The system performance, including the blocking rate and dropping rate of regular calls and the successful access rate of emergency calls, has been studied ...
Proc. Natl. Sci. Counc. ROC(A) Vol. 23, No. 3, 1999. pp. 389-395
(Scientific Note)
Performance Evaluation of Priority Access Procedure over TDMA System J ENN -K AIE LAIN
AND
J YH -H ORNG WEN †
Institute of Electric Engineering National Chung Cheng University Chiayi, Taiwan, R.O.C. (Received March 11, 1998; Accepted September 25, 1998) ABSTRACT Emergency service has become an important application in telecommunications today. System resources can be shared with higher priority given to the emergency users by using a practical priority access procedure though ordinary traffic will suffer some impact. In this paper, a priority access procedure using one of three different priority access algorithms is designed based on the architecture of the Personal Access Communications System (PACS) with consideration given to hardware implementation. The system performance, including the blocking rate and dropping rate of regular calls and the successful access rate of emergency calls, has been studied using a simulation system constructed with the Block Oriented Network Simulator (BONeS). Simulation results show that the proposed priority access procedure using one of the three studied algorithms is suitable for implementation in the PACS system. Key Words: priority access procedure, time division multiple access, cellular system, Personal Access Communications System
I. Introduction Emergency service is an important application in telecommunications today due to the growing number of mobile originating emergency calls (Silventoinen and Rantalainen, 1996). Urgent service should be supported no matter how heavy the traffic load is. In such systems, emergency users have access to system resources with higher priority based on a priority access procedure. However, regular traffic will suffer some impact due to such emergency service, especially when such service is supported unconditionally. A good priority access procedure not only provides resource with higher priority to emergency users, but also reduces the system impact on ordinary traffic. In this study, a priority access procedure using one of three different priority access algorithms based on the Personal Access Communications System (PACS) architecture was designed and investigated. Since this research focused on the effect of parameters related to this procedure and the impact of emergency service on the performance of regular traffic, the power control algorithm was not considered. †
The rest of this paper is organized as follows. An overview of the PACS system is given in Section II. The design of a priority access procedure using one of three different priority access algorithms is described in Section III. The simulation model and performance measure are described in Section IV. Section V presents the simulation results. Finally, conclusions are drawn in Section VI.
II. Overview of PACS PACS is one of the TDM/TMDA (time division multiplexing/time division multiple access) personal communication systems presented by Technical Ad Hoc Group 3 of the T1/TIA Joint Technical Committee (Motorola Inc., 1995) and is open to both public and private system deployment. Its architecture consists of fixed or portable SUs (Subscriber Units) communicating through RPs (Radio Ports) that, typically, have wire-line access via the RPCU (Radio Port Control Unit) and AM (Access Manager) to the PSTN (Public Switched Telephone Network). In PACS, the TDM technique is used in the
To whom all correspondence should be addressed.
− 389 −
J.K. Lain and J.H. Wen downlink while the uplink uses the TDMA technique. Both the uplink and downlink frequencies are in different bands for full duplex operation. Generally, the PACS standard uses the term “time slot” to refer to an ongoing sequence of bursts. Time slots are used to broadcast information or to support user traffic. The time slot numbered 5 in each frame is called the SBC (System Broadcast Channel) and is used to broadcast information. The other slots are called TCs (Traffic Channels) and are used to carry user traffic. Furthermore, traffic channels are distinguished based on their status. TCs are either busy (in use) or idle (not in use), and idle TCs should be marked “available”, “not available”, or “priority access only” (emergency access). Channels marked “available” can be used for regular traffic and emergency service, and those marked “for priority access” are used for emergency service only. Each RP always marks an available channel if at least one idle channel exists while it marks a channel for priority access only when it is necessary. To monitor the quality of a time slot, the SU and RP take radio link measurements on the received signal of each time slot in every frame and detect burst errors via the CRC (Cyclic Redundancy Code). There are three kinds of radio link measurements, (1) RSSI (Received Signal Strength Indicator): the measurement of the received RF signal strength and translation of this measurement into a quantity that can be logically manipulated; (2) QI (Quality Indicator): the estimate of the “eye opening” of a radio signal that is related to the ratio of signal to interference plus noise; (3) WEI (Word Error Indicator): an indication of whether one or more bit errors occur in a burst due to any radio link degradation. With these three indicators, PACS can detect and maintain channel quality.
