GSM Cellular Mobile Network Simulation with CellSim++

George Grozev; Bertil Marksj; Shawn Foo

CSIRO BCE, Graham Rd, PO Box 56, Highett, VIC 3190 [email protected]; [email protected]; [email protected] Abstract. This paper presents a simulation approach to the object-oriented modelling of GSM cellular mobile networks implemented in the GSM CellSim++ simulation tool. Key aspects of this tool include: (i) the modelling of the random nature of radio wave propagation, mobile locations and user calling behaviour; (ii) modelling performance enhancing techniques such as frequency hopping, power control, and advanced antenna systems; and (iii) reporting statistics on the quality of service in the network. GSM CellSim++ was developed by CSIRO Infrastructure Systems Engineering (ISE) in collaboration with Telstra Research Laboratories. It allows cellular radio network designers to perform "what-if" scenarios on existing or proposed future network configurations. This simulation platform is the first in a series of applications under development including CDMA technology. 1 INTRODUCTION The MATLAB simulation toolbox GSMsim [3] works at the device level. It performs evaluation of the transmitter and receiver front-ends of GSM systems. The GSMSIM simulator of digital cellular networks is described in [2] and works at the system level. It takes as input customer density, mobility models, positions and parameters of base stations, and field strength. The simulator calculates the load of the cells, number of handovers, and some power control parameters. It was implemented in C++ and was quite advanced for the time when the first commercial GSM systems were emerging.

The global system for mobile communication (GSM) is now a worldwide-accepted standard for digital cellular systems and is based on the time division multiple access (TDMA) technology [5]. The first commercial GSM system was installed in 1991 in Europe. Today there are hundreds of millions of customers using this technology in their every day communications worldwide. There is a lot of information dedicated to the description of GSM. Helpful introductions to cellular communications and to GSM technology are given in [4],[6]. GSM cellular radio networks are complex systems that cannot be described by exact models or equations. One of the reasons for this is the random behaviour of signal propagation. The fundamental reason behind this behaviour is the complex interactions (reflection, absorption, diffraction, dispersion, etc.) of electromagnetic waves in different environments. This is particularly true for human-created settlements like cities and villages, where the variety of structures, shapes, and materials is huge. Peoples mobility and call behaviours, introduce additional random factors adding to the complexity. Unavoidable interaction for GSM cellular networks is the interference arising from reusing scarce frequencies many times over a network. Computer simulation has proven to be a very productive approach to modelling complex systems like GSM cellular radio networks. A number of simulation tools have been reported, which model different aspects of GSM systems. Examples include: The ParFlow++ software for simulation of radio wave propagation in urban micro-cells is described in [1]. It works at a physical level, simulating electromagnetic waves by spherical wavelets emitted by secondary radiators. This parallel simulator was run on a supercomputer and is obviously not generalpurpose software that can easily be accessed and used.

CSIRO ISE in collaboration with Telstra Research Laboratories have developed the GSM CellSim++ simulator to answer a broad range of what-if questions in planning, design and operation of GSM cellular radio networks. It works at the system level, similar to GSMSIM, however it puts the emphasis on different performance measurements. This tool allows configuration and simulation of complex networks taking into account the random nature of signal propagation and user calling behaviour. Working with the system the user can create and configure the main network components like base station and mobile stations, select frequencies, model antennas, select models for signal propagation, define power control, etc. GSM CellSim++ offers a friendly and easy to use graphical user interface (GUI) in the Windows NT operating system. 2 OBJECT MODEL

GSM CellSim++ has been developed using the objectoriented modelling and design approach. Its high-level class diagram for telecommunication entities is given in Figure 1. Cellular communication networks contain many sites. One site usually has up to three base stations each servicing one adjacent cell. Every base station is assigned a set of different radio frequencies. In order to reuse the spectrum of frequencies and to increase capacity, base

stations that are at sufficient distance may reuse the same frequency. The reuse is possible only if the interference between the same and adjacent frequencies is at or below acceptable levels. The cell geometry class assigned to every base station defines the geometry of every cell. All mobile stations that are placed in a given cell can be connected to the corresponding base station via servicing channels (representing a combination between a frequency and time slot). One mobile station is specially selected as a probe mobile aiding gathering statistics during the simulation.
Network

2. 3. 4.

