Reading Time: 29 minutes

Abstract

The aim of this research paper is to investigate how we can improve the capacity of a cellular system in order to meet the growth of subscribers on a cellular network.  The concept of a cellular design was found to improve the scalability of the wireless network by reducing the frequency reuse distance.  Directional antennas were found to have a lower cell capacity compared to omnidirectional antennas using one calculation method, but a higher grade of service due to reduced co-channel interference.  We also invested smart antennas using SDMA beam forming technology to create lobes which follows the movement of mobile users.  Providing there is sufficient distance between mobile users, the same frequency can be reused within a sector.  A Cell Sim application was created to analyse the performance of various cellular configuration scenarios.

Objectives

1. Cluster size
2. Antenna configuration- omnidirectional & sectorised
3. Investigate and implement suitable cell simulation code to analyse the performance characteristics of various cell configurations
4. Review and critically analyse the current and past primary research on smart antenna systems (switched beam and adaptive array)
5. Investigate modulation techniques used to improve system capacity (FDMA, TDMA, CDMA, SDMA)
6. Identify the operation and performance of current omnidirectional & directional antennas
7. Measurements of SIR
8. Capacity calculations
9. Critically analyse results and form conclusions

Introduction

Mobile communications have evolved significantly since Marconi’s first radio transmission to a tugboat in 1897.  The first land mobile communication system was based on wide area transmission where each base station provided coverage for a large autonomous geographical zone.  Calls from customers leaving a zone had to be dropped and re-established in a new zone.  (Brand & Aghvami, 2002, p. 45), these systems suffered with low capacity.  By the 1970’s cellular mobile communication systems were introduced which allowed the capacity of the systems to be significantly increased and this led to the growth in mobile telecommunications over recent years.  As the growth of mobile communications continues with both voice and data communication, we can expect mobile systems to be pushed beyond their current capacities.  In order to meet this growth, a cost effective infrastructure is required to enhance network capacity and coverage is required.  In addition, 3G will also be prone to problems of spectral congestion as the number of subscriber’s increases and services expand.  (Perini, 2000, pp. 2-875)

Mobile communication is available globally; this is achieved via a cellular framework.  Mobile units are wirelessly communicating to base stations and the base stations are typically hard wired to the Public Switched Telephone Network (PSTN).  Our interest for this assignment is the wireless communication between the mobile units and base station.  In particular the technological advances which has allowed the wireless capacity to grow using directional antennas and digital processing.  This allows efficient use of the frequency spectrum to increase capacity within a cell.  The future of mobile communications is dependent on network planning and mobile radio equipment design that will enable efficient and economical use of the radio spectrum (Lee, 1997, p. 6).

Part I – Secondary research

Cellular Design

Mobile communication is a radio link typically between a base station (BS) being a fixed terminal, and a mobile unit (MU) being a terminal of an undefined location.  An important aspect of radio communication is maintaining a capacity to ensure Quality of Service (QoS) with a blocking probability of less than 2%.

Since the spectrum available to mobile operators is of limited resource there is a demand to ensure that capacity is efficiently managed to maximize the number of subscribers within a single base station.  One method of achieving this is by the cellular concept which is a service area covered with many small zones or cells (Akaiwa, 1997, p. 369).  A cellular design provides the means for users moving at various velocities to maintain mobile communication with a base station.  By dividing geographic regions into hexagonal cells, each cell has its own omnidirectional antenna.  The theoretical hexagonal cells allow us to calculate the distance between cell clusters.  A hexagonal geographic layout, as shown in  Figure 1, is optimal in the sense that the cell cluster size is minimum under the condition of a given co-channel interference (Akaiwa, 1997, p. 370).  Adjacent cells are assigned different frequencies to reduce co-channel interference.  The objective of cellular networks is to allow frequency reuse in nearby cells to increase capacity, whilst minimising impact on the QoS which can be achieved by managing the tradeoffs such as hand over and co-channel interference.

Figure 1 Cellular system  (Dunlop & Smith, 1994, p. 517)

An important objective of frequency reuse is determining the cellular distance between cells using the same frequency.  We can determine the number of hexagonal cells per group (N_g) as shown in Equation 1, where i and j are vectors shown in Figure 1.

R = Radius

D = distance between the centre of two cells

Equation 1 Number of cells per group

Current Antenna Technology – Omnidirectional & Directional Antennas

An ideal half wave dipole antenna will have an omnidirectional radiation pattern and if positioned in the centre of a cell it will radiate power uniformly over a horizontal plane as shown in Figure 2. The design of an omnidirectional antenna is inefficient since the range can only be increased by transmitting power in all directions.  The transmitted power needs to be carefully controlled to minimise co-channel interference between adjacent cells.  As more subscribers use the system there is the risk the base unit will have insufficient frequencies to cope with the demand resulting in new connections being rejected.

A report by (Stallings, 2009, p. 417) states that a number of approaches have been used to increase wireless capacity as listed below; however there is no mention of using adaptive antenna design which would also increase capacity.

• Adding new channels
• Congested cells dynamically borrowing frequencies from adjacent cells
• Cell splitting
• Cell sectoring
• Microcells

Figure 2 Omnidirectional Antenna polar plot

Figure 3 Directional Antenna polar plot

The capacity of the cell can be further increased by replacing the omnidirectional antenna with directional antenna which radiates its signal in one or more directions, thus allowing concentrated signal strength in a given direction as shown in Figure 3.  A base unit can be fitted with three directional antennas each servicing 120 degrees.  Providing the frequency of each transmitter is selectively chosen to reduce co-channel interference the capacity of the network can be increased, as shown in Figure 4.  An omnidirectional antenna would have co-channel interference from 6 adjacent cells; however a directional antenna would only have interference from 2 adjacent cells.

