You are on page 1of 6

GSM-R Radio Planning for the Bucureti Constana Railway Corridor Case study

Lcrmioara Mihaela Nemoi, Corneliu Mihail Alexandrescu, Elena Alina Stanciu

Faculty of Transportation Politehnica University Bucharest, Romania
Abstract The paper describes in detail a radio planning procedure for the design of an EIRENE (European Integrated Railway Radio Enhanced Network) network, which is a railway radio communications network based on the ETSI GSM standard (GSM-R), compliant with all the mandatory requirements specified in the EIRENE Functional Requirements Specification and System Requirements Specification. The approach of the radio planning procedure is highly practical and original, starting with clear indications on why and how to manipulate the digital cartography dataset, in order to accurately represent the railway infrastructure, the service area for the GSM-R network. Furthermore, the procedure provides all the cell and frequency planning deliverables required to accurately dimension a GSM-R network that covers the Bucureti Constana railway corridor. The radio planning tool used throughout the radio planning procedure is ICS Telecom produced by ATDI SA ( Keywords-component: ITU; Land mobile radio cellular systems; Land mobile radio propagation factors; Land mobile radio interference.

Adrian Ludovic Murean

Radio Division Special Telecommunications service Bucharest, Romania



A. Digital cartography From the radio planning perspective, the digital cartography representation, integrated within the radio planning tool, is essential in order to ensure the desired radio planning accuracy. Regardless the types of propagation models used for the radio planning purposes, the representation of terrain and manmade features in the digital cartography set is one of the major inputs, if not the essential one, for the accurate simulation of radio propagation conditions. All the relevant radio propagation mechanisms at VHF and UHF frequency ranges distance, reflection, scattering, refraction, diffraction, absorption are decisively influenced by the environmental data stored in the digital cartography set, which, integrated within the radio planning tool, is further referred to as the Geographic Information System (GIS) Functionality. The GIS Functionality includes the following data types: 1) Terrain elevation data: The terrain elevation data (Digital Terrain Maps DTMs) are usually available without additional obstructions such as buildings and vegetation. In general, DTMs have the longest currency, and it is only subject to change due to the largest natural phenomena or massive engineering projects (building of dams, massive excavations). The DTMs are usually available from military sources, consisting of paper maps contour lines extraction, followed by mathematical interpolation. The DTM data, used throughout the radio planning herein, used as a primary source DTED (Digital Terrain Elevation Data) Level 2 data, with performance specifications as per MILPRF89020B [1]. As the subject of the paper herein is not the GIS Functionality itself, it is enough to specify that the DTM has a resolution of 25 meters, meaning that every pixel of the screen, at the maximum resolution, represents a 25x25 meters square, with the same terrain altitude, expressed in meters relative to the Mean Sea Level.



An accurate radio planning is essential anywhere in the implementation process of an EIRENE network, starting with the preliminary planning phase, meant to correctly assess the system dimensioning and to fundament the equipment and services acquisition, and ending with the optimization phase, which ensures that the required parameters for the quality of services offered by the system are met. As an EIRENE network provides the radio bearer for the train signaling systems (ERTMS European Rail Traffic Management System and ETCS European Train Control System), hence the railway safety relies on transmission link between train-borne and trackside ERTMS/ETCS applications, the EIRENE System Requirements Specification are much more restrictive than those for commercial GSM networks. For example, the coverage level should exceed 41.5 dBV/m (-95 dBm), with a probability of 95%, on lines with ETCS levels 2/3, for train speeds lower than or equal to 220km/h.

2) Radio clutter data:

Radio clutter data is effectively a map of terrain occupancy insofar as it relates to the effect on radio propagation: buildings, vegetation of different types, including forests, different types of infrastructures like roads, railways etc. Buildings are either individually represented, or as belonging to builtup areas of different types, such as rural, suburban, and different categories of urban. The clutter data, used throughout the radio planning herein, were extracted from satellite Landsat 7 sources and topographical maps, both type of data being collected between the years 20002003. The original resolution of the clutter file is 50 meters, which was resampled to match the 25 meters resolution of the DTM. Note that the resampling method does not increase the precision of the original clutter representation within the GIS Functionality. As shown below in this paper, the radio planning for GSMR requires much higher precision for the clutter layer, than the one available directly from commercial digital cartography sources. The most important clutter, for the purpose of GSMR radio planning, is the railway infrastructure, i.e. the railway lines, the railway stations, and generally all the Railway Company property, which represent both the service area for the GSMR network, and the most eligible candidate areas for GSMR base stations deployment (from the site acquisition perspective).

