TELE4652 Mobile and Satellite Communication Systems Lecture 5 Multiple Access Techniques Having studied in the previous lectures how electrical signals can be converted to and from electromagnetic waves using antennae, and then how we can understand and model the propagation of these electromagnetic waves through the radio channel, we now turn our attention to how we can insert data onto these electromagnetic waves efficiently. There are two principle aspect to our discussion: firstly, we will look at techniques with which we can allow many users to share the same radio channel at the same time, known as multiple access; and then we will briefly discuss the theory of modulation, which addresses the problem of inserting and removing our data from the carrier wave in such a way as to minimise the bandwidth we have used while maximising the robustness of the data to channel noise and errors. The radio spectrum is a limited resource, with certain specific frequency bands allocated for mobile cellular use and other bands to other wireless services. Our focus is thus how to use this limited spectral allocation to achieve maximum capacity. We ll consider three forms of multiple access in turn: Frequency Division Multiple Access (FDMA); Time Division Multiple Access (TDMA); and Code Division Multiple Access (CDMA). FDMA means users each transmit at separate, distinct, nonoverlapping frequency bands. TDMA systems involve users occupying a larger, common channel bandwidth, but only for a limited amount of time. In essence the channel is shared and cycled in time around a group of simultaneous users. CDMA is a more sophisticated technique, in which specially designed orthogonal signal sets are used to allow all users to all occupy the available bandwidth simultaneously. The receiver can then recover a particular user s signal by correlating against the corresponding signal of the orthogonal set. It is often helpful to think of our communication channel as offering us three basic finite resources with which to communicate with frequency, time, and power. Students will be familiar with the famous formula for the capacity of an AWGN channel, S C = B log 2 1 + N indicating that the maximum rate at which we can communicate data over the channel is determined by the bandwidth B available and the signal power S available (N is of course the average noise power on the channel). Each of our multiple access techniques is then a different way of dividing the available resources up amongst different users. FDMA divides the channel in frequency, and TDMA divides the channel in time, as illustrated in the diagram below. CDMA is a little different, can be thought of as power division multiple access, since we are effectively dividing the available signalling among users, as they all transmit over the same frequency region at the same time.
The first generation of cellular systems were necessarily FDMA. The data signals were continuous and not sampled, so the user required allocation of the entire frequency channel for the duration of the call. Our archetype first generation system is AMPS, which divided the spectrum into 30 khz channels, with the data modulated onto the carrier using analogue Frequency Modulation (FM). The movement to digital, along with advances in speech compression techniques, allowed second generation systems to be entirely digital and employ TDMA. GSM adopted GMSK modulation and featured eight users sharing a 200 khz channel via TDMA. On the other hand, the USDC (US Digital Cellular) system featured three users sharing a 30 khz AMPS channel, with π/4 DQPSK (differential quadrature phase shift keying) modulation. The steady increase in computing power allowed the adoption of CDMA, and the IS-95 system allowed many users to share a 1.25 MHz channel with π/4 DQPSK modulation. The success and flexibility of CDMA made it the basis for third generation networks, W-CDMA and cdma2000. Duplexing Duplexing refers to the ability of communication to occur in both directions at the same time. The basic problem is that a radio cannot transmit and receive simultaneously at the same frequency, unless separate antennae with widely spaced frequency bands are used. The use of separate antennae is not feasible on a mobile station, and in addition the large difference in transmitted power (in the order of 30 dbm) to received power (possibly in the order of 100 dbm) means that leakage and coupling from the transmitter would likely swamp the received signal anyway. There are two basic methodologies for achieving duplex communication. A system that alternatively transmits then listens is known as half-duplex. Half-duplex schemes are suitable for data communication, but phone conversations typically require fullduplexing, where communication can occur in both directions at the same time. To provide full-duplex communication in a mobile radio system, it is necessary to interleave transmitting and receiving functions in time in such a way that is transparent to users, or to transmit and receive at different frequencies. Frequency Division Duplexing (FDD) provides two separate frequency bands for users. The forward channels that carry information from the BTS to the MS are chosen from one band, while reverse channels are selected from the other. A bandpass filter called a duplexer is used in each mobile unit and in each base station to prevent energy from the transmitter reaching the receiver input. The forward and reverse channels must be separated by sufficient bandwidth so that the duplexer can attenuate transmitted signals in the receiver band, but not sufficiently far in frequency that a common antenna cannot be used for transmitting and receiving. In an analogue system like AMPS FDD must be employed, as the voice signals are continuous and not sampled, and if communication is to occur in both directions simultaneously then this communication must occur at separate frequencies. In AMPS forward and reverse channels were always assigned in pairs spaced by 45 MHz, to standardise the design of these duplexer filters. Primarily as a consequence of the
