Potentials for Application of Millimeter Wave Communications in Cellular Networks

Transcription

1 Wireless Pers Commun (2017) 92: DOI /s Potentials for Application of Millimeter Wave Communications in Cellular Networks Aleksandar Ichkov 1 Vladimir Atanasovski 1 Liljana Gavrilovska 1 Published online: 1 November 2016 Springer Science+Business Media New York 2016 Abstract Future 5G cellular networks will need to deliver significantly increased system capacity and user data rates. This expected growth along with today s shortage of spectrum raises the need for new frequency allocations. Millimeter wave spectrum is emerging as a suitable candidate with a vast amount of available bandwidth (around 60 GHz). Extending cellular networks communications on millimeter wave frequencies requires extensive measurement campaigns and analysis of signals propagation characteristics. This paper gives an overview of recent measurement studies and results used for modeling millimeter wave channel behavior in different propagation environments. Also, the paper provides a preliminary simulation analysis of a hybrid LTE-millimeter wave heterogeneous network, which suggests that Gbps user data rates are achievable with sufficient beamforming gains. However, the millimeter wave cellular extensions will require architectural changes to address the technical issues spanning from the transceivers design to the operational procedures in both access and backhaul network parts. Keywords Cellular networks Millimeter wave & Aleksandar Ichkov ichkov@feit.ukim.edu.mk Vladimir Atanasovski vladimir@feit.ukim.edu.mk Liljana Gavrilovska liljana@feit.ukim.edu.mk 1 Faculty of Electrical Engineering and Information Technologies, Ss. Cyril and Methodius University in Skopje, Skopje, Macedonia

2 280 A. Ichkov et al. 1 Introduction The rapid mobile data growth and users demands are creating unprecedented challenges for wireless networks deployments in terms of massive capacity requirements [1]. Currently allocated spectrum in the sub 3 GHz bands is overly saturated and spread into disjoint frequency bands (0.8, and 2.4 GHz etc.) with different propagation characteristics. LTE-Advanced network concept introduces carrier aggregation of distinct frequency carriers, continuous or non-continuous, with different size. The technique can extend the available bandwidth up to 100 MHz (5 20 MHz). New more efficient sharing methods such as Licensed Shared Access (LSA) [2] and cognitive radio concepts can further extend spectrum opportunities. However, this is still not sufficient for the envisioned massive increase of traffic demands towards 5G. New revolutionary technologies and spectrum opportunities are needed. Ultra-dense heterogeneous deployments, higher frequencies radio access and software-defined network architectures are among the candidate technologies that will lead the way into an overall 5G solution, alongside upgraded and evolved existing technologies such as LTE and/or LTE-A. The massive expected increase in envisioned future cellular networks system capacity ( 1000 times) and data rates ( times) inevitably leads to a need for new massive spectrum bandwidth availability. Recent studies suggest that millimeter wave frequencies, ranging from 30 to 300 GHz, could be used to augment the currently saturated sub 3 GHz spectrum bands for wireless communications [1]. Millimeter wave communications can take advantage of the huge and unexploited bandwidth to cope with the high requirements for next generation of wireless networks, enabling: low-interference ultra-dense millimeter wave heterogeneous networks. multi gigabit per second (Gbps) wireless millimeter wave broadband access. low-cost fiber replacement millimeter wave mobile backhaul links. low-latency high-definition media transfers. The emergence of millimeter wave communications will create need for new signal processing, circuit, antenna, and communication technologies. The convergence of these technologies will inevitably cope with the stringent constraints imposed by the adverse propagation characteristics. Millimeter wave cellular networks will differ from conventional cellular systems due to the distinct channel characteristics at these frequencies. The move to millimeter wave cellular deployments will enable larger available bandwidth (in GHz) and capabilities for high gain steerable multi-antenna systems that will exploit advanced multiple-input multiple-output (MIMO) and beamforming techniques. All this will enhance the spectral efficiency, improve the system capacity and enable multi Gbps user data rates in future cellular networks. This paper presents an overview of the millimeter wave spectrum potentials as a future cellular extension and a 5G-capable spectrum bandwidth. We analyze the main characteristics of millimeter wave frequencies such as atmospheric attenuation (atmospheric oxygen, humidity, fog and rain), penetration and reflection losses and path loss modeling using data from conducted urban propagation measurements. Measurement results have confirmed the feasibility of using millimeter wave in cellular access and backhaul networks [3 6], personal area networking [7] and wireless local area networks [8]. The paper is organized as follows. Section 2 overviews the characteristics of the millimeter wave spectrum bands, atmospheric attenuation and increased path losses, and the

3 Potentials for Application of Millimeter Wave Communications possible advantages of these high frequencies. Section 3 presents the feasibility of millimeter wave wireless communications with an overview of recent urban propagation measurement studies. Section 4 gives the key aspects for enabling millimeter wave cellular communications. Section 5 presents novel channel modeling approaches and simulation models for millimeter wave cellular communications. We present a preliminary simulation study and performance evaluation of a hybrid LTE-millimeter wave heterogeneous network in Sect. 6. Section 7 provides a brief discussion on the potential enabling technologies and research directions for millimeter wave in 5G cellular networks. Finally, Sect. 8 concludes the paper. 2 Characteristics of Millimeter Wave Spectrum Bands The main benefit of frequencies above 10 GHz is the potential availability of large amount of spectrum bandwidth. According to the International Telecommunication Union (ITU) [9], the sub 100 GHz frequency allocation possesses up to 63 GHz of available bandwidth, as seen at Table 1. High millimeter wave frequencies are the key to unlocking wide transmission bandwidths (from hundreds of MHz to GHz) that will enable massive capacity enhancements in cellular network and multi Gbps data rates. Table 1 ITU s frequency allocation for GHz Frequency designation Frequency range (GHz) Available spectrum (GHz) /38/ Total available

