Anda di halaman 1dari 24

Optimizing Coverage and Capacity Using Actix ACP

and Veritune
Actix Americas 05 User Conference
October 11, 2005

presenter Jeff Atkins

Americas User Conference 11-12 October 2005


Contents

Where is there opportunity to improve existing


coverage and capacity optimization tools?
How do different approaches to achieving
improvement compare?
What are the key considerations in developing
practical solutions and processes?
Where is there opportunity to improve
existing coverage and capacity
optimization tools?

Or why do model
trains run on time?
The antenna optimization process

To meet our objectives:


Reduce drive testing Field drive drive 03
Optimized!
Validation 02
01
Reduce antenna changes
and rework
Reduce risk of network
quality degradation
Reduce iterations and time to
market
Antenna
To a lesser degree reduce
changes KPI analysis
analysis time and costs
in field
cost

time Estimate changes to


antenna coverage
Traditional approaches to predicting the
impact of antenna changes

Experience
Dependant upon experience
and local knowledge required
rework can vary between
10% and 80%(!).
Tilt spreadsheets
Indicating range of main lobe
from site for a given antenna
and site configuration
Planning tools
Strategies to improve accuracy
The coin exercise is commonly used to teach
the use of a compass.
The objective is to follow a series of bearings
and distances to find a coin in a field.
Small errors at each stage accumulate, leading
to a potentially large overall error, compared to
that required to find the target.
To improve there are 2 basic strategies
Shortcut Cheat by starting from a later known
point. I.e. simplify the problem. Shortcut

Stage improvement - Improving the accuracy of


one or more stages.
Both of these approaches can be applied to RF
prediction of antenna configuration changes Stage
Improvement
How does a planning tool work? (1)

Antenna Masking
Required inputs: Map database, antenna
performance and configuration

Modified Interfering
Cell Cells

Interference Generation
Link Budget Path Loss Required inputs: Map database, model
Required inputs: Required inputs: Map diffraction/penetration/ reflection assumptions, antenna
Equipment database, model performance and configuration, traffic geographical
performance & diffraction/penetration/ distribution, building penetration, interferer link budget, traffic
configuration. reflection assumptions. modelling, power control, RRM modelling, mobile location in
buildings.
How does a planning tool work? (2)

Antenna
Masking

Traffic

RRM
Calculation of
signal strength or
interference for a
Propagation Link Budget single sector
How does a planning tool work? (3)

Sector 5

Sector 4

Sector 2
Calculation of
signal strength
or interference
for a single
Sector 1 Sector 3 location
Using measured data is the answer

It allows us to improve the accuracy of our solution


identification process
As a result we will be able to:
Reduce rework
Reduce drive testing
Reduce implementation costs
Reduce analysis time
Improve service quality
Reduce number of sub-optimal changes
Reduce drops / blocks due to unfixed issues
How do different
approaches to
achieving
improvement
compare?

Or how do football coaches


decide when to use different
formations?
Automatic per-site propagation model
calibration

Signal Original planning Propagation model Strategy


Strength predicted signal calibrated
automatically on a
Stage improvement - propagation
per- sector basis Accuracy
Expected 8dB
Strengths
Automation reduces time and
costs
Per-site calibration
Automation allows more
parameters to be calibrated
Weakness
No user Interaction to identify
Per-sector input data anomalies.
Measured calibration Care over reasonability of output
Data predicted signal parameters required.
Points
Statistical validity of data.
Curve fitting to drive route/
availability of validation data.
Distance from Sector
Interpolation of predicted data to
measurements
Original planning Original propagation
Signal
predicted signal model interpolated Strategy
Strength
onto measured data
Stage improvement -
propagation
Accuracy
Expected <8dB
Strengths
Automatic, little time or
cost
Per-site consideration of
Measured Interpolated
Data signal measured data
Points
Weakness
Trends in the original
Distance from Sector model may be inaccurate
A route based approach

Antenna Masking
Required Inputs: Map database, antenna
performance and configuration