vector data can be achieved from high precision VMAP2 data, normally available from military sources. The vector representation will be used during the GSMR prospective radio planning procedure (as detailed in Chapter III), to generate subscribers along vector lines (railway tracks), with exactly the desired density and within the whole service area of the GSM R network, and most importantly nowhere outside the service area. This will allow for an optimal placement of the GSMR base stations, to cover only the desired service area, thus minimizing the capital and operations expenditures needed to build and operate the GSMR network. The clutter representation of the of the railway infrastructure (and GSMR network service area) can be generated from its vector representation (a process called rasterization), by means of two functions embedded in ICS Telecom: for polyline vectors (railway tracks) we used the Modify clutter along vector line function (available in the Map Raster operation menu of the ICS Telecom main window); for region vectors (railway stations and depots, shunting areas) we used the Modify clutter from SHP polygons function (also available in the Map Raster operation menu). The clutter representation will be used during the GSMR prospective radio planning procedure (as detailed in Chapter III), to represent both the GSMR network service area, and most importantly, the most eligible candidate areas for GSMR base station deployment. As the railway clutter spans almost exclusively across the Railway Company property, the site acquisition process would be considerably facilitated in such areas.

3) Scanned and geographically referenced paper maps. Photographic images of the service area, used to visually relate the calculated radio coverage to important social features, such as roads, railways and areas important to the end user community. B. Representation of the railway infrastructure, the desired service area for the GSMR network For the purpose of the GSMR radio planning, as detailed in the prospective radio planning procedure (Chapter III) the railway infrastructure must be represented both as a dedicated radio clutter and as vector files, the latter representation being the source for the former. The vector representation of the railway infrastructure (polylines and regions in ESRI SHAPE or other standard formats) is easily available at the desired precision, from several sources: Collection of railway tracks data (vectors of polyline type) through a dedicated activity, that can be carried out by the Railway Company itself, by mounting a professional GPS receiver on a train cab, and collecting the GPS coordinates while driving a locomotive along the railway corridor of interest. Alternatively, depending on the desired accuracy of the radio planning, railway

To conclude, practically, for the purposes of this paper, the vector representation of the Bucureti Constana Railway Corridor was extracted from the vector files included within the digital cartography set Romania 50 meters, acquired with the ICS Telecom software package. The railway clutter was firstly removed from the original clutter file also provided within the aforementioned software package, and then reinserted in the clutter file, through the rasterization process described above. This ensured a perfect coherence between the vector and clutter files representing the railway infrastructure, that is all the subscribers used for the prospective radio planning (generated along vector lines) are positioned on the railway clutter, which effectively cuts through other clutters represented on the map, giving the radio planner the possibility to deploy GSMR base stations, in order to optimally provide radio coverage for the whole service area of the GSMR network, leaving no portion uncovered. C. Radio coverage required by EIRENE SRS (System Requirements Specifications) Extract from the EIRENE SRS for the required radio coverage values:

