AACHENER BEITRAGE ZUR INFORMATIK

Transcription

1

2 AACHENER BEITRAGE ZUR INFORMATIK Herausgeber der Reihe: Prof. Dr. rer. nat. Klaus Indermark, Lehrstuhl fur Informatik II Prof. Dr.-Ing. Manfred NagI, Lehrstuhl fur Informatik III Prof. Dr. rer. nat. Otto Spaniol, Lehrstuhl fur Informatik IV Spaniol Otto - Slavik, Jan - Drobnik, Oswald (Hrsg.): Proceedings of the 2nd Workshop on Personal Wireless Communications (Wireless Local Access), Frankfurt am Main (Gerrnany), December 10-11, 1996, 1 Auflage Aachen Vetlag der Augustmus Buchhandlung, 1996 (Aachener Beitrage zur Informatik, Band 19) ISBN X 1996 bei den Autoren der Beitrage Verlag der Augustinus Buchhandlung PontstraBe Aachen Tel. & Fax: Druck: Druckerei Paletti, Aachen Gedruckt auf chlorfrei gebleichtem Papier

3 Analysis of a new hybrid prma type channel access scheme with frequency hopping for cellular mobile systems Jćzef Woźniak Abstract Faculty of Electronics, Telecomunications & lnformatics Technical University of GDAŃSK, POLAND Narutowicza 11/12 (Street) GDAŃSK, POLAND tel.: fax: jowoz@sunrisepg.gda.pl This paper focuses on characteristic features and advantages of a new Packed Reservation Multiple Access (PRMA) type algorithm combined with the Frequency Hopping (FH) technique. The PRMA type protocols are intensively investigated as possible channel access schemes for the next generation cellular and personal communication systems. The paper presents example simulation results obtained for the PRMA and PRMAJFH algorithms used in channels with selective frequency in fadings. Variations of the packet dropping probability versus the number of voice/data terminals for different station and network parameters are presented. 1 Introduction In discussions on the next generation cellular mobile systems different channel access schemes are considered. They range from TDMA based solutions to different CDMA with DS (Direct Sequence) or FH (Frequency Hopping) techniques. The success of Qualcomm's cellular systems shows that the CDMA based channel access algorithms can become very attractive alternatives for TDMA oriented methods. In the paper we investigate the behaviour of an efficient PRMA type protocol combined with the Frequency Hopping strategy, proposed for the data and voice transmissions. Similarly to the GSM system we define a set of narrow-band frequency channels FCH, with a certain number of logical channels, obtained according to the TDMA principle. The high frequency channels, used for mobile communications, are considered to be channels of rather poor quality. In traditional narrow-band communication a channel sustaining an adequate signal-to-noise ratio is initially selected for communication. When transmission condition, namely, propagation or interference conditions change, a new channel must be selected. In the case of PRMA (Packet Reservation Multiple Access) schemes active mobile stations must reserve one out of the total number of F logical channels, defined within each of narrowband channels FCH, where F corresponds to the TOMA frame length equivalent to the number of logical channels. According to the PRMA schemes a station that has lost the reservation due to longer silence periods, interferences or fadings have to content for another time slot, usually at the same frequency channel (FCH) [1], [2], [3].

4 In the paper we consider a modified PRMA scheme that combines features of both PRMA and the Frequency Hopping techniques. FH has been proposed as a technique to cope with frequency selective channel fadings. In the new PRMAJFH algorithm, a mobile terminal loosing a connection (channel reservation) with a base station selects a new frequency channel (FCH) and starts a reservation procedure for a new time slot in the frame. Throughout the rest of this paper we describe the network model and show the basic principles of stations' operations. Then we present example simulation results obtained for the PRMA/FH algorithm and its PRMA origin. 2 Network Model We model the network as a set of cells, each one of them being a network with the base station playing a role of an intelligent digital switch. The total channel bandwidth is divided into a certain number N of frequency channels (FCH). Each of these channels is slotted and a slot duration is equal to the voice packet transmission time. The slots are grouped into frames of F consecutive slots. Usually the frame duration is calculated with respect to the average transmission rate of voice packets during the active period. During this period the voice terminals generate a stream of packets, called a talkspurt. The speech activity is modelled by a two-state "talk" and "silence" discrete-time Markov chain [1], [2], [3], illustrated in Fig. 1. Figure 1 The speech activity model A terminal generates one packet per frame, each time it enters the active - "talk" - state. If a terminal is switched to the "silence" state it does not generate packets. This period corresponds to a gap in speech and is detected by a Speech Activity Detector (SAD). On the other hand, when a station switches to the "talk" state SAD packetizes the voice and stores these packets in a special buffer until the connection is estabilished again. The procedures performed by stations allow to utilize efficiently the medium by a number of voice and data terminals. In the case of data packets no advanced reservation has to be made, for they are being sent as single/separate messages. After a silence period a voice terminal has to conted for the channel (slot) once more. It is worth to be mentioned that a partial lost of voice packets is tolerable by human perception and droppings up to 1 of voice packets do not cause any significat degradation in speech comprehension. When TDMA based channel access protocols are designed it is necessary to evaluate the frame length in order to avoid frequent silence gaps causing frequent disconnections of terminals and droppings of voice packets. Example measurements show that the speech activity factor (the part of a connection time the station remains in "talk" state) is only 0.43.

