Future Mobile Telecommunications Systems

Future Mobile Telecommunications Systems Peter F. Driessen Department of Electrical and Computer Engineering University of Victoria Victoria, B.C., Canada [email protected] ABSTRACT: In this paper, we review 2nd and The second generation systems such as IS-136 TDMA 3rd generation wireless telephony systems, both in (Time-Division Multiple Access) with CDPD (Cellular terms of technology and service concepts. We point Digital Packet Data), IS-95 CDMA (code-division multi- out some of the technical and service issues that ple access) and GSM (Global System for Mobile)/PCS1900 may aect the functionality and performance from (Personal Communications System) use digital modula- the user point of view. In particular, we focus on tion technology (i.e. voice is digitized using a vocoder one technical and one service concept important for and transmitted as data), provide short messaging service long term success of 3rd generation systems: trans- (SMS), low speed data, call management features in ad- mit diversity to increase capacity (quantity),and dition to voice service, and are marketed to consumers as context-sensitive applications to increase user ac- well as business users. These second generation systems ceptance (quality). are currently being deployed, and have been in service for 1. Introduction only the last few years. In some areas, the rst and second generation systems compete with each other. Wireless telephony and the Internet are both expand- There are also `low-tier' second generation systems with ing rapidly, and nding new users around the world. Fu- limited coverage intended for pedestrian-only (no vehicu- ture mobile telecommunication systems will converge with lar) use. Such systems include PHS (Personal Handyphone the Internet and corporate Intranets to provide integrated System), CT-2 (Cordless Telephone 2), DECT (Digital Eu- voice/data/video (multimedia) `anywhere, anytime' com- ropean Cordless Telephone), and provide primarily voice munications services to people on the move. The market service to consumers. demand for these services is huge, and presents both op- portunities and challenges. The third generation system IMT-2000 (International Mobile Telephone - 2000) is planned to be a single world- The challenge is two-fold: to provide genuinely useful wide standard or group of standards while accommodat- information access and management services which are as ing a variety of terminal types (voice/data/video), using easy to use as a telephone (i.e. quality), and to provide the more advanced digital modulation technologies, provides system capacity to send enough bits per second per Hertz an integrated information access and management service, per unit area to accommodate the peak and average data and will be marketed to specic industries, as well as con- rates needed to support these services (i.e. quantity). In sumers and business users. The third generation standards this paper, we review two key ideas to meet these chal- are currently under development under the auspices of the lenges: context-sensitive services for quality, and transmit ITU-R (International Telecommunications Union - Radio). antenna diversity for quantity. This paper is organized as follows. In Section 2, we dis- Wireless telephone systems may be broadly divided into cuss some technologies which address the ever increasing three generations. The rst generation systems such as need for larger system capacity to serve both high data NMT (Nordic Mobile Telephone) and AMPS (Advanced rates and many simultaneous users, and focus on the con- Mobile Phone Service) use analog FM (Frequency Modu- cept of transmit diversity. In Section 3, we review some lation) technology, provide only voice communications ser- of the service concepts made possible by this capacity, and vice, and were originally marketed to business users. These focus on the need for context-sensitive applications. systems are still widely used today, and can accommodate limited data communications using the audio channel and a modied telephone-type modem. 