ICT STP TUCAN3G Wireless technologies for isolated rural communities in developing countries based on cellular 3G femtocell deployments

Transcription

1 ICT STP TUCAN3G Wireless techologies for isolated rural commuities i developig coutries based o cellular 3G femtocell deploymets D41 UMTS/HA etwork dimesioig Cotractual Date of Delivery to the CEC: 1st Aug 2013 Actual Date of Delivery to the CEC: 11th Nov 2013 Author(s): Josep Vidal, Jaume del Olmo, Adria Agusti, Atoio Pascual, Olga Muñoz (UPC), Omar Tupayachi, Miguel Agel Sachez, Carlos Garcia, Atoio Ueute (TdP) Participat(s): Workpackage: Est. perso moths: 3 Security: Dissemiatio Level: Versio: UPC, TdP WP4 PU PU Total umber of pages: 157 a Abstract: ased o the defiitio of the access etwork architecture ad the voice ad data traffic evolutio forecast, this documet derives egieerig rules for dimesioig the umber of ope-access femtos, as a fuctio of the frequecy bads adopted for access, as well as the solar cells uits. Dimesioig is based o service provisio ad eergy cosumptio evaluatio. Two deploymet cases have bee studied: the case of sigle HN with differet ateas types, ad the case of two co-located HN operatig at differet frequecies with directioal ateas. The cosidered deploymet sites are the five specified i documet D21 i the Napo river ad Paraapura river. The backhaul badwidth is determied for all cases. Keyword list: Cell plaig, etwork dimesioig, WCDMA, eergy supply dimesioig

2 Documet Revisio History DATE ISSUE AUTHOR SUMMARY OF MAIN CHANGES 15/08/2013 a UPC The first draft icludes dimesioig for voice traffic ad problem statemet ad aalysis for mixed voice ad data traffic 23/10/2013 aa UPC Draft versio with voice ad data traffic 30/10/2013 ab UPC, URJC Deliverable versio addressig all remarks by the techical coordiator (URJC) 11/11/2013 ae UPC Fial versio 1

3 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Executive Summary A etwork plaig exercise is doe i this documet for the locatios ad 3G voice ad data services evisioed i TUCAN3G. The etwork items to be decided i each locatio are the umber of HNs ad their cofiguratio, the associated backhaul ad the eergy uits (batteries ad solar paels). We make a umber of costraiig assumptios that iclude: 1) oly three types of HN as provided by ip.access, characterized by the trasmitted power ad umber of chaels 2) measured traffic models for voice ad data i rural areas of Peru, icludig its evolutio alog the day ad through 5 years. I each locatio, up to six scearios (defied with the type of atea ad umber of HN) are tested, with either oe or two HN ad differet atea cofiguratios. It has bee observed that HN deploymets o a sigle high tower (already available i the locatios) provide the coverage ad service, except for Sata Clotilde where 3G data traffic ca be served provided that voice traffic is routed through the existig GSM etwork. For each cofiguratio, three types of HN are assumed. Due to the limitatios i power ad umber of chaels i the provided ip.access equipmet, we have performed a complete study that icludes ot oly the cogestio probabilities associated to the limited umber of but also the impact of coverage associated to radio coverage ad to itracell iterferece both i UL ad DL. From the evaluated results, we draw recommedatios for each locatio, by adoptig the followig criteria. Amog all cofiguratios fittig the specified blockig probability of 2%, select: 1. For each year, choose the cofiguratio, which supposes a miimum umber of HNs, givig preferece to those oes that use HN or (as off-the-shelf products). 2. If several solutios are still valid, choose the oe requirig the least backhaul badwidth. 3. If more tha oe cofiguratio requires the same backhaul, the choose the cofiguratio with lower eergy cosumptio. 4. If more tha oe cofiguratio still survives, choose the oe with the lowest cost. 5. Wheever o cofiguratio usig HN or caot provide service for a particular year, chose HN. 6. Fially, if the cofiguratio ad/or HN has to be chaged startig from the secod year, this cofiguratio ad/or HN for the secod year will also be selected for the first year. Fially, sectios 8.2 cotai the backhaul requiremets for each locatio over the 5 years period, ad sectio 8.3 lists the materials to be purchased for activities i WP6. TUCAN3G_D41UPCae.doc 2

4 DISCLAIMER The work associated with this report has bee carried out i accordace with the highest techical stadards ad the TUCAN3G parters have edeavoured to achieve the degree of accuracy ad reliability appropriate to the work i questio. However sice the parters have o cotrol over the use to which the iformatio cotaied withi the report is to be put by ay other party, ay other such party shall be deemed to have satisfied itself as to the suitability ad reliability of the iformatio i relatio to ay particular use, purpose or applicatio. Uder o circumstaces will ay of the parters, their servats, employees or agets accept ay liability whatsoever arisig out of ay error or iaccuracy cotaied i this report (or ay further cosolidatio, summary, publicatio or dissemiatio of the iformatio cotaied withi this report) ad/or the coected work ad disclaim all liability for ay loss, damage, expeses, claims or ifrigemet of third party rights. 3

5 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Table of Cotets 1 INTRODUCTION STUDY CASES GENERAL DAY-TIME TRAFFIC EVOLUTION MODELS Voice traffic EARTH project voice traffic model Telefóica del Perú voice traffic model Data traffic Earth project data traffic model Telefóica del Perú data traffic model Adopted data traffic model LONG-TERM TRAFFIC EVOLUTION SOCIO-ECONOMIC CHARACTERISTIC OF THE TARGET SITES AND OFFERED TRAFFIC 22 3 SYSTEM ASSUMPTIONS ADOPTED CHANNEL MODELS Path Loss Model Small scale fadig Shadow fadig AIR INTERFACE ASSUMPTIONS Codes Physical chaels Power cotrol procedure Sychroizatio ad cell search procedure IP.ACCESS NODE ECIFICATIONS E-class HN S-class HN TECHNICAL SCENARIOS METHODOLOGY FOR ACCESS NETWORK PLANNING KEY PERFORMANCE INDICATORS RADIO NETWORK PLANNING FLOW CHART RADIO LINK UDGET Radio lik budget WCDMA parameters Chip Rate (W) lock Error Rate (LER) it rate (R) Activity factor (v) Orthogoality factor (α) Spreadig factor ad processig gai Eergy per bit over iterferece plus oise spectral desity E b /N o Eergy per chip over total received power spectral desity E c /I o Power Rise Power Cotrol Headroom WCDMA power allocatio Power allocatio i the uplik Power allocatio i the dowlik Path losses evaluatio Pilot coverage requiremets Lik quality evaluatio TUCAN3G_D41UPCae.doc 4

6 Target E b /N o values Defiitio of the probability of coverage QUEUING MODELS, LOCKING AND CONGESTION PROAILITIES Voice service M/M/m/m queuig system Two parallel M/M/m/m queuig system idimesioal M/M/m/m queuig system Mixed voice ad data service D M/M/m/m queuig system Parallel 2D M/M/m/m queuig system D M/M/m/m queuig system NETWORK PLANNING RESULTS NAPO RIVER REGION Sata Clotilde Atea cofiguratio CPICH coverage lockig ad cogestio probabilities Negro Urco Atea cofiguratio CPICH coverage lockig ad cogestio probabilities Tuta Pisco Atea cofiguratio CPICH coverage lockig ad cogestio probabilities PARANAPURA RIVER REGION Sa Gabriel Atea cofiguratio CPICH coverage lockig ad cogestio probabilities Sa Jua Atea cofiguratio CPICH coverage lockig ad cogestio probabilities EVALUATION OF ACKHAUL REQUIREMENTS ACKHAUL REQUIRED FOR VOICE TRAFFIC Ideal case (uit coverage probability) Realistic evaluatio Sata Clotilde Negro Urco Tuta Pisco Sa Gabriel Sa Jua ACKHAUL REQUIRED FOR MIXED VOICE AND DATA TRAFFIC Ideal case (uit coverage probability) Realistic evaluatio Sata Clotilde Negro Urco Tuta Pisco Sa Gabriel Sa Jua ENERGY PROVISION FOR HN

7 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig 7.1 POWER CONSUMPTION MODELS OF HN RADIATED POWER EVALUATION DIMENSIONING OF ENERGY UNITS Geeral evaluatio methodology Applicatio to TUCAN3G sites Dimesioig for voice traffic Dimesioig for mixed voice ad data traffic CONCLUSIONS AND RECOMMENDATIONS SOLUTIONS FOR EACH LOCATION Sata Clotilde Negro Urco Tuta Pisco Sa Gabriel Sa Jua RECOMMENDATIONS TO WP RECOMMENDATIONS TO WP ANNEX VALUES OF LINK UDGET PARAMETERS PATH LOSS OPTIMISATION RESULTS Sata Clotilde Negro Urco Tuta Pisco Sa Gabriel Sa Jua TUCAN3G_D41UPCae.doc 6

8 Refereces [3GPP TS ] [3GPP TS ] [3GPP TS ] [Awoiyi03] [ertsekas 92] [EARTH-D23] [ETSI TR ] [Holma00] [ITU-R P ] [ITU-R P.527-3] [ITU-R P.832-3] [Kleirock 75] [Laiho01] [NTIA ] [RAD] 3GPP TS ; Techical Specificatio Group Radio Access Network; User Equipmet (UE) radio trasmissio ad receptio (FDD), Release 11, v11.5.0, April GPP TS ; Techical Specificatio Group Radio Access Network; ase Statio (S) radio trasmissio ad receptio (FDD), Release 11, v11.5.0, April GPP TS ; Techical Specificatio Group Radio Access Network; Requiremets for support of radio resource maagemet (FDD), Release 11, v11.4.0, April O. Awoiyi, N.. Mehta, ad L.J. Greestei, Characterizig the orthogoality factor i WCDMA dowliks, IEEE Tras, Wireless commuicatios, vol. 2, o 4, p , D. ertsekas, R. Gallager, Data Networks, Secod Editio, Pretice Hall, New Jersey G. Auer, et al, Eergy efficiecy aalysis of the referece systems, areas of improvemets ad target breakdow, deliverable report D23, ICT EARTH project, Jauary Techical Report. Uiversal Mobile Telecommuicatios System (UMTS); Deploymet aspects (3GPP TR v release 11). October H. Holma, ad A. Toskala, Wcdma for Umts (Vol. 4). New York, Wiley, Recommedatio ITU-R. P , The radio refractive idex: its formula ad refractivity data, P Series, Radiowave progagatio, February Recommedatio ITU-R. P.527-3, Electrical characteristics of the surface of the Earth, P Series, Radiowave progagatio, Recommedatio ITU-R. P.527-3, World atlas of groud coductivities, P Series, Radiowave progagatio, February L. Kleirock, Queueig systems. Volume I: Theory, Joh Wiley & Sos, 1975 J. Laiho, A- Wacker, ad T. Novosad, editors. Radio Network Plaig ad Optimizatio for UMTS. Joh Wiley & Sos Ltd., G.A. Hufford, A.G. Logley, ad W.A. Kissick. A guide to the use of the ITS irregular terrai model i the area predictio mode, techical report , NTIA,

9 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig [Redo11] [SCF ] [Sipila99] [SRTM] [TUCAN3G D21] [Wag00] A. Redó, P. Jeaeth, A. Martíez, Tecologías de la Iformació y las Comuicacioes para zoas rurales. Aplicació a la ateció de salud e países e desarrollo, CYTED 2011 Documet , Extedig rural ad remote coverage usig small cells, Small Cells Forum, Feb 2013 K. Sipila, J. Laiho-Steffes, A. Wacker, M. Jasberg, "Modelig the impact of the fast power cotrol o the WCDMA uplik", Vehicular Techology Coferece, 1999 IEEE 49th, vol.2, o., pp.1266,1270 vol.2, Jul A. Garcia, et al., Socio-ecoomic scearios, techical specificatios ad architecture for the proof of cocept, deliverable report D21, ICT STP TROPIC, May 2013 Y.-P.E. Wag, ad T. Ottosso, "Cell search i W-CDMA", Selected Areas i Commuicatios, IEEE Joural o, vol.18, o.8, pp.1470,1482, Aug TUCAN3G_D41UPCae.doc 8

10 List of abbreviatios & symbols 3GPP 3 rd Geeratio Partership Project AMR Adaptive Multi-Rate AWGN Additive White Gaussia Noise LER lock Error Rate PSK iary Phase-Shift Keyig CPICH Commo Pilot Chael CRC Cyclic Redudacy Check DL Dowlik DPCCH Dedicated Physical Cotrol Chael DPDCH Dedicated Physical Data Chael FDD Frequecy Divisio Duplexig HN Home Node HSDPA High-Speed Dowlik Packet Access HSUPA High-Speed Uplik Packet Access OVSF Orthogoal Variable Spreadig Factor PRACH Physical Radom Access Chael QAM Quadrature Amplitude Modulatio QoS Quality of service QPSK Quadrature Phase-Shift Keyig RF Radio frequecy SF Spreadig Factor SINR Sigal to Iterferece plus Noise Ratio UE User equipmet UL Uplik UMTS Uiversal Mobile Telecommuicatio System WCDMA Widebad Code Divisio Multiple Access Orthogoality factor E b /N o Eergy per user bit divided by oise plus iterferece spectral desity E c /I o Eergy per chip divided by the total received power spectral desity ms G UE atea gai bs G HN atea gai P m Trasmitted power by termial m to termial r P m Received power at termial m from termial P S i Total trasmitted power by HN i r P S Total received power from HN i i ( ) P i Received CPICH power from HN i comch P i Trasmitted power by HN i for dowlik commo chaels R Service rate SM Shadowig margi 2 Thermal oise power v Activity factor 9

11 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig W Chip Rate Received iter-cell iterferece i the dowlik Received iter-cell iterferece i the uplik TUCAN3G_D41UPCae.doc 10

12 1 INTRODUCTION Network plaig is the first stage towards the deploymet of 3G services evisioed i TUCAN3G. This activity is sustaiig busiess case study i WP3 (i terms of the evaluatio of busiess case study), the requiremets for the backhaul i WP5 ad the estimatio of equipmet eeded for the platform i WP6. ased o the defiitio of the access etwork architecture ad the traffic evolutio forecast, as defied i WP2, we derive egieerig rules for dimesioig the HN-based access etwork satisfyig the requiremets of coverage ad grade-of-service. A umber of possible techical solutios (assumig the operator has the exploitatio rights of two carriers) are evaluated for actual voice ad data traffic models provided from etwork measuremets i the Telefóica del Peru etwork, ad for socioecoomic parameters of the commuities ivolved i Napo ad Paraapura river. The positio of the HN is ot optimized, istead it is assumed that HNs are placed o top of the existig towers deployed by PUCP ad EHAS i the regio. The documet is orgaized as follows: sectio 2 describes the traffic (voice ad data) assumptios for the target commuities i rural areas, as well as their geographical ad socio-ecoomical characteristics. Sectio 3 cotais a detailed descriptio of the system assumptios cotemplated i the study, icludig the chael models, air iterface ad HN relevat deploymet specificatios. Sectio 4 describes i detail the methodology used to evaluate the key performace idicators, usig the Radio Mobile [RAD] freeware, the lik level evaluatio priciples ad queuig theory, whe mixed voice ad data traffic are cotemplated. Sectio 5 lists the etwork plaig results for all the target commuities. Sectio 6 cotais a evaluatio of the backhaul requiremets for each site. I sectio 7, a study of the eergy supply uits dimesioig based o solar cells is give, i which eergy cosumptio models are used for the HN cosumptio. Fially, Sectio 8 collects the coclusios ad recommedatios geerated i this activity, to be forwarded to WP5 ad WP6. 11

13 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig 2 STUDY CASES This sectio describes the characteristics of traffic assumed for all sites uder cosideratio. To that ed, we first itroduce some cosideratios o voice ad data usage observed i the EC-fuded EARTH project. This project is focused o rural areas i Europe so the results reported eed to be cosidered carefully, ad adopted wheever the measuremets take by Telefoica del Peru (TdP) i rural areas i Peru are ot cocludig. Figure 1 presets the percetage of populatio o differet areas of Peru that have a cellphoe. I the target rural areas, we assume that 53% of the populatio has oe cell phoe. Additioally, i the rural areas the umber of populatio i itieracy teds to be sigificat, i this regard we ca assume that the total populatio of a give commuity is elarged by 20%. Figure 1. Households with access to TIC by poverty level ad type of residetial area (accordig to INEI reports, 2011) 2.1 Geeral day-time traffic evolutio models Models for the data user activity, voice traffic ad data traffic are provided i [EARTH-D23] for differet areas i Europe. The EARTH-project defies the followig categories: Dese Urba Areas: 3000 citize/km 2 o average Urba Areas: 1000 citize/km 2 o average Suburba Areas: 500 citize/km 2 o average Rural Areas: 100 citize/km 2 o average Sparsely populated & wilderess : 25 citize/km 2 o average Additioally, the EARTH-project characterizes the distributio of the areas associated to each category: Dese Urba Areas: 1% Urba Areas: 2% Suburba Areas: 4% Rural Areas: 36% Sparsely populated & wilderess : 57% TUCAN3G_D41UPCae.doc 12

14 2.1.1 Voice traffic EARTH project voice traffic model Although the voice activity may differ from oe regio to aother i the world, we cosider i this sectio, the model elaborated i the EARTH-project; for completeess ad compariso purposes, that takes ito accout the typical Europea rural user behaviour. The model assumes a voice traffic daily profile preseted i Figure 2 (averaged coveietly over may users ad scearios) ad a voice usage of 180 miutes/moth/subscriber. Note that the peak is ot a sigle hour, but there is a sigificat activity betwee 9 AM ad 19 PM. 0.1 Hourly percetage of the total daily voice traffic (%) Time (hours) Figure 2. Day-log model for the percetage of the total daily voice traffic for rural populatio i Europe The percetage of traffic durig the busy hour show i Figure 2 (10 AM) is the 8% of the total traffic of the day, which correspods to the followig offered traffic i mili Erlags per subscriber: v mi 1 hour 1 moth v h mer moth 60 mi 30 day v We obtai a average arrival call rate of v = calls/s, ad average served call rate of v = calls/s for a average call duratio h = 90 secods, The average daily traffic profile per user ca be obtaied by scalig Figure 2 so that the maximum at the busy hour is 10 mer. The voice traffic characterizatio fits a Poisso arrival law (the time betwee ew icomig calls follows a expoetial distributio) ad expoetial service time Telefóica del Perú voice traffic model The first voice traffic measuremets performed by Telefóica del Peru (TdP) i the rural areas of Peru reveal that the EARTH model could ot be accurate eough for our purposes. TdP provides a estimated voice traffic model for the localities selected i sectio 2.3. These have bee take from real traffic measuremets obtaied i FITEL s project Rural ad Social Preferet Areas Itegratio to the Mobile Service Network, started o Jauary 2012 through Jue I this project, the villages were grouped i four categories: Category 1: Populatio p 500 ihabitats, average is 242 ihabitats. Category 2: Populatio 500 p 1000 ihabitats, average is 717 ihabitats. 13

15 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Category 3: Populatio 1000 p 1500 ihabitats, average is 1227 ihabitats. Category 4: Populatio 1500 p ihabitats, average is 1985 ihabitats. The average voice traffic measuremets i Erlags as a fuctio of the time of the day is depicted i Figure 3, i the first year (year 0), i Jue 2013 (year 2) ad i the 4 th year, see Table 3. Figure 3. Voice traffic measuremets for the four categories of villages With the objective of derivig the hourly daily traffic per mobile termial or user, we have to take ito accout that, accordig to [TUCAN3G-D31]: 1. The mobile phoe peetratio i Peru is 53% 2. The total populatio i the villages cosidered for the voice traffic measuremets ca icrease up to 20% due to the itieracy 3. TdP is the oly providig 2G voice traffic services i the area of measuremets Hece, processig accordigly the measured data show i Figure 3, we ca obtai the hourly daily voice traffic per mobile termial preseted i Figure 4 for the four categories of villages. We ca observe that the voice traffic per user i the villages of categories 1 (p 500), 2 (500<p 1000) ad 3 (1000 < p 1500), i.e. up 1500 ihabitats, is quite similar gettig aroud 12 merl i the busy hour (19.00 PM). I cotrast, i the villages for more tha 1500 ihabitats, the traffic geerated per user teds to be smaller tha i the other categories villages, for example 9.4 merl i the busy hour. TUCAN3G_D41UPCae.doc 14

16 0.012 Hourly daily traffic per user (Erlags) p < p < p 1500 p Time (hours) Figure 4. Hourly daily voice traffic profile per mobile termial I order to provide a simple voice traffic model to be represetative for all rural villages i Peru, we propose a sigle hourly daily traffic per user for all villages i Table 1. It has bee obtaied by scalig the average voice traffic profile of villages with less tha 1500 ihabitats i order to have 10 merl at the busy hour. For a average call duratio of 90 secods, we obtai a average arrival call rate of v = calls/sec, ad average served call rate of v = calls/sec. Time (hours) merl Time (hours) merl Table 1. Voice traffic model per user for the rural areas i TUCAN3G Figure 5 shows the compariso betwee the traffic voice model defied for the TUCAN3G project ad the oe defied i the EARTH project. It is importat to highlight the differeces betwee the EARTH Europea model ad the adopted model i TUCAN3G. 15

17 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Hourly voice traffic per user (Erlags) TUCAN3G-model EARTH-model Time (hours) Figure 5. Compariso of voice traffic models obtaied i TUCAN3G ad i EARTH projects Data traffic Earth project data traffic model With the objective of characterizig the volume of data traffic, it is importat to elucidate the umber of active users at every momet. The EARTH project foud two importat properties of the total traffic evolutio: The daily variatio of the umber of active users is proportioal to the daily variatio of the traffic % of the data subscribers are active i the busy/peak hours i today s etworks. The EARTH project assumes a referece value of 16%. O average over a day, the activity level is 9.64%. Figure 6 depicts the referece curve cosidered by the EARTH project to model the percetage of active data users alog the day. Hourly percetage of the total umber of data users(%) Time (hours) Figure 6. Day-log model for the percetage of the total data users per hour for rural populatio i Europe TUCAN3G_D41UPCae.doc 16

18 The EARTH project also provides the followig estimatio o the type of termials ad services for the year 2015 i the Europea regio (assumed all populatio are subscribers of data services): Mobile PC 20% of the populatio Smartphoes 50% of the populatio Tablet 5% of the populatio The remaiig 25% of populatio hold voice-oly cellphoes. Two differet types of users are cosidered: Heavy user requests a hourly average data rate of 2 Mbps, 250 Kbps ad 1 Mbps for Mobile PC, Smartphoe ad Tablet, respectively. Ordiary user requests a hourly average data rate of 250 Kbps, Kbps ad 125 Kbps for Mobile PC, Smartphoe ad Tablet, respectively. These user average data rate is assumed to be measured as the total umber of bits received by all active users durig the busy hour (3600 secods) divided by the umber of active users, r N N users bits 3600 bits / s y tuig the ratio of heavy ad ordiary users, differet scearios ca be cosidered: Sceario A (sceario #4 i [EARTH-D23]): Cotemporary traffic demad. 10% of the populatio is referece as PC user, of which 10% are classified as heavy users usig 125 Kbps. The remaiig 90% users are ordiary ad cosume 31 kbps. Sceario (sceario #1 i [EARTH-D23]): Most relevat sceario for % of the subscribers are classified as heavy users. Sceario C (sceario #2 i [EARTH-D23]): 50% of the subscribers are classified as heavy users. Sice the data volume per subscriber does ot deped o the deploymet sceario, the area traffic demad for a give deploymet is give by (i Mbps), k k R t p t r s (1) where r k ad s k are the average data rate demad ad the ratio of subscribers for termial type k respectively, α(t) deotes the percetage of active data users (see Figure 6),ad p is the populatio (i umber of users). k Telefóica del Perú data traffic model I early 2013, TdP started to deploy 3G i some specific rural places i Peru. However at the time of writig it was too early to determie a data traffic model for rural areas from the measuremets obtaied by TdP. To this ed, it was decided to build a data traffic model from the measuremets performed over several small urba areas with less tha 3000 ihabitats. From those measuremets we ca obtai a ormalized hourly daily data traffic, which could be used for rural areas. The actual values of the traffic alog the day will be obtaied as a fuctio of the umber of ihabitats, percetage of active users durig the busy hour, user peetratio, ad average traffic per user. 17

