coverage picture is from the
planning tool and road coverage is based on drive tests. If the model is too optimistic, there
might be some coverage problem areas. Some coverage problems area can be seen from
the picture below Figure 9. In some points, the model is pessimistic. In these kinds of
cases, the interference analysis is difficult to do.
16/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

>= -60dBm

-60 to -70dBm

-70 to -80dBm
-80 to -90dBm Interference analysis
-90 to -100dBm

-100 to -110dBm

coverage analysis

Figure 9 Coverage from planning tool vs. drive test measurements. Colors should
be as equal as possible. In some areas the model is not working well so the
coverage analysis and the interference analysis in the planning tool are difficult to
do.

6. Preparation

6.1 Site Location and Route Planning

Site selection is playing a very important role in model tuning and special attention has to
be paid when the site place is selected. Particularly if the digital map used is not very
accurate, site location and route planning are playing really important role.

Main important factors for site location and route planning are given below.
• Antenna installation, no obstacles near the antenna
• Measurement sector must be defined.
• Antenna height as close typical antenna height as possible
• Site location, also measurements far away should be able to do
• Measurements must be done as versatile as possible
• Also narrow streets must be measured
17/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

• Enough total samples and enough samples in every clutter

• Angle distribution must be as constant as possible
Meaning of these bullets are described deeper, see below paragraphs.

Site place should be selected so that the antenna can be installed properly. If old site
places are used, only these site places should be selected which are optimal in terms of
radio network planning. There shouldn’t be any obstacles (for example roof) near the
antenna which could block or attenuate the radiated signal and thus distort the results.

It is preferred to take some photos from the antenna place. Documentation is very
important part of model tuning. If documentation is well done, data is more reliable and
measurements can be reused later. Please write down as much notes as possible and
start it from the very beginning of model tuning project.

It is also recommended to select site place so that measurements over 10 km can be

easily done. If measurements are done only near the site (<3km) may be the model
(related to parameter C, dB /dec) is not working well when the distance is over 10km. If
the C value is not optimised properly, interference analysis, meaning many times
samples in long distances, can’t be done accurate enough. This is the main reason,
why it is very important to measure lots of samples also over 10km or even over 20km.
It is also important to notice that there are enough combinations of clutter types (in
selected distance parameter) when the measurement distance is 10-20km.
For example In Finland there is lots of houses which are located in the countryside just
near the lake, onshore. Many times the signal is coming from the nearest mast, which
is located over 10km from the houses. For example in these kinds of cases, it is really
important that the model is working well in the long distances. When new masts are
planned to the countryside, house coverage is a really important part of planning. If
there are lots of coverage problems in the residential area (due to the bad model) after
new mast was built, it is almost impossible (or really expensive) to repair it.

It is also recommended to define a measurement sector (in the main lobe of antenna)
where all the measurements related to the certain cell are done. If the antenna is
located in the rooftop, it might be that in some direction there are some obstacles near
the radiated antenna. If measurement sector is defined, we can be sure that results are
more reliable, there are no any obstacles, for example roof, causing distortions, see
figure below Figure 10.
18/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 10 Measured antenna (black) and measured sector. Results are more
reliable, if model tuning measurements are measured in the error free sector.
This is playing more important role, if antenna is located for example in the
rooftop.

The height of the selected reference site should be as near the typical local site height
as possible. Particularly this is very important if the map used is not very accurate. 5m
differences in height can affect on results really much, particularly if low antenna heights
are used. Please note that the error may not occur due to incorrect antenna heights but
due to inaccuracies of the height profile.

The transmitter site should fulfill the optimal requirements for radio network planning
o Optimal height (as typical height as possible)
o Clearance of obstacles. Obstacles in the very proximity of the antenna
should be avoided and measurements should be made only inside the
defined sector
o Site should not be on top of a hill or in a dip. It is advisable to select sites in
a flat region.
o The area to be driven around the site should cover the maximum number of
clutter areas and different clutter combinations. Models should work
everywhere, so all the clutters with enough samples should exists. Distance
parameter defines which pixels (clutters) are observed.
o If the model is for certain purpose, for example to urban environment, only
that certain area with certain clutters should be measured. Normally model
is for the big area / whole country.
o All the elements of the transmitter should be known. It is recommended to
use omni antenna with measured radiation pattern. If existing site is
measured, radiation patterns should be measured.

