`
`
`
`APPLICATION OF GSM IN HIGH SPEED TRAINS: MEASUREMENTS AND SIMULATIONS
`
`Manfred Goller
`
`Abstract
`The paper presents results of measurements and simulations concerning the application of the European
`GSM system in high speed trains travelling at up to 500 krn/h. The aim is to answer the question to what
`extent GSM (performance specified up to 250 km/h) can cope with the high velocities which are demanded
`for future railways. Measurements along railway lines have shown that a railway mobile radio channel
`results in better performance (Rice channel) than standard mobile radio channels (Rayleigh or weak Rice
`channel, see GSM-Rec‘s).
`BER and block error rate of GSM traffic channels up to 500 km/h are simulated. Comparison of the results
`at 250 km/h and 500 km/h shows that the GSM high velocity problem can be solved either by increasing
`the SNR by about ZdB or by increasing the Rice parameter c by about 6dB (numerical values for profile
`RA = rural area; railway channel with c = 6dB against standard channel with c = OdB). is. the BER at 500
`krn/h (railway channel) is not worse than the BER at 250 km/h (stande channel).
`A simple example shows that the benefit in the transmission of telegrams consisting of blocks of decoded
`bits can be much higher.
`The desired channel performance. Le. a strong direct path (high Rice parameter), can be achieved by
`careful radio coverage planning along the railway line. This means a GSM standard receiver is sufficient to
`cope with the GSM high velocity problem and no additional means are needed.
`
`1.
`
`Introduction
`
`The present situation of mobile communications in the European railways is characterized by a multitude
`of different networks using various analog technologies. No common solution exists and the systems are
`not compatible. This situation does not satisfactorily reflect the current process of integration within the
`European Community.
`Considering this. the German Railway is working on a solution with GSM which is in correspondence with
`a decision of the UIC (Union internationale de chemin de fer ) [1 1.
`Future high-speed trains in Europe are to reach velocities up to 500 km/h.
`Thus for GSM application the question arises to what extent GSM (performance specified up to 250 km/h)
`can cope with the high velocities.
`Useful but expensive measures against this limitation are antenna diversity or a fast adaptive equalizer.
`Railway radio channels are characterized by a coverage in the form of a line along straight railway
`embankments. cuttings and tunnels. The antennas are situated close to the railway line. their heights are
`relatively small and the coverage range is limited to some few kilometers.
`The environment typical to railways suggests that the railway mobile radio channel is a Rice channel.
`which results in better performance c.g. BER than a standard Rayleigh channel. This assumption led to
`radio channel measurements on lines of German Railway, modeling this channel on a computer and
`simulating the performance of GSM. The results of the investigations reported below contribute towards a
`decision as to the suitability of GSM for high-speed railway communication.
`
`Radio channel measurements
`2.
`The radio channel measurements were carried out on lines of German Railway including new built high
`velocity lines. The radio channel is completely characterized by its complex. time variant impulse response
`h (T. t), with T = delay time. t = real time.
`
`Measuring equipment
`The impulse response was measured with the channel sounder RUSK400.
`ATmin = Sps and °Tmin = 1,4us delay spread which is sufficient for GSM.
`
`It‘s measurement limits are
`
`M. Goller is with the DETECON Gmbl-I. Bonn, Germany
`
`6) 1995 The institution 01 Electrical Engineers.
`Printed and published by the IEE, Savoy Place, London WC2R OBL, UK,
`
`5 / 1
`
`I
`
`Liberty Mutual
`
`Exhibit 1017
`
`Page 000001
`
`
`
`
`
`Measuring conditions
`-
`Frequency band 945 MHz
`-
`Transmitter power 20W 50W
`-
`Transmitting antenna height above rail level
`open terrain 1‘1 = 14m, 20m, tunnel hl = 4m.
`The antenna is situated close to the railway line.
`Polarization vertical
`
`-
`
`-
`-
`-
`
`Receiving antenna height h2 -= 4,3m, omnidirectional. mounted on the top of the measuring coach.
`Measuring distance uansmitter-receiver d s 0, 1, 3,5 km
`Measuring interval so a: 150
`200m
`
`Channel parameter
`The received process
`
`’-
`r0) = r (r) - rm = ,lzlm .0]
`k:l
`r
`l
`k
`
`2
`
`consists of a fast part rs(t) (Rayleigh or Rice fading) and a slow part r1(t) (lognorrnal fading).
`Of interest is the dynamic behaviour of the channel which is determined by the fast fading rs(t).
