Experiment and theoretical analysis of electromagnetic interference on communication systems in multi-rails environment

Purpose – Electromagnetic interference (EMI) on communication systems of unban rail transit can hardly be clari ﬁ ed because of complicated factors around railways. This paper aims to target this issue and extend experimental andtheoreticalanalysis. Design/methodology/approach – This paper take the Nanjing Dashengguan Bridge as an example, because it carries the most tracks in the world and bears three kinds of trains running through, providing a perfect complex environment. First, it investigates the three communication systems, terrestrial trunked radio, communications-based train control (CBTC) and passenger information system (PIS) that Nanjing Metro uses, and select appropriate devices accordingly. Second, it establishes a system level platform and conduct three tests to analyze their respective operating principles and performance difference under common electromagnetic environments. Third, it adopts theoretical formula to verify test results. Findings – The experiment results and theoretical analysis mutually corroborate each other and present practical recommendations: an 8 m or more distance between two tracks will ensure no obvious EMI created by a passing train on communication systems; two certain communication systems should not share the same frequency band; interference level is more related to ﬁ eld strength than weathers and building materials; and CBTC DSSS waveguidemode as well as PISLTEmode are preferred. Originality/value – This research also providesa practical methodof investigatingEMIfor other complex situations.


Introduction
China's high-speed rails have expanded at a world-leading pace for more than a decade, along with the progress of metros and other kinds of passenger rails. Chances are high that different kinds of transit trains are running on adjacent tracks, and create complex electromagnetic environments that may interfere with their respective communication systems (Flyvbjerg et al., 2005;Flyvbjerg et al., 2007).
A typical example is Nanjing Dashengguan Yangtze River Bridge, which carries the most railway lines in the world. It bears six tracks: two for the Beijing-Shanghai High-Speed Railway, two for the Shanghai-Wuhan-Chengdu Railway and two for the Ninghe Intercity Transit (part of the Nanjing Metro). The electromagnetic influences on urban rail transit communication systems when three kinds of trains converge, however, have yet to be clarified, and particular factors that function in system performance need to be identified (Heddebaut et al., 2017).
Therefore, this research intends to establish an experimental method that can effectively acquire operating characteristics of main communication systems under various external environments, so that direct influence factors such as line distance and weather condition will be explicit. We can monitor performance degradation in certain situations. Then, whether mode adjustment will improve work status and whether cochannel interference is severe, can be analyzed in detail.

Prior research
Many researchers have noticed electromagnetic interferences (EMI) around railways and explored different ways to detect them. This paragraph reviews relevant experiments that make an assessment of communication quality to understand widely used experimental facilities as well as design principles. Jun et al. (2016) invested EMI of 350 km h À1 China Standard Electric Multiple Units. They used a quasi-peak detector in receiver and spectrum analyzer to find out whether interference values of nine commonly occupied frequency bands were within the specified range under static and dynamic conditions. The group also found that the traction system was the main EMI source. Xiwei and Feng (2014) targeted Wuhan-Guangzhou high-speed railway and proposed a more practical resolution bandwidth to obtain the railways' EMI from 30 MHz to 1 GHz. Both concerned about standard optimization as well as external interference recognition. Xing Kui et al. (2015) introduced a rapid interference response test for receiver that used automatic dual-frequency test system. Pous et al. (2018) further proposed a full-time EMI measurement system, which could increase measuring time and reduce environmental noise to obtain the most accurate measuring results. Although these attempts all reached satisfying results, their experimental methods and instruments were simple and results were far from enough. In addition to external interference monitor, researchers put effort into finding out the primary factor among various mobile communication systems. Ahmad et al. (2017), for example, designed a cooperative communication scheme to analyze coexistence of passenger information system (PIS)-LTE and LTE-R networks, where indicators such as receiving interference and outage probability were selected to evaluate performance under particular scenarios. They verified that cochannel interference was a common occurrence for highspeed railways, because the communication systems occupy relatively stable frequency bands. Although involved experimental methods can be used for reference, specific equipment installation is extremely difficult to come true, especially for urban rail transits, which lack prior research.
Actually, many researchers have also examined electromagnetic influences on communication systems in theoretical ways. It is convenient to think that communication 1.2 Background on the Nanjing Dashengguan Yangtze River Bridge As typical landmark in railway construction, Dashengguan Yangtze River Bridge sets a successful record for long-span bridges with multiple railway tracks. To improve integrity and smoothness of the bridge structure, the composite slab-truss structure is of the priority. It is also the first attempt that Q420qE structural steel was applied in the bridge. However, whether the material has obvious influence on electromagnetic components of communication systems needs to be further studied.
As shown in Figure 1, the up-and downlink of Ninghe Intercity Transit are, respectively, 8.2 and 8.7 m far from Beijing-Shanghai High-Speed Railway and Shanghai-Wuhan-Chengdu Railway (Mingyang, 2017). Alternating current (AC)-and direct current (DC)electrified railways are close than ever before, which may give rise to dangerous voltage on urban rail transit traction network through electromagnetic induction. Therefore, we choose Ninghe Intercity Transit as the experimental subject and launch field experiment which basically covers main communication systems of it. And through performance comparison, relevant departments can find out the most possible interference cause and consequence from diverse disturbance sources.
However, sophisticated equipment build process is one of the technological difficulties that nobody has conquered so far, but our experimental scheme is perfect enough to analyze almost all EMI problems on the selected objective. Other research teams can later use the paper for reference to draw broad conclusions.
2. Technology roadmap and experiment setup 2.1 Technology roadmap Figure 2 explains our approach that combines theoretical and experimental analysis.
Three systems are selected at the stage of object refinement, for a daily operation of Nanjing Metro needs three kinds of services using radio communications: terrestrial trunked radio (TETRA) for command and dispatch service, CBTC, a railway signaling system and PIS providing public transport information (Peng, 2015), as illustrated in Table 1.
Normally, the three systems work within their respective frequency bands, but signal transmissions malfunction sometime due to external EMI. Thus, we deployed our experiment and analysis to figure out to what extent signal transmissions is affected under different conditions.