Site places should be selected so that the routes around the sites can be selected as
versatile as possible. Different angles in every type of cluster to measured sites should
be quite constant, meaning that there should be enough samples in every angle where
different cluster types exists between site and measured point, see figure below Figure
11.
19/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

measurement angle to measured Site

5000
4500
4000
3500
samples

3000
2500
2000
1500
1000
500
0
<10 degrees

>10 and <20 degrees

>20 and <30 degrees

>30 and <40 degrees

>40 and <50 degrees

>60 and <70 degrees

>70 and <80 degrees

>80 and <90 degrees

>100 and <110 degrees

>110 and <120 degrees

>120 and <130 degrees

>130 and <140 degrees

>140 and <150 degrees

>150 and <160 degrees

>170 and <180 degrees

>50 and <60 degrees

>90 and <100 degrees

>160 and <170 degrees

angle

Figure 11 There should be enough samples and angle distribution should be as

constant as possible.

Zero angles means that we are driving toward the site and angle 180 degrees means that
we are driving away from the site. If there are too much samples near zero or near 180
degrees, the results will not be accurate enough, because the cluster type “way or road,
meaning like open” would be dominant. From figure below the meaning of angle can be
seen Figure 12.
20/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

SiteA

Signal is coming Signal is coming

from this from this direction
direction

SiteB

Figure 12 It is really important to select site place properly. Otherwise the results
are not accurate enough. If the signal is coming from SiteA, measured cluster type
is more “road” than industry area. If the signal is coming from the SiteB, measured
cluster type is more industry area than “road” Due to these differences cluster type
analysis and angle analysis should be made to be sure that results are reliable.

If the signal is coming from the SiteA and we are measuring roads in this industry area.
Maybe the results are not accurate enough, because actually we are measuring cluster
type “road” more than cluster type industry area, because there are only roads between the
measured point and site. If the signal is coming from the SiteB, the situation would be
different, because now there are houses and other buildings between the measured point
and measured site. And the angle to measured site would be also different. Now there
would be other angle samples than near 0 or near 180 angle samples. If the resolution of
the map is not very accurate there might be over 30dB differences within one pixel. That is
the reason, why angle distribution and amount of samples are playing really important role.
The results will be more reliable

Quite often the measurements are done using very wide main streets. This is the quickest
and easiest way to gather the data. However, it is not the most accurate way to measure,
because propagation in the wider streets is not the same as the propagation in the narrow
streets. Measured route should be as versatile as possible.

Usually omni antennas are used in measurements. The omni antenna features a radiation
pattern which has a perfect circular symmetry over the xy plane Figure 13. Directional
antennas have a radiation pattern which presents a certain number of “lobes” the antenna
gain can thus be rapidly variable across the measurement angle. Such variation is
practically unpredictable and may constitute a source of inaccuracy for the measurements
and for the model tuning.
21/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 13 Usually omni antennas are used in measurements which have a perfect
circular symmetry over the xy plane. In this antenna side lobes are occurring.

In real life clutter types varies very much. For example the industry area may vary very
much, some buildings are higher than others and in some points there is even open. Due to
this special attention has to be paid, when site and route planning is done. It is vital that
there are enough samples and there are enough samples in every cluster and the angle
distribution in every cluster type is quite constant. If the maps are possible to get before
measurements, the cluster type analysis should be made and based on this site places will
be selected and route planning will be done.

The measurement route should be planned so, that the route within the defined sector
would be as versatile as possible, see the principle of the route from the figure below
Figure 14. The angle distribution should be as constant as possible. The route should
include narrow wide streets also, not only main streets.

Figure 14 Measured route should be as versatile as possible within the sector.

There should be at least 6 reference sites to measure and the route should cover all the
clutters. Of course this depends on the area the model is planned. If it is planned for the
whole country, there should be more sites to measure. It is really important that there are
enough samples so it is better to measure too much than too less.

