A Study of Wireless Channel on GSM-R Network
JIN Xinyu* JIANG Lurong WU Duanpo
Department of Information Science & Electronic Engineering Zhejiang University Hangzhou 310027, P.
R. China firstname.lastname@example.org* email@example.com firstname.lastname@example.org
Abstract—GSM-R is a railway communication system based on the GSM technology. With the widely develop of high-speed rail, the GSM-R quality of service (QoS) is becoming much more important than ever. We analyze GSM-R by measuring field coverage and radio field coverage model, deducing the Doppler frequency offset, giving the upper and lower limits of overlap in handover at different speed, and studying the data-transmitting error rate of GSM_R network. We propose that increasing the space of two neighbor cells does not significantly weaken the quality of field coverage, but can reduce the chance of ping-pong handover and other network performance. We also discover that the Doppler frequency is decreasing gradually and the error rate is decreasing shakingly when the train is becoming closer to the base station. All the data we used in this paper is comes from drive tests on JJ-DPL and W-G DPL, China. Keywords-GSM-R; Handover; Hata’s prediction method; Overlap; Doppler effect
Wuhan-Guangzhou DPL (W-G DPL). The former is the Chinese first intercity high-speed rail that covers 117 km with top speed of 330 km/h, and the later is the longest high-speed in China which travels 968 km and reaches top speed of 350km/h. This paper is divided into five parts. Section II provides background of handover in GSM-R. Section III studies the Hata’s prediction method and deduces the Doppler shift in high-speed. Section IV analyzes the upper and lower limits of overlap in different speeds. Section V proposes and simulates a error-bit rate model. In Section VI, we make a conclusion. II. BACKGROUND
GSM-R is an international wireless communication standard for railway communication and applications based on the GSM technology. It is developed as a platform for voice and data communications as well as for traffic control system for railway. GSM-R is accepted by many countries, including most of Europe countries, China, and India. In Europe and China, GSM-R is under implementation successfully and replacing the old railway communication system. With the widely develop of high-speed rail, the GSM-R quality of service (QoS) is becoming much more important than ever. Although GSM-R is based on mature technology GSM, the situations of QoS are significant different between GSM-R and GSM. Handover is one of the key features for GSM-R QoS. The line-settled base transceiver station (BTS) next to railway and high-speed moving mobile station (MS) on trains makes the handover in GSM-R more important and difficult than it is in GSM. Handover failure could cause failure in call, data transmission, or connection loss. Therefore, it is necessary to study the handover model in GSM-R. In this paper, we analyze GSM-R by measuring field Coverage and radio field coverage model, deducing the Doppler frequency offset, giving the upper and lower limits of overlap in handover at different speed, and studying the datatransmitting error rate of GSM_R network. All the data we used in this paper comes from drive tests on Beijing-Tianjin Dedicated Passenger Line (JJ-DPL) and
This work was supported by National Science & Technology Pillar Program of China (2009BAG12A08-07), National High Technology Research and Development 863 Program of China (0912JJ0104-TX00-H-HZ-00120100105), and National High Technology Research and Development 863 Program of China (0912JJ0203-WX00-H-HZ-001-20091116).
978-1-4244- 7618-3 /10/$26.00 ?2010 IEEE
Handover happens in speech call or data transmission from one channel connected to the network to another. There would be several handovers in a speech call as the Small cell radius of BTS and train’s high-speed movement. During a speech call or data transmission process, MS periodically reports the signal strength of serving and neighbor cells to the network via the measurement report. If the reports show that signal from serving cell is much weaker than it from neighbor cells, the base station controller (BSC) would initiate a handover. Then the BSC selects a neighbor cell as target, in this target cell the MS would be better served. The target cell provides a new channel for the MS, the MS changes the speech call or data transmission to the target cell. After the handover to the target cell is successful, the network will release original channel. In short, the handover process can be divided into four step of measurement, access burst, select, and act. Relatively work has been undertaken on analysis of handover model in GSM-R, as well as new handover algorithms. Characters of the Propagation Channel are provided in , and it is a reference to research the handover model. Channel model for train to train communication using 400 MHz band is described  which can be extended to GSMR band. Group handover process is discussed in different environments with high-speed is proposed in . A handover algorithm is proposed in  based on sampling of received signal strength. A new handover scheme with relay station in GSM-R network is proposed to increase the probability of successful handover. Extra relay stations can improve radio field coverage but also increase the cost. Two fuzzy logic based handover algorithms using the average handover factor are proposed  to reduce the probability of unnecessary handover in high-speed train environment. The location information is
added to determine the target cell and accelerate the handover process . These researches deal with the characters of channel model or handover algorithms, but not combine them. III. CHARACTERS OF THE CHANNEL MODEL
coverage threshold, which is -92dBm. Fading and noise should be considered in the drive test data. But even though, the radio field coverage follows the Hata’s experimental formula statistically with the distance of 1~7 km and all the drive test data are still much larger than threshold. B. Doppler Shift Doppler shift is the change of radio frequency in the relative motion of transmitter and receiver. In different environment and speed, varying degree of radio frequency shift occurs. As the train’s high-speed, the turning radius of DPL is very large, commonly more than 7 km. Therefore, we could take the curving rail as straight rail when we analyze the Doppler shift in rail communication. We consider the Doppler shift problem in straight rail and multi-BTSs situation. On the basis of the analysis of Straight rail and single BTS, adding consideration of handover, we can analyse Doppler shift in multi-BTSs condition. Almostly, handover is happened between two adjacent circles. Let the position of adjacent BTS including BSk-1 BSk BSk+1 be (xk-1,yk-1) (xk, yk) and (xk+1, yk+1) let the handover place between BSk-1 and BSk is ((xk-1+xk)/2+ k-1,0) let the handover place between BSk and BSk+1 be ((xk+xk+1)/2+ k),0).The symbol k represents the offset distance to the center of BSk and BSk+1.