4 282 A. Ichkov et al. We overview the main propagation characteristics such as atmospheric attenuation and path losses and presents interference and power advantages that are essential for successful and reliable wireless communications at millimeter wave frequencies. 2.1 Atmospheric Attenuation A common myth in the wireless community is that rain and atmospheric attenuation make millimeter wave spectrum useless for mobile communications. The propagation of millimeter waves through the atmosphere depends on atmospheric oxygen (oxygen absorption), humidity and rain attenuation. Figure 1 shows the atmospheric attenuation for millimeter wave propagation. The colored circles denote different spectrum bands in the millimeter wave spectrum, each possessing distinct advantages in atmospheric propagation. The blue ones distinct the pick in the atmospheric attenuation and the green ones denote the lows (the white one is for the microwave sub 10 GHz spectrum bands). For example, at 28 and 38 GHz there is very low attenuation (with and db over 200 m respectively) or for the 60 GHz bands there is high attenuation of 4 db over 200 m [1]. This makes each of these bands suitable for different kinds of communications. For example, due to the high attenuation, the 60 GHz has the capability of becoming the new ISM millimeter wave band, with unlicensed, short-range indoor communications usage. The low atmospheric attenuation for 28, 38 and 70 GHz makes them preferable for cellular communications usage in both access and backhaul networks. Rain (out of all atmospheric conditions) has the highest impact on signal loss in the millimeter wave propagation. This signal attenuation depends on the rainfall rate, usually measured in millimeters per hour. For example, as rain attenuation results show at Fig. 2, the attenuation significantly increases from 0.9 db/km on a light rainfall (1 mm/h) at 70 GHz signals, up to 18.4 db/km on an intense rainfall (50 mm/h) at 70 GHz as the rainfall rate increases. The characteristics of rainfall throughout the world have been well studied. ITU s developed a model for computing the probability of rain rates at various geographical locations [10], which can be used effectively to estimate millimeter wave link s performance and easily overcome rain attenuation in future millimeter wave network links. Considering the fact that cell size in urban network deployments are averaging radiuses of few hundreds of meters, atmospheric attenuation for millimeter wave communications may not be such a limiting factor. Employing proper planning and deploying techniques Fig. 1 Atmospheric attenuation [1]

5 Potentials for Application of Millimeter Wave Communications Fig. 2 Rain attenuation [1] can help overcome the issues with the atmospheric and rain attenuation and enable millimeter wave wireless communications. 2.2 Propagation Losses Propagation losses (path losses) at millimeter wave frequencies were considered to be too high for practical wireless communications until now. Previous generation of network deployments were utilizing network links longer than few hundreds of meters and inter-site distances of few kilometers. Considering the Friis free space path loss equation, by moving from a 2 to a 28 GHz, the free space path loss increases by nearly 14, 19 db by moving to 38 GHz and up to 30 db for moving to 60 GHz frequency carrier. Also, the carrier wavelength k shrinks from 15 cm at 2 GHz to 10.7, 7.9 and 5 mm for 28, 38 and 60 GHz, respectively. Increasing the antenna gains at both sides can mitigate this significant increase in the path loss. Signal wavelength at these frequencies ranges up to several millimeters. As current antenna elements size is proportional with the wavelength size, a higher number of antenna elements can be packed into a given antenna physical size or aperture. Using such multi-antenna arrays, higher gains can be achieved as the wavelength shrinks. The path loss is independent of the frequency as long as receiver antenna physical sizes are the same in the comparing frequencies, as seen from Eq. 1. With the assumption that the receiver antenna size remains constant at any comparing frequency, the expression for the free space path loss becomes: P t ¼ G t G r ð4prþ 2 ð1þ P r where P t and P r denote the transmit and received power respectively, G t and G r denote the gain of the transmit and the receive antennas and R denotes the distance between the transmit and the receive antennas.

