ROUTING ALGORITHMS | WiMAX Mesh Networks

In this section we will investigate routing and scheduling algorithms for WiMAX mesh networks, with the target application of using WiMAX mesh networks to network cellular BSs and gateways to the Internet. The gateways of general cellular networks can be mobile switching center (MSC) in GSM networks or MSC and serving GPRS support node (SGSN) in WCDMA networks. In this network architecture, the gateways act as 802.16 mesh BSs in the 802.16 backhaul network. They manage the backhaul network and allocate bandwidth to the SSs. The cellular BSs act as 802.16 SSs in the backhaul network. In the remaining part of the chapter, the cellular BSs will be called SSs for simplicity. The SSs forward aggregated traffic from the mobile users to Internet in the uplink direction and deliver traffic to the mobile users from the Internet in the downlink direction via the 802.16 BSs. For simplicity, only uplink traffic with QoS requirement is considered in this chapter. However, the approaches of routing and scheduling can be applied to bidirection traffic. It is assumed that the backhaul topology and traffic demands from the SSs will not change frequently. Therefore, the frequency of updating traffic routes and scheduling will be low. It is feasible and beneficial to use high performance algorithms such as the optimal centralize scheduling and routing algorithms.

CLASSIFICATION

As IEEE 802.16 standards specify the MAC and PHY layer protocols, routing protocols are outside the scope of the standard work. We can identify the following ways of solving routing and scheduling (bandwidth allocation) problems for WiMAX mesh networks:

• Centralized routing and centralized scheduling (CRCS): All the traffic, position information of the stations are sent to the BSs. The BSs jointly solve the routes and schedule problems, and send the routes/schedule information back to the SSs.
• Distributed routing and centralized scheduling (DRCS): SSs distributedly find their routes to the BSs. After the routes are found, the SSs send their information to the BSs. The BSs determine collision-free transmission schedule.
• Distributed routing and distributed scheduling (DRDS): Stations distributively find their routes to the BSs. After SSs find their routes to BSs, they work with their next-hop stations toward the BSs to determine collision-free transmission schedules in a distributed way.
• Hybrid routing and scheduling (HRS): Both routing and scheduling algorithms can be implemented in either a distributed or a centralized way.

In WiMAX mesh networks, distributed scheduling can be used by SSs with their neighbor SSs to reserve time slots. However, the success of reserving time slots will depend on the availability of free time slots not reserved by the centralized scheduling and other neighbor stations. As centralized scheduling has priority over distributed scheduling in WiMAX, it will be easier to provide time slots reservation for end-to-end QoS support. Therefore, we will consider only centralized scheduling in this chapter.

Centralized scheduling algorithm will work together with either centralized or distributed routing algorithms. For the centralized routing algorithm, routing, bandwidth allocation, and scheduling problem can be jointly investigated. We will formulate the joint routing, allocation, and scheduling problem with an optimal mathematical model. The input to the optimal model from the routing point of view will be network connectivity matrix CM, CM_ij = 1 if and only if station i is in the communication range of station j. After the optimal problem is solved, the route from an SS to a BS will be determined.

For the centralized scheduling and distributed routing algorithms, routes from SSs to BSs will be determined first. Then we can also derive network connectivity matrix CM for distributed routing. However, in this case, the station connectivity matrix CM will not only be determined by the communication ranges but also the traffic routes. We have CM_ij = 1 if and only if the link e_ij between station i and station j is in any route of SSs to BSs. Then the connectivity matric CM can be input to the optimization model for centralized scheduling. 

DISTRIBUTED ROUTING

In the WiMAX mesh networks, the BSs will periodically broadcast network configuration messages, which are further forwarded by the neighbor stations. The forwarding process continues toward network edge until the predefined maximum number of hops is reached. For each forwarding neighbor stations, it will add the information of hops from itself to the BSs. Although routing algorithm is not specified in the standards, the distance information together with other local information can be utilized by a new SS to find a route to a BS.

