EXPERIMENTAL STUDY ON THE IMPACT OF THE BASE STATION HEIGHT ON THE CHANNEL PARAMETERS. Aihua Hong and Reiner S. Thomae

Transcription

1 EXPERIMENTAL STUDY ON THE IMPACT OF THE BASE STATION HEIGHT ON THE CHANNEL PARAMETERS Aihua Hong and Reiner S. Thomae Technische Universitaet Ilmenau PSF 565, D Ilmenau, Germany Tel: Fax: aihua.hong, ABSTRACT In this paper, we investigated the experimental results about the statistical behavior of the channel parameters, based on the multiple-input multiple-output (MIMO) measurements performed in Ilmenau downtown, Germany. Our special interest focuses on the impact of the base station (BS) height on the statistical properties of the large scale channel parameters in two types of propagation condition: line of sight () and none line of sight (N). Furthermore, the dependence of the inter-sector correlation of the large scale channel parameters on the BS height is investigated in the overlapped area of two sectors of the same BS.. INTRODUCTION In the conventional cellular networks, base station (BS) is highly elevated by mounting on the top of high-buildings or high masks since large coverage by single BS has been expected. However, highly-elevated BS introduces simultaneously significant interference, which impairs the system performance. To find the trade off between coverage and interference, optimized deployment BS height is especially interesting. Different BS heights result in different propagation phenomena. As a consequent, the channel metrics, reflecting the propagation characteristics, show different statistical behavior. In this paper, we consider the impact of the BS height on the statistical properties of the large scale parameters (LSPs), which have been intensively studied in the European WIN- NER project [6] [5] and are the key channel parameters for the system-level channel modeling. The impact of the statistical properties of the large scale fading (LSF) on the system performance has been investigated by Fraile in [7] and by Saunder in []. The results indicate that an accurate modeling of LSF is a key point for the system-level performance assessment. The statistical properties of the LSPs have been intensively studied by authors in [8] in the urban hot-spot scenario, This work was supported by the Deutsche Forschungsgemeinschaft (DFG) within the TakeOFDM project in [9] [] in the indoor scenario, and in [] in the car2bridge scenario. In these comprehensive works, experimental results are presented based on the measurement data. However, the BS height is fixed in each of these measurement campaigns. The primary goal of this paper is to extend these studies to three dimensions. Whereby, the additional dimension is the height of the BS, besides the moving dimension of the mobile station (MS). In this paper, we include propagation loss, LSF, delay spread (DS), and cross-polarization ratio (XPR) [2] [3] in the terminology LSPs. Using the multiple-input multiple-output (MIMO) measurement data gathered in Ilmenau downtown, Germany, we studied the impact of the BS height on the statistical properties and the inter-sector correlation properties of the channel parameters. This paper is organized as follows. Section 2 provides a detailed description to the measurement campaign. Section 3 gives a short description to the definitions and extraction procedure of the LSPs, including propagation loss, LSF, DS, and XPR. Section 4 shows the experimental results, comprised of the propagation loss along the main street and two perpendicular streets, the statistical distributions of the LSPs, and their inter-sector correlation properties in the overlapped area. Finally, Section 5 concludes the paper. 2. MEASUREMENT CAMPAIGN The measurement campaign has been conducted in Ilmenau downtown, Germany, using HyEff sounder [4]. The measurement area can be characterized as a canonical small urban scenario with -3 floor stone buildings with the average height around 2 [m] at both sides of the street. The measurements were performed for downlink with approximately [w] transmitting power at the BS antenna at 5.2 GHz in a bandwidth of 2 MHz. The maximum multipath delay has been set to 3.2 µs. Dedicated antenna arrays are used both at the BS and at the MS. 8-element uniform linear patch array (ULA8) is used as the BS antenna with 9 λ element spacing, dual-polarization, approximate 2 degree directivity. At