`By low-pass filtering of the received process r(l) one gets rim and after division r5(t) - r(t)/rl(t).
`In particular the Rice parameter
`channel
`; Gaufl
`P = 0
`to
`;
`channel
`Rayleigh
`;
`Pd = 0
`;
`0
`pd/Pm
`c
`is defined as the power ratio of the signal in the direct path to the multipath-spread signals. It is estimated
`by the aid of a chi-square-test by comparison of the theoretical Rice-distribution with the measured
`empirical distribution of the fast fading [21.
`Additional characteristics of the fast fading are mean value 56— [dB], standard deviation ox[dB] and mean
`fading depth
`AdeB] = X(90)-X(10),
`where X(90) and X(10) are the 90% and the 10% quantile of the empirical distribution function P(x<X) of
`the signal level x[dB] respectively.
`For Rayleigh fading one gets
`
`f=-—2.51dB,
`
`o =5,57dB, AX=13.39dB.
`x
`
`For further details of channel characterization see [2]. [3} and [4].
`
`Measuring results
`3.
`1 shows as an example results of measurements at Milbertshofen. a Marshalling yard and industrial
`Fig.
`area near Munich (Bavaria).
`the
`As can be seen from the delay»Doppler—spectrum (scattering function) and the Doppler spectrum.
`measuring train moves towards the transmitter (angle of incidence = 0°. maximum Doppler frequency
`+fDmax)- A strong direct path exists with a delay of zero (main impulse) and a spike at +fDmax~ A weak
`reflection behind the train (angle of incidence - 180°, maximum Doppler frequency 'fDmax) occurs at a
`delay of 8.6m related to the main impulse (compare with the impulse response and the delay spectrum). It
`results from a highway bridge crossing the line 1.3km behind the measuring train.
`The multipath propagation is weak and in the order of the measuring limits. At locations where deep
`fadings occur the delay spread and the delay window are higher.
`Mean value PS, standard deviation CS and mean fading depth APS of the fast fading already show that
`ISO) is a Rice—process (compare with the theoretical results above for a Rayleigh process). A high Rice
`parameter corresponds to small standard deviation and small mean fading depth and reversed.
`A summary of measuring results is shown in table 1. For further details see [2], [3] and [4].
`In particular the results show that in the delay plane at all investigated railway lines only weak multipath
`propagation was measured. These channels can be described by a delay spectrum RA (rural area) or TU
`
`5/2
`
`Page 000002
`
`
`
`
`
`(typical urban) as specified in GSM-Rec. 05.05. Exception are lines in the Alps which can be characterized
`by the delay spectrum HT (hilly terrain). RA applies for tunnels.
`It should be noted furthermore that in the Doppler plane for a coverage along the line a direct path was
`always measured. The measured Rice parameter was about c = 8...13dB, see table 1. This means that in
`the Doppler plane railway mobile radio channels (Rice) are better than standard mobile radio channels
`(Rayleigh) as specified in GSM Rec. 05.05.
`The results suggest to model a railway mobile radio channel by
`—
`a delay spectrum as defined in GSM-Rec. 05.05 and
`—
`a Doppler spectrum also as defined in GSM-Rec. 05.05 but superimposed by a single spectral line
`resulting from a direct path. It's strength is determined by the Rice parameter.
`Thus the typical railway mobile radio charmel is a Rice channel superimposed by a slow. lognormal fading.
`
`_Rice parameter
`
`slow fading
`stand. dev.
`0 [dB]
`
`Milbertshofen
`
`Kraichtal, NBL
`Glems. NBL
`Eichstati
`Osterrni‘mchen
`Oberau — Garmisch
`Oberau — Miinchen
`
`Table 1: Summary of channel measuring results under standard measurement conditions: distance d:3lcm;
`transmitter: beam antenna 51° horizontal. gain 10,3dBd. height h1=20m, tunnel h1=4m;
`receiver: omnidirectional antenna, heigth h2=4,3m; NBL=newly built high speed line; (xx)=less frequent.
`
`Simulation results
`4.
`To simulate a railway mobile radio channel a 6-path-model as specified in GSM is used. The direct path is
`superimposed on the first path so that quite arbitrary Rice parameters can be generated.
`The simulations were focused to standard GSM—channels RAx and railway channels RAxy, where RA s
`terrain profile rural area. it = velocity in km/h and y = Rice parameter in dB.
