Measurement Data

Consideration of Measurement Data in WinProp

AWE Communications GmbH Otto-Lilienthal-Straße 36

D-71034 Böblingen [email protected]

Issue Date Changes

V1.0 June 2011 First version of document

Consideration of Measurement Data June 2011

© AWE Communications GmbH. All rights reserved 1

Content 1 Motivation ..................................................................................................... 2 2 Import of Measurement Data ......................................................................... 3 3 Modification of Measurement Data.................................................................. 5 4 Comparison of Prediction and Measurement .................................................... 7 5 Calibration of Prediction Models with Measurement Data.................................. 9 6 Further Information ..................................................................................... 10 7 References .................................................................................................. 10



1 Motivation The most important criterion to measure the performance of a wave propagation model is the accuracy (in comparison to measurements). This accuracy can be analyzed by comparing simulation results with corresponding measurement campaigns. Therefore ProMan offers the possibility to import and compare measurement data with prediction results. Besides analyzing the accuracy of prediction models, measurement data can also be used to calibrate wave propagation models [1] for specific environments.

Figure 1: Comparison of measurement and prediction along a trajectory



2 Import of Measurement Data Measurement data can be imported from an ASCII file in two different ways.

• The first option is to select ‘New Measurement File...’ from the ‘File’ menu of the ProMan application. After choosing the corresponding database(s) to describe the scenario a new dialog opens (see Figure 2). In this dialogue the extension, the resolution as well as the height of the new measurement file have to be specified.

Figure 2: Definition of measurement database

• A second option is to load the measurement data into an existing prediction result. Here the user should select ‘File -> Import -> Import Data’ from the File menu. In this case there is no need to define the extension, the resolution and the height of the measurement file, as this data is taken from the prediction result. The advantage of this second option is that the location and all other settings of the transmitter (e.g. antenna pattern, antenna orientation, etc.) are correctly defined and stored in the measurement file. The disadvantage is that the user has to compute a prediction first (but very often predictions are already available).

Afterwards, the ASCII file containing the measurement values has to be selected. This procedure is identical for both import options listed above.

Figure 3: Measurement Import dialog



As the coordinates used in the ASCII file must match the ones of the selected database, the user has the possibility to convert and adjust the coordinates of the measurement data during the import process. Optionally, measurement values can be imported along a line, which has to be drawn with the mouse. In this case, the ASCII file contains only measurement values without corresponding location coordinates. The data contained in the currently active result view can be deleted automatically before importing the new measurement data by checking the corresponding option in the lower left part of the dialog box. This option should be selected if the measurement data is imported into a prediction result (second import option described above). In the right part of the dialogue, the type and the format of the data contained in the ASCII file can be specified. The measurement data to be imported has to be in a tabular format with longitude (x), latitude (y) coordinates and a corresponding measurement value. Figure 4 gives an example of the required file format. The order of the columns to be imported can be specified within the import dialog (cf. Figure 3). In case the ASCII file contains additional information, such as a header section, keywords indicating the data section to be imported have to be entitled in the lower right part of the import dialog. These keywords can be defined arbitrarily (cf. Figure 3). If only a header section is contained, one keyword indicating the start of the measurement data is sufficient. All data located before this keyword will be ignored. In case additional data is also located after the measurement data, two keywords are required indicating start and end position of the measurement values.

Figure 4: ASCII file with measurement data



Depending on the specified resolution of the measurement data contained in the ASCII file, several measurement points might fit into a single raster point of the created measurement database. In this case ProMan imports only the last measurement point which fits the raster point, whereas all other values for this pixel will be neglected.

3 Modification of Measurement Data Sometimes it might be useful to modify the imported measurement data. For example, if an additional attenuator was used during the measurement campaign or if the measurement data contains errors. A further post-processing of imported measurement data might be required if the coordinates of the measurement points are not as accurate as required by the prediction models (e.g. in urban scenarios measurement points close to buildings might now be inside the buildings due to localization errors during the measurement). Here the location of the measurement points should be modified after the import. Sometimes a few measurement points should be deleted because they were measured at locations which are not modelled correctly in the database of the scenario (e.g. measurement route is under the crossing of railway lines or in tunnels). These modifications of measurement data can be done using ProMan's paint tools or mathematical operations from the Edit menu. In case measurement values are scattered along a measurement route for example, ProMan offers the possibility to align those values along an arbitrary line, which can be drawn with the mouse. The tool 'Align Data Along a Line' can be found in the Edit menu ('Edit -> Align Data Along a Line'). After selecting the tool, the user has to draw a polygonal line along which the data shall be aligned. As soon as the polygonal line is finished with a right click, a dialog opens, where a measurement radius, i.e. a catch radius for the values has to be specified. All values which are contained in the matrix and which are located within this radius will be automatically moved onto the specified line. Single pixels or even whole areas with measurement data can be changed or deleted easily by specifying a new value or ‘Value not computed’, respectively within the paint tool options (see Figure 5). Depending on the selected paint mode, the pixels specified with the mouse are modified accordingly. ‘Add Value to Pixel’ will add the specified value to the already existing values in the measurement database which are selected with the mouse. The drawing mode ‘Cut and Move Pixels’ for example offers the possibility to shift measurement data to a new location within the measurement database. The pixels to be influenced can be selected using the mouse and one of the paint tools, which can be chosen in the lower part of the dialog shown in Figure 5. The different drawing tools offer the possibility to adjust measurement values along a user defined line or within an arbitrary polygonal area for example.



