WCDMA RAADIOVÕRGU PLANEERIMINE LÄHTUVALT OLEMASOLEVAST GSM VÕRGUST WCDMA RADIO NETWORK PLANNING BASED ON EXISTING GSM SITE REUSE

Transcription

1 TALLINNA TEHNIKAÜLIKOOL Raadio- ja sidetehnika instituut Kood: IRT80LT WCDMA RAADIOVÕRGU PLANEERIMINE LÄHTUVALT OLEMASOLEVAST GSM VÕRGUST WCDMA RADIO NETWORK PLANNING BASED ON EXISTING GSM SITE REUSE Jevgeni Okuško Töö on tehtud telekommunikatsiooni õppetooli juures Juhendaja: Avo Ots Kaitsmine toimub Infotehnoloogia teaduskonna kaitsmiskomisjonis Autor taotleb tehnikateaduste magistri nimetust Esitatud: Kaitsmine: Tallinn 2006

2 2 REFERAAT Tänapäeval on klientide ootused ja drastiliselt kasvanud nõudmised erinevate mobiiliside teenuste järele sundinud mobiilside tööstust pöörama oma pilgud teiselt mobiilside generatsioonilt kolmandale. Olenemata faktist, et teise generatsiooni mobiilside tehnoloogia ikka veel areneb ja turule tuuakse uusi teenuseid on suur osa mobiilside operaatoreid alustanud kolmanda generatsiooni võrkude rajamisega. Euroopa Telekommunikatsiooni Standardite Instituut (ETSI) valis kolmanda generatsiooni mobiilside süsteemi raadioliidese standardiks WCDMA Laiaribaline koodidega eraldatud mitmekordne juurdepääs. WCDMA on mõeldud ja arendatud pakkumaks suurematel andmedastuskiirustel põhinevaid multimeedia teenuseid, kui see on võimalik teise generatsiooni süsteemides. Koos muutustega võrgu infarstruktuuris toob WCDMA ka uue lähenemise raadiovõrgu planeerimisse, lisades sinna uusi väljakutseid. Kui teise generatsiooni raadiovõrkude planeerimine on peamiselt tehtud katvuse printsiipi jälgides, siis kolmanda generatsiooni võrkude planeerimise juures on liskaks katvusele oluline ka interferentsi ja mahu analüüs. Teisisõnu on kolmanda generatsiooni raadiovõrkude planeerimisel vaja samaaegselt arvesse võtta palju enam faktoreid. Seega on tugijaamde konfiguratiooni optimeerimine alates tugijaama asukohast ja lõpetades antenni tüübiga, -suuna ja allakallutusega väga oluline täitmaks teenuse ja mahu nõudeid. Käesolev magistritöö on esitatud 68 leheküljel, sisaldab 33 joonist ja 17 tabelit ja on kirjutatud inglise keeles. Märksõnad: WCDMA, GSM, raadiovõrkude planeerimine, mahtuvuse dimensioneerimine, katvuse dimensioneerimine, raadiovõrgu planeerimise tarkvara simulatsioonid.

3 3 PREFACE Nowadays customer expectations and drastically growing demand of different mobile services push the whole mobile communication industry to focus on radio access technology evolution from the second generation towards the third generation. Despite the fact that 2G mobile networks still evolve and bring new services into the market, great number of operators started to deploy 3G networks. European Telecommunications Standards Institute (ETSI) selected Wideband Code Division Multiple Access (WCDMA) as a multiple access technique for the radio interface in the 3 rd generation mobile telephone communication systems. WCDMA has been specified and developed to enable advanced multimedia services using significantly higher bit rates than the current 2 nd generation systems. Together with providing changes in the network infrastructure, WCDMA brings a new approach in the radio network planning process, promising a number of challenges on the way to 3G. If in the case of the 2 nd generation, radio planning is based on coverage optimization, then there is more interference and capacity analysis than just coverage area estimation in 3 rd generation planning. In other words, 3G-network planning is based on the fact that a lot of issues should be considered simultaneously. Thus, the site configurations optimization, starting from site locations and finishing with antenna type, directions and tilts, takes on special significance to meet capacity and service requirements. Current thesis is presented on 68 pages, containing also 33 illustrations and 17 tables and is written in English language. Keywords: WCDMA, GSM, radio network planning, capacity dimensioning, coverage dimensioning, radio planning tool simulations.

4 4 TABLE OF CONTENTS REFERAAT... 2 PREFACE... 3 LIST OF FIGURES... 6 LIST OF TABLES... 7 USED ABBREVIATIONS... 8 INTRODUCTION WCDMA OVERVIEW WCDMA AIR INTERFACE WCDMA RAN ARCHITECTURE RADIO NETWORK PLANNING PLANNING PROCESS RADIO PROPAGATION MODELING Fast and slow fading Properties of wideband radio channel Okumura-Hata model The COST 231 Walfisch-Ikegami model NETWORK DIMENSIONING Traffic estimation Definitions Radio access bearers Activity factor and DTX gain Channel models Radio link budget Margins Uplink link budget Downlink link budget Capacity calculations Uplink capacity Downlink capacity Soft capacity DETAILED PLANNING ANTENNA CONFIGURATION General Antenna height Antenna beamwidth selection and sectorization Antenna downtilt Mechanical downtilt Electrical downtilt Soft and softer handover Softer handover factor Pilot pollution Antenna directions Coverage overlap between adjacent sectors CAPACITY AND COVERAGE PLANNING Planning tool Monte Carlo analysis of TEMS Cell Planner Universal Load estimation STUDY CASE SIMULATION ENVIRONMENT General simulation parameters Traffic density information... 53

5 5 4.2 SIMULATION RESULTS First scenario CPICH coverage Soft handover probability Influence of spacing between adjacent sectors Statistics report Second scenario CPICH coverage Soft handover probability Soft handover probability KOKKUVÕTE CONCLUSION REFERENCES A. LOG-NORMAL FADING MARGINS A.1 SIMULATED LOG-NORMAL FADING MARGINS FOR 3-SECTOR SITES, MULTI-CELL ENVIRONMENT... 68

6 6 LIST OF FIGURES FIGURE 1.1 WCDMA FREQUENCY BAND ALLOCATION [12] FIGURE 1.2 UMTS DEPLOYMENT [12] FIGURE 1.3 WCDMA RAN ARCHITECTURE FIGURE 2.1 SIGNAL STRENGTH VARIATIONS [12] FIGURE 2.2 FADING WIDTH VS. THE PRODUCT OF DELAY SPREAD AND SYSTEM BANDWIDTH FIGURE 2.3 OKUMURA SET OF CURVES FIGURE 2.4 SITE AREA DEFINITION FIGURE 2.5 WALFISCH-IKEGAMI MODEL PARAMETERS FIGURE 2.6 DEFINITION OF STREET ORIENTATION ANGLEϕ FIGURE 2.8 RADIO ACCESS BEARER (RAB) DEFINITION FIGURE 2.9 LINK BUDGET FIGURE 2.10 THE CUMULATIVE NORMAL DISTRIBUTION FUNCTION [12] FIGURE 2.11 COVERAGE AREA ACCORDING TO JAKE S FORMULA [12] FIGURE 2.12 NOISE RISE FIGURE 2.13 SOFT CAPACITY IN WCDMA [16] FIGURE 3.1 FIRST FRESNEL ZONE FIGURE 3.2 PRACTICAL ANTENNA RADIATION PATTERNS. (A)33 HORIZONTAL ANTENNA BEAMWIDTH,(B) 65 HORIZONTAL ANTENNA BEAMWIDTH [26]...41 FIGURE 3.3 EFFECT OF WIDENING OF THE HORIZONTAL RADIATION PATTERN IN CASE OF INCREASING MECHANICAL DOWNTILT VALUE [29] FIGURE 3.4 HORIZONTAL RADIATION PATTERN IN CASE OF USING ELECTRICAL ANTENNA TILT [17] FIGURE 3.5 COVERAGE OVERLAP BETWEEN ADJACENT SECTORS FIGURE 3.6 PLANNING TOOL SIMULATIONS APPROACH FIGURE 3.7 TCPU SIMULATIONS STEPS [35]...48 FIGURE 4.1 FIRST SCENARIO NETWORK PLAN FIGURE 4.2 SECOND SCENARIO NETWORK PLAN FIGURE 4.3 CPICH COVERAGE PLOT (1 ST SCENARIO) FIGURE 4.4 REAL VS. AVERAGE ANTENNA HEIGHT OF THE SITES FIGURE 4.5 IRREGULAR ANTENNA DIRECTIONS FIGURE 4.6 HANDOVER MAP (1 ST SCENARIO) FIGURE 4.7 CORRELATION BETWEEN ADJACENT SECTOR SPACING AND SHO PROBABILITY FIGURE 4.8 DOWNLINK LOADING (1 ST SCENARIO) FIGURE 4.9 CPICH COVERAGE PLOT (2 ND SCENARIO) FIGURE 4.10 HANDOVER MAP (2 ND SCENARIO) FIGURE A.1 LOG-NORMAL FADING MARGINS FOR 3-SECTOR SITES... 68

7 7 LIST OF TABLES TABLE 1.1 WCDMA TECHNICAL CHARACTERISTICS TABLE 2.1 UMTS QOS CLASSES TABLE 2.2 RADIO ACCESS BEARERS TABLE 2.3 DTX GAIN [12] TABLE 2.4 MAPPING OF CHANNEL MODEL TO UE SPEED [12] TABLE 2.5 UPLINK LOG-NORMAL FADING MARGINS FOR 3-SECTOR SITES (HANDOVER GAIN INCLUDED) TABLE 2.6 UPLINK LINK BUDGET TABLE 2.7 DOWNLINK LINK BUDGET TABLE 4.1 SIMULATION PARAMETERS TABLE 4.2 TRAFFIC DATA INPUT PER SERVICE TABLE 4.3 LOAD LEVEL OF SERVICES PER CELL TABLE 4.4 WORST SOFT HANDOVER PROBABILITY CELLS TABLE 4.5 STATISTICAL RESULTS FOR SYSTEM PARAMETERS (1 ST SCENARIO) TABLE 4.6 COMPARISON OF COVERAGE AREAS OF 1 ST AND 2 ND SCENARIOS TABLE 4.7 COMPARISON OF HANDOVER PROBABILITIES OF 1 ST AND 2 ND SCENARIOS TABLE 4.8 STATISTICAL RESULTS FOR SYSTEM PARAMETERS (2 ND SCENARIO) TABLE 4.9 AMOUNT OF SERVED USERS (1 ST VS. 2 ND SCENARIOS)