The process of using directional antennas is called cell splitting, this has the advantage of increasing the QoS within a geographic cell.  However, it should be noted that simply increasing the number of sectors for example from three to six will not necessarily double the capacity since there is an increase of handover zones between the sectors.  A report by (Yea, 2001) states that the size of the handoff zones between sectors is a function of the roll-off characteristics of the antenna used.  Therefore, the sharper the main-lobe roll-off, the smaller overlap between sectors.

During a communication link between the MU and BS; a process known as handover may occur.  There are two common causes for this; when a subscriber is moving between cells or the Signal to Noise Ratio (SNR) has dropped below a threshold.  If the cells are split, then the number of handovers will increase; this will cause an increase in the amount of processing required by the MU and BS to maintain a communication link.

Figure 4 Cellular telephone frequency reuse pattern (Wiki)

Multiple Access Techniques

As discussed, the frequency spectrum is a finite resource, which is directly related to the capacity of the system.  Multiple access techniques allow us to improve upon the efficiencies of the system, allowing us to handle more calls.  We shall investigate four multipath access techniques to determine their impact on the system’s capacity, whilst still maintaining a given Grade of Service (GoS).

It would be reasonable for the reader to ask how we measure the blocking probability of a network, unfortunately there is no straightforward answer.  A technique used by (Yea, 2001, p. 32) is to have a network with two carrier frequencies F1 and F2, and an algorithm to control which frequency the new subscriber is assigned too.  Under normal circumstances users are assigned to F1, however if the load of F1 reached a hypothetical threshold of 67%, then new subscribers are assigned to F2.  The theory is, any calls that are assigned to F2 are to be considered blocked calls (Equation 2), and therefore we can expect the probability of the calls to be dropped as the blocking probability.  A high number for T_F2 would mean that we have a high blocking probability, and vice versa.

Equation 2 Blocking probability

Another area of concern is the traffic volume which is measured as the overall call holding time of a system during a measurable time period.  The traffic density is defined as the traffic volume divided by the measuring of time.  The unit of traffic density is called Erlangs (Akaiwa, 1997, p. 365).  To determine the efficiency of the cell under different sectorisation the GoS can be determined by the general Erlang B formula which is given by Equation 3, where C is the number of trunked channels offered by a trunked radio system and A is the total offered traffic in Erlangs.

Equation 3 Erlang B formula

FDMA (Frequency Division Multiple Access)

FDMA manages the calls by dividing the bandwidth into different frequency channels.  AMPS are an example of a first generation cellular system which used FDMA.  When a connection is set up for a mobile user two channels are assigned at f and  f + for full-duplex communication,  users are allowed to transmit on one or more of these channels, as shown in Figure 5.

One of the main inherent disadvantages of FDMA is that these frequencies are assigned to the user throughout the duration of the conversation and cannot be reused by another user until the conversation has terminated.  This arrangement is quite wasteful, because much of the time one or both of the channels are idle (Stallings, 2005, p. 287).  This protocol suffers both from the limited available spectrum and the finite number of channels available.  A report by (Lawrence, 1998, p. 49) states that by the early 1990’s the demand for cellular systems was reaching capacity limit, therefore there was a demand to develop the second generation cellular system using digital traffic channels to increase capacity.  In order to improve the capacity further a technique to share the channels was researched which led to TDMA.

Figure 5 FDMA (Bates, 2002, p. 306)

TDMA (Time Division Multiple Access)

TDMA operates on the principle of multiplexing and allows several users to share the same frequency.  This is achieved by segmenting the transmission signal into α timeslots and each time slot is allocated to a user.  The efficiency of the channel is increased by a multiple of the number of time slots that are being used (Wheat, 2001, p. 92).  The most successful TDMA system is GSM where each user has access to the channel for 577µs every 4.6ms allowing eight users on the same channel (Webb, 1999, p. 71).  There are a number of advantages of using TDMA over FDMA as it allows the data rate of each user to be controlled by adjusting the number of time slots allocated to a user, we can also transmit both data and voice.  (Stallings, Wireless Communication and Networks, 2005, p. 256) reports that frame periods range from 100µs to over 2ms and consist of from 3 to over 100 slots.  Data rates range from 10 Mbps to over 100 Mbps.

TDMA needs to transmit framing information known as guard times (Figure 7) and slot synchronisation (Figure 6) which ensures that any multipath delay will not cause frames to overlap and ensure that data is sent during the correct time slot.  (Webb, 1999, p. 324) reports that in a typical TDMA burst of 148 data bits, only 116 bits can be used to transmit the user’s data, this represents an efficiency of only 78%.  When a user moves between cells, there is no guarantee that the new cell will have any available time slots.  If no time slots are available the user could be disconnected from the network.

Figure 6 TDMA frames, time-slots and bursts (Brand & Aghvami, 2002, p. 48)Figure 7 FDMA channels and guard bands (Brand & Aghvami, 2002, p. 48)

CDMA (Code Division Multiple Access)

CDMA is a spread-spectrum system which uses linear modulation combined with pseudo-random code (also called pseudo-noise) to generate signals.  These sequences (codes) spread the modulated signal over a larger bandwidth which effectively reduces the spectral density of the signal (Godara, 2002, p. 20) as shown in Figure 8.

