Deployment of Multi-layer TDMA Cellular Network with Distributed Coverage for Traffic Capacity Enhancement Jérôme Brouet, Patrick Charrière, Vinod Kumar* Armelle Wautier, Jacques Antoine** *Alcatel, Corporate Research Center, Radio Communications Dpt., 5 rue Noël Pons, 92734 Nanterre, France **Ecole Supérieure d'electricité, Plateau du Moulon, 91192 Gif sur Yvette, France E-mail: Armelle.Wautier@supelec.fr, jerome.brouet@telspace.alcatel.fr ABSTRACT In second generation TDMA cellular systems, traffic increase is achieved via cell size reduction. But, to handle moving mobile stations (MS) and ensure coverage on the whole served area, a 2-layer network has to be deployed and radio resource management techniques such as directed retry (DR) are implemented for spectrum efficiency. Optimally, fast (respectively, slow) moving MS should be served by the macrocell (respectively microcell) layer. So, in such network organization, traffic performance is sensitive to both coverage quality and accuracy of estimation of MS speed. Conventional microcell layer deployment suffers from coverage holes and uncertainty of speed estimate (making impossible to exploit full capacity of such networks). This paper proposes to replace the usual layer of microcells by a distributed coverage. Strategies for serving base transceiver station (BTS) selection and channel allocation dependant on MS position over the service area, are implemented. Both the communication quality and the accuracy of estimation of MS speed are improved. This makes DR more efficient and enhances the overall traffic performance. I. INTRODUCTION The success of personal mobile communications systems such as GSM900/DCS1800/PCS1900 has lead to a constantly increasing need for capacity. Since the allocated spectrum is limited, various means to enhance the achievable spectrum efficiency have to be progressively introduced. Some of these techniques are: Increase the number of carriers in congested cells. Improve management of co-channel interference to enable the introduction of more compact cellular reuse patterns. Cell splitting and introduction of microcells. Introduction of 2-layer hierarchical networks with mechanisms to handle DR. A) Increase of the Number of Radio Carriers: With a traditional macrocellular network the capacity increase is first of all obtained by increasing the number of carrier frequencies per BTS. This is the most straightforward and inexpensive method. However the achievable capacity enhancement is clearly limited by the allocated spectrum. A capacity gain of few percents is obtainable by using more compact frequency reuse patterns. The subsequent radio degradation in transmission quality can be compensated by implementing interference control techniques such as: Power control (PC). Discontinuous transmission (DTx). Slow frequency hopping (SFH). B) Cell Size Reduction: For further capacity enhancement, the next step is to decrease the size of the cells by splitting and inserting new transmitting sites in the networks. Also the replacement of omni-directional cell sites by sectorized cells and the corresponding adjustment of frequency reuse patterns can bring some capacity enhancement. In a macrocellular network, cell size reduction is however quickly limited by co-channel interference and the traffic gain remains therefore limited. The only way to obtain a real capacity boost is then to deploy superposed layers of micro-cells and macro-cells. Careful segregation of available spectrum between the layers is also essential. C) Traditional Two-Layer Network Architecture: The micro-bts transmit antennas are usually located a few meters above the ground level. The radiated power is low and therefore the signal remains confined in the vicinity of the microbts (a typical cell radius is a few hundreds of meters as compared to 1km for a macrobts). With such an architecture the channel allocation is made using the information on the MS mobility. The microcell capacity is dedicated to the slowest mobile stations whilst the macrocell is used by the faster ones (Fig. 1).
