On the Feasibility of High Speed Railway mmWave Channels in Tunnel...

Research ArticleOn the Feasibility of High Speed Railway mmWave Channels inTunnel Scenario

Guangkai Li1 Bo Ai12 Danping He12 Zhangdui Zhong12 Bing Hui3 and Junhyeong Kim34

1State Key Laboratory of Rail Traffic Control and Safety Beijing Jiaotong University Beijing 100044 China2Beijing Engineering Research Center of High Speed Railway Broadband Mobile Communications Beijing 100044 China3Mobile Application Research Department Electronics and Telecommunications Research Institute (ETRI)Daejeon 34129 Republic of Korea4School of Electrical Engineering Korea Advanced Institute of Science and Technology (KAIST) Daejeon 34141 Republic of Korea

Correspondence should be addressed to Bo Ai boaibjtueducn

Received 21 May 2017 Revised 22 August 2017 Accepted 24 August 2017 Published 11 October 2017

Academic Editor Francesco Benedetto

Copyright copy 2017 Guangkai Li et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Rail traffic is widely acknowledged as an efficient and green transportation pattern and its evolution attracts a lot of attentionHowever the key point of the evolution is how to develop the railway services from traditional handling of the critical signalingapplications only to high data rate applications such as real-time videos for surveillance and entertainmentsThe promisingmethodis trying to use millimeter wave which includes dozens of GHz bandwidths to bridge the high rate demand and frequency shortageIn this paper the channel characteristics in an arched railway tunnel are investigated owing to their significance of designing reliablecommunication systemsMeantime asmillimeter wave suffers fromhigher propagation loss directional antenna is widely acceptedfor designing the communication systemThe specific changes that directional antenna brings to the radio channel are studied andcompared to the performances of omnidirectional antenna Note that the study is based on enhanced wide-band ray tracing toolwhere the electromagnetic and scattering parameters of the main materials of the tunnel are measured and fitted with predictingmodels

1 Introduction

Mobile communication is one of the most essential and suc-cessful technology progress types which becomes an indis-pensable part of more than 5 billion people [1] With themobile data demand exponentially growing [2ndash5] wire-less communication systems working at millimeter wave(mmWave) have attracted much more attention in manymobile communication scenarios For mmWave it poten-tially contains a large amount of spectrum resources forachieving multi-Giga bps data rate for wireless communica-tion systems [6ndash8]

As a typical application scene of mobile communicationhigh speed railway (HSR) with speed of more than 300 kmper hour challenges the constantly improved mobile com-munication systems [9] The most influential challenges forHSRmmWave communication system include frequent han-dovers difficult signal processing as very high speed andhigh

penetration loss of signal from base-station to intrawagonuser [10] However for achieving multi-Giga bps data rate inhigh speed railway (HSR) the existing HSR dedicated com-munication systems are no longer satisfactory For examplethe maximum data rate for the Global System for MobileRadio Communications for Railway (GSM-R) is less than200 kbps Even for the Long Term Evolution for railway(LTE-R) it cannot provide more than 100Mbps data rateTherefore an interest group of High Rate Rail Communica-tion (HRRC) has been established in the IEEE 80215 work-ing group for developing advanced mobile communicationtechnologies and it targets final achievement of very highdata rate for HSR wireless communication system Also adistributed antenna system (DAS) [11] has been designedand it becomes a very promising communication system forHSR andmetro system [10] However all these desiderate fullunderstanding of the channel and an adequate and reliablechannel model [12]

HindawiWireless Communications and Mobile ComputingVolume 2017 Article ID 7135896 17 pageshttpsdoiorg10115520177135896

2 Wireless Communications and Mobile Computing

Since channel sounding measurements are treated as apriority when studying channels in the era of 3G and 4Gcommunication system the channel parameters and modelsare extracted mostly from massive measurement campaigns[13] But the measurement at mmWave band greatly chal-lenges the existent sounding systems especially for high-dynamic channel measurement The HSR mmWave channelmeasurement data is even rare [3] So we propose anapproach that enhances the capability of deterministic chan-nel model for obtaining mmWave HSR channel Then someimportant channel characteristics are studied and modeledfor guiding future HSR mmWave communication systemdesign

In this paper we simulate an important wireless chan-nel by an enhanced ray tracing simulator where the elec-tromagnetic (EM) and scattering parameters of dominantmaterials in the scenario aremeasured and fittedwith suitablemodels The simulation configuration practically follows thecommunication system demands in [15] Meanwhile a time-interpolation method is employed for promoting the simula-tion capability for high-dynamic channels Based on exten-sive simulations the channel characteristics in an archedtunnel are completely exposed in two defined regions More-over the channel feasibilities of using different antennas areinvestigated for determining the most suitable one

The remainder of this paper is organized as followsSection 2 describes principles of the proposed ray tracingtool with material measurement and time-interpolation InSection 3 simulation scenario and system setups will bepresented with more details Section 4 demonstrates thefeasibility of communication system with different antennasetups Section 4 also illustrates the necessity of partition ofregions in radio channel analysis Then characteristics ofradio channel are provided in Section 5 Finally Section 6gives the conclusion

2 Ray-Optical Based mmWave ChannelModeling for High-Dynamic Scenario

The wireless channel is simulated by utilizing ray-opticalbased deterministic propagation model that is ray tracing(RT) model As it differs from other traditional RT theenhanced RT used in this study has been calibrated invarious frequencies and scenarios [16ndash19] We make a spe-cial measurement for extracting electromagnetic (EM) andscattering parameters of the most influential materials in theHSR tunnel The extracted parameters will be implementedinto RT simulator Moreover for investigating the detailedchannel characteristics in such high-dynamic scenario awell-studied RT time-interpolation algorithm is employed toexpose the channel small-scale characteristics

The RT in this study is developed based on ray-opticalalgorithm [20] at Technische Universitat Braunschweig [21]It is three-dimension (3D) channel simulator which is per-formed in 3D digital map For an integrated channel mod-eling several radio propagation mechanisms are taken intoconsideration as line-of-sight (LOS) ray reflection rays andscattering rays The LOS ray is known as free-space wavepropagation between transmitter (TX) and receiver (RX) Its

power is calculated by free-space path loss model in RT TheLOS path if it exists will dominate the received power in theRXThen the reflectionwhich is defined as the incident angleand reflected angle of EMwave from a surface is the same Itspower is calculated by solving the well-known Fresnel equa-tions and the related reflection point in the 3D digital map isobtained by applying the image method [20] Still in the caseof reflection 119899th order multiple reflections will be practicallyconsidered for capturing dominant power contributions oftotal received power For the diffuse scattering the powercontribution of diffuse scattering is evaluated by adoptingthe ldquoeffective roughnessrdquo (ER) scattering models [22] theER models generally include Lambertian Model DirectiveModel and Backscattering Lobe Model for predicting dif-ferent types of scattering patterns [22] Owing to the specialstructure of the tunnels where no obstacles are situatedthe transmission through the materials and diffraction fromedges of obstacles are omitted in the simulation

Based on ray-optical principle and several propagationmechanism models the final output of the RT is a time-variant channel impulse response (CIR) ℎ(120591 119905) which is thesum of powers of all the determined rays [23]

ℎ (120591 119905) = 119873(119905)sum119896=1

119886119896 (119905) sdot 119890119895(2120587119891120591119896(119905)+120593119896(119905)) sdot 120575 (120591 minus 120591119896 (119905)) (1)

where 119886119896(119905) 120591119896(119905) and 120593119896(119905) denote amplitude delay andadditional phase shift of 119896th ray (totally 119873(119905) rays that havebeen found by RT kernel)

Then119860119896(119905) = 119886119896(119905)sdot119890119895(2120587119891120591119896(119905)+120593119896(119905)) is defined as a complexcoefficient formula (1) can be rewritten as

ℎ (120591 119905) = 119873(119905)sum119896=1

119860119896 (119905) sdot 120575 (120591 minus 120591119896 (119905)) (2)

Then (2) attached with the effects of wave polarizationand antenna gains 119860119896(119905) can be further expressed by

119860119896 (119905) = radic119866RX119896 sdot 997888rarr119890 (120593RX119896 120579RX119896)119867 sdot 119875119896 (119905) sdot radic119866TX119896

sdot 997888rarr119890 (120593TX119896 120579TX119896) sdot 119871119896 (119905) (3)

where 120593RXTX119896 and 120579RXTX119896 indicate the angle of arrival(AOA) or angle of departure (AOD) of the 119896th multi-path component Therefore the vectors 997888rarr119890 (120593RX119896 120579RX119896) and997888rarr119890 (120593RX119896 120579RX119896) denote the complex polarization vectors [24]of antennas at RX and TX respectively 119866RX119896 and 119866TX119896indicate the additional antenna gains of the 119896th multipathcomponent in RX and TX (sdot)119867 denotes the Hermitian trans-pose 119875119896(119905) depicts the complex channel polarization matrixwhich indicates the polarization shifting of the ray 119871119896(119905)comprises propagation loss and the phase shift accordingto the delay 120591119896(119905) of the 119896th multipath component Moreinformation can be found in [21]