1..n

Antenna Pattern

Site

Power Control

5.

1..n

Path Loss Model

Base Station

1..n

Channel

1..n

Shadow Fade Model

Mobile Station

Cell Geometry

6. 7. 8.

Figure 1: High Level Class Diagram Every base station is assigned an antenna pattern, a cell geometry, and a set of channels. Different path loss models, shadow fade models and power control parameters may be selected and configured for every base station. The user can define his/her own antenna pattern files. Following some rules, the user can create new path loss and shadow fade models and plug and play with them. There are many more classes in the system but it is not appropriate to describe all of them in this section. One big set of classes is related to the graphical user interface, another is connected to the simulation manager. 3 3.1 CELLULAR MOBILE RADIO FEATURES How GSM CellSim++ Works

Create a base station for this cell site. Create a mobile station and choose it to be a probe mobile. For the created base station, change some properties if necessary. Select a path loss model and a shadow fading model for this base station. Select the uplink (from a mobile station to a base station) and the downlink (from a base station to a mobile station) power control algorithms. Assign frequencies to the base station. Switch on/off the slow frequency hopping feature and set hopping parameters. Change the call blocking probability for the corresponding cell and select discontinuous transmission if desired. Change the azimuth and elevation antenna pattern files. Create other cell sites, base stations and mobile stations if required and change their properties if necessary. If the number of created mobile stations is small, additional mobile stations will be generated for every configured cell and positioned inside the cell using random coordinates. Select visualisation options. Run a specified number of realisations. Analyse the distributions for uplink and downlink performance (C/I) values (see Section 3.4).

Not all steps from this sequence are always necessary. For example, once it is prepared the network configuration can be saved in a file and later loaded from the file. 3.2 Link Budget

The simulation in GSM CellSim++ examines both the downlink and uplink radio signals. The performance parameter of most important interest is the carrier-tointerference ratio C/I for both links. The formulas in this section calculate the wanted signal (the carrier C) and the interference I. An important objective during design and operation of cellular mobile networks is to improve the C/I ratio in all cells. The downlink is calculated based on the following equation (see Figure 2):
m Prx

Ptxb,c

Lbf

b Lb Gant ,tx c

m L p (d ) Ls Gant ,rx

Lm , f

There are four main steps when doing simulation with GSM CellSim++: (i) create the network objects (cell sites, base stations, mobile stations); (ii) configure the network objects by selecting values and models for different properties; (iii) run a number of realisations (system states); and (iv) analyse the resulting statistical distributions of performance values. Usually many realisations are performed to collect valid statistics. Every realisation is a snapshot picture of the mobile network configuration instantiated by drawing random numbers for all variable components (see Section 3.5). Here we explain the basic functionality of GSM CellSim++ following a simple typical scenario of steps for simulation that the user may perform: 1. Create a cell site at a selected location.

where superscript b denotes a property of the base station, superscript m denotes a property of the mobile station, m Prx is the received power by the mobile station m
b measured in dBm, Ptx ,c is the transmit power (in dBm) by the base station b per channel c serving the mobile station, Lbf is the base station feeder loss in dB, Lb is the base c

b station combiner loss in dB, Gant ,tx is the base station antenna gain for transmission in dBd, L p (d ) is the path

loss in dB for a distance d between the base station and the mobile station (for street-scape path loss the value depends on the position as well), Ls is the shadow fading loss describing the performance variation of the signal m due to natural terrain and human-made structures, Gant ,rx

is the mobile station antenna gain for reception in dBd, and Lm is the mobile station feeder loss in dB. f
b Gant ,tx

the total interfering signal generated by all base stations bj mk to the mobile station mk we can use the

bg
b

Base station
Lp (d )

values given by the formula for the downlink in Section 3.21:

antenna

Mobile station
antenna

I mk

I mk bk , ck , mk

bj

m bk g

m Prx k .

