Professional Documents
Culture Documents
i
Contents
Section1 Network Simulation Introduction.................................................................................................5
1.1 Overview of network simulation..........................................................................................................5
1.2 Network Simulation basics...................................................................................................................5
1.2.1 Digital Map ................................................................................................................................5
1.2.2 Universal Transverse Mercator (UTM) ......................................................................................7
1.3 Wireless Propagation Scenarios Definition ..........................................................................................8
1.4 Model Selection .................................................................................................................................12
-iii-
Section1 Network Simulation
Introduction
Knowledge point
General process of network planning and the main tasks in each phase-----level 12
However, network simulators are not perfect. They will NOT perfectly model
the network. They will, however, be close enough so as to give you a meaningful
insight into how your network is working, and how changes will effect its operation.
Simulator uses well-defined and easily obtainable map data formats that can be
ordered from all major digital mapping data suppliers. Conceptually, a pixel
represents the average value over the complete pixel. Each pixel is referred to using
the lower left co-ordinate. Slashes in index file paths can be either / or \ to allow
easy sharing of data with PC and UNIX based file servers. Although sub-folders are
supported with the paths, the up-one-directory symbol is not, for example,”010\..\”.
Height data is used to form the height information of the clutter, which is
formatted in grid database.
The height data - digital terrain model (DTM) is stored in a binary format
where each element of the data represents the height above sea level in meters for a
square area of, for example, 50m x 50m..
Clutter data is used to differ with the characteristics of the location, which is
formatted in grid database.
The clutter data is stored in a binary format with each element of the data
containing a code corresponding to a category of land usage for a square area of, for
example, 50m x 50m.
Line/Vector data contains features such as roads, railways, coastlines, and so on.
Separate mapping features must be in separate data files, however there can be more
than one file for a specific vector. Building vector data is a specialised version of
normal vector data, which contains a particular attribute called 'building' and
describes the shapes and heights of buildings in more detail, which is formatted in
vector data base.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-6-
WiMAX Network Planning
The DEM format is exactly the same as the DTM except that the height values
stored represent building height above ground level, that is, flat earth. Typically the
pixel sizes would be in the range of 1m to 10m, which is formatted in vector data
base and in grid database.
The UTM system divides the surface of the Earth between 80° S latitude and
84° N latitude into 60 zones, each 6° of longitude in width and centered over a
meridian of longitude. Zones are numbered from 1 to 60. Zone 1 is bounded by
longitude 180° to 174° W and is centered on the 177th West meridian. Zone
numbering increases in an easterly direction.
Each of the 60 longitude zones (Figure 1-3) in the UTM system is based on a
transverse Mercator projection, which is capable of mapping a region of large
north-south extent with a low amount of distortion. By using narrow zones of 6° (up
to 800 km) in width, and reducing the scale factor along the central meridian by
only 0.0004 (to 0.9996, a reduction of 1:2500) the amount of distortion is held
-7-
below 1 part in 1,000 inside each zone. Distortion of scale increases to 1.0010 at the
outer zone boundaries along the equator.
In each zone, the scale factor of the central meridian reduces the diameter of the
transverse cylinder to produce a secant projection with two standard lines, or lines
of true scale, located approximately 180 km on either side of, and approximately
parallel to, the central meridian (ArcCos 0.9996 = 1.62° at the Equator). The scale
factor is less than 1 inside these lines and greater than 1 outside of these lines, but
the overall distortion of scale inside the entire zone is minimized.
Among the terrain characteristics, the main factors affecting the radio signal
propagation are the terrain contour and terrain roughness. The degree of terrain
variation can be measured by the "terrain irregularity parameter" h , also called the
interdecile range. The interdecile range indicates the difference between 10% and
90% of the peak height in the vertical contour of the terrain (Figure 1-4).
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-8-
WiMAX Network Planning
From the prospect of radio wave propagation, various terrains can be classified
into the following main types based on the value of h , as listed in Table 1-1.