`Fig. 2 shows as an example simulation results of the half-rate traffic channel TCH/HZ.4 after convolutional
`decoding (code rate r = 1/3. interleaving depth 1= 19. interleaving delay = l85ms).
`With respect to statistical safety and appropriate simulation time only BERBIO‘5 were simulated.
`The results from Fig. 2a Show that [5], [6]
`-
`the BER goes up when the velocity increases from 250km/h to SOOkm/h;
`-
`the rise of BER can be compensated by increasing the SNR by about ZdB (see also [7]);
`~
`the rise of BER can be compensated by increasing the Rice parameter from c = M8
`(GSM channel) to c: 6dB (railway channel);
`the BER at 500km/h with the railway channel RA500,6 is not worse than the BER at 250km/h
`with the standard GSM channel RAZSO.
`
`-
`
`Block error rate
`in data communications K information bits are combined in blocks or telegrams. A block or telegram is
`wrong if it contains 2 to wrong bits. Then P (2 m: K) is the cumulative probability that in a block with a
`length K are at least in wrong bits. Especially m = 1 describes the block error probability and m = g > i the
`probability that a block contains errors with a weight 2 g.
`Figures 2.b...2.d show the cumulative block error rate of TCH/H2,4 for SNR = 5dB. Table 2 shows the
`results for blocks of K = 256 information bits. Automatic train control could be a possible application of
`this example.
`
`5/3
`
`Page 000003
`
`
`
`
`
`It can be stated that with a railway channel not only the BER, but more important, the cumulative block
`error rate is reduced as well, actually the more the higher the error weight is.
`For example, if we assume an additional error detection after the TCH/H2.4—decoding which can detect
`errors up to a weight e -= d -
`l - 7 over K - 256 bits then P (2 8; 256) is the residual block error rate.
`Comparing RA500.6 and RA500.0, the railway channel provides a factor of 1/10 or 1/38 respectively, by
`which the block or residual error rate is smaller.
`
`P (2 64; 256)
`
`l; 256)
`8', 256)
`
`P (2
`P (2
`
`Table 2: Cumulative block error rate of the TCH/HZ.4 at SNR=5dB and a telegram length K - 256.
`Radio channel RAx,y = rural area with velocity xlkm/h] and Rice parameter y[dB].
`
`Example
`5.
`the
`to what extent
`The following simple example illustrates, for the case of automatic train control,
`performance is influenced by the radio channel. The results are to be understood as an upper limit which is
`achievable more effectively by an appropriate error detecting or error correcting code.
`Suppose that
`information blocks (telegrams) of the length K = 256 are transmitted n times in the
`TCH/H2,4 with a block error rate p = P( 2 l; 256) as shown in table 2 for SNR = 5dB.
`Statistical independence assumed. table 3 shows the probability
`'
`p11 s P"
`that all n transmitted blocks are wrong (residual error rate),
`Qn = (1 -p)n
`that all transmitted blocks are right and
`Q(21;n)=1-p“
`that at least one of the transmitted blocks is right (throughput, availability).
`Comparing again RA500,6 (railway) and RA500,0 (standard), the results show that a factor of 1/10 in the
`block error rate of the RA500.6 provides a multiple of powers of ten already for one of the simplest codes.
`the multiple transmission.
`This means that at 500km/h any given code provides a higher transmission security if a railway channel
`can be assumed.
`
`°
`
`°
`
`standard GSM
`
`BER
`P (2 1:256)
`
`0.9225
`
`8.2‘10'3 55-107 14-10“
`0,992
`0.976
`0.921
`0964
`09209
`
`5.7-10‘2 1,8-104 3.6‘10‘13
`0.943
`0,839
`0.556
`0.943
`0.9998
`0.9126
`
`59-103 20-107 5.1-10‘23
`0,994
`0,982
`0,942
`0,994
`0.968
`
`Table 3; Probabilities for n-times transmission of a block of K = 256 information bits in the TCH/H2A.
`
`Comparison of standard GSM and railway channels at SNR = 5dB. (eg. 0938 means 0,9998 or 1-2-10'4).
`
`5/14
`
`Page 000004
`
`
`
`
`
`Summary and conclusions
`6.
`The paper presents results of mobile radio channel measurements in a railway environment The results
`show that in general a railway mobile radio channel is a Rice channel with better quality (higher Rice
`parameter) than standard GSM channels.
`Simulation results of TCH/H2,4 with terrain profile RA show that the high-velocity problem of GSM at
`500km/h can be solved either by increasing the SNR by about ZdB or by increasing the Rice parameter
`from c = OdB (standard GSM channel) to c = 6dB (railway channel). In this case BER and block error rate
`at 500km/h are not worse than at 2SOkm/h with a standard channel.
