This paper was presented at the 2025 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN). ©2025 IEEE. Please cite it as: P. Testolina, E. Beshaj, M. Polese and T. Melodia, “Spectrum Sharing Across Terrestrial and Non-Terrestrial Services in the FR3 Upper Midband,” 2025 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN), London, United Kingdom, 2025, pp. 1-9, doi: 10.1109/DySPAN64764.2025.11115947.
Spectrum Sharing Across Terrestrial and Non-Terrestrial Services in the FR3 Upper Midband Paolo Testolina, Ergest Beshaj, Michele Polese, Tommaso Melodia
arXiv:2606.13511v1 [eess.SY] 11 Jun 2026
Institute for the Wireless Internet of Things, Northeastern University, Boston MA {p.testolina, beshaj.e, m.polese, melodia}@northeastern.edu
Abstract—The frequency bands between 7 and 24 GHz, also known as upper midband or Frequency Range (FR) 3, are being considered as an enabler of 6th Generation (6G) mobile networks. This portion of the spectrum exhibits different propagation characteristics compared to frequencies above 24 GHz, while also offering the potential to provide larger bandwidth allocations for mobile systems than those available in the sub-6 GHz range. 6G technology and spectrum policy, however, will need to guarantee coexistence with the incumbents that already use these frequency bands, which include a variety of services, from radiolocation to satellite-based communications, remote sensing, and radio astronomy. In this paper, we consider the challenge of coexistence between 6G terrestrial systems and satellite incumbents in different portions of the FR3 bands. Using a large-scale 3D model of a terrestrial deployment in the city of Boston and an open-source ray tracing solution, we evaluate the level of Radio Frequency Interference (RFI) that tens of terrestrial Next Generation Node Bs (gNBs) generate toward satellites at different elevation angles. Our model, based on realistic obstruction, clutter, diffraction, and reflections, shows that sidelobes and Non-Line-of-Sight (NLoS) paths can significantly contribute to RFI. Besides directionality, the spatial distribution of gNBs also plays a key role in defining the RFI levels, suggesting that a careful design and operation of terrestrial deployments can create coexistence opportunities.
I. I NTRODUCTION With the 6th Generation (6G) standardization process underway within the 3rd Generation Partnership Project (3GPP), industry, academia, and policymakers are looking into what spectrum should be used to support next-generation wireless systems [1]. While the lower millimeter wave (mmWave) spectrum, or Frequency Range (FR) 2, was seen as a promising enabler of ultra-high data rates in 5th Generation (5G) systems, the challenges associated with supporting efficient and costeffective end-to-end 5G deployments in these bands have limited their application in a limited number of dense, urban markets, leaving the promise unfulfilled [2]. At the same time, providing mobile service over large bandwidths (e.g., 400 MHz carriers, compared to 100 MHz or smaller blocks This work was partially supported by the U.S. NSF under award CNS2332721 and by OUSD(R&E) through Army Research Laboratory Cooperative Agreement Number W911NF-24-2-0065. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the Army Research Laboratory or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation herein.
available at FR1) remains an enticing proposition, as it enables high area capacity as well as support for precise sensing and positioning applications being considered for 6G [3], [4]. For this reason, the upper midband, or FR3, between 7 and 24 GHz, is considered as a candidate for 6G systems as it comes with opportunities to allocate larger chunks of bandwidth to mobile services, and has more favorable propagation characteristics compared to FR2 [5]–[7]. At the same time, this portion of the spectrum comes with its own set of policy and engineering challenges. While being discussed as a single frequency range, different subportions have different design requirements in terms of, for example, antenna and beamforming support. Further, the upper midband spectrum is already allocated and in use for multiple, heterogeneous services, spanning from fixed satellite uplink and downlink connectivity to sensing and radiolocation solutions. Therefore, any solution for allocating spectrum to 6G mobile services will be more effective if it incorporates a spectrum-sharing approach. Sharing allows new services to coexist with incumbents safely and does not lead to complex and extensive spectrum refarming processes, which usually have a significant economic and societal impact [8]. When considering sharing with legacy infrastructure, sharing mechanisms and options are more limited compared to a clean slate design, in particular if the incumbent infrastructure does not support coordination mechanisms [9]. For example, Citizen Broadband Radio Service (CBRS) in the sub-6 GHz spectrum evacuates the communication channels when an incumbent is detected, implementing a time-based sharing without the need for two-way coordination. Another option is spatial sharing, where the different services can coexist thanks to intrinsic limits in the propagation domain. Specifically, when considering communications in higher frequency bands, e.g., FR3 and FR2, the directionality required to close the terrestrial link for mobile services can, in some way, facilitate spatial sharing with satellite or non-terrestrial incumbents. FR3 hosts several different kinds of non-terrestrial incumbents, including communication links (which receive interference from terrestrial networks in their uplink receivers), sensing, and radio astronomy. Each of the services come with different needs and requirements in terms of sensitivity and protection from Radio Frequency Interference (RFI). In this paper, we provide insights on whether spatial sharing
©2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
