Multihop relaying for broadband wireless access systems at 800 and 3500 MHz in rural areas

Authors


Abstract

This paper addresses the coverage enhancement for broadband wireless access (BWA) at 800 and 3500 MHz in a rural scenario in north Germany using multi-hop relay concepts. The investigation is divided into two phases: In phase I, the coverage of a single BWA system in a rural area is predicted and verified with measurement data. In phase II, the coverage from the BWA system is analyzed and enhanced through the deployment of relay stations. The number of relay stations required for each carrier frequency is determined, and the positions of the relay stations are identified via three different relay placement algorithms, namely path-loss-based, distance-based, and hybrid algorithms. At 800 MHz, the path-loss-based algorithm requires seven relay stations to achieve an overall coverage of 95%. The hybrid and distance-based algorithms require 11 and 14 relay stations, respectively, to achieve the same coverage. At 3500 MHz, path-loss based and hybrid algorithms require 19 and 16 relay stations, respectively, to achieve the same coverage. The distance-based requires 20 relay stations to achieve an overall coverage of 82%. Further increase of relay stations has led to higher interference. Lastly, the transmit power of the relay station is optimized via an intelligent power allocation scheme. The results show that 20% of the total transmit powers from 14 relay stations can be saved at 800 MHz whereas 18% of the total transmit powers from the 20 relay stations can be saved at 3500 MHz.

1 Introduction

Fixed broadband wireless access systems have been regarded as a promising solution to extend broadband services to suburban/rural areas where access to fiber optics is not as easy as in urban cities. The system aims to replace cable infrastructures with macrocell-size radio links so that broadband access can be extended to users in such areas. Wave propagation in these remote areas is mainly challenged by dense vegetation with irregular terrain and clutter distribution. Random fading in such areas often results in areas of poor reception or even dead spots within the coverage regions. One possible solution may be to deploy the base stations in a denser manner. However, the enormous deploying and maintenance costs together with the issue of interference have refrained it from becoming a solution in such an environment. Relay stations, which have no direct backhaul connection to the network, seem to be a cost-effective solution to this. The adoption of the relay concept in dense urban scenarios has been studied in Tameh et al. [2003], Esseling et al. [2004], Schultz et al. [2003], Pabst et al. [2004], and Doppler et al. [2008], where relays can be used to improve coverage on the street among high rising concrete jungle, for example the Manhattan-like scenario as described in Esseling et al. [2004], Schultz et al. [2003], Pabst et al. [2004], and Doppler et al. [2008], in order to boost the supporting traffic with minimal cost. In rural areas, residential buildings may be vastly distributed within a dense vegetation area with irregular terrain. Therefore, it is important to identify where the residential buildings (traffic demand) are located, so that relays can serve as repeaters to be installed at the top of residential buildings in order to take advantage of building heights for better link conditions. This paper is to investigate the coverage enhancement for fixed broadband wireless access (BWA) systems deployed in rural areas using relay stations (RS). The investigation is based on a realistic scenario in Hetzwege, a rural township in the county of Rotenburg (Wümme), Germany. The study adopts the relay placement algorithms proposed in the literature [Streng et al., 2003, 2002; Kwak et al., 2008] and demonstrates how these algorithms can be applied to our scenario in Germany. The pros and cons of these algorithms are discussed and eventually a new algorithm which combines the path loss-based and distance-based relay placement algorithms is proposed. The performance of the newly proposed algorithm in our rural area scenario is then compared with the algorithms from the literature, and the advantages are elaborated in the subsequent sections. The study is divided into two phases. In phase I, the coverage of a single BWA system in rural areas located in the northwest of Hetzwege was determined. The coverage prediction takes into account the clutter distribution, diffraction due to irregular terrain, residential buildings, and trees which are typically identified in the region. In order to verify the prediction accuracy and effectiveness, a measurement campaign was conducted on site using the deployed Worldwide Interoperability for Microwave Access (WiMAX) system in Hetzwege [Wimax, 2009].