8 8 USED ABBREVIATIONS 2G 3G 3GPP AGL AMR AS ASC BCCH BPSK BS CCH CDMA CN COST CPICH CPL CS DCCH DL DS-CDMA DTCH DTX ETSI FACH FDD GSM HO HSDPA HSUPA HT KPI LOS MS NLOS PDF QoS QPSK PS P-SCH RA RAB RB RBS RET RNC RF S-SCH SHO SRB - The 2 nd Generation of Mobile Networks - The 3 rd Generation of Mobile Networks - The 3 rd Generation Partnership Project - Antenna Height above Ground Level - Adaptive Multi Rate - Active Set - Antenna System Controller - Broadcast Control Channel - Binary Phase Shift Keying - Base Station - Control Channels - Code Division Multiple Access - Core Network - Cooperative Scientific Research - Common Pilot Control Channel - Car Penetration Loss - Circuit Switched - Dedicated Control Channel - Downlink - Direct-Sequence Code Division Multiple Access - Dedicated Traffic Channel - Discontinuous Transmission - European Telecommunication Standards Institute - Forward Access Channel - Frequency Division Duplex - Global System for Mobile communications - Handover - High Speed Downlink Packet Access - High Speed Uplink Packet Access - Hilly Terrain - Key Performance Indicator - Line-Of-Sight - Mobile Station - Non-Line-Of-Sight - Probability Density Function - Quality of Service - Quadrature Phase Shift Keying - Packet Switched - Primary Synchronization Channel - Rural Area - Radio Access Bearer - Radio Bearer - Radio Base Station - Remote Electrical Tilt - Radio Network Controller - Radio Frequency - Secondary Synchronization Channel - Soft Handover - Signaling Radio Bearer

9 9 TDMA TEMS TCPU TDD TMA TU UE UL UMTS UTRAN WCDMA - Time Division Multiple Access - Test Mobile Station - TEMS Cellplanner Universal - Time Division Duplex - Tower Mounted Amplifiers - Typical Urban - User Equipment - Uplink - Universal Mobile Telecommunications System - UMTS Terrestrial Radio Access Network - Wideband CDMA, Code Division Multiple Access

10 10 INTRODUCTION Nowadays co-siting of 3G sites with the existing 2G sites help to significantly reduce costs and speed up WCDMA deployment. However, very often co-siting can introduce a number of issues that need to be considered. Apparent cost savings can come back as an ineffective network plan with wasted radio resources and money as a consequence. This could be a result of blindly copying an existing 2G network for the WCDMA network planning, which is strongly discouraged due to differences in the planning approaches for the two different technologies. Current paper covers the radio network planning of a new 3G network based on an existing GSM network s radio plan. The main differences between radio planning approaches in GSM and WCDMA in order to avoid dramatic network performance deterioration by following inappropriate old-fashioned radio planning principles are considered. A general overview of the WCDMA technology characteristics and features as well as the WCDMA radio access network (RAN) architecture are provided in Chapter 1. Chapter 2, firstly, highlights the main principles of the WCDMA radio planning process, split into a few major sub-processes described further in more detail. Secondly, the main principles of radio propagation modeling with descriptions of the most common models used in urban environments have been presented. Finally, WCDMA radio link budget calculations together with capacity dimensioning methodology are addressed in this chapter. As mentioned above, GSM radio planning principles can no longer be valid for WCDMA radio planning to certain extent. If in case of low traffic areas, WCDMA radio planning is quite similar with GSM radio planning, then in high traffic scenarios this similarity disappears. The reason is that, there is no clear split between coverage, capacity and interference planning in WCDMA radio planning. Chapter 3 highlights the most important aspects of detailed planning. The output is worked out methodology of radio network design based on initial minimization of the harmful effects of excessive coverage overlap and soft handover probability by means of antenna re-directions, changes of angles between adjacent sectors and electrical tilt optimization. In addition, the main functionality of Monte-Carlo analysis of TEMS Cellplanner Universal (TCPU) planning tool, used for case study simulations, has been described in Chapter 3. Two different radio planning scenarios, performed by TCPU tool, are being investigated in Chapter 4. In the first scenario, the 3G network based purely on the existing GSM network s radio is investigated. In the second scenario, a modified plan, according to the planning methodology worked out in Chapter 3, is presented. Coverage and traffic statistics of both networks as well as simulation results plots, showing network performance, interference and pilot pollution, are presented and analyzed. The influence of proper site configuration choice on network performance is studied. Chapters 5 and 6 conclude the thesis along with providing some assumptions and recommendations.

11 11 1 WCDMA OVERVIEW 1.1 WCDMA AIR INTERFACE WCDMA (Wideband Code Division Multiple Access) is a wideband Direct-Sequence Code Division Multiple Access radio interface technology. It has been selected as a radio transmission technology for UMTS (Universal Mobile Telecommunications System), which is the European third generation mobile Communications system developed by ETSI (European Telecommunications Standards Institute). WCDMA has become the most popular global 3G air interface mode within the last few years. It has been actively implemented by existing GSM operators across the world. time code 5 MHz (WCDMA) frequency Figure 1.1 WCDMA frequency band allocation [12]. The main technical characteristics of WCDMA are shown in the table below: Table 1.1 WCDMA technical characteristics. Multiple Access Scheme DS-CDMA Duplex Scheme FDD/TDD Carrier spacing 5 MHz Chip rate 3.84 Mcps Frame length 10 or 20 ms (optional) Multirate Variable speading and multicode Data modulation QPSK (downlink) BPSK (uplink) Spreading factors Power control Open and fast closed loop Handover Soft and interfrequency handover The main WCDMA features and advantages can be presented as follows: Same frequency time domain for all users Users separated by orthogonal codes Frequency Division Duplex (FDD) and Time Division Duplex (TDD) modes Chip rate of 3.84 Mbps within a 5 MHz frequency band Multi circuit and packet-switched high bit-rate services Simultaneous operation of mixed services

12 12 Highly variable user data rates based on the rate matching procedure (HSDPA and HSUPA) Asynchronous base stations Coherent detection on downlink and uplink Inter-system handover with GSM Soft/softer inter-frequency handover 1.2 WCDMA RAN ARCHITECTURE The main goal of UMTS is to deliver new high data rate services to the user in the mobile domain. This new technology requires a wide frequency band of between 5 MHz and 15 MHz (compared with a 200 khz carrier for GSM), which has an impact on both radio and core network architecture (see Figure 1.2). GSM operator UMTS operator Other UMTS operator Macro layer Micro layer 3 MHz 4.4 MHz 5.0 MHz 5.0 MHz f = 14.6 MHz Figure 1.2 UMTS deployment [12]. A WCDMA network can be overlaid onto an existing GSM network, and in so doing this can help to achieve significant savings on operator infrastructure investment. This paper is focused on the WCDMA radio network planning issues. The WCDMA radio access network (RAN) architecture is shown in Figure 1.3. WCDMA RAN consists of one or more radio network subsystems (RNS), which in turn consists of 2 main network node types: The radio network controller (RNC) exercises radio network control functions on radio base stations (RBSs) and user equipment (UE). It comprises a number of enduser services such as: mobile telephony, packet data (including HSDPA), short message service (SMS), ciphering and mobile positioning. It also provides end-user functions: radio access bearer (RAB) (see Section ), mobility (cell reselection, handover, macro diversity and location update), radio resource management, connection control and capacity management. The radio base station (RBS) provides radio resources for the RAN, i.e. it handles the user data traffic and control communication with the RNC and the UE.

13 13 Communication between UTRAN nodes goes through a number of external interfaces: Iu: Interface between an RNC and the Circuit Switched / Packet Switched Core Networks. The interface is used for traffic related signaling. Iub: Interface between an RNC and an RBS used for traffic related signaling. Uu: Interface between an RNC and a UE. Mub/Mur: Management interface provided by the RBS/RNC. It is used for element management and network management. Iur: This interface is between two different RNCs. The interface is used for user data and control related signaling Figure 1.3 WCDMA RAN architecture. Operation and Maintenance functionality is handled through the embedded management in RNC and RBS as well as software solutions [12]: Radio Access Network Operation Support (RANOS): operates the WCDMA RAN including coordinated handling of tasks on multiple network elements. Tools for Radio Access Management (TRAM): tool-set for planning, design, test and performance monitoring.