For investigating mmWave channel it is still the chal-lenge that the mmWave channel sounding measurements arevery costly and time-consuming even for static scenariosFor measuring high-dynamic mmWave channel it is even

Wireless Communications and Mobile Computing 3

minus90∘

RX

RX

RX

RX

TX

90∘

0∘

Material

Measurement

）Ｈ＝Ｃ

３＝Ｎ

(a)

TXRX

）Ｈ＝Ｃ３＝Ｎ

(b)

Figure 1 Study on EM and scattering parameters of cement wall (a) Geometry of material study where the tunnel image is from [30] (b)Measurement campaign

unrealistic tomeasure ammWave channel sample at a normalHSR speed (around 300Kmh in China) So the RT is apromising tool for revealing the channel characteristics inHSR scenarios It is also a trade-off between the limitation ofmeasurement and demand for the channel data To enhancethe RT capability of accurately predicting channels in HSRtunnel scenario we perform two approaches as mentionedin the initial part of this chapter which are measuring theEMand scattering parameters of themost influentialmaterialand utilizing RT time-interpolation algorithm to extract thesmall-scale channel parameters

21 EM and Scattering Parameters Acquisition of the MaterialTheaccurate energy calculations of the reflected and scatteredrays in RT simulation depend on accurate EM (relativepermittivity 120576119903 in our study) and scattering parameters ofmaterials as well as accurate 3D digital map of the scenarioGenerally RT simulation starts with determining the mate-rials which constitute the scenario then 120576119903 and scatteringparameters of these selected materials are acquired fromchecking literatures (eg ITU-R recommendations [25])performing dedicated EM measurement [26] or derivationfrom channel measured data [27] However apart from ded-icated EMmeasurement these methods are less than ideal as120576119903 and scattering parameters of a material vary in differentscenarios and conditions (eg humidity of environment anddensity of material) Therefore we measure the interferencesof the material on the wave propagation at 325 GHz using aself-built testbed and estimate 120576119903 and scattering parametersaccordingly And the material in this study is sulphoalumi-nate cement which is generally applied in HSR tunnels (cfFigure 1(a)) unlike classic cerement utilized in buildings ithas advantages of quick-drying high strength compactionand cementationwhich requires a dedicatedmeasurement forstudying its unknown 120576119903 and scattering parameters

As it is shown in Figure 1(b) the self-built testbed is on thebasis of two high-accurate rotatable arms which could leadan accurate 2D scanning of the scattering of a material theTX and RX are with horn antennas and the Vector NetworkAnalyzer (VNA) used in the measurement is manufacturedby Keysight Corporation with model N5247A The VNA

measures 119878 minus 119901119886119903119886119898119890119905119890119903 between RF-ports of two cablesafter the end-to-end calibration process In themeasurementthe material is hanged upside the rotating center by arope for eliminating some unwanted interference which isconsistent with the function of the anechoic chamber As itis illustrated in the right of Figure 1(a) the diffuse scatteringdata is obtained by rotating the RX while the TX is fixedand when 120593Inci = 120593Scat we get the reflection data Aftermeasurement the data are postprocessed with proper filterto attenuate some interferences Then we can estimate 120576119903 ofthe material by the method of free-space measurement basedpartly on [28] and the estimated 120576119903 is verified and slightlytuned with the measured refection coefficient accordingly(cf Figure 2(a))Meanwhile the scattering parameters can beestimated by the method similar to [22 26] but additionallywith SimulatedAnnealingAlgorithm [29] to obtain the betterscattering parameters by automatic minimising of the gapbetween fitting results and measurement (cf Figure 2(b)) Itshould be noted that the Directive Model as one of mostimportant ER models was employed to fit the measuredscattering radiation patterns therefore frequency-dependent119878(119891) and 120572119877(119891) of DirectiveModel are the scattering parame-ters that should be extracted from estimation process [22 26]

As it is described above a brick made of sulphoalu-minate cement was measured Figure 2 illustrates the esti-mation processes of relative permittivity (120576119903) and scatteringparameters (119878 and 120572119877) at 325 GHz Figure 2(a) gives thecomparison of reflection coefficients between fitting curveand measurement 120576119903 of the fitting curve is 120576119903 = 347 minus 119895015where the mean error (ME) and standard deviation (Std)betweenmeasurement andfitting curve are 00070 and 00141respectively Furthermore the Directive Model is fitted withmeasurement in various incident positions for exampleFigure 2(b) gives a scattering fitting curve when 120593Inci = 70∘Themost suitable 119878 and120572119877 for the scattering fitting are 000118and 120 respectively The ME and Std between fitting curveand measurement are 216119890 minus 4 and 577119890 minus 4 respectively22 Interpolation Algorithm in RT The main drawbackof the RT is the high computational cost according tothe complexity of 3D digital map For dynamic channel


Fitting results Measurement

）Ｈ＝Ｃ

908070605040302010003

04

05

06

07

08

09

1Re

flect

ion

coeffi

cien

t

(a)


0

2

4

6

8

Am

plitu

de o

f S-p

aram

eter

times10minus3

300 60 90minus60 minus30minus90３＝Ｎ

）Ｈ＝Ｃ = 70∘

(b)

Figure 2 The comparison between fitting curve and measurement of cement brick at 325 GHz (a) Reflection (b) Scattering

characteristics learning especially the HSR channel a hightime resolution (Δ119905 = 119905119899 minus 119905119899minus1) of the channel is requiredfor the study of both large-scale and small-scale channelcharacteristics In this study the distance between TX andRX ranges from zero to one kilometer (km) As the channel at30GHz band is studied the distance interval between ℎ(120591 119905119899)and ℎ(120591 119905119899minus1) should be small enough (normally less thanhalf-wavelength of EM wave which is 5 millimeter (mm)in this study [31]) The geometry-based path interpolationis employed to overcome the impractical simulation timedue to the high computational complexity [32] The basicidea of the interpolation is to obtain the information aboutthe continuous propagation paths between two consecutivescenario snapshots Then the linear interpolation will beperformed between the two continuous paths The detaileddescription of this algorithm is in [32]

For this simulation in an arched tunnel the initial timeresolution is 10 milliseconds (ms) in the simulation so theinitial interval between two sampled snapshots is 1m as thespeed of HSR is assumed to be 360 kmh in this study Thenthe interpolation algorithm is preformed between each twosnapshots By utilizing geometry-based ray information themethod of interpolation will drop time resolution to 2ms(2mm interval in distance) in this studyTherefore with verysmall time resolution the extraction of small-scale fadingparameters can be guaranteed

3 Simulation Scenario and SystemSetups in the Tunnel

31 Tunnel Scenario in Simulation In this study the straightarched tunnel is employed as the HSR tunnel scenarioFigure 3 shows the overview of the tunnel scenario in thesimulation Figure 3(a) illustrated that the cross section oftunnel includes the accurate dimension and the locationsof TX1 and RX1 According to realistic ldquoType IIrdquo tunnel

described in [14] the arched tunnel in this study is withdimension 841m times 687m (119882tunnel times 119867tunnel where119882tunneland119867tunnel are defined as maximum width and height of thetunnel resp) The heights of TX1 and RX1 are 65m and 3mrespectively Both TX1 and RX1 are located in the middleof the tunnel The prior works show that totally 18 smoothsurfaces which constitute the tunnel digital map can provideeffective results and keep the computational complexity atreasonable level

As depicted in Figure 3(c) the train is installed with twoantennas in the head and the tail respectively the distancebetween two adjacent base stations (BSs) is 1 km the trainis assumed to be 200m long As indicated in Figure 3(c)the tail antenna (RX1) communicates with the backward BS(TX1) while the head antenna (RX2) communicates withthe forward BS (TX2) Note that in this study only thechannel characteristics of Link 1 are investigated with thesymmetric manner of Link 1 and Link 2 Furthermoretwo radiation lobes of antennas in Figure 3(c) are used toillustrate the pointing directions of the directional antennaswhich are used in the TXs Moreover Figure 3(b) illustratesa snapshot in the simulation In order to characterize thechannel in this tunnel the simulation scenario and systemsetups in this study follow the real requirements of themobilecommunication system described in [15] The scenario andsystem setups of the simulation are listed in Table 1

32 Antenna Setups As it is widely recognized that thedirectional antenna is indispensable for mmWave communi-cation system the detailed effects of directional antenna onmmWave channels still lack a careful investigation especiallyfor HSR channels In other words though the proper use ofdirectional antenna gives a high compensation to receivedpower the distinctions of the radio channels characteristicsbetween applying directional antenna and applying omnidi-rectional antenna are not clear Therefore in this study three


y-axis

x-axis

TX1 65 m height

RX1 3 m height

85685134170

64

56

48

40

32

24

16

08

0

HＮＯＨＨＦ

(m)

WＮＯＨＨＦ (m)

(a)

RX1

TX 1

Link 1

rarr ＮＬＣＨ

(b)

RX1

RX2TX2

TX1

1 km interval

Link 1

Link 2

(c)

