Research Article  Open Access
Broadband Wireless Channel in Composite HighSpeed Railway Scenario: Measurements, Simulation, and Analysis
Abstract
The rapid development of highspeed railway (HSR) and trainground communications with high reliability, safety, and capacity promotes the evolution of railway dedicated mobile communication systems from Global System for Mobile CommunicationsRailway (GSMR) to Long Term EvolutionRailway (LTER). The main challenges for LTER network planning are the rapidly timevarying channel and high mobility, because HSR lines consist of a variety of complex terrains, especially the composite scenarios where tunnels, cuttings, and viaducts are connected together within a short distance. Existing researches mainly focus on the path loss and delay spread for the individual HSR scenarios. In this paper, the broadband measurements are performed using a channel sounder at 950 MHz and 2150 MHz in a typical HSR composite scenario. Based on the measurements, the pivotal characteristics are analyzed for path loss exponent, power delay profile, and tap delay line model. Then, the deterministic channel model in which the 3D raytracing algorithm is applied in the composite scenario is presented and validated by the measurement data. Based on the raytracing simulations, statistical analysis of channel characteristics in delay and Doppler domain is carried out for the HSR composite scenario. The research results can be useful for radio interface design and optimization of LTER system.
1. Introduction
Over the past few years, the world has witnessed the rapid development of the highspeed railways (HSRs) in China. The GSMR system, which is specially designed and standardized for communication between train and control centers, could provide voice group call, functional addressing, location dependent addressing, enhanced multilevel priority and preemption, train control safety data transmission, and dispatching nonsafety data transmission service. However, the GSMR still has some limitations to meet a diversity of data transmission including security monitoring and maintenance information in real network operation since it is only available to narrowband communications. Thus, the broadband wireless trainground communication on HSR plays an important role in train operation control, monitoring, and maintenance data transmission and makes a great number of applications possible [1, 2]. To satisfy the increasing demand for railway services, HSR faces the challenge that is to deploy broadband communication system called LTER [3–6]. The LTER has been identified as the technical direction of the next generation of railway mobile communication in HSR environments due to its technical advantages, such as increased performance, flexible quality of service (QoS) profile, supporting mission critical service, and highspeed mobility [7, 8]. Currently, a series of studies on HSR broadband communication system are being carried out, while radio propagation is the main research hotspot because the high performance of communication system relies on excellent wireless network coverage. The sustainable development of railway broadband mobile communication system needs to keep abreast of HSR wireless channel characteristics [9, 10]. Establishing reliable and credible HSR channel model is the precondition to realize the efficient, reliable, and safe data transmission under the conditions of high mobility and high data rate.
The researches on broadband and narrowband channels in HSR or subway environments have been conducted over the last few years. The smallscale fading behavior, which is defined as signal variations occurring within very short distances, has been discussed in large numbers of papers for broadband mobile communications. Several wireless channel propagation measurement activities have been performed in different kinds of HSR scenarios, such as the train station [11–13], the viaduct [14–17], the cutting [18, 19], the crossing bridge [20], the railway tunnel [21], the subway tunnel [22, 23], and the passageway tunnel [24]. The empirical path loss model and smallscale fading parameters have been obtained at 930 MHz. The propagation characteristics in a small passageway tunnel for nonlineofsight (NLoS) are studied in [24], and the model is given to predict the path loss, power delay profile (PDP), rootmeansquare (RMS) delay spread, and coherence bandwidth. The tap delay line (TDL) models and Doppler characterizations for HSR hilly scenario are put forward in [25]. In [26], the propagation mechanisms at 2.35 GHz have been investigated for the viaduct scenario, where the path loss, time delay spread, and Kfactor have been obtained to establish the statistical positionbased channel models. In [27], a cooperative antenna system is built on separated train carriages and a Doppler frequency offset estimation and joint channel estimation algorithm exploited multiantenna diversity gains are put forward. These studies mainly concentrate on channel parameters in an individual scenario. However, as a result of the highspeed mobility in the actual HSR environment, scenarios including the tunnel, cutting, and viaduct are continuous for wireless communication. Thus, the mobile station will experience many kinds of scenarios in a short period of time, and these scenarios are inseparable for the mobile station. Call drops and data transmission errors occur in such composite scenarios frequently. In summary, existing studies generally ignore the quick transition from one scenario to another. Thus, the channel model for a composite scenario is still inadequate and the nonstationary characteristic of timevarying and fastfading channels in HSR composite scenario still needs further investigation.
The deterministic modeling technique based on raytracing (RT) simulation is also an important method to model HSR channels [28, 29]. This method takes into account the direct path, reflections, and scattering and models the propagation channel in a specific scenario using the morphological and geographical information. However, in a complicated scenario, the RT method requires a large amount of computations. In general, for the sake of reducing the computational time in the complicated scenarios, only the first and secondorder reflections are taken into consideration. The output of the deterministic channel models for each link is the timevarying channel impulse response (CIR) which characterizes the frequencyselective channel. The 3dimensional (3D) scenario reconstruction is the basis of the RT modeling, which has been studied in the literature, for example, [30, 31].
