Advertisement

Annals of Telecommunications

, Volume 74, Issue 1–2, pp 25–33 | Cite as

Numerical evaluation of human exposure to 3.5-GHz electromagnetic field by considering the 3GPP-like channel features

  • Congsheng Li
  • Chunying Xu
  • Ruixin Wang
  • Lei Yang
  • Tongning WuEmail author
Article
  • 152 Downloads

Abstract

Human exposure to 3.4–3.6-GHz radiofrequency (RF) electromagnetic field (EMF), which is the frequency band utilized by trial test of the fifth-generation mobile communication systems (5G), has been numerically analyzed in the study. The study evaluated the EMF exposure of this frequency band by taking into account of the channel features. Two exposure scenarios were reconstructed according to the technical specification on channel modeling from the 3rd Generation Partnership Project. The channel features of the reconstructed EMF were numerically validated. The equivalent source principle and the finite-difference time-domain method were applied to calculate the RF energy specific absorption (SA) using three human models. The results revealed that the exposure scenarios with various channel features affected whole-body SA (WBSA) by about 50–70%. The variation was mainly introduced by the configuration of the incident waves defined by the channel models. Dosimetric difference between the two exposure scenarios for some tissues has been presented and discussed. The results demonstrated that the anatomy of the model was also a factor influencing SA.

Keywords

Fifth-generation mobile communication systems (5G) Human model Channel modeling Specific absorption Radiofrequency exposure 

1 Introduction

The emergence of each new generation of mobile communication technology has drastically changed the social life since 1980s. Accompanied with the convenience that the technology brought, public concerns on health effects by exposure to the radiofrequency (RF) electromagnetic field (EMF) have been raised. To limit the excessive emission, International Commission on Non-Ionizing Radiation Protection (ICNIRP), has published the guidelines [1]. According to the guidelines, uniform plane-wave exposure scenario is assumed for evaluating the far-field EMF exposure. Exposure to EMF with different frequencies or signal types is supposed to have different RF power absorption, and the dosimetric studies have been intensively conducted for the second-, the third-, and the fourth-generation mobile communication signals [2, 3]. In contrast, realistic wireless signal transmission between a transmitter and receiver is affected by the propagation path (object penetration, reflection, scattering, diffraction, and absorption caused by atmospheric gases, fog, and precipitation). Human under exposure will experience a superposition of signals with differences in attenuation, delay, and phase shift. The situation is in sharp contrast to those prescribed in the ICNIRP guidelines for plane-wave EMF evaluation. As consequence, researchers conducted a series of studies to investigate the dosimetric significance of the fading patterns [4] and the incident directions [5, 6]. The results indicated that these parameters may change the RF power absorption by the human body.

Continuously, increasing demand for higher data rates, larger network capacity, higher energy efficiency, and higher mobility has motivated research for the fifth-generation mobile communication systems (5G) [7]. Since 5G aims to provide the massive connectivity for large coverage with high data transfer rate, it should be applicable to diverse fading scenarios. Moreover, some over-the-air (OTA) techniques, as sending and receiving multiple data signal simultaneously over the same radio channel by exploiting multiple-input and multiple-output (MIMO, [8]) or enhancing the performance of a single data signal (e.g., beamforming and diversity [9]), have become the indispensable components of 5G. The abovementioned elements reveal that the radio propagation channel for 5G is much complicated than ever before and the modeling is essential to the system performance as well as for the network realization [10]. As consequence, there have been several studies on 5G channel characterization [11, 12, 13]. In contrast, there are very few studies on evaluating 5G EMF exposure by considering the channel modeling but it should be investigated because it determines the key features of 5G.

5G utilizes wide frequency spectrum and can spread to millimeter-wave band. In it, the 3.5-GHz band has been selected by many mobile manufactures, operators, and chipset providers for the first turn of 5G trial test due to the availability of the spectrum and the instrumentation [14, 15]. For the frequency band below 6 GHz, international standard organization has defined two canonical three-dimension (3D) channel models for repeatable OTA measurement, i.e., urban macrocells (UMa) and urban microcells (UMi) models [16]. The two models are believed to represent the channel models in densely populated buildings.

