1. Introduction
The mmW frequency band is most promising for overcoming the gigabits per second barrier in the upcoming Wireless Communication System (WCS) [
1]. With reference of global data growth in cellular communication predicted by CISCO, the data growth will be seven times higher by 2021 when compared with the data demand of 2016. The estimated cellular data growth will be 143 Exabyte per quarter by 2021 [
2]. The mmW wave band is uniquely fit to serve the upcoming bandwidth hunger due to wider accessible bandwidth, frequency reutilization, and the minimized size of the base station and the mobile station components [
3]. A single carrier frequency division multiple access extended was an excellent hopeful radio wave access technology for the fourth generation, the third generation partnership project and long term evolution, and the long-term evolution advanced networks cellular system [
4,
5]. Right now for the 5G application, we need to move toward the higher frequency like mmW. The application of this study covers a great research area such as the advent of the Internet of Things and the Fifth Generation (5G) network planning era where optimum WCS channels categorizations have a direct impact on the resource allocation process [
6]. However, vast knowledge is needed on radio wave propagation to design mmW WCS. The frequencies band above 10 GHz not only considers the general propagation feathers like reflection, diffraction, and refraction but also needs to incorporate an environmental atmosphere effect such as rain, snow, storm, and the handover effect for mobility [
7,
8,
9]. This study covers the indoor radio propagation, which is free from rain, snow, storm, and ignorable mobility.
The frequency 38 GHz is most suitable for the 5G networks frequency in WCS from the estimated mmW band [
10,
11,
12]. To get the standard judgment of performance in WCS, a practical analysis of the propagation prediction for mmW is needed as a requirement for the cell deployment. The experimental way to evaluate the radio channels in a particular scenario is to measure the Received Signal Strength Indication (RSSI) and the Path Loss (PL) at multiple positions among the area by the deploying Receiver (Rx). To serve this persistence, an experimental measurement campaign has been performed on new wireless communication centers. The experimental environment is the basement floor of a two-story building in Universiti Teknologi Malaysia (UTM), in Johor Bahru Campus at Malaysia name as NWCC_P15a. The opportunity to run a measurement campaign is limited because a site-specific method is used in measurement, which makes the process time-consuming and costly. In this circumstance, RT simulation is robustly utilized for categorizing radio channel propagation for mmW. The RT is an essential tool for different smart technologies and WCS design [
13]. It significantly reduces the use of time-consuming extensive costly measurement campaigns [
14,
15,
16]. In the RT method, a ray is launched from the base station (TX) based on Uniform Theory of Diffraction (UTD) and Geometric Optics (GO). Furthermore, small-wavelength electromagnetic signal propagation is suitable for explaining the ray-optic estimate method, so RT simulation methods have been extensively used for the mmW in WCS [
17]. Therefore, to replace a time-consuming and expensive measurement campaign, the RT simulation method can be utilized since it incorporated all the geometric principals accurately.
The dependency of scenarios, Ray Launching (RL) continual angular dimension, and environmental objects accurate modeling is the main challenge of RT simulation. Additionally, reflection, diffraction, and refraction features of RT needs to be implemented correctly. In this study, a proposed algorithm developed for 3-D RT is implemented in the simulator. Furthermore, the SBR method is also implemented in the RT simulator for comparison of simulation time, the number of rays launched, and the number of rays received by Rx with a proposed method. As per literature review, SBR is a widely used method for indoor mmW channels modeling [
18]. However, most of the simulation-related articles are rarely found to have comprehensive information about the simulation environment. The simulation environment property is partially found in References [
19,
20,
21]. However, simulation parameters like the highest number of reflection values and different obstacle penetration properties are not deeply analyzed when comparing with measurements.
The manuscript is organized as follows. The description of the measurement campaign is presented in the second section. The RT modeling is mentioned in section three. The validation of RT simulation is done by comparing with measurement data and is shown in section four. Lastly, the conclusion and future directions of this paper are stated in the last section.
