Abstract: Capacity has always been an issue to any generation of mobile communication technologies as it is always inversely proportionate to QoS (quality of service) in nature. In order to increase the mobile communication system capacity without jeopardizing the system QoS, smart antenna is usually being suggested. There are two major types of smart antenna systems; switch beam antenna and adaptive beam antenna. Adaptive antenna which has dynamic beam is better in improving the system capacity but requires more advance beam forming algorithm and requires intensive processing power. In this paper, a novel adaptive downlink beam forming technique for WCDMA system namely Minimum Downlink Power Consumption (MDPC) is introduced. This MDPC beamforming technique is assisted by Genetic Algorithm (GA) to minimize the downlink power usage at Node B. A dynamic radio network simulator was developed in Visual C++® to study the power consumption at Node B and estimate the downlink capacity improvement by implementing this novel algorithm. Simulation was done based on single micro cell environment with considering interference from the first tier. User mobility is taken into account to provide a combined evaluation of Radio Resource Management (RRM). Capacity system under various simulation scenarios was expressed in downlink outage in this study.
INTRODUCTION
Contradict to FDMA and TDMA system, the main limiting factor for capacity expansion in WCDMA system is the interference (De Sousa et al., 2003). Each user acts as interference source to other users in the system. Thus, in order to improve the capacity offered by WCDMA systems, great attention has been drawn in the exploitation of spatial domain using smart antennas (Marikar, 2002; Martinez et al., 2001) to segregate the users into separate domains.
Many studies have shown that smart antennas are capable to provide enhanced coverage through Multiple Access Interference (MAI) cancellation, range extension, increase system capacity and improve link quality (De Sousa et al., 2003; Liberti and Rappaport, 1999). However, most of the studies were done on the uplink. This is due to most of the researchers believed that WCDMA system is reverse link limited (Holma et al., 2001), (Hernandez et al., 2000). Nevertheless, recent researches have shown that the forward link can be the limiting link under certain circumstances (Wan et al., 2000; Eisenblatter et al., 2001).
The capacity of a mobile communication system can be defined in several ways. Typically, maximal number of simultaneous users in the system is used and this method is widely used for designing a call-admission control for the given set of capacity lines. Another method is to determine the peak load that can be supported by the system while maintaining desired service quality. The latter matter can be used to find the Erlang capacity that is used in cell planning and in evaluating capacity expense (Wan and Jin, 2001). In this paper, downlink outage in the aspect of blocking and dropping probability is used to evaluate the Erlang capacity as in Wan et al. (2000).
In this paper, study has been carried up to verify the effect of uplink and downlink power limit towards the system capacity prior to system level simulation. Parameters used are as shown in Table 1. Study result in Fig. 1 shows that downlink power is the limiting power for larger WCDMA cell. Thus, downlink capacity improvement is targeted in this research.
| Table 1: | System parameters Parameters |
| Fig. 1: | WCDMA downlink and uplink capacity in different cell size |
It is expected that by reducing and carefully control the downlink power at Node B, less interference will occur in WCDMA system and hence more users can be accommodated by the system, which subsequently increase the system capacity.
SIMULATOR MODELING
In order to simulate the downlink capacity of a WCDMA system, a dynamic system level simulation tool named as WSAGASIM(WCDMA Smart Antenna Genetic Algorithm Simulator) was developed. The simulator consists of five major parts; initialization, call admission control, beamforming and power control, mobile stations processes and simulation output.
Initialization: Initialization phase is the first part of the simulator, where topography of the simulation such as cell sites, Node B is generated. Besides, the UE traffic parameters such as mean arrival time and mean call duration are generated based on Poisson distribution model. Meanwhile, UE mobility parameters are generated based on ITU-Vehicle A model.
Call admission control: Call admission algorithm measures the received Energy per bit per noise density,
from the UE who requesting admission into the wireless system. Depending on
the position and location of the UE, the algorithm decides which beam is best
selected to serve the mobile station and determine whether the UE can be admitted
to the system without jeopardizing the performance of other UEs currently in
the system. The downlink
| (1) |
where ρt,n is the traffic channel power of UE #n, Gt is Node B antenna gain, Gr,n is UE #n antenna gain, Al,n is lognormal shadowing experienced by UE #n, Lp,n is propagation loss from Node B to UE #n, Lt and Lr is cable loss at Node B transmitter and UE receiver respectively, brt is bit rate of downlink traffic channel, No is thermal noise density at 290 K and It,n is the total interference density experienced by UE #n.
