You are on page 1of 15

Radio Propagation Model Calibration

for

Ericsson 9999 Model

August 2004

1 of 15
Titre: Radio Propagation Model Calibration Results
Version: V1.1
Date: August 2nd , 2004
Auteur: Hatem Mokhtari
Résumé: This document summarizes the radio propagation model calibration results
to be used for RF Planning and Optimization within Wataniya. It also shows
the validity domain along with the limitations of this propagation model.

Revision History
Version Date Auteur Modifications

2 of 15
Table of Content
1 EXECUTIVE SUMMARY...........................................................................................................................................4

2 ACRONYMS..............................................................................................................................................................4

3 INTRODUCTION .......................................................................................................................................................4

4 SURVEY ANALYSIS.................................................................................................................................................4
4.1. CW Measurements………………………………………………………………………………………………..5
4.2. Typical Environments……………………………………………………………………………………………..5
4.3. Data Amounts and Filtering………………………………………………………………………………………5
4.4. Selected Sites for Algiers………………………………………………………………………………………...5
5 RADIO PROPAGATION REQUIREMENTS .............................................................................................................6
5.1 RECEIVE POWER VALUES ..........................................................................................................................................6
5.2 DISTANCE RANGE VALUES .........................................................................................................................................6
5.2 TRANSMIT ANTENNA REQUIREMENTS ...................................................................................................................................... 6

6 TCP PROPAGATION MODEL DESCRIPTION........................................................................................................8

7 MODEL CALIBRATION RESULTS ..........................................................................................................................9

8 CONCLUSION.........................................................................................................................................................10

9 REFERENCES ........................................................................................................................................................11

10 ANNEXES ...............................................................................................................................................................12

3 of 15
1. Executive Summary

This document summarizes the radio propagation model calibration results to be used for RF Planning and
Optimization within Wataniya. It also shows the validity domain along with the limitations of this propagation
model.

2. Acronyms

TCP : TEMS Cell Planner

3. Introduction

The 9999 Model looks like COST231/Okumura-Hata Models except that additional corrective
factors are taken into account. Namely: Spherical Earth Diffraction correction and multiple
knife-edge diffraction algorithms.

Propagation model tuning is not always an easy task as it requires an in-depth knowledge of
Radio propagation mechanisms from one hand, and a reliable drive test measurements utilized
in conjunction with an accurate Clutter database from the other hand.

This document aims at providing a comprehensive overview on propagation model calibration


that will be used in TCP for RF planning, dimensioning and Optimisation purposes. As the
model that is implemented in TCP is a semi-empirical site-general model, the emphasis is put
on the fact that automatic model tuning is used, and a ready-to-use drive test measurements
contributed to make the calibration a fast and fairly accurate process.

Section 4 describes the Data collection process, the sites that have been selected for our
surveys, and especially the reason why such a choice. Section 5 explains and reviews some
propagation fundamentals applied in model calibration and especially inherent to TCP model.
Section 6 refers to Ericsson propagation algorithm as described in TCP reference manual. This
section is merely a reminder for those who want to have a clear idea on the propagation
mechanism within TCP (namely 9999 Model). Section 7 is a bench of results obtained further
to automatic calibration steps as described in Section 6. Section 7 also encompasses the
obtained statistical results, the number of samples used in different clutter classes, but also
contains the final mean error and the standard deviation of the error.

4. Survey Analysis

Most of the planning tools make use of semi-empirical propagation models such as Okumura-Hata (900
MHz band) and COST231-Hata (1800 MHz band). These models are limited in terms of accuracy as
they have been designed for typical European cities and therefore cannot be applied in the Algerian
environment. However, if corrective parameters are taken into account these models can give
reasonable accuracy. To be able to design a GSM network with reliable best server and neighbor plan,
we need to calibrate the semi-empirical model. To do so we need to perform CW measurements in
different clutter types to be able to represent as much as possible the environment that will be
encountered during the deployment.