`A simple example shows that, if a railway channel can be supposed. the benefit of smaller BER and block
`error rate is gained when telegrams are to be transmissioned.
`The desired channel performance, Le. a strong direct path (high Rice parameter), is achievable by carefully
`planning the radio coverage along the railway line. This has the benefit that a GSM standard receiver can
`cope with the high speed problem and no additional, expensive means like antenna diversity or fast adaptive
`equalizer are needed.
`
`Acknowledgement
`The author is indebted to Diplnlng. W. Lautenschlager and Dipl.—Phys. P. Hill from the German Railway,
`who were involved in preparation, organization and carrying-through the measurements and to his
`colleagues Dip1.-Phys. G. Frohlingsdorf, DipI.—Ing. K. D. Masur and Dipl.-Ing. U. Weber for their helpful
`and engaged work in analyzing the measuring results and carrying—out the simulations.
`
`Literarture
`
`[1] M. Geller, D. Miinning, H. Ranch, G. Singer: A future European integrated railway mobile network
`based on GSM.
`Proc. Fifth Nordic Seminar on Digital Mobile Radio Communications. DMRV, 1-3 Dec. 1992
`Helsinki.
`
`[2] M. Goller, K. D. Masur, G. Frohlingsdorf, U. Weber. MeBergebnisse und Parameter zur
`Modellierung von Bahn—Mobilfunkkanalen im 900—MHz-Band.
`Nachrichtentechnik-Elektronik, Berlin 43 (1993) No. 6, pp. 290-295.
`[3] M. Geller, K. D. Masur: Ergebnisse von Funkkanalmessungen im 900-MHz—Bereich auf
`Neubaustrecken der Deutschen Bundesbahn.
`Nachrichtentechnik-Elektronik. Berlin 42 (1992), No. 4, pp. 143-146. No. 5, pp. 206209.
`[4] M. Geller: Radio channel measurements on lines of German Railway (Deutsche Bundesbahn) in the
`900 MHz frequency band.
`COST 231 TD (92) 20. Vienna 7-10 Jan. 1992.
`[5] M. Goller. U. Weber, G. Frohlingsdorf: Results of simulations with railway mobile radio channels for
`high-speed applications of GSM.
`ETSUSTC/SMGZ TDoc 260/ 1993.
`[6] M. Goller. K. D. Masur, U. Weber, G.
`Bahnmobilfunkkanélen und speziellen Bahnszenarien.
`Internal report. 15.11.93.
`[7] Performance of GSM with high vehicular speed. (AEG Mobile Communications)
`ETSl/STC/SMGZ TDoc 97/1992.
`
`Simulationsergebnisse mit
`
`Frohlingsdorf:
`
`5/5
`
`Page 000005
`
`
`
`
`
`.ln5.5.5:,3195.
`
`‘"
`
`
`
`E..5539.x33
`
`E:s...2.35
`
`a?
`
`E.2..e
`
`2.3....»
`
`EE...
`
`IE'PI USDS
`
`on3
`
`.515.3%
`
`e»a.on01onona.eEIt;
`
`.2
`
`
`
`a»:12.3.30!
`
`
`
`5.2.-in
`
`:3Ts.
`
`.5.3-6
`
`:32.“
`
`E0-ancuo»
`ona.oHVoE2....
`
`
`
`.m8.2.3342?mo—W.53.33:32.....sec.-:2.25..I
`
`
`3e.oncuon2a.c.0e.2-.a..
`
`
`5/6
`
`
`
` E..a.u.h_E.312....43....
`
`EE...
`
`E8an2an2ea2.Z..9fi.,
`
`
`rIn:E:._
`
`5-6nfluanena.a
`
`
`
`
`
`
`
`EONI;EMS:.3558;o_m«5.25E39.Exmlu3:520Afiagm5:92.:25—5cue—05:95.52E3.32wctzmuo—z”—.wE
`
`
`
`
`
`
`
`
`
`
`
`Page 000006
`
`
`
`
`
`
`
`
`
`
`
`
`
`r,___‘__,__,
`“SURE/“1‘
`—.—mn.aa.lu-I)W‘
`
`l
`
`
`
`
`
`
`
` Ill 1: "In! CW‘U.‘
`
`
`
`
`
`
`
`
`
`
`__W na-scIu-uw’
`
`Ii Km II "”116
`‘mnhaim-a‘m-7MIVS
`
`
`
`(win...Mann—nuIn.In
`
`Winkllfl-M'llnli
`
`l
`l Fugum : u
`“—‘5—
`l mu: 1.:
`
`Fig. 2: Simulation results of TCH/H2.4 with standard GSM channels RA250 and with railway
`channel RASO0.0 and RASOO.6.