3.2 Coverage

3.2.1 For network planning, the coverage level is defined as the field strength at the antenna on the roof of a train (nominally a height of 4m above the track). An isotropic antenna with a gain of 0dBi is assumed. This criterion will be met with a certain probability in the coverage area. (The target coverage power level is dependent on the statistical fluctuations caused by the actual propagation conditions.) (I) 3.2.2 The following minimum values shall apply: (M) coverage probability of 95% based on a coverage level of 38.5 dB V/m (98 dBm) for voice and nonsafety critical data; coverage probability of 95% based on a coverage level of 41.5 dBV/m (95 dBm) on lines with ETCS levels 2/3 for speeds lower than or equal to 220km/h. 3.2.3 The following minimum values are recommended: (I) coverage probability of 95% based on a coverage level of 44.5 dBV/m (92 dBm) on lines with ETCS levels 2/3 for speeds above 280km/h; coverage probability of 95% based on a coverage level between 41.5 dBV/m and 44.5 dBV/m (95 dBm and 92 dBm) on lines with ETCS levels 2/3 for speeds above 220km/h and lower than or equal to 280km/h. 3.2.4 The EIRENE mobile installation shall be designed to operate in a network meeting the criteria in 3.2.2 and 3.2.3. (M) Note 1: The specified coverage probability means that with a probability value of at least 95% in each location interval (length: 100m), the measured coverage level shall be greater than or equal to the figures stated above. The coverage levels specified above consider a maximum loss of 3 dB between antenna and receiver and an additional margin of 3 dB for other factors such as ageing. (I). Note 2: The values for ETCS levels 2/3 concerning coverage and speedlimitations are to be validated and, if necessary, reviewed after the first operational implementation of ETCS (I). D. Downlink and uplink link budgets for GSMR equipment For the mobile cab radio (8W transmit power) the maximum permissible pathloss in the uplink and the downlink are almost equal (around 152 dB). This helps to ensure that there is good balance between the qualities of reception, at either end of the call. For the portable handheld radio (2W transmit power, similar calculations result in a 6 dB unbalanced link budget, uplink limited (151 dB maximum downlink pathloss, 145 dB maximum uplink pathloss). Each of the BTS and Mobile Station features listed above are represented in the ICS Telecom software, through the radio stations and subscriber parameters. The BTS antenna used throughout the simulation is Kathrein 739623 (65 degrees horizontal HPBW half power beam width, 17 dBi gain, crosspolarized).

An example of mobile subscriber representation (8W cab radio), with all its radio and traffic parameters is presented in Fig. 1.

Figure 1. GSM-R Subscriber representation in ICS Telecom

The radio planning has been carried out for two different values of the network coverage threshold: Minimum design signal level accepted is 92 dBm for cab radio, 8W, receiver at 4 meters above terrain; Minimum design signal level accepted is 77 dBm for handheld outdoor, 2W, receiver at 1.5 meters above terrain.

E. Propagation models used for GSMR radio planning The propagation models of choice are deterministic point topoint models: ITUR P.525 for free space attenuation [2]; ITUR P.526 Deygout for diffraction geometry and subpath attenuation corrections [3]; Rain attenuation (ITUR P.838) [4]; Link reliability calculations (ITUR P.530) [5].

The factors that determine the propagation model selection are explained at large in [6]. In the following, we will only resume our reasoning: When the terrain and clutter can be explicitly described, the ITU-R P.525/526 will typically be the best choice. It has been largely described in the paragraphs above, that both the DTM and radio clutters available (and that can be made available) for the GSM-R radio planning, have the appropriate level of detail required to explicitly describe the path between transmitter and receiver, when the assumption is made that a receiver is at a particular location. The assumption that the receiver is at a particular location is true, as the prospective planning procedure relies on subscriber generation along the railway tracks (represented by vector lines). ICS Telecom function Generate subscribers along vector line

(Subscriber menu in ICS Telecom main window), allows even the modeling of the distance among the generated subscribers (see Fig. 2). 10528 subscribers have been generated along the vector line representing the Bucureti-Constana railway corridor (one subscriber every 25 meters).

II), from each digital map pixel identified as belonging to the candidate areas for the placement of GSM-R base stations. The prospective planning procedure configuration window allows the radio planner to set the following main parameters (see Fig. 3): terrain filter, base station parameters, site and base station deployment constraints, and subscriber parenting rules.

Figure 2. ICS Telecom function Generate subscribers along vector line

In this respect, its worth mentioning that the 95% coverage probability, as understood in EIRENE SRS [7], has no equivalent when using deterministic pointto-point models. In our case, 4 subscribers are generated on each 100 meters of railway track, thus exceeding the granularity required by EIRENE SRS.