Figure 3 (a) Cross section of the tunnel in the simulation (b) One snapshot of the simulation in the tunnel (c) The sketch map for thescenario

Table 1 Computation time (min) of total 1000 snapshots for orders of reflection

1st 2nd 3rd 4th 5th 6th 7th 8th28 (s) 57 (s) 114 (s) 361 (s) 2126 (s) 15 times 103 (s) 11 times 104 (s) 84 times 104 (s)antenna setups are employed under various combinations ofdirectional antenna and omnidirectional antenna refer toFigure 4 They are defined as follows

(1) Direc-Direc TX1 and RX1 are both with the direc-tional antennas The antenna at TX1 is staticallypointing along the tunnel while the antenna at RX1is pointing at opposite direction of TX1 antenna

(2) Direc-Omni TX1 is with directional antenna andRX1 is with omnidirectional antenna The antenna atTX1 is statically pointing along the tunnel

(3) Omni-Omni TX1 and RX1 are both with the omni-directional antennas

The directional antenna is designed by drawing sinusoidalchart in polar coordinate as horizontal and vertical patternsThe following studies will be carried out in describing radiochannel characteristics of three antenna setups

33 Number of Frequencies in Simulation As the simulationconfiguration should follow the real mobile communicationsystemdescribed in [15] the simulation frequency bandwidth

is set to be 125MHzMoreover wewant to expose the channelcharacteristics over a frequency range of 315 GHzsim335 GHzSo the number of the center frequency points 119873119891 is chosensufficiently enough Therefore total 264 center frequencypoints are considered in this study [9 33 34]

34 Order of Multireflection in Simulation The computa-tional complex of RT simulation is significantly affected bymultireflection which requires massive cyclic and traversalsearch of RT kernel Although higher order of reflectiongreatly decreases the efficiency of RT simulations the higherorder of reflection gives more accurate simulation resultsTherefore the order of reflection should be selected verycarefully Before massive RT channel simulations we studiedthe power contributions of each order of reflection in thesame arched tunnel in Omni-Omni case Figure 5 illustratesthe percentages of overall power of each order of reflectioncompared to total received power In the figure on the wholeit is apparent that the overall received power which includespower contributions from LOS to 5th order of reflectionstrikingly reaches 99 of total received power Accordingly


3D omnidirectionalantenna pattern

minus10123

minus2

3D directional antenna pattern with 8∘ beamwidth (3 dB width) and 22 dBi antenna gain

2

20

15

10

5

0

minus15 minus10 minus5 0minus20

(dBi)

Figure 4 The directional antenna and omnidirectional antenna in this study

Percentage ()

lt1

lt01

lt001

0 100 150 200 25050Distance (m)

LOS

1st

2nd

3rd

4th

5th

6th

7th

8th

Pow

er co

ntrib

utio

n

0

10

20

30

40

50

60

Figure 5 Percentages of overall power contribution of each reflec-tion order with distance changes in Omni-Omni case

we recorded the RT computation time of 1000 snapshots forconsidering up to 119899th order of reflection (cf Table 1) Ascan be seen the computation time is of exponential growthwith the increase of reflection order Comparing results ofFigure 5 andTable 1 the percentages of power contribution ofreflection orders higher than 6th are less than 01while theycomputation time are strikingly hundred times larger thanthat of lower reflection orders Therefore we limited maxi-mum reflection order at 5th which gives accurate simulationresults while keeping the computation time acceptable

Finally Table 2 gives an overview of scenario setups andsimulation configuration in this study

4 Study in System Feasibility andPartition for Regions

41 System Feasibility with Different Antenna Setups Inwireless communication systems coverage of the system is

Table 2 Scenario setups and simulation

Tunnel type [14] Arched tunnelAntenna types Directional antenna(Figure 4) Omnidirectional antennaMaterial of tunnel Sulphoaluminate cementMaterial permittivity 120576119903 = 347 minus 119895015Material scattering parameters 119878 = 000118 120572119877 = 120Tunnel length 1 kmHeights of TX and RX 65m and 3mSpeed of HSR 360 kmhSystem bandwidth for a link 125MHzFrequency range investigated 315 GHzsim335 GHzTransmit power 30 dBm [6]Cable loss 6 dBFinal channel sample interval 2mm

defined or controlled by a minimum required signal-noiseratio (SNR) The SNR is calculated by

SNR (dB) = 119875 minus (minus174 + 10 sdot log10 (119882) + 119873119865) (4)

The value 119875 is the received power without small-scale fadingwhich is excluded by averaging received signal with a 40120582slidingoverlapped window [35]119882 is the channel simulationbandwidth it is 125MHz in this study119873119865 is the noise figurewhich is the noise factor expressed in decibel Here 119873119865 ispractically assumed to be 10 dB Further the number minus174 iswidely used as spectral noise power density for 1Hz

As shown in Figure 6 the channel performances ofdifferent antenna setups vary considerably in the near region(which is defined in Section 42) In Direc-Direc and Direc-Omni cases with the RX1 moving away from the TX1 theLOS component gradually enters the illumination of themain lobe (3 dB beam width) of the directional antennaThisprocess causes the SNR obvious increases in the distanceranging from 15m to 50m Afterwards when RX1 movesinto far region the LOS and lots of NLOS components enter


Direc-DriecDirec-Omni Omni-Omni

Near region Far region

minus40minus30minus20minus10

0102030405060

Sign

al-n

oise

ratio

(dB)

50100 200 300 400 500 600 700 800 900 10000Distance (m)

Figure 6 SNRs of different antenna setups The red vertical solidline is the partition for two regions the black horizontal line depictsthe minimum threshold for the available SNR

the illumination of the main lobe of the directional antennaObviously the fluctuations of the SNRs of different antennasetups undergo a similar tendency in far region

According to Figure 6 if the minimum SNR is 10 dB fora reliable detecting [6] that is system minimum availablethreshold this communication system in tunnel can supportmore than 1 km coverage range in the Direc-Direc andDirec-Omni cases But it is difficult to support 1 km signalcoverage when deep shadow fading exists In the Omni-Omni case the system can only support coverage range lessthan 50m Although it is obvious that the directional antennabrings a better performance of signal coverage the detailedchannel characteristics are still under research

42 Definition of Regions for Radio Channel Analysis As it isshown in Figure 6 the received SNRs are obviously differentin different regions The reasons are mainly depending onthe antennas used in simulation for example half-powerbandwidth (HPBW) pointing direction and position [36]The following channel characteristics should be studied indifferent regions Figure 7 gives the sketch of partition forregions 119867BS and 119867Ant are the heights of BS (TX1) and RX1respectively 120579 and 120572 are the inclination angle and HPBWof the directional antenna (TX1) The red solid line indicatesthe pointing direction of the TX1 The two black dotted linesdepict the region which will be illuminated by the antennamain lobe The value 119863 determines the boundary betweennear region and far region which can be calculated by [36]

119863 = 119867BS minus 119867Anttan (120579 + 1205722) (5)

In this study 120572 = 8∘ 120579 = 0∘ 119867BS = 65m and 119867Ant = 3mThe angle (120579 + 1205722) indicates the pointing direction plus halfof the (elevation) HPBW As a result the length of 119863 in thisstudy is 50m

D

RX1 RX2

TX1


= 0∘ (inclination angle)

H３

H３ minus HＨＮ

HＨＮ

= 8∘ (HPBW)

Figure 7 Sketch for region definition

5 The Detailed mmWave ChannelCharacteristics in HSR Tunnel Scenario

Here the radio channel characteristics will be presented intime frequency and polarization domains in order to help indesigning a robust and sophisticatedwireless communicationsystem in HSR scenarios

51 Path Loss and Shadow Fading Extraction The large-scale fading (including path loss and shadowing fading) areobtained by averaging received power with a 40-wavelengthwindow [35] The large-scale fading is generally expressed asa log-distance path loss model 119871(119889)with a path loss exponent(119899)

119871 (119889) = 119871 (1198890) + 10119899 lg ( 1198891198890) + 119883120590 (6)

where 119871(119889) is the function of 119889 which indicates the distancebetween TX1 and RX1 119871(1198890) is the intercept value at refer-enced distance (1198890) 119883120590 is the shadow fading Figure 8 givesone example of extraction process of 119899 at a frequency centerof 325 GHz in far region Note that the following analysesfor path loss and shadow fading are mainly in far regionThis is because in the near region the fluctuation of thereceived power is largely dominated by antenna radiationpattern

In Figure 8 a red solid line indicates the least squarefitting result of the simulated data (marked in blue)The pathloss exponent is the slope of the red solid line It is around11 in far region which indicates the small attenuation ofwave propagation in tunnel This character may stem fromthe waveguide effects caused by tubular structure of thetunnel where the reflection attenuation will be small enoughwhen incident angles of reflected rays are quite large in farregion The path loss exponents of different antenna setupsat frequencies in the range 315 GHzsim335 GHz are calculatedand shown with statistical values in Table 3