6 284 A. Ichkov et al. For fixed physical size antennas, frequency does not matter in the path loss for free space propagation. In principle, path loss can be overcome with beamforming by using multiple antenna arrays with high gain antennas at the transmitter and the receiver. 2.3 Interference and Power Advantages Frequency independent beamforming can be implemented at millimeter wave frequencies using antenna array comprised of a number of smaller individual antenna elements. For a fixed physical antenna size, these arrays will have more elements as the frequency increases (i.e. the wavelength decreases yielding a decrease in the antenna elements size), providing greater gain at higher frequencies. Millimeter wave links possess very narrow beams as illustrated in Fig. 3. High gain antenna arrays with very narrow beams can be used to overcome the perceived propagation loss at millimeter wave frequencies. The transmit energy will be focused in a narrow direction in the radio link from the transmitter to the receiver, lowering the transmitted beam s interference within the area. This will allow usage of multiple independent links in the coverage area with a high level of flexibility and scalability. Multiple link radio transmissions, higher frequency reuse and lower interference levels are among the benefits that will enhance the overall network performances. However, the practical implementation of such high gain multi-antenna arrays in millimeter wave wireless communications is a big challenge, due to hardware constraints. Implementation of these large antenna arrays is not straightforward and will need a change in the system design principle to exploit array gain at both the transmitter and the receiver. Digital beamforming cannot be implemented directly at millimeter wave frequencies due to the high power consumption. We need hybrid designs, combining analog and digital, to increase the flexibility in deploying such antenna arrays and help overcome the power and hardware constraints at these high millimeter wave frequencies. 3 Feasibility of Millimeter Wave Communications Extending wireless communications on millimeter wave frequencies requires measurement studies and extensive analysis to deeply understand the signal propagation characteristics. We provide an overview of recent measurement studies and results used for modeling channel behavior in different propagation environments. Fig. 3 Beam sizes for microwave and millimeter waves [11]

7 Potentials for Application of Millimeter Wave Communications Propagation Measurements: Penetration and Reflection The shorter wavelength of the millimeter wave signals causes significant increase in blockage susceptibility. To understand the millimeter wave signal propagation in different areas, signal penetration and reflection of common building materials are required for both indoor and outdoor cases in different terrain (urban, suburban etc.). Penetration and reflection measurements of millimeter wave propagation were conducted in New York City, for 28 and 38 GHz signals [12 14]. These measurements have indicated that typical outdoor materials show high penetration losses, outlining that outdoor signals penetration through buildings will be difficult using outdoor transmitters. However, common indoor materials exhibit much lower penetration losses than outdoor materials. This will encourage millimeter wave frequencies usage in future indoor wireless networks, such as the new IEEE ad standard [7]. Reflection measurements have also shown that outdoor materials have higher reflection coefficients compared to indoor materials, leading to the conclusion that RF energy can be contained within indoor environments. Outdoor-to-indoor penetration will be very difficult at millimeter wave frequencies. But, the strong outdoor reflectivity and the low attenuation of indoor materials can support separate indoor and outdoor coverage [12]. The highly reflective outdoor building materials can further enhance the outdoor signals coverage and allow a wide range of possible angles of arrival in creating outdoor communication links for different outdoor network deployments. 3.2 Urban Propagation Measurements for 28 and 38 GHz Urban propagation measurements of millimeter wave signals for 28 GHz were conducted at different locations in New York City [1], [14]. Measurement data suggests a rich scattering environment for millimeter wave signals, with high number of multipath components in the observed links. This a result of the high reflectivity of outdoor environments (building materials), for both line-of-sight (LOS) and non-line-of-sight (NLOS) links. Large scale fading modeling for millimeter wave outdoor environments show existence of a reasonable average number of resolvable multipath components within a number of cluster varying from 1 to 4, if the signal at the receiver is detectable. First steps to model detectable clusters are made in [15], implying that the power distribution amongst the clusters follows a log-normal distribution. It is found that exponential distribution can be used to approximate angular dispersion, i.e. the root mean-squared (rms) beamspread in each detectable cluster. The presence of multiple clusters of multi-path components opens the possibility for advanced techniques such as spatial multiplexing and MIMO systems to exploit diversity gains using multi-antenna systems. Combined with beamforming and/or beemsteering, high gains can be achieved to overcome the increased path loss at millimeter wave frequencies. The observed average path loss exponent (PLE) values are similar to microwave frequencies (2 for LOS and for the average PLE for NLOS environments). Major distinction is the presence of high outage probability with 57 % of the locations with distances larger than 200 m in outage, due to the obstructive nature of the channel. This indicates the fact that millimeter wave networks will rely on dense deployments, using cell sizes of few hundred meters. This is not an immersive challenge as today s urban network deployments use similar cell sizes. Future network deployments will be even denser, as network densification has been identified as one of the key paradigms of cellular networks aimed at addressing the increased need for capacity.

8 286 A. Ichkov et al. Another set of propagation measurements were conducted in Austin, Texas, for the 38 GHz spectrum band [14]. The average PLE for LOS environment was found to be around 2.30 and the average PLE for NLOS was measured to be 3.86, as shown in Table 2. Ifwe compare the results from the two sets of propagation measurements, first we note that the PLE values show similar values as the ones in today s microwave cellular networks. Also, the 38 GHz PLE at these measurements has lower values than the ones for the 28 GHz measurements. This can be primarily due to the different measurement environments and conditions, which implies that a higher number of measurement sets is required to generalize results. Both measurement campaigns suggest same average coverage distances of around 200 m as achievable cell size. The data collected at all LOS, obstructed LOS and NLOS environments (summarized at Table 2) can provide a basis for development of spatial channel models, showing that propagation characteristics at millimeter wave frequencies are not too different than the ones used in today s microwave cellular networks. 3.3 Propagation Modeling Empirical path loss models were developed for estimating steerable single beam mobile systems to fit urban propagation measurements presented in the previous section. The propagation models in [16] present a modification of existing microwave path loss models using the standard Friis free-space and SUI model. In [17], five types of large-scale path loss models using omnidirectional antennas are compared with outdoor and indoor measurement data: single-frequency FI model, single-frequency CI model, multi-frequency alpha-beta-gamma (ABG) model, multi-frequency CI model and the multi-frequency close-in free space reference distance with frequency weighting (CIF) model. The analysis provided concludes that the CI model is most suitable for outdoor millimeter wave environment compared to the other models, presented as: 4pfr 0 PL CI ðf ; rþ ¼20 log 10 c þ 10n log 10 ðrþþv CI d ; for r [ 1 m ð2þ where 20 log 10 ð 4pfr0 c Þ denotes the free space path loss (FSPL) with respect to the reference distance r 0, n denotes the path loss exponent, r denotes the RX TX distance and v CI d denotes the log-normal shadowing. Table 2 Path loss exponents with respect to a 1 m free space reference distance using CI path loss model presented in Eq. (2) and standard deviations obtained for arbitrary pointing angles and for the best angles for 28 and 38 GHz Environment NLOS LOS NLOS LOS TX height (m) RX height (m) PLE g all d all (db) PLE g best d best (db) TX gain (dbi) RX gain (dbi) Slope correction factor a