legacy of this first cellular system, FDD with the same 45 MHz spacing is also used in GSM, IS-95 CDMA, and even the third generation networks. The move to digital communication systems opened the possibility to employ time division Duplexing, TDD. Time Division Duplexing achieves duplex communications by having transmission and reception occurring at different time intervals. As long as these slots are sufficiently short in duration and close together the user perceives that two-way transmission is occurring simultaneously. The radio then does not transmit and receive simultaneously, and as such the expensive duplexer is not needed. TDD was adopted in GSM along with the FDD inherited from the first generation system GSM is a TDMA scheme in which each user is assigned a pair repeating time slot on a pair particular frequency channels, one for transmit and one for receive. GSM offsets these timeslots for the forward and reverse channels by three timeslots (there are eight total timeslots in a GSM frame). A GSM mobile and then transmit, wait a couple of timeslots and listen, so the duplexer is not required. The complications to a TDD scheme are firstly that overall synchronisation is needed between the transmitter and receiver, so that the forward and reverse signals do not overlap in time. TDD schemes are only possible for short distance transmission, since significant propagation delay can interfere with the timeslot coordination. Even more difficult is the time-varying propagation delay when either the transmitter or receiver are in motion. Multiple Access To motivate the following discussion of multiple access techniques, let s first describe in general terms the process through which a mobile station gains access to the network. The main points that we will gain from this discussion is, firstly, that there are really several different multiple access methodologies employed in each mobile network, and also that all cellular systems are in their nature circuit-switched, since once a frequency band, time slot, or code has been assigned to a subscriber, the subscriber has exclusive use of that channel for the entire duration of the call.
Each cell in a mobile network will have been assigned a certain number of traffic channels and control channel, where a channel could refer to a frequency band, time-slot, or spreading code, depending on whether it is a FDMA, TDMA, or CDMA system, respectively. One of the forward control channels will always be a sync, or synchronisation, channel (often referred to as the base channel ). The sync channel is like a beacon that transmits the synchronisation and system-dependent information needed by mobile units to connect to the network. When a mobile unit is turn on it scans a pre-programmed set of forward control channels to identify the strongest signal. From this forward control channel it then obtains the required information to connect to this base station, such as the access codes, location of random access and paging channels, and identification of the service provider. The mobile station (MS) must then access the system by transmitting a message on a random access channel, a reverse control channel. This random access channel is shared amongst all subscribers, who are operating in an uncoordinated fashion. A contention-based protocol must be used on this access channel, such as ALOHA or Slotted-ALOHA, described shortly. These schemes involve the MS transmitting its message in packet form and then waiting for the acknowledgement (ACK) from the BTS. If no ACK is received, indicating that some other MS in the cell transmitted at the same time corrupting both messages, the mobile station will attempt to transmit its packet again. When a mobile initiates a call, or responds to a page, it will also transmit on a random access channel. The BTS will then assign the MS a dedicated control channel, on which it does not have to contend with other users, to establish the connection and assign a traffic channel for the call. The generic procedures for a mobile initialised call and mobile responding to a page are shown in the diagrams below. The important points to see from this discussion is the existence of a variety of channel types on any cellular network, and that these channels can use different multiple access strategies. The random access channel will always feature a contention-based multiple access scheme. Let s now consider the principle of FDMA, Frequency Division Multiple Access. The central design issue is maximising the number of independent users that can be simultaneously supported in a given bandwidth. In essence this means deciding on the bandwidth of each channel, B chan. At a minimum, the channel bandwidth B chan must be large enough to accommodate the modulated signal without unreasonable distortion. On the other hand, the channel bandwidth determines how close different frequency signals are placed, and the channels must be separated by a sufficient region to protect against adjacent channel interference. Protecting against interference from neighbouring channels requires the establishment of guard bands between channels. These guard bands are naturally an overhead and reduce the available bandwidth and hence system capacity. There are two types of guard bands: guard bands between channels, which are usually included in calculation of the channel bandwidth; and guard bands at the edge of the spectral allocation, to protect against other systems and schemes, whose power leakage and properties are unknown.