Figure 3. ICS Telecom function Prospective planning from subscribers configuration window

The accuracy of the selected propagation models (ITU-R 525/526) can be highly improved by model tuning. The propagation model tuning has been done using iterative modification of the terrain factors, i.e. clutter attenuations, as recommended by ICS Telecom user documentation, the coverage prediction results being correlated to actual drive test samples (52046 samples measured for five different GSM based stations, covering rural areas, comparable to railway environment). The objective of the propagation model tuning is to achieve a negative mean error and a standard deviation not exceeding 3-4 dB, in other words, the simulation results should be slightly pessimistic when compared with the measurement results (drive tests). III. DESCRIPTION OF THE GSM-R PROSPECTIVE RADIO

Automatic sites deployment procedure result: after several iterations, the procedure determined 14 base station locations (sites), from which 9924 subscribers are parented, out of the total 10528 generated subscribers (94.26%). The automatic planning procedure cannot go further, as it reports Remaining subscribers density too low. The next step in the GSM-R prospective radio planning procedure was the manual deployment of two additional sites to cover most of the remaining subscribers, in the dense urban area of Bucureti railway terminal, shunting and depot areas. This deployment increased at 16 the total number of base station sites. The final outcome of the GSM-R prospective radio planning procedure was: 16 base station locations (sites); 10166 subscribers parented out of the total 10258 generated subscribers (96.56%). IV. GSM-R RADIO NETWORK OPTIMIZATION

The GSM-R prospective radio planning procedure can be initiated after the following steps are accomplished: The digital cartography set is manipulated, as shown in Chapter II, to represent with the maximum possible accuracy, the railway infrastructure, both as vector files and radio clutter. Subscribers are generated along the railway vector lines, and subscribers parameters are set to the values of real GSM-R equipment (see link budgets above).

The following steps were performed in order to optimize the GSM-R network resulted from the prospective radio planning procedure: A. Base station sectorization Each base station was configured with two sectors, oriented back-to-back along railway tracks, configured with the chosen sector antennas (Kathrein 739623). The sector orientation was automatically optimized using the Station azimuth optimizing function (Subscriber Parenting menu from the ICS Telecom

The best base stations positions are determined according to the number of generated subscribers that can be point-to-point parented (using the deterministic models specified in Chapter

main window). This ensured that each sector is optimally oriented to connect the largest number of subscribers in its service area, along the railway tracks. B. GSM-R network radio coverage calculation The radio coverage was calculated based on the same propagation models, and the same radio parameters of the base stations, that were used for the GSM-R prospective radio planning procedure, again in accordance with EIRENE SRS. The radio coverage map is represented using a color palette for different signal levels. Fig. 4 presents a detailed view of radio coverage, showing how the base station position and sector orientation were optimized to exactly cover the railway track, the desired service area.

only. We should always try to configure with only one cell per station where this is possible from a frequency and capacity planning point of view. Also, at high speeds handover between cells on same site (intra-BTS handover) should be avoided, and this is another reason for not splitting cells in rural high-speed areas. The handover between BTSs is easier to perform since the signal strength of the serving and neighbouring cells are more or less equal during a longer period of time. The frequency planning strategy was to apply a reuse of 8, meaning that the minimum frequency spacing between the two channels of the same cell is set at 1.6 MHz. The frequency allocation was done automatically, using the Network assignment Band assignment function (Coverage menu, in the ICS Telecom main window). Channels 955, 971 and 973 were deliberately excluded from the attributable channels, in order to allow for further development and interference mitigation where need may impose that. The spare channels were chosen such that they can all be allocated in the same site. Additional settings for the automatic frequency assignement: C/I mask (as for the GSM systems, C/IC > 12 dB, C/IA > 6 dB; Coverage threshold: 92 dBm; As we configured one cell per station, and each cell is distributed on two sectors facing oppositely along the rail tracks, we should not account for interferences between two sectors of the same cell. In ICS Telecom, this is achieved through setting the same Network ID for both sectors of the same cell, and allowing the allocation of the same frequencies for both sectors of the same cell. Following the frequency assignment, a global interference level of 0.15% is reported by ICS Telecom. This is a negligible value, taking into account that this value is reported for the whole area covered with a radio signal greater than or equal to 92 dBm. GSM-R RADIO NETWORK ANALYSIS

Figure 4. Radio coverage detailed view railway track coverage in a forested area

C. Frequency assignment In accordance with EIRENE SRS, the UIC frequency band allocated for GSM-R is: 876 880 MHz (mobile station transmit); paired with 921 925 MHz (base station transmit).