This study presented the numerical method to reconstruct the EMF by modeling the UMa and UMi channels at 3.5 GHz. We compared the dosimetric difference between these two exposure scenarios using three anatomical models representing different age groups and genders. Whole-body averaged energy specific absorption (WBSA) and the SA for major tissues were to be calculated and compared for the two exposure scenarios.

2 Methods and materials

2.1 Human models

We used the three human models representing different age-groups in the simulation. They were the infant model (28 tissues, 0.74 m and 9.84 kg) [17] and the Chinese adult female (90 tissues, 1.62 m and 53.5 kg) and male (87 tissues, 1.72 m and 63.0 kg) models [18, 19, 20]. Frequency-dependent dielectric properties were from the database of Gabriel et al. [21].

2.2 Computational methods

To analyze the EMF power absorption in heterogeneous tissues using the resolution of cubic millimeter, finite-difference time-domain (FDTD, [22]) is the appropriate numerical method. However, FDTD needs huge memory and time cost when processing a large computational volume with a fine resolution. Previous researchers adopted the principle of equivalent source [23] in relevant studies [24]. The electromagnetic wave propagation between the antennae and the equivalent source surface can be efficiently computed by ray tracing [25], finite element method [26], method of moment [27], or obtained by in-situ measurement campaign [28]. As consequence, the derived E- and H-field values are implemented surrounding the human model by Huygens box [29]. Therefore, FDTD iterations can be updated in the reduced volume to save the computational cost. The key part in the method is to construct the E- and H- field on the equivalent source surface.

2.2.1 Reconstructing the EMF by taking into account of the channel model

To define the field values on the equivalent source surface, the parameters can be referred to the 3GPP description for the 3D channel models ([16], as shown in Table 1 of Supplementary file). In total, there are six clusters of signals to be realized in the simulations. We plot the schema of the six incident clusters in Fig. 1. According to the document, the antenna emits the six clusters of signals alternatively without any overlap on the time domain.
Fig. 1

Configuration of the six incident clusters for UMa (a) and UMi models (b). #1–6 refer to the six clusters of the incident rays. Each cluster is composed of two cross polarized signals. Black arrows indicate the wave factor, while red and green arrows represent the polarization of the E- and H- field, respectively. Degree values in the two subfigures represent angle of arrival (AoA) of each cluster

For each cluster, E-field is described by (1):
$$ \overrightarrow{E(t)}=\sum \limits_{i=0}^2{A}_{\mathrm{vi}}\cos \left(\upomega \left(t-{D}_i\right)\right)\cdot {e}^{-{\left(4\uppi \frac{t-{D}_i-\tau }{\tau}\right)}^2}\cdot \overrightarrow{k_v}+\sum \limits_{i=0}^2{A}_{\mathrm{hxi}}\cos \left(\omega \left(t-{D}_i\right)\right)\cdot {e}^{-{\left(4\uppi \frac{t-{D}_i-\tau }{\tau}\right)}^2}\cdot \overrightarrow{k_{hx}}+\sum \limits_{i=0}^2{A}_{\mathrm{hyi}}\cos \left(\omega \left(t-{D}_i\right)\right)\cdot {e}^{-{\left(4\uppi \frac{t-{D}_i-\tau }{\tau}\right)}^2}\cdot \overrightarrow{k_{hy}} $$
(1)
where, i = 0, 1, 2 represents the three pulses produced by the different propagation paths; D corresponds to the time delay for three components in the cluster; Av and Ah are the amplitudes of the cross-polarized E-field with Ah being further decomposed to two components parallel to X axis (Ahx) and Y axis (Ahy); \( \overrightarrow{k_v}=\left(\mathrm{0,0,1}\right) \), \( \overrightarrow{k_{hx}}=\left(\mathrm{1,0,0}\right) \), and \( \overrightarrow{k_{hy}}=\left(\mathrm{0,1,0}\right) \) are the wave vectors.
Ah and Av satisfy (2):
$$ 10\ \log \left({A}_{\mathrm{vi}}^2/\left({A}_{\mathrm{hxi}}^2+{A}_{\mathrm{hyi}}^2\right)\right)=\Big\{{\displaystyle \begin{array}{c}\ 8.13\ \mathrm{dB}\kern1.5em \left(\mathrm{UMa}\ \mathrm{model}\right)\\ {}0.74\ \mathrm{dB}\kern1.5em \left(\mathrm{UMi}\ \mathrm{model}\right)\end{array}} $$
(2)
and (3),
$$ {P}_i=10\ \log \left({A}_{\mathrm{vi}}^2+{A}_{\mathrm{hxi}}^2+{A}_{\mathrm{hyi}}^2\right)\kern0.5em i=0,1,2 $$
(3)
where, P represents the power of each of the three pulses in the cluster and its relative value is described in Table 1 of Supplementary file.
The signal pattern for each cluster is shown in Fig. 2a, which is composed of 3 Gaussian pulses. The interval for the adjacent pulses is 5 ns. The intensity for each Gaussian pulse was designed according to the relative power value in Table 1 of Supplementary file. According to Provisions of the Peoples Republic of China on Radio Frequency Division [30], 200 MHz from 3400 to 3600 MHz is firstly allocated for 5G service. \( \uptau \overrightarrow{=}2.3\mathrm{e}-8 \) was defined so as the 3-dB bandwidth was 200 MHz (Fig. 2b).
Fig. 2