For a new SS to join the WiMAX mesh network, it is required to listen to the network configuration/synchronization from its neighbor stations. After it hears network configuration message at least twice from a station, it can send request to join the network through this station. With the available local information, such as the signal strength, distance between, the new station can select which station to be its sponsor station to a BS. Therefore the distributed routing algorithm can be reduced to find a sponsor station for a route toward a BS. We simply give four methods.

• Random selection (RS): In this method, a new station will choose a neighbor station as its sponsor station randomly among the candidate stations with the same minimum number of hops to a BS.
• Minimum node ID (MID): Among all the neighbor stations with the same minimum number of hops to any BSs, the station with minimum node ID will be selected as the sponsor station.
• Maximum signal strength (MSS): The station with maximum average signal strength will be selected as the sponsor station among the candidate stations, which can achieve higher transmission data rate.
• Minimum aggregate traffic (MAT): To balance traffic routed over the neighbor stations, a new station can also choose the station with minimum aggregated traffic load among the candidate stations.

LITERATURE SURVEY

An in-depth survey which covers the research activities and open issues from physical layer to applications, and from mobility management to power management, from designs with single channel single radio device to those with multiple channel multiple radio devices. However, most of the existing researches about wireless mesh networks are based on IEEE 802.11 standard. The routing and MAC layer protocols designed for 802.11 standard-based wireless mesh network cannot be used directly or efficiently for WiMAX mesh networks. And not much work has been done on WiMAX mesh networks. In this section, we will introduce the routing and scheduling related work on WiMAX mesh networks.

COORDINATED DISTRIBUTED SCHEDULING

The main idea of the coordinated distributed scheduling is to coordinate the transmission of MSH-DSCH messages over transmission opportunities in a collision-free manner. Through the exchanges of collision-free MSH-DSCH over control subframes, collision-free data slot reservations in the data subframes can be achieved.

To achieve the goal of collision-free MSH-DSCH transmission, nodes will exchange 2-hop or 3-hop neighborhood scheduling information with each other. Because nodes shall run the scheduling algorithm independently, a common algorithm has been specified in the standard for each node in the neighborhood to calculate the same schedule. The algorithm is random and predictable by dynamically constructing the seeds of a random number generator for each node according to a common rule. In particular, the seed for a given node is constructed based on its unique node ID and the index of the candidate transmission opportunity. The most important parameters that have significant impacts on the performance of coordinated distributed scheduling are Xmt Holdoff Exponet (3 bits) and Next Xmt Xm (5 bits). They can be used to control the contention on the transmission opportunities and improve bandwidth utilization.

Due to the importance of the coordinated distributed scheduling on the network performances, an analytical framework is needed to assess the performance of the scheduling scheme. Cao et al. analytically investigated how the channel contention is correlated with the total node number, exponent value, and network topology. With the assumption that the transmit time sequences of all the nodes in the control subframe form statistically independent renewal processes, they developed methods for estimating the distributions of the node transmission interval and connection setup delay. The analytical method will be helpful for evaluating upper layer performance like throughput and delay. They implemented the coordinated distributed scheduling module in NS-2 and showed that their analytical model is quite accurate under various scenarios, including both single hop and multihop networks.

Based on Cao’s analytical model, Bayer et al. presented an enhancement of the model. In particularly, they evaluated the scalability of the coordinated distributed scheduling. A scalability problem was observed that leads to poor performance in dense networks and aggravates QoS provisioning. The problem may result from the election-based transmission timing mechanism for scheduling the transmission of MSH-DSCH messages. They propose a dynamic adaptation mechanism to counteract the scalability problem, in which the parameter Xmt Holdoff Exponet is dynamically and locally adjusted according to the network contention and the status of a node. The Next Xmt Xm is used as a contention indicator. If Next Xmt Xm used by the node or its neighbors exceeds a specified threshold, the Xmt Holdoff Exponet is increased. The status of a node is defined according to its transmission activity, if it is BS or if it is a sponsor node. Significant UDP throughput increase is observed with the application of the adaptation mechanism for both single and multiple hop network scenarios.