External interference test
To learn how the three communication systems normally work, the interference test was conducted in outer space. Figure 3 depicts the connectivity among main equipment.
Antenna input was considered as some degree of interference, which could be calculated by spectrum analyzer. Sometimes low noise amplifier or attenuator was necessary to get more effective data. Then we obtained related parameters such as antenna gain and feeder loss provided by EMI measurement receiver.
We chose seven test points, the distribution of which is exhibited in Figure 4. Points 1-5 were set along the Shanghai-Wuhan-Chengdu Railway every 500 m, while points 6 and 7 on the Beijing-Shanghai high-speed side. However, among them, points 5 and 7 were on the concrete part of the bridge, whereas others on the steel part. Thus, we could pinpoint the main influence factors. Besides, for better accuracy, test duration at every point should be no less than 30 min.

Interference comparison test
We conjecture that AC carried by Beijing-Shanghai High-Speed Railway ( Figure 5) would affect DC traction device of Ninghe Intercity Transit. Hence, the same test was conducted as comparison group when tracks of Beijing-Shanghai High-Speed Railway and Ninghe Intercity Transit converged.

System level electromagnetic compatibility performance test
Apart from interference source search, real performance of the three systems in vehicleground and ground-vehicle communication process need to be judged, so that we can find out the optimal mode, with which service quality will not dramatically drop when faced with interference. Conversely, frequency band occupancy problem was paid much attention in our test, because we regard cochannel interference as poor performance culprit. For research integrity, we made an attempt to take all possible modes into consideration. It consumed time because of lack of prior experience; however, we insisted on completing our goal and managed finally. We also need to say that experimental devices were chosen   Satellite image of Beijing-Shanghai High-Speed Railway, which is taken from Google map SRT according to standards or protocols, such as GB50382-2006 and IEEE 802.11 wireless LAN protocol group, so the experimental data was reliable. They were classified into two categories to achieve two-way communication.
Here is a list of main simulated modes that the three communication systems applied: (1) TETRA: 851-870 MHz (800 MHz) antenna coverage mode.
2.4.1 Bridge deck equipment. On the main bridge, the mode-related antenna and access point (AP) were installed in the following order ( Figure 6), so that the three systems could work, respectively. In detail, seven sets of CBTC (DSSS) devices stood at points 2, 4, 6, 8, 10, 12 and 14 beside the track; three sets of CBTC (FHSS) devices were located at points 2, 4 and 6 beside the track; CBTC (waveguide) was placed in the line section from points 2-4; only one TETRA and PIS (TD-LTE) were deployed, the antenna (or transmitting device) of which was located at point 1; seven sets of PIS (a/g) stood at points 3, 5, 7, 9, 11, 13 and 15 alongside Multirails environment the track. In addition, point 1 was located at the bridgehead on the east bank, other points were arranged every 100 m westward.
Then changing view to dead ahead of Ninghe Intercity Transit (Figure 7), we could see that both trackside AP and vehicle AP were mounted on the metal bridge deck below the cable bracket, while the crack waveguide was placed on the metal platform.
The bridge deck equipment communicated with ground devices via fiber optic cable successfully in our test. We intended to generalize AP layout reference method, by which the three systems could work, respectively.
2.4.2 Ground equipment. Figure 8 depicts the ground equipment, which includes one test server, one ground server, a TETRA base station, switches and power supply equipment. We installed the devices in rentals or makeshift houses.
2.4.3 Equipment collaboration. With collaboration between bridge deck and ground equipment, the overall test equipment installation was accomplished. Figure 9 exhibits the transmission way of the three communication systems, that is, TETRA antenna is Ground equipment installation SRT controlled by the ground base station; LTE-RRU and BBU is a group; signals from CBTC and PIS AP can be recognized through photoelectric conversion; and seven test points are also marked. Finally, the whole system was put into use and got satisfying effect.