We could express the capacity of an FDMA system as Bsys 2Bguard N = Bchan where N is the number of channels available, B chan is the bandwidth of each channel including guard regions, B guard is the spectral region assigned as guard band at the edge of the system spectral allocation of. The extent of the guard bands is B sys determined largely by three main factors: the selectivity of the filters that can be used at the receiver to separate desired from undesired signals; the relative range of variations in amplitudes of desired and undesired signals; and finally the out-of-band spectral occupancy of the transmitted signals, ultimately determined by the filters used on the transmitter side. The need to spatially separate neighbouring frequency channels is another design factor cell cluster assignment. Let s consider AMPS as a good example for the assignment of channels to minimise adjacent channel interference. Many later deployments of AMPS involved a 7 cell cluster deployment with 120 sectoring. The role sectoring was to improve the signal to interference ratio. Each cell was divided into three 120 sectors, with each sector containing one third of the channels assigned to that cell. This minimises co-channel interference since only two base stations in the first-tier will now produce interference, as the other four base stations in that tier transmit their radiation in other directions.
The 832 channels available across the AMPS band are labelled 1, 2, 3, 832. One cell within the cluster would be assigned channels 1,8,15,22,, and another 2,9,16,23,, so that channels within the same cell are separated by at least 7 30kHz or 210 khz. The channels were further sub-divided among the three sectors of the cell, so that channels within a sector were separated by 630 khz. While no guard bands were employed in AMPS, the channel assignment was employed to ensure that channels within a cell were maximally spaced through channel assignment. The main advantages of a FDMA system could be summarised as follows. Firstly, the channel bandwidth tends to be small, since this improves capacity, and is typically less than the channel coherence bandwidth. This means that no complex equalisation is needed at the receiver, and a simple flat filter is sufficient to recover the signal at the receiver. The fact that a channel is assigned to a user for the duration of the call
means that no synchronisation is needed between different subscribers, and that the signal processing and associated circuitry can be simple. A major disadvantage of FDMA is that the assignment of a channel to a user for the entire duration of the call is a significant waste of resources, as a significant portion of a conversation involves silence from either party. However, in an FDMA system the channel is still occupied during the idle periods and cannot be shared to another user. The move to digitisation, and advances in voice detection and speech compression, resulted in a move to TDMA, or Time Division Multiplexing, thereby allowing several users to share a single frequency channel. Other disadvantages of FDMA are the need for complex bandpass and RF filters to minimise adjacent channel interference, and the need for duplexers, since transmission and reception must necessarily occur at the same time. On naïve analysis a Time Division Multiple Access (TDMA) system, in which each user transmits at different time instants in a round robin fashion, should have the same capacity as an FDMA system. The analysis as follows: we know from digital modulation theory that a channel of bandwidth B sys fundamentally gives us 2 Bsys basis functions per second with which to communicate with. If our information signal is of bandwidth B chan then the sampling theorem says that we must sample this signal at a rate of at least 2 Bchan. The number of simultaneous users we could thus support in a round-robin transmission scheme is thus approximately, 2Bsys N users 2Bchan assuming we send a sample on each available basis function, which is the same capacity as the equivalent FDMA system. However, in practise a TDMA will be greatly superior to the FDMA equivalent. The reasons for this are the availability of speech compressors, called vocoders. These devices are able to take telephone quality speech, generally considered to be 8 bit
