Low Cost Network Planning Tool Development using Google Maps

0 downloads 0 Views 2MB Size Report
Keywords-Network Planning tool; WiMAX; NPCET; GIS;. Google Maps; SRTM; Coverage Prediction; Point to Point Link;. Capacity Simulation; MATLAB GUI.
Low Cost Network Planning Tool Development using Google Maps, SRTM Data and MATLAB 1*

Chamal Sapumohotti, 2*Mohamad Yusoff Alias, 3#Ho Chiung Ching, 4#Wee Kuok Kwee *

Faculty of Engineering, #Faculty of Information Technology Multimedia University Cyberjaya, Malaysia 1 [email protected],[email protected], [email protected],[email protected]

the deployment (rural networks emphasis on high coverage requirements as opposed to high capacity requirements usually found in urban deployments) we have emphasized on coverage predictions. NPCET also has very good support for planning point-to-point backhauls as it is very hard to put wired backhauls in remote sites found in rural deployments. There is also some support for capacity planning. The tool uses freely available Google maps static Application Programming Interface (API) [1] as the GIS information source and uses freely available Shuttle Radar Topography Mission (SRTM) 90m [2] data as the elevation data source.

Abstract—Network planning is an essential part of any wireless network deployment. The prices of commercial network planning tools are expensive, making them inaccessible to many researchers and organizations with small networks. However, in house development of network planning tools are not as hard as it was used to be. The freely available web maps and rapid development capability of MATLAB Graphical User Interfaces (GUIs) and recently integrated Object Oriented Programming to MATLAB make it much easier. Shuttle Radar Topography Mission (SRTM) data is a public database which gives access to elevation data for the whole world. In this paper we give the architecture details of our owned Network Planning Cell Tool (NPCET), a network planning tool that was designed for planning rural WiMAX networks in Malaysia. Developing a network planning tool is multi-disciplinary in nature, the scope of the project involves wireless propagation, geographical information systems and maps, network planning and programming. The paper can be used as a guide for researchers on how to integrate necessary data sources and propagation models for building a comprehensive planning tool for almost any type of communication system.

II.

NPCET includes a number of features to allow efficient network planning such as follows: • • • •

Keywords-Network Planning tool; WiMAX; NPCET; GIS; Google Maps; SRTM; Coverage Prediction; Point to Point Link; Capacity Simulation; MATLAB GUI

I.

• •

INTRODUCTION

• •

Fixed WiMAX can be utilized to effectively provide broadband penetration to rural areas. The idea is to use lower frequencies (ex. 450MHz) which have favorable propagation conditions for covering large distances. We designed a low cost network planning tool for such a deployment. Telecommunication network deployment starts off with an extensive planning stage. The initial simulation based analysis is vital to ensure a reasonable outcome from the deployment; otherwise it would result in the need to put additional base stations and even worse, the need to relocate base stations.

• • •

There are many software solutions which aid in performing this process. These tools use Geographic Information System (GIS) information and propagation models to predict the coverage and capacity performance of a network and tune these parameters before an actual deployment. Commercial tools available (Mentum Planet, EDX signalpro etc.) are relatively expensive. In this paper we introduce Network Planning Cell Tool (NPCET), a low cost network planning tool targeting rural fixed WiMAX network deployments. Due to the nature of

978-1-61284-663-7/11/$26.00 ©2011 IEEE

NPCET



• •

13

Coverage planning Capacity Planning Simulate the WiMAX 802.16d air interface. Integrated extensible antenna database which allows to setup base stations with precise azimuth and tilt. APViewer to view antenna patterns Integrate Google geocode API to obtain latitude and longitude for any location. Uses Google Maps as the GIS source. Obtain the elevation of any point in the map and obtain the highest points in the map Manually define the clutter types for models which require clutter information Include SRTM 90m data as the elevation information source. Highly accurate point to point link simulation to identify line of side (LOS) requirements for backhauls as well as for long distance point to multi point connections. Plot the Fresnel zones in the point to point link analysis to identify significant obstacles and to identify the minimum backhaul antenna heights for LOS communication. Many different propagation models such as Friis, Plain Earth, Hata and SUI. User friendly graphical user interface (GUI) to create, save, edit, and simulate WiMAX networks.

ICOIN 2011

Once the location latitude and longitude is obtained from the Google geocode API request, the Google map for the given location can be obtained using the Google static maps API. A sample request is given below.

