ASSET 5.0 Model Tuning Guide v1.0
RADIO ENGINEERING SOLUTIONS
How-To Guide for Model Calibration Summer 2004
Page 1
How-To Guide for Model Calibration
Introduction This document outlines and provides selected details about building a empirical model in ASSET using drive test data. ata collection is not e!tensivel" discussed# as its procedures var" depending on e$uipment used. %n general# it is assumed that T% e$uipment is used to collect &CCH data on a live networ'. An update of the various various data collection collection methodologies# methodologies# data data filtering guidelines# guidelines# and discussion discussion of acceptable formats is presented. %t illustrates how to navigate between various modules of ASSET and suggests useful tips on various user dependent options. %t provides some general guidelines to an ASSET user to better calibrate the propagation model. but do not address ever" possible approach to model calibration.
CW Measurements Traditionall" C( )ield measurements are carried out using a spectrum anal"*er# which measure the output of a test transmitter# which produces a Continues (ave +utput at the desired fre$uenc" and output power. This document does not discuss traditional C( t"pe drive testing# but data preparation# import# and anal"sis is essentiall" the same. %n carr"ing out C( t"pe measurements# the engineer has full control of the transmit facilit" and 'nows with great certaint" site power and antenna parameters. Most often# this is a omni antenna# so a*imuth and downtilt become irrelevant. ,nfortunatel"# in a live networ'# there ma" be some errors associated with the site databases used for this wor'. (hile C( measurements ma" onl" involve or site locations /and 0-1' sample points2# &CCH measurements can utili*e as man" site locations as time permits# and the number of sample points can be magnitudes larger /033 4 33'2. (ith the large diversit" of site locations that ma" be used# it will be of greater difficult" to achieve traditional error limits of 5 d& /std. dev.2 %f this is a limiting factor to "our wor'# reduce the number of sites used in anal"sis.
Live System/CC! Measurements The need to carr" out measurements on modulated &roadcast Channel /&CCH2 arises from the long setup time involved in C( measurements and from the large overhead of data collection over the repeated routes in the same location. Modulated &CCH measurements involves using a Scanner that carries out fast multiple fre$uenc" scanning# and is also able to decode the &ase Station %dentit" Code /&S%C2 and the Transmitter %. The scanning is carried out on 6%7E networ's# and does not use up s"stem resources. The scanner scans scans all the fre$uenc fre$uencies ies that are used as a &roadc &roadcast ast Channel# Channel# and logs logs the position# position# the fre$uenc"# the &S%C and Transmitter %. The ma8or advantage of this method is the near nil setup time and the ease of data collection. This enables the data collection of man" sites# and hence a more accurate model calibration. There is also a fle!ibilit" of choosing an" site to carr" out model tuning# even after the data collection is completed. There There are a few disadvant disadvantages ages in carr"in carr"ing g out Modulat Modulated ed &CCH &CCH measur measureme ements nts for model model calibration9 •
•
The most prominent disadvantage disadvantage being the use of directional antennae with ver" narrow vertical beam widths and having appreciable vertical down tilts. This tends to distort the radiation pattern of the antenna which has a significant effect on the model developed. %n dense urban areas# often antennae are below the surrounding clutter# with the bore sight of the antenna pointing towards the street. This leads to tunneling of the signal through the street# with a ver" high roll off of signal strength of streets perpendicular to the main street.
)or %nternal ,se +nl"
:reliminar" v0.3
:age
Page 2
How-To Guide for Model Calibration
Introduction This document outlines and provides selected details about building a empirical model in ASSET using drive test data. ata collection is not e!tensivel" discussed# as its procedures var" depending on e$uipment used. %n general# it is assumed that T% e$uipment is used to collect &CCH data on a live networ'. An update of the various various data collection collection methodologies# methodologies# data data filtering guidelines# guidelines# and discussion discussion of acceptable formats is presented. %t illustrates how to navigate between various modules of ASSET and suggests useful tips on various user dependent options. %t provides some general guidelines to an ASSET user to better calibrate the propagation model. but do not address ever" possible approach to model calibration.
CW Measurements Traditionall" C( )ield measurements are carried out using a spectrum anal"*er# which measure the output of a test transmitter# which produces a Continues (ave +utput at the desired fre$uenc" and output power. This document does not discuss traditional C( t"pe drive testing# but data preparation# import# and anal"sis is essentiall" the same. %n carr"ing out C( t"pe measurements# the engineer has full control of the transmit facilit" and 'nows with great certaint" site power and antenna parameters. Most often# this is a omni antenna# so a*imuth and downtilt become irrelevant. ,nfortunatel"# in a live networ'# there ma" be some errors associated with the site databases used for this wor'. (hile C( measurements ma" onl" involve or site locations /and 0-1' sample points2# &CCH measurements can utili*e as man" site locations as time permits# and the number of sample points can be magnitudes larger /033 4 33'2. (ith the large diversit" of site locations that ma" be used# it will be of greater difficult" to achieve traditional error limits of 5 d& /std. dev.2 %f this is a limiting factor to "our wor'# reduce the number of sites used in anal"sis.
Live System/CC! Measurements The need to carr" out measurements on modulated &roadcast Channel /&CCH2 arises from the long setup time involved in C( measurements and from the large overhead of data collection over the repeated routes in the same location. Modulated &CCH measurements involves using a Scanner that carries out fast multiple fre$uenc" scanning# and is also able to decode the &ase Station %dentit" Code /&S%C2 and the Transmitter %. The scanning is carried out on 6%7E networ's# and does not use up s"stem resources. The scanner scans scans all the fre$uenc fre$uencies ies that are used as a &roadc &roadcast ast Channel# Channel# and logs logs the position# position# the fre$uenc"# the &S%C and Transmitter %. The ma8or advantage of this method is the near nil setup time and the ease of data collection. This enables the data collection of man" sites# and hence a more accurate model calibration. There is also a fle!ibilit" of choosing an" site to carr" out model tuning# even after the data collection is completed. There There are a few disadvant disadvantages ages in carr"in carr"ing g out Modulat Modulated ed &CCH &CCH measur measureme ements nts for model model calibration9 •
•
The most prominent disadvantage disadvantage being the use of directional antennae with ver" narrow vertical beam widths and having appreciable vertical down tilts. This tends to distort the radiation pattern of the antenna which has a significant effect on the model developed. %n dense urban areas# often antennae are below the surrounding clutter# with the bore sight of the antenna pointing towards the street. This leads to tunneling of the signal through the street# with a ver" high roll off of signal strength of streets perpendicular to the main street.