samples at a rate of 8 khz, for a bit rate of 64kbps, and compress it to bit rates lower than 16 kbps with no loss in perceptual quality. These speech compression systems, along with sophisticated digital modulation techniques and advanced digital signal processing algorithms, mean that the capacity and performance of digital TDMA cellular systems far exceeds the FDMA systems. The basic unit in a TDMA system is a repeating frame. An example of a frame structure is shown in the diagram below. Each frame is divided into timeslots as shown. Each user is assigned a certain timeslot within the frame, and only transmits or receives during that timeslot in a repeating fashion. It is apparent from the figure that the TDMA frames require significant overheads. At the beginning of each frame there is a preamble, allowing receivers to synchronise to the signal and identify the position of each time slot. There are further overheads within each time slot. The guard interval gives transmitters and receivers to switch and power-up. The synchronisation and preamble are used to achieve carrier and bit clock synchronisation, respectively. Aside from indicating the beginning of the data field, the preamble can also contain addressing information. These overhead bits, which are an unavoidable consequence of a digital multi-user radio system, reduce the proportion of the frame that can be assigned for data and hence the capacity of the system. We define the frame efficiency, η f, of a TDMA system as the fraction of bits per frame that are assigned to carry user data. The total overhead bits per frame is b OH = N rbr + N tbp + N tbg + N rbg where N r is the number of reference bursts per frame (synchronisation preambles), N t is the number of traffic burst, or user time-slots, per frame, br is the equivalent number of bits per reference burst, b g is the equivalent number of bits per guard interval, and bp is the number of bits per preamble. The total number of bits per frame is b = T R TOT f b
where T f is the frame duration and Rb is the total air-interface bit rate. The frame efficiency is thus, boh η f = 1 b The major existing TDMA cellular system is GSM. However, since we will devote an entire chapter to a detailed study of GSM, let s instead cite the US Digital Cellular (USDC) here as an example of a TDMA system, in order to get a picture practically of the level of overheads required and the performance of such a cellular network. USDC was introduced in the early 1990s, at the same time as GSM, but was built to coexists with the AMPS network and so allow network providers a path to update their systems. USDC used the same 30 khz channels as in AMPS, but now voice data was digitised. The modulation technique adopted was π/4-dqpsk with a symbol rate of 24.3 kbaud, resulting in an air bit rate of 48.6 kbps. To fit this signal into the allocated bandwidth, Nyquist square-root raised cosine pulse shaping was used with a roll-off factor of 0.35. The USDC frame lasted for 40 ms, and contained 1944 bits of data, divided into 6 timeslots of 324 bits each. A full-rate voice channel would involve a user being assigned two timeslots within a frame, say timeslots 1 and 4. Half-rate voice channels were also possible. The structure of the timeslot is shown in the diagram below for both the forward and reverse channels. For the forward link time slot has 28 synchronisation bits followed by 12 bits of the Slow Associated Control Channel (SACCH). The SACCH is used primarily for power control messages. The 260 data bits per timeslot are divided into two 130 bit segments and placed around the 12 bits of the Coded Digital Verification Colour Code (CDVCC). The CDVCC plays a role a bit like an address and must be echoed back to the base station to verify that the mobile is receiving messages from the correct base station. There are twelve unused bits at the end of the timeslot. The reverse channel time slot has 6 guard bits followed by 6 bits to allow the mobile unit transmitter to ramp up. The 260 data bits in a time slot are broken into three segments and interspersed with the 28 synchronisation bits, SACCH and CDVCC bits as shown. Of the 260 data bits per timeslot, 101 are used to error detection and correction, so there are only 159 data bits per timeslot. This corresponds to a frame efficiency of η USDC f = 159 324 TOT 49.1% The full-rate voice channel would thus correspond to a user