The main contributions of this paper lie in the following aspects:(1)A broadband wireless channel measurement campaign is conducted in HSR composite scenario by using the channel sounding system at the frequencies of 950 MHz and 2150 MHz.(2)Based on the measurement data, we explicate the phenomenon of the variation trends of path loss, PDP, mean excess delay, and the RMS delay spread in 6 typical areas of the composite scenario. Then, the TDL based channel models are formed, which are helpful for the characterization of the HSR timevarying channel.(3)The HSR composite scenario is reconstructed and the 3D RT simulator is validated by the measurement data. Then, the smallscale fading, PDP, Doppler statistical properties, and the correlation coefficients between delay and Doppler domain are investigated. This is helpful for the development of HSR broadband communication system in the foreseeable future.
The rest of this paper is organized as follows. The measurement scenario, channel sounder configuration, and measurement campaign are provided in Section 2. Section 3 presents the measurement results with statistical analysis, and a novel TDL channel model for HSR composite scenarios is established. Section 4 provides the 3D scenario reconstruction of the composite HSR scenario; a 3D rayoptical based channel simulation and stochastic analysis are introduced as well. In Section 5, the deterministic modeling of the channel in HSR composite scenario and the statistical modeling from the measured data in the same scenario are compared and validated. Moreover, the smallscale fading, Doppler statistical properties, and correlation coefficients between delay and Doppler domain are analyzed in detail. Finally, the conclusions are drawn in Section 6.
2. HSR Channel Measurements
2.1. Measurement System Configuration
The measurement system consists of a channel sounder including transmitter (Tx) and receiver (Rx), oscilloscope, and antennas. The measurement specifications are described in detail below.
2.1.1. Channel Sounder
A channel sounder serves as a systematic device to measure the behaviors of wireless signals in the specific environment [32]. The test platform adopts the narrow pulse technology and can be used for the broadband channel sounding, with the parameters listed in Table 1. The Tx is composed of a broadband frequency synthesizer, pulse modulator, pulse generator, and broadband power amplifier. The synthesizer is controlled by the customized software on PC to program the transmission frequency. The transmitted signal is modulated by pulse modulator to get a narrow pulse with selectable pulse widths. The high quality pulse generator is built to synchronize with the clock signal. Eventually, a high power amplifier is used to amplify the modulated sounding signal. The receiver is composed of local oscillator, low noise amplifier (LNA), mixer, intermediate frequency filters, and logarithmic amplifier used as the demodulator. The received signal is sent to the LNA. Then, the amplified signal is mixed with a phaselocked loop (PLL) for superhet receiving to get a +/−150 MHz intermediate frequency, through an addition or subtraction to the transmitted signal frequency. Then, a 100 MHz bandpass filter is set to filter out unwanted parts. Finally, the logarithmic amplifier is used as an envelope detector, which connects to the oscilloscope to provide the magnitude of the received signal [33].

2.1.2. Oscilloscope
The output demodulated signal of the receiver is sent to the digital oscilloscope, which allows capturing digitized PDP in real time and saving data on the hard disk. In order to meet the requirements of high sampling rate, Infiniium MSO9104A from Keysight is used, which offers a bandwidth up to 1 GHz and channels with responsive deep memory to ensure superior viewing of signals under test.
2.1.3. Antennas
The channel sounder is equipped with the LCom HG72714P090 panel directional antenna whose polarization is vertical, with vertical and horizontal beamwidth, as shown in Figure 1(a) and Table 2. The Rx antenna is omnidirectional with 3 meters height plus the bracket. The Rx antenna is LCom HG72107U, which is also vertically polarized and has the gain of 4 dBi and 7 dBi in different bands, as shown in Figure 1(b) and Table 2.

(a) Tx antenna pattern
(b) Rx antenna pattern
2.2. Measurement Scenario
In the construction of railway in mountainous region, the cuttings are often used to help the HSR cross the large obstacles. A standard deep cutting with the slopes of both sides having the similar inclination can produce abundant reflection and scattering components. Viaducts with a height of 10 m to 30 m are also used in HSR to ensure flatness and straightness of rails. The height of the antenna on the roof of the train is increased due to viaduct, and this reduces the number of scatterers and forms a relatively “clear” lineofsight (LoS) propagation channel. Therefore, cutting and viaduct have a pronounced effect on channel propagation characteristics, which provides the motivation for investigating the composite scenario of cutting and viaduct in this paper.
The broadband measurements were performed in Xinzhou section of the “DatongXi’an” HSR in China, as shown in Figure 2. A tunnel with the height of up to 9 m, a cutting of 130 m length, a viaduct of 280 m length, and a cutting of 100 m length are linked end to end. In this case, it is more likely to receive multipath components with a higher height of the upper cutting than that of the receiving antenna. The numerous multipath components caused by the side walls result in serious fading. Moreover, a semiclosed space is generated by the structure of the cutting and the mountain, which may contain a large number of scattering and reflection components.
(a) Scenario from Google Earth
(b) Field measurement scenario
2.3. Measurement Campaign
This paper mainly analyzes and discusses the effects of multipath components in composite scenarios. In order to achieve this goal, a series of broadband measurements at 950 MHz and 2150 MHz bands which may be used for LTER in future were conducted. The channel sounder transmitter was installed on the platform at the top of the cutting, the Tx antenna was 35 m high from the rail surface and 20 m away from the tunnel portal. The pitch angle of transmitting antenna was about . The channel sounder receiver with antenna moved along the track, and measurement data were collected at 50 different locations.