Tests and results
We conducted simulations on the platform we established. We also adjusted the selected points after each test for accuracy, so that the effects of random factors on the test results could be avoided. Totally, we collected data from 14 test points of three experiments, which costs one and a half year. Finally, we obtained a sufficient number of test samples and here are the results.
3.1 External interference test and interference comparison test 3.1.1 Tetra test. Figure 10 shows the widest interference band of TETRA at point 3, that is, only 806.4-807.6, 808.8-810.2, 810.6-811.6, 812.6-817.0 and 817.4-821.0 MHz of up-band and 851.8-852.2 and 856.8-859.2 MHz of down-band were immune to interference. After overall consideration from seven test points, the most suitable frequency band was intended to be 812.6-814.2 and 857.6-859.2 MHz, where existed no signal at all.
Then the comparison test was launched and indicated when Beijing-Shanghai High-Speed Railway passed by, interference signals occasionally were generated but not in fixed band. Figure 11 demonstrates the interfered band at points 4 and 7, which was 827.0-830.0, 832.4-834.6, and 894.0-896.4, 898.6-899.6 MHz. However, maximum interference level was only À88.09 (833.0 MHz) and À83.80 dBm (899.2 MHz), less than À80 dBm.
3.1.2 CBTC test. In practice, CBTC frequency band is divided into 11 channels for signal transmission. Thus, four points (points 1, 3, 4, 6) were selected to detect the number of access devices and maximum interference level of each channel, presented in decibels. Table 2 lists the external interference test results for CBTC. Figure 12 is the typical result for CBTC interference comparison test. The interfered band at test point 3 was 2400.0-2405.8 MHz, while maximum level equaled À90.51 dBm (2402.0 MHz), which was very narrow. And there existed even no interference at test point 7. Nevertheless, the spectrum of point 6 exhibited "glitch" moment, as shown in Figure 13. The interfered band extended to be 2455.  Table 3. When a train of Beijing-Shanghai High-Speed Railway was passing by, BER increased, and no call dropped throughout the testing process.    (Kuang and Liu, 2017), so we selected three test channels to compare PIS antidisturbance performance within different frequency bands (Table 7). During the test, dynamic control was applied.

Multirails environment
Then we adopted PIS LTE mode, as mentioned in Tang et al. (2016). We set RSSI at points 3 and 7 as À49 and À73 dBm, respectively, otherwise field strength coverage from a distance farther than 1 km turned so weak that results might get distorted. For the same purpose, the one-way packet loss rate was also tested under different throughput, that is, 1 Mbps and 400 Kbps, respectively.
As depicted in Figure 14, both vehicle-ground and ground-vehicle process went well. When a high-speed train passing on the Beijing-Shanghai Railway tracks, the vehicleground one-way throughput reached 6.629 and 4.048 Mbps, 0.02 and 0.1 Mbps higher than the throughput in normal conditions, which means the throughput did not reduce at all. Moreover, response time did not increase and no packet loss took place at any test point.  GHz PIS is preferred, it will occupy part of channels that CBTC use. Hence, we desire to find out to which extent the two systems influence each other. Table 8 concludes their independent test results at point 2, in which channel of CBTC were 11 and 1, while PIS channel was 1. Then in full load condition, the throughput of CBTC and PIS were recorded as Figure 15. It implies that CBTC throughput reached 5.062 Mbps, while PIS performance was bad, dropping to 28.728 Mbps. Next, we turned CBTC to light load mode, and traced out PIS performance. We can easily read from Figure 16 the PIS packet loss rate as well as delay decrease.

