Investigations on Millimeter-Wave Indoor Channel Simulations for 5G Networks

Abstract

Due to the extensively accessible bandwidth of many tens of GHz, millimeter-wave (mmWave) and sub-terahertz (THz) frequencies are anticipated to play a significant role in 5G and 6G wireless networks and beyond. This paper presents investigations on mmWave bands within the indoor environment based on extensive simulations; the study considers the behavior of the omnidirectional and directional propagation characteristics, including path loss exponents (PLE) delay spread (DS), the number of clusters, and the number of rays per cluster at different frequencies (28 GHz, 39 GHz, 60 GHz and 73 GHz) in both line-of-sight (LOS) and non-LOS (NLOS) propagation scenarios. This study finds that the PLE and DS show dependency on frequency; it was also found that, in NLOS scenarios, the number of clusters follows a Poisson distribution, while, in LOS, it follows a decaying exponential distribution. This study enhances understanding of the indoor channel behavior at different frequency bands within the same environment, as many research papers focus on single or two bands; this paper considers four frequency bands. The simulation is important as it provides insights into omnidirectional channel behavior at different frequencies, essential for indoor channel planning.

1. Introduction

Today, more than 54% of people live in cities, indicating a significant global development in urbanization. Thanks to the widely used 5G cellular networks’ ultra-reliable low latency, gigantic transfer speed, and ultra-high connection density, our lives are set to change. 5G usage covers three main services: enhanced mobile broadband, mission-critical communications, and massive Internet of Things (IoT) [1]. In addition, for improved smartphones, 5G mobile technology has the potential to bring forth new immersive experiences like virtual reality (VR) and augmented reality (AR) with faster, more consistent data rates, lower latency, and lower cost [2]. In 5G systems, multiple input and multiple output (MIMO) technology is utilized along with beamforming techniques to reduce interference and increase reliability [3].

The demand for high-definition videos and low-latency online gaming is increasing noticeably. Also, mobile data traffic has increased dramatically recently due to the sharp increment in smartphones, tablets, and devices that can be connected/controlled remotely. The emergence of Wi-Fi technology and its associated applications created congestion in the sub-6 GHz band. Therefore, researchers aimed to utilize a different window in the spectrum that can meet the new demands.

Propagation at mmWave frequencies is quite different from microwave frequencies [4], as the higher frequency increases reflectivity from walls and objects as the wave dimensions become even smaller compared to the object’s dimensions; however, diffuse reflection becomes more dominant, while less power can penetrate through objects. Diffraction contributes less to signal strength; therefore, objects create sharp shadows. This is due to the effect of material electrical properties, which change with frequency.

Signal propagation within indoor environments has received attention in the research community, with models being developed to describe the propagation behavior within these environments [5]. Some models are empirical, where the best fit for measurements is used; numerous empirical models are proposed in [6,7]; however, these models should be used in similar environments; others are stochastic, where the channel is random at any time, yet stochastic models describe an indoor channel using its statistical characteristics [8,9,10,11,12]. Furthermore, other models are deterministic, which can estimate channel parameters by solving Maxwell’s equations. Generally, the solvers are either the ray-tracing method [13] or finite difference time domain FDTD-based methods [14]; there are also semi-deterministic models, which are hybrids of deterministic and empirical or stochastic models. These models improve the deterministic model’s performance [15].

This paper presents a study on 5G signal propagation behavior in indoor environments at mmWave frequencies; the study considers the use of directional and omnidirectional antennae in LOS and NLOS scenarios. The investigated parameters include path loss exponent (PLE), Root-mean Square (RMS) delay spread (DS), and some temporal channel parameters. This study offers a comparative analysis to enhance our understanding of indoor channel behavior at various operating frequencies within the same environment. The simulation is important as it provides insights into omnidirectional channel behavior at different frequencies, essential for indoor channel planning. This research is beneficial for network engineering designers to account for the impact of walls on indoor network coverage.

The paper’s layout is as follows: Section 1 contains the introduction; Section 2 discusses the methodology and simulation setup; results and discussions are presented in Section 3; and, finally, conclusions are drawn in Section 4.

