Indian Journal of Radio & Space Physics Vol. 30, December 200 I, pp. 294-299

Mobile train radio measurements in urban western India and comparison with some models

M V S N Prasad Radio & Atmospheri c Sciences Di vision, National Physical Laboratory, Dr K S Kri shnan Road, New Delh i 110 0 12

and

Rajendra Singh Te lecom Regulatory Authority of Indi a, A2/14, Safdarjang Enc lave, Africa Avenue, New Delhi 11 0 029

Received 4 January 2001; revised 30 March 2001; accepted 11 May 2001

Field strength measu rements were conducted in the mov ing trai n in the UHF band in the urban and sub-urban regions of western India. The observed signal levels were compared wi th those predicted from methods like Hata, Wal fish and Ber­toni, Uni form theory of diffract ion, Ibrahim and Parsons, Egli and Blomquist and Ladell and suitable methods were ident i­fied. These studies will be use ful for designi ng future train mobile communication systems and also for assessi ng the inter­fere nce potential of base stations for electromagnetic interference/electromagnetic compat ibility and spectrum management applications.

1 Introduction The tremendous growth of mobile radio requires

the detailed knowledge of radio propagation charac­teri stics and the effect of env ironment and terrain on the wave propagation '-4• The des ign of future mobile communication systems for different land mobile ap­plications such as railways, personal communications in car telephones, transport operators, etc. requires extensive experimental campaigns and comparisons of these investigations with different models. The ex­perimental measurements of signal levels help to de­duce the dimensions of the cells accurate!/ and as­sess the interference levels among the base stations. Another important applicat ion of these investigat ions is the determination of the validity domains of various prediction methods that can help cellular operators to minimize their time and money. Though rigorous models give good accuracy, they consu me lot of time and require hi gher reso lution data base of building, land usage th at can affect budget of operators6

.

Urban radio wave propagation gained a lot of im­portance due to the increas ing demands of mobile communication and ex tensive measurements and modelling techn iques were developed on thi s as­pect7·8. Adapti ve propagation techniques were advo­cated by Stern et a/9. for spec ified geographic region. The UHF propagation in different environments was extensively reviewed by Bertoni 10

.

In the present study, field strength measurements were conducted inside a moving train in western India

using Kurla and Kalyan base stations in di ffe rent en­vironments. The measured signal levels were con­verted into path loss values and compared with vari­ous narrow band prediction methods. This has been carried out to identify the suitable prediction over this region and the results of thi s study can be useful for developing mobile communication network design tools.

2 Experimental details With this objective in view field strength meas­

urements were conducted in a moving train in western India in UHF band utili zing the track side microwave towers as base stations. The base stations utilized in the present study are: Kurl a (part of Bombay) and Kalyan. The effective radi ated powers of Kurl a and Kalyan are + 37 dBm. The base station transmitted continuous carrier. The carrier levels origin at ing from these base stations were monitored inside the moving train. The RF carrier level was recorded in side the coach with a calibrated receiver, chart recorder and computerized data logger. The receiver output vo ltage corresponding to the incoming RF voltage was re­corded on a paper chart. For relating signal level (path loss values) with a locat ion, a counter of wheel rota­ti ons was included to drive the paper chart propor­tional to the distance covered by the tra in accordi ng to its speed. In this way signal level as a function of distance can be obtained. Signal level was received by an omni-directional antenna installed on the roof of

PRASAD & SINGH: MOBILE TRAIN RADIO MEASUREMENTS 295

the coach. The measurements were not repeated. In radio communication signal level is a random variable but median signal level averaged over an interval is pretty useful design !Jarameter and is extensively used in network planning tools. In mobile communications, environmental effects like scattering, diffraction and reflection from buildings, structures, other moving vehicles and trees affect the signal level and give rise to multipath fading. The signal level is not affected by the variability of atmospheric conditions.