9 Potentials for Application of Millimeter Wave Communications The CI model inherently has an intrinsic frequency dependence of path loss embedded within the 1 m free space path loss value. That means that the first meter is included in the FSPL term and the path loss is treated differently for distances greater than 1 m. This model is very simple to compute, has only one parameter to be optimized, the path loss exponent n, and is applicable to both single-frequency and multi-frequency cases. CI model behavior in the 28 and 38 GHz frequency bands, summarized in Table 2, has proven that propagation characteristics of millimeter wave outdoor signals are similar with today s models for microwave outdoor propagation. Yet, we must consider the fact that using fitting parameters from specific measurement data may not be generally correct when applied to different environment scenarios. We need more measurement campaigns for a better channel understanding to provide adequate models that will capture the distinct propagation characteristics of millimeter wave signals. 4 Key Aspects for Millimeter Wave Cellular Networks Cellular networks will need to deliver up to a 1000 times the capacity compared to today s networks in the next decade. The number of connected devices will rapidly increase, up to 50 billion by 2020 [18]. Many of the requirements envisioned for 5G cellular systems, such as multi-gbps peak throughputs and Gbps user data rates seem unattainable using existing cellular technologies. The exponential growth of cellular networks capacity demands with today s shortage of spectrum have emerged the need for new spectrum bandwidths. Millimeter wave spectrum is emerging as a suitable candidate with a vast amount of available bandwidth (around 60 GHz). However, development of millimeter wave frequencies in cellular network faces significant technical challenges. The increased path loss, severe vulnerability to shadowing, blockages, reflections and sparser multipath environment can result in poor channel quality and complete signal outages at certain locations. The design of a cellular system based on millimeter wave communications will need significant changes, much more than just simply scaling the carrier frequency, to reach their full potential. Millimeter wave communication links will rely on highly directional transmissions through the usage of high gain steerable antenna arrays at both the transmitter and receiver s end. A related consequence of highly directional transmissions is that the links become directionally isolated, with interference playing a much smaller role than in current microwave small cell networks. This will revert millimeter wave cellular networks from today s interference-limited regime to a noise-limited regime. Device power consumption to support large number of antennas with very wide bandwidths will also be a key challenge. Recent advances in CMOS RF and digital processing are not able to support low cost and low-power millimeter wave chips suitable for commercial mobiles devices [5]. More efficient RF power amplifiers and combiners will be needed to support large-scale phased array antennas for efficient and reliable millimeter wave cellular communications. The following paragraphs elaborate on some of the most critical items and possible technology enablers towards a successful deployment of a millimeter wave cellular network.

10 288 A. Ichkov et al. 4.1 High Gain Multi-antenna Arrays The physical dimensions of associated antenna elements are reduced due to smaller wavelengths at millimeter wave. This allows a large number of antennas to be packed into the same physical size antenna aperture to be used for directional transmissions using beamforming and beemsteering at both the transmitter and the receiver. The main objective of adaptive beamforming is to shape the beam patterns using high number of steerable antennas at both ends so that the received signal strength is maximized. The transmitted energy is focused into a single direction with small beamwidth signals and very weak power densities radiated in unintended directions. High gain steerable antennas using pencil beams transmissions will extend coverage distances, limit the interference and increase frequency reuse in dense network deployments. 4.2 Path Loss, Blockage and Reflection, Multipath Environment As detailed in Sect. 2, outdoor propagation of millimeter wave signals is not very different from microwave signals. LOS signals at millimeter wave frequencies will propagate as in free space. NLOS signals are much weaker than the LOS ones, with an expected signal coverage range of around 200 m. Both cases require distinct propagation models that incorporate blockages and reflection characteristics. Outdoor signals at millimeter frequencies are more sensitive to blockages. The isolation effects of walls makes it hard for outdoor base stations to cover indoor users, which motivates the deployment of indoor small cells. Indoor blockage effects can also be strong depending on inner walls, number of obstacles and so on. Some indoor locations may require several access points to cover certain area. Channels in the millimeter spectrum bands are sparser in terms of multipath components, with a lower number of significant scatterers, compared to microwave. The angle spreads of reflected millimeter wave signals are also smaller compared to microwave. Still, the number of resolvable multipath components is sufficient to utilize advanced technologies such as spatial multiplexing and multiple input multiple output (MIMO), to increase the radio transmission gains and enable higher capacity links with multi Gbps rates. 4.3 Hardware Constraints and Challenges Millimeter wave transceivers are subject to a set of practical hardware constraints that have a great impact on the transceivers architecture. The traditional approach in microwave transceivers for beamforming/mimo using dedicated RF chain for each antenna (fully digital beamforming) is extremely difficult in millimeter wave. Signal components like analog-to-digital converters (ADCs) have very high power consumption and cost. The analog phase shifters typically used to beamsteer the transmitted signals are affected by a number of limitations as they can only change the phase of the transmitted signal (within a finite set of possible steering angles) and not the amplitude. This hardware constraints will have a great impact on the communication systems design with a need for a new hybrid approach that will incorporate both analog and digital approaches, to enable low cost and energy efficient transceivers for millimeter wave systems. Next, we provide a preliminary study on potential millimeter wave cellular network deployments with simulation results incorporating recent advancements in millimeter wave propagation modeling.