2. Previous Related Work

Channel modeling for 5G indoor networks has been an exciting topic. In [16], measurements were conducted in university campaigns at 28 GHz and 73 GHz to estimate path loss exponents in line-of-sight LOS and non-LOS scenarios. The authors [17] conducted measurements in two buildings and compared indoor path-loss models at 38 GHz; they proposed an enhanced model based on ITU recommendations for indoor propagation. WINNER project [18] presents channel models based on channel measurements performed at 1–6 GHz; the model was later extended to cover LOS and NLOS propagation scenarios at 61–65 GHz [19]. In [16], the authors investigated the behavior of delay spread, path loss, and received signal strength at mmWave frequencies for multi-floor environments using ray-tracing software; a similar study was reported by [17]. The usage of ray-tracing software for mmWave frequencies is highly dependent on the software input parameters and accuracy of environment design; in [20], the authors investigated the effect of using different values of electrical constitutive parameters (permittivity and conductivity) on ray-tracing results. It was found that choosing the accurate values of permittivity and conductivity is vital for obtaining accurate results. In [21], authors studied the millimeter frequency band (26, 28, 32, and 38 GHz) propagation properties in indoor multi-floor stairwell situations; the estimated PLE value was around 7. A similar study was performed at 26 GHz and 38 GHz in corridors and stairwell scenarios; it was found that directional PLE in corridors was less than 2 and 4 for LOS and NLOS, respectively, for both types of linear polarizations (V-V and V-H). At the same time, omnidirectional PLE is lower, especially for V-H polarization cases; in stairwells, the omnidirectional PLE values are slightly bigger than those recorded in corridors. In contrast, the directional PLE values are remarkably bigger than the recorded values in corridors.

In [22], experiments were conducted in a typical indoor office environment at New York University. The study investigated large-scale path loss and temporal statistics at 28 GHz and 73 GHz. The PLEs for LOS scenarios were 1.1–4.1 and 1.3–4.7 at 28 GHz and 73 GHz, while, for NLOS scenarios, PLEs were 2.7–5.1 and 3.2–6.4 at 28 GHz and 73 GHz, respectively. In [23], a wideband double-directional channel sounding campaign at 28 GHz is presented and the investigated environments include an indoor open hall. For LOS and NLOS scenarios, PLE were around 1.9 and 2.8, respectively. PLE and RMS delay spread were examined at 28 GHz within an indoor environment [24]; for LOS scenarios, the PLE and RMS-DS were 1.5 and 5–20 ns, respectively, while, for NLOS scenarios, the PLE and RMS-DS were 2.2 and 10–40 ns, respectively.

A set of wideband directional propagation measurements was conducted at 39 GHz in an indoor hall environment, where the PLE at LOS was around 1.8 [25]. Authors of [26] conducted measurements within an indoor environment at 60 GHz with different antenna heights; in their study, they found that scatter paths contribute more to signal strength than direct paths and that human movements can cause obstructions around 18–36 dB. The PLE for LOS scenarios were in the range of 0.6–1.2, while, for NLOS scenarios, the PLEs were in the range of 2.7–5.4.

Authors in [27] examined PLE and RMS-DS in LOS, obstructed LOS, and NLOS scenarios at 30 GHz and 60 GHz within indoor environments; for 30 GHz, the recorded PLE values for LOS, OLOS, and NLOs were 1.7, 1.8, and 2.3, while, for RMS-DS, the recorded values for LOS, OLOS, and NLOS were 6.7 ns, 27.1 ns, and 28.3 ns, respectively. For 60 GHz, the recorded PLE values for LOS, OLOS, and NLOs were 1.2, 3.7, and 5.7, while, for RMS-DS, the recorded values for LOS, OLOS, and NLOS were 3.4 ns, 23.5 ns, and 22.5 ns, respectively. RMS-DS tends to be smaller for higher frequency, while PLE tends to be larger.

The RMS-DS was examined at 60 GHz within LOS and NLOS scenarios in indoor corridors and office environments [28]; for LOS scenarios, RMS-DS were in the range of 12.3–21.1 ns, while, for NLOS scenarios, the RMS-DS were in the range 18.5–31.7 ns.