In phase II, the coverage map is to be extended with the aid of relay stations. Given a predefined throughput to be achieved in the region (section 'Identification of Service Area'), the coverage map derived from phase I will be analyzed so that areas with poor reception can be identified. The prediction identifies where the demand for traffics is and then returns the optimized number of relays as well as the optimized position where relays would be installed in order to achieve a full coverage. The relay network is assumed to be operated with homogeneous relaying where the same radio access technology for the relay links (Base Station BS to RS) and access links (BS, RS to Mobile Station MS) has been used. The propagation loss within the relay links as well as within the access links takes into account the propagation effects and interference along the propagation path. Furthermore, the relay stations are assumed to be operated with decode and forward scheme (DF) where the received signal is first decoded, then re-encoded before retransmission. This is to avoid the coupling of noise that was picking up along multiple relay links.

The paper is organized as follows. Section 'Scenario in Rural Areas and Measurement Campaign' describes the scenario in rural areas where the relay system will be implemented. A measurement campaign conducted simultaneously at 800 MHz and 3500 MHz is described. Section 'Propagation Loss Modeling' discusses the modeling of the propagation path loss in the area, so that the coverage map of a single base station can be derived. Section 'Multihop Relaying System' describes the implementation of a multihop relay system which includes the identification of the areas with poor reception and different relay placement algorithms which are used to determine the locations where the relay stations should be placed. Discussions are presented along with the numerical results in this section. Section 'Conclusion' concludes the paper.

2 Scenario in Rural Areas and Measurement Campaign

2.1 Environment

Under the framework of the research project “WiMAX Field Trial in Lower Saxony” [Wimax, 2009], an Institute of Electrical and Electronics Engineers (IEEE) 802.16e Orthogonal Frequency-Division Multiple Access (OFDMA)-based WiMAX system has been deployed in the northwest of Hetzwege, in the county of Rotenburg (Wümme) in Germany. The project was funded by Stiftung Zukunfts- und Innovationsfonds [Stiftung, 2007] and aimed to deliver broadband services to rural areas where fiber optics are far to be reached. The area is typically characterized by irregular terrain, with residential buildings vastly distributed among agriculture farmlands and forested areas. Figure 1 shows a geographical representation of Hetzwege. In order to take into account the wave propagation in this rural environment, a self-tailored 3-D digital elevation model (DEM) of Hetzwege with 3 m × 3 m resolution inclusive clutter information has been constructed [Chee and Kürner, 2009].

Figure 1.

Geographical representation of Hetzwege (Scale: 1 km × 1.8 km).

2.2 Measurement Campaign

In order to quantify the wave propagation in this area, a measurement campaign was carried out at Hetzwege using the 3500 MHz base station that delivers internet access via WiMAX to the residents in the region. The four-column array antenna from Andrew (APW435-12014-0N) was erected on a 25 m tower with 115° azimuth and 2° mechanical tilting. The sector antenna has a gain of 23 dBi at the boresight with a horizontal beam width of 25° and vertical beam width of 5°. The transmit signal is vertically polarized with a power of 58 dBm equivalent isotropically radiated power.

Parallel to the 3500 MHz system, a second base station operated at 825 MHz was established next to the existing base station operated at 3500 MHz (Figure 2a). The omnidirectional antenna (XPO2V-0.8-6.0-GF/1441 from Cobham) was mounted to a lifting platform of 25 m height. The 825 MHz system consisted of a signal generator (Agilent E4428C) connected to a laptop equipped with the software package Signal Studio for IEEE 802.16e OFDMA WiMAX. The output power from the signal generator was −10 dBm. It was then boosted by 44.5 dB with an external power amplifier before being connected to the antenna with a gain of 2 dBi. A predefined frame sequence was continuously transmitted from the 825 MHz system. At the receiving end, the TSMW WiMAX scanner from Rohde & Schwarz was used to detect the transmitted frame sequence. The omnidirectional receiving antenna (XPO2V-0.8-6.0-GF/1441 from Cobham) was mounted to the roof of a moving vehicle which is 2 m above ground as shown in Figure 2b. The vehicle was driven through all possible tracks in the region with a constant speed of 40 km/h. An external GPS device (Navilock NL-602U Universal Serial Bus (USB)) with an accuracy of 2.5 m circular error probable was used to associate the detected signal strength with geographical coordinates. All measurement data were streamed to a laptop operating with the dedicated software ROMES version 4.40 for post processing in order to allow further analysis.

