Comparison of Asset and Atoll Cellular Planning tools for LTE Network Planning.pdf

8/9/2019 Comparison of Asset and Atoll Cellular Planning tools for LTE Network Planning.pdf

1/40

Comparison of Asset and AtollCellular Planning tools for LTE

Network Planning

Hormoz Parsian 85014KDepartment of Communication and Networking

Supervisor: Prof. Jyri HmlinenInstructor: Dr. Kimmo Mkelinen

This Masters thesis was conducted at

Nokia Siemens Networks

5/22/2012 1


2/40

Outline of the work:

Introduction

Background Research question

Methodology

Propagation Models Hata Models

Standard Propagation Models in the Tools

Deriving Equivalent Parameters

Coverage Analysis Coverage Analysis with original constant term in Atoll propagation model

Coverage Analysis with optimum constant term in Atoll propagation model

Interference Analysis Analysis of Number of Covering Cells in Asset

Interference Analysis for Outdoor Users

Interference Analysis for Indoor Users

Capacity Analysis Capacity Analysis of LTE services Capacity Analysis for MIMO modes

Capacity Analysis of Scheduling Algorithms

Conclusions and Future Works

5/22/2012 2


3/40

Introduction

Backgrond

Software for network planning:

Astoll of Forsk

Asset of Aircom

Planet of Mentum

Planning tool manufacturers design their tools independently of eachother

Users do not know before testing whether different planning tools

produce comparable performance estimates for a given network

Sometimes a network is planned or investigated in two different planning

tools Companies often change from one planning tool to another

Planning results of the first tool have to be reproduced in the second one

5/22/2012 3


4/40

Introduction

Research Question

Comparing two planning tools, Atoll of Forsk and Asset of Aircom withrespect to LTE network planning

Not trying to find out which of the tools is better than the other one

Trying to investigate whether they provide comparable performance measures in the

test network

Tools performance can be close to each other if two investigated networkconfigurations and other configurations like propagation models are equal

to each other.

Finding equal parameters mean they produce same performance

estimates AND it necessary does not mean numerically equal parameters

LTE input parameters for Asset and Atoll, prepared by NSN to give equalperformance estimates

5/22/2012 4


5/40

Introduction

Methodology To compare the tools different testing scenarios and network

configurations are analyzed.

Propagation Models

Number of Transmitting and receiving Antennas

Services used

MIMO configurations Scheduling methods

Planning tools are compared in terms of three performance estimates:

Coverage

Interference

Traffic capacity

For radio propagation prediction digital map is required.

Digital map of Helsinki with 10 mresolution

5/22/2012 5


6/40

Propagation Models

Hata Model

All computed coverage, interference and capacity results in the cellularnetwork planning tools are based on losses between base stations adpoints on the digital map that computed from propagation predictionmodels

Standard propagation models in Asset and Atoll are based on Okumura-Hata models

Hata model is derived from Okumuras measurement reports The reports are obtained from four different environments in and around Tokyo

The measurements are for limited parameter ranges e.g. frequency, distance and height.

In Hata model frequency, distance, base (BS) and mobile station (MS) antenna heights

are limited.

150 MHz < Frequency < 1500 MHz 1 km < Distance < 20 km

30 m < BS Antenna Height < 200 m

1 m < MS Antenna Height < 10 m

5/22/2012 6


7/40

Propagation Models

Hata Model

Hata model was extended to frequencies higher than 1500 MHz but lessthan 2000 MHz

Also it was extended for distances between 20 km and 10 0km

5/22/2012 7


8/40

Propagation Models

Standard Propagation Models (SPM): SPM in Asset is presented below:

= + +

Where:

= + + +

= +

= +

In Asset distances between BS and MS are in kilometers.

5/22/2012 8


9/40

Propagation Models

Standard Propagation Models (SPM):

SPM in Atoll is presented below:

= + +

Where

= + + +

= +

= + + ,

In Atoll distances between BS and MS are in meters.

Thus coefficients of the terms in models of the tools containing distanceparameter have to be modified.

5/22/2012 9


10/40

Propagation models


SPM in Asset is based on Okumura-Hata model. To achieve equivalent propagation models, corresponding terms in the

tools have to be equated.

The tests are carried out at frequency of 2100 MHz.