11 Potentials for Application of Millimeter Wave Communications Simulation Modeling of Millimeter Wave Cellular Deployment Bai and Heath [4] present a theoretical model for coverage and rate analysis that incorporates blockages and beamforming for millimeter wave cellular communications in. Twostate link model is presented according to the LOS probability of the link. The LOS probability of the link is modeled as an exponential function of the UE to BS distance. Accordingly, different propagation laws are applied for both LOS and NLOS links. Akdeniz et al. present a systematic study of millimeter wave channel modeling in [15]. Using data provided from measurement sets conducted for 28 and 73 GHz, channel modeling processes are presented for both large-scale and small-scale fading. A new statistical channel model has been proposed based on measurement data (detailed in Sect. 3) showing that at many receiver locations over 200 m, the signals cannot be detected. This outage stage occurs due to reflections, blockages and scattering characteristics of millimeter wave signals, as one of the main difference of millimeter wave and microwave. Channel models until now are using a two-state link modeling according to the standard 3GPP statistical model, where channel variations are modeled by lognormal shadowing [19]. Measurement test data show that millimeter wave channels can experience significant outages that are not well-modeled using the standard two-state statistical model. A novel three-state statistical link model has been proposed by Akdeniz et al., as one of the recent advancement in millimeter wave channel modeling, where blockages and reflections of radio signals have been incorporated to account for the increased outage probability: 1. LOS: there are no obstacles blocking the link between the UE and the BS PL LOS ðdþ ¼ð1 p out ðdþþe a losd 2. NLOS: the link between the UE and the BS is blocked by obstacles PL NLOS ðdþ ¼1 p out ðdþ p los ðdþ 3. Outage: link between the UE and the BS cannot be establish, as the path loss is too high PL OUT ðdþ ¼maxð0:1 e aoutdþbout Þ where d is the UE to BS distances and parameters a los, a out and b out values used to fit the data are given at Table 3. A comprehensive stochastic geometry analysis and modeling for multi-tier millimeter wave networks is presented in [20]. It incorporates the presented three-state channel modeling results, extending the work to a multi-tier millimeter wave heterogeneous networks analysis and modeling using stochastic geometry. An approximation for the three state link modeling is given using the multi ball modeling in [21]. The presented two ball model approximation has come to some very intriguing conclusions on the link-state of millimeter wave radio transmissions. Choosing a radius in order of 50 m is expected to provide excellent coverage but results in a much denser network deployments while scaling the radius up to 200 m is expected to provide a good coverage with a particular trade-off for the network nodes Any coverage radius above 200 m will not be sufficient as most of the RX locations will be in outage state. This two-ball model provides a very simplified version of the three-state link model that can be use with ease for system-level simulations. Another key factor for reliable millimeter wave communications is the utilization of directional transmission links using the small beamwidth propagated signals ( pencil ð3þ ð4þ ð5þ

