TECHNIQUES TO INCREASE THE CAPACITY OF A DS-CDMA SYSTEM
David Bozward and R.L. Brewster
The capacity or maximum number of users within a Direct Sequence Code Division Multiple Access (DS-CDMA) system is determined by the amount of noise or interference each user contributes to the overall system. DS-CDMA system design is thus the expertise of producing a spectrum that provides the smallest amount of interference to the other users. The paper first introduces two distinct system architectures; Cellular and Private Mobile Radio (PMR), and discusses how their system requirements differ. The paper also introduces the parameters by which DS-CDMA systems are measured, comparing the requirements of a classical CDMA system with those of the cellular and PMR multi-user schemes. The paper contains a discussion of which appropriate system value should be used to gauge the gain provided by the various capacity increasing techniques. The paper continues with an investigation of the various techniques that have been proposed for the cellular environment and the complexities concerned with their efficient usage. The paper provides an analysis of the gains provided against overall system complexity and other implementation considerations. The paper is concluded with a summary of the further methods that require research and the benefits they may provide.
Spread Spectrum (SS) was first used in the 1950s by the military, who have continued to be its main user to date. This has led to Spread Spectrum techniques taking a low profile due to its distinct military advantages that are based on a robust immunity to interference and jamming. The major non-military use to date has been for space communications because of its ability to provide better immunity against noise and interference than any other modulation technique. The recent upsurge in open discussions of SS techniques occurred when the cellular telephone authorities and service providers looked at better ways of utilising their overcrowded spectrum during the design of the second and third generations of cellular telephone systems. Spread Spectrum systems have advantages and disadvantages over the existing systems, but as the system requirements change so do the advantages. Recent advances have shown promising uses in fibre optics [1] as well as in the cellular and private mobile radio schemes.
This paper is organised as follows: In Section 2, the architectures used within the mobile environment are introduced. In section 3, the parameters which establish the scheme are investigated, these include the processing gain and the capacity of a CDMA channel. Section 4 introduces the techniques that are used to increase the capacity of the scheme.
When considering the capacity equation of a CDMA system, we must first look at the network topology used. Within the radio environment there are only two under development, these being:-
Cellular Where all mobile stations communicate with one or more central fixed base stations. The calls are then trunked to the appropriate base station for communications with the second user.
Private Mobile Radio Where all users within the localised mobile network (net) may communicate with each other over either one channel or a small number of channels.
Figure 1.
Mobile Architecture
These networks are illustrated in Figure 1, A being the cellular case and B the PMR. From these, it can be seen that there are distinct differences between the two topologies. The cellular system may provide cheap, effective spectral usage to a mass market who require to communicate over large mobile areas, using inexpensive equipment. PMR provides a single administrated channel to a licensed set of customers who can afford to pay for this privilege. Communications are confined to a designated area within this pre-defined user group. These two environments pose separate degrees of sophistication to the system designer and network provider. Cellular systems are provided for large numbers of mobile users and thus a larger degree of sophistication can be incorporated into the handset which may then be made in large production environment. PMR is usually provided for small user groups on an individual basis and thus additional amounts of sophistication are directly financed by the service provider and user.
The transition between military communications that is primarily concerned with the prevention of jamming to the commercial systems that is almost solely interested in higher user density for a set bandwidth has been a slow one. The classical system parameter has been the processing gain and from this the jamming margin has been determined, attention has now shifted towards the capacity of the scheme that provides a direct indication of system user density per cell.
The figure of merit usually used in spread spectrum systems is the processing gain [2,3]. This gain is a readily estimated quantity if the bandwidth employed in a system is known and the information rate is available. In spread spectrum processors the process gain available may be estimated by the approximation:-
Gp=B/R (1)
where the RF bandwidth is that of the transmitted spread spectrum signal and the information rate is the data rate in the information baseband channel. This processing gain provides an indication of the number of simultaneous users the system can sustain within a given error rate, this relationship is given in the capacity equation.
The capacity of a cellular CDMA system is bounded due to the base station having to communicate with every active mobile user in its cell. The base station then has an interference limit at which a number of mobile stations may communicate within a given adequate performance signal. Here we introduce a single radio cell capacity equation [4,5] to provide an indication of the parameters that dictate typical system performance. The network to be considered is configured in a star, i.e. each user communicates with the central cell base station. Each user of the CDMA system occupies the entire allocated spectrum, employing a direct sequence Spread Spectrum waveform. The basic equation for capacity in terms of the number of users in the system can be written as:-
(2)
Where is the bit energy to noise spectral density ratio required for adequate performance, Gp is the processing gain and is the background noise-to-signal power ratio. This equation assumes perfect power control and that the cross-correlation function will not significantly degrade the Bit Error Rate (BER). In a large multi-user system the performance is predominantly determined by the noise produced by other users and, as such, the background noise term can be neglected. This equation can be expanded to cover additional system features such as Voice Activity Detection (D), Frequency Reuse (F) and Sectorisation (G) gains.