Equivalent coefficient in the tools are given in the table:

Asset coefficients Atoll coefficients

k1= 158.92 dB K1= k13k2= 24.22

k2= 44.9 K2= k2= 44.9

k3= 0 K3= k53k6= 5.83

k4= 0 K4= k7= 0.8

Asset coefficients Atoll coefficients

k5= -13.88 K5= k6= -6.55

k6= -6.55 K6= k3= 0

k7= 0.8 K7= k4= 0

-- Kclutter= 1

-- Khill.LOS= 0

5/22/2012 10


11/40

Propagation Models

Deriving Equivalent Parameters Correction Factors in Asset and Atoll

Propagation models in the tools have additional terms as correction

factors to take into account terrain height, clutter losses and diffraction

losses.

Equivalent algorithms for antenna height corrections, diffraction

corrections and clutter corrections are chosen in tools.

5/22/2012 11


12/40

Propagation Models


Verifying the derived propagation models Derived propagation models are tested using point analysis

LTE cell coverage is determined using Reference Signal Received Power

which is reported in form of Energy Per Resource Element of Reference

Signal (RS EPRE) in the tools.

Discrete points at different distances were considered in point analysis

Systematically Asset gives 1 - 2dB higher RS EPRE values than Atoll does

Such differences in models should not cause significant errors in coverage,

interference and capacity analysis.

5/22/2012 12


13/40

Coverage Analysis Coverage Analysis with Original Constant Term in Atoll

Equivalent propagation models and path losses do not guarantee

equivalence of coverage results.

Comparing coverage distribution of predicted RS EPRE intervals

RS EPRE range of -55 to -130 dBmin steps of -5 dBare investigated.

RS EPRE Interval (dBm) % Coverage Area in Asset (k1=

158.92)

% Coverage Area in Atoll (K1=

24.22)

-60.00


14/40

Coverage Analysis Coverage Analysis with Original Constant Term in Atoll

Table in the previous page and the plot in this page indicate that RS EPRE

distributions are not similar. Systematically Asset calculates higher RS EPRE

values than Atoll.

Root Mean Square difference and maximum absolute difference are 19.9%

and 2.38%respectively

5/22/2012 14


15/40

Coverage Analysis

Coverage Analysis with Optimum Constant Term in Atoll

To achieve similar RS EPRE coverage distributions, constant term inpropagation model of Atoll is decremented in steps of 0.1 dB.

Optimal value of constant term results in minimum RMS difference

between RS EPRE distributions of the tools

Optimal value results in similar coverage distributions in the tools.

To achieve optimal value 1.4 dB is to be decremented from original

constant term in model of Atoll (K1= 22.82 dB).

RS EPRE Interval (dBm) % Coverage Area in

Asset (k1= 158.92

dB)

% Coverage Area in

Atoll (K1= 22.82 dB)

-60.00


16/40

Coverage Prediction

Observations from table in the previous page and the plot in this page

reveals that reduction of 1.4 dB from original constant term results in similar

coverage distributions .

It also minimizes RMS difference between the RS EPRE distributions.

Optimum value leads to Maximum absolute difference of 0.329%which is

negligible.

5/22/2012 16


17/40


18/40

Interference Analysis Coverage estimation is based on the assumption that the signal of the

serving station is on during the time that received power is observed.

However, the interference estimation is not based on assumption that the

interference were on 100% of time (and frequency). The interference

estimation is based on a specific loading of each interfering cell.

Cell load can be derived through Monte Carlo simulation or fixed by

network planning engineer.

For interference analysis is based on comparison of SINR arrays for

downlink Reference Signal (DLRS) and downlink Traffic Channel (DL TCH) in

Asset and Atoll.

In interference analysis, fixed load of 75% is assumed for all base stations.

SINR arrays are created and analyzed for (-10 dBto 30 dB) range and thisrange is divided in steps of 5 dB.

5/22/2012 18


19/40

Interference Analysis

Analysis of Number of Covering Cells in Asset

Theoretically, number of covering cells is very critical parameter in MonteCarlo simulation.

But it was found that when number of covering cells was varied from 6 to

12, the differences in the SINR of the received signal were insignificant

(below 0.5 dBfor most of the area).

DLRS SINR

Interval (dB)

6

7

8

9

10

11

12

20

-10


20/40


Analysis of Number of Covering Cells in Asset

5/22/2012 20


21/40


Interference Analysis for Outdoor Users (DLRS) In Asset, reference signal pattern affects SINR calculations and

consequently interference analysis.

Downlink cell load level is set to 75%.

Simulation arrays are created without running any real simulation

snapshots.