Figure 2.

(a) Base stations at 825 MHz and 3.5 GHz. (b) Receiving antenna with GPS mounted to the roof of a vehicle.

3 Propagation Loss Modeling

Using the self-tailored 3-D DEM of Hetzwege, the terrain profile between the base station and each receiving position was derived with the Bresenham algorithm [Deliste et al., 1985]. The propagation loss was derived using the Extended Hata model from Spectrum Engineering Advanced Monte Carlo Analysis Tool with a validity in the frequency range 150 MHz ≤ f ≤ 3 GHz [Seamcat, 2010]. In order to take the terrain irregularities and effect of clutter into account, the effective antenna height for each transmission path was derived from the terrain profile with the method described in [Chee and Kürner, 2010]. In the non-line-of-sight scenario (NLOS), the obstructions were first identified from the terrain profile with the method described by Durkin [Edwards and Durkin, 1969]. In the case when more than three obstructions were identified, they were combined depending on the distance separation criteria as described in Chee and Kürner [2010] and the diffraction due to these obstructions was modeled using the Deygout method [Deygout, 1996]. The comparison between the predicted path loss and the path loss derived from measurements at 800 MHz and 3500 MHz. are tabulated in Table 1. At 800 MHz, the predictions achieve a mean error of 3.57 dB and a standard deviation at 5.24 dB for LOS paths. For NLOS paths, the prediction error has a mean error of 0.70 dB and a standard deviation of 7.58 dB. At 3500 MHz, the predictions achieve a mean error of 2.07 dB and a standard deviation of 11.81 dB for LOS paths. However, for NLOS path, the prediction error has a mean error of 0.97 dB and a standard deviation of 10.22 dB. Obviously, the standard deviations of the prediction errors are much higher at higher frequencies due to the scattering effects from trees in the areas. This indicates that for a mobile radio system deployed at 3500 MHz in such an environment, a higher link budget should be taken into account in order to ensure the reliability of the radio systems.

Table 1. Comparison of Path Loss Between Predictions and Measurements
Frequency (MHz)Meana (dB)Stda (dB)
LOSNLOSLOSNLOS
  1. aPrediction–measurement.
800 MHz3.570.705.247.58
3500 MHz2.070.9711.8110.22

4 Multihop Relaying System

4.1 Adequate Coverage

In order to assess whether an area is adequately served by the network, the link quality can be evaluated by determining the carrier-to-noise and interference ratio (CINR). In an interference-free scenario, for example, when there is a single network cell in the rural area, the carrier-to-noise ratio is defined as

display math(1)

where fB is the data rate of the channel and B is the channel bandwidth. Eb is the average energy per bit and No is the noise spectral density No = N/B. N is the average internal noise power of the receiver, which includes the inherited thermal noise and the noise figure.

In an interference-rich scenario, signals from neighboring cells significantly degrade the useful signal from the operating transmitter; hence, interference is a serious issue that should be taken into account. In a realistic communication network, network operators are obligated to operate the system at dedicated frequency bands. This leads to the fact that a Multihop Relaying System (MRS) is likely to be implemented at the same frequency bands, leading to a limited system performance due to cochannel interference. In the MRS, multiple versions of the signal from different relay stations are received at the receiver. All these signals can be combined intelligently using maximum ratio combining to yield a stronger received signal. However, in this paper, a worst case scenario is considered where the receiver takes only the strongest received signal as a useful source; the received signals from all other stations are taken as interference

display math(2)

where C is the strongest received signal and IR is the interference from the relay station R. Considering an area of 500 × 500 m, if 95% of the area has a CINR higher than the predefined threshold CINRTH, the area is deemed to have adequate coverage.