The measurement environment can be classified as four regions in the composite scenario: locations 1 and 2 belong to region I, that is, the tunnel portal; locations 3–15 fall into region II which is cutting 1; locations 16–44 belong to region III, where a viaduct exists, and region IV, which is cutting 2, includes the rest of the locations. The sampling interval is 50 ns and approximately 20000 CIRs for each snapshot are collected. Each neighboring snapshot has a distance of 10 m.
3. Measurements Results and Analysis
The knowledge of wireless channel fading behavior is helpful for designing fading countermeasures. Thus, parameterizing the fading properties and developing channel model for HSR are necessary. This section will make a detailed analysis of the measurement results.
3.1. Path Loss Analysis
The path loss model assumes a linear dependency between the path loss in dB and the logarithm of distance [34], shown aswhere is the frequencydependent free space path loss at the reference distance , is the average path loss exponent (PLE) over distance, and is a zeromean Gaussian distributed random variable with standard deviation describing the random shadowing. We find that the m closein (CI) free space reference path loss model is a simple physically based oneparameter PLE model [35, 36], which is more stable across frequencies and environments than the traditional floatingintercept (FI) leastsquares regression equation line. The fixedintercept is given aswhere and is the carrier wavelength. The CI models for the two carrier frequencies of 950 MHz and 2150 MHz are shown in Figure 3 with the measured data. As can be seen from Figure 3, the PLE of the composite scenario is 1.6 and 1.7 at 950 MHz and 2150 MHz, respectively. Such small PLE values imply the waveguide effect in the propagation of this composite scenario. Apart from the direct path between the Tx and the Rx, numerous reflected components caused by the cutting slopes and mountains also enforce the propagation, and, therefore, their superposition decreases the PLE values. Similar observation at 930 MHz in HSR viaduct scenarios has been reported in [37], in which the PLEs are between 1.5 and 1.7.
3.2. Power Delay Profile (PDP)
The PDP is widely used for modeling multipath fading channel. The extent of multipath effects on radio channel can be clearly determined by the time delay spread. Channel parameters such as the first and second significant multipath components can be extracted from the PDP curve.
Since the HSR base station is always installed on one side of the track and the antenna is higher than the track by at least 25 m except for tunnel part, railway environment is a typical scenario with LoS propagation. Usually, the radio wave coming from LoS path offers the dominant component, and the radio waves coming from other directions are scattered or reflected by the steep walls or hills, causing the time delay spread.
From Figures 4 and 5, it is found that the PDP fluctuates in this composite scenario along the track at the frequency of 950 MHz and 2150 MHz. Generally, waves at these two frequencies have quite similar behaviors, with the distance changes [33]. In the first 150 m, the propagation at both 950 MHz and 2150 MHz is with quite clear delay spread. More significant delay spread is demonstrated when Rx is closer to the mountain tunnel, which is due to the strong reflection from the wall above the tunnel. Then, the delay spread has the tendency to decrease gradually when the Rx moves to the cutting, which is due to the open space. However, there are still rich multipath components. The delay spread decreases when the Rx is close to 150 m from the Tx. This point is exactly the entrance of viaduct in the open area. Thus, a number of the multipath components disappear, and the propagation is dominated by the direct ray.
The typical PDPs on different scenarios are demonstrated in Figures 6(a)–6(f), respectively. Upon the Rx being in region I, that is, the tunnel portal, due to the direct path to the Tx being blocked by the tunnel hall in the shadowed area, it is the NLoS scenario. Radio propagation at both frequencies shows longer and clearer time delays in the NLoS area. In particular, the delay spread at 950 MHz is longer and the severe multipath effect is reflected, compared with 2150 MHz. Upon the Rx being in the region II, there are obvious LoS component and multipath components caused by the cutting steep walls. Upon the Rx being in region III, which is the viaduct scenario, it is difficult to detect the multipath components due to the disappearance of reflector. The contribution of multipath components from distant mountain scattering can also be ignored. The delay spread greatly decreases. Upon the Rx being in region IV, which is the cutting 2 scenario, there is only the direct path because the cutting height is low and the excess delay of reflected signals is close to that of the dominant signal. The power of reflected signal originated by the remote cutting is also small, and, therefore, a clear path cannot be formed.
(a) Tunnel portal
(b) Cutting 1
(c) Viaduct near cutting 1
(d) Viaduct
(e) Viaduct near cutting 2
(f) Cutting 2
3.3. Analysis of Time Dispersion
As a general method for analyzing broadband channel measurements, the PDP can be quantified by the mean excess delay and RMS delay spread, which describe the time dispersion characteristics of the channel in a specific scenario. Delay of the multipath component is estimated relatively to the first detectable peak path [38, 39]. The mean excess delay is the first moment of the PDP and is defined aswhere is the power level of the th path and is the relative time delay of the th path. The commonly quantified RMS delay spread is the square root of the second central moment of the PDP and is presented in (4), which can be used to extract the number of channel taps [40].
In general, the mean excess delay and the RMS delay spread rely on the selection of noise threshold, which is used to distinguish the received multipath components power from the thermal noise power for raw PDPs. In this paper, the noise threshold is set to be 6 dB above the noise level of PDP, as given in [41, 42]. Thus, the peak that is greater than the threshold is considered as a resolvable path. Meanwhile, to reduce the implementation complexity of the channel model, the number of channel taps should be specified. Using RMS delay spread to determine the number of taps is given by [38]where is the RMS delay spread and is the pulse duration, which is 30 ns in this paper.