Figure 8 CDMA signal (Unknown)

The signal bandwidth can be increased in three ways.

1) Switching the frequency in a pseudorandom fashion which is known as frequency hopping.  Adjacent cells can reuse the same frequencies unlike FDMA and TDMA which increases capacity

2) The signal can be transmitted in pseudorandom short bursts known as time hopping.

3) The signal can be pseudo-random coded into a higher frequency known as direct sequence.

In all three cases because the pseudorandom code is deterministic, the receiver is able to find the signal at any point in time or frequency and effectively treat all other transmitted data not using this code as noise by the use of a notch filter.  Adjacent cells can reuse the same frequencies without risk of co-channel interference unlike FDMA & TDMA, which allows an increase in capacity.

Speech naturally has gaps created between words during conversation and is therefore classed as being ‘bursty’.  The background noise during the silent period is roughly the average of transmitted signals from all other users and thus a non-active period in speech reduces the background noise.  Hence, extra users may be accommodated without the loss of signal quality.  This in turn increases the system capacity (Godara, 2002, p. 21).  In addition to (Godara, 2002) report a major disadvantage not discussed is the lack of international roaming capabilities which can restrict a mobile from being used in other countries on GSM networks

Comparison of Radio Capacity between FDMA, TDMA and CDMA

The capacity for FDMA and TDMA systems can be expressed by the channels / cell where C/I is always greater than 1 (Lee, 1997, p. 523).

Equation 4

M₁ is the total number of channels, and K₁ is the frequency reuse factor and is related to the variable C/I, thus

Equation 5

This gives us

Equation 6

The capacity for CDMA systems is expressed as

Equation 7

The frequency reuse factor K₂ is known since D=2R, thus

Equation 8

The total number of traffic channel M₂ is a variable related to C/I for a omnicell system is

Equation 9

For a three sector cell system

Equation 10

Therefore the CDMA capacity shown in Equation 10 indicates that K₂ is known but M₂ and M₃ depend on C/I.

Note: P.G is the processing gain from the spread spread (SS) modulation

SDMA (Space Division Multiple Access)

SDMA uses directional antennas and beam forming technology to produce a narrow beam width lobe to follow mobile users and nulls to cancel out co-channel interference.  Providing there is sufficient distance between two mobiles, the same frequency can also be reused within a sector.

(Brand & Aghvami, 2002, p. 54) reports that SDMA will normally be used on top of other multiple access schemes such as CDMA, TDMA and/or FDMA to increase capacity, as it is not a ‘full’ multiple access scheme in its own right.  Brand continues by ruling out this scheme from the rest of the book on this basis.  However, we believe that SDMA can provide further capacity improvements and warrants further investigation.  A report by (Cooper, 1996, p. 1001) shows that there are a number of benefits to SMDA including, range extension, interference reduction and increased capacity.

The coverage area of the antenna array is greater than that of any element as a result of the gain provided by the array.  When a system is constructed using SDMA, the number of cells required to cover a given area can be substantially reduced.  Since the system can control the radiation patterns to be focused on a user, the average co-channel interference in adjacent cells is reduced; this allows for the channel reuse patterns within a cell to be tighter.  (Cooper, 1996, p. 1001) suggests that moving from a 7-cell to a 4-cell reuse pattern nearly doubles capacity

A ten-element array offers a gain of ten, which typically doubles the range of the cell and thereby quadruples the coverage area (Cooper, 1996, p. 1001).  This is also confirmed in a report by (Bellofiore, Balanis, & Foufz, Smart-Antenna Systems for Mobile Communication Networks Part I: Overview and Antenna Design, 2002, p. 161) who further explains that smart antennas are more directional than omnidirectional and sectorised antennas and thus a range potential increase is possible.  In addition this report explains that smart antennas using SDMA will increase the SIR by simultaneously increasing the useful signal level and lowering the interference level.

A report by (Tachikawa, 2002) finds SDMA provides less spectral efficiency than a conventional system when subjected to high blocking or large re-use distance.  One of the problems with this report is that it only takes into account the forward link and not the download link, and also does not take into account the power control.

Beam forming antennas are a very complex technology.  The benefits of implementing this into a design will yield significant capacity improvements; such as spatially selective transmission and allows the base station to radiate less power.  Another advantage is the possibility of a reduction in the power amplifier size; which would save money.  Research has shown that there is already a significant number of manufactures producing this technology which is called ‘Smart Antennas’.

Part II Primary Research on Smart Antenna System

A smart antenna system can be considered in two parts.  First we have an antenna array with M elements as demonstrated in Figure 9; each element is able to form a radiating signal.  By controlling the amplitude and phase of the signals fed into each element, we can control how many lobes and nulls the antenna will radiate and the azimuth angle and elevation direction of the lobes as demonstrated in Figure 10.

The control of the amplitude and phase can be performed by a butler matrix (switched antenna system) or digital signal processing (adaptive antenna system).  It is the processing of the signal in a smart antenna that actually makes it smart.

The critical advantage of a smart antenna is the easy integration with the existing architectures in base

stations (retro-fit), since only the replacement of the RF front-end is needed (F. E. Fakoukakis, 2005, p. 276).  An added benefit as reported by (Cooper, 1996, p. 1001) is that SDMA is compatible with almost any modulation method, bandwidth, or frequency band including AMPS, GSM, PHP, DECT, IS-54, IS-95, and other formats.  SDMA can be implemented with a broad range of array geometries and antenna types.