In the 802.16 standard and the above analytical work, it is assumed that the control messages can be transmitted without collision in the extended neighborhood (2-hop or 3-hop). However, such kind of interference model may not hold in practice. Zhu and Lu investigate the performance of coordinated distributed scheduling under a realistic interference model [12]. Extensive simulations were conducted to evaluate the reception collision performance of the scheduling mechanism. It was reported that the collision ratio of control messages can be as high as 20 percent for 2-hop extended neighborhood. They studied how to deal with the collision problem by appropriate configuration of parameters such as Xmt Holdoff Exponent.

MESH NETWORK ENTRY MECHANISM

The standard specifies network entry mechanism to find sponsor nodes and establish links with neighbors. In this chapter, the network entry mechanism is used as one of the alternatives to establish traffic routes for WiMAX mesh networks. Upon entering the mesh network, a new SS searches for MSH-NCFG to acquire coarse synchronization with the network. Once the physical layer has achieved synchronization, the MAC layer can acquire network parameters for the MSH-NCFG message. Meanwhile, the SS can builds a physical neighbor list, from which the SS can select a potential sponsor node out of the eligible sponsor nodes. How to select the sponsor node is not specified in the standard. The selection method will be discussed in more details later in this chapter. The SS then synchronizes its time to the potential sponsor node and sends a network entry request message to the potential sponsor node. If the candidate sponsor node accepts the request and opens a sponsor channel, the channel is ready for use to register with the BSs. After the SS is authorized to enter the network by the BS, it can request bandwidth from the BS via the sponsor node and can also establish links with the SSs other than the sponsor node.

BANDWIDTH ALLOCATION AND GRANT MECHANISMS

In the 802.16 standard, flexible bandwidth allocation and grant mechanisms have been defined for PMP mode to guarantee the QoS of various service flows. Those mechanisms include periodically polling, real-time polling, nonreal-time polling, contention-based bandwidth request scheme, poll-me bit, bandwidth stealing, and piggyback. However, the bandwidth request and grant mechanisms can not be used in Mesh mode due to the multihop networking in Mesh mode. In WiMAX Mesh mode, all the communications in the links in the network are controlled by three ways, i.e., using a centralized scheduling algorithm, using a distributed scheduling algorithm within each node’s extended neighborhood, or using a combination of these two types of algorithms.

In the coordinated distributed scheduling algorithm, all stations (BS and SSs) coordinate their transmissions in their extended two-hop neighborhood. Coordinated distributed scheduling does not rely on the operation of a BS and transmissions are not necessarily directed to or from the BS. Within the constraints of the coordinated schedules, uncoordinated distributed scheduling can be used for fast and ad-hoc setup of schedules on a link-by-link basis with directed requests. Grants of the uncoordinated schedules need to ensure that the resulting data transmissions do not cause collisions with the data and control traffic scheduled by the coordinated scheduling algorithms.

Although distributed scheduling algorithms are more scalable, they are inefficient in QoS guarantee. On the other hand, centralized scheduling ensures collision-free scheduling over the links in the network, typically in a more optimal manner than the distributed scheduling method. Therefore better QoS support and network bandwidth utilization can be achieved. In this chapter centralized scheduling will be the only scheduling algorithm studied.

Centralized Scheduling

With the centralized scheduling algorithm, transmission schedule for the SSs is defined by the BS. The BS determines the flow assignments from the resource requests from the SSs. Subsequently, the SSs validate the MSH-CSCH schedule and determine the actual schedule from these flow assignments. The assignments determined by the BS extends to those SSs not directly connected to the BS. Intermediate SSs are responsible for forwarding bandwidth requests for SSs listed in the routing tree that are further from the BS (i.e., more hops from the BS) and the MSH-CSCH message from the BS to their neighbors as required. The SS resource requests and the BS assignments are both transmitted during the schedule control subframe. Centralized scheduling ensures that transmissions are coordinated to ensure collision-free scheduling over the links in the routing tree to and from the BS. The centralized scheduling will persist over a duration that is greater than the cycle time to relay the new resource requests and distribute the updated schedule.

Determination of the flow assignment and routing tree is outside the scope of the standard.