52 Amplitude Distribution of Shadow Fading As expressedin (6) the shading fading 119883120590 can be extracted from thelarge-scale fading 119871(119889) 119883120590 is conventionally modeled as alog-normal distribution [14 36] which is confirmed in thisstudy


Table 3 Channel fading characteristics statistics

Setups Direc-Direc Direc-Omni Omni-OmniPath loss exponent (119899)

Min 106 106 107Mean 110 110 111Max 113 113 113

Shadow fading standard deviation (120590) [dB]Min 332 339 342Mean 343 349 347Max 349 354 351

Decorrelation distance [m]119889cor 119889 119889 119889 119889 119889 119889[119898] (05) (119890minus1) (05) (119890minus1) (05) (119890minus1)10 120 160 090 140 080 10050 203 275 189 260 173 23090 381 501 341 470 334 431Mean 246 336 208 309 189 268

Rician 119870-factor [dB]Regions Near Far Near Far Near Far10 590 minus779 578 minus828 375 minus82950 1948 minus698 1059 minus747 728 minus74890 3657 minus229 2085 minus360 1262 minus464Mean 2031 minus580 1228 minus654 780 minus685

Root mean square delay spread [ns]Regions Near Far Near Far Near Far10 011 040 067 044 262 04450 032 047 194 052 482 05890 572 054 572 069 572 120Mean 212 047 297 054 442 070

Root mean square Doppler spread [kHz]Regions Near Far Near Far Near Far10 040 040 041 040 050 04050 042 041 055 041 094 04190 246 042 246 042 245 042Mean 095 041 104 041 120 041

XPD120579 [dB]Regions Near Far Near Far Near Far10 1044 minus580 807 minus647 544 minus66150 2845 300 1872 220 1417 22690 4111 1293 2903 1263 2345 1277Mean 2684 313 1866 254 1434 249


Figure 9 describes the probability density function (PDF)of shadowing fading in Direc-Direc case at 325 GHz withthe results of 120583 = minus17052 dB and 120590 = 34443 dB althoughas the shadowing fading was extracted by (6) some deep

fading inevitably leads to a no-zero value of 120583 Howeverthe modeled 120590 is still valuable for studying the channelshadowing characteristics in tunnel The statistic value of 120590at whole frequencies can be found in Table 3


Path loss + shadow fadingFitted path loss

50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060

Figure 8 Path loss fitting at 325 GHz in Direc-Direc case in farregion

Simulation

Fitted normal distribution

0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n

0 10 20 30minus10Shadow fading (dB)

= minus17052

= 34443

Figure 9 Fitting for amplitude distribution of shadow fading

521 Autocorrelation of Shadow Fading As shadowingcauses the channel deep fading the communication linkstend to be interrupted (refer to Figure 6) To overcome thepotential communication interruption the autocorrelation ofshadow fading should be well-studied The autocorrelationcoefficient of shadow fading is one important characteristicfor designing distributed antenna system which is defined as

12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)

where 119864sdot denotes the expectation 119878(119889) is the expression ofthe shadow fading at distance 119889 120590(119889) is the expression ofthe standard deviation for the shadow fading at distance 119889

Simulated autocorrelation coefficient of shadow fading

0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding

40 60 8020 100Distance between Tx and Rx (m)

80216J model with d＝ＩＬ(eminus1)

80216J model with d＝ＩＬ(05)Exponential model with d＝ＩＬ(e

minus1)

Exponential model with d＝ＩＬ(05)

Figure 10 Autocorrelation coefficient of shadow fading in Direc-Direc case at 325 GHz in far region

Further two widely used empirical models are employed tofit the autocorrelation coefficient [14] the exponential modeland 80216J model The former is accepted in WINNER IImodel

120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)

the latter is presented in standard IEEE 80216J

120588exp (Δ119889) = 119890(minusΔ119889119889cor)sdotln 2 (9)

In (8) and (9) Δ119889 is the distance between two interestedpositions (1198891 and 1198892) There are mainly two definitions fordecorrelation distance 119889cor 119889cor (05) and 119889cor (119890minus1) Theypresent the correlation coefficient equal to thresholds 05 and119890minus1 respectively [14] Obviously these two models have samestructure Figure 10 gives the autocorrelation coefficient ofshadow fading in Direc-Direc case at 325 GHz in far regionIn conclusion 80216J model is fitting well when threshold is05 The exponential model performs better when thresholdis 119890minus1522 Decorrelation Distance of Shadow Fading Compen-sating the deep shadowing fading of the channel is gener-ally used in multiantennas technology Therefore antennasshould be separated long enough to obtain channel diver-sity gain This distance is so-called decorrelation distanceThe decorrelation distances for each RX1 position alongthe tunnel is extracted by using (7) with two thresholds(05 and 119890minus1) at whole frequencies in the range 315 GHzsim335 GHz Figure 11(a) is an example of decorrelation distanceat 325 GHz in far region Obviously there are rare differencesamong decorrelation distances of different antenna setupsAround a distance of 150m the Direc-Direc case reveals alonger decorrelation distance as shown in Figure 11(a) For


Direc-Direc Direc-OmniOmni-Omni

0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion

200 400 600 800 10000Decorrelation distance (m)

(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

5 10 15 200Decorrelation distance (m)

Direc-Direc with d＝ＩＬ(05)

Direc-Omni with d＝ＩＬ(05)

Omni-Omni with d＝ＩＬ(05)

Direc-Direc with d＝ＩＬ(eminus1)

Direc-Omni with d＝ＩＬ(eminus1)

Omni-Omni with d＝ＩＬ(eminus1)

(b)

Figure 11 (a) Decorrelation distances of three antenna setups at 325 GHz in far region (b) For three antenna setups the figure shows theCDF of the decorrelation distances at whole frequencies with thresholds 05 and 119890minus1 in far region

the whole frequency range Figure 11(b) describes cumulativedistribution function (CDF) of the decorrelation distances ofthree antenna setups Furthermore Table 3 lists their statisticvalues

For all three antenna setups the decorrelation distancescalculated by using threshold 119890minus1 are no doubt longer thanthat using threshold 05 Moreover Direc-Direc performs alonger decorrelation distance value (mean value around 25)than other cases (mean value around 23m and 2m resp)These indicate that the decorrelation distance will becomelonger when directional antenna is employed

53 Rician119870-Factor for the Received Signal The time-varyingfading characteristic of the signal is normally modeledby Rician 119870-factor when LOS component exists [37] theRician 119870-factor is defined as the ratio of the power of LOScomponent to the total power of NLOS components

The Rician119870-factor at 325 GHz is shown in Figure 12(a)In the near region the 119870-factor experiences a rapid changeIn Omni-Omni case this process can be described by thenarrow structure of the tunnel which causes the attenuationof reflected components For other two cases this rapidchangemainly stems from the radiation pattern of directionalantenna In the far region the 119870-factors of all three antennasetups decrease slowly Moreover Figure 12(b) gives the CDFof three antenna setups at whole frequencies in the range315 GHzsim335 GHz in both near region and far region It isobvious that in the near region the119870-factor varies intenselyBut in the far region the differences among three 119870-factorsare fairly small The statistical values are listed in Table 3where we find that the 119870-factor (in dB) is positive in near

region but is negative in the far region This characteris-tic indicates that the dominant power contribution of thereceived signal is changing from the LOS component to theNLOS components

54 Delay Characteristics in the Tunnel The root meansquare (RMS) delay spread is widely known as the singleparameter that can provide a quick overview of channel delaycharacteristics It is defined as the normalized second-ordermoment of the power delay profile (PDP)which characterizeschannel delay dispersion [37] In this study the RMS delayspread is calculated as follows

119878120591 (119905) = radicsum119873(119905)119896=1 119875119896 (119905) sdot 120591119896 (119905)2sum119873(119905)119896=1 119875119896 (119905) minus (sum119873(119905)119896=1 119875119896 sdot 120591119896 (119905)sum119873(119905)119896=1 119875119896 (119905) )2 (10)

where 119878120591(119905) is the RMS delay spread 119875119896(119905) is the power of 119896thray As all the rays are specific with certain delay power andangle information (10) is efficient for calculating the RMSdelay spread directly from rays of RT kernel results

Figure 14 depicts CDFs of the RMS delay spreads whichwere extracted in every snapshot at whole simulation fre-quencies Apparently when directional antennas are used inthe system the RMS delay spread will be decreased especiallyin near regionThese results are in line with the phenomenondisplayed in Figure 13 which specially compares PDPs ofthree antenna setups in near region It is clear that the direc-tional antenna can be a great spatial filter in near region thatattenuates multipath components which are not illuminatedby the main lobe of directional antenna Therefore whendirectional antennas are employed at both TX1 and RX1 the


Direc-DirecDirec-OmniOmni-Omni

minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000

Distance (m)Near region Far region

(a)

Direc-Direc in near regionDirec-Omni in near regionOmni-Omni in near regionDirec-Direc in far regionDirec-Omni in far regionOmni-Omni in far region

0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

minus5 0 5 10 15 20 25 30 35 40minus10Ricean K-factor (dB)

(b)