Figure 9 Smart Antenna System (Perini, 2000, pp. 2-876)

Figure 10 Radiation pattern (Parini, 2006, p. 80)

Switched Beam Smart Antenna System

A switched beam antenna system comprises of a butler matrix which is used to point the beam both in terms of azimuth and elevation direction.  The system allows the antenna elements to generate a limited number of fixed lobe patterns as shown in Figure 11.  The system switches between these fixed beam patterns in order to select a lobe directed towards a user with the best SNR.  Normally each beam has approximately 3-6dB more gain than the traditional sector antenna (Perini, 2000, pp. 2-877).  Since the beam is focused towards the user(s) and not in all directions there is a reduction in co-channel interference, which improves the GoS and conserves capacity.

A report by (Tutorials, p. 18) suggests that a switched beam system can increase the base stations range from 20 to 200 percent over conventional sectored cells.  This is dependent on the environmental circumstances and the hardware/software used; this is best suited where there is minimal to moderate co-channel interference.  Since the beams are predetermined by the butler matrix, the signal strength varies as the user moves around the edge of the lobe.  This results in a significant degradation of the signal strength before the user is switched to another beam (Perini, 2000, pp. 2-878).

Two important disadvantages of switched beam antennas raised by (Cooper, 1996, p. 1001) are

• The peak gain provided by switched beam antennas are generally less than what is provided by SDMA. This means that the range extension is also going to be less than SDMA.
• The switched beam antennas have further limitations in high interference areas since their fixed patterns prevent the system from adaptively rejecting interference.

Figure 12 demonstrates a cell consisting of three 120 degree directional antennas with 30 degree beams; each of the beams is steered towards the users.  Since mobile 1 and 3 do not fall within the same 30 degree beams, they do not act as interferers.  If the system cannot distinguish between the signal on mobile 2 and interference generated from mobile 3, the C/I could be significantly poor quality.  Thus the beam width, (30 degrees) can limit its interference suppression.  This can however be improved by either increasing the number of antenna elements to eight allowing the system to offer a beam width of ~15 degree, or to use adaptive array antenna system.  This theory is also confirmed by a report by (F. E. Fakoukakis, 2005) who details how a 4*4 system (Figure 13a) which only has 4 elements has the disadvantage that the beam width elevation direction is ~30degrees.  However an 8*8 system (Figure 13b) increases the elevation band width to ~15 degrees.

Figure 11 Switched Beam (Hsieh)

Figure 12 Switched beam smart antenna coverage (Perini, 2000, pp. 2-878)

Figure 13 Switched Beam Smart Antenna Systems block diagram (F. E. Fakoukakis, 2005, p. 276)

Adaptive Array Antenna Systems

An adaptive beam former consists of

1) Multiple antenna arrays

2) Complex weights to amplify attenuate and delay the signals to each antenna element

3) A summer to add all the processed signals together.

An adaptive antenna system monitors its geographic area and using digital processing it can control each antenna element to create an infinite possibility of beams; these beams can be dynamically controlled and adapted to follow the motion of mobile users.  A digitally adaptive beam former works by first minimising the signal to interference C/I thereby cancelling as many interferers as possible.  This allows the system to pass the desired signal with minimum distortion which is called null steering.  The beam former then uses the remaining degrees of freedom to steer the desired beam towards the source; this maximises the background signal to noise ratio (SNR) which is called beam steering.(Perini, 2000, pp. 2-879).

An adaptive array system can increase the range coverage.  A report by (Hourani, 2004/2005, p. 2) states that an adaptive antenna system with eight antenna elements can have a gain of eight (9dB) compared to a single element antenna; however there is no evidence to support this claim.  A densely populated area will have the risk of degraded SIR due to the interference generated by other users within the cell.  Adaptive antennas will on average, increase the SIR.  Experimental results report up to 10 dB increase in average SIR in urban areas.  For UMTS networks, a fivefold capacity gain has been reported for CDMA.(Shamim).

A report by (Perini, 2000, pp. 2-877) suggests that by directing the lobes towards the users; there will be an approx 3-6dB increase in gain.  This means that the amplification can be reduced by 3-6dB at the base station and will provide reduced amplification cost.  Another option would be to reduce the transmission by 3-6dB at the mobile allowing improved battery life.  Another area of research is the development of narrow beam width designs, which will allow smart antennas to reject more signals which are not of interest.  (Bellofiore, Balanis, & Foufz, Smart-Antenna Systems for Mobile Communication Networks Part I: Overview and Antenna Design, 2002, p. 151) suggests that whilst this is an attractive design enhancement in order to achieve a narrower beam angle and larger number of nulls a large antenna array would be required.  The paper explains that this has the disadvantage of increasing the costs and the complexity of the hardware since there is larger computational burden on the beam former.  Whilst the cost and complexity would have been an issue when this paper was written in 2000, the recent fabrication enhancements leading to higher density and lower costs of FPGA’s; would mean this claim would no longer be a significant issue.

According to (Yea, 2001, p. 30) in order to obtain a six sector cell with conventional antennas as much as 18 precisely aligned antennas are required in order to operate efficiently; whereas a smart antenna only three antennas would be required.  Therefore, using smart antennas will provide a cost saving.  A case study by (Yea, 2001) shows the average number of completed calls per day using a 3 sector base station compared to a 6 sector cell shows there was only a 5% increase in completed calls (Figure 14).  Whilst these results are lower than expected, the reason for the low percentage increase is due to QoS issues which results from pilot pollution.  This is a type of co-channel interference in CDMA systems where the signal pilot code from a neighbouring cell or base station has sufficient amplitude to create an interference problem.