III. Priority Access Procedure When an emergency call arrives, the SU tries to access the RP in a manner similar to the initial access procedure for regular calls (Motorola Inc., 1995). If the SU fails to seize a channel, it will alert the RPCU using the priority access procedure via the SBC uplink. The priority access procedure is shown in Fig. 1. It should be noted that blocks with dark background are performed on the fixed side (RP and RPCU) while the other blocks are performed on the portable side (SU). The SU alerts the RPCU by transmitting a PRIORITY_ACCESS_REQ message on the SBC uplink and starts a timer, TS3081. If some mistakes occur, like collision or word errors, the SU will request this emergency service again by re-transmitting a
Fig. 1. The flowchart of the priority access procedure.
PRIORITY_ACCESS_REQ message to the RPCU after the TS3081 expires. Upon receipt of a PRIORITY_ACCESS_REQ message, the RPCU transmits a PRIORITY_ACCESS_ ACK message to acknowledge the SU. If the SU fails to receive the PRIORITY_ACCESS_ACK message due to word errors, it must re-initialize the emergency call again after TS3081 expires. On the other hand, if the SU successfully receives the PRIORITY_ACCESS_ ACK message, it stops the TS3081 and starts a timer, TS3031, and it then waits for the PRIORITY_ CHANNEL_ASSIGN message sent from the RPCU. After sending the PRIORITY_ACCESS_ACK message, the RPCU performs the priority access algorithm to select a suitable time slot in which to serve this emergency call. Then, the RPCU informs the SU about the selected time slot ID by transmitting a PRIORITY_CHANNEL_ASSIGN message repeatedly on the downlink in the PRC (Priority Request Channel) until the SU performs an initial access via the selected time slot or until TS3031 expires. This available traffic
− 390 −
Performance of Priority Access Procedure
Fig. 2. Priority access algorithms.
channel, which is selected by the priority access algorithm, is marked as “priority access only” in order to provide emergency service. If no word error occurs in the downlink, the SU transmits the access request on this time slot in accordance with the initial access procedure (Motorola Inc., 1995). In this study, we designed three priority access algorithms, which are shown in Fig. 2. In algorithm 1, the RPCU selects a time slot according to the link quality of each regular call. The RPCU selects the regular call with the minimum signalto-interference ratio (SIR) and forces this regular call to handoff to another RP, i.e., to perform an auto link transfer (ALT) (Motorola Inc., 1995). If the selected regular call performs an ALT successfully, then only the released time slot is marked for priority access. Otherwise, the RPCU should select another regular call with the next minimum SIR to perform an ALT until TN3001 expires or until all the regular calls fail to handoff to another RP. Timer TN3001 is a pre-assigned timer and is used to determine the maximum duration during which the RPCU can perform this algorithm. If no regular call in this RP performs an ALT successfully during the TN3001 period, the RPCU terminates
the regular call with the minimum SIR directly and marks this time slot as “priority access only”. In algorithm 2, the RPCU simply waits for a preassigned time interval, TN3031, for one of the existing regular calls to be completed. During this period, if any regular call can be completed, the RPCU will mark the corresponding first time slot as “priority access only”; otherwise, the emergency call will be blocked. In algorithm 3, the RPCU finds a regular call with the minimum SIR and drops this existing regular call in order to free a time slot for emergency service.
IV. Simulation Model and System Measure The simulation model for the cell configuration is a two-dimensional square service area with 144-RP regular hexagonal locations. Each RP transceiver has eight TDM/TDMA time-slots, operating on any one of K frequencies based on the PACS specification, where K is the frequency reuse factor. A frequency-division duplex scheme is adopted in which separate uplink and downlink frequencies are used (Cox et al., 1987; Bernhardt, 1989, 1991a, 1991b). The frequency as-
− 391 −
J.K. Lain and J.H. Wen signment method used in this simulation is quasi-static autonomous frequency assignment (QSAFA), reported by Chuang (1991). This frequency assignment method can be performed by each RP automatically to determine its dedicated frequency. It consists of signal strength measurements and an algorithm that can cause the RPs to select the frequency with the minimum interference. Other baseline assumptions are as follows: (1) The frequency reuse factor, K, is 7. (2) The RPs are placed in the middle of each cell, with omni-directional antennas, and all the SUs are uniformly distributed over the service area. (3) Call arrivals for both regular traffic and emergency traffic are generated by two independent Poisson processes with mean arrival rates λr and λ e, respectively, where
Fig. 4. Coherent demodulator structure with selection diversity.
