WaveSight TM - see your net work! Technical Presentation

Transcription

1 WaveSight TM - see your net work! Technical Presentation

2 Table of Contents I I. Network Planning Concerns 1. Calculating coverage and reducing interference 2. Rising numbers of subscribers in GSM networks 3. Additional Needs for GPRS 4. Set-up of UMTS networks II. Propagation History 1. Empirical models 2. Semi-empirical models 3. Deterministic models (Ray-Tracing) III. Deterministic models: WaveSight 1. Physical basics of WaveSight 2. Implemented Parameters and algorithms 3. Data Input 4. Simulation 5. Prediction Outcome (Accuracy and Speed) Slide 2

3 Table of Contents II IV. The added value of WaveSight for network planning 1. Reduced need for measurement and drive tests 2. Better knowledge of Coverage 3. Better Input / better results for frequency planning 4. Rising network quality / better opportunities for fine/tuning 5. Increased speed of network set-up 6. Reduction in the number of needed sites V. Integration of WaveSight into Planning Tools 1. Integrated Planning Tools 2. Integration on Windows Systems 3. Integration on Unix Systems 4. License Protection Slide 3

4 I. Network Planning Concerns 1. Calculating coverage and reducing interference 2. Rising numbers of subscribers in GSM networks 3. Set-up of UMTS networks Slide 4

5 1. Calculating coverage In the early 90 s, the initial concern in term of planning was to ensure a coverage for all areas as fast as possible. After rolling-out their network, the operators had to face one problem : in Urban areas you can easily have shadow areas because of the characteristics of buildings. Finally these last years, the most important challenge became to face the important increase of subscribers, which caused problems of saturation in congested areas. Slide 5

6 2. Rising numbers of subscribers in GSM networks The number of subscribers has continuously raised with an explosion in the year 2000, seeing more than 210,000,000 subscribers in Western Europe and an increase of more than 30%. Now as the amount of users has grown, it came close to the point that network capacities are saturated in some congested areas. Simultaneously, users have rising quality expectations and don t accept dropped calls and bad connections. As the bandwidth allocated to operators is not infinite, the only way to raise network capacity is to add new microcells to increase the maximal amount of users in such areas. The important problem that is caused by adding microcells is that the frequency planning became more complex (a good one allows good frequency re-use) Slide 6

7 3. Additional Needs for GPRS On GPRS (General Packet Radio Service) users use remaining free slots of Base Station for data transmission. The number of available slots for GRPS is scalable. It depends on the circuit load of the cell (number of voice channels) The loading of GSM networks will raise seriously!! And so interference will raise as well. The received data throughput is directly related to the ratio received signal/interference. Bad coverage will result in poor data transmission and unacceptable Quality of Service Badly planned GSM networks will have trouble with GPRS and may also experience problems with voice users. Slide 7

8 4. Set-up of UMTS networks Planning of UMTS is a new challenge, as the technology is completely new, the problems in planning will be different Infrastructure costs will be a significant burden for all operators No frequency planning (frequency reuse = 1) Soft handover will be a feature There is a great uncertainty about the availability and attractivity of killer applications Slide 8

9 4.1 Set-up of UMTS networks UMTS-Planning is done through network simulations. This simulation typically consists of spreading users, engaging different services like voice or data transmission, in the environment. The simulating software will then predict the behavior of the simulated network e.g. uplink, downlink, power These characteristics are directly related to interference and so capacity Simulation is based on propagation (power path-loss prediction) Doing simulations on the basis of bad predictions of the coverage will give nonsense results Slide 9

10 Simulation User distribution Cost-Hata WaveSight Green: OK ; Black: inactive; Red: mobile power outage; Pink: BS power outage; Purple: overload (load factor) Slide 10

11 II. Propagation History 1. Empirical models 2. Semi-empirical models 3. Deterministic models (ray-tracing) Slide 11

12 1. Empirical Models Okumura-Hata This model has been developed from measurements made by Okumura intokyo and published [1] in Then the formula was given by Hata [2] in Advantage : - Needs no building data Disadvantages : - Needs expensive and time consuming calibration - Rough prediction (circular shaped) - the model is not appropriate for planning a sophisticated network Slide 12