is possible within terrestrial and non-terrestrial (specifically, Low-Earth Orbit (LEO)) networks at FR3. Compared to recent literature in this area [5], [6], we consider two key elements of next-generation terrestrial cellular systems. The first is the high density of the deployments, with a large number of terrestrial Next Generation Node Bs (gNBs) generating potential RFI to satellite incumbents. The second is directionality, i.e., we model beamforming with Multiple Input, Multiple Output (MIMO) arrays, and its interplay with reflections originating from the ground and obstacle, to analyze how this affects the average and worst-case RFI. We specifically focus on simulating the RFI between the terrestrial system and the in-orbit incumbent of a non-terrestrial network, differently from [10]–[12]. The contributions of this paper are as follows. (i) We answer the question what are the network conditions and configurations that can lead to coexistence opportunities for terrestrial and non-terrestrial networks? (ii) We do so by extending an open-source framework that combines large-scale ray tracing and 3D modeling [13] to the non-terrestrial domain, modeling satellites and their orbits. (iii) We leverage a 3D model based on real-world maps and building elevation to simulate largescale RFI in an urban scenario. This is usually associated with high deployment density but also comes with significant building obstructions to the Line-of-Sight (LoS) between gNBs and satellites, which facilitates the coexistence of terrestrial and non-terrestrial systems. (iv) For the same reason, we show that LoS channel models commonly employed for ground-tosatellite sharing studies and RFI analysis might not capture the rich scattering conditions of the urban environment, where the multipath components play a key role. Our results show that (a) the sidelobes of directional beams can decrease their effectiveness in suppressing the LoS interference to the incumbent, if not carefully designed. (b) The Non-Line-of-Sight (NLoS) components of the channel can deliver a non-negligible amount of interference to the satellite, even when the direct path is successfully attenuated, particularly when considering them in combination with directional communications systems. (c) The spatial distribution of gNBs plays a pivotal role in the resulting RFI, pointing to promising spatial-sharing schemes based on strategic placement or activation of the gNBs. (d) Characterizing the aggregated interference from multiple sources is not trivial and is not always captured by the analysis of a single interferer. II. R ELATED W ORKS In the last decade, RFI evaluation has been a key focus of research, due to the adoption of new frequency bands by commercial stakeholders. Several publications have provided insights [14], [15] to develop solutions for coexistence of current incumbent services and new cellular networks. Many studies on RFI follow the Minimum Coupling Loss (MCL) approach [5], [16]–[18], which analyzes the RFI generated by a single interferer affecting a single victim. Kang et al. [5] focus on the evaluation of the RFI to satellites in the FR3 bands, but consider a single link. Other recent papers
focus on different frequency bands, e.g., above 100 GHz in [17], [19]. Polese et al. [17] review existing regulations for the spectrum above 100 GHz and highlight the importance of developing sharing strategies to optimize the spectrum usage in these frequency bands. The paper studies RFI under a number of different scenarios aligned with ITU standards, such as addressing single backhaul terrestrial links. Similarly, [19] considers an urban scenario, using a rooftop-mounted receiver as a surrogate for a satellite. The results indicate that in a single-link urban setting, the interference safety threshold at 140 GHz is maintained if the transmitter’s beam stays under 15◦ above the horizon. Our prior work [20] extends the above100-GHz analysis to multiple ground transmitters, showing the importance of considering ground reflections—amplified by the main lobe—in the analysis. Other approaches aim to use stochastic tools, ray tracing, and propagation models to estimate the RFI. For example, [21] analyzes aggregated interference in the 47.2 − 50.2 GHz band and highlights the importance of considering the density of terrestrial transmitters, path loss, and stopband attenuation. Similarly, [22] introduces stochastic models that account for LoS, reflections, diffraction, building clutter and foliage to characterize RFI in urban scenarios, in the sub-6 GHz band. [23] estimates the interference from terrestrial Fixed Service stations into aeronautical SATCOM links operating in the 17.7–19.7 GHz band. The paper emphasizes that larger SATCOM antennas and higher flight altitudes can mitigate interference effectively. Similarly, [24] examines the aggregated RFI from a fixed-service network to an aircraft operating at 18 GHz. Also at FR3, [6] investigates terrestrial-to-satellite spectrum sharing, considering the interference caused from cellular downlink transmission to the satellite uplink in the 12 GHz band and proposes a spatial nulling method to reduce RFI. While [6] is based on a ray tracing tool, the evaluation does not consider an urban scenario, as we do in this paper, which may lead to different reflection and obstruction patterns. Other papers that focus on RFI in the upper midband include [10]–[12], [25], [26], which however consider a ground station as the target of the mobile network RFI. Niloy et al. [11] develop a simulation-based framework that incorporates realistic deployment scenarios, beamforming, directionality, and propagation models at 12 GHz, and highlight the importance of exclusion zones around satellite receivers. The authors of [10] dynamically adjust such zones with a closedloop feedback system. Similarly, [12] focuses on adjacent frequency interference and mitigation methods and emphasizes the need for isolation distances. The simulations in [25] and the findings in [26] evaluate the interference aggregated by gNBs of nearby International Mobile Telecommunications (IMTs) networks at 18 GHz and 3.4–3.6 GHz, targeting a Fixed Satellite Service (FSS) terrestrial station. Both studies conclude that coexistence between these systems is feasible at higher frequencies, provided that specific conditions for base station deployment and configuration are met. Authors in [27] analyze the interference caused by automotive radars operating in the 22–27 GHz frequency range on satellite radiometers
6 Atm. Abs. Loss [dB]
Sat. Elevation
10°
30°
60°
13
14
90°
4 2
0 Comm Scientific 8
9
10
11
12
15
16
17
18
19
20
21
22
23
24
Frequency [GHz] Fig. 1: Overview of the spectrum allocations (bottom) and atmospheric absorption loss (top) across the upper-midband.