NPCET is completely written in MATLAB [3,4]. The final product is distributed as a distribution package created by depolytool using MATLAB compiler. A. The GIS of NPCET The GIS (Google maps) has been used to input the network data namely location of base station and point to point links and also to display the simulation results in a meaningful way. We have used the static maps API provided by Google to acquire the maps for the location under consideration. Any other web map source could have been used for this purpose. However, each map consists of a limited resolution. We have combined 4 maps to come up with a larger map since a single static map request only provides 256 × 256 pixel resolution at any zoom level. The absolute distance covered by each pixel is obtained by assuming an arc second as equal to 30m. Since Malaysia is near the equator, this assumption is reasonable.

googleMap=imread(‘http://maps.google.com/maps/a pi/staticmap?center=5.9288,102.8392&size=256x25 6&zoom=10&maptype=roadmap&sensor=false&for mat=jpg&key=ABQIAAAAN5_trHVC2PmdTNXfc BQpqBRxLj2ZrEGdAE-AhCWesxNkxJ3PRSAjKrXYH6CaCWcXjRTwDnh9uJEiQ‘ , 'jpg'); At any zoom level the image returned by an image request is 256 × 256. Although the API allows for requests of up to 512 × 512, the dimensioning is not correct for the higher image sizes. Thus, it was required to tile 4 images of size 256 × 256 to come up with the final image of the map. The width of such a requested 256 × 256 image in arc grids = 360 / (2 ^ (zoom)). The tiled image is a 496 × 512 image to scale to the rectangular display.

The elevation data used by the NPCET GIS is a processed version of the freely available SRTM data. The Consultative Group on International Agricultural Research (CGIAR) Consortium for spatial Information (CGIAR-CSI) [5] allows processed SRTM data sets to be downloaded in GeoTiff format. The elevation data can be easily extracted from the available GeoTiff files. This data is used for point to point link predictions. The downloadable map tiles have a 5 arc degree by 5 arc degree in size with 0.000833333333 arc degree or approximately 90m per pixel resolution. These map tiles have to be combined to create the view of a user defined area. The downloaded tiles have a resolution of 6001 × 6001. Each tile has redundancy for 1st and last column as well as 1st and last row. These redundancies have to be removed when combining tiles row wise or column wise. For the purpose of covering Malaysia, we have used 10 tiles which cover latitudes from N0.0004168 to N9.9995837 and longitude from E95.0004168 to E119.9995837. The extracted pixels for the combined view are given in the Table I. TABLE I.

Vertical Zone A Vertical Zone B

The concept of this grid is centralized to all the operations in NPCET. The resolution of all the calculations is limited to the grid resolution. This grid size was chosen due to the processing time required for large networks. The absolute width and height of a grid point is found by performing an approximate conversion. This approximation is valid for locations near the equator. PixelWidthinMeters = pixelWidthInArcSeconds * 30

The user is able to place the locations in the map by clicking the mouse at the desired location. This will automatically extract the geo location from the map. The user has to configure the sectors and antenna parameters of individual base stations. The List Box of base stations is used to select and edit base stations after initialization.

NUMBER OF PIXELS TO BE EXTRACTED FROM TILES Horizont al Zone 1

Horizont al Zone 2

Horizont al Zone 3

Horizont al Zone 4

Horizont al Zone 5

5999

5999

5999

5999

5999

5999

5999

5999

5999

5999

(1)

B. Coverage Simulations There are four propagation models used for predicting the coverage implemented in NPCET. All of these models start with calculating the receiver noise floor. %boltzman constant. k = 1.3806503 * (10 ^ (-23)); %convert temeperature to kelvin T = (temperature + 273.15); %receiver noise in mW. rcvNoise = k * T *(channelBandwidth * (10 ^ 6)) * (10 ^ (rcvNoiseFigure / 10)) * 1000; %noise floor in dBm noiseFloor = 10 * log10(rcvNoise);

Any network planning tool should have the ability to use geo location information. We have used Google geocode API and static maps API for this purpose. The first step in any project is to use the given location by the user to get the latitude and longitude of the location of the network. The request is concatenated and sent. For example to get the location of “Redang Malaysia” will require the following request to be sent. GeoCodeData=xmlread(http://maps.google.com/ma ps/geo?output=xml&key=ABQIAAAAN5_trHVC2 PmdTNXfcBQpqBRxLj2ZrEGdAE-AhCWesxNkxJ3PRSAjKrXYH6CaCWcXjRTwDnh9uJEiQ &q=Redang Malaysia);