3. Methodology and Simulation Setup

This paper investigates the general indoor propagation channel characteristics at the mmWave. This includes the path loss exponent (PLE), Root-mean Square (RMS), delay spread (DS), and multipath temporal properties. The simulations were conducted in an indoor simulated environment; the environment was constructed using a ray-tracer software named Wireless InSite (WI), version 3.3.5, which has been validated over WLAN [29] and millimeter wave frequencies [20]. The software considers the effects of material electrical properties. It also allows the user to configure waveform, antenna types, and the properties of the transmitter and receivers. Simulations consider line-of-sight (LOS) and non-LOS (NLOS) scenarios. Four millimeter-wave frequencies were adopted, 28 GHz, 39 GHz, 60 GHz, and 73 GHz; their corresponding bandwidths are 0.8 GHz [22] and 2 GHz [30,31]. Figure 1 shows the simulated environment of the Chesham building at the University of Bradford. The access points (APs) transmit power to over 1450–1591 receiver points (RPs) (the number of RPs depends on AP coverage), and both APs and RPs are mounted on 1 m. The study considers three antenna pattern configurations, using directional antennae at both ends of the communications links to find the directional properties of the channel. It also considers the channel’s omnidirectional properties by using an omnidirectional antenna and, finally, the directional antenna used at the transmitters and omnidirectional antenna used at the receivers.

The WI permits the user to adjust a wide range of parameters, including antenna type, transmitted power, operating frequency, signal bandwidth, electrical constitutive parameters, number of reflections, transmissions, and diffractions, propagation model, ray-tracing method, sum complex electric fields, the number of propagation paths, etc.

Table 1. Wireless InSite settings.

Property	Setting
Transmitter antenna	Horn antenna: WR-15 (26.5–40 GHz) WR-15 (50–70 GHz) Omnidirectional
Receiver antenna	Horn antenna: WR-15 (26.5–40 GHz) WR-15 (50–70 GHz) Omnidirectional
Transmitted power	23 dBm
Antenna gain	15 dB (WR-15 26.5–40 GHz) 24 dB (WR-15 50–70 GHz)
Sum complex electric fields	None
Number of reflections	6
Number of transmissions	4
Number of diffractions	3
Number of paths	10
Ray Spacing (°)	0.1
Plane-wave ray spacing	0.5 m
Propagation model	Full 3D
Ray tracing method	SBR
Ray tracing acceleration	Octree

summarizes the WI settings used in the simulations.

Table 2. Building material properties’ relationship with frequency.

Frequency (GHz)		28	39	60	75.3
Concrete	${\epsilon }_r$	5.31	5.31	5.31	5.31
	$\sigma$	0.48	0.633	0.90	1.06
Glass	${\epsilon }_r$	6.27	6.27	6.27	6.27
	$\sigma$	0.23	0.34	0.57	0.72
Wood	${\epsilon }_r$	1.99	1.99	1.99	1.99
	$\sigma$	0.17	0.24	0.38	0.47
Drywall	${\epsilon }_r$	2.94	2.94	2.94	2.94
	$\sigma$	0.12	0.16	0.21	0.24

provides building material properties’ values as a function of the operating frequencies, including conductivity (σ) and relative permittivity (εr); these values are based on the ITU recommendations [32]. Due to losses at high frequencies (path loss (PL) tends to increase by a factor of $20 log\left(f\right)$ dB, where f is the operating frequency), higher antenna gain is used at higher frequencies. More specifications on the antenna used are available in [33].

The WI uses the shoot and bouncing rays (SBR) as a ray-tracing method; it also uses the method of images to correct the SBR paths, which reduces the error in calculating received power.

The Saleh Valenzuela model is widely used to describe the propagation channel environment at UWB frequencies; the model describes multipath behavior in indoor environments, suggesting that rays come in clusters. For indoor environments, if the time difference between two adjacent rays is more than 6 ns, the rays are said to come from 2 different clusters [34]. The raw information is taken from WI; the data are then analyzed using Matlab(R2023a) codes to estimate channel parameters. For example, PLE is estimated by taking the path loss information recorded by WI; then, Equation (5) (below) is used through a Matlab code. The number of clusters is estimated by taking the rays’ arrival time using a Matlab code; if the time delay between two consecutive rays is more than 6 ns, the two rays belong to different clusters.

Figure 2 presents a simulated LOS propagation. The propagation paths include a direct path, a reflected path, and multi-reflected paths, and the rays’ colors denote the relative power of each path.

Received signal strength (RSS) can be computed with and without the consideration of the small-scale effect; the first method takes the power sum of all multipath rays, known as “power sum prediction (PS)” [35,36]:

$$〈P_{PS}〉=\sum _M{P}_M.$$

(1)

where 〈P_PS〉, M, and P_M are the averaged power using the PS method, number of multipath rays, and power of each ray, respectively. This method requires a huge bandwidth and antenna array to exclude the small fading effect.