used for water vapor observations, showing that, under realistic scenarios, vehicle densities are likely to exceed RFI thresholds. Differently from these paper, we consider interference across different kinds of transmitters (terrestrial gNBs) and receivers (satellites). Finally, [28] studies interference from IMT networks to satellite relays in the 25.25–27.5 GHz bands, using LoS propagation and the ITU channel model to aggregate interference over large areas. Similarly, [29]–[31] analyze the coexistence of terrestrial networks and Earth Exploration-Satellite Service (EESS) systems at mmWave frequencies, accounting for RFI from both the satellite’s 3-dB footprint and its broader coverage area. Ground nodes are modeled as geographically clustered 5G systems, with gNBs and User Equipment (UE) antennas randomly placed and oriented. Aggregate RFI is calculated by summing contributions from terrestrial links and clusters, incorporating factors such as free-space path loss, terrain elevation, and atmospheric attenuation. However, the simplified modeling of building blockage, attenuation, and beamforming may not fully capture real-world topologies, which are instead considered in this paper. III. S YSTEM M ODEL Throughout this paper, we consider a dense terrestrial network deployment as the source of RFI, and a non-terrestrial satellite-based system, which is subject to RFI. Terrestrial Network. The terrestrial network represents a next-generation cellular deployment in an urban scenario, e.g., the city of Boston. Each gNB transmits a directional signal to a local UE within its coverage area, which is defined as an annulus with radiuses 5 and 400 m (corresponding to a pathloss of at most 130 dB, in the worst case). The UEs positions are determined as follows: 1) for each gNB, a coverage map is computed; 2) from each coverage map, the first 100 locations where the received power is the strongest are selected; 3) at each iteration, a UE is placed at a random location among the pre-computed list of each gNB; 4) the gNB beam is steered toward to selected UE. With this procedure, we ensure a realistic gNB-UE association and replicate realistic beamforming and beam steering dynamics.
The gNBs are equipped with a single uniform planar array of Ntx elements, where 2 λ 2 ≤ Amax , (1) Ntx = max N s.t. N 2 where N represents the number of antenna elements along one dimension of the array, λ is the wavelength corresponding to the carrier frequency, and Amax is the array area, that we set as 40 × 40 mm2 . Non-Terrestrial Satellite System. We model a satellite receiver transiting over the area of the terrestrial deployment, at different elevation angles, and with a beam that covers the whole area of the terrestrial deployment. International and national regulations reserve several bands for satellite-based services in the 7 − 24 GHz range, as shown in Fig. 1 (bottom). The services can be classified as related to scientific, commercial, or national security missions. The first consists of satellites equipped with scientific instruments, e.g., the EESSs, that observe natural and anthropogenic phenomena for scientific purposes. The other services provide communications capabilities, which can be used to relay commercial, defense, or scientific data across the globe and to/from orbit and space missions, as well as radiolocation and radionavigation solutions. A key difference between the scientific and the other classes of satellites is their sensitivity to RFI. Scientific missions often aim at measuring signals with a very low Signal-to-Noise Ratio (SNR), and thus require extremely sensitive receivers, with a low tolerance for RFI, especially in passive sensing contexts. Conversely, communication systems or radiolocation and navigation generally have a higher tolerance to interference sources. In this work, we propose an analysis that is incumbent-agnostic, i.e., we consider the signal at the input of the incumbent antenna. In this way, we derive general results that can be easily extended to incumbent-specific considerations. IV. L ARGE -S CALE RFI E VALUATION To efficiently evaluate RFI at large scale, we leverage BostonTwin [13], [20], [32], a digital twin of the city of Boston that combines (i) high-accuracy 3D models of the structures in the city; (ii) real-world locations of the gNBs; and (iii) an optimized ray tracing toolchain. With this, it allows studying RFI and other performance metrics over a
Refl. Loss [dB]
10
Inc. Angle
80 °
60 °
Med.-Dry Ground
Concrete
30 °
5 0 Glass
Brick
Plaster
Material Fig. 2: The reflection loss has a negligible dependency on frequency across the upper midband.
large geographical area which includes realistic obstructions and deployment characteristics. Large-scale digital twins are becoming the backbone of holistic network optimization, e.g., optimal gNB deployment, coverage analysis, beam alignment, and interference management. However, they are fundamental when considering any link between a ground and a satellite node, as the footprint of an antenna on board a satellite covers a large geographical area, even if highly directional. The BostonTwin encompasses a 360 km2 area, which includes more than 1000 real gNB locations and about 163000 structures, for a total of more than 11 million mesh triangles. The BostonTwin is fully compatible with the Sionna raytracer [33], a GPU-based, open-source ray-tracer developed by NVIDIA. Reflection of arbitrary order, diffraction of the first order, and scattering on the considered meshes are simulated with a Shooting-and-Bouncing Rays (SBR) algorithm with a number of rays set by the user. Compared to the (exhaustive) Method of Images (MoI), which checks the impinging of the rays on every mesh of the scene by building a complete visibility tree [34], SBR launches a set of rays in different directions, traces the intersection of each ray to the triangles, up to the maximum reflection order and to the receiver. Thus, the complexity of the latter is independent of the number of triangles, i.e., to the size and detail of the scene, but rather by the number of launched rays. This is key to enabling ray-tracing in large-scale scenarios, reducing memory requirements and enabling detailed propagation environments. However, when using the SBR, the appropriate number of rays must be chosen, to guarantee that all the significant paths are explored. Furthermore, the receiver is generally represented by a sphere rather than a point, to increase the probability of intersection and avoid numerical errors [35], [36]. Both the receiver sphere radius and ray count must be chosen based on scene characteristics, such as scattering density, as no standard tuning method exists. To address this, Sionna refines SBR results by applying MoI only to rays that hit the receiver. A. Ray-based Ground-to-Satellite RFI Channel Model As part of this work, we have extended the integration between BostonTwin and the Sionna Ray Tracer (RT) to derive a ground-to-satellite channel model that takes into account different propagation phenomena. In this extension, these include the atmospheric absorption loss, which is dependent on the elevation angle and the frequency of the ground-to-satellite