Cell defined by the the same beacon message active relay Concentrator Fig. 1: Two layer cellular network (DR on) Such a solution has many advantages: The microcellular layer does not have to be continuous since wide area coverage is provided by macrocellular layer. No major change has to be done on the macro layer apart from reconsidering the frequency planning. Geographical variations in traffic densities can be handled with good overall spectrum efficiency in the network. However, in a practical context, reduction of microcell size is limited by the considerations related to interference management and handover. Since the transmit power has to be kept sufficiently high to provide non line of sight coverage (around the street corners), the subsequent interference for the other cells in line of sight is high. Handover tuning problems are due to very small power difference for beacon signals coming from two adjacent cells. The high level of interference further increases the probability of too late / wrong handover decision. Furthermore, as the average number of handovers also increases when cell sizes drop, the probability of call loss / ping-pong / interference increases sharply, leading to very poor grades of service. Fig. 2: Distributed coverage cell B) Optimized Spectrum Utilization: to BSC/MSC Improved interference management is feasible by combining the advantages of distributed coverage [1] to those of intelligent power delivery [3]. After an initial access phase during which all the relays of the cell could catch the access bursts transmitted by an MS, the call is transfered on a traffic channel. On this channel, only a subset of relays is kept active for the communication (Fig. 3). These are typically the two or three server relays the nearest to the MS. This procedure substantially reduces the interference level on the air. Moreover, if slow frequency hoppping is also implemented, it has been shown in [4] that there is a fractionnal loading value for which a frequency reuse patterns as low as 1 (unity) is achievable for the traffic channels in dense urban areas. Very high spectrum efficiency is thus achieved. II. A NOVEL DISTRIBUTED COVERAGE ARCHITECTURE FOR TRAFFIC ENHANCEMENT Zone covered by the two relays left active A) System Architecture: In such a system, radio coverage is provided by a set of synchronized transmitters, also called relays. Multiple relays constituting a cell broadcast a common beacon message on a common channel. Each set of relays is seen as a normal microcell by the mobile station (Fig. 2). In the case of a communication, each relay is capable, on the downlink, of synchronously retransmitting a copy of the information coming from a concentrator (multicast). On the uplink the concentrator collects the information coming from each relay listening to the MS for further processing (macrodiversity). inactive relay Concentrator to BSC/MSC Fig. 3: Distributed coverage and efficient power delivery
C) Mobility Management: During a communication, the best server selection is periodically updated [3]. With such a system, the traditional handover based on downlink measurement reports is replaced by an automatic channel transfer (ACT) [2]. Decisions for ACT are based on the uplink measurements of the signal transmitted by the MS and received by multiple relays-including the present servers and the potential candidate server relays. D) Advantages Compared to the Conventional Microcellular Solution: Several advantages are offered by this solution: First of all spectrum efficiency is optimized and good communication quality is maintained by the use of multicast technique on the downlink, macrodiversity for the uplink channel, and slow frequency hopping. Coverage is improved through multisite illumination. Frequency planning is no longer necessary for the traffic channels. This is a very big step forward for implementing dynamic radio resource management. Such an architecture applies to both indoor and outdoor applications. The estimation of MS speed is much faster and more accurate than with a traditional microcellular architecture. Moreover, the estimation is performed within the cell whilst in the other case it is performed at the boundary of the cell. The probability of a successful handover towards the macrocellular layer or towards a neighboring distributed cell is improved. E) Multi-layer Networks with Distributed Coverage in the Microcellular Layer: As previously mentioned, the hierarchical networks are organized such that a continuous wide area coverage is provided by the macrocells and the hot-spot traffic is handled by local micro-cells. For the present case, the micro-cellular layer is implemented with distributed coverage. All the calls are initiated on the micro-cellular layer. After a first phase during an "active communication", the displacement speed of the MS is estimated and the fast moving MS are re-directed towards the macrocellular layer. The normal procedure is to assign two relays to each communication and to update the assignment as the MS moves within the micro-cellular coverage. The coverage with two relays however, may not always be possible. Also, a conventional HO is performed for the MSs moving between different cells. In order to maximize traffic efficiency, redirection of the call to the macrocellular layer can be performed. Two criteria for redirection are analyzed: With the first criterion (C1), the call is handed-off from the micro layer to the macro layer when the number of ACT, performed for that call is equal to a threshold value Nact. With the second criterion (C2), the call is handed-off if there is not enough available resource to provide coverage with two relays. C1 is mainly based on a MS speed. Calls can be re-directed even if not absolutely essential - say due to reasons of traffic load on the relays. The ACT mechanism are applicable to fast moving MSs. On the contrary, C2 exploits full ability of ACT and waits for a possible congestion problem on the channel allocated to the MS before performing redirection. Redirection from macro to micro is not considered. B) Description of Simulation tool: The re-direction strategies are evaluated through simulations. The simulator is both event and discrete time driven to account