The noise floor is later used for calculating the signal to interference noise ratio (SINR). The simulation is then carried for each base station. The highest power received at a grid

14

point is stored in a matrix. The other base stations add up as interference for the same grid point given that they have sectors which use the same frequency sub band. Now we will look at the calculation for a single grid point with respect to a single base station. The antenna gain is obtained by calculating the azimuth angle (depends on the direction of the grid point from of the antenna) and elevation angle (depends on the height of the base station and subscriber station and the antenna tilt). The angles with the elevation and azimuth antenna patterns of the sector antenna are used to identify the base station antenna gain. The fixed subscriber station is assumed to be pointed at the base station antenna. Thus, subscriber station antenna gain is assumed to be the maximum gain of the subscriber station antenna. The path loss is calculated differently according to the propagation model used.

where A is 20 * log10(4 * pi * d0 / lambda), gamma is (a - (b * Hb) + (c / Hb)), a, b, c are constants that depends on the terrain type, d0 is 100m, feqCorrection is 6 * log10(sectorFeq/ 2000) and antennaHeightCorrection is The rcvHeightCorrectionFactor * log10(Hm). rcvHeightCorrectionFactor also depends on the terrain type. 4) Hata Model The final model that we used is the Hata Model [8]. Hata model uses different propagation effects depending on the Terrain type (Urban, Suburban, open) and City size in urban terrain. Sample implementation for Hata urban is given below. PathLoss = 69.55 + 26.16*log10(sectorFeq) - 13.82*log10(hB) (5) CH + (44.9 - 6.55*log10(hB))*log10(dAct) where CH is a parameter that is calculated differently depending on the city size and the frequency. It is calculated using frequency and the subscriber station height.

1) Friis free space path loss model The first model to be used is the Friis free space path loss model [6]. It is given by:

The received power is calculated using PathLoss = 32.4 + 20 * log10(sectorFeq) + 20 * log10(dAct) (2) rcvPower = transmitSignalPower + BSantennaGain + SSantennaGain – PathLoss

where sectorFeq is the center frequency in use by the sector which is covering the grid point and dAct is the absolute distance between the base station and the subscriber station. The shadow fading is not considered in this coverage prediction model. Thus, only 50% edge coverage probability is assumed. This model is only suitable for high base stations with subscriber stations operating with LOS in open areas.

where rcvPower is the power received by the subscriber station, transmitSignalPower is the power output of the transmitter subtracted by the cable losses. BSantennaGain is the antenna gain in the azimuth and elevation of the direction towards subscriber station. SSantennaGain is the gain of the subscriber station antenna (assume the subscriber station antenna is directly pointed towards base station), PathLoss is the path loss calculated by a selected propagation model.

2) Plane Earth Model The second model is the plane earth model [6] which is given by PathLoss = 40 * log10(dAct) + 20 * log10(Hb) + 20 * log10(Hm)

The shadow fading variance for SUI is between 8.2dB to 10.6dB. The user has to define the shadow variance in other models. The shadow variance is used in order to calculate the fade margin with the user defined outage probability.

(3)

outageProbability = 1 (str2double(coverageProbabilityPrecentage) / 100); shadowStandardDeviation = 8.2; sigmaPoints = qfuncinv(outageProbability); fadeMargin = sigmaPoints * shadowStandardDeviation;

where Hb is the base station height and Hm is the subscriber station height. 3) SUI Model The third model is the Stanford University Interim (SUI) model [7]. SUI Model classifies the terrain in to three types. The three main terrain types described in SUI model are, • type A (hilly with heavy tree density) • type B (hilly with low tree density or flat with high tree density) • type C (flat with low tree density) The path loss for SUI is given by, PathLoss = A + 10 * gamma * log10(dAct / d0) + feqCorrection + antennaHeightCorrection

(6)

The Carrier to Interference-plus-Noise Ratio (CINR) requirement is given a higher value by adding the fade margin to the minimum CINR requirement of a modulation type. C. Point to Point Link Simulation The point to point link simulation is required to identify LOS condition for backhauls and to validate Subscriber station deployment. The SRTM 90m data is used to extract the elevation of the direct path in the point to point link. The earth bulge is added to the terrain to elevate it.