)or %nternal ,se +nl"
:reliminar" v0.3
:age
Page 2
How-To Guide for Model Calibration
•
Calibrating this t"pe of propagation is ver" difficult# if not impossible# using a slope intercept model. Also# the modeling modeling of the data collected in the bac' lobe of the directional antenna antenna is ver" difficult and tends to introduce error into the model. This problem can be addressed b" using appropriate antenna filtering /i.e.2 using a filter to e!clude points outside the d& beamwidth of the antenna.
There are advantages and disadvantages to each method.
"ost#"rocessin$ o% Scanned Modu&ated CC! data :ost :ost process processing ing of the data data involve involves s assign assignmen mentt of a particu particular lar measur measureme ement nt of a particu particular lar fre$u fre$uen enc" c" to its respe respect ctiv ive e trans transmi mitte tters rs using using uni$ uni$ue ue &S%C &S%C4&C 4&CCH CH-T -Tran ransm smitt itter er %ds; %ds;.. )or )or measurements in which the &S%C and?.35@05> ->?.033?51 ->?.055?>@ ->?.3355>@ ->?.0@?103? ->?.05@0@?@ ->?.0>>005
Latitude Drive_Number Setor_!"_#ower C_$ %&! !'()*L ?3.11>>3? 0 -@@. 0.>>1 .?5@30 ? ?3.1?>31 0 -00.> -3.??>51@0 > ?3.130@@ 0 -003. -3.@00?1@ > ?3.1>03?> 0 -5.@@ > 3.0@ 3 ?3.101? 0 -10.@ > 3.0@ 3 ?3 ? 3.105@? 0 -1?.5 > 3.0@ 3 ?3.1?1 0 -@0.>0 > 3.0@ 3
This data is produced# on a per sector basis# in an MS Access format /B.mdb2. %n order to import into ASSET# it must be edited /with a te!t editor# or more simpl"# MS E!cel2 and put into a form that ASSET ASSET will will recogni*e. recogni*e. Although Although there there are several several formats formats that ASSE ASSET T can read# read# a common common one for model tuning is the Signia format. T+e Signia ,ormat The Signia format is used as it is convenient and eas" to create /MS E!cel is the most li'el" editor2.There are two files needed for Signia# a Header /B.hd2 file and a ata /B.dat2 file. The Header and ata files are lin'ed b" an identical file name. Hint9 The header and data file must be in the same folder# and the folder can have an" path.
)or %nternal ,se +nl"
:reliminar" v0.3
:age
Page 3
How-To Guide for Model Calibration
Header file ./+d1 A Header file is a tab-delimited te!t file containing information regarding the individual cells. Although there are man" fields# onl" a few of them are critical for model tuning.
Hint9 %f header file does not load# chec' format# spacing# and E+) mar'er /carriage return2# for errors. =emove an" tags or unit /degrees# feet# etc.2 from the input. Hint9 &e sure ATEADT:E is located in the antenna database file# or an error message will be generated when loading. TFDHE%GHT should be in meters# and TFD:+(E= in d&m. TFD:+(E= is the site E%=:# not hatchplate power. Hint9 To facilitate file management# ma'e the S%TED%# the header file name# and the data file name identical.
)or %nternal ,se +nl"
:reliminar" v0.3
:age ?
Page 4
How-To Guide for Model Calibration
Data ,ile .dat1 The ata file is a tab-delimited te!t file containing 6ong# 6at# and =SS% of each cell as measured b" the receiver device. ecimal 6at-6ong /662 formatting is re$uired and each line represents one measurement location. There is no limitation to the number of measurement points in a ata file. %f MS E!cel is used as the te!t editor# there will be a limit of 1' points. 6ongitude /662 6atitude /662 =eceived Signal Strength %ndicator /=SS%2 d&m
Tags
Single data entry
)or %nternal ,se +nl"
:reliminar" v0.3
:age 1
Page 5
How-To Guide for Model Calibration
Loadin$ Drive Data into ASSET The rive ata is loaded from the Main Menu Tab Tools This opens a pop-up window as shown below.
C(
Measurements Anal"sis Tab.
ot applicable. )or use with =A+:T optimi*ation tool.
Adds<=emoves individual sector drive data files ispla"s sector information contained in header /B.hd2 file :ops-up )iltering and Model Selection (indow :ops-up Graph window for =!6ev 7s istance# or Mean Error 7s istance &egins regression curve fitting and prompts user for Error =eport T"pe :ops-up AutoTuner (indow# showing initial parmeter set.
Hint9 (hen a sector drive file is added# the user is prompted for I&in AveragingI# which averages all the samples found within a map bin. This feature is usuall" not selected# but ma" be applicable for drives with high number of samples# such as in a ense ,rban area where the test vehicle was moving ver" slowl".
Add/Remove uttons This adds or removes individual sector data files for anal"sis. The loaded sectors and file path
)or %nternal ,se +nl"
:reliminar" v0.3
:age
Page 6
How-To Guide for Model Calibration
In%o utton +nce a sector file is loaded# the site information pertaining to that cell can be reviewed. %nformation about the location of the test site# output power# antenna height# cable and connector t"pe and losses# and the antenna t"pe. %f there is cause or need to edit this info# it can be done at through this window.
Hint9 Most often# site parms such as )EEE=D6EGTH# T:E# C+ECT+=D 6+SS# etc. ma" not be 'nown. %nsert the site E%=: value as the transmit power /TFD:+(E=2 value and *ero the cable and connector losses. Hint9 %f site parameter changes are made in these windows# the changes will be applied# but not committed. ou must manuall" change the header file if "ou desire an" permanent parameter changes.
)or %nternal ,se +nl"
:reliminar" v0.3
:age >
Page 7
How-To Guide for Model Calibration
O'tions/(i&ter Ta) This window provides filtering options that the user ma" wish to emplo"# depending on the tas' involved. istance# signal level# 6ine-of-Sight# and Antenna )iltering are shown. Also given is the option of removing specific data points assigned to clutter t"pes. More on the usage of these filters is given in the Tuning and Anal"sis section of this document.