1  2. Technology issues - need for high ca- and no scatterers [4]. Roughly speaking, the wide spacing pacity replicates the eect of scatterers which create images and 2.1. Third generation system technologies thus serves to spread out the apparent source of the sig- nals over a wider angular range. In this way, capacities on The ITU-R conducts workshops to receive and dis- the order of Clin = nlog2(1 + ) bps/Hz can be obtained cuss various proposals for design approaches to IMT-2000. for LOS as well as Rayleigh channels when n transmit and Here we review some proposals presented in Toronto in n receive antennas are used. In contrast, when the trans- September 1997 [1][19]. The FRAMES Multiple Access mit antennas are closely spaced, the number of degrees of (FMA) concept combines CDMA and TDMA techniques freedom on a LOS channel degenerate, resulting in capac- into one harmonized platform also harmonized with GSM. ities of only Clog = log2(1 + n). For other recent work on Thus to build multi-mode terminals, GSM and FMA clock transmit antenna diversity see [5], [6][9] and [7]. rates should be related, and carriers should be placed on 2.3. Capacity expressions a common 200 KHz frequency grid. FMA is proposed in three forms: FMA1 (TDMA without spreading and The fundamental result for the capacity in bps/Hz of a TMDA/CDMA with spreading) with bandwidth 1.6 MHz wireless system with nT transmit antennas and nR receive and FMA2 (wideband CDMA) with 5 MHz bandwidth, antennas with an average received SNR  (independent of and both Frequency Division Duplex (FDD) and Time nT ) at each receive antenna was obtained in [2] as Division Duplex (TDD) operation. TDD is included to adapt to asymmetric services (dierent rates on uplink and C = log2(det[InR + (=nT )HH y ]) (1) downlink) Multiple user data rates are supported in all three forms by multislot, multicode, and variable spread- where the normalized channel matrix H contains complex ing/multicode respectively. The link can be adapted to scalars with unity average power loss, and H y is the com- changing service requirements, interference, channel state plex conjugate transpose of H . The capacity is expressed and system loading. in bps/Hz in the narrowband limit with no frequency de- There are 4 dierent concepts for wideband CDMA, but pendence. Normalization is achieved by dividing out the free space power loss and setting the parameter  to the many of the key parameters have been agreed, e.g. basic desired SNR 1 . This result assumes that H is unknown to chip rate of 4.096 Mchips/s, 10 ms frame length, coherent the receiver but nR and  are known [2][3]. The signal from detection with pilot symbols, and variable spreading for each antenna is dierent. The transmitted data has been low and medium bit rates, multicode for high bit rates. split into substreams which are separately independently FMA2 includes packet mode operation where small infre- coded and modulated, and cycled among the antennas to quent packets are attached to a random access request, ensure that over time each substream experiences a similar whereas large and frequent packets are sent using a dedi- propagation environment. cated channel. FMA2 also proposes support for multi-user detection, interference cancellation and adaptive antenna We consider an environment with only LOS propagation arrays for both the uplink and downlink receivers. and T and R arrays of nT = nR = n antennas. Here we 2.2. Transmit diversity designate the base and subscriber ends of the link as T and R, respectively, but reciprocity applies, and all subsequent The rst premise of this paper is that one method that results apply to both the downlink (base-to-subscriber) can signicantly increase system capacity, but which may and the uplink (subscriber-to-base). not have been considered in the working papers for [1] By replacing HH y by its singular value decomposition is the idea of using multiple transmit antennas (trans- U U y where  is a diagonal matrix of eigenvalues and mit diversity). For xed bandwidth and total radiated the columns of U are the corresponding unit-magnitude power, with Rayleigh fading, a wireless system with multi- eigenvectors of HH y, we can show ple antennas at both transmitter and receiver has a chan- nel capacity which grows linearly (rather than logarith- Yn mically) with the number of antennas [2]. This result, C = log2 ( (1 + (=nT )i ) (2) which assumes separate information is sent out of each i=1 transmit antenna, holds true even though the transmitter Thus C is maximized when all eigenvalues i are equal does not know the complex channel transfer characteris- (since H is normalized, the eigenvalues will all be of mag- tic. By explicitly spreading out the antennas well beyond nitude 1). a wavelength, such capacities can be achieved not only on Rayleigh channels with many scatterers, but also on deter- 1 This avoids the need to compute absolute propagation loss and ministic channels with direct line-of-sight (LOS) paths only then set the transmitted power to obtain the desired SNR.  