A. Propagation Loss In order to determine the parameters of the system, accurate prediction method is needed. Hata’s prediction method is used in system planning for land mobile radio service quality evaluation in Jing-Jin Dedicated Passenger Line (JJ-DPL). Hata’s prediction method consists of several experimental formulas derived from Okumura’ prediction method. The Hata’s prediction method is applicable with the distance of 1~20 km between MS and BTS, but not adapted when the distance is less than 1km. Experimental formula for propagation loss in the open area is
L po ( dBm ) = 69.55 + 26.16 × lg Fc ? 13.82 × lg hb ? a ( hm ) + ( 44.9 ? 6.5 × lg hb ) × lg R ? 4.78 × (lg Fc ) 2 + 18.33 × lg Fc ? 40.94
where Fc is BTS frequency, hb is BTS antenna height, hm is MS antenna high, a(hm) is the correction factor for the MS antenna height, R is the distance between BTS and MS. The correction factor for the MS antenna height in mediumsmall city or open area is
a(hm ) = (1.1 × lg Fc ? 0.7) × hm ? (1.56 × lg Fc ? 0.8)
Therefore, the received signal strength in dBm is
Figure 2. Multi-BTS and train
Lms (dBm) = L power (dBm) ? L po (dBm) where Lpower is the transmit power of BTS.
(3) Doppler shift formula can be defined as below:
f d (k ) = f ? xk ? x v( x) ? c ( x k ? x) 2 + y k 2
x k ?1 + x k x + x2 + δ k ?1 < x < 1 +δk . 2 2
CELL OVERLAP REGION
Figure 1. Comparation between Hata’s experimental formula in open area and drive test data in JJ-DPL
Comparation between Hata’s experimental formula in open area and drive test data in JJ-DPL can be seen in figure 1. The red lines come from 20 times testing data of a cell on JJ-DPL. The black line is computed from experimental formula for propagation loss in the open under the following conditions: frequency 930MHz, transmit power of BTS 25W, BTS antenna height 40m, MS antenna 5m. The dotted line is radio field
A. Upper and Lower Limits of Overlap Overlap coverage is needed to ensure the success of handover. figure 3 shows the cell boundary and overlap in GSM-R system. Lines of signal strength in figure 3 are symmetrical since the symmetrical setting of serving cell and neighbor cell. In GSM-R system, hoMargin is an adjustable parameter influences handover procedure. Handover is available when the difference of signal strength values from serving cell and neighbor cell is less than hoMargin. The train moves from serving cell to neighbor cell. A is the edge of neighbor cell coverage. Handover could be start on the point of B and C. As the figure 3 shows, R is the cell coverage radius, L is the overlap of the two cells, d0 is the area that handover is available, d1 is the handover preparing area.
l ≥ 2vt
(10) Therefore, we can simulate the overlap area for handover in different speed as figure 4 has shown. B. Simulation and Drive Test Result on Overlap The ratio of cell coverage radius to the space between two neighbor cells should also be considered. This radio in J-J DPL is 2, while it is 1.5 a German railway. Over-intensive BTS distribution not only increases the expense of contribution and maintenance, but might lead to ping-pang handover effect, which means handover to and from between two cells frequently.
Figure 3. Cell overlap between two neighbor BTSs
If Fc, hb, and hm have be settled in invariable numbers, We can rewrite equation (1) as L po (dBm) = a + k × lg R (5)
Where both a and k are constant values, where a = 69.55 + 26.16 × lg Fc ? 13.82 × lg hb ? a(hm ) ? 4.78 × (lg Fc ) 2 +18.33 × lg Fc ? 40.94 and k = 44.9 ? 6.5 × lg hb > 0 when hb=20 m.
The radio coverage strength on B is L pB (dBm) = L power (dBm) ? a ? k × lg(r ? l + d1 ) = L power ? a ? k × lg(r ? l + vt ) (6)
Figure 5. Frequency assignment and coverage in J-J DPL
Where v is the train speed, t is the handover preparing time. The radio coverage strength on C is L pC (dBm) = L power (dBm) ? a ? k × lg(r ? vt ) (7)
As handover could be in activation between the point of B and C, we can get
( L pB ? L pC )(dBm ) = ? k × lg ( r ? l + vt ) ≤ hoMargin r ? vt
A drive test on J-J DPL is shown in figure 6(a). The train drives from YL-WQ02 to YL-WQ04 with the speed of 220km/h. The serving cell is YL-WQ03 and the handover target cell is YL-WQ04. Handover is available at 59.3km as the difference of signal strength values from serving cell and neighbor cell is equal to hoMargin. The overlap area is about 2km and the value is between the upper and lower limits of overlap as figure 4 has shown in the speed of 220km/h. Handover happens at 60.2km and the serving cell is changed to YL-WQ04 successfully. Finally, all the values of received signal strength is much greater than -92dBm, the threshold mentioned in .
