MPT 1318 Engineering Memorandum Trunked Systems in the LandMobile Radio Service February 1986 Revised and reprinted January 1994 1
Foreword 1 This Code of Practice has been produced by the Radio Communications Agency 2 For further details please contact: Mobile Services Section Radiocommunications Agency Waterloo Bridge House Waterloo Road LONDON SEl 8UA 2
Contents MPT 1318 1 Revised and reprinted January 1994 1 Foreword 2 1 Introduction 6 2 Application of this memorandum 6 3 Trunked systems 7 4 Benefits provided by a trunked system 7 4.1 Spectrum utilisation 7 4.2 Operation 7 4.3 Reliability 7 4.4 Privacy 8 4.5 Expansion 8 4.6 Example of a practical trunked system 8 5 System considerations 10 5.1 Frequencies and channel separation 10 5.2 Power levels 10 5.3 Portable and mobile equipment 10 5.4 Queueing 10 5.5 Signalling 10 5.5.1 Speed and format 10 5.5.2 Protocol 11 5.6 Loading 11 5.7 Wide area coverage 11 6 Technical consideration and recommendation 11 6.1 The application of Erlang C to performance analysis 11 6.1.1 Grade of service 12 6.1.2 Holding time 12 6.1.3 Assumptions made when analysing a trunked system 13 6.1.4 Grade of service and system capacity calculations using Erlang C formula 14 6.1.5 Mean waiting time 16 3
6.1.6 System performance summary using 18 Erlang C 18 6.2 Practical system performance evaluation by computer simulation 19 6.2.1 Control strategies 19 6.2.2 The model 20 6.2.2.1 Dedicated queueing system 20 6.2.2.2 Scanning queueing system 20 6.2.2.3 Voice channel signalling system 20 6.2.3 Mobile action 21 6.2.3.1 Dedicated queueing 21 6.2.3.2 Scanning queueing 21 6.2.3.3 Voice channel signalling 21 6.2.4 Computer simulation 21 6.2.5 The importance of fast signalling (conclusions) 22 6.2.6 The control channel allocation strategy 29 6.2.7 Other results 29 6.2.8 Other aspects 30 6.2.8.1 Contention control 30 6.2.8.2 Tolerance of propagation errors 30 6.2.8.3 Off-air call set-up 30 6.2.8.4 End of conversation recognition 31 6.2.9 Summary 31 6.3 System failure 31 7 Frequency assignment and intermodulation 31 8 Traffic monitoring considerations 32 8.1 Monitoring periods 32 8.2 Presentation 32 8.2.1 Busy hour detail 33 8.2.2 Overall loading 33 APPENDIX A 35 APPENDIX A - DEFINITIONS 35 1 Blocking System 35 2 Busy Hour 35 3 Call 35 4
4 Contention 35 5 Dedicated System 35 6 Erlang 35 7 First-in/first-out (FIFO) 36 8 Grade of Service (GOS) 36 9 Holding Time 36 10 Load 36 11 Mean Holding Time 36 12 Message Trunking 36 13 Overheads (signalling) 37 14 Polling System 37 15 Protocol 37 16 Queueing System 37 17 Scanning System 37 18 Time-out 37 19 Traffic Intensity 37 20 Transmission Trunking 38 21 Unbalanced System 38 22 Voice Channel Signalling System 38 APPENDIX B 39 APPENDIX B - REFERENCES 39 APPENDIX C 40 APPENDIX C - COMPUTER PROGRAM FOR ERLANG C FORMULA 40 5
Trunked systems in the land mobile radio service 1 Introduction The growth in demand for frequencies for land mobile radio services is currently around 10% a year and shows no signs of diminishing in the near future. This ever increasing demand is proving difficult to satisfy with the spectrum allocated to land mobile radio and therefore new techniques are being sought which can make more use of the available spectrum. One such technique is the application of trunking which allows more use to be made of a channel by making users less aware of the congestion on the channel. With trunking, users share a pool of channels and are allocated temporarily a vacant channel to make a call. It achieves a reduction in waiting time for the users as a consequence of the decreasing probability that all channels will be in use at the same time the more of them there are. Once the number of channels in the pool exceeds three, the gain from trunking compared with single channel operation can be significant. Although trunking has been applied to telephone systems since before the development of radio, the technique may be unfamiliar to radio specialists. This memorandum is intended to give an introduction for radio users, operators and designers to the principles of trunking and the potential benefits. Suggested design methods are given together with recommended parameters for successful system design. Difficulties are discussed, particularly those associated with multiple frequency base stations, and the regulatory requirements are given. Tables of delay probabilities and an explanation of the terms used are provided as appendices. 2 Application of this memorandum This memorandum applies to all the bands reserved for private land mobile radio use. Adherence to this memorandum should help to ensure that full advantage can be taken of the benefits offered by trunking. In some cases, however, there may be licensing conditions associated with particular bands or individual systems which will make some of the following points mandatory. 6
3 Trunked systems "Trunking" has become a generally accepted term for the automatic sharing of several communication channels by a number of users. For the term to have any real meaning the number of users must be many times the number of trunks. The term is well known to telecommunications engineers who apply it to a path for traffic between 2 points. For example, a road between two towns can be considered a trunk in the same way as a wire line between two telephone exchanges. The common factor linking these two examples is the sharing of a resource, the communication path, by a number of users, namely road traffic vehicles on the one hand and telephone subscribers on the other. There is no reason to exclude a radio communication path from the trunking concept and a trunked radio system is a system in which many users share a common pool of radio channels. In this case, for mobile applications, the common resource is the assignment of frequencies for the system and for communication conducted within the common coverage area. Channels from the pool are allocated, on demand, for the duration of a call and as calls are completed the channels are returned to the pool for allocation to other users. The important principle behind this concept is that any user has access to any free channel within the pool. 4 Benefits provided by a trunked system 4.1 Spectrum utilisation For trunked systems more mobiles per channel can be accommodated, or a better grade of service can be provided, i.e. a reduced average time to establish communication, resulting in an improvement in spectrum utilisation. A single trunked scheme will have a higher capacity and better efficiency than is possible with an equivalent number of single channel systems. 