Figure 12 (a) Rician 119870-factor for different antenna setups at 325 GHz (b) CDF of Rician 119870-factor in different antenna setups and regionsat whole frequencies

Omni-Omni Direc-Omni Direc-Direc

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

Figure 13 Partial enlarged views to describe the changing of the PDPs with different antenna setups in near region at 325 GHz

minimum value of RMS delay spread can be obtained whichis around 212 ns in near region and 047 in far region

55 Doppler Characteristic in the Tunnel As discussed pre-viously the HSR channels in tunnel were simulated at speedof 360 kmh Therefore the Doppler effect on the channelsis widely of interest as it gives physical interpretation of thefrequency shift caused by movement [37] As can be seenin Figure 15 in near region the train movement obviouslyspreads the Doppler spectrum in Omni-Omni case but thespectrum shows a stable Doppler frequency shift with limitedfrequency spread in far region The striking variations ofthe Doppler spectrum in near region are partly due to thefact that the incident angles of received rays are sparse andchange rapidly whereas in the far region the incident anglesof received rays are very close and change slowly (referringto the tunnel narrow structure) which leads to a stable

Doppler frequency Meanwhile since the directional antennaattenuates lots of rays in near region Figure 16 illustrates thedetailed effects of directional antenna on Doppler spectrawhere some distinct differences among Doppler spectra areclearly shown To better evaluate the Doppler effects theCDFs of mean Doppler shifts and RMS Doppler spreadsof three antenna setups are studied at whole frequenciesin Figure 17 These two parameters are the moments ofthe Doppler spectra which can be calculated similar to themoments of the PDP [37] According to Figure 17 and thestatistic values listed in Table 3 the same conclusion as thatfrom Figure 16 can be obtained that directional antenna islike a spatial filter which causes largermeanDoppler shift andlower RMS Doppler spread in HSR tunnel

Generally the Doppler effects are studied with othersecond-order fading statistics that also closely related tochannel dynamic characteristics and the quality of received



0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

32 4 6510RMS delay (ns)

Figure 14 CDF of RMS delay in different antenna setups andregions at whole frequencies

20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)

Figure 15 Doppler spectra of Omni-Omni case at 325 GHz intunnel

signals They are level cross rate (LCR) and average fadeduration (AFD) [38] As detailed in Figure 18(a) curves thatstand for LCRs in far region exhibit lower LCRs compared tothat in near region and the result of Direc-Direc case in farregion shows the minimum LCR These phenomena are inline with observation results in Doppler spectra (cf Figures16 and 17) Additionally since the LCRs of far region arefar lower than LCRs of near region the Doppler frequenciesin far region are accordingly more stable with minimumDoppler spread andmean shift However although the AFDsexhibited in Figure 18(b) have no clear distinctions betweendifferent regions the AFD of Direc-Direc case still shows

the minimum value And AFDs of three antenna setupsin far region increase sharply when thresholds for AFDcalculation are around 0 dB which indicate weaker fast-fading characteristics of the channels in far region as well asweaker Doppler spread

56 Polarization Characteristics in the Tunnel The aboveanalyses of channel characteristics describe the vertical polar-ization only However the channel polarization characteris-tics in the tunnel are valuable for system designThe radiatedsignal from a vertically polarized antenna will experienceinteractions that result in horizontal polarized componentbefore it arrives at RX and vice versa [37] This processis so-called depolarization process that has been depictedby a 2 times 2 matrix which is formulated by 119875119896(119905) Thencross-polarization discriminations (XPD including XPD120579and XPD120593) are invited which indicates leakage from onepolarization to another by depolarization effect [39 40]

XPD120579 = 1003816100381610038161003816ℎVV100381610038161003816100381621003816100381610038161003816ℎVH10038161003816100381610038162 XPD120593 = 1003816100381610038161003816ℎHH

100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)

where |ℎVH|2 refers to the power transmitted in verticalpolarization and received in horizontal polarization |ℎVV|2|ℎHH|2 and |ℎHV|2 are defined accordingly

Figure 19(a) illustrates XPD120579 and XPD120593 simulated data inDirec-Direc case accompanying fitting lines The depolar-ization processes of the channel are found to be similar forboth vertical and horizontal polarized signal at 325 GHzThetubular structure of tunnel which results in the depolarizationprocess of the channel does not obviously prefer vertical orhorizontal polarized signal Moreover the values of XPDs innear region are much larger than values in far region As LOScomponent cannot be depolarized and gradually illuminatedby antenna main lobe in near region the numerator of(11) will increase with attached antenna gain to the LOScomponent This causes an apparent change as shown inFigure 19(a)

Figures 19(b) and 19(c) show theCDFs ofXPD120579 andXPD120593of three antenna setups at whole frequencies in differentregions We learn that using directional antenna can baffledepolarization process of the channel in near region Thestatistic values of the results of depolarization process in thetunnel can be found in Table 3

57 Relative Discussions Based on the radio channel char-acteristics and aforementioned analyses Table 3 lists allthe statistic values of the parameters Then the followingdiscussions can be made

(1) Path loss the path loss exponents for each antennasetup are all around 11 in far region The tubularand narrow structure of the tunnel probably causesthe waveguide propagations which compensate thereceived power in far region



minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

Figure 16 Partial enlarged views to describe the changing of the Doppler spectra with different antenna setups in near region at 325 GHz


110008500 9000 9500 10000 1050080007500Mean Doppler shift (Hz)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

500 1000 1500 2500 30000 2000RMS Doppler spread (Hz)

(b)

Figure 17 (a) CDFs of mean Doppler shift in different antenna setups and regions at whole frequencies (b) CDFs of RMS Doppler spreadin different antenna setups and regions at whole frequencies

(2) Shadow fading as the log-normal distribution can fitthe shadow fading well the shadow fading standarddeviations for three antenna setups are similar andaround 34 dB in the far region This number is closeto the measurement results in the tunnel at 57 GHz[14]

(3) Decorrelation distance the deep fading caused byshadow fadingwill decrease received power anddecaysystem capacity Also possibly it may result in ping-pong effect when system is in handover The decorre-lation distances calculated by threshold 119890minus1 are obvi-ously larger than that using threshold 05 The meanvalue forDirec-Direc case is around 25m (threshold

05) which is larger than other cases In conclusionas per directional antenna employed in system thedecorrelation distance increases 02ndash04m

(4) Rician 119870-factor the 119870-factors in two regions showgreat differences especially for Direc-Direc casewhere the difference between the mean values of tworegions is around 25 dB Even if the 119870-factors have agreat variation in different regions the 119870-factors ofdifferent antenna setups have little differences in farregion (around minus6 dB)

(5) RMS delay spread the RMS delay spread will bedeceased when directional antenna is employed Inthe near region the directional antenna performs as


10minus1

100

101

102Le

vel c

ross

rate

(1s

)

0minus10minus20 minus30Amplitude of the fast-fading (dB)


(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)

Figure 18 (a) Results of level cross rates of three antenna setups at 325 GHz (b) Results of average fade durations of three antenna setups at325 GHz

a spatial filter which reduces the RMS delay spreadby 1ndash15 ns In the far region RMS delay spreads ofdifferent antenna setups are less than 1 ns and inthe Direc-Direc case the number will be less than05 ns

(6) RMSDoppler spread there are less than 100Hz varia-tions in RMSDoppler spreads when one or two direc-tional antennas are employed in the system All RMSDoppler spreads in far region are around 041 kHzFurthermore the RMS Doppler spread may be lessinfluenced than RMS delay spread when directionalantennas are employed

(7) XPDs for Direc-Direc and Direc-Omni cases itis interesting to find that the mean values of XPD120579and XPD120593 are very close in near region and have avariation of 17 dB in far region For Omni-Omnicase the values of XPD120579 are 2 dB larger than XPD120593in both the two regions Since the vertical polarizedsignal shows an advantage in resisting depolarizationcompared to horizontal polarized signal in the tunnelit is better to deploy vertical polarized antennas in thearched tunnel

6 Conclusion

In this paper 30GHz band HSR radio channels in an archedtunnel with different antenna setups are investigated by aRT tool To enhance RT for more accurate simulations aspecial material measurement campaign is performed for

the cement which is the main construction material of thetunnel Then an advanced time-interpolation method isdescribed for extracting small-scale channel characteristicsThese two steps work as a promising and solid base forour extensive channel simulation at 315 GHzsim335 GHz Thechannel characteristics are found to be widely different indifferent regions The results show that directional antennasdeployed at both TX and RX can significantly improve thecoverage range of mobile communication system But thedirectional antennas will bring about larger decorrelationdistance in radio channel Therefore the interval betweenantennas at RX (if multiantennas are applied) should beseparated by more than 25m to overcome deep shadowfading Meanwhile the directional antennas play the roleof spatial filter which obviously attenuates the rays out ofthe antenna main lobe especially when RX is close to TXThese effects greatly decrease RMS delay spread in the nearregion Actually comparing to omnidirectional antenna thedirectional antenna mostly affects the channel character-istics in the near region but it rarely affects the channelcharacteristics in the far region Considering the length ofthe near region is around 50m in the study the HSR willget through the near region in a flash (05 second) In thismomentwhen the channel changeswith extreme rapidity it isdifficult to keep a reliable communication link So advancedhandover strategies should be considered before or afterHSR getting through the near region Finally the verticalpolarized antenna is suggested in this arched tunnel ratherthan horizontal polarized antenna