19 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Data traffic model i urba areas The measuremets have bee performed i a sub-urba/urba area where there are 26 ase Statios (Ss). Each S is equipped with both 3G ad 2G techologies. Figure 7 presets the aggregated umber of bits (dowlik plus uplik) at each hour served by all the Ss for 2G ad 3G techologies, respectively. Mbits Mbits Figure 7. Aggregated served data traffic for dowlik ad uplik per hour (over the 26 Ss) i Mbits The followig cosideratios have to be take ito accout for the iterpretatio of the obtaied measuremets: The estimated total populatio that uses 3G is ihabitats while the estimated populatio served usig 2G is ihabitats. The differece is due to larger coverage area of 2G techology. Oly TdP is operatig i 2G techology. However, there are other mobile operators usig 3G techology i the area of measuremets. The peetratio ratio of TdP i that area is aroud 60%, i.e. the measured 3G data traffic correspods just to the 60% of the total mobile traffic i that area. The target populatio for data traffic i urba/sub-urba areas is 25%, i.e. 25% of the mobile termials is estimated to use data traffic. The estimated 3G traffic per mobile termial is i average 15 kbps i dowlik ad 5 kbps i uplik. At the time of writig there is ot a estimatio of the average 2G traffic. The measuremets are give i terms of total bits received/trasmitted at each hour. They also iclude the total raw bits (payload) trasmitted/received i the dowlik ad uplik trasmissio, i.e. without cosiderig the sigalig overhead. Figure 8 compares the ormalized daily traffic profile for 3G ad 2G traffic i the same area. It ca be observed that both traffics follow a similar patter, clarifyig also that the busy hour for data traffic (at PM PM) is differet from the busy hour for voice traffic (aroud 19 PM i Figure 4). For the sake of a complete compariso, Figure 9 presets the data traffic patter profile selected for TUCAN3G ad it is compared with the oe obtaied i the EARTH project. TUCAN3G_D41UPCae.doc 18

20 G traffic 3G traffic ormalized Hourly daily data traffic (%) Time (hours) Figure 8. Normalized hourly daily data traffic profile ormalized Hourly daily data traffic (%) TUCAN3G-model EARTH-model Time (hours) Figure 9. Normalized hourly daily traffic profile for TUCAN3G ad EARTH projects for urba ad suburba areas Regardig the umber of active data termials i the peak hour, i the EARTH project it was observed that this value could rage betwee 10% ad 30%. Whe derivig this value for the Peruvia urba/sub-urba area from 3G traffic measuremets: 1. The total traffic measured i the busy hour is Mbits, i.e. aroud 22 Mbps. 2. The total populatio i the area is ihabitats. 3. It is assumed that 53% of the total populatio has a mobile termial, i.e mobile termials. 19

21 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig 4. The measured traffic oly cosiders the 60% of the whole umber of mobile termials due to the user peetratio of TdP, i.e mobile termials are subscribed to TdP. 5. It is assumed that 25% of mobile termials will use data traffic i urba/semi-urba areas, i.e mobile termials will geerate data traffic. 6. The average data traffic per user i the busy hour is assumed to be 15 kbps for dowlik ad 5 kbps for uplik, i accordace with Small Cells Forum documets [SCF ]. 7. Hece, the average umber of active mobile termials will be 30% of the all mobile termials i busy hour. Data traffic model i rural areas For rural areas we will adopt the same data traffic profile as the oe show i Figure 9. However, the followig aspects will be assumed i cotrast to the urba/semi-urba areas: 1. Target mobiles termials for data traffic 5%. 2. Itieracy 20% Hece, the aggregate data traffic demad for a rural area where oly oe operator is active will be give by, Mbps R t p t r (2) where t is the time i hours, p is the total umber of ihabitats, factor 0.53 is the mobile termial peetratio rate, 1.2 is associated to the icrease of ihabitats due to itieracy, factor 0.05 assumes that the target of mobile termials for data traffic is 5%, factor 0.3 takes ito accout the fractio of active termials durig the busy hour, β(t) is the ormalized hourly data traffic show i Table 2 (also show i Table 2) ad fially r is the average data rate demad per user (15 kbps for dowlik ad 5 kbps for uplik). Table 2. Normalized rural data traffic model i TUCAN3G, (t) Adopted data traffic model Adoptig the values metioed above, FTP model 2 (see sectio i [TUCAN3G D21]) is assumed for the data traffic characterizatio, where the packet arrival fits a Poisso law ad size is costat Kb/packet. Eve though the service time is determiistic, it is still possible to model the traffic usig a birth-death Markov model [ertsekas 92]. The per-user average umber of packets i the busy hour is give by: TUCAN3G_D41UPCae.doc 20

22 C Kbps d (3) Kb/packet where C=15 Kbps for the dowlik ad C=5 Kbps for the uplik. Assumig that packets are delivered usig 128 kbps service i UMTS, the trasmissio time to deliver oe packet is: Kb/packet Ts (4) 128 Kbps I order to determie the packet service rate, the probability of departure from oe state i the Markov model i a time frame with duratio has to be evaluated.. Assume A(t) is a coutig variable that represets the total umber of arrivals that occurred from time 0 to time t, ad hece Pr{A(t T s +) A(t T s ) = 1} is the probability of oe packet arrival i the iterval [t T s, t T s +]. The, Pr Packet delivered i tt, server busy at t t Pr Packet delivered i tt,,server busy at t Prserver busy at PrPacket delivered i tt, PrAtTs AtTs 1 d expd Prserver busy at t Pr At At T 1 1expdTs s (5) If the time frame is small eough compared to 1/ d we ca approximate (5) by: T T d exp d d PrPacket delivered i tt, server busy at t d 1exp 1exp d s d s (6) Note that 1/ d is the average service time per packet, which turs out to be T s uder the coditio of the packet geeratio rate d beig much smaller tha 1/T s. This is also the average delay per packet at low load, which we would like to keep below a reasoable value for a good user experiece: 1expdTs C kbps 1 exp C kbps /128 Kbps d A value of 2.6 secods is associated to a FTP packet size of 512 Kb for the dowlik, ad 387 Kb i the uplik (which we roud up to 512 Kb for simplicity). The d = s -1 ad d = s -1 are calculated for the dowlik, ad d = s -1 ad d = s -1 for the uplik. The assumptio d << 1/T s i (6) is approximately matched, where i i our case d /T s = for the dowlik ad d /T s = for the uplik). 2.2 Log-term traffic evolutio Table 3 describes the evolutio of traffic over the followig 5 years after the deploymet of wireless services, which is calculated usig the actual measuremets from Telefóica del Perú 2G ad 3G etworks aual pla. Year Traffic icrease Commet 2 180% Moth 13th of operatio (maily i the first 6 moths of operatio) 3 5% Moth 25th of operatio (estimated from the first six moths of real traffic) 4 2% Estimated from Telefóica del Perú s Aual Plaig 5 2% Estimated from Telefóica del Perú s Aual Plaig Table 3. Log-term traffic evolutio i rural areas i a 4 years forecast 21

23 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig 2.3 Socio-ecoomic characteristics ad offered traffic of the target sites This sectio defies the locatios where the HN will be deployed i terms of geographical coordiates of towers, areas where populatio live ad the traffic geeratio probabilities depedig o the specific areas for each village, socio-ecoomical parameters icludig populatio, itieracy, approximate average icomes, etc. Fially, expected voice ad data traffic values for each locatio are preseted. The locatios where HN will be deployed are decided to be Sata Clotilde, Negro Urco ad Tuta Pisco i Napo river regio, ad Sa Gabriel ad Sa Jua i Paraapura river regio. Table 4 shows the GPS coordiates of the towers i each locatio. Locatio Tower GPS coordiates Height of towers Sata Clotilde 02º29'22.4''S 73º40'40.7''W 72 m Negro Urco 03º01'23.1''S 73º23'31.5''W 75 m Tuta Pisco 03º06'31.4''S 73º08'17.5''W 57 m Sa Gabriel 05 42'49.99"S 76 24'39.59"W 60 m Sa Jua 05 52'35.13"S 76 21'21.73"W 60 m Table 4. Geographical coordiates of towers for each locatio where HN for access etwork will be istalled Rural data ad voice traffics have bee characterized i previous sectio for rural commuities. Moreover, we assume that the traffic geerated is ot homogeeous geographically so differet areas for each locatio are associated to differet probability of traffic geeratio. These probabilities are idepedet from the assumed traffic i 2.1, rather, they are characterize the desity of houses as observed o the map, ad hece, of active termials. Figure 10 to Figure 14 show the itesity of expected traffic geeratio assumed for each locatio, as well as the positio of the towers (displayed as red poit). Areas with a itesity of 1 refer to maximum expected traffic geeratio (probability of active users), while lower itesities are refereced to this expected maximum value. As a result, differet probability distributios of havig active users are created for each locatio. (a) Sata Clotilde locatio ad tower positio (b) Itesity of traffic assumed Figure 10. Itesity of traffic geeratio assumed over the geographical area i Sata Clotilde, obtaied from the observatio of household locatios o Google Earth images TUCAN3G_D41UPCae.doc 22

24 (a) Negro Urco locatio ad tower positio (b) Itesity of traffic assumed Figure 11. Itesity of traffic geeratio assumed over the geographical area i Negro Urco, obtaied from the observatio of household locatios o Google Earth images (a) Tuta Pisco locatio ad tower positio (b) Itesity of traffic assumed Figure 12. Itesity of traffic geeratio assumed over the geographical area i Tuta Pisco. White (yellow) areas o the left plot correspod to higher (lower) desity of households (idetified from field observatios) 23

25 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig (a) Sa Gabriel locatio ad tower positio (b) Itesity of traffic assumed Figure 13. Itesity of traffic geeratio assumed over the geographical area i Sa Gabriel. White (yellow) areas o the left plot correspod to higher (lower) desity of households (idetified from field observatios) (a) Sa Jua locatio ad tower positio (b) Itesity of traffic assumed Figure 14. Itesity of traffic geeratio assumed over the geographical area i Sa Jua locatio. White area o the left plot correspod to the locatio of households (idetified from field observatios) Fially, the expected aggregated traffic that must be served by the etwork ca be computed through the traffic characterizatio of sectio 2.1 ad the target populatio who is expected to use services provided by the etwork: For Voice Traffic: o Symmetrical badwidth o Mobile traffic per user i rural areas is 10 mili-erlag i busy hour,accordig to the model of Telefoica del Peru (sectio ). o Average percetage of populatio with oe cell phoe is 53% o Itieracy (% of people i trasit or short periods of time stay) is assumed as 20% For Data Services: o Asymmetric per-user offered data traffic: 15 Kbps (dowlik), 5 Kbps (uplik) TUCAN3G_D41UPCae.doc 24

26 o Average of populatio with oe cell phoe: 53% o Itieracy (% of people i trasit or short periods of stay): 20% Potetial market is 5% of the total populatio plus people i itieracy with cell phoe. Table 5 summarizes the average aggregated voice ad data traffic (accordig to equatio (2)) offered. The etwork must be desiged so that this traffic is accommodated with a blockig probability of 2%. Note that the backhaul requiremets that will be computed i sectio 8 are much larger, so that they are able to accommodate istataeous traffic peaks. Locatio ihabitats ihabitats + people i itieracy Average umber of subscribers Expected umber of data users Aggregated voice traffic Aggregated offered data traffic [kbps] Dowlik Uplik Sata Clotilde Er Negro Urco Er Tuta Pisco Er Sa Gabriel Er Sa Jua Er Table 5. Total umber of ihabitats, target populatio that is expected to use wireless services ad aggregated geerated traffic expected i each locatio i year 1 25

27 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig 3 SYSTEM ASSUMPTIONS 3.1 Adopted chael models To model the chael i the cosidered rural localities give i [TUCAN3G D21], we must adopt specific ad realistic chael models, accordig to the specific characterizatio of the locatio where the etworks will be deployed Path Loss Model ITS Irregular Terrai Model [NTIA ], which also kow as Logley-Rice model, is cosidered as the path los model i this documet. It is a semi-empirical model which applies for frequecies betwee 20 MHz ad 20 GHz, which is based o electromagetic theory ad statistical aalyses of both terrai features ad radio measuremets. For a particular area, the ITS model takes ito accout differet radio propagatio coditios to compute the path loss betwee two termials. These coditios are modeled statistically, where the fial path loss is a value with a give level of cofidece ad reliability. Path loss is computed as the sum of fixed losses, which are defied through the free space equatio, ad additioal losses, which vary accordig to the time ad the positio of the uits. The fixed losses are a fuctio of the frequecy ad the distace, while the varyig losses are modelled statistically by meas of empirical models which take ito accout differet physical pheomea: reflectio o the groud, refractio through the atmosphere, diffractio over the obstacles ad tropospheric scatter. These physical pheomea will be more or less importat depedig o the radio wave characteristics (frequecy ad polarizatio), the electrical characteristics of the groud (relative permittivity ad coductivity), the eviromet physical properties (temperature, humidity, desity of atmosphere, etc.) ad the terrai irregularities or obstacles i the propagatio path. To compute the path losses i a particular area, we employ the Radio Mobile software [RAD]. This software works with elevatio maps, obtaied for example from the Space Shuttle Radar Terrai Mappig Missio [SRTM]. These elevatio maps have differet resolutios depedig o the locatio ad the resolutio of the area of iterest i TUCAN3G is about 3 arc secod (90 meters). Radio Mobile allows to calculate coverage maps by givig this iformatio i terms of field stregth for each pixel i the map. It does ot provide the path loss of a particular area explicitly, but resultig coverage maps ca be iterpreted as path loss. To cofigure the propagatio coditios i Radio Mobile, the iputs are: ase statio locatio o the map. The atea heights above the groud. The radiatio patter of ateas. Polarizatio of both ateas. Relative permittivity of the groud Groud coductivity. Surface refractivity. Climate. Variability ad statistical parameters such as cofidece ad reliability. The base statio locatio (or locatios), the atea heights, the radiatio patter ad the polarizatio of both ateas are parameters that shall be determied by the plaig strategies ad the particular solutio adopted for the particular etwork. The rest of iput parameters will deped o geographical area where the etwork will be deployed. The atmosphere refractivity determies the amout of bedig of the radio waves. It ca be TUCAN3G_D41UPCae.doc 26

28 characterized through the surface refractivity, typically expressed i terms of N-uits. This parameter depeds o the atmospheric coditios (temperature, atmospheric pressure ad water vapour pressure). I [ITU-R P ], differet values of the surface refractivity are proposed accordig to the geographical area. The electrical characteristics of the groud may be expressed by two parameters, the permittivity ad the coductivity, ad they have direct impact o the reflectio of the radio wave i the groud. Permittivity ad coductivity values ca be foud i [ITU-R P.527-3] ad [ITU-R P.832-3], respectively, accordig to the type of groud. Climate, i cojuctio with the surface refractivity, characterizes statistically the atmosphere i a particular locatio. The variability mode determies the statistical distributios of the parameters, depedig o the topology of the etwork ad fially, the statistical parameters will defie the level of reliability ad cofidece of the model which is expressed as quatiles. Thus, path loss is obtaied from Radio Mobile software by traslatig the coverage maps expressed i terms of field stregth ito path loss. It is worth otig that this strategy implies that path loss depeds o the adopted atea patters. Table 6 summarizes the cosidered iput parameters to compute path losses i TUCAN3G. Iput parameter Frequecy Uplik: MHz / Dowlik: MHz Polarizatio Vertical Mode of variability Mobile (90% of time ad situatios) Surface refractivity 360 N-uits Groud coductivity 0,02 S/m Relative groud permittivity 25 Climate Tropical Atea height Depedig o the etwork cofiguratio ase statio locatio Depedig o the etwork cofiguratio Table 6. Iput parameters i Radio Mobile software to compute path loss Small scale fadig The small scale fadig is produced by the reflectio of the radio wave i differet objects, such as buildigs. Each reflectio takes a differet path ad fially the sum of these reflectios, where each oe has differet phase ad amplitudes, is received by the receiver. The small scale fadig is modelled by meas of a tapped-delay lie profile, a model based o the COST 259 {Rural Area} [ETSI TR ], whose tapped-delay lie profile is depicted i Table 7. The power of each path is timevaryig ad therefore, the fadig ca be well described by Rayleigh distributed amplitudes varyig accordig to a classical Doppler spectrum. Tap Relative time (s) Average relative power (d) (*) Delay spread: 140 s (*) Determiistic direct path (o-fadig) with Doppler frequecy equal to 0,7 times the maximum Doppler spread Table 7. Cost 259 {Rural Area} tapped-delay lie power profile 27

29 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Shadow fadig Shadow fadig is the atteuatio produced by differet obstacles i the radio path. The adopted shadowig model is a logormal radom variable s (Gaussia if it is expressed i d) with zero mea ad variace var{s}: 2 1 s fs (s) exp (7) 2 Var{ s} 2 var{ s} Logley-Rice path loss icludes differet atteuatios due to objects that are i the radio path ad therefore, adopted shadowig model oly characterizes the atteuatios produced by obstacles that are ot take ito accout i the path loss model (buildigs, trees, etc). The impact of shadow fadig is characterized by meas of a shadow margi SM(d), which depeds o the level of cofidece. I practice, a certai value of SM(d) requires a icrease of trasmitted power so that the received sigal stregth stays above the certai value withi a give level of cofidece. For a X% level of cofidece the shadow margi is give by 2 1 SM exp X % (8) s 2 var{ s} 2 var{ s} For radio etwork plaig, we assume two differet levels of shadow fadig: for idoor ad outdoor locatios. Table 8 summarizes variaces ad shadow margi assumptios for each type of locatios with 99% cofidece level. User locatio Variace (d) Shadow margi Idoor d Outdoor d Table 8. Shadow margis assumed for idoor ad outdoor users 3.2 Air iterface assumptios The objective of this sectio is to summarize the physical layer relevat aspects for Widebad Code Divisio Multiple Access (WCDMA) etwork dimesioig. Most of air iterface assumptios deped o the equipmet provided by ip.access, which is FDD-based (Frequecy Divisio Duplexig) ad supports Rel.99, High-Speed Dowlik Packet Access (HSDPA) ad High-Speed Uplik Packet Access (HSUPA). As metioed i [TUCAN3G D21], the mai air iterface assumptios of TUCAN3G are: Duplexig techique: FDD Chip Rate: 3.84 Mcps Chaelizatio badwidth: 5 MHz Frame legth: o Rel-99: 10 ms o HSDPA ad HSUPA: 10 ms or 2 ms. Slots per frame: 15 Modulatio scheme: o Rel-99: QPSK (DL) ad PSK (UL) o HSDPA: 16-QAM o HSUPA: 16-QAM Codes Direct Sequece Spread Spectrum is used i WCDMA system i order to separate trasmissios i uplik ad dowlik ad also to achieve multi-rate trasmissio. WCDMA uses two differet types of TUCAN3G_D41UPCae.doc 28

30 codes: chaelizatio codes ad scramblig codes. The first are used to separate differet trasmissios from a sigle source ad are resposible for chagig the badwidth of the sigal. The secod are used to separate differet termials or base statios ad the badwidth is kept after the scramblig operatio. Figure 15 shows the spreadig ad the scramblig operatio. it Rate Chip Rate Chip Rate Chaelizatio code Scramblig code Figure 15. Scramblig ad spreadig i WCDMA A further descriptio of these WCDMA codes ca be foud i [TUCAN3G D21]. Table 9 summarizes the mai properties ad utilities of the scramblig ad chaelizatio codes. Chaelizatio Codes Scramblig Codes Usage Uplik: separatio of physical data (DPDCHs ad DPCCHs). Uplik: separatio of termials. Dowlik: separatio of cells. Dowlik: separatio of coectios to differet users withi oe cell. Legth chips (dowlik also 512 chips) chips (*) (*) 256 chips if advaced base statio receivers are used (for example, multiuser detectio) Maximum umber of codes uder oe Uplik: Several millio codes Code family scramblig code: spreadig factor (SF) Orthogoal Variable Spreadig Factor (OVSF) Dowlik: 512 Gold code (for codes of legth of chips) Exteded S(2) (for codes of legth of 256 chips) Spreadig Yes, icreases trasmissio badwidth No, does ot affect trasmissio badwidth Table 9. Fucioality of the scramblig ad chaelizatio codes [Holma00] Physical chaels I [TUCAN3G D21], the frame structure ad differet chaels (logical, trasport ad physical) i a WCDMA-FDD etwork are preseted. This sectio provides the mai assumptios of chael structure i radio etwork plaig. The adopted structure for dedicated physical chaels is the structure provided by 3GPP specificatios i the aexes of [3GPP TS ] ad [3GPP TS ] for each type of service. Table 10 summarizes the spreadig factor assumptios for each type of service of Rel.99, for both uplik ad dowlik. The mai characteristics of physical commo chaels ca be obtaied from [Holma00] ad [Laiho01]. User bit rate (service multiplexig) Uplik SF Dowlik SF Voice service 12.2 kbps + DCCH 3.4 kbps Packet data 64 kbps + DCCH 3.4 kbps Packet data 128 kbps + DCCH 3.4 kbps 8 16 Packet data 284 kbps + DCCH 3.4 kbps 4 8 Table 10. Dowlik ad uplik DPDCH spreadig factors [Laiho01] 29

31 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Power cotrol procedure I a WCDMA etwork, the trasmitted powers by the UEs have a direct impact o the system capacity, sice they icrease the amout of iterferece i a cell. I the absece of a mechaism that cotrols these trasmitted powers, it could happe, for example, that overpowered mobiles close to the HN block the whole cell. This effect is kow as the ear-far problem: i the uplik, received powers from users located close to HN are strog, whereas the received powers from remote users are too weak ad caot be detected. Therefore, oe of the most importat mechaisms i a WCDMA system is the power cotrol, which aims to mitigate the ear-far problem i the uplik. This techique is also used i the dowlik, but the motivatio is differet: if the amout of trasmitted power eeded to serve each user is adjusted accordig to the user distace to the HN, the total trasmitted power by the HN will be lower compared to the case where all powers are equal for each user. Therefore, more users could be accommodated ad the iterferece geerated to other cells will be lowered. Two mai mechaisms are ivolved i the power cotrol, which are implemeted ad maaged by the physical layer: the fast power cotrol ad the outer-loop power cotrol. Fast power cotrol sets the trasmitted powers to keep a target Sigal to Iterferece plus Noise Ratio (SINR) durig the coectio, while the outer-loop power cotrol adjust this target SINR accordig to the eeds of the idividual radio lik ad aims at a costat QoS, defied as a certai target lock Error Rate (LER). I additio, fast power cotrol provides aother mechaism to combat fast fadig, maily whe the mobiles velocity is low ad iterleavig does ot provide eough diversity. It is worth otig that capacity ad coverage i a WCDMA system are closely related: whe the cell load is higher (high system iterferece), higher trasmitted power is eeded to meet the quality requiremets. Thus, users at the cell edge may have ot eough power to establish a particular service coectio. This effect is kow as cell breathig. A further descriptio of power cotrol procedure i WCDMA etwork ca be foud i [Holma01] ad [Laiho01] Sychroizatio ad cell search procedure Sychroizatio ad cell search procedure are briefly described i [TUCAN3G D21]. For radio etwork plaig, these mechaisms are importat from coverage ad cell rage poits of view, sice oe UE ca coect to a particular HN (or cell) through these mechaisms. A further study of cell search ad sychroizatio ca be foud i [Wag00], where it is proved that there exists a trade-off betwee capacity ad acquisitio time: it is desirable that the CPICH power be as lower as possible i order to miimize the itra-cell iterferece, while a lower CPICH power could mea a larger acquisitio time. [Wag00] proves that a good choice for a CPICH power to have a acceptable acquisitio time is allocatig at least 5% of the maximum available power i the HN. The procedure followed i allocatig power for CPICH is explaied i more detail at sectio Its goal is to miimize the allocated power for pilot chael, while the whole area of iterest is covered ad a acceptable acquisitio time is obtaied. 3.3 ip.access ode specificatios This sectio describes those aspects of Ip.access HN which are directly related to etwork plaig. Two classes of Node are cosidered, E-class ad S-class, ad both classes have the followig properties: Services supported: o 12,2 kbps Adaptive Multi-Rate (AMR) voice service TUCAN3G_D41UPCae.doc 30