2. Measurement Details
The measurement of the mmW propagation prediction in 5G networks at 38 GHz is carried out for capturing detailed information. This measurement campaign is conducted to estimate the Line of Sight (LoS) as well as Non-Line of Sight (NLoS) signals. The campaign is done in the NWCC_P15a building ground floor. The interior big rooms are partitioned by using approximately 5 cm thick glass and board. Additionally, some exterior door/windows are used and those are made with wood, plastic, and glass. One base station and 20 mobile stations are used in the measurement campaign. In the NWCC_P15a, mobile stations are placed in several areas to capture the LoS and NLoS signals within 1 to 22.7 m distances. NWCC_P15a height and width were 21 and 3 m. In the measurement campaign, the base station is placed in a fixed location of room 2 in the NWCC_P15a. The measurement is started from a 1-m distance mobile station from the base station. The RSSI and PL are captured in the Rx stationary at that point with respect to the position. Subsequently, the Rx changes location by moving 1 m apart from the TX. For every Rx point, the same process will be re-run until the measurement campaign finishes. An Anritsu MG369xC model signal generator is used to produce continues mmW, which is linked with a horn antenna. For the receiver, an MS2720T model omnidirectional antenna is used to receive frequency. The measurement used hardware configurations, which are given in
Table 1.
In this measurement campaign, a horn antenna is used to transmit a signal at the base station and an omnidirectional antenna is used to receive a signal at the mobile station. For the measurement, some configurable parameters are actively incorporated, which are presented in
Table 2.
The PL is a crucial factor in measuring WCS. PL is the large-scale fading calculation based on 38 GHz frequency depending on signal attenuation in a mobile station placed environment. WCS propagation characteristics are investigated based on empirical, stochastic PL and an empirical method [
22,
23]. Moreover, accurate PL increases the measurement accuracy in WCS [
24,
25]. Generally, PL is calculated in the measurement based on Equation (1).
In this scenario,
is used to express the PL among all TX-Rx distance at the frequency 38 GHz. Additionally,
is used to mean the PL for closing in the (CI) path. The symbol
and
is used for the Zero Mean Gaussian Random Variable (ZMGRV) where standard deviation (SD) is symbolize by
. The sensitivity (XPD) and Cross-Polarization (CP) has an additional effect on the CI PL model during the time of broadcast CP. It is described as the CI path within the XPD (CIX) PL that can be driven by Equation (2) [
26].
The CP factor XPL can be driven by Equation (3).
The term
and
are shown the co-polarization and CP PL, respectively. The XPD is identified using Equation (3), which is an average of cumulative XPL between the total path at 38 GHz frequency of f (see Equation (4)).
The XPD value of Equation (4) is the reverse engineering of Equation (2) and the shadow fading is bring Equation (5).
To calculate the XPD feature, the study incorporated by a modified PL model name is the Frequency Attenuation (FA) PL model. The FA PL can be written by Equation (6).
The term is indicated by PL at the CI path in the meter distance and is the user for 38 GHz frequency. The model defined the is the lowest frequency for the adjustment scenario. The term is the PL exponent at frequency, which is from the CI PL model with respect to vertical-to-vertical and antenna setup. The term express the FA factor, which mean the signal interfere because of reference frequency and is express shadow fading (SF) with an SD of σ.
3. Radio Wave Propagation Modeling by Ray-Tracing Simulation
3.1. SBR Ray Tracing Method
The SBR method is widely used in ray tracing to cover higher frequency radio propagation modeling. The basic SBR method is proposed by H. Ling to detect the “radar cross section of cavities” [
27] and is enhanced to identify the statures of targeted objects [
28]. Currently, SBR is the most renowned method for reflections and scattering modeling in WCS. From the base station, rays are shot in all the directions. As per ray characteristics, launched rays will be reflected from multiple objects before reaching the destination Rx. The Rx is detected based on the radius area sphere with respect to the ray path length and the regulation angle between launched rays. The SBR method is graphically presented in
Figure 1 for the indoor scenario.
The calculation complexity and simulation time of the SBR method directly related to the number of objects in the environment and the RL angles regulations. Since the ray launched in all directions in the simulation area, the ray tracing zone gets wider for higher order of reflection and diffraction. For this wider range, launched rays increase the complexity of the SBR method to implement a huge computational recourse. Moreover, it is good for the crowd indoor environment, but is not suitable for the indoor environment by bearing less numbers of Rx as the number of rays launched are the same.
In
Figure 2, RT has been performed within eight mobile stations and one base station considering both the LoS and NLoS signal using the SBR method.