link. The extension is open-source and available within the BostonTwin repository [13]. In geometrical optics RTs, the propagation of the electromagnetic (EM) field in the environment is represented by rays, which discretize the (continuous) field into the multiple geometric paths that connect the transmitter to the receiver. Each ray carries part of the transmitted signal’s energy and is characterized by its amplitude a ∈ C, which depends on its interaction with the environment, Angle of Departure (AoD) (ϕD , θD ), Angle of Arrival (AoA) (ϕA , θA ), and delay τ . We include the direct ray and the multipath components generated by multiple reflections and by the first-order diffraction on the buildings of the considered area. This leads to a highly accurate channel model, capable of capturing geometry-specific elements of the deployment, reflection across obstacles and the ground. This can be used to map specific transmitter on the ground to their impact on the RFI, as we show in Sec. V. a) Free-space Propagation: As the signal propagates for a distance l, part of its energy is dissipated according to the free space loss coefficient af sp (l, f ) ∈ C. b) Atmospheric Absorption: The gases present in the Earth’s atmosphere can interact with the EM field if its wavelength has a comparable dimension to the gas molecules. In that case, the latter can resonate and dissipate part of the signal energy as thermal energy. The International Telecommunication Union (ITU) recommendations [37] provide a standard way to model this interaction, and the pycraf package offers an excellent Python implementation of the standard1 [38]. The atmospheric absorption coefficient aatm depends on a number of atmospheric parameters, including humidity, temperature, and the resolution to model the atmospheric layers. For this work, we use the mid-latitude summer height profile [39]. To include the atmospheric absorption in our model, we consider the elevation angle and altitude of the satellite and apply the corresponding atmospheric attenuation coefficient to the rays. The atmospheric absorption loss for the upper midband and different elevation angles of a satellite orbiting the Earth at 400 km is reported in Fig. 1. c) Reflection Loss: When an EM wave impinges on a surface, part of the energy is absorbed by the surface itself. The energy loss is modeled through the reflection attenuation coefficient aref l ∈ C, which depends on the material properties [40] and on the incident angle. Note that in general the material’s EM properties, and thus the reflection coefficient, depend on the frequency of the considered signal. However, for the frequencies under consideration, the dependency on the frequency is almost negligible. Figure 2 reports the reflection coefficient for typical materials in the upper mid-band [40]. d) Diffraction Loss: Finally, the EM field can be diffracted by sharp edges present in the environment, creating diffraction patterns behind the obstacle. Diffraction is generally modeled in geometrical-optics RTs by bending the rays around the edges. The diffraction coefficient is adif f . 1 https://github.com/bwinkel/pycraf
Gtx [dB]
TABLE I: Simulation Parameters 20 10 0 −10 −20 −30 −40 −50 −60
# Ant. El. 1 4 16 36 −80
−60
−40
−20
0
20
40
60
Parameter
Array Size
80
Deviation from Boresight [deg] Fig. 3: Transmitter gain for different deviation angles (x-axis) and array size. As the number of elements increases, the pattern becomes more directional, although sidelobes emerge.
The EM field E at any point P in space is thus E(P ) =
R−1 X
ar (P )E0 ,
(2)
r=0
where R is the number of paths connecting the source to P , E0 is the complex amplitude of the electric field at the source, and ar (P ) is the total attenuation coefficient for ray r. The latter is computed as (dropping the dependency on P for readability)
with
ar = ar,f sp ar,atm ar,ref l ar,dif f ∈ C,
(3)
( aref l ar,ref l = 0
(4)
if r is reflected , otherwise
Value
and analogously for ar,dif f . For this work, we consider a time-invariant channel, and we assume the symbol duration is long enough for the received multipath components to be aggregated in a single tap, thus neglecting the time dimension from the equations and from the analysis. V. R ESULTS In this section, we first provide further details on the simulation setup and metrics of interest, and then discuss the results of the large-scale RFI analysis. A. Simulation Setup and RFI Metrics of Interest The simulation parameters are reported in Table I. The simulations sweep across the upper midband, focusing on the frequencies at the intersection between existing allocations and the bands of interest for the communications community, i.e., 7, 8.8, 10, 12.2, 15, 19, and 24 GHz. A BostonTwin scene is defined by the 3D models of the structures and by the antenna locations in a given area. For this work, we selected BOS G 5, one of the default scenes, that encompasses a 1.5×1.5 km2 area and includes 195 gNBs. The pattern of each antenna element follows the 3GPP standard [41]. The resulting array pattern is reported in Fig. 3. As detailed in Sec. III, at each Monte Carlo iteration, each gNB is steered to one UE, to simulate a realistic pointing in a dynamic system.
Array area Amax Array spacing Beamforming Polarization Element patterns Number of transmitters Transmitter Height BostonTwin Scene Number of Monte Carlo iterations Minimum P G UE-gNB distance Satellite Altitude Satellite Elevation Num. of Rays per TX Reflection Max. Refl. Order Diffraction
Freq. [GHz] # Ant. El. 7 1 8.8, 10, 12.2 4 15, 19 16 24 36 40 × 40 mm2 λ/2 Steering Vertical TR 38.901 [41] 195 10 m BOS G 5 [13] 100 −120 dB 5 − 400 m 400 km 10-89° 1e7 Enabled 3 Enabled
The broad spectrum of satellite incumbents, ranging from extremely sensitive, narrow-beam EESS radiometers to highpower radars to communications satellites, brings a correspondingly diverse set of characteristics and requirements. Evaluating the RFI for each incumbent requires accounting for its gain and specific requirements. To provide more general results, we evaluate the path gain at the input of the incumbent antenna, i.e., P Gtx+ch =
R−1 X EI D gtx (ϕD = r , θr )ar , Etx r=0
(5)
where EI and Etx are the electric field at the incumbent and at the transmitter, respectively, and gtx is the transmitter gain. With this definition, we are able to study the interplay between the antenna gain and directionality with the channel and the relevant propagation phenomena. Note that commonly, satellites employ directional antennas to compensate for the increased path loss (in communications) and to reduce the observation area and increase the spatial resolution (both communications and scientific purposes). However, given the satellite’s altitude, even a narrow beam projects a large footprint on the ground. Note that any emission from the footprint area is amplified by the receiver antenna within −3 dB from the peak gain GI = max{gI (ϕ, θ)}, independently of the AoA. Thus, extending the results presented in this work to a specific incumbent with gain GI is straightforward, or considering a transmit power Ptx , as the interference power PI is obtained as PI = P Gtx+ch GI Ptx .
