Characterization of a 60 GHz scattered wireless channel with different antenna polarizations for underground multimedia applications

Natural Sciences and Engineering Research Council of Canada, Grant/Award Number: RDCPJ 460985 ‐ 13 Abstract Large scale (i.e. <10 m) and small scale (i.e. 1 � 1� 1 cm) measurements and characterization of a scattered wireless channel with different antenna polarizations and configurations in the underground mine galleries at 60 GHz are addressed. Results show that the rough surface scattering and the gallery dimensions affect the path loss (PL) exponent and it becomes smaller than the free space. Vertically polarized antennas give a lower value of the PL exponent and root mean square (RMS) delay spread compared to the horizontal one. The small scale 3Dmeasurement results show that the power loss of around 1–5 dBwithin a small scale cubical area. Results also show that the channel is less time dispersive in a wider gallery and observed a higher value of the RMS delay spread compared to a smaller gallery. The statistical results of the small scale multipath amplitude fading provide a better fit with the Rician distribution. The effects such as scattering, polarization, antenna radiation patterns, and waveguide which caused increase and decrease of the value of PL exponent and delay spread are also analysed. Results revealed that a directional narrow beam dual‐ polarized antenna configuration might be a good candidate in this environment.

aerial drones could be a solution for a faster wireless safety system. Those solutions could be implemented at the lower frequency bands such as 2.4, 5.8 GHz, however, those systems may not be enough to provide a high quality of multimedia services to a dense network in a larger mine gallery length due to the lower bandwidth and higher interference. Triband (2.4, 5.8, and 60 GHz) solution, however, could provide a wireless solution to fulfil the mining industry requirements [4,5]. The industry requirements such as the optimum attenuation, lower interference, low latency, and high accuracy of geolocalization in a dense wireless network are highlighted herein to design a wireless system for the underground mining environment.
The 60 GHz band, which has an oxygen absorption feature in the propagation phenomenon and additional features such as low interference, unlicensed spectrum, larger bandwidth, smaller package size, and beamforming capability can possibly fulfil many of the mining industry's technical requirements. Therefore, the 60 GHz bands could be used for the small cell backhaul point to point link for the applications of video monitoring, remote control of vehicles, three-dimensional (3D) mapping, and creating virtual environments, cooperative robotic systems for remote operation of mineral extraction etc. However, the 60 GHz bands may suffer from dispersion, absorption, scattering on the wireless channel because of short wavelength, and the complex structure of the mine surface.
For more than a decade, several measurement campaigns were carried out for the wireless channel characterization in an experimental mine called Canada Centre for Mineral and Energy Technology (CANMET), by the Laboratoire deRecherche Télébec en Communications Souterraines (LRTCS), located in Val-d'Or, QC, Canada. Some of the experimental results are listed in Table 1. Comparative results show that as the operating frequency increases and the antenna half-power beamwidth (HPBW) decreases, a lower value of RMS delay spread is obtained. The values of the path loss (PL) exponent are found to be around 2. The results revealed that the PL exponent and the delay spread particularly depend on the antenna HPBW, topology of the experiment, gallery curvature, and operating wavelength as well as the transmitter receiver separation distance (Tx-Rx). Measurements with a heavy vehicle between inside and outside of the CANMET mine gallery have also been carried out in a frequency range between 2 and 6 GHz [17]. The results showed that the PL exponent is found to be approximately 2.23 and 1.96 when a scoop vehicle was inside and outside of the gallery, respectively. The average values of RMS delay spread of the inside and outside gallery were also found to be around 5.02 and 3.85 ns, respectively.
Other measurements carried out in different mines and tunnels in different frequencies are reported in [18][19][20][21]. Emslie et al. [19] performed a measurement in a coal mine at 200-4000 MHz and noted that the increase of a signal power loss is caused by the surface roughness and the tilt of the tunnel walls. At 900 MHz and 1.8 GHz, Zhang et al. [20] reported that the RMS delay spread is less than 25 ns for a straight empty subway tunnel and increased to 103 ns when vehicles were presented.