Figure 14 Levels of total carried (Yea, 2001)

To analyse the capacity improvements using smart antenna we shall review a simulation by (Bellofiore & al, Smart-Antenna System for Mobile Communication Networks Part 2: Beamforming and Network Throughput, 2002) where the capacity is measured using a variety of antenna patterns and length of training packets.  An ad hoc network consisting of 55 nodes was simulated using OPNET Modeller/Radio tool with the load of each node being poisson distributed.  The values used for the simulation packet lengths and time intervals specified in the protocol are shown in Table 1 and the results from the simulation shown in Figure 15.  The report concludes “The average throughput was measured for the planar arrays of size 8 x 8 and 4 x 4, with tschebyscheff (-26dB side lobe level) and uniform excitation distribution.”  The graph in Figure 15 shows us that the throughput for an 8 x 8 array was higher than a 4 x 4 array.  This confirms that a narrower beam width improves throughput due to reduced co-channel interference.  To gain a further understanding of how we can improve cellular performance we shall in part III perform simulations on an alternative model of a cellular system to estimate the SIR performance characteristics.

		Control Signals			Beam forming		Control Signal		Payload
Packet Type/Interval	DIFS	SIFS	RTS	CTS	TXRTN	RXTRN	ACK	ACK	Data
Length	0.014L	0.003L	0.007L	0.007L	variable	variable	0.007L	0.007L	L

Table 1 Packet lengths and time intervals for simulation (Bellofiore & al, Smart-Antenna System for Mobile Communication Networks Part 2: Beamforming and Network Throughput, 2002, p. 111)

Figure 15 Throughput as a function of load (Bellofiore & al, Smart-Antenna System for Mobile Communication Networks Part 2: Beamforming and Network Throughput, 2002, p. 112)

Part III – Simulation

Simulation Background

Simulation of a cellular system can provide an analytical way of determining system performance.  Performance can be measured through analysis of parameters including: Signal to Interference ration (SIR), outage probability, system capacity, and blocking probability to name a few.  Given the complex nature of a cellular system, designing a cell simulation package from first principles would be very challenging and with limited development time available this would not be feasible.

An important aspect of the simulation is to gather statistical data that can be used to compare system performance of various antenna configurations and cell cluster sizes.  A simulation directly related to the topic of this assignment (i.e. directional antennas with dynamic control) would be ideal; however, it has not been possible to locate the code for such a simulation.  Instead, for the purpose of this assignment, a simulation that could provide data for various cellular system configurations with the option of customising system parameters such as antenna configuration and cell cluster size would be sufficient.  Such a model could then be used to predict the performance of a cellular system operating with dynamically controlled directional antennas.

Much effort was spent researching suitable simulation codes designed to run on a variety of platforms.  The majority of simulations were not made available for student use as the code for such simulations are the work of recent and current research institutions.  Contact was made with Dr Grozev in Australia who is involved in the development of two simulations packages called Cellsim++ and CDMAnet which are referenced extensively in research journals based around antenna development.  Dr Grozev commented that he thought the named tools would be useful for the assignment simulations but unfortunately licensing of the packages lies with Australia’s biggest telecommunications company and is not readily available to students.  A second application called PAASoM: Phased Array Antenna Simulation Tool which appeared ideal for the assignment was found and contact made with the application developers.  A reply was received along with a price list starting at $8k per license!  This prompted the search for a more readily accessible simulation.

A book called ‘Principles of Communication Systems Simulation with Wireless Applications’ by William H. Tranter, K. Sam Shanmugan, Theodore S. Rappaport &

Kurt L. Kosbar was located and after thorough reading a suitable cell simulation example was found.  For convenience the aforementioned book will be referred to as simply ‘Simulation Source’ from here onwards.  A brief overview of the simulation model will be covered in the next section while a very thorough explanation of how the simulation model is formed along with the associated Matlab^® code can be located in chapter 17 of the Simulation Source text.

Simulation Algorithm

The selected simulation uses an iterative Monte Carlo approach to take ‘snapshots’ of a cellular system operating at random locations within a cell.  This technique is used to provide a statistical model to estimate the performance characteristics (SIR) of a simple cellular system operating under various configurations.  Several aspects of the simulation model exist including:

• Location of co-channel cells
• Determination of cell to be simulated
• Determination of distances between mobiles and base stations
• Determination of SIR statistics on both links using various algorithms

Figure 16 (from Simulation Source) shows a flow chart of the Monte Carlo approach to estimate the SIR and outage of a cellular system.

Figure 16: Simulation Flow Chart (Tranter W. H., 2009, p. 689)

As shown in the flow chart Figure 16, there are three main components to the simulation.

1. Cellular configuration scenario to be simulated is predetermined by user

User set parameters includes:

• Antenna configuration – omnidirectional, 60° sectorisation or 120° sectorisation
• Cell Cluster Size – 3, 4, or 7
• Path loss exponent – this is a measure of the reduction in power density of a signal as it propagates through free space, usually has a value between 2 and 5.
• Shadowing Standard Deviation – interference
• Shadowing Standard Deviation – desired signal
• This is the standard deviation of the log-normal fading in dB
• Base Station Antenna front to back ratio

2. An iterative ‘snapshot’ approach is used to collect statistical SIR data for both the forward (base station to mobile transmission) and reverse (mobile to base station transmission) links.