(6)
In the following, we drop the subscript from P Gtx+ch for readability, and we use P G to refer to the quantity defined in Eq. (5).
0.6 0.4 0.2 0.0 0.0
0.2
20 15 10 0
−200
−190
20
40 60 Elevation [deg]
−180
(a) LoS −170 −160 −150 Path Gain [dB]
−160 −150 Path Gain [dB]
1.0
−140
−130
15 10 0
20
−200
−130
−170
0.8
20
80
−140
−180
0.6
Frequency [GHz]
Frequency [GHz]
Frequency [GHz]
−190
0.4
−190
40 60 Elevation [deg]
Reflection −180 (b)−170 −160 −150 Path Gain [dB]
80
−140
20 15 10 0
−130
−200
20
−190
40 60 Elevation [deg]
80
Diffraction −180(c)−170 −160 −150 Path Gain [dB]
−140
−130
Fig. 4: Sample mean of the Path Gain plus transmitter gain P G for the LoS (a), reflected (b), and diffracted rays (c). As the satellite transits over the transmitters, the LoS gets attenuated by the beam pattern, reproducing the trend of Fig. 3. Differently, the multipath components contribute the most to the aggregated interference at other satellite elevations, due to their different AoD. 7 GHz
12 GHz
24 GHz 1.0
0.8
0.8
0.8
0.6
0.6
0.6
F (x)
1.0
F (x)
F (x)
Frequency: 1.0
0.4
0.4
0.4
0.2
0.2
0.2
0.0 −300
−280
−260
−240 −220 −200 x = P G [dB]
−180
−160
−140
(a) Satellite elevation 10°.
0.0 −300
−280
−260
−240 −220 −200 x = P G [dB]
−180
(b) Satellite elevation 40°.
−160
−140
0.0 −300
−280
−260
−240 −220 −200 x = P G [dB]
−180
−160
−140
(c) Satellite elevation 80°.
Fig. 5: Empirical Cumulative Distribution Function (ECDF) for the considered gNBs, obtained with 100 Monte Carlo simulations. The light lines represent the ECDF of each gNB, while the solid ones were obtained by accounting for all the gNB at each frequency.
B. Relevant Propagation Phenomena in a Ground-to-Space Channel Model We first analyze the contribution of the three propagation modes, i.e., LoS, reflection, and diffraction, to the aggregated interference. For tractability, diffuse scattering is not considered in this work. Figure 4 shows the sample mean of the path gain P Gtx+ch for different midband frequencies and satellite elevation angles, for the LoS (P GLoS ), the reflected (P Gref l ), and the diffracted rays (P Gdif f ), computed as follows: NgN B −1
P Gtype =
X b=0
NgN B −1 Rb −1
P Gb =
X
X
b=0
r=0
D gb (ϕD r , θr )ar δtype ,
(7) where Rb is the number of rays (propagation paths) from the D bth gNB to the satellite, gb (ϕD r , θr ) is the gNB gain in the D D th direction (ϕr , θr ) of the r ray. ar is the attenuation factor of the rth ray due to propagation effects, as defined in Eq. (3), and ( 1 if ray type is type δtype = , type ∈ {LoS, refl., diff.}. 0 otherwise (8) The bottom row of Fig. 4 (a-c) represents the path gain at 7 GHz, obtained using a single patch antenna with low directionality [41], as reported in Fig. 3. Thus, it can be used as a baseline to understand the directionality of the channel.
As the satellite rises over the horizon and its elevation angle increases, the distance between the ground transmitters and the incumbent decreases. Thus, the free-space and the atmospheric loss decrease, as the signal travels a shorter distance and crosses fewer layers of the atmosphere. At the same time, when the satellite is low on the horizon, the probability of blockage by buildings is higher than when the satellite is above the city. As the incumbent transits over the city, the interference increases, peaking around 40° elevation for the LoS channel, when more interferers are in LoS and before getting attenuated by the antenna pattern. The reflected and diffracted rays in Fig. 4b and 4c exhibit similar behavior. When using directional arrays, steering the gNB arrays toward the users amplifies the LoS path for lower elevation angles and suppresses it when the satellite is above the network. This is visible in the last column of Fig. 4a, where the aggregated path gain does not cross the −180 dB threshold. Further increasing the directionality, e.g., using 16 (15, 19 GHz) or 36 (24 GHz) antenna elements, effectively reduces the interference peak, the LoS toward a satellite at 30 − 40° elevation is attenuated by the null of the array pattern (Fig. 3). However, the presence of sidelobes limits the effectiveness of the directionality, as the interference around 50° elevation remains significant. The reflected rays (Fig. 4b) exhibit a distinct trend with peaks around 40° elevation, strongly linked to geometric conditions. Specifically, signals directed downward toward
−140
−130 −140
−150
−150 −160 −170
0 20 Elevat 40 60 ion [d 80 eg]
−160 10 Hz] 15 [G 20 ency −170 25 requ F
Path Gain [dB]
−130 Path Gain [dB]
terrestrial users are reflected from scatterers, e.g., buildings and roads, aligning optimally at mid-elevation angles. At low to mid-elevation angles (10° –60° ) and lower frequencies, the signal is diffracted by the building edges in its path, reaching the incumbent with non-negligible power. Moreover, in general, the AoD of the multipath components differs from that of the LoS. Correspondingly, the transmit gain applied to the reflected rays is different than that of the LoS, which results in the different patterns visible in Fig. 4b and Fig. 4c. Notably, more interference is generated by the multipath components than the LoS when the satellite is above the terrestrial network (89° elevation), particularly for highly directional arrays, as they can be amplified by the main lobe before bouncing towards the satellite. This is in line with our previous findings [20], which showed that RFI analysis for very directional networks should take into account secondary paths to the incumbent.