Gra'* utton The tool will also produce a graph of the sample data vs. distance. This graph shows a numerical intercept and gradient value for the data# but does not t"picall" give useful insight for calibration.
Also available is mean error vs. distance. This givesJ. )or %nternal ,se +nl"
:reliminar" v0.3
:age 5
Page
How-To Guide for Model Calibration
Ana&yse utton Generating an anal"sis of the chosen base model versus the actual data points generates an %nitial Statistics of the data loaded. The Anal"se button prompts for Anal"sis =eport +ptions# and based on user needs# the report options are chosen.
ispla" Mode onl" valid for &in %nfo =eport
Anal"sis Tab to generate %nitial Statistics
ote9 The report can either be generated in either MS EFCE6 /best option2 or an" Te!t Editor. The following shows e!ample of =eport +ptions generated upon completion of anal"sis. ,ile Summar - provides summar" on a per-file or per-sector basis /i.e. per drive test2 The file summar" identifies the various sites loaded into the s"stem for anal"sis# along with a site wise brea'up of the data points /Num.Bins2 collected for that sector# the Mean Error # the =oot Mean S$uare Error /RMS Error 2# the Standard eviation /Std.Dev Error 2 and the correlation coefficient /Corr.Coeff 2. %t helps the user in assessing the model on a site b" site basis# and also helps the user# if re$uired# to reclassif" certain sites under a different morpholog" class. E!ample )ile Summar" Site $D Site Name DN0304C DN0304C
Num %in 523
Mean &rror -67
!MS &rror StdDev &rror Corr Coeff 688 60 08292
:verall Summar - gives overall summar" of all drive files loaded. The overall summar" provides the combined statistics of how the model compares with the collected data. The values provided in the overall summar" are the 'e" points b" which the model is evaluated. E!ample Model Summar" Model Num %in D&N_S)_6004_v6 523
Mean &rror !MS &rror StdDev &rror Corr Coeff -67 688 60 08292
Clutter Summar - gives brea'down of error based on clutter t"pe. The clutter summar" provides clutter wise distribution of mean error and standard deviation. This particular table is ver" useful to help tune clutter parameters. E!ample Clutter Summar" Clutter )or %nternal ,se +nl"
Num %in Mean &rror !MS &rror StdDev &rror Corr Coeff :reliminar" v0.3
:age @
Page !
How-To Guide for Model Calibration
,oret :;enLand LowDenit%uilding MediumDenit%uilding Tran;ortation • •
•
•
•
63 393 286 204 2
-630 -208 -87 -630 -290
64 232 609 68 298
78 603 74 50 75
05669 09280 08597 0949 02584
um. &ins - The number of =SS% samples within the sector file# bro'en down b" clutter class. Mean Error - the calculated mean error between the measured and predicted values. A negative value indicates the model is underpredicting. =MS Error - the root-mean-s$uared error. Generall" a measure of the IspreadI of the error between the measured and predicted values. Std. ev. Error - the classic measure of IgoodnessI in model tuning. %t is more a measure of the ImagnitudeI of error between the measured and predicted values. Correlation Coefficient - between 0.3 and -0.3# it is a statistical measure of degree of linear relationship between the measured and predicted values# or how well the sample points fit the model curve. The higher the value# the better the relationship. A value of 3.> is t"pical.
These reports are useful to help tune "our model and guide parameter changes. Error values /high or low2 are not relevant with a small sample si*e /i.e.# less than 33-33 pts.2
Autotune utton +nce Header files are loaded# when the Autotune button is selected# a Model Calibration ,tilit" window will appear. %n the Status 6og# the data files will load individuall" and the tool will compute %nitial Statistics based on the selected model chosen in under I+ptionsI.
%nitial Statistics
Status 6og
)or %nternal ,se +nl"
:reliminar" v0.3
:age 03
Page 10
How-To Guide for Model Calibration
"otes #nsert $i% o& 'odel $ar's (indo(. )is%uss use o& *eig+t Pro&iler tool in analysis
)or %nternal ,se +nl"
:reliminar" v0.3
:age 00
Page 11
How-To Guide for Model Calibration
#re;aring t+e Data (i&e Screenin$ )ile Screening refers reviewing a sector file with Asset and ma'ing a sub8ective call as to acceptance or re8ection of the data set. As a first step# each drive file must be screened. %ndividual sector files are loaded and inspected in the ASSET map window. Screening of individual sectors is performed to chec' for anomalies such as possible bloc'ed antenna or errors in the site database such as9 • • • • •
%ncorrect antenna orientation E!cessive downtilt /greater than 03 deg for a ver" narrow /less than ? deg &(2 antenna 6ow antenna height /less than 03 meters# but dependent on cluster average 6ow E%=: /less than d&m2 6ow number of data points /less than 33 samples2
The Sectors that are discarded are summari*ed on a Mortalit" 6ist# with specific reasons and recommendations are made based on them. E!ample of a )ailed Sector 4 A*imuth Error9
The site data indicates an a*imuth of ?3 degrees# but the plot shows ver" little correlation. This sample set ma" be discarded and added to a mortalit" list. A suggestion was made to the mar'et to chec' for a possible sector cabling and
)or %nternal ,se +nl"
:reliminar" v0.3
:age 0
Page 12
How-To Guide for Model Calibration
Data (i&terin$ After screening each sector file and accepting the data# the sample file needs to be filtered to remove data points that are either unreliable or not desirable for model tuning. These include data points that are within a certain radius of the antenna# be"ond a certain radius of the antenna# data points that have =SS% less than a specified power# and data points that have a =SS% that is considered to be wea'er than the noise floor of the scanner. The filtering process aids in e!cluding data points that lie outside the linear region of the amplifier of scanner and hence the propagation path. The values for power levels and distances are largel" based on e$uipment specs and site specs respectivel". +ther filtering options can be applied based on 6ine-of-sight /6+S2 or on 6ine-of-sight /6+S2 data points. The filtering is based on terrain data# but can also ta'e into consideration building vector heights and clutter heights if the" are assigned. This filtering is used to compute the effect of diffraction. &"lude Clutter This removes samples based on clutter t"pe. +ften# clutter t"pes with an insufficient number of samples /for reliabilit" reasons2 ma" also be e!cluded from anal"sis. This is done b" selecting those clutter t"pes from anal"sis in the filter window.