Distributed Scheduling

Coordinated distributed scheduling ensures that transmissions are scheduled in a manner that does not rely on the operation of a BS, and that is not necessarily directed to or from the BS. In the coordinated distributed scheduling mode, all the stations (BS and SSs) need coordinate their transmissions in their extended two-hop neighborhood. The coordinated distributed scheduling mode uses some or the entire control portion of each frame to regularly transmit its own schedule and proposed schedule changes on a PMP basis to all its neighbors. Within a given channel all neighbor stations receive the same schedule transmissions. All the stations in a network have to use the same channel to transmit schedule information in a format of specific resource requests and grants.

Within the constraints of the coordinated schedules (distributed or centralized), uncoordinated distributed scheduling can be used for fast, ad-hoc setup of schedules on a link-by-link basis. Uncoordinated distributed schedules are established by directed requests and grants between two nodes, and shall be scheduled to ensure that the resulting data transmissions (and the request and grant packets themselves) do not cause collisions with the data and control traffic scheduled by the coordinated distributed nor the centralized scheduling methods.

The major differences between coordinated and uncoordinated distributed scheduling are in the portions where the MSH-DSCH messages are transmitted. In the coordinated case, the MSH-DSCH messages are scheduled in the control subframe in a collision-free manner. In the uncoordinated case, MSH-DSCH messages are transmitted in data subframes and may collide.

Routing and Scheduling for WiMAX Mesh Networks

In addition to the mandatory implementation of point-to-multipoint (PMP) mode, WiMAX networks can be optionally configured to work in Mesh mode, to achieve increased reliability, coverage, and reduced network costs. Although IEEE 802.16 standards specify several quality of service (QoS) schemes and related message formats for WiMAX networks, the problems of scheduling algorithms for both PMP and Mesh mode are left unsolved. Routing algorithms for WiMAX networks are outside the scope of the standard work as well. In this chapter, we investigate the issues of routing and scheduling for WiMAX mesh networks. We overview the mesh mechanisms specified in the IEEE 802.16 standard and survey the existing research on scheduling and routing for WiMAX mesh networks. Then both distributed and centralized routing algorithms are studied, and their effectiveness on alleviating potential network congestion is compared. The scheduling problem is mathematically modeled by taking into account the interference constraints. Solutions are developed which can maximize the utilization of network capacity subject to fairness constraints on the allocation of scarce wireless bandwidth.

INTRODUCTION

In today’s telecommunications, networking and services are changing in a rapid way to support next generation Internet (NGI) user environment. Wireless networks will play an important role in NGI. Wireless broadband networks are being increasingly deployed and used in the last mile for extending or enhancing Internet connectivity for fixed or mobile clients located on the edge of the wired network.

With high data rate, large network coverage, strong QoS capabilities, and cheap network deployment and maintenance costs, WiMAX is regarded as a disruptive wireless technology and has many potential applications. It is expected to support business applications, for which QoS support will be a necessity. Depending on the applications and network investment, WiMAX network can be configured to work in different modes, point-to-multipoint (PMP) or Mesh mode. An illustration of PMP and mesh mode is presented in Figure 1. For example, the network can have a simple base station (BS) working in PMP mode and serving multiple subscriber stations (SSs) if the potential SSs can be covered by the BS. Mesh topology is an optional configuration for WiMAX networks. In the Mesh mode, traffic demands are aggregated at a set of SS nodes which are equipped with 802.16 interfaces. Subsequently, the traffic demands at SS nodes are delivered to a set of BSs nodes which functions in the PMP mode. These BS stations can be connected by a backhaul and connected to Internet access point (IAP) nodes. An amendment to the IEEE 802.16 specifications (where WiMAX is based) is IEEE 802.16j, Multihop Relay Specification. IEEE 802.16j is being developed by IEEE 802.16’s Relay Task Group. It is expected to extend reach/coverage through relaying. 

 

Figure 1: WiMAX (a) PMP network and (b) mesh network architectures.