3.2. Conventional RT Method Limitations
In a conventional RT method, rays are launched in all the possible directions with several angle resolutions without sensing the receiver locations. Therefore, in this circumstance, there is a need to run 3-D RT for every single ray, which is very much time consuming and costly. In the final stage, rays may successfully reach the mobile station sphere or rays may not collide with the receiver spheres. For processing, the high volume of rays, without any pre-assessment, occupied a huge computer resource and had a higher computational time. The fundamental disadvantage of the existing techniques is that the predefined probable zone concept is not incorporated for the receiver. Furthermore, it bounded to RL in all angles through the scenario from TX. For this type of mention, the limitation of the conventional method in simulation time, coverage, PL, RSSI, and RL is very hampering badly.
3.3. Proposed RT Model
In the proposed method, more rays were launched in the predefined zone. Moreover, to increase the accuracy, it does not allow us to permit it to launch rays in all the directions surrounded by the base station.
Proposed 3-D RT method works in several stages. Those are mentioned below.
In phase I, scenario wise 3-D layout design considering the major object with respect to a mobile station and base stations.
In phase II, RL from the base station and incorporate as reflection, refraction, and diffraction as per RT characteristics. Based on layout, used θ = π/60 for the RL angular regulation. This angle is variable. Therefore, a higher angle difference means fewer rays are needed to launch.
In phase III, only some mathematical pre calculations have been performed to sort out the successive angle those rays successfully reached in the destinations.
In phase IV, at least two onward additional directions are added to deliver more rays on the probable zone. The additional angle regulation depends on the scenario such as 0.25, 0.50, 0.75, or 1.0 from the determine angle.
List<double> Probable_VerticalAngles = new List<double> ();
if (Determine_VerticalAngles != NULL)
{
Probable_VerticalAngles.Add(Determine_VerticalAngles + π/240);
Probable_VerticalAngles.Add(Determine_VerticalAngles + π/120);
}
In phase V, again determine the angle wise at least two backward additional directions and add more rays to deliver on the probable zone. The additional angle regulation depends on the scenario such as −0.25, −0.50, −0.75, or −1.0 forward from the determine angle.
if (Determine_VerticalAngles ! = NULL)
{
Probable_VerticalAngles.Add(Determine_VerticalAngles − π/120);
Probable_VerticalAngles.Add(Determine_VerticalAngles − π/240);
}
In phase VI, in this phase, distinctly combine all the probable angles from phase IV and V.
List<double>FinalProbable_VerticalAngles = Probable_VerticalAngles.Distinct(). ToList ();
Lastly, every FinalProbable_VerticalAngles wise shoot rays in the Rx predefine zone.
In phase VII, Trace the launched rays and draw a blue line if the LoS else red line as NLoS. Ray wise calculations are used in the analysis.
Since the ray launched in a predefine zoon in the simulation scenario, the ray launching zone gets narrow with a better level of accuracy in higher order of reflection and diffraction. So, for this predefined zone, launched rays decree the complexity of the proposed method drastically. Moreover, to implement this method with less computational recourse is sufficient for an acceptable output. This method is most suitable for its less computational complexity in the indoor environments including those bearing fewer numbers of mobile stations since, in those scenarios, fewer rays are needed to be launched. For the crowd indoor environment like the indoor stadium, this method suffers computational complexity as it needs to launch rays among the mobile stations in almost all directions.
Figure 3 represents the RT simulation that has been performed using proposed methods among the 1 base station and eight mobile stations. Moreover, in this simulation, all the Rx gain enough LoS and NLoS rays for strong RSSI. It will be mentioned from
Figure 3 that a probable angle wise launched rays have more contribution in mobile stations.
The
Figure 4 represented the Rx3 mobile station from
Figure 3 received multipath signals. The mobile station Rx3 were hit by six LoS and six NLoS ray multi-paths. Those are in the phase so the RSSI will be higher. In conclusion, the proposed method provides strong coverage to the RX, which is a current feature of higher mmW.
3.4. Reflection
The indoor scenario has a larger wall rather than a wavelength of mmW. In this circumstance, calculations of the refection coefficient have been performed using “Fresnel equation” [
29]. According to this mathematical equation, reflection coefficients of both perpendicular and parallel polarizations are expressed by the symbol
and
respectively. The value of
and
are calculated using Equations (7) and (8) with respect for wave impedance.