Preparation is a really important part of model tuning. If route plan is properly done, the
measurement route is long enough, meaning there is enough samples in every clutter type
22/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

and in every angle and the distribution of different clutter type combinations in good
enough. After this the model tuning is easier to do and the accuracy of the result is better.

7. Measurements

The propagation model measurements are normally done using very accurate and
calibrated measurement tools, spectrum analyzer and signal generator. Professional test
equipment has an accuracy of ± 1dB. If other tools are used, for example existing site with
existing antenna, Nemo kit and scanner etc, these devices should be measures very
carefully. Otherwise it is impossible to know, are differences between measured and
predicted sample due to the inaccuracies of map, inaccuracies of devices etc. Right and
stable devices should not take for granted, there might be some inaccuracies, which are
really difficult to find (connectors etc) It is always recommended to use only calibrated
measurement devices.
It is also important to measure frequency, which is as clean as possible. Neither same nor
adjacent frequencies should exist near, otherwise results can be ruined. DL Power control
is not used during model tuning measurements. GPS is used to get location information.
Typical measurement configuration can be seen below Figure 15.

Transmitter

Figure 15 Typical measurement configuration.

7.1 CW signal measurement

During the measurements CW signal or modulated signal is measured. CW signal is a
narrow band signal; in the modulated signal bandwidth is wider. If the narrow CW signal is
used, the noise level is lower so measurement in the lower signal level can be done. This
means that the range of measurements is wider and results are more accurate. CW
measurements are little more time consuming, because own transmitter is used.
Measurements are playing really important role, so budgetary or resources can’t be limited
factors if accurate model is needed.
23/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

If the narrow CW signal is used, also fast fading is greater. This has to be taken into
account when measurement samples are averaged. Measurements should be made based
on Lee Criteria.
Lee Criteria says that samples have to be uncorrelated. They can be seen uncorrelated if
they are at least 0.8λ apart. On choosing 36 samples for 40λ the samples are 1.1 λ apart
from each other. This conclusion is known as a Lee Criteria. If the frequency measured is
900MHz, then λ is approximately 0.33m/cycle. Lee Criteria is fulfilled if 36 samples are
taken over 40λ (13.2m). If the speed of vehicle is 100km/h (27.8m/s) then samples have to
be taken every 0.0132 second ((13.2m / 27.8m/s) / 36). Note that if 500ms sweep time is
used and the speed of the vehicle is 100km/h, the distance the car will go during this 500ms
is approximately 40 λ. If sweep time is used samples are averaged over the time and
distance and effects of the fading are eliminated.
Here are some examples of settings when CW signal is measured using signal generator
and spectrum analyzer.
Settings for the spectrum analyzer:
The centre frequency should be the same as the transmitter frequency.
For a bandwidth narrower than 10 kHz:
• Span = approximately 2.04*bandwidth
• Resolution bandwidth (RBW) = 300Hz
• Video bandwidth (VBW) = 1 kHz
• Sweep time = 500 ms
• Attenuation = 0 dB
• Reference level = -30 dB

If bandwidth is wider that 10 kHz:

• Span = approximately 2.04*bandwidth
• Resolution bandwidth (RBW) = 3 kHz
• Video bandwidth (VBW) = 10 kHz
• Sweep time = 500 ms
• Attenuation = 0 dB
• Reference level = -30 dB

The analyzer does not give accurate results if you operate in bandwidth narrower that 0.3
kHz. Practically Doppler spread increases the minimum bandwidth to 0.75 kHz if we
assume maximum speed of the measurement vehicle during measurements to be 120
km/h. This velocity would result in Doppler shift of approx. 0.2 kHz at 1900 MHz carrier
frequency, when driving directly towards or away from the transmitter, see figure below
Figure 16.
24/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 16 Principle of Doppler Shift. The mobile receives signals from a number of
different directions. Therefore each signal will experience a different Doppler Shift.

On the other hand, using too wide bandwidth increases the amount of received noise.
When the bandwidth doubles, the noise level rises about 3 dB. Hence, the used
bandwidth should be sufficiently narrow to guarantee reliable results. Bandwidth of
about 5 kHz is recommended. It can be safely utilized without fear of results distorted
by Doppler spread at vehicular speeds, and the noise level should remain sufficiently
low (approximately at -118 dBm). Narrower bandwidths are also possible, but it is
strongly advised not to use narrower bandwidth than 1 kHz.