13 1.1 Empirical Models Shortcomings : Despite this model has been developed especially for urban areas, the approach needs heavily calibration which is based on measurements and in some complex terrain like very dense cities it is very difficult to carry out. [1] Y. Okumura, E. Ohmori, T. Kawano, K.Fukura: " Field strength and its variability in VHF and UHF land-mobile radio service", Review of the Electrical Communication Laboratory, Vol. 16 N 9-10, Sept. 68, pp [2] M. Hata: "Empirical formula for propagation loss in land mobile radio services", IEEE Transaction on Vehicular Technology, Vol.29 N 3, Aug. 80, pp Slide 13

14 Empirical Models Expression given by Hata : L = log10 (f) log10 (h) - ( log10 (h)) log10 (d) -A -B -C D Where : f = frequency, d = distance transmitter-receiver, h = Base station height A, B, C & D = relief loss, near obstacles loss, correction for mobile height (1.5m) other (rivers, wood, building density) Slide 14

15 2. Semi-Empirical Models Walfish-Ikegami This model has been published for the first time by Bertoni and Walfish [4,5]. Then only the attenuation caused by buildings was implemented. Several ameliorations were added by different authors to improve the calculation formula. Ikegami has added the last reflection on the opposite wall. Advantage : Takes into account terrain and building data Disadvantage : Still needs expensive and time consuming calibrations Slide 15

16 2.1 Semi-Empirical Models Shortcomings : Despite this model takes the environment into account, it is still needed to calibrate it so the accuracy of the predictions is still not satisfying enough. [4] J. Walfish, H.L. Bertoni: "A theoretical model of UHF propagation in urban area", IEEE Trans. on Ant. and Prop., Dec. 1988, pp [5] H. L. Bertoni, J. Walfish: "A diffraction based theoretical model for predicting UHF path loss", IEEE Trans. on Veh. Tech., Vol. 37, Slide 16

17 Semi-Empirical Models Base station Mobile multi-screen effect from base station to last diffraction = Bertoni-Walfish formula 4 : 4 Pr = C * Pt / d the reality Classical Fresnel diffraction α from terrain data base Reflected term (Ikegami) b w n-1 n W What can be computed by theory w2 h Slide 17

18 3. Deterministic Models Since the early eighties, a lot of work has been carried out on physical models, which take into account all the three-dimensional environment. All of these models are only based on physical laws and use different techniques like ray-tracing or ray-launching. Advantages : - Does not need calibration (frequency, antennas, environment ) - More accurate - One model for all uses (macro small and micro cells). - Allows full channel information (Received power, direction of arrival, impulse response, short term fading) Disadvantages : - Long computation time Shortcomings : Few: As the telecom industry is now looking for better accuracy in prediction results to improve the fine tuning of the network, the physical models look at this time the better alternative. Slide 18

19 Deterministic Models Ray construction principle Ray tracing method Problems involved Virtual source for each reflector Virtual source Reflector Virtual source Reflector Source Receiver Source Receiver Ray launching method Construction imprecision Source Reflector Beam Source Ray 1 Reflector Ray Ray 2 Receiver Slide 19

20 Deterministic Models Base station Mobile Slide 20

21 Deterministic models : WaveSight TM 1. Physical basics of WaveSight 2. Implemented parameters and algorithms 3. Data input 4. Simulation 5. Prediction outcome (accuracy and speed) Slide 21

22 1. Physical basics of WaveSight The propagation of waves is following different modes in the three-dimensional environment. These propagation modes are : Free space Diffraction Transmission Reflection Scattering Slide 22

23 1.1 Physical basics of WaveSight a. Propagation in free space Energy concentration on a sector by antenna Free space Pr = C * Pt / 4pd 2 Antenna gain Pr = P * G / 4pd 2 Slide 23

24 1.2 Physical basics of WaveSight b. Diffraction phenomenon Source Attenuation near shadow boundary Source Obstacle Wave rounds the obstacle Obstacle Mobile Diffraction phenomenon Ray representation of diffraction (GTD, UTD) In shadow area, the received power is proportional to wavelength Slide 24

25 1.3 Physical basics of WaveSight Reflection, scattering and transmission phenomenon Source Scattering Transmission and scattering have no significant impact. So they are not taken into account by WaveSight. Reflection Building Transmission Slide 25