Fig. 6: Aggregated path gain at the satellite incumbent. Directional beams can effectively suppress the interference when the incumbent transits above the transmitter. When the satellite is high over the horizon (30-50°), the sidelobes of the array cause a significant leakage toward the incumbent.
C. Interference by a single gNB Figure 5 shows the ECDF of the path gain for each gNB P Gb , b ∈ {0, .., NgN B } obtained with 100 Monte Carlo simulations at 7 (red), 12 (green), and 24 GHz (blue) for three satellite elevation angles. The semi-transparent lines represent the individual gNB distributions, while the opaque ones represent the aggregated statistics, i.e., the distributions of all the collected samples at the specified frequency and elevation angle. Note that, at each Monte Carlo iteration, the gNBs are steered toward a random user among the selected ones, as explained in Sec. III. At low satellite elevation angles, such as in Fig. 5a, a significant overlap among different frequencies and gNBs is exhibited, suggesting that interference at lower angles is primarily determined by blockage and propagation distance rather than frequency-dependent antenna beamforming. Nevertheless, the probability of the gNB being steered toward the incumbent is greater than at larger elevation angles. For this reason, the rightmost tails of the 24 GHz distributions emerge from the others, corresponding to the cases when beamforming amplifies the interference in the direction of the satellite. With a single antenna (7 GHz, red line), the gNB ECDFs are often step-shaped, indicating a low variability across the Monte Carlo iterations. A low-directivity antenna minimizes variation across iterations, as the primary difference between iterations lies in the direction of the gNBs antenna. Similarly, as the satellite elevation increases, the beamforming used at 12 and 24 GHz effectively suppresses the RFI to the satellite, reducing the breadth of the corresponding ECDFs. At moderate elevations, such as in Fig. 5b, the 12 GHz configuration generates higher interference than the 24 GHz configuration, due to the lower elevation of the sidelobes of the 4 × 4 (12 GHz) than those of the 6 × 6 (24 GHz) array. Conversely, at high elevations (Fig. 5c) the interference at 24 GHz exceeds that at 12 GHz. All the distributions show a plateau lower tail, corresponding to the samples where the generated interference is below the numerical precision of the simulation, i.e., P G ≃ −∞.
The plateau value of F (x) represents the corresponding probability for the single gNB (transparent) and on the whole network (opaque). For the former, there appear to be three clusters of gNBs, characterized by the zero-interference probability of 0.67, 0.38, and 0. As the satellite rises over the horizon, the probability of generating non-negligible interference increases, and thus the gNBs move from the lower to the higher-probability clusters. Similarly, the overall probability of causing zero interference decreases from 0.43 (10°) to 0.23 (40°) and 0.2 (80°). This suggests that there exists gNBs that due to their placement are more prone to generating significant RFI, e.g., those in the first cluster. Site-specific sharing systems can take this feature into account to minimize changes to the ground network and reduce the coordination required to meet the safety interference margins. D. Aggregated Interference Figure 6 reports the aggregated path gain from the gNBs in the considered area, averaged over the 100 Monte Carlo iterations and for different frequencies and satellite elevation angles. The surface mirrors the results presented in Sec. V-B and Sec. V-C. At 0°, the atmosphere and the satellite distance effectively reduce RFI, even when aggregated over multiple sources. The transmitter beamforming plays a similar role for 90° elevations. However, if the transmitter pattern is not directional enough (7 − 8 GHz), or has significant sidelobes (15 − 24 GHz), the transmit power can leak in the direction of the incumbent, peaking at around −125 dB when using 4 antennas at 8.8 − 12 GHz. We remark that the chosen metric does not take into account the transmit power of the transmitters, nor the incumbent gain. Considering the high directivity of the antennas onboard satellites, and the extreme sensitivity of the scientific radiometers used by EESS, the safety levels defined by the ITU recommendations could be easily exceeded. However, narrower beams show promising results, suggesting that a careful beam design, which suppresses the rays in the direction of the incumbent, can enable