7.2 Modulated signal measurement

In the modulated signal the bandwidth is wider, so the noise level is higher compared to
CW signal measurements Figure 17. For example the noise level in the UMTS can be -
104…-105 dBm, so the minimum received signal level is -95…-100 dBm (if demodulation is
not used).

Bandw idth

Figure 17 In the modulated signal the bandwidth is wider compared to CW signal.

Effects of the fast fading are also stronger in the CW signal measurements.

GSM BCCH is also modulated signal. If existing Base station is measured, it is important
that all the cables and output power are measured. If existing antenna is used, exact
antenna radiation pattern should be known and it should be verified that there is not any
obstacles near the antenna.
During measurements, CW or modulated signal , it is recommended to make some notes, if
there are some samples or any distortions, which will affect to the results (LOS samples,
tunnel samples, some corrupted data samples etc). In addition, some points where the
25/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

signal level is varying very strongly within one pixel of the digital map should be written
down, because later these can verify from the results. For example in the industry area,
where the heights and sizes of the buildings are varying really much, within one pixel (if the
resolution is for example 50m) the signal level can also varying really much, even more
than 30dB Figure 18. Due to these differences, the model tuning can be difficult.
Particularly if the map is not very accurate, these notes are playing more important role.

Measured Signal levels:

-60dBm
-96dBm

-64dBm

One 50m x 50m pixel,

yellow square.

Figure 18 Within one pixel of digital map, the signal level can vary even more than
30dB. What is the value of the pixel, when model is tuned? As far as interference
analysis or coverage analysis is concerned, these differences have to be known
and have to be taken into account. One average model can cause negative
surprises, if these differences are not known.

After all the measurements have been done, model-tuning process can be started.

NetAct Planner is using measurement data as follows. If Spectrum analyzer is used, this
format should use as an output from analyzer Figure 19, Figure 20.

17.068671 -22.57308 -40.19541168 Start

17.068678 -22.5731 -41.6576767
17.068683 -22.57312 -46.55962753
17.068686 -22.57313 -39.59311676
17.06869 -22.57315 -44.63028336 Finish

Figure 19 Example of .Dat -file format, NetAct Planner is using. This contains
coordinate information and signal level information.

DATA_FILENAME Filename1.dat
SITE_ID 1
SITE_NAME Sitename1
SITE_GRHEIGHT 100
SITE_LONGITUDE 17.06782318
SITE_LATITUDE -22.573746
SITE_TYPE Mast
26/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

TX_AZIMUTH 0
TX_TILT 0.0
TX_HEIGHT 25.0
TX_POWER 43
ANTENNA_TYPE 738450\738450
FEEDER_TYPE Feeder0
FEEDER_LENGTH 0
CONNECTOR_LOSS 0
FREQUENCY 955
OPERATOR Nokia
COMMENTS
DATE_START 5/9/2007
TIME_START 10:43:42
DATE_END 5/9/2007
TIME_END 12:43:42

Figure 20 Example of .HD file, NetAct Planner is using. This contains information of
reference Site

Based on these formats, logfiles can be imported to NetAct Planner and model tuning can
be started.

8. Model Tuning and Analysis

The propagation model adopted in most of cases is the modified Okumura-Hata model.
There is also some other models which could be used, but in most of cases Okumura-Hata
is used. Here is small comparison between models.
When the propagation models are compared, three major differences which affect the use
of the model can be found. These key parameters are frequency, distance between BTS
and MS and the antenna height range (both mobile and base station). The following graphic
shows the differences between models Figure 21.
27/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Frequency band

Cell range

BS and MS antenna ranges

Figure 21 Comparison between Okumura-Hata, Juul-Nyholm and Walfish-Ikegami can

be seen here. Okumura-Hata model is very suitable for normal cellular network
frequencies. The cell size is also suitable for macro cell modeling. Walfish-Ikegami suits
better for micro cell.