XP＄ data in Direc-Direc caseXP＄ fitting curve in Direc-Direc caseXP＄ data in Direc-Direc caseXP＄ fitting curve in Direc-Direc case

100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)

Cross-polarization discrimination ８０＄ (dB)50403020100minus10minus20

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)

Cross-polarization discrimination ８０＄ (dB)50403020100minus10minus20minus30

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)

Figure 19 (a) Results of XPD120579 andXPD120593 in theDirec-Direc case at 325 GHz Two fitted lines describe the variation tendencies for XPD120579 and119883119875119863120593 respectively (b) CDFs of XPD120579 in different antenna setups and regions at whole frequencies (c) CDFs of XPD120593 in different antennasetups and regions at whole frequencies

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work is supported by Institute for Information amp Com-munications Technology Promotion (IITP) grant funded bythe Korea government (MSIT) (no 2014-0-00282 Develop-ment of 5GMobile Communication Technologies for Hyper-Connected Smart Services)

References

[1] Z Pi and F Khan ldquoAn introduction to millimeter-wave mobilebroadband systemsrdquo IEEE Communications Magazine vol 49no 6 pp 101ndash107 2011

[2] B Ai K GuanM Rupp et al ldquoFuture railway services-orientedmobile communications networkrdquo IEEE CommunicationsMag-azine vol 53 no 10 pp 78ndash85 2015

[3] K Guan G Li T Kuerner et al ldquoOn millimeter wave and THzmobile radio channel for smart rail mobilityrdquo IEEE Transactionson Vehicular Technology vol 66 no 7 pp 5658ndash5674 2016

[4] B Ai X Cheng T Kurner et al ldquoChallenges toward wirelesscommunications for high-speed railwayrdquo IEEE Transactions on


Intelligent Transportation Systems vol 15 no 5 pp 2143ndash21582014

[5] Q Wang B Ai D W Matolak et al ldquoSpatial variation analysisfor measured indoor massive MIMO channelsrdquo IEEE Access2017

[6] T Rappaport S Sun R Mayzus et al ldquoMillimeter wave mobilecommunications for 5G cellular it will workrdquo IEEE Access vol1 pp 335ndash349 2013

[7] T S Rappaport Y Xing G R MacCartney A F Molisch EMellios and J Zhang ldquoOverview of millimeter wave communi-cations for fifth-generation (5G) wireless networks-with a focuson propagation modelsrdquo IEEE Transactions on Antennas andPropagation 2017

[8] J Zhang P Tang L Tian Z Hu T Wang and H Wangldquo6ndash100 GHz research progress and challenges from a channelperspective for fifth generation (5G) and future wireless com-municationrdquo Science China Information Sciences vol 60 no 8Article ID 080301 2017

[9] G Li K Guan B Ai et al ldquoOn the high-speed railwaycommunication at 30 GHz band Feasibility and channel char-acteristicsrdquo in Proceedings of the 11th International Symposiumon Antennas Propagation and EMTheory ISAPE 2016 pp 796ndash799 October 2016

[10] JWang H Zhu andN J Gomes ldquoDistributed antenna systemsfor mobile communications in high speed trainsrdquo IEEE Journalon SelectedAreas in Communications vol 30 no 4 pp 675ndash6832012

[11] X Chen S Guo and Q Wu ldquoLink-level analysis of a multiser-vice indoor distributed antenna system [wireless corner]rdquo IEEEAntennas and Propagation Magazine vol 59 no 3 pp 154ndash1622017

[12] W Fan T Jamsa J O Nielsen and G F Pedersen ldquoOn angularsampling methods for 3-D spatial channel modelsrdquo IEEEAntennas and Wireless Propagation Letters vol 14 pp 531ndash5342015

[13] M V S N Prasad R Singh S K Sarkar and A D SarmaldquoSome experimental and modeling results of widely varyingurban environments on train mobile radio communicationrdquoWireless Communications and Mobile Computing vol 6 no 1pp 105ndash112 2006

[14] K Guan B Ai Z Zhong et al ldquoMeasurements and Analysis ofLarge-Scale Fading Characteristics in Curved Subway Tunnelsat 920 MHz 2400 MHz and 5705 MHzrdquo IEEE Transactions onIntelligent Transportation Systems vol 16 no 5 pp 2393ndash24052015

[15] J Kim H-S Chung I G Kim H Lee and M S Lee ldquoA studyon millimeter-wave beamforming for high-speed train com-municationrdquo in Proceedings of the 6th International Conferenceon Information and Communication Technology ConvergenceICTC 2015 pp 1190ndash1193 October 2015

[16] K Guan Z Zhong B Ai and T Kurner ldquoDeterministic prop-agation modeling for the realistic high-speed railway environ-mentrdquo in Proceedings of the IEEE 77th Vehicular TechnologyConference VTC Spring 2013 June 2013

[17] T Abbas J Nuckelt T Zemen C F Mecklenbrauker and FTufvesson ldquoSimulation and measurement-based vehicle-to-vehicle channel characterization accuracy and constraint anal-ysisrdquo IEEE Transactions on Antennas and Propagation vol 63no 7 pp 3208ndash3218 2015

[18] Q Wang B Ai K Guan Y Li and Z Zhong ldquoRay-based anal-ysis of small-scale fading for indoor corridor scenarios at

15GHzrdquo in Proceedings of the Asia-Pacific Symposium on Elec-tromagnetic Compatibility (APEMC rsquo15) pp 181ndash184 IEEETaipei Taiwan May 2015

[19] S Priebe and T Kurner ldquoStochastic modeling of THz indoorradio channelsrdquo IEEE Transactions on Wireless Communica-tions vol 12 no 9 pp 4445ndash4455 2013

[20] J W McKown and R L Hamilton Jr ldquoRay tracing as a designtool for radio networksrdquo IEEE Network vol 5 no 6 pp 27ndash301991

[21] J Nuckelt M Schack and T Kurner ldquoDeterministic andstochastic channel models implemented in a physical layersimulator for Car-to-X communicationsrdquo Advances in RadioScience vol 9 pp 165ndash171 2011

[22] V Degli-Esposti F Fuschini E M Vitucci and G FalciaseccaldquoMeasurement and modelling of scattering from buildingsrdquoIEEE Transactions on Antennas and Propagation vol 55 no 1pp 143ndash153 2007

[23] W Fan I Carton P Kyosti and G Pedersen ldquoEmulatingray-tracing channels in multiprobe anechoic chamber setupsfor virtual drive testingrdquo IEEE Transactions on Antennas andPropagation vol 64 no 2 pp 730ndash739 2016

[24] R C Jones ldquoA new calculus for the treatment of optical systemsI description and discussion of the calculusrdquo Journal of theOptical Society of America vol 31 no 7 pp 488ndash493 1941

[25] ldquoRecommendation itu-r p1238-8-em propertyrdquo httpswwwituintrecR-REC-P1238en

[26] J Pascual-Garcia J Molina-Garcia-Pardo M Martinez-InglesJ Rodriguez and N Saurin-Serrano ldquoOn the importance ofdiffuse scattering model parameterization in indoor wirelesschannels at mm-wave frequenciesrdquo IEEE Access vol 4 pp 688ndash701 2016

[27] S Priebe M Kannicht M Jacob and T Kurner ldquoUltrabroadband indoor channel measurements and calibrated raytracing propagation modeling at THz frequenciesrdquo Journal ofCommunications and Networks vol 15 no 6 pp 547ndash558 2013

[28] D K Ghodgaonkar V V Varadan and V K Varadan ldquoAfree-space method for measurement of dielectric constants andloss tangents at microwave frequenciesrdquo IEEE Transactions onInstrumentation and Measurement vol 38 no 3 pp 789ndash7931989

[29] A Corana M Marchesi C Martini and S Ridella ldquoMini-mizing multimodal functions of continuous variables with theldquosimulated annealingrdquo algorithmrdquo Association for ComputingMachinery Transactions on Mathematical Software vol 13 no3 pp 262ndash280 1987

[30] K Guan Z D Zhong J I Alonso and C Briso-RodrıguezldquoMeasurement of distributed antenna systems at 24 GHz ina realistic subway tunnel environmentrdquo IEEE Transactions onVehicular Technology vol 61 no 2 pp 834ndash837 2012

[31] A F Molisch Wireless Communications vol 34 John Wiley ampSons 2012

[32] J Nuckelt M Schack and T Kurner ldquoGeometry-based pathinterpolation for rapid ray-optical modeling of vehicular chan-nelsrdquo in Proceedings of the 9th European Conference on Antennasand Propagation EuCAP 2015 pp 1ndash5 May 2015