=emoves samples outside of given distances. 7alues var" given morpholog"# site height and terrain.
=emoves samples based on signal strength. 7alues shown are t"pical.
=emoves either 6+S or on-6+S samples. &ased on terrain data onl". ,seful for evaluation of K>. =emoves samples outside of antenna beamwidth. Chec'ed when using directional antennas.
Antenna eam+idt* (i&terin$ (hen using live-s"stem or &CCH drive data# or when using directional antennas# it is necessar" to filter datapoints outside of the main antenna beamwidth. This removes the sample points outside of the calculated d& beamwidth0 of the antenna# as inclusion of these points will distort the model as
1
The beamwidth is determined b" ASSET b" reading in the antenna pattern and cannot be altered or changed b" the label in the antenna pattern file. )or %nternal ,se +nl"
:reliminar" v0.3
:age 0
Page 13
How-To Guide for Model Calibration
there will be a wide spread of signal values vs. distance. The influence on K0 and K ma" be substantial# as there will be a wider spread of sample points relative to distance from the site.
See e!ample graph. :lot of unfiltered drive data9
Antenna A,i - 340
)or %nternal ,se +nl"
:reliminar" v0.3
:age 0?
Page 14
How-To Guide for Model Calibration
:lot of data file with d& Antenna filter
O'tions/Mode& Ta) Multiple map resolutionsJ Map =esolution at which Model is to be tuned
Model to be tuned. See IAdding a &ase ModelI
ASSET can have igital maps with more than one resolution /t"pical 1m and 033m or 3m and @3m2. Since Model Calibration is done based on bin b" bin basis# selection of the Map =esolution is needed.
)or %nternal ,se +nl"
:reliminar" v0.3
:age 01
Page 15
How-To Guide for Model Calibration
Hint9 %f there is onl" a single Map =esolution# then that resolution is default# otherwise# a selection needs to be made. Choose a Map with a higher resolution# so as to produce a more finel" tuned models# but if there is drive samples in lower resolution bins# these will not be included in the Anal"sis.
)or %nternal ,se +nl"
:reliminar" v0.3
:age 0
Page 16
How-To Guide for Model Calibration
*dding a %ae Model Asset De%au&t Mode& After setup of the Autotuner and )iltering +ptions# the user must define a IdefaultI model. The default will be renamed after tuning to according to mar'et re$uirements and categori*ation of the sample files. This is usuall" based on build-up# with each mar'et defining its morpholog" classes such as ense ,rban# ,rban# Suburban# =ural. The efault model contains all the standard /untuned2 values in the model# such as fre$uenc"# Effective Antenna Height Algorithm# the iffraction Methodolog"# etc. These are seen under IConfiguration# -L I:ropagation ModelsI. ote9 The IMacrocell 3 Model I is used as a base model with its defaults for 0@33 MH*. )or a model in the :CS band# the fre$uenc" is set to 0@3 MH*. The Effective Base Station Antenna Height algorithm used is the !elative algorithm /this is the calculated height between the base station antenna and the mobile antenna and is the most accurate representation2. Diffraction Loss is calculated using &;tein-#eteron method without merging an" of 'nife-edges along the path of the terrain database. Maroell 3 Model Default <6 <2 03.33 ?3.33 <6 /near2 3
<3 <4 -.11 3.33 <2 /near2 3
Effective Antenna Height Algorithm iffraction 6oss Calculation Method Mobile Antenna Height /m2 Clutter T;e (ater )orest +pen 6ow ensit" &uilding Med ensit" &uilding High ensit" &uilding Ma8or Transportation Airport
< -0.5
<9 3.53
!elative &;tein #eteron 6
T+roug+ Clutter Lo .d%=>m1 -.3 .3 3.3 .3 .3 @.3 -.3 -.3
Through Clutter 6oss istance /m2
)or %nternal ,se +nl"
<7 -.11 3.3 'm
Clutter :ffet .d%1 3.3 3.3 3.3 3.3 3.3 3.3 3.3 3.3 533
:reliminar" v0.3
:age 0>
Page 17
How-To Guide for Model Calibration
General Tab
#at+ Lo Tab
Two-;iee model ;arm
)or %nternal ,se +nl"
:reliminar" v0.3
:age 05
Page 1
How-To Guide for Model Calibration
Clutter Tab
)or %nternal ,se +nl"
:reliminar" v0.3
:age 0@
Page 1!
How-To Guide for Model Calibration
Tuning t+e Model Setu' o% AutoTuner "arms (ithout setting limits for the tool# it will return results that provide statistical merit# but are not necessaril" engineering sound. %nitial parameter values# iteration limits# elta =anges /which limit the change it can ma'e to a particular parameter2# and the i! or loc'ing of other parameters which the user does not want changed during the auto routine need to be initiali*ed. These ma" be narrowed as the user progresses towards a final value. )or the initial setup of +ptiimiser :arameters9 Ma! %terations - 033 Conv. Accurac" - 3.330 • •
)or elta =anges of K-parameters K elta =ange - 0.3# changing to 3.0 when narrowing. K> elta =ange - 3.30 • •
• • •
Nero all Through Clutter Settings and )i! )i! K1 to K> to default o not change K or K? from default value %teration 6imiter 7alue 6imiter 6oc' 7alues
)or %nternal ,se +nl"
:reliminar" v0.3
:age 3
Page 20
How-To Guide for Model Calibration
Initia&i,in$ -. and - The first step in model calibration is determining initial /or base2 values of K0 and K. The thin'ing at this point is to determine a best fit line for the most li'el" mobile location /dominant clutter t"pe2. K0 will be changed at several points during the tuning process# but should be ad8usted periodicall" to prevent other parameters from deviating too far from final value. • • • •
6oad Header files for a Single Morpholog" T"pe etermine most significant clutter t"pe for given morpholog" Appl" Autotuner defaults /given above2 Appl" Changes and record initial error values
Turn off all clutter e!cept that which is determined most dominant and Anal"se. The Autotuner will return an initial value for K and K0. %f the value for K is reasonable# then commit the values and continue. =emember these values are 8ust preliminar"# and further twea'ing will be necessar" at the end of the calibration process. Hint9 K0 alwa"s *eros the mean error. (hen mean error is positive# the model is underpredicting compared to the drive data and K0 should be reduced. (hen mean error is negative# the opposite applies.
