The Vault

The Effect of Human Blockage on the Performance of Millimeter-wave Access Link for Outdoor Coverage
Research Paper / Feb 2014

The Effect of Human Blockage on the Performance of Millimeter-wave Access Link for Outdoor Coverage Mohamed Abouelseoud and Gregg Charlton InterDigital, King of Prussia, PA 19406, USA, Abstract— In this paper we study the effect of human blockage on a new cellular architecture that incorporates millimeter wave technology for the access link. Assumptions, methodology, and simulation results are discussed to analyze the effect of human blockage on this type of deployment. A statistical model for human blockage and self blockage of the millimeter wave (mmW) signal is presented and analyzed. Our simulations show that even with a simple modulation scheme, an average network throughput of 400 Gbps/km2 is achieved. The objec- tive is to propose and demonstrate that such an architecture is capable of providing high throughput services to outdoor users even in heavily crowded areas. I. INTRODUCTION The demand for data and corresponding data delivery in- crease have been observed to approximately double every year or two. As we near the Shannon limit, air interface technology improvements provides only incremental gains while advances in networks are expected to provide much higher gains. Small cells are conceptually a way to achieve greater capacity by increasing spatial reuse of the same spectrum, and at the same time help reducing the total power transmitted in the network. The main downsides are the cost which grows with increasing infrastructure nodes, and the interference which needs to be managed. Another strategy to harvest more throughput involves the use of high frequency carriers. A benefit for high frequency carriers (e.g., in the millimeter- wave (mmW) spectrum) is the ability to inexpensively ac- quire significant amounts of spectrum with comparatively low fractional bandwidth (BW). The available bandwidth of the 60 GHz unlicensed spectrum alone is about 7 GHz (depending on country) which is more than all the available bandwidth in the traditional cellular spectrum. This opens the door to power saving as well, where simpler modulation schemes can be used and power is traded for spectrum. There has been a lot of work in both academia and industry investigating the effect of human blockage for 60 GHz indoor coverage, for example [1]–[5]. To our knowledge, no such re- search has considered outdoor coverage. Analysis for outdoor coverage has been generally considered in [6]. There have been other activities in the industrial standards, for example Fig. 1: mmW beams blocked by humans walking around the IEEE 802.11ad, WirelessHD, ECMA-387, IEEE 802.15.3c and the Wireless Gigabit Alliance (WiGig). The 60 GHz band in particular has some concerns. The high molecular oxygen absorption is an issue for long links and can cause around 16 dB/km attenuation. The regulatory landscape is also of some concern: the Federal Communications Com- mission (FCC) equivalent isotropically radiated power (EIRP) limit is only 40 dBm which will likely be the limiting factor in link distance rather than interference issues. Furthermore, the 60 GHz band is unlicensed, making it difficult to predict how well different radio access technologies (RATs) coexist and if operators will consider the band reliable enough. In addition, the mmW carriers have near optical properties with high penetration losses, high reflection losses, little diffraction and high dependence on line of sight (LOS) communications. It is expected that the mmW base stations (mBs) to be deployed in heavily crowded areas where the LOS between the mBs and the user equipments (UEs) can be easily blocked. This makes human blockage one of the biggest concerns for the mmW outdoor deployment, Figure 1. In this paper we consider a mmW based system for outdoor coverage and discuss the effect of human blockage on the performance of the system. A statistical model has been introduced to model the human blockage effect and the perfor- mance of the access link is estimated for several population densities. The architecture of the proposed system includes new small mmW base stations and multiple UEs connected to these base stations. 