Interdecile Terrain
Range (m) Classification
Water surfaces
0-5 or very smooth
surfaces
Smooth
5-20
surfaces
Slightly rolling
20-40
surfaces
Rolling
40-80
surfaces
80-150 Hills
Rugged
300-700
mountain areas
Extremely
Greater
rugged mountain
than 700
areas
In the areas with densely distributed buildings, the radio wave propagation is
greatly affected by the clutter, especially by the sizes and distribution density of
buildings. In general, rural, suburban and urban areas are used to describe the
environmental features qualitatively. The rural areas refer to the open areas with few
buildings, and the forests and farmlands. The suburban areas refer to the areas with
sparse residential communities, parking lots, parks, and so on. The urban areas refer
-9-
to the areas with high-rise buildings, dense office buildings and commercial districts.
However, such qualitative descriptions are not very accurate and different people
may have different definitions on them. For example, an area can be described as
urban area if it is in a small city, while in a large city it may be described as the
suburban area. Therefore, it is necessary to describe the clutter features
quantitatively.
In 1977, Kozono and Watanabe tried to quantitatively describe the Tokyo urban
environment to research the influence of buildings on the mean received signal
strength. They proposed four parameters:
Building-occupying area factor
Building-occupying extended area factor
Building volume in the sampled area
Building volume in the extended area
The sampled area was a circle with the radius of 250 meters based on the Japan
public map. And the extended area indicated that along the line connecting the BTS
and the sampled area extending the sampled area towards the BTS into a 500 m x
500 m area. About the influence of buildings on the mean received signal strength,
they obtained such conclusion: although the correlation between and the mean
received signal was better, was more convenient to use and easier to be extracted
from the map.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-10-
WiMAX Network Planning
At present, from the digital maps, the following land use information can be
obtained:
Locations of buildings
Sizes or base areas of buildings
Total area occupied by buildings
Number of buildings in concerned areas
Parking lots or parks with trees and vegetation
After getting such information, the following parameters can be further acquired:
In the following, three environment types are introduced. Each type further
includes multiple cases:
Table below lists a group of recommended values for the parameters in type 2 and
type 3 environments.
-11-
BSD BHD
Category BAI (%) Standard Standard VI (%)
Mean Mean
Deviation Deviation
2A 12~20 95~115 22~70 2 1 ≥2.5
2B 20~30 100~120 70~90 2~3 1 >5
2C ≥12 ≥500 >90 ≥4 1 ≤2
3A ≥45 200~250 ≥180 ≥4 1 0
3B 30~40 150~200 ≥160 3 1 0
The criteria for propagation model classification are in accordance with those
for classifying geographic areas. In addition, the model classification should also
take the frequency band into account. For example, under the same frequency band,
if the environmental characteristics of the current network are similar to the
conditions (such as the terrain contour, terrain roughness, building density and
average building height) under which the test data for propagation model calibration
are obtained, the current network can directly utilize this propagation model
1. Traffic Distribution
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-12-
WiMAX Network Planning
The CDMA system is a self-interference system, and the traffic volume has a
direct impact on forward and reverse coverage radiuses. Therefore, when
performing network simulation the traffic allocation must accord with the actual
situation. The traffic can be allocated to the BTSs in the form of either subscriber
number or Erlang. Normally, the traffic is allocated to each BTS based on its Erlang
in busy hour.
For the same clutters, they have similar subscriber distribution and traffic
model features. In order to simplify calculation, similar clutters are assigned the
same traffic weight, while different clutters have distinguishing traffic. In digital
maps, this is reflected that different clutters are assigned different traffic weight.
During network simulation, it needs to allocate each clutter a traffic weight. The
traffic weight is just a relative value, and it needs to estimate the detailed number of
subscribers in each clutter based on the real environment features and the number of
potential subscribers.
-13-
Section2 Propagation Model
Tuning
Knowledge point
From above discussion, it can be known that the macro and micro cells
generally adopt empirical or semi-empirical propagation models. In the Section 1.4,
we introduced some typical propagation models that are universally applicable. But
in some special propagation environments the prediction with these models may be
not accurate enough, and this will lead to unsatisfying network design. Therefore, it
needs to modify the parameters in those classical models to enable them to reflect
the real environment more accurately.