Path loss (PL) models describe the propagation environment through a mathematical expression. These models are widely used in environments with similar features to the original one from which the model’s parameters are extracted. The close-in (CI) free space reference distance PL model is a common PL model, which allows simple comparisons with different frequency bands, which is characterized by a single PLE (n) [37,38].

$${PL}_{\left(f,d\right)}^{CI}=FSPL\left(f,d_0\right)+10n {\log }_{10}\left({\displaystyle \frac{d}{d_0}}\right)+X_{\sigma }^{CI}$$

(2)

where FSPL is the free-space path loss, f is the operating frequency, d₀ is the reference distance, usually 1 m from the transmitter, and $X_{\sigma }^{CI}$ is a zero-mean Gaussian random variable due to shadowing with standard deviation (σ) in dB. FSPL is given by Equation (3), where $\lambda$ is the wavelength in m.

$$FSPL\left(f,1\right)=20 {\log }_{10}\left({\displaystyle \frac{4\pi }{\lambda }}\right)$$

(3)

The PLE is estimated using the Minimum Mean-Squared Error (MMSE), which fits the simulated results with the lowest error. From Equation (2):

$$X_{\sigma }^{CI}={PL}_{\left(f,d\right)}^{CI}-FSPL\left(f,1\right)+n\left(10 {\log }_{10}\left(d\right)\right)=A-nB$$

(4)

where A = ${PL}_{\left(f,d\right)}^{CI}-FSPL\left(f,1\right)$ and B = $\left(10 {\log }_{10}\left(d\right)\right)$.

The PLE is estimated using MATLAB as in Equation (5), where A and B are written as column vectors.

$$n=A^T{\left(B^{T}B\right)}^{-1}B$$

(5)

RMS delay spread (${\sigma }_{\tau }$) is estimated from the second moment of the delay power spectrum and describes the multipath time dispersion and coherence bandwidth:

$${\sigma }_{\tau }=\sqrt{\overline{{\tau }^2}-{\left(\overline{\tau }\right)}^2}$$

(6)

where $\overline{{\tau }^2}$ is the 2nd central moment of a power delay profile and $\overline{\tau }$ is the mean excess delay. WI calculates the RMS-DS directly so the user can have this information from the extracted data.

4. Results and Discussion

Figure 3 presents a single-frequency directional PL model at 39 GHz for both LOS and NLOS scenarios using the three antenna pattern configurations; as seen, for the LOS propagation scenario, the PLE is less than the free space value (n = 2) due to the waveguiding effect; however, omnidirectional PLE has the lowest value, which means that path losses are lower since omnidirectional antennae can collect rays that are due to waveguiding more efficiently. In NLOS, similar behavior was recorded, where PLE ranged from 4.88 to 5.8.

Although the directional antenna focuses the beam in a specific direction, its PLs are more dispersed; it is worth mentioning that PL does not consider antenna gain. Figure 4 presents RSS comparisons using an omnidirectional and directional antenna. The RSS is larger using a directional antenna since the beams are focused from both antenna terminals; as distance increases, the difference between RSS values decreases slightly. The purpose of this figure is to show the reader that we use a directional antenna to focus the beams in specific directions and, hence, the RSS will be larger. However, PL is still larger using the directional antenna, as seen in Figure 3 and

Table 3. Omnidirectional–omnidirectional CI path loss model parameters.

	NLOS		LOS
	PLE (n)	STD (dB)	PLE (n)	STD (dB)
28 GHz	4.36	8.89	1.48	0.63
39 GHz	4.88	10.7	1.5	0.73
60 GHz	4.51	10.1	1.54	0.96
73 GHz	4.71	11.69	1.6	1.08

,

Table 4. Directional–directional CI path loss model parameters.

	NLOS		LOS
	PLE (n)	STD (dB)	PLE (n)	STD (dB)
28 GHz	5.29	9.24	1.58	0.63
39 GHz	5.8	9.97	1.63	0.63
60 GHz	6.98	13.07	1.72	0.76
73 GHz	6.75	13.564	1.83	0.99

and

Table 5. Directional–omnidirectional CI path loss model parameters.

	NLOS		LOS
	PLE (n)	STD (dB)	PLE (n)	STD (dB)
28 GHz	4.94	9.58	1.54	0.78
39 GHz	5.16	11.38	1.58	0.58
60 GHz	4.65	7.74	1.65	0.5
73 GHz	4.56	8.31	1.72	0.6

.