Formulation of the incident signal applied in the simulations. Each incident cluster is composed of three Gaussian pulses (a) and the central frequency is 3500 GHz with 200 MHz bandwidth (b)

The six clusters of signals emit alternatively in the time domain. In the simulations, we can compute the dose for each cluster and make summation by taking into account of the time delay in Table 1 of Supplementary file.

2.2.2 Dose calculation

SA was selected as the dose metric in the simulations. We made 1200 time samples for the incident cluster, corresponding to a sampling rate of about 8 GHz (two times of the carrier frequency). Numerical experiments have been performed to ensure that the sampling rate was sufficient to provide a stable result. The procedures can be referred to our previous study [31]. Consequently, \( \overrightarrow{E}\left(\mathrm{x},\mathrm{y},\mathrm{z},\mathrm{t}\right) \) was calculated by FDTD updates and then transformed to frequency domain (\( \overrightarrow{E}\left(\mathrm{x},\mathrm{y},\mathrm{z},\mathrm{f}\right) \)) by Fourier transform. Hence, SA was obtained by (4):
$$ \mathrm{SA}\left(\mathrm{x},\mathrm{y},\mathrm{z}\right)={\int}_{f_{\mathrm{L}}}^{f_{\mathrm{H}}}\frac{P\left(\mathrm{x},\mathrm{y},\mathrm{z},\mathrm{f}\right)}{m\left(\mathrm{x},\mathrm{y},\mathrm{z}\right)} df={\int}_{f_{\mathrm{L}}}^{f_{\mathrm{H}}}\frac{\sigma \left(\mathrm{x},\mathrm{y},\mathrm{z},\mathrm{f}\right){\left|E\Big(\mathrm{x},\mathrm{y},\mathrm{z},\mathrm{f}\Big)\right|}^2}{m\left(\mathrm{x},\mathrm{y},\mathrm{z}\right)} df $$
(4)

Debye model for frequency-dependent material [32] was applied in the simulations.

The simulation volume was 440 × 340 × 880 mm3 for the infant model, 600 × 360 × 1740 mm3 for the female adult, and 610 × 380 × 1840 mm3 for the male adult model. The volume was discretized by uniform voxels of 1.5 × 1.5 × 1.5 mm3. Human models were placed in the center of the simulation volume. The minimum distance between the boundary of the volume and the model was 60 mm (for the male model). SEMCAD-X v14.8 (https://speag.swiss/news-events/news/simulation/) was implemented as FDTD solver.