32 o Data services of Rel /128/384 kbps. o Support of HSDPA to 14,4 Mbps (21 Mbps upgradable) o Support of HSUPA to 1,45 Mbps (5 Mbps upgradable) UTRAN mobility: o Reselectio ad hadover to/from macro layer (two-tier deploymets) o Reselectio ad hadover betwee HNs o Itra-frequecy ad iter-frequecy reselectio ad hadover Noise figure of 8 d Parameters that deped o the specific product class are the maximal theoretical rage due to the umber of chips i the firmware for the PRACH searcher, maximum umber of simultaeous dedicated users, RF output power, supported UMTS bads ad possibility of usig exteral ateas. I the followig sectios, specific parameters depedig o the class of ip.access HN are depicted. Despite Ip.access specificatios, some possible chages i their products are cosidered here so that they ca provide services to the areas uder study, amely: 1. It is assumed that HN S-class 16 supports the use of exteral ateas. 2. A HN E-class 24 with RF output power of 24 dm at UMTS bad 2/5 is cosidered (it will be referred i the sequel as E-class 24*) E-class HN The E-class Node specificatios are show i Table 11. E-class 16 E-class 24 Simultaeous dedicated users RF Output Power 24 / 13 dm (*) 24 / 13 dm (*) UMTS bads 1, 2/5, 4 1, 2/5, 4 Exteral ateas Optioal Optioal Maximal theoretical rage 2.4 km 2.4 km (*) For UMTS bad 2/5 used i the TUCAN3G etworks, the RF output power is 13 dm S-class HN Table 11. E-class HN specificatios The S-class Node specificatios are show i Table 12. All assumed support for exteral atea. S-class 8 S-class 16 Simultaeous dedicated users 8 16 RF Output Power 13 dm 20 dm UMTS bads 1, 2/5, 4 1, 2/5 Exteral ateas No No Maximal theoretical rage 1 km 2.4 km Table 12. S-class HN specificatios 3.4 Techical scearios Oe of the mai advatages of femto base deploymets is that they allow updatig the access etwork topology ad desity i a easy yet uplaed way. However, a uplaed deploymet may geerate a harsh iterferece sceario which has to be maaged i order to provide large spectral efficiecy, low coverage outage, ad reduced eergy cosumptio. Due to the low traffic expected i the rural regios i Amazo (service will maily be limited by coverage), ad the large distaces betwee villages, we will assume that at most two HN will be used for oe locatio, usig the two carrier frequecies Telefoica del Peru is allowed to operate. I this situatio, oise-limited scearios without 31

33 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig itercell iterferece ca be cosidered. A oise-limited sceario is characterized by the fact that the limitig factor i the quality of a give coectio is the oise ad itracell iterferece, i.e. the iterferece produced by other cells is egligible. This allows simplifyig the etwork plaig procedures. Accordig to the features that characterize the locatios, the most suitable deploymets are show i Table 13 (a subset of those defied i [TUCAN3G D21]): Populatio features Cocetrated coverage area Disperse coverage area Multi-HN deploymet Traffic Deploymet aspects Sigle HeN with omi/directioal ateas May HeNs i high positio with omi or directioal ateas Low traffic Low-tomedium traffic - Towers up to 70 m high - Towers up to 70 m high - Load balacig might be eeded - Frequecy plaig is eeded, if more tha two HN are deployed Table 13. Deploymets cosidered for Napo ad Paraapura river sites Accordig to the locatios proposed i [TUCAN3G D21], techical scearios selected for these sites are summarized i Table 14, alog with a descriptio of why these are selected ad i which situatios they are expected to be appropriate The right-most colum shows the tools based o Markov chai traffic modelig used to evaluate the blockig probabilities which are described i sectio 4.4 [Kleirock 75]. These scearios are selected o the basis of the previous etwork facilities (towers) i order to reduce istallatio costs ad therefore, deploymets with HN/s located i the existig towers are assumed i all scearios. Each of these scearios has bee tested with the three types of HNs described i sectio 3.3: 1. S-class 16 (16 chaels, 13 dm output power). 2. E-class 24 (24 chaels, 13 dm output power). 3. E-class 24* (24 chaels, 24 dm output power). The HN S-Class 8 has ot bee tested sice the maximum theoretical rage is ot eough to provide coverage. Results i sectio 5 will elucidate the most suitable techical scearios for each site, based o the target performace idicators show i sectio 4.1. TUCAN3G_D41UPCae.doc 32

34 Sceario #1 #2 #3 #4 #5 #6 Number of HN Sigle HN located i high tower Sigle HN located i high tower Sigle HN located i high tower Sigle HN located i high tower Two HNs located i high tower Two HNs located i high tower Atea patter beamwidths ad gai Dipole Gai: 0d Azimuth: omi Elevatio: 45º (wide) Gai: 2 d Azimuth: omi Elevatio: 20º (arrow) Gai: 7 d dowtilt optimised Azimuth: 180º Elevatio: 20º Gai: 13 d dowtilt optimised Azimuth: 180º Elevatio: 20º dowtilt optimised (for each HN) Azimuth: omi Elevatio: 20º (arrow) Gai: 7 d dowtilt optimised Descriptio Suitable for areas where the available power is ot a critical poit, the amout of itracell iterferece is low due to small traffic ad area to cover is small. Also, it is appropriated whe user distributio aroud HN requires omidirectioal coverage. Same as previous oe, with larger coverage capability. Same as previous oe, with eve larger coverage capability. Atea orietatio depeds o the traffic distributio i the coverage area: suitable whe users are distributed withi a sector of 180º. Also, appropriated whe available power is critical. Atea orietatio depeds o the traffic distributio i the coverage area. Differet frequecy bads are used for each HN. Suitable i areas with high traffic, oe HN does ot provide eough ad the available power is critical. Omidirectioal ateas. Differet frequecy bads are used for each HN. Suitable i areas with high traffic, oe HN does ot provide eough ad the available power is ot critical. Aalysis tool for voice blockig probability M/M/m/m M/M/m/m M/M/m/m M/M/m/m 2 parallel M/M/m/m Twodimesioal M/M/m/m Aalysis tool for voice plus data blockig probability Twodimesioal M/M/m/m Twodimesioal M/M/m/m Twodimesioal M/M/m/m Twodimesioal M/M/m/m 2 parallel twodimesioal M/M/m/m Fourdimesioal M/M/m/m Table 14. Tested techical scearios for proposed locatios ad aalysis tool used for blockig probabilities (i sectio 4.4) 33

35 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig 4 METHODOLOGY FOR ACCESS NETWORK PLANNING 4.1 Key performace idicators I this framework we defie the coverage, blockig ad cogestio probabilities as follows: Coverage probability: probability of a icomig call ca be served, give a certai umber of simultaeous users i the system. Cogestio probability: probability that all the of the HN are occupied, i.e., the HN could ot accommodate a ew call eve if it is geerated i a geographical positio where there is radio coverage. lockig probability: probability that a ew icomig call caot be served, i.e., percetage of geerated calls that caot be served. It ecompasses coverage ad cogestio probabilities. The first probability eeds to be evaluated through simulatios, as described i sectio The mathematical descriptio of the latter two probabilities ca be foud i sectio 4.4. Network performace is evaluated i terms of CPICH coverage ad blockig probabilities assumig the traffic desity i the busy hour. The requiremets imposed to determie whether oe sceario is valid or ot are the followig: 1. Network must be capable to cover, i the worst case, the 95% of area where the itesity of traffic geeratio over the geographical area is 10% higher tha the maximum value. 2. Network must be capable to serve the traffic assumed i the busy hour with a blockig probability lower tha 2%. The fial decisio to adopt the specific sceario will deped also o the aalysis described i sectio 7 i order to miimize the eergy cosumptio. 4.2 Radio etwork plaig flow chart This sectio presets the steps that have bee followed for radio etwork plaig. Radio etwork plaig starts off with the defiitio ad characterizatio of parameters ivolved i a WCDMA etwork, such as path losses, chael models ad specific WCDMA etwork parameters. The, ext step is to ru the etwork simulator tool to obtai the idicators that will be used to determie the performace of a particular etwork cofiguratio i terms of coverage, blockig probability ad eergy cosumptio. Figure 16 shows the flow chart of the etwork plaig procedure followed i TUCAN3G to evaluate differet etwork cofiguratios i each locatio proposed i D21. The radio etwork plaig ca be summarized i the followig poits: 1. Defiitio ad selectio of differet possible techical scearios for each locatio (sectio 3.4). 2. Defiitio ad characterizatio of parameters ivolved i radio etwork plaig (sectios ad 4.3.2) 3. Path loss acquisitio ad optimizatio depedig o differet HN atea cofiguratios, i.e., differet heights above the groud ad dowtilts (sectio 4.3.3). 4. CPICH coverage (sectio 4.3.4). 5. Capacity aalysis ad trasmitted powers computatio (sectio 4.3.2). 6. lockig ad cogestio aalysis (sectio 4.4). 7. Eergy cosumptio aalysis (sectio 7). Steps 1-3 are ecessary to characterize the physical layer etwork properties ad parameters o which the results will rely, while steps 4-7 are the procedures to obtai differet results ad etwork performace. TUCAN3G_D41UPCae.doc 34

36 Figure 16. Radio etwork plaig flow chart 4.3 Radio lik budget The goal of this sectio is to describe the mai parameters of WCDMA ivolved i radio etwork plaig, as well as to determie the trasmitted power expressios i the uplik ad the dowlik depedig o the etwork load (umber of simultaeous users i the same cell). To evaluate the trasmitted powers, a WCDMA etwork with power cotrol i both radio lik directios is assumed. Oce the trasmitted power expressios are obtaied, the followig sectios focus o the last steps of radio etwork plaig procedure described i sectio 4.2, givig details how to aalyze the etwork performace through the trasmitted power expressios obtaied i the radio lik budget Radio lik budget WCDMA parameters This sectio focuses o the descriptio of the mai parameters ivolved i the radio lik budget i a WCDMA system. These parameters are eeded to determie the trasmitted power expressios i WCDMA etwork. I aex 9.1, it ca be see the differet values of these parameters after their characterizatio ad used i radio etwork plaig Chip Rate (W) Chip rate is the rate at which waveforms of WCDMA are geerated. This chip rate is obtaied after spreadig operatio lock Error Rate (LER) LER refers to the average block error rate calculated for Trasport locks. The system ca compute this parameter with very high reliability through the Cyclic Redudacy Check (CRC). This parameter is especially importat for the power cotrol procedure sice the outer-loop power cotrol aims to adjust the target SINR accordig to a particular QoS for a particular service, which is defied i terms of LER it rate (R) it Rate is the rate of iformatio bits, excludig overhead from higher layers, such as CRC bits, 35

37 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig codig ad DPCCH cotrol bits. The bit rate will deped o the type of service Activity factor (v) The activity factor ca take values betwee 0 ad 1 ad characterizes the impact of service rate variatio durig the coectio, resultig i a average bit rate of vr. It is importat i radio etwork plaig especially for voice service, where users seldom trasmit at the same rate durig the coectio Orthogoality factor (α) The impact of loss of orthogoality amog chaelizatio codes due to the multipath propagatio coditios is characterized by meas of the orthogoality factor. It ca take values betwee 0 ad 1 ( = 0 implies complete loss of orthogoality, while = 1 correspods to complete orthogoality). This parameter oly makes sese i the dowlik, where the trasmissios betwee HN ad users at the same cell are sychroous. As metioed previously, the orthogoality factor depeds o the multipath propagatio coditios ad therefore, it will be characterized i radio etwork plaig by the cosidered multipath chael model. I [Awoiyi03], a characterizatio of this parameter accordig to the chael tapped-delay lie profile is proposed Spreadig factor ad processig gai The spreadig factor i WCDMA system is defied as the chip rate divided by the user rate. I this case, the user rate icludes overhead from higher layers. The characteristic processig gai i direct sequece spread spectrum systems are typically defied as the spreadig factor, but i most of WCDMA radio etwork plaig literature the processig gai is defied as W/R, where R does ot iclude the overhead from higher layers ad chael codig Eergy per bit over iterferece plus oise spectral desity E b /N o The eergy per user bit divided by the iterferece plus oise spectral desity is a performace idicator used i radio etwork plaig ad it is related to a certai LER. I a WCDMA system, this parameter has differet expressios depedig o the directio of the lik. I the uplik, it is defied for coectio betwee user j ad HN i as r E b W ji r r 2 N0 vr ji j j Pi Pji i i P (9) r 2 where P ji is the received power at the S i from user j, i is the thermal oise power at the HN i, r P i is the total received power at the HN i from users located at the cell i ad i is the received iterferece received at HN i from users coected to other cells. I the dowlik, the defiitio of E b /N o is slightly differet because the sychroized orthogoal codes reduce the iterferece from the same cell. Thus, it is defied i the dowlik for the coectio betwee user j ad HN i as E P N r b W ij 0 vr ij j j S ) i ij j j j r r 2 P P (1 r r where P ij is the received power at user j from S i iteded to serve this user, P S is the total i received power from S i at this user, 2 j is the thermal oise power at user j ad j is the iterferece received at user j from other base statios. (10) TUCAN3G_D41UPCae.doc 36

38 These requiremets will deped o the particular sceario (chael model ad receiver structure) ad particular etwork cofiguratio (the service data rate, how the user data is mapped ito the WCDMA physical chaels, the chael codig, etc.). The E b /N o parameter, both for the uplik ad for the dowlik, is oe of the most importat i the radio etwork plaig. As metioed above, power cotrol procedure aims to keep a target E b /N o to fulfil a give target LER ad therefore, the amout of power allocated to each coectio i the system will strogly deped o this target lik quality idicator. I sectio 4.3.2, developmet of equatios (9) ad (10) is doe to obtai the ecessary trasmitted powers both for the uplik ad for the dowlik depedig o the cell load (itra-cell iterferece) ad the level of iter-cell iterferece (which will be ull accordig to the scearios defied i sectio 3.4) Eergy per chip over total received power spectral desity E c /I o The eergy per chip divided by the total received power spectral desity is a performace idicator used typically for CPICH i dowlik radio etwork plaig. Thus, i this case the orthogoality effects are ot take ito accout. The expressio of E c /I o for the CPICH is cpich cpich( r) c Pi r 2 r 0 ij S i j j E (11) I P cpich(r) where P i is the received power from CPICH at the user j. This parameter coicides with the SINR because the total received power spectral desity is ormalized by the chip rate. A UE ca be detected oe cell if the E c /I o for CPICH is higher tha a give target value. Thus, this quality idicator parameter is relevat to determie the area covered by each cell Power Rise As metioed previously, fast power cotrol i WCDMA adjusts the trasmitted powers i the uplik ad dowlik to keep a fixed SINR with the objective of reducig etwork iterferece. Also, this procedure is able to compesate fast fadig i the coectio, which ca be fairly well compesated i the case of low UE speeds. Ideal fast fadig compesatio implies that termials bear a AWGN chael (at least at a good degree of approximatio) ad therefore, the average E b /N o eeded for the LER is low. Cosequetly, the amout of power dedicated for oe radio lik is ot costat durig the coectio, resultig i a average trasmitted power rise. I the uplik, this effect oly icreases the iterferece received from surroudig cells because the received itra-cell powers at the S remai the same, these beig the powers which satisfy the particular target E b /N o. However, i the dowlik, the power rise implies a icrease i the amout of iterferece that users see from the same cell ad from eighbourig cells. I [Sipila99], the impact of fast power cotrol i the uplik is described ad the power rise parameter is defied ad characterized depedig o the multipath coditios ad UE speeds. I sectio 4.3.2, a mathematical defiitio of the power cotrol headroom is also preseted, both for the uplik ad for the dowlik Power Cotrol Headroom Whe a low-speed UE is approachig the cell edge, the UE trasmissio power ca reach its maximum value due to fast power cotrol procedure to compesate fast fadig. Thus, a fast fadig margi or power cotrol headroom is eeded i radio etwork plaig i order to esure that these UEs i the cell edge ca be served by the etwork. I the dowlik, this fast fadig margi is eeded i order to esure that the available power i the base statio is eough to compesate fast fadig of all coectios. This parameter is explaied i further detail i [Sipila99] for the uplik. I sectio 4.3.2, a mathematical defiitio of the power rise is also preseted, both for the uplik ad for the dowlik. 37

39 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig WCDMA power allocatio I this sectio, we fid the expressios that describe the trasmitted powers i WCDMA etworks whe power cotrol is assumed, both for the uplik ad for the dowlik. These expressios ca be obtaied through the lik quality requiremets defied i (9) ad (10), solvig these equatios for the trasmitted powers. I scearios cosidered i sectio 3.4 there is o iter-cell iterferece sice almost all are isolatedcell deploymets ad the two-cell deploymet uses differet carriers for each cell. However, i this sectio the geeral case for power allocatio i WCDMA etworks is described, cosiderig more tha oe cell ad iterferece betwee them (deoted i the sequel as ), sice it will be useful for activities 4A2 ad 4A Power allocatio i the uplik I the uplik, the received sigal from user j i the HN i will be X r bs ms ji T Pji PjiGi ( j ) Gj ( i ) PjiGji X ji, (12) L ji ms j bs where P ji is the trasmitted power by the UE j, G ( i ) ad Gi ( j) are the UE ad HN atea gais depedig o their radiatio patter, respectively, L ji deotes the path loss betwee this two uits i the uplik ad X ji deotes the istataeous multipath chael gai. We have defied the parameter T bs ms G G ( ) G ( ) / L, which deotes the effective gai betwee the user j ad HN i. ji i j j i ji From the poit of view of the receptio of user j s sigal, the amout of iterferece power received at the HN i will be the sum of all user received powers coected to the same cell, except that of the 2 user j, plus the thermal oise power at the HN, i, ad the iter-cell iterferece received from all users coected to the other cells. Let M be the total umber of users coected to the cell ad N the total umber of HN i the etwork. The, the SINR for the user j coected to the HN i ca be writte as PG X SINR. T ji ji ji ji Mi N M T T 2 T( ) ( ) PG mi mi Xmi PGX ji ji ji i PG k ki Xki m1 1 k1 i (13) ( ) where the ( ) X ki ad G T ki deote the istataeous gai due to fast fadig ad the effective gai betwee a user k coected to the cell ad the HN i, respectively. The superscript () remarks that, despite user k causes iterferece i cell i, it is coected to HN. As see previously i sectio , lik quality requiremets are typically give i terms of eergy per user bit divided by the iterferece plus oise spectral desity, E b /N o ad differet target E b /N o to keep a particular QoS ca be obtaied by radio lik level simulatios, measuremets or from techical specificatios. Typically, the miimum requiremets to keep a particular LER are give for differet services rates, cofiguratios ad scearios, ad they are defied i the uplik [3GPP TS ] as E L b chip W SINR SINR, (14) N Lif vr 0 target where W is the system chip rate, R is the bit rate for a particular service, L chip is the total umber of chips per frame, L if is the umber of iformatio bits per frame ad v is the activity factor. From equatios (13) ad (14), we ca express the user j coverage coditio as TUCAN3G_D41UPCae.doc 38

40 E N PG X T b W ji ji ji Mi N M 0 vjr ji j T T 2 T( ) ( ) PG mi mi Xmi PGX ji ji ji i PG k ki Xki m1 1 k1 i. ji (15) ms Let P max deotes the maximum available power i the UEs. The, the WCDMA power cotrol i the uplik solves the followig miimizatio problem: miimize P i1,, N jcelli Pij ji PG X subject to ji, i1,, N j cell i vr T W ji ji ji j j Mi N M T T 2 T( ) ( ) PG mi mi Xmi PG ji ji Xji i PG k ki Xki m1 1 k1 i ms P P, i1,, N jcell i ji max (16) Differet users will be coected to oe HN or aother depedig o the dowlik pilot coverage. To fid out how the power cotrol performs ad its impact o the trasmitted powers P ji, it is assumed a fixed iter-cell iterferece i the HN i deoted by, N M T( ) ( ) i PG k ki Xki 1 k1 i i, i1,..., N. (17) Now, the miimizatio problem described i (16) ca be divided ito N less complex miimizatios correspodig to each cell. I the optimum, if ideal power cotrol is assumed, (16) must be fulfilled with equality ad therefore we ca solve the lik quality restrictio i (16) for each P ji i the cell i ad gives M i 2 T i i PG mi mi X mi m1 1 1 i i Pi ji T T G W W ji X ji Gji X ji P, i 1,..., N jcelli, 1 1 vr vr j j ji j j ji (18) where P i deotes the total received power i the HN i ad it will be Solvig (19) for P i gives Mi Mi 2 T i i Pi i PG mi mi mi W m1 m1 (19) P X, i1,..., N. 1 vr M i 2 i i W m1 1 2 ul vr m m i mi i i Pi, i1,..., N Mi ul 1 1i 1 W m1 1 vr m m mi m m mi (20) where ul i is defied as the uplik itra-cell load factor: 39

41 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig M i ul 1 i, i 1,..., N. (21) W m1 1 vr m m mi The itra-cell load factor defied i (21) characterizes the amout of oise rise over thermal oise plus iter-cell iterferece due to the iterferece from users coected to the same cell (itra-cell iterferece). Fially, pluggig (20) ito (18), the trasmitted power by the user j coected to the HN i ca be rewritte as i i Pji, i1,..., N j cell i T ul G W ji X ji 1 1i vr j j ji (22) The trasmitted power by the users is proportioal to the path loss betwee the user ad the HN ad to the iter-cell iterferece plus thermal oise, ad iversely proportioal to the atea gais, to the istataeous chael gai due to fast fadig, to the factor which characterizes the processig gai for ul a particular service ad to the factor 1 i. Fially, for radio etwork plaig, the statistics of trasmitted powers durig the coectio are ecessary. The average trasmitted power of (22) for each UE i the uplik assumig ideal power cotrol will be i i E(P ji ) P ji E (23) X T ul W ji G ji 1 1i v R j j ji where the term E1/ X ji is the average power rise defied i I additio to power rise, it is also importat to characterize the variatio of this trasmitted power i order to determie the power cotrol headroom, as explaied i Power allocatio i the dowlik I the dowlik, the HN i speds part of its power to serve user j. The rest of the HN trasmitted power iteded to cover the other UEs ad the received powers from other HNs are perceived as iterferece power by user j. Hece, the lik quality equatio for the UE j coected to the HN i will be E b W PG X ij, N0 v ij jrj T 2 1 X T ij ij ij N j T Gij ij PS P i ij j PS G j Xj 1 i (24) where P ij deotes the amout of trasmitted power by the HN i iteded to serve the user j, T bs ( ) ms cch Ni Gij Gi j Gj ( i )/ Lij is the effective gai betwee the HN i ad the user j, PS P i i P j1 ij is the total trasmitted power by HN i (the summatio of the powers iteded to serve each user plus N T the total commo chaels power), 1 PS G jx j is the received iterferece from other HNs at i TUCAN3G_D41UPCae.doc 40

42 user j ad j is the orthogoality factor for the dowlik of user j. The lik quality requiremets, ij, ca also be obtaied from measuremets, simulatios or from 3GPP specificatios [3GPP TS ]. Moreover, i the dowlik plaig, the commo chaels must be cosidered. For dowlik commo chaels, such as Commo Pilot Chael (CPICH), the requiremets are usually give i terms of E c /I o [3GPP TS ], where E c is the eergy per chip ad I o is the total received power spectral desity at the UE, as described before i sectio The lik quality requiremet for the CPICH must be fulfilled i ay user j located i ay positio of the cell i, which is equivalet to say that the requiremet must be satisfied i the worst case. The, E c I0 CPICH CPICH T Pi Gip T 2 P ip S G i ip p p cpich, (25) where p deotes a user located i the worst positio i the cell. The worst positio i the cell will be T 2 this positio where the summatio PSG i ip p pis the highest, where p deotes the iter-cell iterferece i this locatio. Notice i (25) that the istataeous chael gai that characterizes the fast fadig is ot icluded i the expressio ad this is because it has bee icluded i the quality requiremet cpich. Fially, otice that allocated power for CPICH is ot ivolved i the power cotrol procedure ad therefore it does ot compesate fast fadig. The requiremets for the other commo chaels ca be set i terms of relative trasmitted power to the CPICH [Holma00]. Therefore, (25) ca be rewritte for the set of commo chaels as E c I0 cch cch T Pi Gip T 2 P ip S G i ip p p cch, (26) cch where P i is the total trasmitted power by the S i for the set of commo chaels ad cch deotes its quality requiremet expressed i terms of E c /I o. Thus, the miimizatio problem i the dowlik for the cell i is cch Pi,{ Pij} Mi cch S P i i ij j1 miimize P P i 1,, N j cell i W PG X subject to ij, i1,, N; j cell i; vr j j T ij ij ij N T 2 T j PS P i ij GijXij j PS G jxj 1 i 1 cch T i p T 2 Si p p p P G P cch i P G cch, i1,, N; Mi bs Pij Pm, i 1,, N. ax j1 (27) Assumig also a fixed iter-cell iterferece i each user of the cell i deoted by j, N T j PS G, j Xj j cell i (28) 41 1 i the miimizatio problem described i (27) ca be divided ito N less complex miimizatios correspodig to each cell. Additioally, if ideal power cotrol is assumed, we ca solve the lik