26 Approach: Horizontal Plane Propagation ray0 Receiver Antenna Slide 26

27 Approach: Vertical Plane Propagation Slide 27

28 Conclusion WaveSight is an enhanced ray tracing algorithm close to ray launching (block ray construction with image source). It is 2,5 D (separate quasi horizontal propagation of vertical propagation) Takes into account : - 2 reflections (Fresnel formula) - 2 diffractions by vertical wedges (building corners) : UTD - 15 diffractions by horizontal wedges (roofs) : UTD - penetration in buildings (constant path loss) Slide 28

29 2. Implemented Parameters and Algorithms Terrain File Building Vector Data Frequency Receiver Height Transmitter Coordinates (x, y, z) Transmitter power Antenna Tilt Antenna Azimuth Antenna Radiation Pattern (horizontal and vertical) Slide 29

30 3. Data Input Data needed is vector data. WaveSight is taking ASCII files as input for geographical data in MSI Planet format (supported by almost all planning tools) Which resolution/accuracy is needed? Common Terrain resolution is 5m, but 25m is sufficient (given that the terrain is not too hilly) Building accuracy +/- 2m Slide 30

31 Actual terrain data of Paris Actual terrain data : (Paris, about 1km x 2km) Slide 31

32 4. Simulation WaveSight s interface is intuitive and easyto-use. Running a prediction can be done in 5 steps : Zooming in to see the area you are looking for Defining the antenna position Setting-up the antenna and base-station parameters Start the prediction Analysing results on the map Slide 32

33 WaveSight of ArcView Building and Terrain Data of a part of the City of Bern in Switzerland (Vector Format) Slide 33

34 WaveSight of ArcView Zooming Slide 34

35 WaveSight of ArcView Zooming II Slide 35

36 WaveSight of ArcView Defining an Antenna Position Slide 36

37 WaveSight of ArcView Setting Antenna And Prediction Parameter Slide 37

38 WaveSight of ArcView Antenna Height, Power, Azimuth, Tilt, Antenna-Type, Prediction-Radius Slide 38

39 WaveSight of ArcView Start of the Prediction Slide 39

40 WaveSight of ArcView Prediction Running Slide 40

41 WaveSight of ArcView Prediction with Radius of 300m (prediction time less than a minute on a PII 400 machine) Slide 41

42 WaveSight of ArcView Shadow Area Shadow Area Shadow Area Slide 42

43 WaveSight of ArcView Canyon-Effect of Streets Canyon-Effect of Streets Canyon-Effect of Streets Slide 43

44 Prediction Outcome (Accuracy and Speed) After the simplicity of WaveSight needing no calibration and being extremely easy to use has been shown before, the two main questions to prove WaveSights value are directed towards accuracy and speed. The accuracy determines wether WaveSight can improve the input for planning and therefore allows higher quality in network setup and fine-tuning. The speed determines wether WaveSight has a practical use at all. So far the big obstacle to the practical use of ray-tracing were computation times up to 12 hours for a microcell. The following slides show, how WaveSight typically performs regarding on these critical decision parameters and prove that WaveSight delivers a unique combination of simplicity, accuracy and speed! Slide 44

45 tested In The Real World Over 80 sites and 1000 km of routes validated: 10 cities / 7 countries 8 operators World class high-tech reference: Bell Labs EPFL KPN Different city types: Paris Munich Bern Fribourg Tampa Manhattan Rotterdam The Hague Brussels Torino 7 years of development Continually growing verification pool! Slide 45

46 Macrocell in Bern Transmitter: Omni directional antenna, 632 m height, Frequency 1800 MHz Building layout: Average building heights 614 m Receivers: located within 1500 m from transmitter on a 20 km route Computing time: 6 minutes on a Pentium II 266 Error with Measurements: Mean and standard deviation are 0 and 7.5 db Power [dbm] Measurements WaveSight Predictions Rx number Measurements and buildings courtesy Swisscom and Istar respectively Slide 46

47 Small Cell In Bern Transmitter: Omni directional antenna, 625 m height, Frequency 1800 MHz Building layout: Average building heights 614 m Receivers: located within 1500 m from transmitter on a 4 km route Computing time: 6 minutes on a Pentium II 266 Error with Measurements: Mean and standard deviation are.5 and 5.4 db Measurements WaveSight Predictions Power [dbm] Rx number Measurements and buildings courtesy Swisscom and Istar respectively Slide 47

