Quasi-synchronous digital trunked TETRA performance

708 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999 Quasi-Synchronous Digital Trunked TETRA Performance Alejandro Morán, Fernando Pérez Fontán, Member, IEEE, Jose M. Hernando Rábanos, Member, IEEE, and Manuel Montero del Pino Abstract— In some private mobile radio/public access mobile radio (PMR/PAMR) applications, there is a stringent need for high-coverage locations probabilities. A spectrally efficient approach in this case is the use of several radio transmitters operating in simulcast mode. There have been several analog mobile radio systems working in this way up to now, but less is known about the performance of digital trunked radio systems operating in simulcast mode. In this paper, predicted digital Trans-European Trunk RAdio (TETRA) system performance results operating in quasi-synchronous mode are presented. These results were obtained by simulation of such a system under a wide range of operational conditions. A comparison is also presented with the European analog standard MPT 1327 currently in operation. It has been concluded that quasi-synchronous techniques well known in analog PMR/PAMR can also be successfully used in digital PMR/PAMR applications. Index Terms— MPT 1327, overlap coverage areas, simulcast, TETRA. I. INTRODUCTION S EVERAL area coverage techniques have been traditionally used in mobile radio communication system design to obtain the wanted coverage area with the required locations probability: 90%, 95%, . The most common approach to wide area coverage is multichannel or frequency reuse as in cellular systems. However, other alternatives are available to radio planners such as voting systems, synchronous and quasisynchronous schemes, etc. [1]. In some private/professional mobile radio (PMR) and public access mobile radio (PAMR) applications, for example, security, field services, utilities, etc., large coverage areas may not be required and, in many cases, these networks may even be limited to a single fixed radio station. However, in irregular terrain or in dense urban areas, adequate local coverage cannot be achieved by using a single transmitter. In these situations, a spectrally efficient coverage solution is the use of several transmitters with the same nominal frequency operating in quasi-synchronous or simulcast mode. Up to now there is some experience in the operation of analog quasi-synchronous transmission systems, but less is known about the performance of digital quasisynchronous systems. In this paper, the quasi-synchronous approach is addressed in the framework of the new pan-European Manuscript received February 14, 1997; revised June 16, 1997. A. Morán and F. P. Fontán are with the Department of Communications Technologies, University of Vigo, E-36200 Vigo, Spain. J. M. H. Rábanos and M. Montero del Pino are with the Telecommunications Engineering School, Polytechnic University of Madrid, Madrid, Spain. Publisher Item Identifier S 0018-9545(99)04021-9. digital trunked system standard Trans-European Trunk RAdio (TETRA) [2]. Performance characteristics can be obtained theoretically by a pure analytical approach, but this is fairly abstract and complex. Thus, it has been decided to carry out simulation studies which are easier to implement and give a physical insight of the simulcast system operation. II. BACKGROUND Quasi-synchronous transmission using two or more radio stations is a technique used to improve area coverage probability when a single transmitter is insufficient. A quasisynchronous transmission system basically consists of the use of two or more radio stations transmitting the same signals toward the desired service area using the same radio channels. Fig. 1 illustrates a six-cell radio system using a multifrequency scheme and where one of the cells is not adequately served by a single radio station due to signal blockage problems. In this case, in order to fill in the coverage gaps left by the main or master radio station in the cell two more slave quasisynchronous stations are used to improve coverage quality. The probability that two or more stations simultaneously experience deep shadowing or blockage is greatly reduced. Small frequency offsets (a few hertz to a few tens of hertz) are allowed between the transmitted radio frequency (RF) carriers at the different coverage radio stations. On the other hand, the uplink (mobile-to-base) operates as a voting system where the best received signal at the different radio stations is selected. The main aim of a quasi-synchronous system will thus be to achieve macrodiversity as is shown in Fig. 2. The use of the same radio channels at the different fixed stations provides the added advantage of frequency economy. No extra frequency assignments are required for local coverage. When one of the received signals is greater than the others, no cochannel