43 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig quality restrictio i (27) for each P ij i the cell i ad gives 2 j j 1 jps i T Gij X ij Pij, i 1,..., N; j celli. W 1 j vr j j ij (29) The total trasmitted power by the HN i will be 2 m m Mi Mi Mi T 1 Gim X cch cch ij S i i im i S i W W m1 m11 m11 m vr m mim1m vmrmim P P P P P, i1,..., N, (30) ad solvig (30) for P S i gives P Si Mi 2 cch 1 1 m m i T G m 1 im X W im 1 m vmrm im dl 1i P, i1,..., N, (31) where dl i deotes the dowlik itra-cell load factor of the cell i ad it is defied as M i dl 1 i, i 1,..., N. (32) W m1 1 vr 1 m m im m Similar to the uplik, the statistics of trasmitted powers by HNs are really importat for radio etwork plaig. The average trasmitted power by differet HNs will be 2 1 Mi cch 1 1 m m EPS PS i P, 1,...,, i dl i E i N (33) T 1 W i X m1 Gim 1 m vr m m im where E(1/X) is the power rise for each dowlik coectio (we assume that multipath coditios are the same i each lik). Fially, as i the uplik, a fast fadig margi is eeded i order to avoid that trasmitted power variatios might block the coectio betwee HN ad UE. I the dowlik, the fast fadig margi will deped o the umber of simultaeous users i the etwork sice the same HN is resposible for compesatig fast fadig through power cotrol procedure i each coectio. If we assume a power cotrol headroom for each lik, PCH oe lik, the total power cotrol headroom eeded by the HN i order to compesate power fluctuatios ca be expressed as PCH (d) PCH ( d) 10log M (34) T oe lik where M i is the umber of simultaeous users i the etwork. Notice that this way to compute the total power cotrol headroom is quite pessimistic sice we assume that all radio liks ca suffer fast fadig at the same istat of time. i TUCAN3G_D41UPCae.doc 42

44 4.3.3 Path losses evaluatio Path losses of each locatio are obtaied by Radio Mobile software. They will be differet depedig o the lik directio (uplik ad dowlik) due to the terrai, operatig frequecy ad differet atea patters at UE ad HN. If a sceario of more tha oe cell is assumed, we require path losses for each cell, both for the uplik ad for the dowlik. Figure 17 shows as a example the path losses i d obtaied with Radio Mobile for differet cofiguratios i Negro Urco locatio. (a) Isolated cell, dowlik ad isotropic atea (b) Isolated cell, dowlik ad atea with azimuth beamwidth of 180º, elevatio beamwidth of 20º ad dowtilt of 10º Figure 17. Path losses i d obtaied for differet atea cofiguratios i Negro Urco for a heigh above groud of 70 m It is importat to otice here that atea patter will have a direct impact o the obtaied path loss byowig to the irregular terrai. Therefore, there will be oe atea cofiguratio (height above groud ad dowtilt) that miimizes the path loss i the coverage area. Thus, a study of the impact of atea cofiguratio is ecessary to optimize the path losses i each locatio ad for each tested sceario. The procedure followed to evaluate this optimum atea cofiguratio at every place is: 1. Path loss extractio with radio mobile for each atea cofiguratios (differet heights ad dowtilts). 2. Compute the 95 th percetile of the path losses iside the area to cover. 3. Fid the optimum cofiguratio of HN height ad atea dowltilt that miimizes this 95 th percetile Pilot coverage requiremets Dowlik commo chaels cause iterferece to dedicated chaels ad also reduce the available power for these chaels, resultig i a dowlik capacity decrease. From the poit of view of capacity, it is desirable that power iteded for commo chaels to be as lower as possible subject to the CPICH quality costrait i the whole area to cover. Thus, the CPICH coverage must be fulfilled i the worst case, this is whe HN are trasmittig maximum power (maximum self-iterferece) ad the expected iter-cell iterferece is maximum. Obviously, if oe user is able to idetify oe cell i the worst case, it will be able to idetify the cell if the traffic load is lower, resultig i a higher coverage area. Also, otice that for the techical scearios proposed, iter-cell iterferece does ot exist. For each locatio, as metioed i sectio 2.3, itesity of traffic geeratio over the geographical area is assumed i order to characterize i which zoes UE desity will be higher (areas with may houses) or lower (areas with few houses). I sectio 4.1, we have defied the area to cover i the worst case as the area where the expected itesity of traffic geeratio is 10% higher tha the maximum value. This area is limited by the maximum theoretical rage (2.4 km for HNs used). Fially, for this 43

45 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig area we defie the maximum path loss to cover as the 95 th percetile, excludig i this way isigificat locatios where path loss ca be very high due to differet physical ad radio electrical reasos. Oce we have the maximum path loss to cover, we ca compute the miimum CPICH power (i d) through equatio (25) as 2 max p CPICH bs p P 10 log cpich Pmax SM( d) T (35) G ip where sub-idex p deotes oe user located i the maximum path loss to cover (95 th percetile), max is the maximum iter-cell iterferece expected i this place depedig o eighborig etworks (it is 0 i the selected scearios where o iterferece is expected), ad SM is the shadow margi defied i Multipath coditios are icluded i the lik quality costrait ad other margis are igored sice CPICH is ot cosidered i the power cotrol procedures. Through the CPICH power computed i equatio (35), the total power iteded for dowlik commo chaels ca be determied by meas of the relative trasmitted powers depicted i Table 15. Power level relative to CPICH power CPICH - P-SCH + S-SCH -3 d P-CCPCH -5 d PICH -8 d AICH -8 d S-CCPCH -5 d Table 15. Dowlik commo chael power levels relative to CPICH power [Laiho01] For the particular case of multi-cell sceario #5 depicted i Table 14 (two HNs at the same tower with 180º sector ateas ad 2 differet carriers), the steps to determie each pilot power are practically the same as the isolated cell case described above. However, i this multi-cell etwork is ecessary to defie the area covered for each cell. For radio etwork plaig i sceario #5, we assume a pessimistic situatio, maitaiig a fixed coverage area for each cell, idepedet from the umber of simultaeous users i each cell (cell loads). The coditio followed to determie whether every poit (bi) withi the theoretical rage belog to oe cell or to the other is based o the miimum dowlik path loss. Moreover, soft hadover is ot possible betwee 2 cells with differet operatig frequecies. Also, there will be some places where path losses betwee a UE ad HNs are practically the same (agular limit of each sector) ad i these places we always assume the worst path loss to determie the HN which will serve the user. This decisio is pessimistic as if oe HN is cogested (o codes left) the UE does ot try to coect to the other HN. For the particular case of multi-cell sceario #6 depicted i Table 14 (two HNs at the same tower with omidirectioal ateas ad 2 differet carriers), the steps to determie each pilot are the same i both cells ad obviously, power eeded for commo chaels will be practically the same at each cell sice the oly differece betwee to cells are the operatig frequecy (same power for commo chaels at each cell are cosidered).. Notice that the uplik is ot take ito accout to determie the coverage area. Usually, the cell rage is limited by the uplik but due to the HNs used i this project this is ot the case sice the total power available by the UEs are higher tha the available power at HN to serve a dowlik coectio. I macro-cells, the available power at the base statios is much higher ad therefore, the etwork dimesioig typically is uplik-limited. Fially, it is importat to otice that i situatios where atea gais are high, path losses are low, etc. it is possible that CPICH power is too weak to detect the cell correctly regardless of the lik quality requiremet. As described i sectio 3.2.4, a trade-off exists betwee acquisitio time ad TUCAN3G_D41UPCae.doc 44

46 received CPICH power stregth. [Wag00] defies the miimum ratio betwee trasmitted power for CPICH ad total trasmitted power by HN of 5% to have a acceptable acquisitio time. Thus, if trasmitted power for CPICH by the miimum path loss is lower tha 5%, we should keep it to 5% Lik quality evaluatio Oce the powers of dowlik commo chaels are determied i radio etwork plaig, the ext step is to geerate differet users withi the coverage area ad determie the ecessary trasmitted powers i both uplik ad dowlik. Equatios (23) ad (33) describes the average trasmitted power eeded depedig o the cell load (umber of simultaeous users i the cell) i a WCDMA system with power cotrol. Also, these trasmitted powers deped o the specific positio of the users aroud the cell ad therefore, differet realizatios are ecessary i order to have a realistic distributio of trasmitted powers by UEs ad HN for a particular cell load. As we cited i previous sectios, the trasmitted powers are ot fixed durig the coectio, but they fluctuate due to fast power cotrol mechaism. I order to determie whether differet UEs ca be served by the HN, we must test if the available power i each UE ad the available power i the HN (or HNs) are eough to serve all users for this particular cell load ad for this particular distributio of users durig all coectio. Thus, oe UE ca be served i the uplik directio if 45 P PCH SM P (36) ms ji ul max where P ji is the average trasmitted power by the UE j coected to HN i described i (23), PCH ul is the power cotrol headroom defied for the uplik ad SM is the shadowig margi. I the dowlik, the coditio is slightly differet because the available power i HN (or HNs i the multi-cell cases) has to be capable of servig all UEs. Furthermore, we must take ito accout the allocated power for commo chaels. The coverage coditio i the dowlik ca be expressed through equatio (33) ad it is M i m m E T X G W m1 im 1 cch m Pi vmrm im PCH ( M ) SM P dl dl 1 1 i i bs dl i max where PCH dl (M i ) is the power cotrol headroom i the dowlik, which depeds o the umber of cch simultaeous users ad P i is the trasmitted power for the set of commo chaels determied i sectio Notice that for trasmitted powers of commo chaels, the shadow margi has already take ito accout previously ad power cotrol headroom is ot eeded sice there is o power cotrol for CPICH. Coditios (36) ad (37) are ecessary i order to determie if there exists a o-coverage situatio for a particular cell load ad user distributio. We cosider a o-coverage situatio i the uplik if at least oe UE does ot have eough power to be served by the HN ad a o-coverage situatio i the dowlik if the HN or at least oe HN (multi-cell case) does ot have eough power to serve all users. There are other reasos for o-coverage situatios which will be discussed i detail at sectio Fially, Mote Carlo simulatios are used to evaluate differet situatios (distributios of users aroud the HN) depedig o the umber of simultaeous users. The procedure is as follows: for each umber of simultaeous users, differet realizatios (radom positios of users i the cell) are geerated ad the o-coverage situatios are determied ad couted for each realizatio. A o- (37)

47 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig coverage situatio is idetified whe there is ot eough power to serve simultaeously those users. To geerate differet users, we assume differet itesities of traffic geeratio over the geographical area aroud the HN, resultig more likely that user will be located i some places where there is a high desity of housigs. Durig this simulatio, differet blockigs are couted but also trasmitted powers are computed ad saved for later processig. Average trasmitted powers by HN are saved without takig ito accout fast fadig margi ad shadow margi for dedicated chaels Target E b /N o values As metioed previously, target lik quality requiremets for each service are oe of the most importat parameters i WCDMA etwork plaig. For low speed mobiles, which are i TUCAN3G locatios, fast power cotrol is able to compesate fast fadig fairly well ad therefore, chael betwee HN ad UEs are practically as a AWGN chael. Thus, thaks to the fast power cotrol, the E b /N o are lower. I radio etwork plaig, we assume that each service eeds a QoS determied by a LER of 1%. For this LER, we ca fid the quality requiremets i 3GPP specificatios both for the uplik [3GPP TS ] ad for the dowlik [3GPP TS ] for a AWGN chael. Table 16 summarizes these requiremets defied i 3GPP specificatios. Uplik Dowlik Voice 12,2 kbps 8,3 d 7,4 d Data 64 kbps 4,8 d 3,98 d Data 128 kbps 4,5 d 3,97 d Data 384 kbps 4,1 d 3,5 d Table 16. Target quality requiremets for a AWGN chael The 3GPP specificatios do ot give a target value for 128 kbps data service, but for 144 kbps. However, it ca be assumed that dedicated chaels (cotrol ad data) have the same cofiguratio i both cases ad target E b /N o for 128 kbps data service ca be expressed as E b 144 E b N0 128 N0 128kbps 144kbps I the dowlik, these requiremets are give i terms of the ratio of the average trasmit eergy per chip to the total trasmit spectral desity (E c /I or ). These ca be iterpreted i terms of E b /N o usig the followig expressio [Holma00]: W Ec Eb R Ior N I / I where I oc deotes the power spectral desity ormalized to the chip rate of a bad limited white oise source simulatig iterferece from cells (the value of this parameters is give together with the differet E c /I or values i the stadard requiremets). Notice that if we assume static coditios (AWGN chael), we must take the orthogoality factor is equal to 1 for the trasformatio from E c /I or to E b /N o. or oc (38) Defiitio of the probability of coverage The goal of this sectio is to defie more precisely the probability of coverage i WCDMA etworks. I equatios (36) ad (37) we have defied the o-coverage coditios, for the uplik ad for the dowlik, respectively. Oe user caot be served by the etwork because the available powers at the UE or at the HN are ot eough, as cited i 4.3.5, but it is ot the oly reaso. The HN must dispose of free spreadig codes for the dowlik coectio. The maximum umber of simultaeous TUCAN3G_D41UPCae.doc 46

48 users limited by the umber of spreadig codes, or by the umber of simultaeous users supported by the HN (16 or 24 as metioed i sectio 3.3) i the dowlik is called hard capacity. O the other had, the maximum umber of users that ca be served before beig limited by the available power is called soft capacity. Soft capacity depeds o the type of service ad it also depeds of the specific locatios of the UEs aroud the HN. Notice i (23) ad (33) that the maximum umber of users ca also be limited simply by the load factor, which does ot deped o the UE distributio aroud the HN, but it oly depeds o the type of service for each active user i the cell. This happes sice beig itra-cell load factor close to 1 causes that the trasmitted powers to serve oe coectio is ifiity. The maximum umber of users limited by the load factor of the cell is called pole capacity. Thus, hard capacity ad pole capacity ca be determied easily for each type of service i the etwork. However, the study of soft capacity i WCDMA etworks is ot easy because it depeds o the specific user locatios. This meas that for a particular umber of simultaeous users, there will be situatios or user distributios where the available powers at UEs ad HNs are eough to establish all service coectios, while there will be user distributios where the available powers are ot eough. For this reaso, simulatio tools are eeded to evaluate capacity i 3G etworks ad it make sese to say that oe HN ca be serve M i simultaeous users with a certai coverage probability. As metioed i sectio 4.1, we defie the probability of coverage as the probability of icomig call ca be atteded give a certai umber of users i the system. With the simulatio tool, we caot evaluate directly this probability, but we ca evaluate the probability of M i simultaeous users ca be atteded by the system, both for the uplik ad for the dowlik through o-coverage coditios cob defied i Now, let P i deotes the probability of i simultaeous users ca be atteded by the HN, thus, the probability of coverage ca be defied by meas of ayes rule as P P P (39) cob cob cob ii, 1 i1 i I sectio 4.4, a further descriptio o evaluatig served traffic by the etwork through the coverage probabilities is give i details. I coclusio, we have described differet reasos that limit the umber of active users coected i the etwork: available codes i the dowlik or maximum umber of simultaeous users supported by HN (hard capacity) The cell reaches a cell load of 1 (pole capacity) The available power by HN is ot eough to serve a certai distributio of users. Fially, for the multi-cell sceario #5, everythig described i this sectio is also valid for each cell. Notice i this two-cell deploymet that, despite the umber of simultaeous users supported by the HN ca icrease by two, the distributio of users aroud these HNs is determiat to elucidate if oe cell ca serve all UEs or, i cotrast, this HN caot serve this distributio (for example, if each HN supports 16 users, whe 32 users are active i the cells, oly a distributio with 16 users i each cell ca be served). The procedures followed to evaluate the blockig probabilities ad the associated for each sceario are described i sectio 4.4. I order to illustrate the cocepts, Figure 18 displays the cob probability of i simultaeous users ca be served ( P i ) for differet techical scearios i Sata Clotilde locatio whe HN S-class 16 is used, both for the uplik ad for the dowlik. Notice that, for the uplik, these probabilities are practically oe, idepedetly of the umber of simultaeous users i the system ad thus, the etwork is dowlik limited. This result is expected as the available power at HN is more striget tha the available powers at the UEs for each cofiguratio ad i each locatio. Note that for the dowlik, sceario #3 provides the best performace i terms of 47

49 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig coverage probabilities sice it has the higher available power to serve users due to the higher atea gai. Figure 18. Dowlik (left) ad uplik (right) probabilities of i simultaeous users ca be served depedig o the umber of simultaeous users before ew icomig call for scearios #1, #2 ad #3 i Sata Clotilde locatio whe HN S-class 16 is used For illustrative purposes, Figure 19 shows the dowlik probabilities of i simultaeous users ca be atteded for each cell i Sata Clotilde locatio if sceario #5 is selected. As metioed i sectio 4.4, each cell of sceario #5 ca be see as two idepedet etworks ad therefore, coverage probabilities for each cell are computed separately. Figure 19. Dowlik probabilities of i simultaeous users ca be served depedig o the umber of simultaeous users before ew icomig call for sceario #5 i Sata Clotilde locatio for differet types of HN 4.4 Queuig models, blockig ad cogestio probabilities This sectio has the aim to provide either the aalytic expressios or a method to calculate the blockig ad cogestio probabilities that will be used to check whether a give etwork dimesioig (distributio of HN s with their correspodig powers) fulfills the miimum quality of service. If such miimum quality is ot attaied, the the etwork dimesioig TUCAN3G_D41UPCae.doc 48

50 should be reiforced by addig more HN s, icreasig their capacity, icreasig their power, adjustig the ateas ad radiatio patters, etc. I covetioal wirelie dimesioig, a give server with N will be able to provide service to a ew icomig call wheever ay of the of the server is available. I the WCDMA case, the coditio uder which a call ca be served is more striget ad twofold: first, the mobile termial geeratig the ew icomig call has to be uder radio coverage (i.e., there should be eough power i the HN to serve it); secod, at least oe of the of the server has to be available ad free Voice service I our case, the N of the server refer to the maximum umber of users that the HN ca support simultaeously. This characteristic should be icluded withi the techical specificatios of the equipmet at had. esides, i a multihn deploymet, differet HNs are allocated differet frequecies, which implies that there will be o iter-cell iterferece ad, therefore, the aalysis ca be carried out usig the coverage probabilities obtaied without iterferece, i.e., o a sigle cell basis. The otatio used i this sectio is give: L: umber of subscribers N: umber of total available of the HN, i.e., total umber of users that the HN could serve simultaeously, P i : probability of havig i active, i.e., i calls beig served by the HN simultaeously, cob Pi i 1 : probability of beig able to accommodate a ew icomig call whe i calls are active, i.e., beig served simultaeously, v : rate of icomig calls geerated by a sigle subscriber, = L v : aggregated rate of icomig calls1/: mea duratio of a active call. We will assume that the time betwee ew icomig calls follows also a expoetial distributio. cob cob cob cob Accordig to ayes rule, the probability Pi i 1 ca be expressed as Pi i 1 Pi 1 Pi, where cob P i ( P cob i 1 ) is the probability that give i (i+1) calls, all of them are uder radio coverage ad ca be served simultaeously (i.e., there is eough power) regardless the availability of at the HN. These probabilities are evaluated umerically usig the priciples described i sectio i a Mote Carlo simulatio test. I this framework we defie the blockig ad cogestio probabilities as follows: Cogestio probability: probability that all the of the HN are occupied, i.e., the HN could ot accommodate a ew call eve if it is geerated i a geographical positio where there is radio coverage. lockig probability: probability that a ew icomig call caot be served, i.e., percetage of geerated calls that caot be served. I the followig we will aalyze two situatios i terms of the umber of subscribers for the derivatio of the expressios of the blockig ad cogestio probabilities. I each case, the probabilities are derived usig a queuig aalysis based o Markov-chai model. The models are referred to as M/M/m/m, followig the covetioal amig that ca be foud i [Kleirock 75] M/M/m/m queuig system This evaluatio is suitable for scearios 1 to 4 as described i Table

51 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig subscribers much higher tha the umber of available I this sceario, we will assume that the umber of ihabitats i the cell is much higher tha the umber of of the HN, which meas that the rate of geeratio of ew calls is idepedet of the umber of curret active calls. Let = L v deote the aggregated rate of icomig calls for the whole set of subscribers. The cotiuous-time Markov chai modelig this sceario is the followig: cob cob cob cob cob P0 1 P 12 P 23 P 34 P N 1 N (1 ) cob cob cob cob P0 1 (1 P1 2) (1 P2 3) P3 4 (1 ) Figure 20. Markov chai characterisig voice users served by a HN of N, assumig a large umber of subscribers The i th state represets the situatio i which i calls are beig served simultaeously by the HN, where its probability is deoted by P i. The total offered traffic (OT) measured i Erlags is give by N 1 1 OT Pi (40) i0 whereas, the carried traffic (CT), i Erlags, which oly takes ito accout the calls that ca be served (i.e., there is radio coverage ad available at the HN), ca be expressed as 1 CT PP PP N1 N1 cob cob i i i 1 i i i 1 i0 i0 (41) The probabilities of the differet states i the Markov chai ca be calculated assumig that we are i a statioary regime so that, for each state, the icomig rate equals the outgoig rate: cob cob State 0: P0P0 1 P1 P1 P0 1P o cob cob cob State i: Pi 1Pi1i Pi 1( i1) Pi i PiPii 1 Pi 1 Pii 1Pi ( i 1) cob cob State N: PN 1PN1N PNN PN PN 1NPN1 N Accordig to the previous expressios: i i1 N 1 cob 1 i 0 kk1 i 1 0 N i i k 0 i0 P P P P P! 1 1 i! Fially, the blockig ad cogestio probabilities are give by i1 i1 k 0 P cob kk1 (42) Cogestio probability: TUCAN3G_D41UPCae.doc 50

52 N N 1 1 cob Pk k 1 N! k 0 C( ) N N i i1 1 cob 1 Pkk 1 i1 i! k 0 P N P (43) lockig probability: N 1 i i1 N N1 cob cob 1 cob 1 cob (1 P 1 0 1) (1 1) N Pi i Pk k 1 P k k 1 cob i 1 i! k 0 N! k0 ( ) i(1 i i 1) N N i i1 i0 1 cob 1 Pkk 1 i1 i! k 0 P N P P P subscribers comparable to the umber of available I this sceario, we will assume that the umber of subscribers i the cell L is barely higher tha the umber of of the HN N, while L>N. The mai cosequece from this fact is that the rate of geeratio of ew calls depeds o the umber of curret active calls. Let o = v deote the rate of geeratio of calls from a sigle subscriber. The cotiuous-time Markov chai modelig this sceario is show i Figure 21. cob cob cob cob cob op0 1 ( L1) op1 2 ( L 2) op2 3 ( L 3) op3 4 ( LN 1) opn 1 N L (44) L (1 ) o cob cob ( 1) (1 ) ( L2) o (1 P2 3) ( L3) (1 P3 4) ( L N) o cob cob P0 1 L o P1 2 Figure 21. Markov chai characterisig voice users served by a HN of N, assumig that the umber of subscribers is comparable to N As i the previous sceario, the state i represets the situatio i which i calls are beig served simultaeously by the HN, where its probability is deoted by P i. The total offered traffic (OT) ca be calculated as i0 o N 1 o OT o( L i) P i ( L i ) (45) whereas, the carried traffic (CT), which oly takes ito accout the calls that ca be served (i.e., there is radio coverage ad free at the HN), ca be expressed as 1 CT Li PP Li PP N1 N1 cob o cob o( ) i i i 1 ( ) i i i 1 i0 i0 (46) The probabilities of the differet states i the Markov chai ca be calculated assumig that we are i a statioary regime so that, for each state, the icomig rate equals the outgoig rate: cob o cob State 0: PL 0 op0 1 P1 P1 LP0 1Po 51

53 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig cob cob o L i cob State i: Pi 1( Li1) opi 1i Pi 1( i1) Pi i Pi( Li) opii 1 Pi 1 Pii 1Pi i 1 cob o L N 1 cob State N: PN 1( LN 1) opn 1N PNN PN PN 1NPN1 N Accordig to the previous expressios: i i1 N o L! cob 1 i 0 kk1 i 1 0 N i k 0 i0 o P P P P P i!( L i)! L! 1 i!( L i)! Fially, the blockig ad cogestio probabilities are give by i1 i1 k 0 P cob kk1 (47) Cogestio probability: lockig probability: P ( N) N 1 i0 i0 P ( N) P N o L! N!( L N)! N o L! 1 i!( L i)! C N i cob Pi( Li)(1 Pi i 1) PN( LN) N P( L i) i N 1 k 0 P i1 i1 k 0 cob kk1 P cob kk1 N 1 i i1 N N1 cob cob o L! cob o L! cob 01 ii1 k k 1 k k 1 i 1 i!( L i 1)! k 0 N!( LN 1)! k0 N i i1 o L! cob L Pkk 1 i0 i!( Li1)! k 0 L(1 P ) (1 P ) P P (48) (49) Two parallel M/M/m/m queuig system This is evaluatio is suitable for sceario 5 as described i Table 14, where we ca assume that two HN, say A ad, give coverage to two areas with directioal ateas whose radiatio patters do ot overlap (a approximatio i practice). Each area is geeratig traffic P HA ad P H respectively. Note that P HA ad P H are the fractio of subscribers covered by HN A ad respectively. Parallel idepedet M/M/m/m chais ca be used (as those i Figure 20 ad Figure 21) sice there is o spill of traffic betwee the two coverage areas whe oe of the HN is blocked. eig this the case, we are iterested i both areas havig the same probability of a call beig blocked which implies A, max A ( A ), ( ) P N N P N P N (50) A A where P ( N ) is the blockig probability for HN A whe a maximum of N A are allowed (equivaletly for HN ) ad ca be computed from equatios (44) or (49), which deped o the umber of subscribers compared to the umber of available. Note that the blockig probabilities of each HN depeds o the umber of ad the coverage probabilities. TUCAN3G_D41UPCae.doc 52