48 Microcell in Bern Transmitter: Omni directional antenna, 613 m height, Frequency 1800 MHz Building layout: Average building heights 614 m Receivers: located within 1000 m from transmitter on a 2.5 km route Computing time: 2 minutes on a Pentium II 266 Error with Measurements: Mean and standard deviation are -1.5 and 8.2 db - Path Loss [db] Measurements WaveSight Predictions Receiver number Measurements and buildings courtesy Swisscom and Istar respectively Slide 48

49 Small Cell in Munich Transmitter: Omni directional antenna, 13 m height, Frequency 900 MHz Building layout: Average building heights 16m Receivers: located within 2500 m from transmitter on a 10 km route Computing time: 7 minutes on a Pentium II 266 Error with Measurements: Mean and standard deviation are -0.2 and 6.4 db - Path Loss [db] Measurements WaveSight Predictions Receiver number Measurements and buildings courtesy Mannesmann Slide 49

50 Microcell in The Hague Transmitter: Omni directional antenna, 8 m height, Frequency 900 MHz Building layout: Average building heights 11 m Receivers: located within 3000 m from transmitter on a 25 km route Computing time: 8 minutes on a Pentium II 266 Error with Measurements: Mean and standard deviation are 3.5 and 6.8 db - Path Loss [db] Measurements WaveSight Predictions Receiver number Measurements and buildings courtesy KPN Slide 50

51 Small Cell in The Hague Transmitter: Omni directional antenna, 11 m height, Frequency 900 MHz Building layout: Average building heights 11 m Receivers: located within 3000 m from transmitter on a 25 km route Computing time: 8 minutes on a Pentium II 266 Error with Measurements: Mean and standard deviation are 4.1 and 6.8 db Measurements WaveSight Predictions - Path Loss [db] Receiver number Measurements and buildings courtesy KPN Slide 51

52 Macrocell in The Hague Transmitter: Omni directional antenna, 15 m height, Frequency 900 MHz Building layout: Average building heights 11 m Receivers: located within 3000 m from transmitter on a 25 km route Computing time: 8 minutes on a Pentium II 266 Error with Measurements: Mean and standard deviation are 2.5 and 8.3 db Measurements WaveSight Prediction - Path Loss [db] Receiver number Measurements and buildings courtesy KPN Slide 52

53 Summarm of Accuracy Bern Bern Bern Munic h Munich The Hague The Hague The Hague Cell Size Macro Small Micro Small Small Micro Small Macro Antenna Height (above building) Mean error (db) Standar d Deviatio n (db) Slide 53

54 Microcell in Paris Area Calculation Time 0,16 km 2 (400m x 400m) 29 sec. / 0,48 min. 1 km 2 (1000m x 1000m) 57 sec. / 0,95 min. 4 km 2 (2000m x 2000m) 153 sec. / 2,55 min. 16 km 2 (4000m x 4000m) 554 sec. / 9,23 min. 36 km 2 (6000m x 6000m) 1099 sec. / 18,32 min. The calculations were performed with a 650 Mhz Pentium III PC with 192 MB RAM. Slide 54

55 Macrocell in Torino Area Calculation Time 0,16 km 2 (400m x 400m) 25 sec. / 0,42 min. 1 km 2 (1000m x 1000m) 54 sec. / 0,90 min. 4 km 2 (2000m x 2000m) 246 sec. / 4,10 min. 16 km 2 (4000m x 4000m) 412 sec. / 6,87 min. 36 km 2 (6000m x 6000m) 951 sec. / 15,85 min. 64 km2 (8000m x 8999m) 5802 sec. / 96,70 min The calculations were performed with a 650Mhz Pentium III PC with 192 MB RAM. Slide 55

56 Microcell in Amsterdam Area Calculation Time 0,16 km 2 (400m x 400m) 17 sec. / 0,28 min. 1 km 2 (1000m x 1000m) 95 sec. / 1,58 min. 4 km 2 (2000m x 2000m) 315 sec. / 5,25 min. 16 km 2 (4000m x 4000m) 960 sec. / 16,0 min. The calculations were performed with a 650 Mhz Pentium III PC with 192 MB RAM. Slide 56