λr =
traffic load of regular call (Erlang) × total RP number , mean call holding time (second)
(1)
λe =
traffic load of emergency call (Erlang) × total RP number . mean call holding time (second)
(2)
Both call holding times are independent and exponentially distributed with a mean holding time of 180 seconds. As for the propagation environment, we only consider co-channel interference arising from the first, second and third tiers. This assumption is applicable for RP placed outdoors to serve users in and around houses in residential areas (Cox et al., 1984). In the propagation model, the average received power decreases with the distance R as R −α, where α=4 and the large-scale shadow fading is a log-normal distribution with a standard deviation of 10 dB. Based on the reciprocity theorem (Lee, 1995), the shadow fading between cell i and cell j, S ij, is supposed to equal S ji. By calculating the carrier strength plus the interference and noise strength in the front end of each
Fig. 3. Receiver block diagram for RSSI measurement.
receiver antenna, the RSSI can be derived from the RF structure as shown in Fig. 3. According to the PACS specification, the SU measures the RSSI over a minimum 55 dB dynamic range from −100 dBm to −45 dBm and the RP over a minimum 65 dB dynamic range from −110 dBm to −45 dBm. As shown in Fig. 3, the NE605 provides a linear range from −120 dBm to −30 dBm at point b. Alternatively, it can operate in the linear interval if the received RSSI ranges from −128 dBm to −38 dBm at point a. Thus, the PACS specification can be satisfied using our RF structure. The received RF level is converted into a value in volts by chip NE605. Finally, this RSSI volt value will be converted into 7 bits by the A/D converter. QI is the estimate of the “eye opening” of a radio signal that is related to the ratio of the carrier to the interference plus noise, including the effects of dispersion. This quality is, in part, dependent on the data contained in the received burst and must be averaged to remove the data dependency. In order to get the QI value and WEI for each burst, we obtained simulation results for QI and WEI for various C/I values based on the finite-word-length digital demodulation structure with a joint frequency offset and symbol timing estimator proposed by Chuang and Sollenberger (1991). The finite-word-length digital demodulation structure block diagram and its frequency offset and symbol timing estimator circuit diagram are shown in Figs. 4 and 5, respectively. The
− 392 −
Performance of Priority Access Procedure
Fig. 5. Block diagram of the frequency/timing estimator (used for QI Measurement).
well-known Jakes’ model was employed in the channel model, and Rayleigh fading with a Doppler frequency of 10 Hz and two-branch antenna diversity was adopted in this simulation. The performance measures include the blocking rate PB, dropping rate PD for regular calls and successful access rate PS for emergency calls, where the blocking rate is defined as the probability that a new regular call cannot be assigned a marked-available time slot, and where the dropping rate is defined as the probability that an established link will be dropped during the conversation because the link quality is below the threshold, SIRd, or because the established regular link is forced to terminate by the priority access algorithm. The successful access rate PS is the probability that an emergency call will get through. Incidentally, the SIR threshold for the call dropping rate in our simulation, SIR d , was equal to 10 dB (Walker, 1990).
the emergency traffic load, and Fig. 7 depicts the system performance PD versus the emergency traffic load for priority access algorithms 1 and 3. From these figures, we observe that provision of emergency service does have a certain impact on regular traffic. Additionally, we find that PB with algorithm 1 is slightly higher than it is with algorithm 3. However, P D with algorithm 1 is lower than it is with algorithm 3. This is because algorithm 3 uses the policy of dropping a regular call directly while algorithm 1 instructs the regular calls to try to execute an ALT. In other words, the effective total traffic load of algorithm 3 is lighter than that of algorithm 1. Therefore, the simulation results obtained using algorithm 1 have a lower PD but a slightly higher PB than do those obtained using algorithm 3. Furthermore, Fig. 8 shows the system performance P S versus the traffic load of emergency calls for algorithms 1 and 3. Although both algorithms, 1 and 3, can find a time slot to provide emergency service no matter how heavy the traffic load is, the emergency service still might
Fig. 6. Blocking rate of regular calls versus emergency call traffic using priority access algorithms 1 and 3.
V. Simulation Results The following simulation results concerning the effect on the system caused by emergency service were obtained using our simulation system based on the PACS architecture, where the values of TS3081, TS3031, and TN3001 were set to be 2, 1, and 1 seconds, respectively, according to the PACS specification (Motorola Inc., 1995). Incidentally, the abbreviation R.C. in Figs. 6-11 denotes the traffic load of regular calls. Figure 6 shows the system performance P B versus
Fig. 7. Dropping rate of regular calls versus emergency call traffic using priority access algorithms 1 and 3.