interference is experienced since this stronger signal dominates over the rest due, for example, to the capture effect in FM systems. However, in the overlapping zones (Fig. 3), i.e., where the mobile receives approximately the same signal levels from more than one transmitter, problems may appear, and, hence, special care must be taken to ensure low error rates. The reception of similar amplitude signals on the same carrier at the mobile causes interference problems. Added to the received multipath structure caused by the surrounding clutter there will be another source of multipath due to the transmission of the same signal on the same nominal RF carrier frequency from two or more radio stations. Fig. 4 illustrates 0018–9545/99$10.00  1999 IEEE MORÁN et al.: QUASI-SYNCHRONOUS DIGITAL TRUNKED TETRA PERFORMANCE Fig. 1. Multifrequency area coverage scheme combined with a simulcast cell. Fig. 2. Macrodiversity gain effect for irregular terrain areas in a quasi-synchronous system. Fig. 3. Capture zones and overlap zones in a quasi-synchronous system. Fig. 4. “Artificial” multipath effect due to the simulcast operation. 709 710 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999 Fig. 5. Two-station simulcast system configuration. the effect of the reception of two equal amplitude carriers on the same radio channel. Multipath-like fades appear on the overall received signal giving rise to noise bursts where the bit error rate (BER) is high. Due to the slight frequency offset allowed between the RF carriers, the deep nulls caused will change position with time. A static terminal will observe a slowly fading signal. The quasi-synchronous technique has been successfully used in numerous applications ranging from simple repeater systems with subaudio tone signaling [3] to analog trunked systems (MPT 1327) [4]. In a quasi-synchronous system, digitized voice and data are sent from the trunking system controller (TSC) to the transmitters via land lines or point-to-point radio links. Such lines may have different lengths as shown in Fig. 5. These line length differences give rise to different delays. Similarly, there will be different propagation delays from the radio stations to the mobile receiver as is also shown in Fig. 5 if the mobile is closer to one of the radio stations. Both line delay and radio propagation delay plus other delay sources are accumulated in the phase of the received signal. These delay differences can be critical to the performance of the system, thus, appropriate delay equalization is required. Another critical parameter that will be paid attention to is the offset between base-station RF carrier frequencies. RF carrier nominal frequencies may slightly drift between maintenance adjustment periods. However, a maximum offset limit must never be exceeded if adequate system performance is to be preserved. In this paper, second-generation European digital trunked system TETRA [5] performance studies are presented for systems using a quasi-synchronous approach to enhance local coverage. Both bit and codeword error rates (CWER’s) have been assessed for the different logical channels defined in the TETRA standard. III. BRIEF OVERVIEW OF THE TETRA SYSTEM [5] TETRA is an European Telecommunications Standardization Institute (ETSI)-defined system for PMR/PAMR applications with far more enhanced features than the existing analog standards, i.e., MPT 1327 (UK DTI). Two versions of the system have been defined: 1) TETRA voice data; 2) TETRA packed data optimized. data version are The characteristics of the TETRA voice reviewed. A large number of bearer services and teleservices have been defined providing the opportunity to implement a wide range of communication applications. The system uses a frequency-division multiple-access (FDMA) structure with 25-kHz RF channels both in the uplink and downlink directions. On the other hand, each RF channel implements a time-division multiple-access (TDMA) structure supporting four logical channels (for voice, data, or signaling). A gross bit rate of 36 kbps and filtered /4shift-DQPSK modulation are used. In order to adequately reject adjacent channel power, limit intersymbol interference, and ease receiver synchronization a raised cosine filter is employed with a rolloff factor of 0.35. RF bursts with the general structure shown in Fig. 6 are fitted into each of the four TDMA time slots. The uplink bursts are preceded by a preamble used for power ramping and power amplifier linearization followed by a postamble for power ramping. In the downlink transmission is continuous, and this means that no power ramping time is needed and the available extra time intervals are used to broadcast an additional training sequence between downlink bursts in order to improve reception quality in the mobiles. In the system there exist seven types of bursts which are fitted into