Data collection
Data processing
Data analysis
Precision assessing
In the following section, we will describe these four stages in more detail.
The CW test is a classic test method used for propagation model calibration. It
adopts the CW transmitter as the signal source and transmits signals through the
omni antenna. Its receiver measures received signal strength on the roads in various
directions and at different distance and gets the signal strength values. The CW test
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-14-
WiMAX Network Planning
is applicable for almost all kinds of environments. Since the test routes have great
effect on the test result, it needs to make a careful design on the test routes. The
general principles for test route design include: testing on the roads along all
directions and at all distances; testing on the roads inside all kinds of clutters;
focusing on the common roads especially the byways (including the byways inside
clutters); and avoiding repeated tests on the same route.
In addition to the CW test, testing the received signal strength in the current
network can also provide test data for propagation model prediction. The advantage
of this method is that there is no need for erecting transmitter or antenna and only
the receiver is required during the test. Hence such method is personnel and
equipment effective, particularly, the measured signal strength can reflect more real
propagation features of the network and the calibrated model can provide the
prediction that is closer to the actual network.
Either for the CW test or the test on the signal strength in the current network,
the test data collection speed should follow the Lee's theorem, which states:
Where, x indicates distance; m(x) indicates the mean received signal strength, which is
the combination of the large-scale fading (slow fading) and the space loss; r0(x) indicates
small-scale fading (fast fading); r(x) indicates the received signal. If there is no fast fading,
m(x) will be the only factor that needs to be considered. However in practice, the fast fading
exists indeed. So when testing the large-scale fading, the influence of the fast fading must be
avoided. Generally speaking, the terrain and clutter in one service area will remain
unchanged for a period of time, so for a given base station or a given location, the mean
received signal strength m(x) is almost unchanging during this period. The field strength test
is just to obtain the mean received signal strength at each test point in the test area, that is,
the fast fading influence is to be filtered from the received signal r(x). Normally, m(x) can be
expressed as:
1 x L
m( x )
2L
x L
r ( y )dy (7.2)
-15-
When averaging a group of measured signal strength r(x), if 2L is too short, the
influence of the fast fading still exists; contrariwise, the slow fading will be ignored.
William C. Y. Lee believes that if 2L is within a 40-wavelength interval, collecting 36 or up
to 50 sampling points can efficiently remove the fast fading influence.
40-wavelength is a reasonable value as a segment of length. If the length is shorter
than 40-wavelength, the weak Rayleigh fading still remains in the final mean value;
if it is longer than 40-wavelength, the local mean received signal strength will be
ignored. The coherent coefficient of two sampling points is less than 0.2 and their
interval should be equal to or larger than 0.8-wavelength, so there should be 50
sampling points within the length of 40-wavelength. With this test method, the
deviation between the test data and the actual mean received signal strength will be
less than 1 dB.
Based on the Lee's theorem, during the data collection for model calibration, it
needs to ensure that within a 40-wavelength interval 36 or up to 50 data points are
sampled, thus the proper vehicle speed can be calculated as follows:
Assume that the vehicle speed is v, n points are sampled per second by the test
equipment and the wavelength is , then:
For example, when the transmit frequency is 800 MHz, = 0.375 m, 30 points
are sampled per second, the required vehicle speed is in the range of 32.4 ~ 45
km/h.
Due to some reasons such as equipment failures, there maybe exists some
anomalous test data. For example, because of the receiver disconnection, some
measured data shows abrupt variation in the strength. Such anomalous data should
be excluded.
During the CW test, in the areas near the transmitter, the received signal
strength is greatly influenced by the building shape and street orientation. Therefore,
the test data collected in these areas should be excluded.
In the CW test, the data collected is the received signal strength at the
transmitted frequency. It needs to first calculate the difference between the strength
of the transmitted and received signals to obtain the path loss at different distances,
and then through curve fitting acquire the median curve of the path loss, thus
achieving the radio propagation model for the tested area.
The calculation formula for the path loss can be expressed as:
L = Pt+Gt+Gr—Lc—Pr (7.4)
Where,
Pt and Pr indicate the transmit power and receive power in units of dBm.