The incident angle symbolized by
and the refracted angle by
. The impedances of first and second media are expressed by
and
. The impedance assessment depends on obstacle permittivity
. Furthermore,
symbolize the blank or non-magnetic media permeability of the building indoor properties. The conductivity σ presumed to be very few for non-conducting building resources such as concrete, brick, and glass.
Figure 5 shows the graphical presentation of the reflection. The symbol σ means the conductibility, which is estimated to be very low for non-conducting indoor objects such as glass.
Moreover, a refracted dimension of media can be exchange by using “Snell’s law” with respect to the equation by incorporating the angle of refractive and incident.
Equations (9) and (10) are used to calculate coefficients values. Moreover, the parameters and estimate the reflection coefficient. The value depends on the frequency band of mmW and the indoor medium properties.
3.5. Diffraction
Vertical and horizontal diffraction coefficients are calculated by the wedge model [
30]. Diffraction rays RSSI value is much lower than LoS even though the diffraction feature helps to reach the rays in the shadowed zone. Therefore, diffraction is considered as an important factor of WCS. The complexity of diffraction is high to implement in RT because diffraction points start producing a subordinate source to produce several diffracted rays. To overcome the complexity issue, estimate the diffraction coefficient by using the final version of the Luebber model [
30,
31]. As per UTD, point wise field
is calculated using Equation (11) [
32].
The term
expresses amplitude of the starting point, k is used to mean the wave number,
expressed the distance between the TX diffraction source point, the distance between Rx and the diffraction edge is expressed by
, and term
expresses the diffraction. The coefficient of non-conducting objects in every polarization has been calculated using Equation (12).
The term and are used to express reflection coefficients for horizontal and vertical polarization. The incident angles and forward directions are expressed by and backward directions by , and (see Equation (12)).
The graphical schematic views of diffraction with respect to several variables are presented in
Figure 6. The simulation incorporated both side diffractions for each ray.
3.6. Ray Tracing Final Mathematical Equations
In simulations as per RT characteristics, reflections and diffraction are accrued with the ray before the ray reached the mobile stations. The term
expresses the electric field intensity, which can be calculated using Equation (13) [
33].
The incident of the electric field for the first scattering is expressed by
. The symbols
are used to express the number of reflections, transmissions, and diffractions, which occurs sequentially. The symbols
are used to express the associate dyadic reflection, the transmission, and diffraction coefficient consecutively. The symbols
are used to express correlated spreading factors. The whole distance travel by the ray is expressed by S
n. Multichannel modeling mobile stations received several rays. The scenario wise total electric field intensity E
total is the subtotal of each ray, which is driven using Equation (14).
The total number of rays that reached the mobile station is through M. The transmit power
, base station pattern, and
in RT is calculated to estimate
by Equation (15).
The
is the electric field intensity, which is estimated as one meter distance from the base station for antenna gain [
33]. The symbol
means the intrinsic impedance,
means antenna directivity,
and The
express antenna gain and polarization in the direction of
, respectively. The
is the length of
from TX. The symbol
expresses the Rx field strength. The estimation of Vrn as voltage is based on the antenna and polarization of Rx. The Vrn is calculated by using Equation (16).
The symbol
expresses the wavelength.
symbolized Rx antenna gain in the ray angle of arrival,
expresses Rx impedance, the symbol
expresses Rx polarization in the ray angle of arrival, and
is the mean fixed-phase shift accessible by Rx. Lastly, the Rx RSSI is calculated by Equation (17) [
34].
The total valid path is express by M.
3.7. Simulation Parameter Configuration by Compilations of Electrical Properties of Materials
The material electrical properties such as permittivity and conductivity can be hard to find since characteristics are mentioned using different groupings of parameters. The permittivity and conductivity may be quoted with respect to frequencies.
The frequency ranges have limits but are suggestive from the measurements to derive the models. The frequency 1–100 GHz limits must not be exceeded. The Permittivity and Conductivity of the NWCC_P15a environment are given in
Table 3.
Table 3 reported the major materials that need to be incorporated in the simulation per measurement.