14 14 2 RADIO NETWORK PLANNING Radio network planning is a very time consuming and labor-intensive procedure. The quality of network planning has a direct impact on an operator s future income. WCDMA technology in particular has set new requirements for radio planning and has created the need for a new planning approach. The reason for that are fundamental differences between 2G and 3G technologies. A WCDMA-based system is a multiservice environment with variable rates for its users using the same frequency simultaneously. The number of active users and type of services used has a direct influence on the coverage of the cell, which means that coverage and capacity planning can t be separated into phases and should be considered at the same time. This well-known phenomenon is called cell breathing and shows strong dependency of coverage and capacity on each other. Despite the differences in the planning approaches of the two technologies, utilization of existing GSM sites by operators is very important in speeding up 3G network deployment. If an operator has an already existing GSM network then use of existing infrastructure can sufficiently minimize WCDMA network deployment expenses. Besides co-siting (using already existing GSM sites locations), similar to existing GSM network cell plan (as similar as possible in WCDMA and GSM, i.e. antenna positions, heights, directions and tilts) will most likely be used in a large extent. However, blind copy of existing GSM1800 network plan, for instance, should be avoided. 2.1 PLANNING PROCESS The overall goal of the radio network planning is to maximize the coverage and capacity while meeting key performance indicators (KPIs) and quality of service (QoS) requirements. In general, a radio network planning consists of three major phases: Dimensioning - area definition - link budgets and coverage analysis - traffic and capacity estimation - estimation of number of base stations required Detailed planning - site configurations (site locations, heights, azimuths, tilts etc.) - traffic estimations - code and frequency planning - coverage thresholds and capacity requirements - parameters settings Optimisation and monitoring - network tuning (parameters settings adjustment) - coverage and capacity verification - traffic measurements It can be easily noticed that the traffic level has to be continuously taken into account at all the radio planning stages. The distribution of the traffic between different services must be determined for each base station coverage area as precisely as possible. Ideally, exact

15 15 location of the users has to be known, which is obviously almost impossible to obtain in reality. However, the more accurate users distribution forecast the better radio network design can be done. Radio base station locations should be placed on the traffic hot spots in order to obtain the best link budget for the mobile users served by a particular RBS. It might significantly reduce power consumption in the network, which decreases interference level and increases capacity. Existing GSM network statistics can be good support in rough traffic estimations in 3G radio planning. As a rule, in the initial dimensioning phase, a fixed predicted during busy hour load defined for all the base stations in the planned area. However, in the detailed planning phase, traffic distribution might be used in order to allocate the traffic to the particular cells. Thus, in some cases the load between the cells can vary dramatically. Coverage targets also should be carefully checked during detailed planning phase. It must be kept in mind that the traffic distribution is assumed to be homogeneous in the particular areas during dimensioning phases. Also, propagation is assumed to be similar for all the cells. However, the fact that traffic density of the particular areas might be quite different must be considered. Moreover, coverage predictions can be very different across the planned network due to propagation environment and traffic distribution. Before we can proceed with the network dimensioning process consideration, the main principles of radio propagation modeling together with well-known models basics, used for dimensioning and predicting in urban environment, should be understood. 2.2 RADIO PROPAGATION MODELING Fast and slow fading Fading is quite typical for urban environments when there is no line-of-sight path between the transmitter and the receiver. Shadowing from obstacles in the propagation path will cause log-normal fading. Due to multiple reflections, radio waves travel along different paths of varying lengths, causing effect of fast fading. High building density causes severe diffraction loss. In addition, the path loss, i.e. the decrease of the global mean value of the signal strength, is one more component that will affect the received signal strength. Further (in sections and ) the two most common models for path loss estimations in terms of urban environments have been considered. Thus, in general, the mobile radio channel can be characterized by three major signal components: Fast fading, e.g. Rayleigh distributed Slow fading or log-normal fading Path loss

16 16 Figure 2.1 Signal strength variations [12] Properties of wideband radio channel The main advantage of a wideband radio channel is that the fast fading decreases with the increase of the system bandwidth. An empirical relationship how the fading width varies with the system bandwidth and the delay spread is shown in Figure 2.2 [6]. Fading width [db] (Delay spread) (system bandwidth) Figure 2.2 Fading width vs. the product of delay spread and system bandwidth. The fading width is directly proportional to the average difference between the values of the peaks and the dips in db. It is defined as the difference between 99 % and 1 % levels of the cumulative distribution function of the measured signal strength values. The delay spread shows the spread of the propagation paths in the time domain. If the delay spread is assumed to be constant (a typical value for urban environment is 0.5 μs), the fading width will decrease with the increase of the bandwidth. Thus, according to

17 17 Figure 2.2, for GSM systems with a bandwidth of 200 khz, the fading width would be 22 db, while it would decrease to 9 db for WCDMA systems with a bandwidth of 5 MHz Okumura-Hata model Usually, the propagation loss is calculated by means of different propagation models. The most common is the Okumura-Hata propagation model, which is based on the Okumura method and considers several parameters such as effective antenna height, terrain type, terrain height, frequency, and so on. The Okumura method is semi-empirical and based on extensive measurements performed in the Tokyo area [9]. The results are expressed in the sets of curves, showing field strength as a function of distance for different frequencies and antenna heights. See example of Okumura set of curves in Figure 2.3. Figure 2.3 Okumura set of curves. The formulas for analytical calculations have been derived from Okumura curves by Hata. Path loss, L, can be found as: p max

18 18 L p max = log f log h a( h ) + ( log h ) log R (2.1) b m b where f - the signal frequency, h b - the effective antenna height of the base station, a ( h m ) - the correction factor, h m - the mobile terminal height, R - the cell range. The cell range, R, is the distance corresponding to the maximum allowed path loss, L pmax. The correction factor, a(h m ), for urban areas (small and medium size cities) can be found as follows [15]: For the large cities, a(h m ) is: a( hm ) = (1.1log f 0.7) hm 1.56 log f (2.2) a ( h ) [ ( 1.54 hm )] [ ( h )] log 1.10 f 200 MHz m = (2.3) log m 4.97 f 400 MHz It must be noted that a(h m ) is fixed to 0 db in case of h m = 1.5m, which is the most common height for cellular radio system calculations. Equation 2.1 can be slightly modified: L p max = A log h + ( log h ) log R (2.4) b b where A - semi-empirical constant (155.1 for urban area, used in calculations, Table 2.6 and Table 2.7) h b - the antenna height of the base station (35m in calculations, Table 2.6 and Table 2.7). According to equation (2.4), the cell range can be found as: α R = 10 (2.5) where L A log h b ( h ) pmax b m α = (2.6) logh + a Thus, the coverage area of the cell can be defined as:

19 Site _ area = 3R (2.7) Figure 2.4 Site area definition. The model is well suited for the following conditions: 150 f 1500 MHz 1 R 20 km 30 h b 200 m 1 h m 10 m Different variants of the Okumura-Hata model are implemented in the various radio planning tools The COST 231 Walfisch-Ikegami model The COST 231 Walfisch-Ikegami model is a combination of the models from J. Walfisch and F. Ikegami. It was further developed in terms of the COST 231 project. COST 231, Evolution of land mobile radio (including personal) communications, is a subgroup of a European Union forum for cooperative scientific research. COST 231 has been later continued by COST 259. The model considers the buildings between the transmitter and the receiver in the vertical plane. Since the propagation over the rooftops (multiple diffractions) is the most dominant part in urban environments, the accuracy of the model is quite high. This model is well suited for the following conditions: 800 f 2000 MHz 0.02 d 5 km 4 h b 50 m 1 h m 3 m Flat ground Uniform building heights and building separations. The parameters used in Walfisch-Ikegami model calculations are depicted on Figure 2.5 [10].

20 20 α ΔH h b H h m d w b Figure 2.5 Walfisch-Ikegami model parameters. In case of free line-of-sight between the base station and the mobile terminal antennas in a street canyon path loss can be calculated as: In case of non-los: L LOS = log R + 20log f, R 20m (2.8) L nlos = L L L (2.9) FS rts msd where L - the free space loss, FS L - the rooftop-to-street diffraction and scatter loss, rts L - the multi-screen loss. msd The rooftop-to-street loss can be found as: L = log w + 10log f + 20 log( H h ) + L (2.10) rts m ori where w - the width of the street (see Figure 2.5), H - the height of the buildings (see Figure 2.5), h - the height of the mobile terminal (see Figure 2.5). m Expression Lori is defined as: o ϕ 0 ϕ 35 o L ori = ( ϕ 35) 35 ϕ 55 (2.11) o ( ϕ 55) 55 ϕ 90 where φ - the angle in degrees between incidences coming from base station and road (see Figure 2.6) [10].

21 21 Figure 2.6 Definition of street orientation angleϕ. In the case, when the base antenna height exceeds the surrounding rooftops, L msd can be defined as: L msd = log ( 1+ h H ) b + 18 log R + f log f log b (2.12) where b - the separation distance between blocks (see Figure 2.5). There are terms added for short distances (R < 0.5 km) and for base antennas lower than rooftops. However it has been shown that the accuracy of the prediction is best for the case h b >> H [12]. Thus, the path loss formula for urban environments with cell size less than 1km can now be expressed as, where L K + 38log R 18 log( h 17) (2.13) = b K = for 1900 MHz for 2050 MHz (2.14) It has to be noted that equation (2.13) is only valid for base station antenna heights above 18 m. α R = 10 (2.15) where L pmax K + 18 log( h 17) α = b (2.16) 38 Different variants of the Walfisch-Ikegami model are implemented in the various radio planning tools.

22 NETWORK DIMENSIONING Purpose of network dimensioning is to estimate amount of node elements needed in order to cover planned area. Thus, this step includes calculation of radio link budgets, capacity and coverage in both uplink and downlink. Normally, traffic estimation and growth forecasts have to be based on market analysis. A common process of network dimensioning is depicted in Figure 2.7. It can be seen that the dimensioning is successfully finished as soon as the balance between calculations in uplink and downlink is reached. Obviously, at the initial stages, when network load is not high, the highest priority is to fulfill set coverage thresholds. Therefore, capacity has not been considered as an issue in further calculations. Figure 2.7 A common process for network dimensioning [12].