Modified Interfering
Cell Cells

Link Budget Path Loss Interference Generation


Measure Signal Strength. Required Inputs: Map database, antenna performance and
Not Required.
configuration of changed interfering cells.
Delta analysis based on measurements

Signal
Strength
Strategy
Shortcut
Accuracy
Antenna masking applied
directly to measured Expected 3.5 to 5dB
data
Strengths
No calibration time or
cost
Accurate
Measured
Data
Points
Weakness
Distance from Sector Only possible where
measurements exist
Veritune

At the last user group we


introduced Veritune and
preliminary validation results.
Over the past year we have
performed extensive validation,
across multiple sites,
environments and countries.
In addition there have been a
number of improvements to the
algorithms in used.
Technical validation approach
Predictions were compared to measurements across as is the
case for propagation model validation.

Validation Control Clusters/Environments


Set Set 1. Dense Urban
Number of 33 20 River valley town in steep valley with significant
water. Buildings predominantly 4-5 stories with
Sectors 10-15 story buildings in centre

Number of 5324 3221 2. Urban


Data Bins Residential/commercial area in relatively flat
topology, typically 2-3 stories
Antenna 12m to 48m
Height Range 3. Rural/Suburban A
Rural/suburban/small town environment with 2
Distance from 100m to 10km story buildings and predominantly agricultural
Site Range fields, in a gently undulating terrain.

Angle from Up to 100 degrees 4. Rural/Suburban B


Azimuth Rural/suburban/small town environment with 2
Range story buildings and predominantly agricultural
fields, in a terrain of rolling hills
The validation results
Standard deviations significantly improve upon typical
propagation model accuracies (benchmark 8dB top of the below
graph).
The mean error was within 1dB of the control set for all clusters.
8

7
4.8dB
Standard Deviation (dB)

6 without 2
outlying
5
sectors
Control
4
Veritune
3

0
Rural/Suburban A Rural/Suburban B Urban Hilly dense urban
How consistent are the results?

A majority of sectors exhibited consistent performance:

20
18
16
Number of Sectors

standard deviation
14
mean
12
10
8 2 outlying
6 sectors
4
2
0
0 to 1 1 to 2 2 to 3 3 to 4 4 to 5 5 to 6 6 to 7 7 to 8 8 to 9 9 to 10
Range (dB)
What are the key
considerations in
developing
practical solutions
and processes?

Is this really the best way to do


things?
(Picture Testing Golf Clubs by Heath-Robinson)
Key components of practical solutions

We have discussed the accuracy of prediction


A number of other key factors influence the
development of practical solutions:
Sampling accuracy
Degree of automation
Use scenario
Complexity/simplicity
Delta and model based solutions compared
Delta Based Model Based
Improvement Approach Shortcut Stage Improvement
Accuracy Typically 4.5dB standard Down to 7dB standard
deviation deviation with improvements
Sampling Like-for-like statistics compared Inherent difference in statistics
to measurements calculation
Information Coverage Certainty about measured Uniform data but less certain
areas
Automatic vs Interactive Interactive Data in uniform grid suitable
Data not uniform - unsuitable for automation
for full automation Possibility of breakout from
Local knowledge and local optima
constraints part of process Local knowledge and
constraints require prior input
and post-validation
Use Scenarios Fine Tuning Initial Planning
Course optimization Coarse optimization
Validation of coarse/break-out Break-out optimization
optimization
Conclusions

Costs are most effectively controlled by reducing the number


of optimization iterations through improving prediction
accuracy.
Where only one iteration is allowed, service quality is
maximized by improving prediction accuracy.
The greatest improvements in accuracy can be achieved by
using delta predictions in tools such as Veritune.
In planning and break-out optimization use cases,
automated model-based approaches such as ACP provide
the most potential for success.
Both ACP and Veritune approaches should be considered to
get the most value out of optimization processes.
Thank you for your attention.
Jeff Atkins
Jeff.atkins@actix.com

Anda mungkin juga menyukai