4.2 Identification of Service Area

It is assumed that the region is guaranteed by the network operator with a high-speed downlink throughput of 12.96 Mbps. Following the characteristics of a typical WiMAX receiver from Runcom, assuming a noise figure of 7 dB at the receiver, using a fix code rate of 3/4 for 16-QAM modulation and bit error rate of 10−5, the CINR needs to be at least 27 dB [Hermann, 2009]. Therefore, areas with CINR < 27 dB are first identified from the coverage map derived from the path loss prediction approach described in section 'Propagation Loss Modeling'. Figures 3a and 3b distinguish the served and underserved area by a single base station in Hetzwege at 800 MHz and 3500 MHz, respectively. Unlike users in urban areas, the residential buildings in rural areas are vastly distributed among agriculture farmlands and forested areas. Covering areas without demand for traffic will be a waste of resources. Therefore, it is important to focus only on those areas where potential users could be located. Owing to this, the definition of well-served and underserved areas can be revised to focus only on those areas where residential buildings are located. The identification of poor reception areas within user communities was achieved with the help of a searching operator which is known as dilation [Gonzalez et al., 2005]. For this, the searching operator first determines if residential buildings are located on each pixel (3 m × 3 m). Those pixels that consist of residential buildings will then be fused together with the eight neighboring pixels to form a housing community. The location of the housing communities correspond to the areas that are to be served by the MRS. Figures 4a and 4b depict the underserved communities at 800 MHz and 3500 MHz, respectively, which should be covered by the proposed relay systems.

Figure 3.

(a) Distinction of well-served (red) and underserved (blue) area by a single base station at 800 MHz. (b) Distinction of well-served (red) and underserved (blue) area by a single base station at 3500 MHz.

Figure 4.

(a) Distinction of area in the communities to be served by relays (white) from the rest of the area (black) at 800 MHz. (b) Distinction of area in the communities to be served by relays (white) from the rest of the area (black) at 3500 MHz.

4.3 Relay Placement Algorithms

Having known that there are significant areas in the communities that are still underserved compared to the predefined target, the cost-effective solution to get those poor reception areas filled is to adopt the concept of relay stations. The relay stations are assumed to serve as repeaters using homogeneous links with the DF scheme. Also, relay stations are assumed to be installed at the rooftops of residential buildings in order to take benefit from the building heights for a better link clearance to the source. This section discusses three relay placement algorithms which are used to optimize the locations of relay stations in this rural area, namely, path loss-based, distance-based, and hybrid relay placement algorithm. The hybrid relay placement algorithm combines the first two approaches in order to benefit from their advantages.

4.4 Path Loss-Based Relay Placement Algorithm

The path loss-based relay placement algorithm represents the most fundamental approach to decide on the position of a relay node where the quality of the access link (source/relay to mobile or vice versa) and the relay link (source to relay or vice versa) can be directly justified from the propagation loss. Following the approach reported in Streng et al. [2003] and Streng et al. [2002], three approaches have been defined

display math(3)
display math(4)
display math(5)

where PLn1 and PLn2 represent the path losses associated with the first and the second hop, respectively, along the nth route, n Є N as shown in Figure 5. The minimum total path loss chooses the relay which results in the lowest total path loss summed up from all links between source and destination. In Least Maximum Path loss (LMP), the relay node which gives the lowest bottleneck in terms of path loss will be chosen. The decision is based on the fact that due to random fading from the environment, the failure rate at either access or relay link with path loss close to the bottleneck will be higher even though the total path loss is low as compared to another route. In order to guarantee the overall functionality of the relay system, a route with the lowest bottleneck will be preferred. In the minimum relaying hop path loss, only the path loss at the link between relay and destination will be considered. For MRS with DF scheme, the path loss at the link between the source and relay node is not critical, as long as adequate coverage can be obtained for correctly decoding at the relay station before signals are regenerated and forwarded to the destination. In this paper the LMP scheme is adopted. Therefore, even though there could be random fading from vegetation which may not be predicted from the propagation loss predictor as discussed in section 'Propagation Loss Modeling', the functionality of the relay system is unlikely to be affected since only the relay stations at positions that result in the lowest bottleneck are chosen.

Figure 5.