4.2 Operation Operation, as far as the user is concerned, may be extremely simple. Selection of channels is automatic and a call could be initiated, for example, by lifting a handset. The call would be placed, without operator intervention, as soon as a tree channel became available. 4.3 Reliability Short term loss of a channel, due to interference or maintenance, will result only in a reduced grade of service and not in a complete loss of communication. 7
4.4 Privacy A degree of privacy is obtained. Conversations between users or groups are not overheard by third parties in the group. Privacy in this context should not be interpreted as secrecy. 4.5 Expansion New users can be accommodated more easily in a trunked system than in one of several single channel schemes, all of which may be nearly full. Similarly, additional channels can be made available to all participants in a scheme, without the need for modifications to existing mobile or portable equipment. 4.6 Example of a practical trunked system The majority of radio system users who have more than one channel could benefit by re-organising their facilities into a trunked radio system. This is true for single-site and multi-site Systems. To investigate the possibility of benefit, it is worthwhile putting each mobile operator, portable operator and controller into one or more user groups. Each operator or controller should have common communication needs with other users in that group. (See Figure 1). There may also be a need for the mobile and portable users to talk to each other as well as to their controller, and for there to be more than one conversation at a time involving different members of the same user group. Each of these needs requires a separate communication channel. Trunking will provide benefit if the number of communication channels required is greater than the number of RF channels assigned to the system. The greatest benefits arise when the number of communication channels needed is many times the number of available RF channels. Trunked systems are unsuitable for applications where everyone using the system needs to be a party to every call, or where all mobile operators need to speak only to the single controller. In those cases, all operators and controllers are in a single user group and only one communication channel is needed. For all other users, trunked systems can provide real benefits. 8
9
5 System considerations 5.1 Frequencies and channel separation Operation of a trunked radio system need not be limited to any specific frequency band or channel separation. The usual constraints to minimise the effects of intermodulation and co-channel interference apply. See also Section 7. 5.2 Power levels Trunked systems operate with power levels equivalent to those used in non-trunked systems. 5.3 Portable and mobile equipment Fast, automatic channel switching and a reliable form of signalling suited to operation in mobile radio environments are of fundamental importance. 5.4 Queueing To obtain maximum benefit from trunking, a queueing system is recommended which is based on First In - First Out (FIFO) queueing of incoming calls rather than a blocking system where callers have to continue to seek access to the system until successful. 5.5 Signalling The signalling scheme is crucial to efficient operation of a system and should be selected only after careful consideration of the topics, which follow. 5.5.1 Speed and format Channels must be allocated with a minimum of delay. Therefore it is probable that a digital signalling scheme will be selected and the speed chosen accordingly. In addition, reliable signalling requires a signal strength no greater than that necessary for acceptable voice communication under the same conditions. Digital signalling should provide advantages over sequential tone signalling of increased flexibility, greater information content and means of applying checks for accuracy (validity) of signalling. 10
The recommended signalling system is described in MPT 1317 - Code of Practice - Transmission of digital information over Land Mobile Radio Systems. 5.5.2 Protocol Attention must be given to the following aspects of any protocol: - the method for allocating the signalling channel, depending whether for a dedicated or scanning system - the method of access to the signalling channel, including contention control - the method of call supervision Section 6 gives additional details. 5.6 Loading System loading is a function of the number of users, the call rate (number of calis per hour), the number of channels and the average duration of a call. Reasonable estimates of performance may be made using these parameters (see Section 6). 5.7 Wide area coverage Multi-site operation is not precluded by trunking. Additional receiving only sites will provide more reliable mobile to base communications particularly in systems employing handportable equipment. Synchronous or quasi synchronous systems can be trunked also, but there will be the usual constraints regarding the frequency stability and audio delay distortion. 6 Technical consideration and recommendation 6.1 The application of Erlang C to performance analysis The formulae for estimating system performance given in this section are derived from Erlang C theory (see ref. 1) which is based on the following assumptions: - an infinite number of users - random intervals between calls - random call duration times - negligible call set-up times 11