Wireless mesh network offers increased reliability, coverage, and reduced network costs. There are extensive research, standardization, and commercial development activities on mesh networks. For example, several IEEE special task groups have been established to define the requirements for mesh networking in wireless personal area networks (WPANs), wireless local area networks (WLANs), and wireless metropolitan area networks (WMANs). A brief description of the standardization activities. Wireless mesh networks can be a prospective solution for broadband wireless Internet access in a flexible and cost-effective manner. However, wireless mesh networks also raise a number of research challenges, e.g., network routing, scheduling, QoS support, network management, etc. Those challenges are faced by WiMAX mesh networks without exception. WiMAX Mesh mode is defined with OFDM for frequency between 2 and 11 GHz and time division multiple access (TDMA) is used in the Medium Access Control (MAC) layer to support multiple users. Unlike the single hop wireless networks, routing algorithms are required to determine routes for the connections between a SS and a BS. As WiMAX networks operate synchronously in a time-slotted mode, it is also necessary to allocate time slots without collision over the network to achieve assigned bandwidth for each connection. More challenging is that the routing and scheduling for WiMAX networks are tightly coupled. The routing and scheduling problem for WiMAX networks is different from 802.11-based mesh networks. In the 802.11-based mesh networks, the MAC layer is contention-based; routing algorithms and MAC layer protocols can be designed and operated separately.

Although IEEE 802.16 standards specify several QoS schemes and related message formats, the problems of scheduling algorithms for both PMP and Mesh mode are left unsolved. Routing algorithms for WiMAX networks are outside the scope of the standard work as well. In this chapter, we will investigate the issues of routing and scheduling in WiMAX mesh networks. Both distributed and centralized routing algorithms will be studied, and their effectiveness on alleviating potential network congestion will be compared. The scheduling problem will also be mathematically modeled by taking into account the interference constraints. Solutions are developed to maximize the utilization of network capacity subject to fairness constraints on allocation of scarce wireless resource among the SSs.

CONCLUSIONS AND RECOMMENDATIONS | Relay-Assisted Mobile WiMAX

Achieving high throughputs with low outage probability in a large cell is a challenge, but this drives effective deployment with MIMO and multihop relay techniques, which have been intensively explored in this chapter. We have analyzed the effective relay efficiency. For relay systems without radio resource sharing, a higher relay SNR gain was required. This implies that the system requires a higher transmit power at the RS. In contrast, radio resource sharing offers high potential for a mobile WiMAX network when relays are deployed. Radio resource sharing is very applicable to MIMO relaying. In addition to multiuser transmission, the relay with radio resource sharing has the potential to achieve optimal multiuser relay gain.

To achieve the practical application of radio resource sharing, a relay system must be based on a topology that fully exploits effective resource assignment based on the spatial separation of nodes. Directional distributed relay topologies were introduced for highly efficient relay deployment. This scheme is fully backward compatible with the current mobile WiMAX standard.

Greater efficiencies in spectrum use can be achieved by coupling channel-quality information in the resource-allocation process. However, this study is heavily dependent upon realistic application environments. Channel modeling becomes critical to reflect application environments. With this purpose, a multihop relay measurement campaign was performed and is presented in this chapter. The measurement results have been analyzed and compared with several statistical and standardized models. It was found that the 802.16d model only matched well to the BS–RS link in terms of path loss and antenna-height correction factor. For the BS–MS link, the WINNER models were a good match to the shadowing, and the COST 231 Hata and WI models were useful for predicting path loss. The RS–MS link was also well matched by the WINNER shadowing values, but none of the path-loss models considered were a good match to any of individual measurement locations.

To demonstrate the potential of the directional distributed relay architecture in a realistic outdoor environment, a site-specific ray-tracing propagation model is used. Results show that, compared to a relay system without resource sharing, the implementation of resource sharing improves capacity significantly. For the OFDMA multiuser transmission, it is ideal to adopt flexible channelization and MIMO LA on different users, and multihop relay with efficient radio resource management.

MULTIHOP STUDY BASED ON RAY-TRACING

As mentioned before, a major issue for radio resource sharing is the interference. Greater efficiencies in spectrum use can be achieved by coupling channel-quality information in the resource-allocation process. Ray-tracing is an effective tool for network setup, evaluation, and optimization. To demonstrate the potential of the proposed resource sharing framework, a site specific ray-tracing propagation model is used in this section to provide realistic environment specific propagation data for the BS–MS, BS–RS, and RS–MS links.