Figure 4. Upper and lower limits of overlap for handover in different speeds
(b) Figure 6. Received signal strength in drive test(a) and after improved(b)
As k>0 and r-vt>0, we could deduce one of the upper limit of overlap
l ≤ r + vt ? 10 r ? vt
hoM arg in k
The values of received signal strength obey the Hata’s prediction equation statistically, but there are still some jumping changes randomly. A random jumping happens at 57.3km in figure 6(a) and this jumping might cause a pingpong handover. Expanding the space between two neighbor cells is a way to reduce the probability of ping-pang handover in the over-
But the overlap should be greater than 2vt for handover preparing, so the lower limit of overlap is
intensive BTS distribution lines. figure 6(b) shows that after expanding the space of YL-WQ02, YL-WQ03, and YL-WQ04, the probability of ping-pang handover reduces and the received signal strength is still greater than -92dBm. V. ERROR-BIT RATE
the base station, which matches the result of simulation on error-bit rate.
A. Error-bit Rate Simulator The GSM-R network applies GMSK as its modulating method, so GMSK modulating module and demodulating module is needed in the simulating module. Owing to the influence of the transmission environment, the radio wave will be reflected, diffracted and scattered, then the radio wave will be faded. So we selected some common channel fading modules including Rician channel, Rayleigh channel and AWGN channel together to simulate. The block diagram of simulating module is described as figure 7.
Figure 9. Drive tes of RxQual on W-G DPL
In this paper, we analyze the measured field Coverage and radio field coverage model, deduce the Doppler frequency offset, give the upper and lower limits of overlap in handover at different speed, and study the data-transmitting error rate of GSM_R network. We propose that increasing the space of two neighbor cells does not significantly weaken the quality of field coverage, but can reduce the chance of ping-pong handover and other network performance. We also discover that the Doppler frequency is decreasing gradually and the error rate is decreasing shakingly when the train is becoming closer to the base station.
Figure 7. Error-bit rate simulator
REFERENCES B. Simulation and Drive Test Result Figure 8 shows the simulate result of error-bit rate at the speed of 350 km/h. We find out that the error-bit is decreasing when the train is becoming closer to the base station. And there is some randomness between the error-bit rate and the distance from train to BTS.
 C. Garcia, T. Strang, A. Lehner. “A Broadcast Vehicle to Vehicle Communication System in Railway Environments”. International Symposium on Vehicular Computing Systems (ISVCS), Dublin, Ireland, July, 2008.  C.R. Garcia, A. Lehner, T. Strang, K. Frank, "Channel Model for Train to Train Communication Using the 400 MHz Band", IEEE Conference on Vehicular Technology Conference, 2008, pp. 3082-3086.  Wu Hao, Wang Qiang, Zhong Zhangdui, "The GSM-R Handover Algorithm Analysis and Research for High-speed Railway", http://www.civil.uminho.pt/ismarti/08ICTI/papers/P648.pdf  Y. Li, Y.S. Yan, "Handoff decision algorithm with dynamic sampling distance in GSM-R network ", IEEE International Conference on Railway Engineering, 2008, pp.1-4.  Zheng Xiang, Fei Yang, "A novel handover scheme with relay technique in GSM-R network", IEEE conference on Communications and Networking in China, 2009, pp. 1-4.  Ye Li, Yusong Yan, "Fuzzy logic based handoff decision algorithm in GSM-R network ", IEEE conference on Wireless, Mobile and Sensor Networks, 2007, pp. 505-508.  K. Kastell, S. Bug, A. Nazarov, R. Jakoby, "Improvments in Railway Communication via GSM-R ", IEEE conference on Vehicular Technology Conference, 2006, pp. 3026-3030.  Masaharu Hata, “Empirical Formula for Propagation Loss in Land Mobile Radio Services”, IEEE Trans. Vehicular Technology, vol. VT29, no. 3, pp.34-39, Aug. 1980.  Ministry of Railways of the People's Republic of China, “Temporary provision of engineering design of railway GSM-R digital mobile communication system”, 2008.  3rd Generation Partnership Project, “3GPP TS 05.08 V8.9.0” (2001-04), http://ftp.3gpp.org/specs/html-info/0508.htm
Figure 8. The error-bit rate simulation at the speed of 350 km/h
In drive test we could take RxQual as the error-bit rate approximately. RxQual is received signal quality, which has the range from 0 to 7. The value of RxQual is opposite to error-bit rate, which means the larger RxQual is, the lower error-bit rate is. Figure 9 shows a drive test of RxQual on W-G DPL. We find out that the RxQual is increasing shakingly (error-bit is decreasing) when the train is becoming closer to