Aircom_UMTS Advanced Cell Planning and Optimisation PS-TR-005-O036_v5.0

8/13/2019 Aircom_UMTS Advanced Cell Planning and Optimisation PS-TR-005-O036_v5.0

1/292

Copyright 2003 AIRCOM International LtdAll rights reserved

AIRCOM Training is committed to providing our customers with quality instructor ledTelecommunications Training.

This documentation is protected by copyright. No part of the contents of this documentation may bereproduced in any form, or by any means, without the prior written consent of AIRCOM International.

Document Number: P/TR/005/O036/v5

This manual prepared by: AIRCOM InternationalGrosvenor House65-71 London RoadRedhill, Surrey RH1 1LQENGLAND

Telephone: +44 (0) 1737 775700Fax: +44 (0) 1737 775770Web: http://www.aircom.co.uk

UMTS Advanced Cell Planning and

Optimisation

O036


2/292


3/292

Contents1 Introduction 7

1.1 Course Overview 7

2 Optimisation Overview 10

2.1 What is Optimisation? 10

3 Network Dimensioning and Planning 17

3.1 Introduction 173.2 Simulating the Effect of Imperfect Site Location and HighSites 253.3 Provisioning for Asymmetric Traffic 323.4 Using More Appropriate Path Loss Models 363.5 Serving Very High Traffic Densities 423.6 Evaluating Simulator Results 453.7 Pilot Pollution 46

Simulation Examples 51

4 Site Location Issues 554.1 The Ideal Situation 55

4.1.1 Mis-placed sites. 584.2 Hot Spots 604.3 Site Density 644.4 High Sites 70

5 Factors Limiting Capacity 74

5.1 Cell Throughput 745.2 The Effect of Mobility on Capacity 75

5.3 Maximising Frequency Re-use Efficiency 785.4 Downlink Capacity and Orthogonality 815.4.1 Pilot SIR as an indicator of downlink capacity 85

5.5 The Noise Rise Limit 87

6 Antenna Selection 89

6.1 Antenna Gain & Coverage 896.2 Repeaters 916.3 Roll-out Optimised Configuration (ROC) 93

7 Soft Handover Issues 97

7.1 Macro-diversity & Maximal Combining Gain 977.1.1 Exercise 1 101


4/292

7.1.2 Exercise 2 1017.2 Optimising Soft Handover Parameters 102

8 Parameter Planning 107

8.1 Introduction 107

8.2 Pilot Channel Power 1088.3 Maximum Power per User 1108.4 Common Channel Powers 111

9 Multi-frequency Planning 112

9.1 Network Performance 112

10 Micro-cell Planning 114

10.1 Introduction 11410.2 Micro-cells targeting hot spots 115

11 Coverage & Capacity 125

11.1 Introduction 12511.2 Exercise Link Budgets 12911.3 Downlink Limited 13211.4 Predicting the Capacity of the Downlink 136

11.4.1 Example 149

12 Analysis, Prediction and Optimisation of DownlinkCapacity. 151

12.1 Analysis of Identical Users 15212.1.1 Verification Using A Monte Carlo Simulator. 15412.2 A Rapid Method for Estimating Downlink Capacity 155

12.2.1 Isolated Cells 15512.2.2 An Evenly Loaded Network 15712.2.3 Uplink-downlink balance. 159

12.3 Interim Conclusion 15912.4 Simulator-aided Prediction for Unevenly-loadedNetworks 15912.5 Optimisation Issues 16112.6 Conclusions 162

13 Masthead Amplifiers 177

13.1 Introduction 17713.2 MHA example 178

14 Diversity Antennas 181

14.1 Introduction 18114.2 Definition of Fading 18214.3 Receive Diversity 18314.4 Transmit Diversity 18514.5 Multi-User Detection MUD 192

14.6 Predicting the Effect of Different Coverage and CapacityEnhancement Devices 196


5/292

15 Smart Antennas 203

15.1 Introduction 203

16 Practical Simulation 211

16.1 Exercise using MHAs 21116.1.1 Network with no MHAs 21216.1.2 Insert MHA at centre of network 21316.1.3 All sites with MHAs 214

16.2 Downlink Limited case MHAs 21516.2.1 Downlink limited case for network without MHAs 21516.2.2 MHA applied to all sites 216

16.3 Transmit Diversity 21716.3.1 Voice Traffic NO Tx diversity 21716.3.2 64kbps Service NO Tx diversity 21816.3.3 Tx Diversity Voice Service 21916.3.4 Tx Diversity 64kbps Service 22016.3.5 Tx and Rx Diversity Applied Voice Service 221

16.3.6 Tx and Rx Diversity Applied to 64kbps Service 222

17 Measuring Success 223

17.1 Customer Focus 22317.2 Key Quality Indicators KQIs 22417.3 Key Performance Indicators KPIs 225

17.3.1 Exercise 17.3 22617.4 Measurements 226

18 Drive Test Measurements 236

18.1 The concept of Drive Testing 23618.2 Test mobile Measurements 23818.3 Interpretation of Measurements 241

18.3.1 Using Measurements to Validate Improvements 24518.3.2 Comparing Uplink and Downlink Capacity 245

18.4 Using Measured Data 246

19 Cluster Identification 250

19.1 Procedure and Measurements 250

20 Scrambling Code Example 255

20.1 Case Study 255

21 Neighbour Planning 261

21.1 Neighbour Lists 26121.1.1 Initial Neighbour List Generation 26321.1.2 Optimisation of Neighbour lists: 26421.1.3 Inter-freq & Inter-system Neighbour Planning: 266

22 Automation Topics 268

22.1 Modelling 268

22.2 Total Power Targets 271


6/292

23 Future Impact of Standards 273

23.1 Observations of Release 5 and beyond 273

24 From Initial Roll-Out to Mature Network 275

24.1 Introduction 27524.2 Initial Roll-Out 27624.2.1 The Initial Plan 276

24.3 Evolution of the Network 27724.4 Concluding Remarks 288

25 Appendix 291

25.1 Amplificadores MHA 291


7/292

1 Introduction

1.1 Course Overview

The objective of this three day course is to provide delegates withknowledge of optimisation methods and techniques which will enablethem to plan, improve and optimise UMTS 3g networks. Exercises andexamples via software and a state-of-the-art 3g simulator will be providedto aid in the understanding of concepts and theories used in optimisation.

Aims of CourseAims of Course

To deepen the understanding of UMTS networks so as to

plan a network with greater confidence and allow specific

required improvements to be targeted.

To be able to evaluate the benefits that can be obtained

from fitting capacity enhancing devices to the UMTS

infrastructure.

To attain an understanding of the optimisation procedures

available within UMTS.

The function and purpose of optimisation.

To understand how to maximise the benefit of making drive-test measurements.

The use of simulation to aid in optimisation.

Introductory Session

UMTS Advanced Cell Planning and Optimisation 7

AIRCOM International Ltd 2003


8/292

Course ScheduleCourse Schedule

A module may take more than one session

to complete.

During each module, questions and

exercises are provided.

Introductory Session




9/292


10/292

2 Optimisation Overview

2.1 What is Optimisation?Depending upon your position within your organisation this question will meanquite different things. Whilst business is about making money, the engineers goal isusually focused on network efficiency. These two issues are linked but the strategyfor change and time scales can be, and very often are, different.

Business will benefit if the quality of service experienced by customers improves. Theengineer should be focused on obtaining the maximum performance and hence

delivering the optimum customer experience from a given resource.




11/292

What is Optimisation ?What is Optimisation ?