The ray-tracing tool used in this work was verified with measurement data, and has been used in many previous WLAN and WiMAX system evaluations. The ray-tracing model takes individual buildings, trees, and terrain contours into account and determines specific multipaths based on scattering and diffraction. Mobile transitions from LoS to NLoS are naturally handled by the algorithms. In this study, we can easily utilize the ray tracer to establish the typical multihop relay application scenarios, such as coverage hole and cell edge. Also, it can directly demonstrate interference paths and strengths which is more suitable for studies presented in previous sections.

SIMULATION SCENARIO

Because mobile terminals are allowed to move freely at street level in the MWA scenario, local surrounding large and small obstructions and terrain contours result in path loss, shadowing, and multipath fading. Interferences from other signals also distort the transmitted signal in an unpredictable and time-varying fashion.

A realistic MWA scenario is now analyzed based on a region of Central Bristol, United Kingdom (Figure 10.16a). A single BS location was chosen on the roof top of a tall central building (30 m above ground level). Eighty-five MS units were distributed over this geographic area at street level with heights of 1.5 m. The BS with 3-sectors was assigned an EIRP of 57.3 dBm (based on a 15 dBi 1200 sector BS antenna). First, the raw multipath components (MPCs) are created using the above ray-tracing tool. Based on isotropic ray-traced channel data, the ETSI specific antenna beam patterns are then incorporated via spatial convolution. Figure 1 presents the distribution of received power from the BS to the surrounding region at street level. A severe coverage hole is clearly visible, and this is mainly caused by variations in the terrain height. Two methods could be used to achieve acceptable WiMAX coverage in this macrocell. First, a second BS could be deployed in the coverage hole. However, this would add to the infrastructure costs and hence it may be more effective to deploy an RS node.

 

Figure 1: Simulated macrocell in Bristol, United Kingdom.

Figure 2 shows the locations of MSs and their affiliations to either the BS or RS. The locations of BS, RS and MSs are overlaid on a terrain map of Bristol city-centre. For each MS a line is drawn to indicate whether communication occurs via the BS or RS. The choice of connection type depends on the received SINR. At a given location, signal power is obtained by sum of all received MPCs, which are transmitted by an assigned BS sector. Interferences are caused by MPCs which comes from either RS (for BS access) or BS (for RS access) if interfering users occupy same frequency resource as the detect user. It should be noted that in NLoS conditions the MS may connect to an adjacent sector since this is determined by the direction of the strongest path.

 

Figure 2: BS and RS radio resource location.

Figure 2 also demonstrates subchannel allocation for a 3-users OFDMA operating with a 5 MHz channel bandwidth, based on 512-FFT DL PUSC (Partial Usage of Sub-Channels) OFDMA TDD profile as specified in 802.16e. For the 512-FFT DL PUSC OFDMA, a total of 15 subchannels are mapped (after renumbering and permuting) in one OFDMA symbol. Each sector has access to one-third of the total number of subcarriers. There are 3 groups (one per sector), with 5 subchannels in each group. Each subchannel comprises 2 clusters (14 physical subcarriers in each cluster). This results in a total of 420 subcarriers in each OFDMA symbol (360 data bearing carriers and 60 pilot carriers). When the RS is used to connect to an MS, a certain number of timeslots (or alternatively subchannels) must be assigned in the covering sector to support the BS–RS link. Here we assume that BS–RS link and BS–MS link (taken from the sector covering the RS) are in group #3. When radio resource sharing is applied, a number of RS–MS and BS–MS links are supported simultaneously using the same resources. For all the RS–MS links, and also the BS–MS links where the antenna beam is steered away from the RS, it is possible to share group #1. Group #2 is used for those MSs that connect to one of the BS sectors (i.e., but not the sector coving the RS); these can be located near to the RS if required.

CALCULATION OF MEAN PATH LOSS AND SHADOWING

With one measurement approximately every 2 cm at a frequency of 3.5 GHz, the fast fading is captured. To extract the slow fading for calculation of mean path loss and shadowing, the fast fading was averaged through a sliding window of 40 wavelengths with successive windows overlapping by 30 wavelengths. An example of this process is shown in Figure 1 for the BS–MS link of route 6.