If the matrix elements Hij are random variables, then C transformation, the capacity is also a random variable. In this case, we dene an outage det[W W y + (=nT )W HH yW y ] threshold x (say 0.01), and dene Cx to be that capacity C = log2 (5) for which P robfC > Cx g = 1 , x. For a Rayleigh chan- det[W W y] nel model, we dene HRayleigh as a matrix of normalized If W is square p and unitary, then C is unaected. By choos- (unit magnitude) uncorrelated complex gaussian variates. ing W = nU1 where U1 = U but with k < n columns, If H = HRayleigh , then from [3], in the limit of large n and then the capacity is given by (2) with sum to k instead of large , the capacity lower bound approaches n. Thus if the eigenvalues beyond k are negligible, then we can use W to reduce the processing complexity, while C = nlog2(=e) (3) still retaining most of the capacity. 2.5. Practical implications where e = 2.718.... This result is close to the maximum capacity Clin , and thus implies that there is a high prob- Capacities approaching Clin = nlog2 (1 + ) can be ability that a particular realization of HRayleigh has close achieved in a deterministic LOS (non-Rayleigh) environ- to n signicant (non-negligible) eigenvalues. This result is ment. This is achieved by spreading out the elements of as expected, since in the Rayleigh model HRayleigh using T either explicitly (by placing one element of T at each independent complex gaussian variates, it is implicit that of n sites), or implicitly (by adding re ectors which create signals arrive from many directions 2 i.e. the apparent images of T ). This result suggests that in the absence of transmitter locations are widely spaced. re ectors, the T array can be explictly spread apart, thus duplicating the eect of images. To nd the maximum capacity which may be realized in practice in a LOS environment (no Rayleigh fading), we The increase in capacity resulting from transmit diver- seek to construct geometric arrangements of the T and R sity results in a signicant benet for services requiring arrays such that all i in (2) are non-negligible. high speed data. To illustrate, we consider a 3-element transmit and re- 3. Service issues - need for sensitivity to ceive array, where the R array is linear with interelement context spacing yr = =2, and the T array comprises 3 elements 3.1. Market Demands for services evenly spaced around a circle (e.g. 120 degree sector an- `Anyplace, anytime communications with anyone' is a tennas at 3 separate cell sites in a conventional hexagonal common expression of the service concept for future mobile cell layout, all facing inward [10]. For the R array at the communications systems. However, it may be very dicult center of the cell, to anticipate exactly what kind of services the marketplace 0 e,j 1 1 ej1 1 will demand and accept [16]. Service concepts proposed for HLOS = @ 1 1 1 A (4) IMT-2000 may be divided into two broad classes: exten- ej1 1 e ,j1 sions and renement of existing second generation services, and entirely new services. can be used to evaluate the capacity at this point. From The primary existing service of second generation sys- a histogram of the capacity for all positions in the cell, tems is plain old (wireless) telephony, which may be ex- assuming a path loss exponent of either 2 or 4, we nd that tended with call management features such as call forward- for most locations, the capacity is near Clin = nlog2 (1 + ing, voice mail, call waiting, 3-way calling and automatic ) for n = 3, so that the benet of transmit diversity is call handling according to user preferences depending on realized. the caller ID. Another existing service is two-way text mes- 2.4. Reduced complexity implementation saging (short message service), an extension of one-way paging towards wireless email service. To realize the high capacity from transmit diversity re- New services and capabilities for IMT-2000 include [18] quires the processing of signals from a n-element signal provision for new audio, video and data services, including vector, as outlined in [3]. To reduce the complexity, the packet data and multimedia services, high service qual- received signal may be premultiplied by a transformation ity and integrity comparable to the xed (wired) network, matrix W having fewer rows than columns, thus reduc- bandwidth on demand to support data rates ranging from ing the length of the signal vector to be processed. After low speed for paging to high speed for video or le trans- 2 For a given interelement spacing at R, see [8] for the angular fer, support for asymmetrical data with high speed in one spread of signals from T required in order to to achieve zero correla- direction and low speed in the other, security and ease of tion between R elements. operation.  