the sharing of the upper mid-band between future terrestrial networks with existing satellite services. VI. C ONCLUSIONS AND F UTURE W ORK The wireless community is analyzing the risks and opportunities associated with spectrum sharing and coexistence with existing services in the upper midband, a candidate for nextgeneration 6G systems. In this work, we considered a realistic deployment of a terrestrial network working at upper midband frequencies (7−24 GHz), along with a generic satellite incumbent. Using the Sionna RT and BostonTwin, we modeled the reflections and diffractions of the transmitted signals on the 3D models of the buildings in the considered 1.5 × 1.5 km area in Boston, USA. Additionally, the effect of atmospheric absorption were introduced. We analyzed separately the contribution of the multipath components generated by the interaction with the environment elements, and the interplay with directional antennas. The results showed that the power reaching a LEO incumbent from a ground network and a realistic deployment can sum up to a significant level of interference. Directional antennas and beamforming are shown to mitigate the RFI for some satellite elevation angles, but the power leaking through the sidelobes remains significant. Therefore, we concluded that directionality alone was insufficient to meet international interference requirements, highlighting the need for carefully designed beam patterns and coordinated deployment strategies. Furthermore, the scalability of our framework, enabled by Sionna, allows for large-scale evaluation of coexistence scenarios. This feature is crucial for assessing the feasibility of 6G deployments across diverse urban environments and ensuring that findings generalize to realistic network conditions. As future work, we plan to further improve the channel model, by accounting for other elements such as foliage loss and scattering, as well as different satellite orbits and terrestrial network configurations. Furthermore, the obtained results indicate the possibility of designing terrestrial networks strategically to minimize interference with existing satellite services, paving the way to a safe, coexistence-oriented 6G deployment. R EFERENCES [1] M. Matinmikko-Blue, S. Yrjölä, and P. Ahokangas, “Spectrum management in the 6G era: The role of regulation and spectrum sharing,” in 2020 2nd 6G Wireless Summit (6G SUMMIT). IEEE, 2020, pp. 1–5. [2] A. Narayanan, M. I. Rochman, A. Hassan, B. S. Firmansyah, V. Sathya, M. Ghosh, F. Qian, and Z.-L. Zhang, “A Comparative Measurement Study of Commercial 5G mmWave Deployments,” in IEEE Conference on Computer Communications, May 2022, pp. 800–809. [3] M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Zorzi, “Toward 6G Networks: Use Cases and Technologies,” IEEE Communications Magazine, vol. 58, no. 3, pp. 55–61, March 2020. [4] M. A. Uusitalo, P. Rugeland, M. R. Boldi, E. C. Strinati, P. Demestichas, M. Ericson, G. P. Fettweis, M. C. Filippou, A. Gati, M.-H. Hamon, M. Hoffmann, M. Latva-Aho, A. Pärssinen, B. Richerzhagen, H. Schotten, T. Svensson, G. Wikström, H. Wymeersch, V. Ziegler, and Y. Zou, “6G Vision, Value, Use Cases and Technologies From European 6G Flagship Project Hexa-X,” IEEE Access, vol. 9, pp. 160 004–160 020, 2021.
[5] S. Kang, M. Mezzavilla, S. Rangan, A. Madanayake, S. B. Venkatakrishnan, G. Hellbourg, M. Ghosh, H. Rahmani, and A. Dhananjay, “Cellular wireless networks in the upper mid-band,” IEEE Open Journal of the Communications Society, 2024. [6] S. Kang, G. Geraci, M. Mezzavilla, and S. Rangan, “Terrestrial-satellite spectrum sharing in the upper mid-band with interference nulling,” in IEEE International Conference on Communications. IEEE, 2024, pp. 5057–5062. [7] Z. Cui, P. Zhang, and S. Pollin, “6G Wireless Communications in 7-24 GHz Band: Opportunities, Techniques, and Challenges,” arXiv preprint arXiv:2310.06425, 2023. [8] Y. Furuya, J.-i. Takada, K. Fukawa, N. Nakajima, T. Fujii, and Y. Takeda, “Radio spectrum vision in 50 years from now: Consideration on long term spectrum refarming mainly below 3 ghz,” in IEEE Conference on Standards for Communications and Networking (CSCN), 2015, pp. 7–12. [9] S. Bhattarai, J.-M. J. Park, B. Gao, K. Bian, and W. Lehr, “An Overview of Dynamic Spectrum Sharing: Ongoing Initiatives, Challenges, and a Roadmap for Future Research,” IEEE Transactions on Cognitive Communications and Networking, vol. 2, no. 2, pp. 110–128, June 2016. [10] T.-s. R. Niloy, S. Kumar, A. Hore, Z. Hassan, C. Dietrich, E. W. Burger, J. H. Reed, and V. K. Shah, “ASCENT: A Context-Aware Spectrum Coexistence Design and Implementation Toolset for Policymakers in Satellite Bands,” arXiv preprint arXiv:2402.05273, 2024. [11] T.-S. R. Niloy, Z. Hassan, N. Stephenson, and V. K. Shah, “Interference Analysis of Coexisting 5G Networks and NGSO FSS Receivers in the 12-GHz Band,” IEEE Wireless Communications Letters, vol. 12, no. 9, pp. 1528–1532, 2023. [12] L. Liu, X. Meng, and W. Xie, “Research on 5G Base Station to Satellite Earth Station Interference and Mitigation Methods,” in 2023 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (BMSB). IEEE, 2023, pp. 1–6. [13] P. Testolina, M. Polese, P. Johari, and T. Melodia, “Boston Twin: the Boston Digital Twin for Ray-Tracing in 6G Networks,” in Proceedings of the 15th ACM Multimedia Systems Conference, ser. MMSys ’24. New York, NY, USA: Association for Computing Machinery, 2024, p. 441–447. [Online]. Available: https://doi.org/10.1145/3625468.3652190 [14] M. J. Marcus, “Harmful interference and its role in spectrum policy,” Proc. of the IEEE, vol. 102, no. 3, pp. 265–269, Feb. 2014. [15] J. Chamberlain, D. Starobinski, and J. T. Johnson, “Facilitating Spectrum Sharing with Passive Satellite Incumbents,” IEEE Journal on Selected Areas in Communications, 2024. [16] J. Park, E. Lee, S.