However, measurements carried out at 60 GHz bands in indoor environments are reported in [22][23][24][25][26][27][28]. Smulders et al. [22,23] reported a PL exponent of less than 2 with the signal fading between 0.14 dB and 2.14 dB and an RMS delay spread of 1 ns with a fan-pen antenna configuration. Channel fading and dispersion effects in an indoor environment were also well studied by Geng [24], where the obtained PL exponent was found to be less than the free space value due to the waveguide effect.
As well, 60 GHz measurement results with different antenna polarizations showed that the use of a circular polarization of an antenna can outperform than the use of a linear polarization in terms of bit error rate in the library, hallway, and residential environments [25]. Results also showed that the number of strong reflected waves depends on the types of polarization such as horizontal and vertical [26]. Moreover, polarization diversity and dual polarization transmission systems may suffer from the antenna polarization mismatch. The angle of arrivals (AoAs) of the strong reflected waves may also depend on the polarization types due to the variation in the value of reflection coefficients [26]. Manabe et al. [27] found that the RMS delay spread value with horizontal (H) polarization is slightly lower than with vertical (V) polarization. Zhao et al. [28] noted that the vertical polarization of a 60 GHz transceiver resulted in a higher received power than with the horizontal polarization in the shadowing region. The IEEE 802. 15.3c task group also demonstrated that the mismatch of polarization characteristics between transmitting and receiving antennas at 60 GHz could result in 10-20 dB received power degradation [29].
In the analysis of the polarization effect in underground mines and tunnels, Cawley et al. [30] found that different polarizations have different coupling losses, attenuations, and bend losses. Zhang et al. [31] also described the received power losses with different polarization types at 900 MHz. Bashir et al. [32] also found that the PL is more sensitive to antenna position and polarization than the time dispersion in underground mines and tunnels over 2.4-4 GHz. In contrast, the antenna polarizations have a significant impact on the channel characterizations at lower frequency bands. Therefore, at higher frequencies such as 60 GHz, this phenomenon might have more effect on the channel. Previously, the 60 GHz measurements and analysis in the CANMET mine were carried out in different aspects such as identifying the PL exponent and delay spread values with single polarization, investigating the link reliability based on the antenna diversity gain, implementing the deterministic ray-tracing modelling using reflection and diffraction phenomenon, and the scattering analysis reported in [11][12][13][33][34][35][36]. Hence we take into account different antenna polarizations in the present analysis of the underground mine wireless channel. Consequently, the polarization and scattering effects are necessary to be addressed in the underground mine channel model but are not reported in the open literature on this subject so far. Some parts of this article are archived in the first author's Ph.D. thesis reported in [35]. Some results and analysis of this article are summarized from the author's studies reported in [36][37][38][39]. In this article, the measurements and characterization are examined by considering the facts of polarization, scattering, waveguide, and antenna radiation patterns dealing with the different magnitudes of the floor, wall, and ceiling roughness in different gallery dimensions.

| MEASUREMENT SETUP AND EXPERIMENTAL PROTOCOL
A 60 GHz frequency domain measurement setup was used to perform the experiment in two measurement depth zones (i.e. 40 and 70 m levels) of the CANMET mine and is shown in

| Underground mine environment
The dimensions of the 40 m level are approximately 5 m in both height and width. From the surface roughness measurements at the 40 m gallery, the maximum, mean, and standard deviation of the roughness of the side walls were found to be around 37, 20, and 6 cm, respectively [37]. The 70 m gallery, which was narrow, had a height of around 2-2.5 m and a width of 3 m. The 70-m gallery ceiling was mostly covered with metallic nets. Some metallic pipes were present in the corner of that gallery. The floor was mostly covered with water puddles at 70 m, whereas it was drier and dustier at 40 m. The humidity in both galleries was approximately 100%. The temperature was around between 6°C and 7°C and a lower temperature was observed at 70 m compared to 40 m. There were large amounts of dust in the air at 70 m but such a large amount did not exist at 40 m gallery. A loud noise caused by the air ventilation system was also observed in both galleries. This kind of mine environment may cause a significant impact on the radio channel particularly at millimetre wavelength. Hence, the analysis of the scattering effect is described herein.