Wireless multimedia services are based on some kind of things' without requiring too much data transfer between one or two way video, high speed internet access and web the network and the wireless access device. browsing capability. These service concepts arise from the A simple example of a context-sensitive service is the convergence of broadband access and wireless technologies. automatic conguration, upon entry, of the user's current However, genuinely useful service concepts using these ca- environment (e.g. temperature, background music) to ac- pabilities remains to be specied and tested in the market- commodate user preferences [11], i.e. to create a personal place. space, in hotel rooms or shared oces. User demands will inevitably drive the evolution of new An instructive example of a context-sensitive service is service concepts. The key demand will be for immediate, a tourist (or real estate) application on a PDA, where the integrated information access and management, which re- tourist is provided with text information about buildings sults from the information explosion, mostly available on and other landmarks in the city depending on his/her cur- corporate intranets as well as the Internet. Since the in- rent location and orientation [15]. This example may be formation exists, users will expect it to be available `any- implemented using a software form of Post-it notes rep- where, anytime', and thus will demand that their portable resented in Standard Generic Markup Language (SGML) wireless device has the same power as their oce desktop. 3 . using tags such as <at>(location), <with> (person or Users will demand multiple types of wireless access devices, object), <during> (time), <facing> (orientation). The ranging from a small cellphone to a large-screen laptop. software Post-it note consists of the context (tag) plus a A convenient user interface for all device types is critical body (text), and a collection of such notes comprises a to the evolution and acceptance of new service concepts. complete document. Notes are triggered dynamically de- The challenge is to develop user interfaces which can oer pending on the current context. A related idea is ConTexts the simplicity of a desk telephone and fax machine and (Conversational Texts) [14] in which a complete text is dy- still meet the more complex needs of integrated informa- namically assembled from a collection of `chunks' which tion access and management. Web browsing using search are triggered depending on the context (parameters) cho- engines is too complicated for mobile users. sen by the reader (e.g. expertise level, amount of detail, language). Each chunk (equivalent to the software Post- It is useful to be reminded of service concepts that were it note) is tagged with the relevant parameters. In both not successful, such as the videophone [16] or that had cases, information (text chunks or a complete document) is only limited success, such as the personal digital assistant accessed as appropriate depending on the desired context. (PDA). In the rst case, the service did not meet a real need and in fact was considered an invasion of privacy by Ideally, using a Context-sensitive information access and some. In the second case, the PDA functionality and user management service which is aware of the user's current interface have not been suciently complete and conve- state will be like having a secretary/butler available at all nient to provide the intended 'assistant' service. times to respond to the user's current as well as anticipated information needs, as well as to prompt/remind the user 3.2. Context-sensitive service concepts of tasks to be done or items requiring a user decision. The second premise of this paper is that a key factor 4. Summary to the success and acceptability of a new wireless service In this paper, we have pointed out one technical and concept is that it be context- sensitive, i.e. aware of the one service concept thought to be useful for 3rd genera- circumstances (state) of the user, and sensitive to the antic- tion wireless systems, but which may not be referred to in ipated needs when in this state. Perhaps the best context- current IMT-2000 documents. Transmit diversity may be