4 of 15
4.1. CW Measurements

It is a common sense to calibrate the propagation model using a single pure RF carrier because the
signal strength is the unique figure of merit of the received power. GMSK modulation is thus useless to
be used during calibration measurements. Most of the time we do CW measurements using a narrow
beamwidth of 30kHz.

4.2. Typical Environments

As we seek to produce a site-general propagation model, measurements have to be performed in all


possible clutter types. Besides, as most of the sites will be deployed mainly in Dense Urban, Urban and
Suburban the measurements will focus on these environments without neglecting rural and highway
areas. This simply means that we identify the following environments :

a) Dense Urban (DU)


b) Urban (U)
c) Suburban (S)
d) Rural (RU)
e) Highways (H)

Generally CW measurements for calibration should have higher amount of data in the first three
categories as mentioned above (i.e. DU, U, and S) and few data for RU and H categories.

4.3. Data amounts and Filtering

To cope with the fast fading requirements and avoid the signal to be faded by the short-term behavior
we need to filter data falling into the fades. According to Lee Criterion the fast fading can only be filtered
out if the sampling rate is higher than 40 times the wavelength. In other words, if we use the 900 MHz
frequency band this should lead to a minimum of 13,33 m and 6,66 m for 1800 MHz band. All the data
between the filtered samples are thus discarded and considered as non representative of the long-term
(slow fading) propagation data. The sampling rate can, in most of the time, be set during the set-up of
the drive test equipment and tool. In our case, we used 50 samples in each 40 wavelength window.

In terms of the amount of data to be collected, we estimate to about 10 km of route for dense urban
areas and up to 30 km for rural environment. The reason is that we do need to have the maximum
clutter types represented as much as possible to be able to do a proper calibration. In terms of number
of samples, we will have to deal with 750 samples in the 900 MHz band for Dense Urban (10 km) and
about 2250 samples in Rural (30 km)l. We need more samples/site in rural areas because we have less
sites selected in that environment. Besides, rural areas have mostly free-space like propagation
behavior, so we need more data far away from the sites as the coverage is deemed to have larger
extent compared to dense urban or urban areas.

4.4. Selected sites for Algiers

We have to bear in mind that calibration requires also evaluation afterwards to confirm the adopted
clutter losses. In other words, CW measurements have to be used in the following proportions :

a) Calibration : About 70% of the overall data have to be exclusively used for the iteration process
during calibration. In this step, the mean error has to be centered (i.e. as close to zero as possible)
and the standard deviation of this error between a reasonable range of 6-8 dB.

5 of 15
b) Evaluation : About 30% of the overall data (not the ones used for calibration) will be utilized to
validate and verify the adopted parameters during calibration.

The list of sites selected for Calibration Drive Tests is summarized in Annex 1, whereas Annex 2
contains the map of Algiers with the location of these sites.

5. Radio Propagation Requirements

5.1. Receive Power Values

As the radio propagation channel is never static, it recommended that the measurements are filtered out
from the fast fading according to the abovementioned Lee Criterion. Also, to avoid the use of wrong
values for the receive downlink signal, it is further recommended to discard values below -110 dBm and
above -40 dBm. The advantage is thus twofold :

We avoid noise-like levels in the cell-edge from one hand and we also avoid values close to the receiver
saturation domain.

5.2. Distance Range Values

There are a number of criteria, regarding the clutter and distance categories, to be taken into account :

5.2.1. Maximum Cell Radius Range : 10 km in urban areas, as a general rule of thumb, and
about 30 km in rural areas. Beyond these values, the propagation model requires
dual-slope modelling, which is well-known a time consuming calculation.
5.2.2. In terms of minimum distance, it is obvious that measurements in the vicinity of the
site are wrong because of the near-field influence. The theoretical near-field region is
given in the literature and it reads : Dmin = 2D2/λ (D is the antenna size in meters and
λ is the wavelength in meters), which is about 1,5 m in GSM900 for a 50 cm antenna
size. As a conservative approach, we used a minimum distance of 100 m far away
from the site to avoid local shadowing and scattering objects in the immediate vicinity
of the transmitting point.