for statistical behavior of call arrivals/terminations and MS mobility according to deterministic patterns. The environment is assumed to be a Manhattan-like service area with inter-street distance of 100m. The lower layer consists of 9 microcells (i.e. 9 concentrators) with 9 relays each (Fig. 4). Each relay can provide an equivalent of one TRx or 8 Traffic time slots. Each concentrator is capable of managing the equivalent of Nseq traffic TRx. The umbrella macrocell covers all the service area and has traffic TRx. III. IMPACT OF CHANNEL REDIRECTION CRITERIA The focus is now on the interaction between the traffic on the micro and macro-cellular layers. Since only the re-directed traffic is served by the macrocellular layer, the optimization of the DR strategy is therefore essential to the overall system performance. A) Re-Direction Strategies: At the call set-up, the incoming calls are allocated to the microcellular layer. Several possible resource allocation/reservation strategies are described in [2]. Fig. 4: Serving area (9 microcells + 1 macrocell) The propagation model is based on line-of-sight (LOS) and nonline-of-sight (NLOS) models [2] regularly used to characterize micro-cellular environments. The transmit power levels and inter-relay distances are adjusted such that every available time-slot on a relay provides an
interference-free traffic channel [4]. The best server selection is assumed to be error free and performed every second for each active MS. A seamless macrocellular coverage is assumed to be available in the service area. Only voice traffic is considered. Call arrivals follow a Poisson distribution while call durations follow a negative exponential distribution. An MS is assumed to generate an average load of 1 call per hour lasting 120s (33 merl/user). The simulated population is a mix of fixed and moving MS (70% of users are fixed). For each simulation run, all moving MS have same speed. The speed is expressed in average number of zones, h, that a MS crosses during a communication (a zone can be defined as the area in between two adjacent relays). Furthermore, at each street corner a moving MS is assumed to have an equal probability to go straight on, turn left or right. 180 turns are not admissible. Pb=1% Pfct=0,1% 1 2 3 4 Nact Fig. 5: C1: traffic per relay at Pb=1% and Pfct=0.1% (=3/Nseq=4/h=3) The offered traffic per relay is simulated for a grade of service (GoS) of 2% (target value for outdoor cellular systems). The GoS is a function of both the blocking probability, Pb, and call drop rate, Pfct. Pb=1% Pfct=0,1% GoS= Pb + 10. Pfct (1) A call is assumed blocked when no physical resource is available for an incoming call: the relays in the range of sensitivity and the macrocell are fully loaded. Call drops may be caused by any failure of procedures related to ACT, HO or DR. The call drop events must be kept less than 0.1% to meet the GoS requirements. In the following, the offered traffic is the traffic figure obtained for Pb 1% and Pfct 0.1%. C) Results: Criteria C1 and C2 are analyzed as a function of: - Nseq, -, - h. System performance with DR criteria C1 is directly linked to Nact. The effect of Nact on the traffic per relay is shown in Fig. 5. If Nact is set to 1, the macro-layer is overloaded with MSs that could be handled by the lower layer. For Nact 3, calls are lost before DR can be applied. Optimal traffic figure (1.55 Erl/relay) is obtained for Nact=2. Indeed, MS speed, h, has an impact on the offered traffic. Fig. 6 depicts the evolution of the traffic per relay with the MS speed. A 40% reduction in achievable traffic capacity per relay is observed with h=15 (MS speed = 45 km/h) compared to h=3 (9 km/h). 3 5 7 9 11 13 15 h Fig. 6: C1: Effect of speed on offered load (Pb=1% and Pfct=0.1%) (Nseq=4/=3/Nact=2) Besides, has to be adjusted to maximize the offered traffic on the lower layer. Simulations indicate that optimal value depends on Nseq (Fig. 7). With the architecture described in III.B, optimal values for are: =Nseq for Nseq 4, =4 for Nseq 4. Nseq=2 Nseq=3 Nseq=4 Nseq=5 0 1 2 3 4 5 Fig. 7: C1: traffic per relay vs. (Nact=2/h=3)
If criteria C2 is used for DR, the maximal offered traffic per relay is a function of Nseq and. With =2, optimal traffic per relay (1.6 Erl/relay) is observed (Fig. 8) whatever Nseq may be. 1,8 1,6 1,4 1,2 1 0,8 0,6 0,4 Nseq=3 Nseq=4 Nseq=5 0 1 2 3 4 Fig. 8: C2: Offered traffic per relay vs. (h=3) MS speed has also an impact on the offered traffic per relay. However, if is correctly adjusted, the system performance are stable. If =3, for h=15, a 6% loss is observed compared to h=3 (Fig. 9). And, if is set to 4, no loss is experienced for the MS speed ranging from 0 to 60 km/h (Fig. 10). IV. CONCLUSION An original multi-layer cellular system with distributed coverage and intelligent power delivery on the lower layer has been analyzed focusing on the performance of DR strategies. This new architecture enhances the traffic capacity through secured DR compared to traditional multi-layer systems. Two DR criteria suitable for such system architecture have been proposed. These cover the range of application environment that can be expected in dense micro-cellular areas. The first one gives better traffic capacity but is sensitive to MS speed. Such DR strategy could apply in ultra dense pedestrian areas. System performance stability versus MS speed is obtained with second DR criterion. Such strategy should be implemented in dense downtown areas. REFERENCES [1] S.Aryavisitakul et al., "Performance of simulcast wireless technique for personal communication systems", IEEE JSAC, vol. 14, No 4, pp.632-643, May 1996. [2] A.Wautier et al., "Performance of a distributed coverage SFH-TDMA system with mobility management in a high traffic density network", to appear in PIMRC'98, Sept. 1998. [3] P.Charrière, J.Brouet, "Optimum channel selection strategies for mobility management in high traffic TDMA based networks with distributed coverage", ICPWC'97, Dec. 1997. [4] Y.Bégassat, V.Kumar, "Interference analysis of an original TDMA-based high-density cellular radio network", VTC'98, May 98. =1 =3 3 9 15 20 h Fig. 9: C2: Impact of speed (Nseq=4) 1,8 1,6 1,4 1,2 1 0,8 0,6 0,4 0,2 0 h=3 h=20 0 1 2 3 4 Fig. 10: C2: vs. traffic vs. speed (Nseq=4)