Finally DLRS SINR arrays are calculated. According to following table and plot, DLRS SINR distributions in Atoll and

Asset are not close to each other.

Systematically Asset calculates higher DLRS SINR values than Atoll. And

Maximum difference belongs to SINR interval of -5to 0 dBwhere absolute

value of difference is 18.82%.

DLRS SINR

Interval (dB)

Asset DLRS SINR

distribution

Atoll DLRS SINR

distribution

Difference

25


22/40


Interference Analysis for Outdoor Users (DLRS)

Besides reference signal power, the tools takes into account power received

from control and traffic channels for reference signal interference power

calculation.

5/22/2012 22


23/40


Interference Analysis for Outdoor Users (DLRS) For reference signal interference calculations Asset considers interference

power received from a single antenna of the interferer.

Atoll considers interference power received from all transmitting antennasof the interferer.

Thus, the difference between SINR results of Asset and Atoll is mainly

because Atoll considers number of transmitting antennas in referencesignal interference calculation while Asset does not.

For interference analysis of the tools, a conversion factor between theirSINR values has to be calculated.

=

= .

Table and plot next page represent DLRS SINR of the tools when theconversion factor in incorporated.

5/22/2012 23


24/40


Interference Analysis for Outdoor Users (DLRS)

DLRS SINR Interval (dB)

Asset DLRS SINRdistribution

Atoll DLRS SINRdistribution

Difference

25


25/40


Interference Analysis for Outdoor Users (DL TCH)

Same procedure is carried out for analysis of SINR distributions for DL TCH. To compensate the difference in calculation of interference in the tools, a

conversion factor is derived.

=

= .

This plot represents the DL TCH SINR

distributions of the tools after

incorporating the conversion factor inAsset.

5/22/2012 25


26/40


Interference Analysis for Indoor Users (DL TCH) For interference analysis of indoor users, SINR results of Asset and Atoll

are compared for downlink traffic channels (DL TCH).

In case of indoor users, indoor penetration loss is included in downlink

losses in addition to path loss. For DL TCH SINR analysis same conversion

factor,

= . , is included in Asset to achieve similar distributions

in the tools.

5/22/2012 26


27/40

Capacity Analysis

Capacity analysis is the most difficult part of performance estimation in a

radio network planning tool. Coverage and interference analysis can be carried out as straightforward

deterministic calculations.

But capacity can only be analyzed statistically by using Monte Carlo

simulation of connection of terminals to cells of the network.

The Monte Carlo simulation requires as its input realistic assumptions ofthe network traffic.

Results of capacity studies are analyzed by comparing number of served

users and total peak RLC throughput.

Traffic Layers

The investigated area has uniform density of 15 users/km2. To have more

realistic image of the network, further three vectors within this area are

introduced with each vector has additional density of 5 users/km2.

Vector User per km2

V1

15

Vector User per km2

V1

20

V2

20

V3 205/22/2012 27


28/40

Capacity Analysis

Capacity Analysis by Investigating LTE Services Service parameters affecting simulation results

Traffic type (Real time/Non real time)

Minimum Guaranteed Bit Rate (GBR)

Maximum Bit Rate (MBR)

Highest and lowest bearers in downlink and uplink

For non real time services (data), Asset puts a constraint on Maximum Bit

Rate.

MBR is set to be equal to GBR for data services.

Data service in Asset is LTE service.

Data service in Atoll is peak performance service as it covers minimum and

maximum possible rates of all bearers.

Traffic Type Non Real Time

GBR / MBR (DL) 1000 / 1000 kbps

GBR / MBR (UL) 1000 / 1000 kbps

Highest and lowest bearer (DL) LTE_DL_12 and LTE_DL_1

Highest and lowest bearer (UL) LTE_UL_12 and LTE_UL_1

Type

Data

Minimum throughput demand (DL) 10 kbps

Maximum throughput demand (DL) 100000 kbps

Minimum throughput demand (UL) 10 kbps

Maximum throughput demand (UL) 40000 kbps

Highest and lowest bearer (DL) LTE_01 and LTE_12

Highest and lowest bearer (UL) LTE_01 and LTE_125/22/2012 28


29/40

Capacity Analysis

Capacity Analysis by Investigating LTE Services

With default services assigned to users, simulation snapshots are run. Offered traffic, carried traffic (served users) and total peak RLC throughput

of the network are obtained from simulation reports.

Results produced by Asset and Atoll are considerably different.