Table 3, Table 4 and Table 5 present the PLE values extracted using the MMSE method for LOS and NLOS scenarios and the three antenna pattern configurations. Omnidirectional PLE has the lowest values due to more paths between the transmitter and receiver than other configurations. In contrast, directional PLE values tend to be the biggest values. Using both directional and omnidirectional antennae, the PLE values are in between the values of the previous two antenna configurations. For all configurations, PLE values show a dependency on frequency. In LOS scenarios, the PLE values are less than two due to the waveguiding effect; it tends to increase with frequency.

Table 6. PLE comparison with the published literature.

		28 GHz	39 GHz	60 GHz	73 GHz
NLOS	Our paper	4.36	4.88	4.51	4.71
	[22]	4.4	-	-	5.3
	[23]	2.8	-	-	-
	[24]	2.2	-	5.7	-
	[39]	3.3	-	-	-
LOS	Our paper	1.48	1.5	1.54	1.6
	[22]	1.7	-	-	1.6
	[23]	1.9	-	-	-
	[24]	1.5	-		-
	[25]	1.8	-	-	-
	[26]	-	1.2	-	-
	[27]	-	-	1.2	-
	[40]	1.6	-	-	-
	[41]	2.1	-	-	-

compares recorded PLE in this research and other published values from the literature, and the comparison shows good agreement. As seen, this paper presents a comprehensive study of many frequency bands.

Table 7. Mean RMS delay spread (ns).

Freq. (GHz)	Dir–Dir		Omni–Omni		Dir–Omni
	NLOS	LOS	NLOS	LOS	NLOS	LOS
28	2.57	2.24	6.76	10.66	5.38	11.66
39	2.21	1.46	4.11	8.37	3.31	9.31
60	1.95	1.65	3.27	4.46	2.57	4.94
73	2.25	2.87	3.64	9.36	2.38	9.98

presents the mean RMS delay spread (DS) over the investigated frequencies; mean RMS-DS is inversely proportional/decreases as frequency increases; this is because, as frequency increases, the signal power drops rapidly, especially for NLOS scenarios, due to higher path loss and high dispersion and penetration losses. Therefore, signals travel shorter distances and fall below receiver sensitivity. The directional antenna focuses the radiation in specific directions, which reduces the effect of multipath and, therefore, the traveling time will be shorter. In LOS scenarios, using a directional and omnidirectional antenna will have bigger RMS-DS results since the directional antenna provides signals that travel further distances and the omnidirectional antenna can collect rays from the surroundings.

Figure 5 presents the probability plot of RMS-DS for NLOS scenarios using different antenna configurations, as seen in the fact that RMS-DS decreases at higher frequencies, as observed in [27]. The difference between recorded values using a directional and omnidirectional antenna is observable. For example, at 28 GHz, the probability for DS to be less or equal to 5 ns is 0.9 (using directional channels (Dir–Dir)) and 0.4 (using omnidirectional channels (Omni–Omni)); similarly, the probability is 0.95 (Dir–Dir) and 0.6 (Omni–Omni) at 39 GHz, 0.85 (using Dir–Dir) and 0.75 (using Omni–Omni) at 60 GHz, and 0.85 (using Dir–Dir) and 0.65 (using Omni–Omni) at 73 GHz.

Omnidirectional channels have more propagation paths and, hence, longer DS. This difference is visible in LOS scenarios, as seen in Figure 6. Figure 6 displays the probability of the omnidirectional and directional RMS-DS for LOS scenarios; as seen, using directional antennae at both ends, the RMS-DS is less than 5 ns for almost 90% of the cases, while, for omnidirectional channels, the percentages were less than 30% for most of the examined frequencies.

Figure 7 and Figure 8 present the number of clusters and the number of rays/cluster for NLOS scenarios, respectively, using omnidirectional, directional, and directional–omnidirectional radiation patterns. The number of ray clusters follows the Poisson distribution; while the number of clusters follows a decaying exponential distribution. For omnidirectional channels, the majority of rays come in three clusters or less, while, for directional channels, the majority of rays come in five clusters or less. When using directional–omnidirectional antenna patterns, rays come in four clusters or fewer, which is more than the case of the omnidirectional channels and less than the case of the directional channels. By inspection of Figure 8, it can be seen that, for the Dir–Dir case, the probability of having 1 ray/cluster ranges from 0.4 to 0.6. In contrast, the probability does not exceed 0.25 for the Omni–Omni case, which explains why the Dir–Dir case has more clusters than others.