2.2.3 Verification of the reconstructed EMF

In order to ensure that the channel models are correctly implemented and hence capable of generating the propagation environment, 3GPP provides four methods to validate the model, i.e., power delay profile, cross polarization ratio, spatial correlation, and Doppler/Temporal correlation. We realized the first three approaches in the numerical simulations.
  1. a)

    Power delay profile

     
It is used to verify the sequence of the impulses. The evaluation point is in the center of the volume where we recorded the time sequence of the power. In 3GPP requirement, the reference for power delay profile is defined and the limits for the tolerance are provided, e.g., ± 0.85 dB for maximum power deviation per cluster (UMi) and ± 0.11 ns for maximum excess delay deviation per cluster (UMi).
  1. b)

    Spatial correlation

     
The evaluation checks whether the E-field in the realized EMF environment follows the theoretical curve. According to the 3GPP requirement, E-field values were recorded at the middle height level along Z direction, on 11 positions distributing in the centerline of the simulation volume (perpendicular to the direction of AoA = 0°, as shown in Fig. 1), with a sampling interval of 0.1 wavelength. The correlation can be calculated by (5):
$$ {\rho}_i=\frac{\mathrm{Cov}\left({E}_i(t),{E}_j(t)\right)}{\sqrt{\mathrm{Dev}\left({E}_i(t)\right)}\sqrt{\mathrm{Dev}\left({E}_j(t)\right)}} $$
(5)
where, i, j = 0, 1, 2, …, 10; Cov is covariant; Dev is standard deviation;\( {\mathrm{E}}_{\mathrm{i}}\left(\mathrm{t}\right)=\sqrt{{\mathrm{E}}_{\mathrm{i}\mathrm{x}}^2\left(\mathrm{t}\right)+{\mathrm{E}}_{\mathrm{i}\mathrm{y}}^2\left(\mathrm{t}\right)+{\mathrm{E}}_{\mathrm{i}\mathrm{z}}^2\left(\mathrm{t}\right)} \).
3GPP prescribed the limits of spatial correlation for different testing methods.
  1. c)

    Cross polarization

     
It evaluates the cross-polarization ratio (V/H) at the center point of the simulation volume by (6):
$$ \mathrm{V}/\mathrm{H}=10\times \log \left(\frac{\Sigma {E}_{\mathrm{z}}^2}{\Sigma \left({E}_{\mathrm{x}}^2+{E}_{\mathrm{y}}^2\right)}\right) $$
(6)

3GPP regulated that the V/H maximum deviation should not exceed ± 0.9 dB.

3 Results and discussions

3.1 Verification of the reconstructed EMF field

In our simulation, we applied the time delay, relative power value, and AoA as described by 3GPP (Table 1 of Supplementary file) for power delay profile analysis. There was no noise or reflection added in the computational domain. Therefore, the obtained time delay and the relative power values for the six clusters of the incident waves were no doubt the same as those in 3GPP document (Fig. 3). The deviation from the referenced spatial correlation value was calculated and the case with j = 0 is presented in Fig. 4 for an example. The simulated spatial correlation (UMi) was in the upper and lower limits by 3GPP documents (Fig. 4a). Although 3GPP document did not define the tolerance range for UMa, the maximal deviation from the referenced spatial correlation value by UMa was even less, below 0.04 (Fig. 4b), thus complying with the practical requirement in OTA testing [33]. The cross-polarization ratio, very close to the 3GPP technical specifications, has been obtained from the simulations (Table 1). The V/H maximum deviation was far less than ± 0.9 dB, satisfying the requirement of 3GPP document.
Fig. 3

Power delay profile for UMa model (a) and UMi model (b) by simulations

Fig. 4

Spatial correlation results for UMi (a) and UMa (b) models. The tolerate of limit prescribed by 3GPP was indicated for UMi. No limits from 3GPP available for UMa but the maximum deviation was less than 0.1. The correlation was calculated with j = 0

Table 1

Comparison for the results of cross polarization ratio (V/H)

Channel model

V/H by 3GPP technical requirement

Simulated V/H

Deviation

UMa

8.13

8.096

0.034

Umi

0.74

0.739

0.001

Among them, power delay profile determined the sequence of the signals as well as the relative power values, which characterized the exposure duty cycle and the intensity. Spatial correlation assessed the distribution of the field in the simulation volume, which was essential to local RF power absorption. Cross polarization ratio described the components of incident signals and it was known that the polarization of the EMF significantly influenced the power absorption [20]. In contrast, Doppler (temporal) correlation was not considered in the simulations because the motion with a speed of 100 km/h would introduce in a frequency shift of less than 10 kHz (less than 0.001% variation in terms of the carrier frequency). The resultant dielectric change was marginal and the dosimetric results would be unchanged.

