oppler Shit Estimation or TETRA Cellular Networks Javad Ashar Jahanshahi, Seyed Ali Ghorashi epartment o Electrical and Computer Engineering, Shahid Beheshti University, Tehran, Iran j.ashar@mail.sbu.ac.ir, A_ghorashi@sbu.ac.ir Abstract Estimation o mobile terminal speed at base station in cellular networks helps BTS in many aspects including channel estimation, adaptive reception, anti-jamming and handover operations. In this paper, we introduce a new algorithm to estimate the channel oppler shit seen by BTS, using the measured received signals at base station. We have improved an LCR based oppler shit estimation algorithm by using only inherent inormation which are available in common receivers without any excessive hardware. The application o the proposed algorithm in a TETRA network is modeled and simulation results have shown a good perormance in a wide range o velocities. Key words: TETRA (Terrestrial Trunked Radio), oppler shit, I. Introduction The speed o a mobile terminal in wireless communication networks is an important piece o inormation that can improve system perormance in many ways. Knowing the speed o the mobile terminal enables the receiver to perorm more eicient channel estimation. Similarly, in adaptive transmission, it helps the transmitter to adjust a suitable modulation/coding scheme according to the channel condition. In addition, based on the speed inormation o the mobile terminals, handover time can be determined more accurately. Speed inormation can also be used in anti-jamming techniques, when the receiver tries to dierentiate between signal attenuations caused by jamming and channel eects [1]. eanwhile, speed estimation by additional sensors like gyroscopes or accelerometers, and systems like GPS (Global Positioning System), increases the complexity and overall costs o the user terminals, and urthermore, reduces the handset battery lie time. Thereore, several techniques have been proposed in literature or mobile terminal speed estimation based on channel oppler shit measurement, and some o them have been implemented in existing mobile communication systems. Covariance estimation schemes estimate oppler requency shit by computing covariance value between training received samples [2-6]. Other schemes or oppler requency shit estimation have used spectral analysis and variance [7], estimation o channel envelope and angle [8], statistical inormation o channel phase variations [9], Eigen based spectral estimation [10], spectrum estimation method based on channel power spectrum density [11], multi vector test by using maximum likelihood approaches [12], wavelet analysis by tracking changes in the temporal scale [13] and channel autocorrelation [14]. In [15], authors proposed a LCR based algorithm that estimates terminal s speed over each speed estimation window, and consequent windows do not overlap each other. They also used a single threshold or signal power comparisons. However, the proposed algorithm in [15] cannot ollow the mobile terminal s variation in low SNR conditions. In this
paper the oppler shit estimation algorithm is improved by utilizing a speed estimation window that slides over bursts with overlaps and by introducing two dierent low and high thresholds or power level comparisons. These thresholds are updated or each speed estimation window s movement to better tracking the oppler shit variations even in low SNR conditions. This algorithm uses only inherent cellular system inormation, which means there is no need or any hardware modiication o the user terminal and also cellular network signaling structure. The proposed algorithm is modeled in a TETRA network and simulation results show an acceptable accuracy in oppler shit estimation in Rician channels. This paper is organized as ollows. Section II gives a brie overview o TETRA networks particularly ocusing on its basic rame structure. In section III we present the proposed algorithm or oppler shit estimation in details, and in section IV, we present simulation results showing the perormance o the proposed algorithm in a TETRA Network and compare our proposed algorithm with Hong Algorithm [15]. Finally, in section V, we provide our concluding remarks. I.TETRA Frame Structure Radio transmission in TETRA network utilizes time division multiple access (TA) as the channel access method and a 14.17m sec time slot is allocated to each user [16]. The rame, ultirame and Hyperrame structures are shown in Fig 1. Fig. 1 Transmission in TETRA Networks [16]. Four successive time slots are grouped to a TA rame. In each time slot, one type o three possible transmission bursts is transmitted in the uplink and downlink terminals. Each burst contains 255 bits. Each time slot may contain one or two bursts according to burst s ormat. The Table 1 lists the main types o the bursts and provides a brie summary o their properties and purposes. Table 1: Basic Bursts Format o TETRA Purpose and irection Uplink control burst Uplink normal burst Uplink linearization burst ownlink normal burst ownlink synchronization burst ownlink dummy burst Format and Speciic Task Hal slot For random and reserved access Normal S-to-BS ull slot ormat burst ater initial system access during ongoing calls Empty ull slot ormat, For S transmitter Power Ampliier (PA) linearization, needed only while tuning to a new carrier requency Normal ull slot ormat BS to S burst during ongoing calls Full slot ormat burst For S synchronization and RF ine tuning Full slot ormat burst For enabling mobile terminals to detect that a requency is in use The so-called normal burst (NB) is used or active transmission. I there is nothing to transmit, BTS has to send a so-called dummy burst (B), enabling mobile terminals to detect that a requency is in use. For time synchronization purpose, the so-called synchronization burst (SB) is sent to allow an accurate timing recovery. Linearization burst is used by Ss to linearize their transmitter. The last burst type, the so-called requency correction burst (FB) consists o an unmodulated carrier with a deined
requency oset. Both last two bursts can only be transmitted in the irst time slot o a rame [17]. Fig. 2 the structure o speed estimation window. Note that speed estimation can only be carried out ater receiver synchronization. Ater synchronization process, mobile station color code (CC), rame numbers (FN) and time slot number (TN) are reached. Temporal position o the bursts and identity o the chosen training sequences are obtained by considering CC, FN, TN. Thereore speed estimation can be carried out. III. Proposed Algorithm Fig. 2 shows the structure o received complex samples over one speed estimation window (i.e. 0.5 or 1 second). This window slides over samples o received signal. Each window divided into N groups o samples where: N = [ ] (1) here [.] is the rounding down operator, is the number o samples within a speed estimation window and is the segmentation actor. The segmentation actor will be updated or each received burst. The lowchart o the proposed oppler shit estimation algorithm is illustrated in Fig.3. At the irst stage, the power o the received signal is calculated. This power is measured within a ixed size speed estimation window. Then, the power meter computes group powers S ( i), i = {1, 2,3,..., } i* 1 2 S( i) =. s( z) ( n ) (2) z= ( i 1)* + 1 where S ( i) is the power o samples over the i th group. In the third stage, RS meter computes the root mean square o group powers during speed estimation window as: 1 RS =. S ( i ) i= 1 2 (3) The calculated RS is then used to determine low and high thresholds or level crossing calculations. Then, high and low level crossing thresholds T H and T L are calculated. These thresholds should be
ractions o the RS value calculated in previous stage: T T H L = y. RS = x. RS o < x < y < 1 (4) ρ where T + T RS 2 k + ( k + 1). ρ LR. e = ρ. 2 π ( k + 1). I (2 ρ. k( k + 1)) 0 H L = (6) denotes the oppler shit, k denotes the Rician ading actor, I 0 represents the modiied zero order Bessel unction, and e is Euler s number. The oppler shit o each received burst is given by its angle o arrival θ, the carrier requency, the n propagation speed C (which is the speed o light), and the mobile terminal speed v. It can be calculated as: c c = v..cos( θn ) C (7) For the maximum value o the oppler shit, the mobile terminal speed is given by v C =. max (8) c For example, when signaling is done with = 400Hz in a typical value or a TETRA c system, 100 km h terminal speed results in maximum Figure 3 Signal Flow o the Proposed Algorithm. In the ourth stage, level crossing counter counts level crossing requency L, which indicates how many times group powers S ( i ) cross thresholds T H and T L in positive slope. In Rician ading channel with 2-dimensional isotropic scattering, the oppler shit is given by [18]: R oppler shit o max = 37Hz. The estimated speed is reported in the ith stage. In the last stage, segmentation actor or the next incoming burst is updated as: s = [ ] scaling actor * (9) max where [.] is again a rounding down operator, sampling requency, max s is is maximum oppler requency and scaling actor which is dependent on the channel type. In rapidly changing channels, since the amplitude o the signal varies more rapidly during a burst, the limits o scaling actor cannot be set as