In Table 3, the average number of multipath components, the mean excess delay, and RMS delay spread are calculated. The RMS delay spreads at the two frequencies are illustrated in Figure 7. The values of RMS delay spreads for region III and region IV lie in the range of 0–100 ns. This is due to the presence of the dominant LoS component, with almost no more other distinguishable multipath components. For the region close to the cutting area, RMS delay spread becomes larger, whereas the RMS delay spread is mainly in the range of 100–300 ns in the cutting scenario, which reflects the severe fading effects. In this area, there are approximately 56 multipath components due to the fact that there are more reflection and scattering components from the cutting wall and remote hills. The RMS delay spread values of different frequencies are close to each other. Note that similar results about the delay statistics are shown in [43] for the approximatively 950 MHz and 2150 MHz cases.

3.4. TDL Channel Model
The TDL based channel model is helpful for analyzing the timevarying channel [26]. The TDL based models are developed according to the hypothesis; namely, the distinguishable channel taps are discrete and have different delays in HSR composite scenario. In this subsection, the paper parameterizes the tap number, the relative delay spread, and the average gain in different scenarios on the basis of a detailed analysis of the measurement data.
3.4.1. TDL Channel Model under Tunnel Portal
Table 4 puts forward the TDL channel model under tunnel portal scenario with Tx on the roof of the cutting near the tunnel. It can be concluded that the number of multipath components is large, and the tunnel portal is very consistent with the NLoS scenario. The 2.4 GHz TDL model in [25] is built based on the similar hilly scenario, where a tap corresponds to a group of multipath components with similar propagation delays and gains as we modeled for our 2150 MHz data set.

3.4.2. TDL Channel Model under Cutting 1
Table 5 deduces the TDL channel model under cutting 1. It can be inferred that there are also several multipath signals; however, the LoS path plays a more dominant role in this scenario and the RMS delay spread is smaller than that of tunnel portal.

3.4.3. TDL Channel Model under Viaduct
Table 6 gives the TDL channel model parameters in the two subsections, viaduct near cutting 1 and viaduct near cutting 2. We can observe that when the Rx is near the cutting 1 or cutting 2, a two or threeray model is applied. This is because the scattering and reflection of cutting 1 and cutting 2 contribute to the second and third paths. When the Rx is in the middle part of the viaduct, that is, far from the cutting 1 and cutting 2, there is only one LoS path which is dominant and almost no multipath component is detected.

3.4.4. TDL Channel Model under Cutting 2
As can be seen from Figure 6(f), only LoS path exists in cutting 2 area, because the Rx is far from the Tx, the distances of the direct path and the reflection path from cutting 2 are substantially equal, and the reflected signals from the remote mountain are quite weak.
4. Scenario Reconstruction and RT Simulation
4.1. Reconstruction of the Scenario
As a deterministic model, the propagation is greatly affected by the modeling of the scenario. As a result, precisely reconstructing the 3D propagation environment is the key factor for deterministic channel modeling. With the help of 3D modeling, the radio channel behavior of the composite scenario has been simulated by the deterministic propagation model. To begin with, the dimensions of the viaduct, cutting, and buildings are measured manually. Then, the latitude and longitude information of the test line is collected and the scenario around the test line is obtained by the satellite image of Google Earth. Finally, by gathering all the information, a complete 3D RT channel model library can be established.
In light of different radio propagation mechanisms, all the major objects in HSR scenario could be divided into the largescale and the smallscale structures. The largescale structures, including ballastless track, tunnel, cutting, and viaduct, are modeled as triangles and polygons with different electromagnetic parameters. The smallscale structures, including pylons and vegetation, are modeled as a conductive cylinder with a limited length. The scenario is composed of tunnel, mountain, cutting, viaduct, and ballastless track. It is essential to know the dielectric properties of materials for improving the accuracy of simulation. The reflection coefficients and scattering coefficients of all the materials above can be determined by the following approaches. First of all, the standardized parameters are defined in ITUR recommendation, like ITUR P.2040, ITUR P.1411, ITUR P.12387, and so on. These parameters can be set to the original inputs. Furthermore, for specific environments where materials are unavailable or insufficient in standards, measurements will be used to calibrate the coefficients. In this work, we apply the proposed approach in [44, 45], and the calibrated material parameters of cutting, viaduct, and others are implanted into the RT simulator after measurements. Thus, the RT simulator can be used for simulation to perform an indepth analysis of the channel characteristics in specific scenario. Reconstruction of the composite HSR scenario for the RT simulation is shown in Figure 8.
4.2. 3D RayOptical Based Channel Simulation
The 3D rayoptical based channel simulation is used to establish 3D deterministic channel model. Several propagation mechanisms are taken into account to model the multipath nature of HSR channel. Firstly, the direct path between the Tx and Rx is calculated based on the distance and path loss. Secondly, specular reflection and diffuse scattering are calculated, with the bouncing order of reflection up to 2 for reducing the computational complexity. Furthermore, the contribution of diffraction is neglected because there is no obstacle between the Tx and Rx. Thus, the information of each ray can be obtained, including the type of path, path loss, path delay, and 3D angular properties. The positions of Tx/Rx, the directions of the antennas, and a snapshot of the RT result are illustrated in Figure 9. In this figure, the lines of different colors represent different received power level.