`
`Radio Channel RAx.y = rural area with velocity x [km/h] and Rice - parameter y [dB].
`a) BER as a function of SNR
`
`b)
`
`d) cumulative block error rate as a function of information block length K
`
`5/7
`
`Page 000007
`
`
`
`DECLARATION OF ANDREW F. WILSON
`
`1, Andrew F. Wilson, make the following Declaration pursuant to 28 U.S.C. § 1746:
`
`1.
`
`I am the Head of Governance and Legal Affairs at The Institution of
`
`Engineering and Technology (“IET”).
`
`l have worked at lET, known as the Institution of
`
`Electrical Engineers (“IEE") prior to 2006, since 1980.
`
`I am familiar with the conference
`
`publications published by lET and its predecessor lEE.
`
`2.
`
`Unless otherwise stated. the facts stated in this Declaration are based on my
`
`personal knowledge.
`
`3.
`
`The IET keeps records. and those of its predecessor the IEE, concerning
`
`conference publications in the ordinary course of business.
`
`I have consulted these records in
`
`making this Declaration.
`
`4.
`
`Based on my personal knowledge and the business records 1 have reviewed,
`
`the document attached at Tab A hereto provides a true and accurate copy of the article
`
`entitled “Application of GSM in High Speed Trains: Measurements and Simulations” by
`
`Manfred Goller, which was published by the IEE in 1995 in Colloquium Digest 1995/105,
`
`entitled “Radiocommunications in Transportation.” This document
`
`is in circulation and
`
`publicly available.
`
`I declare under penalty of perjury under the laws of the United States of America that
`
`the foregoing is true and correct.
`
`Executed on September 12, 2012
`
`9/44
`
`Andrew F. Wilson
`
`at Savoy Place. London WC2R OBL, United Kingdom
`
`Page 000008
`
`
`
`Tab A
`
`Page 000009
`
`
`
`
`
`APPLICATION OF GSM IN HIGH SPEED TRAINS: MEASUREMENTS AND SIMULATIONS
`
`Manfred Goller
`
`Abstract
`The paper presents results of measurements and simulations concerning the application of the European
`GSM system in high speed trains travelling at up to 500 krn/h. The aim is to answer the question to what
`extent GSM (performance specified up to 250 km/h) can cope with the high velocities which are demanded
`for future railways. Measurements along railway lines have shown that a railway mobile radio channel
`results in better performance (Rice channel) than standard mobile radio channels (Rayleigh or weak Rice
`channel, see GSM-Rec‘s).
`BER and block error rate of GSM traffic channels up to 500 km/h are simulated. Comparison of the results
`at 250 km/h and 500 km/h shows that the GSM high velocity problem can be solved either by increasing
`the SNR by about ZdB or by increasing the Rice parameter c by about 6dB (numerical values for profile
`RA = rural area; railway channel with c = 6dB against standard channel with c = 0dB). is. the BER at 500
`km/h (railway channel) is not worse than the BER at 250 km/h (stande channel).
`A simple example shows that the benefit in the transmission of telegrams consisting of blocks of decoded
`bits can be much higher.
`The desired channel performance. Le. a strong direct path (high Rice parameter), can be achieved by
`careful radio coverage planning along the railway line. This means a GSM standard receiver is sufficient to
`cope with the GSM high velocity problem and no additional means are needed.
`
`1.
`
`Introduction
`
`The present situation of mobile communications in the European railways is characterized by a multitude
`of different networks using various analog technologies. No common solution exists and the systems are
`not compatible. This situation does not satisfactorily reflect the current process of integration within the
`European Community.
`Considering this. the German Railway is working on a solution with GSM which is in correspondence with
`a decision of the UIC (Union internationale de chemin de fer ) [1].
`Future high-speed trains in Europe are to reach velocities up to 500 km/h.
`Thus for GSM application the question arises to what extent GSM (performance specified up to 250 km/h)
`can cope with the high velocities.
`Useful but expensive measures against this limitation are antenna diversity or a fast adaptive equalizer.
