Chapter 4: Frequency Utilisation and System Profiles
4.1 The Cellular Concept
The global objective of a wireless network is rather simple: to connect wireless users to a core network and then to the fixed network. Figure 4.1 illustrates the principle of a Public Land Mobile Network (PLMN), as defined for second-generation GSM networks.
Figure 4.1: Illustration of a Public Land Mobile Network (PLMN) offering a cellular service
The first wireless phone systems date back as far as the 1930s. These systems were rather basic and had very small capacity. The real boost for wireless networks came with the cellular concept invented in the Bell Labs in the 1970s. This simple but also extremely powerful concept was the following: each base station covers a cell; choose the cells small enough to reuse the frequencies (see Figure 4.2). Using this concept, it is theoretically possible to cover a geographical area as large as needed!
Figure 4.2: The cellular concept. simple and so powerful!
The cellular concept was applied to many cellular systems (mostly analogue) defined in the 1980s: AMPS (US), R2000 or Radiocom 2000 (France), TACS (UK), NMT (Scandinavian countries), etc. These systems being incompatible, a unique European cellular system was invented, GSM, which is presently used all over the world.
WiMAX applies the same principle: a BS covers the SSs of its cell. In this section, some elements of the cellular concept theory needed for WiMAX dimensioning are provided. First sectorisation is reminded.
A base station site represents a big cost (both as investment, CAPEX, and functioning, OPEX) for a network operator. Instead of having one site per cell, which is the case for an omnidirectional BS, trisectorisation allows three BSs to be grouped in one site, thus covering three cells (see Figure 4.3). These cells are then called sectors. Three is not the only possibility. Generally, it is possible to have a sectorisation with n sectors. Yet, for practical reasons, trisectorisation is very often used. Sectorisation evidently needs directional antennas. Trisectorisation needs 120° antennas (such that the three BSs cover the 360°).
Figure 4.3: Omnidirectional antennas and trisectorisation
For economical reasons, sectorisation is almost always preferred to omnidirectional antennas for cellular networks unless in rare specific cases, e.g. for very large cells in very low populated geographic areas. WiMAX is not an exception as sectorisation is also recommended.
4.1.2 Cluster Size Considerations
Cellular networks are based on a simple principle. However, practical deployment needs a complicated planification in order to have high performance, i.e. great capacity and high quality. This planification is made with very sophisticated software tools and also the ‘know-how’ of radio engineers.
Frequency reuse makes room for an interference that should be kept reasonably low. As illustrated in Figure 4.4, in the downlink (for example) and in addition to its useful signal (i.e. its corresponding BS signal), an SS receives interference signals from other BSs using the same frequency. Signal-to-Noise Ratio (SNR) calculations or estimations are used for planification. The SNR is also known as the to-Carrier-Interference-and-Noise Ratio (CINR). The term SNR is used more for receiver and planification considerations, while CINR is used more for practical operations. In this book, SNR and CINR represent the same physical parameter. An appropriate cellular planification is such that the SNR remains above a fixed target value (depending on the service, among others) while maximising the capacity.
Figure 4.4: Illustration of a useful signal and interference signal as used for SNR calculations
A parameter of cellular planification is the cluster size. A cluster is defined as the minimal number of cells using once and only once the frequencies of an operator (see Figure 4.5). It can be verified that having a small cluster increases the capacity per cell while big clusters decrease the global interference and then represent high quality. The choice of the cluster size must be done very carefully. In the case of GSM networks, the value of the cluster size was initially 12 or 9. With time and due to different radio techniques (such as frequency hopping, among others), the cluster size of the GSM may be smaller (down to 3, possibly).
Figure 4.5: Example of a (theoretical) regular hexagonal network. Cluster size = 3. In each cluster all the operator frequencies are used once and only once
The regular hexagonal grid (as in Figure 4.5) is a model that can be used for first estimations. For this model, a relation can be established between the minimal SNR value of the network and the cluster size :
where n is the cluster size and α a value depending on the radio channel (of the order of 4). This is only an approximated formula used for a nonrealistic channel (too simple) and cell shapes used only for a first estimation of n. Practical deployment uses more sophisticated means but still the minimal SNR (used for cells planification or dimensioning) increases with n.
What about WiMAX frequency reuse? WiMAX is an OFDM system (while GSM is a Single Carrier, SC, system) where smaller cluster sizes can be considered. Cluster size values of 1 or 3 are regularly cited. On the other hand, it seems highly probable that sectorisation will be applied. This gives the two possible reuse schemes of Figure 4.6. Consequently, an operator having 10.5 MHz of bandwidth will have 3.5 MHz of bandwidth per cell (respectively 10.5 MHz) if a cluster of 3 (respectively 1) is considered. Yet, it is not sure that a cluster of 1 will lead to a higher global capacity. With a cluster of 1, reused frequencies are very close and then the SNR will be (globally) lower, which means less b/s Hz if link adaptation is applied (as in WiMAX, see Chapter 11). The choice of cluster size is definitely not an easy question.