Path loss-based relay placement algorithm.

Figures 6a–6d show the coverage map of an MRS at 800 MHz and 3500 MHz and their corresponding aggregated coverage versus the relay number at different transmit powers. For a transmit power of 0.1 W, nine relay nodes are required to achieve coverage of 95% at 800 MHz while 20 relay nodes are only able to provide a coverage of 80% at 3500 MHz. Further increases of the relay number at 3500 MHz for the same transmit power lead only to serious interference, and hence, the coverage of 95% can never be achieved. When the transmit power increases to 0.2 W, eight relay nodes are required to achieve a coverage of 95% at 800 MHz and 22 relay nodes are required for a coverage of 95% at 3500 MHz. When the transmit power increases to 0.3 W, only 7 relay nodes are required at 800 MHz, while 19 relay nodes are required to achieve 95% coverage in the communities. The performance of the system strongly depends on the transmit power especially at 3500 MHz; this shows the necessity of intelligent power allocation schemes for efficiency improvement.

Figure 6.

(a) Coverage map of LMP-based MRS at 800 MHz. (b) The aggregated coverage versus relay number for LMP-based MRS at 800 MHz at three different transmit powers. (c) Coverage map of LMP-based MRS at 3500 MHz. (d) The aggregated coverage versus relay number for LMP-based MRS at 3500 MHz at three different transmit powers.

4.5 Distance-Based Relay Placement Algorithm

The distance-based relay placement algorithm (DIS) serves as an alternative option to decide where relay nodes could be located. It has been reported in Streng et al. [2003] that path loss-based relay placement algorithms incur much higher signaling overheads, mainly due to the path loss estimation technique used. Unlike an infrastructure network, where the signaling can be fixed after the network deployment, the signaling for ad hoc relay stations may not be predefined. Therefore, signaling can be included in the overheads to specify where the packets should be forwarded. The LMP scheme requires the path loss at relay and access links at all potential relay positions to be determined before a decision is made, thus resulting in large overheads for signaling. This disadvantage has prevented it as an optimal solution, especially in a dense scattering mechanism in rural areas. Unlike conventional distance-based relay placement algorithm which takes only the physical distance and no shadowing effect into account, a modified DIS which also takes the diversity gain from the direct link between source and destination into account in order to optimize the error probability of the DF scheme is introduced in Kwak et al. [2008]. The algorithm assumes that source, relays, and destination lie on a straight line that aims to minimize the distance, hence the propagation loss. By determining the underserved position with largest distance from the base station (the further edge of the underserved region, denoted as “Target”), the first relay position can be determined. The relay position can later be split into two with two different azimuth angles, so on and so forth until the targeted coverage is reached. It should be emphasized that the distance between relay and destination is minimized so that the error probability at the receiver can be reduced, while the distance between source and relay is maximized so that only adequate signal is received at the relay for decoding before being forwarded to the destination.

Figures 7a–7d show the aggregated coverage for DIS-based MRS at 800 MHz and 3500 MHz for a fixed relay transmit power of 0.3 W. At 800 MHz, 75 % of the areas can be covered when six relay nodes are deployed on a straight line (one-target direction). A further increase of relay nodes will increase only the interference level in the channel. In order to achieve a higher coverage of 95%, 2 target directions are required, with a total of 14 relay nodes lying on two straight lines, representing the 2 target directions. At 3500 MHz, 20 relay nodes are required to achieve coverage of 82% with 4 targets. Further increase of the relay number will introduce higher interference; hence, coverage of 95% can never be achieved.

Figure 7.

(a) Coverage of DIS-based MRS at 800 MHz. (b) The aggregated coverage versus relay number for DIS-based MRS at 800 MHz with one and two targeted directions. (c) Coverage of DIS-based MRS at 3500 MHz. (d) The aggregated coverage versus relay number for DIS-based MRS at 3500 MHz with one to four targeted directions.