Figure 9 and Figure 10 present the number of clusters and the number of rays/cluster for LOS scenarios, respectively, for different antenna configurations. The number of clusters follows a decaying exponential distribution, where most rays fall within two clusters, especially in directional and omnidirectional cases. In the directional scenarios, the probability for rays to arrive in a single cluster is around 0.8 for all frequencies, while, in the omnidirectional scenarios, it ranges from 0.7 to 0.9. It is worth mentioning that, in the Dir–Omni case, the majority of arrays are coming in four clusters or less; this is due to the combined effect of using a directional antenna and omnidirectional antenna; this can be seen in the number of rays per cluster as depicted in Figure 10.

Figure 11 shows the arrival times of multipath at a receiver point. There are three clusters, each with a different number of intra-cluster paths. Considering the first ray of each cluster, it was found that the best fit for the amplitude of these rays follows a negative exponential distribution. In contrast, the inter-arrival times of each cluster follow a modified Poisson distribution and their corresponding amplitudes follow a negative exponential distribution [42].

Figure 12 shows penetration losses through a concrete wall using Dir–Dir configuration; as seen, as operating frequency increases, penetration losses increase; this is expected, as walls become more reflective when increasing the frequency, with penetration losses through concrete at 28 GHz, 39 GHz, 60 GHz, and 73 GHz are 43 dB, 48 dB, 61 dB, and 72 dB, respectively. Also, as seen in the figure, when no obstacle is in the LOS (the first 3.5 m), free-space path gain ($20 {\log }_{10}\left({\displaystyle \frac{\lambda }{4\pi d}}\right)$) is different for each frequency due to the effect of frequency, where λ is the wavelength and d is the distance between the transmitter and receiver.

Penetration losses depend on the type of the wall material and thickness of the wall; in a separate simulation where there is no reflection from ground or ceiling and the only source of signal strength is transmission through walls and diffraction around them, the penetration losses through an 11.5 cm drywall at 28 GHz, 39 GHz, 60 GHz, and 73 GHz were 4.22 dB, 5.35 dB, 6 dB, and 6.6 dB, respectively, while transmission losses through a 30 cm concrete wall, using the same set of frequencies were 74.83 dB, 86.58 dB, 90.9 dB, and 113.2 dB, respectively. As seen, penetration losses tend to increase monotonically with increasing frequency for different materials.

Additionally, diffraction becomes less prominent as frequency increases; for example, diffraction loss around a concrete wall increases by 6 dB as frequency increases from 28 GHz to 39 GHz, while it increases by 25 dB as frequency increases from 60 GHz to 73 GHz.

The effect of wall penetration losses and diffraction losses gives an indication for network designers when deploying access points in the environment, for example, in an environment like offices where most of the partitions and walls are made of drywalls, the number of the required hot spots that operate at 60 GHz and 73 GHz will be less compared to an environment that has concrete walls.

5. Conclusions

In this paper, extensive simulations were conducted at mmWave frequencies (28 GHz, 39 GHz, 60 GHz, and 73 GHz) within an indoor environment to investigate the propagation behavior at LOS and NLOS scenarios using different antenna configurations. The investigated parameters include PLE, RMS-DS, number of clusters, and number of rays per cluster. It was found that PLE shows dependency on frequency, the omnidirectional PLE was found to be less than the directional ones in both LOS and NLOS scenarios, and the LOS PLE was less than two due to the waveguiding effect. DS generally is lower at higher frequencies; in LOS scenarios, omnidirectional RMS-DS is bigger than directional ones due to having more propagation paths. The number of clusters follows a Poisson distribution for NLOS scenarios, while, for LOS scenarios, it follows a decaying exponential distribution; omnidirectional channels tend to have fewer clusters. The number of rays per cluster for NLOS scenarios follows a decaying exponential distribution. As seen from the literature, depending on the type of indoor environment (i.e., offices, halls, corridors, etc.), PLE, RMS-DS, and other channel parameter values will vary significantly. This work presents a comparative study that will enhance our understanding of indoor channel behavior at different operating frequencies within the same environment. This simulation is important as it provides the omnidirectional channel behavior at different frequencies, which is important for indoor channel planning. The work is helpful to network engineering designers in considering the effect of walls on network indoor coverage. In the future, the work can be extended to consider more indoor environments and operating frequencies.