Tunin$ %or -0 The ne!t step is tuning for diffraction loss and shadowing effects caused b" the terrain. %n urban or flat terrain areas# this ma" not be a significant factor# but must be investigated. K> is a multipl"ing factor that alters the impact of the diffraction loss and its value is alwa"s less than 0.3 To assess the effect of the diffraction effect# data points with on-6+S with the transmitting antenna are chosen. This is done be deselecting the 6+S data points in the filter options.
,nchec' 6+S Chec' on-6+S
The on-6+S data is then auto-tuned to return a value for K> alone# b" loc'ing A66 other parameters in the Auto-Tune Module.
)or %nternal ,se +nl"
:reliminar" v0.3
:age 0
Page 21
How-To Guide for Model Calibration
6oc'ed :arameters
The elta value for K> can be set to 3.30# and the %teration limits to 033. %f the returned value for K> is acceptable# the value is manuall" applied to the model. To chec' the effects of the change in K> on the overall model# the data points with 6+S are included and the anal"sis is rerun. A change in the standard deviation# and or a change in mean error and correlation coefficient are observed. %f the statistics show improvement# then the changes are committed. ote9 The change in K> value ma" result increase mean error in the anal"sis report. o not worr" about changing K0 until later in the process as other changes are still necessar". The mean error is *eroed out as a final step of model calibration.
Tunin$ -1 # -2 Model coefficients K 4 K are constants which alter the effect of the &S Effective Height Gain and the MS Antenna Gain. K and K? are used to modif" the effect of the mobile antenna height on the received signal strength. %n most mobile networ's# the mobile height is considered to be fi!ed at 0.1m above the terrain height. The default values for 0@33MH* s"stems are K -.11 and K? 3.33. These values are not altered when model tuning. K1 and K are used to modif" the effect of the base station antenna height gain on the received signal strength. Since the Effective Height Algorithm used is the Relative Method # the effect of the terrain data is more prominent than the absolute base station antenna height. The default values for K1 and K are hence not generall" altered. )or %nternal ,se +nl"
:reliminar" v0.3
:age
Page 22
How-To Guide for Model Calibration
Tuning for Clutter C&utter T*ru#Loss (hen using high-resolution clutter data# a more accurate model can be developed utili*ing the Thru-6oss algorithm within the Macrocell model. A reduction in std. dev of 0- d& can usuall" be achieved if applied properl". Tuning for Thru-6oss# li'e the model constants# is partl" a manual and iterative process# but the Autotuner can help the user ma'e initial assignments. As with other model parameters# the user must help guide the autotune process through use of range deltas and fi!ing of parameters. After a sanit" chec' of values and noting cause and effect the values can be applied. =eview the Clutter Summar" to get the number of sample points used to ma'e an assignment b" the Autotuner. Some clutter t"pes will have a ver" low number of samples and will need some manipulation b" the user. Clutter t"pes with a high number of samples are generall" reasonable for assignment. ,se these values to help guilde manual assignment to the ones with few sample points# as there should be a trend in the values. 6astl"# round-up or down values to maintain simplicit" /e!9 1.@? to .32. ,se the table below for sanit" chec'ing of assignments. %t will var" slightl" from model to model# but will maintain a trend as mentioned /values will be higher for an urban# or more built-up area# and lower for a more rural or open area2. Clutter Default ?alue <6 <2 03.33 ?3.33 <6 /near2 3
<3 <4 -.11 3.33 <2 /near2 3
Effective Antenna Height Algorithm iffraction 6oss Calculation Method Mobile Antenna Height /m2 Clutter T;e (ater )orest +pen 6ow ensit" &uilding Med ensit" &uilding High ensit" &uilding Ma8or Transportation Airport
< -0.5
<7 -.11 3.3 'm
<9 3.53
!elative &;tein #eteron 6
T+roug+ Clutter Lo .d%=>m1 -.3 .3 3.3 .3 .3 @.3 -.3 -.3
Through Clutter 6oss istance /m2
Clutter :ffet .d%1 3.3 3.3 3.3 3.3 3.3 3.3 3.3 3.3 533
+nce all the Thru-6oss assignments have been made# Thru-6oss istance is e!amined. Thru-6oss istance is based on morpholog"# but is also influenced b" the average antenna height. T"pical values for Thru-6oss istance range from 133 to 0333 meters. ,se the Autotuner results for initial guidance and finali*e based on error results. )inall"# twea' the Thru-6oss values b" e!amining the Clutter Summar".
C&utter O%%sets 6astl"# clutter offsets are assigned. ,nli'e Thru-6oss# Clutter offsets have no trend and will often be ver" close to *ero when using Thru-6oss. This is a final offset made b" the Autotuner to reduce the )or %nternal ,se +nl"
:reliminar" v0.3
:age
Page 23
How-To Guide for Model Calibration
mean error# but li'e above# is meaningless for clutter t"pes with ver" few sample points. ,se these returned values with the same discretion as all other values. Clutter +ffsets are end-point offsets associated with each clutter t"pe. Clutter +ffsets are based on statistical anal"sis that ma'es the final ad8ustments to the Through Clutter 6oss 4 Slope<%ntercept model. This value is used as a balancing mechanism to minimi*e the mean error. Hence# it values ma" not appear to be intuitive or follow the trend of values for Through Clutter 6oss. Clutter offsets wor' best to characteri*e +ceans# 6a'es# and =ivers /or (ater2. An assignment with deviates from the Autotuner value is most often re$uired# as it will mis-characteri*e the cross-water effect. A value of -.3 d& is t"pical.