978-1-4673-6337-2/13/$31.00 ©2013 IEEE (a) 3D view (b) mBs deployments Fig. 2: University college campus mBs deployments II. STUDIED ENVIRONMENT AND MBS/UE PLACEMENT We study a mmW system deployed in a public university central campus area, Figure 2a. The abstracted deployment contains buildings with various heights and trees. The effect of people walking around the campus area (human blockage) has been modeled through a statistical model, discussed in details in section III-D. The proposed mBs placement, Figure 2b, employs 4 mBs placed at the college campus corners with the intent to provide coverage to the main quadrangle and an additional 5th mB placed at the center of the quadrangle itself. The WINPROP propagation simulator from AWE Commu- nications [7], a commercially available cell planning tool, is used to design and collect power measurements in the test environment. It allows for the creation of outdoor scenarios in which the user is able to specify the RF properties of the various building materials and vegetation areas. WINPROP generates data from each transmitter to a set of points on a user-defined dense square grid at a specified elevation above the ground. The UE locations are randomly drawn from the set of pre-generated grid points. The exercise is repeated for as many drops as are needed to ensure that the statistics reliably estimate network performance. Because of the directionality of the channel and the effect of blockage from a specific direction, the orientation of the UE is a matter of concern. Each of the dropped UEs is randomly oriented at an azimuth angle between 0◦ and 360◦ and an elevation angle between 0◦ and 90◦ where the horizon is considered to be 0◦. III. SIMULATOR DESIGN Our system simulator comprises two main modules. The first is the cell planning tool, WINPROP, which is capable of making detailed propagation predictions based on ray tracing while taking transmit antenna patterns into account. The tool determines the delay, signal strength, and type of interaction the path rays encountered from transmitter to each of the grid points. This allows the creation of accurate multi-path channel models between each transmitter and receiver pair. The second module is written in Matlab. It uses the channel models based on the data created by WINPROP to estimate the throughput achieved by each UE in the system by modeling the impacts of transmit power, inter-cell interference, receive antenna gain, modulation scheme, scheduler parameters, etc. The statistics are then collected to generate an estimate of the overall network performance. A. Channel Model Individual rays (LOS, reflected, and diffracted) are traced between each transmitter and the entire set of user-defined grid points in a deterministic fashion. The signal strength, delay and angle of arrival for each ray are functions of the re- flection and diffraction angles, building materials, propagation model, transmit antenna pattern etc. The received power/delay combination allows for the creation of a tapped delay profile. A fading profile is applied to each of the identified rays to capture the fast fading environment. The fading profile follows a Rician distribution with k-factor of 10 because of the LOS characteristics of the channel. B. Transmitter and Receiver Antenna Design To generate the mB antenna patterns, we use a uniform rectangular planar phased array (URA) with 12x12 elements and 0.32λ element spacing. Each URA is assumed to cover 90◦ in azimuth; consequently, the array pattern beamwidth and spacing is set to support the number of beams required to span the 90◦ coverage. The simulation constrains the arrays to use one of a subset of the beams that it could possibly generate. The array generates 5 beams each separated by 18◦, the antenna gain at the crossover point between two adjacent beams is set to be -6 dB. The beams have a maximum gain of 21.6 dBi, a 3-dB beamwidth of 6.5◦ and a crossover point between the neighboring beams at 8.9◦. The transmit power is adjusted to meet the FCC regulation of 40dBm equivalent isotropically radiated power (EIRB). Two types of receive antennas have been tested, an ideal omni-directional antenna and a directional antenna. The omni- directional antenna has a uniform 0 dBi gain in all directions. The directional antenna is a 4x4 planar phased array antenna with element spacing equal to half the wavelength. The antenna array generates 5 beams in the zero degree elevation angle plane with around 60◦ beamwidth and overlapping at the 3 dB point. This is equivalent to approximately 30◦ beam separation. The receiver beam that provides the maximum total received power is selected and used for reception. A set of down-tilted beams and a set of up-tilted beams are used to form a 5x3 beam array to accommodate the various orientations of the UE antenna. Also, two back-to-back planar phased array antennas are used to cover the front and the back of the device to allow reception all around the UE. C. Transmit Beam Assignment and receive Antenna Pattern Selection UE-mB association is a two stage process. In the first stage, the UE measures the received power from all antenna beams of all