Quality of Service

Applications

Network Coverage/Capacity

Competitive Advantage

Radio Propagation

Measurement Signals

UE UTRAN

Optimisation Overview

Optimising a UMTS network is distinctly different from the optimisationof a GSM network. The fact that we have a single frequency on a cell

layer poses challenges for the network planner. For example, it is nolonger possible to use a frequency plan to help reduce the impact ofpoorly position sites. Further, there is no fixed capacity of a TRX in aUMTS network. The throughput possible depends on the services beingutilised and the radio environment.

The high level of mutual interference between users and cells leads to atrade-off between capacity and coverage. As use of the networkincreases, so does interference. This higher level of interference reducesthe maximum path loss over which a connection can be satisfactorilymade. Optimising for coverage and optimising for capacity will entail adifferent approach, both to planning and to infrastructure investment.

When optimising any network, it is vital that any improvements can beconfirmed by means of measurements made on the network. Feedbackfrom drive-test measurements and OMC reports must be incorporatedinto a continuous cycle of optimisation and monitoring.




12/292

Why is Optimising different for UMTS ?Why is Optimising different for UMTS ?

Single Frequency

Cannot frequency plan around problemscaused by rogue sites.

Flexible structure sensitive to small changesin performance

Air interface performance directly affectscapacity and coverage.


Air Interface affecting Network Performance ?Air Interface affecting Network Performance ?

Suppose we are able to reduce the target Eb/No onthe uplink by 1 dB.

Capacity increased by 25%

Range increased by

(7% if exponent equals 3.5)

Area increased by 14%

So both Capacity and Coverage Increase


exponent1.0

10




13/292

Coverage/Coverage/CapacityCapacity TradeTrade--off ?off ?

We do not have to accept a 25% capacity increase and a 14%coverage area increase. We could make capacity our goal. Inthis case we could increase the Noise Rise limit by 1 dB, andreduce Eb/No by 1dB

The improvement in capacity depends on the pre-existing NoiseRise Limit. ( Eb/No 4.8dB, voice 12.2kbps i=0.6)


10-NR

10-1FactorLoading =

634.4

390.4

158.5

Throughputkbps

841.8

597.8

366

Throughputkbps

33%

53%

131%

TotalCapacityIncrease

8 dB

4 dB

2 dB

New NoiseRise Limit

19.47 dB

32.843 dB

42.721 dB

Coverage

km2

OriginalNoise RiseLimit

CoverageCoverage/Capacity Trade/Capacity Trade--off ?off ?

We could make coverage our goal. In this case we could fix thecapacity, this will reduce the loading factor and reduce Eb/No by1dB

The improvement in coverage depends on the pre-existingloading factor. ( Eb/No 4.8dB, voice 12.2kbps i=0.6)


( )= 1log10NR

634.4

512.4

317.2

Capacity

kbps

19.43

26.75

36.39

Coveragekm2

162%

138.5%

123%

TotalCoverageIncrease

4.31

3.08

1.64

NewNoiseRise dB

31.52

37.05

44.79

Coveragekm2

62.93%

50.83%

31.46%

NewLoadingFactor

6.9980%

4.5665%

2.2240%

NoiseRisedB

ExistingLoadingFactor




14/292

Capacity/Coverage Enhancing DevicesCapacity/Coverage Enhancing Devices

Because of facts such as this, devices to give improvements of a

few dB can be purchased. These include:

Mast Head Amplifiers

Diversity (uplink and downlink)

Multi-User Detection

Smart Antennas

Any one device will usually give improvement in the uplink or thedownlink but not both.

Important to be able to determine which direction is limiting networkperformance.

Improvement in performance must be predicted so that the bestway forward can be implemented.


Measuring and MonitoringMeasuring and Monitoring

Any improvement in quality must be measurable.

Improvement should be:

User experience:

fewer blocked calls;

new services being offered;

greater coverage.

Revenue generation:

greater network capacity;

higher revenue services offered.

Optimisation is part of the overall quality cycle.





15/292

Course StructureCourse Structure

How to Plan a Network - Effectively

Getting the most out of a network with conventional equipment.

Analysing and Comparing devices that will enhance

performance

Selecting the item that will provide the most benefit.

Network Measurements and Optimisation Procedures.





16/292


17/292

3 Network Dimensioning andPlanning

3.1 Introduction

It is necessary to be able to apply all the understanding of the technologyand capacity, dimensioning and link budget calculations in a practicalsituation. Accordingly, it is imagined that a network is to be plannedproviding a certain capacity over a certain area. Initially, certainparameters will be over-simplified when compared with what can beexpected to be encountered in practice. For example, the first assumptionis that the terrain is flat, the traffic distribution is uniform and that thenetwork will be offering only a single service. After dimensioning andexamining the predicted performance of such a network, the effects ofproblems such as high sites and being unable to position base stationsexactly where required will be demonstrated. After that, more realisticterrain data is introduced together with the need to be able toaccommodate varying traffic density.




18/292

Planning a UMTS NetworkPlanning a UMTS Network

We will assume that a coverage area is defined. We have mapping data.

We have a traffic forecast (in this case asingle voice service with uniform distribution.)

Planning a UMTS Network

The PhilosophyThe Philosophy

A strategy needs to be defined.

For this environment, continuous coverage for voice services coulddefine the high level approach.

Other issues: Path Loss; Cell Range





19/292

Link BudgetLink Budget

Crucial to the planning process.

Derived assuming a particularNoise Rise.

Combined with Path Loss modelto determine cell range.

Voice Service

Eb/No 5 dBPower Control 2 dB

Shadow Fade 4 dB Noise

Rise 3 dB

Antenna Gain 18 dBi

Proc Gain 25 dB

Mobile Tx Pwr 21 dBm

Cell Noise Floor -100 dBm

Max Path Loss 150 dB

Range 2.35 km


Iterative Spreadsheet DimensioningIterative Spreadsheet Dimensioning

Carry out link budget todetermine range (remember

link budget assumes a NR)

Assess loading of cell andpredict Noise Rise. This will

differ from assumed Noise

Rise.

Re-calculate range usingpredicted Noise Rise.

Re-assess the loading of thecell and re-predict the Noise

Rise.

Keep Calculating Range andre-assessing Noise Rise.

Finally, the iterations shouldconverge so that the assumed

and predicted values of Noise

Rise agree.





20/292

Graphical ExplanationGraphical Explanation

Increasing Range causes more traffic to be gathered. Gathering More traffic increases Noise Rise and reduces Range.

Range/PathLoss

Number of active users

Intersection gives the

operating point


A complicationA complication

Range/PathLoss

Number of active users

Intersection gives the

operating point

Range calculated from averagenumber of users.

Noise Rise predicted fromestimated peak use of cell.

Additionally, soft capacity must beconsidered.





21/292

Spreadsheet MethodSpreadsheet Method

All relevant parameters (Eb/No, Tx Power etc.) known. From traffic forecast and coverage area, calculate density.

Make initial estimate of the number of trunks required per cell.

Estimate Noise Rise and hence Cell Range 1

Using Erlang B and considering soft capacity estimate Erlangs served.

Estimate area and hence Cell Range 2

Adjust number of trunks until Range 1 = Range 2



Spreadsheet MethodSpreadsheet Method

All relevant parameters (Eb/No, Tx Power etc.) known.

From traffic forecast and coverage area, calculate density.

Estimate Number of

Simultaneous

Connections per Cell

Estimate Noise Rise

Estimate Maximum

Path Loss (Uplink)

Estimate Number of

Erlangs Served per Cell

From Traffic Density

forecast, estimate

cell range

Estimate Maximum

Path Loss (using

Propagation model).

Path Losses Equal?

No

The method outlined above was used to dimension a network given thefollowing input parameters:

Voice Service

Data Rate: 12200 bps

Eb/No 5 dB

Power Control Margin 2 dB

Antenna Gains 18 dBi




22/292

other to own interference ratio 0.6

Shadow Fade Margin 4 dB

Coverage Area 1000 km2

Traffic to be Served 4000 ErlangsMobile Transmit Power 21 dBm

Cell Noise Floor -102 dBm

Path Loss Model: Loss = 137 + 35log(R) dB

The result is that 82 sites would be required. The Noise Rise limit shouldbe set to 3.9 dB in order to maintain continuous coverage.