In conclusion, the reconstructed EMF environment followed the UMa and UMi models. The simulation results, thus, not only represented the exposure scenarios with a certain incident energy density but also took into account the radio propagation channel features.

3.2 RF energy absorption

WBSA and the SA by five major tissues are shown in Fig. 5. The results were calculated with 1 J/m2 incident energy density (in 0.1 s). It revealed that the UMi featured EMF introduced enhanced WBSA for 50–70% compared with that of UMa. Elevated SA of about 50–70% has also been observed for the skin, fat, and eyes under the case of UMi channel.
Fig. 5

SA for the whole body and five tissues. The results are calculated with the incident energy density of 1 J/m2 in 0.1 s

The main reason for SA difference was the configuration of the clusters in each exposure scenarios. A slice view at the chest level (Fig. 6) indicated that the respective incident cluster, although with the same serial number in UMa and UMi models, resulted in different energy absorption patterns. Table 2 shows the comparison of SA by the corresponding cluster between the two exposure scenarios. UMi model usually showed a higher dose. For UMi model, AoA of five clusters (#1, #2, #4, #5, and #6) located between around − 30 and 0°, while only one cluster from UMa model had AoA between − 30 and + 30°. It indicated that the frontal incidence was dominant for the UMi model. It can also be found in Fig. 6 that under UMi model, much more frontal part of the human body was illuminated. The previous study [5] demonstrated that the frontal plan-wave incidence (both V and H polarizations were considered) can induced about 40% higher whole-body specific absorption rate than the lateral ones [5]. Therefore, it was reasonable to see UMi model induced higher WBSA. Besides, the abovementioned five clusters of UMi model, which produced a frontal dominant exposure, had higher power repartition compared to the clusters in the UMa model (refer to relative power values in Table 1 of Supplementary file). As a consequence, 50–70% higher WBSA was observed for the UMi model. To note, in reality, for UMa, the base station installation height is 25 m and its transmitted power is 46 dBm, while for UMi, the base station installation height is 10 m and with its transmitted power as 41 dBm. These factors should also be taken into consideration for realistic exposure analysis and hence, the relative distance between the exposed subject and the base station is also an essential factor for the energy absorption.
Fig. 6

Comparison of energy absorption pattern by each incident cluster for UMa and UMi models (slice view at the chest level)

Table 2

Difference of the SA for two scenarios. The results are calculated by: \( \%=\frac{{\mathrm{SA}}_{\mathrm{UMi}}-{\mathrm{SA}}_{\mathrm{UMa}}}{{\mathrm{SA}}_{\mathrm{UMa}}}\times 100\% \)

 

#1

#2

#3

#4

#5

#6

Infant model

 Whole body (%)

44.78

24.66

7.63

7.83

75.08

0.43

 Skin (%)

32.46

27.76

16.97

14.79

869.09

− 8.44

 Fat (%)

50.51

25.56

4.45

6.28

72.79

4.39

 Muscle (%)

57.62

− 28.63

4.66

2.61

86.33

4.27

 Brain (%)

21.60

9.78

6.27

4.14

− 78.32

− 0.28

 Eye (%)

57.10

18.00

25.10

− 5.18

272.80

− 8.74

Female model

 Whole body (%)

58.63

32.54

10.15

12.71

102.48

1.32

 Skin (%)

58.34

48.00

15.92

21.75

90.33

− 1.31

 Fat (%)

58.56

23.61

6.53

6.97

97.44

3.52

 Muscle (%)

62.66

22.92

6.19

5.67

129.41

5.08

 Brain (%)

8.33

11.16

12.77

8.15

− 11.15

3.58

 Eye (%)

68.21

− 2.55

46.51

2.78

248.73

17.24

Male model

 Whole body (%)

46.89

24.72

13.26

16.60

125.66

7.36

 Skin (%)

32.88

26.98

19.29

25.51

119.05

4.82

 Fat (%)

47.61

20.46

7.31

11.01

89.05

8.49

 Muscle (%)

68.74

26.23

9.39

10.51

164.31

11.64

 Brain (%)

18.12

9.66

− 1.94

0.39

− 0.55

− 2.12

 Eye (%)

83.27

23.25

26.82

− 1.80

410.28

3.13

Since the energy absorption for individual tissue was dependent on the incident angle, the rotation of the human model can change the dosimetric results. Nevertheless, the presented configurations gave out much conservative WBSA because of the abovementioned reasons (frontal plan-wave incidence prevailed over the lateral incidence and most of the energy was convoyed by the current front incident clusters).