2 the MS, 24 x 4-element stacked dual-polarized uniform circular patch array (SPUCPA24x4) is used with radius about λ and omni-directivity. The MS antenna array is put on the top of a trolley which is about.8 [m] above the ground. During the measurement campaign, the height of the BS has been changed from [m] to 6 [m] with the help of a lifting ramp. Figure shows the measurement routes and the orientation of the BS antenna sector in a 2-dimensional map. The names BS2 and BS5 stand for the [m] BSs while BS3 and BS6 stand for the 6 [m] BSs. Two BS sectors, orthogonal with each other, have been set to the each BS height, as indicated in Fig. as arrow. Since the surrounding buildings are around 2 [m] in height in the measurement area, the two BS heights correspond to two kinds of propagation. The one with [m] BS height, below rooftop, represents the micro-cell propagation while the other with 6 [m] BS height, above rooftop, represents the marco-cell propagation. The MS moves both along the main street and along two perpendicular streets. The former is show as a solid curve while the latter is shown as dashed curves in Fig.. They correspond to the and N propagation conditions, respectively. 3. PARAMETER DEFINITIONS AND EXTRACTION PROCEDURE At the beginning of this section, a short description to the definitions of the LSPs studied in this paper is given. After this, the methods used to extract the LSP values from the measurement data are described. 3.. Definitions of propagation loss and LSF The received signal power in mobile communication is often modeled as a produce of four factors (seen Eqn. ): the transmitted power (in [dbm]), the distance dependent path loss (in [db]) [2], log-normal distributed LSF (in [db]) [4], and small scale fading (in [db]) due to the superposition of multipath propagation. Small scale fading is also called as fast fading. P Rx = P T x P L + LSF + F F. () The difference between the transmitted signal power P T x and the received signal power P Rx is defined as propagation loss. Note that the antenna gain is included in propagation loss in this paper. The path loss characterizes the dependence of the signal attenuation on the distance between the transmitter and the receiver. As a result of multiplication of large number of random attenuating factors in the propagation, the LSF has a log-normal distribution with zero mean value and σ 2 variance [4]. Even though both the BS and the MS antenna arrays are dual-polarized, only the horizontal (H) polarization is used in the measurement data postprocessing due to its larger receive power compared with the vertical (V) polarization Definition of DS Generally speaking, the root mean square (RMS) of a given probability density function (PDF) f(x) is computed as the second central moment of f(x). The RMS can be expressed as σ = (x x) 2 f(x)d x, (2) f(x)d x where x is the first order moment of f(x) and can be calculated as x = xf(x)d x. (3) f(x)d x The DS is computed over the delay power spectral density or over the power delay profile (PDP) Definition of XPR XPR is defined as the ratio of the power between the co-polar link and the cross-polar link. Since the H-polarization is considered in the measurement data post-processing, the co-polar link stands for the HH-polarized link while the cross-polar stands for the HV-polarized link Extraction procedure of the LSP values from measurement data To get the values of propagation loss, LSF, DS, and XPR from the measurement data, the same method and extraction procedure as presented in [] are used in this paper. At first, a space-time averaging is performed to the PDP, squaring of the channel impulse response, to remove the small scale effect. In this paper, the time duration that the MS moves a distance of λ, corresponds to the time averaging length while the averaging over 8 BS antennas together with 24x4 MS antennas is the space averaging. After that, the measurement thermal noise is removed from the averaged PDP. The value of the averaged PDP at delay bin will be set to zero when it is smaller than the noise level margin. 9 [db] is used as the threshold of the noise level margin. The averaged PDP after removal of noise is the basis for the calculation of the Propagation loss, LSF, DS, and XPR values. The sum of this PDP in delay domain is the propagation loss. After removal of the distance dependent path loss values from the propagation loss as described in [], LSF values can be returned. The difference between the propagation losses of the HH-polarized link and the HV-polarized link is the XPR value. Substituting the averaged PDP after removal of noise into Eqn. 2, the DS value can be obtained.