time slots making up a frame structure. The bursts MORÁN et al.: QUASI-SYNCHRONOUS DIGITAL TRUNKED TETRA PERFORMANCE 711 Fig. 6. TETRA system general burst scheme. Fig. 7. Error protection schemes for the different TETRA logical channels. types defined in the TETRA standard are the following: 1) control uplink; 2) linearization uplink; 3) normal uplink; 4) normal continuous downlink; 5) synchronization continuous downlink; 6) normal discontinuous downlink; 7) synchronization discontinuous downlink. Additionally, a multiframe structure of 18 frames is defined which allows the introduction of associated control channels together with their corresponding traffic channels and a hyperframe to facilitate the monitoring of adjacent cells by the mobile and accommodate a cryptographic scheme. Two basic types of logical channels have been defined: 1) traffic channels, carrying speech or data in circuit switched mode; 2) control channels, carrying signaling messages and packet data. Different traffic subchannels are defined for speech or data applications with several data rates: 1) speech traffic channel (TCH/S); 2) speech or data traffic channels a) 7.2-kbps net rate (TCH/7.2); b) 4.8-kbps net rate (TCH/4.8); c) 2.4-kbps net rate (TCH/2.4). There are five categories of control channels. 1) Broadcast control channel (BCCH), comprising the following. a) Broadcast network channel (BNCH). b) Broadcast synchronization channel (BSCH). 2) Linearization channel (LCH), with two subchannels. a) Common linearization channel (CLCH) shared by all the mobiles in the uplink direction. b) Base-station linearization channel (BLCH) downlink used by the base station. 712 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999 Fig. 8. Detailed TETRA simulcast system simulation layout. 3) Signaling channel (SCH) shared by all the mobiles, which is further divided into three categories depending on the size and direction of the messages. 4) Access assignment channel (AACH) downlink. It is used to indicate the assignment of the uplink and downlink slots. 5) Stealing channel (STCH), which is bidirectional. It is associated with a TCH and temporarily “steals” a part of the TCH capacity in order to transmit control messages when fast signaling is required. In Fig. 7, the coding schemes used for the different channel types are summarized. IV. SIMULATION SCHEME In order to carry out simulations, the transmitter and receiver blocks were implemented following the TETRA specifications [2]. Error probabilities were computed by means of an errorcheck block. In order to account for multipath propagation effects and quasi-synchronous transmission, a channel block was introduced in the simulations as is shown in Fig. 8. Two quasi-synchronous transmitters were simulated and only the downlink (mobile reception) was studied. The transmitter block was fed by a PRN sequence generator. Several block and convolutional coders were implemented according to Fig. 7. The modulation scheme was /4-shift DQPSK with a gross bit rate of 36 kbps in a 25-kHz RF channel bandwidth. A square-root-raised cosine filter with a rolloff factor of 0.35 was placed after the modulator. At the receiver, a pass-band filter was placed right before the demodulator. A simple receiver without an equalizer was implemented. The phase of the incoming signal was integrated over a symbol period in order to compute its change during TABLE I the last s (symbol period). Only four-phase change values are allowed in the /4-shift-DQPSK modulation scheme. In order to compute the received bits, a decision device with the appropriate thresholds was placed after the sampler. Ideal synchronization was assumed. In the simulations, a 10-m/s speed (36 km/h) was considered for the mobile receiver. Quasi-synchronous operation is defined by three parameters, namely, the relative delays between modulating signals due to land line and propagation delays, the relative received amplitudes, and the frequency offset of the carriers. The transmitter output was fed to two channel blocks representing the signals received from two fixed radio stations. These signals were added at the receiver. Finally, Gaussian noise was also added. Narrow-band channel propagation conditions were considered. This assumption is acceptable when the multipath delay spread is much smaller than the inverse of the signal bandkHz and 1/BW s. Typical width, in this case, BW delay spread values will hardly reach a few microseconds as shown in Table I [5]. From the table, it is seen that the narrowband assumption holds at least for rural and typical urban areas. The channel amplitude and phase variations were generated using a simple geometrical model [6], [7]. The model produces the complex envelope variations due to multipath. Typically, a Rayleigh probability density function will be followed by the received amplitude when the direct signal