(3)
The equation in this form shows that to increase the capacity of the system, we must either decrease the Adequate Performance Signal required or increase the system gains. This has classically been done by increasing the processing gain of the system, but as indicated above there are other methods that may be used.
In section 3.2 we discussed the capacity of a CDMA system in terms of N, the number of active users per cell. The measure taken to determine or compare systems can be one of many:-
The classical method with SS systems has been to use the processing gain and jamming margins as the system value, this is because the primary aim has been their use in military schemes. The capacity equation shown calculates the 'number of user' calls simultaneously available for a given adequate performance signal. If we multiply N by the user information rate, the system capacity in terms of bits-per-second is calculated. Within the commercial telecommunications industry the measure used to state the amount of calls simultaneously available has been the Erlang. It is also widely used in the calculation of capacity for conventional cellular systems, and provides a direct measure that may be used within the telecommunication networks. One Erlang is the traffic intensity in a channel that is continuously occupied and thus can be derived from the previous equation.
There are a number of ways in which the capacity, however measured, may be increased in a DS-CDMA system. All the methods discussed here reduce the perceived user interference by one means or another.
This is provided by two means within a digital communication system [6,7]. Firstly by interleaving, which forces the errors to be in the correct statistical groupings and secondly, by Forward Error Correction (FEC). Using FEC the transmission error rate can be reduced to a desired level, without increasing the transmission power above the value for which the capacity was calculated. The coding gain of the system is the difference between the required to achieve a specified BER without coding and that required to achieve the same with coding. This increases the complexity and bandwidth in terms of bit rate utilised by the system.
Performance of the error correction codes is closely related to the noise characteristics of the channel. These characteristics range from error rate to the statistical relationship between errors, and as a basis can be split into a bursty or random error channel. Thus FEC codes are chosen on the channel characteristics over which they are to be used. The predominant factor for error generation in a CDMA channel is the presence of the other users' contributions to the noise floor, unlike other channels that are primarily Gaussian in nature. This contribution may be assumed to be Gaussian when bit interleaving and FEC are employed.
Power Control is important in a multi-mobile station environment to keep user interference within the cell and neighbouring cells to a minimum. Each communicating pair controls their transmitter power for optimum propagation and BER characteristics. Perfect power control is assumed in the capacity equation and as such, if this is not implemented, is capable of reducing the capacity of the system. Experience has shown that this may require a dynamic range of control of the order of 80 dB.
In a typical full-duplex two-way voice conversation [4,8], the duty cycle of each voice is approximately 35% to 40%. If the mobile station does not transmit during these periods then the received interference to other stations will be reduced. In practice this may be slightly less since code synchronisation and background comfort noise are sent over the air interface. This provides a capacity increase through a VAD gain of approximately 2.5, plus a reduction in the transmitter average power by the same factor. Depending upon the system and the type of information to be conveyed, this factor may change. As an example, a reduction in system performance would occur when security measures are taken to ensure traffic security which requires a constant data stream to be conveyed. Modern Codecs are capable of speech transmission using channel rates as low as 8 kbps, tolerating an error rate up to 0.1%. These are likely to reduce with further research and rates of 3.8 kbps have been reported.
The application of frequency reuse and sectorisation in the first generation of analogue mobile telephones have been the major advancement in cellular technology. A gain is achieved by splitting a cell into sectors, reducing the total user interference per sector and thus increasing the cell capacity by the number of divisions. In practice this figure is reduced by around 15% due to overlap and sidelobe anomalies within the antenna array of the cell. This means a cell divided into three has a sectorisation gain G of 2.55. In the first generation of mobile cellular systems there are two types of division, three proposed by Ericsson and six proposed by Motorola. Each had their own benefits [9], the trunking efficiency is better in the three sector systems due a smaller division in system resources, catering for 27% more mobile stations. The six sector system provides an adaptive correction to traffic loads due to antenna overlap and a constant system gain relative to the sector antenna pointing direction.
The concept of frequency reuse in the first generation of analogue mobile telephones was used to control the amount of co-channel interference to an acceptable limit using spatial separation while increasing system capacity. In a CDMA system the frequency reuse factor F is defined as the amount of interference received in a cell from stations operating in neighbouring cells. This contribution to noise from all neighbouring cells has been calculated [5] to be equal to approximately 65% of the interference due to the mobile stations operating within the cell. The received interference is determined by the inverse fourth-power law relationship between signal strength and distance. This is approximated most closely at some distance from the transmitter rather than near to it, and in turn this means that in small cells the carrier-to-interference ratio deteriorates, as does the frequency reuse factor F. This relationship at a distance is determined by the propagation terrain and the multiple reflected paths available to the Rake receiver.