Gt and Gr indicate the transmit antenna gain and receive antenna gain in units of dBi.
Lc indicates the total loss of the antenna and feeder system in units of dB.
After calculating the path loss at each test point, it can be obtained the
parameters in the propagation model through curve fitting. For various propagation
models, they have different parameters to be corrected and these parameters can be
-17-
selected according to the actual conditions. At present in the telecommunications
industry, many wireless network simulation software tools can support the functions
of automatically processing the CW test data and calibrating propagation models,
which greatly improves the efficiency of test data analysis.
For a running network, the collected data is the co-operating result of all the
cells in the network, as shown in Figure 2-1.
As seen from above discussion, the CW test method calibrates the propagation
model through curve fitting on the basis of the corresponding relationship between
the distance d and the received signal power Pr. If other parameters such as antenna
height are considered, the model calibration through CW testing can be taken as a
multiple linear fitting course. However, with respect to the model calibration
method by testing the current network, the model parameters are determined by the
corresponding relationship between a group of di and Pr; meanwhile during the data
analysis the power is in units of watt and the propagation model is a non-linear
formula. Therefore, the model calibration through testing the current network is far
more complex than that through CW testing. At present, the method of model
calibration through testing on the current network commonly adopts the multiple
non-linear regression method. And the most two important methods solving the
multiple non-linear regression are the Gauss-Newton algorithm and
Newton-Raphson algorithm, both of which adopt the iterative method to calculate
the parameters.
Once the data analysis is completed, the calibrated parameters including all
correction factors in the model and the clutter correction factor are generated. Take
the ZXPOS CNS1 software as an example, the model that it attends to calibrate is as
follows:
L = K1 + K2·lgd+ K3·Hms + K4·lgHms+ K5·lgHeff + K6·lgHeff·lgd+ K7Ld + Kclutter
(7.5)
Where,
Heff and Hms indicate the effective antenna height of the BTS and of the mobile station.
K1, K2, K3, K4, K5, K6 and K7 indicate the model parameters to be calibrated.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-18-
WiMAX Network Planning
Kclutter indicates the correction factor of clutter loss, which includes the loss in local
clutter and other clutters on the propagation path.
Figure 2-1 Testing the Current Network for Propagation Model Calibration
For each test point, it has two values of received signal strength: one is the
predicted value obtained by using the calibrated model and the other is the data
measured in practical test. Compare these two values. Assume that their difference is C,
and then the formulas for calculating the mean and the standard deviation of C can be
expressed respectively as follows:
n
C i
E (C ) i 1 (7.6)
n
1/ 2
1 n (7.7)
(C )
i 1
n
(Ci E (C )) 2
Where,
Ci indicates the difference between the predicted value and the measured value of the
received signal strength at the ith test point.
-19-
Section3 Simulation Flow
Knowledge point
In the data preparation stage, it needs to collect the information including the
requirement analysis, digital maps, traffic parameters, network parameters, etc. The
stage of propagation model calibration is not certainly necessary. When the area being
simulated can select a proper propagation model from the model database, the model
calibration stage can be ignored. But if there is no suitable model in the model database
for the simulated area, the model calibration is needed. For more details about model
calibration, please refer to the Section 7.3. Other stages in the network simulation flow
also include the path loss prediction, traffic distribution creation and wireless network
simulation analysis. Finally, output the wireless network simulation report.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-20-
WiMAX Network Planning
Requirement Analysis
First, consult the operator for the network design objectives and simulation
requirements, including the system frequency band, coverage range, coverage rate, etc.
Digital Maps
The higher resolution the digital maps have, the closer the simulation result is to
the actual situation. On the other hand, the higher resolution the digital maps have, the
more expensive they cost and the longer the simulation process takes. Therefore, it
needs to select the digital maps with proper resolution comprehensively considering the
demands on simulation precision and the simulation running time to optimize the
performance-to-price ratio. For example, for dense urban areas, the digital maps with
-21-
higher resolution should be adopted. Particularly for the simulation in micro cells, the
digital maps containing building information should be used. While for the wide areas
in suburban or rural areas, the resolution of the digital maps can be relatively lower.