− 393 −
J.K. Lain and J.H. Wen The system with algorithm 2 behaves like a queuing system for emergency service. Under an equivalent state, the mean departure time for regular calls is equal to the mean inter-arrival time. Thus, with 5-Erlang regular call traffic, the mean departure time is calculated to be approximately 36 seconds using Eq. (3). From Figs. 9 and 10, we find that both P B and P D increase slightly with the increased time setting of the TN3031 timer. This is because the provision of emergency service will increase the total traffic in the system. However, since algorithm 2 allows the emergency call to wait for the completion of a regular call, the impact of algorthm 2 on regular calls is not as serious as it is when algorithms 1 and 3 are used. Additionally, a higher PS can be achieved with a longer TN3031 setting Fig. 8. Successful access rate of emergency calls versus emergency call traffic using priority access algorithms 1 and 3.
Fig. 9. Blocking rate of regular calls versus the TN3031 setting using priority access algorithm 2 under a 0.1 Erlang emergency traffic load.
Fig. 10. Dropping rate of regular calls versus the TN3031 setting using priority access algorithm 2 under a 0.1 Erlang emergency traffic load.
be blocked due to collisions when it tries to access the channel marked for priority access. When the mean inter-arrival time, T , of emergency calls is calculated using the following equation:
T=
mean call holding time (second) , traffic load (Erlang)
(3)
it is found to be 1800 seconds if the emergency call traffic is 0.1 Erlangs. With such a long mean interarrival time, failure of emergency service due to collision will occur so seldom that the successful access rates for emergency calls for algorithms 1 and 3 will reach 100%. Figures 9-11 show the system performance P B, PD, and PS versus the TN3031 setting for priority access algorithm 2 under a 0.1 Erlang emergency traffic load.
Fig. 11. Successful access rate of emergency calls versus the TN3031 setting using priority access algorithm 2 under a 0.1 Erlang emergency traffic load.
− 394 −
Performance of Priority Access Procedure as shown in Fig. 11.
R.O.C., under Grant NSC 84-2221-E194-003.
VI. Conclusions
References
In this study, we designed a priority access procedure and three different priority access algorithms. The simulation environment was set up based on the PACS architecture with consideration given to prototype hardware implementation. The simulation results show that regular traffic suffers somewhat due to the provision of emergency service. From the simulation results, both algorithms 1 and 3 can provide emergency service under heavy traffic load conditions. But the impact on regular traffic due to algorithm 1 is smaller than that due to algorithm 3 if the effects on P B and P D are both considered. Although priority access algorithm 2 can provide emergency service with less impact on the regular calls compared to algorithms 1 and 3, the corresponding average waiting time is too long. With the goal of providing fast and reliable emergency service while minimizing the impact on regular traffic, priority access algorithm 1 employed in the priority access procedure is the best among the three studied algorithms for implementation in the PACS system.
Bernhardt, R. C. (1989) User access in portable radio systems in a co-channel interference environment. IEEE J. Select Areas Commun., 7, 49-58. Bernhardt, R. C. (1991a) Call performance in a frequency reuse digital portable radio systems. IEEE Trans. on Veh. Tech., 40, 777-785. Bernhardt, R. C. (1991b) Time-slot management in digital portable radio systems. IEEE Trans. on Veh. Tech., 40, 261-272. Cox, D., H. Arnold, and P. Porter (1987) Universal digital portable communications : a system perspective. IEEE J. Select Areas Commun., SAC-5, 764-773. Cox, D., R. Murray, and A. Norris (1984) 800 MHz attenuation measured in and around suburban houses. Bell Labs. Tech. J., 921-954. Chuang, J. C. I. (1991) Autonomous adaptive frequency assignment for TDMA portable radio system. IEEE Trans. on Veh. Tech., 40, 627-635. Chuang, J. C. I. and N. R. Sollenberger (1991) Burst coherent demodulation with combined symbol timing, frequency offest estimation, and diversity selection. IEEE Trans. Commun., 39, 1157-1164. Lee, C. Y. (1995) Mobile Cellular Telecommunications, p. 60. McGraw-Hill, Inc., New York, NY, U.S.A. Motorola Inc. (1995) PACS Air Interface Specifications. PN3418, Joint Technical Committee. Silventoinen, M. I. and T. Rantalainen (1996) Mobile station emergency locating in GSM. ICPWC’96, pp. 232-238. New York, NY, U.S.A. Walker, J. (1990) Mobile Information Systems. Artech House, Boston, MA, U.S.A.
Acknowledgment This research was supported by the National Science Council,
− 395 −