(4)

15

The Fresnel zones are plotted on top of the LOS plot. This plot can be used to identify obstruction which comes in to the first Fresnel zone. The antenna heights can be dynamically adjusted to identify the minimum height required for clearance.

numOfUsers = (sectorDLCapacity / commitedInformationRate) * overSubscribtionRate * 1000;

D. Capacity Simulation The WiMAX Air interface parameters are used to initially calculate [9] the data rate for a single direction with only binary phase shift keying (BPSK) modulation used for modulation. This is the raw BPSK bit rate.

III.

SAMPLE SIMULATIONS

A. Coverage Simulation The result of a coverage simulation in Terengganu Malaysia is given below. In Fig. 1, the network to be simulated is shown. Three frequency bands have been allocated among the four base stations (BS). BS1 and BS4 have single sector directional antennas while BS2 and BS4 uses 3 sector directional antennas. The network layout in the order of Azimuth, tilt, frequency (MHz) and transmit power (dBm) for each sector is detailed in Table 2 below. The antenna pattern used in all sector antennas is given in Fig. 2. The results of propagation simulation returns three matrices, • The sector with the best SINR for a grid point • The SINR of the grid point • The frequency of the sector covering the grid point

dataCarrierRatio = dataSubCarriers / fftSize BPSKrate = channelBandwidth * dataCarrierRatio * (1 - cyclicPrefix) The BPSK rate for uplink and downlink are calculated according to the duplexing system used by the system. Half Duplex Frequency Division Duplex (HFDD) is the duplexing used in low cost subscriber station systems. Semi-HFDD is used in situations where the spectrum is not paired (uplink and downlink frequency separation is low). Example for semi-HFDD, %semi-HFDD (HFDD with short channel seperation.) %equivalent to TDD in 2 frequency. rawBPSKrateDownLink = BPSKrate * (1 ULtoTotalratio); rawBPSKrateUpLink = BPSKrate * ULtoTotalratio; Using the results generated by the coverage analysis, for each sector the number of grid points covered by different modulation schemes are determined. Subsequently the ratios of points with the different modulation types are derived. %multiply the rawBPSK rate by the modulation ratios and the number of user bits carried at each modulation level. QAM64_FEC_3by4Rate = (QAM64_FEC_3by4Area / totalSectorArea) * 6 * (3/4); QAM64_FEC_2by3Rate = (QAM64_FEC_2by3Area / totalSectorArea) * 6 * (2/3); QAM16_FEC_3by4Rate = (QAM16_FEC_3by4Area / totalSectorArea) * 4 * (3/4); QAM16_FEC_1by2Rate = (QAM16_FEC_1by2Area / totalSectorArea) * 4 * (1/2); QPSK_FEC_3by4Rate = (QPSK_FEC_3by4Area / totalSectorArea) * 2 * (3/4); QPSK_FEC_1by2Rate = (QPSK_FEC_1by2Area / totalSectorArea) * 2 * (1/2);

Figure 1 Layout of the network with four base stations

where QAM is the quadrature amplitude modulation, FEC is the forward error correction and QPSK is the quadrature phase shift keying.

Figure 2 The sector antenna pattern using built in APViewer tool

A different color code is assigned to each sector of each base station. This is combined with the RGB (color values) of the map jpg image. Fig. 3 shows the serving sector map. Fig. 4 shows the frequency map. This is the frequency of the sector

The total capacity is the sum of all modulation types. sectorDLCapacity = (QAM64_FEC_3by4Rate + QAM64_FEC_2by3Rate + QAM16_FEC_3by4Rate + QAM16_FEC_1by2Rate + QPSK_FEC_3by4Rate)

16

which is covering a given grid point. The modulation map Fig. 5 overlays different colors for different modulation types. The light red overlay indicates 64QAM modulation with ¾ coding. The areas with purple overlay indicates the lowest modulation used in the system which is QPSK with ½ coding. If the signal strength is not sufficient to sustain this lowest modulation, it will be displayed without any overlay. TABLE II. Base Station BS1 BS2 BS3 BS4

Sector1 315, 1, 445, 30 30, 1, 450, 35 33, 1, 455, 25 135, 1, 445, 35

NETWORK LAYOUT Sector2

Sector3

N/A

N/A

135, 1, 450, 30 150, 1, 455, 30

270, 1, 450, 30 270, 1, 455, 30

N/A

N/A

Figure 5 Modulation map of the area overlaid on a Google satellite map.