mBs. A quasi-omni antenna pattern is used to measure the received power when UEs use directional antennas, otherwise an ideal-omni antenna pattern is used. The quasi-omni antenna pattern is the pattern of a single element patch antenna which effectively constrains the possible pointing direction of the array. The mB array and corresponding beam providing the largest received power is selected. In the second stage, only for directive antennas, all possible receiver beams at the UE are considered to determine the best antenna pattern. D. Human Blockage Model One of the biggest concerns regarding the use of mmW outdoors is the need to establish LOS links. In a dynamic environment where people are continuously moving around the receiver, there is a high likelihood that the people completely block the LOS and non-LOS (NLOS) components. 1) Blockage from Other Users: We define a probability for each path to be blocked by one or more persons. This probability depends on two main factors: the path length of the ray from the mB to the UE and the height of both the mB and the UE. For a specific population density, the longer the path length, the higher the probability of a blocker in that path. On the other hand, the higher the mB, the lower the probability that the ray will be blocked. When the mB antenna is higher than the UE antenna, only blockers close enough to the UE can actually block the signal. For an mB mounted at 4 m, a 1.5 m UE height and blockers 1.75 m tall, using simple trigonometry, one can show that only those blockers in the last 1/10 of the ray path length block the path. The rays are traveling above the rest of the blockers and are not affected by their existence. In order to estimate a ray’s blockage probability, we simu- late an open area where blockers are randomly dropped. The blockers have a specified uniform density around a UE that is dropped at a specific distance away from the mB. Each blocker and the UE are assumed to occupy a 30 cm by 30 cm area and the rays are assumed to be narrow enough that one blocker is sufficient to block the ray path. For each ray path length and for a specific blocker density, 10,000 Monte Carlo runs are carried out. At each run, blockers are dropped randomly and both the probability of blockage and the PDF of the number of blockers in case of blockage are estimated. Figure 3 shows the probability of blockage for various densities of blockers between .01 human/m2 to 1 human/m2 and various ray path lengths. Figure 4 shows the probability of a ray with specific length to be blocked by 0, 1, 2, 3, 4 and 5 or more blockers at an example density of 0.1 human/m2. Figure 3 and Figure 4 are used during the system simulation to estimate the event of an existence of a blocker. For each dropped UE, according to each ray path length, the probability of blockage is extracted from Figure 3. Based on our own lab experiments, and matching [1], it is assumed that a specific ray blocked by a human standing in its path suffers attenuation of 20 dB. For simplicity, it is assumed that a blocker will only affect the received power of the blocked ray and would not change the angle of arrival. If multiple humans are located in the ray path, each will result in 20 dB attenuation. 0 0.2 0.4 0.6 0.8 1 0 200 400 600 0 0.2 0.4 0.6 0.8 1 Blocker density− (Blker/m2) mB height=4m,UE height=1.5m Ray path length (m) Pr ob ab ilit y of b lo ck ag e Fig. 3: The blockage probability as a function of the ray path length and the blockers density 0 50 100 150 200 250 300 350 400 450 500 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Ray path length (m) Blocker density = .1Blk/m2, mB height=4m,UE height=1.5m Prob of 0 blocker/path Prob of 1 blocker/path Prob of 2 blocker/path Prob of 3 blocker/path Prob of 4 blocker/path Prob of 5 blocker/path or more Fig. 4: Multiple blockers probability for .1 blocker/m2 2) Self Blockage: Another source of blockage is self block- age. The person who’s holding the UE is blocking the signal from certain directions. It is assumed that the signal coming through the back side of the UE is attenuated. For an average blocker 175 cm tall and 60 cm in width holding the UE 30 cm away, the blocking area is a 90◦ azimuth angle centered at the back of the UE with elevation angles between 62.8 and -90 degrees. Any signal coming from the direction of the self blocker is assumed to be attenuated by 40 dB. E. SINR and Throughput Calculation The instantaneous signal interference noise ration (SINR) is calculated for each UE in the network. SINR is mapped to its corresponding throughput according to the Shannon capacity formula, R = B log2(1 + SINR) which provides an upper bound on performance or via a QPSK SINR to throughput