The recorded carrier levels were averaged over 100 m section. The base station Kurla is situated in a densely crowded city with multistorey buildings ranging from 10 to 15 storeys. The urban environment prevails up to 12 km on either side of base station. The base sta­tion Kalyan comes under sub-urban category. Both the base stations are separated by a distance of 36.6 km. The height of Kurla base station is 32 m and that of Kalyan is 49 m above ground level. The terrain is flat between the base stations. The carrier level is measured at a distance of every metre and is averaged over a distance of 100 m. This averaged signal over 100 m is converted into path loss values and its stan­dard deviation, maximum and minimum values are deduced.

3 Environmental description Railway environment is different from the tradi­

tional road environment. A 25 kV line runs along the track generating harmonics of radio waves causing electromagnetic interference. The height of the coach is at a higher elevation compared to the height of tra­ditional road based vehicle. This may give added ad­vantage for signal levels using height gain character­istics of antenna. The day-to-day environment con­sists of plane earth, side-walls, terrain cuttings, over bridges, masts, wires, tunnels, crossing trains, etc. 11

Propagation path in thi s environment is complicated due to different propagation mechanisms like reflec­tion , diffraction , sca ttering, etc. Railway environ­ments are first divided between stations or station yards. In station yard, as there are some rai lway structures, multiple interference occurs. In between the stations, env ironment can be divided into tradi­tional classification . Also in railway mobile environ­ment, user density is less compared to tradit ional road based cellular environment. Hence, propagation char­acteristics in this environment would be different as compared to the road based environment. The envi ­ronment up to 12 km on either side of Kurla base sta­tion could be categorized as urban environment,

whereas in the case of Kalyan base station, the envi­ronment towards Yangani side up to 2 km is sub­urban and the rest is open environment. On the Kurla side it is sub-urban environment.

4 Results and discussion The observed carrier levels were converted into

path loss values and were compared with different prediction methods like Hata12, Walfisch and Ber­toni13, Ibrahim and Parson14, Uniform theory of dif­fraction15 (UTD), Egli 16, Blomquist and Ladell 17 and ITU-R (Ref.18). The main advantage of the compari­son with various prediction methods whether empiri­cal or semi- empirical is the possible extrapolation of these measurements to higher frequencies and differ­ent base station antenna heights.

Figure 1 shows the comparison of path losses de­duced from the observations of Kurla base station with those predicted from Hata, Walfisch and Bertoni, Ibrahim and Parsons, UTD, Egli and Blomquist and Ladell. The base station Kurla is situated at a distance marked zero and the observed path losses plotted on the left side correspond to those of measurements conducted when the train was moving towards Kalyan side. The observed path losses plotted on the ri ght side correspond to those of measurements when the train was moving away from Kurla base station on the opposite side. The coverage area of this base station is 8 km on either side. The observed path loss increases steadily on either side and a maximum path loss of 145 dB is seen at a distance of 6.5 km on Kalyan side. Hata's method overestimates the observed values till 6 km with a maximum devi ation of 15 dB at 4 km on the Kalyan side. On the other side the devi ation is more till 4 km. At 3 km distance, the deviation is 9 dB and at 6 km the deviation remains within I dB. It ap­pears that on both sides the deviation is more till 4 km and the deviation decreases beyond 4 km. In the case of Ibrahim and Parsons' method on the Kalyan side the deviation is high of the order of 12 dB at 3 km and the agreement is better between 4 and 5 km. Beyond 5 km, the predicted path loss does not follow the ri se in observed path loss. On the other side, Ibrahim and Parsons' method underestimates the path loss by 15 dB . On both sides of the base station, Blomquist and Lade II' s method undcresti mates urban path loss by as much as 20-25 dB. This method cannot be app lied to predict urban radio wave propagation unless coiTec­tion factors are incorporated. On both sides, Egli's method gives good agreement up to 2-3 km and , thereafter, the predicted curve does not follow the

296 INDIAN J RADIO & SPACE PHYS, DECEMBER 200 1

170,-------------------------------------------------------~

(/)

(/) 130 0

_J

----- ............ , I ...... ~ 120 <t CL

110 ...........

' 100

BASE STATION KURLA

......

' '

...... ......... ......

' ' .........