23 Traffic estimation The traffic estimation for speech in the network can be done based on live GSM network statistics. Although, even for speech the amount of traffic is strongly linked to how aggressively 3G handsets are pushed into the market. UMTS specific services such as video service, high speed packet services are very hard to predict, and will be to a certain extent affected by the pricing model for video calls, for instance. Likewise, for packet the usage will be strongly linked to the pricing model in conjunction with the pricing of other (mobile) broadband services available. All in all, based on the combination of 2G voice and the number of 3G handsets planned to be sold (with contracts or on pre-paid) the voice traffic, and to a certain extent the packet traffic can be estimated. Despite on obvious difficulties with the UMTS services traffic estimation, it is very important at least roughly to predict potential growth of the traffic capacity in the near future Definitions Radio access bearers The transport through the radio network in UTRAN is provided by a Radio Access Bearer (RAB) service. A RAB is a logical connection between the Core Network (CN) and the User Equipment (UE). It consists of a Radio Bearer (RB) connection between the UE and RNC, and an Iu bearer connection between RNC and core network, and is used to provide a connection for a UMTS service via UTRAN (Figure 2.8). Radio access bearer (RAB) Radio bearer Iu Bearer UE RBS RNC Uu Iub Iu Core network Figure 2.8 Radio access bearer (RAB) definition. The radio bearers can be mapped onto both dedicated channels as the Dedicated Traffic Channels (DTCH) and common channels as the Random Access Channel (RACH)/Forward Access Channel (FACH). The RABs are divided into different traffic classes, which are optimized for different application types with different QoS requirements. Table 2.1 presents types of traffic classes with their characteristics and example applications [1]:

24 24 Traffic class Conversational Table 2.1 UMTS QoS classes RAB configuration Preserved time variation between information units in data stream (low delay) Example application Voice Streaming Preserved time variation between information units in data stream Streaming audio Interactive Background Request response pattern (preserved payload content) Destination is not expecting the data within a certain time (preserved payload content) Web browsing Background download of s More details about the RAB attributes and QoS concepts can be found in [1]. Table 2.2 shows configuration information of some WCDMA system radio access bearers (RABs) [12]. The total bit rates in both UL and DL is defined as the sum of the service rates and signaling rates, i.e. the total bit rate in the case of the speech service 12.2 kbps (service) and 3.4 kbps (signaling) is 15.6 kbps. The higher bit rate connections have lower Eb/No requirements. The robustness against interference provides a processing gain, PG, defined as: Rchip PG = 10 log( ) (2.17) R inf o where R chip - the chiprate (3.84 Mcps), R inf o - the data rate. The higher the data rate, the lower the processing gain value. The reason is that a fixed chip rate of 3.84 Mcps is used in WCDMA systems. Hence the coverage area will be smaller for high data rate connections.

25 25 Table 2.2 Radio access bearers. Traffic class Signaling radio bearers Conversational Interactive Streaming Mixed RAB configuration SRB for BCCH SRB for PCCH SRBs for BCCH, CCCH and DCCH (FACH) SRBs for CCCH and DCCH (RACH) 13.6/13.6 kbps SRB for DCCH AMR 12.2 kbps RAB 28.8 kbps CS RAB 64 kbps CS RAB 64/64 kbps PS RAB 64/128 kbps PS RAB 64/384 kbps PS RAB 57.6 kbps CS RAB 8/54 kbps PS RAB + Interactive [64/64 or 64/128 or 64/384] kbps PS RAB AMR 12.2 kbps RAB + + Interactive [64/64 or 64/128 or 64/384] kbps PS RAB AMR 12.2 kbps RAB + + Interactive [64/64 or 64/128 or 64/384] kbps PS RAB + + Interactive [64/64 or 64/128 or 64/384] kbps PS RAB Interactive [64/64 or 64/128 or 64/384] kbps RAB + + Interactive [64/64 or 64/128 or 64/384] kbps RAB Activity factor and DTX gain A radio link between the UE and the RBS consists of a Dedicated Traffic Channel (DTCH) and a channel used for radio resource control signaling sent on the Dedicated Control Channel (DCCH). When one of the channels (DTCH or DCCH) is not transmitting, less interference is generated on the air interface and more capacity can be expected. The capacity gain depends on the activity factor of the DCCH(s) and DTCH(s).

26 26 Table 2.3 DTX gain [12]. RB configuration DTX gain UL DL Speech/AMR 12.2 kbps RB kbps SRB 2 69% 102% AMR 7.95 kbps RB kbps SRB 2 69% 102% AMR 5.95 kbps RB kbps SRB 2 69% 102% AMR 4.75 kbps RB kbps SRB 2 69% 102% 28.8 kbps CS RB kbps SRB 7% 5% 64 kbps CS RAB kbps SRBs on DPDCH/DPCH 2 7% 5% 8 kbps PS RB kbps SRB 2 5% 5% 16 kbps PS RB kbps SRB 2 5% 5% 32 kbps PS RB kbps SRB 2 5% 5% 64 kbps PS RB kbps SRB 2 5% 5% 128 kbps PS RB kbps SRB 2 2% 384 kbps PS RB kbps SRB 2 1% 57.6 kbps CS RAB kbps SRBs for DCCH 2 6% 6% Streaming 16 kbps PS RB kbps PS RB kbps SRB 2 6% Streaming 64 kbps PS RB kbps PS RB kbps SRB 2 6% 1 50% activity factor 2 10% activity factor 3 100% activity factor during packet transmission Channel models Bit energy over noise, Eb/No, is an important parameter in WCDMA coverage and capacity planning as a part of quality requirements for the WCDMA bearers. It depends on the radio environment and the speed of the mobile terminal. The 3GPP specifications standardize channel models [5] in order to reflect the broadband channel properties of a WCDMA system and simplify radio environment modeling in network planning tools. Table 2.4 shows an overview of some 3GPP-specified channel models applicable for WCDMA network planning. Table 2.4 Mapping of channel model to UE speed [12]. Channel model Typical Urban (TU-3, TU-50) Rural Area (RA-3, RA-50, RA-120) Hilly terrain (HT-120) UE speed 3 km/h, 50 km/h 3 km/h, 50 km/h, 120km/h 120km/h

27 Radio link budget The link budget is used to calculate the maximum allowed path loss and the maximum cell range. After that the cell coverage area can be calculated. Antenna gain G Pathloss ant L p max Feeder loss L RBS power P RBS f RBS sensitivity RBS sens Noise rise I UL UE sensitivity UE sens UE power P UE Figure 2.9 Link budget. The link budget includes a number of margins [16], [12], which are needed in order to compensate various uncertainties and losses during radio network design Margins Interference margin The interference margin is the increased noise level caused by greater load in a cell. It shows dependency between coverage and capacity dimensioning - the larger the cell load, the smaller the cell coverage area. Fast fading (log-normal) margin The signal strength value calculated by means of wave propagation algorithms can be considered as a mean value of the signal strength in a small area, which is determined by the accuracy and resolution of the model. The deviation of the local mean has nearly a normal distribution in db in comparison with the predicted mean. That is the reason why variation is called log-normal fading. As a rule, radio propagation models used for pathloss predictions in WCDMA planning tools do not capture the log-normal fading. As a result, the coverage probability reaches only about 50%, whereas most operators have requirements of 95-99,99%. In order to improve percentage of probability of signal strength above the prediction value, a lognormal fading margin, LNF design process. m arg, is added to the link budget calculation during the radio LNFm arg can be expressed by means of Jakes formula [12]:

28 28 LNFm arg = [ Pt ( Lpath ( d 0) + 10 log( R / d 0 ))] γ (2.18) where P - the transmitted power, t L path ( d 0 ) - the average reference path loss, d 0 - the distance at which path loss is referred to (in meters), R - the radius from transmitted antenna, γ - the desired received signal threshold. The relation between LNF and the perimeter coverage, P (γ ), which denotes the m arg probability of having a received signal strength value above a certain threshold on the cell border, can be expressed as [33]: 1 1 LNF = + m arg LNF = Φ m arg P ( γ ) erf (2.19) 2 2 σ 2 σ where Φ (t) - the normal distribution function with its variable t, σ - the standard deviation. If P(γ ) is known then LNF marg can be found as [12]: LNF m arg = t σ (2.20) The variable t can be found in a normal distribution function table. Figure 2.10 The cumulative normal distribution function [12]. According to Figure 2.10, if the perimeter coverage equals 75 %, t will be In case of standard deviation value of 8 db, LNF marg equals 5.4 db, according to equation (2.20).

29 29 Thus, if a guaranteed availability of 75 % is required at the cell border and 90 % within the cell area, then an extra margin of 5.4 db has to be added to the signal threshold. Correlation between distance dependence exponent n (1/R n ) and standard deviation σ according to Jakes formula is depicted on Figure ,98 0,96 98 % 96 % 94 % 92 % 90 % Area coverage [%] 0,94 0,92 0,9 0,88 Border coverage 88 % 86 % 84 % 82 % 80 % 78 % 76 % 0,86 74 % 72 % 0,84 70 % 0,82 0 0,5 1 1,5 2 2,5 3 3, σ/n [db] Figure 2.11 Coverage area according to Jake s formula [12]. Jakes formula doesn t take the effect of many servers into account. The presence of many servers at the cell border causes handover, which results in the presence of a handover gain and lower required log-normal fading margin as a consequence. Thus, handover gain, G HO, must be taken into consideration during LNF marg calculations for multi-cell system: LNFm arg = t σ G HO (2.21) The fading margin values for the most common area coverage probabilities are presented in Table 2.5 [12]. More values presented graphically in Appendix A. Table 2.5 Uplink log-normal fading margins for 3-sector sites (handover gain included). Environment σ LNF Area coverage [%] [db] Rural, suburban Urban Dense urban; suburban indoor Urban indoor Dense urban indoor