4.6 Hybrid Relay Placement Algorithm

wHaving acknowledged the DIS takes no shadowing into account, for areas with serious shadowing, for example, trees in a typical rural area, this results in many new target directions to be established within a small region, which indicates a serious waste of resources. On the other hand, it has been discussed in section 'Distance-Based Relay Placement Algorithm' that the shadowing effect can be better combated using the LMP scheme. In this section, a hybrid algorithm that combines DIS-based and path loss-based algorithm is proposed. The hybrid algorithm begins with the distance-based algorithm with a lower system demand of 70%. Further improvements at areas with deep shadowing are achieved with relays established following the LMP scheme. The coverage fraction value of 70% is justified based on the observation in the DIS algorithm where at least 70% coverage can be achieved at both 800 and 3500 MHz using solely DIS algorithm with a minimum number of target directions before the efficiency of the DIS algorithm started to be degraded by interference (see Figures 7b and 7d, with one target). Since further deployment of relay stations using DIS algorithm will lead to higher interference with only minor improvement on the overall coverage fraction, the LMP algorithm which deploys relay stations directly at locations where path loss is high, has become a much efficient scheme to further improve the overall coverage fraction in the hybrid algorithm. The LMP algorithm aims at those areas that undergo shadowing effects from the map and then distribute the additional relay stations accordingly. By doing so, the disadvantage of the DIS algorithm can be avoided. This approach avoids large overheads that are needed from the LMP scheme since at least 70% of the coverage were achieved using the DIS-based algorithm which can be implemented without large overheads, meanwhile allowing the areas with deep shadowing to be efficiently covered with the LMP scheme. The flowchart of the proposed hybrid algorithm is summarized in Figure 8.

Figure 8.

The flowchart of hybrid algorithm.

Figure 9a shows the coverage map of hybrid algorithm at 800 MHz and Figure 9b shows the comparison of aggregated coverage versus relay number at 800 MHz for the three different mechanisms, namely, LMP scheme (0.3 W), DIS-based algorithm (two targets), and Hybrid algorithm (one target DIS-based for 70% followed by LMP scheme at 0.3 W). The path loss-based algorithm gives the best performance with seven relay nodes needed to achieve coverage of 95%. The hybrid-based needs 11 relay nodes while the DIS-based algorithm (two targets) needs 14 to achieve the same coverage of 95%. Though the required number of relays in the hybrid algorithm is slightly higher than in the LMP scheme, the simulation running time shows great advantage. Figure 9c shows the coverage map of hybrid algorithm at 3500 MHz and Figure 9d shows the comparison of aggregated coverage versus relay number at 3500 MHz for LMP scheme (0.3 W), DIS-based (four targets), and hybrid algorithm (one target DIS followed by LMP 0.3 W). In order to achieve coverage of 95%, the LMP scheme needs 19 relay nodes, while the hybrid algorithm needs only 16 relay nodes. Here the performance of LMP is limited by interference due to the high transmit power of 0.3 W. For coverage below 85%, the LMP needs few relay nodes to achieve the same coverage as compared to the hybrid algorithm. For coverage higher than 85%, a further increase of relay nodes supposedly should improve the coverage in the underserved area. However, the interference introduced to the previously well-served areas degrades the overall performance. In contrast, the hybrid algorithm adopts the DIS-based algorithm to achieve a basic coverage of 70%, while the LMP-based relay nodes are only deployed at regions with deep shadowing. Thus, the interference-limited behavior has been avoided. The DIS-based algorithm needs 20 relay nodes (four targets) to achieve coverage of 82%. Any further increase of relay nodes leads to higher interference; hence, coverage of 95% can never be reached.

Figure 9.

(a) Coverage of hybrid relay placement algorithm at 800 MHz. (b) The comparison of aggregated coverage for different relay placement algorithms at 800 MHz. (c) Coverage of hybrid relay placement algorithm at 3500 MHz. (d) The comparison of aggregated coverage for different relay placement algorithms at 3500 MHz.

Generally speaking, any increment of number of relay stations for LMP algorithm will increase the overall aggregated coverage at 800 and 3500 MHz; however, any increment of number of relay stations for hybrid and DIS algorithms may not always lead to larger aggregated coverage. Some of the relay stations in the DIS algorithm (and also the first 70% coverage of the Hybrid algorithm) merely serve as intermediate source for coverage extension meanwhile introducing interference to the existing relay stations. Hence, the efficiency of these algorithms is reduced.