Pixels which are undefined can be excluded from the drawing operations by un-checking the option at the bottom of the dialog. Please refer to the ProMan Manual [2] for more details about the drawing options.

Figure 5: Modification of measurement data in ProMan



4 Comparison of Prediction and Measurement Predicted result values can be easily compared to measured data with ProMan. First of all, the measurement data has to be imported into a measurement file for the corresponding simulation project (cf. chapter 3). In order to do a comparison between predicted results and the corresponding measurement, measurement values can be subtracted from the predicted ones. Therefore ‘Subtract Data -> Value (File, linear)’ has to be selected from the Edit menu after the prediction result has been opened. The file containing the values to be subtracted from the predicted values, i.e. the imported measurement data has to be selected afterwards. ProMan will show a new result file with the difference between simulation and measurement then. Figure 6 gives an example for a measurement comparison in an urban environment.

Figure 6: Prediction result (left), measurement data (centre) and resulting difference (right) for an urban city

Measurement data and prediction results can be compared very easily using external applications, too. WinProp's result data can be exported along an arbitrary polygonal line via File menu ‘File -> Export Data (Polyline)’. This makes it possible to export the predicted results along the measurement route. The polygonal route can be defined in ProMan with the mouse. The user can therefore either choose to export the currently active result directly by selecting ‘File -> Export -> Export Data (Polyline) -> Export data along polyline’ from the File menu or to create a polygonal mask for multiple exports using ‘Export Polyline (Write to Polyline to File)’ in the same submenu. For the second case the defined polyline is stored in an ASCII file which can be applied for several results. To draw the polyline press the left mouse button at the starting point and keep it pressed until the last point of the polyline is reached. To insert further points in-between click the right mouse button. The resulting ASCII file (*.epl) containing the specified polygonal track has the following format.

Figure 7: File format of polygonal measurement route



If the coordinates of the measurement route are already available in an ASCII file which has the same format as the ProMan polyline (cf. Figure 7) the extension of the file can be named to *.epl in order to use the available file directly to export the predicted data along the given measurement route. Once both measurement data and predicted data are available for the same coordinate positions along the measurement track, the user can do comparisons with arbitrary applications. The following figure depicts a graphical comparison of measurement and prediction data using a two dimensional graph generated with Microsoft Excel®.

Figure 8: Comparison of measurement and prediction along a track with MS Excel®

Further statistical information about the comparison can be obtained from the Information Page by selecting ‘Information’ from the Analysis menu.

Figure 9: Statistical information of measurement comparison



Additional graphical evaluations can be done in terms of histograms, probability density functions (PDF) and cumulated probability density functions (CDF) by using the Analysis menu of the ProMan application. An example showing the distribution of the discrepancies is depicted in Figure 10.

Figure 10: Histogram (left), PDF (centre) and CDF (right) of the difference between

prediction and measurement data

Values of the graphical results depicted in Figure 10 can be exported to ASCII files by selecting ‘File -> Export -> Export Data (Area)… -> Export ASCII (All)’ from the File menu or by right clicking and choosing ‘Save Values in ASCII-File’ from the context menu. This offers the possibility to post-process the evaluation results with other applications.

5 Calibration of Prediction Models with Measurement Data In order to achieve better prediction results, especially for empirical wave propagation models, it is often required to calibrate the models. The equations of empirical propagation models are based on the evaluation of numerous measurement campaigns. Thus these models do, in contrast to deterministic propagation models, not reflect physics but rely on empirical values, which have to be calibrated for specific environments. This calibration can be done automatically based on measurement data imported and stored in a measurement file. More details about the automatic calibration of prediction models can be found in a separate application note [1].



6 Further Information For further information you are invited to visit AWE Communications’ website http://www.awe-communications.com or to send an e-mail to [email protected]

7 References [1] Application Note, “Automatic Calibration in WinProp”, AWE Communications

GmbH, November 2010. [2] User Manual, “ProMan Manual”, AWE Communications GmbH, June 2011.

Documents

Measurement Data