Carried traffic in Atoll is more than in Asset

Network total peak RLC throughput is much higher in Atoll than in Asset. Their total peakRLC throughput in downlink differs by 340.25%.

Simulation outputs Asset Atoll difference % difference

Offered traffic 435.10 436.23 -1.13 0.26

Carried traffic 384.35 436.13 -51.78

13.47

Total peak RLC throughput of network DL

(kbps)

384350 1692084.1 -1307734 340.25

In realistic scenarios, a packet based service has varying data rates, i.e. in reality

maximum throughput demand is not necessarily same as minimum throughput

demand.

Asset definition of non-real type of service does not represent a non-real time

service in practical cases.5/22/2012 29


30/40

Capacity Analysis

Capacity Analysis by Investigating LTE Services Thus a real type of service has to be assigned to the users in Asset because

those services do not put such a constraint on Maximum Bit Rate.

In Asset, real type of service assigned to users have following parameters:

Traffic Type

Real Time

GBR / MBR (DL) 10 / 100000 kbps

GBR / MBR (UL) 10 / 40000 kbps

Highest and lowest bearer (DL) LTE_DL_12 and LTE_DL_1

Highest and lowest bearer (UL) LTE_UL_12 and LTE_UL_1

With new service assigned to users in Asset, simulation snapshots are run.


Offered traffic 435.45 432.55 2.9 0.67

Carried traffic 428.55 432.4 -3.85 0.9


(kbps)

1399025.89

1740693.71

-341668

24.42

Carried traffic in the tools are pretty comparable.

Total throughput in Atoll is 24.42%higher with respect to total throughput in

Asset.5/22/2012 30


31/40

Capacity Analysis

Capacity Analysis by Investigating MIMO Modes The planning tools provide adaptive switching between different antenna

configurations, resulting in considerable improvement in system

performance.

SINR requirements for bearers are adjusted in such a way that they include

effect of adaptive switching between multiplexing and diversity.

Thus in Asset, multiplexing and in Atoll SU-MIMO are selected overadaptive switching, respectively. These modes effectively implements

switching between multiplexing and diversity.

Spatial multiplexing in Asset and SU-MIMO in Atoll are implemented

differently.

In Asset spatial multiplexing is not implemented by increasing the bearerrate but rather by reducing an offset from SINR requirements of bearers.

In Atoll SU- MIMO is realized by multiplying bearer rate with an offset

obtained from measurements.

5/22/2012 31


32/40

Capacity Analysis

Capacity Analysis by Investigating MIMO Modes Modified SINR requirements for bearers in Asset are obtained from

following formula:

,,

= ,

+ ,

+ ,

In the original Asset parameters for LTE downlink bearers, the SU-MIMOSINR Delta values,

, , were given with the wrong sign. This

error was corrected in Asset.

5/22/2012 32


33/40

Capacity Analysis

Capacity Analysis by Investigating MIMO Modes This error was corrected in Asset and simulation snapshot are run.

The results are presented in following figure.

Differences between total peak RLC throughputs and carried traffic of the

network simulated by the tools have been reduced.

Simulation outputs

Asset

Atoll

difference

% difference

Offered traffic 440.95 435.40 5.55 1.25

Carried traffic 433.25 435.35 -2.1 0.48

Total peak RLC throughput of network DL (kbps) 1531568 1740693.71 -209126 13.65

5/22/2012 33


34/40

Capacity Analysis

Capacity Analysis by Investigating Scheduling Methods

Proportional Fair (PF) is a conventional scheduling method used in cellular

planning.

It is chosen for eNodeBs in the tests.

This method distribute resources among connected users based on theirchannel conditions

Asset and Atoll implement proportional fair algorithm differently.

Asset implements it according to the definition mentioned above, i.e. inAsset users receive unused resources according to their channelconditions.

For proportional fair, Atoll considers gains due multiuser diversity whichare functions of number of users considered for scheduling in a cell and

multiuser SINR threshold which is set manually in Atoll. Atoll includes multiuser diversity gain (MUG) in peak RLC channel

throughput calculation of the connected users with maximum throughputdemand.

5/22/2012 34


35/40

Capacity Analysis


To reveal the differences in implementation of proportional fair algorithmsin the tools, scheduling methods are changed to Round Robin (RR).

Both of the tools implement this algorithm similarly.

Round robin method distributes unused resources equally among all

connected users.

Simulation snapshots are run. The results are presented below.