Total interference density experienced by UE #n, It,n is obtained by considering both interference within the home cell and also interference from first tier cells.
| (2) |
where Bw is bandwidth in Hz, fr is frequency reuse factor which representing interference caused by first tier neighbour cells and PT is the total ERP power from Node B.
Path loss is calculated based on channel characteristics and in this simulation, macrocell propagation model (Eq. 3), which is applicable for both urban and suburban areas, with assumption of nearly uniform height of buildings is used (3G TR 25.942.2.1.3/2000-03).
| (3) |
where Lp is the path loss in dB, r is the distance between base station and mobile station in kilometers, f is the carrier frequency in MHz and h is the base station antenna height, in meters, measure from the average rooftop level.
UE mobility: In this dynamic simulator, mobile stations are moving at different speeds. UE speeds with 3, 20 and 120 km h-1 were used in the simulation. Thus, the mobile stations conditions need to be updated from time to time at certain time steps. New locations of mobile stations are obtained by predefined mobility model during initialization phase.
As mobile stations moving around, mobile stations may move from one antenna beam coverage to another and thus, intra-handover occurs between antenna beams. Consequently, beam ownership of every mobile station need to be updated from time to time. Besides, new states of mobile stations will be checked in every time step to determine each mobile station in progress, whether in active or idle mode.
Beamforming and power control: Three types of antennas were studied in this paper. They are 60° sectorized antenna, 9-element switch beam smart antenna, 9-element adaptive beam smart antenna. For antenna elements, it is assumed that identical antenna elements arranged with equal spacing in circular are used. 9 elements antenna array was used to simulate for smart antenna system (fixed 40° degree beamwidth for switch beam and variable 6°~120° for adaptive beam). Two types of adaptive beam forming techniques were used in simulation. They are Maximum Angle of Arrival Separation (MAOAS) and Minimum Downlink Power- Controlled (MDPC), which is assisted by Genetic Algorithm (GA) to optimize the total power consumption at Node B to the acceptable lowest level. Conventional MAOAS beam forming technique is used as the benchmark to be compared with new proposed MDPC beam forming technique.
Power transmitted by any particular antenna array at Node B depends on UE location
and interference faced by UE. Closed loop power control is used to continuously
monitor
Simulation outputs: Output of the simulator is system outage that describes the capacity of the system. According to Anand et al. (2002) and Annalisa et al. (2004), the outage probability is expressed as below;
| (4) |
Where ρo is the Eb/No threshold value.
According to Mahamod et al., 1995, Pout of a radio cellular system with N number of interfering channels can be expressed as in Eq. 5.
| (5) |
| where; | |
| p(.) | is the probability density function (pdf) of received signal strength. |
| Pw | is the wanted signal |
| Pin | is the interfering signal (n = 1 to N) |
| RTH | is interference protection ratio |
| PTH | is threshold power required |
In approximation, system outage can be obtained by summing the blocking and
dropping probabilities. Blocking of a new UE will occur if the
| (6) |
| (7) |
| (8) |
Besides, average of total power consumption all antenna arrays at Node B is obtained as below.
| (9) |
Where An is Antenna Array n at Node B.
GENETIC ALGORITHM OPTIMIZATION METHOD
The first operation of MDPC adaptive beam forming algorithm is activating all antenna arrays to cover whole micro cell in order to get the quantity of instantaneous active UE at a particular moment. Communication will be carried out through pilot channel in WCDMA system.
| Fig. 2a: | MDPC adaptive beam forming main program flowchart |
A simplified MDPC adaptive beam forming algorithm is shown in Fig. 2.
Fitness function developed to evaluate the fitness in the aspect of power usage at Node B is shown in Eq. 10.
| (10) |
where,
where ωm,max dan ωm,min are maximum coverage angle and minimum coverage angle for smart antenna element #m and ρ ialah minimum separation angle between antenna beam.
| Fig. 2b: | MDPC adaptive beam forming beam fitness evaluation flowchart |
Where δ = 360/M and M is number of smart antenna elements and where δ = 360/M and M is number of smart antenna elements and;
| (11) |
where pm is poynting power transmitted, ε is the weight factor for pm (ε = 0.001 in this study), wm is solid angle of antenna beam and assume that the cable loss at antenna element is 2.5 dB.