Example OutputExample Output

For voice service over an area of 1000 km2 offering 4000 Erlangs of Traffic:

82 sites with 246 cells were required.

Noise Rise Limit of 3.9 dB was required to maintaincontinuous coverage.


It is possible at this stage to place sites on a map such that continuouscoverage can be maintained. However, it is highly likely that the actuallocation of sites will not be as required. Further, assumptions made when

creating the spreadsheet may not be accurate in practice. For thesereasons, and for other including those listed below, it is necessary toutilise a planning tool that will consider practical variations from theinitial broad assumptions made.




23/292

The need for a toolThe need for a tool

If this can be done using a simple calculator, why do we need a planning tool?

Planning tool can validate the strategy.

We need to be able to simulate the effect of imperfections. Sites not placed perfectly

terrain/environment factors

Uneven traffic distribution

Some parameters (for example interference ratio, i) have been assumed. Mixed services will have different coverage areas.


Using the 3G Planning ToolUsing the 3G Planning Tool

The coverage area was filled with the correct number of sites and trafficwas spread across the region.

Coverage was checked to be in accordance with requirements.





24/292

Summary of Initial ResultsSummary of Initial Results

Parameters: Eb/No = 7 dB (Incorporating Eb/No and Power Control) S.D. = 7 dB 4000 Terminals NR limit 3.9 dB

Results: Coverage Probability 98.0% Almost all failures due to Noise Rise


Action takenAction taken

3.9 dB NR limit provides continuous coverage even when all cells aresimultaneously at their maximum load.

In reality not all cells would besimultaneously at their maximumloading. The neighbour can oftenassist an overloaded cell.

Noise Rise limit can be raised. Noise Rise was raised to 5 dB.





25/292

Summary of ResultsSummary of Results

Parameters: Eb/No = 7 dB (Incorporating Eb/No and Power Control) S.D. = 7 dB 4000 Terminals NR limit 5.0 dB

Results: Coverage Probability 99.7% (c.f. 98.0%) Even split of failures between NR and UL Eb/No


Next StepNext Step

As Noise Rise limit was raised without any apparent gaps in coverageappearing, it should be possible to raise the amount of traffic served.

Traffic spread raised to 4600 terminals.

Results: Coverage Probability 98.7% (c.f. 99.7%) 83% NR and 17% UL Eb/No.


3.2 Simulating the Effect of Imperfect SiteLocation and High Sites




26/292

Simulating the Effect of ProblemsSimulating the Effect of Problems

Imperfect location of sites.

50% of sites moved randomly byup to 1 km from ideal position.

Gaps appear in coverage.


Summary of ResultsSummary of Results Parameters:

Eb/No = 7 dB (Incorporating Eb/No and Power Control) S.D. = 7 dB 4600 Terminals NR limit 5.0 dB

Results:

Coverage Probability 97.5% (c.f. 98.7%)

78% NR and 22% UL Eb/No

Uneven distribution of failures

Results: Problem area gives 95%

coverage probability (c.f. 97.5% forwhole area).





27/292

Action takenAction taken Antennas were re-pointed in an attempt to restore coverage. Improvement was marginal (96.0% c.f. 95.8%)

Problem is uneven distribution of loaddue to improper placement of sites.Those sites with largest area sufferedNoise Rise failures.

NR failure occurs if more than approx.29 terminals attempt to access a cell.

Average is 19 terminals.


Problems caused by High SitesProblems caused by High Sites

15% of sites made high sites with apath loss 10 dB less than that ofnormal sites at a given range.





28/292


Uneven loading causesdisastrous results.

Coverage probabilityreduced from 98.7% to78.6%.



Probability of NR failurevery high in high site area.

FRE for high site ~ 48%(63% average)

Throughput for high site ~26 E (18 E average)





29/292

Action takenAction taken

Excess coverage area reduced by down-tilting the antennas of the

high-sites.

Result:

Coverage probability increased to 95.1% (c.f. 78% beforedown-tilting and 98.7% with perfect sites).


Alternative ActionAlternative Action

Instead of down-tilting, reduce pilot power of high sites by 10 dB to equalise service areas.

Result:

Problem made worse! This is because terminals still caused Noise Rise even though theywere not connected. Reduction of High Site service area causes an increase in Mobile Tx

power hence aggravating the problem.

Pilot Power Equal

Mobile Connects to High Site

Pilot Power scaled to equalise service areas.

Mobile Connects to Low Site - Tx Power increased





30/292

Alternative ActionAlternative Action

Increased NR Limit of High Site by 10 dB

Decreased Max Tx power, Common Chan power and Pilot power by10 dB.

Result:

A dramatic improvement. Performance of networkindistinguishable from ideal case.

High NR experienced by High Site but continued to performsatisfactorily.

Detecting the existence of High Sites is crucial.


Spotting a High SiteSpotting a High Site

Examining the Best Server by

Pilot array is informative. Spreading a traffic terminal and

examining traffic captured ispossibly more informative as itconsiders traffic distribution.

Site35C: 18.0946 Site36A: 18.2301 Site36B: 19.5065 Site36C: 18.4447 Site37A: 13.9719

Site37B: 14.4915 Site37C: 18.2414 Site38A: 37.0476 Site38B: 38.7644 Site38C: 36.72 Site39A: 10.6173 Site39B: 18.9417 Site39C: 10.1203

High Site





31/292

High SitesHigh Sites -- a final worda final word

There is no single definition of a high site.

Do not think that it is wrong to place UMTS base stations onhilltops.

High sites tend to gather uplink interference generated by otherusers.

Problems occur as area becomes more heavily loaded (if the trafficis reduced from 4000 terminals to 2000 terminals, coverage isexcellent even with untreated high sites).

If coverage area is very lightly loaded - no problem.





32/292

3.3 Provisioning for Asymmetric Traffic

It is common to find that the downlink is not being required to transmit atfull power. In fact there is often about 10 dB extra power capacity onaverage in the downlink direction. This can be utilised to serviceasymmetric (downlink only) traffic requirements. It is possible toestimate the amount of traffic possible by attempting to establishapproximate values for Noise Rise before at the current average basestation transmit power (as obtained from the cell reports) and at themaximum transmit power. Then the extra loading possible can bedetermined.

Because no two mobile stations are likely to experience exactly the sameNoise Rise, the approximate values of traffic calculated should bevalidated by using a planning tool with a UMTS simulator.

In the case being studied it was noted that the Uplink was approximatelyhad a 60% loading factor on average. Because of the effect oforthogonality, it is expected that the loading on the downlink for the sameamount of traffic would be approximately 40%. Thus the mobile stationscould expect to experience a Noise Rise of 2.2 dB on average. It is notedthat the average base station transmit power was 34 dBm. The maximumpower available is 42 dBm. We need to be able to establish the Noise Risethat would be caused if the transmit power rose to 42 dBm given that atransmit power of 34 dBm causes a Noise Rise of 2.2 dB. The necessaryequations are:

Noise Rise increase on downlink on increasing Node B transmit power fromBdBm to C

dBm

New Noise Rise ( )10/

10 101log10 AC

+= where

A

=

110

10log10

10

10

10 X

B

whereXis the noise rise in dB with transmit

powerBdBm.

The above equations suggest that the new Noise Rise will be 7.1 dB (aloading factor of 80%). Thus the loading factor on the downlink can beexpected to increase from 40% to 80% if the transmit power is increased




33/292

from 34 dBm to 42 dBm. This represents an increase in downlink trafficby a factor of 2.