being allocated 2 159 = 318 bits per frame for data, and with a frame duration of 40 ms this meant the voice data rate was 7.95 kbps. This was exactly the data rate out of the vocoder. The advantages of a TDMA system are as follows. Firstly, the fact that data transmission is achieved in regular bursts takes advantage of the natural idleness in speech, and generally lowers the demands on battery power. TDMA also facilitates MAHO (Mobile Assisted Hand-Off), since the MS can monitor the signal strength received from other base station during the timeslots when it is idle. Offsetting the forward and reverse timeslots does away with the need for the duplexer, a
considerable saving in cost. Finally, TDMA offers the prospect of bandwidth on demand, as potentially a user could be assigned more timeslots per frame to increase the data rate available to a high-demand user. The first obvious disadvantage of a TDMA system is the need for synchronisation across users on the network, as any loss in synchronisation will result in overlapping timeslots and inter-user interference (IUI). As demonstrated in the above example from the USDC, these systems require a considerable amount of overhead information to achieve reasonable performance and synchronisation, greatly lowering the effective capacity. A major disadvantage of these systems is that the air data rate of these systems tends to be high, meaning that the air data rate is comparable to the channel coherence bandwidth, and as such equalisation is needed at the receiver. This is the case in GSM, where the air data rate is 270.833 kbps over a 200 khz channel. The capacity of a TDMA system could be expressed as m( Bsys 2Bguard ) N = Bchan where m is the number of users that can be accommodated per channel, and N is number of simultaneous users that can be supported on the system. Having discussed the basic principles of FDMA and TDMA we ll now turn our attention to Spread Spectrum Communications. The are generically two basic types of Spread Spectrum Communication: Frequency-Hopped Multiple Access (FHMA) and Direct Sequence Spread Spectrum, which is usually referred to as CDMA (Code Division Multiple Access). Spread Spectrum Communication was developed during the Second World War as a technique for providing clandestine communication that is resistant to jamming attacks. The basic principle is to take a signal of a certain bandwidth and use a certain signature sequence (or code) to spread this information over a much larger bandwidth. The ratio of the bandwidth of the spread signal to the original signal is known as the spreading factor, which is typically in the order of a hundred to a million, depending
on the application. The spreading signal must be independent of the information signal and known at the receiver, which must use it to recover the information, possibly after some fairly complex synchronisation operations. The point is that the signal power is now spread out over a much wider bandwidth, lowering the power within any particular frequency band. At the extreme the signal can appear below the ambient noise floor, and so remain essentially undetectable to eavesdroppers. The first and simplest form of Spread Spectrum Communication to be developed was Frequency Hopping (FHSS). It is a simple idea of randomly hopping the carrier frequency around some wide spectral region. As long as the receiver knows the hopping sequence it can track and recover the signal. The form of a FHSS transmitter and receiver are illustrated in the diagrams below. FHSS systems are not common in mobile cellular networks, but have been used in some wireless standards, such as Bluetooth. The key element in a FHSS is the Pseudo-random noise (PN) sequence generator, which generates a binary sequence that appears random unless you know the structure for generating the sequence. This is often implemented as binary shift register, of length n, which when excited with a certain bit sequence will produce a seemingly random bit sequence of period 2 n 1. An example is shown in the diagram below, of a linear shift register system that generates a maximal length sequence depending on its initial state. The PN sequence drives a frequency synthesiser that changes its carrier frequency in response to the bit sequence, hopping the narrowband signal to different carrier frequencies across the spectrum.
