As in all kinds of networks, nodes in a wireless network have to share a common medium for signal transmission. Multiple Access Control (MAC) protocols are algorithms that define the manner in which the wireless medium is shared by the participating nodes. This is done in a way that maximizes overall system performance. MAC protocols for wireless networks can be roughly divided into three categories: Fixed assignment (e.g. TDMA, FDMA), random access (e.g. ALOHA, CSMA/CA) and demand assignment protocols (e.g. polling). The large number of MAC protocols for wireless networks that have appeared in the corresponding scientific literature (a good overview appears in Ref. [6]) demands a large amount of space for a comprehensive review of such protocols. In this section, we present some basic wireless MAC protocols.
Figure 2.28 8-level QAM constellation encoding 3 bits/baud
2.6.1 Frequency Division Multiple Access (FDMA)
In order to accommodate various nodes inside the same wireless network, FDMA divides the available spectrum into subbands each of which are used by one or more users. FDMA is shown in Figure 2.29. Using FDMA, each user is allocated a dedicated channel (subband), different in frequency from the subbands allocated to other users. Over the dedicated subband the user exchanges information. When the number of users is small relative to the number of channels, this allocation can be static, however, for many users dynamic channel allocation schemes are necessary.
In cellular systems, channel allocations typically occur in pairs. Thus, for each active mobile user, two channels are allocated, one for the traffic from the user to the Base Station (BS) and one for the traffic from the BS to the user. The frequency of the first channel is known as the uplink (or reverse link) and that of the second channel is known as the downlink (or forward link). For an uplink/downlink pair, uplink channels typically operate on a lower frequency than the downlink channel in an effort to preserve energy at the mobile nodes. This is because higher frequencies suffer greater attenuation than lower frequencies and conse- quently demand increased transmission power to compensate for the loss. By using low frequency channels for the uplink, mobile nodes can operate at lower power levels and thus preserve energy.
Due to the fact that pairs of uplink/downlink channels are allocated by regulation agencies, most of the time they are of the same bandwidth. This makes FDMA relatively inefficient since in most systems the traffic on the downlink is much heavier than that in the uplink. Thus, the bandwidth of the uplink channel is not fully used. Consider, for example, the case of web browsing through a mobile device. The traffic from the BS to the mobile node is much heavier, since it contains the downloaded web pages, whereas the uplink is used only for conveying short user commands, such as mouse clicks.
The biggest problem with FDMA is the fact that channels cannot be very close to one another. This is because transmitters that operate at a channel’s main band also output some energy on sidebands of the channel. Thus, the frequency channels must be separated by guard bands in order to eliminate inter-channel interference. The existence of guard bands, however, lowers the utilization of the available spectrum, as can be seen in Figure 2.30 for
Wireless Communications Principles and Fundamentals 55
Figure 2.29 Illustration of FDMA
the first generation AMPS and Nordic Mobile Telephony (NMT) systems covered in Chapter 3.
2.6.2 Time Division Multiple Access (TDMA)
In TDMA [7], the available bandwidth is shared in the time domain, rather than in the frequency domain. TDMA is the technology of choice for a wide range of second generation cellular systems such as GSM, IS-54 and DECT which are covered in Chapter 4. TDMA divides a band into several time slots and the resulting structure is known as the TDMA frame. In this, each active node is assigned one (or more) slots for transmission of its traffic.
Nodes are notified of the slot number that has been assigned to them, so they know how much to wait within the TDMA frame before transmission. For example, if the bandwidth is spread intoNslots, a specific node that has been assigned one slot has to wait forN21 slots between its successive transmissions. Uplink and downlink channels in TDMA can either occur in different frequency bands (FDD-TDMA) or time-multiplexed in the same band (TDD- TDMA). The latter technique obviously has the advantage of easy trading uplink to downlink bandwidth for supporting asymmetrical traffic patterns. Figures 2.31 and 2.32 show the structure of FDD-TDMA and TDD-TDMA, respectively.