This Okumura-Hata model is mainly in used by Nokia. The traditional Okumura-Hata model
is based on a series of parameters (K1, K2, etc) to be tuned in order to have the model
matching the measurements.
The Nokia model is based on the use of a single constant “C” in the formula which
describes how fast the signal is attenuating with the distance.
In the propagation model wizard in NetAct Planner Nokia Propagation model should be
selected

Parameters to be optimized fall into two distinctive categories

• Morpho correction factors which can be optimized using the build-in parameter fit
function in Net Act Planner
• A number of options and parameters which can be set in the Propagation Model
Wizard. Some of these options can be considered as fixed, some others can be
changed iteratively during the tuning process.

A common practice is sometimes to have two or more propagation models for city area and
rural/village area. Such choice improves the efficiency and the accuracy of the radio
28/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

network planning avoiding to struggle to develop a model which is suitable both for big
cities and small villages. The practical process for tuning should follow the diagram below
Figure 22. Some of the options available in the propagation model wizard are to be
considered fixed:

• Model type: Okumura-Hata

• LOS check: Off
• Morpho correction: Function
• Topo: Diffraction

Some others are to be considered parameters, which can be iteratively changed

• Constant “C”
• Effective antenna height: Method A or B
• Calculation length
• BS and MS Weight Factors
• Diffraction weight factor
• Diffraction options: Largest obstacles or Deygout (N<=3)
On top of this, we have the morpho correction factors.
.
29/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Creating the new

model

Checking the
fixed options

Choosing values
for parameters

Optimizing
morpho
Choosing values correction factors
for morpho and parameters
correction factors

Standard deviation minimum

reached (6-8dBm).
NO

YES

Final result

Figure 22 Process flow of Model Tuning

30/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

8.1 Parameter Selection

In this section the tuning parameter of the Nokia model are described. Parameters and
particularly Morpho correction factors are playing really important role, when model tuning
is done. Parameters described here are:
• General setting parameters Figure 23
• Okumura-Hata parameters Figure 24
• LOS parameters Figure 26
• Antenna height parameters Figure 27
• Morpho correction parameters Figure 28 , Figure 29
• Topo correction parameters Figure 32, Figure 33

Figure 23 Tab “General”. General setting can be selected here.

In this first tab a name should be assigned to the model and the Okumura-Hata box should
be selected. The parameters in the lower part of the window refer to the Walfish-Ikegami
model and can thus be ignored. The Walfisch-Ikegami is not reliable, at least when
following the process and criteria’s here described.
31/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 24 Tab “Params”. For example, C parameters can be selected here.

The main parameter of interest here is the constant “C”. Suitable values can vary from 39 to
48. This parameter needs to be tuned manually trying different values until a good match is
reached. The value affects the steepness of the curve, which means the basic attenuation
due to distance Figure 25. For cities with high density of buildings and narrow streets, we
can expect higher values (43… 48). For countryside areas with few scattered houses, the
value should be lower (39… 42). It is therefore obvious that it in general preferred to have
more than one propagation model since it is quite challenging to come up with a single
model, which provides the same accuracy in dense urban area and in the countryside,
particularly if model is used for coverage analysis and for interference analysis.

Figure 25 The value C indicates what the basic average attenuation due to
distance is. It is important to discover that the parameter is also working when
measurements are done far away from the measured site. Only one C
parameter value can be used even normally propagation slope increases as the
cell coverage increases.
32/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

The box “City Type” has to be chosen according to the type of city we are dealing with. It is
anyway not relevant if the ms antenna height is 1.5m.
The parameters in the lower part of the window refer to the Walfish-Ikegami model and can
be ignored.

Figure 26 “LOS Check”. LOS check can be selected here.

As previously mentioned, the basic model is not accurate in condition of Line Of Sight
(LOS). Nevertheless, the model implemented is capable to take into account the LOS. This
is only possible, if such information is retrievable from the digital map in use. The map
should be supplied with the “building height” information. That means that the simulator is
capable of estimating whether the receiver is in condition of Line Of Sight from the RF
antenna. For relying on such feature anyway, we should have a very accurate map (2m-5m
resolution).
The preferred choice in normal condition is to select Off in this tab. During model tuning
measurements, it is recommended to make notes if LOS situation is occurring or not to
measure LOS situations. Preparation part is playing important role in model tuning.
33/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 27 “Eff.Height”. Effective antenna height method can be selected here.