57 Microcell in Bern Area Calculation Time 0,16 km 2 (400m x 400m) 15 sec. / 0,25 min. 1 km 2 (1000m x 1000m) 34 sec. / 0,57 min. 4 km 2 (2000m x 2000m) 74 sec. / 1,23 min. 16 km 2 (4000m x 4000m) 196 sec. / 3,27 min. The calculations were performed with a 650Mhz Pentium III PC with 192 MB RAM. Slide 57

58 Macrocell in Paris Area Calculation Time 0,16 km 2 (400m x 400m) 27 sec. / 0,45 min. 1 km 2 (1000m x 1000m) 73 sec. / 1,22 min. 4 km 2 (2000m x 2000m) 201 sec. / 3,35 min. 16 km 2 (4000m x 4000m) 492 sec. / 8,20 min. 36 km 2 (6000m x 6000m) 915 sec. / 15,25 min. The calculations were performed with a 650 Mhz Pentium III PC with 192 MB RAM. Slide 58

59 Macrocell in Paris Area Calculation Time 0,16 km 2 (400m x 400m) 30 sec. / 0,50 min. 1 km 2 (1000m x 1000m) 105 sec. / 1,75 min. 4 km 2 (2000m x 2000m) 355 sec. / 5,92 min. 16 km 2 (4000m x 4000m) 750 sec. / 12,50 min. 36 km 2 (6000m x 6000m) 990 sec. / 16,50 min. The calculations were performed with a 650 Mhz Pentium III PC with 192 MB RAM. Slide 59

60 Summary of Speed Area of Calculation Microcell in Paris Macrocell in Torino Microcell in Amsterdam Microcell in Bern 0.16 km min 0.42 min 0.28 min 0.25 min 1 km min 0.90 min 1.58 min 0.57 min 4 km min 4.10 min 5.25 min 1.23 min 16 km min 6.87 min 16.0 min 3.27 min 36 km min min km Slide 60

61 IV. The added value of WaveSight for network planning 1. Reduced need for measurement and drive tests 2. Better knowledge of Coverage 3. Better Input / better results for frequency planning 4. Rising network quality / better opportunities for fine/tuning 5. Increased speed of network set-up 6. Reduction in the number of needed sites Slide 61

62 Reduced need for measurement and drive tests WaveSight is a deterministic model. The input consists of a rough picture of the physical environment and the calculation process is based on a picture of the physical behavior of the rays (UTD, Ray-Tracing, Maxwells-Theory of Rays). Because all signficant geographic information is taken into account in the calculation process, the model needs NO TUNING. Therefore a lot of money, time and energy which is going into drive testing can be saved by using WaveSight. Slide 62

63 Better knowledge of Coverage It is clearly visible that the coverage in the empirical prediction is only very crudely estimated (red area), whereas the WaveSight prediction takes the physical characteristics of the city into account (buildings, streets, terrain) to compute a precise map. The canyon effect of streets is visible. Okumura-Hata Model (macro) + Pseudo-Ray Tracer (micro) WaveSight Model Slide 63

64 Coverage by signal level Cost-Hata WaveSight Coverage by signal >=-75 dbm >=-81 dbm >=-85 dbm >=-93 dbm >=-96 dbm >=-102 dbm Measurements Slide 64

65 Better Input / better results for frequency planning To show the impact of WaveSights predictions on frequency planning the Wavecall team has performed a study based on the GSM-technology (download: The main objective of this case study was to show that using a sophisticated prediction model reduces cost and time in frequency planning. ILSA, a frequency-planning tool from Aircom, was used to compare frequency planning obtained by using classical propagation models and the ray-tracing model WaveSight. The tests presented were performed on a 4.5 km 2 area comprising 17 sites (36 cells) actually used in the Bouygues Telecom network of Paris. This study demonstrates that using WaveSight can reduce the number of carriers needed to provide the same network quality from 47 carriers to 40. This is significant not only because it can reduce the cost of fine-tuning the network, but also because extra carriers can be used to increase traffic capacity. By using, WaveSight, the network interference area can be reduced by 80%. Slide 65

66 Rising network quality / better opportunities for fine/tuning Changing a minor parameter can have astonishing impact on the coverage. The single change made is the antenna s height. Initial coverage (left), Antenna 2m higher (above) Slide 66