Network Parameters
BTS parameters include the BTS RF parameters such as the accurate latitude and
longitude of BTS location, antenna height relative to the ground, antenna type, azimuth,
downtilt and pattern files of each sector. They also include BTS equipment parameters
such as forward/reverse radio configuration, receive/transmit link loss, numbers of
various channels and their transmit power, handoff threshold, power control parameters,
carrier, and BTS noise figure, etc.
(2) MS Parameters
Protocol parameters include the forward/reverse data rate supported by the system,
corresponding relationship between demodulation signal-to-noise ratio (SNR)
threshold and frame error rate (FER), coding mode, etc.
Traffic parameters indicate the traffic type, traffic model and traffic volume
proposed by the network design.
The time that the path loss prediction takes is related to some factors including the
computer hardware configuration, the range set to be predicted, algorithms contained in
the propagation model such as the algorithms respectively for effective BTS antenna
height, for diffraction loss, for clutter height, etc. Generally speaking, the larger the
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-22-
WiMAX Network Planning
predicted range is and the more complex the related algorithms are, the longer time the
path loss prediction will take.
The first step is to determine the traffic types, traffic models and traffic volume in
the network. The traffic types include the voice service, data service, and so on. The
traffic models are used for simulating the features of each traffic type with statistical
methods and thus achieving data processing in the computer. The traffic volume refers
to the traffic amount of each traffic type.
Finally, create the traffic distribution file. The wireless network simulation can
simulate the actual network performance under a certain status on the computer
platform. However, in practical networks, the distribution of various types of
subscribers is random and not predictable. To simulate such feature of randomness, the
simulation takes the subscriber presence probability in each bin as a traffic distribution
file, that is, a random distribution about subscriber presence probability is generated
and this distribution is related to the traffic weight assigned to the type of the clutter
that the bin belongs to. During the simulation analysis, a subscriber may appear in any
bin of the network at random, however, his presence probability obeys the traffic
distribution file.
-23-
network simulation result can be obtained. The dynamic analysis refers to sampling the
network statuses continuously and simulating the continuous variation on processes
such as subscriber movement, resource arrangement, power control, and so on, thus
getting the simulation result on the network. Although the dynamic analysis can
simulate the network operation more comprehensively with higher precision, it is
seldom applied in the practical network planning and optimization due to its complex
implementation, a lot of calculation and huge time consumption. In terms of the static
analysis, despite the fact that the snapshots used by the analysis are separate
discontinuously, it can still reflect the real network performance relatively accurately as
long as the count of samplings is large enough. Therefore, in the current wireless
communications industry, the static analysis is the more widely used simulation
method.
The Monte Carlo method is a common analysis method used in static analysis.
The fundamental idea of this method is to transfer the issue being calculated into a
probability model and enable the model’s numerical characteristics (e.g. mean value) to
be the calculation result required, and then calculate the estimated values of these
numerical characteristics through sampling test and statistical methods. The statistical
experiment is such a method that generates a large number of required sampling values
of the random variables (i.e. random number) via the computer, and then performs the
calculation and obtains the evaluated values of the numerical characteristics.
In brief, the Monte Carlo method is to obtain the required approximate solution by
simulating the function composed of random variables x1, x2, ···, xp:
and then getting the sampling values 1, 2, ···and at last getting the probability
distribution of and statistical estimate of each matrix through statistical processing.
As we can see from above description, the Monte Carlo method is a unique method in
wireless network analysis.
Adopting the Monte Carlo method, the static analysis generates a large number of
instantaneous network snapshots in the computer and analyzes to acquire the network
performance results represented by all of these snapshots, and then, through statistical
assessment on these snapshot results gets the estimated network performance. The
static simulation analysis flow is shown in Figure 3-2.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-24-
WiMAX Network Planning
One snapshot
Yes
Continue snapshot?