54 idimesioal M/M/m/m queuig system This evaluatio is suitable for sceario 6 as described i Table 14, where we assume that two colocated HN with omi ateas are givig coverage to a certai populatio. Each UE is coected to oe HN or the other with equal probability. If oe HN caot provide service, the traffic is spilled to the other HN. Traffic balacig is ot cosidered amog the HN, i.e., each user directs iitially its call request to ay HN without takig ito accout the curret load of the HNs. I this way, we ca defie a bidimesioal Markov chai as show i Figure 22, where the first (secod) idex i each state is related with the umber of active calls beig atteded by HN A (HN ): H 0 1, N H 1 2, N N HA 1 N HA, N H H H 0, N 1 N H H 1, N 1 N HA H H N 1, N 1N HA H H N, N 1 N 0 1, N 1 1 2, N 1 HA HA N 1 N, N1 H H 0, N 2 N 1 H H 1, N 2 N 1 0,1 2 HA HA N 1,1 2 N,12 1, ,1 12,1 HA HA N 1 N,1 0,0 1 1,0 1 HA N 1,01 HA N,01 0 1,0 1 2,0 HA HA N 1 N,0 Figure 22. idimesioal Markov chai characterisig the voice traffic served by two HN of N, assumig a large umber of ihabitats where the call geeratio rates at each state are defied as the sum of the birth rate associated to a give HN (assumig that both HNs are placed o the same tower, oe half of the total birth rate is associated to each HN) plus the rate associated to the traffic spilled from the other HN due to the lack of coverage: 53

55 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig P 1 P P 2 P P j 0,..., N P 1 P P 2 P P i0,..., N cob A cob cob A cob cob A i i 1, j i i 1 jj1 i i 1 jj1 i i 1 cob cob A cob cob A cob A i, jj1 jj1 ii1 jj1 ii1 jj1 ad = L v is the total birth rate for all subscribers. Note that the traffic is ot spilled whe oe of the HN has occupied all its, which implies the followig formulatio for the call geeratio rates at j=n ad i=n A H i i 1, N HA N, jj1 P P cob A i i 1 cob jj1 Note that, sice the coverage area is the same, these birth rates are equal as log as the HNs have the same available power ad umber of chaels. The aalysis of this chai is doe by settig the icome ad outcome flows i each state as follows [Kleirock 75]: 1 1 P i j P i P P j P i, j i i 1, j i, jj1 i1, j i1, j i 1 i, j i, j1 i, j1 i, j 1 j A i 1,0 i,01 i,1 i1,0 i1 i,0 i1,0 P P P P 0,0 0 1,0 0,0 1 0,1 1,0 i,0 i i A A A A A A A N,0 N,01 N,1 N 1,0 N 1 N,0 1 0, j 0 1, j 0, j j 1 1, j 0, j 1 0, j 1 j 0, j 1 i1,, N 1 j 1,, N 1 A P P + P i1 i1,..., N 1 P N P P P j P P P j P N P P 0, N 0 1, N 0, N 1 0, N 1 N 1, N P N 1 P P + P i 1 in, i i 1, N i 1, N i 1 in, in, 1 in, 1 N i 1, N P N N P P A A A A A A A N, N N 1, N N 1 N, N N, N 1 N, N 1N A A A N, j N, jj1 P N j P j1 P P A A A A A A N, j1 N, j1 N, j1j N 1, j N 1 N, j together with a additioal equatio to ormalize the probabilities: j 1,..., N 1 A i1,..., N 1 j 1,..., N 1 (51) (52) (53) A N N Pi, j1 (54) i0 j0 All these equatios are liear ad ca be collected i the form: p e (55) where is a ((N+1) 2 +1) (N+1) 2 sparse full-colum rak matrix, p cotais all state probabilities, ad e = [ ] T. Note that, although matrix is ot square, (55) always admits a sigle solutio. Fially, the blockig ad cogestio probabilities are give by Cogestio probability: P P (56) 2D C A N, N TUCAN3G_D41UPCae.doc 54

56 lockig probability: will joitly deped o N A ad N, that is, the umber of allowed i each HN A N 1 N 1, A A,, 1 1, 1 N N N j i N 1 P N N P P P P P 2 D A cob, cob, A j j i i j0 i0 A N 1N 1 i0 j0 cob, A cob, 1 1 P P P i, j i i 1 j j 1 (57) It is importat to metio that the state probabilities P i,j which appear i equatios (56) ad (57) deped implicitly o N A ad N, as for each value of the pair (N A, N ) a flows balace equatio similar to the oe show i (55) eeds to be resolved. Overall, give a certai blockig probability (see sectio 4.1), we eed to evaluate the miimum umber of chaelizatio codes N A ad N so that the backhaul requiremets are miimized. A certai HN ad atea cofiguratio (from Table 14) is suitable for deploymet if N A ad N are less tha or equal to the umber of chaelizatio codes available at each HN Mixed voice ad data service It is assumed that two types of users are preset i the system: voice ad data users. Data users are costraied to be served at 128 Kbps, accordig to the specificatios of ip.access HN (see sectio 3.3). A total amout of N is available which ca be divided betwee voice ad data, depedig o the requiremets or demads at a give time istat. I the WCDMA case, the coditio for a voice or a data call request to be served is twofold: first, the mobile termial geeratig the ew icomig call has to be uder radio coverage (i.e., there should be eough power i the HN to serve it); secod, at least oe of the of the server has to be available. Note that the first coditio is related ot oly to pathloss but also to the itra-cell iterferece geerated by ogoig calls. The otatio used i this sectio is the followig: L v : umber of subscribers for voice services, L d : umber of subscribers for data services, N: umber of total available of the HN, i.e., umber of total users that the HN could serve simultaeously, P i,j : probability of havig i voice ad j data active, i.e., i + j calls beig served by the HN simultaeously, cob Pi i 1, j : probability of beig able to accommodate a ew icomig voice call request whe i voice calls ad j data calls are active, i.e., beig served simultaeously, cob Pi, j j 1: probability of beig able to accommodate a ew icomig data call request whe j data calls ad i voice calls are active, i.e., beig served simultaeously, v : rate of icomig voice calls geerated by a sigle subscriber, d : rate of icomig data calls geerated by a sigle subscriber, V= L v v : aggregated rate of icomig voice calls, D= L d d : aggregated rate of icomig data packets, 1/ v : mea duratio of a active voice call (we will assume that the duratio of each active call follows a expoetial distributio), 1/ d : mea duratio of a active data call (see equatio (6)). We will assume that the arrival times of voice calls ad data packets are idepedet ad that the time betwee cosecutive arrivals also follows a expoetial distributio characterized by the rate of geeratio of the icomig calls. We will also assume that the umber of subscribers i the cell is 55

57 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig much higher tha the umber of of the HN, which meas that the rate of geeratio of ew calls is idepedet of the umber of curret active calls. Let us adopt N=16 available for traffic delivery, beig the spreadig factors used 16 (for data traffic), 128 (for voice traffic) ad 256 (for commo chaels). For data traffic, it is clear that a maximum of 15 codes of legth 16 ca be used so that commo chaels ca be accommodated. cob Accordig to ayes rule, the probabilities Pi i 1, j ca be expressed as cob cob cob cob Pi i 1, j Pi 1, j P, where i, j P i, j is the probability that give i voice calls ad j data calls, all of them are uder radio coverage ad ca be served simultaeously (i.e., there is eough power at the HN to serve all). Likewise, the cob probabilities Pi, j j 1 ca be expressed as P cob cob cob i, jj 1 Pi, j 1 Pi,. These probabilities are evaluated j umerically usig the priciples described i sectio i a Mote Carlo simulatio test. I this framework we defie the blockig ad cogestio probabilities separately for voice ad data calls D M/M/m/m queuig system This evaluatio is suitable for scearios 1 to 4 as described i Table 14. Havig this i mid, the cotiuous-time Markov chai modelig this sceario is show i Figure 23. Note that the two dimesios are associated to voice chaels occupied (i the horizotal display) ad data chaels occupied (i the vertical display). Agai, state P 0,N caot exist if oly N-1 data ca be accommodated. Ufortuately the values of the state probabilities P i,j caot be obtaied i closed form. Rather we eed to adopt the theory of statioary cotiuous Markov chais to assess that the probability of reachig ay state ca be computed from the probability of the eighbor states ad the trasitio rates. We eed to equatio iput ad output flows i each state [Kleirock 75]. I this way it is possible to write dow the followig relatioships: P i P j P cob cob i, j v V i i 1, j d D i, jj1 cob cob cob cob 1 0, 1 0, 1 cob cob cob P i1 P P P j1 P P i1,, N 2 j 1,..., N i1 cob i1, j v i1, j V i1 i, j i, j1 d i, j1 D i, j1j P P j P P P 0, j V 0 1, j d D 0, j j 1 1, j v 0, j i,0 v V i i 1,0 D i,0 1 i,1 d i 1,0 V i 1 i,0 i 1,0 v cob cob i, Ni v( ) d i, Ni1 D i, Ni1Ni i1, NiV i 1 i, Ni cob cob 0,0 V 01,0 D 0,01 0,1d 1,0v cob N,0 v N1,0V N1 N,0 cob cob 0, N1 1d V 01, N1 0, N2D 0, N2N1 1, N1v TUCAN3G_D41UPCae.doc 56 P P j1 j 1,..., N 2 D j j j d P i P P P P P + P i1 i1,..., N 1 P i N i P P P P P P P P P N P P P N P P P P Also, we eed a additioal equatio to ormalize the probabilities: N1 N Ni P P 0, j i, j j0 i1 j0 P i1,..., N 1 All these equatios are liear ad ca be collected i the form: p e (58) where is a ((N 2 +3N)/2+1) (N 2 +3N)/2 sparse full-colum rak matrix, p cotais all state probabilities, ad e = [ ] T. Note that, although matrix is ot square, the previous set of equatios always admits a sigle solutio. 1

58 cob VP0 1, N1 cob DP0, N2N1 cob cob VP 01, N2 VP12, N2 cob DP0, N3N2 P cob D 0,23 P cob P cob V 01,2 V 1 2,2 P cob V 2 3,2 P cob D 0,1 2 P cob P cob V 01,1 V 1 2,1 P cob cob V 2 3,1 VPN 2N1,1 P cob D 0,01 P cob P cob D 1,01 D 2,01 P cob P cob P cob cob V 01,0 V 12,0 V 23,0 VPN 1 N,0 Figure 23. Markov chai characterisig mixed voice ad data users served by a HN of N, assumig a large umber of subscribers I the defiitio of blockig probabilities, we icorporate the parameters N v ad N d. N v (N d ) is the maximum umber of simultaeous voice (data) calls that ca be served. Takig this ito accout, a ew icomig voice (data) call will be served wheever the curret umber of voice (data) calls beig served is lower tha N v (N d ) ad the total umber of calls beig served is lower tha N (N-1). The objective of this work is to determie the miimum umber of voice (N v ) ad data (N d ) that are eeded to guaratee a give blockig probability. Depedig o their relative values, we may ecouter situatios as those preseted i Figure 24. Note that i the sloppy border N d + N v = N, beig N the total umber of codes available at the HN. The plot o the left (right) represets the case where N d + N v < N (N d + N v > N). The top-right corer (deoted as (N v *, N d *)) represets the state geeratig the largest voice plus data bitrate (ote the larger bitrate per chael associated data chaels as compared to voice chaels). 57

59 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig R R Figure 24. Examples of regios R(N v, N d ) from the bidimesioal Markov chai (i Figure 23) used to capture the umber of codes eeded to set the blockig probability at a give value. The top-right corer (deoted as (N v *, N d *)) represets the state geeratig the largest voice plus data bitrate Accordig to Figure 23 ad Figure 24, we defie the blockig ad cogestio probabilities as follows: Cogestio probability for voice: probability that all the of the HN are occupied, i.e., the HN could ot accommodate a ew call eve if it is geerated i a geographical positio where there is radio coverage NNv Nd NvN1 C, v v, d Nv, i NNd i, Nd i i0 i0 (59) P N N P P It should be oted that the secod term i the summatio oly appears as a cosequece of the slope i the regio R (see Figure 24), that is, oly if N d + N v > N. Cogestio probability for data: probability that all the of the HN are occupied, i.e., the HN could ot accommodate a ew data call eve if it is geerated i a geographical positio where there is radio coverage NNd Nd NvN (60) P N, N P P Cd, v d in, d NNd in, d i i0 i1 It should be oted that the secod term i the summatio oly appears as a cosequece of the slope i the regio R (see Figure 24), that is, oly if N d + N v > N. lockig probability for voice: probability that a ew icomig voice call caot be served. It is give by NN 1 d cob cob,, 1, 1 d d (61) P N N P N N P P P N N P, v v d C, v v d i, N i i 1, N i, j v d i i 1, j i0 i, j where regio R( v, d ) is the covex subset of states i Figure 23 defied by 0 v < N v, 0 d < N d ad v + d < N (ote that the orther, ortheaster ad easter borders are ot icluded i the regio). Figure 24 illustrates the cocept. lockig probability for data: probability that a ew icomig data call caot be served. It is give by TUCAN3G_D41UPCae.doc 58

60 NN 1 v cob cob,, 1, 1 v v (62) P N N P N N P P P N N P, d v d C, d v d N, i N, i i 1 i, j v d i, j j 1 i0 i, j where the regio R( v, d ) is defied above. Equatios (61) ad (62) are used i sectio 6 to determie the umber of codes for a target blockig probability. A certai HN ad atea cofiguratio (from those i Table 14) is suitable for deploymet if N v ad N d are less tha or equal to the umber of chaelizatio codes available. It is importat to metio that the probabilities i equatios (59) to (62) deped implicitly o N v ad N d, as for each value of the pair (N v, N d ). For each possible pair (N v, N d ) the flows balace set of equatios that has to be solved to fid the state probabilities is that obtaied from (58) by keepig oly the equatios correspodig to states (i,j) such that 0 i Nv, 0 j Nv, ad 0 i j N. The other equatios have to be skipped. I additio, those flows whose origis were i the states that are kept i the ew set of equatios ad that were directed towards states that have bee skipped, have to be elimiated i the ew set of equatios. Oce all N 2 possible regios R have bee tested, those fittig the target blockig probabilities for voice ad data (see sectio 4.1) are kept. For each of those regios, the backhaul is computed as sum of the cotributio of the N v * voice chaels plus the cotributio from the N d * data chaels of the top-right corer of the regio (the state geeratig the highest bitrate demads). Amog all regios kept, we decide the oe with the lowest badwidth demads Parallel 2D M/M/m/m queuig system This evaluatio is suitable for sceario 5 as described i Table 14, where we ca assume that two HN', say HA ad H, give coverage to two areas with directioal ateas whose radiatio patters do ot overlap (a approximatio i practice). Each area is geeratig traffic P A V, P A D ad P V, P D respectively. Note that P A ad P are the fractio of subscribers covered by HN A ad respectively. Parallel idepedet 2D M/M/m/ chais ca be used (as those show i Figure 23) sice there is o traffic spill betwee the two coverage areas whe oe of the HN is blocked. Whe this is the case, we are iterested i both areas havig the same probability of a call beig blocked which implies where A, A, A v v d A A A A A,,, max,,, A A A A A,,, max,,, P N N N N P N N P N N, v v d v d, v v d, v v d P N N N N P N N P N N, d v d v d, d v d, d v d P N N has bee defied i sectio (likewise for all other probabilities). Note that the blockig probabilities of each HN differ i the umber of but also o the coverage probabilities icluded i equatios (44) ad (49). Overall, give a target blockig probabilities for voice ad data, we eed to determie the miimum umber of chaelizatio codes N v ad N d that joitly match the target blockig probabilities, so that the backhaul requiremet is miimized. A certai HN ad atea cofiguratio (from Table 14) is suitable for deploymet if N v ad N d are less tha or equal to the umber of chaelizatio codes available. (63) D M/M/m/m queuig system This evaluatio is suitable for sceario 6 as described i Table 14. Havig this i mid, the aalysis tool is a 4D M/M/m/m Markov chai that ca be aalyzed usig the priciple of flows balace as i previous cases. The flows amog states are depicted i Figure 25, where the idexes of the states are defied as: i A ad i, refer to the umber of voice calls beig served simultaeously at HN A ad 59

61 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig respectively, while j A ad j deote the umber of data coectios beig served simultaeously at HN A ad, respectively. P i, j 1, i, j A A Pi 1, j, i, j A A D i, j 1 j, i, j A A A i A v V i 1 i, j, i, j D i, j, i, j j 1 A A A A A (j +1) d P i, j, i, j 1 A A D i, j j 1, i, j A A A (j A +1) d P i, j 1, i, j A A j A d V i, j, i 1 i, j A A P i A, j A, i, j V i i 1, j, i, j A A A P i, j, i 1, j A A i v D i, j, i, j 1 j A A (i A +1) v V i, j, i i 1, j A A Pi 1, j, i, j A A j d (i +1) v P i, j, i, j 1 A A P i, j, i 1, j A A Figure 25. Liks amog the states of a 4D Markov chai characterisig mixed voice ad data users served by two colocated HN with omidirectioal ateas The expressios for the birth rates ca be derived i a straightforward way from those i assumig that voice ad data traffic spill betwee HNs are possible. It is expressed as: V V cob cob A A i 1,,, 2 1, 1, ad ad ad AiA ja i j P ii j P iaia j i A Nv i j N ia Nv ia ja N 2 V V cob A cob A i,, 1, 2 1, 1, ad ad A ja ii j P iaia j P A ii j i A Nv ia ja N i Nv ad i j N 2 D D cob cob A A i, 1,, 2, 1, 1 ad ad ad A jaja i j P i jj PiA jaja j Nd i j N ja Nd ia ja N 2 D D cob A cob A i,,, 1 2, 1, 1 ad a A ja i jj P ia jaj P A i jj j A Nd ia ja N d j Nd ad i j N 2 At the border regio the birth rates are defied with the coditio of a HN ot beig able to accept ew calls: (64) TUCAN3G_D41UPCae.doc 60

62 P i N or i j N ad i N ad i j N P i N or i j N ad i N ad i j N cob A A P j 1 or ad ad A j Nd i j N ja Nd ia ja N P j N or i j N ad j N ad i j N V coba A iaia1, ja, i, j V iaia1, ja v A v A A V cob A ia, ja, ii1, j V ii1, j A v A A v D ia, jaja1, i, j D ia, ja D cob A ia, ja, i, jj1 D i, jj1 A d A A d As usual, the equilibrium equatios for flows at each state ca be collected i a liear set of equatios: (65) p e (66) where ow is a (((N 2 + 3N)/2) 2 + 1) ((N 2 + 3N)/2) 2 sparse full-colum rak matrix, p cotais all state probabilities, ad e = [ ] T. The defiitio of the cogestio ad blockig probabilities is as follows: Cogestio probability for voice: probability that oe of the HNs ca accommodate ew voice calls eve if it is geerated i a geographical positio with radio coverage. This is the situatio i the regio,,,,,, A i i j j i i j j i N or i j N ad i N or i j N (67) v A A A A A v A A v P N, N, N, N P A A C, v v d v d i, j, i, j i, i, j, j A A v A A Cogestio probability for data: probability that oe of the HNs ca accommodate ew data packet coectios eve if it is geerated i a geographical positio with radio coverage coverage. This is the situatio whe,,,,,, A i i j j i i j j j N or i j N ad j N or i j N (68) d A A A A A d A A d P N, N, N, N P A A C, d v d v d i, j, i, j i, i, j, j A A d A A lockig probability for voice: probability that a ew icomig voice call caot be served. It is give by the probability of cogestio plus the probability of a usuccessful voice call due i, i, j, j (defied i (67)): to the lack of coverage for states ot belogig to regio v A A, A A ia, i, ja, jv P N, N, N, N P N, N, N, N P 1 P i, i, j, j A A A A cob A, v v d v d C, v v d v d i, j, i, j A A where A A cob A cob,,, 1 1 A A A A A A A A ia, i, ja, jv cob A cob PiA, ja, i, j 1P 1,,,,,, 1, 1,, iaia i ja j PiA ja i j P ia ii ja j P N N N N P P P C, v v d v d i, j, i, j i i 1, i, j, j i, i i 1, j, j A i, i, j, j i, i, j, j A A v A A v i, i, j, j i, i, j, j i N ad i j N ad i N ad i j N A A v A A A A A v A A v deotes the ier states for the voice services, ad correspod to those states with two outgoig voice flows (see Figure 25), i, i, j, j i, i, j, j i N ad i j N ad i N or i j N A A v A A A A A v A A v deotes the states with icreasig i A ad oe departig voice flow (see Figure 25), ad i, i, j, j i, i, j, j i N ad i N or i j N A v A A A A v A v A A 61

63 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig i, i, j, j i, i, j, j i N ad i j N ad i N or i j N A v A A A A v A v A A deotes the states with icreasig i ad oe departig voice flow (see Figure 25). lockig probability for data: probability that a ew icomig data packet coectio caot be served. It is give by the probability of cogestio plus the probability of a usuccessful data packet coectio due to the lack of coverage for states ot belogig to regio i, i, j, j (defied i (68)): d A A P N, N, N, N P N, N, N, N P 1 P i, i, j, j A A A A cob A,, d v d v d C, d v d v d ia, ja, i, j A A ia, i, ja, jd A A cob A cob PC, d Nv, Nd, Nv, Nd Pi,,, 1,, 1, 1 A ja i j P ia i jaja j P ia, i, ja, jj1 A ia, i, ja, jd cob A cob PiA, ja, i, j 1P,, 1,,,, 1,,, 1 ia i jaja j PiA ja i j P ia i ja jj A i, i, j, j i, i, j, j where A A d A A d i, i, j, j i, i, j, j j N ad i j N ad j N ad i j N A A d A A A A A d A A d deotes the ier states for the data services, ad correspod to those states with two departig data flows (see Figure 25), i, i, j, j i, i, j, j j N ad i j N ad j N or i j N A A d A A A A A d A A d deotes the states with icreasig j A ad oe departig data flow (see Figure 25), ad i, i, j, j i, i, j, j j N ad i j N ad j N or i j N A d A A A A d A d A A deotes the states with icreasig j ad oe departig data flow (see Figure 25). TUCAN3G_D41UPCae.doc 62

64 5 NETWORK PLANNING RESULTS The objective of this sectio is to preset the etwork performace for each sceario depicted i Table 14 ad for each locatio cosidered i TUCAN3G where the etwork will be deployed. Network performace is evaluated i terms of CPICH coverage ad blockig probabilities as metioed i sectio 4.1. First, traffic desity assumptio over the geographical area i each locatio is described. As metioed previously, heterogeeous desity of users aroud the HN is assumed i order to characterize those places where is more likely that a user tries to establish oe coectio to the cell. This distributio depeds practically o the desity of housigs i the zoe. Oce the traffic distributio over the area is defied, we ca evaluate the ecessary power for CPICH ad the rest of commo chaels to cover the whole area of iterest, as detailed i The rest of HN power that has ot bee allocated to commo chaels may be available for dedicated chaels, i.e., to serve differet users. I this step, some scearios may ot be valid due to, for example, all the available power is eeded to trasport commo chaels ad o power is left for dedicated chaels, which typically happes for low power HNs. The, blockig ad cogestio probabilities are computed accordig to the expressios described i 4.4 assumig oly voice users (12.2 kbps AMR). The maximum blockig probability for cosiderig the valid performace of a particular sceario is 2%, as metioed i 4.1. These probabilities are evaluated for each case described i 4.4: Case A: umber of ihabitats is much higher tha the umber of available (codes). Case : umber of ihabitats ot much higher tha the umber of available. The, blockig ad cogestio probabilities are computed whe mixed voice ad FTP traffic is assumed, followig the guidelies of sectio To avoid showig excessive results, the dimesioig of voice ad data as well as the evaluatio of the backhaul are displayed i sectio Napo river regio Locatios i Napo river regio are: Sata Clotilde, Negro Urco ad Tuta Pisco. The followig sectios show the results for each locatio Sata Clotilde I Sata Clotilde, scearios #1, #2, #3, #4, #5 ad #6 depicted i Table 14 are tested. Sceario #4 is ot possible to test i this locatio sice the traffic distributio aroud the HN implies omidirectioal coverage. Table 4 ad Figure 10 show the locatio ad tower coordiates ad the itesity of traffic geeratio assumed over the geographical area, respectively. I Figure 26, two cells of sceario #5 are idetified by meas of the miimum path loss criteria described i Notice that the resultig coverage areas deped o the orietatios of each atea ad these orietatios are selected i such a way that each cell covers the approximately the same desity of traffic assumed over the area. Notice that Figure 26 does ot depict the coverage area for each cell, but i which cell oe user will try to establish a coectio depedig o its positio by followig the miimum path loss criteria described i No-coloured area represets the zoe outside the rage of HNs. 63