The effective antenna height calculation is necessary in order to keep into account the
height gap due to the different ground height. If the terrain were completely flat, then the
suitable choice for this option would be constant. Such choice would make the prediction
faster. If the terrain profile has to be kept into account, which is the most common case,
then the preferred option is Method B. This is therefore one of the parameter which can be
iteratively changed during the tuning process.
In method A height is defined as the height of the base station antenna plus height
difference at both ends. Method B is based on the mean height of the ray line between MS
& BS above the ground. Okumura method specifies that the effective antenna height is the
difference between the base station antenna height and the average terrain height between
3 and 15 km
34/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 28 Morpho Correction method can be selected here.

Morpho corrections are based on the terrain type or usage profile generated for each path.
The preferred option is always Function. With this option, the morpho correction takes into
account attenuation of the signal due to the different clutter type area. Such factor can be
fed in the Morpho Correction Factors window by clicking on the button pointed from the
arrow. Function NLOS is otherwise equal to Function method, but line-of-sight situation is
rejected.
35/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 29 “Morpho Correction Factors”

In the picture above some typical value for correction factors are used. The correction
factors represent the different attenuation due to different clutter areas. It is important to
click the box “water” when the clutter type refers to a water surface (sea, lake, river, and
wetland). The suitable value for different clutter type can vary noticeably in models
developed for different cases and in different project. One critical aspect is the map vendor.
The quality and accuracy of the map is a critical issue in this case. A higher number of
clutter types can result in a better accuracy of the propagation model, but at the same time,
it implies a heavier work in terms of data collection and correction factor tuning.
Every map vendor might have different definition and criteria when it comes to classify
some areas for example as “urban” or “dense urban”. Likewise, for rural area we can have
“forest” or “dense forest” and so on. This is why the values in the pictures have to be taken
as a rule of thumb and their suitable value can be rather far from the one displayed here.
The box uses map data is to be left unchecked. This is an option to be chosen with certain
types of maps, which are not commonly used.
36/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Calculation Length
The Calculation Length value indicates how far from the receiving station clutter areas are
taken into account. Typical values vary from 250 to 2000. This is one of the parameters to
be iteratively tuned in a manual way over the tuning process. This is also the main
parameters, when combinations of clutter types are combined, how many clutters are taken
into account.

Figure 30 Calculation length and BS/ MS weights. The longer is calculation

distance; the lower should be BS end weight in comparison with MS end weight.

A good criterion is to start from a calculation length of 2000 m. Such criteria can take a
bigger portion of the terrain into account. If there are some problem areas, this parameter
can be the main reason for inaccuracy. For example if the site is located behind the lake
and there is thick forest between the lake and the car, see figure below. Now most of the
samples are lake samples, so the coverage predicted is really good even there might be
some coverage problems. If the defined distance would be 200m instead of 2000m, both
clutters would be weighted equally. Note that the longer is calculation distance; the lower
should be BS end weight in comparison with MS end weight!

2 km

Lake = 1.9km
forest = 0.1km
Figure 31 An example of inaccuracy of the distance parameter. Coverage is too
optimistic, because clutter lake is mostly calculated.
37/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

BS and MS end are weight factor for the calculation of the pathloss. Increasing BS End the
contribution for the attenuation in the proximity of the base station is taken into account
more heavily. What is taken into account is the ratio between the two values; which means
that choosing 1 and 2 as values has the same effect as to choose 2 and 4. Normally the
behaviour of the propagation environment in the proximity of the mobile station is to be
taken more into account. Which means that suitable value as a start could be:
• BS end: 2
• MS end: 1

Figure 32 “Topo” tab. Topo Correction can be selected here

The preferred choice in this case is Diffraction. This option takes into account the diffraction
which occurs when the mobile is shadowed by a hill.
The desired factor can be selected clicking the button
38/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