4. Validation of Proposed RT Results
The proposed RT method simulation output validation is compared with the actual measurement and is presented in this section. Additionally, the proposed method is compared with the highly used conventional SBR method. The aim of this work is validated with the proposed method by the similar arrangement of measurements with respect to comparison results. The layout design incorporated walls including indoor furniture and more. The simulation has been performed on the layout using SBR and the proposed method by Telecom Malaysia, which included research and development through a funded in house developed software tool. Based on RT characteristics, most of the rays hit within the indoor objects. As rays hit with obstacles, reflection and tracing continued within the reflections limit. The simulation estimates the number of received rays by the receiver, the simulation time, RSSI, and PL of the mobile station [
38]. The receiver sensitivity limit is −150 dB and it is configured in the simulation. The ray line between the Rx and TX will be removed depending on some conditions: the ray dose not reached the receiver spare, the number of reflections reaches the limit, and RSSI is below the receiver sensitivity.
The measurement is conducted at NWCC_P15a. The base on the measurement layout scenario simulations have been done using SBR and proposed RT methods. Mainly, windows transparent windows, doors, and walls are incorporated in the simulation layout. The RT simulation has been done using the standards values of permittivity and conducting regarding the object mentioned in
Table 3. Some issues have been found to estimate the permittivity and conducting of objects such as a different wooden wall, different concrete, and ceiling. In this situation, estimated permittivity and conducting values of obstacles are given in Reference [
39] and are used in the simulation. The base station is placed inside room 2 and several Rx are positioned in several places as per the measurement scenario. The different frequency has a different effect on obstacles and materials in WCS [
40,
41,
42]. In this research, the standard values conductivity and permittivity are used for 38 GHz.
The graphical representation of the NWCC_P15a by the SBR method is presented in
Figure 7. RT visualization in
Figure 7 bear several reflections. Consequently, its PL and propagation times are higher and RSSI is lower. It is also noticeable that the coverage is not good because a few rays contributed in the Rx with most of them bearing higher reflections. If fewer rays contribute in Rx, then RSSI will be lower. Furthermore, Rx received several rays but most of them are because of reflections, which bear PL that reduce the RSSI. By observing the output of the SBR and measurement, we found that RSSI and PL has moderate agreement.
The graphical representation of the NWCC_P15a by the proposed method is presented in
Figure 8. Since RT visualization in
Figure 8 bear less reflections, its PL and propagation times are lower and RSSI is higher. The coverage is good because a large number of rays contributed in the Rx. Furthermore, Rx received several rays, which helped increase the RSSI. By observing the output of the proposed method and the measurement, RSSI and PL had considerable agreement.
The RSSI Standard Deviation(SD) of the SBR method with the measurement is 1.79 for the NWCC_P15a scenario. The RSSI lowest and highest dissimilarity of the SBR method with measurements are 0.6 and 7.99 dBm, respectively. The highest dissimilarity RSSI is found to have an obstacle or the received rays bear several reflected rays. Correspondingly, the RSSI SD of the proposed method with the measurement is 1.46 for the same scenario. The SD value of the simulation with the measurement 1.46 is expressing the proposed method acceptable for RT modeling. The RSSI lowest and highest dissimilarity of the proposed method are 0.27 and 5.58 dBm, respectively. The highest dissimilarity RSSI is found due to an obstacle. Nevertheless, in the proposed method even in the worst-case scenario, the RSSI dissimilarity is considerable between the measurement and the proposed method simulation.
Figure 9 represents the comparison line graph of RSSI for the measurement, the SBR, and the proposed methods data. For this research, the measurement RSSI data are taken into account as standard data. The SBR method’s mobile stations Rx62, Rx69, Rx59, Rx68, and Rx71 shows the moderate relationship trend in RSSI with the measurement, respectively. The RSSI SD of SBR method and standard data is 0.86. Reversely, The SBR method’s mobile stations Rx66, Rx53, Rx26, Rx57, Rx18, Rx67, Rx61, Rx29, Rx60, Rx63, Rx30, Rx58, Rx22, Rx24, and Rx70 shows the distance relationship trend with the measurement, respectively. The RSSI SD of the SBR method with standard data is 1.38. Similarly, for the proposed method, mobile stations Rx71, Rx69, Rx59, Rx29, Rx57, Rx53, Rx67, Rx60, Rx63, Rx68, Rx26, Rx66, Rx61, and Rx70 show the good relationship trend in RSSI with the measurement. The RSSI SD of the proposed method with standard data is 0.75. Reversely, the proposed method’s mobile stations Rx22, Rx58, Rx18, Rx30, and Rx24 show the moderate relationship trend with the measurement. The RSSI SD of the proposed method with standard data is 0.83. Moreover, the proposed method maximum number of mobile stations except Rx24 shows very good agreement with standard data. With the comparative analysis,
Figure 7,
Figure 8 and
Figure 9 for RSSI founded that the proposed method represented better agreement with the measurement when compared to the SBR method.