5. Validation of RT Simulation and Further Analysis
5.1. Validation of RT Simulation
For the purpose of improving RT algorithm efficiency, the objects involved in the deterministic channel model are dominant for the HSR, such as the viaduct, tunnel, and cutting. Hence, the model has excluded the objects that are remote, unobservable, and lower than the track. In RT algorithm, is taken as the sampling interval, where refers to the wavelength. The other parameters used in the simulations are the same as those used in the actual measurements. Assuming that the train runs at 310 km/h, the time axis indicating the train location is illustrated in Figure 8.
Figure 10 shows the comparison of path losses between the measured and the simulated results. Due to the limitation of RT simulation, for example, the limited order of reflection, no diffraction mechanism modeling, and rough scene reconstruction, there is certain difference between RT simulated and measured path loss. However, the changing tendency is in agreement qualitatively at different frequency bands. Figure 11 shows the PDP based on the measurement data and the PDP predicted by the deterministic channel model in one snapshot for 950 MHz and 2150 MHz. The prediction and measurement results are well consistent for the normalized power and multipath delay, except a few points (delay over 400 ns) which do not match so well due to simplification of the remote scattering object. Indeed, the capability of capturing the dense multipath component (DMC) is the weakness of the RT simulation methods, which has been pointed out by numerical publication. In the scenario of this paper, the tunnel and cutting are typical closed or semiclosed environments, and, therefore, usually strong multipath components with extremely long delay are not expected. This is well reflected in the measurement data as well. Thus, for the target scenario, the RT simulator is applicative for further analysis.
(a) 950 MHz
(b) 2150 MHz
5.2. SmallScale Fading Analysis
Based on the above 3D scenario, smallscale fading characteristics are investigated through the channel simulation in this subsection. The signal bandwidth, power, antenna parameters, and location are the same as field measurement. Figure 12 illustrates the smallscale fading depth and maximum fading value as the function of time at 950 MHz and 2150 MHz, respectively. Different time points reflect the different geographical locations in the composite scenario. The significantly different characteristics are found for smallscale fading in different locations which belong to different scenarios, given in Table 7. It can be observed that the fading depth in zone 2 is greater than in zone 3 for both 950 MHz and 2150 MHz, whereas the two zones are both viaduct region. The reason is that zone 2 is even closer to the cutting. In zone 4, the maximum fading depth for 2150 MHz is about −59 dB, which is deeper than other zones, while the same phenomenon is not found so serious for 950 MHz. Similar results about the smallscale fading are obtained for the 900 MHz data set in [43], but the maximum fading depth in [43] is lower than that in this paper. This result indicates that the composite scenario leads to more severe smallscale fading.

(a) 950 MHz
(b) 2150 MHz
5.3. Stochastic Analysis
Wireless communication system must be developed for various propagation conditions, which motivates stochastic characteristic analysis that treats a linear timevarying channel as a random quantity. Since the channel has no great change in the stationarity interval which is assumed to be 10 ms [46], the hypothesis of widesense stationary uncorrelated scattering (WSSUS) is rational. According to WSSUS hypothesis in the stationary interval, the stochastic analysis is conducted subsequently.
5.4. PDP Properties Analysis
Based on the RT simulation, the PDP is studied in this subsection. Figure 13 gives PDP versus scenario time at two frequencies. In general, the changing trends and fluctuation from the deterministic model are similar to the measurements. The PDPs of 950 MHz and 2150 MHz are similar. When the train is near the transmitter position where the scenario is the tunnel portal and high cutting, more diffuse multipath components exist due to the rich scattering rays from tunnel external wall and the higher cutting. Dense diffuse multipaths are demonstrated; see Figure 13. The delay is mainly distributed in 0–200 ns, which is below the measured delay and the power of multipath is higher than the measurement data because of the simplification of the scenario.
(a) 950 MHz
(b) 2150 MHz
5.5. Doppler Statistical Properties Analysis
Based on the RT simulation, Doppler statistical properties are analyzed in this subsection. The mean Doppler shift is defined aswhere is the Doppler power spectral density, is Doppler shift, and is frequency. The RMS Doppler spread is defined as the square root of the second central moment of the Doppler power spectral density and is presented in (7), which can be used to characterize the time selectivity of the fading [41].The Doppler power spectrum, mean Doppler shift, and RMS Doppler spread of the two frequencies are given in Figures 14 and 15. When the train passes through the transmitter, the mean Doppler shift drifts drastically from 200 Hz to −600 Hz for 2150 MHz and from 100 Hz to −260 Hz for 950 MHz. When the RMS Doppler spread is relatively large, the smallscale fading depth becomes larger, indicating that the dense scattering components from the cutting lead to severe Doppler spread. Moreover, if the delay spread is larger, the fading would be deeper. For 950 MHz, the Doppler spread is not so large as 2150 MHz in simulations, so the serious fading does not occur. Therefore, the mobile station and base station should be designed to estimate the Doppler shift in real time and dynamically adjust the system parameters according to the Doppler shift.
(a) 950 MHz
(b) 2150 MHz
(a) 950 MHz
(b) 2150 MHz
5.6. The Correlation Coefficients between Delay and Doppler Domain
Based on the RT simulation, the crosscorrelation coefficient between delay and Doppler domain is estimated in this subsection, which is defined as [47]where is delay spread, is the Doppler spread, and and are the mean value of the data sets and with length , respectively. The results of the extracted crosscorrelation coefficients are given in Table 8. There is a high correlation between the delay spread and the Doppler spread at the cutting area for different frequency bands. Meanwhile, the delay and Doppler spreads exhibit a low correlation when the train moves on the viaduct as sparse scatterers are on both track sides.