Offered traffic 435.95 433.98 1.97 0.45

Carried traffic 428.59 433.8 -5.21 1.21

Total peak RLC throughput of network DL (kbps) 1191385.20 1165339.91 26045.29 2.18

Difference between DL total peak RLC throughputs of the networks is 2.18%

which is relatively small comparing with the differences obtained when

proportional fair was selected as scheduling method in the tools.

5/22/2012 35


36/40

Capacity Analysis Capacity Analysis by Investigating Scheduling Methods

To reveal effects of multiuser diversity gain (MUG) in total peak RLC

throughput of the network simulated by Atoll, two tests are carried out.

In one test, maximum multiuser gain thresholdis set high enough so that

throughput of all users is magnified by corresponding MUG.

Simulation results of this test are presented in second column of following

table.

In the other test, maximum multiuser gain threshold is set low enough

which effectively disables multiuser diversity capability of proportional fair

method in Atoll and consequently throughput of users would not be

modified by MUG.

Simulation results of this test are presented in third column of following

table.

Simulation outputs with MUG without MUG difference %difference

Offered traffic 431.93

429.1

2.83

0.66

Carried traffic 431.82

429

2.82

0.66


(kbps)

1913199.47 1162241.99 750958.48 64.61

5/22/2012 36


37/40

Capacity Analysis


As it can be observed from the table, multiuser diversity has significant

effect on total peak RLC throughput of the network in Atoll.

In Atoll when MUG is applied to the users total throughput of the networkis 64.61%higher than the case when no MUG is applied to any of theusers.

Comparing throughput results when multiuser diversity is disabled with

the results when round robin is used for scheduling indicates that totalthroughput in both of the cases are comparatively similar to each other.

Proportional fair scheduling in Atoll exploits fast fading characteristics ofthe channel to maximize total throughput of the network.

When number of scheduled users is large, the probability that some usersare in good channel state is high and these users can be scheduled first.

in long term total cell throughput is increased by taking advantage offading channels. This is called multiuser diversity.

The dissimilarity in implementation of proportional fair method in Assetand Atoll causes relatively significant difference in resulting total peak RLCthroughputs of the tools.

5/22/2012 37


38/40

Conclusions and Future Work

The first stage of comparison was derivation of equivalent propagation

models, since all computed coverage, interference and capacity results in a

cellular network planning tool are based on path losses that are computedfrom a propagation prediction model.

The second stage of tool comparison was comparison of coverage prediction

results. The result of comparison was that the tools produce very similar RS

EPRE coverage arrays.

The third stage of tool comparison was comparison of interference predictionresults.

Atoll computes systematically about 3 dBlower downlink SINR levels for both

DLRS and DL TCH than Asset when two transmit antennas were used in base

stations with 75%loading.

The reason is that Atoll multiplies the received interference power by the

number of transmit antennas in the interfering cell, which Asset does not do.

Atoll also applies a cyclic prefix correction to the received interference power,

which Asset again does not do.

To compensate for these differences in computation, a conversion factor for

converting a SINR value in one tool into a SINR value in another was derived.5/22/2012 38


39/40

Conclusions and Future Work

The last stage of tool comparison was traffic capacity comparison.Capacity analysis requires Monte Carlo simulation, which requires trafficlayer as its input. The capacity comparison was based on number ofserved terminals and total peak RLC throughput.

Aircom's implementation of a non-real-time service is such that itsmaximum bit rate (MBR) demand is set equal to its minimum guaranteedbit rate (GBR) demand.

It was also found out that the SU-MIMO SINR Delta values in the Assetimport files had wrong signs, which had to be corrected before the realcapacity comparison.

The result of capacity comparison was that Atoll showed almost 13%higher total peak RLC throughput for the whole network than Asset.

Implementation of the proportional fair scheduling in the tools was thelikely reason for differences in the capacity estimates.

During this study, no reliable procedure for simulating equivalent capacityestimated from the two tools was found when proportional fair schedulingis used.

Since the proportional fair algorithm is the most commonly usedscheduling algorithm in LTE, this is a serious drawback, and further studieson workarounds for achieving approximately comparable capacityestimation results would be necessary.

5/22/2012 39


40/40

Thank You

Any Questions?

5/22/2012 40

Comparison of Asset and Atoll Cellular Planning tools for LTE Network Planning.pdf

Documents

Transcript of Comparison of Asset and Atoll Cellular Planning tools for LTE Network Planning.pdf