There are two types of FHSS systems. A fast frequency hopped system means that the period of a hop, T hop, is less than the symbol duration, T s. This provides good frequency diversity. The other case is known as slow FHSS, in which the signal will remain at one carrier frequency for a duration of at least several symbols.
Not that frequency hopping can be done with any form of modulation, ASK, PSK, or FSK. The key performance metric of a FHSS is the probability of a collision, by which we mean that two different users will randomly hop onto the same carrier frequency at the same time, resulting in the interference of their signals. Given the bandwidth occupied B sys, the number of users K, and the signal bandwidth B signal, this could be approximated as Bsignal p 1 1 coll Bsys As the number of users on the system increases, the probability of a collision increasing and the system performance degrades. There is no absolute limit to the number of users that a FHSS can accommodate, like there is for TDMA or FDMA systems. Instead the performance of a FHSS system drops as more users are added, so it is a question of what level of collisions the system can tolerate. This is then known a soft-capacity limit. K 1 Code Division Multiple Access (CDMA) IS-95 CDMA, and the third generation networks, W-CDMA and cdma2000, use Direct Sequence Spread Spectrum (DSSS) to achieve multiple access. DSSS systems, which we will henceforth refer to as CDMA, achieve the spreading of the user s signal across the wide bandwidth in a very different way to frequency hopping. The basic structure of a CDMA transmitter and receiver are shown in the diagram below. To generate a CDMA signal, the users data is first encoded using a polar line code, where the bit duration is T b. Raised-cosine pulses are typically used to control the spectral roll-off but we ll ignore their effect here. The Spreading Code generator creates PN sequence, a duration T c, where T c < Tb (and commonly T b is a multiple of T c ). The data rate of the spread sequence is known as the chip rate, to distinguish from the data rate of the information signal. The Spreading code signal is a serial stream of the PN sequence as a polar NRZ line code. This is multiplied to the data
signal to produce the spread signal. The resultant signal thus has the same effective bit period as the spreading code signal, T c. The basic idea then is that the bandwidth of a signal is inversely proportional to its bit period. Thus, through the application of the spreading sequence the bandwidth occupied is increased by a factor, known as the processing gain, of Tb Rc Bcode G = = = Tc Rdata Bdata where R is known as the chip rate and R is the data rate of the information signal. c data The data signal could then be recovered at the receiver by multiplying the received signal by the same spreading code, assuming that the receiver has been able to synchronise its version of the code to that of the transmitter. Including noise, the signal received at the transmitter could be written as r() t = Am( t) c( t) cos ( 2πf c t) + n( t) where m () t is the information signal, c ( t) is the spreading code signal, and n () t is the noise, which we will usually consider to be AWGN. The signal to noise ratio, the ratio of the signal power to noise power we could write as A 2 2 SNRspread = N0Bcode where N 0 is the noise power spectral density, assumed flat, in Watts/Hz. Stripping off the carrier and multiplying by the code, we obtain A 2 r () t = m() t c () t + ni ()() t c t 2
The baseband equivalent noise and the spreading code should be uncorrelated, so this apparent noise should still be uniformly distributed over the wider bandwidth B code. 2 Since the code is Polar NRZ, c ( t) 1(it is a binary sequence that is either 1 or - 1 ). We can then pass this received signal through the baseband channel filter of bandwidth B data to recover the message, and so the signal to noise ratio at the output is A 2 2 SNRout = = G SNRspread N 0Bdata It is enhanced by that factor, the processing gain, G. The signal to noise ratio on the channel can be very low but still we can be able to reconstruct the signal. The issue then is how to accommodate multiple users on the radio channel. The key idea here is to assign each orthogonal codes. Two codes are orthogonal if they are uncorrelated. In other words, two codes are orthogonal if when we multiply them together, given any relative shift, the result sums to zero. That is, R C ( ) = 1( ) 2 ( ) 0 1 C τ 2 c t c t τ dt for two different codes, c 1 () t and c 2 ( t), and for some relative time shift τ. If my received signal contains the information of both spread users, r() t = Am1 () t c1 () t cos( 2πf t) + Am2 ( t τ ) c2 ( t τ ) cos( 2πf ( t τ )) n( t) c c +