As per the aforementioned references, the carrier frequency is designated by the absolute radio frequency channel number (ARFCN). For carriers in the UIC frequency band, the following convention shall be used, where Fl(n) is the frequency value of the carrier ARFCN n in the lower band, and Fu(n) the corresponding frequency value in the upper band (frequencies expressed in MHz): Fl(n) = 890 + 0.2*(n-1024) 955 n 973 Fu(n) = Fl(n) + 45


The resulting frequency table contains a total of 19 channels, with 200 kHz bandwidth and 45 MHz duplex spacing. For all stations along the rail track in rural areas and for other stations where the capacity demand is low, i.e. maximum 2 carrier units, there shall be one cell

A. Interference analysis The interference analysis was carried out using the Network interference C/I mode function (Coverage menu of the ICS Telecom main window). The interference analysis reports that 1.3% from the desired service area (railway clutter) is subject to interference, a fraction which is also negligible. A more detailed interference report is also available in tabular format, specifying the percentage of the service area interfere, for each base station and each radio

channel. The highest interference percentage for one sector is 0.56%, again negligible from the radio planning perspective. B. Handover map Handover maps for each cell can be produced using the Coverage Handover function (available in the popup menu following the selection of a base station). The result is available both in graphical and tabular formats, and indicates each base stations neighbors, and the surface of the handover area (see Fig. 5).

D. GSM-R network radio planning for the -77 dBm coverage threshold Following step-by-step the radio planning procedures described herein, the resulted network can be dimensioned at 25 sites, with at least 50 sectors. Such increase in network dimensions was to be expected, because of the 15 dB increase of the coverage threshold (in order to have a balanced uplink/downlink link budget), and the decrease of the subscriber antenna height, from 4 meters for the cab radio, at 1.5 meters for the handheld portable radio. VI. CONCLUSIONS AND FUTURE WORK

Figure 5. Handover map graphical representation

Radio planning for the GSM-R networks requires a much higher precision for the digital cartography set, than the one required for commercial GSM networks, mainly because of the extremely stringent radio coverage requirements specified by EIRENE SRS. The manipulation of the digital cartography to represent as exactly as possible the railway infrastructure, both as a set of vectors, and as a clutter layer, is described in detail in the paper herein, providing an original and easy to follow procedure and the solutions to generate the required geographic data. Although the radio planning carried out throughout the paper is just a case study for the Bucureti Constana Railway Corridor its results can be considered accurate enough at least to dimension an equipment and services acquisition for the implementation of such network. The radio planning deliverables are original by themselves, their accuracy being guaranteed by the use of deterministic models, accurate cartography and a procedure that determines the best site locations/ sectors orientation from the subscribers locations perspective. Future work is intended to refine the radio planning procedure in order to include tunnels coverage, as well as to produce even more accurate geographic datasets, to represent the Romanian railway infrastructure. ACKNOWLEDGMENT The Romanian Special Telecommunications Service, which provided the authors free access to its ICS Telecom and geographic data resources. REFERENCES
[1] Performance Specification Digital Terrain Elevation Data (DTED), MIL-PRF-89020B, USGS National Imagery and Maping Agency, 23.05.2000. International Telecommunications Union, Recommendation ITU-R P.525-2: Calculation of free space attenuation, 1994. International Telecommunications Union, Recommendation ITU-R P. 526-11: Propagation by diffraction, 2009. International Telecommunications Union, Recommendation ITU-R P. 838-3: Specific attenuation model for rain for use in prediction methods, 2005. International Telecommunications Union, Recommendation ITU-R P. 530-13 (10/09): Propagation data and prediction methods required for the design of terrestrial line-of-sight systems, 2009. A. Graham, N. Kirkman, P. Paul, Mobile Radio Network Design in the VHF and UHF Bands, John Wiley and Sons, Ltd, pp. 37-71, 2007.

The handover map indicates that the handover area is correctly positioned, between the base stations successive along the railway tracks, and is large enough (e.g. 47.27 square kilometers between BS7 and BS8, representing 23.41% of the BS7 coverage area, for a 5 dB handover margin). Such parameters guarantee the handover success rate of 99.5%, as required by EIRENE SRS. C. Railway tracks coverage analysis The percentage covered from the desired service area (clutter 11 railway tracks) is reported to be 98.21% (see Fig. 6). More sophisticated reports, for each point along railway tracks, can be generated using the Vector layer FS received on vector lines function (Map menu of the ICS Telecom main window).

[2] [3] [4]


[6] Figure 6. Railway tracks (clutter representation) percentage covered