65 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Figure 26. Idetificatio of each cell (ot ecessarily showig coverage areas) i sceario #5 blue: cell 1; black: cell 2 Sata Clotilde Atea cofiguratio Table 17 summarizes the best atea cofiguratio (height ad dowtilt) depedig o the selected sceario of Table 14. The best cofiguratio is selected as the oe that miimizes the 95 th percetile of path loss of the area to cover, as metioed i sectio The differet 95 th percetiles of path loss obtaied depedig o the particular sceario are displayed i aex 9.2. Height Dowtilt Sceario #1 70 m - Sceario #2 70 m - Sceario #3 70 m 10º Sceario #4 - - Sceario #5 Cell 1: 70 m Cell 2: 70 m Cell 1: 10º Cell 2: 10º Sceario #6 Cell 1: 70 m Cell 2: 70 m Cell 1: 10º Cell 2: 10º Table 17. Optimum atea cofiguratios (Sata Clotilde) CPICH coverage Results obtaied for CPICH coverage evaluatio described i sectio are detailed i Table 18. This table shows the total trasmitted powers for commo chaels eeded to cover the area defied i sectio 4.1 i the worst case ad the available power for dedicated chaels. It ca be see that the HN does ot have eough power i scearios #1 ad #2 ad therefore, the available power for dedicated chaels is zero. However, atea gais i scearios #3,#5 ad #6 are higher ad thus, HN ca cover the whole area ad provide power to serve users. Notice i Table 11 ad Table 12 that HN has higher available power to serve users after CPICH plaig. Fially, it ca be see that scearios with higher atea gais (#3 ad #5) provide better performace (lower power for commo chaels ad higher power for dedicated chaels). For the particular case of sceario #6, power for commo chaels is equal at all HNs sice both must cover the same area. Thus, the available power for dedicated chaels after commo chaels plaig determies the capacity of the etwork to serve users i the dowlik ad therefore, the results obtaied are clearly related to the capacity results showed i sectio Fially, i order to illustrate how the type of HN ad the sceario impacts o the covered area, differet coverage maps are plotted i Figure 27 TUCAN3G_D41UPCae.doc 64

66 for differet situatios. Notice that the covered area would be the same i all scearios as log as the resultig power for commo chaels were more tha 13% of the total available power at HN. HN Sceario Percetage of area covered (%) #1 95 #2 95 #3 95 S-Class # #6 #1 #2 #3 #5 E-Class 24 #6 #1 #2 #3 #5 E-Class 24* #6 95 (cell 1) 95 (cell 2) (cell 1) 95 (cell 2) (cell 1) 97.2 (cell 2) Maximum path loss icludig shadow margi (d) (sector 1) (sector 2) (cell 1) (cell 2) (sector 1) (sector 2) (cell 1) (cell 2) (sector 1) (sector 2) (cell 1) (cell 2) TX power for commo chaels (dm) (sector 1) (sector 2) (cell 1) (cell 2) (sector 1) (sector 2) (cell 1) (cell 2) (sector 1) (sector 2) (cell 1) (cell 2) Available power for dedicated chaels (dm) (sector 1) (sector 2) (cell 1) (cell 2) (sector 1) (sector 2) (cell 1) (cell 2) (sector 1) (sector 2) (cell 1) (cell 2) bs Pi cch / Pmax (%) (sector 1) (sector 2) (cell 1) (cell 2) (sector 1) (sector 2) (cell 1) (cell 2) (sector 1) (sector 2) (cell 1) (cell 2) Table 18. CPICH coverage results (Sata Clotilde) (a) Sceario #1 ad HN S-class 16 (b) Sceario #5 ad HN E-class 24* Figure 27. Coverage areas for two scearios ad HN type (Sata Clotilde) lockig ad cogestio probabilities I sectio , blockig ad cogestio probabilities are evaluated whe oly voice traffic is assumed, while sectio is devoted to aalyse mixed traffic Voice Traffic Cogestio ad blockig probabilities for the dowlik obtaied for each cofiguratio i Sata Clotilde are preseted i Table 19: situatios where the target blockig probability (2%) is fulfilled are depicted i gree, while situatios where it is ot fulfilled are depicted i red. It ca be see that both cases for traffic model (case A ad case ) gives practically the same results i isolated-cell scearios 65

67 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig sice the umber of ihabitats is high i Sata Clotilde. I multi-cell scearios, oly case A traffic model is assumed to compute blockig ad cogestio probabilities, as it is the most pessimistic. Also, otice that HN E-class 24 provides a blockig probability of 1 i scearios #1 ad #2 due to the lack of available power for dedicated chaels as proved i Empty cells for certai scearios i Table 19, Table 20, Table 23, Table 24, Table 27, Table 28, Table 31, Table 32, Table 35 ad Table 36 idicate that case has ot bee evaluated. HN Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 0,0392 0,0379 0,3231 0,3205 #2 0,1191 0,1162 0,2514 0,2479 #3 0,1778 0,1740 0,2226 0,2186 #5 0,0151 0,0226 #6 0,0003 0,0013 # # # ,4838 0,4822 #5 1,27E-06 0,0566 #6 2,52E-16 0,1415 #1 0,0037 0,0034 0,1138 0,1114 #2 0,0122 0,0111 0,0638 0,0617 #3 0,0173 0,0159 0,0482 0,0462 #5 3,34E-05 0,0019 #6 3,49E-10 2,64E-05 Table 19. lockig ad cogestio probabilities for the dowlik (Sata Clotilde) As see i previous sectios, the blockig ad cogestio probabilities deped o the coverage probabilities. Notice that i all sigle-cell scearios, if HN is used, cogestio probabilities are zero sice it is ot possible to serve the maximum umber of simultaeous users (coverage probabilities are low due to HN does ot have eough power). If HN E-Class 24* or HN are used, the blockig probabilities for sigle-cell scearios are higher tha 2%. Cogestio probabilities for these cases are ear to 2% which meas that capacity is limited practically by the maximum umber of simultaeous users that the HN ca serve. Sceario #5 does ot provide blockig probabilities less tha 2% eve though multi-cell deploymet doubles the umber of available codes ad the available power i DL comparig with the isolated-cell scearios. However, sceario #6 provides cogestio probabilities less tha 2% if HN or HN E- Class 24* are used. This happes because i sceario #5, although the atea gais are higher at each cell, a ew user ca oly be served by oe HN (the coverage areas are ot overlapped); while i sceario #6, if oe HN caot provide service, the traffic is spilled to the other. I localities where the performace is limited by hard capacity, ad ot by available power at HNs, sceario #6 always provides lower blockig probabilities. Fially, Table 20 shows the blockig ad cogestio probabilities for the UL. Results obtaied for HN ad HN are the same sice the limitatio i the UL is associated to the available power at the UE ad to the maximum umber of codes at HNs. It ca be see that UL for all scearios has a better performace. Notice that the blockig ad cogestio probabilities are practically the same i all scearios whe ay HN is used. This meas that coverage probabilities are practically always oe, idepedetly of the umber of simultaeous users ad therefore, the capacity of the UL is basically limited by the maximum umber of simultaeous users supported by HNs. TUCAN3G_D41UPCae.doc 66

68 HN ad E-class 24* Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 0,2070 0,2030 0,2089 0,2046 #2 0,2073 0,2033 0,2086 0,2043 #3 0,2080 0,2040 0,2083 0,2040 #5 0,0167 0,0168 #6 0,0004 0,0004 #1 0,0240 0,0221 0,0263 0,02427 #2 0,0241 0,0222 0,0257 0,0237 #3 0,0242 0,0222 0,0249 0,0229 #5 3,8E-05 8,05E-05 #6 4,38E-10 6,72E-08 Table 20. lockig ad cogestio probabilities for the uplik (Sata Clotilde) Mixed voice ad data traffic Mixed voice ad data traffic results have ot bee icluded. Rather, the dimesioig of voice ad data as well as the evaluatio of the backhaul are displayed i sectio 6.2. Notice that the power allocatio for commo chaels is ot deped o the type of UE services ad therefore, the available power i the dowlik will be the same both for the case of oly voice users ad the case of mixed voice ad data users Negro Urco I Negro Urco, all scearios depicted i Table 14 are tested to evaluate the etwork performace. Sceario #4 is possible to test i this locatio sice the traffic distributio aroud the HN is cocetrated withi a sector of 180º. Table 4 ad Figure 11 shows the locatio ad tower coordiates ad the itesity of traffic geeratio assumed over the geographical area, respectively. I Figure 28, two cells of sceario #5 are idetified by meas of the miimum path loss criteria described i Notice that the resultig cells deped o the orietatios of each atea ad these orietatios are selected i such a way that each cell covers the approximately the same desity of traffic assumed over the area. Figure 28. Idetificatio of each cell (ot ecessarily showig coverage areas) i sceario #5 blue: cell 1; black: cell 2 Negro Urco 67

69 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Atea cofiguratio Table 21 summarizes the best atea cofiguratio (height ad dowtilt) depedig o the selected sceario of Table 14. The differet 95 th percetiles of path loss obtaied depedig o the particular sceario ca be see i aex Height Dowtilt Sceario #1 70 m - Sceario #2 70 m - Sceario #3 70 m 10º Sceario #4 70 m 10º Sceario #5 Sector 1: 70 m Sector 2: 70 m Sector 1: 10º Sector 2: 10º Sceario #6 Cell 1: 70 m Cell 2: 70 m Cell 1: 10º Cell 2: 10º Table 21. Optimum atea cofiguratios (Negro Urco) CPICH coverage Table 22 summarize the CPICH coverage results for differet scearios i Negro Urco. It ca be see that, whe is used, oly scearios #4 ad #5 are valid sice these provide eough power for dedicated chaels. O the other had, all cofiguratios whe HN ad HN E-Class 24* provide eough power for users. Fially, i order to illustrate how the type of HN ad the sceario impacts o the covered area, differet coverage maps are plotted i Figure 29 for differet percetages of covered area, which depeds o the specific HN type ad the selected sceario. HN S-Class 16 E-Class 24 E-Class 24* Sceario Percetage of area covered (%) Maximum path loss icludig shadow margi (d) TX power for commo chaels (dm) Available power for dedicated chaels (dm) cch bs P / P (%) # # # # # (sector 1) (sector 2) (sector 1) (sector 2) (sector 1) (sector 2) (sector 1) (sector 2) #6 95 (cell 1) 95 (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) # # # # # (sector 1) (sector 1) (sector 1) (sector 1) (sector 2) (sector 2) (sector 2) (sector 2) #6 95 (cell 1) (cell 1) (cell 1) 0 (cell 1) (cell 1) 95 (cell 2) (cell 2) (cell 2) 0 (cell 2) (cell 2) # # # # # (sector 1) (sector 1) (sector 1) (sector 1) #6 95 (cell 1) 95 (cell 2) (sector 2) (cell 1) (cell 2) (sector 2) (cell 1) (cell 2) (sector 2) (cell 1) (cell 2) Table 22. CPICH coverage results (Negro Urco) i max (sector 2) (cell 1) (cell 2) TUCAN3G_D41UPCae.doc 68

70 (a) Sceario #1 ad HN S-class 16 (b) Sceario #5 ad HN E-class 24* (c) Sceario #4 ad HN S-class 16 Figure 29. Coverage areas for three scearios ad HN types (Negro Urco) lockig ad cogestio probabilities Voice Traffic Cogestio ad blockig probabilities obtaied for each cofiguratio i Negro Urco are preseted i Table 23 ad Table 24. It ca be see that HN E-class 24 provides a blockig probability of 1 i scearios #1, #2,#3 ad #6 because there is o power available for dedicated chaels after CPICH plaig. Notice that i all scearios, the cogestio probabilities are very low. This happes sice the umber of ihabitats i Negro Urco is low ad therefore, the aggregated traffic to be served by the etwork is small. However, blockig probabilities are high i may cases which meas that etwork performace is practically limited by the available power at HN. Scearios that fulfil the requiremets for the DL are scearios #4, #5 ad #6 if HN is used, ad scearios #3, #4 ad #6 whe HN E- Class 24* is used (scearios with higher atea gais ad HNs with higher available power). For the UL all scearios, except the sceario #1 which has a lower atea gai at HN, provides a blockig probability less tha 2% ad otice that blockig ad cogestio probabilities are practically the same i valid scearios sice the etwork performace is limited by the hard capacity i the UL. lockig ad cogestio probabilities for the UL are expected to be always lower tha DL sice the coverage probabilities are always higher (each UL coectio has always more available power tha the DL). As a cosequece, UL results are irrelevat from the selected cofiguratio. 69

71 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario Cogestio Probability lockig Probability Case A Case Case A Case # ,2916 0,2906 #2 7,63E-15 3,66E-15 0,1375 0,1361 #3 3,98E-13 1,91E-13 0,0317 0,0307 #4 5,23E-12 2,51E-12 0,0006 0,0005 #5 2,5E-15 0,0006 #6 0 7,81E-05 - # # # # ,0598 0,0585 #5 9,35E-27 0,0453 # # ,0541 0,0529 #2 2,1E-23 3,75E-24 0,0147 0,0140 #3 3,66E-22 6,52E-23 0,0007 0,0007 #4 3,35E-21 5,97E-22 7,54E-05 6,62E-05 #5 3,07E-26 2,96E-05 #6 0 1,91E-09 Table 23. lockig ad cogestio probabilities for the dowlik (Negro Urco) HN ad Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 1.82E E #2 3.06E E E E-06 #3 3.37E E E E-11 #4 3.37E E E E-11 #5 2.62E E-15 #6 0 0 #1 1.94E E #2 4.49E-20 8E E E-06 #3 6.96E E E E-16 #4 6.95E E E E-06 #5 3.12E E-26 #6 0 0 Table 24. lockig ad cogestio probabilities for the uplik (Negro Urco) Mixed voice ad data traffic Mixed voice ad data traffic results have ot bee icluded. Rather, the dimesioig of voice ad data as well as the evaluatio of the backhaul are displayed i sectio Tuta Pisco I Tuta Pisco, scearios #1, #2 ad #3 ad #6 from Table 14 are tested i order to evaluate the etwork performace. Sceario #4 is ot cotemplated sice the traffic distributio aroud the HN implies the usage of omi-directioal ateas. As ca be see at Figure 12, traffic itesity o the geographical area implies that sceario #5 is ot useful. Table 4 ad Figure 12 show the locatio, tower coordiates ad the itesity of traffic assumed over the area. TUCAN3G_D41UPCae.doc 70

72 Atea cofiguratio Table 25 summarizes the best atea cofiguratio (height ad dowtilt) depedig o the selected sceario of Table 14. The differet 95 th percetiles of path loss obtaied depedig o the particular sceario ca be see i aex i sectio 9. Height Dowtilt Sceario #1 50 m - Sceario #2 50 m - Sceario #3 50 m 10º Sceario #6 50 m (cell 1) 50 m (cell 2) 10º (cell 1) 10º (cell 2) Table 25. Optimum atea cofiguratios (Tuta Pisco) CPICH coverage Table 26 summarizes the CPICH coverage results for differet scearios i Tuta Pisco. It ca be see that, whe is used, oly sceario #3 ad #6 are valid sice these provide eough power for dedicated chaels. O the other had, all cofiguratios whe HN ad HN E-Class 24* provide eough power to serve users. Coverage maps are plotted i Figure 30 for differet percetages of covered area, which depeds o the specific HN used ad the selected sceario. HN S-Class 16 E-Class 24 E-Class 24* Sceario Percetage of area covered (%) Maximum path loss icludig shadow margi (d) TX power for commo chaels (dm) Available power for dedicated chaels (dm) cch bs P / P (%) # # # # (cell 1) 98.9 (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) 13 (cell 1) 13 (cell 2) # # # #6 95 (cell 1) (cell 1) (cell 1) (cell 1) (cell 1) 95 (cell 2) (cell 2) (cell 2) (cell 2) (cell 2) # # # #6 100 (cell 1) 100 (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) Table 26. CPICH coverage results (Tuta Pisco) i 13 (cell 1) 13 (cell 2) max (a) Sceario #3 ad HN S-class 16 (b) Sceario #1 ad HN E-class 24* 71 Figure 30. Coverage area for two scearios ad HN types (Tuta Pisco)

73 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig lockig ad cogestio probabilities Voice Traffic Cogestio ad blockig probabilities obtaied for each cofiguratio i Tuta Pisco are preseted i Table 27 ad Table 28. It ca be see that HN E-class 24 provides a blockig probability of 1 i scearios #1, #2 because there is ot power available for dedicated chaels after CPICH plaig. Notice that i all scearios, the cogestio probabilities are very low. This happes sice the umber of ihabitats i Tuta Pisco is low ad therefore, the aggregated traffic to serve by the etwork is small. Scearios that fulfil the requiremets for the DL are scearios #3 ad #6 for HN ad scearios #1, #2, #3 ad #6 for HN. HN does ot have eough power to provide blockig probabilities less tha 2% i ay sceario. Fially, blockig ad cogestio probabilities are lower for UL tha for DL sice the coverage probabilities are always higher (each UL coectio has always more available power tha the DL). Therefore, UL results are irrelevat i the choice of a particular cofiguratio ad HN for each locatio. Notice that for the UL, all scearios are valid regardless of HN used. HN Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 6,26E-15 3,2E-15 0,0825 0,0811 #2 8,33E-14 4,27E-14 0,0278 0,0269 #3 4,94E-12 2,53E-12 0,0027 0,0025 #6 3,06E-16 5,51E-08 # # # ,2247 0,2232 #6 0 0,0253 # ,0059 0,0056 #2 2,43E-22 5,05E-23 0,0007 0,0006 #3 5,47E-21 1,13E-21 0,0002 0,0002 #6 0 6,48E-11 Table 27. lockig ad cogestio probabilities for the dowlik (Tuta Pisco) HN ad Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 1,16E-10 5,97E-11 0,0003 0,0003 #2 1,16E-10 5,96E-11 0,0001 0,0001 #3 1,17E-10 6E-11 1,17E-10 5,53E-11 #6 1,49E-16 1,49E-16 #1 4,7E-19 9,76E-20 0,0003 0,0003 #2 4,82E-19 1E-19 0,0001 0,0001 #3 4,86E-19 1,01E-19 4,86E-19 8,85E-20 #6 0 0 Table 28. lockig ad cogestio probabilities for the uplik (Tuta Pisco) Mixed voice ad data traffic Mixed voice ad data traffic results have ot bee icluded. Rather, the dimesioig of voice ad data as well as the evaluatio of the backhaul are displayed i sectio 6.2. TUCAN3G_D41UPCae.doc 72

74 5.2 Paraapura river regio Selected locatios i Paraapura river regio are Sa Gabriel ad Sa Jua. The followig sectios show the results for each locatio i this regio Sa Gabriel I Sa Gabriel, scearios #1, #2, #3, #5 ad #6 depicted i Table 14 are tested i order to evaluate the etwork performace. Sceario #4 is ot a suitable solutio to test i this locatio sice the traffic distributio aroud the HN implies a omidirectioal coverage. Table 4 ad Figure 13 show the locatio ad tower coordiates ad the itesity of traffic geeratio assumed over the geographical area, respectively. I Figure 31, two cells of sceario #5 are idetified by meas of the miimum path loss criteria described i Notice that the resultig cells deped o the orietatios of each atea ad these orietatios are selected i such a way that each cell covers the approximately the same desity of expected traffic over the area. Figure 31. Idetificatio of each cell i sceario #5 blue: cell 1; black: cell 2 Sa Gabriel Atea cofiguratio Table 29 summarizes the best atea cofiguratio (height ad dowtilt) depedig o the selected sceario of Table 14. The differet 95 th percetiles of path loss obtaied depedig o the particular sceario ca be see i aex Height Dowtilt Sceario #1 30 m - Sceario #2 30 m - Sceario #3 30 m 10º Sceario #4 - - Sceario #5 Sector 1: 40 m Sector 2: 20 m Sector 1: 10º Sector 2: 10º Sceario #6 30 m (cell 1) 30 m (cell 2) 10º (cell 1) 10º (cell 2) Table 29. Optimum atea cofiguratios (Sa Gabriel) CPICH coverage Table 30 summarize the CPICH coverage results for differet scearios i Sa Gabriel. It ca be see that all scearios are valid for each type of HN sice the available power to cover the worst path loss is eough. Moreover, the fial power allocated by commo chaels is the miimum power 73

75 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig for a acceptable acquisitio time (13% of total radiated power of HN or equivaletly a CPICH power of 5% of total radiated power of HN). It happes because the covered area is small ad resultig worth path loss to cover is low. Differet coverage maps are plotted i Figure 32 for differet percetages of covered area, which depeds o the specific HN ad the selected sceario. HN S-Class 16 E-Class 24 E-Class 24* Sceario Percetage of area covered (%) Maximum path loss icludig shadow margi (d) TX power for commo chaels (dm) Available power for dedicated chaels (dm) cch bs P / P (%) # # # # (sector 1) (sector 2) (sector 1) (sector 2) (sector 1) (sector 2) 13 (sector 1) 13 (sector 2) # (cell 1) 99.7 (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) 13 (cell 1) 13 (cell 2) # # # # (sector 1) (sector 1) (sector 1) 13 (sector 1) (sector 2) (sector 2) (sector 2) 13 (sector 2) # (cell 1) (cell 1) (cell 1) (cell 1) 13 (cell 1) 99.2 (cell 2) (cell 2) (cell 2) (cell 2) 13 (cell 2) # # # # (sector 1) (sector 1) (sector 1) 13 (sector 1) #6 100 (cell 1) 100 (cell 2) (sector 2) (cell 1) (cell 2) (sector 2) (cell 1) (cell 2) (sector 2) (cell 1) (cell 2) Table 30. CPICH coverage results (Sa Gabriel) i max 13 (sector 2) 13 (cell 1) 13 (cell 2) (a) Sceario #1 ad HN S-class 16 (b) Sceario #1 ad HN E-class 24 Figure 32. Coverage areas for two scearios ad ad HN types (Sa Gabriel) lockig ad cogestio probabilities Voice Traffic Cogestio ad blockig probabilities obtaied for each cofiguratio i Sa Gabriel are preseted i Table 31 ad Table 32. It ca be see that all cofiguratios provide blockig probabilities lower tha 2% for each type of HN. Notice that despite sceario #3 uses a higher atea gai, scearios #1 ad TUCAN3G_D41UPCae.doc 74

76 #2 may provide better performace if HN or HN are used. It happes because ateas with wide elevatio beamwidth may provide overall better power budget i this coverage area. est sceario may also deped o the specific HN because coverage area is differet if power for commo chaels is fixed as the 13% of total radiated power. For the UL, blockig probabilities are always lower tha blockig probabilities i the DL ad furthermore, these coicide with cogestio probabilities i all scearios sice i the UL, etwork performace is totally limited by the umber of available chaelizatio codes. HN Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 5,04E-05 4,06E-05 0,0002 0,0001 #2 5,14E-05 4,14E-05 6,84E-05 5,56E-05 #3 5,09E-05 4,1E-05 0,0001 0,0001 #5 5,17E-08 1,37E-06 #6 6,79E-16 4,75E-12 #1 1,09E-13 6,36E-14 0,0183 0,0174 #2 3,03E-12 1,78E-12 0,0047 0,0043 #3 1,34E-10 7,85E-11 0,0007 0,0006 #5 7,74E-15 1,77E-06 #6 5,47E-17 2,3E-10 #1 6,94E-10 4,07E-10 2,68E-07 2,22E-07 #2 7,09E-10 4,16E-10 1,13E-09 6,98E-10 #3 6,46E-10 3,79E-10 4,89E-05 4,53E-05 #5 7,54E-15 9,45E-14 #6 0 0 Table 31. lockig ad cogestio probabilities for the dowlik (Sa Gabriel) HN ad Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 5,18E-05 4,18E-05 5,18E-05 4,08E-05 #2 5,18E-05 4,18E-05 5,18E-05 4,08E-05 #3 5,18E-05 4,18E-05 5,18E-05 4,08E-05 #5 5,18E-08 5,18E-08 #6 7,02E-16 7,02E-16 #1 7,1E-10 4,17E-10 7,1E-10 4,01E-10 #2 7,1E-10 4,17E-10 7,1E-10 4,01E-10 #3 7,1E-10 4,17E-10 7,1E-10 4,01E-10 #5 8,24E-15 8,24E-15 #6 5,18E-17 5,18E-17 Table 32. lockig ad cogestio probabilities for the uplik (Sa Gabriel) Mixed voice ad data traffic Mixed voice ad data traffic results have ot bee icluded. Rather, the dimesioig of voice ad data as well as the evaluatio of the backhaul are displayed i sectio Sa Jua I Sa Gabriel, scearios #1, #2, #3, #5 ad #6 depicted i Table 14 are tested i order to evaluate the etwork performace. Sceario #4 is ot possible to test i this locatio sice the traffic distributio aroud the HN implies omidirectioal coverage. Table 4 ad Figure 14 shows the locatio ad 75