30 30 Power control margin Approaching the cell border UE causes the path loss increase. The power control loop of the system compensates it by increasing the UE power until the loop saturates and transmits with maximum power. At this point power control is not operational anymore, which leads to reduced uplink system sensitivity. Power control margin, PC marg, takes this effect into consideration. Margin values for different radio environments (channel models and UE speeds) can be found in [12]. Body loss The human body has negative impact on the UE performance. The reason is that it absorbs energy causing reduction of antenna efficiency. The body loss margin recommended by ETSI is 3 db for 1900 MHz [14]. For data services users normally don t keep the mobile terminal positioned by the ear, so body loss is not applied for them and is set equal to 0 db. Car penetration loss If the UE is placed in the car without external antenna, an extra 6 db margin has to be added in order to compensate for the penetration loss [19]. Building penetration loss Building penetration loss is expressed as the difference between the average signal strength immediately outside the building and the average signal strength over the ground floor of the building. A typical value for urban and dense urban areas in link budget calculations is 18 db [8]. However, building penetration loss can vary significantly, depending on specific environments. Feeder loss Normally the feeder loss parameter includes all the losses between the top of the RBS and the antenna, i.e. includes possible jumper, duplexer, diplexer, etc. losses. In the case of ASC/TMA presence, feeder loss parameter includes all the losses between the top of the RBS and the ASC/TMA in the DL. In the UL the ASC compensates for feeder losses. In addition, ASC insertion loss has to be added to the calculations Uplink link budget Calculations start from the uplink (reverse link) direction, because typically uplink interference is the limiting factor in WCDMA systems. The uplink budget can be calculated as follows [12]: L p max = PUE RBSsens PCmarg IUL LNFm arg LBL LCPL LBPL + Gant L f L j (2.22) where L p max - maximum path loss due to radio wave propagation, P UE - the maximum UE output power, RBS sens - the RBS sensitivity. Depends on the RAB and channel model, PC m arg - the power control margin, I UL - the noise rise or uplink interference margin, section,

31 31 LNF m arg - the log-normal fading margin, L BL - the body loss, L - the car penetration loss, CPL L BPL - the building penetration loss, G - the sum of RBS antenna gain and UE antenna gain, ant L - the feeder loss f L j - the jumper loss. The RBS sensitivity can be expressed as [12]: RBS 10 + L (2.23) sens = N t + N f + log Rinf o + Eb N 0 where N t - the thermal noise power density (-174 dbm/hz), N f - the noise figure (typical value 2.3dB), R inf o - the information bit rate [bps], E b N 0 - the bit energy divided by noise spectral density, L f - the feeder loss. f The rough estimation of macro cell range as well as coverage area, without taking into account terrain specifics, in Table 2.6 and Table 2.7 is obtained by means of the Okumura- Hata propagation formulas , Section The input values for the calculations have been taken from Sections and 4.1 as well as from [12] and [35]. The number of sites, N site, required in order to fulfill presented in the link budget tables requirements is found as: N site Total _ area = (2.24) Site _ area where Total _ area - the total planned area (13.63 km2 in calculations, Section 4.1). A reference uplink link budget for different service types used in the planning tool simulations is shown in Table 2.6.

32 32 Table 2.6 Uplink link budget. Parameter Service type Units Voice CS64 PS128 Probability of coverage % RBS Receiver Sensitivity, RBS sens dbm Subscriber Maximum Transmit Power, P UE dbm Power control margin, PC m arg db UL Max Loading, η UL % Noise rise or the uplink interference margin, IUL db Lognormal fading margin (outdoor) 1, Lognormal fading margin (indoor) 1, Building Penetration Loss, Body Loss, RBS Antenna Gain, Feeder and Jumper losses, Maximum Path Loss (Outdoor), Maximum Path Loss (Indoor), LNF m arg db LNF m arg db L BPL db L BL db G ant dbi L + L db f j L p max db L p max db Cell range (Indoor), R km Area (Indoor) km Number of sites According to formula (2.32) uplink noise rise from the table is calculated as: 1 I UL = 10 log = 1. 5 db (2.25) As already noted, the Okumura-Hata formula can be normally used only for rough estimations at initial stages. For more precise results, a network planning tool must be used. Since the formula is valid for 1 R 20 km then (see sections and ), it can give unreliable results for small cells, e.g. in dense urban environment. In such cases, formula based on the Walfisch-Ikegami model should be used, see Section NB! The calculation results in both uplink and downlink link budget tables (Table 2.6 and Table 2.7) have been considered as input for planning tool simulations presented in Chapter 4. In order to keep consistency, the Okumura-Hata model has used in both the link budget calculations and the planning tool simulations. However, it must be kept in mind

33 33 what presented margins should be considered as example values and can be not applicable for real networks dimensioning. In conclusion, it must be noted that packet service has the lowest maximum allowable pathloss, which makes it the limiting service in the uplink direction Downlink link budget Downlink link budget calculations are similar to uplink link budget calculations. However, downlink dimensioning is a more complex procedure for a few reasons. The most important reason is that all the UE s share the same transmitter, i.e. use the same power source. In addition, interference levels are strongly dependant on user distribution and traffic load. As a consequence, downlink dimensioning is determined by the average total output power and the average CCH/DCH power. The downlink link budget can be expressed by means of following formula [12]: L p max = PTX UEsens PCmarg I DL LNFm arg LBL LCPL LBPL + Gant L j (2.26) where L p max - maximum path loss due to radio wave propagation, P TX - the maximum transmitter output power at the reference point, UE sens - the user equipment sensitivity, PC m arg - the power control margin, I DL - the noise rise or downlink interference margin, section , LNF m arg - the log-normal fading margin, L BL - the body loss, L CPL - the car penetration loss, L BPL - the building penetration loss, G ant - the sum of RBS antenna gain and UE antenna gain, L - the jumper loss. j The UE sensitivity can be expressed as [12]: UE sens N t + N f + log Rinf o + = 10 E N (2.27) where N t - the thermal noise power density (-174 dbm/hz), N f - the noise figure (typical value 7 db), R inf o - the information bit rate [cps], E b N 0 - the bit energy divided by noise spectral density. b 0

34 34 Total output power P tot, ref at the reference point (at the antenna) can be calculated as the sum of the powers allocated to each individual UE, the power allocated to synchronization and common control channels. Normally about 25% of total power allocated for control channels P,, leaving 50% for dedicated channels to traffic and about 25% as CCH ref headroom for mobility needs. Assuming that activity factors for synchronization channels (P-SCH and S-SCH) are considerably low, the average value of the total output power at the system reference point can be expressed as [12]: P = P + H L CCH, ref sa tot, ref (2.28) 1 ηdl where η DL - the downlink system loading, P CCH, ref - the average level of the power allocated to the all common control channels at the system reference point (at the antenna), H - the factor related to the path loss distribution of the UEs L sa - the signal attenuation at the system reference point to the UE at the cell border. The signal attenuation at the system reference point to the UE at the cell border: L + sa = Lp max + PCmarg + LNFm arg + LBL + LBPL Ga L j (2.29) Downlink noise rise for UE located at the cell border can be found as [12]: I DL P = 1 + K CCH, ref L 1 η sa DL + H (2.30) where K + F N N R c c = α (2.31) t f chip where α c - the non-orthogonality factor at the cell border, F c - the average ratio between received inter-cell and intra-cell interference at the cell border, R chip - the system chip rate. The downlink noise rise DL I defines margin for the noise floor at the UE in the loaded WCDMA system, which is obviously higher than in an unloaded system. As can be seen from formulas ( ), downlink noise rise depends on the transmitter output power and the location of the users in the cell.

35 35 A reference downlink link budget for different service types used in the planning tool simulations is shown in Table 2.7. The input values for the calculations have been taken from Sections and 4.1 as well as from [12] and [35]. Table 2.7 Downlink link budget. Parameter Service type Units Voice CS64 PS128 Probability of coverage % UE Receiver Sensitivity, UE sens dbm Subscriber Maximum Transmit Power, P UE dbm Power control margin, PC m arg db DL Max Loading, η DL % Noise rise or the downlink interference margin, I DL db Lognormal fading margin (outdoor), LNF m arg db Lognormal fading margin (indoor), LNF m arg db Building Penetration Loss, L BPL db Body Loss, L BL db RBS Antenna Gain, G ant dbi Feeder and Jumper losses, L + L db f j Maximum Path Loss (Outdoor), L p max db Maximum Path Loss (Indoor), L p max db Cell size (Indoor), R km Area (urban) km Number of sites As can be seen from Table 2.6 and Table 2.7, the balance between UL and DL has been reached. According to the dimensioning results, 21 sites required in order to fulfill requirements for the service with the lowest maximum path loss (PS128). However, it must be taken into account that the rough estimations obtained during the dimensioning phase can be considered only as starting point for further detailed planning by means of the radio planning tool.

36 Capacity calculations This section introduces well-known load equations for calculating load of WCDMA cell in both downlink and uplink directions Uplink capacity First, the definition of bit energy per noise ratio, E b N 0, has to be established: E N b 0 j = R chip R j i own p j p j + i other + P N (2.32) where R - the system chip rate, chip R j - the bit rate of j th user, p j - the received power from j th user, i own - the interference power received from the own cell, i - the interference power received from the other cell, other P - the thermal noise power. N The capacity per cell depends on the amount of interference per cell; hence it can be defined by means of throughput-based load factor η UL [16]: N 1 η UL = ( 1+ i) (2.33) R j= 1 chip 1+ E b R j j N υ 0 j where N - the number of active users in the cell, υ j - the activity factor of the service for j th user, i - the ratio of other cell and the own cell interferences (other-to-own-cell interference). The given formula shows a limit for the uplink load, which is contributed by fractional load of each particular active user. Obviously, in case of fixed target uplink load factor, lower other-to-own-cell interference helps to increase the number of served users, i.e. capacity. A radio network planning parameter, interference margin also known as noise rise, can be found using the uplink load factor. IUL = log(1 η ) (2.34) 10 UL

