Professional Documents
Culture Documents
Training Programme
1. SPM Calibration Concepts 2. Guidelines for CW Measurement Surveys 3. Working with CW Measurements 4. Automatic Calibration Method 5. Analysing the Calibrated Model 6. Calibration Process Summary
Forsk 2010
Slide 2 of 54
Requirements
Quality target
Forsk 2010
Slide 3 of 54
Goal
SPM
Macrocell statistical propagation model Based on empirical formulas + set of parameters Default values in new projects
Propagation depends on
Frequency Area type (urban, suburban, rural, etc.) Geography (relief, vegetation, climate, etc.)
Forsk 2010
Slide 4 of 54
Requirements (1)
Accurate and recent geo data
DTM and clutter resolution <= 25m for urban areas DTM and clutter resolution <= 50m for rural areas Vector map with major roads
CW measurement surveys
Selection of representative areas for different area types Site selection for each (area type frequency band) 8 recommended (minimum 6) sites for calibration, 2 sites for verification Survey route study Perform CW surveys by fully following guidelines
Forsk 2010
Slide 5 of 54
Requirements (2)
Drive Test data
Possible but not recommended Conversion to CW measurements is needed
Downside
Real network is measured interferences
Directional antennas: accuracy of pattern, only a few points are relevant Several frequencies are measured Low sampling rate for each measured station (lee criterion cant be met) Signal measured over a short distance from the transmitter (model will not be calibrated for interference evaluation)
Quality Target
Calibration sites
Global mean error on calibration sites < 1 dB Global standard deviation on calibration sites < 8 dB Mean error on each calibration site < 2.5 dB Standard deviation on each calibration site < 8.5 dB
Verification sites
Global mean error on verification sites < 2 dB Global standard deviation on verification sites < 8.5 dB
Forsk 2010
Slide 7 of 54
Training Programme
1. SPM Calibration Concepts 2. Guidelines for CW Measurement Surveys 3. Working with CW Measurements 4. Automatic Calibration Method 5. Analysing the Calibrated Model 6. Calibration Process Summary
Forsk 2010
Slide 8 of 54
Radio criteria
Forsk 2010
Slide 9 of 54
Inspection on site
Possibility to rig omnidirectional antenna (no obstacle on any side) Panoramic photographs Report site details: precise height, coordinates
Forsk 2010 Confidential Do not share without prior permission Slide 10 of 54
Clutter
Routes through major clutter classes Avoid forests and lakes between transmitter and receiver
Maps
Supply vector maps of survey routes to import in Atoll Check that survey routes match roads (vector data or scanned maps)
Forsk 2010
Slide 11 of 54
Equipment data
Antenna patterns + downtilt + azimuth (if not perfectly omnidirectional) Antenna height + transmit power + transmission gain and losses Receiver height + sensitivity + reception gain and losses
Forsk 2010
Slide 12 of 54
45 68 90 113
Training Programme
1. SPM Calibration Concepts 2. Guidelines for CW Measurement Surveys 3. Working with CW Measurements 4. Automatic Calibration Method 5. Analysing the Calibrated Model 6. Calibration Process Summary
Forsk 2010
Slide 14 of 54
Forsk 2010
Slide 15 of 54
Coordinates of points
Altitude, Clutter classes and heights, Distance, etc. read from the Geo data
Forsk 2010
Slide 16 of 54
For predictions along the CW measurement path, you can either use Existing path loss matrices or recalculate them by choosing a specific Propagation model
Forsk 2010
The points can be displayed according to any data contained in the measurement Table Confidential Do not share without prior permission
Slide 17 of 54
To compare statistics between measured and predicted signal levels. Note: This can also be run from top folders.