(ina& Tunin$ o% -.3 - and C&utter O%%sets After all the Thru-6oss values and Thru-Clutter istance are twea'ed# finali*ed# and loc'ed down# final ad8ustments need to be made to K0 and K. =epeating the process for the initial ad8ustments of K0 and K returns the final values. This ensures a null mean error and null clutter mean errors# for the best slope possible. %n ver" few cases does K re$uire a change. %t is more li'el" if the thru-loss of the dominant clutters was changed greatl" in the above and the distribution of those changes /positivel" or negativel"2. &ecause clutter offsets cannot be fi!ed /or loc'ed2# the Autotuner assigns or updates them each time. The offsets will not be valid for clutter t"pes that have ver" few data points for a statisticall" reasonable assignment. %n these cases Clutter +ffset must be manuall" assigned and reviewed based on the trend shown.
)or %nternal ,se +nl"
:reliminar" v0.3
:age ?
Page 24
How-To Guide for Model Calibration
&rror *nali etermining if "our model is sound and reasonable is difficult. (hat if "our results give a standard deviation greater than 5 d&O (here to "ou loo' for errorsO (hat if none are foundO %f error values are not acceptable# consider possibilit" of a two-piece model. Eventuall" "ou will have to stop "our anal"sis as ever" possible parameter will have been twea'ed and modified. Effective Height Algorithm 4 Select a different effective height algorithm and recalculate the K1 and K parameters. iffraction 4 Choose a different diffraction algorithm and retune the diffraction parameter /'>2. Also investigate merging 'nife-edges. The Height :rofile window and the drive test signal and signal error on the Map 7iew provide valuable visual aids to identif"ing possible areas where merging ma" be re$uired and b" how much. +ther parameters that ma" be changed are Clutter Heights# Separation and Mobile Heights. Adding clutter heights a separation value /must be L 32 can be of occassional aid when modeling urban environments. Clutter Separation has the effect of modeling the Iurban can"onI situation of a mobile being at street level. 6astl"# mobile height models the situation of the mobile being at a specified height within the clutter.
)or %nternal ,se +nl"
:reliminar" v0.3
:age 1
Page 25
How-To Guide for Model Calibration
Com'arison Test ,sing comparison plots to change the model parameters is difficult# but is effective to develop more reliable models. Changes are made manuall"# the results noted and its effect of other parameters. %f the changes are reasonable to other tests and show an improvement in statistics# the changes are committed. E!ample of a Comparison :lot9
Hint9 (hen producing a comparison plot# after "ou have determined the number and signal level for the respective bands# it is convenient to represent the bands with a color that is a lighter shade of that used in the drive test. This helps ma'e the comparison more intuitive and easier to visuali*e.
)or %nternal ,se +nl"
:reliminar" v0.3
:age
Page 26
How-To Guide for Model Calibration
Re&ative "arameter Test Are the individual K parms and Thru-6oss values reasonableO How do the value compare from model to model. %t is reasonable to assume K0 and K are larger for urban environments than rural environments. oes this trend holdO
Error Test +verall# mean error# =MS error# and std dev are used in regression anal"sis to $uantif" the results. %ndividuall"# the results are not significant# but have depth when viewed collectivel". Mean Error %n all cases# overall mean error should be# or ver" close# to *ero. (hen not# it gives indication the model is overpredicting or underpredicting cell coverage. %n most cases it will range from P01 to -01 when viewed on a per cell basis. %f a cell mean error is significant# for e!ample -3# it ma" indicate an operational problem with the cell and the site should be removed from the anal"sis list. Also# the morpholog" of the cell could be mis-classified compared to the other cells and it is 8ust not a good fit. Std. ev. and =MS Error Std. ev and =MS error are almost the same and it is usuall" user preference on which one is most important. Error stats alone are not sufficient# as 5d& std dev ma" be impossible given environment and number of sites
Sin$&e S&o'e Mode& vs4 Dua& S&o'e Mode& Sometimes it is more appropriate to model the data distribution with a -piece model. A two-piece will fill in coverage near the site if the drive data shows this trend and occasionall"# can improve error results /0 to d& Std.2 %t is applicable for rural environments# as man-made reflections mas' this in urban settings. The characteristics of the radio propagation differ at the near-end and the far-end of the site. This model has a second K0 and K# which serve to characteri*e near antenna coverage# and then# after some brea'point distance# trends a line with a shallower K value. This is demonstrated in the picture below.
)or %nternal ,se +nl"
:reliminar" v0.3
:age >
Page 27
How-To Guide for Model Calibration
Two :art Models
$ntere;t 6 Slo;e 6
%rea> #oint $ntere;t 2 Slo;e 2
l e v e L e v i e e !
Di-tane from %a-e Station Hint9 )or the above graph to be theoreticall" valid# K0near has to be less than K0far# and Knear has to be greater /-1d&2 than Kfar.
There is no clear wa" to tell if the drive data is single or dual-slope other than close visual and anal"tical inspection of the data. Start with a single slope model and if it does not give the results desired# a two-piece should be investigated. The brea'point distance is best determined b" inspection of the drive data for a good number of sectors under test. Most often it is seen that the brea' point distance is between 0.1 and 'm for t"pical cells# however it will var" based on antenna height# E%=:# and morpholog" class.
Deve&o'in$ a #"iece Mode& A base model ma" be retuned to achieve the desired error statistics and at the same time concentrating on a best fit between the drive data and the propagation at the far-end of the site. Having calculated the various K values# Clutter Thru-6oss and Clutter +ffsets# proceed to develop a model for the near-end b" 8ust twea'ing K0 and K values b" specifi"ing K0 /near2 and K /near2 to achieve a best fit for the near-end of the site. The near-end of the site is determined b" a factor called brea' point distance /2. • • • •
Anal"*e data on whol as normal# and come to stopping point based on final error stats. &rea' data into two parts# near and far. )ilter the data on brea'point distance. Appro!imate using ?H0H
Hint9 %t also important to develop smooth transition from the near end to the far end. There should be no abrupt changes in signal level and IfeatheringI of the transition must be ta'en to ensure satisfaction with coverage plot. Most often this is apparent when Knear and Kfar get too far apart /L03d& t"pical2. To smooth# slight ad8ustments to K0 and K ma" be needed after inspection.