3 4.. Propagation loss 4. EXPERIMENTAL RESULTS The experimental results of propagation loss from 4 BSs are presented in Fig. 3. Figure 3(a) shows the results along the main street in the propagation condition while Fig. 3(b) presents the results along two perpendicular streets in the N propagation condition. By comparing Fig. 3(a) with Fig. 3(b), it can be observed that the MS undergoes significant propagation loss when it disappears from the main street and enters into the perpendicular streets. It is due to the disappearance of the propagation path. This loss is defined as corner loss. It is around 2 [db] in the studied environment. By comparing the propagation losses between two BS sectors, it is found that the MS experiences almost the same propagation loss from two BS sectors both along the main street and along two perpendicular streets. Even though the BS antenna has been set to two sectors, the main street is located in the main beam of the both BS antenna sectors. This is the reason why there is no difference in propagation loss between two sectors along the main street. As the MS enters the perpendicular streets, the signal from the BS is reflected or diffracted to the perpendicular streets by scatters around the cross-road on the main street. These scatters work as virtual sources. Therefore, the MS receives also the same signal strength from two BS sectors along two perpendicular streets due to the same virtual source position on the main street, even though the perpendicular streets are located in the side beam of one BS sector but in the main beam of another BS sector. By comparing the propagation loss results of [m] and 6 [m] BS heights, it can be observed that the propagation loss with 6 [m] BS is higher than the propagation loss with [m] BS both in the and in the N propagation conditions, but not at a remarkable degree. In the propagation condition, the difference is higher when the MS is located near to the BS. It is due to the difference in propagation distance. As the MS moves aways from the BS, the 6 [m] BS height difference contributes fewer to the total final propagation distance. As a consequence, the propagation losses of two BS heights show more similarity Statistical distributions of the LSPs Figure 2 shows the cdf curves of the LSPs from the measurement data for the and N propagations. As stated in Subsection 2, the height of the BS varies from [m] to 6 [m] during the measurement campaign. Figure 2 shows the dependence of the LSPs on the BS height. It is observed in Fig. 2 that, both in the and in the N propagations, the higher the BS antenna array, the higher the LSF and DS values are, whereas, the lower the XPR values are. The reason is that the propagation between the BS and the MS varies from a two-ring or ellipse model to a one-ring model [2] [3] when the BS antenna array goes from below rooftop to above rooftop. When the BS antenna array is below rooftop, both the BS and the MS are surrounded by the local objects such as buildings and traffic lights. The wave propagation between the MS and the BS interacts with these objects. As a consequence, a two-ring or ellipse model is formed with the BS and the MS being the two centers of the two rings or two foci of the ellipse. The geometric range of the two-ring or ellipse model is limited by the local objects. As the BS height increases, the number of the local objects around the BS decreases. In the extreme case, there is no object around the BS when the BS is highly elevated. However, the MS is surrounded by objects which interact with the propagation waves. Furthermore, more objects are involved into the propagation including some far objects. Therefore, a onering model is formed between the MS and the BS with the MS being the center of the ring. The geometric range of the ring goes from the local objects to the far objects. Therefore, the propagation between the BS and the MS includes multipath components with long propagation delay which can not happen in the two-ring or ellipse model. As a consequence, both the LSF and the DS values are increased with increasing BS height and their cdf curves shift to the right hand side. Simultaneously, the additional multipath components with long propagation delay lead to the reduction of the XPR values Inter-sector correlation of the LSPs Table shows the experimental results of the inter-sector correlation of the LSPs. The sector pair BS2, BS5 has [m] BS height while the sector pair BS3, BS6 has 6 [m] BS height. It is found in Table that all LSPs are highly correlated in the overlapped area with correlation coefficient around.9. Furthermore, in the overlapped area, the intersector correlation coefficients of the LSPs with 6 [m] BS height are a little bit higher than these with [m] BS height. The reason is that more common scatters will be involved into the propagation when the BS antenna array becomes higher. As a consequence, the similarity in the propagation from two sectors in the overlapped area is increased. However, the inter-sector correlation could not be one, even though in the overlapped area where botho sectors have main beam. The fact is that the MS can still receive the signal coming from one sector s main beam area which is outside the overlapped area by reflection /diffraction /refraction. This signal reflected /diffracted /refracted by the objects outside the overlapped main beam leads to the reduction of the inter-sector correlation. The larger the ratio of the overlapped main beam area to the whole main beam width of two sectors, the larger the inter-sector correlation is. The inter-sector correlation could be one only if two sectors are fully overlapped to each other.