6. Conclusion
In this paper, the broadband wireless channel measurements using a customized channel sounder in a real HSR composite scenario at 950 MHz and 2150 MHz are reported and analyzed. The path loss exponent and parameters of time dispersion containing the number of paths, the mean excess delay, and the RMS delay spread are estimated. The TDL channel model is proposed for the HSR composite scenarios containing tunnel portal, cutting, and viaduct based on the actual measurement data. The analysis reveals some important phenomena for the composite scenario: the difference of RMS delay at two frequencies is very small while the propagation loss is more serious; the cutting and tunnel portal will bring greater RMS delay; the multipath in viaduct is more influenced by the two cuttings that are connected to the viaduct, because the interference from the reflection and scattering of the cuttings produces not only deep fading, but also Doppler dispersion and delay spread. The deterministic channel model of the composite scenario, which is reconstructed using the 3D RT method, is put forward in this paper. The channel model is compared and validated with measurement data. The transition regions of different scenarios could be clearly identified by the normalized smallscale fading and Doppler characteristics analysis. Our future work will continue to analyze cluster birth and death behavior of the multipath components in this scenario. As radio channels have profound impacts on the field strength coverage, reliability, and quality of service of HSR mobile communication system, the results are useful for providing guidance on the broadband HSR communication system design, network planning, and optimization.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Acknowledgments
The authors express their thanks to the support in part from the National Key R&D Program of China under Grant no. 2016YFB120010204, the National Natural Science Foundation of China under Grants 61501021, U1334202, 61501020, and 61771037, the Project of China Railway Corporation under Grants 2016X009E, 2016X003L, and 2016X003O, the Fundamental Research Funds for the Central Universities under Grant 2016JBM076, State Key Lab of Rail Traffic Control and Safety Project under Grant RCS2016ZJ005, and Beijing Natural Science Foundation under Grant L161009.
References
 L. Lei, J. Lu, Y. Jiang et al., “Stochastic delay analysis for train control services in nextgeneration highspeed railway communications system,” IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 1, pp. 48–64, 2016. View at: Publisher Site  Google Scholar
 C. BrisoRodríguez, C. F. López, J. R. O. Fernández et al., “Broadband access in complex environments: LTE on railway,” IEICE Transactions on Communications, vol. E97B, no. 8, pp. 1514–1527, 2014. View at: Publisher Site  Google Scholar
 Y.S. Song, J. Kim, S. W. Choi, and Y.K. Kim, “Long term evolution for wireless railway communications:testbed deployment and performance evaluation,” IEEE Communications Magazine, vol. 54, no. 2, pp. 138–145, 2016. View at: Publisher Site  Google Scholar
 J. CalleSanchez, M. MolinaGarcia, and J. I. Alonso, “Top challenges of LTE to become the next generation railway communication system,” in Proceedings of the COMPRAIL 2012. 13th International Conference on Design and Operation in Railway Engineering, pp. 85–96, Wessex Institue of Technology, New Forest, UK, September 2012. View at: Publisher Site  Google Scholar
 L. Brunel, H. Bonneville, and A. Charaf, “Throughput performance of 3GPP LTE system in railway environment,” in Communication Technologies for Vehicles, vol. 10222 of Lecture Notes in Computer Science, pp. 60–71, Springer International Publishing, Cham, Switzerland, 2017. View at: Publisher Site  Google Scholar
 R. He, B. Ai, G. Wang et al., “HighSpeed railway communications: from GSMR to LTER,” IEEE Vehicular Technology Magazine, vol. 11, no. 3, pp. 49–58, 2016. View at: Publisher Site  Google Scholar
 J. CalleSánchez, M. MolinaGarcía, J. I. Alonso, and A. FernándezDurán, “Long term evolution in high speed railway environments: feasibility and challenges,” Bell Labs Technical Journal, vol. 18, no. 2, pp. 237–253, 2013. View at: Publisher Site  Google Scholar
 H. Bonneville, L. Brunel, and D. Mottier, “Standardisation roadmap for next train radio telecommunication systems,” Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Preface, vol. 9669, pp. 51–61, 2016. View at: Publisher Site  Google Scholar
 3rd Generation Partnership Project, “Technical specification group radio access network; evolved universal terrestrial radio access (EUTRA); base station (BS) radio transmission and reception,” Tech. Rep., 3GPP, Sophia Antipolis, France, 2016, TS 36.104 V14.0.0. View at: Google Scholar
 M. Zhao, M. Wu, Y. Sun et al., “Analysis and modeling of the LTE broadband channel for trainground communications on highspeed railway,” in Proceedings of the 2013 IEEE 78th Vehicular Technology Conference, VTC Fall 2013, USA, September 2013. View at: Publisher Site  Google Scholar
 K. Guan, Z. Zhong, B. Ai, and T. Kurner, “Empirical models for extra propagation loss of train stations on highspeed railway,” IEEE Transactions on Antennas and Propagation, vol. 62, no. 3, pp. 1395–1408, 2014. View at: Publisher Site  Google Scholar
 T. Zhou, C. Tao, S. Salous, L. Liu, and Z. Tan, “Channel characterization in highspeed railway station environments at 1.89 GHz,” Radio Science, vol. 50, no. 11, pp. 1176–1186, 2015. View at: Publisher Site  Google Scholar