Sources o% Error The =) environment is ver" strange behaving at times. irect and ground-reflected waves# as well as reflections from buildings all impinge upon the mobile and produce a signal that is widel" )or %nternal ,se +nl"
:reliminar" v0.3
:age 5
Page 2
How-To Guide for Model Calibration
var"ing. iffraction# shadowing# tunneling# and cross-water effects contribute as well. %n combination is the speed of the mobile and its movement across a sector face# with var"ing gain and an "ou can produce a incredibl" var"ing random environment. )ortunatel"# this randomness follows some sort of order and can be $uantified b" regression anal"sis and statistics. The 5d& figure mentioned above is a starting point# but it attempts to identif" the randomness described as well as the )%T of a slope
Conluion Model tuning is an iterative process that re$uires time and patience# but most importantl" - a deliberate approach. This document attempts to give one such approach that the authors have found successful. The overall strateg" to maintain# regardless of approach is to find a middle point and then appl" successive twea'ing# tr"ing to improve the results. Chec' "our results# and twea' again. %f "ou go off course# then "ou revert bac' to a 'nown good result and tr" again# this time in another direction or with another parameter. This document has tried to mention# if not discuss# ever" parameter available for model tuning# and give some insight on how best to appl" it. Model tuning is part science# part art. The science is 'nowledge of radio behavior and statistical merits. The art comes with ad8ustment of some parms and seeing their effect on others. Trial and error is the onl" wa" to become adept.
)or %nternal ,se +nl"
:reliminar" v0.3
:age @
Page 2!
A""ENDI5 . # A)out t*e ASSET Macroce&& 1 Mode& $ntrodution The form and parameters of the base Macrocell model is based on the ETS%-HataBiffn P Closs (here :loss is :athloss# Hms is mobile station height# Heff is base station effective height# iffn is diffraction loss# and Closs is clutter loss. istance /d2 is in >ilometer. K0 is the intercept and is thought of as the amount of pathloss encountered at the 0'm point. K is the slope in d& is a modifier to the calculated diffraction loss# and is usuall" less than 0.3 • • • • •
Each clutter can also be assigned an associated "hru#loss in d&<'m and is used in con8unction with a Thru-6oss istance. A clutter-offset parameter is utili*ed as a final ad8ustment to minimi*e the mean error associated with a clutter t"pe. +ther parameters associated with clutter are clutter heights and separation /average distance from obstruction to mobile2. See below for a detailed e!planation about Thru-6oss algorithm. &ffetive Heig+t There are four Effective Antenna Height Algorithms within ASSET# each suited to different terrain and networ' characteristics. •
The Absolute method is not widel" used in cellular networ's but is in certain broadcast s"stems.
•
The Average method wor's well in flat or gentl" rolling terrain.
•
The =elative method wor's well in rolling-hill" terrain where the base station is normall" above the mobile.
•
The Slope method wor's well in hill" and severel" hill" areas where the other algorithms consistentl" over-estimate the Heff.
Diffration The diffraction algorithm determines how a loss figure is calculated when multiple 'nife-edges are detected along the terrain profile from base station to mobile. There are four methods within ASSET9
Page 30
How-To Guide for Model Calibration
• • • •
Epstein-:eterson - the loss from each 'nife edge is calculated and then summed together. &ullington - this method replaces the real terrain with a single 'nife edge. e"gout - loss is calculated relative to the main obstruction. Qapanese Atlas - similar to Epstein-:eterson# but height of transmitter is altered.
Hint9 T"picall"# Epstein-:eterson or &ullington are the most popular# but is user-preferenced. (hen using high-resolution terrain# merge 'nife-edges less than *ero.
A''endi6 # More on C&utter T*ru#Loss and T*ru#Loss Distance Clutter offsets are a fi!ed end point correction factor that improves the correlation between measured and predicted pathloss. This improves the Standard eviation appreciabl"# but ta'es into consideration onl" the clutter t"pe that the pathloss is being computed for and doesn;t ta'e into consideration the loss due to the different clutter t"pes in the path of propagation. Through Clutter 6oss is the additional loss attributed to the clutter t"pe that the signal propagates through. The total thru-loss for a prediction point is calculated b" e!amining the clutter l"ing between the mobile towards the base over a defined distance# the Through 6oss distance# d. (hen calculating thru-loss# the individual bins are weighted linearl" so that the ones closest to the mobile have greatest effect# and the ones at point d have a minimum. The value of Through Clutter 6oss would var" for different environments# and depends largel" on the clutter through which the signal has alread" passed through. The effects of the clutter t"pe in the path tend to have a residual effect on the value of the Through Clutter 6oss parameter. %t is not a $uantitative measure of the additive loss associated with the clutter t"pe# but rather represents a value that could shape the straight line better in order to fit the measured data# and hence ma" not be intuitivel" assigned or predicted. Through Clutter istance represents the distance from the mobile towards the base station# through which the signal penetrates through the clutter. The remaining distance# the signal is assumed to propagate above the clutter. The Through Clutter 6oss and istance algorithm wor's as follows9 Through Clutter 6oss is added to the computed pathloss after appl"ing a weighing factor.
How-To Guide for Model Calibration
• • • •
Epstein-:eterson - the loss from each 'nife edge is calculated and then summed together. &ullington - this method replaces the real terrain with a single 'nife edge. e"gout - loss is calculated relative to the main obstruction. Qapanese Atlas - similar to Epstein-:eterson# but height of transmitter is altered.
Hint9 T"picall"# Epstein-:eterson or &ullington are the most popular# but is user-preferenced. (hen using high-resolution terrain# merge 'nife-edges less than *ero.
A''endi6 # More on C&utter T*ru#Loss and T*ru#Loss Distance Clutter offsets are a fi!ed end point correction factor that improves the correlation between measured and predicted pathloss. This improves the Standard eviation appreciabl"# but ta'es into consideration onl" the clutter t"pe that the pathloss is being computed for and doesn;t ta'e into consideration the loss due to the different clutter t"pes in the path of propagation. Through Clutter 6oss is the additional loss attributed to the clutter t"pe that the signal propagates through. The total thru-loss for a prediction point is calculated b" e!amining the clutter l"ing between the mobile towards the base over a defined distance# the Through 6oss distance# d. (hen calculating thru-loss# the individual bins are weighted linearl" so that the ones closest to the mobile have greatest effect# and the ones at point d have a minimum. The value of Through Clutter 6oss would var" for different environments# and depends largel" on the clutter through which the signal has alread" passed through. The effects of the clutter t"pe in the path tend to have a residual effect on the value of the Through Clutter 6oss parameter. %t is not a $uantitative measure of the additive loss associated with the clutter t"pe# but rather represents a value that could shape the straight line better in order to fit the measured data# and hence ma" not be intuitivel" assigned or predicted. Through Clutter istance represents the distance from the mobile towards the base station# through which the signal penetrates through the clutter. The remaining distance# the signal is assumed to propagate above the clutter. The Through Clutter 6oss and istance algorithm wor's as follows9 • •
•
Through Clutter 6oss is added to the computed pathloss after appl"ing a weighing factor. The weighting is linearl" applied# with a weighting factor of 0.3 for the bin closest to the Mobile Antenna and a weighting factor of 3.3 at the bin that is at a distance defined b" Through Clutter istance. The Clutter +ffset is used as an end point correction factor to balance the effect of Through Clutter 6oss in order to minimi*e the mean error.