| Scattering effect
Scattering may create specular and diffuse reflections during the interaction between an electromagnetic (EM) wave and a rough surface which may lead to a lower or higher value of the received power at the receiver. Objects larger than the wavelength cause reflections, sharp rock edges may cause diffractions, and scattering happens when an object dimension is smaller compared to the incident wavelength. During scattering, the incident energy may distribute in coherent and incoherent ways in all angular directions.
Any object of the rough mine surface larger than 5 mm can be considered as a flat surface. Therefore, within a small scale area (i.e. 1 cm 2 ), an object could be larger than λ or lower than the λ or larger than λ with sharp edges. As well, the surface roughness is irregular and having higher inclinations. Moreover, the galleries consist of metallic pipes, nets, arches and water puddles. Consequently, three propagation phenomena could function randomly. Thus, a single and multiple bounced reflections, diffractions, and scattering could occur during the interaction between an EM wave and the mine surface [35]. If the scattered wave with a strong specular reflection is constructively added at the receiver, then the Rx power will increase. This phenomenon depends on the phase of the signal that could change randomly due to the diffuse scattering phenomenon.
A strong reflection in a specular direction was observed during the scattering in the galleries reported in [37]. The results show that the incident power is scattered within a range of about 10°along the specular direction of the reflected wave. Moreover, the PL difference between the reflected and the line of sight (LOS) signals is found to be around 19 dB reported in [39] where the surface material absorption and incident power dissipation may occur along with the propagation loss [40,41]. Thus, the channel scattering and the constructive addition of the scattered paths along with the specular reflected path may increase the Rx power thereby leading to a lower value of n compared to the lower frequency bands such as 2.4 GHz.

.41
Note: n is the PL exponent; τ rms is the mean RMS delay spread; A tx and A rx are the transmitter (Tx) and receiver (Rx) antenna types, respectively; A tg and A rg are the Tx and Rx antenna gains in dBi, respectively; Tx-Rx is the approximate maximum transmitter receiver separation distance; and fc and BW are the centre frequency and bandwidth, respectively.
Abbreviation: CANMET, Canada Centre for Mineral and Energy Technology; PL, path loss; RMS, root mean square.

| Measurement setup
The setup was designed and implemented in the CANMET underground mine in cooperation with the LRTCS laboratory. An illustration of the measurement setup is shown in Figure 2. A vector network analyzer (VNA) with a frequency band range of 40 MHz to 70 GHz was used as a two-port network measurement system for transmission and reception of the radio signal.
The number of frequency points of 2000 was used for the measurement. The power amplifier was used with a gain of 30 dB. Since a high loss (almost 36.8 dB) in the 4.6 m cable was observed, even with a low noise amplifier (LNA) with 30 dB gain, the level of the signal was not strong enough to have a reflected path, therefore a second LNA was added with a gain of 18 dB to enhance the signal power. The frequency range was then selected, following the IEEE Standard 802.15.3c, between 57.24 GHz and 59.4 GHz (channel 1). Other channels in this standard were also taken into account during the measurements, but high attenuation of the radio signal was observed due to the 61 GHz upper frequency range of the QuinStar LNA. The transmit power was set at 4 dBm for the measurements. Horn and omnidirectional antennas were used to make directional and omnidirectional propagation scenarios measurements between the Tx and Rx. The Horn antenna radiation patterns were measured in the anechoic chamber as shown in Figure 3. Results showed an HPBW of about 12°in both azimuth and elevation planes. Since the horn antenna was pyramidal, the EM field radiation was mostly directive in the HH (Tx as horizontal and Rx as horizontal) polarization, whereas some strong ripples in the side lobes were observed in the VV (Tx as vertical and Rx as vertical) polarization. The vertically polarized omnidirectional antenna had an HPBW of 360°in azimuth and around 40°in the elevation direction. In order to find the polarization effect, the horizontal polarization of the omnidirectional antenna was used to observe any influence between horizontal and vertical polarizations in correlation to the communication links between the moving miners. A similar experimental setup was used for all antenna configurations. The polarization of the antenna was changed from vertical to horizontal and vice versa, manually using a shift of 90°. A laser and a camera tripod were used in order to have an accurate placement of the Tx and Rx in LOS condition. Since the channel considered between two miners body to body link, the Tx and Rx were positioned in the middle of a gallery with a width and height of approximately 1.5 m. A list of the measurement parameters is shown in Table 2. The data acquisition was completed by connecting a computer to a VNA via a general purpose interface bus interface. A Lab-VIEW programme was used to control the measurement procedure and a MATLAB programme was used to move the VELMEX 3D positioning table.