the signal obtained when I de-spread for user 1, removing the carrier, multiplying by the user 1 spreading code, and then summing using my matched filter, one obtains T A A b r () t = m () t ( f ) c ( ) c ( ) d n 1 + cos 2π c τ 1 λ 2 λ τ λ + ( t) 2 2 0
and so if the above orthogonality condition is met then there is no interference from the signal of user 2. Note that if we can somehow synchronise the transmission of the two users the orthogonality condition is somewhat easier. The unknown time offset τ can be set to zero and we merely require that, T b c 0 () t c () t dt 0 1 2 = This could be achieved for the forward link, where the BTS is transmitting all signals to each user, so can synchronise all these transmissions in time. A popular choice of codes in this case is the Walsh codes, found as rows of the Hadamard matricies, 1 1 H N H N H 2 = and H 2 N = 1 1 H N H N Any two rows of a matrix from this set are orthogonal. The Walsh codes for order 4 are shown in the diagram below. Notice orthogonality refers to the two signals agreeing in one half of their bit positions, and disagreeing for the other half. Thinking of our matched filter receiver our bit decisions are made by integrating over the bit interval, so the contribution from the terms of two orthogonal sequences will sum to zero. When synchronisation cannot be ensured, such as in the uplink of a cellular mobile network, Walsh codes cannot be used. In this situation we usually use pseudo-random (PN) sequences. A well -formed PN sequence should look like it is equally probable that the value of any bit will be 1 or 0, hence two un-related PN sequences will be uncorrelated, regardless of their relative orientation in time. CDMA is likewise a soft-capacity system. The only limitation on how many users can be put on the channel is the number of orthogonal codes that can be generated, which at a naïve level, is unlimited. However, when more users are added to the system the performance degrades. In DSSS the effect of another user s signal is to appear as noise for the user of interest. We the more users we add onto the network the higher the noise level at the output of each receiver. When there are many users on the system the principle component of noise at the output of the receivers comes from other users. This means that the signal power from
other users ultimately determines the noise floor. This makes power control of prime importance in CDMA, and our previous discussion of the near-far effect means that the network tries to ensure that all users have the same received power level at the base station. Recall that the near-far effect refers to the signal from high power MS, particularly those near to the base station, introducing a too-high noise level at the detection of a weaker signal of a far away mobile station. Assuming all MS are transmitting at the same power, let s then consider the signal to noise and interference ratio at the output of a receiver for some user. As illustrated in the diagram below, the signal power from the other K 1 users on the network will be spread over the wider bandwidth, B code, and considering that their signals are orthogonal and agree and disagree in half of the chip positions, they will appear as random to noise at the output of the detector. The fraction of interfering signal power at the output of the receiver, after the low pass filter of bandwidth B data, is B B = G. The signal to noise and interference ratio is then, data code 1 Ps SNR1 SINR = = Ps SNR1 N 0Bcode + ( K 1) 1+ ( K 1) G G where SNR 1 is the signal to noise ratio at the output of the receiver when there is only one user on the system, and so no interfering users. This clearly shows the degradation of performance as more and more users are added to the network. The soft-capacity limit is certainly one attractive feature of CDMA, along with the flexibility of channel cell assignment. Fundamentally channels are merely codes that
can be easily distributed to different cells in the network. CDMA networks are able to implement soft handovers, where the channels can move cell with the mobile, greatly reducing the level of signalling overheads associated with maintaining mobility in the network. By the very nature of spreading the bandwidth of the transmitted signal is much larger than the channel coherence bandwidth. This makes the signals robust to small-scale fading, but, on the other hand, the multipath delay spread is much larger than chip period, so equalisation at the receiver is required. By carefully designing the spreading codes to have low autocorrelation, a clever equaliser receiver called a RAKE receiver can be used. A RAKE receiver is able to identify these multi-path components and combine them to obtain a stronger output signal. We