Forsk 2010 Confidential Do not share without prior permission Slide 18 of 54
Forsk 2010
Slide 19 of 54
Tool to filter the data path in an more advanced way than in the Filter dialogue available at the folder level (previous slide)
Tool to exclude some points from the measurement path according to a drawn polygon (all points within the polygon will be filtered out)
Forsk 2010
Slide 20 of 54
AFTER
Smoothing can be used to limit fading effect Smoothing keeps the number of measurement points unchanged Smoothing cannot be used to average gross CW measurements
Forsk 2010 Confidential Do not share without prior permission Slide 21 of 54
Forsk 2010
Slide 22 of 54
Training Programme
1. SPM Calibration Concepts 2. Guidelines for CW Measurement Surveys 3. Working with CW Measurements 4. Automatic Calibration Method 5. Analysing the Calibrated Model 6. Calibration Process Summary
Forsk 2010
Slide 23 of 54
Initial model
Calibration wizard
Final model
Forsk 2010
Slide 24 of 54
CW Measurements Pre-processing
Correspondence between Measurements and Geo data
Projection checking Check that CW measurements match roads
Surrounding checking Check with panoramic photographs that there is no obstacle Option of setting an angle filter to avoid attenuation due to obstacles
Forsk 2010
Slide 25 of 54
CW Measurements Pre-processing
Filtering
Available at the Folder level for each site Will be applied to all the measurement paths in that folder
Clutter Classes filtering Distance, Measurements values and Azimuth filtering
CW Measurements Pre-processing
Distance filtering (Min Distance / Max Distance)
Typical min value: 200 m (not representative of mean propagation) Typical max value: 10 km (rural area)
Azimuth filtering
To remove points in a certain angle
Filtering assistant
In addition to the Filter located at the Folder level, you can define more precise filtering depending on the CW measurement file
Forsk 2010
Slide 27 of 54
CW Measurements Pre-processing
Filtering assistant (1/2)
Display of M = f ( 10log(D) ) Selection rectangle simultaneous Signal/Distance filtering
Signal/Distance filtering according to the selection rectangle Possibility to keep the selected points or to exclude them
Selection Rectangle
Forsk 2010
Slide 28 of 54
CW Measurements Pre-processing
Filtering assistant (2/2)
Remaining points after the Distance, Signal level, Azimuth and Clutter classes filtering
Forsk 2010
Slide 29 of 54
CW Measurements Pre-processing
Final filtering (1/2)
Display each CW measurement according to their Measured signal level Check that propagation loss is spatially homogeneous
Forsk 2010
Slide 30 of 54
CW Measurements Pre-processing
Final filtering (1/2)
Removing which points? Sudden drop of signal level Suspicious areas How? Delete from the CW measurement table Draw Filtering zones
Forsk 2010
Slide 31 of 54
Verification stations
Stations so that measurements are inside covered area (not at edges) Major part of their covered areas are also covered by calibration stations
How many
If 8 measured stations If < 8 measured stations 6 for calibration; 2 for verification all stations used for calibration verification performed with same stations
Forsk 2010
Slide 32 of 54
Initial Model
General SPM formula Lmodel = K1 + K 2 .log (d ) + K 3 .log (HTxeff ) + K 4 .Diffraction Loss + K 5 .log (d ).log (HTxeff ) + K 6 .(H Rxeff ) + K7 .log (H Rxeff ) + K clutter .f (clutter ) + K hill, LOS Ki values
Let K6 = 0 Others will be calibrated
Initial Model
Max distance
Forced to 0 during calibration If >0 no continuity ensured
Kclutter
= 1 is recommended Multiplying factor of clutter losses
Minimum loss
= Free space loss Avoid unrealistic values
Profiles
Radial optimisation quicker
Forsk 2010
Slide 34 of 54
Initial Model
Heights of Clutter taken or not into account in Diffraction If you have a Clutter Heights file
Then put 1-YES in the box
Initial Model
Clutter Classes Losses can be calibrated
You need to define the max distance from the Receiver (towards the Transmitter) for which the different clutter classes will be considered Choice between 4 types of Weighting functions (Uniform, Triangular, Logarithmic, Exponential)