(1)

Ls
feeder

m Gant ,rx

Lb f

Lb c
b Ptx,c

Lm f
m Prx

If the base stations are not synchronised, then GSM CellSim++ assigns a random time offset t (between zero and one time slot length) for every base station. In this case, the set of interfering base stations mk is expanded to accommodate some additional base stations having co-channels and adjacent channels that use adjacent time slots to the channel ck , and which are at least partially concurrent with the time slot of channel ck .

bg

combiner

feeder

Figure 2: Downlink budget The uplink is calculated based on a similar equation:

b Prx ,c m Ptxm Lm Gant ,tx f b L p (d ) Ls Gant ,rx

The interference for the uplink channel ck connecting the mobile station mk to the base station bk is generated by all mobile stations m j bk , m j mk , where bk is

bg

bg

Lbf

Lb , c

b where Prx ,c is the received power by the base station b

measured in dBm for the serving channel c, Ptxm is the transmit power by the mobile station m in dBm, L f , Lc , L p (d ) are the corresponding feeder loss, combiner loss, and path loss, all in dB, and Gant is the corresponding antenna gain in dBd. 3.3 Calculating the Interference

the set of mobile stations having the same or adjacent frequency and the same or adjacent time slot. Using the values given by the formula for the uplink in Section 3.2, the total interfering signal is2:

I cbkk

I cbkk mk , ck , bk

mj

b bk g

b Prxk, c k .

(2)

Signal interference is an essential unavoidable factor in GSM cellular mobile networks and is present because a given frequency is reused many times in different cells/base stations. For example, a mobile station receiving the wanted signal from a base station close by, may also receive interfering (unwanted) signals from other base stations that transmit on the same channel (the same frequency and time slot). Let us now describe the interference that a given mobile station mk receives from other base stations. This mobile station is connected to the wanted base station bk via the channel ck defined by a specific frequency and time slot. In order to do this let us first assume that all base stations are synchronised, i.e. they start transmission of the radio signal for the first time slot at the same time and as a result all consecutive time slots are also synchronised between different base stations. The interference is of two types co-channel interference (the same frequency and the same time slot) and adjacent channel interference (the adjacent frequencies and the same time slot). So, all base stations, except the wanted, which have active cochannels or adjacent channels contribute to the interference of the wanted signal for the given mobile station mk . Lets denote this set by mk . To calculate

We have discussed synchronised and non-synchronised base stations so far in this section. GSM CellSim++ also allows simulation of site-synchronised base stations. In this case all co-sited base stations are synchronised between themselves, however there is no synchronisation between sites. 3.4 Carrier-to-Interference ratios

The C/I ratio distributions for the uplink and downlink are the main simulation output from GSM CellSim++. For a mobile station m the downlink C/I ratio is given by the formula:

C FI GJ HK I
m

m Prx Im Nm

(3)

m where Prx is the wanted signal from Section 3.2 in W, I m is the interference according to formula (1) in W, and N m is the mobile station noise in W. For a given base station b and channel c the uplink C/I ratio is given by the formula:

C FI GJ HK I
b c

b Prx ,c I cb N b

(4)

bg

1 2

m Prx k is converted from dBm to Watts. bk Prx , c k is converted also from dBm to Watts.

b where Prx ,c is the wanted signal from Section 3.2 in W,

I is the interference according to formula (2) in W, and N b is the base station noise in W. C/I values from these
two formulas are converted to dB for final display and analysis. 3.5 Random Nature of the Signal Propagation

b c

characteristics of this behaviour. One of the GSM features is the discontinuous transmission (DTX), which is a technique to transmit a signal only when the user is talking. Typically, a person speaks for less than 40% of the time in a normal conversation. This technique reduces the interference and also helps to conserve power. The mobility of mobile phone users is modelled through random coordinates that are generated for every mobile station for every realisation. There is an option to fix a mobile station at a desired position for the whole duration of the simulation. Height variation of the mobile station can be randomly generated for every realisation. The traffic level on each cell is described by the call blocking probability. It allows for some mobile stations to fail to be connected because all channels are already busy. GSM CellSim++ uses the Erlang B model [5] a model of serving without queuing, where all blocked calls are lost. Based on a specified blocking probability for every base station, the number of active mobile stations in the corresponding cell is calculated. 3.7 Conventional and Smart Antennas

The mobile radio environment is very complex, especially so when close to the ground in urban and suburban areas [5]. There are many theoretical models that try to describe and evaluate radio signal propagation especially for cellular mobile networks. GSM CellSim++ is designed and implemented to enable simulations for different environments and using different theoretical models. For example, for the median path loss (often called the path loss) the following different models [7] are implemented: free space path loss model Hata model Egli model street-scape path loss model.