- first-in, first-out queuing A practical radio system will not fulfil all the foregoing criteria, but Erlang C can quickly provide approximate performance figures. Section 6.2 presents the results using a more accurate, but time consuming, method. 6.1.1 Grade of service Grade of service is a measure of the degree of congestion that is of the condition where the immediate establishment of a new connection is impossible owing to the unavailability of communications channels. Users having sole access to a lightly loaded radio channel obtain a very good service but seldom use the channel heavily for very long. The practice of uncontrolled sharing of a channel between many independent users results in better utilisation of the channel but also in a worse grade of service as users sometimes have to wait for access. A useful way to define the grade of service is the probability that a user will experience a specified delay in the busy period (e.g. one hour) of an ordinary day. Such delays get worse as channel, loading increases and the channel is deemed to be saturated when the grade of service has become unacceptable to the user. It should be noted that low values of grade of service are the most acceptable to the user. When planning a radio system, therefore, a reasonable approach is to choose a delay which may cause some annoyance (e.g. 20 seconds) and to specify that only a small proportion of calls (e.g. 5%) should be delayed more than this. It may also be useful to guard against occasional long waits (e.g. 1% probability of waiting more than 1 minute) In trunked systems, users have access to more than one radio channel. If the system is properly designed this can improve dramatically the grade of service for the same channel loading. This improvement may be desirable in itself and/or can be used to achieve higher channel loadings for a given grade of service. No estimate of the traffic capacity of a trunked system, however, can be made without first deciding what grade of service is to apply. 6.1.2 Holding time The channel holding time is defined as the total period for which a channel is engaged by the user for the purpose of sending, and/or receiving speech/data, following a channel assignment. This does not include the overhead time required to set up and clear down a call. In a trunked system, the time required for setting up and clearing down a call will have been specified by the system designer. This time taken for a typical radio message or conversation, however, depends on users and will have been either defined with due consideration to their specific requirements, or evaluated by monitoring their transmissions. 12
In practice, long channel holding times are unlikely to impede access to the system during periods of low activity. As the number of calls increases and traffic builds up, holding times become significant because their duration affects directly the channel loading and so also the grade of service. When planning a system, the holding time used is the mean value of holding time per mobile in the busy hour of an ordinary day, this value having been either estimated or evaluated. Additionally, it is necessary that the number of calls which each mobile is likely to make in the same period be either assumed or evaluated. The traffic capacity of the system, the minimum number of channels required and the maximum number of mobiles in the system can then be calculated for the desired grade of service. 6.1.3 Assumptions made when analysing a trunked system In the preceding paragraphs, the need has been emphasized for quantified parameters, which are required when planning a trunked system. In the absence of specific user instructions based on knowledge and experience, the recommended assumptions and values are as follows: Grade of service Waiting time (W) in busy hour 20 s Probability of W being exceeded: system design value 5% Grade of service (continued) acceptable value 10% system overloaded value 30% Holding time Mean channel holding time per mobile in busy hour 20 5 Number of calls per mobile in the busy hour 1 13
6.1.4 Grade of service and system capacity calculations using Erlang C formula If the necessary conditions are assumed to be fulfilled, the probability (PD) of a call experiencing a delay can be obtained from: PD = AC K=C-1 (1) AC+C! (1-A/C) k = 0 where: C = number of trunked channels A total traffic load in erlangs (see ref 1) Values of ~D as a function of A and C, are given in many standard reference books and can be found also in Appendix D. If necessary, the expression (1) can be re-arranged for calculation of intermediate values using a pocket calculator. Alternatively, the Basic computer program given in Appendix C may be used. The Probability P'(W>t) that delayed calls must wait for a periods in excess of a specific time t is given by: where: P'(W>t) = exp [-(C-A) t/h] where: (2) H = mean holding time per mobile in the busy hour. Therefore the probability P(W>t) that any call will be delayed more than t seconds can be obtained from (1) and (2): P(W>t) = PD exp [-(C-A) t/h] (3) Equation (3) is effectively an expression for grade of service as a function of system load A and number of channels C. In figure 2, P is plotted as a function of channel load 'a' (a=a/c) for different values of C. The ratio of t/h has been set to 1 in this case, but if this ratio is increased, i.e. a longer waiting time can be accepted, then the complete 14