Discussions
We further analyzed the experiment results to summarize practical recommendations that communication systems would take in such complex environment. The following sections include data processing and empirical formula supporting.
4.1 External environment and additive factor analysis 4.1.1 External environment analysis. Through external interference test, we learned that the best frequency band for TETRA was 812.6-814.2 and 857.6-859.2 MHz.

SRT
For CBTC, maximum power in each channel was less than À80 dBm, although the exact values in corresponding and adjacent channels rose when more access devices joined in the system. And moreover, AP number and maximum interference level of the system exhibited the same tendency, as shown in Figure 17, which exemplifies results at point 3. 4.1.2 Multirails environment. Compared with those in normal conditions, performances of three communication systems usually experienced reduction when a high-speed railway passed by. However, the change was not so huge as to cause abnormal signal transmissions. Specifically speaking, TETRA and PIS were not interfered at all. While for CBTC, "glitch" phenomenon emerged at point 6, but the probable interference source was the nearby WLAN signal used in FHSS rather than the adjacent track.
Hence, we hypothesize that the insignificant interferences owe to adequate distances between the tracks, and here is the theoretical verification: to decrease interference, autotransformer power-supply mode has been adopted by Beijing-Shanghai High-Speed Railway and Shanghai-Wuhan-Chengdu Railway, so that the Dashengguan overhead line is a four-wire system. The distances between the four lines and our experimental objective (Ninghe Intercity Transit communication cable) were measured to be 7.9, 12.9, 22.9, 27.9 m, respectively. 4.1.2.1 Magnetic hazard potential. Under normal operating conditions, it can be described as equation (1):

Multirails environment
where M i is the ith mutual inductance coefficient between overhead lines and the communication cable at 50 Hz; I p = 3.7 km is the approximate bridge span; I d = 1,200 A is the equivalent traction current of overhead contact network; l , T, K are the rail shielding factor, the communication cable shielding coefficient, and the magnetic integrated shielding coefficient of ground conductor. The potential turns out to be 26. 73, 23.463, 18.117, 16.335 V, so E = 36.19 V according to superposition method. Obviously, E is under maximum 100-150 V. Considering about short circuit faults, I s = 20-30 kA replaces I d , so the potential is calculated to be 668.327, 586.64, 452.98, 408.42 V. Thus % 1,078.33 V, which is also less than 1,080 V. (2)].