`Railway radio channels are characterized by a coverage in the form of a line along straight railway
`embankments. cuttings and tunnels. The antennas are situated close to the railway line. their heights are
`relatively small and the coverage range is limited to some few kilometers.
`The environment typical to railways suggests that the railway mobile radio channel is a Rice channel.
`which results in better performance c.g. BER than a standard Rayleigh channel. This assumption led to
`radio channel measurements on lines of German Railway, modeling this channel on a computer and
`simulating the performance of GSM. The results of the investigations reported below contribute towards a
`decision as to the suitability of GSM for high-speed railway communication.
`
`Radio channel measurements
`2.
`The radio channel measurements were carried out on lines of German Railway including new built high
`velocity lines. The radio channel is completely characterized by its complex. time variant impulse response
`h (T. t), with T = delay time. t = real time.
`
`Measuring equipment
`The impulse response was measured with the channel sounder RUSK400.
`ATmin = Sps and °Tmin = 1,4115 delay spread which is sufficient for GSM.
`
`It‘s measurement limits are
`
`M. Goller is with the DETECON Gmbl-I. Bonn, Germany
`
`6) 1995 The lnstitutlon 01 Electrical Engineers.
`Printed and published by the IEE, Savoy Place, London WC2R OBL, UK,
`
`5 / 1
`
`Page 000010
`
`
`
`
`
`Measuring conditions
`-
`Frequency band 945 MHz
`-
`Transmitter power 20W 50W
`-
`Transmitting antenna height above rail level
`open terrain 1‘1 = 14m, 20m, tunnel hl = 4m.
`The antenna is situated close to the railway line.
`Polarization vertical
`
`-
`
`-
`-
`-
`
`Receiving antenna height h2 -= 4,3m, omnidirectional. mounted on the top of the measuring coach.
`Measuring distance uansmitter-receiver d s 0, 1, 3,5 km
`Measuring interval so a: 150
`200m
`
`Channel parameter
`The received process
`
`’-
`r0) = r (r) - rm = ,lzlm .0]
`k:l
`r
`l
`k
`
`2
`
`consists of a fast part rs(t) (Rayleigh or Rice fading) and a slow part r1(t) (lognorrnal fading).
`Of interest is the dynamic behaviour of the channel which is determined by the fast fading rs(t).
`By low-pass filtering of the received process r(l) one gets rim and after division r5(t) - r(t)/rl(t).
`In particular the Rice parameter
`channel
`; Gaufl
`P = 0
`to
`;
`channel
`Rayleigh
`;
`Pd = 0
`;
`0
`pd/Pm
`c
`is defined as the power ratio of the signal in the direct path to the multipath-spread signals. It is estimated
`by the aid of a chi-square-test by comparison of the theoretical Rice-distribution with the measured
`empirical distribution of the fast fading [21.
`Additional characteristics of the fast fading are mean value 56— [dB], standard deviation ox[dB] and mean
`fading depth
`AdeB] = X(90)-X(10),
`where X(90) and X(10) are the 90% and the 10% quantile of the empirical distribution function P(x<X) of
`the signal level x[dB] respectively.
`For Rayleigh fading one gets
`
`f=-—2.51dB,
`
`o =5,57dB, AX=13.39dB.
`x
`
`For further details of channel characterization see [2]. [3} and [4].
`
`Measuring results
`3.
`1 shows as an example results of measurements at Milbertshofen. a Marshalling yard and industrial
`Fig.
`area near Munich (Bavaria).
`the
`As can be seen from the delay»Doppler—spectrum (scattering function) and the Doppler spectrum.
`measuring train moves towards the transmitter (angle of incidence = 0°. maximum Doppler frequency
`+fDmax)- A strong direct path exists with a delay of zero (main impulse) and a spike at +fDmax~ A weak
`reflection behind the train (angle of incidence - 180°, maximum Doppler frequency 'fDmax) occurs at a
`delay of 8.6m related to the main impulse (compare with the impulse response and the delay spectrum). It
`results from a highway bridge crossing the line 1.3km behind the measuring train.
`The multipath propagation is weak and in the order of the measuring limits. At locations where deep
`fadings occur the delay spread and the delay window are higher.
`Mean value PS, standard deviation CS and mean fading depth APS of the fast fading already show that
`ISO) is a Rice—process (compare with the theoretical results above for a Rayleigh process). A high Rice
`parameter corresponds to small standard deviation and small mean fading depth and reversed.
`A summary of measuring results is shown in table 1. For further details see [2], [3] and [4].