| Measurement protocol
A VNA with 1 kHz intermediate frequency bandwidth (IFBW) allowed a single channel (S 21 ) full sweep time of around 6 s for 2000 points, which provided a stable state of the frequency response of the channel. The ANRITSU VNA noise floor was −107 dBm with 10 Hz of IFBW. Note that, to have the smooth operation of the measured samples within one full sweep, the IF bandwidth can be increased, as a result, the noise floor level and the measurement time will increase accordingly. Therefore, with 1000 Hz of IFBW, the VNA noise floor was −87 dBm (by increasing the IFBW to a factor-of-10 where the noise floor increases by 10 dB). Consequently, the dynamic range (i.e. the difference between the maximum output power and the specified noise floor) of the VNA was around 87 dB. A SOLT (short-open-load-thru) calibration with a 2.16 GHz bandwidth and 2000 sweep points spaced by 1.08 MHz was used [3].
The antenna's far-field distance (d f ) from the Tx antenna aperture was calculated using the equation as follows: where D is the physical dimension (diametre) of the antenna and λ is the wavelength. The d f of the horn and omnidirectional antennas was found to be 0.83 m and 0.58 m, respectively. Therefore, for the postprocessing of the channel impulse responses, 1 m measurements were used for each configuration. Two fixed references such as a 2 m long plane of wood on the floor and a rope outside the channel were used in order to ensure the maximum accuracy of LOS. Significant time was spent to place two pieces of wood on the ground floor, which were used as rails to move the trolley (where the Velmex3D

| Large scale
In the large scale case, 32 measurement locations were considered with a separation of 40 λ (i.e. 20 cm), covering 1-7.2 m, as shown in Figure 4. This separation distance can be of any arbitrary value greater than λ. At each location, three consecutive measurements were taken, separated by λ/2 (2.5 mm) to have a local average of the channel which eliminates the spatial amplitude fading. Elimination of fading requires a separation distance of less than λ. The measurement configurations are listed in Table 3. For each configuration, 96 frequency responses were recorded for the postprocessing. The Tx position remained fixed and the Rx position moved by a distance of λ/2 and 40λ using the Velmex 3D positioning table.

| Small scale
In the small scale case, the study of the 3D channel characteristics was considered in order to compensate the antenna misalignment (caused by the roughness of the floor) and to eliminate the effect of spatial amplitude fading. The measurement procedure is shown in Figure 5. Two measurement locations such as at 1 m and 3 m were considered. The measurement configurations are listed in Table 4. The Velmex stepping controller was used to move the Rx in x, y, and z directions. It was observed that the floor of the galleries was inclined largely within a 1 m 2 area, particularly at 40 m. Although the LOS is aligned in the x direction, the alignments in the y and z directions were also required in order to get the maximum power between the Tx and Rx antennas. This should be noted that it was difficult to know at which point the maximum power of Rx was reached due to the rough surface. Therefore, at each distance, a 5 � 5 � 5 grid points separated by a distance of λ/2 was considered for the measurement. As well, Tx and Rx were aligned at the (λ, λ, λ) grid point and the 5 � 5 � 5 grid points were recorded following a square wave trajectory (defined by the Velmex) as illustrated in Figure 5. For each antenna configuration, 250 frequency responses were recorded for the postprocessing where 125 responses were taken at each distance.