5.3. Transmitter System Description

The transmit system comprises the following :

• TEMS Transmitter of 17 dBm EiRP used a signal generator


• RF Power Amplifier of 28 dB supplied by ETSA (www.etsa.fr)
• A 10 m 7/8” Feeder Cable
• An omnidirectional Antenna supplied by ETSA of 2 dBi Gain. Vertical and Horizontal
antenna patterns are available in the annexes.
• VSWR Measurements – Thanks to IRIS - have also been performed to check both the
supplied power to the antenna, from one hand, and to check that the VSWR is less than
1.15 (acceptable value)

Photographs of the test-bed equipment are available in the annexes.

The measured output power (@ antenna connector) is 39 dBm, which leads to an EiRP of 41 dBm. This
value is quite low as we need at least 50 dBm. This leads to a restriction in terms of maximum range,
which we could not reach because of this power limitation.

5.4. Selected Sites for Calibration

6 of 15
The following site locations has been used for both calibration and evaluation of the model :
Cell Id COMMUNE
AL1654e OUED SMAR
AL1632a OUED SMAR
AL1645c BAB EZZOUAR
AL1629a BENI MESSOUS
AL1625a BOUZEREH
AL1672a BIR MOURAD RAIS
AL1671b GUE DE CONSTANTINE(AIN NAADJA)
AL1628b CHERAGA
AL1677b EL MADANIA
AL1620b BEN AKNOUN
AL1679d MOURADIA
AL1604a ALGER CENTRE

These have been selected to represent as much as possible the different clutter classes in Algiers.

7 of 15
CLUTTER MAP LEGEND

Dense Urban
Urban
Low Urban
Forest
Residential
Park
Cultivated
Industrial
Semi-Open

6. TCP Propagation Model Description

TCP propagation model is called 9999 Model and is thoroughly described in the IRIS document
referred to as ”Propagation Model Tune Methodology & Criteria”, June 4th, 2004.

8 of 15
The model contains an Automatic Tuning kernel that speeds up the calibration process.
Although it is accurate, one may need to do manual checking such as evaluation of the model
on measurements other than those used for calibration.

7. TCP Propagation Model Calibration Results

As the model tuning in TCP is an automatic process, we did not do any manual fine-tuning as
we quickly reached the convergence criterion. The model is optimized (i.e. Mean Error = 0 dB)
and the standard deviation is within a reasonable range.

RMS
Model_100_500 Num. Bin Mean Error Error Std.Dev. Error
EIRP:41TUNED 8860 -0.0 8.3 8.3
Num. RMS
Clutter Bins Mean Error Error Std.Dev. Error
high residental 445 -0.0 8.5 8.5
urban 677 0.0 7.3 7.3
low urban 2018 0.0 7.5 7.5
residental 1969 -0.0 7.5 7.5
industrial 196 0.0 6.2 6.2
park 370 0.0 7.5 7.5
semi-open 2980 -0.0 9.6 9.6
forest 150 -0.0 7.4 7.4
plantation 5 0.0 2.2 2.4
cultivated 50 0.0 6.1 6.2

9 of 15
Data Clutter Distribution

35,00%

30,00%

25,00%

20,00%

15,00%

10,00%

5,00%

0,00%
high urban low urban residental industrial park semi-open forest plantation cultivated
residental
Clutter Category

According to the histogram above, no ”Dense Urban” data were collected because the unique
site in the Casbah was rejected. Normally this should be avoided but, in the case of Algiers, a
very small red spot (Dense Urban category) appears on the map. Therefore, this should not
lead to major problems.
The Clutter losses to be used in TCP are summarized in the table below :

Clutter Loss dB
high residental 18
urban 22
low urban 17
residental 21
industrial *19
park 18
semi-open 13
forest *16
plantation **10
cultivated **13
*, **, and *** marks mean that data collected are much less than the other clutter classes according to the drive
test results.

8. Conclusion

The calibration results (i.e. Mean Error and Standard Deviation) are very acceptable and lie
within a reasonable range. However, we should bear in mind that this model has to be

10 of 15
evaluated first before using it in the neighbouring cities/towns surrounding Algiers. Subsequent
work to be done is to test the Frequency Planning using this model, load it in the OSS, and
then perform dedicated mode drive tests to check the interference.