The central base station that is receiving all communications within the cell can cancel out individual mobile transmissions since it knows their PN codes and all transmissions are on the same frequency. This may be done in the forward and reverse directions at the cell base station. In the forward direction, (cell to mobile) the base station may subtract its own transmitted signal from that received to provide a signal that only contains those transmissions from other stations. In the reverse direction, (mobile to cell) once communications have begun, each known user's signal may be subtracted from the incoming transmissions to leave only the one required for demodulation. This could be extended in the base station network to include the neighbouring base station transmissions, since this accounts for 36% of the interference contributed. The second and greater tiers of cells only contribute less than 2% and would not provide a gain required for implementation. To a limited extent this could be done by the mobile station since processing power here is limited by the battery life and size of the unit. The mobile unit could cancel out the pilot, synchronisation, paging and access channels that are known to it from the base stations it currently monitors.
When a mobile station is communicating within a cellular environment, the wanted signal from the base station may be considered to be in front and the majority of unwanted signals are radiated from other directions. This is indicated in a frequency reuse factor of 0.65. An advantage would be gained by using a directional antenna at the mobile station to reduce received user interference. These could be provided using an adaptive antenna array, [10,11] that would track in real time the movement of the mobile in-relation to the base station. These arrays may also provide such features as nulling, in order to cancel out high powered near interference signals and unwanted base stations. Adaptive processing is already used to a limited extent in CDMA systems with the Rake receiver design [5] which provides a multiple path receiver to combat multi-path fading effects. The introduction of more complicated processing would be slow in the mobile unit since both size and battery capacity are important marketing points in the cellular industry. Simple processing-constrained algorithms that will provide a positive gain for the mobile user, (in terms of the perceived noise) and for the service provider, (in terms of additional capacity) are likely to be the first applied to these systems. There are problems with physically accommodating in a hand portable a practical antenna design with the frequencies currently under use, but a vehicle mounted device could provide the starting point required for this technology.
In an asynchronous DS-CDMA network the received signal-to-noise ratio is determined by the noise contributed by each user in the system and the cross-correlation properties of the spreading codes. As indicated in section 3.2, to increase the capacity we may either decrease the required adequate performance signal or increase the system gains. The improvement in the adequate performance signal is achieved by increasing the amount of FEC available to the system, and by deterring modulation schemes that require smaller energy-per-bit per noise bandwidth for a given BER. The techniques discussed in this paper reduce the perceived noise in the system. This is done either by reducing the transmitted interference, e.g. VAD and Power Control or by reducing the interference to the receiver, e.g. Error Control, Sectorisation and Interference Cancellation.
CDMA is ideally placed to utilise the benefits available with adaptive noise reducing techniques. These provide benefits in the user perceived noise and increased user densities. Since the base station is interference limited when considering capacity, it would seem to be the easiest and most cost effective place to implement adaptive processing. There will be a trade-off between acceptable outlay costs for equipment and higher capacities. Techniques that reduce the correlation function seem to be an area that has been overlooked and require further research.
The authors would like to express their thanks to Roke Manor Research for their financial support.
[1] Parham M.J., Smythe C., Weiss B.L. "Code Division Multiple-Access Technquues for use in Optical-Fibre Local Area Networks" Electronics & Communication Engineering Journal pp203-212 Aug 1992
[2] Dixon R.C. Spread Spectrum Systems John Wiley & Sons 1976
[3] Pickholtz R.L., Schilling D.L., Milstein L.B. "Theory of Spread Spectrum Communications - A Tutorial" IEEE Transactions on Communications Vol.30 No.5 pp855-884 May-82
[4] Gilhousen K.S., Jacobs I.M., Padovani R., Viterbi A.J., Weaver L. A., Wheatley C. E. "On the Capacity of a Cellular CDMA System" IEEE Transactions on Vehicular Technology Vol.40 No.2 pp303-312 May-91
[5] Qualcomm An Overview of the Application of Code Division Multiple Access (CDMA) to Digital Cellular Systems and Personal Cellular Networks May-92
[6] Berlekamp E.R. "The Technology of Error-Correcting Codes" Proceedings of the IEEE Vol. 68, No 5 pp564-592 May-80
[7] Wiggert D. Codes for Error Control and Synchronization Artech House 1988
[8] Pickholtz R.L, Milstein L.B., Schilling D.L. "Spread Spectrum for Mobile Communications" IEEE Transactions on Vehicular Technology Vol. 40, No.2 May 1991
[9] Graziano V. "Increasing Analog Capacity" Motorola Inc 1989
[10] Torrieri D.J. Principles of Secure Communication Systems 2nd Ed.Artech House 1992
[11] Widrow B., Stearns S.D. Adaptive Signal Processing Prentice-Hall 1985