-H. Park, S.-S. Raymond, S. Pyo, and H.-S. Jo, “Modeling and Analysis on Radio Interference of OFDM Waveforms for Coexistence Study,” IEEE Access, vol. 7, pp. 35 132–35 147, 2019. [17] M. Polese, X. Cantos-Roman, A. Singh, M. J. Marcus, T. J. Maccarone, T. Melodia, and J. M. Jornet, “Coexistence and spectrum sharing above 100 GHz,” Proc. of the IEEE (to appear), Jul. 2023. [18] S. Yuan, F. Hsieh, S. Rasool, E. Visotsky, M. Cudak, and A. Ghosh, “Interference Analysis of HAPS Coexistence on Terrestrial Mobile Networks,” in 2022 IEEE Wireless Communications and Networking Conference (WCNC), 2022, pp. 2494–2499. [19] Y. Xing and T. S. Rappaport, “Terahertz Wireless Communications: CoSharing for Terrestrial and Satellite Systems Above 100 GHz,” IEEE Communications Letters, vol. 25, no. 10, pp. 3156–3160, Oct. 2021. [20] P. Testolina, M. Polese, J. M. Jornet, T. Melodia, and M. Zorzi, “Modeling Interference for the Coexistence of 6G Networks and Passive Sensing Systems,” IEEE Transactions on Wireless Communications, vol. 23, no. 8, pp. 9220–9234, 2024. [21] B. Lim and M. Vu, “Interference Analysis for Coexistence of Terrestrial Networks With Satellite Services,” IEEE Transactions on Wireless Communications, vol. 23, no. 4, pp. 3146–3161, 2024. [22] R. Aghazadeh Ayoubi, D. Tagliaferri, F. Morandi, L. Rinaldi, L. Resteghini, C. Mazzucco, and U. Spagnolini, “IMT to Satellite Stochastic Interference Modeling and Coexistence Analysis of Upper 6 GHz-Band Service,” IEEE Open Journal of the Communications Society, vol. 4, pp. 1156–1169, 2023. [23] S. P. Winter and A. Knopp, “Modeling of Fixed Service Interference in Aeronautical SATCOM Channels,” IEEE Transactions on Aerospace and Electronic Systems, vol. 58, no. 2, pp. 942–961, 2022. [24] ——, “Statistics of Terrestrial Fixed Service Interference in the Aeronautical SATCOM Channel,” in IEEE International Conference on Communications (ICC), 2019. [25] F. Guidolin, M. Nekovee, L. Badia, and M. Zorzi, “A study on the coexistence of fixed satellite service and cellular networks in a mmWave
scenario,” in IEEE International Conference on Communications (ICC), 2015, pp. 2444–2449. [26] C. Su, X. Han, X. Yan, Q. Zhang, and Z. Feng, “Coexistence Analysis between IMT-Advanced System and Fixed Satellite Service System,” in IEEE Military Communications Conference, 2014, pp. 1692–1697. [27] A. J. Gasiewski, C. S. Ruf, M. Younis, and W. Wesbeck, “Impacts of mobile radar and telecommunications systems on Earth remote sensing in the 22-27 GHz range,” in IEEE International Geoscience and Remote Sensing Symposium, vol. 3. IEEE, 2002, pp. 1679–1681. [28] Liyuan Zhong et al., “The Feasibility of Coexistence between IMT2020 and Inter-Satellite Service in 26 GHz band,” in Int. Wireless Communications and Mobile Computing (IWCMC), 2020, pp. 1006– 1011. [29] Y. Cho, H. Kim, E. E. Ahiagbe, and H.-S. Jo, “Spectral Coexistence of IMT-2020 with Fixed-Satellite Service in the 27-27.5 GHz Band,” in International Conference on Information and Communication Technology Convergence (ICTC), 2018. [30] Y. Cho, H. Kim, D. K. Tettey, K.-J. Lee, and H.-S. Jo, “Modeling Method for Interference Analysis between IMT-2020 and Satellite in the mmWave Band,” in IEEE Globecom Workshops, 2019. [31] Y. Cho, H.-K. Kim, M. Nekovee, and H.-S. Jo, “Coexistence of 5G With Satellite Services in the Millimeter-Wave Band,” IEEE Access, vol. 8, pp. 163 618–163 636, 2020. [32] P. Testolina, M. Polese, and T. Melodia, “Sharing Spectrum and Services in the 7–24 GHz Upper Midband,” IEEE Communications Magazine, vol. 62, no. 8, pp. 170–177, 2024. [33] J. Hoydis, F. A. Aoudia, S. Cammerer, M. Nimier-David, N. Binder, G. Marcus, and A. Keller, “Sionna RT: Differentiable ray tracing for radio propagation modeling,” in 2023 IEEE Globecom Workshops (GC Wkshps). IEEE, 2023, pp. 317–321. [34] M. Lecci, P. Testolina, M. Polese, M. Giordani, and M. Zorzi, “Accuracy versus complexity for mmwave ray-tracing: A full stack perspective,” IEEE Transactions on Wireless Communications, vol. 20, no. 12, pp. 7826–7841, 2021. [35] E. Egea-Lopez, J. M. Molina-Garcia-Pardo, M. Lienard, and P. Degauque, “Opal: An open source ray-tracing propagation simulator for electromagnetic characterization,” Plos one, vol. 16, no. 11, p. e0260060, 2021. [36] J.-M. Molina-Garcı́a-Pardo, L. Rubio, M.-T. Martinez-Inglés, E. EgeaLopez, A. Mateo-Aroca, V. M. R. Peñarrocha, and J. Reig, “Wireless channel characterization from 2 to 28 ghz in an outdoor parking lot,” IEEE Antennas and Wireless Propagation Letters, 2024. [37] ITU, “Attenuation by atmospheric gases and related effects,” Rec. ITU-R P.676-13, 2022. [38] B. Winkel and A. Jessner, “pycraf: Spectrum-management compatibility,” Astrophysics Source Code Library, record ascl:1810.008, Oct. 2018. [39] ITU, “Reference standard atmospheres,” Rec. ITU-R P.835-6, 2017. [40] ——, “Effects of building materials and structures on radiowave propagation above about 100 MHz,” Rec. ITU-R P.2040-2, 2021. [41] 3GPP, “Study on channel model for frequencies from 0.5 to 100 GHz,” Technical Report (TR) 38.901, Jun. 2018.