[33] G Li B Ai K Guan et al ldquoChannel characterization formobile hotspot network in subway tunnels at 30 GHz bandrdquo inProceedings of the 83rd IEEE Vehicular Technology ConferenceVTC Spring 2016 May 2016

[34] G Li B Ai K Guan et al ldquoPath loss modeling and fadinganalysis for channels with various antenna setups in tunnels at


30 GHz bandrdquo in Proceedings of the 10th European Conferenceon Antennas and Propagation (EuCAP rsquo16) April 2016

[35] W C Y Lee ldquoEstimate of local average power of a mobile radiosignalrdquo IEEE Transactions on Vehicular Technology vol 34 no1 pp 22ndash27 1985

[36] R He B Ai Z Zhong A F Molisch R Chen and Y YangldquoAmeasurement-based stochasticmodel for high-speed railwaychannelsrdquo IEEE Transactions on Intelligent Transportation Sys-tems vol 16 no 3 pp 1120ndash1135 2015

[37] A F Molisch Wireless Communications John Wiley amp Sons2007

[38] T S RappaportWireless Communications Principles and Prac-tice vol 2 Prentice Hall PTR Upper Saddle River NJ 1996

[39] S Priebe ldquoTowards THz Communications Propagation Stud-ies Indoor Channel Modeling and Interference Investigationsrdquo2013

[40] A Panahandeh F Quitin J M Dricot F Horlin C Oestgesand P De Doncker ldquoMulti-Polarized Channel Statistics forOutdoor-to-Indoor and Indoor-to-Indoor Channelsrdquo in Pro-ceedings of the IEEE 71st Vehicular Technology Conference pp1ndash5 Taipei Taiwan May 2010

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of


International Journal of

RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal of

Volume 201

Submit your manuscripts athttpswwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



Since channel sounding measurements are treated as apriority when studying channels in the era of 3G and 4Gcommunication system the channel parameters and modelsare extracted mostly from massive measurement campaigns[13] But the measurement at mmWave band greatly chal-lenges the existent sounding systems especially for high-dynamic channel measurement The HSR mmWave channelmeasurement data is even rare [3] So we propose anapproach that enhances the capability of deterministic chan-nel model for obtaining mmWave HSR channel Then someimportant channel characteristics are studied and modeledfor guiding future HSR mmWave communication systemdesign

In this paper we simulate an important wireless chan-nel by an enhanced ray tracing simulator where the elec-tromagnetic (EM) and scattering parameters of dominantmaterials in the scenario aremeasured and fittedwith suitablemodels The simulation configuration practically follows thecommunication system demands in [15] Meanwhile a time-interpolation method is employed for promoting the simula-tion capability for high-dynamic channels Based on exten-sive simulations the channel characteristics in an archedtunnel are completely exposed in two defined regions More-over the channel feasibilities of using different antennas areinvestigated for determining the most suitable one

The remainder of this paper is organized as followsSection 2 describes principles of the proposed ray tracingtool with material measurement and time-interpolation InSection 3 simulation scenario and system setups will bepresented with more details Section 4 demonstrates thefeasibility of communication system with different antennasetups Section 4 also illustrates the necessity of partition ofregions in radio channel analysis Then characteristics ofradio channel are provided in Section 5 Finally Section 6gives the conclusion

2 Ray-Optical Based mmWave ChannelModeling for High-Dynamic Scenario

The wireless channel is simulated by utilizing ray-opticalbased deterministic propagation model that is ray tracing(RT) model As it differs from other traditional RT theenhanced RT used in this study has been calibrated invarious frequencies and scenarios [16ndash19] We make a spe-cial measurement for extracting electromagnetic (EM) andscattering parameters of the most influential materials in theHSR tunnel The extracted parameters will be implementedinto RT simulator Moreover for investigating the detailedchannel characteristics in such high-dynamic scenario awell-studied RT time-interpolation algorithm is employed toexpose the channel small-scale characteristics

The RT in this study is developed based on ray-opticalalgorithm [20] at Technische Universitat Braunschweig [21]It is three-dimension (3D) channel simulator which is per-formed in 3D digital map For an integrated channel mod-eling several radio propagation mechanisms are taken intoconsideration as line-of-sight (LOS) ray reflection rays andscattering rays The LOS ray is known as free-space wavepropagation between transmitter (TX) and receiver (RX) Its

power is calculated by free-space path loss model in RT TheLOS path if it exists will dominate the received power in theRXThen the reflectionwhich is defined as the incident angleand reflected angle of EMwave from a surface is the same Itspower is calculated by solving the well-known Fresnel equa-tions and the related reflection point in the 3D digital map isobtained by applying the image method [20] Still in the caseof reflection 119899th order multiple reflections will be practicallyconsidered for capturing dominant power contributions oftotal received power For the diffuse scattering the powercontribution of diffuse scattering is evaluated by adoptingthe ldquoeffective roughnessrdquo (ER) scattering models [22] theER models generally include Lambertian Model DirectiveModel and Backscattering Lobe Model for predicting dif-ferent types of scattering patterns [22] Owing to the specialstructure of the tunnels where no obstacles are situatedthe transmission through the materials and diffraction fromedges of obstacles are omitted in the simulation

Based on ray-optical principle and several propagationmechanism models the final output of the RT is a time-variant channel impulse response (CIR) ℎ(120591 119905) which is thesum of powers of all the determined rays [23]

ℎ (120591 119905) = 119873(119905)sum119896=1

119886119896 (119905) sdot 119890119895(2120587119891120591119896(119905)+120593119896(119905)) sdot 120575 (120591 minus 120591119896 (119905)) (1)

where 119886119896(119905) 120591119896(119905) and 120593119896(119905) denote amplitude delay andadditional phase shift of 119896th ray (totally 119873(119905) rays that havebeen found by RT kernel)

Then119860119896(119905) = 119886119896(119905)sdot119890119895(2120587119891120591119896(119905)+120593119896(119905)) is defined as a complexcoefficient formula (1) can be rewritten as

ℎ (120591 119905) = 119873(119905)sum119896=1

119860119896 (119905) sdot 120575 (120591 minus 120591119896 (119905)) (2)

Then (2) attached with the effects of wave polarizationand antenna gains 119860119896(119905) can be further expressed by

119860119896 (119905) = radic119866RX119896 sdot 997888rarr119890 (120593RX119896 120579RX119896)119867 sdot 119875119896 (119905) sdot radic119866TX119896

sdot 997888rarr119890 (120593TX119896 120579TX119896) sdot 119871119896 (119905) (3)

where 120593RXTX119896 and 120579RXTX119896 indicate the angle of arrival(AOA) or angle of departure (AOD) of the 119896th multi-path component Therefore the vectors 997888rarr119890 (120593RX119896 120579RX119896) and997888rarr119890 (120593RX119896 120579RX119896) denote the complex polarization vectors [24]of antennas at RX and TX respectively 119866RX119896 and 119866TX119896indicate the additional antenna gains of the 119896th multipathcomponent in RX and TX (sdot)119867 denotes the Hermitian trans-pose 119875119896(119905) depicts the complex channel polarization matrixwhich indicates the polarization shifting of the ray 119871119896(119905)comprises propagation loss and the phase shift accordingto the delay 120591119896(119905) of the 119896th multipath component Moreinformation can be found in [21]

For investigating mmWave channel it is still the chal-lenge that the mmWave channel sounding measurements arevery costly and time-consuming even for static scenariosFor measuring high-dynamic mmWave channel it is even


minus90∘

RX

RX

RX

RX

TX

90∘

0∘

Material

Measurement

）Ｈ＝Ｃ

３＝Ｎ

(a)

TXRX


(b)









）Ｈ＝Ｃ

908070605040302010003

04

05

06

07

08

09

1Re

flect

ion

coeffi

cien

t

(a)


0

2

4

6

8

Am

plitu

de o

f S-p

aram

eter

times10minus3


）Ｈ＝Ｃ = 70∘

(b)










y-axis

x-axis

TX1 65 m height

RX1 3 m height

85685134170

64

56

48

40

32

24

16

08

0

HＮＯＨＨＦ

(m)


(a)

RX1

TX 1

Link 1

rarr ＮＬＣＨ

(b)

RX1

RX2TX2

TX1

1 km interval

Link 1

Link 2

(c)













minus10123

minus2


2

20

15

10

5

0


(dBi)


Percentage ()

lt1

lt01

lt001

0 100 150 200 25050Distance (m)

LOS

1st

2nd

3rd

4th

5th

6th

7th

8th

Pow

er co

ntrib

utio

n

0

10

20

30

40

50

60
















0102030405060

Sign

al-n

oise

ratio

(dB)

50100 200 300 400 500 600 700 800 900 10000Distance (m)





119863 = 119867BS minus 119867Anttan (120579 + 1205722) (5)


D

RX1 RX2

TX1



H３

H３ minus HＨＮ

HＨＮ

= 8∘ (HPBW)





119871 (119889) = 119871 (1198890) + 10119899 lg ( 1198891198890) + 119883120590 (6)







Min 106 106 107Mean 110 110 111Max 113 113 113












50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060


Simulation


0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n


= minus17052

= 34443



12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)