sensitive service is that provided by a human secretary or used to signicantly increase the capacity (bits per second butler backed up with instant access to the desktop com- per unit bandwidth) of wireless systems, thus enabling the puter. A practical (non-human) approximation may be high data rates needed for integrated information access provided by autonomous agents [13][12] working with a and management and desktop functionality while mobile. wireless access device. However, a complete and conve- Context-sensitive applications may be used to signicantly nient user interface for the device is essential. An interface ease the task of information access and management. which is sucient for the desktop (e.g. type-point-click), may not be practical for a portable device. Context-sensitive services may be particularly eective for group projects where there is awareness of the state of all users. Such services may also help to reduce the de- 3 HTML, Hypertext Markup Language used to represent web mand for capacity on the wireless system by `taking care of pages is one instantiation of SGML.  Acknowledgements [10] F. Borgonovo, A. Acampora, `Capture division multi- The author thanks G.J. Foschini for discussions of ple access: a new cellular access architecture for future [2][3], R.L. Kirlin for suggesting the transformation W , S. PCN's', IEEE Comm. Mag., Vol.34, No.9, September Catreux for deriving (5) and the gures in Section 2, and 1996, p. 154-162. m.c. schraefel for the key ideas in Section 3. [11] Bill Gates, The Road Ahead, Microsoft Press 1996. [12] K. Decker, A. Pannu, K. Sycara, M. Williamson, "De- References signing behaviours for information agents", Proceed- ings of the First International Conference on Au- tonomous Agents, Marina del Ray, CA, February [1] www.itu.int/imt/ 1997, ACM Press, 197, pp. 404-412. [2] G.J.Foschini, M.J. Gans, `On limits of wireless com- [13] M.N. Huhns, M.P Singh, "Conversational Agents", munication in a fading environment when using mul- IEEE Internet Computing Magazine, March-April tiple antennas', Wireless Personal Communications, 1997, pp. 73-75. accepted for publication. [14] m.c. schraefel, "ConTexts, Intensional Document Cre- [3] G.J.Foschini, `Layered space-time architecture for ation, Delivery and Retrieval", Proc. IEEE Pacic wireless communication in a fading environment when Rim Conference on Cmmunications, Computers and using multi-element antennas', Bell Labs Technical Signal Processing, Victoria, BC August 1997, pp. 417- Journal, accepted for publication. 419. [4] P.F. Driessen, G.J. Foschini, `Deploying multiple [15] P.J. Brown, J.D. Bovey, X. Chen, "Context-Aware transmit antennas to greatly enhance capacity in wire- Applications: From the Laboratory to the Market- less systems', IEEE Trans. Comm., under review. place", IEEE Personal Communications Magazine, Vol. 4, No. 5, October 1997, pp. 58-64. [5] V. Weerackody, `Characteristics of a simulated fast [16] R.W. Lucky, "New Communications Services - What fading indoor radio Channel," Proceedings 43rd IEEE Does Society Want?", Proc. IEEE, Vol. 85, No. 10, Vehicular Technology Conference, pp. 231-235, Secau- October 1997, pp. 1536- 1543. cus, NJ, May 1993. [17] M. Sha, et al, "Wireless Communications in the [6] V. Weerackody, `Diversity for the Direct-Sequence Twenty-First Century: A Perspective, Proc. IEEE, Spread Spectrum System Using Multiple Transmit Vol. 85, No. 10, October 1997, pp. 1622-1638. Antennas', Proceedings IEEE International Confer- ence on Communications, pp. 1775-1779. Geneva, [18] K. Buchanan et al, "IMT-2000: Service Provider's Switzerland, May 1993 Perspective", IEEE Personal Communications Maga- zine, Vol. 4, No. 4, August 1997, pp. 8-13. [7] N. Seshadri and J. H. Winters, Two Signalling [19] R.D. Carsello et al, "IMT-2000 Standards: Radio Schemes for Improving the Error Performance of Aspects" IEEE Personal Communications Magazine, Frequency-Division-Duplex (FDD) Transmission Sys- Vol. 4, No. 4, August 1997, pp. 30-40. tems Using transmitter Antenna Diversity, Interna- tional Journal of Wireless Information Networks, Vol. 1, No. 1, 1994, pp 49-60. [8] J.Salz, J.H.Winters, \Eect of fading correlation on adaptive arrays in digital mobile radio", IEEE Trans. Vehic. Tech., vol. 43, no. 4, pp. 1049-1057, November 1994. [9] S. Ariyavisitakul, T. E. Darcie, L. J. Greenstein, M. R. Phillips and N. K. Shankaranarayanan ` Performance of simulcast wireless techniques for Personal Commu- nication Systems', IEEE Sel. Areas Commun., Vol. 14, No. 4, May 1996, pp. 632-643.