`In particular the results show that in the delay plane at all investigated railway lines only weak multipath
`propagation was measured. These channels can be described by a delay spectrum RA (rural area) or TU
`
`5/2
`
`Page 000011
`
`
`
`
`
`(typical urban) as specified in GSM-Rec. 05.05. Exception are lines in the Alps which can be characterized
`by the delay spectrum HT (hilly terrain). RA applies for tunnels.
`It should be noted furthermore that in the Doppler plane for a coverage along the line a direct path was
`always measured. The measured Rice parameter was about c = 8...13dB, see table 1. This means that in
`the Doppler plane railway mobile radio channels (Rice) are better than standard mobile radio channels
`(Rayleigh) as specified in GSM Rec. 05.05.
`The results suggest to model a railway mobile radio channel by
`—
`a delay spectrum as defined in GSM-Rec. 05.05 and
`—
`a Doppler spectrum also as defined in GSM-Rec. 05.05 but superimposed by a single spectral line
`resulting from a direct path. It's strength is determined by the Rice parameter.
`Thus the typical railway mobile radio charmel is a Rice channel superimposed by a slow. lognormal fading.
`
`_Rice parameter
`
`slow fading
`stand. dev.
`0 [dB]
`
`Milbertshofen
`
`Kraichtal, NBL
`Glems. NBL
`Eichstati
`Osterrni‘mchen
`Oberau — Garmisch
`Oberau — Miinchen
`
`Table 1: Summary of channel measuring results under standard measurement conditions: distance d:3lcm;
`transmitter: beam antenna 51° horizontal. gain 10,3dBd. height h1=20m, tunnel h1=4m;
`receiver: omnidirectional antenna, heigth h2=4,3m; NBL=newly built high speed line; (xx)=less frequent.
`
`Simulation results
`4.
`To simulate a railway mobile radio channel a 6-path-model as specified in GSM is used. The direct path is
`superimposed on the first path so that quite arbitrary Rice parameters can be generated.
`The simulations were focused to standard GSM—channels RAx and railway channels RAxy, where RA s
`terrain profile rural area. it = velocity in km/h and y = Rice parameter in dB.
`Fig. 2 shows as an example simulation results of the half-rate traffic channel TCH/HZ.4 after convolutional
`decoding (code rate r = 1/3. interleaving depth 1= 19. interleaving delay = l85ms).
`With respect to statistical safety and appropriate simulation time only BERBIO‘5 were simulated.
`The results from Fig. 2a Show that [5], [6]
`-
`the BER goes up when the velocity increases from 250km/h to SOOkm/h;
`-
`the rise of BER can be compensated by increasing the SNR by about ZdB (see also [7]);
`~
`the rise of BER can be compensated by increasing the Rice parameter from c = M8
`(GSM channel) to c: 6dB (railway channel);
`the BER at 500km/h with the railway channel RA500,6 is not worse than the BER at 250km/h
`with the standard GSM channel RAZSO.
`
`-
`
`Block error rate
`in data communications K information bits are combined in blocks or telegrams. A block or telegram is
`wrong if it contains 2 to wrong bits. Then P (2 m: K) is the cumulative probability that in a block with a
`length K are at least in wrong bits. Especially m = 1 describes the block error probability and m = g > i the
`probability that a block contains errors with a weight 2 g.
`Figures 2.b...2.d show the cumulative block error rate of TCH/H2,4 for SNR = 5dB. Table 2 shows the
`results for blocks of K = 256 information bits. Automatic train control could be a possible application of
`this example.
`
`5/3
`
`Page 000012
`
`
`
`
`
`It can be stated that with a railway channel not only the BER, but more important, the cumulative block
`error rate is reduced as well, actually the more the higher the error weight is.
`For example, if we assume an additional error detection after the TCH/H2.4—decoding which can detect
`errors up to a weight e -= d -
`l - 7 over K - 256 bits then P (2 8; 256) is the residual block error rate.
`Comparing RA500.6 and RA500.0, the railway channel provides a factor of 1/10 or 1/38 respectively, by
`which the block or residual error rate is smaller.
`
`P (2 64; 256)
`
`l; 256)
`8', 256)
`
`P (2
`P (2
`
`Table 2: Cumulative block error rate of the TCH/HZ.4 at SNR=5dB and a telegram length K - 256.
`Radio channel RAx,y = rural area with velocity xlkm/h] and Rice parameter y[dB].
`
`Example
`5.