No
Although snapshots are discrete in time, the analysis process on each snapshot is
the same. It normally includes the following steps:
-25-
According to the traffic distribution file, determine the location of terminals. The
terminal location in one snapshot differs from those in other snapshots. However, in
one snapshot, once the location of terminals has been determined, it will not change
any longer. The number of terminals in each snapshot follows the Poisson distribution
that takes the input terminal number as the mean value.
Initialize the system and terminal parameters such as channel usage, power usage,
terminal transmit power, etc.
Iterate the forward and reverse links until the iteration convergence condition is
satisfied. The iteration convergence condition can be either the iteration times or the
change rate of the two adjacent iterations.
Record all network performance indexes after iteration and determine whether the
next snapshot is necessary.
Make statistics of all snapshot results and output the final network performance
index.
The prediction on network coverage mainly includes the following two cases:
And the commonly predicted performance indexes include: system soft handoff
percentage, system call drop rate, etc.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-26-
WiMAX Network Planning
Knowledge point
From above discussion, it can be seen that the wireless network simulation design
is related to multiple factors, such as propagation model, traffic distribution, system
parameters, and so on. The simulation design quality is the cooperating effect of these
factors. A good network simulation design is able to truly reflect various performance
indexes of the network, guide the network planning and optimization and improve their
efficiency, and is cost-effective. Contrarily, a poor network simulation design will fail
to achieve these functions. Here we will introduce a method on accessing the quality of
network simulation design. The main steps of this method are: First, driving test on the
current network to measure its performance data; then comparing the test data with the
data designed in network simulation process; finally, according to the comparison
result assessing the network simulation design quality. The flow of this method is
shown in Figure 4-1.
-27-
Network simulation Test data on current
design result network
Comparing
Comparison result
Quality assessment on
comparison result
Deducing
The basic idea of this method is to deduce the design quality in the entire network
through assessing the design quality on the tested roads. It can be considered as a
means of deducing population quality through assessing sample quality. This method
involves issues such as sampling, inferential methods, and so on, which will be
introduced in detail in the following section.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-28-
WiMAX Network Planning
Figure 4-2 Deduction on Simulation Quality in Entire Network (from Sample to Population)
The test routes should cover the areas with both good and poor coverage in the
network.
The test routes should cover the areas of all types of clutters as fully as possible.
Meanwhile, in order to remove the influence of fast fading, the data collection
should also meet the Lee's theorem. About details of Lee's theorem, please refer to the
Section 7.3.1.
The quantity of test data is an important factor affecting the deduction accuracy,
the more the test data is, the higher the confidence of the deduction result, but
-29-
meanwhile the more the test workload. Therefore, it needs to make a tradeoff between
test result confidence and test workload and try to minimize the workload under that
the confidence can be guaranteed.
Record the grids which contain the test data, and read the simulation data from
these grids. Then calculate the difference between the two kinds of data one by one:
X ( x) F ( x) f ( x) (7.9)
Where,
X(x) indicates the error between the test data and the simulation data.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-30-
WiMAX Network Planning
Conduct above calculation in all the grids containing test data. The calculated X(x)
is a random variable following the normal distribution. Make statistics of the mean and
standard deviation of X(x):
1 N
E( X )
N
X
i 1
i
(7.10)
1/ 2
1 N
S n1 ( X ) X i E ( X )2 (7.11)
N 1 i 1
It has been proved that E ( X ) and S n1 ( X ) are unbiased estimates for the real
mean error E and the standard deviation , and they can be used to assess the quality
of network simulation on the test routes.
In here, the principle of deducing the mean and standard deviation confidence
interval of the entire network simulation design through the mean and standard
deviation of the data on test routes is based on the random sampling theory in the
statistics science.
According to the statistics theories, when the sample mean E and standard
deviation are known, if the population follows the normal distribution, the
population’s mean and standard deviation confidence interval under a certain
confidence level can be deduced based on the sample mean and standard deviation.