This prediction was verified by simulating an additional load on the

downlink only equal to the original load. The simulator reported nosignificant effect on the existing traffic due to the extra load.

Further Work: Adding traffic onto theFurther Work: Adding traffic onto thedownlinkdownlink

Examining the Simulation Reportsreveals that the average Node BTx Power is approximately 34

dBm.

The maximum Tx power is 42dBm.

This extra power can be used tosend uni-directional data.


Further Work: Adding traffic onto theFurther Work: Adding traffic onto thedownlinkdownlink

Amount of extra data possible depends on the effect that increasing thetransmit power will have on Noise Rise at the mobile.

NR at 34 dBmNR at 42 dBm

Increase in throughput





34/292


35/292

Campbells Theorem Example(1)Campbells Theorem Example(1)

Consider 2 services sharing the same resource: Service 1: uses 1 trunk per connection. 12 Erlangs of traffic.

Service 2, uses 3 trunks per connection. 6 Erlangs of traffic.

In this case the mean is:

The variance is:

=+=== 3063121Erlangs iiii aab

=+=== 6636112Erlangs 2222 iiii aab



Capacity Factor c is:

Offered Traffic for filtered distribution:

Required Capacity for filtered distribution at 2% GoS is 21

2.230

66===

c

63.132.2

30TrafficOffered ===c





36/292


Required Capacity is different depending upon target service for GoS (inservice 1 Erlangs):

Target is Service 1 C1=(2.2 x 21) + 1 = 47

Target is Service 2, C2=(2.2 x 21) + 3 = 49

Different services will require a different capacity for the same GoS. In otherwords: for a given capacity, the different services will experience a slightly

different GoS.


Calculating the Relative AmplitudeCalculating the Relative Amplitude

What is the resource? Bitrate - no

Loading of individual user - yes

Calculate traffic analysis using the ratio of single channel loading for differentservices

Loading is affected by bitrate andEb/N0

1amplitudefor1amplitudeforratebit

serviceforserviceforratebitamplitudeRelative

0

0

NE

NE

b

b

=


3.4 Using More Appropriate Path Loss Models

The path loss model used so far is too simple to be realistic. More widelyused models reduce to similar equations if the height of the mobile isfixed and, also, the terrain is flat. However, incorporation of the moresophisticated models is essential if terrain height variations are to beconsidered.




37/292

A typical Okumura-Hata style of equation was used to predict the pathloss over a terrain that included substantial variations in height. Thevariation in height caused coverage gaps to appear in the shadows of thehills. These were filled by the provisioning of additional base stationssuch that almost 95% of the areas covered to the required level of 146 dBpath loss. It was found that some of the base stations fell into the categoryof high site and caused excessive blocking. The level of blocking couldbe reduced by careful re-pointing of the antennas.

Incorporating more sophisticated Path LossIncorporating more sophisticated Path LossModelsModels

( ) ( ) )log()log()log()log()log()log()log()log(logLoss

625431

654321

dhkkhkhkhkk

dhkhkhkhk(d)kk

effeffmsms

effeffmsms

+++++=

+++++=

Cost 231 - Hata

If hms is fixed then variations are only dependent on heff. Usingtypical default parameters:

Antenna Ht Model

15 140.0 + 32.3 log(d)

20 138.2 + 31.5 log(d)

25 136.9 + 30.8 log(d)

30 135.8 + 30.3 log(d)


A More Challenging TerrainA More Challenging Terrain

154 km2. Heights vary from zero to 135 m a.s.l.





38/292

The ChallengeThe Challenge

Challenge is to serve 2000 Erlangs of demand for voice service.

Even spread of traffic across the whole area.

13 E/km2

With 20 m antenna heights, initial calculation suggests 25 sites.

Max path loss should be 146 dB, range 1.8 km.

Peak Noise Rise will be 8.7 dB.


Placing the SitesPlacing the Sites

Due to irregular outline, 31 sites were required to provide

continuous coverage at a range of 1800 metres.





39/292

Coverage AnalysisCoverage Analysis

Initial site placing leads to 80% ofarea being covered to requiredlevel.

UMTS simulation suggestscoverage probability of 87% withfailures split between uplinkEb/No and Noise Rise.


Increasing Percentage CoverageIncreasing Percentage Coverage

Adding four more sites (35 intotal) resulted in 94.3% coveragebased on pathloss and 92%

coverage probability from UMTSsimulator.

Again failures split betweenEb/No and Noise Rise.





40/292

Analysing Reason forAnalysing Reason for EbEb/No Failures/No Failures

Eb/No failures follow high path loss areas. If the path loss is too great therequired Eb/No cannot be achieved.

Coverage Eb/No Failures


Analysing Reason for NR FailuresAnalysing Reason for NR Failures

Noise Rise failures concentrated on High Sites. An example is shown.

Coverage Strongest Pilot





41/292

Action taken to decrease NR failures.Action taken to decrease NR failures.

Starting statistics: Throughput 382 kbps(approx 31 connections); 20 blockedconnections due to NR.

Action: Height reduced to 10 m; antennadown-tilted by 3 degrees.

Result: Throughput 294 kbps; 0.65blocked connections due to NR; nonoticeable increase in failures onneighbouring cells.

Coverage

For the cell being

investigated:


Covering an Urban Area.Covering an Urban Area.

2000 Erlangs over 154 km2 is not a verybig density.

New challenge is to serve 2000 Erlangs

of voice service generated by users withinan area of 2.36 km2.

This Urban area is not flat (zero to 50 ma.s.l.) or regularly shaped, posingsignificant challenges.





42/292

3.5 Serving Very High Traffic Densities

In practice, it is possible to encounter traffic densities far in excess of the13 Erlangs per km2examined in the last simulation. Accordingly, a small(2.4 km2) urban area was investigated with a view to servicing 2000Erlangs of voice traffic: a density of approximately 800 Erlangs per km2.

The main finding was that the other to own interference ratio tends tobe much higher when the cells are packed closely together. Rather thanthe assumed value of 0.6, values of 1.5 were encountered. This reducesthe capacity per cell. Lowering the antenna heights and down-tiltinghelped improve the situation but not to the extent where the assumedvalue of 0.6 was realised. Thus it seemed impossible in the first instanceto service the level of traffic with the number of cells first calculated. Thenetwork provided good coverage for 1600 terminals as opposed to therequired 2000 terminals. Increasing this level to 2000 would entail re-starting the dimensioning exercise assuming a more realistic value for theinterference ratio (unity being a suggested value for such situations).

This is another example of a simulation tool being required to validatespreadsheet calculations.

Spreadsheet Dimensioning.Spreadsheet Dimensioning.

Initial dimensioning exercise predicts thatcoverage can be achieved by 22 siteseach of range 240 metres.

Low path loss means that very high (20dB+) Noise Rise can be tolerated.

Cell capacity effectively become PoleCapacity.

Coverage prediction suggests that pathloss will not be a problem.





43/292

UMTS Simulation.UMTS Simulation.

Only 65% Coverage Probability achieved.

All failures due to Noise Rise. Estimation of Pole Capacity of a cell is

erroneous.

Cell Reports indicate very low FRE(~40%) suggesting a value for theinterference ratio, i, of 1.5 (c.f. 0.6assumed).

Increasing FRE is crucial to increasingcapacity.Coverage Probability


Optimisation Procedures.Optimisation Procedures.

Lowering antenna heights and making thedowntilt as high as 10 degrees improvedmatters.

Coverage probability now 86% (c.f. 65%). FRE still only 50%.

Initial estimate of 32 Erlangs per cellunachievable in first instance.

Reduce traffic to more realistic levels.

