A comparison between received power level estimation methods for practical indoor radio network planning

Transcription

1 A comparison between received power level estimation methods for practical indoor radio network planning Supervisor: Professor Sven-Gustav Häggman Master s thesis made for Siemens Osakeyhtiö

2 Contents Introduction Presentation of tools and methods Results Summary of own contribution Conclusions

3 Introduction Problem statement & Research question Current coverage planning methods are focused on Macro- and micro cell planning What kind of method(s) should be used for predicting cell coverage in indoor environment. Scope of work This master s thesis study concentrates on evaluating different received power level prediction methods for practical indoor radio network planning

4 Presentation of tools and methods Test receiver TEMS Light 3.0 Commercial available tool Developed by Ericsson Erisoft Used in field trial performed in Siemens corporate campus Munich GSM Pico/micro cell trial operating on 900 MHz Indoor radio propagation tool WinProp Used in computer based received power prediction trial Commercial available tool Developed by AWE communications Propagation models under study COST Multi-Wall Model Intelligent Ray Tracing Two building databases generated for same building WinProp default wall properties Custom wall properties (used wall properties are from WinProp material database)

5 Indoor propagation models used in study RX L 2 L 1 r TX COST Multi-Wall model Considers direct ray between transmitter and target pixel Different wall transmission losses are considered RX TX Intelligent Ray Tracing A variation of standard ray tracing Building database is pre-processed by dividing building in to tiles and visibility relationships are determined before calculation. Considers different wall properties

6 Results (field trial) Received power measurement results were obtained from 520 individual measurement points on 9 th floor respectively 699 measurement points on 10 th floor.

7 Results (computer based prediction) Example 10th floor predicted with COST Multi-Wall Model WinProp default wall properties used

8 Comparison of models Prediction accuracy of the prediction models has been evaluated and visualised in the following way Average accuracy Average offset (measured - calculated received power) Standard deviation of the offset Correlation between measured and calculated results Difference plots measured - calculated received power plotted on building layout Visualises local variation of the offset Comparison plots The measured power level for each pixel is sorted in ascending order and plotted. The corresponding prediction results are plotted in the same graph. The comparison plot visualises how the prediction accuracy fluctuates over the dynamic range of the measured received power.

9 Comparison of models 9th floor default wall properties CMWM IRT332 IRT442 IRT552 IRT663 Average offset [db] Standard deviation of the offset [db] Correlation between results th floor custom wall properties CMWM IRT332 IRT442 IRT552 IRT663 Average offset [db] Standard deviation of the offset [db] Correlation between results

10 Comparison of models Received power difference plot for 10 th floor (measurements prediction CMWM default values).

11 Comparison of models Received power level [dbm] TEMS Light Prediction CMWM default wall properties Comparison plots

12 Comparison of models 9th floor accuracy vs. calculation time Standard deviation [db] Calculation time [s] Standard deviation of the offset [db] default wall properties Standard deviation of the offset [db] custom wall properties Calculation time [s] default wall properties Calculation time [s] custom wall properties CMWM IRT332 IRT442 IRT552 IRT663 Prediction model 0

13 Preparations for measurement [h] Comparison of methods Measurement on site CMWM default wall prop. 3 h - Measurement [h] 3 h - Building database generation [h] Inclusion of custom wall properties [h] Method IRT442 default wall prop. CMWM custom wall prop. IRT442 custom wall prop Calculation time [h] Total time to perform [h] Accuracy (average offset) [db] Accuracy (standard deviation of the offset) [db] ±

14 Summary of own contribution Performed field measurement campaign in hard partitioned office environment individual measurement points was recorded on two floors Created a 10 storey office building database for prediction tool Evaluated two prediction models Compared predictions with measurements

15 Conclusions Intelligent Ray Tracing model achieves better prediction accuracy than COST Multi-Wall model. The accuracy comes with the cost of longer calculation time. The calculation time with COST Multi-Wall was in the range of seconds while the Intelligent Ray Tracing calculation time was in the range of hours. However, the main cost in terms of time was the building database generation. The building database used in this study took 20 hours to complete. Prediction accuracy was improved by introduction of customised wall properties for both prediction models. However, the addition of custom wall properties to the building database is time consuming.

16 Conclusions (continued) The measurement method on site can be combined with site survey for installation purposes to reduce overall costs for implementing indoor networks. On the other hand calculation based method enables networks to be designed for buildings under construction. The measurement method and calculation method are not mutually exclusive, instead the methods complement each other. The decision to use one of the methods should be made depending on situation.