Though Clutter algorithm provides a smooth transition# or averaging effect of patloss between clutter areas. )or e!ample# consider a water edge ne!t to a tree line. The loss does not 8ump from ?d& to -d& immediatel"# but graduall" decreases bin-b"-bin. Through-Clutter models this more accuratel" than offset effects. %t is similar to path profile algorithms found in other propagation tools.
)or %nternal ,se +nl"
:reliminar" v0.3
Page 31
:age 0
How-To Guide for Model Calibration
The following e!ample illustrates the operation of Clutter Through 6oss correction. Assume the following9 &in Si*e9 033!033 meters Through Clutter istance9 0333 meters Total Through Clutter 6oss correction 3 P 3.3 P 3.0 P 3. P 3.? P 3.1 P 3. P 3.? P 3.?5 P 3.@ P 0.3 ?.1? d&.
T
Moile
0
0.
0.
/
/
/
/
/
0. 0 4 0 5 0 6 0 7 0 8 0 9 1 0
lutter Ty$e
Weig
How-To Guide for Model Calibration
The following e!ample illustrates the operation of Clutter Through 6oss correction. Assume the following9 &in Si*e9 033!033 meters Through Clutter istance9 0333 meters Total Through Clutter 6oss correction 3 P 3.3 P 3.0 P 3. P 3.? P 3.1 P 3. P 3.? P 3.?5 P 3.@ P 0.3 ?.1? d&.
T
Moile
0. 0. 2 1 0 0.0 0.12 6 0
- orest 6 d/'
/
/
/
/
/
0. 0.4 0.5 0.6 0.7 0.8 0.9 1.0 3
lutter Ty$e
Weig ht
0. 0.4 0.5 0.360.420.480.9 1.0 Through Loss 3
Loss= 6 dB/km * 100/1000 mts * 0.7 = 0.42 dB
/ - /uildings 10 d/' T+roug+ lutter )istan%e - 1' )or %nternal ,se +nl"
:reliminar" v0.3
:age
Page 32
Gloar of Term •
•
•
•
•
&ins - the mapping resolution of the tool# usuall" 1!1 meters. The smallest measurable unit in the modeling tool. Several sample points could be located in a $in. C( Measurements - a generic term used to describe propagation testing done with an unmodulated carrier# supported from a temporar" TF facilit"# i.e.# usuall" low power and omni transmit. )iltering - removal of select data points within a sector file. %ntercept
Gloar of Term •
•
•
•
•
•
•
•
•
•
•
•
•
&ins - the mapping resolution of the tool# usuall" 1!1 meters. The smallest measurable unit in the modeling tool. Several sample points could be located in a $in. C( Measurements - a generic term used to describe propagation testing done with an unmodulated carrier# supported from a temporar" TF facilit"# i.e.# usuall" low power and omni transmit. )iltering - removal of select data points within a sector file. %ntercept
Page 25
(ui> Guide for *SS&T Model Tuning
To calibrate a macrocell model# perform these initial steps9 •
•
•
%nspect the drive test data to verif" its validit" and filter out an" erroneous data. Ensure that sufficient data points are available for each clutter class. %n most situations it is desirable for the data to be evenl" distributed with respect to log /distance2 from the site# clutter classes and the sectors. Enter a set of default values as an initial step.
=ough Calibration of the Standard Macrocell Model Having performed the initial recommended steps# use these recommended steps as a guide to roughl" calibrating the standard macrocell model9 •
•
•
•
•
•
•
•
•
6oad one or more drive test files and use the filtering to remove $uestionable data and get an unbiased data set. )or e!ample# filter out readings with a signal level below the noise floor or clutter t"pes with too little data to be statisticall" meaningful. erive a estimate of Slope 7alue /K2 from a plot of the =eceived 6evel vs. the 03 log /distance2 using the Measurement Graph facilit". Then fine tune this value. Ad8ust the '0 parameter to a value# which will lower the mean error to 3. (hen the anal"sis report shows a positive mean error# it means the propagation model is pessimistic when compared to the drive test data b" the reported value. %n this case# "ou should lower the '0 value b" the reported amount. (here a negative value is reported# the opposite applies. iffraction effects /'>2 occur onl" when there is no 6ine of Sight from the site to the mobile. Therefore to determine the '> parameter# filter the dataset to include onl" the non-6+S and a value determined using the process described in the above section. As a rule of thumb if the mean error is greater than 3# decrease '> otherwise increase it. Modif" the filter to its original setting /to include 6+S data as well in the anal"sis2. =ead8ust the '0 value if the reported mean in the anal"sis report has increased or decreased after the '> change. Ad8ust the ' value# again using the process in the above section. %t is useful to view the graphs and the Signal Error plot on the Map 7iew to identif" trends with successive parameter changes. =ead8ust the '0 value if the reported mean in the anal"sis report has increased or decreased after the '> change. Ad8ust each clutter offset in turn tr"ing to get the mean error of that particular clutter to 3.
•
Modif" the '# '? and '1 parameters until the reported error is lowered.
•
ow "ou can fine tune the model.
)ine Tuning the Standard Macrocell Model (hen "ou have performed the initial and rough tuning steps# use these recommended guidelines when fine tuning the standard macrocell model. The ob8ective is to identif" what ma" be causing the differences between the propagation model and the actual drive test data and act on minimi*ing the error. ,se the anal"sis# filtering and graph features to help "ou. %nvestigate9