Coverage Probability





44/292


45/292

3.6 Evaluating Simulator Results

When examining the prediction made by a simulator it is important to beclear regarding exactly what you are simulating. Essentially, a MonteCarlo style of static simulator will provide a prediction of the outcome ofattempts to establish a connection to the network. Noise Rise failuresgenerally indicate a failure to connect because of over demand. It is veryuseful to gain an estimate of the likelihood of a call being dropped once aconnection has been established.

If the network becomes under stress from overloading, or capacitybeing reduced due to external interference, there are various load controlmeasures that can be introduced. These include tolerating a lower Eb/Novalue and also reducing the bit rate provided on a particular service.Simulations of network performance with these lower quality targetsshould be made and evaluated.

In these circumstances the lower values of Eb/No and bit rate shouldresult in Noise Rise failures being eradicated. The location of areas where

the likelihood of failure is high should then be identified. These willgenerally be areas where the path loss to the best server is too high toallow the required Ec/Io and Eb/No conditions to be met. Theirseriousness can be evaluated and remedial action taken.




46/292

Evaluating Simulation ResultsEvaluating Simulation Results

The simulator provides aprediction of the outcome of

attempts to establish aconnection to a network.

Of special interest is theprobability of a call beingdropped.

Load control in times of stress will involve reducing Eb/No and reducingbit rates. The performance of the network under such circumstancesshould be evaluated.


Evaluating Simulation ResultsEvaluating Simulation Results

With reduced Eb/No and bit rates(e.g. Eb/No 2 dB below target andvoice bit rate reduced to 7.95

kbps), Noise Rise failures shouldbe extremely rare (ideally zero).

Eb/No and Ec/Io failures willprobably be confined to smallproblem areas which will usuallybe related to high path loss.

Location of Failures


3.7 Pilot Pollution

The term pilot pollution is used in various texts to describe a number ofrelated yet distinct problems. Essentially, they all relate to the situationwhere a similar path loss exists from a mobile to many (four or more)cells. It is possible under such circumstances for the total received powerto be so high that Ec/Io failures are recorded due to the high level of Io.




47/292

Pilot PollutionPilot Pollution

If a mobile experiences comparable path loss to a number of cells,problems can arise through no single cell dominating.

Problems include: low Ec/Io; low capacity on downlink; frequentupdates to membership of the active set.


The value of Ec/Io at a point depends on the pilot power of the best

server, Pp, the other power transmitted by the best serving cell, T1 (thatwill benefit from orthogonality ), the link loss to best serving cell, LL1,the transmit powers of interfering cells (T1, T2, T3 etc..) and the link lossto these interfering cells (LL2, LL3, LL4 etc.).

( )( )

++++

=

....3

3

2

2

1

111log10

0

LL

T

LL

TP

LL

PTLL

P

I

E

NP

Pc

dB




48/292

EcEc/Io/Io

In the above situation the pilot power would be received at a level of -97 dBm.

Total of interference plus noise would be -89.5 dBm giving a value forEc/Io of -7.5 dB.

Pilot Power: 33 dBm

Interference Power: 40 dBm

Link Loss 130 dB

Noise Floor: -99dBm


The power from a neighbouring site would add to the totalinterference and noise power. In the above situation this total powerwould become -86.2 dBm and Ec/Io would be reduced to -10.8 dB

Pilot Power: 33 dBm


Link Loss 130 dB


Link Loss 131 dB

EcEc/Io/Io





49/292


50/292

The Pole Capacity in the Downlink Direction is approximately

kbps.

Downlink CapacityDownlink Capacity

( )iN

Eb +1

3840

0


In the situation above it is possible for the value of i to be as high as

4, thus reducing the downlink capacity. Network capacity may

become downlink limited.

Downlink CapacityDownlink Capacity





51/292

Simulation Examples

A small network of 7 omni-directional sites was loaded with traffic.

For a mean of 200 voice terminals 100% success rate was achieved.

Simulation ExamplesSimulation Examples


Removing the central site caused pilot pollution in the central arearesulting in Ec/Io failure (coverage probability now 78%).


Ec/Io failures





52/292

Increasing the pilot power from33 dBm to 38 dBm resulted in

Ec/Io failures being eradicated

but downlink Eb/No failures now

occur in the same region.


Downlink Eb/No failures


With a SHO margin of 5 dB, the mean size of the active set wasdetermined for the 6-cell and 7-cell configurations.

Identifying Problem Areas.Identifying Problem Areas.

Active Set Size: 7 cells Active Set Size: 6 cells





53/292

The maximum active set size was 3 with the 7-cell configuration but as high as6 in the 6-cell case.

This indicates the number of cells with a low loss path to a mobile.

Active Set Size: 7 cells Active Set Size: 6 cells

Identifying Problem Areas.Identifying Problem Areas.





54/292


55/292


56/292

connection to be made is equal to the required probability. However, in UMTSsystems the phenomenon of Noise Rise will affect the link budget. It is normal to adda Noise Rise (or interference) margin into the link budget. It is also usual to set thisto the limit for a particular cell. This means that the calculations made will result in apath loss being output that will give a 90% connection (uplink Eb/No) probabilityeven if the cell is fully loaded. At lower loading levels the probability will be greater.Thus, if an average probability is required, a lower value of Noise Rise should beused. This value of Noise Rise could be equal to that produced under averagerather than peak loading conditions. The difference that this will produce willagain depend on the desired capacity of the cell. The table shows the differencebetween peak and average Noise Rise and, further, provides an estimate in thedifference this would make in the estimate of coverage area.

Peak Noise Rise Average Noise Rise % coverage area difference

2 dB 1.3 dB 9%

5 dB 3.4 dB 19%

10 dB 5.8 dB 43%

Assumptions: NR caused by voice traffic on cell with pole capacity of 65 connections.Cell provisioned for average traffic on a 2% blocking probability. Propagationexponent assumed to be 3.5.

The Uniform NetworkThe Uniform Network

An ideal situation.

Issues

Capacity/Coverage

Trade-off

Ability to place sites

where they are

required.

Site Location Issues




57/292

Capacity/Coverage TradeCapacity/Coverage Trade--OffOff

Uplink Pole Capacity in kbps

Typical values give a pole capacity(voice 12200 bps, Eb/No 4.8 dB, i= 0.6)as approximately 65 voice connections(100% activity).

Taking the coverage possible for 5 voiceconnections as a reference it is possibleto compare coverage for other

capacities.

Try the calculator, record area for 5users, then increase users and note areareduction.

( ))1

3840

0

iN

Eb +


100%

50

20

5

Number ofConnections

85%

45%

Dimensioning Procedures for UMTSDimensioning Procedures for UMTS

General Rule is to add in a slowfading margin.

In UMTS a NR margin mustalso be included.

Result will be that coverageprobability is achieved at NR

limit.

At lower values of NR,coverage is better thancalculated.

Perhaps coverage probabilityshould be calculated at averagecell loading levels.


x0 -

x0 -

P(connect)

P(connect)

5.6

0

50%

76% 90%

75%

Point Location Probability the probability of beingable to connect at thispoint

Area Location Probability the probability ofconnecting in the area ofinterest




58/292

Dimensioning Procedures for UMTSDimensioning Procedures for UMTS

Ratio of Peak to Average Noise Rise is not constant.

If voice traffic (pole capacity 65 connections as before)is considered with an Erlang B distribution, differencesin coverage area predictions can be calculated fordifferent values of NR limit.