TDMA is essentially a half-duplex technique, since for a pair of communicating nodes, at a specific time, only one of the nodes can transmit. Nevertheless, slot duration is so small that the illusion of two-way communication is created. The short slot duration, however, imposes strict synchronization problems in TDMA systems. This is due to the fact that if nodes are far
Figure 2.30 Total and usable channel bandwidths for AMPS and NMT systems
Figure 2.31 Illustration of FDD-TDMA
from one another, the propagation delay can cause a node to miss its turn. This is the case in GSM. Each GSM slot lasts 577 ms, which poses a limit of 35 km on the range of GSM antennas. If this range were to exceed 35 km, the propagation delay becomes large relative to the slot duration, thus resulting in the GSM phone losing its slot. In order to protect inter-slot interference due to different propagation paths to mobiles being assigned adjacent slots, TDMA systems use guard intervals in the time domain to ensure proper operation. Further- more, the short slot duration means that the guard interval and control information (synchro- nization, etc.) may be a significant overhead for the system. One could argue that this overhead could be made lower by increasing the slot size. Although this is true, it would lead to increased delay which may not be acceptable for delay-sensitive applications such as voice calls.
Dynamic TDMA schemes allocate slots to nodes according to traffic demands. They have the advantage of adaptation to changing traffic patterns. Three such schemes are outlined below [8]:
† The first scheme was devised by Binder. In this scheme, it is assumed that the number of stations is lower than the number of slots, thus each station can be assigned a specific slot.
The remaining slots are not assigned to anyone. According to their traffic demands, stations can contend for the remaining slots using slotted ALOHA, which is presented in the next paragraph. If a station wants to use a remaining slot to transmit information, it does so at the start of the slot. Furthermore, a station can use the home slot of another station if it monitors this home slot to be idle during the previous TDMA frame, a fact that means that the slot owner has no traffic. When the owner wants to use its slot, it transmits at the start of the slot. Obviously a collision occurs which notifies other stations that the slot’s owner has traffic to transmit. Consequently, during the next TDMA frame, these stations defer from using that slot which can thus be used by its owner.
† The second scheme was devised by Crowther. In this scheme, it is assumed that the number of stations is unknown and can be variable. Thus, slots are not assigned to stations which contend for every available slot using ALOHA. When a station manages to capture a slot, it transmits a frame. Stations that hear this transmission understand that the station has successfully captured the slot and defer from using it in the next TDMA frame. Thus, a station that captures a slot is free to use it in the next TDMA frame as well.
Wireless Communications Principles and Fundamentals 57
Figure 2.32 Illustration of TDD-TDMA
† The third scheme, due to Roberts, tries to minimize the bandwidth loss due to collisions.
Thus, a special slot (reservation slot) in the TDMA frame is split into smaller subslots which are used to resolve contention for slots. Specifically, each station that wants to use a slot transmits a registration request in a random subslot of the reservation slot. Slots are assigned in ascending order. Thus, the first successful reservation assigns the first data slot of the TDMA frame, the second successful reservation assigns the second data slot, etc.
Stations are assumed to possess knowledge of the number of slots already assigned, so if the reservation of a station is completed without a collision, the station is assigned the next available slot.
2.6.3 Code Division Multiple Access (CDMA)
As seen above, FDMA accommodates nodes in different frequency subbands whereas TDMA accommodates them in different time parts. The third medium access technique, CDMA [9], follows a different approach. Instead of sharing the available bandwidth either in frequency or time, it places all nodes in the same bandwidth at the same time. The transmission of various users are separated through a unique code that has been assigned to each user.
CDMA has its origins in spread spectrum, a technique originally developed during World War II. The purpose of spread spectrum was to avoid jamming or interception of narrowband communications by the enemy. Thus, the idea of spread spectrum was essentially to use a large number of narrowband channels over which a transmitter hops at specific time intervals.
Using this method any enemy that listened to a specific narrowband channel manages to receive only a small part of the message. Of course, the spreading of the transmission over the channels is performed in a random pattern defined by a seed, which is known both to the receiver and transmitter so that they can establish communication. Using this scheme the enemy could still detect the transmission, but jamming or eavesdropping is impossible with- out knowledge of the seed.
This form of spread spectrum is known as Frequency Hopping Spread Spectrum (FHSS) and although not used as a MAC technique, it has found application in several systems, such as an option for transmission the physical layer of IEEE 802.11 WLAN. This can be justified by the fact that spread spectrum provides a form of resistance to fading: If the transmission is spread over a large bandwidth, different spectral components inside this bandwidth fade independently, thus fading affects only a part of the transmission. On the other hand, if narrowband transmission was used and the narrowband channel was affected by fading, a large portion of the message would be lost. FHSS is revisited in Chapter 9.
CDMA is often used to refer to the second form of spread spectrum, Direct Sequence Spread Spectrum (DSSS), which is used in all CDMA-based cellular telephony systems.