`the
`to what extent
`The following simple example illustrates, for the case of automatic train control,
`performance is influenced by the radio channel. The results are to be understood as an upper limit which is
`achievable more effectively by an appropriate error detecting or error correcting code.
`Suppose that
`information blocks (telegrams) of the length K = 256 are transmitted n times in the
`TCH/H2,4 with a block error rate p = P( 2 l; 256) as shown in table 2 for SNR = 5dB.
`Statistical independence assumed. table 3 shows the probability
`'
`p11 s P"
`that all n transmitted blocks are wrong (residual error rate),
`Qn = (1 -p)n
`that all transmitted blocks are right and
`Q(21;n)=1-p“
`that at least one of the transmitted blocks is right (throughput, availability).
`Comparing again RA500,6 (railway) and RA500,0 (standard), the results show that a factor of 1/10 in the
`block error rate of the RA500.6 provides a multiple of powers of ten already for one of the simplest codes.
`the multiple transmission.
`This means that at 500km/h any given code provides a higher transmission security if a railway channel
`can be assumed.
`
`°
`
`°
`
`standard GSM
`
`BER
`P (2 1:256)
`
`0.9225
`
`8.2‘10'3 55-107 14-10“
`0,992
`0.976
`0.921
`0964
`09209
`
`5.7-10‘2 1,8-104 3.6‘10‘13
`0.943
`0,839
`0.556
`0.943
`0.9998
`0.9126
`
`59-103 20-107 5.1-10‘23
`0,994
`0,982
`0,942
`0,994
`0.968
`
`Table 3; Probabilities for n-times transmission of a block of K = 256 information bits in the TCH/H2A.
`
`Comparison of standard GSM and railway channels at SNR = 5dB. (eg. 0938 means 0,9998 or 1-2-10'4).
`
`5/14
`
`Page 000013
`
`
`
`
`
`Summary and conclusions
`6.
`The paper presents results of mobile radio channel measurements in a railway environment The results
`show that in general a railway mobile radio channel is a Rice channel with better quality (higher Rice
`parameter) than standard GSM channels.
`Simulation results of TCH/H2,4 with terrain profile RA show that the high-velocity problem of GSM at
`500km/h can be solved either by increasing the SNR by about ZdB or by increasing the Rice parameter
`from c = OdB (standard GSM channel) to c = 6dB (railway channel). In this case BER and block error rate
`at 500km/h are not worse than at 2SOkm/h with a standard channel.
`A simple example shows that, if a railway channel can be supposed. the benefit of smaller BER and block
`error rate is gained when telegrams are to be transmissioned.
`The desired channel performance, Le. a strong direct path (high Rice parameter), is achievable by carefully
`planning the radio coverage along the railway line. This has the benefit that a GSM standard receiver can
`cope with the high speed problem and no additional, expensive means like antenna diversity or fast adaptive
`equalizer are needed.
`
`Acknowledgement
`The author is indebted to Diplnlng. W. Lautenschlager and Dipl.—Phys. P. Hill from the German Railway,
`who were involved in preparation, organization and carrying-through the measurements and to his
`colleagues Dip1.-Phys. G. Frohlingsdorf, DipI.—Ing. K. D. Masur and Dipl.-Ing. U. Weber for their helpful
`and engaged work in analyzing the measuring results and carrying—out the simulations.
`
`Literarture
`
`[1] M. Geller, D. Miinning, H. Ranch, G. Singer: A future European integrated railway mobile network
`based on GSM.
`Proc. Fifth Nordic Seminar on Digital Mobile Radio Communications. DMRV, 1-3 Dec. 1992
`Helsinki.
`
`[2] M. Goller, K. D. Masur, G. Frohlingsdorf, U. Weber. MeBergebnisse und Parameter zur
`Modellierung von Bahn—Mobilfunkkanalen im 900—MHz-Band.
`Nachrichtentechnik-Elektronik, Berlin 43 (1993) No. 6, pp. 290-295.
`[3] M. Geller, K. D. Masur: Ergebnisse von Funkkanalmessungen im 900-MHz—Bereich auf
`Neubaustrecken der Deutschen Bundesbahn.
`Nachrichtentechnik-Elektronik. Berlin 42 (1992), No. 4, pp. 143-146. No. 5, pp. 206209.
`[4] M. Geller: Radio channel measurements on lines of German Railway (Deutsche Bundesbahn) in the
`900 MHz frequency band.
`COST 231 TD (92) 20. Vienna 7-10 Jan. 1992.
`[5] M. Goller. U. Weber, G. Frohlingsdorf: Results