| Post processing
To obtain a power delay profile (PDP) of each Rx position, at first, an average of 15 snapshots of the frequency response was performed. Second, to remove the antenna effect and the cable loss, over the air calibration (i.e. a division between the 1 m complex frequency response and all complex frequency responses) was performed. Finally, an inverse fast fourier transform (IFFT) of the calibrated frequency responses was performed and the normalized PDPs were obtained. A multipath detection process was performed by the procedure described in [15]. The 1.08 MHz sweep frequency gave a temporal range of 925 ns (corresponding to the inverse of the sweep frequency), which was sufficient to observe the largest multipath excess delay (i.e. approximately 30 ns) through all sets of measurements. Figure 6 shows an example of the postprocessed normalized PDP. The strongest peak of the PDP corresponds to the first observable path. As expected, the first direct path corresponds to a delay of about 9.7 ns at 3 m Tx-Rx distance. A decrement of multipath component powers with various propagation delays is seen in Figure 6. The threshold value for the multipath detection process was set to −25 dB. The dynamic range of the measurement was around 40 dB.

| LARGE SCALE (i.e. <10 m) CHARACTERIZATION
Results of the large scale characterization have been analyzed by PL exponent and signal fading. The PL at a distance d is calculated by adding all incoming multipath. The free-space PL at a reference distance of 1 m (i.e., d 0 ) with 58.32 GHz centre frequency is set to 67.76 dB [42][43][44] as follows: PLðdBÞ ¼ 67:76 ðdBÞ þ 10log 10 1 where K and L are the total number of consecutive measurement points along the x direction and the total number of multipath, respectively. a i and τ i are the i th path amplitude and delay at j th location, respectively. By computing the deviation of the PL in dB with respect to the linear regression line, the value of the signal fading was estimated. The values of the deviations were then fitted with a zero mean normal distribution to obtain the fading parameter.
The average PL for any arbitrary distance between Tx and Rx can be modelled by a lognormal (normal in dB) random distribution [42]: free space PL at reference distance d 0 and S σ is a fading parameter with a zero mean Gaussian random variable in dB with a standard deviation of σ dB .

| Path loss
An example of a scatter plot of the PL according to the Tx-Rx distance with different antenna polarizations is shown in Figure 7. The values obtained are summarized in Table 5. The linear least square regression method was used to find the value of the PL exponent n and the fading parameter σ dB . The PL exponent values obtained are close to the value found in free space (i.e. n = 2). The results reported in [23,[45][46][47][48][49] also show that the PL exponents are ranging between 1.4 and 1.8 in different kinds of indoor and underground mine environments at 60 GHz. The obtained results can be explained due to the facts of scattering, polarization, antenna radiation patterns, and waveguiding effects. Those effects are explained later in section 5. Particularly, a differential value of n between VV and HH polarization obtained may be the cause of the different antenna radiation patterns. Due to the different roughness magnitudes of the wall, ceiling, and floor (which consist of higher inclination and sharp edges), the coherent and incoherent scattering phenomena may occur in the channel. As shown in Figure 3, the horn-horn antenna configuration may provide the ) at the VV receiver resulting in a higher number of multipath when compared to the HH one (i.e. 40°). Therefore, the VV polarization provides more multipath components into that experimental topology, correspondingly, the constructive addition of the multipath at the receiver leads to a lower value of PL exponent compared to the HH polarization.

| Interference fading
Interference fading can be characterized by increasing and decreasing the signal power losses when the Rx is driven at an LOS direction in the presence of the fixed objects (e.g. rough walls, metallic pipes, nets, ventilation systems, water puddles etc.) beside a propagation path. Sharp-edged rocks that are larger than the wavelength (5 mm) can increase the effects of large scale radio signal fading due to the reflection, diffraction, and scattering, in which the signal power variations along a wireless link can be characterized by the statistical distributions. In order to quantify this, the deviations of the PL values from the linear regression line were calculated and the standard deviation of the variations (σ dB ) was estimated. The obtained results of σ dB are summarized in Table 5. Results show that once the distance increases the value of the fading increases a bit. The value of the fading is not severe with the 60 GHz bands overall, compared to the 2.4 GHz bands, due to the use of the lower value of HPBW of antennas. An example of the cumulative distribution function (CDF) of the amplitude fading between the fitted and the measured data is plotted in Figure 8. Large scale fading in underground mines is usually found to be normally distributed [16] and was seen to all of our configurations.