family of curves will move to the right. A typical example of the usefulness of these curves is as follows. A 5 channel trunked system is planned and a 5% grade of service is required. (It is assumed that the 5 channels are all voice channels and no allowance has been made for signaling requirements). From the 5% level on the y-axis, a line parallel to the x-axis is drawn until the 5 channel characteristic is intersected. This occurs at a channel loading of 0.645erlang, which corresponds to a system load of 5 x 0.645 = 3.22 erlang. The required system parameters are as follows: Number of channels 5 Grade of service 5% System load 3.22 erlang In order to determine how many mobiles such a system will support, it is necessary either to know or to make assumptions on the traffic statistics. In the absence of other inputs the recommendations for mean holding time (H = 20s) and a call rate (λ = 1) can be used to assess the system capacity in terms of the number of users, or mobiles, that can be supported. By definition, load is the product of call rate and mean holding time. Therefore: A = M λ H (4) 3600 where: A = total system load in erlangs λ= number of calls per mobile in the busy hour H =mean channel holding time per mobile, in seconds, in the busy hour M = number of mobiles Note that as the call rate λ is given in calls per mobile per hour, this requires that the mean holding time, normally given in seconds, be converted to hours, hence the factor of 3600 in the denominator. Rearranging the expression (4) in terms of mobiles, M: M= 3600A (5) λ H With A = 3.225 erlang and the recommended values for H and λ, the total number of mobiles is: M = 3600 x 3.22 = 580 1 x20 15
6.1.5 Mean waiting time Grade of service has been defined in terms of the percentage of calls that must wait more than a specified time, but a further useful figure for assessment of system performance is the mean waiting time. Mean waiting time for delayed calls (WD) can be found by integration of equation (2): WD= exp [-(C-A) 't/h] = H/(C-A) (6) Hence the mean waiting time for all calls (WA) will be given by: WA= PD H/(CA) (7) Example for a 5 channel system with a 5% grade of service: C A H = 5 channels = 3.22 erlangs = 20 seconds PD = 0.2941 (From Appendix C) WA = 3.3 seconds 16
17
6.1.6 System performance summary using Erlang C In section 6.1.3 the design value, the acceptable value and the overload value for grade of service were set out as 5%, 10% and 30% respectively. In Table 1 the corresponding traffic loads, numbers of mobiles and mean waiting times are tabulated for 5, 10,15 and 20 channel systems. Number of Channels, C 1 (for comparison) 30 5 5 10 30 10 5 10 30 15 5 10 30 20 5 10 30 Table 1 Grade of Service (%) 5 10 Traffic (erlang) Number of Mobiles Per channel, a 0.120 0.220 0.500 0.645 0.719 0.846 0.793 0.839 0.914 0.853 0.886 0.939 0.885 0.911 0.953 Total, A 0.12 0.22 0.50 3.22 3.59 4.23 7.93 8.39 9.14 12.79 13.29 14.09 17.70 18.22 19.06 Per channel 22 40 90 116 129 152 143 151 165 153 159 169 159 164 172 Total M 22 40 90 580 645 760 1430 1510 1645 2300 2390 2535 3185 3280 3430 Mean Waiting Times, (s) 2.7 5.6 20 3.3 5.8 16.8 3.8 6.2 16.5 4.1 6.5 16.4 4.3 6.7 16.4 Note W=20s; H=20s; λ = 1; a=nc The values given in this Table are derived from theoretical calculations and do not make allowance for call set up time. 18
The numbers in Table 1 strengthen conclusions which may be drawn from examination of the curves in Figure 2. Note for example, that a single channel can withstand only a light loading of 0.12 erlang before the 5% grade of service threshold is reached. However 1 for a trunked system the allowable channel loading increases with the number of channels, reaching a value close to 0.9 erlang for 20 channels, for the same 5% grade of service. Therefore, under a given set of conditions, such as a 5% grade of service, a mean call rate (λ) of 1 and a mean holding time (H) of 20 seconds, the number of mobiles will increase from 22 for a single channel to 159 per channel for 20 trunked channels. This represents a considerable improvement in spectrum utilisation. Also in Figure 2 it is apparent that the "knee" of the curves becomes sharper as the number of channels (C) increases. This is emphasized by the values in the column listing the number of mobiles per channel. For a single channel, addition of 68 mobiles will reduce the grade of service from 5% to 30%. Put another way, a single channel can accommodate 4 times as many mobiles for a 30% grade of service than for a 5% grade of service. A 20 channel system on the other hand can tolerate only 13 additional mobiles per channel (about 8%) for the same reduction in the grade of service. 6.2 Practical system performance evaluation by computer simulation Whilst the previous section is based on theoretical assessment, this section presents the results obtained from the computer simulation of a model based on practical considerations and designed as an evaluation tool for non-specific trunking systems. The results show the effect of signaling speed and of different control strategies on the system performance and comparison is made with an ideal queueing system. In addition, other aspects of the trunking system design which can affect efficiency are discussed. 6.2.1 Control strategies It is assumed that the data signalling necessary to assign channels to callers takes place on a control channel. Two strategies for allocating the control channel are considered: scanning and dedicated. A dedicated system has a control channel permanently available for signalling; a scanning system uses the control channel for speech if all other channels are in use. In the latter case, the next channel to become free is marked as the control channel and the mobiles find it by scanning. When a scanning system reassigns the control channel, time must be allowed for mobiles to scan the channels and find it. This scanning time has been assumed to be 100 ms multiplied by the number of channels. The radio system mainly considered is one based on a queueing strategy in which calls that cannot be serviced immediately are queued; then, as channels become free, they are assigned to mobiles in First-in, First-out (FIFO) order. However, some results are given for a blocking system which rejects call requests when all channels are busy and relies on users to try again later. 19