5 Additionally we assume that the uplink channels, i.e., channels from terminals to the base stations are channels with selective frequency fadings, that can be modelled by the Gilbert channel model presented in Fig. 2, with "good" and "bad" transmission states. Figure 2 Two state Gilbert HF channel model We also assume that there either is, or is not a crosscorrelation between different frequency channels. In the first case there is no interdepedence between separate FCH (the state transitions, in separate FCH channels, are independent) while in the second case there is a relatively high probability that fadings will occur in two or more FCHs, at the same time. Frequency hopping has been proposed as a technique to cope with multipath fadings and interference problems in high frequency (HF) channels. In the paper we utilise the concept of selecting frequency channels (FCH) out of a set of available narrowband channels in order to improve the PRMA protocol performance. The narrow-band channels can be characterised by different signal - to - noise ratios. In some of the HF channels transmitted signals may be subject to severe distortions, whereas other channels exhibit favourable conditions. Therefore, in our simulation studies we model a narrow-band channel behaviour using the Gilbert channel model, very often applied to systems with time varying channels, in particular VHF channels for mobile communications [4]. In Gilbert models we define "bad" and "good' channel states. The first (a "bad" state) state describes the situation when channel fadings or signal interferences occur, while the second one (a "good" state) corresponds with good transmission conditions. 3 PRMA Algorithm with FH In the paper we investigate the behaviour of PRMA protocols [1], [2], [3]. According to the PRMA algorithms a1l slots constituting one frame are classified as free-available or busyreserved. Information on slot states is always broadcasted by the base station and is known to all terminals at the beginning of a new frame. The voice terminal entering the " talk" state starts to contend for an available slot according 10 the S-ALOHA procedure [5]. A voice terminal that has a reserved slot in a frame transmits a packet over that time slot. This means that the first successful transmission in a slot reserves the same slot in all subsequent frames for the exclusive use by the station (see Fig.3), until the station moves to "silence" state detected by SAD, or the connection is lost due to channel fadings. In our PRMA/FH procedure we shall assume that when the station looses its reservation because of fadings it must contend for slot reservation in a different or FCH.

6 Figure 3 Illustration of operation of PRMA protocols When the reservation is lost due to the station switching to the "silence' state-detected by SAD - the next reservation attempt will be undertaken after moving, once again to the. 'talk" state. However, this attempt can be arranged in the same FCH, as the previously occupied one. All terminals that have packets to transmit and have no reservation must contend for free slots (in one out of accessible FCHs), undertaking their actions with a certain probability P, treated as a system parameter. This so called retransmission probability is identical for all cell terminals. In the case of data packets no advanced reservation has to be made; they are sent as single, separate messages. The service of both types of traffic in one system significantly influences both the dropping probability of voice packets and 'the transmission delay of data packets. In the case of an excessive number of data packets offered to the common channel, a certain degradation of the quality of service of voice packets can be observed. The probability of dropping of voice packets can go beyond the usually accepted level of In the paper we study the cellular mobile system that allows to extend the set of accessible TDMA channels, implementing the FH strategy. 4 Example results In this Section we discuss the behaviour of the PRMA and PRMAJFH protocols for different channels and station parameters. In Figures 4-7 we present example results obtained by simulation. In all simulation experiments we have accepted the warm-up period equal to 1000 frames The statistical analysis of output data for steady-state simulation requires elimination of initial nonstationarity caused by the existance of the initial transient period. In order to limit the influence of this period on the final results and to make simulation running as short as possible, simulations were initialised with a randomly chosen number of active stations. Ali simulation results were verified accepting the confidence level of We have considered the efficiency

7 and flexibility of operation of both protocols. Figure 4 Variations of the packet dropping probability vs. the number of voice terminal s for PRMA and different retransmission probabilities Before we start to investigate the behaviour of the PRMAJFH algorithm we shall shortly analyse some properties of the original PRMA scheme. When designing any channel access strategy we are interested in the optimum transmission and retransmission conditions. In majority of papers, devoted to the analysis of PRMA schemes, it is assumed that the highest protocol effectiveness (the biggest value of the channel throughput) is usually achieved for the retransmission probability P equal to In Figure 4 (a, b) we present simulation results, showing the influence of the probability P on the packet dropping probability. From these results it follows that the probability value P=0.35, obtained for F=48 (6 independent FCH channels with 8 logical channels F per each FCH) really ensures very high system effectiveness. For this reason, in the following, we shall accept this probability value as the optimum one. Under this assumption we shall investigate the system utilisation for both PRMA and PRMAJFH algorithms. Figure 5 depicts variations of the average utilisation of one logical channel for PRMA and for different frame lengths F. We can observe that for relatively long TDMA frames the channel utilisation saturates.