| Delay spread
The mean and standard deviation of the excess delay (τ), RMS delay spread (τ rms ), and the maximum excess delay (τ max ) of different antenna configurations from the large scale measurements (between 1 m and 7.2 m distance) were computed from the PDPs of each of the grid points followed by the Equation (5) and summarized in Table 5. Results show that the use of vertical polarization provides a higher value of τ rms than the horizontal ones. Moreover, a higher value of τ rms and τ max is observed in a wider gallery. Ben et al. [11] noted that the RMS delay spread is approximately 7 ns at 60 GHz in CANMET mine. Since this experiment was with MIMO channels with 60°HPBW of antennas in a different environmental topology, a higher number of multipath may arrive at the Rx with higher amplitudes and delays. On the other hand, an RMS delay spread of  around 2 ns by using a horn antenna of 16°HPBW in the same mine is reported in [13]. The difference between the obtained results herein is due to the value of HPBW, and the experimental topology. Hence, the time dispersion results of the polarimetric measurements depend on a specific topology and antenna radiation patterns as well as the factors described later in the section 5.
Therefore, for a position d x,y,z , the averaged measured channel impulse response can be calculated as follows: where M is the number of snapshots taken at each time instance t, τ is the path delay, and f = 1, 2,…, 2000. The PDP can be defined as follows [42,50]: where N is the number of multipath components, a i and τ i are the i th path amplitude and delay, respectively. The time dispersion parameters of the channel such as the mean excess spread (τ), the maximum excess delay (τ max ), and the RMS delay spread (τ rms ) have been extracted from the PDPs followed by the Equation (5) [42,51]. The τ rms can be expressed as: where E d {⋅} is defined as the mean value of τ rms collected from every grid points. The τ rms (d x,y,z ) at each point is mathematically defined as: where τ is the mean excess delay and its first and second moments (n = 1, 2) expression is given below: where N is the total number of paths, τ i and a i are the i th path delay and amplitude, respectively.

| Fading
The normalized multipath total power (e.g. the power summation of all multipath components) of each 3D grid points was extracted from the PDPs. An example of the spatial distribution of multipath total power of 2D grid points is shown in Figure 9. The signal power was mostly distributed throughout the grid points at 70 m compared to the 40 m gallery, caused by the narrow passage gallery. At 40 m, the power was mostly directive and a lower signal power fluctuation was observed due to the larger gallery dimension. The average small scale fading ranges were found to be from 1 to 2 dB and 3 to 5 dB for horn-horn, and horn-omni configurations, respectively. This implies that the Rx may experience around an additional 1-5 dB power loss when it moves by a particular cubical area (i.e. 1 cm � 1 cm � 1 cm). Therefore, the Rx displacement of λ/2 was not found to have a significant effect at 60 GHz in the mine environment.
To find the statistical characteristics of the amplitude fading, different known distributions were compared. The Kolmogorov-Smirnov test (KST) was used to identify which distribution matched best with the experimental results corresponding to the lower value of the KST score. Figure 10 shows an example of the measured CDF of multipath amplitude fading, which fits with the Rician distribution. The distribution of small scale multipath amplitudes of the LOS channel can be modelled as a Rician distribution which characterizes a stronger LOS path and scattered multipath. A comparative study of all configurations showed that Rician distribution was matched, since the Tx was directive horn antenna in the LOS direction, and a strong dominant direct path and scattered multipath arrived at the Rx [52].