22.72

31.37

40.99

ResultingArea

5.8 dB

3.4 dB

1.3 dB

AverageNoise RiseusingErlang B

74%

24%

9%

AreaIncrease

13.0710 dB

25.245 dB

37.472 dB

ResultingArea

Peak NoiseRise

Breathing

4.1.1 Mis-placed sites.

An evenly loaded network is fairly straightforward to provision with evenly spacedsites. A problem arises if you are not able to place the sites where you would likethem to be. In this situation a particular cell may have a coverage area larger than theaverage. Consider the example where a cell must increase its range from 1300 metresto 2000 metres (a 50% increase). This increase would raise the path loss at the celledge by approximately 6 dB. This would reduce the probability of connection at thecell edge from an initial value of, typically, 75% to less than 50% when the noise riseexperienced on the cell is close to the limit. Action that can be taken would be toreduce the Noise Rise limit. Perversely, this reduces the capacity of the cell when its

coverage area has been increased but it would increase the probability of connectionat the cell edge. Additionally, it would be possible to offer reduced bitrate services atthe cell edge. With voice services, that would lead to an increase in the maximumpath loss tolerated of up to 4 dB.

Another possibility to consider is the use of an active repeater station.




59/292

MisMis--placed Sitesplaced Sites

If sites cannot be placed

exactly as required, thecoverage area of some siteswill increase.

50% range increase leads topath loss increase of ~ 6 dB.

( 35 * log ( 1.5 ) = 6.2 dB )

Location probability willreduce, possibly to less than50%.


Location Probability ~75%

Location Probability ~ 50%

MisMis--placed Sites: possible actionplaced Sites: possible action Coverage helped by reducing NR limit.

This will reduce capacity of the cell.

Reduction of bit rate of services offered

12200 bps reduced to 4750 bps gives a 4 dB benefit.

Deploy a repeater station.


Repeaters can improve coverage to remote parts of the cell.




60/292

4.2 Hot Spots

The fact that a single frequency is used to provide continuous coverage ina UMTS system means that cells act together much more intimately thanin a GSM system for example. One object of system optimisation can bethought of as the minimisation of transmit powers. That will lead in turnto a minimisation of interference and therefore a maximisation ofcapacity. If hotspots exist, the impact on the network is very dependenton their location within the network.

As an example let us suppose that a hotspot exists that demands serviceson the uplink that will generate a 3 dB noise rise on that uplink. If thebackground noise level is -102 dBm, the power from the users will alsoequal -102 dBm. However, the transmit power of the users, and hence the

amount of inter-cell interference generated will be dependent on thelocation within the cell. We will consider a simplified situation where thehotspot is only 100 metres from the serving cell and 2000 metres from thenearest neighbour. The difference in path loss has been measured to be 45dB. That means the neighbouring cell will experience an interferencelevel of -147 dBm. This would cause a noise rise of only 0.0001 dB(equivalent to a loading factor of 0.003% - negligible). However, if thehotspot was in a location where the path loss to the neighbouring cell wasonly 4 dB more than that to the serving cell, the power level at theneighbouring cell would be -106 dBm and the noise rise generated would

be 1.5 dB which is equivalent to a loading factor of approximately 30% -definitely not negligible.

Enhancement techniques such as uplink diversity and mast headamplifiers will be examined in detail later in this course. Suffice to say atthe moment that they improve coverage and will allow the noise rise limitto be increased whilst maintaining the cell range. It is understandabletempting to employ such techniques on a cell that has a hotspot.However, this can lead to worse consequences for neighbouring cells.

Suppose the hotspot was such that it generated a noise rise of 8 dB at theserving cell and path loss to the victim neighbour was only 4 dB greaterthan that to the serving cell. The signal level received at the neighbouringcell would be -98 dBm, generating a noise rise of 5.5 dB. This may well beabove the noise rise limit of this cell. That means that adjacent cellinterference alone has effectively fully loaded the neighbouring cellmaking it incapable of accepting any more traffic.

As a final note, it is worth remembering that, in sectored sites, the cell

edges do not necessarily occur at large distances from the cell.




61/292

The location of hot spots within cells and in relation to other cells requiresunderstanding. Techniques for handling these hot spots therefore variesdepending upon the location of own cell base station and neighbouring

base station location.

Traffic HotspotsTraffic Hotspots

Accommodati ng Hot spots

In a UMTS network the

affect of a hotspotdepends on its location in

relation to the network

cells.




62/292

Traffic Hotspots: an analysisTraffic Hotspots: an analysis

Acc ommodati ng Hotspo ts

Thermal Noise Power = -102 dBm Power from Hotspots = -102 dBm NR = 3 dB

PL=100 dBPL=145 dB

Noise Power= -102dBm

Power fromHotspots = -147 dBm

Negligibleinterference


affect of a hotspot

depends on its location in


cells.


Acc ommodati ng Hotspo ts

Thermal Noise Power = -102 dBm Power from Hotspots = -102 dBm NR = 3 dB

PL=121 dB

PL=125 dB

Noise Power= -102 dBm Power from Hotspots =-106 dBm

NR = 1.5 dB (loadingfactor of 30%)


affect of a hotspot

depends on its location in


cells.




63/292


Suppose the servingcell employs, for

example, uplink

diversity and then

allows the NR to

reach 8.6 dB.

Interference is now

very damaging.


Thermal Noise Power = -102 dBm Power from Hotspots = -94 dBm NR = 8.6 dB

PL=121 dBPL=125 dB

Noise Power= -102 dBm Power from Hotspots =-98 dBm

NR = 5.5 dB(loading factor of 72%)

Traffic Hotspots: sectored sitesTraffic Hotspots: sectored sites


A hotspot located near to theedge of a cell has a moreserious effect than one with avery dominant serving cell.

In sectored sites, note that celledges can be located near tothe site.

Soft or softer handover willimprove the situation. A nicely

locatedhotspot

A badlylocatedhotspot




64/292

4.3 Site Density

Many ballpark figures are quoted regarding the capacity of a UMTS cell.This has affected the configuration offered by vendors when supplyingNode Bs. Perhaps a cell is configured such that its hardware willaccommodate 32 simultaneous voice connections. If this is accepted forthe moment, the factor that limits the user density that can be served thenbecomes cell density. The problem is that, as we increase the site densityof a site, cell interference increases. If the pole capacity of a cell is

considered as equal to that given by the expression ( )( )iNEb +13840

0

kbps, the parameter itends to increase with site density. This is due totwo main reasons:

Employing normal levels of down-tilt leads to the main beam

penetrating neighbouring cells.The dual-slope nature of the signal strength vs. distancecharacteristic makes the propagation exponent lower the shorter thedistance.

Antenna downtilt. Suppose that the general rule was to down-tileantennas by 2 degrees. That will lead to the main lobe of the antennabeing directed towards the ground at a distance approximately equal to30 times the antenna height. The strength of the signal in the horizontaldirection will be a few dB less than the main lobe (the exact value will

depend on the vertical beamwidth of the antenna). This helps to reduceinterference between cells. If the cells are packed more closely together,the main lobe of the antenna will penetrate adjacent cells.

Dual slope propagation Characteristic. It has been found that for aparticular environment and site configuration the path loss can be

approximated by the expression dkk log21+ where d is the distance in

kilometres and k1and k2are constants for a particular environment andconfiguration. The parameter k2is related to the propagation exponent

and is crucial in determining the level of inter-cell interference. The lowerthe value of k2the worse the inter-cell interference. In a GSM networkthat delivers a typical C/I value of 10 dB when the value of k 2is 35 (anexponent of 3.5), the value of C/I will reduce to approximately 7 dB if thevalue of k2reduces to 25 (an exponent of 2.5). In a UMTS network, thevalue of the relative interference parameter I, will increase if the exponentreduces. A typical value for I in an evenly loaded network is 0.6 for anexponent of 3.5. For an exponent of 2.5, it can exceed 1.




65/292

The fact is that signal strength is more accurately modelled by a dual-slope graph where k1 and k2 have two different values: near and far.The value of k2NEARwill be less than that of k2FAR. The point of transitionbetween the two values is referred to as the knee of the graph. The

distance at which the knee occurs depends upon the height of the cellantenna (it increases with antenna height). Typically, the knee could beexpected to occur at about 500 metres for an antenna height of 15 metresand 1100 metres for an antenna height of 30 metres. This means that themore densely you attempt to pack the sites, the lower the exponent andthe worse the inter-cell interference.