Calculation of the SIR for forward and reverse links takes shadowing effects and path loss into consideration.  Figure 17 is a diagram taken from the Simulation Source that depicts two of the 6 co-channel cells interfering with the centre cell.  Calculation of the mean and standard deviation for  and are computed using Wilkinson’s method.

Figure 17: Snapshot for 120° Sectoring (Tranter W. H., 2009, p. 696)

3. System performance calculated from collated data

The performance of the cellular system cannot be judged primarily on calculation of theoretical link quality alone.  The Simulation Source states “When we consider both the spatial distribution of mobiles and the effects of shadowing, SIR becomes a random variable… The performance of the cellular system must then be measured through the outage probability…”

Outage probability defines a channel that is unserviceable for a period of time.  The outage probability of the cellular configuration is computed by measuring the proportion of SIR results that are below a minimum  threshold.

Simulation Development

The Simulation Source contains the necessary Matlab^® code for the user to re-create the source code to perform individual simulations.  This code was copied direct from the text and formed into an ‘.m’ file for interaction with Matlab^® (complete program code has been included for reference in Appendix 3).  Unfortunately, it was discovered that the raw code would not execute due to multiple errors.  After careful studying of the Simulation Source and armed with an understanding of the operation of the code, the errors were corrected and the standard code executed correctly.  The Simulation Source provides three methods of deriving the SIR for the forward and reverse links: Method 1, Method 2A, and Method 2B.  Matlab^® code to calculate the outage probability was not provided in the Simulation Source so three custom functions were created to plot the results.

In order to develop and tailor the simulation towards the needs of the assignment, it was decided that with the use of National Instruments software package LabVIEW 2009 it would be possible to include additional functionality.  New simulation features include:

• Input of multiple scenarios
• Selection of SIR calculation method
• Calculation of outage probability for a given
• Calculation of traffic (Erlang) using Erlang-B formula
• Calculation of system capacity with frequency reuse factor taken into account
• Automated plotting of all scenario and capacity results

The LabVIEW application is based around the use of Active-X control to automate the underlying Matlab^® simulation code.  The next section provides a brief overview of the Simulator Graphical User Interface (GUI) while a more detailed breakdown of the program code can be located in Appendix 2.

LabVIEW Cell Sim GUI

A screen shot of the LabVIEW application GUI tab 1 has been included in Figure 18 for reference.  This tab enables the user to have overall control of the simulation process.  Multiple configurations can be entered and stored in the scenario configuration array.  The user has the option of selecting which method is used to calculate the forward and reverse link SIR by simply clicking the desired process.  The threshold value can be set which is used for analysis of outage probability in the Simulation Output tab 2.  Lastly, the value of Erlang traffic consumed per user can be defined for capacity calculation purposes.

Figure 18: LabVIEW Cell Sim GUI Tab 1

Cellular Simulation Scenarios

The primary focus of the simulations is to analyse various cellular simulation scenarios to compare the effects of antenna configurations and cell cluster size versus the overall cellular performance with shadowing and path loss taken into consideration.  Table 2 includes the configuration for each scenario simulated in addition to other parameters which remained constant throughout.

Scenario	Antenna Type	Cluster Size	Other Parameters
			Constant Throughout
0	Omnidirectional	3	Simulation loops = 1000 Number of channels = 395 Cell Radius (m) = 1000 Path Loss = 4 – Interference = 8dB – Desired Signal = 8dB Antenna FTB Ratio = 30dB Traffic per User = 0.02 Erlang SIR Threshold = 18dB
1	Omnidirectional	4
2	Omnidirectional	7
3	120° Sectoring	3
4	120° Sectoring	4
5	120° Sectoring	7
6	60° Sectoring	3
7	60° Sectoring	4
8	60° Sectoring	7

Table 2: Cellular Simulation Scenarios

After entering the simulation scenarios into the Cell Sim application the run button is pressed to begin operation of the program.  Results are published in the form of graphs and tables as shown in the following section of the report.

Simulation Results & Analysis

Prior to conducting these simulations, manual calculations were used to estimate the total co-channel interference of a system with and without cell sectoring.  The results of the calculations showed that by utilising a sectorised antenna within a cell it is possible to improve link quality significantly over a standard omnidirectional antenna.  As a consequence, cell sectoring reduces the cell capacity due to a reduction in trunking efficiency.  The manual calculations did not take the important effects of shadowing and path loss into consideration in addition to the frequency reuse factor.  For these reasons, the results of the simulations provide essential data to analyse.

Figure 19 is a graph of SIR vs. outage probability plotted by the LabVIEW Cell Sim application (a full scale version can be located in Appendix 1).  The results are generated for the cellular scenarios displayed in Table 2.  SIR calculation Method 2B was used to compute the SIR as this process does not require an assumption as to whether the co-channel interference is lognormal or normally distributed.  The graph shows a trace for each scenario simulated providing a convenient way of comparing system performance.

Analysis of the graph shows that ‘Scenario 0’ (omnidirectional antenna, cluster size of 3) has the highest outage probability throughout while ‘Scenario 8’ (60° sectorised antenna, cluster size of 7) has the lowest outage probability throughout.  An interesting comparison is that the second best performance is from ‘Scenario 5’ (120° sectorised antenna, cluster size of 7) which outperforms ‘Scenario 7’ (60° sectorised antenna, cluster size of 4).  This reinstates the fact that cellular performance is dependent on both the level of sectoring and the number of cells within the cluster (due to the frequency re-use distance).