| Delay spread
The time dispersion parameters extracted from the 3D grid points are summarized in Table 6. For H-H configuration, τ rms is higher in the 40 m gallery than at 70 m gallery, because the 40 m gallery is wider and, hence, the path delays are higher. For the 70 m H−O V V case, a higher number of multipath components is observed compared to all configurations that provide a higher value of τ rms . This is due to the use of the omnidirectional antenna at the Rx, the presence of narrower gallery dimension, and metallic pipes on the channel beside the LOS direction, which may add a higher number of multipath with higher value of amplitudes at the Rx. The empirical CDF of the τ rms values is illustrated in Figure 11. It seems that the standard deviation of τ rms does not vary significantly and overall τ rms lies between 0 ns and 4 ns.
Overall, the large-and small-scale results for both galleries show that the horn-omni and omni-horn configurations provide a higher time dispersion of the channel compared to the horn-horn configuration due to the use of the omnidirectional antenna which provides a higher fluctuation of multipath amplitudes. Around two to three multipath components were observed in horn-horn configuration versus around four to five in horn-omni or omni-horn scenarios [35]. Moreover, with a horn-horn configuration, a higher time disperse channel was observed at 70 m compared to 40 m due to the narrow gallery dimension with different magnitudes of the roughness. The directional antenna in both Tx and Rx has a major advantage to reduce the number of multipath components in the underground mine gallery and consequently maximizes the data transfer rate. Large scale and small scale results show that the signal undergoes a frequency selective fading due to the smaller value of the coherence bandwidth compared to the signal bandwidth, meaning an equalizer (adaptive tapped delay filter) will be needed at the Rx [42]. The obtained values in both galleries imply that the horn-horn antenna configuration may fulfil the mining industry requirements such as precise geolocalization and high speed connectivity due to the antenna directivity and lower signal power degradation. In addition, the use of both polarizations in the wireless system may help to enhance the system capacity. Omnidirectional antennas are not very effective at 60 GHz because of the lower directivity and gain compared to the directional antennas, but they might be appropriate for the wider signal coverage.

| COMPARISONS AND EFFECTS
The difference between this experiment and others performed in the CANMET underground mine is listed in Table 7. The wireless scattered channel measurements with different antenna polarizations provide an explanation to determine whether a dual or single polarization would be best suited, as well as which polarization type in underground mine. Different antenna configurations may contribute to the choice of using a directional or omnidirectional antenna in underground mine galleries. A combination of gallery depths, antenna configurations, and polarizations, as well as 3D small scale measurements, provides more channel characteristics information in terms of scattering effect compared to the other works. Consequently, some effects which caused the increment and decrement of the value of n and τ are identified. These are summarized below in order to get into depth of the scattered channel characterization.

| Waveguide effect
The waveguide effect creates a constructive addition of the multipath after having multiple bounced reflections in a tunnel, corridor which phenomenon leads to a higher value of the received power at the receiver. The waveguide effect defines where a direct path and a slightly delayed path (or paths) from the surface are combined at the receiver in such a way where the waves are being guided along a particular direction. Generally, this effect occurred at a larger Tx-Rx distance where the reflects waves can travel along the propagation direction after having multiple bounces from the surrounding surface.
This effect is seen in CANMET underground mine galleries at larger Tx-Rx distance (i.e. >10 m) and at lower frequency bands (i.e. 2.4 GHz) [3,[6][7][8]. Results of the angular dispersion measurements in CANMET at 60 GHz also show Note: τ rms , τ, and τ max are in nanosecond (ns). HPBW is the half power antenna beamwidth in azimuth plane.

-
that the angular spreads of multipath are proportional to the gallery dimensions and inversely proportional to the Tx-Rx distances [36]. The results revealed that no reflected paths were received at the receiver after around 4 m. Therefore, this waveguide effect might not be the only explanation for the results obtained herein at a 10 m Tx-Rx distance. At this short distance, the specular reflections may travel until one or two bounces. Since the surface is highly rough, the 5 mm wavelength may additionally create multiple bounces of reflections due to the scattering phenomenon. Consequently, the received power may increase due to the constructive addition of the scattered multipath which leads to having a lower value of n compared to the free space. The results also show that the waveguide phenomenon is more in pronounced in the narrower gallery which also implies a lower value of n compared to the wider one.