SA of specific tissues also depended on their distribution in the body. At 3.5 GHz, the RF power attenuated drastically in the human body (e.g., the skin depth was 1.5 cm for skin, 7.8 cm for fat, 1.5 cm for muscle, and 2 cm for bones). Accordingly, deeper tissues such as the brain were less affected by the environmental channel features compared with the superficial tissues (the eyes, skin, and fat). The same reason can apply to muscle (locating beneath fat): difference of SA by muscle under the two exposure scenarios was 30–40% for infant and adult female models, less significant than the whole-body level. Male adult had 70% SA difference for muscle mainly because it was from a lean individual [18]. In that case, less fat existed in the body and muscle distributed superficially in the body.

It was not easy to find out the reason why the infant and the female adult had similar WBSA solely from the point view of the physical parameters because their physical parameters (height and weight) and the derived values (e.g., body mass index and body surface area) were obviously different. It might be attributed to the individual anatomical structure. For example, the skull of the infant was much thinner compared with that of the adults and it resulted in higher WBSA for the brain. The infant had the smallest dimensions and the waves penetrated deep in the body but at the same time, infant had higher fat content compared with the adult female. In such a case, the energy absorption by the muscle layer from the two models was at the similar level. And the energy absorption of fat for the infant was higher (Fig. 5). Fat and muscle accounted for 61% and 67% of the total weight for infant and adult female, respectively. Consequently, the derived WBSA was similar for the two models. In addition, the combination of the incident clusters (and the corresponding cross-section area) could also introduce the effect.

At 3.5 GHz, the minimum wavelength in the human tissues is at the order of 1 cm. 1.5 mm is, at least, less than 1/6 of the minimal wavelength in the tissue. We also reduced the resolution to 1 × 1 × 1 mm3 and conducted the comparative simulations using an 8-core CUP workstation with 256 GB RAM. About 5% difference in WBSA has been obtained and the maximal difference in the specific tissue was about 18% (blood and significant difference usually appeared in the tiny tissues). We believed that the current spatial resolution for FDTD was sufficient to derive the results. On the other hand, the GPU accelerated workstation used in the simulations can only solve the field distribution in the human models with the finest resolution at 1.5 × 1.5 × 1.5 mm3.

4 Conclusion

The study numerically reconstructed 3.5-GHz EMF environment with 3GPP-like channel features. Using the reconstructed exposure scenarios, human exposure to the 3.5-GHz EMF has been evaluated. The results indicated that 50–70% difference in terms of WBSA could be identified between the two cases. The effect was mainly due to the configuration of the incident clusters defined by the individual channel model. Besides the channel features, the dose for the studied tissues also relied on the anatomy of the model.

Notes

Funding information

The work is supported by grants from National Natural Science Foundation Project (Grant Nos. 61371187 and 61671158) and National Science and Technology Major Project (No. 2018ZX100301).

Supplementary material

12243_2018_682_MOESM1_ESM.docx (16 kb)
ESM 1 (DOCX 15 kb)