· -.. .. ' '

--OBSERVED - ·- ·- HATA - ... _BLOMQUIST ---- EGLI ~IBRAHIM S ~RSON <>---<> WALFISH- BERTONI --..UTD

-/

... / /

/

90~--L-~---J--~--~~~---~--~--~--~~~--~--L---L--L~ 8 7 6 5 4 3 2 0 2 3 4 5 6 7 8

KALYAN ~- DISTANCE, km

Fig. ! -Compari son of observed and predicted path losses for urban (Kurla) base statio n

steep rise in path loss. This could be probably due to line-of-sight propagation up to thi s distance and, thereafter, multiple diffractions and reflections of the diffracted field contribute substantially to the ob­served path loss. Walfisch and Bertonis' method gives very good agreement up to 5 km on the Kalyan side and, thereafter, the deviation increases. At 6 km the dev iation is 13 dB. On the other side, the agreement is moderate up to 3.5 km and, thereafter, it deviates by I 0-15 dB between 4 and 8 km di stance. Also, an at­tempt is made to deduce path loss exponent at differ­ent distances. This vari es from 2.7 to 5.0 on Kalyan side of the base station and varies from 3.6 to 7.7 on the opposite side of the ase station. The higher path loss exponent on the other side of the base station is due to the dense urban env ironment.

According to ITU-R draft revi sion of recommen­dation ITU-R P.ll45 [Ref. l8] in urban situations the following formula has been employed to calculate the loss relative to free space.

L = 82.550955 _ 402.4117 + 754.939291n (/) urb In(/ ) J

664 1.8739 !1.5 ... (l)

where, f is the frequency in MHz. With Eq. (I) the urban loss relative to free space comes to 25.25 dB, whereas the observed results showed that the urban loss rel ative to free space is 2 1 dB at 2 km and in­creases to 43 dB at 6 km distance. Probably the ITU­R value becomes closer to the value of 21 dB ob­served at shorter distances and can give some idea of the urban loss without going into the details of urban structure like width of the street and height of the buildings, etc . The ITU-R method gives the expres­sion to calcu late the urban loss based on the frequency and provides a first hand information to the design engineers in the absence of elaborate prediction meth­ods. We have incorporated this to see how thi s method agrees closely with other methods. The value of Lurb (value of urban loss relative to free space) cal­culated from this is constant irrespective of the dis­tance and when added to free space loss gives the to­tal loss. The total path loss varies with distance due to the variation of distance-dependence in free space loss term. Hence, it is not plotted in the Fig. l .

A UTD based propagation model deri ved by Zhang 15

has been utilized to deduce the path loss in the urban region. Here, the path loss in the case of narrow band system is

PRASAD & SINGH: MOBILE TRAIN RADIO MEASUREMENTS 297

• 00 (2)

where L0 is the free space loss, L 1 is the path loss due to local row of buildings closing the mobile, L mct the path loss due to multiple diffraction caused by several rows of buildings between base station and mobile, and L, is the reflection of diffracted electric field. Here, the rows of buildings have been taken as knife edges with an equal spacing in the range of 30-60 m . In the present case, the spacing of 30 m is taken. Here, the loss due to reflection of the diffracted elec­tric field comes to 3 dB and remains more or less the same with distance. The agreement of UTD deduced path loss with observed values becomes better beyond 4 km onwards. If the loss due to reflection, i.e. 3 dB , is not taken into account the agreement at distances below 4 km improves moderately . At 6 km the devia­tion of UTD on the right hand side is 6 dB, whereas towards Kalyan side it is 1 dB. The agreement of UTD depends on the precise choice of parameters d,, height of buildings, estimation of loss due to reflec­tion of diffracted electric field, etc. At closer distances the agreement is not that good.