CDMA can be understood by considering the example of various conversations using differ- ent languages taking place in the same room. In such a case, people that understand a certain language listen to that conversation and reject everything else as noise.
The same principle applies in CDMA. All nodes are assigned a specificn-bit code. The value of parameternis known as the system’s chip rate. The various codes assigned to nodes are orthogonal to one another, meaning that the normalized inner product3 of the vector representations of any pair of codes equals zero. Furthermore, the normalized inner product
3The normalized inner productPof two vectors A and B is essentially the cosine of the angle formed between A and B. Thus, it is a metric of similarity of the two vectors, since for A and B being orthogonal,Pẳ0.
of the vector representation of any code with itself and the one’s complement of itself equals 1 and21, respectively. Nodes can transmit simultaneously using their code and this code is used to extract the user’s traffic at the receiver. The way in which codes are used for transmission is as follows. If a user wants to transmit a binary one, it transmits its code, whereas for transmission of a binary zero it transmits the one’s complement of its code.
Assuming that users’ transmissions add linearly, the receiver can extract the transmission of a specific transmitter by correlating the aggregate received signal with the transmitter’s code.
Due to the use of then-bit code, the transmission of a signal using CDMA occupiesntimes the bandwidth that would be occupied by narrowband transmission of the same signal at the same symbol rate. Thus, CDMA spreads the transmission over a large amount of bandwidth and this provides resistance to multipath interference, as in the FHSS case. This is the reason that, apart from an channel access mechanism, CDMA has found application in several systems as a method of combating multipath interference. Such a situation is the use of CDMA as an option for transmission the physical layer of IEEE 802.11 WLAN.
An example of CDMA is shown in Figure 2.33 where we map the transmission of the ones and zeros in stations’ codes to11 and21, respectively. For three users, A, B and C andnẳ 4, the figure shows the users’ codesCA,CBandCC, these stations bit transmissions and the way that recovery of a specific station’s signal is made.
CDMA makes the assumption that the signals of various users reach the receiver with the same power. However, in wireless systems this is not always true. Due to the different attenuation suffered by signals following different propagation paths, the power level of two different mobiles may be different at the BS of a cellular system. This is known as the near-far problem and is solved by properly controlling mobile transmission power so that the signal levels of the various mobile nodes are the same at the BS. This method is known as power control and is described in the next section. Furthermore, as FDMA and TDMA, CDMA demands synchronization between transmitters and receivers. This is achieved by assigning a specific code for transmission of a large sequence by the trans- mitter. This signal is known as the pilot signal and is used by the receiver for synchroniz- ing with the transmitter.
2.6.4 ALOHA-Carrier Sense Multiple Access (CSMA)
The ALOHA protocol is related to one of the first attempts to design a wireless network. It was the MAC protocol used in the research project ALOHANET which took place in 1971 at the University of Hawaii. The idea of the project was to offer bi-directional communications without the use of phone lines between computers spread over four islands and a central
Wireless Communications Principles and Fundamentals 59
Figure 2.33 CDMA operation
computer based on the island of Oahu. Although ALOHA can be applied to wired networks as well, its origin relates to wireless networks.
The principle of ALOHA is fairly simple: Whenever a station has a packet to transmit, it does so instantaneously. If the station is among a few active stations within the network, the chances are that its transmission will be successful. If, however, the number of stations is relatively large, it is probable that the transmission of the station will coincide with that of (possibly more than one) other stations, resulting in a collision and the stations’ frames being destroyed.
A critical point is the performance of ALOHA. One can see that in order for a packet to reach the destination successfully, it is necessary that:
† no other transmissions begin within one frame time of its start;
† no other transmissions are in progress when the station starts its own transmission; this is because stations in ALOHA are ‘deaf’, meaning that they do not check for other transmis- sions before they start their own.
Thus, one can see that the period during which a packet is vulnerable to collisions equals twice the packet transmission size. It can be proven that the throughputT(G) for an offered load ofGframes per frame time in an ALOHA system that uses frames of fixed size is given by
TðGị ẳGe22G ð2:16ị
Equation (2.15) gives a peak ofT(G)ẳ0.184 atGẳ0.5. This peak is, of course, very low. A refinement of ALOHA, slotted ALOHA, achieves twice the above performance, by dividing the channel into equal time slots (with duration equaling the packet transmission time) and forcing transmissions to occur only at the beginning of a slot. The vulnerable period for a frame is now lowered to half (the frame’s transmission time) which explains the fact that performance is doubled. The throughputTs(S) for an offered load ofGframes per frame time in a slotted ALOHA system that uses frames of fixed size is given by
TsðGị ẳGe2G ð2:17ị
which gives a peak ofTsðGị ẳ0:37 atGẳ1.