4 Table. Inter-sector correlation of the LSPs with different BS heights BS pair BS height LSF DS XPR BS2, BS5 [m].78 4 BS3, BS6 6 [m] CONCLUSIONS In this paper, based on the field measurement data collected in an urban area of Ilmenau, Germany, we studied the impact of the BS height on the propagation loss, on the statistical distributions of the LSPs, and on the inter-sector correlation properties of the LSPs. The experimental results show that the BS height has impact on the propagation loss in the case when the MS is located in the near of the BS. Both in the and in the N propagations, higher BS antenna array results in higher LSF and DS values but lower XPR values. The intersector correlation of the LSPs is around.9 in the overlapped area. Furthermore, the inter-sector correlation coefficients of the LSPs become higher with increasing BS height. Acknowledgment The authors would like to thank the colleagues from Technische Universitaet Ilmenau for their efforts in conducting the measurement campaign. 6. REFERENCES [] S. Saunders, Antenna and propagation for communication systems concept and design, Wiley, 999. [2] T. S. Rappaport, Wireless Communications, Principles and Practice, 2nd ed, Prentice Hall, New Jersey, 22. [3] M. Steinbauer, The radio propagation channel - a nondirectional, directional, and double-directional point-of-view, Technische Universitt wien (TUW), Austria, PhD thesis, Sep. 2. [4] [5] [6] P. Kysti et al., WINNER II Channel Models, IST WINNER II D..2 V.2, Sep. 27. [7] R. Fraile, J. Monserrat, N. Cardona, and J. Nasreddine, Impact of shadowing modelling on TD-CDMA system-level simulations, In Proc.3rd International Symposium on Wireless Communication Systems (ISWCS), Valencia, Spain, Sep. 26. m BS5/BS6 BS2/BS3 Fig.. Measurement routes in 2-dimensional Ilmenau city map [9] A. Hong, C. Schneider, G. Sommerkorn, M. Milojević, R. S. Thomae, and W. Zirwas, Experimental Evaluation of Correlation Properties of Large Scale Parameters in Indoor Pico-cell Environments, In Proc. 3rd International Symposium on Wireless Communication Systems (ISWCS), Valencia, Spain, Sep. 26. [] A. Hong, M. Narandzic, C. Schneider, and R. S. Thomae, Estimation of the correlation properties of large scale parameters from measurement data, In Proc. IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Athens, Greece, Sep. 27. [] N. Jalden, A. Hong, P. Zetterberg, P. Ottersten, and R. S. Thomae, Correlation Properties of Large Scale Fading Based on Indoor Measurements, In Proc. IEEE Wireless Communication and Networking Conference (WCNC), Hongkong, China, Mar. 27. [2] D-S. Shiu, G.J. Foschini, M.J. Gans, and J.M. Kahn, Fading correlation and its effect on the capacity of multielement antenna systems, IEEE Transactions on Communications, Vol. 48, Num. 3, pp , Mar. 2. [3] Y. Kai, Modeling of multiple-input multiple-output radio propagation channels, Royal institute of Technology (KTH), Sweden, PhD thesis, Oct. 22. [4] J. Salo, L. Vuokko, and P. Vainikainen, why is shadow fading lognormal?, In Proc. the 8th International Symposium on Wireless Personal Multimedia Communications, Aalborg, Denmark, pp , Sep. 25. [8] A. Hong, G. Sommerkorn, R. S. Thomae, and W. Zirwas, Considerations on the relationship between path loss and spatial characteristics based on MIMO measurements, In Proc. ITG Workshop on Smart Antennas, Ulm, Germany, Mar. 26.

5 N BS m BS 6m (a) LSF (db) N. BS m BS 6m (b) DS (ns) Propagation loss [db] m BS 6m BS BS3/BS2 BS6/BS Moving distsance of MS [m] (a) Propagation loss along the main street ( propagation condition) Propagation loss [db] m BS 6m BS BS3/BS2 BS6/BS5.5.3 N Moving distsance of MS [m] (b) Propagation loss along two perpendicular streets (N propagation condition). BS m BS 6m 5 5 Fig. 3. Propagation loss with two BS heights and two BS sectors in the and N propagation conditions (c) XPR (db) Fig. 2. Cdf curves of the LSPs from the measurement data in the and N propagation conditions with two BS heights