lookup table based on LTE [8]. F. Simulation Parameters The total number of UEs per drop is 50 which corresponds to a UE density of 730 UE/km2. We consider 10 UE drops, 0 500 1000 1500 2000 2500 3000 3500 0 10 20 30 40 50 60 70 80 90 100 User Throughput (Mbps) CD F (% ) RR, no self blockage RR, self blockage PF, no self blockage PF, self blockage Fig. 5: Self blockage effect on the user throughput CDF 15 20 25 30 35 40 45 0 10 20 30 40 50 60 70 80 90 100 Total Cell Throughput (Gbps) CD F (% ) RR, no self blockage RR, self blockage PF, no self blockage PF, self blockage Fig. 6: Self blockage effect on the network throughput CDF each of 1000 transmission time interval (TTIs). A TTI duration is assumed to be 1 msec. The carrier frequency is 60 GHz. Following 802.11ad, 2640 MHz bandwidth is used and divided into 512 sub-carriers of which 336 are used for data so the BW assigned for data is 2640/512*336 MHz. The UE noise figure is 6 dB. The users are allocated resources via a round robin scheduler. The default for the receiver antenna is ideal omni at 1.5 m height. Each mB can have more than one array. The mBs antennas down-tilt angle is 0◦ with EIRP of 40 dBm and 4 m height. IV. PERFORMANCE EVALUATION A variety of scenarios have been created to estimate the likelihood of a human blocking the rays arriving at a receive grid point. The idea is to quantify the probability of human blockage so that a single, i.i.d. blockage probability can be assigned to each ray rather than go with a much more complicated approach in which actual human blockers are placed in the ray tracing simulator. For the scenarios studied, the blocking probability has been found to be dependent on the population density, the transmitter-receiver distance, and height difference between the mB and the UE. (a) .01 blocker/m2 (b) .3 blocker/m2 Fig. 7: University campus blocker density A. Self Blockage Effect Self blockage is defined as the blockage of the signal by the UE’s human operator. Figures 5 and 6 show the performance of a system where the self blockage is taken into account and where there is no self blockage. A UE that is not held by a user is an example of a case where self blockage is not present. The self blockage effect is seen by the reduction of both the cell edge user throughput and the network throughput. Even with the use of a scheduler that depends on the channel conditions (proportionally fair scheduler (PF) [9] with β = 1), the effect of the self blockage remains significant. On the other hand, the effect on the network throughput is smaller in the case of the PF scheduler compared to the simple round robin scheduler which is due to the ability of the PF scheduler to smartly distribute the channel resources. B. Human Blockage The system performance has been measured at various human densities using ideal omni directional antennas at the UE and a round robin scheduler with the self blockage effect taken into account. Results have been generated for an empty environment (no blockers), low density (0.01 blocker/m2), crowded environment (0.1 blocker/m2 ) and very crowded environment (0.3 and 0.5 blockers/m2). Figure 7 shows the distribution of the blockers for two of the studied human blocker densities in the 76000 m2 university campus. The .01 blocker/m2 results in a total of 760 blockers and the .3 blocker/m2 results in a total of 22800 blockers distributed uniformly around the university campus. Results are shown in Table I using two SINR to throughput mapping techniques, the QPSK SINR to throughput lookup table and Shannon formula. The QPSK represents a simple modulation scheme while Shannon formula puts an upper bound on the throughput that can be achieved. The follow- ing metrics have been calculated: 10th and 50th percentile throughput, the average UE throughput, the percentage of UEs with no service and the average network throughput. It is noted that in low blockage density environments, the system performance is almost the same as no blockage at all. This is because in such case, the probability of a ray to be blocked is low and if one path happened to get blocked the mB and the UE can adjust their transmit and receive beams pattern to avoid SINR to TP mapping and human blocker density 10th perc. TP (Mbps) 50th perc. TP (Mbps) avg UE TP (Mbps) 0 TP UEs (%) avg Net. TP (Gbps) QPSK 0 blk/m2 87.8 477.3 596.8 3.8 29.8 QPSK .01 blk/m2 85.5 464.6 590.3 4.2 29.5 QPSK .1 blk/m2 8.1 449 540 7.8 27 QPSK .3 blk/m2 0 322.6 441.4 19.4 22 QPSK .5 blk/m2 0 186.8 368 32.2 18.4 Shannon 0 blk/m2 88.2 621.8 933.3 3.8 46.7 Shannon .01 blk/m2 85.8 606.3 923.5 4.2 46.2 Shannon .1 blk/m2 8.1 552.8 847.7 7.8 42.4 Shannon .3 blk/m2 0 350.4 710.8 19.4 35.5 Shannon .5 blk/m2 0 189.7 