Shadow fading (variation in the signal level due to the mobile moving) is another component of the link budget. The following options for shadow fading are provided in GSM CellSim++: correlated uncorrelated.

The selection of a path loss model and a shadow fading model can be specific for every base station. Some of these models have parameters that can be further changed. The fast (multi-path fading) is modelled by the Rayleigh distribution and this type of fading is often called Rayleigh fading. GSM CellSim++ performs many realisations to build distributions of C/I ratios of the links. Every realisation is a particular instance of signal propagation. The following components of the network configuration vary from realisation to realisation due to the random nature of signal propagation: locations of mobile stations variations to physical parameters that occur accross the network, e.g. arrival angle for the base stations, mobile antenna height, etc. shadow fading loss Rayleigh fading loss.

GSM CellSim++ allows a different conventional antenna (antennas of unchanging antenna gain) to be used for every base station. Two ASCII files (for azimuth and elevation planes) provide the definition of the antenna radiation patterns. The selection of a conventional antenna for a given base station is equivalent to the selection of appropriate azimuth and elevation pattern files via the base stations GUI property page. For the elevation pattern file, the sensitivity of the antenna is specified for angles in the interval [-90, +90] degrees, while for the azimuth pattern in the interval [0, 360] degrees. GSM CellSim++ also allows simulation of switched beam antennas. These are instances of smart antennas having a predefined and fixed number of narrow beams (or patterns). For any call only one beam is active depending on the signal environment. Smart antennas, in general, combine antenna arrays with digital signal processing providing adaptive transmission and reception in different directions. They provide the benefits of reduced interference and increased antenna gain. Only traffic channels (TCH carriers in the GSM technology) can be transmitted/received by a switched beam antenna. For the broadcast control carriers in GSM (BCCH carriers in the GSM technology), conventional antennas must be used.

Values for the above components are selected every realisation by drawing random numbers from corresponding distributions. 3.6 Users calling behaviour

The variety and unpredictability of users calling behaviour is an important factor to take into account. GSM CellSim++ models some of the most common

cellular mobile network. In reality the power control is constantly changing the power levels. Our implementation cannot fully simulate this dynamics and it allows only a certain number of power control iterations before collecting statistics. 4 IMPLEMENTATION

Figure 3: Switched beam antenna In GSM CellSim++ it is possible to work with a variety of different switched beam antennas. Azimuth and elevation pattern files, direction, and gain value describe every switched beam. The number of switched beams is also configurable. An example of visualisation of a switched beam antenna in GSM CellSim++ is shown in Figure 3. There is one cell site that has three base stations one base station for every sector of the cell site. The switched beam antenna is shown in the right sector pointing to the probe mobile station. 3.8 Frequency Hopping

GSM CellSim++ is implemented using the C++ programming language and the Microsoft Visual Studio environment. It runs under Windows NT operating system. The main components of the system are classes and they can be grouped into three main sets. Every set can well represent a separate layer according to the idea of layering architecture of software systems. The main sets of classes are simulation classes, mobile network objects, and GUI classes. The mobile network objects have some specific methods for file input/output. The object-oriented design and implementation of GSM CellSim++ allows new features to be easily accommodated. New types of path loss models, shadow fading models, and power control can be added straightforwardly. The mobile network objects can be used as a core for other simulation systems, for example event-driven simulation of GSM.