6.2.2 The model The system is modelled as a single-server queue representing the control channel, followed by a multi-server queue representing the speech channels. The service time of the single-server queue corresponds to the total reassignment time of a speech channel and is modelled as having a negative exponential distribution with a given minimum time (T1) and a given mean (T2). (T2 is the mean prior to the minimum limit distortion). This distribution represents the variation in access signalling times caused by the need for message repetition because of propagation error and message collision. The second (multi-server) queue models the use of channels for speech traffic where the number of servers corresponds to the number of speech channels. The service time of this queue represents the duration of a call (the holding time) and is assumed to have a negative exponential distribution with a given minimum (taken as 1 second). New calls are generated with a Poisson distribution (the mean rate corresponding to the traffic level) and join the first queue. 6.2.2.1 Dedicated queueing system In a dedicated queueing system the number of servers in the second queue is one fewer than the number of channels. New calls join the first queue and are served in FIFO order by the second queue. 6.2.2.2 Scanning queueing system In a scanning queueing system the number of servers in the second queue is equal to the number of channels. When all servers are busy, the first queue is inhibited from serving, but call requests are not blocked. Instead they form a pool of pending requests which is effectively another, random service, queue feeding the first queue. When any one of the servers in the second queue becomes free, the first queue is inhibited from serving for the scanning time and then service can resume. 6.2.2.3 Voice channel signalling system A voice channel signalling system is a special case of a scanning queueing system. A free channel is marked as the control channel and this channel is allocated to the next call request. The control channel is then moved to the next free channel if one is available. There is no FIFO queue of pending requests in the system controller; all queueing is random as mobiles wait to access the system. The number of servers in the second queue is equal to the number of channels. New calls join the first queue and are served in random order by the second queue, as long as at least one server is free. The first queue is inhibited from serving, for the scanning time, following every channel assignment and following a period when all servers in the second queue have been busy. 20
6.2.3 Mobile action In order to give a more practical description of the three types of control strategy that have been considered, the basic signalling action of a mobile operating in these hypothetical systems is described. Real systems, of course, will be more complicated than the systems considered here due to time-outs, area coverage schemes, fail-soft subsystems and many other factors. 6.2.3.1 Dedicated queueing A mobile operating in a dedicated queueing system monitors continuously the control channel for instructions to switch to a speech channel. If such an instruction is received, the mobile changes to the directed channel and conversation can begin. At the end of the conversation the mobile returns to the control channel. If the mobile wishes to initiate a call, it sends a request on the control channel (using a multi-access protocol). Then either it will be given a speech channel immediately, or it will be queued and directed to a speech channel at a later time. 6.2.3.2 Scanning queueing A mobile operating in a scanning queueing system tracks the control channel continuously by scanning. If the control channel has been 'lost then the mobile scans cyclically all the channels in the system, pausing on each channel for sufficient time to detect whether or not control channel data messages are present. When the control channel has been found then the mobile dwells on it and operates in the same way as a mobile in a dedicated queueing system. If a call request is initiated while the mobile is scanning, then the request remains pending in the mobile until the control channel has been found. 6.2.3.3 Voice channel signalling A mobile in a voice channel signalling system operates in the same manner as a mobile in a scanning queueing system. However, the mobile is never queued (in the system controller) for allocation of a speech channel at a later time. 6.2.4 Computer simulation The computer simulation was validated by comparison with theory and by comparison with the results obtained from other simulations. Comparison with theory was possible with the dedicated queueing system by setting the service time of the first queue to zero (T 1 =T 2 =0); the model now becomes a pure MIMIN FIFO queueing system which is analytical by the Erlang C formula or by reference to published tables (2). For 21
comparison with other simulations, T 1 and T2 were chosen to correspond as closely as possible to known signalling protocols of which extensive simulations have been performed (3). These simulations have models which include radio propagation path loss, shadowing and fading, receiver capture, geographical distribution of mobiles, mobile fleet structures and many other effects. The model described was found to over estimate the traffic carried on 20 channels (worst noted case was a difference to 0.04 erlang/channel, the main cause of the discrepancy being that the model takes insufficient account of the contention control problem at higher call rates). Below 15 channels the model was found to be reasonably accurate. 