Telecommunication lines interference voltage [equation
where: In equation (3), I pr is the equivalent interference current at 800 Hz; I z is the total load current of overhead contact network; Z c = 600 X is the characteristic impedance of the communication line. M e equals 330, 250, 160, 120 m H/km; U e equals 0.081, 0.0619, 0.0396, 0.0297 mV. Thus, E = 0.113 mV, much less than the maximum interference value.
Therefore, as long as the space between the overhead wire and communication cable is more than 8 m, the surrounding electromagnetic environment is basically good. It holds up SRT in Dashengguan section for the shortest track distance equals 8.2 m (between Ninghe Intercity Transit uplink and Beijing-Shanghai High-Speed Railway). Moreover, if the lines are shielded by rows of trees, absorption apparatus, etc., the electromagnetic radiation influence will be much smaller.
4.1.3 Additive factor analysis. Figure 18 explains that during TETRA performance test (take TETRA as an example), BER at point 7 reached 3%, while at point 5, it was only 1%. However, the latter growth rate was bigger when a railway passed by. It means TETRA performance is closely related to field strength. Like point 7, signals were weakened by the bridge's shielding effect there, so its service quality degraded.
In addition, our research proves no obvious relevance between the communication system performance and weather condition, and moreover, the shielding effects of steel as well as concrete structure on field strength are on the whole the same.
4.2 Mode comparison 4.2.1 CBTC mode selection. CBTC mode selection is regarded as research emphasis during the system level performance test, on account of its transform convenience. Table 9 lists CBTC performance under antenna redundant and nonredundant mode at points 3 and 7. When a high-speed railway was running on the close track, the short-term throughput decreased under redundant mode, and delay was 3 ms. However, switching to the nonredundant mode, CBTC throughput experienced significant growth at point 3, which attributed to available bandwidth reduction caused by received signal strength drop. Then we closed neighboring AP at point 7, the field strength turned out to be stable. Nevertheless, packet loss rate under nonredundant mode still increased when CBTC was interfered.
Next, we make a comparison between two modes, throughput shrank in nonredundant condition, meanwhile, response time (delay) and packet loss rate had significant growth. Thus, we deem that CBTC performs better to some extent under antenna redundant mode.
Furthermore, CBTC had very low packet loss rate under waveguide mode in our test, which validates waveguide application advantages. While under FHSS mode, CBTC throughput was constant, and the loss rate could also be accepted. However, WLAN signals from FHSS devices might interfere with CBTC. Therefore, CBTC DSSS waveguide mode is preferred.
4.2.2 PIS mode selection. PIS also has two commonly used modes, namely, the mesh mode and LTE mode. We tested PIS performance under mesh mode within 2.4 and 5.8 GHz frequency bands, and found more than once that secondary interference was involved in the In comparison, PIS under LTE mode was much more stable: almost no delay was detected in the system level performance test. It should conform to characteristics of LTE, that is, the dedicated frequency band and interference rejection combining technology (Liu et al., 2016).
To sum up, appropriate modes should be selected for different needs, so that communication systems can work even well.
4.3 Frequency band usage 4.3.1 PIS 2.4 GHz versus 5.8 GHz. Nowadays, 5.8 GHz PIS gains the popularity among scholars, because it contains abundant information and is able to meet the high requirements of wireless communication. However, we need to take interference level into account, for the equal importance of communication quality. Our test supplemented that 5.8 GHz PIS experienced a larger attenuation than PIS with 2.4 GHz frequency band, that is to say, the response time and packet loss rate grew a lot when interference emerged. Thus, we should judge the two frequency bands more comprehensively.
4.3.2 CBTC and PIS-shared system. If PIS system occupies 2.4 GHz frequency band, cochannel interference will emerge when it works together with CBTC. The independent performance test showed that PIS throughput was nearly tenfold of CBTC throughput, and the PIS packet loss rate was a little higher. Then the frequency-sharing test was conducted, as a comparison. It indicated that CBTC performance did not decreased significantly, for channel 11 was only occupied by CBTC, so that CBTC could maintain normal operation. While performance of PIS dropped a lot. Finally, we changed CBTC to light load, delay and loss rate decreased to normal level again. Hence, PIS is considered normal when CBTC only does its basic job.
The result implied that if two communication systems share the same channel, performances will turn worse. The assumptions can be verified as follows.
The channel overlap interference degree can be described by interference factor jspec(f)j 2 , and the power spectrum density of OFDM signal is the sum of power spectrum subdensity on N subcarrier [equation (4)]:

SRT
The OFDM signal can be affected by filter of IEEE802.11 standard wireless network, thus, the power spectrum density should greatly reduce. Moreover, IF filter frequency response function can be described as equation (5): where v is the angular frequency (18). Hence, the overlap of two channels at a specified frequency is calculated as equation (6) where ch i; j ð Þ ¼ f À 2412 À 5 i À 1 ð Þ

:
This function can be integrated for a certain frequency band, and then we will obtain the interference factor between the two channels, namely, olf i,j . It verifies that cochannel interference really exists.

Conclusions
This study gains valuable information from three integral tests: external interference test for communication systems working on characteristics identification; interference comparison test confirming main disturbance factors; and system level electromagnetic compatibility performance test considering the overall performance of three communication systems (TETRA, CBTC and PIS), so as to dig into underlying principles. Through unremitting efforts, our one-to-one scale analog equipment installation method can also be extended to other projects, including sites where communication devices are especially intricate. After scientific analysis, we make following conclusions: External EMI has no obvious influence on main functions of urban rail transit communication systems, as long as the field strength coverage stays good. A high-speed passing railway can create large electromagnetic field, and AC traction devices can affect DC operation, resulting in communication equipment malfunction, etc. But an enough distance, 8 m or more, between a passing railway and the tested transit can keep the communication system from being influenced. Plus, the electromagnetic hazard under normal or short-circuit fault conditions are acceptable. The performance of urban rail transit communications is not related to bridge materials or weather conditions. Furthermore, appropriate electromagnetic shielding measures, such as rows of trees along the track and protection over communication cables, can suppress the shielding effect. Different communication modes may present various performances: DSSS waveguide and LTE mode are more stable for CBTC and PIS. It is better that CBTC and PIS do not share a same frequency band, otherwise communication quality will reduce due to cochannel interference.
Our approach, including experiment setup and theoretical analysis, successfully helped us find out crucial facts of EMI on communication systems. Given the pioneering structure and load of the Nanjing Dashengguan Bridge, our study also provides convincible recommendations and solid reference to other complex rail transit systems.