Frequency hopping [7] is an efficient way to combat the effects of signal fading and interference, providing both frequency and interference diversity. When frequency hopping is applied, the carrier hops from one frequency to another in a predefined sequence. The slow frequency hopping method is implemented in GSM CellSim++. Multiple-frame simulation in a single realisation provides functionality to study the impact of slow frequency hopping on the quality of the radio link. 3.9 Power Control Figure 4: Main screen display The GUI provides a friendly and flexible environment for modelling of cellular radio networks by using standard Windows user interface features. By point-and-click operations you can easily create cell sites, base stations, or mobile stations and edit their properties. One view of the GSM CellSim++ main screen is given in Figure 4. It contains the main window (the network layout) and the C/I histograms window on the right. The main window has three cell sites (represented as small towers) visible on the screen. One of the cell sites (the left top) has three base stations and the other two have only one. The base stations are represented by multi-line patterns indicating signal radiation. The small objects, scattered in the cells are the mobile stations. There are two histograms and the corresponding cumulative graphs for the downlink and

GSM CellSim++ provides the following options for power control: power control based solely on the received signal level. It takes into account only the wanted signal and increases or decreases step-wise the corresponding transmit power so the level of the wanted signal falls into a certain desirable interval. power control based on the received signal level and the quality of the radio link. It takes into account the quality of the received signal (both the wanted signal and the interference) and corrects the transmit power in the desired direction (up or down).

The options for the uplink and downlink are independent. GSM CellSim++ assumes a certain steady-state of the

uplink C/I distributions as measured for the probe mobile station during the simulation. The attributes of every object (network, cell site, base or mobile station) can be viewed and edited via corresponding dialog boxes. The main property page of the base station class is shown in Figure 5 as an example. It has Apply Selected and Apply All buttons that are very useful for assigning the same attribute values to a subset or to all other base stations. Separate property pages are provided to select, view, and configure antenna patterns, carriers, slow frequency hopping, teletraffic, propagation model (path loss, shadow fading, and Rayleigh fading), and power control.

recent simulation platform in the series. Possible future extensions are in the area of 3G technologies and for more extensive simulations based on the event-driven approach. A CellSim++ brochure is available at:
http://www.dbce.csiro.au/ind-serv/brochures/cellsim/cellsim.htm

ACKNOWLEDGEMENTS

GSM CellSim++ has been developed under a research and development contract with Telstra. The authors are thankful to John Campbell from Telstra Research Laboratories for his key input in the development of the system. The authors would like to thank Robert Parker, Steven Oaks, Cheryl McNamara, Jessica Guo, and Tony Spehar for their substantial contribution in the design and implementation of the GSM CellSim++ simulation tool and Miles Anderson for his review of the initial draft of the paper. REFERENCES
1. Calgari, P., Guidec, F., Kuonen, P. (1997) Urban Radio Network Planning for Mobile Phones, EPFL Supercomputing Review, No. 9, Nov. Drozdy, G., Niemel, J., Vlimki, J., et all. (1992) Study of GSM System Performance by GSM Network Computer Simulator, IEEE vol 2, pp. 336-339. Ekstrm, A.N., Mikkelsen, J.H. (1997) GSMsim a MATLAB Implementation of GSM Simulation Platform, Technical report R-97-1004, Aalborg University, Denmark. International Engineering Consortium, Global System for Mobile Communication (GSM), www.iec.org/tutorials/gsm; Cellular Communications, www.iec.org/tutorials/cell_com. Lee, W.C. (1993) Mobile Communications Design Fundamentals, second edition, John Wiley & Sons: New York. Scourias, J. Overview of the Global System for Mobile Communications, http://ccnga.uwaterloo.ca/~jscouria/GSM/gsmreport.htm l Yacoub, M.D. (1993) Foundations Engineering, CRC Press: Boca Raton. of Mobile

2.

3.

4.

Figure 5: Base station controls main property page

5.

SUMMARY AND FUTURE WORK

6.

In this article we have presented the GSM cellular mobile radio simulation tool CellSim++. This system allows modelling and simulation of a wide range of characteristics of GSM networks. It can be used to quantify the performance of different radio network features (e.g. slow frequency hopping, switched beam antennas, power control, etc.) and the settings of the respective radio parameters. It offers a friendly and easy to use graphical user interface in the Windows NT operating system environment. GSM CellSim++ is also effective training tool for GSM networks. Telstra Research Laboratories, a key contributor to the technical knowledge within GSM CellSim++, have run many simulations exploring different technologies and features for Telstras GSM networks. CellSim++ is a series of applications for simulation of cellular radio technologies. CDMA CellSim++ is the most

7.