6.2.5 The importance of fast signalling (conclusions) Figures 3 to 5 show the traffic which can be carried by scanning and dedicated queueing systems with different reassignment times. The results are shown for a fixed grade of service, the grade of service criterion being that the probability of waiting for a channel for more than 20 seconds is 10%. The mean speech channel holding time is 20 seconds, with a minimum holding time of 1 second. The figures also show the traffic that could be carried for the same grade of service by an ideal queueing system (in which no time is needed for signalling so that there is infinitely fast' channel reassignment); the differences represent the signalling overheads of the practical systems. Figure 3 shows the traffic that can be carried by a trunking system with a high performance protocol (T 1 =300 ms, T2=500 ms). The overheads at 20 channels for the dedicated system are less than 0.1 erlang/channel and there is not much to be gained by signalling faster. Further simulation has shown that there is no decline in speech carrying capacity for up to 35 channels. The reassignment speed of T1 =300 ms and T 2 =500 ms can be met with 1200 bit/s signalling (3). Figure 4 shows the traffic that can be carried when using a less sophisticated protocol (T 1 =600 ms, T2=900 ms). The overheads of the dedicated system are a minimum of 0.15 erlang/channel at 10 channels. The increase for larger systems is caused by channels becoming free at a faser rate than the control channel can assign them. Figure 5 illustrates the dangers of signalling too slowly (T1 =1 S, T 2 =1.5 s). At 20 channels almost half the system capacity has been wasted, because there is now a control channel speed bottleneck with a constant backlog of pending callers. The overheads of the dedicated system have a minimum of 0.22 erlang/channel at 8 channels. 22
23
24
25
26
27
28
6.2.6 The control channel allocation strategy Figures 3 to 5 illustrate also the relative performances of the scanning and dedicated control channel strategies. With few channels, dedicated systems are inefficient as the control channel carries little signalling traffic and so is idling for most of the time. As the number of channels is increased, the time taken for mobiles to locate the control channel in a scanning system (the scanning time) becomes a significant overhead to the signalling. This explains the scanning/dedicated crossover effect observed in the results. Figures 3, 4 and 5 have crossover points at 14,13 and 13 channels respectively. Below these figures a scanning system is more efficient and above it a dedicated system is more efficient but, owing to the gradual nature of the transition, the exact crossover point is not critical. The scanning/dedicated crossover points were generated using the premise that the scanning period is 100 ms multiplied by the number of channels. This is derived by assuming that a scanning mobile has to pause on each channel for sufficient time to recognise a data message sync word, and that in order to allow more than 95% of mobiles to find the control channel before signalling begins, we must multiply the 'pause time' by the number of channels. At first sight the 'pause time' should be: (number of bits in message slot + number of bits in sync word) I bit rate (for MPT 1317 this would be about 120 ms), but in fact this time can be reduced by sending concatenated sync words on the control channel during the scanning period, and/or by using hardware which can detect quickly the presence of data. 100 ms has been chosen as a compromise, but Appendix F gives some results for other pause times. 6.2.7 Other results Appendix F tabulates further results from the simulation for a wider range of signalling speeds and other conditions. Also, Figures 6 to 8 show the delay probabilities for 1, 5 and 10 channel scanning systems, to illustrate the advantages of using a queueing rather than a blocking approach. It is assumed that, in the blocking system, requests for service are rejected when all channels are busy and are regenerated after a random retry time (with a mean of 10 seconds). This approach is inefficient because when a channel becomes free it remains unused until one of the waiting users requests it. 29
6.2.8 Other aspects Other design considerations which have a bearing on system efficiency are considered in the subsections which follow. 6.2.8.1 Contention control One of the problems in radio trunking systems is that many users can try to access the system at the same time. These access attempts can clash and, in the absence of any control, can produce an unstable situation where the requests continue to clash and no requests are received successfully. Controlled multi-access protocols, which impose a discipline on users trying to gain access, are used to alleviate this contention problem. The level of sophistication of the chosen multi-access protocol determines the maximum message throughput, which has a direct bearing on system efficiency. Contention control may take the form of simple schemes such as pure Aloha or slotted Aloha (4), or may be more sophisticated and use feedback control to avoid instability (5,6). It is difficult to provide quantitative performance comparisons of these options in a mobile radio system, as the contention control algorithm is embedded in the signalling protocols and is system-specific. The only reliable way of assessing the performance of a protocol is by simulation of the specific system under consideration. Evidence that the simpler schemes suffer from instability is available in the literature (5,7). 6.2.8.2 Tolerance of propagation errors Errors may be caused by receiver noise in the demodulated data messages which are sent between mobiles and base station to establish conversations, and which are affected by RF path loss, fading and shadowing (8). Errors can be introduced also by co-channel interference effects and automotive ignition noise. This necessitates the use of selective retransmission protocols (9), error detection/correction codes and modems with adequate noise performance. A poor design will lose system efficiency because of unnecessary message redundancy and may give a poor grade of service to mobiles at the edge of the service area. 6.2.8.3 Off-air call set-up f a call to an individual unit is initiated, a check should be made before a channel is assigned to ensure that the called mobile is switched on, within range and not engaged in conversation with another party. Failure to do so will lead to the loss of system capacity as speech channels are unnecessarily assigned. 30