| Polarization effect
The polarization effect may introduce a differential value between V and H polarized antenna received power. Further, a single polarized wave may change its polarization states during the interaction with the rough surface.
The horizontal and vertical polarization orientations of an antenna can be used to measure the Fresnel reflection coefficient parameters denoted as |Γ ⊥ | and |Γ ∥ |, respectively [39,53]. The results show that the values of the reflection coefficients of limestone, glass, and brick materials are different at 1.9 and 4 GHz [54]. Therefore, the value of the reflection coefficients depends on the electric field orientations, the plane of the incident, operating frequency, and dielectric properties of the materials.
At 60 GHz, results show that the value of reflection coefficient varies severely as the incident angle varies [53]. As well, the AoAs of the reflected waves depend on the value of reflection coefficients as reported in [26]. Results reported in [27,29] also show that the polarization types have a significant impact on the 60 GHz channel. This kind of inhomogeneous, irregular rough surface of the galleries may consist of materials such as rocks, sand, minerals etc. where those can be reflective and dispersive at 60 GHz. Therefore, the V and H polarized antennas with different values of roughness of the floor, wall, and ceiling may provide a differential value of |Γ ⊥ | and |Γ ∥ |.
In addition, the incident wave could change the polarization types during the interaction between the EM wave and the rough surface. Consequently, the new polarized signal may not receive at the single previously polarized antenna state at the receiver. This phenomenon may provide a lower value of the multipath total power which may lead to have a lower value of τ. However, the cross-polarization experiment could provide a more in-depth analysis of this effect in the future.

| Antenna beam pattern effect
Different antenna beam patterns may increase and decrease the number of multipath at the receiver due to the antenna HPBMs and side lobes. The vertical configuration of the horn antenna beam pattern has strong ripples (i.e. around 20 dB compared to the horizontal) in the side lobes between ±20°a nd ±60°as shown in Figure 3. This kind of beam pattern may provide a higher value of the multipath total power at the receiver. Consequently, a constructive addition of the multipath provides a higher value of τ which in turn leads to having a lower value of n compared to the horizontal configuration. This effect is also observed in other measurements as reported in Table 1 which shows that a higher value of HPBW causes a higher number of multipaths at the receiver which also leads to a higher value of the delay spread.

| CONCLUSIONS
The results and analysis of the 60 GHz channel measurements for underground mine applications are presented in this work. The channel characteristics with different polarizations and different types of antennas in different gallery depths are highlighted. The comparative study of the large scale (i.e.  <10 m) results show that the PL exponent is close to one of the free-spaces because of the constructive combination of the multipath components at the Rx where the channel is impacted by the scattering, polarization, antenna beam pattern, and waveguide effects. The results also show that the RMS delay spread lies between 0 and 4 ns. A 3D small scale (i.e. 1 cm � 1 cm � 1 cm) measurement has been carried out for compensating the antenna misalignment issues where the amplitude fading lies between 1 dB and 5 dB within a small scale area. The channel presents less time dispersion in wider gallery compared to the narrower one. The Rician distribution provides a better fit with the results of the small scale multipath total power.
The use of vertical polarization on directional and omnidirectional antennas has provided a higher number of multipaths at the Rx implying a higher value of the delay spread when compared to the horizontal one in this particular experimental topology. The irregular rough surface, which generates scattering, has a direct impact on the amplitude, phase, and delay of the channel severely compared to the lower frequency bands. In addition, polarization, antenna beam pattern, and waveguide effects may create a significant change in the channel characteristics at 60 GHz bands. In order to deal with the scattering, the system designer may define a lower antenna beam resolution to have a highly directive propagation links. To compensate with the polarization effect, the use of dual polarization at the antenna would be a good choice. Therefore, by taking into account the beamforming feature deployment in the point to point communication links, a directive narrow beam less than 10°HPBW with dualpolarization of the antenna could be the good choice for the wireless system design in this particular environment at 60 GHz. These results can be used to understand the amplitude and time dispersion, and the effects such as scattering, polarization, and antenna beam pattern of the scattered channel which can support the design and development of a high-speed wireless system in this particular underground environment.