To reach the receiver, the ray has to travel some distance. Therefore, it bears PL, which reduce the RSSI. In WCS, PL has great impact in terms of modeling.
The PL SD of the SBR method with the measurement is 1.96 for the NWCC_P15a scenario. The PL lowest and highest dissimilarity of the SBR method with measurements are 0.71 and 7.99 dB, respectively. The highest dissimilarity PL is found to quickly appear or the received rays bear several reflected rays.
Correspondingly, the PL SD of the proposed method is 1.74 for the same scenario. The SD value 1.74 of the simulation is expressing the proposed method acceptable for RT modeling. The PL lowest and highest dissimilarity of the proposed are 0.32 and 6.69 dB, respectively. The highest dissimilarity PL is found due to an obstacle. Nevertheless, in the proposed method even in the worst-case scenario, the PL dissimilarity is considerable between the measurement and the proposed method simulation.
Figure 10 represented the comparison line graph of PL for measurement, SBR, and proposed methods data. For this research, the measurement PL data takes into account the standard data. The SBR method’s mobile stations Rx61, Rx63, Rx66, Rx26, Rx67, Rx29, and Rx57 show the moderate relationship trend in PL with measurements. The PL SD of SBR method with standard data is 0.62. Reversely, The SBR method’s mobile stations Rx22, Rx68, Rx24, Rx30, Rx70, Rx58, Rx53, Rx59, Rx69, Rx60, Rx62, Rx71, and Rx18 show the distance relationship trend with measurements, respectively. The PL SD of the SBR method with standard data is 1.26. Similarly, for the proposed method, mobile stations Rx60, Rx59, Rx66, Rx26, Rx67, Rx63, Rx57, Rx53, Rx29, Rx22, Rx24, Rx58, Rx30, and Rx70 show the good relationship trend in PL with the measurement, respectively. The RSSI SD of the proposed method with standard data is 0.71. Reversely, the proposed method’s mobile stations Rx69, Rx71, Rx68, Rx62, Rx18, and Rx61 show the moderate relationship trend in PL with the measurement. The PL SD of the proposed method with standard data is 1.35. Moreover, the proposed method maximum number of mobile stations except Rx61 shows very good agreement with standard data. With respect to the good similarity of simulation by using PL SD with standard data less than 1: seven numbers of Rx are found in SBR and fourteen numbers of Rx are found in the proposed method. On the other hand, fewer similarity of simulation by using PL SD with standard data greater than 1: thirteen numbers of Rx are found in SBR, and six Rx are found in the proposed method. With the comparative analysis in
Figure 7,
Figure 8, and
Figure 10 for PL, the proposed method represented better agreement with the measurement than the SBR method.
Figure 11 represented the line graph regarding the receiver and the number of contributed rays in the SBR and the proposed method. This type of data is not available in the measurement, so a comparison has been performed between the SBR and the proposed method data. The mobile station Rx61 received the maximum number of rays in both simulations. For the case of the SBR method, this receiver received 44 rays, which is similar to the proposed method in which this receiver received 63 rays. As per the overview from
Figure 11, in the proposed method, a higher number of rays reached the mobile station rather than the SBR method.
Figure 12 represented the line graph of the mobile station and the simulation time for both the SBR and proposed methods. The SBR method bears high computational time when compared to the proposed method. This simulation has been run using high configuration with a good graphics server. In a normal computer, this simulation time will be much higher than in this scenario.
In the proposed method, the ray-launching phase has significant contribution in the simulation. Consequently, RL is the first phase of RT [
43]. Proper utilization of RL makes the proposed method more capable, suitable, and intelligent [
44]. Few ray launchings are done in the proposed method rather than SBR. To get sufficient coverage using few numbers of launched rays in the predefined potential region is the outstanding contribution of the proposed method. Lastly, the proposed method has a very significant contribution in the WCS base on the coverage, RL, simulation time, PL, and RSSI.