Action that needs to be taken to minimise the adjacent cell interferenceincludes reducing the cell height (to make the distance to the knee

smaller) and to down-tilt cell antennas, quite severely.

When sites are densely packed, coverage becomes a secondary issuebecause of the small cell boundaries. The capacity of the cell effectivelyequals its pole capacity. Crucial in this pole capacity is the value of inter-cell interference. If the cell range is 100 metres and the cell antenna heightis 15 metres, a down-tilt of 8.5 degrees will point the main lobe at the edgeof the cell. Additional down-tilt will reduce the signal strength at the celledge but will have the advantage that it will also reduce the amount ofinterference gathered from adjacent cells. Down-tilts of greater than 10degrees can be expected in such circumstances.




66/292

Increasing Site Density: ProblemsIncreasing Site Density: Problems

Increasing Site Density

What is the capacity of a UMTS

Cell?

Ballpark figures suggest thatapproximately 32 simultaneousvoice connections can beaccommodated. ( 100% voiceactivity )

This has influenced thehardware configuration of cells.

But, how densely can sites bepacked?

32

32

32

32

3232

3232

32

3232

32

3232

32

32

Increasing Site Density: ProblemsIncreasing Site Density: Problems


As sites become moredensely packed, inter-cellinterference increases.

Frequency reuse efficiency,FRE, decreases

Two main reasons:

Using normal levels of

down-tilt leads to the mainbeam penetratingneigbouring cells

The dual-slope nature ofpath loss against distance.

32

32

32

32

32 32 32

3232

32

32

i ~ 0.6

i ~ 1.0




67/292

Antenna DownAntenna Down--tilttilt


For a given level ofdown-tilt, the level ofinter-cell interferencewill increase when thecell size reduces.

Large cells: low level of inter-cell interference

Small cells: high level of inter-cell interference

DualDual--Slope CharacteristicSlope Characteristic


Propagation exponents quoted usually refer to the far slope.

The exponent near to the base station is usually considerablyless (2.0 being a typical value).

Remember that up until now we have had 3.5

Pathloss = 137 + 35 x log (R)

distance

Loss (dB)

Near slope

Far slope

Break point




68/292


The lower the value of theexponent, the greater thepenetration from one cellinto the next.

Generally, the lower theexponent, the greater thevalue of i.

For an evenly loadednetwork:

-exponent

26i


0.384

0.533.5

0.753

1.12.5

1.52

iExponent


The position of the breakpoint is influenced by theheight of the base station.

The higher the base station,the greater the distance.

At 2 GHz, the distance tothe breakpoint isapproximately

BTSh30


Distance to

Break point (m)

BS height (m)

10

20

30

300 600 900




69/292


Increasing the exponent reduces interference and

increases network capacity.

Reducing antenna height will lead to an increase inthe exponent. Far value of exponent typically (urbanenvironment)

Also reducing antenna height brings the break pointcloser to the BTS. (break point ~ 30 x hBTS)

Down-tilting antennas will also reduce mutualinterference.


( )BTSh10log66.05.4

Implications for Increasing Site DensityImplications for Increasing Site Density

As site density increases, interference will increase.

Reducing Base Station height and down-tilting antennaswill help to reduce this.

For very small cell ranges, the amount of down-tilting willbe severe.


At a range of 100 metres, a 15 metre mastwill have to be down-tilted by 8.5 degrees topoint at the cell edge.

tan-1(15/100) = 8.5o

Additional down-tilting will reduce fieldstrength at the cell edge but also reducemutual interference.

Vertical beamwidth of antenna is significant.




70/292

Vertical AntennaVertical Antenna BeamwidthBeamwidth

Sectored antennas with a typical horizontal beamwidth of85 degrees can have a variety of vertical beamwidths.

A high gain (e.g. 18 dBi) antenna will have a verticalbeamwidth of only about 3 degrees.

A lower gain (e.g. 13 dBi) antenna will have a verticalbeamwidth of about 12 degrees.

Lower gain antennas are likely to cause fewer problemswhen severely down-tilted.


18 dBi antenna 13 dBi antenna

4.4 High Sites

A high site is a site that has a lower path loss at a particular distance thanis normal for cells in the network. If GSM legacy is used forestablishing a UMTS network any GSM umbrella site is likely to act as ahigh site in a UMTS network. Remember that it is not possible tofrequency plan your way out of trouble in a UMTS network. You cannotisolate a rogue site by allocating a separate carrier to it. The fact that thepath loss is lower to a high site means that it will tend to dominate anetwork unless action is taken. When a mobile selects a cell it measures

the pilot channel power. The high site will have a larger area of coveragethan its neighbours if they are all transmitting the same pilot channelpower. That will lead to the cell becoming overloaded very quickly withits Noise Rise limit being reached. Further, the neighbouring cells will beaffected by downlink interference generated by the high site.

Many of the problems associated with a high site can be reduced bycareful parameter planning. Sometimes, however, the results can be alittle strange. It is tempting to reduce the pilot power of the high site in

order to reduce its traffic. However, this would lead to mobilesincreasing their transmit power as they connect to cells with a higher path




71/292

loss. In this way, the high site in particular would experience a highernoise rise than if it actually served the traffic. One aspect that must beconsidered is, because the path loss is low, the noise rise limit can beraised whilst still maintaining coverage. If this is done, the pilot powercan be reduced without severe consequence. The same argument can beused to allow the reduction in downlink power (total cell power andcontrol channel power). This will reduce the downlink interferenceexperienced by neighbouring cells.

The presence of the high site remains undesirable however. It will beoperating very near its pole capacity with the value of inter-cellinterference reducing this pole capacity to modest levels. Additionally,because of its high position, the propagation exponent is likely to be lowwith the result that mutual interference is further enhanced. It is clear

that controlling the radiation from the cells is crucial in getting themaximum performance from a network. The performance of a networkcontaining a high site may be improved by careful down-tilting of theantennas.

Summarising, the following parameters on a high site can be modified toimprove the network performance

Noise Rise Limit Increase

Cell Power DecreasePilot Channel Power Decrease

Common Channel Power Decrease

Antennas Consider Down-tilting

The amount of increase or decrease that should be made is approximatelyequal to the amount by which the path loss at a particular distance islower than normal. However, this is not a constant and someexperimentation will be necessary in order to determine the optimumsolution.




72/292

High SitesHigh Sites

The higher the site:

The lower the loss to a particular point (offset ~ 14 log10[hBTS])

Remember that it is the effective height of the BTS that is

considered.

The lower the propagation exponent (offset ~ 0.66 log10[hBTS])

Result:

High Sites suffer interference on the uplink and generate

interference on the downlink.

High Sites


An umbrella site from a legacy GSM network will generally act as a

high site.

You cannot frequency plan your way out of trouble in a UMTS

network.

High Sites




73/292


Lower path loss can be rectified by either

Inserting extra loss

Parameter planning: pilot power, noise rise limit, downlink Tx

power.

The lower exponent causes greater problems; careful control

of radiation (e.g. by down-tilting) can help with mutual

interference.

An untreated high site will cause severe problems with

localised capacity reductions of 50% to be expected.

Note that high sites cause capacity rather than coverage

problems.

High Sites




74/292

5 Factors Limiting Capacity

5.1 Cell Throughput

Factors Limiting CapacityFactors Limiting Capacity

Cell Throughput is given by the simplified expressions for polecapacity in kbps multiplied by the loading factor

Crucial parameters are Eb/No, inter-cell interference i,

orthogonality and loading factor (which is affected by the

Noise Rise limit).