Figure 2 shows the comparison of observed path losses deduced from Kalyan base station with those

predicted from different prediction methods. The cov­erage area of this base station is 19.24 km . Here Hata's method with sub-urban correction factors has been used on both the sides. This method follows the trend of observed curves reasonably on both the sides. On the Yangani side (it is the base station beyond Kalyan whose measurements are not discussed here) the agreement is very good till 4 km and a deviation of around 8 dB is noticed at 6 km distance. From 12 km onwards the agreement becomes very good on both the sides. The deviation at these distances is con­fined between 3 and 4 dB. A moderate decrease in path loss by 20 dB at 13 km distance and a subsequent rise in path loss by 40 dB is noticed on the Kalyan side. This might have occurred when the train was taking a turn or the vehicle might have been ob­structed by an obstacle at these distances.

Egli's method also follows the similar trend on both the sides. On Vangani side the deviation is be­tween 2 and 3 dB ti II 12 km and increases to 10 dB beyond this . On the Kurla side the deviation is high only at 16 km where steep rise in observed path loss is noticed. Blomquist and Ladell's method clearly un­derestimates on both sides by 15-20 dB. The variation

170~-----------------------------------------------BASE STATION KALYAN ---OBSERVED

--·- HATA -·· ·-BLOMQUIST ---- EGLI

co u

(/) (/)

0 130 __J .--:X: 1-

_. ... / c:t Q.

110 /

16 12 8 4 0 4 8 12 16 20 24 ........__ KURLA VANGANI ~

DISTANCE, km

Fi£. 2-Comoarison of observed and predicted oath losses for sub-urban (Kalvan) base station

298 INDIAN J RADIO & SPACE PHYS, DECEMBER 2001

of observed path loss for Kurla base station between 8 and 7 km for every I 00 m is shown in Table I. Table I depicts mean, maximum, minimum and standard deviation path loss. This gives an idea about the vari­ability of the fading range. Table 2 depicts the similar parameters for Kalyan base station.

Lazarakis et a/.19 proposed a point-to-point mobile

radio prediction model based on effective height esti­mation at various points of interest in urban and sub­urban zones of Athens. The drawback of this method is the requirement of huge data base of height of ter­rain, building heights, and the need of geographical information system to correlate the given points with the measurements . They have developed different equations based on measurements done at 900 MHz for positive effective height and negative effective height of the antenna. The effectiveness of their model is demonstrated by comparing their model with well-known models. L lace et a/.2 have used the mod­els of Saunders and Bonar20 and L Jun Iacer and Cardona21 to compute multiple diffraction over regu-

Table 1-Variability of observed path loss in dB with distance for Kurla base station

Distance Mean value Max. value Min. va lue Standard km deviation

8. 1 140 153 127 4.8

7.9 138 151 125 5.7

7.8 142 155 129 4.8

7.7 145 155 131 4.5

7.6 142 155 127 4.8

7.5 142 155 127 4.8

7.4 143 153 129 4.9

7.3 144 155 127 5.2

7.2 141 155 13 1 5.1

7.0 140 153 129 5.7

Table 2-Variability of path loss in dB with di stance for Kalyan base station

Distance Mean Max. Min. Standard km value value value deviation

19.24 143 161 123 4.8

19.14 152 157 137 4.2

19.04 ISO 159 135 4.2

18.94 148 157 14 1 4.0

18.84 150 161 139 4.0

18.74 150 159 14 1 4.0

18.64 151 159 127 4.3

18.54 146 161 135 5.0

18.44 140 153 127 6 .9

18.24 138 153 97 7.2

18.14 139 151 125 5.3 18.04 138 151 125 5.9

lar profiles of buildings and flat edge model of Saun­ders and Bonar 22 for irregular profiles of buildings .

5 Summary and conclusions Field strength measurements were conducted in a

moving train and the observed results were compared with different models. In the past many workers have compared experimentally observed results with em­pirical models taken from road based vehicles. Very few results are reported in the literature on train based measurements and the present study fills thi s gap. In the urban region Hata's method gives better agree­ment beyond 5 km on Kalyan side and beyond 4 km on the opposite side. Walfisch and Bertoni's method gives good agreement at closer distances. The UTD gives reasonably good agreement at all distances and the deviation is slightly high at shorter distances. Ibrahim and Parson, Egli and Blomquist and Ladells ' methods deviate by large values. In the sub-urban re­gion, both Hata and Egli's methods follow the trend at all distances, and Blomquist and Ladel ls' method un­derestimates the observed path losses by large values. These studies would be useful for the design of future railway mobile communication systems especially in this region of the world and for considering EMI/EMC aspects into spectrum management appli­cations. This work in the area of train mobile commu­nications is the starting point and more experiments in this line can confirm the conclusions of this study and give further insight into the mobile radio propagation mechanisms.