References

  1. 1.
    Ahlbom, A., Bergqvist, U., Bernhardt, J. H., Cesarini, J. P., Grandolfo, M., Hietanen, M., Mckinlay, A. F.et al. (1998). Guidelines for limiting exposure to time-varying electric, magnetic and electromagnetic fields (up to 300GHz). Health Phys, 74, 494–522Google Scholar
  2. 2.
    Martínez-Búrdalo M, Martin A, Sanchis A, Villar R (2009) FDTD assessment of human exposure to electromagnetic fields from WiFi and bluetooth devices in some operating situations. Bioelectromagnetics 30(2):142–151CrossRefGoogle Scholar
  3. 3.
    Krayni A, Hadjem A, Vermeeren G, Sibille A, Roblin C, Joseph W et al (2017) Modeling and characterization of the uplink and downlink exposure in wireless networks. Int J Antennas Propag.  https://doi.org/10.1155/2017/8243490
  4. 4.
    Li C, Yang L, Lu B, Xie Y, Wu T (2016) A reverberation chamber for rodents’ exposure to wideband radiofrequency electromagnetic fields with different small-scale fading distributions. Electromagn Biol Med 35(1):30–39CrossRefGoogle Scholar
  5. 5.
    Conil E, Hadjem A, Gati A, Wong MF, Wiart J (2011) Influence of plane-wave incidence angle on whole body and local exposure at 2100 MHz. IEEE Trans Electromagn Compat 53(1):48–52CrossRefGoogle Scholar
  6. 6.
    Vermeeren G, Joseph W, Martens L (2013) Statistical multi-path exposure method for assessing the whole-body SAR in a heterogeneous human body model in a realistic environment. Bioelectromagnetics 34(3):240–251CrossRefGoogle Scholar
  7. 7.
    Sharma S, Chakraborty M, Singhal M, Jha M (2014) A review of 4G and 5G in context of future of wireless communication. Int J Comput Appl 95(22):1–4Google Scholar
  8. 8.
    Farhang A, Marchetti N, Figueiredo F, Miranda JP (2014) Massive MIMO and waveform design for 5th generation wireless communication systems. In: 1st international conference on 5G for ubiquitous connectivity, pp 70–75Google Scholar
  9. 9.
    Jang J, Chung M, Hwang SC, Lim YG, Yoon HJ, Oh T et al (2016) Smart small cell with hybrid beamforming for 5G: theoretical feasibility and prototype results. IEEE Wirel Commun 23(6):124–131CrossRefGoogle Scholar
  10. 10.
    Mondal B, Thomas TA, Visotsky E, Vook FW, Ghosh A, Nam YH et al (2015) 3D channel model in 3GPP. IEEE Commun Mag 53(3):16–23CrossRefGoogle Scholar
  11. 11.
    Agiwal M, Roy A, Saxena N (2016) Next generation 5G wireless networks: a comprehensive survey. IEEE Commun Surv Tutor 18(3):1617–1655CrossRefGoogle Scholar
  12. 12.
    Rappaport TS, Sun S, Shafi M (2017) 5G channel model with improved accuracy and efficiency in mmWave bands. IEEE 5G Tech Focus 1(1):1–6Google Scholar
  13. 13.
    Medbo J, Börner K, Haneda K, Hovinen V, Imai T, Järvelainen J, et al. (2014). Channel modelling for the fifth generation mobile communications. In 8th European Conference on Antennas and Propagation (EuCAP), 219–223Google Scholar
  14. 14.
    Qualcomm. (2017). Qualcomm, ZTE and China mobile announce collaboration on 5G NR trials at 3.5 GHz to accelerate wide-scale 5G deployments in China. Qualcomm press release. https://www.qualcomm.com/news/releases/2017/02/21/qualcomm-zte-and-china-mobile-announce-collaboration-5g-nr-trials-35-ghz. Accessed 24 Nov 2018
  15. 15.
    Sue M. Why the 3.5 GHz CBRS band could be a breakthrough for 5G. sdxcentral. https://www.sdxcentral.com/articles/news/3-5-ghz-cbrs-band-breakthrough-5g/2017/03/. Accessed 24 Nov 2018
  16. 16.
    Study on 3D channel model for LTE. (2017). 3GPP TR 37.977 V14.1, 2017. 3rd Generation Partnership Project Google Scholar
  17. 17.
    Li C, Chen Z, Yang L, Lv B, Liu J, Varsier N et al (2015) Generation of infant anatomical models for evaluating electromagnetic field exposures. Bioelectromagnetics 36(1):10–26CrossRefGoogle Scholar
  18. 18.