The obvious advantage of ALOHA is its simplicity. However, this simplicity causes low performance of the system. Carrier Sense Multiple Access (CSMA) is more efficient than ALOHA. A CSMA station that has a packet to transmit listens to see if another transmission is in progress. If this is true, the station defers. The behavior at this point defines a number of CSMA variants:
† P-persistent CSMA.A CSMA station that has a packet to transmit listens to see if another transmission is in progress. If this is true, the station waits for the current transmission to complete and then starts transmitting with probabilityp.Forpẳ1 this variant is known as 1-persistent CSMA.
† Nonpersistent CSMA.In an effort to be less greedy, stations can defer from monitoring the medium when this is found busy and retry after a random period of time.
Of course, if more than two stations want to transmit at the same time, they will all sense the channel simultaneously. If they find it idle and some decide to transmit, a collision will occur and the corresponding frames will be lost. However, it is obvious that collisions in a
CSMA system will be less than in an ALOHA system, since in CSMA stations ongoing transmissions are not damaged due to the carrier sensing functionality. When a collision between two CSMA nodes occur, these nodes keep collided packets in their buffers and wait until their next attempt, which occurs after a time intervaltẳks.sis the system’s slot time, which for wired networks equals twice the propagation delay of signals within the wire. For wireless networks,sis defined in another way, as described in Chapter 9. The value ofkis given by the exponential backoff algorithm, which uniformly selects a number from the interval [0…,2I21], whereiẳminðc;cwị. cis the number of consecutive collisions encoun- tered by the frame andcwis a system parameter that governs the maximum random number produced,k.
CSMA has found use in wireless networks, especially WLANs. In wired networks, CSMA is the basis of the IEEE 802.3 protocol, known to most of us as Ethernet. Ethernet is superior to simple CSMA due to the fact that it can detect collisions and abort the corresponding transmissions before they complete, thus preserving bandwidth. However, collision detection cannot be performed in WLANs. This is due to the fact that a WLAN node cannot listen to the wireless channel while sending, because its own transmission would swamp out all other incoming signals. Since collisions cannot be detected, CSMA-based protocols for wireless LANs try to avoid them; thus, CSMA schemes for such networks are known as CSMA- Collision Avoidance (CSMA-CA) schemes. CSMA-CA protocols are the basis of the IEEE 802.11 and HIPERLAN 1 WLAN MAC sublayers which are presented in Chapter 9.
2.6.5 Polling Protocols
Polling protocols are centralized. For a polling protocol to be applied, a central entity (Base Station, BS) is assigned responsibility for polling the stations within the network. If the BS decides that a specific station grants permission to transmit, it polls this station, meaning that it sends to the station a small control frame notifying it that it can transmit one or more frames. After the transmission of this station, the BS proceeds to poll the other stations of the network. If a station is polled but has no traffic to transmit, it notifies the BS of this fact, the procedure continues and the next station is polled.
Polling is an appealing MAC option, however, it demands that the BS possesses knowledge regarding the network topology (the nodes under its coverage) in order for the network nodes to be polled. Such knowledge is difficult to achieve for a wireless network since topology changes occur frequently due to the mobile nature of nodes and the fading wireless links.
Several polling protocols tailored to the characteristics of wireless networks have appeared.
In this section, we present the Randomly Addressed Polling (RAP) and Group RAP (GRAP) protocols [10–12], which were designed for Wireless LANs (WLANs) as an example. RAP and GRAP also use CDMA in one of their stages.
RAP combats the imprecise knowledge regarding the network topology by working, not with all the nodes under the coverage of the BS, but only with the active ones seeking uplink communication. In RAP, a station is said to be active, if it has a packet to transmit. The RAP protocol assumes an infrastructure cellular topology, which is presented in later paragraphs.
Within each cell, multiple mobile nodes exist that compete for access to the medium. The cell’s BS initiates a contention period in order for active nodes to inform their intention to transmit packets. The operation of RAP constitutes a number of polling cycles. When a collision between two or more stations occurs, these stations keep the collided packets and
Wireless Communications Principles and Fundamentals 61