References I Kukushkin A, Aust Telecommun Res (Australia), 28 ( 1994) I. 2 Juan -L lace L, Ramos L & Cardona N, IEEE Trans Veh Tech­

not (USA).48 (1999) 1463. 3 Har D. Watson A M & Chadney A G, IEEE Trans Veh Tech­

not (USA), 48 ( 1999) 1451. 4 Perera S C M, Williamson A G & Rowe G B, Electron Lett

(UK), 135 ( 1999) 1135 . 5 Bernardin P, Yee M F & Ellis T, IEEE Trans Veh Techno/

(USA),147 ( 1998) 1215. 6 Carciofi C, Daniele P, Esposti V D & G razioso P, Identifica­

tion of the validity domains of beloiV roof top and over roof top micro cellular prediction tools, presented at the 7'h Vir­g inia Tech'/MPRG Symposium on Wireless Personal Com­munications, Paper no.3.1 , Black burg, U.S.A ., June 1997.

7 Balis P G & Hinton 0 R, /££ Proc Micro Antennas & Propag (UK), 144 (1999) 367.

8 Kim S C, Guarino B J (Jr), Willis T M, Erceg V, FortuneS J, Valenzuela R A, Thomas L W, Ling J & Don Moore J, IEEE Trans Veh Techno/ (USA), 48 ( 1999) 931 .

9 Stern H P, Hermes J G & Subrahmanyam D, 1£££ Trans Broadcast (USA), 43 ( 1998) 56.

10 Bertoni H L, Hancharanko W, Maciel L R & Xia H 1-1, Proc /£££(USA) , 82 (1994) 1333.

PRASAD & SINGH: MOBILE TRA I RAD IO MEASUREME TS 299

II Hattori T, Shiokawa T & Abe K, Measure111e111 of 111iCro 1vw•e, 111illillleler-wave band propagaiion charaCierisiics in e111•i­rowne111s along railway /racks, in Proc. of !SAP, Aug 2000, Yol. l.pp 32 1-324.

12 Hat a M, IEEE Trans Veil Techno/ (USA), YT-29 ( 1980) 317.

13 Walfish J & Bertoni H L, IEEE Trans An1ennas & Propag (USA), 36( 1988) 1788.

14 Ibrahim M F & Parsons J D, Proc /nsf Eleclr Eng (UK), 130 ( 1983) 377.

15 Zhang W, IEEE Trans. Awennas & Propag (USA), 45 ( 1997) 1669.

16 Egli J J, Proc IRE£ Ausl (A usiralia), 45 ( 1957) 1383.

17 Blomquist A & Ladcll L, Prediciions and calculalions of

Irans111issionloss in dijferenl Iypes of ierrain, NATO AGARD Conf, Stockholm. Sweden, 1974, pp 3211-32117 .

18 Propagaiion e.ffeus i111pac1ing land 1110bile sen•ice in Ihe VHF and UHF bands. drafl revision of Ih e reCOIII IIIendaiion ITU-R P. /1 45, Doc111nen1 3 K/5-£, 7 ov 1996. pp. 1- 14.

19 Lazarakis F, Dangakis K. Alexandridis A A & Tombras G S. A poilll -Io-poinl mobile radio prediction 111ode/ based on ef­feciive height esiimation for Ihe region of Aihens in /£££ PIMRC Conf, 1994, pp. 258-262.

20 Saunders S R & Bonar F R, £/eciron Lett (UK), 27 ( 1991 ) 1276.

2 1 Jun L Iacer L & Cardona N, Electron Lett (UK), 33 ( 1997) 92. 22 Saunders S R & Bonar F R, IEEE Trans Antennas & Propag

(USA), 42 ( 1994) 137.