Figure 19: Graph of SIR Vs Outage Probability

A threshold = 18dB is normally accepted as a good level of link performance for a cellular system (e.g. AMPS).  Figure 20 shows a bar chart of the outage probability for each scenario at 18dB.  The results show that as the cluster size increases the outage probability is reduced while operating with a sectorised antenna also improves link performance over an omnidirectional antenna.  At the SIR threshold, ‘Scenario 8’ (60° sectorised antenna, cluster size of 7) has the lowest outage probability of 10%, while ‘Scenario 0’ (omnidirectional antenna, cluster size of 3) has the highest with a value of 54%.

This outcome confirms the predictions from the manual calculations: an improvement in SIR as the cluster size increases and sectorised antennas are used.  The improvement in SIR is due to an increase in the frequency reuse distance (cluster size increase) and a reduction in co-channel interference from neighbouring cells (sectorisation).  Sectorisation reduces the number of co-channel interference sources by a factor of 3 for 120° sectoring and a factor of 6 for 60°.  Second tier interference is usually neglected if the path loss exponent is equal to 4 or greater.

Figure 20: Bar Chart of Outage Probability at SIR₀ = 18dB

Derivation of the cellular system capacity is very important as this dictates if the system will have enough resource to serve users of a particular geographical area.  Examples of calculating the capacity of a standard cell based system appear frequently in textbooks, the Simulation Source is no exception.  Page 702 of the text includes worked calculations for deriving system capacity.  For the initial calculations a similar approach has been utilised for the purpose of this assignment.

During the configuration of the simulation scenarios, the user is able to input the number of channels available and the value of traffic produced per user in Erlangs.  A default value of 395 channels and 0.02 Erlangs per user was used throughout.  The cellular capacity in terms of the number of serviceable users has been estimated by first calculating the maximum traffic handled per cell in Erlangs at a 2% blocking probability, (assuming blocked calls are cleared), then secondly calculating the number of users possible if on average each user produces 0.02 Erlang of traffic.  A number of observations can be made from the results shown in rows 1-3 of Table 3.

 

Table 3: Simulation Statistics

Comparisons of scenarios 0, 1 & 2 (all utilise omnidirectional antennas) show a reduction in capacity as the cluster size increases.  This is due to an increase in the frequency reuse distance, resulting in spectral inefficiencies.  On the contrary, an increase in cluster size improves the corresponding link quality, as illustrated by a reduction in outage probability of 25%.  When comparing the capacity of a cell utilising a sectorised antenna (scenario 3) with that of an omnidirectional antenna (scenario 0) one would expect to see an increase in capacity and SIR performance.  Secondary research conducted in previous sections of this assignment reveals that sectorisation is used as a method of increasing system capacity.  In fact as a result of the calculation applied, the capacity of the sectorised system is significantly less than that of the omnidirectional system.  The outcome was not as expected and therefore a more thorough understanding was required.

On paper (row 3 of Table 3) the capacity of the sectorised system is less than that of the omnidirectional system because a decrease in trunking efficiency is experienced (fewer channels per sector).  However, an important factor which has not been considered is the effects of an increase in SIR performance.  Due to this improvement it is possible to reduce the frequency reuse distance and thus increase the spectrum efficiency resulting in an increase in capacity.  The Simulation Source does not factor this into the calculations so further enhancements were made to aid the prediction of cellular capacity.

To account for the capacity improvement due to the increase in SIR and spectral efficiency, a term referred to as Frequency Reuse Factor (FRF) was used.  This factor is calculated as follows:

Equation 11 Frequency Reuse Factor

Where:

– Cluster size is either 3, 4 or 7

– Sectors per cell is 1 for omnidirectional, 3 for 120° sectoring and 6 for 60° sectoring

So as an example: Scenario 7 (cluster size of 4, 60° sectoring) has an FRF of:

Equation 12 Scenario 7 FRF

Therefore, to calculate the new system capacity this factor must be multiplied by the previous system capacity.  In order to provide more realistic capacity results it was decided that the outage probability of the system at the SIR threshold should also be factored into the calculation.  For this reason, it is assumed that the percentage of successful calls in a cell is the percentage of calls that do not get blocked (i.e. 1 – outage probability).  This is multiplied by the new cell capacity to generate a value which takes not only trunking efficiency into consideration but also outage probability and the frequency reuse factor.  Row 5 of Table 3 contains the FRF and outage probability compensated cell capacity results.  The entire computation process has been automated within the Cell Sim application.

Without consideration for the FRF and outage probability, the capacity of a cell reduces where sectoring is used and cluster size is increased.  However, in these examples the corresponding outage probability is much lower leading to an improvement in link quality and SIR.

Factoring FRF and outage probability into the calculation of cell capacity results in the highest sectorised system with the fewest number of cells (per cluster) having the highest capacity (scenario 6).  A comparison of the two capacity calculation methods is presented in Cell Sim GUI tab 2B as shown in Figure 21.

Figure 21: Cell Capacity Comparison

Part IV – Conclusions

As mobile communications continue to grow and the relative affordability of mobile phones decreases, we can expect a rise in the number of subscribers using this technology.  The demand for data transfer has risen in recent years due to technological advances in mobile handsets allowing multimedia applications, such as the internet becoming commonplace.  Efficient utilisation of the limited spectral resource available will be required in order to meet this growing demand.  Therefore, we can expect network operators to evaluate all possible techniques to increase capacity.  Based on the reports evaluated it would be reasonable to say that Smart Antennas using SDMA will make efficient use of the available spectrum by increasing capacity and reduced co-channel interference.  Smart antennas are very complex and require state-of-the-art processing to control the beams from each antenna element.  In the past this has made them more expensive than conventional antennas.  Whilst this may still be the case; commercial FGPAs are now becoming relatively inexpensive and should no longer be a significant issue.