Figure 4.6: Possible frequency reuse schemes in WiMAX
Other cellular frequency reuse techniques can be also be used. Reference  mentions fractional frequency reuse. With an appropriate subchannel configuration, users operate on subchannels, which only occupy a small fraction of the whole channel bandwidth. The subchannel reuse pattern can be configured so that users close to the base station operate with all the subchannels (or frequencies) available, while for the edge users, each cell (or sector) operates on a fraction of all subchannels available. With this configuration, the full-load frequency reuse is maintained for centre users in order to maximise spectral efficiency and fractional frequency reuse is implemented for edge users to assure edge user connection quality and throughput. The subchannel reuse planning can be dynamically optimised across sectors or cells based on network load and interference conditions on a frame-by-frame basis. Figure 4.7 shows an example of operating frequencies for each geographical zone in a fractional frequency reuse scheme. In the OFDMA PHYsical Layer (Mobile WiMAX), flexible subchannel reuse is facilitated by the subchannel segmentation and permutation zones. A segment is a subdivision of the available OFDMA subchannels (one segment may include all subchannels).
Figure 4.7: Example of operating frequencies for each geographical zone in a fractional frequency reuse scheme. The users close to the base station operate with all subchannels available. (Based on Reference )
The tools for efficient radio resource use and other radio engineering considerations for WiMAX are described in Chapter 12.
Handover operation (sometimes also known as ‘handoff’) is the fact that a mobile user goes from one cell to another without interruption of the ongoing session (whether a phone call, data session or other). Handover is a mandatory feature of a cellular network (see Figure 4.8). Many variants exist for its implementation. Each of the known wireless systems have some differences. WiMAX handover is described in Chapter 14.
Figure 4.8: Illustration of handover in a cellular network
Lee, W. C. Y., Mobile Cellular Telecommunications: Analog and Digital Systems, McGraw Hill, 2000.
WiMAX Forum White Paper, Mobile WiMAX - Part I: a technical overview and performance evaluation, March 2006.
Chapter 10: Uplink Bandwidth Allocation and Request Mechanisms
10.1 Downlink and Uplink Allocation of Bandwidth
Downlink and uplink bandwidth allocations are completely different. The 802.16 standard has a MAC centralised architecture where the BS scheduler controls all the system parameters, including the radio interface. It is the role of this BS scheduler to determine the uplink and downlink accesses. The uplink and downlink subframe details were given in Chapter 9.
The downlink allocation of bandwidth is a process accomplished by the BS according to different parameters that are determinant in the bandwidth allocation. Taking into consideration the QoS class for the connection and the quantity of traffic required, the BS scheduler supervises the link and determines which SS will have downlink burst(s) and the appropriate burst profile. In this chapter, the uplink access mechanisms of WiMAX/802.16 are described. Chapter 11 describes scheduling and QoS.
In the uplink of each BS zone or, equivalently, WiMAX cell, the SSs must follow a transmission protocol that controls contention between them and enables the transmission services to be tailored to the delay and bandwidth requirements of each user application. This is accomplished while taking into account five classes of uplink service levels, corresponding to the five QoS classes that uplink transmissions may have.
Uplink access and bandwidth allocation are realised using one of the four following methods:
unsolicited bandwidth grants;
piggyback bandwidth request;
unicast polling, sometimes simply referred to as polling;
contention-based procedures, including broadcast or multicast polling, where contention based bandwidth request procedures have variants depending of the PHYsical Layer used:OFDM or OFDMA (see below).
The standard states that these mechanisms are defined to allow vendors to optimise system performance by using different combinations of these bandwidth allocation techniques while maintaining consistent interoperability. The standard proposes, as an example, the use of contention instead of individual polling for SSs that have been inactive for a long period of time. Next, the realisation of these methods is described, but first two possible differentiations of an uplink grant-request are introduced.
10.2 Types of Uplink Access Grant-request
The BS decides transmissions in the uplink and the downlink. For uplink access, a grant is defined as the right for an SS to transmit during a certain duration. Requests for bandwidth must be made in terms of the number of bytes needed to carry the MAC header and payload, but not the PHY overhead. For an SS, bandwidth requests reference individual connections while each bandwidth grant is addressed to the SS's Basic CID, not to individual CIDs. It is then up to the SS to use the attributed bandwidth for any of its CIDs. Since it is nondeterministic which request is being honoured, when the SS receives a shorter transmission opportunity than expected due to a scheduler decision, the request message loss or some other possible reason, no explicit reason is given.
Grants are then given by the BS after receipt of a request from an SS. Two possible differentiations can be made for this request. These differentiations are now described.
10.2.1 Incremental and Aggregate Bandwidth Request
A grant-request (by an SS) may be incremental or aggregate:
When the BS receives an incremental bandwidth request, it adds the quantity of bandwidth requested to its current perception of the bandwidth needs of the connection.
When the BS receives an aggregate bandwidth request, it replaces its perception of the bandwidth needs of the connection with the quantity of bandwidth requested.