SIMULATION RESULT
The main simulation parameters are as shown in Table 2 and the simulation results are shown in Fig. 3-7. Figure 3 to 6 shows the outage probability in the system implementing different types of antenna. Simulation was carried out with different traffic load condition to obtain the capacity performance under various traffic load condition. The composition of blocking probability and dropping probability are shown in the figure.
From Fig. 3 to 6, it is shown that adaptive antenna gives better capacity to the system as outage probability is lower with adaptive antenna.
| Fig. 3: | Outage probability with sectorized antenna |
| Fig. 4: | Outage probability with switch beam antenna |
Relatively, blocking probability is higher as compared to dropping probability in both sectorized antenna and switch beam antenna system. However, dropping probability is relatively higher than blocking probability in adaptive antenna system. It is due to power usage at Node B that utilizing sectorized antenna and switch beam antenna are not intelligently optimized.
| Table 2: | Simulation parameters |
| Fig. 5: | Outage probability with MAOAS adaptive antenna |
| Fig. 6: | Outage probability with GA assisted MDPC adaptive antenna |
| Fig. 7: | Power consumption for different type of antenna systems |
| Fig. 8: | Comparison of outage probability with different types of antenna systems |
As a result, power consumption in the system is higher and thus, higher possibility
of a new UE to be blocked due to insufficient power to improve the new UE's
Eb
Figure 7 shows the downlink power consumption for different types of antenna system. Simulation was carried out to obtain the power consumption at Node B under different traffic load conditions. Meanwhile Fig. 8 shows the comparison of system capacity by utilizing different antenna system under different traffic load condition.
| Fig. 9: | Average number of UE under different outage probability with different types of antenna system |
From Fig. 7, it is shown that adaptive antenna systems use
lower power as compared to sectorized and switch beam antenna. From Fig.
8, it can be seen that the outage probability by using sectorized antenna
is higher than smart antenna systems. As the system capacity is inversely proportional
to outage probability, thus the system capacity with sectorized antenna is lower
compared to any smart antenna systems. It is due to the higher power usage at
Node B, the least number of simultaneous active UEs can be supported by the
system. Power consumption of sectorized antenna and switch beam antenna that
are not intelligently optimized causing higher possibility of a new UE to be
blocked due to insufficient power to improve the new UE's
Figure 9 shows the average number of active UE in the system under different outage probability. It can be seen that at the same outage probability, different antenna system will support different number of instantaneous active UE. Again, it can be observed that adaptive antenna systems can support more instantaneous active UEs as compared to other antenna systems. GA assisted MDPC adaptive antenna gives the best performance in improving the system capacity as compared to three other antenna systems.
Figure 10 shows that power consumption at Node B is increasing with cell size.
| Fig. 10: | Outage probability by using different types of antenna systems at different cell size |
| Fig. 11: | Outage probability by using different types of antenna systems at different cell size |
When the cell size grows larger, more UE will be blocked due to greater propagation loss occurs between the Node B and UE, causing insufficient power at Node B to maintain every UE signal above the threshold level. GA assisted MDPC shows the least power consumption at different cell size.
Figure 11 shows the outage probability for different antenna systems at different cell radius. Simulation results show that the outage probability increases when cell radius increases. As expected, outage probability by using smart antennas is lower as compared to sectorized antenna. At the cell radius with 1 km, the outage probability is 0.03 for sectorized antenna and 0.00 for all of the smart antenna systems. At the cell radius with 2 km, the outage probability of sectorized antenna increases to 0.51. Meanwhile, the outage probability of switch beam antenna, MAOAS adaptive antenna and GA assisted MDPC adaptive antenna at cell radius with 2 km are 0.37, 0.36 and 0.31, respectively. Again, MDPC adaptive antenna provides the highest capacity to the system at any cell size with radius between 1 to 2 km.
CONCLUSIONS
The simulation results have shown that different types of antenna systems consume different level of power at Node B and at all conditions, adaptive antennas use lower power as compared to sectorized antenna and switch beam antenna. Besides, simulation results show that all three types of smart antenna systems were capable to provide downlink capacity improvement to WCDMA system and the new GA assisted MDPC beamforming technique based on the downlink power consumption is superior in improving the capacity of WCDMA system in all simulated conditions. Furthermore, GA assisted MDPC adaptive antenna system has additional advantage such as does not require a reference signal level like MMSE adaptive beam forming technique.