12 290 A. Ichkov et al. Table 3 Simulation parameters Macro layer Femto layer Observed area (A) 25 km 2 (5 5) Carrier frequencies (GHz) Available bandwidth (MHz) Transmit power (dbm) Average cell radius (km) Density of PPP (BS=km 2 ) Log-normal shadowing r nlos ¼ 4 r los ¼ 5:8, r nlos ¼ 8:7 Noise power N½dB ¼ 174 þ 10 log 10 ðbwþþf with F ¼ 10 db NLOS-LOS-outage probability According to the three-state link model (Eqs. 3 5) a out ¼ 0:0334 m 1, b out ¼ 5:2, a los ¼ 0:0149 m 1 Beamforming gains G max BS ¼ 20 db, G max UE ¼ 5dB beams ). Beamforming and beamsteering are essential in providing high antenna gains to overcome the increased path loss. Modeling beamforming in millimeter wave networks is critical for precise characterization of propagation links. Assuming a simple beamsteering approach that maximizes the received SNR, the actual array pattern can be approximated by a sectored pattern where constant directional gains are assumed for both the main lobe and side lobes [3]. Using the sectored pattern, a simple form for directional beamforming gains at the receiver and the transmitter can be derived for mathematical tractability: ( ðmaxþ G i;k ; if jhj - q G i;k ðhþ ¼ G ðminþ ð6þ i;k ; if jhj [ - q where i, k denotes the pair BS-UE, h is the angle off the boresight direction, - i;k is the beamwidth of the main lobe, G ðmaxþ i;k and G ðminþ i;k are the array gains of main and side lobes, respectively. The UE and the serving BS estimate the angles of arrival to adjust their steerable antenna arrays accordingly to exploit maximum directivity gain of the intended link (in case of absence of beamforming alignment errors). In such case, assuming no alignments errors and beamforming capability and both the transmitter and the receiver, the maximum (according to Eq. 6). The use of highly directional links combined with the short cell coverage radius results in high SINR links with little interference (very small power leakages in unintended directions). This allows millimeter wave cellular networks to move from interference-limited regime (in today s microwave networks) to a noise-limited regime. gain of the intended link will be G ðlinkþ ¼ G ðmaxþ BS G ðmaxþ UE 6 Performance Analysis of Millimeter Wave Cellular Networks We have incorporated the analysis used in [15 21] to set up a hybrid heterogeneous cellular networks with two tiers, using LTE on the macro layer and utilizing millimeter wave frequencies on the pico layer. This simulation scenario is preliminary, not considering any

13 Potentials for Application of Millimeter Wave Communications MAC layer extensions. The macro layer base stations (MBSs) are randomly distributed following Poisson Point Process (PPP) with density k m (MBSs=km 2 ). The pico layer base stations (PBSs) are also randomly distributed following Poisson Point Process with density k p (PBSs=km 2 ). The intensity of each PPP is calculated according to the average cell radius p R c ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1=ðpkÞ. The average number of BSs at each tier is N i ¼ Ak i, where A denotes the observed area and k i denotes the density of the particular tier. The users are also randomly distributed according to a PPP with density k u (users=m 2 ). The performance analysis of the proposed two tier cellular network model was conducted in MATLAB using the parameters given in Table 3. For the macro layer path loss modeling we are using the UMi NLOS path loss model: PL macro ðf c ; rþ ¼26 log 10 ðf c Þþ36:7 log 10 ðrþþ22:7 ð7þ For the pico layer path loss modeling we are using empirical derived single-frequency FI model incorporating the three-state link model given at Eqs. (3 5): PL pico ðrþ ¼a þ 10b log 10 ðrþþv r ð8þ where v r is the log-normal shadowing and parameters a and b are used for fitting the data (different values for LOS and NLOS): a los ¼ 72 db; b los ¼ 2:92; a nlos ¼ 61:4dB; b nlos ¼ 2:92; PL outage ¼þ1 We are using the maximum received power user association, where the associated BS per user is calculated according to the associate rule: n o j ¼ argmax jh s j 2 v s d a s ij ; for s 2ðLOS; NLOSÞ ð9þ 8j2MBS;PBS where jh s j 2 denotes the channel gain, v s is the log-normal shadowing and d a s ij denotes the path loss at distance d between the ith user and the jth BS with PLE a s. We are evaluating the network performances with the respect to the rate distribution W for the association users, or equivalently, the probability that certain percentage of users achieve rate higher than a predefined threshold d, defined as: W ¼ Pr½R [ d; R [ 0 ð10þ The rate R for the ith user in the system, with respect to the received SINR from the associated jth BS, is calculated as: R ij ¼ W ij log 2 ð1 þ SINR ij Þ ð11þ where W ij denotes the occupied bandwidth by the user. The use of directional beamforming on the pico tier significantly lowers the other-cell interference. Simulation scenario shows that noise levels at majority of the receivers locations exceed the aggregate other-cell interference. Therefore we assume a noise-limited regime at the millimeter wave pico tier, which eases the network analysis. Figure 4 gives the user rate distribution in the network. Results show a significant increase of the probability for achieving rates higher that 25 Mbps for the pico users compared to the macro users. A big part of this gain comes from the larger available bandwidth for the millimeter wave pico BSs, 25 times larger than the microwave. Another factor is the smaller number of associated users on the pico tier compared to the macro tier. That is due to the lower average received power for the millimeter wave compared to the microwave received signals.

14 292 A. Ichkov et al. Fig. 4 Rate distribution W for the associated users Table 4 Rate comparison for different network deployments BS average cell radius (m) Average rate per user (Gbps/user) Average system spectral efficiency (bps/hz/cell) Average area spectral efficiency ðbps/hz/km 2 Þ Millimeter wave: Millimeter wave: Millimeter wave: Microwave: Simulation results for different average cell radiuses of the millimeter wave BSs are summarized at Table 4, presenting the average rate per user, the average system spectral efficiency per cell (the sum of the average user data rates per bandwidth at a cell site) and the average area spectral efficiency (the sum of the average user data rates per bandwidth at area unit). Results show that Gbps user data rates are achievable with sufficient beamforming gains. Additional gains can be achieved with a better usage of the whole available bandwidth at millimeter wave frequencies through more advanced beamsteering techniques. 7 Research Directions and Challenges for Millimeter Wave 5G The potential usage of millimeter wave communications in cellular networks requires addressing of several key topics ranging from transceiver design to appropriate operational access and backhaul procedures. This section highlights the most important features that are currently foreseen as pillars towards the successful millimeter wave 5G deployments. The increased number of antenna elements in a single phased array offers many dimensions in the channel that can be leveraged for both spatial multiplexing and beamforming capabilities. Such benefits can be achieved with proper hardware architecture to exploit the millimeter wave channel opportunities. The authors in [22] propose a hybrid transceivers design that combines analog and digital beamforming as a compromise between the high power consumption of digital BF and the lower cost of analog BF transceivers. This design offers a possibility for combining multiple-beams beamforming that maximizes the received SNR in power-limited situations with unique data streams on each of the beams on the same carrier frequency to increase the user data rates.