will study more of the details and finer points of CDMA when we study the systems IS-95, W-CDMA, and cdma2000 in later chapters. Contention-based Multiple Access The multiple access techniques we have thus far studied, FDMA, TDMA, FHSS, and CDMA, are all based on the idea from circuit-switched systems that a user is granted a channel for the entire duration of the communication link, and the user has exclusive access to this channel for the duration. We ll now briefly describe multiple-access methodologies that are suitable for packet-switched systems. In mobile networks these multiple access methodologies are the domain of the paging and random access control channels. A mobile unit, when placing a call or responding to a page, sends to the base station a packet of data containing identification and authorisation information, and perhaps a destination phone number. These packets are sent only during call set-up and so are infrequent. It is wasteful to assign a channel on a exclusive long term basis for a user to send such infrequent short bursts of data. The common strategy to enable a large number of users to transmit and receive packets of data in an infrequent and uncoordinated manner is to have these users share a single common communication channel. User data is presented as packets, with each packet having a source and destination address so it can be delivered correctly. These schemes are commonly encountered in computer networks, both wire-line and wireless, though our interest in this course is in their use on the paging and random access control channels in cellular networks. The major issue in these common channels is to establish the protocol, or order, by which these random and uncoordinated users can use the channel to transmit their packets. The basic idea is that only one user can transmit its packet on the channel at any one instant, so how to decide who uses the channel when? In a sense all these users contend for use of this common channel, hence the name contention-based multiple access. The original contention based multiple access methodology is called ALOHA the name coming from the place it was first developed, the University of Hawaii. The protocol is trivial to conceive. Whenever a user is ready to transmit a packet of data it does so and sets a timer running. If a collision occurs, and two or more users attempt
to transmit data packets simultaneously corrupting transmission, no acknowledgement will be received. When the timer runs out the user then waits a random amount of time before attempting transmission again. The throughput of these ALOHA schemes, essentially the amount of traffic that such a channel can pass, is determined by establishing the probability that a collision occurs. Assume all packets have the same duration, τ. As shown in the diagram below a collision will occur if another packet is generated within a vulnerable period of 2τ, extending one packet duration before our desired packet. Assuming that packets are generated with a Poisson distribution with average arrival rate λ, n { } ( λτ ) λτ Pr n packets in τ = e n! we find the probability of no collision to be the same as the probability that no packets are generated during the vulnerable period, 2 Pr no collision = e { } λτ The throughput of the system is the average offered traffic multiplied by the probability of successful transmission, after identifying the offered traffic as 2 T = R Pr success = R e { } R R = λτ, The throughput of an ALOHA system as a function of offered traffic is sketched below. At low R throughput is low due to under-utilisation of the channel, while at high R throughout is low due to excessive collisions. The maximum throughput is achieved for an offered traffic of 0.5 Erlang, and is only 0.184 Erlang. This means that at best an ALOHA channel can successfully pass data less than19% of the time. This is makes this ALOHA scheme fairly inefficient.
The throughput can be increased with Slotted ALOHA, where time is divided into timeslots and each user can only begin transmitting at the beginning of a timeslot. This reduces the vulnerable period to only τ, as only packets generated in the previous timeslot can disrupt our packet. This doubles the effective maximum throughput of such a channel to 0.368 Erlang. This slotted ALOHA protocol is used for the paging and random access channels in GSM, USDC, IS-95 CDMA, W-CDMA, and cdma2000. Still more advanced than ALOHA systems are Carrier-Sense Multiple Access (CSMA). These feature users listening to the channel to establish if it is in use and only transmitting if the channel is vacant. There are many ways this can be done, and variants of these CSMA protocols are used in the Ethernet, Wifi, and wireless computer networks in general.