is not available. In the frequency domain, the classical U-shaped Doppler MORÁN et al.: QUASI-SYNCHRONOUS DIGITAL TRUNKED TETRA PERFORMANCE 713 (a) Fig. 9. BER versus Eb =No . Rayleigh channel (- - (a) Delay = 0 s. (b) Delay = 3:5 s (6% of Ts ). (b) ), simulcast operation: spectrum is present. If the direct ray is considered, a Rice distribution will characterize the amplitude variations. The model assumes that the mobile is surrounded by a crown of point scatterers with a uniform azimuth distribution. These scatterers are illuminated by a distant transmitter. The signal is scattered at each point in the crown before it reaches the mobile receiver through multiple paths. A direct ray may be considered at will, thus, producing Rice or Rayleigh distributions. The total received signal is the coherent sum of all scattered rays plus the direct ray, if it exists. Channel samples along the mobile route are produced with a spatial , a typical value would be separation of less than This spacing provides sufficiently close samples so that no received signal deep nulls are lost in the sampling process. In order to carry out transmission system performance simulations a travelled distance-to-time conversion must be carried out according to the assumed vehicle speed. Rayleigh and Rice fading channels are slowly variant for moderate vehicle velocities when relatively low binary transmission rates are considered. This means that a number of transmitted symbols will “see” approximately the same channel amplitude and phase. Eight samples per symbol were used in the simulations. In order to match this sample rate, the same number of channel (amplitude and phase) samples per symbol were produced by means of simple interpolation out of the simulated channel samples having a spatial separation of The simulations presented in this paper were carried out using the Ptolemy (University of California at Berkeley) simulation package environment. The carrier frequency was set to 400 MHz close to one of the bands internationally foreseen for allocation to the TETRA system [5]. The mean amplitude ratio in decibels of the two incoming signals was A = 10 dB (- - -), A = 3 dB (- - - -), and A = 0 dB (-1-1-1-1). Fig. 10. BER versus frequency offset of a Eb =No = 35 dB. A = 0 dB (- - - -), A = 3 dB (- - - -), A = 5 dB (-1-1-1), and A = 10 dB (1 1 1 1 1): 714 Fig. 11. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999 BER versus frequency offset of the 2.4-kbps TCH. A = 0 dB (- - - - -), TABLE II TABLE III set to a range of to dB in order to simulate the system performance in the overlapping coverage areas of two transmitters. RF carrier offsets up to 3000 Hz and delays up to one half of the symbol period were studied. The assumption was made that the location of the receiver was midway between the two transmitters. This means that the delay difference in the plots is due fundamentally to unequalized line delay differences. A = 3 dB (- - -), and A = 10 dB (1 1 1 1): V. RESULTS levels for Fig. 9 shows BER values for different a TETRA system channel before error protection decoding assuming a Rayleigh channel. Two cases are shown in the figure [see Fig. 9(a)], where the relative line propagation delays of the two received signals are zero and [see Fig. 9(b)], where the relative delay is 3.5 s (6% ). The parameter in the figure is the relation in logarithmic units between the two quasi-synchronous signals at the input of the receiver. From ratio grows, Fig. 9, it can be observed how, as the the curves corresponding to the Rayleigh channel ( dB, i.e., one signal dominates) depart from those belonging to quasi-synchronous operation ( 10 dB: and dB). The reason for this is that as the transmitted power is increased, the Rayleigh effects are partially mitigated, but not the interference effects due to the simulcast operation. ratio For the simulation results shown below, the was set to a “good” working level of approximately 35 dB, MORÁN et al.: QUASI-SYNCHRONOUS DIGITAL TRUNKED TETRA PERFORMANCE Fig. 12. BER versus frequency offset of the 4.8-kbps TCH. - -), A = 3 dB (- - - -), and A = 10 dB (1 1 1 1): A 715 = 0 dB (- - a point where the transmissions are mainly impaired by the quasi-synchronous operation effects. This selection is justified due to the fact that interference between the two stations will be more likely in mutual visibility areas. A Rayleigh model was assumed in the simulations since it represents the worst case situation. In Fig. 10, results are presented for different delay (Table II) frequency offset combinations. In the figure it is clearly shown how the BER deteriorates dramatically with increasing frequency offsets and delays, especially when the two received dB). signals present similar amplitudes ( A. Traffic Channels (TCH) Results TETRA traffic channels (TCH’s) use data protection schemes based on punctured convolutional