Figure 33 “TopoCorrection: Diffraction”

Suitable values for diffraction are in the range 40% - 80%. This value must be tuned
manually trying different values. If it seems that height differences, little hills or dips are
causing signal errors this weight factor parameter can be tuned. It must be remember that
such changes affect the results only when the measurements taken are passing behind an
obstacle in the terrain. The higher the value, the better the model can spot coverage dips. If
there are problems with this parameter, it is recommended to check that height of the BS
antenna is correct. Please note again that the error may not occur due to incorrect antenna
heights but due to inaccuracies of the height profile
It is convenient to try the effect of both options: Largest Obstacle and Deygout (N<=3).

8.2 Morpho correction auto tuning

Before model tuning can be started, measurement files have to be imported to the Planner.
To load measurement files select menu Tools -> CW Measurements Figure 34

Figure 34 Before tuning measurement files have to be exported to Planner

39/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

After loading, model tuning can be started. In statistical analysis, CW measurements are
compared to the values produced by the prediction model. When model is tuned, some
basics have to keep in mind.
• Change only one property at the time
• Perform analysis after each change

By clicking Analyse textual analyse of the comparison can be seen.

Figure 35 Auto tuning window. Correction factors and results can be analyzed here.

The tuning of morpho correction factor can be done automatically in NetAct Planner. This is
not anyway the preferred way to proceed. Sometimes; values which result from this process
are outside of the proper range. Two tricks can be adopted in order to avoid this:
• Defining some upper and lower limit for the desired range
• Deselecting all correction factor and leaving free one at the time
The best way to evaluate every correction factor is to tune it manually taking into account
measurements for which the vector between MS and BS crosses only the clutter area of
interest. So, values have to be tuned manually and evaluated after every change.
When values are tuned, it is good to know what this model is for. If the model is for
coverage analysis, it is important that coverage problems are taking more seriously
account. Mean error 0 is not always the optimised value; some margin of error is good to
have. More important is the standard deviation, it should be 6-8 dB.
40/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

If the model is for interference analysis, peak values (not LOS values) should take account.
Also samples in long distances should take account to make sure that C parameter is
working also in long distances, meaning many times interfering samples. One average
model can be very dangerous, unless all the errors are understood properly. During tuning
when the predicted model values are compared to measured propagation data, average
values are just for information that we are heading up to the better results. Problems must
be found out to make sure that the model is accurate enough and it can be used as part of
radio network planning.

After tuning results can be displayed in the 2D view Figure 36. Signal error display is really
useful to analyse if there are some areas where the model is not working accurate enough.
In some area, even the same clutters are taken account, signal error can be positive and in
some point it can be negative. This may happens if the map is not accurate enough and if
the signal level is varying very much within the pixel or clutter area. Also good notes can
help, if some inaccuracies were found in some certain area.

After measurements it is strongly recommended to make clutter analysis, angle analysis

and clutter combination analysis to make sure that there are data enough and the there are
enough different clutter combinations to solve these kinds of problems out.

Figure 36 CW data can be seen in the 2D view. Signal error analysis is really
useful to analyse where the biggest problems are occurring.
41/41 Document Type Document Title
Author Date, Version
Unit/Dept. For internal use only

9. Report preparations
The propagation model(s) must be submitted to the customer along with a report which
describes the main issues which has been dealt with in the tuning process. The constraints
for the model validity must be explained (no LOS etc.). If more than one model is submitted,
then the criteria of suitability must be remarked.
Even if the model is to be provided in XML format, a textual description must be presented
in the documentation. This is basically a list of all the parameters of the model.
Some examples of model matching with the measurements must be presented and a
reasonable explanation for the mismatches must be provided.
The model tuning report should include follows:
• Measurement configuration
• Used map information, resolution etc
• Measured log files and the route
• Results, standard deviation and mean errors.
• Model parameters for certain frequency
• Clutter area definitions
• Purpose of the model(s)
• Description – executive summary
• Conclusions and comments if there are some issues, which are affecting to the
model (inaccurate map, validity of the map etc). These should be mentioned in the
report.
If it is possible, it is recommended to arrange a meeting with customer to discuss about
the model.