77 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig tower coordiates ad the itesity of traffic geeratio assumed over the geographical area, respectively. I Figure 33, two cells of sceario #5 are idetified by meas of the miimum path loss criteria described i Notice that the resultig cells deped o the orietatios of each atea ad these orietatios are selected i such a way that each cell covers the approximately the same desity of expected traffic over the area. Figure 33. Idetificatio of each cell i sceario #5 blue: cell 1; black: cell 2 Sa Jua Atea cofiguratio Table 33 summarizes the best atea cofiguratio (height ad dowtilt) depedig o the selected sceario of Table 14. The differet 95 th percetiles of path loss obtaied depedig o the particular sceario ca be see i aex Height Dowtilt Sceario #1 20 m - Sceario #2 20 m - Sceario #3 20 m 10º Sceario #4 - - Sceario #5 Sector 1: 20 m Sector 2: 20 m Sector 1: 20º Sector 2: 10º Sceario #6 20 m (cell 1) 20 m (cell 2) 10º (cell 1) 10º (cell 2) Table 33. Optimum atea cofiguratios (Sa Jua) CPICH coverage Table 34 summarize the CPICH coverage results for differet scearios i Sa Jua. It ca be see that all scearios are valid for each type of HN sice the available power to cover the worst path loss is eough. Moreover, the power allocated by commo chaels is the miimum power for a acceptable acquisitio time (13% of total radiated power of HN or equivaletly a CPICH power of 5% of total radiated power of HN). It happes because the covered area is small ad resultig worst path loss to cover is low. Also, otice that for all scearios the whole area is covered (100%). TUCAN3G_D41UPCae.doc 76

78 HN S-Class 16 E-Class 24 E-Class 24* Sceario Percetage of area covered (%) Maximum path loss icludig shadow margi (d) TX power for commo chaels (dm) Available power for dedicated chaels (dm) cch bs P / P (%) # # # # (sector 1) (sector 2) (sector 1) (sector 2) (sector 1) (sector 2) 13 (sector 1) 13 (sector 2) #6 100 (cell 1) 100 (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) (cell 1) (cell 2) 13 (cell 1) 13 (cell 2) # # # # (sector 1) (sector 1) (sector 1) 13 (sector 1) (sector 2) (sector 2) (sector 2) 13 (sector 2) #6 100 (cell 1) (cell 1) (cell 1) (cell 1) 13 (cell 1) 100 (cell 2) (cell 2) (cell 2) (cell 2) 13 (cell 2) # # # # (sector 1) (sector 1) (sector 1) 13 (sector 1) #6 100 (cell 1) 100 (cell 2) (sector 2) (cell 1) (cell 2) (sector 2) (cell 1) (cell 2) (sector 2) (cell 1) (cell 2) Table 34. CPICH coverage results (Sa Jua) i max 13 (sector 2) 13 (cell 1) 13 (cell 2) lockig ad cogestio probabilities Voice Traffic Cogestio ad blockig probabilities obtaied for each cofiguratio i Sa Jua are preseted i Table 35 ad Table 36. All cofiguratios yield blockig below 2% for each type of HN. Cogestio probabilities coicide with blockig probabilities practically i all scearios usig ay HN sice the etwork performace is limited by the umber of available chaelizatio codes at HN. Notice that despite sceario #3 uses a higher atea gai, scearios #1 ad #2 may provide better performace. It happes because ateas with wide elevatio beamwidth may provide overall improved lik budget i this coverage area. I the UL, all sigle-cell scearios provide the same results regardless of the used HN type. 77 HN (13 dm) (24 dm) Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 1,33E-17 1,62E-18 1,33E-17 1,22E-18 #2 1,33E-17 1,62E-18 1,33E-17 1,22E-18 #3 1,33E-17 1,62E-18 3,93E-15 1,27E-15 #5 6,68E-22 6,68E-22 #6 0 0 #1 1,02E-29 6,1E-32 2,78E-26 9,11E-28 #2 1,02E-29 6,11E-32 1,02E-29 3,8E-32 #3 7,65E-30 4,58E-32 3,99E-08 2,96E-08 #5 3,06E-36 4,9E-17 #6 0 0 #1 1,02E-29 6,11E-32 1,02E-29 3,8E-32 #2 1,02E-29 6,11E-32 1,02E-29 3,8E-32 #3 1,02E-29 6,11E-32 8,73E-28 1,93E-29 #5 3,14E-36 3,14E-36 #6 0 0 Table 35. lockig ad cogestio probabilities for the dowlik (Sa Jua)

79 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario Cogestio Probability lockig Probability Case A Case Case A Case #1 1,33E-17 1,62E-18 1,33E-17 1,22E-18 #2 1,33E-17 1,62E-18 1,33E-17 1,22E-18 #3 1,33E-17 1,62E-18 1,33E-17 1,22E-18 #5 6,68E-22 6,68E-22 #6 0 0 #1 1,02E-29 6,11E-32 1,02E-29 3,8E-32 #2 1,02E-29 6,11E-32 1,02E-29 3,8E-32 #3 1,02E-29 6,11E-32 1,02E-29 3,8E-32 #5 3,14E-36 3,14E-36 #6 0 0 Table 36. lockig ad cogestio probabilities for the uplik (Sa Jua) Mixed voice ad data traffic Mixed voice ad data traffic results have ot bee icluded. Rather, the dimesioig of voice ad data as well as the evaluatio of the backhaul are displayed i sectio 6.2. TUCAN3G_D41UPCae.doc 78

80 6 EVALUATION OF ACKHAUL REQUIREMENTS This sectio evaluates the backhaul badwidth demaded by the access etwork i the busy hour, accordig to the populatio distributio ad traffic demads as it was aalysed i ad sectio 2.1. I accordace to previous sectios, voice-oly service is studied first ad the mixed voice ad data services. 6.1 ackhaul for voice traffic The voice traffic geerated i each locatio is obtaied as described i sectio 2.3. The umber of voice N v for the traffic at the peak hour ad a blockig probability of 2% is obtaied umerically accordig to what is described i sectio As the voice traffic is symmetric we adopt the highest amog DL ad UL i the peak hour. Results for each locatio are give i the followig tables kbps of backhaul for each voice coectio would be used for UL+DL Ideal case (uit coverage probability) Here it is assumed that all coverage probabilities are oe, idepedet o the traffic desity or available trasmitted power. These results may be used as a referece of performace if larger power at the HN were available. Table 37 to Table 41 show the backhaul whe uit coverage probabilities are supposed for each locatio ad year. Those situatios where the target blockig probability caot be achieved are marked with dash symbol -. Table 37 shows that the etwork ca oly satisfy the requiremets i the first year i Sata Clotilde. Oly multi-cell scearios ca fulfill the requiremets ad it ca be see that ay HN ca provide eough to serve users. From the secod year ad o, the umber of available chaelizatio codes at HNs is ot eough to serve the expected icrease i the voice traffic. For the other localities, the performace requiremets ca be fulfilled durig first five years (assumig ideal coverage probabilities), ay HN class ca provide eough umber of voice, with the exceptio of Sa Gabriel, where oly HN ad HN ca offer eough chaelizatio codes to serve voice users i sigle-cell scearios from the secod year ad o. Table 38 shows the miimum umber of voice codes eeded i Negro Urco to serve the icreasig expected voice traffic durig five years. It ca be see that ay HN ca provide eough umber of for all scearios. Table 39 summarizes the miimum umber of i Tuta Pisco, where ay HN has eough chaelizatio codes to serve voice users regardless of the chose sceario. The umber of voice i Sa Gabriel ca be see i Table 40 ad, as metioed above, all HNs have eough codes to serve the expected voice traffic if multi-cell sceario is assumed; while oly HN ad HN ca provide eough for sigle-cell scearios. Fially, Table 41 shows the miimum umber of i Sa Jua ad i this case, all scearios usig ay HN are capable to serve the expected voice traffic alog the five years. 79

81 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Year 1 Year 2 Year 3 Year 4 Year 5 Year Sceario lockig probability #1,#2,#3,# #5 14 (sector 1) 0, (sector 2) #6 25 0, #1,#2,#3,# # # #1,#2,#3,# # # #1,#2,#3,# # # #1,#2,#3,# # # ackhaul (Kbps) Table 37. codes ad resultig block probabilities for the uplik ad dowlik whe ideal coverage is assumed (Sata Clotilde) Year 1 Year 2 Year 3 Year 4 Year 5 Year Sceario lockig probability #1,#2,#3,#4 6 0, #5 4 (sector 1) 0, (sector 2) #6 6 0, #1,#2,#3,#4 10 0, #5 7 (sector 1) 0, (sector 2) #6 10 0, #1,#2,#3,#4 10 0, #5 7 (sector 1) 0, (sector 2) #6 10 0, #1,#2,#3,#4 10 0, #5 7 (sector 1) 0, (sector 2) #6 10 0, #1,#2,#3,#4 11 0, #5 7 (sector 1) 0, (sector 2) #6 11 0, ackhaul (Kbps) Table 38. codes ad resultig block probabilities for the uplik ad dowlik whe ideal coverage is assumed (Negro Urco) TUCAN3G_D41UPCae.doc 80

82 Year 1 Year Year 1 Year 2 Year 2 Year 3 Sceario lockig probability #1,#2,#3,#4 6 0, #6 6 0, #1,#2,#3,#4 11 0, #6 11 0, #1,#2,#3,#4 11 0, #6 11 0, #1,#2,#3,#4 11 0, #6 11 0, Year 4 #1,#2,#3,#4 11 0, #6 11 0, ackhaul (Kbps) Table 39. codes ad resultig block probabilities for the uplik ad dowlik whe ideal coverage is assumed (Tuta Pisco) Year 1 Year 2 Year 3 Year 4 Year 5 Year Sceario lockig probability #1,#2,#3,#4 10 0, #5 6 (sector 1) 0, (sector 2) #6 10 0, #1,#2,#3,#4 22 0, #5 12 (sector 1) 0, (sector 2) #6 22 0, #1,#2,#3,#4 22 0, #5 12 (sector 1) 0, (sector 2) #6 22 0, #1,#2,#3,#4 23 0, #5 12 (sector 1) 0, (sector 2) #6 23 0, #1,#2,#3,#4 23 0, #5 13 (sector 1) 0, (sector 2) #6 23 0, ackhaul (Kbps) Table 40. codes ad resultig block probabilities for the uplik ad dowlik whe ideal coverage is assumed (Sa Gabriel) 81

83 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Year 0 Year 1 Year 2 Year 3 Year 4 Year Sceario lockig probability #1,#2,#3,#4 4 0, #5 3 (sector 1) 0, (sector 2) #6 4 0, #1,#2,#3,#4 6 0, #5 4 (sector 1) 0, (sector 2) #6 6 0, #1,#2,#3,#4 6 0, #5 4 (sector 1) 0, (sector 2) #6 6 0, #1,#2,#3,#4 6 0, #5 4 (sector 1) 0, (sector 2) #6 6 0, #1,#2,#3,#4 6 0, #5 4 (sector 1) 0, (sector 2) #6 6 0, ackhaul (Kbps) Table 41. codes ad resultig block probabilities for the uplik ad dowlik whe ideal coverage is assumed (Sa Jua) Realistic evaluatio Actual results are displayed here, assumig the three types of ip.access HNs, ad simulated (adoptig the models i ad hece lower tha 1) coverage probabilities that take ito cosideratio the itracell iterferece. Those situatios where the target blockig probability caot be achieved are marked with dash symbol Sata Clotilde Table 42 ad Table 43 summarize the miimum umber of voice to serve the voice traffic expected i Sata Clotilde durig the first five years. I Table 42 it ca be see that oly sceario #6 ca satisfy the target blockig probability if HN E-Class 16 is used, while sceario #5 ad sceario #6 ca fulfill the requiremets with HN is use. Obviously, as give i sectio 6.1.1, startig from the secod year, there is o available HN type with eough codes to serve the expected voice traffic. TUCAN3G_D41UPCae.doc 82

84 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # # #6 25 0, , # # # # # # # # #5 14 (sector 1) 14 (sector 1) 16 (sector 2) 0, (sector 2) 0, #6 25 0, , Table 42. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sata Clotilde) Year 1 HN Sceario Uplik lockig probability Dowlik lockig probability # # # # # # # # # # # # # # # ackhaul (Kbps) Table 43. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sata Clotilde) Year 2 to year Negro Urco Table 44, Table 45, Table 46, Table 47 ad Table 48 summarize the miimum umber of voice to serve the voice traffic i Negro Urco durig the first five years. Durig the first year, it ca be see that scearios #4, #5 ad #6 ca fulfill the requiremets if HN is used, while if HN is used, the requiremets are satisfied by scearios #2, #3, #4, #5 ad #6. From the secod year, oly sceario #6 ca fulfill the target blockig probability if HN S-Class is assumed, ad scearios #4, #5 ad #6 with the HN is used. Due to the rise i voice traffic startig from the secod year, sceario #4, which has the atea with higher gai, is the oly isolated-cell sceario capable to serve this expected traffic. 83

85 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #4 6 0, , #5 4 (sector 1) 4 (sector 2) 0, (sector 1) 4 (sector 2) 0, #6 6 0, , # # # # # # # #2 6 0, , #3 6 0, , #4 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , Table 44. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Negro Urco) Year 1 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # # # #6 10 0, , # # # # # # # # # #4 10 0, , #5 7 (sector 1) 7 (sector 1) 6 (sector 2) 0, (sector 2) 0, #6 10 0, , Table 45. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Negro Urco) Year 2 TUCAN3G_D41UPCae.doc 84

86 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # # # #6 10 0, , # # # # # # # # # #4 10 0, , #5 7 (Sector 1) 7 (Sector 1) 7 (Sector 2) 0, (Sector 2) 0, #6 10 0, , Table 46. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Negro Urco) Year 3 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # # # #6 10 0, , # # # # # # # # # #4 10 0, , #5 7 (Sector 1) 7 (Sector 1) 7 (Sector 2) 0, (Sector 2) 0, #6 10 0, , Table 47. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Negro Urco) Year 4 85

87 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # # # #6 11 0, , # # # # # # # # # #4 11 0, , #5 7 (Sector 1) 7 (Sector 1) 7 (Sector 2) 0, (Sector 2) 0, #6 11 0, , Table 48. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Negro Urco) Year Tuta Pisco Table 49, Table 50, Table 51, Table 52 ad Table 53 show the miimum umber of voice to serve the expected voice traffic i Tuta Pisco durig the first five years. Durig first year scearios #3 ad #6 ca fulfill the requiremets with HN, while all scearios ca satisfy the blockig probability with HN. From the secod year, oly sceario #6 ca fulfill the target blockig probability with HN S-Class, ad scearios #3 ad #6 with HN E-Class 24*. HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # #3 6 0, , #6 6 0, , # # # # #1 6 0, , #2 6 0, , #3 6 0, , #6 6 0, , Table 49. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Tuta Pisco) Year 1 TUCAN3G_D41UPCae.doc 86

88 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #6 11 0, , # # # # # # #3 11 0, , #6 11 0, , Table 50. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Tuta Pisco) Year 2 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #6 11 0, , # # # # # # #3 11 0, , #6 11 0, , Table 51. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Tuta Pisco) Year 3 87

89 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #6 11 0, , # # # # # # #3 11 0, , #6 11 0, , Table 52. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Tuta Pisco) Year 4 HN Sceario Number Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #6 11 0, , # # # # # # #3 11 0, , #6 11 0, , Table 53. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Tuta Pisco) Year Sa Gabriel Table 54, Table 55, Table 56, Table 57 ad Table 58 show the miimum umber of voice to serve the expected voice traffic i Sa Gabriel durig the first five years. Durig the first year all HN have eough chaelizatio codes to serve the expected voice users usig ay sceario. From the secod year, multi-cell scearios (#5 ad #6) satisfy the target blockig probability regardless of the HN type. However, oly HN has eough codes ad available power to serve the expected voice traffic satisfyig the requiremets for ay isolated-cell sceario. TUCAN3G_D41UPCae.doc 88

90 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) #1 10 0, , #2 10 0, , #3 10 0, , #5 6 (sector 1) 7 (sector 2) 0, (sector 1) 7 (sector 2) 0, #6 10 0, , #1 10 0, , #2 10 0, , #3 10 0, , #5 6 (sector 1) 6 (sector 1) 7 (sector 2) 0, (sector 2) 0, #6 10 0, , #1 10 0, , #2 10 0, , #3 10 0, , #5 6 (sector 1) 6 (sector 1) 7 (sector 2) 0, (sector 2) 0, #6 10 0, , Table 54. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Gabriel) Year 1 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #5 12 (sector 1) 12 (sector 1) 14 (sector 2) 0, (sector 2) 0, #6 22 0, , # # # #5 12 (sector 1) 12 (sector 1) 14 (sector 2) 0, (sector 2) 0, #6 22 0, , #1 22 0, , #2 22 0, , #3 22 0, , #5 12 (sector 1) 12 (sector 1) 14 (sector 2) 0, (sector 2) 0, #6 22 0, , Table 55. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Gabriel) Year 2 89

91 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #5 12 (sector 1) 12 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 22 0, , # # # #5 12 (sector 1) 12 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 22 0, , #1 22 0, , #2 22 0, , #3 22 0, , #5 12 (sector 1) 12 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 22 0, , Table 56. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Gabriel) Year 3 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #5 12 (sector 1) 12 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 23 0, , # # # #5 12 (sector 1) 12 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 23 0, , #1 23 0, , #2 23 0, , #3 23 0, , #5 12 (sector 1) 12 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 23 0, , Table 57. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Gabriel) Year 4 TUCAN3G_D41UPCae.doc 90

92 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) # # # #5 13 (sector 1) 13 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 23 0, , # # # #5 13 (sector 1) 13 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 23 0, , #1 23 0, , #2 23 0, , #3 23 0, , #5 13 (sector 1) 13 (sector 1) 15 (sector 2) 0, (sector 2) 0, #6 23 0, , Table 58. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Gabriel) Year Sa Jua Table 59, Table 60, Table 61, Table 62 ad Table 63 show the miimum umber of voice to serve the expected voice traffic i Sa Jua durig the first five years. Alog the first five years, all HN types have eough chaelizatio codes ad available power to serve the expected voice users usig ay sceario. 91

93 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario Uplik lockig probability Dowlik lockig probabilit y ackhaul (Kbps) #1 4 0, , #2 4 0, , #3 4 0, , #5 3 (sector 1) 3 (sector 2) 0, (sector 1) 3 (sector 2) 0, #6 4 0, , #1 4 0, , #2 4 0, , #3 4 0, , #5 3 (sector 1) 3 (sector 1) 3 (sector 2) 0, (sector 2) 0, #6 4 0, , #1 4 0, , #2 4 0, , #3 4 0, , #5 3 (sector 1) 3 (sector 1) 3 (sector 2) 0, (sector 2) 0, #6 4 0, , Table 59. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Jua) Year 1 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 2) 0, (sector 1) 4 (sector 2) 0, #6 6 0, , #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , Table 60. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Jua) Year 2 TUCAN3G_D41UPCae.doc 92

94 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 2) 0, (sector 1) 4 (sector 2) 0, #6 6 0, , #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , Table 61. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Jua) Year 3 HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 2) 0, (sector 1) 4 (sector 2) 0, #6 6 0, , #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , Table 62. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Jua) Year 4 93

95 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario Uplik lockig probability Dowlik lockig probability ackhaul (Kbps) #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 2) 0, (sector 1) 4 (sector 2) 0, #6 6 0, , #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , #1 6 0, , #2 6 0, , #3 6 0, , #5 4 (sector 1) 4 (sector 1) 4 (sector 2) 0, (sector 2) 0, #6 6 0, , Table 63. codes ad resultig block probabilities for the uplik ad dowlik whe oly voice users are assumed (Sa Jua) Year ackhaul for mixed voice ad data traffic The voice ad data traffics geerated i each locatio are obtaied as described i sectio 2.3. For the evaluatio of the umber of chaels associated to voice traffic (N v ) ad data traffic (N d ) that fit a certai blockig probability, we follow this methodology: 1. Adopt a HN type + atea cofiguratio providig the grade of service, both for voice ad data. 2. Evaluate the umber of chaelizatio codes to meet the blockig probability for voice ad data traffic (for UL ad DL), call it N v ad N d respectively, accordig to the guidelies of sectio Evaluate the voice plus data traffic for these chaels. For voice chaels assume 25 kbps of backhaul. For data traffic, assume 128 kbps for both UL ad DL, plus additioal overhead due to sigallig of 20%. For the mixed voice ad data case we caot defie a peak hour (otice that traffic i Figure 4 ad Figure 8 peak at differet hours). y aalysig the traffic through the day, three peak hours ca be idetified where both voice ad data traffic are larger tha the rest. The umber of has bee evaluated o each hour. For voice, the umber of (the same for the UL ad DL sice symmetric desig) will be the maximum umber of amog the three hours both for the UL ad for the DL. The umber of data will be the maximum umber of amog the three most loaded hours, differetiatig betwee UL ad DL (asymmetry is assumed for data traffic). The dimesio of backhaul will be the maximum amog the three hours, differetiatig betwee UL ad DL. Notice that the dimesio of backhaul is ot determied by the resultig total umber of for voice ad data traffic sice it may happe that the maximum umber of voice ad maximum umber of data are i differet hours. Therefore, despite access etwork must have eough codes, ot all will be used at the same time. TUCAN3G_D41UPCae.doc 94

96 6.2.1 Ideal case (uit coverage probability) Here it is assumed that all coverage probabilities are oe, idepedetly o the traffic desity or available power. Results are a referece of performace if larger power at the HN were available. Table 64 to Table 69 show the backhaul whe uit coverage probabilities are supposed for each locatio ad year. Those situatios where the target blockig probability caot be achieved are marked with dash symbol -. Table 64 shows that the expected voice ad data traffic i Sta Clotilde caot be served with the desired blockig probability usig ay of the scearios after the secod year, cofirmig what has already bee see i sectio whe oly voice traffic is assumed. This is why a aalysis for dataoly traffic is show i Table 65, assumig that the existig GSM etwork could serve all voice traffic. This study is ot repeated for the other localities. Notice that all HNs provide eough chaelisatio codes ad available power usig ay sceario cosiderig oly data users are assumed. Whe voice ad data users are assumed, the blockig probability depeds o the umber of maximum simultaeous users supported by the HN if N v plus N d is higher tha this umber of maximum users, although coverage probabilities are always oe. This happes because the cogestio probability will be higher for HNs with lower umber of maximum users. Therefore, i those places ad years where N v plus N d is higher tha the maximum umber of simultaeous users supported by ay of HNs, we must distiguish differet cases. Table 66 ad Table 67 summarize the umber of voice ad data i Negro Urco ad Tuta Pisco cosiderig the ideal coverage probabilities. It ca be see that the expected traffic is low eough for all HNs to fulfill the requiremets i ay sceario. Table 68 shows the voice ad data i Sa Gabriel. I this locatio, It is expected that HN does ot have eough codes to serve the expected traffic startig from the secod year with the assumptio of isolated-cell sceario. Startig from the third year, other type of HNs either do ot have eough codes i the same type of scearios. Fially, Table 69 shows the umber of voice ad data i Sa Jua. It ca be see that the expected traffic is low eough for all HNs to fulfill the requiremets i ay sceario. 95

97 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Year HN Sceario Year 1 Year 2 Year 3 Year 4 Year 5 Year 1 ad All HNs All HNs All HNs All HNs Uplik (VOICE) (DATA) ackhaul (kbps) Dowlik (VOICE) (DATA) ackhaul (kbps) #1.#2,#3,# # # , ,2 #1.#2,#3,# #5 14 (sector 1) 3 (sector 1) 14 (sector 1) 5 (sector 1) 16 (sector 2) 3 (sector 2) 1671,6 16 (sector 2) 5 (sector 2) 2236 # , ,2 #1.#2,#3,# # # #1.#2,#3,# # # #1.#2,#3,# # # #1.#2,#3,# # # Table 64. codes for the uplik ad dowlik whe voice ad data users are assumed ad ideal coverage (Sata Clotilde) Year Year 1 Year 2 Year 3 Year 4 Year 5 Uplik Dowlik ackhaul ackhaul Sceario lockig lockig Probability (kbps) Probability (kbps) (DATA) (DATA) #1.#2,#3,#4 4 0, ,4 7 0, ,2 #5 3 (sector 1) 5 (sector 1) 3 (sector 2) 0, ,6 5 (sector 2) 0, #6 4 0, ,4 7 0, ,2 #1.#2,#3,#4 7 0, , #5 5 (sector 1) 8 (sector 1) 5 (sector 2) 0, (sector 2) 0, ,8 #6 7 0, , #1.#2,#3,#4 7 0, , #5 5 (sector 1) 9 (sector 1) 5 (sector 2) 0, (sector 2) 0, ,4 #6 7 0, , #1.#2,#3,#4 8 0, ,8 15 0, #5 5 (sector 1) 9 (sector 1) 5 (sector 2) 0, (sector 2) 0, ,4 #6 8 0, ,8 15 0, #1.#2,#3,#4 8 0, ,8 15 0, #5 5 (sector 1) 9 (sector 1) 6 (sector 2) 0, ,6 10 (sector 2) 0, ,4 #6 8 0, ,8 15 0, Table 65. codes for the uplik ad dowlik whe oly data users are assumed ad ideal coverage (Sata Clotilde) TUCAN3G_D41UPCae.doc 96