9. References

[1] D. Parsons, The Mobile Radio Propagation Channel, London, UK, Pentech, 1992

[2] C. A. Balanis, Advanced Engineering Electromagnetics, New York, John Wiley & Sons,
1989.

[3] COST 231 – “Urban Transmission Loss Models for Mobile Radio in the 900 and 1800 MHz
Bands (Rev. 2), COST 231 TD90) 119 rev2, The Hague, The Netherlands, Sept. 1991.

[4] Epstein J. Peterson D. W.; „An Experimental Study of Wave at 850 Mc/s”, Proc. IRE, Vol
41, pp 595-611, May 1953.

[5] Bullington K., “Radio Propagation at Frequencies above 30 Mc/s”, Proc. IRE, Vol 35, pp
1122-1136, October 1947.

[6] Ikegami F., Yoshida S., Takeuchi T., Umehira M.; “Propagation Factors Controlling Mean
Field Strength on Urban Streets”, IEEE Transactions on Antennas and Propagation, Vol. AP-
32, N°28, August 1984.

[7] Hata M., “Empirical Formula for Propagation Loss in Land Mobile Radio Services”, IEEE
Transactions on Vehicular Technology, Vol VT-29, N°3, August 1980.

11 of 15
ANNEX 1
TRANSMIT ANTENNA PATTERN

To see the real vertical antenna pattern, we should squeeze the diagram below 90° clock-wise
or 90° anti clock-wise. Horizontal Pattern is just a circle as the power is the same in all
directions.

Vertical Antenna Pattern EiRP

1
36 37 50 2 3
35 4
34 40 5
33 30 6
32 20 7
31 10 8
30 0 9
29 -10 10 EiRP
28 11
27 12
26 13
25 14
24 15
23 16
22 21 18 17
20 19

12 of 15
ANNEXE 2

PATH LOSS VARIATION VERSUS DISTANCE BEFORE CALIBRATION PER CLUTTER CLASS
(Few Sites)

Path Loss Vs Distance (Residential) AL1625a

160,00
140,00
Path Loss (dB)

120,00
100,00 PLMes(dB)
80,00 PLPred(dB)
60,00 Free Space
40,00
20,00
0,00
0,500
0,530
0,592
0,644
0,656
0,672
0,713
0,761
0,819
0,883
0,981
1,229
1,522
1,640
Distance (km)

Path Loss vs Distance (Low Urban)


AL1625a

160,00
140,00
Path Loss (dB)

120,00
100,00 PLMes(dB)
80,00 PLPred(dB)
60,00 FreeSpace
40,00
20,00
0,00
0,66
0,72
0,79
0,93
0,98
1
1,07
1,14
1,21
1,26
1,35
1,5
1,52
1,67
1,91

Distance (km)

13 of 15
Path Loss vs Distance (Industrial)
AL1625a

160,00
140,00

Path Loss (dB)


120,00
100,00 FreeSpace
80,00 PLMeas(dB)
60,00 PLPred(dB)
40,00
20,00
0,00
0,98
1
1,03
1,04
1,06
1,07
1,1
1,13
1,16
1,18
1,19
1,22
1,24
1,26
Distance (km)

Path Loss Vs Distance (Semi-Open)


AL1632a

160,00
140,00
Path Loss (dB)

120,00
100,00 FreeSpace
80,00 PLMes(dB)
60,00 PLPred(dB)
40,00
20,00
0,00
0,741
0,860
0,989
1,115
1,181
1,199
1,225
1,272
1,322
1,355
1,471
1,569
1,688
1,818
1,985
2,102
2,196
2,482

Distance (km)

14 of 15
ANNEX 3
TRANSMITTER UNIT AND ANTENNA PHOTO

Signal Generator :

TEMS Transmitter 17 dBm EiRP

Amplifier :

ETSA 28 dB Gain

Telescopic Mast : 5,7 m

Feeder : 10m 7/8”

15 of 15

You might also like