Capacity Limi ting Factors

( )

( )

+

+

iN

E

iN

E

b

b

1

3840

1

3840

0

0

Uplink

Downlink




75/292


76/292


77/292

Factors Limiting Capacity: EFactors Limiting Capacity: Ebb/N/N00

If the mobile cannot respond to power control commands,the UE will notice a variation in the received signal.

This will lead to BER variations that will cause the networkto require a higher target Eb/No (a fast fading margin orpower control margin will be required).

The effect can be to increase the target Eb/No from anormal value of perhaps 4 dB to 10 dB or more for fastmoving mobiles.

This will reduce the capacity of a cell from typically 32simultaneous connections to only 8 a dramatic reduction.

Lesson: the multipath environment and user mobility canaffect the target Eb/No and hence cell capacity.

Capacity Limiting Factors

Factors Limiting Capacity: FREFactors Limiting Capacity: FRE

Frequency re-use efficiency is the name given to theproportion of received power that comes from a cells ownusers rather than from all users including other cells.


11

1

1

cellintra

cellinter1

1

cellintercellintra

cellintra

=

+=

+=

+=

FREi

iFRE




78/292

5.3 Maximising Frequency Re-use Efficiency

Frequency Re-use Efficiency is a key indicator of how well the availablespectrum is being used. It shows the percentage of power that is receivedat the cell (excluding thermal noise) that comes from users of that cell.Any cell has a limit to the amount of Noise Rise that will be toleratedbefore new admission attempts are refused. Out of cell power will causenoise rise just as much as in-cell power. Interference from outside the cellwill therefore reduce the capacity of that cell. In fact the Frequency Re-use Efficiency factor (FRE) indicates the fraction of cell resource that isavailable for users of that cell. For an evenly loaded network, a value of65% is thought to be typical.

In a UMTS network, the problem is made worse by the fact that mobilesnear the edge of a cell will be both transmitting at near maximum powerand also in a location where the path loss to the adjacent cell is aminimum.

Unevenly loaded networks yield surprisingly low values of FRE. If areasof high traffic density are near the edge of a cell, they will produce highlevels of interference leading to a reduction in the capacity of

neighbouring cells. Similarly, a cell providing coverage over a largegeographic area that is surrounded by smaller cells serving areas of highuser density will experience high levels of interference with the result thatits capacity is adversely affected.

The most effective methods of increasing FRE are: down-tilting cellantennas where appropriate and planning the location of Node Bs to tryand ensure that they are placed as close as possible to any predictedtraffic hotspots.




79/292


The ideal situation is where the receiving antenna can onlysee its own users but not those of other cells. ie FRE = 1

The power from neighbouring mobiles close to the cellborder cause the biggest problems.


High power mobiles close to

Cell border cause FRE reduction


A large cell serving a low subscriber density surrounded byseveral smaller cells serving high subscriber densities willexperience a low value of FRE.


A Large cell will experience low FRE

Because it is surrounded by

many users of other cells




80/292


Hotspots near the cell border will cause more problems thatevenly distributed neighbouring cells

A quantitative analysis is not always possible. A simulatoris extremely valuable in helping to develop a feel for theseriousness of potential problems.


Hot spots near cell border causeFRE reduction


Increasing FRE: the main weapon is to down-tiltantennas.

This is most effective when there is a large anglebetween the line from the antenna to the cell edgeand the horizontal.

In the case of large cells, planning to avoid hotspotsnear the cell border will reduce the incidence of lowFRE.





81/292

5.4 Downlink Capacity and Orthogonality

The parameter known as orthogonality describes the amount ofmutual interference that will be experienced between users of the samecell. The fact that downlink transmissions within a cell aresynchronised means that the OVSF codes used provide interferencerejection that is not possible on the uplink. In fact, an isolated cell withperfect orthogonality will have an infinite pole capacity (although theneed for multiple scrambling codes would compromise this). Theamount of interference rejection decreases in multipath environments.We shall now analyse two situations in which only orthogonality ischanged. The effect on downlink power requirements and downlink

capacity will be examined.

Factors Limiting Capacity: OrthogonalityFactors Limiting Capacity: Orthogonality

Dramatic effect on downlink capacity.


( )i+=

1N

E

3840CapacityPole

0

b




82/292


Example: Eb/No = 4 dB, i = 0.6, 12200bps


2568

1.0

1926

0.8

154112841100963Pole Capacity

0.60.40.20Orthogonality

Pole Capacity(kbps)

1000

2000

Orthogonality0.5 10


The Loading factor deliverable on the downlink depends upon

the link loss, maximum transmit power and noise performance

of the mobile.

Example: Tx Power 43 dBm; Noise Floor of Mobile -100 dBm.

Deliverable loading factor can be expected to exceed 75%.

Pole capacity is crucial.


( ){ }( )

( ){ }

( )

( )

=

+=

+=

+

=

+=

1log10

110

11

110log1010

1010log10NR

dBm1010log10PowerRxMobile

10143

10143

10/100

10/1001043

10/1001043

orth

orthLL

orthLLorthLL

LL




83/292


Question:

Suppose a group of users of a 64kbps service in an isolatedcell experiencing a link loss of 138.4 dB are demanding a total

data throughput of 1.024 Mbps at an Eb/No of 4 dB.

What is the downlink loading factor at this throughput if the

orthogonality is i) 0.4 and ii) 0.8?

Further, what is the traffic channel power demanded and what

is the maximum throughput possible at that path loss if the

maximum traffic channel power is 42.7 dBm?

Assume a noise level at the mobile of -102 dBm before noise

rise.



Answer:

At an orthogonality of 0.4, the pole capacity is 2568 kbps. 1024 kbps represents a loading factor of 39%.

Hence the Noise Rise would be approximately 2.1 dB.

The effective received traffic power would be -104.06 dBm

Actual received traffic power is 2.2 dB higher (-101.86 dBm)

indicating a transmit power of 36.5 dBm (link loss 138.4 dB).

42.7 dBm would be able to deliver almost 72% loading factor

and hence the throughput possible should be approximately

1846 kbps.





84/292


85/292

5.4.1 Pilot SIR as an indicator of downlink capacity

The power of the pilot channel does not vary with time. The value of theratio Pilot to Noise plus Interference (Pilot SIR) indicates the quality ofthe downlink channel at a particular location. For example, if the pilotSIR is 6 dB then the SIR experienced by a traffic channel of the samepower as the pilot would also be -6 dB. If the required Eb/N0for thebearer carried on the traffic channel was, say, +4dB then a processing gainof 10 dB would be needed. This limits the throughput to 384 kbps. Thevalue of the pilot SIR indicates the capacity of the downlink per unitpower (bits per second per milliwatt). The pilot SIR will vary withlocation and with orthogonality and the amount of loading on the

network.

Factors Limiting Capacity: Pilot SIRFactors Limiting Capacity: Pilot SIR

Although a minimum value of pilot SIR is a pre-requisite for any

communication between the UE and the Node B, its level will

indicate the quality of the user environment at any location.

The SIR of the pilot channel indicates the SIR (Ec/Io) of any

channel with the same transmit power.

If the target Eb/No of such a channel was known then the

required processing gain, and hence user bitrate, could be

determined.





86/292

Factors Limiting Capacity: Pilot SIRFactors Limiting Capacity: Pilot SIR

Example, at a particular location the pilot SIR is found to be -12 dB.

A service with a target Eb/No of 4 dB is to be delivered to a user at

this area. If the same power as the pilot is available for the traffic

channel then the processing gain necessary would be 16 dB. This

limits the throughput to 96 kbps.

If the maximum channel power was 3 dB greater than the pilot

power then this

Aircom_UMTS Advanced Cell Planning and Optimisation PS-TR-005-O036_v5.0

Documents