600.4 32.2 30 TABLE I: Statistics for various human blockage densities 0 500 1000 1500 2000 2500 3000 3500 0 10 20 30 40 50 60 70 80 90 100 User Throughput (Mbps) CD F (% ) Ideal omni, 0blker/m2 Ideal omni, 0.1blker/m2 Ideal omni, 0.3blker/m2 Ideal omni, 0.5blker/m2 2 dir planar array 5x3, 0blker/m2 2 dir planar array 5x3, 0.1blker/m2 2 dir planar array 5x3, 0.3blker/m2 2 dir planar array 5x3, 0.5blker/m2 Fig. 8: Rx antenna effect on the UE throughput CDF that path. In more crowded environments, the percentage of UEs that get no throughput at all is increasing and the network throughput is decreasing, 32.2% of UEs with zero throughput for the 0.5 blocker/m2 compared to 3.8% for the no blockers. However, even with very high blocker densities, the average UE throughput as well as the average network through are still very high compared to any other cellular architecture. The main problem in this case is the high percentage of UEs with zero throughput. This can be solved if these UEs have access to another network (LTE for example) deployed in the same region or by introducing high gain directional antenna at the receiver as shown in the next section. C. Blockage and Receiver Directional Antennas Figure 8 and Figure 9 show a comparison between the per- formance with ideal omni receive antenna and two directional back-to-back planar array receive antennas, each producing 5x3 beams under various human blockage densities. The QPSK SINR to throughput lookup table is used to calculate the achieved throughput. It is noted from Figure 8 that with the directional antenna, the percentage of unserved UEs in high density blockage scenarios is less than that for the ideal omni antenna case. The use of directional antennas resulted in around 14% of the UEs with no coverage at 0.5 blocker/m2 while 32.2% of UEs have no service in the ideal omni case. 0 5 10 15 20 25 30 35 40 45 50 0 10 20 30 40 50 60 70 80 90 100 Total Cell Throughput (Gbps) CD F (% ) Ideal omni, 0blker/m2 Ideal omni, 0.1blker/m2 Ideal omni, 0.3blker/m2 Ideal omni, 0.5blker/m2 2 dir planar array 5x3, 0blker/m2 2 dir planar array 5x3, 0.1blker/m2 2 dir planar array 5x3, 0.3blker/m2 2 dir planar array 5x3, 0.5blker/m2 Fig. 9: Rx antenna effect on the network throughput CDF This shows that directional antennas are essential for the operation of mmW systems in crowded areas. This is due to the higher gain that can be applied to the best path. V. CONCLUSION An approach was outlined in which the effects of human blocking can be modeled as a random process rather than specifically creating human blockers in the simulation tool. The effect of self blockage on the signal as well as the blockage of the signal in a dense environment was simulated. It was demonstrated that dense concentrations of human blockers can have significant impacts on mmW communications. This impact can be reduced by the use of directional antennas at the receiver. However, the system still provides throughput that is superior to the current wireless technologies. REFERENCES [1] W. Jing, R. Prasad, and I. Niemegeers, “Analyzing 60 GHz radio links for indoor communications,” IEEE Transactions on Consumer Electronics, vol. 55, no. 4, pp. 1832–1840, Nov. 2009. [2] S. Singh, F. Ziliotto, U. Madhow, E. Belding, and M. Rodwell, “Blockage and directivity in 60GHz wireless personal area networks: from cross- layer model to multihop MAC design,” IEEE J. Select. Areas Commun., vol. 27, no. 8, pp. 1400–1413, Oct. 2009. [3] H. Lee, J. Bok, B. G. Jo, G. H. Baek, and H. G. Ryu, “Indoor WPAN communication system using 2-dimensional array antenna in 60GHz frequency band,” in Computing, Communications and Applications Con- ference (ComComAp), 2012, jan. 2012, pp. 158 –161. [4] X. Zhang, L. Lu, R. Funada, C. S. Sum, and H. Harada, “Physical layer design and performance analysis on multi-gbps millimeter-wave wlan system,” in Communication Systems (ICCS), 2010 IEEE International Conference on, nov. 2010, pp. 92 –96. [5] X. An, C. Sum, R. Prasad, J. Wang, Z. Lan, J. Wang, R. Hekmat, H. Harada, and I. Niemegeers, “Beam switching support to resolve link- blockage problem in 60 GHz WPANs,” in Personal, Indoor and Mobile Radio Communications, 2009 IEEE 20th International Symposium on, sept. 2009, pp. 390 –394. [6] Z. Pi and F. Khan, “System design and network architecture for a millimeter-wave mobile broadband (mmb) system,” in Sarnoff Sympo- sium, 2011 34th IEEE, may 2011, pp. 1 –6. [7] AWE Communications, “WinProp software suite,” http://www. [8] “IEEE 802.16m evaluation methodology document (EMD),” IEEE 802.16 Broadband Wireless Access Working Group, Jan 2009. [9] D. Tse and P. Viswanath, Fundamentals of Wireless Communication. Cambridge University Press, 2005.