15 Potentials for Application of Millimeter Wave Communications Higher gains can be achieved implementing the massive MIMO concept, using a massive number of BS antennas (e.g. 100 or more) [23]. Millimeter massive MIMO has flexibility for spatial multiplexing, beamforming and diversity that can further enhance spectral and energy efficiency [24]. High gain multi-antenna arrays offer interference isolation that increases the frequency reuse opportunities. Incorporating self-backhauling BSs that utilize shared frequency allocations for both access and backhauling links will enable backhauling among different BSs without significant loss in throughput [25]. In-band backhauling and advanced BF/MIMO access will enable ultra-dense network deployments with massive capacity capabilities. Additional incentive towards the millimeter wave in 5G cellular networks is the latest decision of FCC proposing new flexible rules for four different bands of high-band spectrum above 24 GHz designed to lay the foundation for 5G networks in the U.S. market [26]. The specific bands that will be studied for 5G services include the GHz, also known as the 28 GHz band; the GHz band, also known as the 37 GHz band; from 38.6 to 40 GHz, known as the 39 GHz band; and the GHz band. The FCC is keen in establishing a regulatory framework to proceed in parallel with technological development to keep pace and help facilitate so-called Fifth Generation (5G) mobile services. 8 Conclusion Millimeter wave cellular networks have a potential for high coverage and capacity as long as the infrastructure is densely deployed, with the use of directional antennas and advanced beamforming for both access and backhaul. However, many open challenges remain for a successful development of mature millimeter wave technology for mobile communications. Nevertheless, the preliminary analysis confirms that the successful integration of the millimeter wave in wireless networks will allow fulfillment of the expected traffic growth and need for higher capacity and user data rates in future 5G cellular networks. References 1. Rappaport, T. S., Sun, S., Mayzus, R., Zhao, H., Azar, Y., Wang, K., Wong, G. N., Schulz, J. K., Samimi, M., & Gutierrez, F. (2013). Millimeter wave mobile communications for 5G cellular: It will work!, IEEE Access, May. 2. Faussurier, E. (2014). ANFR, Introduction of new spectrum sharing concepts: LSA and WSD, ITU-R SG 1/WP 1B Workshop: Spectrum management issues on the use of white spaces by cognitive radio systems, Geneva, Jan. 3. Bai, T., Alkhateeb, A., & Heath, R. W, Jr. (2014). Coverage and capacity of millimeter-wave cellular networks. IEEE Communications Magazine, Sept 4. Bai, T., & Heath, R. W, Jr. (2015). Coverage and rate analysis for millimeter wave cellular networks. IEEE Transactions on Wireless Communications, 14(2), Rappaport, T. S., Murdock, J. N., & Gutierrez, F. (2011). State of the art in 60-GHz integrated circuits and systems for wireless communications. Proceedings of the IEEE, 99(8), Baldemair, R., et al. (2015). Ultra-dense networks in millimeter-wave frequencies. IEEE Communications Magazine, 53(1), Agilent Technologies, Wireless LAN at 60 GHz IEEE ad Explained, May literature.agilent.com/litweb/pdf/ en.pdf. 8. Baykas, T., et al. (2011). IEEE c: The first IEEE wireless standard for data rates over 1 Gb/s. IEEE Communications Magazine, 49(7), ITU-R. Working document towards a preliminary draft new report ITU-R M [IMT above 6 GHz], Feb Recommendation ITU-R P.838-3, Specific attenuation model for rain for use in prediction methods, March