6.2.8.4 End of conversation recognition Capacity can be wasted by tailing to recognise that a conversation has finished, and having to rely on an inactivity time-out before the channel can be reassigned. This necessitates a design of mobile equipment which encourages operators to go 'on-hook' (or indicate in some other way) when a conversation has ended, so that the system controller can be signalled that the channel is free. 6.2.9 Summary There are several important design aspects of a mobile radio trunking system which have a direct bearing upon its spectral efficiency. These include: - the speed of re-assignment signalling - the control channel allocation strategy - the type of service - the contention control algorithm - the tolerance of propagation errors - the use of 'off-air' call set-up for selective calls - the reliability of 'end-of-conversation' recognition. 6.3 System failure Failure of the trunking control logic should not result in total failure of the radio system. In this event, communications should be maintained even at the expense of a lower grade of service. 7 Frequency assignment and intermodulation Where a number of transmitters share an antenna or operate with closely spaced antennas, coupling between transmitters will occur. Due to non-linearities in transmitter output stages, intermodulation products will be generated and radiated by the transmitter antenna. In trunked systems employing closely spaced channels, these products may cause interference to channels within the group assigned to the trunked system. These intermodulation products may affect also users of other radio Systems. Due to the continued demand for new radio services making severe demands on the available radio spectrum, there can be no guarantee that the channels assigned to a trunked system will be free of intermodulation products. In order to minimise such interference, a likely requirement for trunked radio systems is that radiated intermodulation products should not exceed a level of 80 db below the radiated carrier power. 31
8 Traffic monitoring considerations Trunked radio systems represent a large investment in capital equipment for the system operators and in radio spectrum for the regulatory authority. It is recommended, therefore, that all systems be monitored. Actual system parameters such as channel utilisation, call rate and mean holding time etc. can be determined and compared with values used for planning of the system. Major deviations from the planned values may require system adjustments, for example reduction of time-out periods to achieve an acceptable balance between grade of service offered to users and spectrum utilisation. From time to time the regulatory authority may require system operators to carry out monitoring and provide information about the traffic carried by the system both in total and for individual user groups or mobile fleets. The monitoring equipment may be installed locally or remotely from the trunked system and should be capable of providing the information described in sections 8.1 and 8.2. 8.1 Monitoring periods Traffic occupancy measurements may be required for 12 or 24 hour periods, recorded at 15 minute intervals (commencing on the hour) or the preceding 15 minute period. 8.2 Presentation The information should be presented on A4 size sheets for the overall system loading and for the user group and fleet loading. The information should include: - NAME of the trunked system operator (name of mobile fleet operator) - LOCATION of the trunked system (location of the fleet control point) - DATES of the beginning and end of the monitoring period - AVERAGE TIME OCCUPANCY for each 15 minute period, presented as a time/percentage occupancy histogram for a 12 or 24 hour period. The percentage occupancy for each 15 minute period should be represented by an asterisk (or similar non-alphanumeric character) for each percentage point. A scale should be provided for the percentage occupancy axis. Alternatively, lines representing 40 and 80 per cent occupancy values may be drawn on the histogram. The time axis should show the time (in GMT) of the end of each 15 minute period. An example of the presentation is given in Figure 9. 32
8.2.1 Busy hour detail The percentage occupancy for successive groups of four 15 minute periods should be examined in order to determine the busy hour. The number of calls made in the busy hour and their mean duration should be shown. 8.2.2 Overall loading The mean percentage occupancy of the 12 or 24 hour period should be calculated. The ratio of busy hour (peak) occupancy to the mean 12 or 24 hour occupancy should be calculated. 33
34
APPENDIX A APPENDIX A - DEFINITIONS 1 Blocking System Also known as a "pure loss" system. Users wishing to place calls when all channels are busy are turned away and must try again later. With telephony it is usually assumed that such calls are lost, because the caller would not try again. With mobile radio this is generally not the case: users repeat call attempts until successful. 2 Busy Hour This is the continuous one-hour period with the highest average traffic intensity, usually expressed in "erlangs". There may be more than one busy period in a day. Traffic intensity during the busy hour is the main factor for assessing the capacity of the system and for identifying of the resources required for a specific "grade of service" (GOS). 3 Call A call is defined as a discrete engagement of a traffic path. 4 Contention This is the situation which arises if, as is usual, all mobiles must use a common channel for initiating calls; it is the problem of giving each contender a fair chance of gaining access to a free channel in accordance with an established rule. 5 Dedicated System A dedicated system has a control channel permanently available for signalling. 6 Erlang Unit of telephone traffic; one erlang is the traffic intensity in a continuously occupied traffic path. Logically, a channel occupied for only half the time in a given period will have a load of 0.5 erlang in that period. 35