high as what it is in static channels. A rapidly changing channel may appear in some situations, i.e. where the speed o the user terminal is high. The algorithm interrupts until receives new bursts. Algorithm started again (go to stage two) by using this new value o when a new burst arrives. oppler shit estimation algorithm improves when SNR increases. IV. Simulation Results and Analysis In order to evaluate the perormance o the proposed algorithm or estimating the user speed by BTS, simulations are perormed according to condition that is reported in Table 2. In the simulation, we used the simulated S generated data in TETRA system and passed them through a Rician ading channel. The initial value or is chosen to be 40. Table 2 Simulation Parameters Parameters Values Carrier Frequency 400 Hz odulation ode π / 4 QPSK Access ethod TA with 4 timeslots per carrier Channel odel Rician Fading Channel Speed o obile 0-120 km/h Length o Speed Estimation Window 500 msec (34 Bursts), 1 Burst= 14.17 msec Rician Factor ( k ) 1 Sampling Frequency Simulation length 8 khz 1000 Bursts Figure 4 normalized relative error in various oppler shits. In Fig. 5, the accuracy o algorithm in tracking the speed o users is shown. As mentioned beore, ater receiving 34 burst (34 burst = 0.5 sec), the proposed algorithm starts and initial estimation o user s speed is perormed. Ater that, a speed is estimated or each next burst. As you seen the proposed algorithm in [15] cannot ollow the mobile terminal s variation in low SNR conditions. For the simulation results, η is deined to indicate the normalized relative estimation error: η ˆ max max = (10) where, max is estimated aximum oppler shit. ˆmax Fig. 4 illustrates the perormance o proposed algorithm in three dierent maximum oppler shits versus SNR. It shows how the perormance o Figure 5 Comparison between Proposed ethod and Hong ethod V. Conclusion
In this paper, a level crossing based algorithm or oppler shit estimation is improved. It is shown that the perormance o the improved algorithm in moderate SNRs (i.e., SNR=5 db) or TETRA users is acceptable. The estimated mobile terminal speed based on oppler shit can be used in several processes, including handover time estimation, or anti-jamming techniques, when the receiver tries to dierentiate between signal attenuations caused by jamming and channel eects. Acknowledgement The authors would like to thank epelmaan Pardaaz Ltd. or partially supporting o this research. Reerences [1]. A. Sampath and J. Holtzman, Estimation o maximum oppler requency or hando decisions, Proc. IEEE VTC, pp.859 862, 1993. [2]. K. E. Baddour and N. C. Beaulieu, Robust oppler Spread Estimation in Nonisotropic Fading Channels, IEEE Transactions on Wireless Communications, vol. 4, no. 6, pp. 2677-2682, November 2005. [3]. K.. Anim-Appiah, On Generalized Covariance-Based Velocity Estimation, IEEE Transactions on Vehicular Technology, vol. 48, no. 5, pp. 1546-1557, September 1999. [4]. J.. Holtzmann and A. Sampath, Adaptive Averaging ethodology or Handos in Cellular Systems, IEEE Transactions on Vehicular Technology, vol. 44, no. 1, pp. 59-66, February 1995. [5]. C. Tepedelenlio glu, A. Abdi, G. B. Giannakis, and. Kaveh, Estimation o oppler spread and signal strength in mobile communications with applications to hando and adaptive transmission, John Wiley & Sons, Ltd., 2001. [6]... Austin and G. L. St uber: Velocity Adaptive Hando Algorithms or icrocellular Systems, IEEE Transactions on Vehicular Technology, vol. 43, no. 3, pp. 549-560, August 1994. [7]. ottier, Castelain, A oppler estimation or UTS-F based on channel power statistics, IEEE Vehicular Technology Conerence, 1999, 5, pp. 3052-3056. [8]. C. Tepedelenliogl,G. B. Giannak, Estimation o oppler Spread and Signal Strength in obile Communications with Application to hando and adaptive transmission, Wireless Communications and obile Computing, 2001, 1(2): 221-242. [9]. Hua Jingyu, You Xiaohu, Sheng Bin, et al. A scheme or the oppler shit estimation despite the power control in mobile communication systems, IEEE Vehicular Technology Conerence. Spring, 2004, 1(17), pp. 284-288. [10]... Austin, Gordon. L. Stuber, Eigen-based oppler estimation or dierentially coherent CP, IEEE Trans on Vehicular Technology, 1994, 43(3), pp. 781-785. [11]. H. Jingyu, H. Han,. Qingmin, Y. Xiaohu, A Scheme or the SNR estimation and its application in oppler shit estimation o obile Communication Systems, Proc. o IEEE 2004 spring, 2004, vol.65, no.1: 244-248. [12]. KRASNY L. oppler spread estimation in mobile radio systems, IEEE Communication Letters, 2001, 5(5): 197-199. [13]. R. Narashimhan and. C. Cox. "Speed Estimation on Wireless Systems Using Wavelets", IEEE Trans. On Communications. Vo1.47. N0.9. Sep. 1999. [14]. Yi Sha, Na Yao, Xiaojing Xu, Improvement and Perormance Analysis o A Scheme or the aximum oppler Frequency Estimation, IEEE-WiCome Conerence, 12-14 October 2008. [15].. G. Cho,. Hong, Velocity Estimation Apparatus and ethod Using Level Crossing Rate, US patent 0235479 A1, 2004. [16]. Hans-Peter, A.Ketterling, Introduction to igital Proessional obile Radio, Artech House, INC, 2004. [17]. P.Stavroulakis, Terrestrial Trunked Radio- TETRA, Springer-Verlag Berlin Heidelberg 2007. [18]. Theodore Rappaport, Wireless Communications: Principles and Practice, 2nd Edition, Prentice Hall, 2001.