67 Best Server (-102dBm) Cost-Hata WaveSight Slide 67

68 Overlapping (- 80dBm) Cost-Hata WaveSight Intercell interference => noise increase => reduced capacity Slide 68

69 Soft Handover No handoff area Cost-Hata WaveSight Slide 69

70 Simulation User distribution Cost-Hata WaveSight Green: OK ; Black: inactive; Red: mobile power outage; Pink: BS power outage; Purple: overload (load factor) Slide 70

71 Simulation 232 active users Simulation with empirical prediction: 4 rejections (1.7 %), due to load factor Simulation with deterministic prediction: 47 rejections (20.3%), 5.2 % due to power, 4.3 % due to pilot problem, 10.8% due to load factor. 18% of speech users, 22% of 64kbps users, 28% of 144kbps users. Slide 71

72 Effective service area - Speech Cost-Hata WaveSight 100% of area 90.3% of area Slide 72

73 Effective service area - 64kbps Cost-Hata WaveSight 100% of area 87.1 % of area Slide 73

74 Service area kbps Cost-Hata WaveSight 99.3 % of area 83.1 % of area Slide 74

75 DL Service area 64kbps different Eb/N0 threshold Cost-Hata WaveSight 64kbps Service area ( Eb/Nt maxi >= 10 db Eb/Nt maxi >= 8 db Eb/Nt maxi >= 5 db Eb/Nt maxi >= 1 db Slide 75

76 Conclusion & future UMTS simulations and planning are based on the path loss prediction => if it s wrong the results are nonsense. Ray tracing models are also able to deliver other interesting feature for UMTS: impulse response, delay spread, direction of arrivals. Slide 76

77 Increased speed of network set-up In the actual situation at the UMTS-frontier speed is desperately needed. For acquiring their licenses operators had to invest tremendous amounts of money, raising their debts to almost intolerable limits. Every day of delaying the start of their new networks and making money with their licenses can be said to cost millions of in interest rates. Because WaveSight is easy to use, calculates accurate predictions extremely fast and needs no calibration to (non-existing) measurements it can significantly speed-up the planning phase and help to shorten the time to start the technical and commercial operation of the network saving millions of in interest rates. Slide 77

78 Reduction in the number of needed sites In the planning phase, a propagation fading margin is taken in the link budget in order to raise the probability of good coverage. This margin is directly related to the standard deviation error of the propagation prediction model. The more accurate the propagation model is, the lower the standard deviation is and the lower the needed margin is. A gain in the link budget results in a lower needed base station density. An improvement of 1dB in the link budget corresponds to 12% reduction in the number of needed sites. 2dB corresponds to a 23 % reduction and 3dB to 32%. (based on a propagation path loss exponent of 3.52) 1 1 (ref WCDMA for UMTS, H.Holma A.Toskala) Slide 78

79 Integration of WaveSight into Planning Tools 1. Integrated Planning Tools 2. Integration on Windows Systems 3. Integration on Unix Systems 4. License Protection Slide 79

80 Integrated Planning Tools Integration Completed : Enterprise Aircom (Partnership with one-stop-shop and support) Totem Nokia (Partnership with one-stop-shop and support in negotiation) Atoll Forsk (Partnership with one-stop-shop and support) Odyssey Logica (Partnership with one-stop-shop and support) Ellipse CRIL Planet MSI Integration work in progress: Astrix Teleplan (Partnership with one-stop-shop and support) Celplanner Suite Celplan (Partnership with one-stop-shop and support in negotiation) Wizard Safco/Agilent (Partnership with one-stop-shop and support in negotiation) Companies with a positive stance towards integration Quantum Quotient Decibel Planner Northwood Slide 80

81 Integration on Windows Systems Planning Tool Calc Array() Create input files for WaveSight WaveSight Interface 1 2 Call Generates Results Files.txt Read DLL Library Winsight.exe w2c_wsal.dll Slide 81

82 Integration on Unix Systems Planning Tool Calc Array() Create input files for WaveSight WaveSight Interface 1 2 Call Generates Results Files.txt Read WS.exe Slide 82

83 License Protection The licensing is made in 2 different ways : 1. Completely integrated into the planning tools, with whom we have a partnership agreement (protection with the dongle solution) 2. In the other case, with a node-locked license where a serial key is generated in link with the D-drive serial number (NT Workstation) or the Host-ID (Unix Workstation) Slide 83