 K. Guan, Z. Zhong, B. Ai, and T. Kurner, “Propagation measurements and analysis for train stations of highspeed railway at 930 MHz,” IEEE Transactions on Vehicular Technology, vol. 63, no. 8, pp. 3499–3516, 2014. View at: Publisher Site  Google Scholar
 T. Zhou, C. Tao, L. Liu, and Z. Tan, “A semiempirical MIMO channel model in obstructed viaduct scenarios on highspeed railway,” International Journal of Antennas and Propagation, vol. 2014, Article ID 287159, 10 pages, 2014. View at: Publisher Site  Google Scholar
 R. He, Z. Zhong, B. Ai, G. Wang, J. Ding, and A. F. Molisch, “Measurements and analysis of propagation channels in highspeed railway viaducts,” IEEE Transactions on Wireless Communications, vol. 12, no. 2, pp. 794–805, 2013. View at: Publisher Site  Google Scholar
 R. Zhao, M. Wu, X. Xiang, and J. Yang, “Measurement and modeling of the LTE trainground channel for highspeed railway in viaduct scenario,” in Proceedings of the 80th IEEE Vehicular Technology Conference, VTC 2014Fall, Canada, September 2014. View at: Publisher Site  Google Scholar
 R. He, Z. Zhong, B. Ai, and J. Ding, “An empirical path loss model and fading analysis for highspeed railway viaduct scenarios,” IEEE Antennas and Wireless Propagation Letters, vol. 10, pp. 808–812, 2011. View at: Publisher Site  Google Scholar
 R. He, Z. Zhong, B. Ai, J. Ding, Y. Yang, and A. F. Molisch, “Shortterm fading behavior in highspeed railway cutting scenario: Measurements, analysis, and statistical models,” IEEE Transactions on Antennas and Propagation, vol. 61, no. 4, pp. 2209–2222, 2013. View at: Publisher Site  Google Scholar
 J.H. Qiu, C. Tao, L. Liu, and Z.H. Tan, “Research on measurement and modeling of wireless channel multipath propagation properties for Ushape cutting,” Tiedao Xuebao/Journal of the China Railway Society, vol. 36, no. 1, pp. 40–48, 2014. View at: Publisher Site  Google Scholar
 K. Guan, Z. Zhong, B. Ai, and T. Kürner, “Propagation measurements and modeling of crossing bridges on highspeed railway at 930 MHz,” IEEE Transactions on Vehicular Technology, vol. 63, no. 2, pp. 502–517, 2014. View at: Publisher Site  Google Scholar
 C. BrisoRodriguez, J. M. Cruz, and J. I. Alonso, “Measurements and modeling of distributed antenna systems in railway tunnels,” IEEE Transactions on Vehicular Technology, vol. 56, no. 5, pp. 2870–2879, 2007. View at: Publisher Site  Google Scholar
 J. Alonso, S. Capdevila, B. Izquierdo, and J. Romeu, “Propagation measurements and simulations in tunnel environment at 5.8GHz,” in Proceedings of the 2008 IEEE International Symposium on Antennas and Propagation and USNC/URSI National Radio Science Meeting, APSURSI, USA, July 2008. View at: Publisher Site  Google Scholar
 K. Guan, Z. D. Zhong, J. I. Alonso, and C. BrisoRodríguez, “Measurement of distributed antenna systems at 2.4 GHz in a realistic subway tunnel environment,” IEEE Transactions on Vehicular Technology, vol. 61, no. 2, pp. 834–837, 2012. View at: Publisher Site  Google Scholar
 J.M. MolinaGarciaPardo, J.V. Rodríguez, and L. JuanLlácer, “Wideband measurements and characterization at 2.1 GHz while entering in a small tunnel,” IEEE Transactions on Vehicular Technology, vol. 53, no. 6, pp. 1794–1799, 2004. View at: Publisher Site  Google Scholar
 Y. Zhang, Z. He, W. Zhang, L. Xiao, and S. Zhou, “Measurementbased delay and doppler characterizations for highspeed railway hilly scenario,” International Journal of Antennas & Propagation, vol. 2014, no. 4, pp. 1–8, 2014. View at: Google Scholar
 L. Liu, C. Tao, J. Qiu et al., “Positionbased modeling for wireless channel on highspeed railway under a viaduct at 2.35 GHz,” IEEE Journal on Selected Areas in Communications, vol. 30, no. 4, pp. 834–845, 2012. View at: Publisher Site  Google Scholar
 Y. Yang and P. Fan, “Doppler frequency offset estimation and diversity reception scheme of highspeed railway with multiple antennas on separated carriage,” Journal of Modern Transportation, vol. 20, no. 4, pp. 227–233, 2012. View at: Publisher Site  Google Scholar
 K. Guan, Z. Zhong, B. Ai, and T. Kürner, “Deterministic propagation modeling for the realistic highspeed railway environment,” in Proceedings of the 2013 IEEE 77th Vehicular Technology Conference, VTC Spring 2013, deu, June 2013. View at: Publisher Site  Google Scholar
 J. Yang, B. Ai, K. Guan et al., “Deterministic modeling and stochastic analysis for channel in composite highspeed railway scenario,” in Proceedings of the 83rd IEEE Vehicular Technology Conference, VTC Spring 2016, China, May 2016. View at: Publisher Site  Google Scholar
 X. Yin, Y. Ji, and H. Yan, “Measurementbased characterization of 15 GHz propagation channels in a laboratory environment,” IEEE Access, vol. 5, pp. 1428–1438, 2017. View at: Publisher Site  Google Scholar
 D. He, G. Liang, J. Portilla, and T. Riesgo, “A novel method for radio propagation simulation based on automatic 3D environment reconstruction,” Radioengineering, vol. 21, no. 4, pp. 985–992, 2012. View at: Google Scholar
 J. Nuckelt, T. Abbas, F. Tufvesson, C. Mecklenbräuker, L. Bernadó, and T. Kürner, “Comparison of ray tracing and channelsounder measurements for vehicular communications,” in Proceedings of the 2013 IEEE 77th Vehicular Technology Conference, VTC Spring 2013, Germany, June 2013. View at: Publisher Site  Google Scholar
 L. Zhang, Channel measurement and modeling in complex environments [Ph.D. thesis], Technical University of Madrid, Madrid, Spain, 2016.