 

Figure 1: Example of BS–MS fast- and slow-fading along route 6, with RS at 5 m.

The shadowing on a wireless link is the signal fluctuations around the local mean level. For the purposes of this part, the local mean level will be obtained by fitting a least-squares regression curve to the data at each location and taking this as the mean path loss around which the shadowing fluctuations occur. For the BS–MS link, the data comprising all of the runs together will be treated as a single variable, but for the RS–MS link, the data collected at each different RS height will be treated separately. An example of the resulting shadowing distribution, on normal probability axes, so that a normal distribution with the same mean and variance as the actual data would be exactly a straight line. This is evidently a normally distributed variable, and calculation shows that it has zero mean and standard deviation of 3.5 dB. Similar distributions were found for the other routes and for the RS–MS link.

MULTIHOP RELAY CHANNEL MEASUREMENTS

Recently, several channel models have been developed for various environments and system topologies such as the COST 231, Stanford University Interim (SUI), and WINNER models. Unfortunately, statistical models of this type make a number of general assumptions that are not always met in practical WiMAX scenarios. For example, the SUI models are suitable for FWA (rather than MWA) and they are based only on cellular-type measurements in the 1.9 GHz band. The COST 231 models do not formally apply beyond 2 GHz. The WINNER models do reach to frequencies at 5 GHz, but does not include frequency correction factors to enable their use in the 3.5 GHz band envisaged for WiMAX. Furthermore, there is little by way of empirical data collected in specifically multihop scenarios in the 3.5 GHz making an evaluation of these models in their intended deployments difficult. This part presents the results of a measurement campaign conducted in outdoor city environment to subject these models to empirical scrutiny.

1 MEASUREMENT CAMPAIGN

Measurements were carried out along seven routes in Bristol city centre, United Kingdom, comprised of five RS locations as shown in Figure 1. These routes were identified from a ray-tracing analysis as being likely to exhibit deep shadowing from the BS (which was mounted on the roof of a building approximately 30 m above the top of a hill). The system under consideration is a 2-hop DL, as shown schematically in Figure 2. A commercial 3.5 GHz panel antenna was used with a beamwidth of 90° in azimuth and 10° in elevation with 17 dBi gain. Since the BS antenna was not omnidirectional, its orientation and downtilt are important. Prior to commencing the main campaign, power measurements were taken at each location for downtilts between 0° and 10° following which during measurement the antenna was set to the downtilt which maximized the received power. The antenna was oriented approximately southwesterly for routes 1–5 and northeasterly for routes 6 and 7.

 

Figure 1: Map of Bristol, United Kingdom, showing the BS, the 7 MS routes, and the 5 RS locations.

 

Figure 2: Schematic of multihop measurement system. f₁ = 2.58 GHz, f₂ = 3.467 GHz.

The RS was on a portable pump-up mast that was left fixed in place once on location. Two dipoles were mounted on top of the mast at either end of a beam that could be rotated in a plane. One end was fixed to act as the RS–MS transmitter, and the other end rotated around it to act as the BS–RS receiver; this rotation was to permit measurements of the local variation of the BS–RS signal. The BS transmitted 20 W before antenna gain and after cable loss, etc., on 3.59 GHz and the RS 3.67 W on 3.467 GHz.

The MS was constructed on a trolley that was pulled or pushed along the selected routes. A laptop on the trolley provided “command and control.” The two wheels on one axle of the trolley were equipped with electronic pulse counters that incremented the counts on the laptop approximately every 4 mm. Also connected to the laptop was a spectrum analyzer which received the signal from the antenna. The antenna (mounted at 1.65 m above ground) was the same type of dipole as used at the RS. The mean length of a route was about 90 m, and each route was repeated with the RS at a selection of heights of {2,3,4,5} m.

Path lengths were determined by using survey grade GPS equipment to obtain positional fixes for the start and end of each route and the trolley’s distance pulses to estimate the route in between. Elevation was assumed to change linearly along a route. Path loss at a particular point was determined as the difference between the transmitted and received power after taking account of the system gains and losses in cables, amplifiers, etc. The effect of antenna patterns was approximated as being the gain in the LoS direction between the transmitter and receiver.