Assume that the sample mean is E and the standard deviation is , and for a
given value of 0 1 , the confidence interval of the population mean under the
1 confidence level is:
-31-
E / 2 , E /2 (7.12)
n n
Where,
Where,
For example: In some one city, the quality indexes of the simulation design on
the test routes are:
E(X) = 1.0 dB
Sn-1(X) = 7.72 dB
Assume = 4.55, then the quality index of the entire network simulation design is:
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-32-
WiMAX Network Planning
Statistical analysis
Analysis result
-33-
Section5 Simulation in Network
Planning and Optimization
Knowledge point
During network planning, the network simulation can help planning engineers in
many aspects such as determining BTS location, predicting forward and reverse
coverage performance, designing an appropriate network planning scheme to meet the
expected objectives. The simulation can be applied to planning a new network or a
network to be capacity-expanded.
In the network optimization stage, the network simulation can help network
optimization engineers analyze the problems existing in the network, compare various
solutions and determine the optimal resolution, thus avoiding the influence of repeated
adjustments on the network in practice.
In Figure 5-1, the dotted line outlines the target coverage scope. In this network,
the entire area is divided into two types of sub-areas based on their respective
environmental characteristics. In these two sub-areas, two propagation models are
respectively adopted and two different coverage radiuses are respectively obtained. It
should be noted that during the network topology design, the value of cell radius is
dependant upon the capacity demand as well as the coverage requirement.
After the preliminary network topology design is completed, output the major
simulation results such as forward and reverse coverage through the simulation
software tool. If the simulation results show that the coverage and capacity
requirements cannot be satisfied, it needs to modify the network topological structure
or adjust relevant parameters of antenna and feeder system until the desired goals are
achieved. Figure 5-2 shows the forward received power (Mean Received Power) plot
obtained by simulation. Figure 5-3 shows the reverse transmission power (UL Required
Tx Power) plot. The simulation results also include the useful information such as the
transmit power and the number of accessing subscribers under each sector, which can
be utilized by capacity analysis.
-35-
Figure 5-2 Simulation in the Planning of a New Network – Mean Received Power
Figure 5-3 Simulation in the Planning of a New Network –UL Received Tx Power
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-36-
WiMAX Network Planning
optimization. For a long time, how to evaluate whether the network planning quality
reaches the KPI demand have been a problem puzzling network planning engineers.
Before a network is built, the network simulation is the only way to predict the
network’s future coverage and operational performance. In this stage, through network
simulation the network planning scheme can be evaluated. At the early stage of
network planning, the coverage is the key measurement index, so in this section we
will focus on introducing the evaluation on network coverage performance through
simulation.
Here is an example: In the early stage of network planning in one project in some
one city, the simulation is utilized to predict the network coverage. When performing
the simulation, the following parameters are input:
Digital maps: The digital maps are relatively accurate with a 20-meter resolution.
System parameters: The system parameters used in simulation are set according to
the latest system parameter setting scheme from the research department.
Propagation model: The propagation model used in simulation is obtained through
the CW test.
Traffic distribution: It is considered that there is no subscriber in the network after
the network has just been put into operation.
The simulation statistics about the coverage percentage of forward received power
(Mean Received Power) is shown in Table 5-1.
After the network is put into formal operation, conduct the test on the scene and obtain
the statistics about actual coverage percentage of Mean Received Power, as shown in
Table 5-2.
-37-
Compare above simulated and tested data and we can obtain the following graph:
Tested data
Simulation data
From Figure 5-4, it can be seen that the simulated data and tested result on coverage
percentage of Mean Received Power is very close.
In addition, by combining the simulation analysis and actual survey on the sites,
the rationality of the newly added sites can be fully demonstrated. This will help ensure
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-38-
WiMAX Network Planning
these new sites can on the one hand achieve the expected targets and on the other hand
will not introduce new problems to the network.
From the analysis in this chapter, it can be known that the network simulation can
be widely applied during network planning and optimization. By imitating the actual
operation of the network, the network simulation can provide the guide for network
planning design and optimization adjustment, hence preventing the influences of
repeated parameter adjustment on the network performance in practice.
-39-
Section6 Static Simulation
Algorithm
Knowledge point
The Absolute method uses the height of the base station antenna above ground as
the effective antenna height.
Heff = Hb
Where:
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-40-
WiMAX Network Planning
The Average method calculates the effective antenna height as the base station
antenna height above the average terrain height across the area of the prediction.