codes. In a Gaussian channel, the performance of a convolutional code can be obtained as a function of the raw BER (before channel coding) and the properties of the code. Although the channel found in a mobile communications environment is not Gaussian at all, the interleaving scheme employed, allows the use of this approach to quantify BER’s in the TCH’s. The approximation used here was [8]–[10] Fig. 13. WER versus offset plots for relative delays T = 0; T s=16; T s=8; and A = 0 dB (- - - - -), A = 3 dB (- - - -), A = 5 dB (-1-1-), and A = 10 dB (1 1 1 1): where is a constant depending on the code and is the values used are free distance of the code. In Table III, the shown. The results obtained for the 2.4- and 4.8-kbps traffic channels are shown in Figs. 11 and 12 for the delay values in Table II. The solid horizontal line represents the threshold performance level set by the TETRA standard for each channel and ). A strong influence of the type (BER relative delay (line delay) can be observed. Once a given is exceeded, large delay value of approximately 10% of BER values were observed no matter what the carrier offset value was. It was also observed that for moderate values of the delay the BER performance degrades rapidly when the frequency offset exceeds 100 Hz. B. Access Assignment Channel Results Separate simulations were carried out for the access assignment channels (AACH) since, instead of a convolutional code, it uses a Reed–Muller (30,14) block code. This code presents a minimum Hamming distance of eight, and, thus, it can detect up to seven errors and correct up to three errors. The word error ratio (WER) was studied first for different frequency offsets and delays. Fig. 13 shows three WER versus 716 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999 Fig. 14. (1 1 1) Error probabilities versus number of errors in the 30-b codewords for offsets: 10 Hz (- - - - -), 30 Hz (- - -), 100 Hz (-1-1-1), and 300 Hz for the different delay-relative amplitude ratios in Table IV. TABLE IV frequency offset plots for different delays and amplitude relation parameter values. It can be observed that for offsets smaller than 100 Hz, WER values remain practically constant. It can also be observed that the delay is the most sensitive parameter when setting up a simulcast radio network. The WER is not, however, a relevant parameters provided that the number of errors in a codeword do not exceed the detection capabilities of the block code used. This can be seen from Fig. 14 where the error probabilities in a 30-b codeword are given for the parameters shown in Table IV. From Fig. 14, it can be concluded that as the frequency offset increases the probability of having a larger number of errors in the 30-b codewords decreases. That is, as the offset grows the WER also grows, however, these codewords will contain a smaller number of bits in error. Fig. 15 clarifies this. In the figure, two error time series are presented. These series correspond to the transmission of 20 000 30-b codewords for a relative delay of 7 s (12.5% ), dB, and frequency offsets of 10 and 300 Hz. It can be observed that for a 300-Hz offset, errors are distributed more uniformly among transmitted codewords, and although the number of erroneous words increases, the average number of errors per codeword is smaller. In this way, the probability of exceeding seven errors in a codeword is smaller for larger offsets. In the simulated case, the probability of exceeding 10 for a 300-Hz offset while for a seven errors is 1.5 10 . 10-Hz offset is 5.2 The transmission performance of the AACH channel must be assessed in terms of two error parameters: the message erasure ratio (MER) or probability of receiving and detecting an erroneous message and the probability of undetected MORÁN et al.: QUASI-SYNCHRONOUS DIGITAL TRUNKED TETRA PERFORMANCE Fig. 15. Error time series for T = 7 s  12.5% T s, A = 10 717 dB, and frequency offsets of 10 and 300 Hz. Fig. 17. PUEM versus offset for relative delays T = 0; T s=16; T s=8; and A = 0 dB (- - - - -), A = 3 dB (- - - -), and A = 10 dB (-1-1-). Fig. 16. MER versus offset for relative delays T = 0; T s=16; T s=8; and A = 0 dB (- - - - -), A = 3 dB (- - - -), and A = 10 dB (-1-1-). erroneous message (PUEM) or probability of receiving an erroneous message and mistaking it for a legitimate one. Figs. 16 and 17 show several MER and PUEM versus frequency offset plots for different delays and amplitude rations. It can be observed how the MER and the PUEM present opposite behaviors as the frequency offset increases. 718 Fig. 18. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999 MPT 1327 forward control channel time slot structure and codeword structure. For offset values below 100 Hz, the MER increases slowly whereas the PUEM does the opposite. For higher offset values, sharper variations of both parameters are observed. As for the relative delay, it is again verified that this is the most sensitive parameter. Values on the order of can still be tolerated, but for higher values performance is seriously impaired. TABLE V C. MPT 1327 System Forward Control Channel Results (Analog Trunked System) It is interesting to compare the performance of a digital system such as TETRA with an analog system such as the MPT 1327 system which is considered as a de facto standard in Europe for trunked radio networks. In order to present similar simulations to those already presented for the TETRA systems, first a brief overview of the MPT 1327 standard must be made. This is an analog system using FM modulation in the traffic channels which are designed specially for voice applications. As for the control channels the data flow with a bit rate of 1200 b/s undergoes a double modulation process. First, an audio frequency carrier is frequency-shift-keyed (FSK) modulated (a binary “0” is represented by the frequency 1800 Hz and the binary “1” by 1200 Hz), and then a FM modulation is followed. Other modulation characteristics can be found in references such as (UK) DTI’s MPT 1327, MPT 1343, and other associated documents. Two-frequency channels of 12.5-kHz bandwidth have been assumed in the simulations since this is the most common channel separation in the bands around 400 MHz for trunked applications. For TETRA, however, a channel bandwidth of 25 kHz was considered. The MPT 1327 forward control channel provides the time reference for a slotted Aloha multiaccess scheme. Transmissions in the downlink are arranged in 106.67-ms time slots. Two 64-kb codewords are sent in each time slot. The first word is called control channel system codeword (CCSC) which identifies the system to the mobile terminals and provides the required synchronization for the reception of the second word in the slot: address codeword (AC) (Fig. 18). Additionally, if required, a time slot may contain two data codewords which are used to send short data messages across the control channel. Also, in Fig. 18 the codeword structure used in the MPT 1327 system is presented. Bits 2–48 carry information whereas bits 49–64 carry data protection bits generated with a (63, 48) block code using the following generating polynomial: where the 64 codeword is completed by adding one single bit in order to produce an even parity codeword. The decoding algorithm is not specified in the standard, however, several options are possible. Table V lists the options studied in this paper. Similarly to what was done for the TETRA system, simulations are shown where the influence of three major installation parameters are assessed: the frequency offset, the relation between the received amplitudes , and the relative delay which, expressed as a percentage of the symbol period in this case is 833 s. A preliminary BER study was carried out for different parameter (10, 3, and 0 dB) and the delay values of the s ). These simulations are parameter ( and and shown in Fig. 19, where the Rayleigh case is compared with three simulcast Rayleigh situations. It can be observed how the simulcast operation greatly deteriorates the performance of the system. As for the TETRA case, the working point used to carry out the evaluation of the performance reduction effects for different offsets, delays and relative received levels was set to approximately 35 dB. Again, it was deemed that simulcast effects are more important in the mutual visibility areas from both transmitters where a high signal level is received from both transmitters. In Fig. 20, the simulation setup for the MPT 1327 analog trunked system is presented. Before proceeding to analyze the MORÁN et al.: QUASI-SYNCHRONOUS DIGITAL TRUNKED TETRA PERFORMANCE 719 (a) Fig. 19. BER versus C/N. Rayleigh channel ((b) Delay T = 26 s (3% Ts). Fig. 20. - -), A = 10 (b) dB (- - - - -), A = 3 dB (- - -), and A = 0 dB (-1-1-). (a) Delay T = 0 s. Detailed MPT 1327 simulcast system simulation layout. composite behavior of the modulation plus channel coding scheme used in the forward control channels, row BER studies were performed. The simplest receiver structure was assumed for the receiver (Fig. 21). The phase-locked loop (PLL)-based FM demodulator was modeled by a differential equation [11]. After the PLL, a symbol period integrator was implemented providing at its output the phase increment produced. A comparison device was placed at the end of the receiving chain. Ideal synchronization recuperation was assumed. In Fig. 22, results for different offsets and relative amplitudes are shown. These results were calculated for the working Fig. 21. MPT 1327 receiver block diagram. dB and for the delays given in point selected of C/N Table VI. From the observation of Fig. 22, it can be concluded that the BER is strongly dependent on the frequency offset. Values 720 Fig. 22. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999 BER versus frequency offset. A = 0 dB (- - - - - -), A = 3 dB (- - - -), A = 5 dB (- - - ), and A = 10 dB ( TABLE VI higher than 30 Hz produce a great deterioration of the BER As for the delay, the BER with values well above 10 maintains acceptable levels if the delay does not exceed 10% of , and past this value transmission performance is greatly impaired. The BER parameter does not allow a complete understanding of the transmission chain performance under simulcast conditions. It is important to study the error distribution within each 64-b codeword. Fig. 23 presents the word error , distribution for the following conditions: BER dB, and delay s In spite of the high BER value considered, the probability of having more than three 1 1 1 1 1 1 1 1 1) for the delays in Table VI. errors in each 64-b codeword is smaller than 6%. For the (64, 48) code used, 3-b error words can be handled without difficulty. The WER was also studied for different conditions. The results are shown in Fig. 24. However, the WER is not a significant parameter since the capabilities of the decoder must be accounted for. In spite of having large WER values, if the number of errors in each codeword is small enough overall performance will be maintained. The next step was to perform a study including the coder properties. Table V gives several decoder implementation options. Fig. 25 presents CWER simulation results for the different coder implementations in Table V and for several simulcast conditions (offsets, delays, and relative signal levels). The left column plots in Fig. 25 represent the probability that a codeword in error is detected and the right column plots represent the probability of mistaking a codeword in error for a correct one (false alarm). In each plot, four lines are shown presenting the following s ), (5 Hz, s offset-delay combinations: (5 Hz, ), and (50 Hz, 26 s) and (50 Hz, 52 s). It can clearly be observed how as the correcting power of the decoder is MORÁN et al.: QUASI-SYNCHRONOUS DIGITAL TRUNKED TETRA PERFORMANCE Fig. 23. 721 Distribution of the number of error in a codeword. (a) (b) Fig. 24. WER versus frequency offset for (a) and A = 10 dB (1 1 1 1): T = 0 s and (b) T = 52 s. A = 0 dB (- - - - - -), A = 3 dB (- - - -), A = 5 dB (-1-1-1-), 722 Fig. 25. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999 CWER probability versus A parameter for offsets: 5 Hz (- - -), 50 Hz (- - - - - - -), and delays increased (from top to bottom of the figure), the probability of words in error decreases. On the other hand, the probability of mistaking an erroneous word for a good one increases. Only for the last case (soft decoding) much more powerful and complex a good balance between both error conditions is dB only achieved: for example, for an amplitude ratio 2% of the transmitted codewords are in error and only 0.5% are mistakenly interpreted as legitimate codewords. It must be remembered that these results correspond to nonequalized conditions. From the analysis of the figure, it is clear that as the frequency offset increases the error distribution amongst transmitted codewords becomes more uniformly distributed, that is, although the number of codewords in error increases the number of these words that present a number of errors above the detection threshold is smaller. Thus, moderate offset values facilitate the detection process. However, it has also been observed that for offset values above 50 Hz, the system performance deteriorates rapidly. Other factors to be borne in mind are the usual conditions under which radio transmissions will be performed. If termi- T = 26 s and T = 52 s. nals are usually in motion when making calls other factor to be taken into account which increases the relative offset between simulcast transmissions is the Doppler shift (that is velocity dependent). The worst case will be when a mobile travels along the straight line defined by both stations. Another factor to be taken into account is that FM-modulated voice transmissions (traffic channels) present different sensitivity to delays and frequency offsets. It has been experimentally observed that for equalized paths carrier offsets must never exceed a few Hertz [12]. A possibility is to setup the installation of analog systems with different offsets for control and traffic channels. VI. CONCLUSIONS In this paper, digital pan-European trunked system standard TETRA performance has been evaluated when a quasisynchronous transmission scheme is used to improve area coverage probability. Two relevant conclusions may be drawn from this study. For one, the delay differences due to different transmission line lengths from the trunking system controller MORÁN et al.: QUASI-SYNCHRONOUS DIGITAL TRUNKED TETRA PERFORMANCE to the simulcast stations severely degrade the performance of quasi-synchronous