    Wu T, Tan L, Shao Q, Zhang C, Zhao C, Li Y et al (2011) Chinese adult anatomical models and the application in evaluation of RF exposures. Phys Med Biol 56(7):2075–2089CrossRefGoogle Scholar
  19. 19.
    Wu T, Tan L, Shao Q, Li Y, Yang L, Zhao C et al (2012) Slice-based supine to standing postured deformation for Chinese anatomical models and the dosimetric results by wide band frequency electromagnetic field exposure: morphing. Radiat Prot Dosim 154(1):26–30CrossRefGoogle Scholar
  20. 20.
    Wu T, Tan L, Shao Q, Li Y, Yang L, Zhao C et al (2012) Slice-based supine-to-standing posture deformation for Chinese anatomical models and the dosimetric results with wide band frequency electromagnetic field exposure: simulation. Radiat Prot Dosim 154(1):31–36CrossRefGoogle Scholar
  21. 21.
    Gabriel C, Gabriel S, Corthout E (1996) The dielectric properties of biological tissues: I. Literature survey. Phys Med Biol 41(11):2231–2249CrossRefGoogle Scholar
  22. 22.
    Taflove A, & Hagness SC (2005) Computational electrodynamics: the finite-difference time-domain method. Artech house Google Scholar
  23. 23.
    Mix J, Dixon J, Popovic Z, Piket-May M (1999) Incorporating non-linear lumped elements in FDTD: the equivalent source method. Int J Numer Model Electron Netw Devices Fields 12:157–170CrossRefzbMATHGoogle Scholar
  24. 24.
    Wu T, Hadjem A, Wong MF, Gati A, Picon O, Wiart J (2010) Whole-body new-born and young rats’ exposure assessment in a reverberating chamber operating at 2.4 GHz. Phys Med Biol 55(6):1619–1630CrossRefGoogle Scholar
  25. 25.
    Bernardi P, Cavagnaro M, D’Atanasio P, Di Palma E, Pisa S, Piuzzi E (2002) FDTD, multiple-region/FDTD, ray-tracing/FDTD: a comparison on their applicability for human exposure evaluation. Int J Numer Model: Electron Netw Devices Fields 15(5–6):579–593CrossRefzbMATHGoogle Scholar
  26. 26.
    Rylander T, Bondeson A (2002) Application of stable FEM-FDTD hybrid to scattering problems. IEEE Trans Antennas Propag 50(2):141–144CrossRefzbMATHGoogle Scholar
  27. 27.
    Chakarothai J, Wang J, Fujiwara O, Wake K, Watanabe S (2014) A hybrid MoM/FDTD method for dosimetry of small animal in reverberation chamber. IEEE Trans Electromagn Compat 56(3):549–558CrossRefGoogle Scholar
  28. 28.
    Ilvonen S, Toivonen T, Toivo T, Uusitupa T, Laakso I (2008) Numerical specific absorption rate analysis and measurement of a small indoor base station antenna. Microw Opt Technol Lett 50(10):2516–2521CrossRefGoogle Scholar
  29. 29.
    Roy JE (2012) On using a closed box as the integration surface with the FDTD method. IEEE Trans Antennas Propag 60(5):2375–2379CrossRefGoogle Scholar
  30. 30.
    Ministry of Industry and Information Technology of the People’s Republic of China (2013). Provisions of the People’s Republic of China on radio spectrum allocationsGoogle Scholar
  31. 31.
    Li C, Yang L, Li CH, Xie Y, Wu T (2015) Dosimetric variability of the rats’ exposure to electromagnetic pulses. Electromagn Biol Med 34(4):334–343CrossRefGoogle Scholar
  32. 32.
    Gandhi OP, Gao BQ, Chen JY (1993) A frequency-dependent finite-difference time-domain formulation for general dispersive media. IEEE Trans Microw Theory Tech 41(4):658–665CrossRefGoogle Scholar
  33. 33.
    Fan, W., De Lisbona, X. C., Sun, F., Nielsen, J. O., Knudsen, M., & Pedersen, G. (2013). Emulating spatial characteristics of MIMO channels for OTA testing. IEEE Trans Antennas Propag, 61(8), 4306–4314Google Scholar

Copyright information

© Institut Mines-Télécom and Springer Nature Switzerland AG 2018

Authors and Affiliations

  • Congsheng Li
    • 1
  • Chunying Xu
    • 1
  • Ruixin Wang
    • 1
  • Lei Yang
    • 1
  • Tongning Wu
    • 1
    Email author
  1. 1.China Academy of Information and Communications TechnologyBeijingChina

Personalised recommendations