8 Figure 5 Load per slot VS. the number of TDMA slots On the other hand the diagrarns presented in Figure 6a show variations of channel utilisation versus the number of active voice terminals for different numbers of logical channels (numbers oftime slots F in one TDMA frarne). The maximum values ofthe channels utilisation, for different frame lengths are almost the same. However, these values are obtained for different numbers of active terminais (and without taking into account any specific value of the packet dropping probability). Assuming the packet dropping probability of 0.01 we obtain the maximum numbers of active voice terminal s that can effectively utilise one time slot in a frame. We can observe a growing effectiveness of the PRMA system with the increase of the number of slots F (See Fig. 6b). In the case of the original PRMA scheme we do not consider FH procedures, i. e., we do not implement selection of a new FCH

9 Figure 6 Packet dropping probability vs. the number of terminals for PRMA and different frame lengths All the above results were obtained assuming noise1ess transmission conditions. The only cause of errors and packet retransmissions were packet interferences. Now, we shall accept that apart from interferences, also selective frequency fadings can occur. In the presence of fadings the transmission quality can drop down dramatiea1ly. Moreover, when fadings occur relatively often, it may turn out that the usually required level of the packed dropping probability ofo.01 cannot be satisfied. Assuming that transmissions, initialised at one frequency channel (FCH) must be continued at the same band and using one out of F logical channels, it becomes evident that in the presence of selective fadings the system will behave badly; transmission quality in the ease of fadings will be very poor. Therefore, in order to improve the transmission quality when the fadings occur, a new FCH will be selected. Figure 7 Packet dropping probability vs. the number of terminals for different values of the cross-correlation between FCH From the results presented in Figure 7 it becomes evident that assuming fading effects in FCH the PRMA algorithm without FH cannot be used. In order to combat fadings and interference problems in HF channels the FH technique must be implemented. Studying the behaviour of our mobile cellular system with the PRMA!FH scheme we

10 assumed: the number of narrow-band channels (FCH) in one celi - 6, the number of slots per frame in one FCH - 8, the frame length ms, the average fading interval ms (10 frames), the probability of channel fading occurence , the voice packet dropping probability , the average talkspurt duration 1000 ms, the average silence period duration ms, the retransmition probability The results presented in Fig. 7 show that for zero value of the crosscorelation coefficient, between separate narrow-band FCH channels, the PRMA/FH strategy with 6*8 logical channels, avaliable in one cell, allows for aproximately 95 simultaneous connections between mobile users. For the probability of fadings occurrence equal to 0.01, and for zero cross-correlation between different FCH channels, the PRMA/FH can ensure entirely normal system operations, while the PRMA scheme cannot satisfy the requirement of less than 1 of droppings of the voice packets (see Fig. 7a). Accepting slightly different model of channel statistical interactions, namely the existence of cross-correlation between separate FCH channels the advantages of the PRMA/FH algorithm becomes less evident. For the cross-correlation function value of 0.5 we obtain the simulation results showing a certain degradation of the system quality (see Fig. 7b). 5 Conclusions In the paper a modified version of the PRMA protocol has been proposed and verified by simulation. We have investigated the behaviour of the PRMA type protocol as well as some properties of a cellular mobile system assuming: selective frequency fadings in HF radio channels and implementation of the FH technique in procedures of selection of new logical channels. The simulation results show that the proposed PRMA/FH algorithm ensures proper system operation even when the fadings cause the total transmission break in one frequency channel (FCH). Possibility of switching from one FCH to another improves the flexibility of stations and the whole system operation. Accepting 0.01 probability of fadings and zero crosscorrelation between the FCH channels we obtain in fact no degradation of the system quality. The probability of droppings of voice packets under these conditions, does not change significantly. Contrary to these results one can see that in the case of the pure PRMA algorithm, i.e., without switching to another FCH, the packet dropping probability becornes in the case of selective fadings occurrence high er than the commonly accepted threshold of O. O l. 6 References [I] Meracos., Shrirang J.: "Voice Packet Losses and Integration in Reservation Random Access Protocols for Wireless Accsee Systems with Call Control". Proc. IEEE GLOBECOM 1992, pp

11 [2] Rubin l., Shambayati S.: "Performance Evaluation of Reservation Random Access Scheme for Packetized Wireless System with Cali Control". Proc. IEEE GLOBECOM 1992, pp [3] Paterakis M., Cleary A.c.: "On the voice Data Integration in Thrid Generation Wireless Access Communication Networks". European Transactions on Telecommun. Vol. 5, 1994, pp [4] Zander 1., Malmgren G.: "Adaptive Frequency Hopping in HF Communications". lee Proc. Commun. vol. 142, [5] Abramson N.: "The Throughput of Packet Broadcasting Channels". IEEE Trans. Commun., voi.com-25, 1977, pp.1l7-128.