37 Noise ris Load Figure 2.12 Noise rise. The interference margin used in the link budget must be equal to the maximum planned noise rise. It must be noted that the recommended maximum uplink load is around 60-70% [12]. A higher system load can lead to uncontrollable noise rise effect. Obviously, 100% load would cause infinite noise rise, which can be considered as infinite interference Downlink capacity Notwithstanding that the downlink principles differ from uplink ones, the downlink load factor η DL can be defined based on a similar approach used for the uplink load factor definition, although the parameters are slightly different [16]: E b N η 1 W R j [( + ) + i ] N 0 j DL = j = υ j α 1 j j (2.35) where new parameters are: i j - other-to-own-cell interference of j th user, α j - the orthogonality factor of the channel of j th user. Orthogonality factor depends on radio channel conditions and mobile user s speed. It equals 1 in an ideal radio environment (without fading). However, in reality the code orthogonality of the user is partially destroyed by multipath propogation. Typically, the orthogonality is between 0.4 and 0.9 in multipath channels. It should be noted that since

38 38 other-to-own-cell interference in the downlink is unique for each user, depending on the user s location, it is practical to use average values in capacity dimensioning. The interference margin can be found by means of the downlink load factor. I DL = log(1 η ) (2.36) 10 DL It should be noted that whenη DL 1, noise rise, i.e. system approaches its pole capacity. Clearly, the downlink coverage is more dependent on load then the uplink coverage. The reason is that transmitted power is constant and shared between all mobile users in the cell, while in the uplink each terminal has its own power amplifier. Therefore, downlink coverage decrease is directly related to the number of mobile users in the cell. At the same time, in order to fulfill a user s following criteria has to be satisfied [26]: E b requirement in the downlink, the N 0 E b W p j = N R P j L 0 m, j (1 α ) j L m m, j 1 + N P L n n= 1, n m n, j + P N (2.37) where p j - the average transmitted power for j th user, L m, j - the path loss from the serving base station sector m to the j th user, L, - the path loss from the other base station sector n to the j th user, n j P m - the total transmit power from the serving base station sectors m, P - the total transmit power from the other base station sectors n. n Since WCDMA downlink capacity is interference limited, transmitted power control accuracy becomes very crucial (see equations (2.35), (2.37)). The power control is based on carrier-to-interference ratio (C/I) and can be defined as [26]: C I 1 γ γ = R R = N N 1 D γ D n= 1 n n= 1 γ n (2.38) where R - the distance between serving base station and mobile, D n - distance between nth interfering base station and mobile, γ - the exponent of the attenuation of the radio wave. In GSM a certain (C/I)-ratio has to be achieved for a proper quality of speech. It is mainly determined by the frequency reuse factor of the network. In WCDMA all cells are using

39 39 the same frequencies. As a consequence, base stations-interferers can locate within the same distance as the serving base stations [26] Soft capacity According to [4] Erlang capacity can be obtained from the Erlang B model only for hardware limited systems, also known as hard blocked systems. However, there is no single fixed value for the maximum capacity in the case of interference limited systems such as WCDMA, because interference and capacity can be shared between adjacent cells. Thus, the Erlang B model is not valid any longer because of too pessimistic results. Such a phenomenon is known as soft capacity. The idea is that the less interference is coming from neighboring cells, the more capacity is available in the middle cell (see Figure 2.13). And if neighbor cell s loading is considerably low, it means that extra capacity is available in this cell, which can be borrowed by other adjacent cells. Thus, interference sharing provides soft capacity. Figure 2.13 Soft capacity in WCDMA [16]. As a rule, in the dimensioning phase, equally loaded cells are assumed, but using radio planning tool (and in reality) different load in each cell.

40 40 3 DETAILED PLANNING In WCDMA systems, mobile users select a cell to camp on by estimating the quality of the common pilot control channel (CPICH) from each base station in the network [2]. Therefore, cell size depends on power level of CPICH, radio environment conditions of the network and loading. On the one hand, the higher CPICH transmit power means larger cell coverage and more mobile terminals camping on the cell, whereas on the other hand, it means larger soft handover areas and less power for traffic channels. In addition, it is very important that traffic distribution is fairly even in the network in order to utilize radio resources efficiently. Careful selection of possible site locations with acceptable antenna heights, azimuths and tilts is catered for in the detailed planning phase. 3.1 ANTENNA CONFIGURATION General The coverage and capacity in a network are influenced by the antenna characteristics and configuration: azimuth, beam width, side lobe level, number of sectors per site, number of antennas per sector, height above ground and tilt. A Tower Mounted Amplifier (TMA) can improve the uplink radio performance and ensure a balanced link budget. The requirements on azimuth and beam-width are set depending on traffic load and on the number of sectors for the site. It is very difficult to define the optimum antenna site configuration for each environment and each traffic load profile. The perfect antenna configuration for a sector is strictly related to the specific position of the site, to the target coverage and to the location of possible obstacles Antenna height As a general rule, the cell size is determined by capacity considerations in an urban environment. Thus, there is no need to maximize coverage by high antenna masts. In order to avoid unnecessary interference, the antenna should be placed to limit the coverage area to the desired target. Therefore, the antenna should be mounted as low as possible, reducing coverage further by tilting, if necessary. In rural environments and for road coverage, coverage may have to be maximized by erecting high poles or masts. There may not be any nearby or close obstacles that are higher than the transmitter antenna. In theoretical terms, the first Fresnel zone should be free of obstacles in the vertical direction and the sector covered by the antenna should be free of obstacles in the horizontal direction [12].

41 41 Figure 3.1 First Fresnel zone. This leads to a rule to mount the antenna height at least 5 meters above the closest obstacle Antenna beamwidth selection and sectorization The main network planning goal is to achieve an overall sufficient coverage, which can be achieved by means of higher antenna gain in the first place. At the same time, power transmission in unwanted directions has to be minimized. Interference level in both downlink and uplink directions can be controlled using a proper antenna configuration as antenna horizontal beamwidth [21], [18], [28]. Figure 3.2 shows examples of horizontally 33 and 65 antenna beamwidths, respectively. Figure 3.2 Practical antenna radiation patterns. (a)33 horizontal antenna beamwidth,(b) 65 horizontal antenna beamwidth [26]. The effect of antenna beamwidth and sectorization on capacity and coverage is explored in [22] and [21]. As research shows it is very difficult to control cell overlapping in case of high sectorization (for instance, 6 sectors sites). The amount of users increase is not proportional to the number of sectors. The number of soft handover (SHO) connections increases simultaneously with interference leakage from neighbor sectors. The described process can become uncontrollable in terms of urban environments.

42 Antenna downtilt In dense urban environments, the antennas are frequently installed at a height greater or equal to the average rooftop level. Clearly, a higher antenna position causes larger coverage overlapping areas increasing soft handover probability. In such cases it might be necessary to tilt the antenna in order to reach the coverage target and reduce interference. The effect of antenna downtilt has been widely studied [17], [29]. It is a well known fact that a wise approach in antenna tilting allows one to increase overall system capacity in WCDMA. There are two different kinds of tilt: Electrical tilt, and Mechanical tilt Mechanical downtilt Mechanical downtilt is achieved by directing the antenna towards the ground. It depends on two main factors: geometrical factor, θ GEO, and antenna vertical beamwidth factor, θ VER,BW. The geometrical factor takes into account heights of the base station, h b, and mobile station, h m, and the size of the dominance area or cell size, R. So, the optimum mechanical downtilt angle, ν m, can be found as a function of the geometrical factor and vertical beamwidth factor [29]: ( θ,θ ) ν = (3.1) m f GEO VER, BW The geometrical factor can be calculated using the relation of the height difference between the base station antenna and mobile station antenna, and sector dominance size [29]: hb hm θ GEO = arctan (3.2) R Mechanical antenna downtilt has been observed and found as an effective solution for reducing inter-cell interference by confining the signal to its dominance area [23]. Typical capacity gain varies from 15% to 20% [13], [25], [7], [29], but strongly dependant on network and antenna configurations [27]. However, it must be noticed that the angle between base station mechanical downtilt and the effective downtilt angle is the same in the horizontal plane only in the main lobe direction. With higher downtilt angles, the radiation pattern effectively shrinks in the main lobe direction, but widens in the side lobe directions, which clearly means increase in the overlapping between adjacent sectors, hence increasing the softer handover probability. In addition, potentially harmful effect of back lobe uptilt has to be taken into account. All in all, high attention has to be paid to extremely high mechanical downtilt values in order to avoid a negative effect on the network performance.

43 43 Figure 3.3 Effect of widening of the horizontal radiation pattern in case of increasing mechanical downtilt value [29] Electrical downtilt Electrical tilt means an in-built phase-shift of all antenna elements that lowers the vertical beam in all horizontal directions. Antennas, supporting the Remote Electrical Tilt (RET), allow modifying the electrical down tilt via remote control system, simplifying the network optimization process. The need for a common interface for base station equipment to support RET has been also highlighted by the 3GPP specification body, which is currently specifying the RET concept for the Release 6 [3]. As for mechanical downtilt, it depends on 2 factors: geometrical factor θ GEO and antenna vertical beamwidth factor θ VER, BW [17]: ( θ,θ ) ν = (3.3) e f GEO VER, BW Obviously, vertical beamwidth factor, BS antenna height and site spacing are as important as in the case of mechanical tilt for network capacity performance. Since all directions are tilted evenly, the efficiency of electrical downtilt in the struggle against inter-cell interference is very high. In general, EDT is able to provide slightly better results than mechanical downtilt in terms of system capacity improvements and mitigation of pilot pollution [30]. However, it should be kept in mind that excessive tilt values can drastically deteriorate coverage in the main lobe direction causing a significant drop in network coverage performance.