4.7 Power Allocation

In an MRS system, the trade-off between the coverage improvement and additional power required to establish the relay nodes have been a concern. To explore the additional energy required in exchange for overall coverage improvement, this paper adopts a power allocation scheme reported in Zuari et al. [2009]. The performance of the scheme is optimized in a channel model which assumes an additive white Gaussian noise and multiplicative flat fading with Rayleigh distributed amplitudes. The fading coefficients are constant over block and uncorrelated. The average transmitted energy per symbol Es, when all R relays are selected to assist in the MRS system is given as [Zuari et al., 2009]

display math(6)

where Er is the average received energy per symbol. The energy transmitted per information bit is expressed as [Zuari et al., 2009]

display math(7)

where h is the ratio between the number of bits and modulation symbols with code rate Rc. In the uniform power allocation scheme, all relays transmit the same power as the source. In an ideal power allocation scheme, the power among source and relay nodes are balanced so that the average received power at the destination is the same. The transmit power at the source is [Zuari et al., 2009]

display math(8)

While the transmit power at the source is [Zuari et al., 2009]

display math(9)

where inline image, with inline image as the distance between relay Rr and destination, dSD is the distance between source and destination. The path loss exponent is β. In this approach, the transmission power of the relay decreases as it is close to the service target; thus, the frame error rate is less sensitive to the positions of relay nodes. Figure 10a compares the aggregated coverage for DIS-based relay placement algorithm with uniform and ideal power allocation schemes at 800 MHz; also, the corresponding transmit power of each relay node is given in Figure 10b. It has been observed that, with an ideal power allocation scheme, the transmit power at relays 8, 10, and 13 can be reduced while maintaining the number of relays required to achieve the targeted coverage. The total power saving corresponds to 20 % of the total transmit powers of 14 relay stations when no power allocation scheme is adopted.

Figure 10.

(a) Comparison of coverage with ideal and uniform power allocation at 800 MHz. (b) The corresponding transmit power for ideal power allocation scheme at 800 MHz. (c) Comparison of coverage with ideal and uniform power allocation at 3500 MHz. (d) The corresponding transmit power for ideal power allocation scheme at 3500 MHz.

Figure 10c compares the aggregated coverage for DIS-based relay placement algorithm with uniform and ideal power allocation schemes at 3500 MHz; also, the corresponding transmit power of each relay node is given in Figure 10d. At 3500 MHz, the transmit power of relays 3, 4, 8, 15, and 20 can be reduced while allowing the number of relays required to achieve the targeted coverage remain unchanged. The total power saving corresponds to18 % of the total transmit powers when no power allocation scheme is used.

5 Conclusion

In this paper, a multihop relaying system for BWA at 800 MHz and 3500 MHz in rural areas has been discussed. The MRS takes into account the propagation effects present in the environment in order to provide a coverage that reflects the realistic channel links in these areas. Under the assumption that all relay stations are established at the rooftops of residential buildings in order to utilize the better link condition, three relay placement algorithms, namely, path loss-based, distance-based, and hybrid have been discussed. With the path loss-based algorithm, the number of relays required is the smallest, while the distance-based algorithm needs the highest number of relay nodes. Though the path loss-based algorithm needs the smallest number of relay nodes, the drawback of large overheads needed for path loss prediction has prevented it as the best choice for the realization of an MRS system, especially in a delay sensitive BWA system. In such a system, a hybrid algorithm that combines path loss-based and distance-based algorithms has appeared to be a better choice with smaller overheads and also able to effectively combat shadowing, which appears to be a problem for the distance-based algorithm. The number of relays needed in the hybrid algorithm is small and comparable to the path loss-based algorithm. Also, it has been shown in section 'Power Allocation', with an intelligent power allocation scheme, the transmit power of the relay nodes can be reduced by 18 to 20% without compromising the coverage achieved with the same number of relay nodes.

Ancillary

Advertisement