f (clutter ) = Li w i
n i =1
Uniform Triangular Logarithmic Exponential
wi=f(d'i)
wi
d'i
Forsk 2010
Slide 36 of 54
Initial Model
Reference model
Create a Reference model containing all the previous settings Duplicate this Reference model for each calibration, and give it a relevant name When duplicated, choose an appropriate name and pay specific attention to:
Methods used for Diffraction and Effective Antenna Height calculation value of Kclutter Hilly terrain correction Heights of Clutter considered or not in Diffraction Clutter Range and associated Weighting function
Forsk 2010
Slide 37 of 54
Calibration Wizard
Automatic calibration overview
Algorithm based on solving a least-squares problem Calculation of the best solution in terms of root mean square : Simple, fast and reproducible procedure
RMS = 2 + M 2
First Step
Selection of calibration stations
Forsk 2010
Slide 38 of 54
Calibration Wizard
Second step (1/2)
Selection of the Parameters to calibrate Possibility to modify their ranges
Forsk 2010
Slide 39 of 54
Calibration Wizard
Second step (2/2)
Recommended ranges Constant K1 K2 K3 K4 K5 K7 It is recommended to leave K6 to 0 Min 0 20 -20 0 -10 -10 Max 100 70 20 1 0 0
Forsk 2010
Slide 40 of 54
Calibration Wizard
Final step
Display of Before and After Parameters values and Statistics Commit will update the model you are calibrating with the new values of Ki, height and diffraction methods as well as the Clutter Losses
Forsk 2010
Slide 41 of 54
Final Model
Extrapolate non-calibrated clutter losses (1/2)
Non-calibrated clutter classes must not have their clutter losses left to 0 Could lead to high error where these classes are present Must be extrapolated from: Calibrated clutter losses (from other propagation model) Typical losses (here centred on the Urban class) Clutter class
Dense Urban Woodland Urban Suburban Industrial Open in urban Open Water
Typical loss
from 4 to 5 from 2 to 3 0 from -5 to -3 from -5 to -3 from -6 to -4 from -12 to -10 from -14 to -12
Forsk 2010
Slide 42 of 54
Final Model
Extrapolate non-calibrated clutter losses (2/2)
Centre clutter losses Relative difference between clutters kept unchanged Use K1 to balance Example:
After calibration, model centred on suburban: K1=17.4 Losses: Dense Urban = 6.5 Wood = 5.7 Urban = 3.5 Suburban = 0 After centring, new values: K1=20.9 Losses: Dense Urban = 3 Wood = 2.2 Urban = 0 Suburban = -3.5
Apply scaling factor Adapt typical losses (or calibrated ones coming from other model) to the calibrated model
-12 C 0 e n t Urban r e d 0
calibrated
Open
-8
Extrapolated
Forsk 2010
Training Programme
1. SPM Calibration Concepts 2. Guidelines for CW Measurement Surveys 3. Working with CW Measurements 4. Automatic Calibration Method 5. Analysing the Calibrated Model 6. Calibration Process Summary
Forsk 2010
Slide 44 of 54
Forsk 2010
Slide 45 of 54
Possibility to run the Statistics on all the Measurement paths, or on specific ones
Forsk 2010
Slide 46 of 54
The calibration wizard will attempt to bring the correlation as close to zero as possible. The results will be a correction value that will be added or subtracted to the initial Ki value in the model
Commit will apply the Correction values to the corresponding Ki values Notes: This will not take into account the Ki Ranges
Forsk 2010 Confidential Do not share without prior permission Slide 47 of 54
Forsk 2010
Slide 48 of 54
Forsk 2010
Slide 49 of 54
Forsk 2010
Slide 50 of 54
Training Programme
1. SPM Calibration Concepts 2. Guidelines for CW Measurement Surveys 3. Working with CW Measurements 4. Automatic Calibration Method 5. Analysing the Calibrated Model 6. Calibration Process Summary
Forsk 2010
Slide 51 of 54
Measurements preparation
Sites selection Survey roads Fulfil radio criteria
Make & Average measurements Create Transmitters used for measurements in the Atoll document
With exact configuration (coordinates, antenna type & height, EIRP, losses)
Pre-processing)
Forsk 2010
Slide 53 of 54
THANK YOU!