Where:
Where:
-41-
Knife-Edge Based Method (Effective Antenna Height)
The Knife-Edge Based method calculates the effective antenna height based on the
difference between the area under the radio path and the area under the height profile
between the mobile station (MS) and the base station (BTS). The radio path is the
convex path from the mobile, over the knife-edges to the base station.
A RP AHP
Heff = 2( d )
Where:
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-42-
WiMAX Network Planning
Where:
(H0b + Hb) is the ground height at the base station + height of antenna above
ground .
(H0m + Hm) is the ground height at the mobile + mobile antenna height above
ground .
You can specify the distance ds and a minimum Heff minand maximum Heff
maxfor the effective antenna height, which are used to clip the calculated values.
K is calculated as follows:
H 0 m H ds
K= ds
Where:
If the mobile is less than a distance ds from the base station, then:
H 0 m H 0b
K= d
Where:
-43-
6.2 Diffraction Loss Calculation (Knife Edge) Techniques
When adding a Standard Macrocell propagation model, on the Diffraction tab, you
can choose from one of these knife-edge techniques to calculate the diffraction loss:
Epstein-Peterson,Bullington,Deygout,Japanese Atlas,Giovaneli
The Epstein Peterson diffraction loss technique is based on the assumption that the
total loss can be evaluated as the sum of attenuation due to each respective significant
obstruction. Referring to the picture below, the diffraction loss from the obstacle is
calculated by assuming that the receiver is at the second obstruction (T-O1-O2). The
loss from the second obstacle is then calculated assuming the transmitter is at the first
obstruction and the receiver at the third (O1-O2-O3). Finally, the loss from a
transmitter at the second obstacle to the receiver (O2-O3-R) is calculated. The total
Epstein-Peterson diffraction loss is given by the sum of all the losses calculated.
Advantages:
This method does not ignore important obstacles in the way that the Bullington
method might, but to some extent it still suffers from underestimating the path loss.
When the obstacles are widely separated, this method gives more accurate values than
Deygout.
Disadvantages:
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-44-
WiMAX Network Planning
This method often gives too small values for obstructions loss where there are
obstacles close to each other and inside the first Fresnel zone.
The Bullington diffraction loss technique calculates the loss over multiple
obstructions by considering a single equivalent knife-edge positioned at the point of
intersection of the transmitter and receiver horizon paths. The total diffraction loss is
taken as that over the equivalent knife-edge obstruction.
Advantages:
Disadvantages:
-45-
Deygout Diffraction Loss Technique
The Deygout diffraction loss technique calculates a „v-parameter for each edge,
then the one with the largest is termed the main edge and its loss is calculated in the
standard way. Additional losses for other obstructions are calculated between the main
edge and the obstructed terminal. The total Deygout loss is given by the sum of all
losses calculated. In order to extend the technique to many obstructions, it is necessary
to employ sub-main edges. These are the next most significant edges at either side of
the main edge. The loss from the sub-main-edge is calculated assuming a hypothetical
terminal located at the main-edge (ignoring any less significant edges).
Advantages:
This method provides accurate results where there are two obstructions, with one
being clearly dominant. For three or four obstructions, the Deygout method gives the
best results of any of the approximate methods.
Disadvantages:
Where there is no dominant edge, the Deygout method tends to overestimate the
loss. Where there are several equal obstacles inside the first Fresnel zone or close to
each other, this method underestimates the obstruction loss.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-46-
WiMAX Network Planning
the diagram, the total loss is given by the sum of losses for T-O1-O2, T -O2-O3 and
T -O3-R.
Advantages:
This method gives improved results when the obstructions are closely spaced.
Disadvantages:
To some extent this method still suffers from underestimating the path loss.
-47-
between the main edge and the receiver, the start point is the main edge and the end
point is the receiver. This process is repeated recursively until losses have been
calculated for all knife-edges. The total Giovaneli loss is given by the sum of all these
losses.
本文中的所有信息均为中兴通讯股份有限公司秘密信息,不得向外传播
-48-