16 294 A. Ichkov et al. 11. Adhikari, P. (2008). Understanding millimeter wave wireless communication. Hawaii: Loea Corporation. 12. Zhao, H., Mayzus, R., Sun, S., Samimi, M., Schulz, J. K., Azar, Y., Wang, K., Wong, G. N., Gutierrez, Jr., F., & Rappaport, T. S. (2013). 28 GHz millimeter wave cellular communication measurements for reflection and penetration loss in and around buildings in New York City. In 2013 IEEE International Conference on Communications (ICC), June. 13. MacCartney, G. R., Zhang, J., Nie, S., & Rappaport, R. (2013). Path loss models for 5G millimeter wave propagation channels in urban microcells. In IEEE Global Communications Conference, Exhibition & Industry Forum (GLOBECOM), Dec. 14. Rappaport, T. S., MacCartney, G. R., Samimi, M. K., & Sun, S. (2015). Wideband millimeter-wave propagation measurements and channel models for future wireless communication system design. IEEE Transactions on Communications, 63(9), Akdeniz, M. R., et al. (2014). Millimeter wave channel modeling and cellular capacity evaluation. IEEE Journal on Selected Areas in Communications, 32(6), Sulyman, A. I., Nassar, A. T., Samimi, M. K., Maccartney, G. R., Rappaport, T. S., & Alsanie, A. (2014). Radio propagation path loss models for 5G cellular networks in the 28 GHZ and 38 GHZ millimeter-wave bands. IEEE Communications Magazine, 52(9), Sun, S. et al. (2016). Propagation path loss models for 5G urban micro- and macro-cellular scenarios, In 2016 IEEE 83rd Vehicular Technology Conference (VTC2016-Spring), May. 18. Rangan, S., Rappaport, T. S., & Erkip, E. (2014). Millimeter-wave cellular wireless networks: Potentials and challenges. Proceedings of the IEEE, 102(3), LTE: Evolved universal terrestrial radio access (E-UTRA): Radio resource control (RRC); Protocol specification (3GPP TS version Release 9) 20. Di Renzo, M. (2015). Stochastic geometry modeling and analysis of multi-tier millimeter wave cellular networks. ieee transactions on wireless communications, 14(9), Lei, W., & Renzo, M. (2015). Stochastic geometry modeling of cellular networks: Analysis, simulation and experimental validation. In MSWiM 15 Proceedings of the 18th ACM international conference on modeling, analysis and simulation of wireless and mobile systems. 22. Sun, S., Rappaport, T. S., Heath, R. W., Nix, A., & Rangan, S. (2014). MIMO for millimeter-wave wireless communications: Beamforming, spatial multiplexing, or both? IEEE Communications Magazine, 52(12), Swindlehurst, A. L., Ayanoglu, E., Heydari, P., & Capolino, F. (2014). Millimeter-wave massive MIMO: The next wireless revolution? IEEE Communications Magazine, 52(9), Gao, Z., Dai, L., Mi, D., Wang, Z., Ali Imran, M., Shakir, M. Z. MmWave massive MIMO based wireless backhaul for 5G ultra-dense network, accepted by IEEE wireless communications magazine Singh, S., Kulkarni, M. N., Ghosh, A., Andrews, J. G. Tractable model for rate in self-backhauled millimeter wave cellular networks Proposed Rule by the Federal Communications Commission (FCC). Use of spectrum bands above 24 GHz for mobile radio services, January Aleksandar Ichkov has received his B.Sc. from Ss Cyril and Methodius University in Skopje in He is currently purchasing his M.Sc. degree in Wireless cellular communications at the Faculty of Electrical Engineering and Information Technologies, Ss. Cyril and Methodius University in Skopje. Aleksandar Ichkov has participated in international projects in his areas of interest (e.g. NATO funded SfP ORCA project) and is currently involved in the FP7 funded ewall project. Aleksandar Ichkov is an author/co-author of 6 research journal and conference publications. His major research interests lie in the areas of heterogeneous networks, resource management for future cellular networks, software-defined networking and virtualization.

17 Potentials for Application of Millimeter Wave Communications wireless sensor networking. Vladimir Atanasovski currently holds the positions of associate professor at the Institute of Telecommunications and Vice Dean for Finances at the Faculty of Electrical Engineering and Information Technologies, Ss. Cyril and Methodius University in Skopje. He has received his B.Sc., M.Sc. and Ph.D. from Ss Cyril and Methodius University in Skopje, in 2004, 2006 and 2010, respectively. Prof. Atanasovski participated in numerous international and domestic projects in his areas of interest (e.g. FP5 PACWOMAN, FP6 MAG- NET, FP7 ARAGORN, FP7 ProSense, FP7 QUASAR, FP7 FAR- AMIR, FP7 ACROPOLIS, NATO funded SfP RIWCoS and ORCA projects etc.). He is currently involved in the FP7 funded ewall and SCOPES funded ERTCON projects. Prof. Atanasovski is an author/coauthor of more than 120 research journal and conference publications and technical papers. His major research interests lie in the areas of resource management for future wireless networks, cognitive radio networks, traffic analysis and modeling, cross-layer optimizations and Liljana Gavrilovska currently holds the position of full professor at the Faculty of Electrical Engineering and Information Technologies, Ss. Cyril and Methodius University in Skopje. She is also Head of the Center for Wireless and Mobile Communications (CWMC) and Head of the research WINGroup (Wireless Networking research Group) working in the area of the wireless and mobile communications. She has received her B.Sc, M.Sc. and Ph.D. from Ss Cyril and Methodius University in Skopje, University of Belgrade and Ss Cyril and Methodius University in Skopje, respectively. Prof. Gavrilovska participated in numerous EU funded projects such as ASAP, PAC- WOMAN, MAGNET, MAGNET Beyond, ARAGORN, ProSense, FARAMIR, QUASAR, ACROPOLIS, CREW and ewall, SCOPES funded ERTCON projects, NATO funded projects such as RIWCoS and ORCA and several domestic research and applicative projects. Her major research interest is concentrated on cognitive radio networks, future mobile systems, wireless and personal area networks, crosslayer optimizations, broadband wireless access technologies, ad hoc networking, traffic analysis and heterogeneous wireless networks. Prof. Gavrilovska is author/co-author of more than 250 research journal and conference publications and technical papers and several books. For her work she was awarded several recognition among them the highest national award Goce Delchev. She is a senior member of IEEE.