B. Point to Point Backhaul Simulation In this example the possibility of establishing a backhaul between BS1 and BS2 is investigated. The arrow between BS1 and BS2 in Fig. 6 indicates the backhaul under analysis. The Google map relief is taken as the GIS image background. The SRTM layer is displayed in Fig. 7. The Red color indicates higher elevations. The BS1 is situated in the mountains while, BS2 is located in low ground. The result of the analysis is indicated in Fig 8. The green line is the earth bulge which is the obstruction created by the earths curvature. This is superimposed on the local terrain height extracted from SRTM data files (the dark blue line). The visual LOS is indicated by the light blue line. The red ellipse represents the first Fresnel zone. From the results it is clear that the LOS condition is not satisfied. Figure 3 The serving sectors overlaid on top of Google satellite map.

Figure 6 Backhaul between BS1 and BS2 on a Google relief map Figure 4 Frequency map overlaid on the Google road map.

C. Capacity Simulation A sample capacity simulation report is given below. A 3.5MHz channel has been used for the air interface with 256

17

[3]

Fast Fourier Transform (FFT) size, 1/16 cyclic prefix, 192 data subcarriers, semi-HFDD duplex, uplink to total ratio = 0.25, channel impulse response (CIR) = 200 kbps and oversubscription rate of 10.

[4]

[5]

Base Station BS1 Sector[1] Down Link Capacity 8.269673 Mbps Up Link Capacity 2.756558 Mbps Maximum Number of users 413 --------------------------------------Base Station BS2 Sector[1] Down Link Capacity 8.289821 Mbps Up Link Capacity 2.763274 Mbps Maximum Number of users 414 Sector[2] Down Link Capacity 7.871173 Mbps Up Link Capacity 2.623724 Mbps Maximum Number of users 394 Sector[3] Down Link Capacity 7.757612 Mbps Up Link Capacity 2.585871 Mbps Maximum Number of users 388 --------------------------------------Base Station BS3 Sector[1] Down Link Capacity 7.640334 Mbps Up Link Capacity 2.546778 Mbps Maximum Number of users 382 Sector[2] Down Link Capacity 7.649957 Mbps Up Link Capacity 2.549986 Mbps Maximum Number of users 382 Sector[3] Down Link Capacity 8.286235 Mbps Up Link Capacity 2.762078 Mbps Maximum Number of users 414 --------------------------------------Base Station BS4 Sector[1] Down Link Capacity 8.223883 Mbps Up Link Capacity 2.741294 Mbps Maximum Number of users 411

IV.

[6] [7] [8]

[9]

Patrick Marchand, “Graphics and GUIs with MATLAB third edition”, Chapman and Hall/CRC, 2003 Object-Oriented Programming in MATLAB, http://www.mathworks.com/products/matlab/object_oriented_programm ing.html The CGIAR Consortium for spatial Information (CGIAR-CSI) SRTM 90m Digital Elevation Data; http://srtm.csi.cgiar.org/ Simon R. Saunders, “Antennas and propagation for wireless communication“. John Wiley & Sons, 1999. IEEE 802.16 Broadband Wireless Access Working Group, “Channel Models for Fixed Wireless Applications,” 2001 M. Hata, "Empirical Formula for Propagation Loss in Land Mobile Radio Services", IEEE Transactions on Vehicular Technology, Vol, VT29, No, 3, 1980 Jeffrey G. Andrews, Arunabha Ghosh, and Rias Muhamed, “Fundamentals of WiMAX: Understanding Broadband Wireless Networking”, Prentice Hall, 2007

Figure 7 The SRTM layer which is used for extracting elevation data

CONCLUSION

In this paper we have discussed NPCET, a low cost network planning tool which uses freely available GIS information and is implemented using rapid application development capabilities of MATLAB. This work can be easily expanded to cover other areas of wireless communication planning using similar data sources. ACKNOWLEDGMENT This project is one part of a project that is funded by the Ministry of Science, Technology and Innovation, Malaysia under the TechnoFund scheme.

Figure 8 Point to Point link analysis

REFERENCES [1] [2]

Google Static Maps API V2 Developer Guide; http://code.google.com/apis/maps/documentation/staticmaps/ NASA Shuttle Radar Topography Mission; http://www2.jpl.nasa.gov/srtm/

18

Suggest Documents