systems. This means that adequate delay equalization is required. Maximum relative delays of 12 s (20% ) and carrier frequency offsets of 100 Hz were shown to be acceptable. Second, it has also been shown that an appropriate selection of the carrier frequency offset value makes it possible to counteract the adverse characteristics of the fast fading channel and the quasi-synchronous interference. Adequate offsets will whiten the sequence of errors and facilitate the task of the error protection code. A maximum frequency offset of about 100 Hz can be permitted in typical quasi-synchronous installations. A similar study was carried out for the forward control channel of the European analog MPT 1327 system using 12.5kHz-bandwidth channels and a double modulation process for FM. This system the signaling data: audio subcarrier FSK was shown to be more vulnerable to simulcast transmission conditions than the TETRA system. Delays exceeding 10% or carrier frequency offsets exceeding 50 Hz produce unacceptable system performances. It can be concluded that quasi-synchronous techniques which have been successfully implemented in the past for analog PMR applications including analog trunked systems, i.e., MPT 1327, can also be applied to second-generation digital pan-European trunked systems in order to benefit from the improved coverage probabilities and enhanced spectral efficiency obtained by this macrodiversity scheme. REFERENCES [1] R. J. Holbeche, “Area coverage techniques,” in Land Mobile Radio Systems, R. J. Holbeche, Ed. (IEEE Telecomms. Series no. 14), ch. 2, 1995. [2] Trans-European Trunk Radio (ETSI), “Documents 05.01, 05.02, 05.03, 05.04, 05.05, and 05.08,” Nov. 1993. [3] G. D. Gray, “The simulcasting technique: An approach to total-area radio coverage,” IEEE Trans. Veh. Technol., vol. VT-28, pp. 117–125, May 1979. [4] BS 770 MPT 1327 VHF-UHF Repeater-Base Station, Bosch, 1996. [5] F. Gourgue, “Air interface of the future European fully digital trunk radio system,” in IEEE Veh. Technol. Conf., 1993, pp. 714–717. [6] F. Perez-Fontan, A. V. Castro, and J. P. V. Poiares Baptista, “A simple numerical propagation model for nonurban mobile applications,” Electron. Lett., vol. 31, no. 25, Dec. 1995. [7] F. P. Fontan, J. Pereda, M. J. Sedes, M. A. V. Castro, S. Buonomo, and P. Baptista, “Complex envelope three-state Markov chain simulator for the LMS channel,” Int. J. Satel. Commun., vol. 15, no. 1, pp. 1–15, Jan./Feb. 1997. [8] S. Benedetto, E. Biglieri, and V. Castellani, Digital Transmission Theory. Englewood Cliffs, NJ: Prentice-Hall, 1987. [9] D. Haccoun, “High-rate puncturing convolutional codes for Viterbi and sequential decoding,” IEEE Trans. Commun., vol. 37, pp. 1113–1125, Nov. 1989. [10] J. Bibb, G. C. Clark, and J. M. Geist, “Punctured convolutional codes of rate n-1/n and simplified maximum likelihood decoding,” IEEE Trans. Inform. Theory, vol. 25, Jan. 1979. [11] M. C. Jeruchin, P. Balaban, and K. S. Shamugan, Simulation of Communications Systems. New York: Plenum, 1992. [12] F. Bueno and F. P. Fontan, “Simulcast systems engineering,” Mundo Electŕonico, no. 261, Nov. 1995 (in Spanish). 723 Alejandro Morán was born in La Corũna, Spain, in 1972. He received the Telecommunications Engineer degree from the University of Vigo, Vigo, Spain, in 1996. He is currently working towards the Ph.D. degree in the field of multicarrier spreadspectrum communications at the University of Vigo. Fernando Pérez Fontán (M’96) was born in Vilagarćia de Arousa, Spain, in 1959. He received the Ph.D. degree from the Polytechnic University of Madrid (UPM), Madrid, Spain, in 1992. He is a Lecturer at the Telecommunications Engineering School, University of Vigo, Vigo, Spain. Jose M. Hernando Rábanos (M’94) received the Ph.D. degree in telecommunications engineering from the Polytechnic University of Madrid (UPM), Madrid, Spain. He is a Professor in the Signals, Systems and Radiocommunications Department, UPM. He has written two textbooks on radio transmission and mobile communications. His professional activity is devoted to teaching and research in radiocommunications. He led research projects in radio planning and radio propagation. He has published several technical papers in international journals and cooperates with radio propagation and mobile services ITU-R study groups. Manuel Montero del Pino was born in LilloToledo, Spain, in 1936. He received the Telecommunications Engineering degree and the Ph.D. degree from the Polytechnic University of Madrid (UPM), Madrid, Spain. He was with Iberia Airlines as a Communications Department Head and with Telefonica as Systems Department Director. He is the author of numerous telecommunications and control articles, books, and patents. He has participated in several ITU-R and ITU-T working groups. Currently, he is a Professor at the Telecommunications Engineering School, UPM.