A new technique of calculating cell capacity with FRF and outage probability taken into consideration has been formulated.  While these calculations only provide an estimation of cell capacity, they do support the evidence of capacity improvement due to sectoring covered in the first half of the assignment.  The calculated improvements in performance of sectorised systems do result in extra cost as multiple antennas are required, in addition to the complexity of handling an increase in call handovers.  The challenge faced is balancing the number of users along with the link quality (QoS) to achieve the most cost effective and efficient cellular network.

In summary, with the knowledge gained from an in-depth study of current and past research papers it is apparent that the suggested benefits of obtaining high link quality while maintaining cellular capacity will be possible in the near future through the use of beam forming using Smart Antennas.  This technology can greatly reduce the co-channel interference while allowing greater user capacity through intense frequency reuse.

Reference List

Akaiwa, Y. (1997). Introduction to Digital Mobile Communication (First ed.). John Wiley & Sons Inc.

Bates, R. J. (2002). Broadband Telecommunications Handbook (Second ed.). McGraw-Hill Telecom.

Bellofiore, S., & al, e. (2002). Smart-Antenna System for Mobile Communication Networks Part 2: Beamforming and Network Throughput. Department of Electrical Engineering.

Bellofiore, S., Balanis, C. A., & Foufz, J. S. (2002). Smart-Antenna Systems for Mobile Communication Networks Part I: Overview and Antenna Design. Department of Electrical Engineering.

Brand, A., & Aghvami, H. (2002). Multiple Access Protocols For Mobile Communications Gprs, Umts And Beyond. John Wiley & Sons.

Cooper, M. (1996). Intelligent Antennas: Spatial Division Multiple Access.

Dunlop, J., & Smith, D. G. (1994). Telecommunicatiotns Engineering (Third edition ed.). Chapman & Hell.

1. E. Fakoukakis, S. G. (2005). Development of an Adaptive and a Switched Beam Smart Antenna System for Wireless Communications. Electromagnetics Research.

Godara, L. C. (2002). Handbook Of Antennas In Wireless Communications (First ed.). CRC Press LLC.

Hourani, H. (2004/2005). An Overview of Adaptive Antenna Systems.

Hsieh, S.-C. (n.d.). 安天娜 Antenna. Retrieved 27th March, 2010, from 安天娜 Antenna: http://www.wireless-lab543.ee.ccu.edu.tw/Antenna.html

Lawrence, H. J. (1998). IS-136 TDMA Technology, Economics and Services (First ed.). Artech House Publishers.

Lee, W. C. (1997). Mobile Communicaton Engineering (Second ed.). McGraw-Hill Telecommunications.

Parini, P. C. (2006). Semi-Smart Antenna Technology Project Ofcom report. Department of Electronic Engineering.

Perini, A. U. (2000). An Overview of Smart Antenna Technology for Wireless Communication1. IEEE Xplore.

Shamim, A. (n.d.). Adaptive Antenna System. Retrieved 13th April, 2010, from 3gtech.info: http://www.3gtech.info/adaptive-antenna-system.html

Stallings, W. (2009). Data and Computer communication (Eighth ed.). Pearson International edition.

Stallings, W. (2005). Wireless Communication and Networks (Second ed.). Pearson Hall.

Tachikawa, K. (2002). W-CDMA Mobile Communications System (First ed.). John Wiley & Sons Ltd.

Tranter, W. H. (2009). Principles of Communication Systems Simulation with Wireless Applications. Prentice Hall.

Tranter, W. H., Shanmugan, K. S., Rappaport, T. S., & Kosbar, K. L. (2009). Principles of Communication Systems Simulation with Wireless Applications. Prentice Hall.

Tutorials, W. P. (n.d.). Smart Antenna Systems. Retrieved 26th March, 2010, from Internationnal Engineering Consortium: http://www.iec.org/online/tutorials/

Unknown. (n.d.). Code division multiple access. Retrieved 27th March, 2010, from Wikipedia: http://en.wikipedia.org/wiki/Code_division_multiple_access

W.S. AU, R. M. (2001). Comparison between the Spectral Efficiency of SDMA Systems and Sectorized Systems. Kluwer Academic Publishers.

Webb, W. (1999). The Complete wireless communications professional : a guide for engineers and managers (First ed.). Artech House Inc.

Wheat, J. (2001). Designing a Wireless Network (First ed.). Syngress Publishing.

Wiki. (n.d.). Cellular network. Retrieved 12th March, 2010, from Wikipedia: http://en.wikipedia.org/wiki/Cellular_network

Wikipedia. (n.d.). Wiki Cellular network. Retrieved 5th March, 2010, from Wikipedia: http://en.wikipedia.org/wiki/Cellular_network

Yea, J.-H. (2001). Smart antennas for multiple sectorization in CDMA cell sites.

Zhuang, W. (n.d.). Retrieved 6th April, 2010, from Fundamentals Of Cellular Communications: http://bbcr.uwaterloo.ca/~wzhuang/414/

Note: Special thanks for K.Collard at UWE for his joint contribution to this research