7 First-in/first-out (FIFO) In a queueing system, FIFO is the preferred order of service to users waiting in a queue for channel assignment. It is a method which is both fairer to the user and more efficient for the trunked system operation. 8 Grade of Service (GOS) The grade of service is a measure of the quality of the service to the user. There are many possible measures of GOS, but the one adopted in this document is the probability that a user has to wait for more than some specified duration; e.g. GOS of 10% means that one in ten users has to wait for more than 20 s for their call to be set up. 9 Holding Time The channel holding time is defined as the total period for which a channel is engaged by a user for the purpose of sending and/or receiving speech/data, following a channel assignment. This does not include the signalling time required to set up and clear down a call. 10 Load Is a basic concept of telephone traffic theory, which quantifies traffic intensity; it is the product of call rate and mean holding time. It is dimensionless and is measured in erlangs. Channel or system loading is a numeric value quantifying the amount of traffic carried by a given channel or system. It is also measured in "erlangs". 11 Mean Holding Time The mean value of "holding time" during the "busy hour" is used for system planning for a given grade of service "GOS" to the user. 12 Message Trunking The trunking control logic detects the end of a transmission and a short time-out period begins. Any other user receiving on the voice channel at this time can respond, for example, by pressing "push-to-talk". Operation on the assigned voice channel can continue indefinitely in this manner (as long as the time-out is not allowed to expire) until a call participant signals termination of the conversation. 36
13 Overheads (signalling) This is the wasted capacity of the system caused by signalling and waiting for signalling. 14 Polling System A trunked system in which each user is assigned a particular time slot in which to signal if there is a message to be sent. It is an Inefficient method of operation for a large number of users. 15 Protocol An agreed set of conventions in a communication system governing the format and relative timing of messages or signalling, mainly used for setting-up and clearing down a call (see "overheads" above). 16 Queueing System A trunked system in which users wishing to place calls (when all channels are busy) are first placed in a queue and then assigned channels as they become available. The order of service to waiting users in the queue may be random, or it may be according to some special law, or it may be "first-in/first-out" (FIFO). 17 Scanning System A scanning system uses its control channel for speech if all other channels are in use. When this happens, the next channel to become free is marked as the control channel, and users' mobiles find it by scanning. 18 Time-out A time-out is the automatic termination of an activity after a set period of time. For example, this activity may be a stage in the protocol, or it may be the use of a voice channel by the user. 19 Traffic Intensity Traffic intensity is traffic quantity, in one or more traffic paths, per unit time; it is usually expressed in erlangs. 37
20 Transmission Trunking This is a method of reassignment of a channel to another user during idle periods which occur during each message. It is made possible because generally, when speech and data messages are transmitted, there are pauses for a few seconds. If the system signalling speed is fast enough the system processor can recognise these pauses in messages and reclaim the idle channel for reassignment to other users. When the original user is ready to continue, a new channel will be assigned, or the request will be queued if no channels are available. The resultant "choppiness" in the case of speech in this form of operation may be unacceptable to some users, and message trunking" (see above) is then the alternative. 21 Unbalanced System Fleet operation may result in the number of users on a system being small, although the number of individual mobiles may be large. A system in which large and small users are mixed is termed "unbalanced" and in such a system the large users can receive a much better grade of service than the small users. 22 Voice Channel Signalling System In this system, a tree channel, if available, is marked to be the control channel; it is then allocated to the next requested call. 38
APPENDIX B APPENDIX B - REFERENCES 1 Bear, D: Principles of Telecommunication Traffic Engineering. 2 Kuhn, P: 'Tables on Delay Systems', Institute of Switching and Data Techniques, University of Stuttgart, 1976. 3 Ball, D M and Stein, P J: 'A Rapid Access Protocol for Trunked Mobile Radio Systems', IEEE International Communications Conference, Amsterdam, May 1984. 4 Kleinrock, L: 'Queueing Systems, Volume II: Computer Applications' and references therein, John Wiley and Sons, 1976. 5 Schoute, F.C.: Control of Aloha Signalling in a Mobile Radio Trunking System', IEEE conf. on Radio Spectrum Conservation Techniques, London, July 1980. 6 Capetanakis, J I: 'Tree Algorithms for Packet Broadcast Channels', IEEE Transactions on Information Theory, Vol. IT-25, No 5, September 1979. 7 Kleinrock, L and Lam, S S: 'Packet Switching in a Multi-Access Broadcast Channel: Performance Evaluation', IEEE Transactions on Communications, Vol. COM-23, No 4, April 1975. 8 French, R C: 'Error Rate Predictions and Measurements in the Mobile Radio Data Systems'. IEEE Trans. on Vehicular Technology, 1978. 9 Mabey, P J.' 'Predicting the Range and Throughput of Mobile Data Systems' IEEE Vehicular Technology Conference, San Diego, May 1982. 39
APPENDIX C APPENDIX C - COMPUTER PROGRAM FOR ERLANG C FORMULA The Erlang C formula, equation (1) of this memorandum, can be evaluated on a computer by the following short program in "BASIC": 10 DIM F(20) 20 PRINT "NUMBER OF CHANNELS"; INPUT C. 30 PRINT "SYSTEM LOAD (ERLANGS)"; INPUT A 40 LET F(1)=1 50 FOR J=2 TOC 60 LET F(J) =J*F(J-1) 70 NEXT J 80 LET S=1 90 FOR K=1 TO C 100 LET T=A K/F(K) 110 LET S=S+T 120 NEXT K 130 LET P=A C/(A (C+1 )/C+F(C)*(1-A/C)*S) 140 PRINT. DELAY PROBABILITY ";P 150 LET X=EXP(-C-A)) 160 LET PO=P-X 170 PRINT"P(W>20S) =";PO 180 GOTO 20 190 END The program calculates also the probability of any call being delayed for a period in excess of the mean holding time. 40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
End of document 63