44 44 Figure 3.4 Horizontal radiation pattern in case of using electrical antenna tilt [17] Soft and softer handover Soft and softer handovers (SHO) are an essential part of WCDMA network functionality. While in soft handover a mobile terminal can be served by a few cells from different radio base stations at the same time. In the case of softer handover a terminal can be connected simultaneously to a few cells from the same base station. The set of cells to which a mobile terminal is simultaneously connected is called an active set (AS). The typical size of an AS is 3. In certain situations, soft and softer handovers give additional gain to the link due to macro diversity. Gain size is proportional to the differences in path losses between participating in handover cells. Smaller difference helps to achieve greater gain. The effect of soft handover in the WCDMA uplink direction for different level differences and radio channels has been analyzed in [32]. The output of research is SHO gain can be up to 4.5dB in the case of slow moving terminals and lower than 1dB in case of high (4-6 db) level difference depending on mobile speed. In softer handover the gain could be up to 1dB higher. According to SHO simulation result in the downlink in [20], the maximum gain of about 3dB can be observed. However, since 2-3 simultaneous transmission links are needed, the gain can also be negative. This proves the importance of having optimal cells in the AS. Besides, it can t be noticed that there is always tradeoff in SHO, the gain is achieved by means of extra power transmitted (mobile connected to a few base stations instead of one) in the network, which introduces additional interference to the network, and causes excess overhead on system resources. Therefore, the amount of mobile terminals in soft/softer handover must be kept to a reasonable number Softer handover factor As it has been already noted in sections , possible overlap between adjacent sectors of one site and other neighboring sites should be taken into account during antenna configuration selection. Generally, the overall coverage overlap of cells, or the soft handover area, is required to be kept below 30-40%, because excessive soft handover probabilities might decrease downlink capacity [16]. The reason is that each SHO connection increases interference in the network. When the generated interference level exceeds the diversity gain, the soft handover does not provide any gain for system performance anymore. Besides, every connection reserves additional system resources (in

45 45 RBS and RNC) in both downlink and uplink and transmission capacity in Iub link. At the same time it must be noted that low handover probability values might inhibit the system ability to support UE mobility. Thus, the task of radio planning and optimization is to keep the soft handover probability values below desired values Pilot pollution Pilot pollution is observed in areas in which there are more CPICH signals received at the mobile station s RAKE receiver than it is capable of processing, or none of the received CPICH signals is dominant enough [24]. Too many means that the number of MS CPICH signals received exceeds the active set size (see chapter 3.1.5). Every additional pilot polluter increases the interference level in the cell, reduces the received energy per chip over the power density ( E c N 0) from the serving cell, causing degradation in the quality of the active connection. In order to avoid pilot pollution, the number of unnecessary CPICH signals in dominant area should be kept as low as possible. In general, pilot pollution can be reduced by means of CPICH power tuning [34]; antenna configuration adjustments, i.e. antenna direction changing, downtilting. Implementation of multi-carrier solutions [11] or repeaters [31] can be also considered Antenna directions The requirements on antenna directions are set depending on traffic load and sectorization of the site (number of sectors on site). If in a nominal cell plan the sectors of a site are normally evenly spaced, i.e. in case of 3-sector site, the angle between the main directions of the antennas is 120, then in reality it is not always possible to configure the site with equally spaced sectors. In general, the choice of the direction is dependant on the particular radio environment and areas with potential traffic concentration. Nevertheless, it is very often recommended not to point an antenna directly towards a hill or a big obstacle like buildings and avoid small angle difference between adjacent sectors in order to avoid high interference levels in the cell Coverage overlap between adjacent sectors As has already been highlighted in sections , the antenna configuration has a great impact on network performance. Suitable antenna height position, beamwidth, sectorization, downtilt can aid to avoid high overlapping between cells and quality degradation in the network, as a consequence. However, in reality, one of the most widespread scenarios of coverage overlap arises, when antennas with quite wide horizontal antenna beamwidth (as a rule, 65 ) are in use and the azimuths of two sectors are too close to each other. In such a case, it is almost impossible to avoid huge coverage overlap, which leads to excessive soft handover probabilities, pilot pollution and ping-pong effect.

46 46 Figure 3.5 Coverage overlap between adjacent sectors. In order to minimize the coverage overlap between adjacent cells, it could be recommended to keep the separation between the adjacent sectors less than a certain angle, dependant on the horizontal antenna beamwidth. If delta in azimuth between adjacent cells is less than 100, switching to antenna with 44 horizontal beamwidth should be considered [12]. 3.2 CAPACITY AND COVERAGE PLANNING Intuitively, capacity and coverage planning cannot be separated into 2 different processes because of their strong inter-correlation, especially in high traffic areas. Capacity and coverage planning activities dependency is expressed in cell size variation in WCDMA networks, also known as the cell breathing phenomenon. When a mobile user is located near the cell edge, its transmitting power will increase according to the load and interference level in the cell, and soon reaches the maximum allowed power per connection. The other users closer to the BS location will overshoot a user at the cell edge, preventing him from connection establishment and, thus, decreasing cell size. In other words, cell size will vary depending on the number of the active users in the cell and their location, services and data rates they are using. Then the objective of coverage planning is to ensure availability of the chosen services in the entire planning area. The maximum capacity in the network depends on the coverage area and network element s configuration (radio hardware configuration, site and antenna configuration). Radio network planning cannot be properly done without using sophisticated planning tools Planning tool Since WCDMA systems are interference limited systems, the interference and capacity estimation play important role during coverage prediction process. Each user influences other users, causing everyone s transmit power to change continuously until they stabilize, i.e. users cannot be separated during analysis and must be considered as a single whole. To compare, in the second generation systems detailed planning concentrated mostly on coverage planning. As a consequence, mobile users speed, radio channel profile, type of services play a much more important role in 3G system planning than in 2G.

47 47 One more major difference between the 2 nd and 3 rd systems planning is the importance of the traffic layer allocation. In addition, QoS targets for all the particular services have to be met during the simulation process. In order to handle complicated behavior of mobile users in 3G, WCDMA iterative analysis based on Monte Carlo simulations can be used for detailed coverage and capacity planning. Figure 3.6 shows the common radio planning approach by means of WCDMA planning tool simulations in case of existing GSM network as an input for 3G network. Coverage/QoS/Capacity/ requirements Re-use of an existing 2G sites as an input for 3G network WCDMA planning tool simulations Requirements fulfilled? No Remove/add sites, reconfigure existing ones END Yes Figure 3.6 Planning tool simulations approach. After initial dimensioning is done, coverage thresholds set and traffic/capacity figures estimated, the number of the most appropriate existing GSM sites has to be chosen as an input for 3G co-sites and WCDMA planning tool simulations in order to meet target requirements (coverage and capacity). During the simulation process necessary configuration and parameters adjustments are to a certain extent being made. The reason is that very often 2G/3G co-site share the same multiband antenna, so antenna configuration change (azimuth re-direction, mechanical tilt) can affect an optimized live GSM network, which could be unacceptable in many cases. As a result of continuous simulation activities, part of the chosen existing candidate sites can be eliminated from the plan as unacceptable for the 3G network or replaced with new candidates from existing GSM network sites or replaced with totally new sites. The process continues until the set targets are met.

48 Monte Carlo analysis of TEMS Cell Planner Universal The Monte Carlo simulator of TEMS CellPlanner Universal (TCPU) is designed to reflect as closely as possible the UTRAN behavior in terms of setting up, managing and releasing user connections [35]. Figure 3.7 TCPU simulations steps [35]. In short, the process of Monte Carlo simulations, shown on the Figure 3.7 contains following steps: Depending on the traffic demand defined for each WCDMA bearer, users are generated at random locations according to Poisson distribution. Connection attempt for specific MS and WCDMA bearer is made according to the list of cells ranged in order by the selected ranking algorithm. The achieved C/I for each user is calculated based on the uplink and downlink power settings and the interference level known from the previous iteration.

49 49 Over several iterations the transmitted power of the cells and the users is updated in order to match as closely as possible the achieved C/I to the target C/I, which is calculated from the user-defined uplink and downlink Eb/Io values and spreading factors used of particular WCDMA bearer. Users, exceeded user-defined thresholds, such as: maximum number of DL or UL ASE, maximum UL interference (noise rise), maximum DL power limit, maximum number of users on particular spreading factors etc. are being disconnected. Once a possible overload situation is resolved for all the cells, the system must verify if a stable state in the network has been achieved, i.e. the power changes are minimal between the iterations. The result of Monte Carlo simulations is expressed in a number of plots and statistic reports reflecting the planned network status. The process of simulations is described in [35] in more detail Load estimation Before starting the WCDMA analysis, the expected load estimation with respect to the given traffic demand has to be done. The total traffic demand per service can be defined as follows: CS services where Traffic Traffic _ demand N sub sub _ demand cell = (3.4) N cell Traffic _ demand cell - the traffic demand per cell in E/cell, Traffic _ demand sub - the traffic demand per subscriber in me/sub, N sub - the total amount of subscribers, N - the total amount of cells. cell PS services where Total Traffic _ demand 8bits / byte sub _ traffic _ demand area = (3.5) (3600s / hour) N sub Total _ traffic _ demand area - the total traffic demand in planned area in kbps, Traffic _ demand sub - the traffic demand per subscriber in kb/hour/sub.

50 50 Traffic Total _ traffic _ demand area _ demand cell, kbps = (3.6) N cell where Traffic _ demand cell, kbps - the traffic demand per cell in kbps/cell. Traffic _ demand cell Traffic _ demand cell, kbps = (3.7) Service _ throughput where Service _ throughput - the service throughput in kbps. It must be noticed that since in the Monte Carlo simulation of the TCPU, each user is assumed to be using the network continuously, one Erlang is supposed to represent one user.

51 51 4 STUDY CASE 4.1 SIMULATION ENVIRONMENT Study case comprises two different scenarios. A planned network based on existing GSM sites locations and antenna configurations of one of operators in Guanzhou province has been observed in the first scenario. In other words, 3G sites antenna configurations (height, antenna directions and tilts) of the WCDMA network fully coincide with existing GSM sites antenna configurations. The idea of such an approach is quite common to use existing GSM network s radio plan entirely as an input for future 3G network, which, however, could very often be inappropriate in case of 3G network planning. The network consists of a 20 3-sector base stations with 750m site-to-site distance in average and covers a total area of km2 (Figure 4.1). The antenna installation height is on average 34.5m AGL. According to the coverage calculations presented in Section and , the chosen amount of sites should allow us to fulfill 95% probability of coverage requirements for the speech and CS64 services. Figure 4.1 First scenario network plan. In the second scenario a more advanced and extensive radio network planning approach has been considered. Antenna configurations of sites have been adopted for the existing environment and WCDMA specifics have been taken into account. Moreover, 3 sites have been discarded and only one new site has been added from/to the initial plan based on the GSM sites. Thus, the modified radio plan with 18 3-sector sites covering the same area as the first scenario, but still fulfilling requirements is presented (Figure 4.2).