 S. Piersanti, L. A. Annoni, and D. Cassioli, “Millimeter waves channel measurements and path loss models,” in Proceedings of the 2012 IEEE International Conference on Communications, ICC 2012, pp. 4552–4556, Canada, June 2012. View at: Publisher Site  Google Scholar
 S. Sun, T. S. Rappaport, T. A. Thomas et al., “Investigation of Prediction Accuracy, Sensitivity, and Parameter Stability of LargeScale Propagation Path Loss Models for 5G Wireless Communications,” IEEE Transactions on Vehicular Technology, vol. 65, no. 5, pp. 2843–2860, 2016. View at: Publisher Site  Google Scholar
 M. K. Samimi and T. S. Rappaport, “3D millimeterwave statistical channel model for 5G wireless system design,” IEEE Transactions on Microwave Theory and Techniques, vol. 64, no. 7, pp. 2207–2225, 2016. View at: Publisher Site  Google Scholar
 P. Liu, A. Bo, D. W. Matolak, J.W. Ding, R. Sun, and Y. Li, “Waveguide effects in urban viaduct propagation channels for highspeed railway at 930 MHz,” in Proceedings of the 8th European Conference on Antennas and Propagation, EuCAP 2014, pp. 3479–3483, Netherlands, April 2014. View at: Publisher Site  Google Scholar
 I. Sen and D. W. Matolak, “Vehiclevehicle channel models for the 5GHz band,” IEEE Transactions on Intelligent Transportation Systems, vol. 9, no. 2, pp. 235–245, 2008. View at: Publisher Site  Google Scholar
 R. He, A. F. Molisch, F. Tufvesson, Z. Zhong, B. Ai, and T. Zhang, “Vehicletovehicle propagation models with large vehicle obstructions,” IEEE Transactions on Intelligent Transportation Systems, vol. 15, no. 5, pp. 2237–2248, 2014. View at: Publisher Site  Google Scholar
 L. Zhang, C. Briso, J. R. O. Fernandez et al., “Delay spread and electromagnetic reverberation in subway tunnels and stations,” IEEE Antennas and Wireless Propagation Letters, vol. 15, pp. 585–588, 2016. View at: Publisher Site  Google Scholar
 A. F. Molisch and M. Steinbauer, “Condensed parameters for characterizing wideband mobile radio channels,” International Journal of Wireless Information Networks, vol. 6, no. 3, pp. 133–154, 1999. View at: Publisher Site  Google Scholar
 M. S. Varela and M. G. Sánchez, “RMS delay and coherence bandwidth measurements in indoor radio channels in the UHF band,” IEEE Transactions on Vehicular Technology, vol. 50, no. 2, pp. 515–525, 2001. View at: Publisher Site  Google Scholar
 B. Chen, Z. Zhong, B. Ai, K. Guan, R. He, and D. G. Michelson, “Channel characteristics in highspeed railway: a survey of channel propagation properties,” IEEE Vehicular Technology Magazine, vol. 10, no. 2, pp. 67–78, 2015. View at: Publisher Site  Google Scholar
 T. Abbas, J. Nuckelt, T. Zemen, C. F. Mecklenbräuker, and F. Tufvesson, “Simulation and measurementbased vehicletovehicle channel characterization: accuracy and constraint analysis,” IEEE Transactions on Antennas and Propagation, vol. 63, no. 7, pp. 3208–3218, 2015. View at: Publisher Site  Google Scholar  MathSciNet
 D. He, K. Guan, A. Fricke et al., “Stochastic channel modeling for kiosk applications in the terahertz band,” IEEE Transactions on Terahertz Science & Technology, vol. PP, no. 99, pp. 1–12, 2017. View at: Google Scholar
 A. Ghazal, C.X. Wang, H. Haas et al., “A nonstationary geometrybased stochastic model for MIMO highspeed train channels,” in Proceedings of the 2012 12th International Conference on ITS Telecommunications, ITST 2012, pp. 7–11, Taiwan, November 2012. View at: Publisher Site  Google Scholar
 M. Zhu, G. Eriksson, and F. Tufvesson, “The COST 2100 channel model: parameterization and validation based on outdoor MIMO measurements at 300 MHz,” IEEE Transactions on Wireless Communications, vol. 12, no. 2, pp. 888–897, 2013. View at: Publisher Site  Google Scholar
Copyright
Copyright © 2017 Jianwen Ding et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.