0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding




minus1)




120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)






0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










minus90∘

RX

RX

RX

RX

TX

90∘

0∘

Material

Measurement

）Ｈ＝Ｃ

３＝Ｎ

(a)

TXRX


(b)









）Ｈ＝Ｃ

908070605040302010003

04

05

06

07

08

09

1Re

flect

ion

coeffi

cien

t

(a)


0

2

4

6

8

Am

plitu

de o

f S-p

aram

eter

times10minus3


）Ｈ＝Ｃ = 70∘

(b)










y-axis

x-axis

TX1 65 m height

RX1 3 m height

85685134170

64

56

48

40

32

24

16

08

0

HＮＯＨＨＦ

(m)


(a)

RX1

TX 1

Link 1

rarr ＮＬＣＨ

(b)

RX1

RX2TX2

TX1

1 km interval

Link 1

Link 2

(c)













minus10123

minus2


2

20

15

10

5

0


(dBi)


Percentage ()

lt1

lt01

lt001

0 100 150 200 25050Distance (m)

LOS

1st

2nd

3rd

4th

5th

6th

7th

8th

Pow

er co

ntrib

utio

n

0

10

20

30

40

50

60
















0102030405060

Sign

al-n

oise

ratio

(dB)

50100 200 300 400 500 600 700 800 900 10000Distance (m)





119863 = 119867BS minus 119867Anttan (120579 + 1205722) (5)


D

RX1 RX2

TX1



H３

H３ minus HＨＮ

HＨＮ

= 8∘ (HPBW)





119871 (119889) = 119871 (1198890) + 10119899 lg ( 1198891198890) + 119883120590 (6)







Min 106 106 107Mean 110 110 111Max 113 113 113












50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060


Simulation


0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n


= minus17052

= 34443



12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)



0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding




minus1)




120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)






0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











）Ｈ＝Ｃ

908070605040302010003

04

05

06

07

08

09

1Re

flect

ion

coeffi

cien

t

(a)


0

2

4

6

8

Am

plitu

de o

f S-p

aram

eter

times10minus3


）Ｈ＝Ｃ = 70∘

(b)










y-axis

x-axis

TX1 65 m height

RX1 3 m height

85685134170

64

56

48

40

32

24

16

08

0

HＮＯＨＨＦ

(m)


(a)

RX1

TX 1

Link 1

rarr ＮＬＣＨ

(b)

RX1

RX2TX2

TX1

1 km interval

Link 1

Link 2

(c)













minus10123

minus2


2

20

15

10

5

0


(dBi)


Percentage ()

lt1

lt01

lt001

0 100 150 200 25050Distance (m)

LOS

1st

2nd

3rd

4th

5th

6th

7th

8th

Pow

er co

ntrib

utio

n

0

10

20

30

40

50

60
















0102030405060

Sign

al-n

oise

ratio

(dB)

50100 200 300 400 500 600 700 800 900 10000Distance (m)





119863 = 119867BS minus 119867Anttan (120579 + 1205722) (5)


D

RX1 RX2

TX1



H３

H３ minus HＨＮ

HＨＮ

= 8∘ (HPBW)





119871 (119889) = 119871 (1198890) + 10119899 lg ( 1198891198890) + 119883120590 (6)







Min 106 106 107Mean 110 110 111Max 113 113 113












50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060


Simulation


0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n


= minus17052

= 34443



12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)



0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding




minus1)




120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)






0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










y-axis

x-axis

TX1 65 m height

RX1 3 m height

85685134170

64

56

48

40

32

24

16

08

0

HＮＯＨＨＦ

(m)


(a)

RX1

TX 1

Link 1

rarr ＮＬＣＨ

(b)

RX1

RX2TX2

TX1

1 km interval

Link 1

Link 2

(c)













minus10123

minus2


2

20

15

10

5

0


(dBi)


Percentage ()

lt1

lt01

lt001

0 100 150 200 25050Distance (m)

LOS

1st

2nd

3rd

4th

5th

6th

7th

8th

Pow

er co

ntrib

utio

n

0

10

20

30

40

50

60
















0102030405060

Sign

al-n

oise

ratio

(dB)

50100 200 300 400 500 600 700 800 900 10000Distance (m)





119863 = 119867BS minus 119867Anttan (120579 + 1205722) (5)


D

RX1 RX2

TX1



H３

H３ minus HＨＮ

HＨＮ

= 8∘ (HPBW)





119871 (119889) = 119871 (1198890) + 10119899 lg ( 1198891198890) + 119883120590 (6)







Min 106 106 107Mean 110 110 111Max 113 113 113












50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060


Simulation


0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n


= minus17052

= 34443



12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)



0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding




minus1)




120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)






0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











minus10123

minus2


2

20

15

10

5

0


(dBi)


Percentage ()

lt1

lt01

lt001

0 100 150 200 25050Distance (m)

LOS

1st

2nd

3rd

4th

5th

6th

7th

8th

Pow

er co

ntrib

utio

n

0

10

20

30

40

50

60
















0102030405060

Sign

al-n

oise

ratio

(dB)

50100 200 300 400 500 600 700 800 900 10000Distance (m)





119863 = 119867BS minus 119867Anttan (120579 + 1205722) (5)


D

RX1 RX2

TX1



H３

H３ minus HＨＮ

HＨＮ

= 8∘ (HPBW)





119871 (119889) = 119871 (1198890) + 10119899 lg ( 1198891198890) + 119883120590 (6)







Min 106 106 107Mean 110 110 111Max 113 113 113












50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060


Simulation


0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n


= minus17052

= 34443



12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)



0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding




minus1)




120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)






0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













0102030405060

Sign

al-n

oise

ratio

(dB)

50100 200 300 400 500 600 700 800 900 10000Distance (m)





119863 = 119867BS minus 119867Anttan (120579 + 1205722) (5)


D

RX1 RX2

TX1



H３

H３ minus HＨＮ

HＨＮ

= 8∘ (HPBW)





119871 (119889) = 119871 (1198890) + 10119899 lg ( 1198891198890) + 119883120590 (6)







Min 106 106 107Mean 110 110 111Max 113 113 113












50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060


Simulation


0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n


= minus17052

= 34443



12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)



0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding




minus1)




120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)






0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












Min 106 106 107Mean 110 110 111Max 113 113 113












50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060


Simulation


0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n


= minus17052

= 34443



12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)



0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding




minus1)




120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)






0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











50

60

70

80

90

100

Path

loss

(dB)

18 2 22 24 26 2816ＦＩＡ10(d) (m)

n = 1060


Simulation


0

002

004

006

008

01

012

014

Prob

abili

ty d

ensit

y fu

nctio

n


= minus17052

= 34443



12058812 = 119864 119878 (1198891) 119878 (1198892)120590 (1198891) 120590 (1198892) (7)



0minus02

0

02

04

06

08

1

12

Auto

corr

elat

ion

coeffi

cien

t of s

hado

w fa

ding




minus1)




120588exp (Δ119889) = 119890(minusΔ119889119889cor) (8)






0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











0

2

4

6

8

10

12Cu

mul

ativ

e dist

ribut

ion

func

tion


(a)

0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n








(b)













minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











minus10

0

10

20

30

40

Rici

anK

-fact

or (d

B)

50 200 400 600 800 10000


(a)


0010203040506070809

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)



300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020

300

250

200

150

100

50

0

Del

ay (n

s)

Time (s)

0

minus20

minus40

minus60

minus80

minus100

minus120

Nor

mal

ized

pow

er (d

B)

080604020








0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n



20

10

0

minus10

minus20

Dop

pler

freq

uenc

y (K

Hz)

2 4 6 80Time (s)

minus100

minus80

minus60

minus40

minus20

0

Nor

mal

ized

pow

er (d

B)






100381610038161003816100381621003816100381610038161003816ℎHV10038161003816100381610038162

(11)








minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)

minus12

minus10

minus8

minus6

minus4

minus2

0

2

Dop

pler

freq

uenc

y (K

Hz)

Time (s)080604020

minus80

minus70

minus60

minus50

minus40

minus30

minus20

minus10

0

Nor

mal

ized

pow

er (d

B)




0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)








10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










10minus1

100

101

102Le

vel c

ross

rate

(1s

)



(a)


10minus5

10minus4

10minus3

10minus2

10minus1

100

Aver

age f

ade d

urat

ion

(1s

)


(b)





6 Conclusion






100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












100 200 800 1000500 600 700300 9004000

Distance (m)

minus40

minus30

minus20

minus10

0

10

20

30

40

50

Cros

s-po

lariz

atio

n di

scrim

inat

ioＨ

at 3

25

GH

z (dB

)

(a)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(b)


0

01

02

03

04

05

06

07

08

09

1

Cum

ulat

ive d

istrib

utio

n fu

nctio

n


(c)




Acknowledgments


References














































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation


















































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









On the Feasibility of High Speed Railway mmWave Channels in Tunnel...

Documents

Transcript of On the Feasibility of High Speed Railway mmWave Channels in Tunnel...