52 52 Figure 4.2 Second scenario network plan. Monte Carlo simulations from the TEMS Cellplanner Universal (TCPU) planning tool have been used in both cases in order to perform WCDMA analysis [35]. The digital map of 20m x 20m resolution including main terrain types and elevation data has been used for predictions. In spite of the fact that it is recommended to use Walfisch- Ikegami model in case of cell size less than 1km (Section ), propagation results have been computed by means of Ericsson 9999 propagation model [12] based on Okumura- Hata propagation principles (Section ). The reason for that is that the model was tuned particularly for this medium urban environment based on prior radio measurements and gave relatively precise results in comparison with the other models.

53 General simulation parameters General simulation parameters can be found in Table 4.1. Table 4.1 Simulation parameters. Parameter Value Maximum Tx power 44.7 dbm CPICH Tx power 28 dbm BS noise figure 5 db UE maximum Tx power (speech/data) 21/24 dbm UE minimum Tx power -50 dbm Minimum E b N0 margin -16 dbm Speech service bit rate 12.2 kbps Video service bit rate 64 kbps Packet service DL/UL bit rate 128/64 kbps Channel profile ITU Vehicular A Mobile station speed 3 km/h SHO window 3 db Downlink orthogonality 0.67 Maximum active set size 3 BS antenna 65, 18 dbi Feeder cable losses 3 db Number of trials/iterations 20/ Traffic density information Random homogenous distribution of users has been used in predictions. Total amount of users in planned area was A service mix of voice users (12.2 kbps), video service users (circuit switched 64 kbps) and packet data users (128 kbps) has been assumed. Following traffic load as per service type has been used. Table 4.2 Traffic data input per service. Service type Speech Video Packet (128 kbps) Load per user 30 me 2 me 100 KB/hour/sub As a result of calculations in (Equations (3.4)-(3.7)), following load for each service per cell has been used during simulations. Table 4.3 Load level of services per cell. Parameter Load (E/cell) Speech service Video service 0.98 Packet service (128 kbps) 1.53

54 54 As can be seen from the Table 4.3, the average cell load is far below recommended values (see Section ). Overall cell load per cell (all services included) in the simulations seats around 30% of loading. 4.2 SIMULATION RESULTS First scenario CPICH coverage The first result that has to be checked is CPICH coverage. Figure 4.3 CPICH coverage plot (1 st scenario). The plot displays the downlink CPICH coverage quality in terms of Ec/Io, considering chip energy from the best serving cell and total DL interference received by a mobile terminal in a specific position in the network. Values are calculated as a margin of the minimum CPICH quality threshold set for WCDMA analysis with the parameter Min.DL Ec/Io for CPICH (db) [35]. The threshold value is the minimum CPICH value required for UE to detect and decode the pilot channel of the best serving cell. The plot shows the margin remaining until the minimum threshold value is reached. A positive margin means that a specific position in the network provides CPICH coverage with sufficient quality. A negative margin means that quality requirements for CPICH are not fulfilled in the position [35]. In our case, areas with negative or slightly above threshold positive values can be observed close to the edge of investigated area. The major reason is that sectors of sites located at

55 55 the border are over down tilted in order to avoid coverage overlap with sites out of the marked area. Since the possible effect of coverage overlap with sites located out of the investigated area is not part of our scope, insufficient CPICH coverage areas can be partially eliminated by means of reasonable uptilting of antennas of sectors pointing towards the edge of the observed area Soft handover probability Obviously, sites with higher antenna positions (see Figure 4.4) can create larger coverage overlapping areas as well, if antenna tilt values are not sufficiently optimized. Therefore, even more attention must be paid to the sites with high (higher then average) antenna heights Site-001 Site-002 Site-003 Site-004 Site-005 Site-006 Site-007 Site-008 Site-009 Site-010 Site-011 Site-012 Site-013 Site-014 Site-015 Site-016 Site-017 Site-018 Site-019 Site-020 Site name Figure 4.4 Real vs. average antenna height of the sites. In addition, it can be easily noticed that sectors of many sites are directed towards each other. Also, the angle between adjacent sectors of many sites is below recommended values (see Section ), (Figure 4.5), which logically results in large coverage overlapping areas, and hence increased soft handover probability (Figure 4.6).

56 56 Figure 4.5 Irregular antenna directions. Thus, one of the most important harmful outputs of non-optimized antenna configurations is expressed in high soft handover probability areas. Figure 4.6 Handover map (1 st scenario). Intuitively, the less site-to-site distance, the higher impact of tilt values on soft handover probability and network performance as a consequence. With lower downtilt angles, this correlation can be emphasized more. Relative dependency of height, downtilt values, spacing between adjacent sectors and soft handover probability for the worst performing cells are shown in Table 4.4.

57 57 Table 4.4 Worst soft handover probability cells. SectorID Antenna height, m MDT, angles EDT, angles Angle spacing, degrees HO probability, % Site-001A Site-003A Site-003B Site-003C Site-006C Site-008C Site-010B Site-010C Site-012C Site-013A Site-013B Site-014C Site-016A Site-016C Site-017B Site-020B Although, each component in the table plays its own role, it must be noticed that the effect of irregular antenna directions and angle spacing between adjacent sectors as well as antenna position height play the most significant role in high soft handover probability achievement. It is difficult to overestimate the importance of proper antenna directions and spacing between sectors, because further adjustments in operational network are very complicated and can lead to uncontrollable results, drastically deteriorating network performance. That is why proper radio network planning has to be done prior to network roll-out. Required actions for minimizing negative influence of this effect have been taken in the 2 nd scenario Influence of spacing between adjacent sectors Influence of spacing between adjacent sectors on SHO probability percentage based on worst performers results from Table 4.4 can be seen on the Figure 4.7. Notwithstanding that negative effect of small angle spacing between adjacent sectors is obvious, it is difficult to estimate precisely the share of particular bad spacing case in obtaining of excessive handover probability levels in terms of real environment. Individual approaches towards all the particular cells have to be taken in order to minimize harmful effect of excessive handover probability on network performance. However, the general recommendation about spacing between adjacent sectors for 65 antennas has to be followed as much as possible in the first place.

58 % 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% Spacing between sectors Figure 4.7 Correlation between adjacent sector spacing and SHO probability Statistics report WCDMA statistics report can be used to analyze statistics on number of user attempts, users served, users blocked and power consumption. Overall WCDMA statistics report per service can be found in Table 4.5 Table 4.5 Statistical results for system parameters (1 st scenario). Service #attempts %served #Insuff.CPICH #Insuff.DL #Blocked High Interf Video Speech Packet Summary The statistics can be used to verify the reason why users were blocked, which is very helpful for troubleshooting. As a result, possible ways of improving network design can be found. Since speech service priority is the highest, and then the blocked speech users can be used as the first indicator of problems in the network. Coverage holes on Figure 4.3 results in number of blocked users for users of different services due to insufficient CPICH coverage. According to the statistics results, there are no serious problems with high interference level in the network because of considerably small average DL and UL load used in predictions (around 30%).

59 59 Insufficient DL indicates lack of downlink transmission power for some cells. The reason is that excessive amount of active links have been performed in some particular cells (Figure 4.8). Figure 4.8 Downlink loading (1 st scenario). More detailed studies of this type of statistics per cell might give us even more valuable feedback. In general, analysis results of 1 st scenario network plan give us a valuable feedback for 2 nd scenario radio planning Second scenario The purpose of 2 nd scenario has been to improve network performance in comparison with 1 st scenario results with same or less amount of sites. Reasonable antenna azimuths re-direction, electrical tilt optimization has been as working tool in order to achieve set targets. As a result of consecutive antenna configuration adjustments and planning tool simulations, optimized radio plan have been worked out. Amount of used base stations has been decreased from 20 sites in original plan to 18 sites (3 sites discarded: Site-006, Site-014, Site-016 and one new added: Site-021). Since shared multi-band antenna usage can be a practical solution from an installation point of view and cost-effective from a financial point of view in the initial stage of network roll-out, the original antenna configurations of the existing GSM network have been kept on as many sites as possible.

60 CPICH coverage It can be seen that CPICH coverage has been slightly improved in comparison with first scenario CPICH coverage in the border areas. As a consequence, best server area for services has been improved as well (see Table 4.6). Figure 4.9 CPICH coverage plot (2 nd scenario). Table 4.6 Comparison of coverage areas of 1 st and 2 nd scenarios. Service type Coverage area, % Coverage area, % (1st scenario) (2nd scenario) Speech Video Packet The most significant coverage area improvement has been achieved for video and packet services. It is also important that CPICH coverage improvement did not lead to additional excessive soft handover probability.

61 Soft handover probability Figure 4.10 Handover map (2 nd scenario). There are only couple areas with considerably high soft handover probabilities can be found on the Figure 4.10, which is very important from saving of system resources point of view. It must be admitted that soft handover probability has significantly decreased after a few inappropriate sites removal from the original plan, some antennas re-direction and electrical tilt optimization % SHO probability in average in 2 nd scenario plan instead of 23.99% in original plan has been obtained. SHO probability values for worst performance cells before and after can be found in Table 4.7. There are no after results presented for sites Site-006, Site-014 and Site-016 because mentioned sites have been removed from the plan during 2 nd scenario radio planning process. Cells with higher than 30% SHO probability could be further investigated.