98 Year Year 1 Year 2 Year 3 Year 4 Year 5 Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) #1.#2,#3,# , ,8 #5 4 (sector 1) 4 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 4 (sector 1) 4 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 # , ,8 #1.#2,#3,# , ,4 #5 7 (sector 1) 2 (sector 1) 7 (sector 1) 3 (sector 1) 6 (sector 2) 2 (sector 2) 939,4 6 (sector 2) 3 (sector 2) 1246,6 # , ,4 #1.#2,#3,# , ,4 #5 7 (sector 1) 2 (sector 1) 7 (sector 1) 3 (sector 1) 7 (sector 2) 2 (sector 2) 964,4 7 (sector 2) 3 (sector 2) 1271,6 # , ,4 #1.#2,#3,# , ,4 #5 7 (sector 1) 2 (sector 1) 7 (sector 1) 3 (sector 1) 7 (sector 2) 2 (sector 2) 964,4 7 (sector 2) 3 (sector 2) 1271,6 # , ,4 #1.#2,#3,# , ,4 #5 7 (sector 1) 2 (sector 1) 7 (sector 1) 3 (sector 1) 7 (sector 2) 2 (sector 2) 964,4 7 (sector 2) 3 (sector 2) 1271,6 # ,4 Table 66. codes for the uplik ad dowlik whe voice ad data users are assumed ad ideal coverage (Negro Urco) Year Year 1 Year 1 Year 2 Year 2 Year 3 Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) #1.#2,#3,# , ,8 # , ,8 #1.#2,#3,# , ,4 # , ,4 #1.#2,#3,# , ,4 # , ,4 #1.#2,#3,# , ,4 # , ,4 Year 4 #1.#2,#3,# , ,4 # , ,4 Table 67. codes for the uplik ad dowlik whe voice ad data users are assumed ad ideal coverage (Tuta Pisco) 97

99 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig Year HN Sceario Year 1 Year 2 Year 3 Year 4 Year 5 Year 1 Year 2 Year 3 All HNs ad ad ad ad Uplik Dowlik (VOICE) (DATA) ackhaul (kbps) (VOICE) (DATA) ackhaul (kbps) #1.#2,#3,# , ,4 #5 6 (sector 1) 2 (sector 1) 939,4 6 (sector 1) 3 (sector 1) 7 (sector 2) 2 (sector 2) 7 (sector 2) 3 (sector 2) 1246,6 # , ,4 #1.#2,#3,# #5 12 (sector 1) 14 (sector 2) 3 (sector 1) 3 (sector 2) 1546,6 12 (sector 1) 14 (sector 2) 4 (sector 1) 5 (sector 2) 1957,4 # , ,2 #1.#2,#3,# , ,2 #5 12 (sector 1) 3 (sector 1) 1571,6 12 (sector 1) 4 (sector 1) 14 (sector 2) 3 (sector 2) 14 (sector 2) 5 (sector 2) 1982,4 # , ,2 #1.#2,#3,# #5 12 (sector 1) 3 (sector 1) 1546,6 12 (sector 1) 4 (sector 1) 15 (sector 2) 3 (sector 2) 15 (sector 2) 5 (sector 2) 1957,4 # , ,2 #1.#2,#3,# #5 12 (sector 1) 3 (sector 1) 1596,6 12 (sector 1) 4 (sector 1) 15 (sector 2) 3 (sector 2) 15 (sector 2) 5 (sector 2) 2057,4 # , ,2 #1.#2,#3,# #5 12 (sector 1) 3 (sector 1) 1546,6 12 (sector 1) 4 (sector 1) 16 (sector 2) 3 (sector 2) 16 (sector 2) 6 (sector 2) 2086 # , ,2 #1.#2,#3,# #5 12 (sector 1) 3 (sector 1) 1596,6 12 (sector 1) 4 (sector 1) 15 (sector 2) 3 (sector 2) 15 (sector 2) 5 (sector 2) 2057,4 # , ,2 #1.#2,#3,# # # , ,2 #1.#2,#3,# #5 13 (sector 1) 15 (sector 2) 3 (sector 1) 3 (sector 2) 1621,6 13 (sector 1) 15 (sector 2) 5 (sector 1) 5 (sector 2) 2186 # , ,2 Table 68. codes for the uplik ad dowlik whe voice ad data users are assumed ad ideal coverage (Sa Gabriel) TUCAN3G_D41UPCae.doc 98

100 Year Year 1 Year 2 Year 3 Year 4 Year 5 Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) #1.#2,#3,# #5 3 (sector 1) 3 (sector 2) 1 (sector 1) 1 (sector 2) 457,2 3 (sector 1) 3 (sector 2) 2 (sector 1) 2 (sector 2) 764,4 # #1.#2,#3,# #5 4 (sector 1) 2 (sector 1) 814,4 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 4 (sector 2) 2 (sector 2) 814,4 # #1.#2,#3,# #5 4 (sector 1) 2 (sector 1) 814,4 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 4 (sector 2) 2 (sector 2) 814,4 # #1.#2,#3,# #5 4 (sector 1) 2 (sector 1) 814,4 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 4 (sector 2) 2 (sector 2) 814,4 # #1.#2,#3,# #5 4 (sector 1) 2 (sector 1) 814,4 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 4 (sector 2) 2 (sector 2) 814,4 # Table 69. codes for the uplik ad dowlik whe voice ad data users are assumed ad ideal coverage (Sa Jua) Realistic evaluatio Results with more realistic assumptios are displayed here, assumig the three types of ip.access HNs, ad simulated (adoptig the models i ad hece lower tha 1) coverage probabilities. Those situatios where the target blockig probability caot be achieved are marked with dashed symbol -. I all cases, results are show i years 0 through 5, i differet tables Sata Clotilde Table 70 shows the umber of voice ad data i the first year. If we assume realistic coverage probabilities, sceario #6 fulfils the requiremets with HN is used, while scearios #5 ad #6 satisfy the requiremets with HN. From the secod year, it has bee see as ay HN does ot have eough chaelizatio codes to serve the expected voice traffic. Thus, Table 71, Table 72, Table 73, Table 74 ad Table 75 display the case of data-oly traffic as oe of the studied scearios studied is able to accommodate mixed voice ad data with the blockig probability. It is certai that that the existig GSM etwork ru by Telefoica del Peru could serve all voice traffic. I the first year, HN has eough codes ad available power to satisfy the requiremets for scearios #5 ad #6 are assumed. O the other had, HN has eough codes ad available power to satisfy the requiremets for sceario #6, while ad HN E-Class 24* has eough codes ad available power to satisfy the requiremets for scearios #3,#5 ad #6 are assumed. Startig from the secod year, oly sceario #5 with HN ad scearios #5 ad #6 for the HN ca fulfil the requiremets. 99

101 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # # , ,8 # # # # # # # # #5 14 (sector 1) 3 (sector 1) 14 (sector 1) 5 (sector 1) 17 (sector 2) 3 (sector 2) 1671,6 17 (sector 2) 7 (sector 2) 2568,2 # , ,8 Table 70. codes for the uplik ad dowlik whe voice ad data users are assumed (Sata Clotilde) Year 1 HN Sceario (DATA) Uplik lockig probability ackhaul (kbps) (DATA) Dowlik lockig probability ackhaul (kbps) # # # #5 3 (sector 1) 3 (sector 2) 0, ,6 5 (sector 1) 6 (sector 2) 0, ,6 # # # # # #6 4 0, ,4 10 0, # # #3 4 0, ,4 10 0, #5 3 (sector 1) 5 (sector 1) 3 (sector 2) 0, ,6 5 (sector 2) 0, # Table 71. codes for the uplik ad dowlik whe oly data users are assumed (Sata Clotilde) Year 1 TUCAN3G_D41UPCae.doc 100

102 HN Sceario (DATA) Uplik lockig probability ackhaul (kbps) (DATA) Dowlik lockig probability ackhaul (kbps) # # # # #6 7 0, ,2 15 0, # # # # # # # # #5 5 (sector 1) (sector 1) (sector 2) 0, (sector 1) 0,0155 #6 7 0, ,2 14 0, ,4 Table 72. codes for the uplik ad dowlik whe oly data users are assumed (Sata Clotilde) Year 2 HN Sceario (DATA) Uplik lockig probability ackhaul (kbps) (DATA) Dowlik lockig probability ackhaul (kbps) # # # # #6 7 0, ,2 15 0, # # # # # # # # #5 5 (sector 1) (sector 1) (sector 2) 0, (sector 1) 0,0177 #6 7 0, ,2 15 0, Table 73. codes for the uplik ad dowlik whe oly data users are assumed (Sata Clotilde) Year 3 101

103 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario (DATA) Uplik lockig probability ackhaul (kbps) (DATA) Dowlik lockig probability ackhaul (kbps) # # # # #6 8 0, ,8 16 0, ,6 # # # # # # # # #5 5 (sector 1) (sector 1) (sector 2) 0, (sector 1) 0,0187 #6 8 0, ,8 15 0, Table 74. codes for the uplik ad dowlik whe oly data users are assumed (Sata Clotilde) Year 4 HN Sceario (DATA) Uplik lockig probability ackhaul (kbps) (DATA) Dowlik lockig probability ackhaul (kbps) # # # # #6 8 0, ,8 16 0, ,6 # # # # # # # # #5 5 (sector 1) 1689,6 9 (sector 1) (sector 2) 0, (sector 1) 0,0198 #6 8 0, ,8 15 0, Table 75. codes for the uplik ad dowlik whe oly data users are assumed (Sata Clotilde) Year Negro Urco Table 76, Table 77, Table 78, Table 79 ad Table 80 show the miimum umber of voice ad data to serve the expected traffic i Negro Urco durig the first five years. I the first year, HN has eough ad available power if sceario #6 is assumed ad HN E-Class 24* has eough ad available power if scearios #5 ad #6 are assumed. From secod year ad o, oly HN ca satisfy the requiremets if sceario #6 is assumed. TUCAN3G_D41UPCae.doc 102

104 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # # # , ,8 # # # # # # # # # # , ,8 #5 4 (sector 1) 2 (sector 1) 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 814,4 4 (sector 2) 2 (sector 2) 814,4 # , ,8 Table 76. codes for the uplik ad dowlik whe voice ad data users are assumed (Negro Urco) Year 1 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # # # # # # # # # # # # # # # , ,4 Table 77. codes for the uplik ad dowlik whe voice ad data users are assumed (Negro Urco) Year 2 103

105 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # # # # # # # # # # # # # # # , ,4 Table 78. codes for the uplik ad dowlik whe voice ad data users are assumed (Negro Urco) Year 3 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # # # # # # # # # # # # # # # , ,4 Table 79. codes for the uplik ad dowlik whe voice ad data users are assumed (Negro Urco) Year 4 TUCAN3G_D41UPCae.doc 104

106 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # # # # # # # # # # # # # # # , ,4 Table 80. codes for the uplik ad dowlik whe voice ad data users are assumed (Negro Urco) Year Tuta Pisco Table 81, Table 82, Table 83, Table 84 ad Table 85 show the miimum umber of voice ad data to serve the expected traffic i Tuta Pisco durig the first five years. I the first year, HN has eough ad available power if sceario #6 is assumed ad HN E-Class 24* has eough ad available power if scearios #3 ad #6 are assumed. Startig from the secod year, oly sceario #6 ca be used for both HNs. HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # , ,8 # # # # # # # , ,4 # , ,8 Table 81. codes for the uplik ad dowlik whe voice ad data users are assumed (Tuta Pisco) Year 1 105

107 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # , ,4 # # # # # # # # , ,4 Table 82. codes for the uplik ad dowlik whe voice ad data users are assumed (Tuta Pisco) Year 2 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # , ,4 # # # # # # # # , ,4 Table 83. codes for the uplik ad dowlik whe voice ad data users are assumed (Tuta Pisco) Year 3 TUCAN3G_D41UPCae.doc 106

108 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # , ,4 # # # # # # # # , ,4 Table 84. codes for the uplik ad dowlik whe voice ad data users are assumed (Tuta Pisco) Year 4 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # , ,4 # # # # # # # # , ,4 Table 85. codes for the uplik ad dowlik whe voice ad data users are assumed (Tuta Pisco) Year Sa Gabriel Table 86, Table 87, Table 88, Table 89 ad Table 90 show the miimum umber of voice ad data to serve the expected traffic i Sa Gabriel durig the first five years. I the first year, HN ad HN have eough ad available power if ay sceario is assumed, while HN has eough ad available power if scearios #3, #5 ad #6 are assumed. From secod year, oly sceario #6 ca be used for both HNs. it ca be see that due to the high expected traffic i this locality, oly multi-cell scearios ca fulfill the requiremets for 5 years: sceario #5 if HN or HN is used ad sceario #6 for ay HN. 107

109 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # , ,4 # , ,4 # , ,4 #5 6 (sector 1) 7 (sector 2) 2 (sector 1) 2 (sector 2) 939,4 6 (sector 1) 7 (sector 2) 3 (sector 1) 3 (sector 2) 1246,6 # , ,4 # # # , #5 6 (sector 1) 2 (sector 1) 6 (sector 1) 3 (sector 1) 7 (sector 2) 2 (sector 2) 939,4 7 (sector 2) 3 (sector 2) 1246,6 # , ,4 # , ,4 # , ,4 # , ,4 #5 6 (sector 1) 2 (sector 1) 6 (sector 1) 3 (sector 1) 7 (sector 2) 2 (sector 2) 939,4 7 (sector 2) 3 (sector 2) 1246,6 # , ,4 Table 86. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Gabriel) Year 1 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # #5 12 (sector 1) 14 (sector 2) 3 (sector 1) 3 (sector 2) 1546,6 12 (sector 1) 14 (sector 2) 5 (sector 1) 5 (sector 2) 2086 # , ,2 # # # #5 12 (sector 1) 3 (sector 1) 5 (sector 1) 14 (sector 2) 3 (sector 2) 1571, (sector 2) 2161 # , ,2 # # , ,2 # #5 12 (sector 1) 3 (sector 1) 5 (sector 1) 14 (sector 2) 3 (sector 2) 1571, (sector 2) 1982,4 # , ,2 Table 87. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Gabriel) Year 2 TUCAN3G_D41UPCae.doc 108

110 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # #5 12 (sector 1) 15 (sector 2) 3 (sector 1) 3 (sector 2) 1546,6 12 (sector 1) 15 (sector 2) 5 (sector 1) 6 (sector 2) 2214,6 # , ,2 # # # #5 12 (sector 1) 3 (sector 1) 12 (sector 1) 5 (sector 1) 15 (sector 2) 3 (sector 2) 1596,6 15 (sector 2) 5 (sector 2) 2186 # , ,2 # # # #5 12 (sector 1) 3 (sector 1) 12 (sector 1) 5 (sector 1) 15 (sector 2) 3 (sector 2) 1596,6 15 (sector 2) 5 (sector 2) 2186 # , ,2 Table 88. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Gabriel) Year 3 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # # , ,2 # # # #5 12 (sector 1) 3 (sector 1) 12 (sector 1) 5 (sector 1) 15 (sector 2) 3 (sector 2) 1596,6 15 (sector 2) 5 (sector 2) 2186 # , ,2 # # # #5 12 (sector 1) 3 (sector 1) 12 (sector 1) 5 (sector 1) 15 (sector 2) 3 (sector 2) 1596,6 15 (sector 2) 5 (sector 2) 2186 # , ,2 Table 89. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Gabriel) Year 4 109

111 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # # # # # , ,2 # # # #5 13 (sector 1) 3 (sector 1) 13 (sector 1) 5 (sector 1) 15 (sector 2) 3 (sector 2) 1621,6 15 (sector 2) 5 (sector 2) 2186 # , ,2 # # # #5 13 (sector 1) 3 (sector 1) 13 (sector 1) 3 (sector 1) 15 (sector 2) 3 (sector 2) 1621,6 15 (sector 2) 3 (sector 2) 2186 # , ,2 Table 90. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Gabriel) Year Sa Jua Table 91, Table 92, Table 93, Table 94 ad Table 95 show the miimum umber of voice ad data to serve the expected traffic i Sa Jua durig the first five years. Due to the traffic expected i this locality is low (few ihabitats), all HNs have eough codes ad available power for five years to serve all data ad voice users for ay sceario. HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # , ,2 # , ,2 # , ,2 #5 3 (sector 1) 3 (sector 2) 1 (sector 1) 1 (sector 2) 457,2 3 (sector 1) 3 (sector 2) 2 (sector 1) 2 (sector 2) 764,4 # , ,2 # , ,2 # , ,2 # , ,2 #5 3 (sector 1) 1 (sector 1) 3 (sector 1) 2 (sector 1) 3 (sector 2) 1 (sector 2) 457,2 3 (sector 2) 2 (sector 2) 764,4 # , ,2 # , ,2 # , ,2 # , ,2 #5 3 (sector 1) 1 (sector 1) 3 (sector 1) 2 (sector 1) 3 (sector 2) 1 (sector 2) 457,2 3 (sector 2) 2 (sector 2) 764,4 # , ,2 Table 91. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Jua) Year 1 TUCAN3G_D41UPCae.doc 110

112 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # , ,8 # , ,8 # , ,8 #5 3 (sector 1) 3 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 3 (sector 1) 3 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 # , ,8 # , ,8 # , ,8 # , ,8 #5 3 (sector 1) 2 (sector 1) 3 (sector 1) 2 (sector 1) 3 (sector 2) 2 (sector 2) 814,4 3 (sector 2) 2 (sector 2) 814,4 # , ,8 # , ,8 # , ,8 # , ,8 #5 3 (sector 1) 2 (sector 1) 3 (sector 1) 2 (sector 1) 3 (sector 2) 2 (sector 2) 814,4 3 (sector 2) 2 (sector 2) 814,4 # , ,8 Table 92. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Jua) Year 2 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 4 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 4 (sector 1) 4 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 # , ,8 # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 2 (sector 1) 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 814,4 4 (sector 2) 2 (sector 2) 814,4 # , ,8 # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 2 (sector 1) 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 814,4 4 (sector 2) 2 (sector 2) 814,4 # ,8 Table 93. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Jua) Year 3 111

113 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 4 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 4 (sector 1) 4 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 # , ,8 # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 2 (sector 1) 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 814,4 4 (sector 2) 2 (sector 2) 814,4 # , ,8 # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 2 (sector 1) 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 814,4 4 (sector 2) 2 (sector 2) 814,4 # ,8 Table 94. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Jua) Year 4 HN Sceario (VOICE) Uplik (DATA) ackhaul (kbps) (VOICE) Dowlik (DATA) ackhaul (kbps) # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 4 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 4 (sector 1) 4 (sector 2) 2 (sector 1) 2 (sector 2) 814,4 # , ,8 # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 2 (sector 1) 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 814,4 4 (sector 2) 2 (sector 2) 814,4 # , ,8 # , ,8 # , ,8 # , ,8 #5 4 (sector 1) 2 (sector 1) 4 (sector 1) 2 (sector 1) 4 (sector 2) 2 (sector 2) 814,4 4 (sector 2) 2 (sector 2) 814,4 # ,8 Table 95. codes for the uplik ad dowlik whe voice ad data users are assumed (Sa Jua) Year 5 TUCAN3G_D41UPCae.doc 112

114 7 ENERGY PROVISION FOR HN 7.1 Power cosumptio models of HN EARTH project provides a power cosumptio model for various types of LTE ase Statios i [EARTH-D23]. The model is useful for quatifyig the eergy losses of differet S compoets: atea iterface, power amplifier, radio frequecy trasceiver, basebad iterface, AC/DC power supply, coolig ad mai supply. Thus, the power cosumed by a base statio is modeled as a fuctio of the radiated power by [EARTH-D23]: P i N P P 0 P P NTRX Psleep PRF 0 TRX 0 p RF RF max where P max is the maximum RF output power, P 0 is the liear model parameter to represet power cosumptio at zero RF output power, p is the slope of the load depedig o the radiated power, N TRX deotes the umber of trasmitter ad receivers preset at the base statio ad fially, P sleep deotes the power cosumed i the sleep mode. These parameters deped o the type of base statio. For example Table 96 presets these values for Macro-, Micro-, Pico- y Femto- base statios. S type N TRX P max (W) P 0 (W) p P sleep (W) Macro 6 20 (43 dm) Micro (38 dm) Pico (21 dm) Femto (17 dm) Table 96. Power model parameters for differet S types Figure 34 presets the cosumed power i terms of dw by the differet Ss cosidered i Table 96 as a fuctio of the RF output power. We ca observe that the differeces i cosumed power at low/medium trasmitted power ad at full-load of the Ss are up to 2.3 d, 0.9 d, 0.31 d, ad 0.22 d for macro-, micro-, pico- ad femto-s, respectively. Moreover the savigs i cosumed power at the S whe i sleep mode with respect to the active mode are aroud of 2 d. (69) 7.2 Radiated power evaluatio The objectives of this sectio are to aalyse the radiated power by the HN, or HNs i multi-cell sceario, ad to determie the expressio of total radiated eergy per day by HNs. This radiated eergy will be used o eergy uit dimesioig. First, we ca obtai the distributio of trasmitted powers by HNs through the simulatio tool. For differet umber of simultaeous users, a differet trasmitted power distributio ca be obtaied. We are iterested i the average trasmitted powers to evaluate the radiated power ad therefore, the power cotrol headroom ad shadowig margi are ot take ito accout. Figure 35 shows the average trasmitted power for each umber of simultaeous voice users whe the sceario #2 ad sceario #3 are assumed for Negro Urco locatio ad HN S-class 16 are used. It ca be see that if a cofiguratio with a higher atea gai is used, the average trasmitted power by the HN is lower. 113

115 ICT STP Documet umber: D41 Title of deliverable: UMTS/HA etwork dimesioig 31.5 Macro S Micro S Cosumed Power (d) Cosumed Power (d) RF Output Power (dm) Pico S RF Output Power (dm) Femto S Cosumed Power (d) Cosumed Power (d) RF Output Power (dm) RF Output Power (dm) Figure 34. Total cosumed power (dw) as a fuctio of the RF output power (dm). Topleft: Macro S, Top-right: Micro S, ottom-left: Pico S ad ottom-right: Femto-S 9.8 Figure 35. Average radiated powers (dm) depedig o the umber of simultaeous voice users i Negro Urco ad HN S-class 16. Scearios #2 (left) ad #3 (right). Figure 36 shows the probability desity fuctio of the radiated power for the sceario #2 whe there are 6 simultaeous users o the etwork. TUCAN3G_D41UPCae.doc 114

116 Figure 36. Probability desity fuctio of the radiated power (dm) coditioed to the umber of simultaeous users is equal to 6. Sceario #2 i Negro Urco, HN S-class 16 Notice that the desity fuctio of the radiated power coditioed to the umber of simultaeous users is ot take ito accout the variatio of traffic desity per hour alog the day, but oly depeds o the traffic desity assumptios o the geographical area. I followig sectios, the traffic variatio alog the day is importat ad it must be cosidered for eergy uit dimesioig. First, let f R P i deotes the desity fuctio of the radiated power coditioed to i umber of chaels (or simultaeous users). This desity fuctio is depicted i Figure 36 for 6 simultaeous users. Let pih deote the probability of the umber of simultaeous users beig i, for a certai hour h of the day. This last probability is the probability of state i i the Markov chai described i 4.4, for particular desity traffic,, which depeds o the specific hour of the day (differet traffic desities alog the day are characterized i sectio 2.1). Through differet traffic desities per hour alog the day ad through the coverage probabilities defied i 4.3.7, we ca compute all the probabilities of beig i simultaeous users for each particular hour. The, the desity fuctio of the total radiated power coditioed to a particular hour of the day ca be expressed as f ( P h) f( P i) p( i h) (70) R I i1 where I deotes the maximum umber of simultaeous chaels supported by the etwork, as obtaied from the target blockig probability, ad specified i sectio 6 for each locatio, sceario ad HN type. Fially, we compute the average radiated eergy by the HN alog the day. Let PR ( h ) deote the average radiated power by the HN depedig o the hour of the day: PR ( h) PR f( PR h) dpr (71) ad the average radiated eergy by the HN alog the day i W h will be give by 23 h0 R L PR ( h) (72) 115