The self-correcting nature of the request-grant protocol requires that the SSs should periodically use aggregate Bandwidth Requests. The standard states that this period may be a function of the QoS of a service and of the link quality, but do not give a precise value for it.
The grant-request may be sent in two possible MAC frame types that are described in the following subsection. Only the first one (the standalone bandwidth request) can be aggregate or incremental.
10.2.2 Standalone and Piggyback Bandwidth Request
The two MAC frame types of the 802.16 standard, already defined in Section 8.2, can be used by an SS to request bandwidth allocation from the BS. Specifically, Section 8.2.3 details MAC headers and gives two types of request that are now described.
The standalone bandwidth request is transmitted in a dedicated MAC frame having a Header format without payload Type I, indicated by the first bit of the frame, the Header Type bit, being equal to 1. A Type field in the bandwidth request header indicates whether the request is incremental or aggregate (see Table 8.3). In the bandwidth request header, a 19-bit long bandwidth request field, the Bandwidth Request field, indicates the number of bytes of the uplink bandwidth requested by the SS for a given CID, also given in this header.
The standalone bandwidth request is included in the two main grant-request methods: unicast polling and contention-based polling.
For any uplink allocation, the SS may optionally decide to use the allocation for:
requests piggybacked in data.
The piggyback bandwidth request uses the grant management subheader which is transmitted in a generic MAC frame (then having a generic MAC header). This is indicated by the first bit of the frame, the Header Type bit, being equal to 0. This can avoid the SS transmitting a complete (bandwidth request MAC header) MPDU with the overhead of a MAC header only to request bandwidth. The grant management subheader, in a generic MAC frame, is used by the SS to transmit bandwidth management needs to the BS. The Type bit in the generic MAC frame header (see Table 8.2) indicates the possible presence of a grant management subheader.
The piggyback bandwidth request (grant management subheader) is a lightweight way to attach a request uplink bandwidth without having to create and transmit a complete MPDU with the overhead of MAC headers and CRCs. The grant management subheader is two bytes long. It is used by the SS to transmit bandwidth management needs to the BS in a generic MAC header frame in addition to other possible data transmitted in the same MAC frame. Depending on the class of QoS of the connection, three types of grant management subheader are defined and used:
Grant management subheader (see Figure 10.1). This is the case for QoS class = UGS. The UGS (Unsolicited Grant Services) is a QoS class designed to support real-time service flows, where the SS has a regular uplink access (for more details on the UGS class, see Chapter 11). For this class of QoS, the PM (Poll-Me) bit in the grant management subheader can be used by the SS to indicate to the BS that it needs to be polled in order to request bandwidth for non-UGS connections (see below).
Figure 10.1: Grant management subheader for the QoS class = UGS (Unsolicited Grant Services)
The grant management subheader contains only two useful bits, the SI (Slip Indicator), used by the SS to indicate a slip of uplink grants relative to the uplink queue depth, and the PM (Poll-Me) bit, used by the SS to request a bandwidth poll, probably for a needed additional uplink bandwidth with regard to the regular access this UGS SS has. The Slip Indicator bit use by the SS is the following: the BS may provide for long-term compensation for possible bad conditions, such as lost maps or clock rate mismatches, by issuing additional grants. The SI flag allows prevention of the provision of too much bandwidth by the BS. The SS sets this flag once it detects that the service flow has exceeded its transmit queue depth. Once the SS detects that the service flow transmit queue is back within limits, it clears the SI flag. No precise values for these limits are given in the standard.
Piggyback grant-request subheader (see Figure 10.2). This is the case for the QoS classes ≠ UGS. Since the piggybacked bandwidth request subheader does not have a Type field, it will always be incremental. The piggyback request field is the number of bytes of the uplink bandwidth requested by the SS. The bandwidth request is for the CID indicated in the MAC frame header (see Section 8.2).
Figure 10.2: Grant management subheader for the QoS class = UGS (Unsolicited Grant Services). The piggyback request field is the number of bytes of the uplink bandwidth requested by the SS
Extended piggyback request. This is defined by the 16e amendment along with and for the (newly defined) ertPS class. The number of bytes of the uplink bandwidth (piggyback) requested by the SS is on 11 bits (instead of 16). This request is incremental. In the case of the ertPS class, if the MSB (Frame Latency Indicator, or FLI) of the grant management subheader is 1, the BS changes its polling size into the size specified in the LSBs of this field (Frame Latency (FL) field).
The standard (16e) states that FL and FLI fields may be used to provide the BS with information on the synchronisation of the SS application that is generating periodic data for UGS/extended rtPS service flows. The SS may use these fields to detect whether latency experienced by this service flow at the SS exceeds a certain limit, e.g. a single frame duration. If the FL indicates inordinate latency, the BS may shift scheduled grants earlier for this service flow (taking into account the FL).
The standard states that capability of the piggyback request is optional. This probably includes the PM grant management subheader.